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.1004590 | Identification of High-Impact cis-Regulatory Mutations Using Transcription Factor Specific Random Forest Models | Cancer genomes contain vast amounts of somatic mutations, many of which are passenger mutations not involved in oncogenesis. Whereas driver mutations in protein-coding genes can be distinguished from passenger mutations based on their recurrence, non-coding mutations are usually not recurrent at the same position. Therefore, it is still unclear how to identify cis-regulatory driver mutations, particularly when chromatin data from the same patient is not available, thus relying only on sequence and expression information. Here we use machine-learning methods to predict functional regulatory regions using sequence information alone, and compare the predicted activity of the mutated region with the reference sequence. This way we define the Predicted Regulatory Impact of a Mutation in an Enhancer (PRIME). We find that the recently identified driver mutation in the TAL1 enhancer has a high PRIME score, representing a “gain-of-target” for MYB, whereas the highly recurrent TERT promoter mutation has a surprisingly low PRIME score. We trained Random Forest models for 45 cancer-related transcription factors, and used these to score variations in the HeLa genome and somatic mutations across more than five hundred cancer genomes. Each model predicts only a small fraction of non-coding mutations with a potential impact on the function of the encompassing regulatory region. Nevertheless, as these few candidate driver mutations are often linked to gains in chromatin activity and gene expression, they may contribute to the oncogenic program by altering the expression levels of specific oncogenes and tumor suppressor genes.
| Precise regulation of gene expression is controlled by cis-regulatory modules (CRM) containing binding sites for transcription factors (TF). The genome-wide location of all TF binding sites can often be obtained by ChIP-seq (chromatin immunoprecipitation followed by deep sequencing), yet in most cases only a minority of the binding peaks actually represent functional CRMs that control the transcription initiation of a bona fide TF target gene. Here, we investigated for 45 cancer-related TFs how machine-learning approaches can be used to predict functional TF target CRMs. After careful evaluation of their performance, we used these TF-target classifiers to predict which cis-regulatory mutations may have a significant impact on gene regulation by evaluating whether the mutation causes a significant gain or loss in the probability that the CRM is a functional TF target. We found that Random Forest classifiers can achieve more than 100-fold higher specificity for mutation prediction compared to the simple approaches based on scanning with position weight matrices. By scanning somatic mutations in breast cancer genomes and in the HeLa genome, we finally show that our TF-target classifiers can identify high impact non-coding mutations that are associated with concordant TF binding, gene expression changes and chromatin activity. In conclusion, TF-specific Random Forest classifiers can be used to prioritize cis-regulatory mutations in cancer genomes with high accuracy.
| Gene regulation determines the identity and behaviour of all cells, and perturbations of gene regulatory programs can cause cells to change their identity or become transformed into cancer cells. Such perturbations of gene regulatory networks can be caused by driver mutations in signalling molecules, transcription factors (TF), and chromatin modifiers [1]. In addition, driver mutations can also occur within the non-coding genomic regions that control transcription, the cis-regulatory modules (CRM). CRMs harbour recognition sites for one to many transcription factors and regulate the transcription initiation rate at one or more nearby target genes. Recently two cancer-related CRM mutations have been discovered, namely: a highly recurrent mutation in the TERT promoter that is found in many cancer types [2–5]; and a more distally located enhancer mutation upstream of the TAL1 gene in T-cell acute lymphoblastic leukemia (T-ALL) [6]. These two examples of driver mutations generate de novo binding sites for oncogenic transcription factors. Particularly, the TERT promoter mutations create new ETS-like binding sites (GGAA), while the TAL1 mutation creates a MYB binding site. Interestingly, the latter is associated with a very significant gain of the activating histone modification H3K27Ac, indicating that the neomorphic enhancer actively regulates TAL1 expression.
To analyze cis-regulatory variation on a genome-wide scale and to prioritize candidate driver mutations, several types of information can be exploited and integrated [7–11]. A first class of methods is based on filtering all candidate variants, such as single nucleotide variants (SNV) and small indels, to retain only those that affect “interesting” nucleotides. For example, a method called FunSeq retains mutations that affect “sensitive” genomic positions (FunSeq also combine other types of data [8]). Sensitive positions are determined by FunSeq as positions that are significantly infrequently substituted in the normal human population. Other methods, like OncoCis [9] and RegulomeDB [11], retain mutations that are located in candidate regulatory regions, as determined by publicly available regulatory data (e.g., from ENCODE [12]). The disadvantage of this approach is that regulatory activity observed in a cancer sample may not correspond to any of the available annotation, particularly when the mutation creates a gain-of-function CRM, or in other words, publicly available regulatory annotation is not always indicative for the function of the CRM in the cancer sample under study. A solution to this problem could be to profile chromatin states in the actual cancer sample itself, but the currently available biochemical methods (mainly open chromatin profiling and ChIP-seq) still require relatively large amounts of input material, which is often not available for tumor biopsies. A second class of approaches is based on QTL analyses, whereby DNA variants are correlated with DNA methylation, chromatin accessibility, or gene expression. These methods have been mostly applied to identify variation in the normal population [13–16] but when larger cohorts of more than 200 cancer samples become available (full genome, methylome, and transcriptome for each sample), they can, in principle, also be applied to identify cancer driver mutations. A related approach is to select mutations that cause allelic shifts in ChIP-seq reads, which was shown to identify functional SNPs that change enhancer activity in HepG2 cells [17]. A third class of approaches, which can be used in combination with the first two, investigates the mutated sequence itself, using information about TF recognition motifs and selects mutations that affect transcription factor binding sites. This can be achieved by scoring the reference and mutated sequences with a position weight matrix (PWM) of a particular TF, assessing the impact of the mutation by the difference of the scores for the reference and mutated sequence. For example, FunSeq calculates “motif maker” and “motif breaker” scores for PWMs and returns a list of all affected PWMs, for each mutation. A limitation of these methods is that PWM-scanning methods are notorious for generating high amounts of false positive predictions, which can affect the accuracy of PWM-based mutation scoring, yielding excessive amounts of false-positive mutations. The prediction of cis-regulatory mutations using PWMs would therefore benefit from more advanced models of TF target prediction, so that the impact of a mutation can be assessed more accurately, in the context of an entire CRM. By incorporating CRM context into a predictive model, we may achieve a higher accuracy for predicting functional cis-regulatory mutations. When using CRM prediction and classification methods to assess mutations, we can build on a large body of previous methods, using various kinds of features such as TF motifs, other (higher-order or structural) sequence features, sequence conservation, or chromatin related data. CRM prediction methods that are based on motif scanning usually score (sliding) sequence windows for the presence of clusters of TF binding sites, either for the same TF (i.e., homotypic clusters) or for different co-regulatory TFs (i.e., heterotypic clusters) [18–21]. CRM classification methods applying machine learning, using a training set of positive CRMs, are more flexible in terms of the types of features, and once a model is trained, it can be used to predict similar CRMs in the genome. For example, Narlikar et al., employed a Lasso model with a collection of 701 position weight matrices (PWMs), de novo discovered motifs and Markov models and were able to predict heart enhancers [22]. kmer-SVM [23] or IMM [24] use a PWM-blind approach whereby the features are entirely learned from the sequence of training CRMs, as over-represented k-mers or Markov chains. Classifiers can also be trained using chromatin data, such as Chromia [25] which uses chromatin data such as histone modification profiles as features in its model, trained on TF binding sites defined by ChIP-seq. It was shown for Chromia that such models, when combined with a PWM, can yield accurate genome-wide predictions of TF targets. More recent methods for enhancer classification use multiple layers of epigenomic data, such as chromHMM [26].
We reasoned that such complex CRM models, when trained on sets of CRMs targeted by specific oncogenic or tumor suppressor TFs, could provide an interesting approach to score putative cis-regulatory mutations, and to assess whether the mutation may cause a gain or loss of a functional TF target. To this end, we developed 45 Random Forest classifiers for more than forty different TFs, each trained on subsets of functional CRMs (i.e., regions bound by the TF that actively regulate target gene expression). We validate these models by cross-validation and genome-wide scoring, and apply them to identify PRIME mutations (mutations with high PRIME score: Predicted Regulatory Impact of a Mutation in an Enhancer), both using simulated substitutions and real somatic mutations in a large breast cancer cohort from TCGA and in the HeLa genome.
Chromatin immuno-precipitation coupled with sequencing (ChIP–seq) allows identifying genome-wide locations of TF binding, however usually only a fraction of observed ChIP-seq peaks (0.9%-54.6%) are functional, in the sense of being actively involved in regulating target gene expression [27]. Here, we wanted to develop TF-specific enhancer models by training them only on functional target CRMs (Fig 1). To identify such training sets of functional ChIP-seq peaks we searched for peaks that are located near up- or down-regulated genes in response to a perturbation of the TF, or that are located near tightly co-expressed target genes with the TF (see Methods). To obtain statistically significant correlations between ChIP-seq data and co-expressed gene sets, we applied a procedure called “track discovery”, whereby ChIP-seq peak sets from ENCODE and other resources are tested for their enrichment on a gene set [28]. Particularly, we compared 344 sets of TF target genes against 1000 ChIP-seq tracks. This led to the identification of 45 sets of positive training CRMs for 41 distinct transcription factors, most of which are related to cancer. The average size of the training sets ranges between 6 (POU5F1) and 3901 (YY1) positive samples (S1 Table).
For each set of positive CRMs we trained Random Forest (RF) classifiers each consisting of 151 decision trees that optimally distinguish the positive CRMs from sets of negative sequences. As negative sequences we used randomly sampled regions from the human genome with the same size and GC content as the positives, in a 1:20 ratio. We trained different types of RF models depending on the type of features used in the decision trees (Fig 1B). The first model, M1, uses ten motifs of the TF and ten motifs of co-regulatory TFs. These twenty motifs are selected by motif discovery on the training CRMs, out of a collection of nearly 10.000 candidate position weight matrices. The second model, M2, uses as features the fifteen most representative regulatory tracks: five open chromatin tracks, five active histone modification tracks, and five ChIP-seq tracks of potential co-regulatory TFs. Model M3 combines all features of M1 and M2, in total twenty motifs and fifteen tracks. Similarly to motif features, these tracks were selected by track discovery (see Methods). To avoid over-fitting, ChIP-seq tracks of the query TF itself were excluded as candidate features. The performance of each of the 45 TF models (for the three different RF types) was evaluated using the area under the precision-recall (AuPR) and area under the receiver operating characteristic (AuROC) curves, as achieved by the model in a five-fold cross-validation (Fig 2, S1 Fig). We compared the performances to a baseline model (M0) that predicts TF targets by simple PWM-scanning using the PWM of the query TF; and to a previously published alternative classifier based on Support Vector Machines trained on k-mers (Mk) [29]. Collectively, M1 (the RF classifiers utilising only motif information) achieved on average across the 45 datasets an AuPR of 0.62; similar to Mk (kmer-SVM: AuPR = 0.61), and both are much higher than M0 (PWM-only: 0.37). Note that we prefer the AuPR since the AuROC is less reliable for imbalanced training sets with high numbers of negative sequences [30]. The best performing M1 models are for SRF, GABPA, CEBPB, STAT2, and YY1. In total, thirty of the RF classifiers achieved an AuPR greater than 0.5 (Fig 2A, S2 Table). Additional quality control and robustness analysis revealed that most models show stabilization of cross-validation performance (S2 Fig); that Random Forests outperform other machine-learning approaches on the same data such as Support Vector Machines or Logistic Regression (S3 and S4 Figs); and that the performance of the models does not depend on the size of the training set (S5 Fig) nor on the information content of the main PWM (S6 Fig) of the query TF.
Next, we investigated the performance of M2 models that use data tracks instead of motifs. These models have drastically higher AuPRs than motif-based models (average AuPR = 0.87), with all 45 models having an AuPR above 0.5. For M2, the best performing models are for SRF, E2F4, JUNB, NFE2L2, and TP53. Interestingly, several TFs with ill-performing M1 models have a much improved AuPR score; for example TAL1 has an M1 model with AuPR = 0.13, whereas M2 with tracks achieves an AuPR of 0.69 (Fig 2A). Finally, for combined models the performance increases even further, although not much beyond M2 (average AuPR = 0.9). Interestingly, TFs can be grouped into different classes, where each class has different types of features contributing to the classifier, as determined by Gini impurity (see Methods) (S7 Fig). For 20/45 models, the TF PWMs contribute more than 20% of all features (e.g., TP53 in Fig 3). For another class of 7/45 models the co-regulatory factor PWMs contribute more than 20% of all features and dominate over the TF PWMs (NANOG in Fig 3); and for 39/45 models the sum importance of the three data tracks groups was dominant providing more than 50% of the feature importance (e.g., MYC in Fig 3).
Fig 3 also shows an example decision tree from the ensemble for the TP53, NANOG, and MYC examples. By investigating the feature importance we can obtain more insight into the CRM code; for example that TCF12 and ATF2 tracks are important to predict NANOG targets; or that SIN3A ChIP-seq peaks in MCF-7 are important to predict MYC binding. Note that this does not mean that SIN3A and MYC necessarily co-bind in the same cell.
In conclusion, we trained multiple well performing models for the classification of TF-specific regulatory target regions. The results suggest that not all information in a CRM can be captured by TF and co-regulatory TF motifs. The track-based M3 models yield an upper-limit to the classification performance based on sequence/motif information alone. Ultimately however, to identify cis-regulatory mutations (see further below), we will rely on sequence/motif-based models because those are generally applicable, as they do not depend on the availability of multiple regulatory tracks in the cancer and normal sample.
To further validate our trained CRM classifers we applied them genome-wide to predict new functional TF target CRMs (including M1, M3, and Mk models). To this end, we split the genome into overlapping sliding windows with sizes corresponding to the average lengths of the sequences in the training set (ranging from 400 bp for NANOG to 2350 bp for GLI3, with an average of 900 bp). The number of newly predicted functional binding sites for M1 models ranges from several hundreds to several tens of thousands. To assess the accuracy of new predictions, we calculated the enrichment of the TF ChIP-seq peaks among newly predicted CRMs, excluding training CRMs. We found a significant recovery for 31 of the 45 models using a RF classifier with motifs only (M1). The five best performing models regarding genome-wide predictions, as measured by the Normalized Enrichment Score (NES) given by i-cisTarget [28], are TP53 (NES = 31.1), IRF1 (NES = 21.5), STAT2 (NES = 17.45), POU5F1 (NES = 16.25), and SPI1 (NES = 14.15). Interestingly, although the cross-validation performances of the motif-only RF and k-mer SVM models were highly similar, the genome-wide prediction accuracies are overall much higher for the RF models (Fig 4A). Particularly, 31 of the 45 M1 models show significant recovery of the correct ChIP-seq peaks, compared to only 17 of the 45 Mk models (Fig 4A). We also performed genome-wide predictions for M3 models, which incorporate regulatory data tracks as features in the model. Although the cross-validation performance of M3 models is much better than M1, the M3 models did not result in more TFs with high-confidence genome-wide scoring, since again 31 models show significant recovery of correct ChIP-seq peaks. This indicates that M1 models with motif-information alone are already very performant in genome-wide predictions, and this is confirmed by inspecting the correlation between the validation scores (i.e., TF track enrichment scores), being very high between M1 and M3 (0.876), while they are both better than Mk (S8 Fig). We also analyzed whether predicted CRMs show enrichment for active chromatin marks, such as H3K27Ac. Indeed, for 26 to 40 models this is the case for M1 and M3 models respectively. More generally, for the majority of TFs (39/45 models) the newly predicted CRMs are enriched for regulatory active chromatin states, as determined by chromHMM segmentations [26] from ENCODE (S9 Fig), with the strongest models overlapping with promoter states being E2F1, TAF1, YY1, E2F7, and KLF5, and the strongest models overlapping with enhancer states being E2F7, TCF12, and FOSL1. In conclusion, we evaluated the quality of the TF-target classifiers in an alternative way, independent of cross-validation performances and found that most RF classifiers are enriched for ChIP-seq peaks of the query TF and active chromatin marks.
Whereas current methods for the prediction of changes in TF binding sites assess local changes in the actual TF binding site, for example using a change in the PWM score [8–10], here we wanted to assess whether TF-specific enhancer models allow identifying cis-regulatory mutations that have an impact on the global CRM score. Firstly, we simulated mutations by creating substitutions in gene promoters. Particularly, we selected the 900 bp promoter of 752 curated cancer driver genes [31–35] and changed at each position the sequence into each of the three alternative nucleotides. To measure the impact of each possible single nucleotide variation (SNV) we introduce a score, called PRIME, that is calculated as the difference between the RF classifier scores for mutant and reference sequences. PRIME values range between -1.0 to 1.0 and allow capturing both gains and losses of CRM function. To evaluate the quality of PRIME, we hypothesized that nucleotides with higher PRIME scores should be more conserved. Indeed, nucleotides tend to be under higher constraint with increasing absolute PRIME score (Fig 5A). There is one caveat to this analysis however: low PRIME scores can represent a mixture of sites that are not bound by either allele, and bound sites where the variant does not change binding. To distinguish between these, we simulated substitutions inside ChIP-peaks (true sites) versus substitutions outside ChIP-peaks (not bound by either allele) in terms of conservation (S10 Fig). The results demonstrate that although nucleotides belonging to real binding sites tend to be more conserved with increasing PRIME score, also high-scoring mutations outside ChIP peaks are enriched for high phastCons scores. We performed a similar validation experiment using open chromatin data and found that substitutions with high PRIME score tend to be more located in accessible regions than low PRIME substitutions, suggesting their potential involvement in CRM function (S11 Fig). As an example, we show in Fig 5B the promoter of the E2F1 gene, where the E2F4 model identifies a hotspot of high PRIME substitutions. Convincingly, these positions overlap with the summit of an E2F4 ChIP-seq peak and cover the entire E2F4 consensus site. We expected an increased specificity (rather than sensitivity) of mutation detection with Random Forest models (M1) compared to the simple PWM model (M0), because PWMs are known to suffer from high false positive rates [36]. To test whether this is indeed the case for the E2F1 promoter, we scored all possible substitutions in this promoter with several E2F4 PWMs, and indeed found many non-functional positions that show a change in PWM score (Fig 5B). This suggests that random forest classifiers are better suited to detect cis-regulatory variation than PWMs.
We then scored a large collection of real non-coding somatic mutations collected from three cancer whole genome sequencing studies: 50 AML samples (N = 19797) [35], 21 breast cancer samples (N = 183703) [37], and 25 melanoma samples (N = 1875157) [38]. Similarly to the simulated substitutions, we found that predicted high-impact mutations are more conserved than mutations with low predicted regulatory impact (S12 Fig). Also, mutations with high absolute PRIME score (greater than 0.4) are enriched for chromatin states corresponding to functional regulatory elements such as active promoters, weak promoters, and strong enhancers (S13 Fig). When compared to measuring the impact of a mutation by the change in PWM score, also on this set we find that the Random Forest models show greater specificity than PWMs (S14 Fig).
In conclusion, the TF specific classifiers can identify regulatory variation affecting the activity of functional CRMs, making this a feasible strategy for the prediction of cancer driver mutations.
To test whether the Random Forest CRM models may be suitable to identify cancer driver mutations we examined in detail a recently published cis-regulatory mutation in the TAL1 promoter in T-cell Acute Lymphoblastic Leukemia (T-ALL) [6]. Particularly, a recurrent (5.5% of patient T-ALL samples) mutation is caused by a short insertion that creates one or two de novo binding sites (depending on the length of the insertion) for the MYB transcription factor, a well-known regulator involved in T-ALL. Our 45 models do not contain a MYB-specific model (only a MYBL2 model), and none of the 45 models predicted a high PRIME score for this site. However, when we trained a MYB-specific M1 model, using MYB target CRMs as training set (obtained by anti-MYB ChIP-seq in the Jurkat T-ALL cell line [6]), the TAL1 mutation yields a very high PRIME score (from 0.054 in the reference to 0.3774 in the mutated CRM). Thus, only the MYB model identifies this gain-of-function mutation. In contrast, when we used the PWM for MYB, which yields an increase in PWM score of 0.1 for the actual driver mutation compared to reference (from 0.844 to 0.949), we also find two other PWMs of the 45 tested M0 models (GABPA and CEBPB) that yield a similar PWM score increase (more than 0.1) and that have a high PWM score (>0.9) for the mutant sequence. In other words, although the MYB PWM can identify the mutation, it is also falsely predicted by other PWMs, but not by other Random Forest models.
For a MYB model to prioritize this mutation in the genome, out of all possible somatic mutations, the model also needs to be specific. To test this, we scored a large set of control somatic mutations (both SNVs and insertions) with the same MYB model (Fig 6A and 6B). These control mutations were selected from breast cancer somatic mutations from TCGA. Since MYB is not known to be involved in breast cancer, we could argue that each mutation with a high PRIME score for the MYB model would be a false positive prediction. This analysis shows a remarkable specificity, with only 2/19796 SNVs and 0/7323 insertions predicted as high-impact mutation for MYB (PRIME>0.3). For comparison, using the MYB PWM identifies 179 SNVs and 354 insertions with a delta of 0.1 or more in the control set. For the TAL1 promoter mutation we can conclude that the predicted high impact corroborates the gain of CRM activity observed in the Jurkat cell line that harbors this mutation, as measured by H3K27ac (Fig 6D).
The empirical distribution of background PRIME scores for the MYB model allows estimating the significance of this PRIME score using a z-score (see Methods), which is 26.5 for the Jurkat insertion. A similar but shorter insertion was found in the MOLT-3 cell line and in several patient samples; these insertions generate only one new MYB binding site and yield z-scores between 1.41 and 21.45 for the MYB model. Note that we used these thresholds based on the MOLT-3 insertion, determined from the empirical distribution of PRIMEs for SNVs or insertions thresholds (9.65 and 14.03, respectively) to determine model-specific PRIME thresholds for other models, further below.
To investigate why the Random Forest model for MYB achieves such high specificity compared to the PWM, we analysed the feature importance within the MYB model and found that both MYB motifs and co-regulatory TF motifs contribute significantly to the classification decision. Interestingly, the most important co-regulatory motif is RUNX, a known co-regulatory factor of MYB (Fig 6C). The combination of MYB motif clusters and co-regulatory motifs allows assessing the impact of a mutation taking the context of a CRM into account. To illustrate this, we tested whether inserting exactly the same sequence at random position does not always produce a similar gain of function. Indeed, when we inserted the same sequence into 100 randomly chosen genomic loci having the same 3 bp flanking nucleotides we found that the PRIME score strongly depends on the surrounding sequence context. For example, the Jurkat insertion generates a PRIME score equal or higher than 0.32 (the observed PRIME in the TAL1 enhancer) in only 10/100 locations, indicating that most genomic locations are not susceptible to this insertion, in terms of MYB-dependent activity (S15 Fig).
We also performed this analysis for another well-known promoter mutation, in the TERT promoter. The TERT promoter harbors two recurrent mutations and these are among the highest recurring mutations in cancer (between 33% and 85% in melanoma [2]). The original articles reporting this mutation suggested that this mutation generates an ETS-like binding site (GGAA) and that ETS family members might cause an up-regulation of the TERT gene due to this gain of function binding site mutation. More recently, these mutations were linked to de novo binding by GABPA, which also binds to a GGAA motif [39]. However, our GABPA model did not result in a significant PRIME score (PRIME = 0.026; Z-score = 0.99). We constructed four alternative models for different ETS-like factors using their respective top ChIP-peaks as training set (see Methods), namely ELF1, ELK1, ELK4, and ETS1. For two of these models, namely for the EFL1 and the ELK1 model, we found significant PRIME scores (z-score = 2.83 and 6.49, respectively), although the PRIME score was much lower than for the TAL1 mutation (the highest PRIME is 0.097 for ELK1 = >TERT, compared to 0.32 for MYB = >TAL1). Remarkably, looking at promoter activity data by H3K27Ac, across a cohort of melanoma samples we generated before [40], we could not observe any gain of activity in the samples that harbor the mutation (Fig 6D). We can conclude for the TERT promoter that the predicted impact scores are significant but modest and that they corroborate with low observed impact at the promoter activity level.
Next we used the TF-specific random forest models to prioritize cis-regulatory mutations in 498 re-sequenced breast cancer genomes from TCGA, for which gene expression data is available [41]. We specifically scored all SNVs and insertions located in promoters (see Methods). To evaluate whether mutations with high PRIME scores could have a functional impact on gene expression, we evaluated the expression level of the target gene in the sample with the mutated promoter, compared to all other samples (using z-scores). Indeed, this shows a clear association of changes in gene expression with predicted impact of promoter mutations (S16 Fig). Moreover, the median absolute z-score values of gene expression increases with increasing PRIME score. When we focused on promoters of cancer related genes (the list of 752 curated cancer driver genes), we found only 36 genes having single nucleotide mutations with absolute PRIME score > 0.3 (Fig 7A, S3 Table). Using the model-specific z-scores (with a cutoff of 9.65 for SNVs and 14.03 for insertions), 84 genes are found with significant mutations. When we applied our models to small insertions in promoters, we found only three high impact insertions, namely in the SOX9 promoter (gain for E2F1), the METTL14 promoter (YY1 loss), and the NLGN2 promoter (PAX5 gain). Interestingly, two of these three mutations are recurrently mutated across the TCGA cohort (Fig 7A). Expanding our search to 10 kb, and focusing only on breast-cancer related transcription factors as targets (along the lines of the MYB-TAL1 gain), we found an additional 91 SNVs and 11 insertions with high impact (S4 Table), including a gain of TP53 CRM upstream of SOX5, and a loss of a SIX5 site upstream of NR3C1. Interestingly, these two latter insertions are recurrent across the TCGA cohort (39 and 59 samples, respectively). Furthermore, expression of SOX5 target gene is significantly higher in the samples with the insertion, compared to the samples without the insertion (Fig 7B). Overall, we thus found a limited number of potentially harmful cis-regulatory mutations, given that in Fig 7A we pooled together all the results across 498 breast cancer genomes.
Finally, we reasoned that if a mutation really causes a gain of CRM activity, this should be directly visible as a change in chromatin activity, such as increased chromatin accessibility, increased H3K27Ac signal, or decreased DNA methylation. Unfortunately, none of these data are available at the genome-wide level for the TCGA cohort (DNA methylation is currently available for 450K probes, which is too sparse for our low number of high-impact mutations). To test a potential correlation between mutations with high PRIME scores and chromatin, we therefore used the HeLa genome [42,43], for which H3K27Ac data is available from ENCODE (GSM733684) [12]. Scoring all 13923 small insertions located in 10 kb regulatory space around TSS of the HeLa genome, for our 45 models, we found 141 variations with significantly high PRIME scores, based on the model-specific z-scores (S17 Fig). A small subset of these are indeed located in regions with H3K27Ac signal that is specific, or semi-specific for the HeLa cell line (compared to H3K27Ac data for 108 other samples, see Methods), possibly indicating that these mutations have a local effect on the activity of the enhancer (Fig 8A). To test whether any particular TF has more mutated CRMs, we compared the amount of gains and losses for each TF model stratified on whether the variation is a known polymorphism from dbSNP or not, the latter representing possibly somatic mutations (Fig 8B). Interestingly, this shows that oncogenic TFs that are important for HeLa, namely MYC, E2F7, JUND, and STAT1, have more gains than losses, specifically for variations not in dbSNP. Vice versa, YY1, a known repressor related to cancer, has almost no gains in non-dbSNP variations, while dbSNP variations have an almost equal amount of gains and losses. We believe that such skews towards “relevant TFs” strongly indicate a cis-regulatory effect for this group of mutations. AP-1 (JUN/FOS dimer) is indeed a relevant factor for HeLa, because it is the most enriched motif and track among all the HeLa-specific H3K27Ac peaks (AP-1 motif rank = 1, NES = 7.36; FOS ChIP-seq NES = 7.20). A clear example of a SNP with a cis-regulatory effect is shown in Fig 8C, where a heterozygous SNP that is predicted to generate a gain in JUN binding (PRIME = 0.21; z-score = 16.28), indeed shows a gain of JUN binding in the HeLa ChIP-seq data for JUN (Fig 8D). All the reads in the JUN ChIP-seq peak contain the alternative (non-reference) allele, which generates an AP-1 binding site in a favorable CRM context. Note that compared to the MYB-TAL1 interaction (see above), which generates a de novo super-enhancer that is unique to the Jurkat and MOLT-3 samples, for the HeLa genome we did not identify such strong effects in H3K27ac gains. Indeed, only four insertions are located in a H3K27Ac peak that is unique to HeLa. One of these four has an absolute PRIME score close to 0.3 (-0.295) (i.e., observed frequency is 0.25). Interestingly, this predicted mutation is located near CDH10, a gene that is specifically expressed in HeLa, compared to other cell types of the human body map, as determined by Landry et al. [43] (S17 Fig).
In conclusion, TF-specific random forest classifiers can identify cis-regulatory variation with potential impact on the function of a promoter or enhancer.
Whole genome sequencing of cancer samples has revealed that cancer genomes harbour thousands to hundreds of thousands of non-coding mutations. Sifting through all these mutations to identify mutations that contribute to the oncogenic process is a key challenge in cancer genomics, as it is yet unclear to what extent regulatory mutations can be actual driver mutations. For coding mutations, driver mutations are usually identified by their significant recurrence across a patient cohort. For example, TP53 is mutated in 37% of breast cancer samples in the TCGA and Sanger cohorts [41,44]. Thus far, although non-coding mutations are more numerous than coding mutations, very few recurrent cis-regulatory mutations have been found, and recent pan-cancer analyses concluded that in fact only one potential cis-regulatory mutation, in the TERT promoter is highly recurrent [2,3]. TERT promoter mutations have been identified in 6 of 14 cancer types where they occur in 3 to 62% of cancer samples, depending on the type of cancer [45]. They are associated with higher expression of TERT, both in promoter-luciferase assays [3], and in patient samples [46]. Because these mutations generate GGAA sites, it was hypothesized that this could lead to increased activation by TFs of the ETS-family, which recognize GGAA consensus sites. Recently, it was shown that the TF is in fact GABPA. While the TERT mutation has relatively low PRIME scores and no gain of H3K27Ac, the TAL1 promoter mutation, which generates a de novo MYB binding site, causes neomorphic/ectopic enhancer activity as seen by a very strong and broad H3K27Ac signal spanning a large region encompassing the mutation. Interestingly, our enhancer models for MYB predict the TAL1 mutation as a high-impact mutation.
We have applied CRM prediction methods to the reference genome and to cancer genomes, and calculated the differential CRM score between the reference sequence and the sequence carrying a single nucleotide variant or a small insertion. CRM prediction methods are computational techniques to predict regulatory regions (e.g., promoters, enhancers) based on their sequence content and usually take advantage of transcription factor motifs [47,48]. Whereas CRM prediction methods have often been applied to identify tissue-specific enhancers (e.g., human heart enhancers [22], Drosophila tissue-specific enhancers [24], etc.), their application to identify TF-specific target CRMs is relatively limited [49–51]. Here we specifically train models on training sets of functional TF ChIP-seq peaks. We define functional peaks as the signficant subset (or “leading edge” [28]) of peaks that are located near genes that are up- or down-regulated upon perturbation of the TF. Compared to previous methods that often rely on k-mers, Markov chains, or de novo discovered motifs in the training set [24,29], we have here assessed the power of using large PWM collections. Since we know (to a large extent) the identity of the TF for each PWM, this strategy allows selecting a set (we choose 10) of specific PWMs for the query TF, and a set of PWMs for potential co-regulatory TFs. We furthermore believe that the power of using PWMs for CRM predictions will further increase, given the recent progress in high-throughput determination of TF binding specificities [52]. Interestingly, we found that for a subset of TFs the co-regulatory transcription factor motifs have a higher feature importance than the motifs of the query TF itself. An important example of this category is the cancer-related TF FOXM1, which requires a Random Forest model with co-regulatory factor motifs to identify FOXM1 targets in the genome. This is also corroborated by the fact that FOXM1 ChIP-seq peaks are not enriched for any FOXM1 motif [53]. Therefore, when potential cis-regulatory mutations are scored for their potential motif-breaking or motif-making effects, using the FOXM1 motif would render meaningless results.
As an alternative we have also trained CRM models using regulatory data as features, such as histone modifications and chromatin accessibility. Corroborating previous work by others [25,49,54], such models have a higher performance compared to sequence-based prediction methods. This likely implies that CRM function/output cannot entirely be captured by sequence and motif content of the CRM itself. In this respect, we consider the models using experimental regulatory data to represent an upper limit to the CRM prediction problem. Although for some TFs the sequence-based models reach an accuracy close to their respective data-based model, the performance of most TF models is still far from perfect (15 TFs with AuPR<0.5). To evaluate our models and to compare different approaches we used standard cross-validation. Importantly however, we included a complementary evaluation approach, namely the genome-wide prediction of CRMs. We then tested the performance of each model by assessing the overlap between predicted TF-specific CRMs and TF ChIP-peaks, excluding the ChIP-peaks used in the training set. This allowed to functionally validate our predictions, and to compare our models with alternative modelling approaches (namely, a simple PWM, a k-mer SVM approach, and a gapped k-mer SVM), showing that Random Forest classifiers outperformed these alternative methods (S18 and S19 Figs). In addition, this experiment showed that the predicted CRMs using sequence-based models represent regions with typical characteristics for CRMs, such as cross-species conservation and enhancer/promoter-related chromatin states, including DNAse I hypersensitivity and H3K27Ac enrichment.
Encouraged by the high CRM prediction performance, we then applied the optimized TF-specific CRM models to mutated cancer genomes, to predict cis-regulatory mutations with potential impact on CRM function. Using thresholds of the PRIME score based on a z-score, which is calculated on a TF model-specific empirical distribution, by scoring that model on 20000 variants from TCGA; we found relatively few mutations, with only 0.1%– 1.2% high impact mutations with PRIME>0.3 (comparable to the MYB-TAL1 mutation) per cancer genome, on average. This was true for the large TCGA breast cancer cohort, but also for smaller cohorts of melanoma (25), another breast cancer (21), and AML (50) genomes. Nevertheless, the high-scoring mutations, as well as simulated substitutions we introduced in promoters of cancer genes, overlap significantly with conserved nucleotides and with enhancer/promoter chromatin states, indicating that these predictions are valuable. The low number of high-impact mutations was again confirmed when we scored insertions found in the HeLa genome (scoring 10 kb regions around TSS) [42,43], where we found only one insertion (near CDH10) with high impact and associated with a gain in H3K27Ac signal (as the reported insertion in the TAL1 enhancer). Note that although the thresholds we have applied to PRIME scores are based on model-specific z-scores, the stringent z-score cutoffs are largely inspired by the TAL1 examplars, and could be fine-tuned or relaxed in the future if more experimentally validated cis-regulatory mutations are discovered. Another reason why we identify few mutations may be partly due to the limited number of models we have built (currently 45; with an additional nine models specifically added for the HeLa genome), but we speculate that even with more models, the total number of high impact cis-regulatory mutations will be low. This is indeed corroborated with the low number of cis-regulatory mutations that are found to be recurrent across cancer samples [7,45]. Importantly, when PWMs are used to score cis-regulatory mutations, more than 100-fold excess of false-positives are predicted. This is mostly due to the context of the CRM, for example, when multiple binding sites of the same factor are present in a CRM (i.e., a homotypic cluster [55]), adding or deleting a single binding site may not have any dramatic effect on the CRM. This is measured in the first layer by the CRM score, but not by the individual PWM scores. In a second layer, different features are combined via optimized parameters in an ensemble of decision trees, further increasing the specificity (S20 Fig). Recently, a similar method was published, called deltaSVM [56], which also uses a machine-learning approach to train a model and score reference and variant sequence, to calculate a delta score. Although deltaSVM was mainly applied to GWAS data to score natural variation, it could in principle also be applied to cancer mutations. This method is complementary to our Random Forest models because it is trained on a different type of training set (all open chromatin regions of a sample, rather than TF-specific models) and it uses entirely different features (k-mers for the deltaSVM, compared to PWMs and data tracks for our PRIME scores). Therefore these two approaches are complemenatary and both can predict the impact of mutations in an enhancer, as shown in S21 Fig, on a data set of synthetic enhancer sequences. Nevertheless, since our RF models are often more specific in genome-wide scorings, they may also yield less false-positive predictions on genomic variation (S22 Fig). A future challenge will be to use M3 models to score mutations, incorporating epigenomic data tracks into the model. To this end however, besides a fully re-sequenced cancer genome and germline control, also a cancer and control epigenome would be required. As an alternative, if full genomes can be phased into haplotypes, M3 models could be exploited to score the allele-specific impact of heterozygous variants.
In conclusion, we presented an approach to model CRM context allowing to predict and prioritize candidate cis-regulatory mutations in cancer genomes that could affect CRM function, and provide a solution to the excess of false-positive predictions obtained by approaches using position weight matrices. Our predictions on cancer genomes furthermore suggest that the majority of non-coding mutations may be passenger mutations, and that only few top-scoring mutations may contribute to the oncogenic program as cis-regulatory drivers.
TFs target genes were selected either from curated TF perturbation gene signatures (MSigDB) [57] or from a GENIE3 [58] inferred co-expression network focused on melanoma (skin (77): GSE7553 [59], GSE28914 [60], GSE13355 [61]; primary melanoma (90): GSE7553 [62], GSE19293 [63], GSE23376; metastasis (71): GSE7553, GSE10282, GSE22968 [64]. As parameters of GENIE3 we used as input list of transcription factors 2245 items (combined from TRANSFAC® Professional database and MSigDB collection (v4.0)), and a threshold of 0.005 (2041 regulatory modules were identified). In total, we selected 224 curated sets and 120 predicted sets based on the availability of TF ChIP-seq data.
For each target gene set we performed “track discovery” using i-cisTarget [28], on all available TF ChIP-seq tracks. If the corresponding TF ChIP-seq track was significantly enriched, the leading edge was selected as optimal target CRMs. As parameters of i-cisTarget we used a search space size of 20 kb around TSS. For four TFs (E2F1, FOXM1, ESRRA, MYC) we found two different studies that provided a target gene set for this TF and for which i-cisTarget found the ChIP-seq of the factor enriched, thus for which we could identify a training set of functional target CRMs. These models are named as E2F1_2, FOXM1_2, ESRRA_2, MYC_2.
Besides the 45 models trained using the high-throughput procedure above, we trained a MYB model using the top 500 peaks from ChIP-seq data from [6] and several models for ETS-like factors, namely ELK1, ELF1, ELK4, and ETS1, each time using the top 500 ChIP-seq peaks from the ENCODE data [12].
To select DNA motifs and regulatory tracks enriched in the set of training CRMs we again used i-cisTarget, but now using regions as input. i-cisTarget uses a large collection of motifs (9,713 PWMs) and human regulatory tracks (2,046) derived from different resources [28]. Two groups of motifs where selected: the top ten enriched motifs of the query TF and the top ten motifs of co-regulatory TFs. In addition, for M2 and M3 models, three groups of the most representative regulatory tracks were selected: up to five open chromatin tracks, five histone modification tracks (active marks), and five ChIP-seq tracks of potential co-regulatory TFs selecting only enriched tracks.
We performed 5-fold cross-validation. The selection of features using i-cisTarget was performed only once, on the entire training set. This does not affect the cross-validation performance because this filtering step is performed in an unsupervised way (without using the negative samples) [65]. We confirmed this by performing i-cisTarget on every fold, without using the left-out samples, thereby using different features during each fold, but as expected this had no influence on the the AuPR values for cross-validation (S23 Fig). Note that for small training sets (e.g., POU2F1 has only 6 positive CRMs in the training set, the 5-fold cross-validation leaves out only 1 or 2 samples, thus making it more a leave-one-out cross-validation.
Selected enriched PWMs and tracks were used for numerical representation of the DNA regions. For the motif scores we used Cluster-Buster (with default parameters except option -c was set to zero to obtain a score for every sequence) employing a Hidden Markov Model to score CRM sequences for clusters of binding sites [18]. We consider the PWMs as features and for each PWM we calculate on a CRM (which is a sample, so positive or negative) the total Cluster-Buster motif-cluster score for that PWM. This means that for each feature (PWM) we have one score per CRM (so per window). The final M1 models thus contain only 20 features, and each region’s feature vector contains 20 Cluster-Buster scores. For M2 and M3 models we also include data tracks as features. For their scores we assigned the maximum score of broad or narrow peaks (corresponds to signalValue column in the bed file format) overlapping with the scoring region (the overlap was obtained using BEDtools [66]). As negatives we used 20x more sequences, randomly selected from the genome without restriction on genomic locations, with the same length and GC distribution as the positives. As Random Forest implementation we used the scikit-learn Python package [67] 151 decision trees were used for each classifier. Changing the number of trees can be indicative of the stabilisation of the cross-validation performance (S1 Fig). The parameter max_features (responsible for number of features to consider when looking for the best split) was set to sqrt(number of features). To calculate the feature importance we used the Gini impurity criterion averaged across trees, using the whole training data, again with the implementation from scikit-learn library [67].
The performance of the RF classifiers was compared with simple PWM matching (M0) and with another supervised machine learning methods, namely kmerSVM (Mk) [29] and gapped kmerSVM (Mgk) [51]. The performance of the Mk, Mgk, M1, M2, M3 models where evaluated in 5-fold cross-validation. To evaluate performance of the M0 we obtained AuROC and AuPR curves varying the motif score threshold. For M0 we used as PWM matching tool MotifLocator [68] with default parameters except option -t was set to zero. For each TF we selected the PWM that was most enriched PWM in the training set. As a background model we used a first order Markov model with nucleotide transition probabilities estimated using human genome (hg19) sequence.
Genome-wide predictions were performed by segmenting the genome in overlapping sliding windows. The size of the window is chosen specifically for each TF as the average length of the regions used for training, and the overlapping segment between windows is equal to 200 bp.
We selected a set of 752 known cancer drivers from different sources (MSigDB, TCGA, COSMIC). In the regions 900 bp upstream of these genes we replaced every nucleotide to each possible variant and scored with M1 models; PRIME score was calculated (difference between M1 classification score in mutant versus reference sequenc) to estimate contribution of the location and type of nucleotide substitution on the CRM score.
The sequence around each mutation was scored with M1 models. Several sliding windows around each mutation were taken into account using a shift equal to 10% of the region. For each mutation the window with the maximum score of the classifier is taken into account.
Bigwig file with phastCons scores [69] based on alignment of 46 placental mammal species was downloaded from UCSC Genome Browser. We used a custom Python script and bigWigToWig [70] tool to calculated the score for each position.
All chromatin states identified across nine human cell lines (HSMM, GM12878, HUVEC, H1-hESC, K562, HepG2, NHEK, HMEC, NHLF) using ChromHMM were downloaded from the UCSC browser [71] and combined into one dataset. We calculated the enrichment of positively predicted functional TF binding sites in different chromatin states using the GAT tool [72]. Only values where enrichment or depletion is significant (pvalue<0.05) are taken into account.
From VCF files provided by TCGA consortium we selected non-coding somatic mutations (SNVs and insertions passed filtering criteria) falling in 500 bp regions around TSS. This yielded 51117 SNVs and 7323 insertions combined from 498 full-genome sequenced breast cancer samples. Z-scores of gene expression across samples were calculated using RPKM values (max value per gene) as derived from processed RNA-seq data for 768 breast cancer samples.
Processed full genome sequencing results of the HeLa cell line (CCL-2 and Kyoto cells) were downloaded as VCF files. Only insertions located in +- 10 kb non-coding regions around TSS and identified in both studies [43,73] were selected for scoring (N = 13923) and all heterozygos HeLa falling in H3K27Ac data (N = 89451). For the HeLa H3K27Ac data we used broadPeak formatted data generated by ENCODE (on the HeLa-S3 cells) [12] from which signalValue was used for creating z-scores as follows. Candidate regulatory regions (that we defined before [28]) were scored by a large collection of 109 H3K27Ac ChIP-seq data across different cell types including HeLa (46 datasets from Blueprint project [74,75], 23 from ENCODE [12], 3 from DEEP [76], 33 from McGill EMC (http://epigenomesportal.ca) and 4 in-house generated datasets). The acetylation score was multiplied by the fraction of the peak length that overlaps with the candidate regulatory region. If more than one peak overlaps with the same regulatory region then the average value was used. Finally, each regulatory region had a score for all the 109 acetylation datasets and z-scores were computed across all the samples.
Python scripts are available at https://github.com/aertslab/primescore.
|
10.1371/journal.pgen.1001119 | The Metabolic Enzyme ManA Reveals a Link between Cell Wall Integrity and Chromosome Morphology | Synchronizing cell growth, division and DNA replication is an essential property of all living cells. Accurate coordination of these cellular events is especially crucial for bacteria, which can grow rapidly and undergo multifork replication. Here we show that the metabolic protein ManA, which is a component of mannose phosphotransferase system, participates in cell wall construction of the rod shaped bacterium Bacillus subtilis. When growing rapidly, cells lacking ManA exhibit aberrant cell wall architecture, polyploidy and abnormal chromosome morphologies. We demonstrate that these cellular defects are derived from the role played by ManA in cell wall formation. Furthermore, we show that ManA is required for maintaining the proper carbohydrate composition of the cell wall, particularly of teichoic acid constituents. This perturbed cell wall synthesis causes asynchrony between cell wall elongation, division and nucleoid segregation.
| The bacterial cell is resistant to extremes of osmotic pressure and protected against mechanical damages by the existence of a rigid outer shell defined as the cell wall. The strength of the cell wall is achieved by the presence of long glycan strands cross-linked by peptide side bridges. The cell wall is a dynamic structure continuously being synthesized and modified to allow for cell growth and division. Damaging the cell wall leads to abnormal cellular morphologies and cell lysis thereby making it a key target for antibiotics. Here we describe the involvement of ManA enzyme in cell wall construction of the rod shaped bacterium Bacillus subtilis. In cells lacking ManA the normal extension of the cell wall is blocked. Consequently, manA mutant cells display abnormal morphologies and fail to properly package and segregate their DNA resulting in the formation of large cells containing disorganized multiple chromosomal copies. Based on these findings, we propose that appropriate cell wall synthesis is necessary for synchronizing chromosome architecture with cell growth.
| The bacterial cell wall is a key determinant of cellular morphology that provides structural support and mechanical protection. The structural dynamics of the bacterial cell wall enable elasticity, growth and division. The cell wall of the Gram positive bacterium Bacillus subtilis (B. subtilis) is primarily composed of peptidoglycan (PG), a net-like polymer of glycan strands cross-linked by peptide bridges, and anionic phosphate-rich polymers. Both PG and anionic polymers play critical roles in maintaining the structural integrity and viability of the bacterial cell [1].
PG strands comprise alternating units of N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc). PG synthesis initiates in the cytoplasm with fructose-6-phosphate and proceeds through a linear pathway to generate the precursors UDP-GlcNAc and UDP-MurNAc-pentapeptide [2]. The following steps are membrane bound and carried out using a special recycled lipid carrier, undecaprenyl phosphate. Initially, the MraY transferase attaches MurNAc-pentapeptide to the lipid carrier thereby yielding lipid I. Consequently, GlcNAc is added to lipid I by the MurG enzyme producing lipid II. Lipid II carries the basic PG monomer composed of GlcNAc and MurNAc-pentapeptide. Next, lipid II is flipped across the membrane, the PG monomer is cleaved of the undecaprenyl phosphate and incorporated into the growing chain [2]–[4].
Labeling of the newly synthesized PG revealed that it is inserted in a helical pattern along the lateral cell wall [5]–[8]. Accordingly, atomic force microscopy of the B. subtilis cell wall exposed helical PG cabling arrangement with glycan strands up to 5 µm in length, longer than the bacterium cell itself [9]. A protein implicated to play a key role in inserting new PG is the actin homolog MreB. MreB forms a dynamic helical scaffold that serves as a platform onto which the cell wall machinery localizes [5], [10]–[16]. B. subtilis has three MreB isoforms, called MreB, Mbl and MreBH, which have been demonstrated to colocalize in a single helical structure [17]. Mutations within the genes encoding these isoforms, as well as in other essential PG components, induce severe morphological defects (e.g.: [6], [13], [15], [18]–[21]).
In B. subtilis the anionic polymers can be either bound to PG, wall teichoic acid (WTA), or anchored to the cytoplasmic membrane, lipoteichoic acid (LTA). The major form of WTA comprises glycerol phosphate polymer [1], [22]–[24] and the minor form is a polymer of glucose (Glc) and N-acetylgalactosamine (GalNAc) [1], [22], [25], [26]. Fluorescence analysis has revealed that WTA enzymes are localized at division sites and along the lateral sides of the bacterial cells [27]. Similar to the PG pathway, WTA biosynthesis begins with formation of nucleotide sugars in the cytoplasm and proceeds with a membrane step that utilizes the same lipid carrier undecaprenyl phosphate. Notably, mutations in the WTA and/or LTA pathways lead to loss of rod shape and non-uniform thickening of the PG layer [28]–[31], suggesting coordinated biogenesis of the cell wall components.
Here we show that the sugar metabolic enzyme ManA (mannose phosphate isomerase), which is part of the mannose phosphotransferase system, is unexpectedly necessary in rich medium, when mannose is not utilized as a carbon source. In the absence of ManA, cells display abnormal morphologies and fail to properly package and segregate their chromosomes. Furthermore, we demonstrate that these abnormal phenotypes are due to a role played by ManA in cell wall construction. We show that the lack of ManA perturbs proper cell wall carbohydrate composition and thereby causes asynchrony between cell growth, division and nucleoid segregation.
To identify new components required for cell division and chromosome segregation, we performed a transposon mutagenesis and screened for B. subtilis mutants exhibiting growth defects (Materials and Methods). The selected mutants were then subjected to a visual microscopy assay. One of the slow growing mutants had a striking phenotype with the cells exhibiting a severe shape defect and atypical nucleoid morphologies (Figure 1). Evidently, mutated cells lost the characteristic rod shape typical of wild type B. subtilis cells and instead appeared as elongated spheres, which were significantly larger than normal. This spheroid like morphology resembles the phenotype described for mutants defective in cell wall synthesis (e.g.: [6], [13], [15], [18], [20], [21], [29], [30]). In addition, we observed internal membrane invaginations in many of the mutant cells indicative of inappropriate cell divisions (Figure 1F). Moreover, DAPI staining revealed a variety of abnormal nucleoid structures, which have lost their spatial organization in comparison to the well-organized wild type chromosomes (Figure 1).
Cloning and sequence analyses revealed that the transposon was inserted within the coding region of the manA (mannose phosphate isomerase) disrupting its function. Accordingly, deletion of manA was sufficient to confer the observed defects and ectopic expression of manA fully complemented the mutant phenotype (Figure S1A-S1C). manA encodes a conserved enzyme that catalyzes the reversible isomerization of fructose-6-phosphate (Fru-6-P) and mannose-6-phosphate (Man-6-P) [32]. Introducing point mutations into the predictable ManA active site abolished the ability of the protein to complement the null manA phenotype (Materials and Methods; Figure S1D and S1E). Surprisingly, the manA mutant phenotype was displayed by cells grown in rich LB medium when mannose is not exploited as a carbon source. Nevertheless, ManA was found to be produced at significant levels under such conditions (Figure S2). Notably, the B. subtilis genome contains a homologue of manA, named pmi (56% identity), which encodes a second mannose phosphate isomerase. However, a strain bearing a knock out of pmi had no observable phenotype and unlike ManA, Pmi was undetectable in rich LB medium (Figure S2), suggesting that the two homologues have non-identical roles. Thus, besides its traditional role as a metabolic enzyme, ManA possesses an additional, previously unrevealed, crucial cellular activity that relies on its enzymatic activity.
The observation that ΔmanA cells display altered nucleoid morphologies suggested that this mutant is perturbed in organizing and segregating the chromosomes. To examine in more detail the nature of ΔmanA nucleoids, we visualized the replication origin region using GFP fused to Spo0J, a protein that binds near the origins [33]–[35]. Wild type cells producing Spo0J-GFP were observed to contain 2–4 fluorescent foci (Figure 2A and 2C), while the number of origins within the larger ΔmanA cells frequently exceeded 4 (Figure 2B). Figure 2D exemplifies a characteristic ΔmanA cell containing as many as 12 copies of origin. Quantification analysis of the number of Spo0J-GFP foci per cell showed that the wild type population exhibited a narrow distribution with more than 99% of the cells containing 1–4 foci, whereas the number of foci in the mutant cells varied widely from 1 to 12, with the majority of cells (55%) bearing more than 4 foci per cell (Figure 2G).
The increased number of origin foci observed within ΔmanA cells raised two possibilities: either the cells are polyploids containing multiple chromosomal copies, or alternatively these mutant cells erroneously reinitiate replication without the actual completion of chromosome synthesis. To distinguish between these possibilities, we visualized the terminus, which is the last chromosomal region to be replicated. We generated wild type and ΔmanA strains that harbor TetR–GFP and carry repeated tetO units inserted in proximity to the terminus region [36]. Consistent with the idea that ΔmanA cells are polyploids, the ΔmanA strain exhibited more TetR–GFP foci than the typical 1–2 foci observed in the wild type strain (Figure 2E and 2F). Nevertheless, to verify that the multiple copies of origin and terminus represent whole chromosomes, we applied DNA microarray to compare the DNA content of wild type and ΔmanA cells [37] (Materials and Methods). Equal amounts of genomic DNA extracted from wild type and mutant cells were labeled and hybridized to a B. subtilis DNA chip. This analysis indicated no significant difference between the two samples implying that amplification of specific DNA regions is not a feature of ΔmanA cells. Thus, ΔmanA cells most likely contain several copies of fully replicated chromosomes within each cell.
Further examination of the ΔmanA mutant strain revealed that the defects in cell shape and chromosome morphology are growth rate dependent. When growth was slowed, either by reducing the temperature or by growing ΔmanA cells in minimal medium, an almost normal phenotype was observed (Figure 3A; Figure S3). Accordingly, the levels of ManA production were significantly higher in rich medium (Figure S3). A prominent feature of rapid bacterial growth is the ability to perform multifork replication but still allocate the daughter chromosomes properly into progeny cells [38]. In light of the growth dependence of the ΔmanA phenotype, we speculated that ΔmanA cells are defective specifically in performing this complex task and as a consequence become polyploids. In order to test this premise, we investigated whether reducing the occurrence of multifork replication suppresses the ΔmanA phenotype. To that end, we combined the ΔmanA allele with a temperature sensitive allele of the replication initiation factor dnaB (dnaBts) [39], [40]. Incubating this strain at the restrictive temperature leads to inactivation of DnaB and therefore reduces the rate of multifork replication. In line with our expectation, when the ΔmanA dnaBts strain was grown at the restrictive temperature (42°C) the ΔmanA phenotype was partially suppressed, as manifested by the frequent appearance of almost wild type cell chains (Figure 3C). Importantly, these “wild-type like” cells were absent when the ΔmanA dnaBts strain was incubated at the permissive temperature (37°C), or when ΔmanA strain was incubated at the restrictive temperature (Figure 3B). Notably, by the time of transfer to the restrictive temperature the majority of the cells already acquired the ΔmanA phenotype and were polyploid, a phenotype which at some point may become irreversible. This occurrence may account for the partial nature of the suppression. To confirm that the appearance of “wild-type like” ΔmanA dnaBts cells correlates with reduced chromosomal copy number, we visualized the number of replication origins in the ΔmanA dnaBts strain by expressing Spo0J-GFP. Indeed, the number of Spo0J-GFP foci observed within the suppressed cells was similar to that in wild type cells (2–4 foci, Figure 3D and 3E). Thus, we conclude that ManA is crucial for proper DNA organization and segregation during rapid growth.
As mentioned above, in association with abnormal nucleoids, ΔmanA cells display morphological defects reminiscent of the phenotypes exhibited by cell wall mutants (e.g.: [6], [13], [15], [18]–[21], [29], [30]). This observation raised the idea that ManA is required for cell wall construction. To explore this possibility we visualized the cell wall architecture of ΔmanA cells. Wild type and ΔmanA cells were labeled with a fluorophore (fluorescein isothiocyanate - FITC) conjugated to wheat germ agglutinin (WGA), which is a carbohydrate binding protein that recognizes mainly GlcNAc. Living wild type cells stained with WGA-FITC exhibited bright midcell bands and fainter helical sidewall staining (Materials and Methods; Figure 4A). This non-uniform pattern is similar to the patterns reported when fixed B. subtilis cells are stained with WGA [9], [41] and when nascent PG is labeled with fluorescent vancomycin [5]. In contrast, living ΔmanA cells stained with WGA-FITC displayed a much more homogeneous pattern (Figure 4B). The ΔmanA cell wall appeared significantly thicker than the wild type cell wall and lacked the characteristic helical staining indicative of nascent PG. Examining the effect of ManA on the subcellular localization of the two cell wall components Mbl and TagO corroborated these findings. Mbl that is typically localized in a helical pattern [15] became dispersed, while the localization of the membrane-associated TagO [27] was hardly modified (Figure S4).
To substantiate that ManA is required for cell wall synthesis, we took advantage of the antibiotic tunicamycin to artificially interfere with cell wall construction and compare the resultant phenotypes with that of ΔmanA cells. Tunicamycin is a uridine nucleoside analog that specifically binds to and blocks the first membrane-associated step of both PG and WTA biosynthesis and thus, actively inhibits cell wall synthesis [42], [43]. When wild type B. subtilis cells were grown in the presence of tunicamycin and subsequently visualized by fluorescence microscopy, a dramatic change in their cell shape was readily detected (Figure S5). Surprisingly, the tunicamycin treated cells resembled the ΔmanA cells not only in their cell shape but also in their nucleoid morphologies (Figure 4D and 4E). Moreover, when wild type cells carrying the Spo0J-GFP fusion were treated with tunicamycin, a significant increase in the number of origins per cell was observed, indicating the transition to polyploidy (Figure 4F and 4G). Consistent with the finding that tunicamycin reproduces characteristic ΔmanA phenotypes, wild type cells treated with tunicamycin and labeled with WGA-FITC displayed an architecture similar to the one exhibited by ΔmanA cells (Figure 4C). Namely, the cell wall appeared thicker than usual and lacked typical sidewall cylindrical structures. Taken together, our data demonstrate that blocking cell wall synthesis per se by adding tunicamycin is sufficient to recapitulate both the cell shape and chromosome defects characteristic of ΔmanA cells. Therefore, we surmise that ManA plays a significant role in cell wall construction.
Since ManA is primarily classified as a sugar metabolic enzyme, we reasoned that it could affect the carbohydrate composition of the cell wall. To test this possibility, cell walls of wild type and ΔmanA cells were isolated, hydrolyzed and their glycosyl composition determined using HPAEC neutral monosaccharide analysis (Materials and Methods). The profile of wild type cell wall material was found to include high amounts of GlcNAc and Glc, medium levels of GalNAc and Gal, and relatively small quantities of Fucose (Fuc) (Figure 4H). These carbohydrates, though in different proportions, have been reported as cell wall constituents of pathogenic Bacilli species (i.e.: B. cereus, B. anthracis, and B. thuringiensis) [44], suggesting that these sugars are common to bacilli. The ΔmanA cell wall contained the same carbohydrates, however, a significant decrease in the amounts of Glc (∼4 fold) and GalNAc (∼5 fold) was monitored, with a milder decrease in the levels of Gal and Fuc (6-deoxy-L-Galactose); these four carbohydrates are characteristic components of WTA [1]. In contrast, the major PG sugar GlcNAc was found to be present at the same level in wild type and ΔmanA cells. Thus, ManA is required for proper formation of WTA and to a lesser degree for PG synthesis. Nevertheless, the modified architecture of the PG observed in the absence of ManA (Figure 4A and 4B) implies a tight coordination between WTA and PG construction.
To better understand the connection between cell wall integrity and chromosome morphology we followed both components simultaneously. We took advantage of the observation that ΔmanA cells exhibit almost normal phenotypes at a low temperature (23°C) but gradually evidence defects when shifted to a high temperature (37°C). Cell wall and nucleoid morphologies were followed by WGA-FITC labeling and DAPI staining, respectively, as the ΔmanA cells were temperature shifted. In wild type cells the chromosome appeared mostly helical, exhibiting morphologies typical of replicating forms (Figure 5A, 5C and 5E) [45]. Notably, the helicity of the chromosome in the majority of the cells seemed to follow the sidewall staining of the cell wall. In a way, it seems that the cell wall restricts the nucleoid spatial localization by caging it. This feature was highlighted when stacks of optical sections were deconvolved, and three-dimensional structure was reconstructed (Figure 5G and Video S1). In comparison, the morphology of the nucleoids in ΔmanA cells altered distinctively upon temperature shift. At the lower temperature the nucleoids appeared similar to those displayed by wild type cells (Figure 5B). However, after the temperature was shifted, concurrently with the cell bulging and loss of helical sidewall staining, the nucleoid lost its helical shape and became less compact and structured (Figure 5D and 5F). Moreover, as rods became spheres, the nucleoid expanded in a pattern associated with the new cell geometry, as if constraints confining its structure were being released (Figure 5H and Video S2). It is possible that the incapability to confine the chromosome leads to defects in DNA segregation, ultimately resulting in the formation of polyploid cells.
Consistent with a putative relationship between cell wall integrity and chromosome morphology, polyploidy was observed in L-form mutant cells, which completely lack a cell wall [46], [47]. Additionally, an examination of chromosome copy number in the cell wall mutant mreB in E. coli revealed multiple chromosomal copies per cell [21]. In accord, B. subtilis cells harboring mutations in mreB or mbl, or wild type cells treated with tunicamycin showed multiple chromosomal copies (Figure 4F and 4G; Figure S6). Thus, an intimate connection between cell wall integrity and chromosome morphology exists. Perturbing this connection gives rise to the formation of polyploid cells.
Cell wall is responsible for shape determination and cellular viability for most bacterial species. Here we demonstrate that the carbohydrate metabolic enzyme ManA, which is conserved among prokaryotes and eukaryotes, participates in cell wall construction of the Gram positive bacterium B. subtilis. Several lines of evidence demonstrate a direct connection between ManA and cell wall synthesis: 1) the shape defect and perturbed cell wall architecture exhibited by ΔmanA cells, 2) the resemblance between the phenotype of ΔmanA cells and that of wild type cells treated with the cell wall synthesis inhibitor tunicamycin and, 3) the decreased amount of specific carbohydrates coating the cell surface observed in the absence of ManA.
What could be the role of ManA in cell wall synthesis? ManA is a cytoplasmic enzyme that catalyzes the reversible isomerization of Fru-6-P and Man-6-P [32]. Both products could affect directly cell wall synthesis. Man-6-P is a substrate for generating GDP-mannose, which is an important precursor of many nucleotide sugars, such as GDP-fucose [48], whereas Fru-6-P is a component of a pathway that leads to the formation of the nucleotide sugar UDP-GlcNAc, basic for PG assembly [2]. Thus, the absence of ManA enzyme could interfere with the equilibrium of several pathways producing nucleotide sugar reservoirs for PG and WTA synthesis. Consistently, deleting the pgi gene encoding an enzyme that produces Fru-6-P [49] resulted in phenotypes similar to ΔmanA (Figure S7). This notwithstanding, according to our observations the absence of ManA has a specific effect on the composition of cell wall carbohydrates: it causes reduced abundance of Glc and GalNAc without perturbing the levels of the major PG precursor GlcNAc. Since, both Glc and GalNAc are components of the teichoic acid pathway [1], our data suggest that ManA is specifically involved in the WTA synthesis rather than in the PG pathway. Moreover, although the overall composition of the PG components was unaffected, visualization of the PG revealed a modified architecture implying that the balance between PG and WTA is crucial for proper cell wall construction.
Interestingly, in some cases mannose phosphate isomerases have been reported to affect cellular structures in other microorganisms. For example, the ManA enzyme of the bacterium Helicobacter pylori participates in capsular biosynthesis [48], while the eukaryotic fungus Aspergillus fumigatus displays altered cell wall synthesis and morphogenesis under mannose starvation conditions [50].
We have previously shown that the bacterial nucleoid adopts a helical morphology during DNA replication [45]. Here we demonstrate that this nucleoid architecture is strictly dependent on cell wall integrity (Figure 5; Video S1 and S2). Interestingly, it has been proposed that the bacterial chromosome structure is largely affected by the transcription of rRNA operons and the transcription-translation-insertion (transertion) of membrane proteins that fasten the chromosome to the membrane [51], [52]. Thus, it is conceivable that the chromosome is anchored to, and coordinated with both components cell wall and membrane. The association between DNA morphology and the cell wall is strongly manifested by the high chromosome copy number observed in L-form [46], [47], tunicamycin treated cells (this work), mreB and mbl mutants [21] (this work). We propose a model whereby the insertion of new cell wall material in a helical pattern dictates chromosome helicity (Figure 6A). In wild type cells the nucleoid is attached to the newly synthesized cell wall (see below) and its organization and segregation are coordinated with cell wall synthesis and elongation. In the absence of ManA the normal extension of the cell wall is blocked, as indicated by the disappearance of helical sidewall staining. Consequently, the nucleoid is detached from cell wall components and the synchronization is lost between cell growth and DNA replication and segregation, resulting in the formation of polyploid cells. When detached, the nucleoid loses its compact helical structure and adopts a looser conformation that seems to follow the overall cellular geometry.
It still remains to be resolved how the nucleoid is attached to the newly synthesized cell wall. Cell wall synthesis initiates on the cytoplasmic side of the membrane and then cell wall precursors are translocated to the outer membrane side [2]. This flip-flop reaction could be an active mechanism for coordinating DNA organization with cell wall elongation. We propose that the DNA is attached to the newly synthesized lipid intermediates on the cytoplasmic side of the membrane. When lipid II flips across the membrane the bound DNA is released, and the detached DNA is now free to bind a new cytoplasmic cell wall precursor (Figure 6B). These attaching/detaching cycles serve to coordinate DNA replication, segregation and cell wall formation. When cell wall construction is blocked the amount of new cytoplasmic cell wall precursors is reduced. Therefore the detached chromosome falls off the membrane and the linkage between cell wall and DNA is lost.
The interaction between the cell wall machinery and the DNA could be also mediated by linker membrane proteins that bind to both DNA and the cell wall. An interesting candidate proposed to possess such a capability is RodZ [53], a bacterial cell morphogenesis protein identified recently in E. coli [54], [55], C. crescentus and B. subtilis [56]. RodZ contains a transmembrane domain and a cytosolic helix-turn-helix (HTH) DNA-binding motif and was shown to form helical structures and associate with MreB [54]–[56]. An additional membrane protein reported to colocalize with MreB and affect nucleoid morphology is SetB in E. coli [57]. Importantly, the primary role of SetB is metabolic, as it was shown to act as a lactose and glucose efflux transporter [58]. It is possible that metabolic proteins such as ManA and SetB through their role in cell wall synthesis operate as sensors that synchronize cell wall elongation with metabolite availability. Indeed, a similar ‘coupling’ function between cell mass and cell division has been attributed to the metabolic protein UgtP, a sugar transferase in B. subtilis that acts both in WTA biosynthesis and inhibits assembly of the cell division protein FtsZ [59].
The similarity between cell wall biosynthesis in bacteria and protein N-linked glycosylation in eukaryotes, a process which is utilized to determine the rate of protein folding [60], [61], raises the possibility that cell wall elongation serves to time cellular activities. In eukaryotes, N-linked glycan chains are added in the endoplasmic reticulum to growing nascent polypeptides and promote proper protein folding [62]. In both N-linked glycosylation and cell wall biosynthesis the initial step is transfer of a sugar nucleotide to a lipid carrier, undecaprenyl phosphate in bacteria and dolichol phosphate in eukaryotes [61], reactions that are inhibited by the antibiotic tunicamycin [42], [63]. During N-linked glycosylation the sugar chain, containing mainly mannose, is transferred from the lipid carrier to the unfolded protein. This sugar tree then undergoes cycles of cleavage and synthesis that serve as a timer for reporting the state of protein folding [64]. In a similar manner, we propose that the cell wall elongation rate in bacteria serves as a timer that coordinates cell growth with critical cellular activities such as chromosome segregation and cell division.
Plasmid and primers used for this study are described in Text S1 and Table S1. B. subtilis strains were derivatives of the wild type strain PY79 [65] and are listed in Table S2. All general methods were carried out as described previously [66]. Cultures were inoculated at 0.05 OD600 from an overnight culture and growth was carried out at 23°C, 37°C or 42°C, in LB medium or in S7 minimal medium [67] as indicated.
Mini-Tn10 transposon was inserted into the B. subtilis (PY79) chromosome as described previously [68], [69]. Out of 2700 colonies that were screened, 260 that showed atypical colony morphologies (i.e.: smooth colonies, transparent colonies) and/or growth defect (i.e.: small colonies) were selected for further analysis.
Using UniProtKB/Swiss-Prot (http://www.uniprot.org/uniprot/O31646), the active site of ManA was determined largely by comparison with the crystal-solved ortholog Pmi from Candida albicans. Accordingly, two amino acids were chosen to be examined: histidine 97 which is located within a Zinc ion binding site, and the highly conserved arginine 192, which is predicted to be part of the catalytic domain.
Fluorescence microscopy was carried out as described previously [70]. Briefly, samples (0.5 ml) were taken during logarithmic phase, centrifuged and resuspended in 10 µl of PBS×1 supplemented with the fluorescent membrane stain FM1-43 or FM4-64 (Molecular Probes, Invitrogen) at 1 µg/ml and the DNA stain 4,6-Diamidino-2-phenylindole (DAPI) (Sigma) at 2 µg/ml. For cell wall labeling, cells were harvested, gently centrifuged, resuspended in 100 µl of T-Base×1 supplemented with WGA-FITC (5 µg/ml, Sigma), incubated for 15 min at room temperature, and washed twice with T-Base×1 before imaging. Stained cell walls and GFP fused proteins were visualized by placing the cells on thin T-Base×1 1% agarose pads. Cells were visualized and photographed using an Axioplan2 microscope (Zeiss) equipped with a CoolSnap HQ camera (Photometrics, Roper Scientific) or an Axioobserver Z1 microscope (Zeiss) equipped with a CoolSnap HQII camera (Photometrics, Roper Scientific). System control and image processing were performed using MetaMorph software (Molecular Devices). For deconvolution microscopy, samples of cells grown in rich LB medium were labeled with WGA-FITC (30 µg/ml) and DAPI (2 µg/ml), applied to an agarose pad, and then subjected to deconvolution microscopy. Optical sections (20–40) were collected at a spacing of 0.2 µm. Images were deconvolved through 50 iterations and then visualized as SFP volume render by using the Huygens Professional Software (Scientific Volume Imaging b.v., Netherlands).
Cell walls of wild type (PY79) and ΔmanA (ME37) cells (triplicates of 250 ml late logarithmic phase cultures) were isolated as described previously [44]. The analysis was performed by GlycoSolutions Corporation (MA, USA). Briefly, each of the lyophilized samples was resuspended in 250 µl sterile double distilled water, rinsed with 250 µl of 4 M trifluoroacetic acid. Of note, by using this method acidic carbohydrates such as MurNAc cannot be indentified. After 3 hours of hydrolysis samples were examined using GlycoSolutions SOP HPAEC Neutral Monosaccharide Analysis. Equal volumes of each sample were injected at various dilutions: undiluted, x10, x100. ManNAc and glucose standard curves were included to normalize the values.
DNA Microarray analysis was preformed to determine the DNA content of wild type (PY79) and ΔmanA (ME37). Digested genomic DNA (0.5 µg, HaeIII) of each strain was amplified by random primer and then labeled indirectly with cy3 or cy5 dyes. Equal amounts of the labeled samples were mixed and hybridized to a DNA chip containing Sigma B. subtilis OligoLibrary, which represents the entire B. subtilis ORFs. Arrays were scanned using a Genepix 4000B scanner (Axon Ltd). Fluorescence intensities were quantitatively analyzed using GenePix Pro 4.1 software (Axon).
|
10.1371/journal.pntd.0006628 | Nucleocapsid protein-based vaccine provides protection in mice against lethal Crimean-Congo hemorrhagic fever virus challenge | Crimean-Congo hemorrhagic fever (CCHF) is an acute, often fatal viral disease characterized by rapid onset of febrile symptoms followed by hemorrhagic manifestations. The etiologic agent, CCHF orthonairovirus (CCHFV), can infect several mammals in nature but only seems to cause clinical disease in humans. Over the past two decades there has been an increase in total number of CCHF case reports, including imported CCHF patients, and an expansion of CCHF endemic areas. Despite its increased public health burden there are currently no licensed vaccines or treatments to prevent CCHF. We here report the development and assessment of the protective efficacy of an adenovirus (Ad)-based vaccine expressing the nucleocapsid protein (N) of CCHFV (Ad-N) in a lethal immunocompromised mouse model of CCHF. The results show that Ad-N can protect mice from CCHF mortality and that this platform should be considered for future CCHFV vaccine strategies.
| Crimean-Congo hemorrhagic fever (CCHF) is a tick-borne disease that can manifest as a viral hemorrhagic fever syndrome. The CCHF virus is widely spread throughout the African continent, the Balkans, the Middle East, Southern Russia and Western Asia where it remains a serious public health concern. Currently, there are no licensed treatments or vaccines available, and medical countermeasures are urgently needed. We developed an adenovirus vector vaccine based on the conserved structural nucleoprotein (N) as the antigen. A prime-boost approach showed promising efficacy in the most widely used immunocompromised mouse model. This vaccine approach demonstrates a role for N in protection and suggests its consideration for future CCHFV vaccine strategies.
| Crimean-Congo hemorrhagic fever (CCHF) is an acute infectious disease with a wide geographic distribution and an average case fatality rate of approximately 20–30% [1, 2]. The etiological agent, CCHF orthonairovirus (CCHFV), belongs to the Orthonairovirus genus of the Nairoviridae family. The CCHFV genome consists of tri-segmented, negative-sense RNA referred to as the small (S), medium (M) and large (L) segments encoding the nucleocapsid protein (N), the glycoprotein precursor (GPC) and the viral RNA-dependent-RNA-polymerase (L), respectively [2, 3].
CCHFV is primarily maintained in and transmitted by ticks in the Hyalomma genus of the Ixodidae family [2]. The virus has a wide host range and causes a transient viremia in many wild, domesticated and laboratory mammals [1, 4, 5]. Humans usually acquire infection by tick bite or through unprotected contact with body fluids of infected animals or humans; additionally, several nosocomial outbreaks have been reported [1, 2]. In contrast to humans, adult immuno-competent mammals have not yet been reported to develop signs of disease [1, 2, 6]. This has impaired animal model development and hampered the testing of medical countermeasures against CCHF. CCHFV is an interferon-sensitive virus and its replication is highly reduced by treatment with interferon in interferon-signaling competent cells [7–9]. These observations led to the discovery that adult mice with gene knockouts in interferon signaling pathways, such as the signal transducer and activator of transcription-1 (STAT1-/-) and the interferon α/β receptor (IFNAR-/-) mouse strains, are highly susceptible to CCHFV infection mimicking some hallmarks of human disease [10–12].
CCHF is considered endemic in more than 30 countries throughout the African continent, the Balkans, the Middle East, Southern Russia and Western Asia [1, 4]. Over the past 20 years, CCHF has emerged or re-emerged in several countries often with dramatically increased case numbers. Furthermore, there has been a marked increase of imported cases of CCHF in European and South Asian countries [13–17]. The root causes of the increased CCHF incidence rates are not fully understood, but factors such as shifts in climate patterns, animal and human migratory patterns, population increase in livestock and wildlife accompanied by changes in agricultural practices and increased land use may be responsible [18, 19]. Thus, there is a pressing need to develop prophylactic and therapeutic countermeasures against this significant emerging zoonosis.
Over the past few years several vaccine candidates have been evaluated for protective efficacy in pre-clinical studies. Among those were vaccines based on DNA, viral subunits, whole inactivated virus, virus-like particle (VLP) and viral vectors such as modified vaccinia virus Ankara (MVA) [20–25]. Full protection against lethal CCHFV challenge in a mouse model was achieved with DNA plasmids expressing GPC subunits and N, as well as a MVA vector expressing the GPC open reading frame. Partial protection was observed with whole inactivated virus preparations, a DNA plasmid expressing GPC, VLPs (consisting of N, GPC and L), and a combined plasmid DNA (GPC subunits and N)/VLP vaccination approach [23–25]. No protection was achieved with a soluble glycoprotein subunit vaccine and MVA expressing CCHFV N. Overall, these studies implicate immune responses to the glycoproteins as most important for protection, however the glycoprotein subunit platform has not shown any protection despite inducing a significant antibody response [22]. Due to the high genetic diversity of the surface glycoproteins (69–99% amino acid identity between CCHFV strains), a broadly efficacious vaccine may need an additional, more genetically conserved CCHFV antigen such as N (91–99% amino acid identity between CCHFV strains) [26, 27].
Here we constructed and characterized an experimental CCHF vaccine vector based on human adenovirus type 5 (Ad) expressing the CCHFV N (Ad-N). Following infection with the recombinant Ad-N the CCHFV N protein was detected in cell lysates. Mice immunized with Ad-N developed an anti-N humoral immune response. A single dose of Ad-N resulted in 30% protection of IFNAR-/- mice against lethal CCHFV challenge. This could be further improved by a prime-boost regimen to 78% protection. These results indicate a significant role of N as a protective component of a CCHFV vaccine.
Animal experiments were approved by the Institutional Animal Care and Use Committee of the Rocky Mountain Laboratories (RML) and were performed following the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care, International (AAALAC) by certified staff in an AAALAC approved facility (#A4149-01). All procedures involving infectious CCHFV were performed in the RML Biosafety Level (BSL) 4 facility and all standard operating procedures (SOPs) including sample inactivation were approved by the Institutional Biosafety Committee (IBC).
CCHFV Strain IbAr 10200 (kindly provided by the University of Texas Medical Branch, Galveston, TX, USA; at that time Michael Holbrook), was propagated in Scott and White No. 13 (SW13) cells maintained in Leibovitz’s L-15 medium (both from ATCC) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 mM L-glutamine, and 50 U/mL penicillin, 50 μg/mL Streptomycin (Sigma-Aldrich) in an environment not enriched in CO2.
The adenovirus (Ad) expressing the complete open reading frame (ORF) of the N of CCHFV Strain IbAr 10200 (Ad-N) (NCBI Ref seq U88410.1 nucleotide 56–1504) or wild type Ad (Ad-wt) were constructed and rescued using the Adeno-X Adenoviral System 3 and titered using the Adeno-X Rapid Titer Kit according to manufacturer’s instructions (both from Clontech). Ad were propagated in 293 cells (ATCC) maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% FBS, 100 mM L-glutamine, and 50 U/mL penicillin, 50 μg/mL Streptomycin (obtained from Sigma-Aldrich).
293 cells were infected with Ad-wt and Ad-N with a multiplicity of infection (MOI) of ~5. Cell lysates were harvested (scraped into PBS-Tween 0.05% [PBST]) on day 2 post infection for immunoblot analysis. For SDS-PAGE, samples were mixed with 2× SDS-PAGE loading buffer (1:1, v/v), boiled for 10 minutes, centrifuged and applied to 10% SDS-PAGE. Proteins were blotted onto a nylon membrane (GE healthcare) followed by blocking overnight in 5% milk-PBST solution at 4 C°. The membrane was incubated with a rabbit anti-N (N1028) peptide polyclonal serum (Thermo Fisher Scientific Inc.) [12] followed by a goat anti-rabbit HRP conjugated antiserum. Detection was performed using Pierce ECL Plus Western Blotting Substrate (both from Thermo Fisher Scientific Inc.) according to the manufacturer's protocol.
IFNAR-/- mice (6 to 12-week-old; on C57BL/6 background) were obtained from an in-house breeding colony. Groups of IFNAR-/- mice (n = ranging from 3–18 animals per group) were vaccinated with 1.25×107 infectious units (IFU) of Ad-wt or Ad-N by the intramuscular route delivered into the hind-leg musculature (day -28 for prime-only and day -56 for prime/boost group; 50 μL total volume). IFNAR-/- mice in prime/boost experiments were boosted 4 weeks (day -28) post vaccination with 108 IFU of the homologous Ad construct by the intranasal route (25 μL per nostril). Mice were boosted by the intranasal route as this vaccine regimen has previously been shown to bypass pre-existing Ad vector immunity and stimulate a more robust immune response against the Ad encoded antigens [28]. Vaccinated mice were acclimatized to the BSL4 environment for 5–7 days prior to CCHFV challenge. Four weeks after vaccination or boost (day 0) IFNAR-/- mice were challenged with 1000 LD50 (50 TCID50) of CCHFV by the subcutaneous route delivered to the intrascapular region (50 μL total volume). Animals were monitored daily for clinical signs with group weights being recorded. On day 3 post infection with CCHFV, 9 mice from the distinct groups were exsanguinated by cardiac puncture under anesthesia with whole blood collected into EDTA tubes (BD Biosciences) for RNA extraction and frozen at −80°C for virus isolation. Liver and spleen specimens were collected for pathological evaluation or immediately frozen at -80°C for virus isolation. The remaining mice (n = 3, 6 or 9) per vaccine group were monitored daily for 30 days post infection.
For antibody (IgG) detection we used an ELISA based on whole CCHFV particle antigen. Supernatants from CCHFV infected (positive antigen) and mock-infected (negative antigen) SW13 cells were harvested, cleared from cell debris through low-speed centrifugation, diluted 1:200 in 0.05% PBST and treated by gamma-irradiation (8 MRads). Maxisorp plates (96-well; Thermo Fisher Scientific Inc.) were coated with positive and negative antigen in 5% skim milk in PBST overnight at 4°C. Mouse serum was serially diluted two-fold starting at a dilution of 1:50, added to the plates and incubated for 1 hour at room temperature. Detection was performed using a goat anti-mouse peroxidase-conjugated IgG (KPL) at a 1:1000 dilution followed by treatment with ABTS peroxidase substrate system (KPL) as per manufacturer’s instructions. The cut-off was set at >3 standard deviations above the reading of negative samples. The data are reported as inverse dilutions.
CCHFV S segment specific quantitative RT-PCR and tissue titration were carried out as previously described [12].
Tissue samples were treated and fixed in 10% formalin according to approved standard operating procedure. Fixed samples were processed and either stained with hematoxylin and eosin (H&E) or N1028 polyclonal antiserum for histopathology or IHC, respectively, as described previously [12]. Slides were examined by a veterinary pathologist and scored as follows: 0 = no obvious pathological changes; 1 = minimal increase in the number of inflammatory cells and hepatocellular necrosis; 2 = mildly increased numbers of inflammatory cells, hepatocellular necrosis or lymphocytolysis; 3 = moderately increased numbers of inflammatory cells, and hepatocellular necrosis or lymphocytolysis; and 4 = highly increased numbers of inflammatory cells and multifocal hepatocellular necrosis or lymphocytolysis.
All physiological parameters were compared and analyzed using one-way or two-way analysis of variance (ANOVA) with Dunnet’s posttest on GraphPad Prism v5.00 (GraphPad Software).
Confluent 293 cells were infected with Ad-wt and Ad-N at a MOI of 5. Expression of CCHFV N was verified only in cell lysates from Ad-N-infected cells by immunoblot using an N-specific antiserum. (S1A Fig). To determine the immunogenicity of CCHFV N in an immunocompromised host, IFNAR-/- mice (n = 3) were vaccinated intramuscularly with Ad-N (1.25×107 IFU) followed by an intranasal boost (108 IFU) four weeks later; seroconversion was assessed by IgG ELISA. All three mice developed detectable IgG antibodies responses to CCHFV N with titers ≥1:6400 (S1B Fig).
IFNAR-/- mice were vaccinated intramuscularly with the recombinant adenoviruses (1.25×107 IFU) 28 days before CCHFV challenge (1000 LD50). Mice vaccinated with Ad-wt (n = 6) rapidly lost weight and succumbed to infection within 6 days (Fig 1A and 1B). Vaccination with Ad-N (n = 9) resulted in partial protection against lethal CCHFV challenge (33% survival, p>0.05) with reduced clinical signs (i.e. weight loss) and increased survival times (8.5 days vs 5 days survival, Ad-N vs Ad vaccinated, p<0.001) in those mice that succumbed to infection (Fig 1A and 1B).
A prime-boost strategy was employed next with the prime being administered intramuscularly and the boost intranasally, a strategy that has resulted in enhanced protective efficacy before [28]. For this, IFNAR-/- mice were immunized at day -56 (1.25×107 IFU) and boosted on day -28 (108 IFU) with recombinant adenoviruses, challenged on day 0 with CCHFV (1000 LD50) and monitored for survival. As with the single-dose vaccination, Ad-wt vaccinated mice (n = 3) rapidly lost weight and succumbed to infection by day 6 post infection (Fig 1C and 1D). Ad-N vaccinated animals (n = 9) showed increased protection from lethal CCHFV challenge (78% survival, p<0.0001) with reduced clinical signs including weight loss and increased survival times in those mice that succumbed to infection (Fig 1C and 1D).
At day 3 post CCHFV challenge, mice (n = 9) were euthanized and their liver, spleen and blood were sampled for virus load titrations and histopathology. In contrast to Ad-wt vaccinated animals, IFNAR-/- mice vaccinated with Ad-N either by a single-dose or a prime-boost approach did not show detectable viremia as analyzed by quantitative RT-PCR and virus infectivity assay (Fig 2A and 2B). Notably, liver and spleen viral loads on day 3 post infection of animals vaccinated with a single dose of Ad-N were similar to those detected in Ad-wt vaccinated animals (Fig 2C and 2D). In contrast, liver and spleen viral loads were significantly reduced following prime-boost vaccination with Ad-N. This observation was most prominent in the spleen for yet unknown reasons (Fig 2C and 2D).
Ad-wt vaccinated mice developed multifocal to coalescing hepatocellular necrosis with infiltration of viable and degenerate neutrophils in the liver (Fig 3A), and the spleens demonstrated mild to marked acute necrotizing splenitis with loss of lymphocytes (S2A Fig) as shown previously for CCHFV-infected IFNAR-/- mice [12]. IFNAR-/- mice vaccinated with a single dose of Ad-N had less severe hepatic lesions compared to Ad-wt vaccinated animals consisting of mild focal necrosis and infiltration of small numbers of viable and degenerate neutrophils (Fig 3B). There were no splenic lesions detectable (S2B Fig). This result was improved by prime-boost vaccination which resulted in further reduction of hepatic lesions (Fig 3C) and absence of splenic lesions (S2C Fig). IHC demonstrated high amounts of CCHFV N antigen in the liver and spleen of those animals immunized with Ad-wt vector (Fig 3D; S2D Fig), whereas mice immunized with Ad-N showed strongly reduced numbers of CCHFV antigen-positive cells in liver and spleen with the lowest numbers for those animals vaccinated with the prime-boost regimen (Fig 3E and 3F; S2E and S2F Fig). Antigen-positivity was scattered throughout the liver and spleen and was associated with cells morphologically consistent with hepatocytes, Kupffer cells, macrophages and endothelial cells, as previously reported [12].
Due to high case fatality rates, potential of human-to-human transmission, increasing likelihood of imported cases and expanding endemic region, CCHFV is a serious threat to public health. This has been recognized by the World Health Organization, which added CCHFV to the list of priority pathogens (http://www.who.int/blueprint/priority-diseases/en/). Therefore, development and evaluation of countermeasures, especially vaccines, is of critical importance in mitigating the detrimental impact of CCHF.
The only CCHF disease models are certain immunocompromised and/or humanized mouse strains [2]. The mouse strains include the interferon signaling deficient IFNAR-/- and STAT1-/- mouse strains and the humanized Hu-NSG-SGM3 mice [10–12, 29]. The Hu-NSG-SGM3 express certain human cytokines and human leukocytes, however the disease progression is somewhat atypical as the disease is primarily neurological [29]. Both the STAT1-/- and IFNAR-/- mice possess complete sets of murine immune systems, but their cells either have altered response to interferon signaling or do not respond to type I interferon signaling, respectively, leading to rapid disease more reminiscent of human hemorrhagic fever [10–12]. IFNAR-/- mice are highly susceptible to severe disease caused by several viral agents [11, 12, 30–31] due to lower and/or altered immune responses to infections compared with wild type mice [32–35]. The IFNAR-/- mice frequently do not respond as quickly to infection as wild type mice and, furthermore, CCHFV replicates to higher levels than in wild type mice [12]. The increased difficulty in protecting IFNAR-/- mice from infection seems associated with insufficient immune responses due to inadequate cross-priming of antigen presenting cells [32]. Thus, efficacy testing of vaccines in this model can be difficult as protective vaccines must elicit proper immune responses by circumventing IFNAR-/- dampened antigen priming, and must stimulate an effective adaptive immune response that can compensate for the lack of antiviral state activation and reduction of CCHFV replication normally mediated by type I interferon signaling. Therefore, IFNAR-/- mice should be considered a “higher bar” for efficacy testing of vaccines than fully immunocompetent rodent models.
CCHFV vaccine development is further complicated by limited information on both B- and T-cell epitope requirements for the development of an effective immune response; and the type(s) of immune responses necessary for protection from disease. The MVA, plasmid DNA and VLP vaccine platforms have had success in protecting mice from lethal CCHFV challenge, and the protection afforded was dependent on both humoral and cell-mediated immunity [20, 21, 24, 25, 36]. In addition, all the platforms strongly suggests that immune responses directed against the GPC, whether antibody or T cell driven, are essential for protection in rodents [20–25, 36]. While vaccination regimens that focus on a single antigen have been successful, an ideal vaccination candidate would facilitate immune response against multiple antigens and achieve protection with as few doses as possible.
Ad vectors are known to elicit strong humoral and cell mediated immune responses in IFNAR-/- mice [37]. Therefore, an Ad-based platform was utilized here to address whether IFNAR-/- mice can be protected from lethal CCHFV infection using only the more conserved CCHFV N as the antigen. In this study, IFNAR-/- mice vaccinated with Ad-N developed IgG responses to CCHFV N (S1 Fig) and were partially protected from CCHFV challenge (Figs 1–3; S2 Fig). Due to a lack of GPC antigen in the vaccine preparations protection is unlikely to be mediated by neutralizing antibodies but rather due to priming of CD4+ and CD8+ T-cells and/or non-neutralizing antibody responses as has been previously reported for the Ad platform in other IFNAR-/- vaccine studies [38]. Additional experiments are required to demonstrate which of the CD4+, CD8+ T cell and/or antibody responses after vaccination/infection are responsible for protection. Several studies have evaluated correlates of protection in IFNAR-/- mice to CCHFV infection. The role of adaptive immune responses following MVA immunization was evaluated using the transfer of antibodies and/or T-cells to naïve IFNAR-/- mice [36], while vaccine studies employing DNA and VLP vaccination were evaluated using circulating cytokine profiling [24] and/or IgG subtype ratios [24, 25]. These or alternate experiments, such as selective depletion of B- and T-cell populations following vaccination, could also determine the importance of each of the two arms of the adaptive immune system in response to Ad vaccination. However, it is important to caution that despite their utility and appropriateness for early stage vaccine testing, IFNAR-/- mice are severely immuno-compromised, inbred mice and therefore may generate different protective immune responses than immuno-competent outbred populations of humans and/or animals [12].
While DNA and MVA platforms have demonstrated the protective potential of GPC, vaccine studies utilizing N as the sole antigen are rare and have been non-protective [20, 21, 24, 25]. Thus, protection by Ad-N as demonstrated here provides critical information for future development of CCHF vaccines and suggests that Ad-N protects by mechanisms distinct, and perhaps complimentary, to that of the MVA and DNA platforms. The identification of a protective Ad-based vaccine could therefore be critical in developing an effective CCHFV combination vaccine regimen, in addition to uncovering other immune targets and mechanisms of protection.
Due to the zoonotic nature of CCHF and the limited potential for person-to-person transmission, mass-scale vaccinations are unlikely to be used in case of CCHF. However, as has been recently reported for Ebola, having an efficacious vaccine on hand for emergencies can help to contain outbreaks and protect medical staff attending patients [39]. To this end having readily available vaccine stockpiles which can be efficiently produced, have an acceptable shelf-life, can be efficiently delivered to patients and clinical staff in outbreak regions, and have undergone at least clinical phase II testing would be of great importance against sporadic, life-threatening, emerging infectious diseases such as CCHF. The Ad platform fulfils these criteria and should, in addition to other vaccines, be considered for vaccine stockpiles.
In conclusion, here we report partial protective efficacy of an Ad-based vaccine vector expressing the CCHFV N (Ad-N) against lethal CCHFV challenge in a highly susceptible, immunocompromised mouse model. Partial efficacy of up to 78% following a prime-boost vaccination strategy mediated by this more conserved CCHFV antigen demonstrates its critical role in protection and suggest its future consideration for CCHFV vaccine strategies.
|
10.1371/journal.pcbi.1000939 | Molecular Dynamics Simulations Suggest that Electrostatic Funnel Directs Binding of Tamiflu to Influenza N1 Neuraminidases | Oseltamivir (Tamiflu) is currently the frontline antiviral drug employed to fight the flu virus in infected individuals by inhibiting neuraminidase, a flu protein responsible for the release of newly synthesized virions. However, oseltamivir resistance has become a critical problem due to rapid mutation of the flu virus. Unfortunately, how mutations actually confer drug resistance is not well understood. In this study, we employ molecular dynamics (MD) and steered molecular dynamics (SMD) simulations, as well as graphics processing unit (GPU)-accelerated electrostatic mapping, to uncover the mechanism behind point mutation induced oseltamivir-resistance in both H5N1 “avian” and H1N1pdm “swine” flu N1-subtype neuraminidases. The simulations reveal an electrostatic binding funnel that plays a key role in directing oseltamivir into and out of its binding site on N1 neuraminidase. The binding pathway for oseltamivir suggests how mutations disrupt drug binding and how new drugs may circumvent the resistance mechanisms.
| Oseltamivir (Tamiflu) is the main antiviral drug used to fight viral influenza outbreaks such as the recent swine flu (H1N1pdm) global pandemic and avian (H5N1) outbreak in Asia. Oseltamivir inhibits a protein on the surface of flu viruses called neuraminidase, which is responsible for releasing newly formed viruses. The rapid emergence of drug resistance in H5N1 avian flu (and recently the H1N1pdm strain) has already motivated numerous studies to understand how mutations render oseltamivir ineffective, but no focused investigation has yet elucidated the specific mechanism behind mutation-induced drug resistance. Here, large scale computer simulations are employed to study both H5N1 and H1N1pdm neuraminidase to answer the questions: how does N1-subtype neuraminidase bind oseltamivir, and how would mutations alter this process? The key finding revealed in our simulations is the discovery of oseltamivir binding to neuraminidase by a charged pathway on the protein surface. We suggest that point mutations may disrupt drug binding by interfering with this pathway. Our results explain the fundamental mechanism behind oseltamivir resistance and pave the way for the design of drugs that circumvent viral drug resistance.
| Oseltamivir, better known by its commercial name Tamiflu, is currently the most important antiviral drug employed to combat the flu virus [1]. Oseltamivir functions by competitively binding, against a natural substrate on cells called sialic acid (SA), to a flu protein called neuraminidase N1 subtype, which is responsible for mediating the release of newly synthesized virion particles from an infected cell [2]. Of grave concern however, is the emergence of oseltamivir-resistant strains of N1-subtype influenza (including H5N1 [3], seasonal H1N1 [4], and H1N1pdm [5]–[7]). Understanding the mechanism behind mutation-induced drug resistance in neuraminidase N1 subtype is critical for the development of effective therapies.
The rapid emergence of oseltamivir resistance in H5N1 avian flu has motivated already numerous studies, both experimental and theoretical, to uncover how point mutations to neuraminidase alter drug binding [8]–[12]. The recent elucidation of crystal structures for both wildtype and mutant H5N1 neuraminidases have opened up a door for an investigation of drug resistance mechanisms and structure-based drug design at the atomic level [12], [13]. From these structures it has been suggested that oseltamivir resistance due to point mutations arise from a destabilization of the hydrophobic packing that binds oseltamivir tightly within the neuraminidase active site [12]. Crystal structures represent a time frozen snapshot into a possible conformation of drug-protein interaction. Drug binding, however, is a dynamic process and computational studies used the crystal structures as starting points to shed light on exactly how protein flexibility and point mutations influence drug-protein endpoint interactions [8]–[11]. Despite these initial inroads of studies based on molecular dynamics (MD) simulations, the current understanding of the mechanism behind drug resistance remains incomplete and some conclusions are conflicting. For example, in one study [10] it was reported that the H274Y mutation disrupts E276-R224 salt bridges that accommodate the hydrophobic pentyl group of oseltamivir, while in another study [8] the same salt bridges were observed to be stable. Up to this point, all proposed mechanisms for oseltamivir resistance have focused mainly on effects of mutations on the SA binding site and equilibrium drug binding affinities.
Since neither H274Y nor N294S are active site mutations [14], studies which only focus on end-point interactions between drug and protein are unable to elucidate if these mutations impact the actual drug binding process, i.e., affect binding kinetics. In this study, we not only employ molecular dynamics (MD) on oseltamivir-bound forms of wildtype and known drug resistant mutants (H274Y, N294S) of avian H5N1 and swine H1N1pdm neuraminidases, but also steered MD (SMD) simulations on avian H5N1 to investigate oseltamivir binding/unbinding pathways. The electrostatic potentials calculated reveal a distinct negatively charged column of residues, bridging the SA binding site and the edge of the binding cavity mouth, that apparently functions as a drug binding/unbinding funnel. Oseltamivir is observed to diffuse, in our simulations, into and out of the neuraminidase active site via this funnel. During drug passage, our simulations did not reveal any specific interactions beyond an obvious electrostatic attraction between drug and protein, suggesting a binding pathway governed by an electrostatic funnel. We also suggest a role that the drug resistant mutations H274Y and N294S play in disrupting this funnel and in altering the binding process between oseltamivir and N1-subtype neuraminidase. Given that the drug binding rates have been found to be significantly diminished for H5N1 mutants [12], it is possible that the stated mutations (H274Y, N294S) induce drug resistance by disrupting the drug binding kinetics as well as by active site endpoint interactions affecting binding affinity. We suggest that our observations regarding the oseltamivir binding behavior of H5N1 should apply for H1N1pdm (associated with the recent pandemic), since the two proteins have very high sequence identity (91.47%) and share a conserved drug binding site.
The following results are based on simulations, summarized in Table 1 (see Methods section), of drug bound N1-subtype neuraminidases, including avian H5N1 and H1N1pdm, both as wildtype and as two oseltamivir-resistant mutant systems. The individual simulations will be referred to by the designations listed in the “Name” column of Table 1. We first describe interactions that stabilize oseltamivir in six neuraminidase systems observed from equilibrium (EQ) simulations (simEQ1–6). Then we present our observations of electrostatics calculated from these simulations. Next, we discuss characteristics of drug binding and unbinding seen in both steered molecular dynamics (SMD) (simSMD1) and subsequent relaxation simulations (simFEQ1–10). Finally we relate the results of our simulations to oseltamivir resistance.
Equilibrium simulations of six oseltamivir-neuraminidase complexes were carried out, including avian H5N1 and swine flu H1N1pdm proteins, for one wildtype (WT) neuraminidase (simEQ1–2) and two mutants, namely, H274Y (simEQ3–4) and N294S (simEQ5–6). The root mean square deviations (RMSD) of the proteins shown in Figure S1 demonstrate the stability of the simulated models and the RMSDs of oseltamivir in Figure S2 and Figure S3 show that the drug, over 20 ns, binds stably to the SA binding pockets for both WT and mutants (figures and text labeled “S” are in Supplementary Materials).
SimEQ1–6 reveal that hydrogen bonds form the bulk of the interactions which stabilize oseltamivir in the SA binding pocket (see Text S1 for a discussion of protein stability and Text S2 for detailed discussion of hydrogen bond networks). In both wildtype and mutant simulations, hydrogen bonds are well conserved between oseltamivir and binding site residues E119, D151, R292, and R371. Specifically, R292 and R371 are observed to hydrogen bond with oseltamivir's carboxylate moiety, and E119 and D151 with oseltamivir's amino group. The H274Y mutation, however, is seen to disrupt the hydrogen bonding of oseltamivir's acetyl group with R152, a stabilizing interaction in the wildtype and N294S systems. The carboxylate group of oseltamivir exhibits a weak hydrogen bond with Y347 of the avian H5N1 neuraminidases (Figure S4), which corresponds to the N347 mutation in the H1N1pdm proteins (Figure S5). Histograms listing the frequency of hydrogen bonds between oseltamivir and neuraminidases are shown in Figure S4A and Figure S5A, with schematic views of the specific residues involved in the drug-protein hydrogen bond pairings provided in Figure S4B, Figure S4C, and Figure S4D for simEQ1, simEQ3, and simEQ5, respectively, and in Figure S5B, Figure S5C, and Figure S5D for simEQ2, simEQ4, and simEQ6, respectively. In the case of the H274Y mutant, we observed a loss of the hydrogen bond between oseltamivir and R152, as well as a decrease in favorable hydrophobic packing between oseltamivir's pentyl group and neuraminidase's hydrophobic subsite (I222-R224-A246-E276). Figure S6, illustrating an increase in solvent accessible surface area (SASA) for the H274Y systems, is discussed in Text S3, along with further description of oseltamivir's hydrogen bonding network observed in our simulations as compared to previous studies [8], [12]. While the H274Y mutant clearly showed decreased binding stability for oseltamivir's pentyl group within the SA binding pocket, disruption to endpoint interactions alone within simulation timescales may not cover all possible effects of the two non-active site mutations H274Y and N294S, on oseltamivir binding. Therefore we turn our attention to other physical characteristics which may also govern drug binding kinetics.
The electrostatic potential of neuraminidases serves as an important driving force both to direct the diffusion of ligands into the SA active site [15], [16] and to stabilize the end point interactions between ligand and the proteins [17], [18]. For example, computational studies investigating the quantitive free binding energy associated with neuraminidase inhibitor binding reveal that the local electrostatic potential of the drug binding pocket significantly impacts the final binding pose and stability of the drug [18]. In regard to neuraminidase, a previous study [16] using Brownian dynamics simulations suggested that electrostatic steering guides SA and neuraminidase inhibitors to enter the primary SA active site of N1 neuraminidase via a secondary SA binding site proximal to loop-430 [15]. A second, well known example in which a cationic substrate is drawn through a short channel to enter a narrow active site due to electrostatic steering is seen in the case of acetylcholinesterase [19], [20]. Even though a possible role of the electrostatic potential for drug binding in neuraminidases had been discussed before, the extensive electrostatic calculations required for fully characterizing the role of electrostatic steering had not yet been carried out.
In order to address the shortcoming, the electrostatic surface potentials of the equilibrated systems were calculated and averaged across every trajectory frame in simEQ1–6 employing GPU-accelerated multilevel summation [21] (see Methods). The resulting electrostatic maps are shown in Figure 1A for H5N1 and in Figure 1B for H1N1pdm. The maps reveal that the binding pocket possesses a negative potential (colored red), and that it is surrounded by a positive potential ring (colored blue). The electrostatic surface potentials of the neuraminidases simulated show, in particular, a column of negatively charged residues that form a pathway, 10Å in length, between the primary SA binding site and the mouth of the binding cavity. Since oseltamivir itself has a positive electrostatic surface potential, as illustrated in Figure 1C, the question arises whether the negatively charged surface column in N1 neuraminidases plays a role in the binding kinetics of oseltamivir. To answer this question, and following insight from earlier studies of binding processes [22]–[25], we employed simulations (described in Methods) to pull oseltamivir out of the SA binding site and probed unbinding in order to reveal actually the properties of the binding pathway.
To simulate binding of oseltamivir would be the most natural approach to identify the drug binding pathway. Unfortunately, the needed computations are impossible since the duration of binding is too long. Unbinding enforced through external forces in so-called steered MD (SMD) simulations requires much less time such that the calculations are feasible. Fortunately, unbinding simulations can reveal features characteristic for the reverse process of binding, as demonstrated many times before, e.g., in [22]–[25]. We note, however, that while SMD simulations may reveal potential entry and exit pathways for drug binding, they do not provide information regarding the thermodynamic feasibility of binding (which correlates to drug inhibition power) except when used with sampling methodologies [25]–[27] not applied in the present case.
In simSMD1, a pulling force was applied to rupture all stabilizing hydrogen bonds between H5N1 and oseltamivir, and draw the drug away from the SA binding site. The results of simSMD1 show that the response of oseltamivir to the pulling force evolves in three distinct stages: 1) from 0 to 8 ns, a buildup of force during which hydrogen bonds between oseltamivir with E119, D151 and R152 are ruptured; 2) at 8 ns when the remaining stable hydrogen bonds with R292 and R371 rupture; 3) after 8 ns when the drug is pulled out of the binding pocket. Figure 2 shows the force dependent rupture of hydrogen bonds in simSMD1 presenting both hydrogen bond length and (in the inset) the force vs. time curve. We carried out simulations at different pulling velocities (simSMD2–3) that all exhibited similar unbinding behavior.
Despite application of force straight out of the binding pocket, oseltamivir surprisingly did not unbind along the direction of the force, following, instead, a lateral unbinding path. This path is characterized through strong interaction of the drug with the negatively charged column of residues identified in the electrostatic potential seen in Figure 1. The path taken by the drug is shown in Figure 3A–D, and in Video S1 (all videos are provided in Supplementary Materials). Tracing the relative position of oseltamivir and all its possible hydrogen bonding pairs with residues located along this pathway, it was recognized that nonspecific electrostatic attractions are the predominant interactions between drug and protein. The lateral pathway taken by the unbinding drug suggests strongly that it functions as a binding funnel which directs oseltamivir into the SA binding pocket.
A key observation from simSMD1 is that oseltamivir undergoes a rotation within the SA binding pocket before unbinding. This rotation, clearly discernible in Video S1, is the result of the rupture of hydrogen bonds between oseltamivir and residues E119, D151, and D152, while hydrogen bonds between oseltamivir's carboxyl functional group and residues R292 and R371 remain intact. The rotation, then, appears crucial for orienting oseltamivir into a position which permits it to more easily dissociate from the SA binding pocket. Oseltamivir in H5N1 is shown in its bound state before and after rotation in Figures 4A1–2 and 4B1–2, respectively. A comparison of the relative orientation of oseltamivir between the two states is provided in Figure 4A3 and 4B3, the latter showing the rotated state.
SMD simulations are capable of capturing drug unbinding by accelerating the event through an applied force. It is desirable to verify SMD results through simulations without applied force. It was observed in simSMD1 that following the transition to the rotated state, a much lower applied force is required to subsequently draw oseltamivir out of its binding pocket. This observation suggests that one may be able to probe the unbinding pathway without applied force, if oseltamivir is already in its rotated state. We performed, therefore, ten additional equilibrium simulations (simFEQ1–10) beginning with oseltamivir already in this state.
From simulations simFEQ1–10, two distinct outcomes were observed, namely the escape of oseltamivir from the SA binding pocket through favorable interactions with the charged binding funnel and a return of oseltamivir, not unexpectedly, to its pre-rotation bound state. Each simulation was carried out with sufficient duration to observe either outcome, with the exception of simFEQ5. In simFEQ5, oseltamivir, after following the binding funnel to escape the protein, actually rebound to the SA binding pocket through the same binding funnel, the dramatic return being captured in Video S6. A summary of observed outcomes from these simulations is shown in Table 2.
Oseltamivir freely, i.e., without external force applied, diffused out of the SA binding pocket by following the electrostatically charged binding funnel (described above) in five out of ten simulations (simFEQ1–5). In four cases (simFEQ1–3, 5) oseltamivir diffused along the full length of the binding funnel before separating from neuraminidase. Snapshots from a representative simulation (in this case simFEQ1) illustrating the trajectory that oseltamivir follows along our proposed binding funnel are shown in Figure 5A. This trajectory and those from simFEQ2,3, and 5 can be inspected in Video S2, Video S3, Video S4, and Video S6. In simFEQ4, oseltamivir was observed to briefly interact with the binding funnel, but dissociated from neuraminidase through an alternate path, namely, through interaction with the “430-cavity” identified in earlier computational studies [15], [28]. This trajectory is shown in Video S5. Our simulation revealed that hydrogen bonds between oseltamivir's carboxylate group and the guanidino group of R430 appear to stabilize the transition of the drug along this alternate pathway. The 430-cavity is believed to function as a secondary binding site for SA, and our simulation (namely, simFEQ4) shows that it may (not surprisingly) also serve as a viable conduit for the binding/unbinding of oseltamivir. Snapshots from simFEQ4 illustrating the trajectory of oseltamivir through the 430-cavity are shown in Figure 5B, with 430-cavity-specific residues colored in green.
In simFEQ1–4, once oseltamivir separated from neuraminidase, the drug diffused into the surrounding solvent environment and away from the protein. However in simFEQ5, we observed not only a diffusion of oseltamivir through the charged binding funnel, but also the reentry of the drug through the same pathway after it had diffused already away from neuraminidase. Specifically, the sequence we observed in simFEQ5 was: 1) between 0 to 25 ns, oseltamivir diffused out of neuraminidase's SA binding pocket by following the charged binding funnel, 2) between 25 and 35 ns, the drug unsuccessfully attempted to rebind from an unsuitable direction through hydrogen bond interactions with R152 in the so-called flexible 150-loop [29]; 3) between 35 and 45 ns, the drug again diffused away from neuraminidase; 4) between 45 and 50 ns, the drug approached the binding funnel again, drawing itself back into the funnel and binding at the SA binding pocket for 5) at least the next 50 ns (100ns total simulation time). Snapshots from these events are shown in Figure 6. Analysis of interactions of the newly rebound oseltamivir with binding pocket residues during the 50 to 100 ns interval revealed that the drug was stabilized by hydrogen bonds with Y406, R292, D151, E119, and R118, even though oseltamivir's pentyl group had not yet moved to its requisite hydrophobic pocket (I222-R224-A246-E276) [30]. The full simFEQ5 trajectory is provided in Video S6, illustrating the strongest evidence observed thus far that the electrostatic funnel serves a crucial role in binding of oseltamivir. We note that while simSMD1–3 and simFEQ1–10 simulated the H5N1 systems, the results should also apply for the H1N1pdm systems which share a highly conserved drug active site and a very high overall sequence homology.
Our study has shed light on the important role of the electrostatic surface potentials in directing the diffusion of oseltamivir into the SA binding site of neuraminidase. The simulations yield strong evidence that the negatively charged funnel identified in this study serves as an unbinding pathway for oseltamivir in the H5N1 and, due to sequence and structural identity, also H1N1pdm wild type systems. The presence of a binding funnel raises an obvious question: would it be possible for the drug resistance mutations, in addition to their effects on destabilizing the hydrophobic packing of oseltamivir, to disrupt or otherwise alter this binding funnel? The conspicuous location of residue 294, which maps directly onto this negatively charged pathway, may play a key role in the N294S mutation for disrupting the proper guidance of the drug into its binding pocket. The H274Y mutation may also have a similar effect on drug binding, besides disrupting a hydrogen bond with R152. In all four simulations of the mutants (simEQ3–6), the positions of residues 274 (shown in Figures 7A and 7C) and 294 (shown in Figures 7B and 7D) lie directly on the charged binding funnel. An allosteric contribution due to the flexibility of the channel or that of the drug cannot be ruled out, i.e., the presence of the drug within the binding funnel may induce local conformational changes that bring the mutant residues into play, in which case electrostatic interactions may also be altered.
While active site interactions, including hydrogen bonds, hydrophobic packing, and solvent permeation, of oseltamivir resistance have been thoroughly studied [8],[10]–[12],[31] little is known at the atomic level about the kinetics of drug binding in mutants. The idea that drug resistant mutants actually disrupt entry of oseltamivir into the SA binding site of neuraminidase through disruption of an electrostatic binding funnel is in part supported by experiments which demonstrated altered drug binding kinetics in H5N1 H274Y and N294S mutants. Specifically, the reported association rate constants () of oseltamivir with H5N1 neuraminidases were 2.52 in WT, 0.24 in H274Y and 1.1 in N294S [12]. Even though the oseltamivir-resistant mutations were seen located in or adjacent to the funnel, additional study is still needed for a full understanding of whether the H274Y and N294S mutations weaken the binding of the drug. Our observations of an electrostatically active binding pathway for oseltamivir provides guidance for further investigations.
Our simulations were restricted to studying six subtypes of N1 neuraminidase; expanding the scope of investigation to encompass N2–9 neuraminidases may reveal similar electrostatic funnels as seen in H5N1 and H1N1pdm neuraminidases. While the N1 subtype neuraminidases are different from other subfamilies (N2–N9) [13], there is evidence that electrostatic interactions play also a key role in mediating ligand recognition in N2 [16] and in ligand binding in N9 [17]. Thus, even if the charge distribution pattern in the N2–N9 families turns out to differ significantly from that of N1, an approach similar to our study (or those of previous studies employing complementary methods to calculate electrostatic potentials) could shed light on mechanisms of drug binding or resistance mechanisms in these subtypes.
Previous studies [17], [29] suggest that two flexible loops (termed “150” and “430” due to their residue positions) play a role in guarding drug access to the SA binding pocket of N1 neuraminidases. A discussion on the positions of these loops relative to the binding funnel are provided in Text S4, and illustrated in Figure S7. Our simulations (simEQ3 and simEQ4) reveal a loss of hydrogen bonds to loop 150 (namely between oseltamivir's acetyl group and R152) in the H274Y mutant systems. However, the disruption of endpoint interactions between oseltamivir and loop 150 do not appear to appreciably alter overall drug binding within the primary SA active site. This observation does not rule out what role loop 150 (and its associated D151 and R152 hydrogen bonds to oseltamivir) or loop 430 may play if the drug is occupying a secondary SA binding site (the “430 cavity”) suggested in [15], [16] and seen to play a role in simulation simFEQ4.
Electrostatic maps suggest a drug binding/unbinding pathway for oseltamivir along the charged binding funnel to the primary SA binding site, but simFEQ4 suggests also a role of the “430-cavity” (via interaction between oseltamivir's carboxylate group and R430's guanidino group). Since our simulations, from which electrostatic maps were derived, modeled oseltamivir in its primary binding site, a secondary charged binding channel exit path may emerge when the drug occupies this secondary site which should be investigated further.
The H274Y mutation induces drug-resistance to peramivir (a phase-III candidate), but neither H274Y nor N294S alter significantly the binding affinity for another antiviral drug, zanamivir [12], [32] or for sialic acid (SA), neuraminidase's natural substrate. The mechanism behind the efficacy of zanamivir against oseltamivir-resistant neuraminidase is not well understood. Examination of the crystal structure in mutants suggests that even in the case of the H274Y mutation, the zanamivir bound system is capable of maintaining a stabilizing hydrogen bond with residue 276 which is lost in the oseltamivir-bound mutant system [12]. Zanamivir's potency against oseltamivir-resistant strains also may be due to the high structural similarity it shares with SA. Specifically, zanamivir and SA share a hydrophilic glycerol group which is replaced by a hydrophobic pentyl group in oseltamivir and peramivir. Furthermore, there is evidence from prior studies [15], [16] that SA may follow a pathway into the neuraminidase active site which differs from that of oseltamivir; zanamivir may follow the same pathway as SA, and, hence, alteration of the oseltamivir binding funnel may not affect zanamivir, despite zanamivir's increased polarization relative to oseltamivir. In contrast, peramivir's ineffectiveness in the case of the H274Y mutant may be due to a change in hydrophobic packing of its pentyl group as witnessed in our simulations for oseltamivir. Recent efforts to derive new drugs that are effective against neuraminidase have also suggested inhibitors with remarkably different scaffolds than oseltamivir, zanamivir, and peramivir [33], [34]. Therefore, rational drug design for neuraminidase inhibitors will certainly benefit from studies that focus on binding kinetics behind selective drug resistance and also consider electrostatic steering [15]–[18].
Our simulations of wild type and mutant H5N1 and H1N1pdm neuraminidase systems bound to oseltamivir suggest that while the H274Y and N294S mutations appear to mainly disrupt the hydrophobic packing of oseltamivir's pentyl sidegroup, but not necessarily its conserved hydrogen network, drug-resistance may also arise from disruption of the binding process, i.e., from altered kinetics. Our simulations reveal a charged pathway in both H5N1 and H1N1pdm neuraminidase, that functions as a binding conduit for oseltamivir in which drug passage is controlled primarily by electrostatic attraction between drug and protein. The insight gained should assist in the rational design of neuraminidase inhibitors that exploit this binding pathway, but also avoid drug resistance.
The amino acid sequence of H1N1pdm neuraminidase was obtained from Genbank Locus ID CY041156, and that of H5N1 neuraminidase from Protein Data Bank entry 2HU4 [13]. A sequence alignment performed using Multiseq in VMD [35] showed that H1N1pdm has the highest percent of sequence identity (91.47%) with H5N1 among all neuraminidases with high resolution structural data. At the drug binding pocket, the notable difference between H5N1 and H1N1pdm neuraminidases is the replacement of Y347 by N347. Therefore, a homology model of H1N1pdm was built using H5N1 as the starting point and by mutating corresponding residues to match the wild type H1N1pdm.
The coordinates for H5N1 neuraminidase bound with oseltamivir (Tamiflu) was taken from a monomer of Protein Data Bank (PDB) structure 2HU4 (tetramer) [13], while those of mutants H274Y and N294S were taken from structures 3CL0 (monomer) and 3CL2 (monomer), respectively [12]. Even though the tetrameric form of neuraminidase has been employed in previous simulations [15], [17], its monomer contains a functionally complete active site and has been shown to yield satisfactory results in prior studies [8], [18], [31]. Therefore to conserve computational resources, only single monomers were used in our simulations. The position for oseltamivir bound to H1N1pdm was adopted from its corresponding location in H5N1, as the two proteins' binding pockets differ only by residue 347, located on a loop at the periphery of the active site. Oseltamivir-mutant complexes of H1N1pdm were built by mutating H274Y and N294S of the H1N1pdm wild type model. An representative simulated system is illustrated in Figure S8.
The simulations carried out are listed in Table 1. Six systems were modeled and simulated, namely oseltamivir bound H5N1, and H1N1pdm wild type, and H274Y as well as N294S mutants. The first six simulations involved equilibration of each oseltamivir bound neuraminidase structure (simEQ1–6). In simSMD1, steered molecular dynamics (SMD) simulations were used to remove oseltamivir from its stable binding site in H5N1 neuraminidase. In simFEQ1–10, equilibration simulations used a starting point generated from simSMD1 in which oseltamivir had undergone a rotation which partially displaced the drug from the binding site. In total, 680 ns of simulation were carried out for a system size of about 35,000 atoms.
Crystallographically resolved water molecules and a structurally relevant calcium ion near the native binding site for sialic acid (SA) were retained and modeled in all systems simulated. The protein complexes were then solvated in a TIP3P water box [36] and ionized by NaCl (0.152M) to mimic physiological conditions. The solvated H1N1pdm system with bound oseltamivir and active site calcium ion is shown in Figure S7A, a schematic view of the buried drug in the SA binding site is shown in Figure S7B.
Simulation parameters for oseltamivir were developed under the CHARMM force field parameter scheme, to complement the CHARMM31 force field for proteins with CMAP correction [37], [38]. Parameters for ligands were prepared using Paratool [39] in VMD [35]. Structure optimization and frequency calculations were performed at the HF/6-31G* level of Gaussian03 [40] and subsequently imported into Paratool. The quantum mechanics frequency calculation of the optimized geometry produced Hessian matrices which were transformed into the set of internal coordinates describing the drug bonded interactions. Atom types and charges already described in the existing CHARMM force field were assigned using the existing parameters. The atomic charges in oseltamivir's six member ring were newly parameterized by dividing the ring into several small fragments, and recalculated based on the total charge of each fragment. Fragments not explicitly defined in the CHARMM force field were modeled using analogs in the CHARMM force field extensions which fit closely to the fragment being parameterized. The dihedral angle potentials, which render the respective torsions highly rigid due to electron delocalization, were generated from the Hessian matrices. The drug's parameter and topology files are included in Supplemental Materials, as Protocol S1 and Protocol S2, respectively.
All simulations were performed using NAMD 2.7b2 [41] and the CHARMM31 force field with CMAP correction [37], [38]. The ionized systems were minimized for 10,000 integration steps and equilibrated for 20 ns with 1 fs time stepping. Following this, a 20 ns unconstrained equilibration was performed for subsequent trajectory analysis, with frames stored each picosecond. Constant temperature (T = 300 K) was enforced using Langevin dynamics with a damping coefficient of 1 ps. Constant pressure (p = 1 atm) was enforced through the Nosé-Hoover Langevin piston method with a decay period of 100 fs and a damping time constant of 50 fs. Van der Waals interaction cutoff distances were set at 12 Å, (smooth switching function beginning at 10 Å) and long-range electrostatic forces were computed using the particle-mesh Ewald (PME) with a grid size of less than 1 Å.
SMD simulations [42]–[47] fixed the center of mass of neuraminidase -carbons and applied a force to the center of mass of oseltamivir, along a vector connecting the two center of masses. In simSMD1–3, a constant velocity protocol was employed, with a pulling velocity of 0.5 Å/ns, 0.10 Å/ns, and 0.25 Å/ns, respectively. For the SMD spring constant [22], [48], we chose = 3kT/Å which corresponds to an RMSD value of 0.6 Å.
Trajectory frames were saved every 1000 integration timesteps (every picosecond). Analysis included the calculation of an averaged electrostatic potential field over all frames of a trajectory using RMSD-aligned structures. Maps of the electrostatic potential field were calculated on a three-dimensional lattice. The long-range contributions to the electrostatics were calculated employing the multilevel summation method (MSM), which uses nested interpolation of the smoothed pairwise interaction potential, with computational work that scales linearly with the size of the system [49]. The calculation was performed using the molecular visualization program VMD [35] that provides a GPU-accelerated version of MSM to produce the electrostatic potential map [21]. The GPU acceleration of MSM provided a significant speedup over conventional electrostatic summation methods such as the Adaptive Poisson Boltzman Solver (APBS) [50]. In fact, we achieved a processing time of 0.2 s per frame, versus 180 s per frame (on a conventional CPU) using APBS for a 35,000 atom system, which corresponds to a speedup factor of about 900. The use of GPU acceleration permitted averaging the electrostatic potential field over all frames of our simulation trajectories.
The root mean square deviation (RMSD) for the position of atoms within the simulation systems were used to access protein stability and state of equilibration. The RMSD calculations took into account a total frame alignment for the -carbons of either drug only or protein only, depending on the value reported (Figure S1 and Figure S2A). In accessing the stability of oseltamivir, we also aligned our coordinates against active site residues only (117–119, 133–138, 146–152, 156, 179, 180, 196–200, 223–228, 243–247, 277, 278, 293, 295, 344–347, 368, 401, 402, and 426–441, taken from [15]) for an additional drug-only RMSD calculation (Figure S3). For hydrogen bond analysis, a distance and angle cutoff of 3.5 Å and 60 degrees were employed, respectively. The change to the amount of solvent accessible surface area was used to assess alterations of hydrophobic packing interactions. The presence of salt bridges was assessed by taking a nitrogen-oxygen cutoff distance of 3.2 Å between charged residue side chains. Any close contacts that fell outside of the bound of molecular bonds or electrostatic (hydrogen bond or salt bridge) interactions were then examined frame-by-frame to distinguish between nonspecific surface or charged contacts.
|
10.1371/journal.ppat.1000783 | Transit through the Flea Vector Induces a Pretransmission Innate Immunity Resistance Phenotype in Yersinia pestis | Yersinia pestis, the agent of plague, is transmitted to mammals by infected fleas. Y. pestis exhibits a distinct life stage in the flea, where it grows in the form of a cohesive biofilm that promotes transmission. After transmission, the temperature shift to 37°C induces many known virulence factors of Y. pestis that confer resistance to innate immunity. These factors are not produced in the low-temperature environment of the flea, however, suggesting that Y. pestis is vulnerable to the initial encounter with innate immune cells at the flea bite site. In this study, we used whole-genome microarrays to compare the Y. pestis in vivo transcriptome in infective fleas to in vitro transcriptomes in temperature-matched biofilm and planktonic cultures, and to the previously characterized in vivo gene expression profile in the rat bubo. In addition to genes involved in metabolic adaptation to the flea gut and biofilm formation, several genes with known or predicted roles in resistance to innate immunity and pathogenicity in the mammal were upregulated in the flea. Y. pestis from infected fleas were more resistant to phagocytosis by macrophages than in vitro-grown bacteria, in part attributable to a cluster of insecticidal-like toxin genes that were highly expressed only in the flea. Our results suggest that transit through the flea vector induces a phenotype that enhances survival and dissemination of Y. pestis after transmission to the mammalian host.
| Bubonic plague cycles depend on the ability of Yersinia pestis to alternately infect two very different hosts—a mammal and a flea. Like any arthropod-borne pathogen, Y. pestis must sense host-specific environmental cues and regulate gene expression accordingly to produce a transmissible infection in the flea after being taken up in a blood meal, and again when it exits the flea and enters the mammal. We examined the Y. pestis phenotype at the point of transmission by in vivo gene expression analyses, the first description of the transcriptome of an arthropod-borne bacterium in its vector. In addition to genes associated with physiological adaptation to the flea gut, several Y. pestis virulence factors required for resistance to innate immunity and dissemination in the mammal were induced in the flea, suggesting that the arthropod life stage primes Y. pestis for successful infection of the mammal.
| Arthropod-borne transmission of bacterial pathogens is somewhat rare but has evolved in a phylogenetically diverse group that includes the rickettsiae, Borrelia spirochetes, and the gram-negative bacteria Francisella tularensis and Yersinia pestis, the plague bacillus. Y. pestis circulates among many species of wild rodents, its primary reservoir hosts, via flea bite. As it alternates between fleas and mammals, it is postulated that Y. pestis regulates gene expression appropriately to adapt to the two disparate host environments, and that different sets of genes are required to produce a transmissible infection in the flea and disease in the mammal.
Many important Y. pestis virulence factors that are required for plague in mammals have been identified, and most of them are induced by a temperature shift from <26°C to 37°C, which mimics the transition from a flea to the warm-blooded host [1]. To date, only three transmission factors (genes specifically required to produce a transmissible infection in the flea) have been characterized. One, the yersinia murine toxin (ymt) gene, encodes a phospholipase D that is required for survival in the flea midgut [2]. The other two, (hmsHFRS and gmhA), are responsible for an extracellular polysaccharide and a lipopolysaccharide (LPS) core modification that are required for normal biofilm formation and blockage in the flea [3],[4]. Biofilm development in the flea digestive tract is important for biological transmission [5],[6],[7]. After being taken up in a blood meal, Y. pestis proliferates in the lumen of the flea midgut to form cohesive multicellular biofilm aggregates. In some infected fleas, the proventricular valve between the midgut and esophagus is colonized. The subsequent growth and consolidation of the adherent Y. pestis biofilm amongst the rows of cuticle-covered spines that line the proventriculus interferes with normal blood feeding, resulting in regurgitation of bacteria and transmission. Fleas with a completely blocked proventriculus make prolonged, repeated attempts to feed, increasing the opportunities for transmission.
Formation of a Y. pestis biofilm in vitro and in the flea proventriculus depends on synthesis of an extracellular polysaccharide matrix (ECM) that is synthesized only at temperatures below 26°C [3],[7]. In common with many other bacteria, ECM synthesis in Y. pestis is controlled by intracellular levels of cyclic di-GMP, which are determined by competing activities of the hmsT diguanylate cyclase and hmsP phosphodiesterase gene products [8],[9]. Bacterial adhesins are typically required for initial adherence and autoaggregation in biofilm development [10], but such factors have yet to be identified in Y. pestis.
In a previous study, we reported the in vivo gene expression profile of Y. pestis during bubonic plague in rats [11]. In this study, we characterized the Y. pestis transcriptome in blocked Xenopsylla cheopis rat fleas, an important vector of plague to humans. Comparing the Y. pestis gene expression profile in the flea to those of in vitro biofilm and planktonic cells cultured at the low temperature typical of the flea implicated several genes in a flea-specific adaptive response and in proventricular blockage. In addition, comparing the gene expression patterns in the flea and in the rat bubo confirmed that distinct subsets of genes are differentially expressed during the Y. pestis life cycle. Notably, several genes with known or predicted roles in protection against the mammalian innate immune system and in pathogenesis were upregulated in the flea, suggesting that transit through the insect vector preinduces a phenotype that enhances Y. pestis survival and dissemination in the mammal after flea-borne transmission.
Little is known about the environmental conditions in the flea digestive tract, how Y. pestis adapts to them, or the physiological state of the bacteria at transmission when they exit the flea and enter the mammal. Adult fleas are obligate blood feeders and take frequent blood meals, consisting primarily of protein and lipid with relatively little carbohydrate. Flea proteases, lipases, and other digestive enzymes begin to process the blood meal in the midgut immediately after feeding, yielding amino acids and peptides, glycerol, fatty acids, and simple carbohydrates [12]. This provides the “medium” for Y. pestis growth, but these and other factors such as pH, oxygen tension, osmolarity, and flea antibacterial immune components are poorly defined. During the first week after being ingested in an infectious blood meal, Y. pestis grows rapidly in the flea midgut to form large bacterial aggregates. Bacterial load peaks at about 106 cells per flea as the Y. pestis biofilm accumulates in the proventriculus to cause blockage, and then plateaus [2],[3].
In this study, we determined the Y. pestis gene expression profile in infective, blocked fleas, in which the proventriculus was occluded with a mature bacterial biofilm. Y. pestis KIM6+, which lacks the 70-kb virulence plasmid that is not required for flea infection or blockage [3] was used for this analysis. Blockage occurred between 1.5 and 3.5 weeks after the initial infectious blood meal, during which time the fleas fed on uninfected mice twice weekly. The Y. pestis in vivo biofilm transcriptome was compared to the transcriptomes of in vitro biofilm and planktonic cultures grown at 21°C, the same temperature at which the fleas were maintained.
Expression of 55% of Y. pestis ORFs was detected in the flea samples; and 74 to 79% in the in vitro biofilm, exponential phase planktonic and stationary phase planktonic cultures. Principal component analysis to visualize overall clustering of the microarray data showed that the transcriptional profiles were reproducible and discrete for the in vitro and in vivo conditions (Fig. 1A). Profiles of the exponential and stationary phase planktonic cultures clustered most closely, whereas the profiles from in vitro and in vivo biofilm growth were more distinct from each other and from the planktonic culture profiles. There were 214 Y. pestis genes whose expression was significantly upregulated and 56 genes downregulated in the flea compared to all in vitro growth conditions (Fig. 1B; Tables S1 and S2). Quantitative RT-PCR analysis of a subset of Y. pestis genes differentially expressed in the flea was confirmatory of the microarray results (Fig. S2).
Of the 214 genes upregulated in the flea gut compared to all in vitro conditions, 78 are metabolic genes, 60 of which are involved in uptake and catabolism of amino acids and carbohydrates (Table S1). In particular, genes involved in transport and catabolism of the L-glutamate group of amino acids (Gln, His, Arg, and Pro) were specifically upregulated in the flea (Fig. 2). The degradation of these amino acids gives rise to L-glutamate and the TCA cycle intermediates succinate, formate, and α-ketoglutarate. The gabD and gabT genes involved in the production of succinate from γ-aminobutyrate (GABA), another member of the L-glutamate group, were also highly induced in the flea. The gabD gene functions to produce succinate from both GABA and hydroxyphenylacetate (HPA), an aromatic degradation product of Tyr and Phe; and the HPA transport (hpaX) and catabolism (hpaCBIFHDE) genes of Y. pestis were also highly upregulated in the flea gut (Table S1, Fig. 2). As Y. pestis does not have homologs of genes required to produce GABA or HPA, these metabolites may be taken up from the flea digestive tract. Alternatively, the gabD and gabT gene products might act in the reverse direction to synthesize GABA, which has osmoprotective properties [13]. The central role of the L-glutamate family of amino acids may also confer this advantage in the flea gut, because Glu and Pro are osmoprotectants. Interestingly, both glutamate and GABA are important neurotransmitters at the neuromuscular junction of insects, and the concentration of glutamate is very low in insect hemolymph, suggesting that it is converted to glutamine before it is absorbed [14]. Insect midgut epithelium is typified by multiple amino acid transporters with specific substrates and rapid absorption kinetics, but different amino acids enter the hemocoel at different rates and amounts [14],[15]. Thus, Y. pestis metabolism in the flea may reflect the available pool of amino acids in the midgut.
In contrast to the amino acids, hexoses do not appear to be an important energy source during infection of the flea. Only the genes encoding for chitobiose phosphotransferase (PTS) uptake and utilization systems (chbBC; chbF), and for a PTS system of unknown specificity (frwBCD) were significantly upregulated in the flea [16],[17]. Chitobiose could be present in the flea gut due to turnover of the chitin layer on the proventricular spines. Expression of the glucose PTS system was only slightly increased relative to LB cultures, and other PTS systems were downregulated (Table S2). Glycolytic pathways were not upregulated in the flea; instead, available hexoses and the gluconeogenesis pathway may be used to synthesize polysaccharide components required for cell growth. Upregulation of the actP and acs genes in the flea, which direct the uptake of acetate and its conversion to acetyl-CoA, also suggests that insufficient acetyl-CoA is produced by glycolysis to potentiate the TCA cycle. The switch from acetate secretion to acetate uptake is typical of growth in a glucose-limited, amino acid rich environment [18]. In contrast to hexose uptake systems, Y. pestis genes that encode permeases for the pentoses ribose, xylose, and arabinose were induced in the flea gut. Acquisition of pentoses from the environment may be important because Y. pestis does not possess glucose 6-phosphate dehydrogenase activity, the first step of the pentose phosphate pathway [19].
Although the flea gut contains lipid derived from the blood meal, Y. pestis does not appear to use it as a major energy source. None of the fatty acid uptake or catabolism genes were upregulated in the flea compared to growth in LB. However, genes for glycerol and glycerol-3-phosphate uptake and utilization were upregulated, suggesting that flea digestion products derived from blood glycerolipids may be used by Y. pestis. In summary, Y. pestis appears to use amino acids, particularly the L-glutamate family, as primary carbon, nitrogen, and energy sources in the flea. Amino acid carbon is presumably funneled into the TCA cycle, the genes for which are highly expressed in the flea (Table S3).
Because blockage of the flea vector is essentially a biofilm phenomenon, Y. pestis genes whose expression patterns are significantly upregulated in the flea and flowcell biofilms relative to planktonic cultures (Table S4) might indicate that they are transmission factors. Several studies comparing the transcriptional profiles of Escherichia coli and other gram negative bacteria during biofilm and planktonic growth in vitro have been published [20],[21],[22],[23]. Certain genes whose mutational loss resulted in an altered biofilm phenotype have been identified in these studies; but in general a consistent, distinct biofilm gene expression profile has not emerged. This is probably because different media and experimental systems have been employed and the fact that a biofilm consists of a physiologically heterogeneous community [24],[25]. Nevertheless, common biofilm-related adaptations include the repression of motility and the induction of specific adhesins, an extracellular polysaccharide matrix (ECM), and an envelope stress response (ESR) [10],[23]. However, Y. pestis is constitutively nonmotile, and synthesis of the Hms-dependent biofilm ECM is regulated post-translationally [26]. The ymt gene was among the most highly expressed genes in the flea (Table S3), but neither it nor the known transmission factors (hmsHFRS, hmsT, hmsP, and gmhA) showed significantly higher expression in the flea than in vitro at 21°C, indicating that they are induced primarily by low temperature, and not by environmental factors specific to the flea gut. Y. pestis homologs of two genes with previously identified roles in biofilm, yidE, which encodes a hyperadherence factor in E. coli [27], and cpxP, a member of the cpxPAR ESR system, were upregulated in the flowcell; but predicted adhesin genes were not upregulated.
The transcriptional profile of Y. pestis in blocked fleas showed greater similarity to the transcriptional profile reported for E. coli in mature, four-day-old in vitro biofilms [23]. In addition to yidE and cpxP, other Y. pestis predicted adhesins and components of an ESR were upregulated in the flea. The Y. pestis homologs of Pseudomonas aeruginosa cupA1 and cupA3 in a predicted fimbrial biosynthesis operon and yapL, a predicted autotransporter adhesin similar to E. coli tibA, were specifically upregulated in the flea (Table S1). The cupA fimbrial locus and tibA are important for surface adherence and for biofilm formation in P. aeruginosa and E. coli, respectively [28],[29]. Evidence for induction of an ESR in the flea included the high expression levels of rpoE, the gene for the alternate transcription factor σE (as well as the anti-σE negative regulator genes rseA and rseB), cpxP; and pspA and pspG, components of the phage-shock protein (Psp) response (Tables S1 and S3). These genes were also found to be upregulated in mature E. coli biofilms [23], suggesting that the three prominent ESR systems are important for integrating signals required for survival in a biofilm.
Because homologs of the yidE, cpxP, tibA (yapL), cupA fimbriae, and pspABC genes were upregulated in the flea and have been shown to be involved in biofilm formation in other bacteria [23],[27],[28],[29], we made a series of Y. pestis strains containing deletions of these loci. However, the single loss of any of these genes did not result in a noticeable defect in biofilm formation in vitro, or in flea infection or blockage (data not shown). These genes may contribute to biofilm formation, but are not individually essential for this phenotype. Although genes in the polyamine transport gabTpotDBC locus are among the most highly induced genes in the flea (Table S1) and polyamines are essential for Y. pestis biofilm formation [30], we have previously reported that a Y. pestis Δpot mutant has no defect in flea infection or blockage [31]. This is likely due to the fact that Y. pestis is able to synthesize polyamines de novo.
With this study, the in vivo transcriptome of Y. pestis in blocked fleas and in the rat bubo [11] have now both been characterized. A comparison of normalized gene expression levels from the two data sets provides insight into the biology of the flea-mammal life cycle. About 15% of Y. pestis genes showed significantly higher relative expression levels or expression only in the flea than in the bubo; 24% were more highly expressed in the bubo than in the flea; and 61% were not differentially expressed in the two hosts (Fig. 3).
Several virulence factors were differentially regulated in the two hosts, but others were not (Table 1). In addition to the known temperature-induced virulence factors, iron acquisition systems, including the ybt and yfe operons that are required for virulence; and oxidative and nitrosative stress response genes, including the hmp virulence factor, are highly upregulated in the rat bubo, but not the flea. The analysis also reinforces the model that Y. pestis produces a hexaacylated lipid A in the flea, and that the change to the less immunostimulatory tetraacylated form occurs only after transmission [32]. Other virulence and transmission factors were not differentially regulated, including the hms genes; and the Y. pestis plasminogen activator (pla), critical for dissemination from extravascular tissue at the fleabite site [33], and ymt were highly expressed in both hosts (Table S3 and [11]). The Y. pestis outer surface protein gene yadB, recently shown to be required for dissemination and bubonic plague pathogenesis from a subcutaneous inoculation site [34], was significantly upregulated in both the flea and the bubo compared to in vitro conditions (Tables 1, S1).
Expression of genes in the pH 6 antigen locus (psaEFABC), responsible for the synthesis and transport of the PsaA fimbriae that enhance resistance to phagocytosis by macrophages [35],[36], were higher in the bubo than the flea, although the usher protein gene psaC was upregulated in the flea compared to in vitro growth (Tables 1, S1). The psa locus is regulated by RovA [36]. Consistent with these findings, rovA expression was downregulated in the flea; whereas expression of rovM, a negative regulator of rovA [37], was upregulated.
The transcriptional regulator gene phoP of the PhoPQ two-component regulatory system and the PhoP-regulated mgtC gene were expressed at levels >2-fold higher in fleas than in any other condition (Tables 1, S1, S3). PhoP and MgtC are established virulence factors known to be important for survival of Y. pestis and other gram-negative bacteria in macrophages and for resistance to cationic antimicrobial peptides (CAMPs) of the mammalian innate immune response [38],[39],[40]. The PhoPQ system is induced in low Mg2+ or low pH environments, or by exposure to CAMPs [41],[42],[43]. The Mg2+ concentration and pH of the flea digestive tract have not been defined, so the inducing stimulus is unknown, but CAMPs are induced and secreted into the gut by blood feeding insects when they take a blood meal containing bacteria [44],[45]. X. cheopis fleas encode homologs of the insect CAMPs cecropin and defensin, and mount an inducible antibacterial response to infection (unpublished data). Thus, the PhoPQ regulatory system may be induced by the flea's immune system in response to Y. pestis in the midgut. Despite the upregulation of phoP in the flea, with the notable exception of mgtC there was little correlation between predicted PhoP-regulated genes in vitro and genes upregulated in the flea [39],[46],[47]. Differential regulation of members of the PhoP regulon may occur depending on the inducing stimulus, however [48].
Soon after transmission, Y. pestis would be expected to encounter rapidly-responding phagocytic cells in the dermis. To assess the overall effect of the flea-specific phenotype on this encounter, we compared the interaction of Y. pestis recovered from infected fleas and from in vitro cultures with murine bone marrow macrophages. Bacteria from fleas showed significantly lower levels of phagocytosis (Fig. 4A). We have previously reported analogous findings using human polymorphonuclear leukocytes (PMNs) [7].
The yit and yip genes in a Y. pestis locus (y0181–0191) that encode predicted insecticidal-like toxins of the toxin complex (Tc) family and three linked phage-related genes were upregulated 4- to 50-fold in the flea midgut (Tables 1 and S1). We previously reported that the genes for these Tc-like proteins are highly expressed in fleas, but that their products are nontoxic to fleas [49]. yitR, a LysR-type regulator that activates the Tc-like yit genes [50], was upregulated >10-fold in the flea, but its expression was not detected in the rat bubo (Table 1). The specific induction in the flea of yitR and genes in the adjacent Tc-like yit and yip loci suggests that they are involved in adaptation to and colonization of the flea. However, deletion of yitR or yitA-yipB (y0183–y0191) does not affect the ability of Y. pestis KIM6+ to infect or block fleas (data not shown). These observations, and the fact that the Yersinia Tc proteins have toxicity to certain eukaryotic cell lines in vitro [50],[51], prompted us to investigate a possible post-transmission antiphagocytic role for these proteins in the mammalian host.
To determine if the insecticidal-like toxins were involved in resistance to phagocytosis, we repeated the macrophage experiments with a Y. pestis ΔyitR mutant, which as expected showed greatly reduced expression of the yit and yip genes in vitro and in the flea (Fig. 4B). Loss of yitR significantly reduced the increased resistance to phagocytosis of Y. pestis isolated from infected fleas (Fig. 4C).
Since the yit and yip genes are not required for Y. pestis to produce a transmissible infection in fleas, it was possible to compare the virulence of wild-type and ΔyitR Y. pestis following transmission by fleabite. The incidence rate and time to disease onset were identical for both Y. pestis strains, demonstrating that expression of yit and yip is not essential for flea-borne transmission or disease (data not shown). On average, the mice challenged with Y. pestis ΔyitR-infected fleas, both those that developed disease and those that did not, received a higher cumulative number of bites from blocked fleas than the mice challenged with Y. pestis-infected fleas, but this difference was not statistically significant (Fig. 5). However, it was not possible to detect any relatively minor difference in LD50 because the number of bacteria transmitted by a blocked flea varies widely [1],[52]. Even a small decrease in LD50 provided by the Yit-Yip proteins would be significant at the ecological level in the maintenance of plague transmission cycles, because the transmission efficiency of blocked fleas is very low– often only a few or no bacterial cells are transmitted in an individual fleabite [52]. Because phoP is required by Y. pestis to produce a transmissible infection in fleas (unpublished data), it was not possible to similarly assess the effect on disease transmission of phoP induction in the flea.
When Y. pestis is transmitted into the dermis by an infected flea, it is immediately exposed to the mammalian innate immune system. The most important antiphagocytic virulence factors, the cytotoxic Yersinia outer proteins (Yops), part of the T3SS encoded by the Y. pestis virulence plasmid and the F1 capsule encoded by the pMT1 plasmid, are not present at this initial stage of infection. Their expression is strictly temperature-regulated and are not produced in vivo until 3–5 hours after the temperature shift to 37°C that accompanies transmission [1],[3],[53],[54]. Consequently, Y. pestis grown at <28°C in vitro are initially susceptible to in vivo uptake and killing by phagocytes until the Yop and F1 virulence factors are produced, effectively preventing further phagocytosis [53],[54]. Our results indicate that Y. pestis entering the mammal from an infective flea is relatively resistant to macrophages, as well as PMNs [7]; a vector-specific phenotype that is not related to the T3SS or capsule.
Coming from the flea, Y. pestis is also associated with the biofilm ECM, identical or closely related to the poly-β-1,6-N-acetyl glucosamine ECM of staphylococcal biofilms, which has been shown to provide protection from innate immune components [55],[56]. In addition, although the antiphagocytic F1 capsule and Psa fimbriae do not appear to be produced in the flea, upregulation in the flea of most F1 genes in the cafRcaf1M1A1 locus and the Psa usher protein gene psaC (Tables 1, S1) suggests that components of the F1 and Psa translocation system are made, which may prime Y. pestis for rapid secretion of these extracellular virulence factors after transmission. The upregulation of the innate immunity resistance genes phoP and mgtC suggest that those Y. pestis that are phagocytized may be prepared for resistance to CAMPs and intracellular survival while still in the flea vector. Finally, the major essential virulence factors yadBC and pla, essential for Y. pestis dissemination from the dermis, were maximally or very highly expressed in the flea (Tables 1, S3). Besides degrading plasminogen, the Pla protease may also inactivate CAMPs, particularly when the F1 capsule is not present [57], which matches the phenotype of Y. pestis in the flea.
In summary, Y. pestis appears to be prepared for pathogenesis in the mammal while still in the flea vector. The biofilm phenotype of Y. pestis and the virulence factors upregulated or highly expressed in the flea may enhance the earliest stages of plague pathogenesis while the full complement of temperature-shift-regulated virulence factors is still being induced. Increased resistance to innate immunity that is preinduced in the flea vector may be critical to productive transmission because blocked fleas transmit relatively few bacteria, often below the LD50 of Y. pestis grown in vitro at <28°C [1],[52].
All animals were handled in strict accordance with good animal practice as defined by NIH animal care and use policies and the Animal Welfare Act, USPHS; and all animal work was approved by the Rocky Mountain Laboratories Animal Care and Use Committee.
Y. pestis KIM6+, which lacks the 70-kb virulence plasmid that is not required for flea infection or blockage, was used for gene expression analyses. A KIM6+ strain with an in-frame deletion that eliminated amino acids 28–281 of the predicted 291 amino acid residue yitR (y0181) gene product was produced by allelic exchange, using the pCVD442 suicide vector system [11]. This mutant was complemented by electroporation with a recombinant pWKS130 plasmid containing the wild-type yitR promoter and orf. The ΔyitR mutant was also transformed with pWKS130 alone to generate an empty vector control strain. For in vitro planktonic samples, bacteria were grown from frozen stocks in brain heart infusion (BHI) medium at 28°C, followed by two successive transfers in Luria Bertani broth supplemented with 100 mM MOPS, pH 7.4 (LB/MOPS) at 21°C. An inoculum of 104 cells/ml was added to 50 ml of LB/MOPS and incubated at 21°C with shaking at 250 rpm until exponential (OD600 = 2.5) or stationary phase (OD600 = 4.5). Approximately 0.5 ml of the exponential phase culture and 0.25 ml of the stationary phase culture was resuspended in 1 ml and 0.5 ml, respectively, of RNAprotect bacterial reagent (Qiagen; Valencia, CA), incubated for 5 min at room temperature, and centrifuged at 21°C for 5 min prior to RNA extraction.
For in vitro biofilms, 400 µl of a 107/ml bacterial suspension was injected into a flowcell (Stovall; Greensboro, NC) that was connected to a reservoir of LB/MOPS at 21°C. Following a 30 min incubation period to allow the bacteria to adhere to the glass surface of the flow cell, LB/MOPS was pumped through the flow cell at a rate of 0.3 ml/min. After 48 hours, the flowcell was disconnected and the thick Y. pestis biofilm was harvested and treated with 0.5ml of RNAprotect similarly to the planktonic cultures.
X. cheopis fleas were infected with Y. pestis KIM6+ by using a previously described artificial feeding system [3]. The infectious blood meal was prepared by growing Y. pestis KIM6+ overnight at 37°C in BHI medium, without aeration. A cell pellet containing 109 bacterial cells was resuspended in 1 ml PBS and added to 5 ml heparinized mouse blood. Fleas that took a blood meal were maintained at 21°C and 75% relative humidity, fed twice weekly on uninfected mice, and monitored for proventricular blockage as previously described [3]. On the day blockage was diagnosed, the digestive tract was dissected out and macerated in RNAprotect, a process that required about 1 min. Thirty midguts from blocked fleas were pooled for each of the two biological replicates. Midguts from 60 uninfected fleas were also collected as controls to assess background hybridization of flea RNA to the microarray.
A flea-borne transmission model [58] was used to determine Y. pestis infectivity after challenge by flea bite. Fleas were infected with Y. pestis 195/P, a fully virulent wild-type strain, or with a Y. pestis 195/P ΔyitR mutant constructed as described above. Between 2–3 weeks after infection, the time required for Y. pestis to block fleas with a proventricular biofilm, groups of 20–40 fleas were applied to a restrained mouse and allowed to feed for 60 min. The fleas were then recovered and examined under a dissecting microscope to determine how many had taken a normal blood meal (unblocked or non-infective fleas) and how many were blocked (infective fleas). After challenge, mice were monitored and euthanized upon the appearance of signs of terminal illness. Mice that did not develop any symptoms after one week following a challenge were re-challenged. A total of 9–10 BALB/cAnN and 10 RML Swiss-Webster mice were challenged with each strain.
RNA was isolated from six independent samples from in vitro and flow cell cultures and two independent samples from pooled blocked fleas (Fig. S1) using the RNeasy Mini Kit (Qiagen). Flea-derived RNA samples were secondarily split into three technical replicates each. RNA integrity was verified on a Bioanalyzer 2100 (Agilent Technologies; Santa Clara, CA). Total RNA (100 ng) was amplified and labeled with modified biotin-11-CTP (Perkin Elmer; Waltham, MA) and biotin-16-UTP (Roche Molecular Biochemicals, Pleasanton, CA) by using the Message-Amp II-Bacteria amplified antisense RNA (aRNA) kit (Ambion; Austin, TX). Amplified RNA was then fragmented using Ambion's Fragmentation reagent (Applied Biosystems), hybridized to the RML custom Affymetrix GeneChip that contains sequences for all Y. pestis predicted ORFs, and scanned. The amplification step did not affect the relative transcript signals obtained by microarray (data not shown).
Affymetrix GeneChip Operating Software (GCOS v1.4, GEO platform GPL2129, http://www.affymetrix.com) was used for initial analysis of the microarray data at the probe-set level. All *.cel files, representing individual biological replicates, were scaled to a trimmed mean of 500 using a scale mask consisting of only the Yersinia pestis KIM6+ probe-sets to produce the *.chp files. A pivot table with all samples was created including calls, call p-value and signal intensities for each gene. The pivot table was then imported into GeneSpring GX 7.3 (http://www.chem.agilent.com), where hierarchical clustering (condition tree) using a Pearson correlation similarity measure with average linkage was used to produce the dendrogram indicating that biological replicates grouped together. The pivot table was also imported into Partek Genomics Suite software (Partek Inc.; St. Louis, MO) to produce a principal components analysis (PCA) plot as a second statistical test for the grouping of biological replicates. ANOVA was run from this data set to produce a false discovery rate report producing false positive reduced p-values for each comparison of interest.
The correlated replicates of all test conditions and controls were combined, and quality filters based upon combined calls and signal intensities were used to further evaluate individual gene comparisons. Present and marginal calls were treated as the same whereas absent calls were negatively weighted and eliminated from calculations. Ratios of test/control values and associated t-test and ANOVA p-values values of all individual genes passing the above filters were determined using GeneSpring, SAM, and Partek software. The microarray data have been deposited in the NCBI GEO public database (accession number GSE16493).
To compare differential in vivo gene expression patterns in the flea and the rat, the average hybridization signal for each individual Y. pestis gene was divided by the average signal of all 4,683 genes on the microarray for both the flea microarray (this study) and the rat bubo microarray [11] data sets. Gene by gene comparisons of these normalized expression data sets were used for Fig. 3 and Tables 1, S5, and S6).
Murine bone marrow-derived macrophages were prepared as described [59],[60] and cultured in Dulbecco's Modified Eagles medium (DMEM) supplemented with 5 mM L-glutamine, 25 mM HEPES, 10% heat-inactivated fetal bovine serum, 5 mM non-essential amino acids, and 10 ng/ml CSF-1 (PeproTech; Rocky Hills, NJ). 1-ml suspensions of Y. pestis KIM6+ containing pAcGFP1 (Clontech; Mountain View, CA) from 21°C stationary phase LB/MOPS cultures, or from triturated midguts dissected from fleas 2 to 3 weeks after infection were treated for 15 sec in a FastPrep FP120 using lysing matrix H (Qbiogene; Carlsbad, CA) to disrupt bacterial aggregates, quantified by Petroff-Hausser direct count, and diluted in DMEM to ∼1×106 bacteria/ml. 0.1 ml of bacterial suspension was added to tissue culture plate wells containing ∼1×105 differentiated primary macrophages cultured on 12 mm glass coverslips in 1 ml DMEM. The plates were not centrifuged after addition of the bacteria, and midgut triturate from an equivalent number of uninfected fleas was added to the in vitro-derived bacterial suspensions used for these experiments. After 1 h incubation at 37°C and 5% CO2, the medium was removed and the cells washed, fixed in 2.5% paraformaldehyde for 10 min at 37°C, and then rewashed. Extracellular bacteria were labelled by indirect immunofluorescence as described [60] using a 1∶50,000 dilution of hyperimmune rabbit anti-Y. pestis polyclonal antibody [7] and a 1∶400 dilution of AlexaFluor 568-conjugated goat anti-rabbit antibody (Invitrogen; Carlsbad, CA). The percentage of extracellular bacteria was determined by dividing the number of red-fluorescent bacteria by the total number (red- and green only-fluorescent) bacteria associated with individual macrophages. To calculate differential resistance to phagocytosis for a given strain, the average percent extracellular LB-grown bacteria was subtracted from the average percent extracellular flea-derived bacteria. Results from 2–3 independent experiments performed in triplicate were analyzed by unpaired two-tailed t-test.
Independent RNA samples were prepared from blocked fleas and in vitro biofilm and planktonic cultures as described for the microarray experiments, except that the RNA was not amplified. Samples were treated with rDnase I (Ambion) and confirmed by PCR to be free of genomic DNA contamination. cDNA was synthesized from the RNA and used for quantitative PCR on an ABI Prism 7900 sequence detection system (Taqman, Applied Biosystems). The reactions contained oligonucleotide primers and probes designed using Primer Express version 2.0 software (Applied Biosystems) and the Taqman Universal PCR Master Mix (Applied Biosystems). For each primer-probe set assay, a standard curve was prepared using known concentrations of Y. pestis KIM6+ genomic DNA and used to transform CT values into relative DNA quantity. The quantity of cDNA for each experimental gene was normalized relative to the quantity of the reference gene crr (y1485), and the ratio of the normalized quantity of each gene in the flea samples to the normalized quantity in the in vitro samples was calculated (Fig. S2). Primer and probe sets used are listed in Table S7.
|
10.1371/journal.pgen.1004572 | Admixture in Latin America: Geographic Structure, Phenotypic Diversity and Self-Perception of Ancestry Based on 7,342 Individuals | The current genetic makeup of Latin America has been shaped by a history of extensive admixture between Africans, Europeans and Native Americans, a process taking place within the context of extensive geographic and social stratification. We estimated individual ancestry proportions in a sample of 7,342 subjects ascertained in five countries (Brazil, Chile, Colombia, México and Perú). These individuals were also characterized for a range of physical appearance traits and for self-perception of ancestry. The geographic distribution of admixture proportions in this sample reveals extensive population structure, illustrating the continuing impact of demographic history on the genetic diversity of Latin America. Significant ancestry effects were detected for most phenotypes studied. However, ancestry generally explains only a modest proportion of total phenotypic variation. Genetically estimated and self-perceived ancestry correlate significantly, but certain physical attributes have a strong impact on self-perception and bias self-perception of ancestry relative to genetically estimated ancestry.
| Latin America has a history of extensive mixing between Native Americans and people arriving from Europe and Africa. As a result, individuals in the region have a highly heterogeneous genetic background and show great variation in physical appearance. Latin America offers an excellent opportunity to examine the genetic basis of the differentiation in physical appearance between Africans, Europeans and Native Americans. The region is also an advantageous setting in which to examine the interplay of genetic, physical and social factors in relation to ethnic/racial self-perception. Here we present the most extensive analysis of genetic ancestry, physical diversity and self-perception of ancestry yet conducted in Latin America. We find significant geographic variation in ancestry across the region, this variation being consistent with demographic history and census information. We show that genetic ancestry impacts many aspects of physical appearance. We observe that self-perception is highly influenced by physical appearance, and that variation in physical appearance biases self-perception of ancestry relative to genetically estimated ancestry.
| Understanding the basis of a variation in human physical appearance has been a topic of long-standing research interest. However, little is known about the genetic basis of most of this variation. An exception is pigmentation, which has been the focus of considerable research, particularly in Europeans [1]–[4]. Refining our knowledge on the genetics of physical appearance in human populations is of considerable evolutionary, biomedical and forensic importance. This research is also of broad social interest due to its bearing on debates around notions of self-identity, ethnicity and race.
Latin America provides an advantageous setting in which to examine the impact of genetic variation on physical appearance. The region has a history of extensive admixture between three continental populations: Africans, Europeans and Native Americans [5], [6]. Latin America also provides an informative context in which to explore the perception of variation in physical appearance. The region has a unique history relating to the social and cultural politics of ethnicity, race and nation [7]–[9]. A considerable number of genetic studies have examined admixture in Latin America [10]–[14]. However, these analyses have mostly been based on relatively small samples and focused mainly on describing patterns of variation in admixture proportions between individuals and countries/regions. Few studies have examined the impact of genetic ancestry on physical appearance or the relationship of these to individual notions of ethnicity and ancestry [15], [16].
In this paper we present the first phase of a research program focused on the genetics of physical appearance in Latin Americans. We base this program on a sample of over 7,000 individuals ascertained in five countries: Brazil, Chile, Colombia, México and Perú. Information was obtained for a range of socio-demographic variables, physical attributes and self-perception of ancestry. Here we report analyses based on individual mean genome admixture proportions. Coordinate-based spatial analyses illustrate the significant variation in ancestry existing across Latin America, in agreement with demographic history and census information. Significant effects of ancestry were detected for most of the phenotypes examined, and the direction of these effects agrees with the phenotypic differentiation of Africans, Europeans and Native Americans. Finally, we observe that certain phenotypes have a strong impact on self-perception and that these phenotypes bias self-perceived relative to genetically estimated ancestry.
Summary descriptive statistics for the study sample collected are presented in Table 1.
We estimated individual African/European/Native American admixture proportions with data for 30 highly informative SNPs using the ADMIXTURE program [17]. These markers were chosen from the 5,000 proposed by Paschou et al (2010) [18] as highly informative for continental ancestry estimation (see Methods). The selected set of markers produced individual ancestry estimates in 372 Colombians, included in a recent genome-wide association study [19], with correlations of ∼70% (for the three continental ancestries) compared to estimates obtained with 50,000 markers (LD-pruned), and identical sample means. Although we estimated individual ancestry with a relatively small number of markers, we verified that the inferences drawn are robust to the level of uncertainty of the estimates obtained (see below).
Consistent with previous studies, we observe extensive variation in ancestry between countries (Table 1) as well as between individuals within countries (Text S1) and between socioeconomic strata (Text S2) [12], [13], [20]–[22]. In order to obtain a spatial representation of variation in ancestry we obtained interpolated maps based on the geographic coordinates for the birthplaces of research volunteers. The geographic distribution of these birthplaces (Figure 1 and Figure S4) overlaps with regional population density from national census data (Figure S5). Consistent with this pattern, the number of volunteers for each birthplace correlates with census size for these localities: Brazil (r = 0.32, p-value <10−5), Chile (r = 0.51, p-value <10−4), Colombia (r = 0.54, p-value <10−13), Mexico (r = 0.44, p-value <10−8), Perú (r = 0.41, p-value <10−4). Few volunteer birthplaces were thus located in sparsely populated regions (e.g. Amazonia) and geographic interpolation of ancestry in those regions should be regarded with special caution.
The Brazilian sample (Figure 1A) shows widespread European ancestry with the highest levels being observed in the south. African ancestry is also widespread (except for the south) and reaches its highest values in the East of the country. Native American ancestry is highest in the north-west (Amazonia). The Chilean sample (Figure 1B) shows the least regional variation, with low levels of African ancestry throughout the country. European and Native American ancestry are relatively uniform, although somewhat higher European ancestry is seen around the main urban areas of the north and centre, Native ancestry predominating elsewhere, particularly in the south. The Colombian sample (Figure 1C) shows highest African ancestry in the coastal regions (particularly on the Pacific) and highest European ancestry in central areas. Native ancestry appears highest in the south-west and in the east of the country (Amazonia) but interpolations in these areas are based on few data points. In the Mexican sample (Figure 1D) Native American ancestry is highest in the centre/south of the country with the north showing the highest proportion of European Ancestry. African ancestry is generally low across Mexico except for a few coastal regions. The Peruvian sample (Figure 1E) shows substantial Native American ancestry throughout the country, particularly in the south, European ancestry appears highest around northern/central areas. African ancestry in Peru is generally low, except for parts of the northern coast.
To evaluate the statistical significance of the observed spatial variation in ancestry we calculated Moran's Index (I) of association between each individual ancestry component and spatial location. These were significant for the three ancestries in all countries (p-values <0.02). Since the three ancestry components are not independent, we also calculated canonical correlation coefficients between ancestry and geographic location. These were also significant for all countries (p-values <0.001). The variation in ancestry seen in the admixture maps of Figure 1 also result in highly significant correlations of the three ancestries with altitude of birthplace (p-values <2×10−16 for the three ancestries): African and European ancestry decreases with altitude (r of −0.24 and −0.39, respectively), while Native American ancestry increases (r = 0.48).
The Kriging interpolation scheme used in building the maps of Figure 1 uses the mean ancestry at each birthplace and does not provide information on the extent of individual variation in ancestry at each map location. In the 102 birthplaces with 10 or more individuals sampled we observe that the standard deviation in the three individual ancestry estimates extends over a wide range: African (0.012–0.022), European (0.046–0.273) and Native American (0.039–0.274). We evaluated the correlation of this variation in individual ancestry with the census size of these localities and found a significant positive correlation for all ancestries (r>0.3, p-values <0.01).
Regression of phenotypic variation on genetic ancestry (taking Native American as reference) demonstrates a significant effect for most of the traits examined (p-value <10−3 using a conservative Bonferroni multiple testing correction, Table 2). Among the non-facial phenotypes (accounting for sex, country, age, educational attainment and wealth) higher European ancestry is associated with: increased height, lighter pigmentation (of hair, skin and eyes) (Figure S6), greater hair curliness and male pattern baldness. Hair graying approaches statistical significance (p-value 10−2). Higher African ancestry is associated with: increased height, higher skin pigmentation and greater hair curliness. The proportion of phenotypic variance explained by ancestry is highest for skin pigmentation (19%) followed by hair shape (8%) and color of eyes and hair (4% and 5%, respectively) but at most 1% for the other phenotypes.
We also observed highly significant effects of educational attainment (p-value 3.87×10−13) and age (p-value <2×10−16) on height, with height increasing for individuals born more recently at a rate ∼1 cm every 10 years (Text S3).
Genetic ancestry also has a range of effects on facial features, both in terms of size and shape, after accounting for height and BMI (in addition to the other covariates). Higher European ancestry is associated with reduced eye fold and an overall smaller face (centroid size).
Face size and shape effects were also evaluated through the analysis of all pair-wise inter-landmark distances (Table S3). Amongst these distances, 133 and 2 show significant effects of European and African ancestry, respectively (p-values 10−6 assuming a conservative Bonferroni multiple testing correction; Table S3). The most significant effects of European ancestry (P<10−10) involve mainly distances between landmarks placed on the lips and nose. Face shape variation, independent of size, was assessed via Principal Components (PCs) of procrustes 3D coordinates. Significant effects of European ancestry were detected for PCs 1 and 3–5, while African ancestry impacts on PCs 1, 2 and 4 (Table 2, Text S5 and Figure S3). These 5 PCs account for ∼55% of the variation in face shape captured by the 36 landmarks placed on the facial photographs, with ancestry explaining up to 5% of the variance in PC scores (for PC4). Examination of the correlation between inter-landmark distances and facial PCs, indicates that the highest correlation of distances between landmarks of the lips and nose is with PC4 (results not shown), consistent with this PC showing the largest proportion of variance explained by ancestry (Table 2).
Four ethno/racial categories (“Black”, “White”, “Native” and “Mixed”) are commonly used across Latin America in national censuses and other population surveys. We contrasted genetic ancestry and skin pigmentation (as measured by the melanin index) across these four self-estimated categories for the countries sampled (Figure 2 and Table S4). Within each country there is a gradient of decreasing European ancestry (and increasing pigmentation) for the “White”, “Mixed” and “Native/Black” categories. Across countries, skin pigmentation is relatively uniform within ethnicity categories, except for “Black”. For “White”, “Native” and “Mixed” the mean melanin index across countries varies within ∼2 units, while the range for “Black” is ∼25 units. By contrast, genetic ancestry varies greatly between countries for all ethnicity categories. For example, European ancestry varies across countries by about 40% for “White”, “Mixed” and “Native” and about 20% for “Black” (Figure 2; estimates for African and Native American ancestry are shown in Table S4).
Contrasting self-perceived (ranked into five bands at 20% increments) and genetically estimated continental ancestry we observe a moderate, but highly significant, correlation: America: r = 0.48, P<2.2×10−6, Europe: r = 0.48, P<2.2×10−6, Africa: r = 0.32, P<2.2×10−6. However, there is a trend for higher self-perceived Native American and African ancestry to exceed the genetic estimates (Figure 3). Similarly, there is a trend for lower self-perceived Native American and European ancestry to underestimate the genetic ancestry (Figure 3). To explore these trends further we performed a multiple linear regression of the difference between self-perceived and genetically estimated ancestry (i.e. the bias, see Methods), using genetic ancestry and covariates as predictors (Table 3). As expected, we observe that genetic ancestry has a highly significant effect (<2×10−16 for all ancestries) and the negative sign of the regression coefficients reflects the orientation of bias seen in Figure 3. At increasing European genetic ancestry, there is greater underestimation in self-perception (a more negative bias). By contrast, with increasing African genetic ancestry there is less overestimation (less positive bias). For Native American ancestry, there is an overestimation (positive bias) at low levels, and an underestimation at high levels of ancestry (negative bias).
Most of the phenotypic traits that show ancestry effects (Table 2) also have a significant effect on self-perception bias (Table 3). There is a particularly strong effect of pigmentation: individuals with lower skin pigmentation tend to overestimate their European ancestry while individuals with higher pigmentation overestimate their Native American and African ancestries. Similarly, lighter eye and hair color lead to an overestimation of European ancestry and an underestimation of Native American ancestry (but not African ancestry). Hair type is strongly associated with an overestimation of African ancestry. Marginally significant associations are seen with other phenotypes, including facial features such as eye fold (leading to an underestimation of European ancestry) and landmark coordinate PCs (Table 3). An effect of social factors on perception bias is evidenced by the observation that greater wealth is significantly associated with an overestimation of European ancestry and that there is significant variation in bias between countries (Table 3). We examined the impact on these results of the uncertainty associated with the ancestry estimates by repeating the regression analyses using ancestry estimates obtained with a subset of 15 markers (Methods). We found that the same covariates had significant effects and that the regression coefficients were not significantly different in the two sets of regression analyses.
Since the late 15th century, the population of what is now called “Latin America” has undergone major demographic changes within the context of a highly diversified physical and social environment [6], [23]. These changes include the occurrence of waves of immigration from various parts of Africa and Europe, the resulting decline of the Native populations most exposed to the immigrants and a variable admixture between these groups. There have also been a number of noticeable population movements in the region. For example, in recent generations there has been an extensive migration to the cities, Latin America now being the most urbanized region of the world (about 80% of its population is currently considered urban) [24]. Three of the countries we sampled (Brazil, Mexico and Colombia) are the most populous in the region and the combined population of the five countries examined here account for ∼70% of Latin Americans. Although ours is a convenience sample, the individuals studied show considerable variation in birthplace and for a range of biological and social variables, illustrating the extensive heterogeneity of Latin Americans.
The interpolated ancestry maps obtained (Figure 1) are consistent with other genetic studies [20], [21], [25], [26] and with census information on the distribution of the main ethnicity groups within each country (available at www.ine.cl; geoftp.ibge.gov.br; www.igac.gov.co; www.censo2010.org.mx, www.indepa.gob.pe). Altogether, these data underline the extensive genetic structure existing within and between Latin American countries. It is possible to relate this genetic heterogeneity to well documented historical factors [6], [23], [27]. Broadly, Native American ancestry is highest in areas that were densely populated in pre-Columbian times (particularly Meso-America and the Andean highlands) as well as in regions that received relatively little non-native immigration and which currently have relatively low population densities (e.g. Amazonia). During the colonial period Africans were brought to Latin America as forced labour mainly to coastal tropical areas, particularly in the Caribbean and Brazil [28]. That country was the main recipient of African slaves in the region (representing about 40% of all African slaves brought to the Americas [29]). Early (mostly male) Iberian immigrants settled across the continent, admixing extensively with Native Americans and Africans [5]. These were followed by further currents of European immigration, including individuals from various parts of Europe (often arriving as a result of governmental initiatives) and resulting in the dense settlement of specific geographic regions (such as the south of Brazil). The larger variance in individual ancestry observed for larger urban centres is consistent with the increasing urbanization of Latin America seen recent generations, the cities absorbing immigrants with diverse genetic backgrounds. Other than demographic history, it is possible that assortative mating has also contributed to shaping population structure across Latin America. The Iberian “Conquest” (i.e. the first century of settlement) was characterized by extensive admixture between Natives and immigrants (driven by the highly predominant immigration of males) [5]. However, during the subsequent colonial period society became increasingly stratified, including the instauration during the 18th century of a caste system regulating marriages [6], [27]. These restrictions were mostly abolished with the establishment of republican governments in the 19th century [6]. However, a number of studies have documented continuing assortative mating in Latin America, in relation to genetic ancestry, physical appearance and a range of social factors [30]–[34].
The pattern of variation we observe between physical appearance and genetic ancestry is consistent with information on the variation in frequency of the traits examined in Native Americans, Europeans and Africans. Constitutive skin pigmentation (i.e. in areas not exposed to light), hair and eye color and hair type are traits with little environmental sensitivity and show large differences between continental populations [35]. As expected, increased European ancestry shows a highly significant association with lighter skin, hair and eye pigmentation. A number of allelic variants impacting on these traits have been identified in Europeans and certain of these show large allele frequency differences between Europeans and non-Europeans [1], [2], [36]. We also found a highly significant effect of ancestry on hair type, individuals with higher Native American ancestry showing greater frequency of straight hair, a phenotype that is essentially fixed in Native Americans. Recent studies in East Asians implicate a p.Val370Ala substitution in the EDAR gene in hair morphology [37]–[39]. One of the ancestry informative markers typed here (rs260690) is located in the first intron of EDAR, is in high linkage disequilibrium with the p.Val370Ala variant in the HapMap dataset and is strongly associated with hair type in our sample, after accounting for ancestry (Text S4), suggesting that variants at EDAR could be impacting on hair morphology in Latin Americans. Greater European ancestry also correlates significantly with higher rates of male balding and (marginally) with hair greying (our sample is perhaps underpowered to detect these effects due to its relatively young age; Table 1). Although no thorough comparative data is available, classical population studies indicate that hair greying and androgenetic alopecia are rarer, less severe and of later onset in Native Americans than in other continental populations [40] and our data points to the existence of loci influencing the continental distribution of these traits. Studies in Europeans have recently identified loci associated with androgenetic alopecia [41], [42], but no similar analyses have been performed for hair greying.
Recent genome-wide association analyses in Europeans have implicated loci for variation in height and related anthropometric traits [43], [44]. However, these traits are also strongly influenced by environmental factors, including nutrition [45]. In the sample studied here we find that Native American ancestry correlates significantly with lower height and we also detect a significant effect of socioeconomic position (Text S3), lower socioeconomic position correlating with decreased height. The significant effect of age on height, with younger individuals tending to be taller than older ones suggests that the two socioeconomic indicators examined here (education and wealth) capture only part of the environmental variation impacting on height. The rate of increase in height for individuals born more recently (∼0.1 cm/year) estimated here is similar to that obtained from extensive longitudinal surveys in Latin America (∼1 cm per decade in the last century), an observation that has been interpreted as resulting from the historical improvement in living standards across the region [45], [46]. It is thus possible that the ancestry effect on height that we detect could be influenced by environmental factors that correlate with ancestry that are not captured by the socioeconomic variables examined here.
The ancestry effects that we detect for facial features (eye fold, face shape and size), but not for head circumference, agree with the notion of a greater developmental and evolutionary constraint on neuro-cranium than on facial variation. This is also in line with proposals that human facial features include a range of environmental adaptations [47]–[49]. Aspects of face shape variation captured by principal components analysis that are influenced by genetic ancestry include mainly, width and height of the face, facial flatness, position of the glabella and fronto-temporal points, extent of eye fold and the relative size and position of lips and nose (a fuller description of face shape variation associated with each PC is presented in Text S5 and Figure S3). Two genome-wide association scans in Europeans have identified a few loci associated with aspects of face shape [50], [51] but these results are pending confirmation by further studies. No genetic variants have yet been implicated in intercontinental differentiation for facial features.
Our joint analysis of genetic, phenotypic and self-perception variation emphasizes the strong impact of physical appearance on self-perception. Comparison of skin pigmentation across self-perceived ethno/racial categories shows remarkable consistency between countries, underlining the weight given to this trait in self-perception [52]. The large variation in genetic ancestry between countries for each ethnicity category illustrates the relatively low predictive power of physical appearance for genetic ancestry. Although we detected highly significant effects of ancestry on many of the phenotypes examined, the observed correlations are relatively low (Table 2). The poor reliability of physical appearance as an indicator of genetic ancestry likely relates to the impact of environmental variation on some of these traits, and to their specific genetic architecture. Particularly, a few genetic variants could have relatively large phenotypic effects (as documented for pigmentation [2], [36]). The impact of physical appearance on self-perception of ancestry likely relates to admixture in Latin America largely occurring many generations ago and the frequent unavailability of reliable genealogical information. The contrast between self-perceived and genetically estimated admixture proportions confirms the impact of physical appearance on self-perception and shows how certain traits, particularly but not exclusively related to pigmentation, can bias self-perception of ancestry. This biased perception of physical attributes is likely to be influenced by social and individual factors shaping the interpretation of phenotypic variation. The effect of such factors is illustrated by the observation of differences in bias across countries and the positive correlation between wealth and European ancestry (Table 2). An effect of wealth on self-perception of ancestry has also been the subject of study in the sociological literature on Latin America [52].
In conclusion, our study sample illustrates the extensive geographic variation in genetic ancestry seen across Latin America, reflecting the heterogeneous demographic history of the region. The highly significant impact of genetic ancestry on physical appearance is consistent with some of the phenotypic variation seen in Latin Americans stemming from genetic loci with differentiated allele frequencies between Africans, Europeans and Native Americans [53]. Further analysis of the study sample collected here should enable the identification of such loci. The significant correlation between self-perceived and genetically estimated ancestry is consistent with the observed effects of genetic ancestry on physical appearance. However, self-perception is biased, possibly due to non-biological factors affecting the perception of phenotypic variation and to the genetic architecture of physical appearance traits. Our findings exemplify the informativity of Latin America for studies encompassing genetic, phenotypic and sociodemographic information and the interest of a multidisciplinary approach to human diversity studies.
Recruitment took place mainly in five locations: México City (México), Medellín (Colombia), Lima (Perú), Arica (Chile) and Porto Alegre (Brazil). With the exception of Chile, most subjects recruited in these cities were students and staff from the universities participating in this research. In Chile about 2/3 of the subjects recruited were professional soldiers. In Brazil ∼10% of samples were collected in smaller towns of the states of Rio Grande do Sul, Bahia and Rondonia. Adult subjects of both sexes were invited to participate mainly through public lectures and media presentations. Maps showing the number of volunteers in each unique birthplace are presented in Figure S4. Being a convenience sample, the main collection sites are overrepresented on these maps for each country. We obtained ethics approval from: Escuela Nacional de Antropología e Historia (México), Universidad de Antioquia (Colombia), Universidad Perúana Cayetano Heredia (Perú), Universidad de Tarapacá (Chile), Universidad Federal do Rio Grande do Sul (Brazil) and University College London (UK). All participants provided written informed consent. Blood samples were collected by a certified phlebotomist and DNA extracted following standard laboratory procedures.
A physical examination of each volunteer was carried out by the local research team using the same protocol and instruments at all recruitment sites. We obtained: height, weight, head, hip and waist circumference, cheilion-cheilion width and sellion-gnation height. All measurements were obtained in duplicate and the mean of the two measurements retained for further analyses. We recorded eye colour into five categories (1-blue/grey, 2-honey, 3-green, 4-light brown, 5-dark brown/black), and natural hair colour into four categories (1-red/reddish, 2-blond, 3-dark blond/light brown or 4-brown/black). Balding in males was recorded using a modified Hamilton scale as: 0) no hair loss, 1) frontal baldness only, 2) frontal hair loss with mild vertex baldness, 3) frontal hair loss with moderate vertex baldness, and 4) frontal hair loss with severe vertex baldness. Similarly, graying was recorded along a five point scale: 0) for no greying, 1) for predominant non-graying, 2) for ∼50% graying, 3) for predominant greying and 4) for totally white hair. Due to the small number of individuals in categories 1–4 for male pattern balding and greying, we pooled these categories so as to contrast only two categories (presence or absence of the trait). Macroscopic hair type was categorized by visual inspection as 1-straight, 2-wavy, 3-curly or 4-frizzy. A quantitative measure of constitutive skin pigmentation (the Melanin Index) was obtained using the DermaSpectrometer DSMEII reflectometer (Cortex Technology, Hadsund, Denmark). Measurements were obtained from both inner arms and the mean of the two readings used in the analyses.
Five digital photographs of the face: left side (−90°), left angle (−45°), frontal (0°), right angle (45°), right side (90°) were taken from ∼1.5 meters at eye level using a Nikon D90 camera fitted with a Nikkor 50 mm fixed focal length lens. The frontal facial photographs were used to score (by visual inspection) the presence of an eye fold along the upper eye lids using a three point scale: 0) absence 1) partial (interior, middle or outer fold) and 2) full (along the entire eye lid).All photographs were annotated manually with 36 anatomical landmarks and 3D landmark coordinates extracted using the software Photomodeler (http://www.photomodeler.com/ Eos Systems Inc, Vancouver, Canada) (Figure S1). Landmark configurations were superimposed by Generalized Procrustes Analysis [54] and Principal Components (PCs) of the 3D landmark coordinates extracted using the software MORPHOJ [54]. To ease visualization of the 3D shape changes associated with each PC we obtained deformation surfaces via a thin plate spline algorithm.
A structured questionnaire was applied to each volunteer. We obtained information on two indicators of socioeconomic position (Text S2). The first indicator is highest education level attained, categorized as: (1) none/primary/technical, (2) secondary and (3) university and post-graduate. The second indicator is a wealth index obtained from a list of items used to assess living standards. These items were: home ownership, number of bathrooms at the place of residence, ownership of household items (cars, bicycles, fridge/freezer/dishwasher, TVs, radios, CD/DVD players, vacuum cleaner, washing machine) and availability of domestic service. We used polychoric principal component analysis to examine the variability of each country sample and retained the first principal component as an indicator of wealth. To allow comparisons across countries we converted an individual's wealth score to decile within each country.
The questionnaire included items exploring self-perception of ethnicity in the categories: “Black”, “Native”, “White” and “Mixed”, and self-perception of African, European, and Native American ancestry proportions. This was explained as a personal estimation of the proportion of ancestors that had a particular continental origin. We proposed a five point scale, expressed in 20% per cent brackets (and in words): 1) 0–20% (none or very low), 2) 20–40% (low), 3) 40–60% (moderate), 4) 60–80% (high) and 5) 80–100% (very high or total). The questionnaire also recorded information on the place of birth of the volunteer.
In order to select 30 markers highly informative for estimating African/European/Native American ancestry, we started from the list of 5,000 markers, highly informative for world-wide continental ancestry estimation, identified by Paschou et al (2010) [18] using the approach of Rosenberg et al. (2003) [55] based on the worldwide CEPH-HGDP cell panel genotyped with Illumina's Human610-Quad beadchip (including data for about 600,000 SNPs [56]). The full list of these 5,000 markers is at: http://www.cs.rpi.edu/~drinep/HGDPAIMS/WORLD_5000_INFAIMs.txt. Of these, allele genotype data is available in Native Americans for 3,848 markers [57], of which 2,392 have been placed on subsequent Illumina bead-chip products. This subset of markers was retained for selection of those to be typed here so as to facilitate subsequent data comparison and integration. We ranked these 2,392 markers based on allele frequency differences in European-Native American or European-African samples. Amongst markers with the highest inter-continental allele-frequency differences we selected those with lowest heterozygosity in Native Americans (so as to reduce the effect of variable allele frequencies between Native Americans on ancestry estimation). Of the final set of 30 markers retained, 13 are monomorphic in 408 Native Americans (from 47 populations from México Southwards [57]), the rest have minor allele frequencies ranging from 0.01 to 0.15 (median 0.06) in that group of populations. The list of markers typed is provided in Table S1. Genotyping was carried out by LGC genomics (www.lgcgenomics.com/). In a sample of Colombians recently included in a genome-wide association study that used Illumina's 610 chip [19], this set of 30 markers produced individual ancestry estimates with correlations of ∼0.7 (for all the three ancestries) compared with ancestry estimates obtained using an LD-pruned set of 50,000 markers from the chip data, and identical mean estimates. We compared the accuracy of these estimates with estimates obtained using markers from the list of 446 proposed by Galanter et al. (2012) [58], specifically for studying admixture in Latin Americans. From this list, 152 markers are present on Illumina's 610 chip (i.e. ∼5 times the number of markers that we used) and produced estimates with correlations of ∼0.85 with the ancestry estimates from the 50,000 marker set. By contrast, when the set of markers we selected was reduced to 15, the resulting ancestry estimates had a correlation of ∼0.6 with the 50,000 marker set estimates, again showing that there is a diminishing return in accuracy when one increases the numbers of SNPs used in ancestry estimation.
Individual African, European and Native American ancestry proportions were estimated using the ADMIXTURE program [17] using supervised runs where African, European and Native American reference groups (K = 3) were provided (see below). Unsupervised runs at K = 3 produced very similar estimates (Figure S2), confirming our choice of ancestry-informative markers and parental populations. Standard errors of the individual ancestry estimates were obtained by bootstrap using the program's default parameters (200 replication runs). Data from a total of 876 individuals sampled in putative parental populations were used in ancestry estimation and specified in the supervised ADMIXTURE runs. These were selected from HAPMAP, the CEPH-HDGP cell panel [56] and from published Native American data [57] as follows: 169 Africans (from 5 populations from Sub-Saharan West Africa), 299 Europeans (from 7 West and South European populations) and 408 Native Americans (from 47 populations from México Southwards). The full list of the putative parental population samples (and their sizes) is provided in Table S2.
The birthplace names of all individuals was consolidated into a list of unique locations organized into three fields: city/municipality, region/state and country. Geographic coordinates (and altitude) for each placename were obtained via the Google Maps Geocoding API (https://developers.google.com/maps/documentation/geocoding/). The GeodesiX software (http://www.calvert.ch/geodesix/) was used for the geocoding query. We used the Global Rural-Urban Mapping Project version 1 data set (GRUMPv1; http://sedac.ciesin.columbia.edu/data/set/grump-v1-settlement-points) to attribute census size to these localities (see Supp. Text S6) [59]. We use census data for 1990, as the median age in our country samples ranges between 20 and 25.
Geographic maps displaying spatial variation in individual admixture were obtained with Kriging interpolation using the software ArcGis 9.3 (http://www.esri.com/software/arcgis). The cartographic database was geo-referenced to the SIRGAS geodesic system (Geocentric Reference System for the Americas, www.ibge.gov.br/home/geociencias/geodesia/sirgasing/index.html) using a Universal Transverse Mercator projection. Corel-DRAW X3 (Corel Corporation, Ottawa, Canada) was used to edit the map images. When a geographic location had multiple data entries (i.e. volunteers), the Kriging interpolation scheme uses the mean ancestry at that location. The correlation between the standard deviation of individual ancestry variation (at locations with more than 10 samples) and census size was tested using Spearman's rank correlation (as population sizes are generally non-normal but rather distributed exponentially). Statistical significance was obtained via permutation of individual birthplaces.
We tested the null hypothesis of ancestry being spatially uniformly distributed using two approaches. Firstly, we obtained Moran's ‘I’ index for each ancestry component (African, European, Native American) separately. This index tests for spatial uniformity of a variable using standard autocorrelation models and we evaluated significance by permuting birthplace locations for every individual maintaining constant the number of individuals sampled per location. To assign a single value to each location we used the average ancestry, recalculating this average after every permutation. Secondly, we used canonical correlation analysis. A disadvantage of Moran's method is that the three ancestry variables are not independent, complicating the interpretation of p-values. Canonical correlation allows one to combine the three ancestries into a single variable: it is the maximal correlation between two sets of linear combinations of multiple variables. In our case, the three ancestries constitute one set and the geographical coordinates (latitude & longitude) constitute the second set. Adding quadratic and cubic powers of the geographic coordinates improved the fit, consistent with the curved shape of the ancestry gradients and the existence of regions with markedly different ancestry. Adding a fourth power did not improve the fit any further. P-values were obtained by permutation as above.
To evaluate the effect of ancestry on phenotype we used multivariate regression models including basic covariates (age, sex, country, education, wealth, and optionally BMI and height). Depending on the trait we used multiple linear (for continuous and ordinal traits) or logistic (for binary traits) regression. The categorical traits in Table 2 were considered ordinal variables (converted into four or five integer levels as specified in Table 1). The justification for doing so is the convention that for an ordinal variable with several categories there is little difference in fitting a linear regression model or an ordered probit model [48]. This is true because for these traits we can assume an underlying continuous variable (for eye or hair colour it could be the amount of pigment, for hair shape it could be the curvature of hair). Since an underlying continuous variable converted into ordered categories is the main assumption for the development of a probit model, this similarity in the two analysis holds. We verified this by examining both models and verifying that the results are similar.
Regression results corresponding to the ancestry variables are presented in Table 2 along with R2 from this full model. A baseline regression model with only the covariates was also performed, leaving out ancestry, and the difference in R2 in the two models was taken to be the proportion of variance in the phenotype explained by ancestry. Standard errors of the individual ancestry estimates (provided by the ADMIXTURE software) were incorporated in the multivariate regressions via the errors-in-variables model [49]. This adjusts the estimated regression coefficients and p-values for all covariates. The error in estimating a variable generally leads to an underestimation of the regression coefficients. However, the p-value still approaches zero under the alternative hypothesis, provided samples sizes are sufficiently large. For the ancestry estimates, the error in estimation was relatively low (∼1–5%), consequently for our large sample sizes the reduction in effect size for each variable was modest (∼5–10%).
To evaluate the relationship between self-perceived and genetically estimated ancestry we performed a bias analysis. This bias was defined as self-perception minus the estimated genetic ancestry. Overestimation therefore means that self-perception exceeds the genetic estimate, while underestimation indicates that self-perception is lower than the genetic estimate. Each genetic ancestry estimate was obtained as a percentage (proportion), while self-perception was recorded into five bands at intervals of 20%. The bias in self-perception was therefore considered zero if the percentage of genetic ancestry fell within the chosen self-perception interval. Otherwise, bias was measured to be the distance of the closest boundary of the self-perception interval to the genetic ancestry percentage. We then performed multivariate linear regression of the bias on the genetic ancestry estimates and other variables (Table 3). The advantage of analysing the bias is that the regression model is easily interpretable. If self-perception was accurate (bias of zero) all the regression coefficients would be non-significant. If the bias is non-zero and some variables show significant effects, the signs of the coefficients are interpretable as leading to overestimation (positive coefficients) or underestimation (negative coefficients) of ancestry, as indicated above.
All statistical analyses were performed using R (www.r-project.org) [60] or MATLAB [61].
|
10.1371/journal.pbio.0060238 | Regulatory Hotspots in the Malaria Parasite Genome Dictate Transcriptional Variation | The determinants of transcriptional regulation in malaria parasites remain elusive. The presence of a well-characterized gene expression cascade shared by different Plasmodium falciparum strains could imply that transcriptional regulation and its natural variation do not contribute significantly to the evolution of parasite drug resistance. To clarify the role of transcriptional variation as a source of stain-specific diversity in the most deadly malaria species and to find genetic loci that dictate variations in gene expression, we examined genome-wide expression level polymorphisms (ELPs) in a genetic cross between phenotypically distinct parasite clones. Significant variation in gene expression is observed through direct co-hybridizations of RNA from different P. falciparum clones. Nearly 18% of genes were regulated by a significant expression quantitative trait locus. The genetic determinants of most of these ELPs resided in hotspots that are physically distant from their targets. The most prominent regulatory locus, influencing 269 transcripts, coincided with a Chromosome 5 amplification event carrying the drug resistance gene, pfmdr1, and 13 other genes. Drug selection pressure in the Dd2 parental clone lineage led not only to a copy number change in the pfmdr1 gene but also to an increased copy number of putative neighboring regulatory factors that, in turn, broadly influence the transcriptional network. Previously unrecognized transcriptional variation, controlled by polymorphic regulatory genes and possibly master regulators within large copy number variants, contributes to sweeping phenotypic evolution in drug-resistant malaria parasites.
| Development of the malaria parasite, Plasmodium falciparum, in the blood is driven by a number of different genes expressed at different times and at different levels. Exactly what influences such transcriptional changes remains elusive, particularly in regard to important phenotypes like drug resistance. Using cDNA microarray hybridizations from the progeny of a Plasmodium genetic cross, we mapped gene expression quantitative trait loci (eQTLs) in an experimental population of malaria parasites. Each gene's transcript level was used as a segregating phenotype to identify regions of the Plasmodium genome dictating transcriptional variation. Several regulatory hotspots controlled the majority of gene expression variation, mostly via trans-acting mechanisms. One, influencing the largest number of transcripts, coincided with an amplified region of the genome traditionally associated with multiple drug resistance (MDR). Overall, integration of two functional genomic tools (gene mapping and transcript quantitation) has revealed a system-wide rewiring of the parasite transcription network: pleiotropic phenotypic variation, driven by drug selection on genome structure that may be attributed in large part to adaptive copy number polymorphisms in the parasite. These structural variants alter the expression of genes within the amplicon as well as many genes scattered across the Plasmodium genome.
| Plasmodium falciparum is an apicomplexan parasite that causes the most severe and lethal form of human malaria. Parasites isolated from patients across the globe exhibit a wide range of phenotypic variation, including drug responses, growth rates, and a variety of virulence factors. Until recently, inter-strain variation had been studied primarily at the DNA sequence and phenotype levels. Since the P. falciparum genome was fully sequenced [1], several large-scale gene expression studies [2–8] have provided the malaria research community with detailed insights into gene expression across the parasite's life cycle. The P. falciparum transcriptome is expressed as an unusual continuous cascade across the distinct stages of the parasite's erythrocytic cycle [4]. However, the lack of traditional DNA-binding proteins raises important questions regarding the nature of transcriptional regulation, characteristically dictated by specific transcription factors in other eukaryotic systems [9,10]. While malaria parasites must have a complex regulatory architecture to control the precise waves of gene expression during erythrocytic development, it is not known whether natural allelic diversity in this species includes variations in the regulatory network itself. P. falciparum is not amenable to many of the standard tools employed to study model organisms. However, combinations of the few available genome-wide methods, including classical genetics, offer novel opportunities to dissect layers of regulatory complexities such as DNA copy number variation (CNV) and transcription.
Transcription in malaria parasites is rigidly programmed through the erythrocytic cycle and largely unresponsive to specific perturbation [11,12]. Recently, it was concluded that the parasite “lacks ubiquitous heritable transcriptional variation” [13], based primarily on recent work comparing gene expression profiles between three unrelated, lab-adapted parasite strains (i.e., parasite clones): 3D7, HB3, and Dd2 [6]. If true, it follows that divergent phenotypes between strains, such as drug resistance, do not result from variation in the transcriptional profile. This interpretation runs contrary to small datasets on direct co-hybridization of cDNA from malaria parasites [3], recent evidence that malaria parasites display distinct physiological states in their in vivo transcriptional profiles [14], and numerous observations that modifications in transcriptional regulation underpin complex adaptations in human, fly, worm, and yeast clonal populations [15–18].
Genetic mapping of genome-wide RNA levels as traits [19,20] has been applied to a variety of organisms to map regulators of transcription [16,21–27]. As has been observed for other classical phenotypes, transcript levels are inherited as complex traits that can be mapped to their causal genetic polymorphisms and, consequently, can broaden our understanding of the regulatory mechanisms underpinning adaptation to fluctuating environments.
Mapping the regulatory determinants of gene expression can be particularly useful in malaria parasites. The P. falciparum genome encodes much of the basal eukaryotic transcriptional machinery, including RNA polymerase II [1], and putative orthologs to general transcription factors have been identified [28]. However, for the most part, the functional roles of even these standard transcriptional regulatory components have not been experimentally confirmed. Comprehensive searches of Plasmodium spp. proteomes for specific transcription factors have yielded few candidates [29], with the exception of the identification of members of a candidate transcription factor ApiAP2 gene family [30]. A recent study by De Silva et al. (2008) [31] provides empirical support for ApiAP2 transcription factors and their cognate binding sequences as a potential source of developmental regulation in this species. The extremely AT-rich (> 80%) P. falciparum genome may obscure signatures of upstream or downstream regulatory motifs (e.g., [32]), as well as regulatory proteins, increasing the difficulty of identifying regulatory determinants in the genome. To compliment in silico searches for potential regulatory domains, an unbiased genetic approach can reveal components of the seemingly unique transcriptional regulation network in this important eukaryotic pathogen.
By mapping genome-wide variations in transcript levels, it is possible to describe a broad architecture of regulatory variation. Mapping transcript level traits to gene expression quantitative trait loci (eQTL) identifies “local” or “distant” genetic contributions for which a regulatory polymorphism resides near the target transcript's gene, or the regulatory variation is displaced from the gene's position, respectively [13]. On the genome scale, many different transcripts mapping to a single distant eQTL suggests a transcriptional regulator with multiple targets. Multiple expression traits that map to a common local eQTL point to either a local cis mechanism (e.g., polycistronic-like transcription) or a local structural determinant of gene expression (e.g., chromatin organization or a CNV such as a chromosomal amplification or deletion that alters the dosage of genes in a given locus).
Traditional quantitative trait loci (QTL) mapping in the HB3 × Dd2 genetic cross of malaria parasites identified the major candidate gene responsible for chloroquine resistance [33] and multiple QTL and candidate genes contributing to quinine susceptibility [34]. HB3, isolated from Central America (Honduras), represents a “wild-type” parasite sensitive to the quinoline line of drugs [35], while Dd2 was derived from a Laotian patient in whom chloroquine (CQ) therapy failed [36]. Dd2 is also resistant to pyrimethamine, as is typical of multidrug resistant (MDR) parasites. Moreover, prior to its use in the genetic cross, Dd2 was further selected in the laboratory for resistance to mefloquine (MQ) [37,38]; consequently the Dd2 genome has been reshaped by sequential drug selections and carries a genetic signature of, and effectively models, southeast Asian MDR parasites. The progeny from the HB3 × Dd2 genetic cross present a unique population in which to study the effects of drug selection on the parasite genome and its impact on gene expression. Understanding how a parasite's drug selection history has impacted its genetic plasticity has been the focus of persistent research [39] to devise sustainable drug therapies.
Here, we use genome-wide expression profiling and linkage analysis in the segregating population of P. falciparum derived from the Dd2 × HB3 genetic cross to locate regions of the genome contributing to heritable levels of transcriptional variation that distinguish parasite strains. We show that transcript level variation is strongly influenced by parasite genotype, and the controlling eQTLs are distributed throughout the genome. Regulatory loci are observed both proximal to and distant from the genes they regulate. They reveal the profound influence of structural chromosomal polymorphisms underlying the identified eQTLs. Overall, we have taken the first step toward understanding how groups of genes are co-regulated. By combining genomic methods with classical genetics, we illuminate previously unrecognized transcriptional complexity and variation in P. falciparum, including a prominent role for drug selection and CNVs in shaping the regulation of transcript levels and their downstream constituents.
Using DNA microarrays, relative gene expression levels were measured across a genetic dimension in 34 progeny from a genetic cross of two laboratory adapted parasite strains. A high-resolution linkage map is available for this cross, and previous efforts have led to the mapping of genes involved in simple and multi-gene drug resistances [33,35,40]. For each progeny clone, relative transcript levels were determined for 7,665 probes representing 5,150 putative genes (open reading frames [ORFs]). Experimental precision derives from nested biological replication: each parental allele for each probe was represented on approximately 18 different microarrays given the predominant 1:1 mendelian segregation of markers genome-wide [40]. Malaria parasites are haploid during the erythrocytic stages and the impact of genotype on phenotype is more direct than for organisms in which influence from heterozygous loci must be taken into consideration. Extensive expression level polymorphisms (ELPs) appeared to be segregating in this population based on the range of relative transcript level variation observed for the vast majority of genes (Figure 1A).
Relative gene expression levels, reported as log2(test/HB3 reference), for all 34 progeny clones were used as expression traits for mapping eQTL. Transcript levels that vary due to nongenetic sources, such as biological or experimental noise, would not generate significant eQTL; however, performing the large number of tests required in eQTL studies increases the chance of obtaining a type I error. To account for the 329 microsatellite markers tested in the linkage analysis, 1,000 permutations were conducted for each eQTL scan (see Methods) to establish corrected genome-wide significance levels of p ≤ 0.05 and p ≤ 0.01 for each expression trait. All nominal and corrected p-values are provided in Table S1. A total of 874 ORFs (981 probes) was differentially regulated by 1,063 significant eQTLs at a genome-wide significance level of p ≤ 0.05 (Tables 1 and S1); approximately 18% of all P. falciparum genes displayed a significant genetic component leading to variation among the progeny at 18 h post-erythrocyte invasion (hpi). We also considered a genome-wide significance level of p ≤ 0.01 and found 315 expression traits with significant linkage (Table 1). In addition to correcting for multiple tests across the large number of markers for each expression trait, we computed false discovery rates (FDRs) associated with the genome-wide (corrected) p-value thresholds to account for the testing of 7,665 expression traits (Figure S1). The 981 (p ≤ 0.05) and 315 (p ≤ 0.01) expression traits regulated by eQTLs corresponded to FDRs of 24% and 14%, respectively. Although this study was performed in a non-model organism with relatively few progeny, the observed FDRs are consistent with observations from model genetic systems (e.g., [25,27]).
Nine hundred seven expression traits mapped to a single locus, and 74 traits mapped to multiple loci. Mapping in a small progeny population of 34 individuals might be expected to have limited power to detect multiple QTL for a given trait; however, this methodology was previously implemented for these same progeny to map five QTL contributing to quinine sensitivity and genetic effects accounting for as little as 10.5% of the phenotypic variation [34]. Nevertheless, the 74 traits for which multiple loci were identified could be disproportionately represented in the pool of expected false positives. In the present study, the detection of eQTLs was not obviously biased for particular phenotype characteristics. For example, eQTLs were detected across a range of relative expression levels among the progeny and span the range of transcript abundances (Figure 1A and 1B, blue squares). This illustrates the strength of the segregation filter and the high degree of nested replication among these progeny. Indeed, we detected loci for expression traits that varied by as little as 1.65-fold between the lowest and highest expressing progeny. Most eQTLs were detected for traits whose relative expression levels in the progeny varied 1 ≥ log2(test/HB3 reference) ≥ 2 (see Methods). As observed for the range in relative expression levels in the progeny, a large majority of eQTLs was derived from low abundant transcripts, and there was no skewing of eQTL detection toward the higher-abundance transcripts (Figure 1B, inset). This observation suggests that technical issues, such as labeling efficiency bias, were not influencing the global picture of eQTLs. As expected, traits with the widest range of variation among the progeny were more likely to have a significant eQTL (Figure 1C).
By considering eQTL numbers, positions, and effect sizes, it was possible to gain a sense of the potential regulatory complexity controlling expression traits. Local regulatory variation, wherein the causal locus of differential gene expression overlapped with the gene being regulated, accounted for 23.6% of the observed eQTLs (Table 1). The remaining majority (76.4%) of the eQTLs denoted mutations that regulated distant transcripts. A comprehensive list of genes and their eQTLs detected at different thresholds is provided in Table S2. Of the traits mapping to a single eQTL, 24% were local effects, commensurate with the overall eQTL pool. These were also the strongest genetic effects, as discerned by examination of the p ≤ 0.01 genome-wide eQTL threshold for detection (Table 1). Conversely, of the 74 probes that mapped to multiple loci, the most common (62% of the 74 probes) mapped to at least two distant eQTLs. It should be noted that such patterns of multi-“trans” factor regulation is much more difficult to detect in a small mapping population; therefore, while our data suggest complex “trans” regulators, we were limited to detecting only the largest genetic effects, no doubt underestimating the genetic complexity of transcriptional regulation in this species.
In addition to numbers and effect sizes, the genome-wide distribution of eQTLs can identify the regulatory architecture driving expression variation. Regulatory loci resided on each chromosome, ranging from as few as 17 eQTLs on Chr 6 to 513 on Chr 5 (Figure 2; Table 2). Of the 329 informative positions in the genome defined by recombination in the combined progeny pool [40], 203 loci harbored at least one eQTL, and 122 had multiple eQTLs (Figure 2). The 81 genome positions with a single eQTL (singletons) influenced expression of 32 local genes and 49 distant genes. Local effects were more common than distant effects in the singleton group (40% versus 24% of the total eQTLs, respectively). This could be because local eQTLs contributed to larger genetic effects and were thus more readily detected.
Several loci influenced the expression of a very large number of genes across the P. falciparum genome. Using permutation tests (p ≤ 0.05; n = 1,000), we identified 12 regulatory hotspots (see Methods), each of which drove transcriptional changes of as few as 14 genes or as many as 182 genes (Figure 2; Table 3). These hotspots accounted for 63% of the detected eQTLs. More than half of the eQTL hotspots (seven of the 12) were found on Chr 5 (Table 3). These included the two largest eQTL hotspots that mapped to adjacent positions on Chr 5 (68.8 cM and 65.9 cM, respectively). The remaining eQTL hotspots were on Chr 3 (one hotspot), Chr 7 (two), Chr 9 (one), and Chr 12 (one). eQTL positions are defined with respect to their nearest independently mapped microsatellite marker; however, for any one expression trait, the resolution of a locus is a function of the genetic recombination resolution (i.e., cM distance). Although the high recombination rate (15 kb/cM) in P. falciparum [40] can facilitate physical mapping resolution to within tens of kilobases, it was not possible to know the exact degree to which neighboring loci overlap physically. Some overlap was expected, and this overlap influenced the designation of the total number of hotspots but did not influence the overall numbers of transcripts regulated by the adjacent segments. Consequently, under different criteria, neighboring loci would combine to generate fewer regulatory hotspots with more associated expression traits per hotspot. For example, when regulatory hotspots at adjacent markers were grouped together, the number of hotspots decreased to six, with hotspots 5_0.0 and 5_5.7 (Chr number_cM distance of genetic marker), 5_11.4 and 5_20.0, and 7_20.2 and 7_28.9 coalescing to three hotspots. Because this distribution of hotspots was critical to our interpretation of novel regulatory features of P. falciparum, we further compared the distribution of all eQTLs identified at the p ≤ 0.05 genome-wide significance threshold to those at identified at p ≤ 0.01and found a highly similar pattern of genome-wide eQTL clusters (R2 = 0.9229) and retained seven of the 12 total eQTL hotspots (3_0.0, 5_0.0, 5_11.4, 5_20.0, 5_65.9, 5_68.8, and 12_103.3). Notably, some eQTL hotspots coincided with structural copy number variations that define their probable physical limits (described below).
eQTL hotspots consisting of distant regulatory effects point to pleiotropic regulators [16]. Alternatively, eQTL hotspots composed of many local eQTLs likely result from chromosomal structural events, such as sequence amplifications or deletions that impact the expression of resident genes. Cis- and trans-acting hotspots may coincide if, for example, a deletion includes transcription factors that act at distant sites, i.e., the genes within the deletion will have reduced transcription as will the distant genes they regulate. The vast majority (86%) of the “linked” eQTLs in P. falciparum hotspots regulated the transcription of unlinked genes. Only a single hotspot (Chr 12) was found to correspond to mostly local regulation of transcripts. Furthermore, when considering a reduced eQTL hotspot threshold (ten or more co-mapping expression traits), we uncovered one additional locus with a high proportion of local genes (Chr 2), compared to an additional five loci corresponding to distant gene regulation (Figure 2, inset).
It was of interest to compare the eQTL hotspots with previous comparative genome hybridization studies that identified chromosomal amplifications and deletions in the parental P. falciparum clones. The two eQTL clusters in the present study that predominantly regulated local gene expression (Figure 2, inset, see asterisks) included a Chr 12 hotspot coinciding with a CNV in the underlying DNA segments [41]. Three hotspots aligned with sequence amplification events (eQTL hotspots 5_65.9, 5_68.8, and 12_103.3) [41–43] and one hotspot (9_97.7) corresponded to a deleted segment from the HB3 parent [44]. Four hotspots were linked to loci rich with cytoadherence and highly polymorphic surface antigen genes such as cytoadherence linked asexual protein (CLAG) genes, rifin genes, and var genes (3_0.0, 5_0.0, 5_5.7, and 7_28.9, respectively) [1], of which three were located in the sub-telomeres and one (7_28.9) at an internal (chromosomal) var cluster [45,46]. One hotspot (7_20.2) was observed near the pfcrt drug-resistance locus, and the remaining three (5_11.4, 5_20.0, and 5_48.7) did not coincide with a previously identified CNV, drug resistance locus, or other highly polymorphic region of the genome.
To assess the genetic basis for transcriptional variation mapping to the Chr 5 amplicon (markers 5_65.9 and 5_68.8), we evaluated the relative gene expression levels within the progeny of the HB3 × Dd2 cross (Figure 3A). These loci span the previously reported amplification event involving the multiple drug resistance gene, pfmdr1 (PFE1150w), found in Dd2 but not in HB3 [35,47]. We merged the genes whose transcripts map to these prominent eQTLs into a single set of genes for further analysis, because the posterior probability for the probes mapping to each of these markers generally spanned both markers and the entire amplicon.
Of the 269 transcripts mapping to the amplicon, 85% (228) were expressed at higher levels in the Dd2 parent compared to the HB3 parent; furthermore, these genes were expressed at higher levels in the progeny that inherited the Dd2 alleles at these loci, as expected for an overexpressed positive transcriptional regulator located within the amplicon. The remaining 15% (41) of the transcripts regulated by these hotspots were generally expressed higher in those individual genotypes inheriting the HB3 alleles at these loci. In the simplest scenario, this would argue for a trans-acting, negative regulator located in the amplified region. Clear segregation of the two allelic effects was evident among these transcripts (Figure 3A), demonstrating strong regulation associated with the number of copies of this amplicon carried by individual progeny. We were surprised to observe clear subsets of up-regulated and down-regulated genes due to this amplification. While there is no need to presume that the underlying mechanisms regulating these genes would necessarily lead to solely up-regulation, we considered the possibility that the down-regulated genes were more likely false positives. We observed no statistical difference between the p-values associated with genes regulated in each direction, and also found the same proportion of genes up- and down-regulated at the highly significant p ≤ 0.01 genome-wide eQTLs (unpublished data; see Table S1), supporting the validity of these two regulatory directions of transcription.
In Dd2, the pfmdr1 amplicon contains 14 ORFs [41] and is repeated three times [47] (Figure 3B). This eQTL hotspot indicates that a polymorphism(s) in or linked to the amplicon is regulating genes across the genome; therefore, we closely examined genes in the amplicon for possible transcriptional regulators (Figure 3C). Of the 14 genes, nine are of unknown function, i.e., hypothetical proteins. We used the predicted amino acid sequences for each of the nine hypothetical proteins to search the PfamA database [48] for protein domains characteristic of DNA binding function, a domain function common to transcription factors. Three genes with significant hits to known protein domains were identified: PFE1130w had a hit to a Duf803 domain common to the drug metabolite transporter superfamily of genes (E-value = 0.0001), PFE1135w had a hit to a domain common to iron-sulfur cluster biosynthesis genes (E-value = 2.9 × 10−24), and PFE1145w had hits to two tandem zinc-finger CCCH-type domains (E-value = 0.18 and 0.00084), spaced ten amino acids apart. Two additional genes with unknown function (PFE1095c and PFE1110w) were identified with weaker hits to nucleic acid binding domains (YL1 nuclear protein and transcription initiation factor TFIID 23–30 kDa subunit, respectively). PFE1110w previously has been hypothesized to be a P. falciparum ortholog for TFIID TAF10, a basal transcription factor associated with RNA polymerase II activity in other eukaryotic systems [28]. These three genes, whose proteins encode putative DNA binding domains (PFE1110w, PFE1110w, and PFE1145w), are primary candidate genes with a role in transcriptional regulation requiring experimental validation.
Distant regulatory variation accounted for most of the transcripts mapping to the eQTL hotspots, suggesting possible master regulators residing within the hotspots. To ascertain co-regulation of possible functionally related genes, we sought Gene Ontology (GO) enrichment (Table 3). Focusing on the Chr 5 pfmdr1 amplification hotspot (5_65.9 and 5_68.8), we found that protein modification via the proteasome complex was a prominent, differentially regulated GO category within our experimental population (Table 4). Notably, four additional hotspots contained differentially regulated genes involved in protein modification (eQTL hotspots at 5_48.7, 7_28.9, 9_97.7 and 12_103.3), implying that transcriptional regulators of post-transcriptional machinery may play a prominent role in phenotypic differences between HB3 and Dd2 (Tables 3 and S2). This strong presence of genes involved in protein modification highlights a biological pathway that is alternatively regulated in the parental parasites' genetic backgrounds. Complete results for calculated GO enrichment values for each of the eQTL hotspots are provided (Table S3).
HB3 and Dd2 are known to diverge in the duration of their erythrocytic cell cycle [6,49]; consequently, we evaluated the possibility that eQTLs corresponded to events of the cell cycle. We first binned genes represented on the microarray by their peak expression time in the life cycle using the same microarray platform [4,6]. We then compared these peak-time bins with genes regulated by eQTLs (Figure 4). No bias toward stage-specific gene expression was present in the eQTL pool.
An emerging theme in evolutionary biology recognizes that mutations in regulatory sequences can account for major physiological differences between strains even when coding genes are relatively unchanged [18]. While in numerous species variation in gene expression serves as a storehouse for phenotypic variation [16,21–23,50], it has been argued that P. falciparum is unusual in exhibiting remarkably little heritable, strain-to-strain variation [13]. This conclusion was based on a study by Llinas et al. that compared broad expression “cascades” across the complete erythrocytic cell cycle for three different parasite clones [6], including HB3 and Dd2. Their approach assessed transcriptional profiles by measuring a parasite's transcript levels against pooled samples from the same parasite clone and was not designed to directly assess the relative abundance of transcripts between strains. In contrast, the current study measured relative transcript abundance of the Dd2 parent and each progeny clone against a common reference, HB3, the other parent from the genetic cross. The two studies are not necessarily contradictory, but rather illuminate different features of transcription in this lethal malaria parasite species emphasizing both the robust transcriptional program that has been so well characterized in this species and the subtle but abundant variation that exists between strains. Viewing transcription across a genetic rather than a developmental dimension allows us to tease out variations in transcriptional regulation that could have important implications for the Plasmodium regulatory network and its role in adaptive evolution.
Until now, no study of malaria parasites (to our knowledge) has specifically examined the genetic inheritance of transcriptional variation. More generally, our work illustrates the concept that genome-wide readouts of segregating natural variation point to polymorphic regulatory loci, a finding particularly relevant in light of recent observations by Daily et al. (2007) [14] that transcriptional profiles associated with distinct metabolic states in blood stage forms of malaria are observed in parasites isolated directly from patient blood; these metabolic states are proposed to influence the course of infection and virulence. In the present study involving cultured parasites under controlled laboratory conditions, we find heritable variation in expression levels to be as extensive as that reported for other organisms. We also find transcription levels to be regulated by few, predominantly distant, eQTL hotspots. These co-regulated genes underpin altered biological processes of the regulatory network and provide an evolutionary path to phenotypic change that is distinct from potentially deleterious coding mutations that alter protein function and would thus be poorly tolerated in haploid parasites. Given the relatively small number of progeny available from the HB3 × Dd2 cross, the eQTLs presented probably capture only the largest genetic effects and certainly underestimate the total number and complexity of regulatory polymorphisms.
We find that a large proportion (76.4%) of the genes with variable transcript levels is influenced by distant regulators, dominated by several regulatory hotspots. Distant regulatory variation is often associated with trans-acting transcription factors. While P. falciparum has been described as having a paucity of transcription factors [1], De Silva et al. (2008) recently experimentally validated a family of specific transcription factors and their DNA binding sites [31], suggesting that the machinery for complex regulation in P. falciparum is present yet difficult to discern by standard homology searches. Our data support the presence of at least a few additional potent regulatory factors, and eQTLs can be dissected to locate candidate transcriptional regulating genes in apicomplexans [30]. Regulatory loci can be identified through this method irrespective of the specific biochemical functions (e.g., DNA binding, nuclear localization, or protein phosphorylation) and is therefore uniquely suited to identify loci harboring genetic determinants that act as traditional and atypical or unique modes of genome regulation.
Although most eQTLs identified in this study are trans-acting, the strongest genetic effects are due to local regulatory polymorphisms (Table 1). These polymorphisms are more likely to occur individually than in clusters. Local linkages arise from classical cis-acting mechanisms, e.g., a polymorphism in the gene regulatory region; various other scenarios are also possible, including CNVs, splicing, mRNA decay, regional chromatin structure [23], or even unconventional autologous protein–nucleic acid interactions [51]. Such cis elements would be expected to have strong genetic effects due to the direct molecular control of the transcript levels. We are aware that eQTL mapping has the potential to overestimate local regulatory variation if substantial sequence variation is present between alleles due to reduced hybridization efficiency, effectively mimicking lower gene expression levels [52]. However, for the divergence between HB3 and Dd2, less than 10% of the probes (expression traits) with eQTLs arose from the highly polymorphic antigenic gene families including vars, rifins, and stevors, nearly the same proportion of these genes represented in the overall set of 7,665 probes (10% versus 8%, respectively). In fact, we detected a relative paucity of local eQTLs compared to a major role for trans-acting mechanisms controlling the majority of the observed ELPs in the HB3 × Dd2 genetic cross. We find that genetic effects due to distant regulatory variation are smaller and more likely partnered, indicative of genetic complexity of the regulatory network. Regulatory hotspots have been hypothesized to contain “master regulators” with the effective mutations having pleiotropic effects [24,26], wherein one DNA sequence variant impacts multiple expression traits with potential broad impact on downstream classical phenotypes.
Another striking feature of the regulatory architecture is the prominent role of CNVs in directing transcription variation, particularly from regions previously associated with drug resistance traits. Originally, the HB3 × Dd2 cross was generated to study chloroquine resistance in malaria parasites [35]; fortuitously, the differing geographic and drug selection histories of these genomes represent independent solutions for survival in their respective environments. It is also likely that adaptations to intensive, long-term drug selection by CQ, and subsequent selections by pyrimethamine and MQ, broadly affect biological processes, perhaps pointing to physiological compensation of mutations in drug resistance genes. Our analysis demonstrates GO term enrichment for genes regulated by each of the eQTL hotspots. Genes related to post-transcriptional protein modification are, intriguingly, enriched in six of the 12 eQTL hotspots, including the two hotspots coinciding with the pfmdr1 amplicon on Chr 5 (Tables 3 and 4), suggesting that physiological differences between the parental parasite clones, HB3 and Dd2, may be at least partially due to post-translational modification/regulation of proteins. This is interesting given the recent findings in yeast that protein levels were unchanged in individuals with increased aneuploidy despite increased gene expression levels, suggesting active regulation at the protein level [53]. These functional differences are prime candidate processes driving phenotypic differences derived from the genetic adaptations associated with multiple drug resistance in the Dd2 parent line and may shed light on post-transcriptional regulation of the genome key to adaptation in the malaria parasite.
DNA sequence amplification events, occurring as multiple tandem copies, can influence gene expression both locally and distantly through dosage effects. Previous laboratory drug selection studies demonstrated that parasites exhibited amplification of the genomic region carrying pfmdr1 on Chr 5 that coincided with a loss of sensitivity to quinine and MQ [54]. Additionally, work performed on parasite isolates from MDR populations confirmed amplification of the pfmdr1 locus is the primary contributor for resistance to MQ [55]. The focus of previous studies on this locus has been the pfmdr1 gene itself, and the roles of coding mutations and gene dosage in influencing drug sensitivities [54–56]. Our data emphasize a potentially critical role of neighboring genes, e.g., increased copy numbers of the entire pfmdr1-containing locus not only coincides with decreased sensitivities to common antimalarial compounds but significantly alters gene expression throughout the genome. In addition, genes regulated by the Chr 5 amplicon are both positively and negatively regulated, potentially indicating multiple regulators or multiple mechanisms for gene regulation contained in this locus. Figure 3C illustrates transcriptional regulator candidates, PFE1095c, PFE1110w, and PFE1145w, residing on the pfmdr1 amplicon, including a tandem zinc-finger domain protein (PFE1145w). Notably, eQTL hotspots are not confined to transcription factors but could also point to novel regulatory mechanisms.
The degree to which drug selection on malaria parasites can impact the expression of genes across the genome is evidenced by the sweeping affects the Chr 5 pfmdr1-containing amplicon has on genome-wide gene expression. In general, Chr 5 has been a primary target for genomic adaptations in the HB3 and Dd2 parental parasites; determining whether Chr 5 is unique to this parasite population or whether it is a key player in all natural parasite adaptations is relevant to understanding the propensity of certain parasite clones to rapidly become resistant to multiple drugs. Identifying the mechanisms of transcriptional regulation and the functional relationships among the co-regulated genes will provide crucial information regarding potential antimalarial targets. It is likely that the drug-selection history of Dd2 affecting relatively few “resistance” genes has had a broader impact with significant implications for parasite biology in the form of drug sensitivity modulators, compensatory mechanism in physiology, and/or simple hitchhiking effects that could themselves acquire an adaptive role in subsequent selection.
In light of historical selection bottlenecks and its impact on multi-drug resistance and virulence, even “simple” resistance mechanisms, e.g., the point mutation in the chloroquine resistance transporter gene pfcrt, conferring resistance to CQ, may elicit a complex expression signature. Here, we illustrated how eQTL scans can uncover nontraditional patterns of transcriptional regulation underlying strain-level variation and regulatory hotspots in P. falciparum. This study bolsters a significant and unexpected role for divergent transcription as a source of phenotypic variation and evolution in malaria parasites and elevates a role for structural changes (e.g., CNVs) as having potentially prominent consequences in the cell, perhaps contributing to adaptive evolution.
Parent and progeny parasites (i.e., clones) of the HB3 × Dd2 genetic cross were obtained from the original cloned stocks [35]. The HB3 × Dd2 genetic cross consists of 35 haploid progeny (34 of which were available for this study), mimicking, in effect, recombinant inbred lines for linkage analysis. Each progeny was previously genotyped for the 329 informative microsatellite markers spanning the 14 chromosomes (23 Mb) at a resolution of approximately 17 kb/cM [40]. Parasites were cultivated in human erythrocytes (RBCs) by standard methods [57,58] utilizing leukocyte-free human RBCs (Indiana Regional Blood Center, Indianapolis, Indiana) suspended in complete medium (CM) [RPMI 1640 with L-glutamine (Invitrogen, Carlsbad, California), 50 mg/l hypoxanthine (Sigma-Aldrich, St. Louis, Missouri), and 25 mM HEPES (Calbiochem, San Diego, California); 0.5% Albumax II (Invitrogen), 10 mg/l gentamicin (Invitrogen), and 0.225% NaHCO3 (Invitrogen)] at 5% hematocrit. Cultures were maintained independently in sealed flasks at 37 °C under an atmosphere of 5% CO2, 5% O2, and 90% N2. Parasitemia was monitored and maintained at 5%–7%. Parasites were synchronized using two consecutive 5% sorbitol treatments for two generations, followed by one cell cycle without treatment. The 18 hpi time point was determined by light microscopy. Samples were flash frozen in liquid nitrogen prior to extraction of total RNA.
Total RNA was isolated as previously described using TRIzol reagent [4]. cDNA was synthesized using the Ovation Aminoallyl RNA Amplification and Labeling Kit (cat. # 2101–12; NuGEN Technologies, San Carlos, California) and prepared for hybridization to microarrays as previously described [4]. Microarrays contained 7,665 70mer oligonucleotide probes representing 5,150 Plasmodium falciparum ORFs kindly provided by Dr. Joseph DeRisi (University of California, San Francisco, San Francisco, California). The majority of the oligonucleotides for printing the array corresponded to the Qiagen Operon set (Operon Biotechnologies, Huntsville, Alabama), but some additional sequences matched ORFs in the original P. falciparum sequence reads. The oligonucleotides were printed on polylysine-coated slides using a new generation, ultra fast, linear servo driven DeRisi microarrayer. Slides were post-processed and hybridized in 3× SSC at 63 °C for 12 h as previously described [4]. Slides were scanned in an Axon GenePix 4000B microarray scanner (Axon Instruments, Union City, California) with 532 nm (17 mW) and 635 nm (10 mW) lasers. Data were collected as an image file, gridded, and converted into a text file using Genepix 3.0 software (Axon Instruments). Each parental and progeny cDNA (labeled with Cy5) were hybridized to a common reference sample of the HB3 parental clone (labeled with Cy3), allowing for direct comparisons between the results from each microarray; we tested 34 progeny clones, for a total of 36 hybridizations.
Lowess smoothing of the raw expression data was used to normalize across all microarray slides using TIGR's MultiExperiment Viewer (MeV) v4.0 and Microarray Data Analysis System (MIDAS) v2.19 software (http://www.tm4.org/). Gene expression values are reported as log2(sample/HB3 reference) of each normalized fluorescent signal. The DecisionSite software (Spotfire, Somerville, Massachusetts) was used to generate relative gene expression heat maps.
For each expression trait, interval mapping linkage analysis was performed using the Bayesian approach implemented in the Pseudomarker v2.03 package [59] executed in MATLAB (The MathWorks, Natick, Massachusetts). Marker genotypes were used directly (i.e., imputation of “pseudomarkers” was not necessary given the high-resolution, uniform marker coverage of the P. falciparum genome [40]), and the marker corresponding to the highest LOD score for each expression trait was retained as the eQTL position. The empirical genome-wide significance was determined for each trait by permutations [60,61] in which progeny haplotypes were randomly associated with expression trait values and linkage analysis performed 1,000 times. The nominal p-values for each trait were converted to genome-wide corrected p-values using the permutation distribution of the maximum LOD score for each trait. This approach corrects the nominal p-values for the multiple testing of 329 markers. LOD score thresholds corresponding to p ≤ 0.05 and p ≤ 0.01 genome-wide significance thresholds were determined for each expression trait. To account for the multiple testing of 7,665 expression traits, we followed the work of Petretto et al. (2006) [27] in which we use the lowest corrected p-value for each expression trait to calculate the FDR, using the q value method of Storey and Tibshirani (2003) [62]. This approach was motivated by the fact that the underlying procedure to estimate the FDR works well under “weak” dependence of the considered features. To be conservative, an eQTL was classified as “local” if the gene was located on the same chromosome as the linkage or as due to “distant” regulatory variation if located elsewhere in the genome.
To locate regulatory hotspots within the genome, the total number of expression trait mapping to eQTL at the p ≤ 0.05 genome-wide significance threshold, specifically 1,063, were randomly assigned to the markers (loci) in the genome. The number of linkages observed at the locus with the most linkages was retained and this process was repeated 1,000 times creating a random distribution. From this distribution, linkage hotspots were identified as the number of traits mapping to a given locus exceeding the 95% distribution frequency derived from the 1,000 permutations (i.e., at p ≤ 0.05 for n = 1,000). Genetic markers harboring 14 or more traits with an eQTL were not expected by chance. Genes comprising the groups of genes at the eQTL hotspots were sifted for enriched GO terms to identify possible co-regulated genes contributing to the same biological processes. To obtain statistically significant measurements of overrepresented GO terms, we applied a Fisher's exact test as implemented in GOStat [63]. GO terms significant at p < 0.01 for each regulatory hotspot were retained and reported.
The peak hours of gene expression for all genes represented on our microarrays were collected from prior microarray experiments using the same platform (DeRisi Lab Malaria Transcriptome Database; http://malaria.ucsf.edu/). Genes were binned at 5 h intervals based on their respective peak hour of expression across the HB3 and Dd2 parasites' life cycles. The percentage of genes whose probes showed significant linkage within each of the 5 h bins was calculated. Regression analyses were performed using GraphPad Prism v4.0 (GraphPad Software, La Jolla, California).
Relative transcript abundance was qualitatively estimated for each probe on our microarray by using the sum of median (SOM) intensity from each array, similar to a previous study using the same microarray platform [6]. For each probe, the highest 18 SOM intensity values across all progeny were retained and the median was calculated. Each probe was then binned according to its corresponding relative transcript abundance. The percentage of probes showing significant linkage within each of the relative abundance bins was then calculated.
The range of relative gene expression levels for each transcript was calculated from the progeny with the highest and lowest relative expression levels compared to the reference HB3 samples, excluding outliers. Outliers were determined as those progeny clones expressing at levels ± 2 standard deviations from the mean, calculated for each probe independently. Probes were then binned according to their relative expression level range in the progeny. The percentage of probes with significant linkage within each of the expression range bins was then calculated.
|
10.1371/journal.pntd.0006000 | Ontogeny of the B- and T-cell response in a primary Zika virus infection of a dengue-naïve individual during the 2016 outbreak in Miami, FL | Zika virus (ZIKV) is a mosquito-borne flavivirus of significant public health concern. In the summer of 2016, ZIKV was first detected in the contiguous United States. Here we present one of the first cases of a locally acquired ZIKV infection in a dengue-naïve individual. We collected blood from a female with a maculopapular rash at day (D) 5 and D7 post onset of symptoms (POS) and we continued weekly blood draws out to D148 POS. To establish the ontogeny of the immune response against ZIKV, lymphocytes and plasma were analyzed in a longitudinal fashion. The plasmablast response peaked at D7 POS (19.6% of CD19+ B-cells) and was undetectable by D15 POS. ZIKV-specific IgM was present at D5 POS, peaked between D15 and D21 POS, and subsequently decreased. The ZIKV-specific IgG response, however, was not detected until D15 POS and continued to increase after that. Interestingly, even though the patient had never been infected with dengue virus (DENV), cross-reactive IgM and IgG binding against each of the four DENV serotypes could be detected. The highest plasma neutralization activity against ZIKV peaked between D15 and D21 POS, and even though DENV binding antibodies were present in the plasma of the patient, there was neither neutralization nor antibody dependent enhancement (ADE) of DENV. Interestingly, ADE against ZIKV arose at D48 POS and continued until the end of the study. CD4+ and CD8+ T-cells recognized ZIKV-NS2A and ZIKV-E, respectively. The tetramer positive CD8+ T-cell response peaked at D21 POS with elevated levels persisting for months. In summary, this is the first study to establish the timing of the ontogeny of the immune response against ZIKV.
| Zika virus (ZIKV) is an emerging viral disease that has the potential to negatively impact future generations by causing birth defects in infected pregnant mothers. While there have been many studies performed in animal models of ZIKV infection, there have only been a limited number of reports studying the immune responses in humans. Ricciardi et. al. analyzed the immune response of a primary ZIKV infection in a dengue virus (DENV) naïve individual during the 2016 outbreak in Miami, Florida. B- and T-cell responses were assessed over multiple time points. Cross-reactive antibodies against DENV, a virus that the patient was never infected with, were generated during the ZIKV infection, but these antibodies failed to neutralize any of the DENV serotypes. Furthermore, while these DENV-cross-reactive antibodies might be expected to cause antibody dependent enhancement (ADE) of DENV infection, they did not. Interestingly, ADE of ZIKV infection was seen at approximately 1 ½ months after infection. Together, these results establish the timing of the ontogeny of the immune response against a primary ZIKV infection in a DENV-naïve individual.
| The sudden emergence of Zika virus (ZIKV) cases in the Americas and the growing concern over the birth defects in ZIKV-infected pregnant mothers, led the World Health Organization (WHO) to declare ZIKV infection to be a Public Health Emergency of International Concern (PHEIC) on February 1st, 2016 [1–5]. ZIKV is a member of the Flavivirdae family along with other viruses including dengue virus (DENV), West Nile virus (WNV), and Yellow fever virus (YFV) [6, 7]. ZIKV, DENV, and YFV all share a common vector for transmission, the mosquito Aedes aegypti [8], and autochthonous human ZIKV infection was limited to Africa and mainland Asia until 2007 [4, 9]. Recent and continuous travel of infected humans has spread and established ZIKV infection to the Americas from Micronesia [3, 10, 11]. After infection with any of these flaviviruses, cross-reactive antibody responses are common [12, 13]. The cross-reactive antibody responses associated with primary and secondary DENV infections have been studied in depth [14–16]. The study of the ontogeny of cross-reactive antibodies after primary ZIKV infection is limited in flavivirus-naïve humans, however there have been several studies examining cross-reactive responses of ZIKV and DENV infections at single time points [17, 18].
ZIKV was first reported in the contiguous United States (US) on July 29th, 2016, when the Centers for Disease Control and Prevention (CDC) confirmed four locally acquired ZIKV infections in Miami, Florida (FL) [7, 19, 20]. Local, mosquito-borne ZIKV transmission, however, likely started in FL 2–3 months prior to detection [21]. Due to its warm and humid climate, Miami is conducive to year round breeding of the primary ZIKV vector, Aedes aegypti [8, 11, 22, 23]. This, along with the constant influx of tourists from ZIKV-endemic and ZIKV-naïve populations around the world, will most likely facilitate future ZIKV outbreaks in Miami [21].
Approximately 20% of ZIKV-infected individuals exhibit symptoms, making it difficult to study primary ZIKV infection without the complication of other co-circulating tropical diseases and flaviviruses [3, 24]. ZIKV-infected patients with symptoms often experience a mild febrile illness with fever and a rash, while other less common symptoms include pruritus, myalgia, and retro-orbital pain [3, 24, 25]. These patients will rarely visit a clinic for diagnosis or treatment, leaving the actual number of infected individuals unknown. In cases involving pregnant women, perinatal transmission to the fetus has been documented to result in microcephaly and other fetal complications [26–29]. Moreover, as the virus replicates in the brain tissue of the fetus, a wide array of cognitive developmental symptoms may potentially develop over time. Furthermore, ZIKV-infection has been associated with an increased risk of developing Guillain-Barré syndrome; however, the exact mechanism is not yet fully understood [30–32]. While there are many factors that contribute to disease outcomes, understanding the ontogeny of the immune response to primary ZIKV infection could help the evaluation of diagnostics, vaccines, and other therapies.
Early during the outbreak, we discovered an individual with symptoms consistent with ZIKV infection in Miami. She had no international travel history and was thus suspected of locally acquiring ZIKV. We confirmed ZIKV infection in this individual using RT-PCR from blood and saliva, and we followed the development of her immune response from day (D) 5 post onset of symptoms (POS) to D148 POS. Here, for the first time, we describe a detailed ontogeny of the plasmablast, antibody, and T-cell immune responses in one of the first locally acquired ZIKV infections in the contiguous US.
Research on human subjects was conducted in compliance with DoD, NIH, federal, and state statutes and regulations relating to the protection of human subjects and adheres to principles identified in the Belmont Report (1979). All human subjects were consented in writing and all specimens, data, and human subject research were gathered and conducted for this publication under University of Miami IRB-approved protocol study number 20160127. Written consent was provided for the use of photos from the patient.
Blood samples were collected from volunteer Hu0015, a 34-year-old woman who reported a skin rash that started five days prior to the first blood draw. Plasma and PBMCs were obtained from blood samples collected in ethylenediaminetetraacetic acid (EDTA) tubes, as well as saliva in sterile 50 mL conical tubes, at day (D) 5, 7, 15, 21, 28, 48, 56, 70, 91, 106, 116, and 148 post onset of symptoms (POS). ZIKV infection was confirmed using reverse-transcriptase (RT) PCR assay for ZIKV RNA in the D5 POS saliva sample as well as plasma samples collected at D5 and D7 post onset of the first rash symptoms. No previous history of DENV infection was reported by the volunteer. YFV vaccination status was unknown to the patient, but because she had never traveled outside of the United States, it is unlikely that she was ever vaccinated against YFV. Whole blood was sent to HistoGenetics LLC (http://www.histogenetics.com) for MHC typing.
Viral RNA was extracted using the RNAeasy kit (QIAGEN). ZIKV genome equivalents (GE) were calculated using a qRT-PCR assay targeting the nonstructural protein 5 region (9014–9123 nt), as described [21]. ZIKV sequencing was performed using an amplicon-based approach [21, 33]. Briefly, cDNA was reverse transcribed from 5 μl of RNA using SuperScript IV (Invitrogen). ZIKV cDNA (2.5 μl/reaction) was amplified in 35 × 400 bp fragments from two multiplexed PCR reactions using Q5 DNA High-fidelity Polymerase (New England Biolabs). The amplified ZIKV cDNA fragments (50 ng) were prepared for sequencing using the Kapa Hyper prep kit (Kapa Biosystems) and SureSelect XT2 indexes (Agilent). Agencourt AMPure XP beads (Beckman Coulter) were used for all purification steps. Paired-end 251 nt reads were generated on the MiSeq using the V2 500 cycle kit (Illumina). Demultiplexing was performed by the Illumina instrument. The primer sequences were removed from the reads and bases with Phred quality scores < 20 were removed by Trimmomatic [34]. The reads were then aligned to the complete genome of a ZIKV isolate from the Dominican Republic, 2016 (GenBank: KU853012) using Novoalign v3.04.04 (www.novocraft.com). Samtools was used to sort the aligned BAM files [35]. ZIKV-aligned reads were visually inspected using Geneious v9.1.5 before generating consensus sequences [36]. The consensus sequence for sample Hu0015Sa is available on GenBank (KX832731).
Published ZIKV genomes (195) from the Pacific and the Americas (Asian genotype, from 2013–2016) were retrieved from GenBank and from recent large sequencing projects [21, 37]. The protein-coding sequences were aligned together with the Hu0015 genome using MAFFT [38]. A maximum likelihood phylogenetic tree was reconstructed with RAxML using the general time-reversible (GTR) nucleotide substitution model and gamma distributed rates amongst sites [39, 40]. The phylogenetic tree was annotated using ETE Toolkit [41].
We determined the frequency of plasmablasts in circulation by flow cytometric analysis of PBMCs obtained from blood collected in EDTA tubes and used a Ficoll-Paque (GE Lifesciences) gradient for separation. Briefly, we stained fresh PBMC samples (1 x 106 cells, room temperature, in the dark), with 100 μl of a surface antibody cocktail (S1 Table). We also included the fixable viability dye LIVE/DEAD Fixable Red Dead Cell Stain Kit (Life Technologies) in the staining mix, in order to discriminate between live and dead cells. After 30 min, we washed the cells twice with FACS buffer (PBS, 0.5% FBS, 2 mM EDTA) and resuspended with a PBS 1x solution. Samples were acquired the same day using either a SONY SH800 or a BD FACSAria IIu flow cytometer and analyzed using FlowJo 9 (Tree Star Software).
The DENV1 (strain West Pac74; GenBank U88535.1), DENV2 (strain New Guinea C; GenBank AF038403.1), DENV3 (strain Sleman/78; GenBank AY648961), DENV4 (strain Dominica/8129; GenBank AF326573.1), and ZIKV (strain Paraiba/2015; GenBank KX280026) were propagated in Vero cells (ATCC). Virus stocks were used for both binding virus capture assays (VCA) and neutralization assays as described below. ZIKV (strain PB-81) was gifted from The World Reference Center for Emerging Viruses and Arboviruses (WRCEVA) at The University of Texas Medical Branch (UTMB). ZIKV (strain PB-81) was used for all antibody dependent enhancement experiments and was also propagated in Vero cells.
Antibody and plasma binding was determined in a side-by-side DENV1, DENV2, DENV3, DENV4, and ZIKV VCA ELISA. The ELISA plate was coated with the mouse-anti-flavivirus monoclonal antibody 4G2 (clone D1-4G2-4-15, EMD Millipore) diluted 1:1,000 in carbonate binding buffer and incubated overnight at 4°C. The next day, the plate was washed 5-times with PBS-Tween20 and the wells were blocked with 5% skim milk in PBS for 1 h at 37°C. Following the block, the plate was washed and each virus was added to the corresponding VCA wells, respectively, and incubated for 1 h at room temperature. Subsequently, the plate was washed with PBS and plasma from different time points diluted in 5% skim milk were added to designated wells and incubated for 1 h at 37°C. Following sample addition, plates were washed and detection was carried out using the antibody goat anti-human IgG HRP (SouthernBiotech, 2045–05) diluted 1:10,000, was added to all wells and incubated for 1 h at 37°C. The plate was washed and the wells were developed with the TMB substrate at room temperature for 3–4 min. The reaction was then stopped with the TMB solution, and absorbance was read at 450 nm.
The neutralizing potency of human plasma was measured using a flow cytometry-based neutralization assay (NEUT) [42, 43]. In brief, human EDTA-plasma was diluted and pre-incubated with the reference ZIKV or DENV serotypes in a final volume of 220 μl for 1 h at 37°C. The virus and plasma mixture (100 μl) was added onto wells of a 24-well plate of 100% confluent Vero cell monolayers in duplicate. The inoculum was incubated in a 37°C incubator at 5% CO2 for 1 h with agitation of the plates every 15 min. After 1 h, the virus and plasma inoculums were aspirated and the wells were washed with media. Fresh media was then added and the plates were incubated for a total of 24 h. Cells were trypsinized with 0.5% trypsin (Life Technologies), fixed (Cytofix; BD), and permeabilized (Cytoperm; BD). Viral infection was detected with the 4G2 antibody (clone D1-4G2-4-15, EMD Millipore) recognizing ZIKV or DENV, followed by staining with an anti-mouse IgG2a APC fluorophore-conjugated secondary reagent (clone RMG2a-62; Biolegend). The concentration to achieve half-maximal neutralization (NEUT50) was calculated using a nonlinear regression analysis with Prism 7.0 software (GraphPad Software, Inc.).
Plaque reduction neutralization tests (PRNTs) were conducted as previously described [44]. Briefly, human plasma was serially diluted in OptiMEM supplemented with 2% human serum albumin, 2% fetal bovine serum, and gentamicin. Virus was diluted to a final concentration of approximately 500–1,000 PFU/mL in the same diluent and was added to equal volumes of the diluted plasma and mixed. The virus/plasma mixture was incubated at 37°C for 30 min. Cell culture medium was removed from 90% confluent monolayer cultures of Vero cells on 24-well plates and 100 μl of the virus/plasma mixture was transferred onto duplicate cell monolayers. Cell monolayers were incubated for 60 min at 37°C and overlaid with 1% methylcellulose in OptiMEM supplemented with 2% FBS, 2 mM glutamine, and 50 μg/mL gentamicin. Samples were incubated at 37°C for four days after which plaques were visualized by immunoperoxidase staining, and a 50% plaque-reduction neutralization titer (PRNT50) was calculated.
Human plasma was serially diluted in RPMI media containing 10% FBS and was mixed with either 1.5 x 103 PFU of DENV2 or ZIKV. The plasma and virus mixture, in a total of 50 μl, was incubated for 1 h at 37°C. The mixture was then added to 5 x 103 K562 cells in 50 μl of RPMI containing 10% FBS in a 96-well plate. Three days later, the cells were fixed, permeabilized, and stained with a pan-flavivirus antibody 4G2 as previously described [45]. Samples were analyzed with an Accuri C6 flow cytometer.
15-mer peptides overlapping by 10 amino acids (aa) spanning the whole ZIKV polyprotein were synthesized (A & A/Synthetic Biomolecules). Subsequently, peptides were combined in 10 Mega-pools according to the ZIKV protein from which they were derived as previously described [46]. Pools that elicited an IFN-γ response were subsequently deconvoluted to identify the individual peptide inducing the response.
The ZIKV 15-mer peptide EPRTGLDFSDLYYLT was identified to induce a ZIKV-specific CD8+ T-cell response. To establish a minimal optimal epitope, all of the possible 8-, 9-, 10-, and 12-mer peptides that had potential to bind any of the MHC class I molecules present in the donor were synthesized and tested. Epitope predictions for class I were made using the consensus prediction methods publicly available through the IEDB Analysis Resource (www.iedb.org).
PBMCs (1 × 106 cells/well) were incubated with ZIKV mega-pools (1 μg/mL) for 6 h. After 2 h of stimulation, Brefeldin A (1 μg/mL; BD Bioscience) was added. At the end of the stimulation, cells were washed, and stained for 30 min with primary antibody cocktail. After primary surface staining, cells were washed, fixed with 4% paraformaldehyde, permeabilized, blocked with normal human sera (Gemini), and stained for intracellular IFN-γ and Granzyme B. A BD LSR-II flow cytometer was used for data acquisition (BD) and data were analyzed with FlowJo X software (Tree Star Software). A full list of antibodies used in the ICS staining are shown in supplementary (S1 Table).
The GLDFSDLYY 9-mer sequence was used to synthesize an A*01:01 tetramer by the National Institute of Health (NIH) Tetramer Core Facility. For tetramer staining, PBMCs (2 × 106 cells/well) were incubated with the tetramer (1:100 dilution) for 90 min at room temperature. After 1 h of incubation with tetramer, surface antibodies were added (S1 Table) for half an hour at room temperature. Cells were then washed, fixed, acquired, and analyzed as previously described [47, 48].
2 × 105 PBMCs were incubated in triplicates with 0.1 mL complete RPMI 1640 in the presence of individual peptides at different final concentrations [1, 0.1, 0.01, and 0.001 μg/ml]. Following a 20 h incubation at 37°C, the cells were discarded, the wells were incubated with biotinylated IFN-γ mAb (clone 7-B6-1; Mabtech) for 2 h, and the wells were developed as previously described [47, 48].
During the ZIKV outbreak in Miami, FL in 2016, a female subject (Hu0015) noticed an abnormal rash, which started on the torso and continued to spread centrifugally, eventually reaching her extremities. The rash (Fig 1) was neither itchy nor sensitive. Mild general malaise was also reported. She did not report pruritus, myalgia, fever, retro-orbital pain, or GI symptoms and was not hospitalized. Immediate speculation of ZIKV infection prompted the study team to enroll Hu0015 in a University of Miami protocol approved by the Human Subjects Board to characterize ZIKV infection. In total, we collected blood and saliva at D5, 7, 15, 21, 28, 48, 56, 70, 91, 106, 116, and 148 POS. We separated whole blood into peripheral blood mononuclear cells (PBMCs) and EDTA-plasma for all time points. While most of the samples were immediately frozen for later analysis, we stained fresh PBMCs from D5, 7, 15, 21, and 148 POS for the human plasmablast (Pb) phenotype in order to track Pb development.
We extracted RNA from plasma at D5, 7, and 15 POS to isolate and quantify virus. We saw RT-PCR amplification with ZIKV-specific probes in the D5 and D7 POS plasma samples in the range of 150 genome equivalents (GE)/mL, but amplification from the D15 POS plasma sample was not detected. The D5 POS saliva sample yielded 21,610 GE/mL, but ZIKV was undetectable in saliva collected at D7 and D15 POS. We sequenced the D5 saliva and produced > 4.5 million 250 nucleotide (nt) sequences aligning to the ZIKV genome [33]. These sequences formed a 10,609 nt contig that covered 100% of the protein-coding sequence and 98.2% of the ZIKV genome at an average depth of 93,103 nt. The results have been deposited into GenBank (Accession ID KX832731). We reconstructed a maximum likelihood phylogenetic tree using the ZIKV genome from Hu0015 and from 195 published genomes from the Pacific and the Americas since 2013 (Fig 2). The placement of the ZIKV genome from subject Hu0015 with other ZIKV genomes recovered from the Miami outbreak confirms that this was a locally-acquired infection.
Subject Hu0015 expressed the MHC-I alleles A*01:01:01, A*02:01:01, B*08:01:01, and B*41:02:01 (Table 1). The MHC-C alleles were not typed. Hu0015’s MHC-II alleles were DRB1*13:03:01, DRB1*15:02:01, DQB1*03:01:01, DQB1*06:01:01, DPB1*11:01:01, and DPB1*13:01:01 (Table 1).
We first determined the timing of the emergence of the Pb response after ZIKV infection. In the case of other flavivirus infections, the induction of massive Pb responses D5 to D14 POS has been reported [49, 50]. We stained Hu0015 PBMC for the human Pb phenotype (CD3−/CD19+/CD20−/low/CD38high/CD27high) and expressed this as a fraction of the total CD19+ B-cells. Pb frequency was measured at D5, 7, 15, 21, and 148 POS along with samples from a ZIKV- and DENV-naïve individual as a control. At D5 POS, 7.23% of the CD19+ B-cells expressed the Pb phenotype while only 0.18% Pbs were present in the naïve control (Fig 3). As the immune response progressed, the Pb frequency increased to 19.6% at D7 POS (Fig 3). The Pb frequency contracted to a baseline level of 0.85% and 0.54% at D15 and D21 POS respectively (Fig 3). The frequency of circulating Pbs during the convalescent phase, measured at D148 POS, was 0.36%. This measurement served as a baseline control for Hu0015.
The non-structural 1 (NS1) protein is secreted into the circulation early on during flavivirus infection [6]. We, therefore, assessed plasma from subject Hu0015 for IgM binding against recombinant ZIKV-NS1 protein (Fig 4A). At D5 POS we observed low levels of ZIKV-NS1-specific IgM antibodies. However, at D7 POS, there was a dramatic rise. The peak of the ZIKV-NS1-specific IgM response occurred between D7 and 15 POS, and then the ZIKV-NS1-specific IgM response declined over time, reaching baseline levels by D148 POS. We then assessed the ZIKV-NS1-specific IgG response (Fig 4B). There was no ZIKV-NS1-specific IgG reactivity at D5 POS. At D7 POS, ZIKV-specific NS1 IgG level had increased by over 300% from D5 POS and continued to increase until D15 POS. From D15 to D48 POS, ZIKV-specific NS1 IgG antibodies were relatively stable in the plasma. At D91 POS there was a decrease, with another slight decrease at D148 POS.
Considerable cross-reactivity among flavivirus-specific antibodies and the various members of the flavivirus family has been reported [13, 17, 51–55]. We sought to understand the exact timing of the appearance of this cross-reactivity after ZIKV infection of a previously DENV-naïve patient. Hu0015 had no recent travel history outside the US and there was no evidence that she had any detectable pre-existing anti-DENV IgG binding antibodies at either D5 or D7 POS. We analyzed DENV-cross-reactive IgM and IgG antibodies from Hu0015 using whole virus binding enzyme-linked immunosorbent assays (ELISAs). We assessed DENV1-4 binding and demonstrated that the initial IgM response was already cross-reactive by D5 (Fig 4C). Even though cross-reactive DENV IgM responses were present at D5 POS, the levels of these IgG antibodies were less than those that were directed against ZIKV. The peak of the IgM response directed against the four DENV serotypes occurred between D15 and D21 POS and it diminished after that. The peak of the DENV IgM response was lower than that against ZIKV. As expected, the initial IgG response against both DENV and ZIKV was delayed in comparison to the IgM response. The DENV-cross-reactive IgG response against serotypes 1–4 was low at the early time points (Fig 4D). However, with time, cross-reactive DENV1-4 IgG responses continued to rise, albeit not as rapidly as the response against ZIKV. Interestingly, antibodies in the plasma from Hu0015 appeared to bind DENV1 and DENV3 less than DENV2 and DENV4.
We then determined when ZIKV-specific neutralization capacity developed in our patient. We conducted neutralization experiments using flow cytometry (NEUT) to assess the ability of plasma to neutralize the ZIKV-Paraiba strain (Fig 5A). We used Hu0002, a ZIKV- and DENV-seronegative subject, and Hu0004, a ZIKV- and DENV-seropositive individual as controls. Hu0015, at D5 POS, neutralized ZIKV, but this only occurred at low dilutions. At D7 POS, the NEUT 50% neutralization point (NEUT50) increased almost 10-fold to 1:817. The plasma NEUT50 titer peaked at D15 POS at 1:2,858 and the titer remained high through D48 POS. Plaque reduction neutralization tests (PRNTs) were also performed against the ZIKV-Paraiba strain out to D148 POS. The calculated 50% neutralization point for plaques (PRNT50), yielded similar titers to the NEUT50 titers. NEUT50 titers were slightly higher than the PRNT50 titers. PRNTs were also performed for all DENV serotypes for the selected time points (Fig 5B). Surprisingly, Hu0015 plasma did not neutralize any of the DENV serotypes at any time point, despite the presence of binding antibodies (Fig 5B).
Since there was no neutralization of DENV with the patient’s plasma, we sought to determine whether the patient’s plasma could mediate antibody dependent enhancement (ADE) using K562 cells. These immortalized monocyte lineage cells are not permissive to DENV or ZIKV infection, but express the Fc-gamma receptor (FcR), thereby facilitating flavivirus infection. Patient plasma from D5 and D21 POS showed no enhancement of ZIKV (Fig 6A). However, at D48 POS, the patient’s plasma enhanced ZIKV infection and this continued until the last time point measured (D148) (Fig 6A). Peak ADE occurred at dilutions of plasma between 1:160 and 1:640. Surprisingly, we saw no enhancement of DENV2 infection with the ZIKV-infected patient’s plasma at any time point measured (Fig 6B).
We were then interested in the ontogeny of the T-cell response against ZIKV. We used pools of overlapping 15-mer peptides spanning the entire ZIKV proteome in both intracellular cytokine staining (ICS) and IFN-γ ELISPOT analysis. Using ICS, ZIKV-specific CD4+ T-cell responses were detected against the ZIKV-NS2A protein (Fig 7A), and CD8+ T-cell responses were detected against the ZIKV-Envelope (E) protein (Fig 7B). Deconvolution of all ZIKV-E derived peptides identified the 15-mer EPRTGLDFSDLYYLT as a target of the ZIKV-specific CD8+ T-cell responses (Fig 7C). To establish the minimal optimal epitope, all possible 8-, 9-, 10-, and 12-mer peptides that had potential to bind any of the MHC class I molecules present in the donor were synthesized and tested with ELISPOT. The 9-mer GLDFSDLYY peptide was shown to elicit the strongest response of all peptides tested and the sequence of this peptide was consistent with the peptide binding motif for HLA-A*01:01 (Fig 7D). This 9-mer minimal-optimal peptide was then synthesized with a HLA-A*01:01 tetramer at the NIH Tetramer Core Facility. Interestingly, the first four amino acids of this 9-mer minimal-optimal peptide were conserved in all four DENV serotypes and ZIKV. The second half was unique to ZIKV, and was identical for all four DENV serotypes (Fig 7E). We then used this tetramer to track the ZIKV-specific response in the CD8+ T-cell compartment (Fig 7F). The initial tetramer response appeared at D7 POS, peaked at D21 POS, and was still present, albeit at a very low level, at D148 POS. The tetramer positive CD8+ T-cells over time can also be seen overlaid on top of the naïve, effector memory, central memory, and effector memory CD8+ T-cells based on expression of CCR7 and CD45RA markers (S1 Fig). Phenotypic analysis of the responding cells revealed that most tetramer positive cells were contained within the effector memory (CCR7-/CD45RA-) T-cell subset. An in-depth characterization of the phenotype of CD8+ T-cells was also performed by ICS (S2 Fig). We also analyzed the ontogeny of the CD8+ T-cell response by IFN-γ ICS against ZIKV protein-specific mega-pools. We saw very low levels (borderline to undetectable) of IFN-γ production at all time points except for D21 POS (S3 Fig), which matched the peak of our tetramer data. The ICS of the CD8+ IFN-γ+ T-cells at D21 POS revealed an overexpression of cytotoxic markers PD1 and Granzyme B, and a downregulation of CX3CR1. The controls for the flow cytometry experiments in Fig 7 were also performed with a ZIKV-naïve individual (S4 Fig).
Local transmission of ZIKV has only recently been reported in the continental US [19]. From January, 2015 to May, 2017, the CDC has confirmed the diagnosis of over 5,000 ZIKV cases in the contiguous US, with 256 of these cases being local transmissions in Miami, FL [56]. Early during the outbreak, we discovered a patient exhibiting the hallmarks of ZIKV infection. This individual was confirmed by RT-PCR to have been infected with ZIKV and we followed her immune response until D148 POS. For the first time in humans, we describe a longitudinal study of the ontogeny of the plasmablast, antibody, and T-cell immune responses in one of the first locally-acquired ZIKV infections in the contiguous US (Fig 8).
The ZIKV isolated from this patient’s saliva was one of the first sequenced ZIKV genomes from an autochthonous mosquito-borne transmission in the US. Sexual transmission was ruled out, as the sole sexual partner of Hu0015 was also tested for ZIKV, and he was found to be negative for ZIKV by RT-PCR and anti-ZIKV antibodies by VCA. The phylogenetic placement of this ZIKV genome puts it close to other genomes sampled from the Wynwood and Miami Beach transmission zones later identified by the Florida Department of Health during the 2016 outbreak in Florida. Interestingly, the ZIKV isolate from Hu0015 was distinctly different from those circulating elsewhere in the Americas [21], which strongly suggests that the patient was indeed infected in Florida. Phylogenetic analysis indicates that the current Zika virus outbreak in Florida was the result of an imported infection (human or mosquito) from the ongoing epidemic in the Caribbean [21]. The Hu0015 ZIKV genome also clusters with genomes obtained from infected Aedes aegypti mosquitoes collected in Miami Beach, suggesting that the route of infection was likely mosquito-borne [21].
ZIKV infection, in mice and non-human primates (NHPs), has been well characterized, allowing investigators to follow the ontogeny of the immune response as well as analyze tissues not previously analyzed in humans [18, 57–61]. Mouse models can be useful in understanding cell signaling pathways in ZIKV infection as well as for rapidly screening drugs that control or prevent ZIKV replication [62, 63]. In addition, because ZIKV infection can induce fetal abnormalities in mice, they provide a more economical option than NHPs when initially developing prophylactic and therapeutic approaches to counteract the impact of ZIKV impact on fetuses [60, 64]. ZIKV infection of NHPs, specifically rhesus macaques, appears to closely resemble the viral kinetics and immune response found in humans [18, 58, 59, 61]. However, no animal model is a perfect substitute for an actual human infection, as disease characteristics such as rash and fever are often not always associated with the NHP and mouse models [58, 59, 61].
We here present the first human study of the ontogeny of the immune responses against ZIKV. We demonstrate that the virus is no longer detectable from the patient’s plasma by D15 POS. This matches the viral kinetics seen in NHP infections which generally have undetectable virus in the plasma by D7 post-infection [18, 58, 59, 61]. This seems appropriate given the suspected 3–14 day incubation period in mosquito-borne transmission in humans [65]. This short detection window of viral nucleic acid in plasma underlines the difficult reality physicians will face when attempting to diagnose ZIKV at the end of an acute infection and the need for a post-acute diagnostic test [66]. Previously, it was thought that ZIKV was more readily detectable in the saliva than in plasma, but the window for detection, on average, remained the same [67]. Moreover, it was found that saliva alone was insufficient for diagnosis as replicate saliva samples varied greatly in virus titer when compared to the consistency of being able to detect ZIKV in plasma [67]. Only recently has it been shown that ZIKV persists in whole blood substantially longer than in plasma, which may allow for more sensitive testing and a greater window of diagnosis [68]. Furthermore, it is unclear as to what ZIKV persistence in whole blood may mean for the potential safety of whole blood and blood products from potential asymptomatic donors that were in ZIKV endemic areas. It has, however, been shown that these blood transmission events were rare in French Polynesia [69].
The Pb response in Hu0015 appears to match the kinetics of previously published Pb studies in ZIKV-infected NHPs; elevated at D5 and peaking at D7 post-infection, before rapidly contracting [59]. Furthermore, the kinetics of the Pb response in a human primary ZIKV infection appear to occur at a similar rate when compared to that of human primary DENV Pb responses [49, 50, 54, 55]. This information may be useful for the isolation of ZIKV-specific monoclonal antibodies, since the ability to clone antibodies from this acute Pb population has led to the isolation of several potent monoclonal antibodies directed against ZIKV which may be pivotal in ZIKV diagnostics and therapies [17, 55, 70, 71].
Some of the specificities of the antibodies generated in this patient were unexpected. Cross-reactive DENV binding antibodies were observed soon after infection in Hu0015’s plasma. However, these antibodies did not appear to neutralize DENV. While Stetter et. al. [17] have shown that some of the mAbs isolated from acute ZIKV infection (without prior DENV exposure) can neutralize DENV, we have found distinctly the opposite result in the plasma from Hu0015 at all time points tested. It is possible, therefore, that while some DENV neutralizing antibodies may be present at low frequencies, the bulk of the antibodies in the plasma of our ZIKV-infected patient do not neutralize DENV.
While the concept of ADE in DENV infections has been well characterized, less is known as to how ZIKV seropositive status might contribute to enhancement of DENV or ZIKV infection [17, 51, 53]. ADE was first described by Halstead in 1973 in which he performed a series of in vitro and in vivo experiments addressing the issue of enhancement of DENV infection [72–75]. Interestingly, Hu0015’s plasma did not enhance DENV infection in vitro, despite the presence of DENV binding antibodies. Perhaps this lack of DENV ADE is related to the lower levels of DENV-cross-reactive antibodies which did not reach a sufficient threshold to induce ADE in our patient. By contrast, ADE of ZIKV infection was present at D48 POS, when ZIKV IgG antibodies were at higher concentrations than they were during the acute phase (Fig 8).
The ontogeny of the T-cell response against ZIKV has yet to be carefully defined in humans, although there are some animal models that may be instructive [76]. The Hu0015 ZIKV-specific CD8+ T-cell immune response is one of the first to be analyzed using a ZIKV-specific MHC class I tetramer. It has previously been shown, that the CD8+ T-cell responses against DENV primarily target the non-structural proteins [47, 77]. Surprisingly, in Hu0015, the T-cell response against primary ZIKV infection was directed entirely against ZIKV-E. New results clearly show that the CD8+ T-cell response in predominately against ZIKV-E in several other cases of primary ZIKV infection in humans, thus further differentiating the CD8+ T-cell response from DENV infection [78]. Previous CD8+ T-cell YF-specific MHC class I tetramer studies have shown that the T-cell response peaks at D30 post-vaccination [79]. With the case of ZIKV infection in Hu0015, it is interesting to note that the peak of tetramer positive CD8+ T-cells occurred at D21 POS. The most likely explanation for the difference is the estimated incubation period of 3–14 days in which the virus is replicating in the patient before the onset of symptoms [65]. Another explanation may also be that the previously mentioned YF vaccination study only measured D14 and D30 post-vaccination and the peak may have occurred earlier than D30 post-vaccination. Additionally, the CD8+ tetramer response after DENV infection shows remarkably similar peaks between D7 to D14 POS [80–82].
In conclusion, the ZIKV outbreak in Miami, FL in 2016 highlighted that ZIKV can pose a severe threat to public health in the US and may become a re-occurring reality. Importantly, the degree to which our results from this single DENV-naïve individual can be extrapolated to the entire DENV-naïve population is unknown. However, a longitudinal analysis of the B- and T-cell ontogeny of a large cohort will require considerable resources. Here, our detailed longitudinal analysis provides the groundwork for larger studies in the future. We followed the evolution of the B- and T-cell response against ZIKV and saw peak immune responses between D15 and D21 POS (Fig 8). The highest neutralization activity in plasma occurred during the peak of the ZIKV-specific IgM and IgG antibody response. We detected a rapid Pb expansion and described the ontogeny of the generation of cross-reactive antibodies against DENV after primary ZIKV infection. We observed ADE of ZIKV with Hu0015’s plasma at D48 POS, but there was no ADE of DENV2 infection despite the presence of cross-reactive antibodies against DENV. For the first time, we have described the ontogeny of the plasmablast, antibody, and T-cell immune responses in one of the earliest locally acquired ZIKV infections in the contiguous US.
|
10.1371/journal.pmed.1002776 | Risk score for predicting mortality including urine lipoarabinomannan detection in hospital inpatients with HIV-associated tuberculosis in sub-Saharan Africa: Derivation and external validation cohort study | The prevalence of and mortality from HIV-associated tuberculosis (HIV/TB) in hospital inpatients in Africa remains unacceptably high. Currently, there is a lack of tools to identify those at high risk of early mortality who may benefit from adjunctive interventions. We therefore aimed to develop and validate a simple clinical risk score to predict mortality in high-burden, low-resource settings.
A cohort of HIV-positive adults with laboratory-confirmed TB from the STAMP TB screening trial (Malawi and South Africa) was used to derive a clinical risk score using multivariable predictive modelling, considering factors at hospital admission (including urine lipoarabinomannan [LAM] detection) thought to be associated with 2-month mortality. Performance was evaluated internally and then externally validated using independent cohorts from 2 other studies (LAM-RCT and a Médecins Sans Frontières [MSF] cohort) from South Africa, Zambia, Zimbabwe, Tanzania, and Kenya. The derivation cohort included 315 patients enrolled from October 2015 and September 2017. Their median age was 36 years (IQR 30–43), 45.4% were female, median CD4 cell count at admission was 76 cells/μl (IQR 23–206), and 80.2% (210/262) of those who knew they were HIV-positive at hospital admission were taking antiretroviral therapy (ART). Two-month mortality was 30% (94/315), and mortality was associated with the following factors included in the score: age 55 years or older, male sex, being ART experienced, having severe anaemia (haemoglobin < 80 g/l), being unable to walk unaided, and having a positive urinary Determine TB LAM Ag test (Alere). The score identified patients with a 46.4% (95% CI 37.8%–55.2%) mortality risk in the high-risk group compared to 12.5% (95% CI 5.7%–25.4%) in the low-risk group (p < 0.001). The odds ratio (OR) for mortality was 6.1 (95% CI 2.4–15.2) in high-risk patients compared to low-risk patients (p < 0.001). Discrimination (c-statistic 0.70, 95% CI 0.63–0.76) and calibration (Hosmer-Lemeshow statistic, p = 0.78) were good in the derivation cohort, and similar in the external validation cohort (complete cases n = 372, c-statistic 0.68 [95% CI 0.61–0.74]). The validation cohort included 644 patients between January 2013 and August 2015. Median age was 36 years, 48.9% were female, and median CD4 count at admission was 61 (IQR 21–145). OR for mortality was 5.3 (95% CI 2.2–9.5) for high compared to low-risk patients (complete cases n = 372, p < 0.001). The score also predicted patients at higher risk of death both pre- and post-discharge. A simplified score (any 3 or more of the predictors) performed equally well. The main limitations of the scores were their imperfect accuracy, the need for access to urine LAM testing, modest study size, and not measuring all potential predictors of mortality (e.g., tuberculosis drug resistance).
This risk score is capable of identifying patients who could benefit from enhanced clinical care, follow-up, and/or adjunctive interventions, although further prospective validation studies are necessary. Given the scale of HIV/TB morbidity and mortality in African hospitals, better prognostic tools along with interventions could contribute towards global targets to reduce tuberculosis mortality.
| HIV-associated tuberculosis (TB) is very common in hospitals in sub-Saharan Africa, and is a major cause of morbidity and mortality.
There is a lack of tools to identify which patients are more likely to die early; therefore, these patients cannot be targeted for more intensive clinical care or other treatments in addition to TB antibiotics and antiretroviral drugs.
A new urine TB diagnostic test (detecting a substance called lipoarabinomannan [LAM]) can identify patients at higher risk of dying and, along with other simply measured clinical signs or symptoms, may be useful for predicting which patients are most likely to do poorly.
We used data from HIV-positive adults admitted to hospital in Malawi and South Africa and diagnosed with TB to develop a simple clinical risk score to identify patients with a 50% chance of dying within 2 months.
The score included 6 factors that can be measured at admission to hospital, including the results of the urine TB test, presence of anaemia, and some demographic factors.
We then tested the performance of the risk score using data from studies of different patients from sub-Saharan Africa, and it performed similarly.
The score was able to identify patients at higher risk of dying during admission to hospital, and after discharge from hospital.
This score could be used to identify patients admitted to hospital and diagnosed with HIV-associated TB who may benefit most from more intensive clinical care, additional treatments, and/or closer follow-up after discharge.
It could also be used as a research tool to study new drugs or strategies aimed at reducing mortality from with HIV-associated TB.
This is the first such tool to our knowledge in this patient population, and future studies could optimise such predictive tools, in particular if appropriate new interventions and/or diagnostics become available.
| Tuberculosis (TB) is the leading infectious disease killer globally, causing an estimated 1.7 million deaths globally in 2017 [1]. This burden lies disproportionately in people living with HIV, who account for approximately 1 in 4 TB deaths. The case fatality rate of HIV-associated TB (HIV/TB) is particularly high in hospitals, estimated at 29% in a recent meta-analysis [2]. This may be an underestimate, given that post-mortem studies from sub-Saharan Africa have demonstrated that a high proportion of HIV-positive deaths in facilities have evidence of undiagnosed TB [3].
Interventional studies aiming to reduce mortality in this patient population have demonstrated mortality reductions with improved TB diagnostics [4,5] and appropriately timed initiation of antiretroviral therapy (ART) [6,7]. However, mortality remains substantial despite these interventions, and adjunctive interventions are likely to be needed to further impact mortality. Currently, predictors for mortality are poorly defined. Being able to identify patients at the highest risk of mortality could inform the development and assessment of new interventions, and also identify which patients would benefit most from interventions beyond TB therapy and appropriately timed ART [8].
Clinical decision tools and risk scores are used widely in clinical practice to simplify the identification of patients at highest risk for poor health outcomes. Predictor scores for mortality have been developed for HIV-associated cryptococcal meningitis and pneumonia, and are used to guide management in Pneumocystis jiroveci pneumonia [9–11]. Although scores have been developed to predict risk of TB disease in various populations, including TB bacteraemia in hospitalised patients [12], to our knowledge no externally validated scores exist to predict outcomes of TB disease among hospitalised patients with HIV [13,14]. Scores developed to predict TB mortality in settings with low HIV prevalence are also of limited use in people living with HIV due to differences in clinical presentation, pathogenesis, and outcomes [15–18]. A recent study from the US developed and internally validated a score to predict mortality in HIV/TB in low-prevalence settings (US), but this would not be applicable to hospitalised patients in Africa given that many of the variables are not routinely available [19].
We have previously shown that detection of lipoarabinomannan (LAM) in the urine of HIV/TB patients using a cheap (approximately US$3) and quick (testing takes 25 minutes) lateral flow assay is independently associated with a 2- to 3-fold increased risk of mortality [20]. We therefore aimed to investigate if urinary LAM detection, along with other clinical variables readily available in high-burden settings, could be used to predict which HIV-positive patients admitted to hospital and diagnosed with TB were at high risk of early mortality, and to externally validate the predictive tool.
We used data from the STAMP (‘rapid urine-based screening for tuberculosis in HIV-positive patients admitted to hospital in Africa’) trial for the clinical risk score derivation [5,21]. The STAMP trial recruited HIV-positive adults (aged 18 years or more), irrespective of symptoms or clinical presentation, who were admitted to medical wards of 2 hospitals in Malawi and South Africa between 26 October 2015 and 19 September 2017. On admission, patients were screened for TB using Xpert MTB/RIF (Xpert; Cepheid) on sputum in both study arms, and Xpert and Determine TB LAM Ag (TB-LAM; Alere) assays on urine in the intervention arm.
Exclusion criteria in the trial were already taking TB treatment and inability to give consent. The clinical teams managing the patients were masked to which TB tests were positive; therefore, management of TB patients should not have differed between arms. The management of HIV/TB in the study hospitals was representative of their local settings and followed local and national guidelines, with no input from the study team (beyond TB diagnostic tests).
Patients diagnosed with TB in the standard-of-care arm had stored urine tested with Xpert and TB-LAM retrospectively. Data were collected at baseline (at or close to admission) on demographics and clinical characteristics, and subsequently on TB investigations and treatment, and clinical events, including death or discharge from hospital. Patients discharged alive were followed up at 2 months by outpatient attendance, home visit, or telephone for vital status. The derivation cohort included all patients (from both trial arms) with laboratory-confirmed TB. The outcome was mortality risk at 2 months after admission. Patients lost to follow-up were assumed alive at 56 days.
Laboratory-confirmed TB was defined as any 1 of a positive smear microscopy, mycobacterial culture, Xpert from any site, or urinary TB-LAM. TB-LAM assay was positive if recorded as ‘grade 1’ or higher on the manufacturer’s (post-2014) reference card. Ability to walk unaided was assessed by healthcare workers (not self-reported by patients), and was equivalent to a Karnofsky functional score below 40 points [22]. WHO danger signs were heart rate > 120 beats per minute, respiratory rate > 30 per minute, temperature > 39°C, and being unable to walk unaided. ‘ART experienced’ was defined as receiving ART at the time of enrolment to the study.
Candidate predictor variables were identified for inclusion in the predictive model based on a priori clinical knowledge, previous literature, and the need for variables to be objective, reproducible, and available in resource-constrained settings [23]. We considered variables known to be associated with mortality in HIV/TB, including age, sex, ART experience, physiological measurements at admission, weight and/or body mass index, CD4 cell count, functional status (being unable to walk unaided), and haemoglobin [24–28]. Time on ART was not considered as not all patients take ART, and because of challenges in accurately ascertaining duration. Where 2 or more predictors were highly correlated, only 1 was selected, to simplify the prognostic model, as inclusion of all would contribute little additional predictive information [23]. Analyses were planned prospectively (see S1 Appendix) except where indicated as post hoc.
Continuous variables were assessed for non-linearity using fractional polynomials, and categorised based on previously established cutoffs (e.g., CD4 cell count and haemoglobin) or associations with mortality (e.g., age and weight, using the fp plot command in Stata). Complete case analysis was chosen for the derivation score as few data (<5%) were missing. We first performed univariable analyses assessing the association of each variable with mortality risk using logistic regression. We then used a backward elimination, stepwise approach to create a multivariable predictive model, starting with all candidate variables, and excluding variables sequentially if p > 0.1 using likelihood ratio tests and the Akaike information criterion. Given that there were 94 deaths, we did not want to estimate more than 9 candidate predictors (various studies have shown each candidate predictor studied requires a minimum of 10 events) [29]. Interactions were also assessed using likelihood ratio testing. All analyses were done using Stata version 14, and all p-values were 2-sided.
Regression coefficients from the final multivariable model were multiplied by the smallest possible constant and then rounded to the nearest integer, and then assigned as ‘points’ to each variable. The clinical risk score was derived by combining the points based on each patient’s characteristics. High-, medium-, and low-risk groups for mortality were then arbitrarily defined after plotting risk score against observed mortality such that the high-risk group accounted for most (>50%) deaths and the low-risk group accounted for as few deaths as possible.
Mortality risk at 2 months and 95% confidence intervals (CIs) were calculated for each risk group, as were odds ratios (ORs) and 95% CI for mortality. In exploratory analyses, inpatient and outpatient (post-discharge) deaths were also compared between risk groups by restricting analyses to deaths occurring during hospital admission or to deaths occurring after discharge in the subset of patients who were discharged alive from hospital. CD4 cell count and TB-LAM grade were also compared between risk groups. Mortality risk was compared between groups using chi-squared tests.
We assessed the model discrimination (ability to differentiate patients who would die within 2 months and those who would survive) by calculating the concordance index (c-statistic) (also known as the area under the receiver operator curve), assuming a c-statistic < 0.6 showed poor discrimination [30]. Model calibration was assessed by plotting the probability of mortality predicted by the model against observed mortality in the derivation dataset using a calibration plot and the Hosmer-Lemeshow test, assuming a p < 0.05 indicated poor calibration. In post hoc analysis, in response to reviewer request and to better understand the utility of the score, the sensitivity and positive predictive value of the score were calculated.
To externally validate the clinical risk score, we used data collected independently from 2 studies: (1) a multicentre diagnostic clinical trial of adjunctive urine TB-LAM testing in HIV-positive patients with TB symptoms who were admitted to hospitals in 4 sub-Saharan African countries (South Africa, Zambia, Zimbabwe, and Tanzania) (LAM-RCT) [31] and (2) a prospective cohort study assessing the diagnostic yield of TB-LAM in HIV-positive patients with TB symptoms in Kenya (Médecins Sans Frontières [MSF] cohort study) [32]. Patients were included in the validation cohort if they were adults and had laboratory-confirmed TB (as previously defined). Patients from the LAM-RCT in the ‘no TB-LAM’ arm were excluded, as were outpatients (i.e., patients not admitted to hospital) from the MSF cohort study.
The validation cohort sites were all in settings in sub-Saharan Africa with high HIV prevalence and TB incidence, but differed from the derivation cohort in that all patients had at least 1 TB symptom (cough, fever, weight loss, or night sweats). The LAM-RCT recruitment occurred between 1 January 2013 and 2 October 2014, and the MSF cohort recruitment between 22 October 2013 and 20 August 2015. Mortality outcomes were assessed at 2 months in both studies.
The clinical risk score for mortality was calculated by assigning the same ‘points’ to variables as for the derivation cohort, and the same cutoffs were used to define high-, medium-, and low-risk groups for mortality. Patients with missing observations were excluded, for a complete case analysis. However, sensitivity analyses were done for score performance using multivariate multiple imputation with chained equations for missing data as 42% of patients had missing data in the validation cohort. Data were assumed to be missing at random, and were imputed for missing candidate predictor variables using mortality risk, other candidate predictor variables, and other baseline demographic variables, with 100 imputations.
Evaluation of the score in the validation dataset was done using the same statistical methods as the internal evaluation, with calculation of mortality risk at 2 months, ORs for mortality, and survival curves. Discrimination was assessed using the c-statistic, and calibration with a calibration plot and the Hosmer-Lemeshow test.
The study is reported in concordance with TRIPOD guidance for multivariable prediction models (see S2 Appendix) [33]. Ethical approval for each of the source studies was obtained from the relevant ethics committees in the country of data collection and from the trial sponsors (see S3 Appendix for list of ethics committees). All patients provided informed written consent.
Of 506 HIV-positive patients diagnosed with TB in the STAMP trial derivation cohort, 322 had laboratory-confirmed TB. Seven patients were excluded from the complete case analysis for missing data (Fig 1). The median age of TB patients included in the derivation cohort was 36 years (interquartile range [IQR] 30–43), 172 (55%) were men, 53 (17%) were newly diagnosed with HIV, and the median CD4 cell count was 76 cells/μl (IQR 23–206; Table 1). In all, 209 (65%) patients were positive on urine TB-LAM testing, indicating probable disseminated TB disease. Anaemia was common and median haemoglobin was 86 g/l (IQR 67–108). Patients presented with advanced disease: 133 (42%) had 1 or more WHO danger signs, and 71 (23%) were severely disabled or unable to walk unaided.
In the derivation cohort, 94 (30%) patients died within 2 months, with 66 (70%) dying during their hospital admission; 29 (31% of deaths) patients died by 1 week, and 52 (55%) by 2 weeks, after admission. In unadjusted analyses, mortality risk was higher in patients aged 55 years or older, men, ART-experienced patients, those unable to walk, patients with severe anaemia (haemoglobin < 80 g/l), patients with CD4 cell count < 100 cells/μl, and those with positive urine TB-LAM tests (Table 2). Six out of 322 (2%) patients were lost to follow-up after hospital discharge.
The final multivariable logistic regression model for mortality at 2 months included age, sex, ART experience, haemoglobin, functional status (being unable to walk unaided), and urine TB-LAM result (Table 2). For associations of linear continuous variables with mortality see S1 Table. CD4 count and weight were dropped from the final predictor score model as their relationship with mortality was mediated by functional status and urine TB-LAM result. We found no significant interactions between variables in the final model. The c-statistic for the predictive model in the derivation dataset was 0.70 (95% CI 0.63–0.76), showing moderate discrimination. Calibration of the predictive model was good, as shown by the calibration plot (see S1 Fig) and a Hosmer-Lemeshow statistic p = 0.78.
The clinical risk score for mortality, based on the regression coefficients, is outlined in Fig 2. Observed and predicted mortality risks for the risk score are reported in S2 Fig. Mortality risk groups were defined as low risk (10 points or fewer), medium risk (11 to 20 points), or high risk (more than 20 points) (Fig 3). Therefore, in the derivation cohort, 48 (15%) patients were deemed low risk, 142 (45%) were deemed medium risk, and 125 (40%) were deemed high risk. Median risk score was 19 (IQR 13–22, range 0–42). Observed mortality risk by 2 months was 12.5% (95% CI 5.7%–25.4%), 21.1% (95% CI 15.1%–28.7%), and 46.4% (95% CI 37.8%–55.2%) in the low-, medium-, and high-risk groups, respectively (p < 0.001). ORs for mortality were 6.1 (95% CI 2.4–15.2) in the high-risk group and 1.9 (95% CI 0.7–4.8) in the medium-risk group compared to low-risk patients (p < 0.001).
As the regression coefficients and points in the clinical risk score were similar for all 6 variables, we created a simplified version of the score by assigning each variable within the score 1 point if present (age 55 years or over, male sex, ART experienced, severe anaemia, being unable to walk unaided, or urine TB-LAM positive; see S3 Fig). A high mortality risk was defined as 3 or more points, medium risk as 2 points, and low risk as 0 or 1 point.
In the derivation cohort, patients with 3 or more points (high risk) had a mortality of 40.0% (70/175, 95% CI 33.0%–47.5%), compared to 19.6% (18/92, 95% CI 12.6%–29.0%) mortality in those with 2 points and 12.5% (6/48, 95% CI 5.7%–25.4%) in those with 0 or 1 point (p < 0.001) (see Fig 3 and S2 Table). The sensitivity of the risk score for mortality in was 0.75 (the score correctly identified 70/94 deaths), and the positive predictive value was 0.4.
The clinical risk score was useful in predicting deaths that occurred during inpatient admission (50 [28.6%, 95% CI 22.3%–35.8%] in the high-risk group compared to 6 [10.9%, 95% CI 5.9%–19.1%] in the low-risk group, p = 0.001) as well as deaths occurring after discharge (20 [16.0%, 95% CI 10.5%–23.6%] in the high-risk group compared to 0 [0%] in the low-risk group, p = 0.015; Fig 4). More patients in the high-risk group were TB-LAM positive and had higher grades of positive result, but CD4 cell count did not differ by risk group (see S4 Fig). Survival curves by risk group are presented in Fig 5.
The external validation cohort included 644 HIV-positive patients with laboratory-confirmed TB, of whom 372 (58%) patients had no missing data for the risk score and were therefore included in the complete case analysis (Fig 1). Baseline characteristics were similar between cohorts, although fewer patients reported taking ART and more patients presented with severe functional impairment and 1 or more WHO danger signs in the validation cohort (Table 1). A similar proportion of patients were positive on urine TB-LAM testing (65% in the derivation cohort compared to 66% in the validation cohort). Mortality at 2 months was lower in the validation cohort (22.8%) compared to the derivation cohort (29.8%). Loss to follow-up was 4% in the validation cohort (15/372).
In complete case analysis (n = 372), the observed mortality risks in the validation cohort were 8.7% (95% CI 4.6%–16.0%) in the low-risk group, 19.1% (95% CI 13.2%–26.8%) in the medium-risk group, and 35.5% (95% CI 27.9%–43.9%) in the high-risk group (see Fig 4). Median risk score was 16 (IQR 10–22, range 0–42). The ORs for mortality by risk group were similar to those in the derivation cohort (5.8 [95% CI 2.7–12.3] for the high-risk group and 2.5 [95% CI 1.1–5.5] for the medium-risk group compared to the low-risk group). The risk score was also useful in predicting both inpatient and post-discharge deaths (the high-risk group had a 20% risk of post-discharge death compared to 5% in the low-risk group). The simplified risk score performed similarly to the full score in the validation cohort.
The predictive model had similar calibration and discrimination in the validation cohort as in the derivation cohort: the c-statistic was 0.68 (95% CI 0.61–0.74; see S3 Table), and the Hosmer-Lemeshow statistic had p = 0.13 (see S5 Fig for the calibration plot). In a sensitivity analysis using multiple imputation for missing data in the validation dataset (n = 644), the c-statistic for the predictive model was 0.64 (95% CI 0.60–0.69), and the Hosmer-Lemeshow statistic had p = 0.67. ORs for mortality were 5.3 (95% CI 2.2–9.5) for the high-risk group and 2.1 (95% CI 1.0–4.6) for the medium-risk group compared to low-risk patients (p < 0.001).
In this study, we developed and externally validated a pragmatic clinical risk score to predict early mortality in HIV-positive patients admitted to hospital and diagnosed with laboratory-confirmed TB. Our score used 6 clinical and laboratory factors that could be readily collected at admission to hospital in settings with high HIV and TB burden. The score was able to categorise patients into 3 risk groups. One-third of the high-risk group died during hospital admission, and almost 50% had died by 2 months. A simplified ‘3 of 6 predictors’ version of the score performed similarly. This is the first study to our knowledge to derive and externally validate a risk score to predict mortality in this patient population.
We found older age, being male, being ART experienced, having severe anaemia, being severely disabled or unable to walk unaided, and being urine TB-LAM positive were all risk factors for mortality. These factors have been established as being associated with outcome in HIV/TB in previous studies [24–28], and most likely reflect more advanced HIV-related immunosuppression, late presentation to healthcare services, and/or poorer underlying physiological reserve. Positive urine diagnostic tests (including LAM detection and Mycobacterium tuberculosis nucleic acid detection) in the context of HIV infection are thought to represent haematogenously disseminated renal TB with high mycobacterial burden, which may explain why it is associated with a worse prognosis [34]. Interestingly, clinical signs and symptoms (such as WHO danger signs) were not predictive of mortality.
In contrast to previously published data, ART-experienced patients had a higher mortality risk in our study [28,35]. This likely reflects a high burden of unrecognised ART failure, due to either poor adherence or drug resistance among patients admitted to hospital. Another potential cause is immune reconstitution inflammatory syndrome (IRIS) in patients who have recently started ART. The relationship between ART and mortality is likely to be more complex, representing different groups of patients with different mortality risks, but for this pragmatic tool we have not been able to explore this further. CD4 cell count, which has been previously shown to be associated with mortality in HIV-positive patients, dropped out of our final multivariable predictive model due to mediation by other variables. Furthermore, in the era of test and treat for HIV and use of quantitative HIV viral load for monitoring, CD4 testing services are being scaled back, and are often not available in resource-limited settings.
Mechanisms and causes of mortality in advanced HIV/TB are still not well understood. Co-pathologies, including other opportunistic infections and bacterial pneumonia or sepsis, are commonly detected post-mortem [36,37]. High-risk patients could be prioritised for screening for co-infections, for example using cryptococcal antigen point-of-care tests, or empirical prophylactic treatment with antibacterial agents, an approach that has been shown to reduce mortality in advanced HIV infection [38].
Whilst this clinical risk score can identify patients with the highest risk of mortality, there remains an absence of proven interventions (beyond TB therapy and appropriately timed ART) to reduce mortality in this population. Therefore, we propose this score could be used as a clinical tool to alert clinicians to patients at high risk of mortality who should be reviewed before discharge and/or flagged for early clinical follow-up in settings where urine TB-LAM scale-up is occurring. The score could also be used as a research tool to aid evaluation of intensified or optimised TB treatment regimens or adjunctive interventions aimed at reducing high mortality in this population.
Possible interventions include rapid viral load testing with ART adherence support and early switching for those with virological failure. Host-directed therapies, which target host immune responses, are in clinical trials for TB, including some specifically for HIV/TB [39,40]. Patients identified as being at highest risk for mortality could also be offered more intensive monitoring or supportive care, for example better management of severe anaemia [41], although optimal strategies of supportive care are not clear [42]. Enhanced treatment and prophylaxis for co-infections have been shown to reduce early mortality in patients with advanced HIV initiating ART [38], and may also benefit those with HIV/TB disease. Interventions will likely need to be instituted rapidly after TB diagnosis to alter outcomes.
The risk score was able to highlight patients at highest risk of death post-discharge, in addition to those at high risk of death during hospitalisation, and could be used to prevent too early discharges. Enhanced community support, including home visits, has been shown to reduce mortality after starting ART in advanced HIV [43], and could have a similar impact for HIV/TB patients. Current services in high-burden settings take a public health approach to service delivery, whereas prognostic risk scores can identify patients suitable for differentiated care [8].
The main aim of this risk score was to detect patients at high risk of early mortality who may benefit from interventions in addition to TB treatment. Although the discrimination of the model was not perfect, the sensitivity of the simplified score was 75%; the score did not identify 25% of patients who died within 2 months, and such patients would still receive standard-of-care management of HIV/TB. Proposed interventions to reduce mortality would have limited adverse events, so those deemed as ‘high risk’ by the score but surviving to 2 months are unlikely to come to significant harm from such interventions. However, if adjunctive interventions are found to reduce early mortality, better predictive biomarkers or more accurate predictive tools would allow more efficient use of resources through targeting of patients.
Limitations of our study include the potential for selection bias. In the STAMP trial standard-of-care arm, only patients started on TB treatment for clinical/radiological criteria or following a positive sputum Xpert result had stored urine retrieved for TB testing. Patients with otherwise undiagnosed TB who would have been urine test positive if they had been tested were not included in this study. Patients unable to provide consent, mostly due to being severely unwell and having altered consciousness, were also excluded. Although our risk score did not have optimal discrimination and calibration, performance was adequate and similar to that of other prognostic scores widely used in clinical practice (e.g., the Framingham cardiovascular risk score) [15,44]. Performance may have been reduced by categorising continuous variables for simplicity. TB drug resistance was not a predictor of mortality in this cohort; however, prevalence of rifampicin resistance was low in these settings. Not all established risk factors for mortality were characterised, leaving potential to improve on performance. Future studies could assess more detailed markers of physiology, as well as social and more distal risk factors.
Whilst the score is pragmatic and its constituent factors are widely available in hospitals in African regions with high HIV and TB burdens, it does rely on access to the TB-LAM lateral flow assay. There is now good evidence to support mortality reductions with the use of TB-LAM in HIV-positive patients admitted to hospital [4,5], and its use as a screening test has been incorporated into the latest guidelines in Malawi and South Africa. The assay has also been scaled up nationally in eSwatini, Kenya, and Uganda [45]. Missing data were common in the validation cohort. However, sensitivity analyses using multiple imputation gave similar results as the complete case analysis. We assumed patients lost to follow-up were alive at 2 months, although only 2% in the derivation cohort and 4% in the validation cohort were not followed up after hospital discharge. Our cohort did not include patients treated for TB without a positive diagnostic test, which remains common in HIV-positive patients admitted to hospital, and this patient group may be an important group for whom to apply risk stratification and predictive scores. The biomarkers studied are imperfect predictors of mortality, and further research is needed to focus on better biomarkers to predict outcome.
Strengths of this study include that the derivation cohort and the LAM-RCT external validation cohort were nested within randomised controlled trials. Our predictive model had similar discrimination and calibration in the validation cohort, and was able to identify groups of patients with similarly increased odds of mortality. This was despite the validation cohort being from geographically distinct locations, collected at different times by different investigators, and with a lower overall mortality risk at 2 months. The factors required for the score can be obtained rapidly after admission.
In conclusion, we have developed and externally validated a clinical risk score capable of identifying, among patients admitted to hospital in settings with high HIV/TB burden, those with the highest risk of early mortality. This score could be a useful clinical and research tool, and could prove beneficial in identifying patients who would gain most from adjunctive interventions to reduce mortality. Further work to assess the impact of such risk scores, and to identify which interventions could potentially reduce mortality, is urgently needed if ambitious global targets to reduce TB mortality are to be met by 2025.
|
10.1371/journal.pbio.1000530 | iCLIP Predicts the Dual Splicing Effects of TIA-RNA Interactions | The regulation of alternative splicing involves interactions between RNA-binding proteins and pre-mRNA positions close to the splice sites. T-cell intracellular antigen 1 (TIA1) and TIA1-like 1 (TIAL1) locally enhance exon inclusion by recruiting U1 snRNP to 5′ splice sites. However, effects of TIA proteins on splicing of distal exons have not yet been explored. We used UV-crosslinking and immunoprecipitation (iCLIP) to find that TIA1 and TIAL1 bind at the same positions on human RNAs. Binding downstream of 5′ splice sites was used to predict the effects of TIA proteins in enhancing inclusion of proximal exons and silencing inclusion of distal exons. The predictions were validated in an unbiased manner using splice-junction microarrays, RT-PCR, and minigene constructs, which showed that TIA proteins maintain splicing fidelity and regulate alternative splicing by binding exclusively downstream of 5′ splice sites. Surprisingly, TIA binding at 5′ splice sites silenced distal cassette and variable-length exons without binding in proximity to the regulated alternative 3′ splice sites. Using transcriptome-wide high-resolution mapping of TIA-RNA interactions we evaluated the distal splicing effects of TIA proteins. These data are consistent with a model where TIA proteins shorten the time available for definition of an alternative exon by enhancing recognition of the preceding 5′ splice site. Thus, our findings indicate that changes in splicing kinetics could mediate the distal regulation of alternative splicing.
| Studies of splicing regulation have generally focused on RNA elements located close to alternative exons. Recently, it has been suggested that splicing of alternative exons can also be regulated by distal regulatory sites, but the underlying mechanism is not clear. The TIA proteins are key splicing regulators that enhance the recognition of 5′ splice sites, and their distal effects have remained unexplored so far. Here, we use a new method to map the positions of TIA-RNA interactions with high resolution on a transcriptome-wide scale. The identified binding positions successfully predict the local enhancing and distal silencing effects of TIA proteins. In particular, we show that TIA proteins can regulate distal alternative 3′ splice sites by binding at the 5′ splice site of the preceding exon. This result suggests that alternative splicing is affected by the timing of alternative exon definition relative to the recognition of the preceding 5′ splice site. These findings highlight the importance of analysing distal regulatory sites in order to fully understand the regulation of alternative splicing.
| Pre-mRNA splicing is catalysed by small nuclear ribonucleoprotein particles (snRNP) that recognise the splice sites on pre-mRNA and remove the introns with great precision. U1 and U2 snRNPs recognise the core motifs present at the 5′ and 3′ splice sites, respectively [1]. These core splice site motifs, however, contain only about half of the information required to define exon/intron boundaries [2]. Additional sequence elements can recruit regulatory RNA-binding proteins either to enhance or silence splice site recognition depending on their position relative to the splice sites [3],[4].
T-cells intracellular antigen 1 (TIA1) and TIA1-like1 (TIAL1, also known as TIAR) are closely related RNA-binding proteins. They have three RNA recognition motifs (RRMs) and a carboxyl-terminal glutamine-rich region [5],[6]. RRM2 is the major domain binding to uridine-rich sequences, RRM3 is thought to bind to RNA with no specificity, and RRM1 has no detectable RNA binding affinity in vitro [7]. Instead, RRM1 and the C-terminus interact with U1 snRNP to enhance its recruitment to the 5′ splice site of alternative exons [8]–[11].
TIA1 and TIAL1 are involved in multiple aspects of RNA metabolism. They are present in both the cytoplasm and the nucleus and shuttle between these two compartments in a manner that requires the RRM2 and RRM3 domains [12],[13]. In the nucleus, TIA1 and TIAL1 regulate alternative splicing by binding to U-rich sequences adjacent to the 5′ splice site and recruiting U1-C to promote exon inclusion [8],[10],[11],[14],[15]. They also regulate the splicing of their own mRNAs, and the resulting two major isoforms have different splicing activity [16],[17]. In the cytoplasm, TIA1 and TIAL1 function as translational silencers by binding to the 3′ untranslated region (3′ UTR) of mRNAs [18],[19]. They were also implicated in stress-induced translational silencing in stress granules [12],[20]. In addition, TIA1 and TIAL1 were shown to promote apoptosis [21], and depletion of both proteins promotes cell proliferation [22].
The role of cis-regulatory RNA motifs located close to alternative exons has been widely investigated, but recent studies suggest that distal regulatory motifs might also play an important role [4],[23],[24]. For instance, Nova1 and Nova2 proteins can silence inclusion of an alternative exon when binding downstream of the preceding exon [4],[24]. In contrast, Nova proteins enhance inclusion when binding directly downstream of an alternative exon [4],[24]. This suggested that the local and distal effects of Nova binding downstream of a 5′ splice site are reciprocal [25]. Since the function of TIA proteins in recruiting U1 snRNP to 5′ splice site is well characterised, these proteins offered a unique opportunity for a comprehensive study of the distal splicing effects of changes in 5′ splice site recognition.
Ultraviolet (UV)-crosslinking and immunoprecipitation (CLIP) was first developed to identify RNA sites bound by the splicing regulators Nova1 and Nova2 in brain tissue [26]. The traditional CLIP cDNA library preparation protocol suffers from a potential loss of cDNAs due to truncation immediately before the “crosslink site,” where at least one amino acid remains covalently attached after proteinase K digestion [27]. Therefore, we used a modified cDNA library preparation protocol that was recently developed (iCLIP), which identifies truncated cDNAs by introducing the second adapter to cDNAs after reverse transcription [28]. In addition, iCLIP introduces a random DNA sequence (barcode) to cDNAs during reverse transcription to differentiate between unique cDNA products and PCR duplicates. Since the first nucleotide of resulting cDNA sequences most likely locates directly downstream of the crosslink site, iCLIP enables quantitative and high resolution analysis of protein crosslinking to the target RNAs [28].
Here, we used iCLIP to identify the RNA crosslink sites of TIA proteins. iCLIP showed a high density of TIA crosslinking in 3′ UTRs of mRNAs and in non-coding RNAs (ncRNAs). Intronic TIA binding clusters were restricted to positions immediately downstream of 5′ splice sites. TIA binding at the 5′ splice site of an alternative exon and/or the preceding exon predicted its dual splicing effects. TIA binding enhanced inclusion of proximal upstream alternative exons and usage of upstream alternative splice sites but silenced distal downstream alternative exons if these lacked direct TIA binding. Interestingly, TIA proteins also regulated distal alternative 3′ splice sites, suggesting that by enhancing 5′ splice site recognition, they can indirectly silence downstream alternative exons.
iCLIP was used to identify the crosslink sites of TIA1 and TIAL1 in HeLa cells in a transcriptome-wide manner. Briefly, cells were UV-irradiated, lysed, and RNA was digested with RNase I to a size of approximately 40–100 nucleotides (nt). The proteins were immunoprecipitated using TIA1 or TIAL1-specific antibodies, and the protein-RNA complexes were subjected to 5′ labelling using 32P-γ-ATP for visualisation after SDS-PAGE separation. The specificity of each antibody was determined by overexpression of TIA1 or TIAL1 in HeLa cells (Figures 1A, 1B, S1A and S1B). TIA1 and TIAL1 antibodies each detected RNA-protein complexes of correct size only from UV-crosslinked cells and only if an antibody was used for immunoprecipitation. Furthermore, the signal decreased in TIA1/TIAL1 double knockdown (KD) cells and increased when the corresponding protein was overexpressed (Figure 1A and 1B). There was a slight cross-reactivity of the TIA1 antibody to the TIAL1 protein (Figures 1A and S1A, when TIAL1 was overexpressed). However, during immunoprecipitation, the TIA1 antibody mainly recognised the TIA1 protein, as no increase was seen when TIAL1 was overexpressed (Figure S1B).
To assess the RNA sequence specificity of TIA1 or TIAL1 without the use of antibodies, we also developed crosslinking and affinity purification (iCLAP), a method to purify Strep/His double-tagged TIA1 and TIAL1 proteins using stringent affinity purification. This method circumvents any cross-reactivity of antibody that would identify the same crosslink sites for both proteins. TIA1 and TIAL1 with the Strep/His tag on the N- or C-terminus were overexpressed in HeLa cells. After UV crosslinking and RNase I digestion, the protein-RNA complexes were first purified with magnetic streptavidin bead before ligation to the 3′ RNA adaptor. Cobalt beads were then used to further purify the protein-RNA complexes under denaturing conditions (8 M urea, Figure S1E). iCLAP detected protein-RNA complexes only if cells were transfected with an appropriate construct, and no signal was detected in vector-transfected or non-crosslinked cells (Figure S1F). In summary, the analysis of radioactive protein-RNA complexes indicated that both iCLIP and iCLAP isolated specific protein-RNA complexes without contamination from other co-purified proteins or RNAs.
To amplify the co-purified RNAs, these were dephosphorylated and ligated to the 3′ RNA adaptor on beads during immunoprecipitation (iCLIP) or affinity purification (iCLAP). After SDS-PAGE and nitrocellulose transfer, the protein-RNA complexes of 70–150 kDa were excised from the membrane (Figure S1C) and subjected to proteinase K digestion. The RNA was reverse transcribed with a primer complementary to the 3′ RNA adaptor, which contained a second half complementary to the 5′ Solexa sequencing primer separated by a BamHI digestion site. The cDNA was then self-circularised, digested with BamHI, giving a product with corresponding adaptors at both ends, and amplified by PCR (Figure S1D). PCR products were sequenced using single-end 44 nt reads on the Illumina GA2 system.
Three independent replicate iCLIP experiments and one iCLAP experiment were performed for both TIA1 and TIAL1 (Table S1). In total, 18.4 million iCLIP sequences were generated, 74% of which aligned to the human genome by allowing only single genomic hits and one nucleotide mismatch (Table S2). Unique cDNAs were identified based on random barcodes, and the crosslink site was mapped to the first nucleotide preceding the start of the cDNAs (Figure 1C). Together, iCLIP produced 869,782 unique cDNA reads for TIA1 and 2,966,801 unique cDNA reads for TIAL1 (Table S2). The iCLIP no-antibody controls, performed in parallel with two of the iCLIP experiments, did not generate detectable PCR products. When submitted for sequencing, they generated 1,074 and 7,798 unique cDNAs mapping to the human genome (Table S2). Since TIA1 or TIAL1 iCLIP generated 100-fold more cDNAs than controls, we estimated that over 99% of cDNAs from the iCLIP experiment represent RNA sites specifically crosslinked to TIA1 or TIAL1.
The random barcode introduced into iCLIP cDNAs allowed us to analyse the distribution of TIA1 and TIAL1 on human RNAs in a quantitative and reproducible manner (Figure S2). Only 1.7% of cDNAs mapped in antisense orientation to annotated genes, confirming the high strand specificity of iCLIP. Only 10% of cDNAs mapped to intergenic regions (Figure 2A). The highest cDNA density was seen in 3′ UTRs and ncRNAs, which together contained 22% of all cDNAs (Figure 2A and 2B). 2,277 ncRNAs and 8,602 3′ UTRs had a higher cDNA density than the whole-genome average, and the cDNA enrichment correlated between TIA1 and TIAL1 iCLIP (Pearson correlation coefficient r = 0.95 and r = 0.90, respectively; Figure S3G and S3H). The ncRNA and 3′ UTR sites with the highest cDNA counts mapped to highly expressed RNAs such as tRNAs and histone mRNAs (Figure S4B). Interestingly, cDNA enrichment in 3′ UTRs was 5-fold higher than in the coding sequence (Figure 2B), in agreement with past findings that TIA proteins bind 3′ UTR to regulate translation [18],[29]–[31].
Fifty-eight percent of TIA1 and 60% of TIAL1 cDNAs mapped to introns (Figure 2A). 67,002 introns had a cDNA density higher than the whole-genome average, and the cDNA enrichment correlated between TIA1 and TIAL1 iCLIP (r = 0.81; Figure S3I). The cDNA density in introns was on average 18-fold lower than in 3′ UTRs and ncRNAs (Figures 2B and S3I). Past studies have shown that TIA1 and TIAL1 regulate alternative splicing of exon 6 of FAS mRNA [14],[17],[32],[33]. Both TIA1 and TIAL1 crosslinked to previously characterised intronic binding sites in FAS pre-mRNA (Figure S5A).
Past studies suggested that TIA1 and TIAL1 have different RNA binding specificities [7],[34]. However, the two proteins can regulate alternative splicing of the same exons [17]. We therefore analysed the in vivo RNA specificity of the two proteins using our iCLIP data. As a control, the iCLIP positions were randomised within the co-expressed genomic regions. The 21 nt sequence surrounding the crosslink sites was compared to randomised positions to identify the pentamers enriched at the TIA1 and TIAL1 crosslink sites. Pentamer enrichment in TIA1 and TIAL1 iCLIP data was highly correlated, and UUUUA and AUUUU were the two most common pentamers (r = 0.99; Figures 2C and S3A, Table S3). Comparing replicate iCLIP experiments of either protein verified the high reproducibility of the observed sequence specificity (Figure S3C and S3D). Similarly, iCLAP experiments with both proteins were also highly correlated and were enriched for the same pentamers as iCLIP, independently supporting the determined sequence specificity of both proteins (Figure S3B). These results demonstrated that TIA1 and TIAL1 share the same in vivo RNA binding specificity.
Due to the high stringency of purification of protein-RNA complexes, iCLIP purified RNA sites that directly interact with TIA proteins. However, it is possible that some of these sites represent transient and low-affinity TIA-RNA interactions. To specifically analyse the high-affinity RNA binding sites, we determined clusters of TIA1 or TIAL1 crosslink sites with a maximum spacing of 15 nt containing a significant cDNA count when compared to randomised positions (FDR <0.05). This identified 12,048 TIA1 and 34,058 TIAL1 crosslink clusters. Uridine represented 82% of TIA1 and 75% of TIAL1 clustered crosslink sites (i.e., crosslink sites that located within these clusters) and was also the most common nucleotide at all positions up to 10 nt away from these crosslink sites (Figure S3E). This agreed with the past studies showing that TIA proteins bind to uridine-rich motifs [8],[14],[17],[19].
To compare the overlap between binding of TIA1 and TIAL1 to the same sites, we analysed the proportion of crosslink clusters identified by both proteins. Eighty-three percent (10,021 / 12,048) of TIA1 crosslink clusters overlapped with a TIAL1 crosslink site, and 59% (20,047 / 34,058) of TIAL1 crosslink clusters overlapped with a TIA1 crosslink site. The overlap depended on the number of cDNAs that defined a cluster. Ninety-nine percent of the crosslink clusters that were defined by five or more TIA1 cDNAs overlapped with a TIAL1 crosslink site (Figure 2D). We also assessed the distances between clustered TIA1 and TIAL1 crosslink sites. TIAL1 crosslink sites overlapped with TIA1 crosslink sites 15-fold more common than with randomised TIA1 iCLIP positions (Figure S3F). These analyses indicated that the binding sites of TIA1 and TIAL1 largely overlap.
Since both proteins showed a redundant binding behaviour, the iCLIP data of TIA1 and TIAL1 were merged to increase the reliability of cluster definition, which depends on the number of unique cDNA sequences. 46,970 crosslink clusters were identified in the joint TIA1/TIAL1 data with a maximum spacing of 15 nt (FDR<0.05). The cDNA counts within these crosslink clusters were highly correlated between TIA1 and TIAL1 data, indicating that they had similar affinity to their common RNA binding sites (r = 0.85; Figure 2E). Furthermore, the cDNA counts were correlated even at the single-nucleotide level, as evident in the 3′ UTR of MYC mRNA, which is a functionally validated translational target of the TIA proteins (r = 0.73; Figures 2F and S5B) [30],[35]. Taken together, iCLIP data suggested that TIA1 and TIAL1 have similar affinity for their common RNA binding sites.
In order to identify the candidate positions where TIA proteins bind to regulate splicing, we compared the distribution of TIA1/TIAL1 iCLIP clustered crosslink sites close to constitutive and alternative cassette exons (Figure 3A). Thirty-fold enrichment was seen at positions 10–28 nt downstream of exon/intron boundaries compared to the last 20 nt of both type of exons. Surprisingly, the constitutive and alternative exons showed a similar extent of crosslinking in this region. Approximately 5% of both types of exons contained a TIA crosslink cluster in this region.
Since TIA proteins were previously described as splicing enhancers [8],[14], we hypothesised that TIA crosslink clusters could be used to predict proximal enhanced exons (Figure 3B). Therefore, a regulatory logic was defined where TIA crosslink clusters 10–28 nt downstream of an alternative exon (region α in Figure 3B) predicted enhanced exon inclusion. This predicted 1,620 alternative cassette exons as enhanced by TIA proteins. Furthermore, since binding of Nova proteins downstream of the preceding exon could silence distal alternative exons [4], we defined a second regulatory logic for distal silencing. TIA crosslink clusters 10–28 nt downstream of the preceding exon (region β in Figure 3B) predicted silenced alternative exon inclusion if TIA crosslink clusters were absent in the region α. This predicted 1,962 alternative cassette exons as silenced by TIA proteins.
To assess the iCLIP predictions in an unbiased way, the splicing changes in TIA1/TIAL1 KD HeLa cells were analysed using a high-resolution splice-junction microarray. Microarray data were analysed with the ASPIRE 3 software [28]. The microarray detected splicing changes in 1,213 cassette exons (|ΔIrank|≥1), 46 of which were further assessed using reverse transcription and PCR (RT-PCR) and capillary electrophoresis (Text S1). RNA was isolated from cells treated with three different siRNA oligonucleotide pairs targeting TIA1 and TIAL1, and with control siRNA oligonucleotides (for knockdown efficiency, see Figure S6). Primer pairs generated a PCR product for 40 of the 46 tested splicing events, with 30 detecting two or more splicing isoforms. Among these, RT-PCR validated 86.7% (26/30) of the splicing changes, confirming the high accuracy of the microarray data (Figure S9A, Table S4).
We assessed the accuracy of iCLIP predictions by comparing them to the splicing changes identified by the microarray. Exons were divided into subsets according to the confidence of the detected splicing change (ΔIrank), and the number of exons correctly predicted by iCLIP was identified in each subset. iCLIP predicted approximately 5% of false positives in control exons (|ΔIrank|<0.1; Figure 3C). However, iCLIP predicted a significantly higher number of exons among those with a detectable splicing change (p<0.05, Fisher's Exact Test; Figure 3C). For these exons, iCLIP correctly predicted the direction of splicing change for 105 of 123 exons (85%), out of which 87 were enhanced (ΔIrank ≥2) and 18 were silenced (ΔIrank ≤−2). For example, a TIA crosslink cluster downstream of exon 23a in NF1 pre-mRNA located to a previously described functional TIA binding site [36], and it correctly predicted the enhancing effect of the TIA proteins (Figure 3D). In contrast, a TIA crosslink cluster downstream of the exon preceding the alternative exon in LRRFIP2 pre-mRNA correctly indicated the TIA-dependent silencing (Figure 3E).
To assess whether TIA proteins bind at additional positions to regulate splicing, we comprehensively analysed the positions of TIA crosslinking in target RNAs with respect to the observed splicing changes (Figure 4A). Each clustered crosslink site in an individual RNA was considered as one crosslink event. The RNA map showed an increase in the number of crosslink events downstream of the exons, confirming that the predictive code included the primary positions where TIA proteins regulate splicing. There was a decrease in the number of crosslink events downstream of the silenced exons, combined with a significant increase downstream the preceding exon (Figure 4A). Similarly, mutually exclusive exons such as in FYN pre-mRNA displayed TIA crosslinking downstream of the enhanced exon but not in the vicinity of the silenced exon (Figure S8A). Since silenced exons had a reduced proximal TIA binding compared to control exons, TIA binding at the preceding exon was the most likely cause for the silencing effect.
The RNA map demonstrated a significant increase in the number of crosslink events downstream of the enhanced exons. To further validate the function of TIA binding at this position, we constructed a reporter minigene containing the alternative exon 5 and the flanking introns and exons of OGT pre-mRNA (Figure 4B). In the iCLIP data, the only TIA crosslink cluster present in this region was located downstream of the 5′ splice site. Overexpression of either TIA1 or TIAL1 in HeLa cells increased exon inclusion, whereas TIA KD HeLa cells decreased exon inclusion (Figure 4B). Overexpression of either TIA1 or TIAL1 in the KD cells was able to restore exon inclusion. This confirmed that both TIA1 and TIAL1 could enhance inclusion of the exon.
To directly test whether the RNA sequence underlying the TIA crosslink cluster is necessary for the ability of TIA proteins to enhance exon inclusion, the 40 nt of intronic sequence downstream of the exon were replaced with a sequence from CDC25C pre-mRNA, which did not contain any TIA crosslink sites (Figure 4B). Splicing of the mutant minigene did not change in response to the increased or decreased TIA protein levels (Figure 4B). Thus, the TIA crosslink clusters located downstream of enhanced exons identified the RNA sites necessary to mediate the enhancing effect of TIA proteins.
Extensive TIA crosslinking downstream of constitutive exons suggested that in addition to regulating alternative splicing, TIA proteins might also play a role in maintaining splicing fidelity. Microarray analysis detected increased retention of 143 introns and decreased retention of 102 introns in TIA KD cells. We tested 18 of these introns using real-time PCR with a 94% validation rate (Figure S7B, Table S5). An example of an intron retained in KD cells is shown in PIAA2 pre-mRNA, which contains TIA crosslink sites downstream of the 5′ splice site (Figure S7A). The introns that were inefficiently spliced in KD cells had a significantly increased number of crosslink events downstream of the 5′ splice sites compared to control introns, indicating that TIA1 negatively regulates intron retention (Figure 4C). This is consistent with a past study using the msl-2 reporter minigene, which showed that TIA1 binding to the uridine-rich track prevents intron retention [37]. Our results indicate that maintaining splicing fidelity at 5′ splice sites of constitutive exons is a widespread function of TIA proteins.
In addition to regulating splicing of cassette exons, TIA proteins can also regulate the usage of alternative 5′ splice sites [38]. We therefore hypothesised that TIA crosslink clusters could be used to predict regulation of variable-length exons (Figure 5A). Since the variable regions are often very short, precise identification of binding sites is crucial. To test whether the resolution of iCLIP was sufficient to resolve dual regulation of alternative 5′ splice sites, a regulatory logic was defined where TIA crosslink clusters 10–28 nt downstream of the intron-distal 5′ splice site (position α in Figure 5A) predicted silenced variable exons (i.e., increased usage of the intron-distal 5′ splice site). Conversely, TIA crosslink clusters 10–28 nt downstream of the intron-proximal 5′ splice site (position β in Figure 5A) predicted enhanced variable exons if TIA crosslink clusters were absent downstream of the intron-distal site. This logic predicted TIA-dependent silencing and enhancing for 84 and 172 variable-length exons, respectively.
The microarray detected a splicing change in 213 variable-length exons in KD cells (|ΔIrank|≥1), 147 of which were a result of alternative 5′ splice site use. The accuracy of microarray data was assessed by analysis of seven alternative 5′ splice sites with a 100% validation rate (Figures 5D, 5E, and S9B; Table S4). iCLIP predicted approximately 3% of false positives in either direction among the control exons (|ΔIrank|<0.1; Figure 5B). However, iCLIP predicted a significantly higher number of true positives among the exons that had splicing change in KD cells (p<0.05, Fisher's Exact Test; Figure 5B). Among these exons, iCLIP correctly predicted 18 out of 19 exons (95%), of which 8 were enhanced exons (ΔIrank ≥2) and 10 were silenced (ΔIrank ≤−2). For example, a crosslink cluster downstream of the intron-distal alternative 5′ splice site was associated with silencing of the variable portion of exon 11 in CLIP4 pre-mRNA (Figure 5D). In contrast, a crosslink cluster downstream of the intron-proximal alternative 5′ splice site was associated with enhanced inclusion of the variable portion of exon 33 in CHD9 pre-mRNA (Figure 5E).
To further assess the predictive value of iCLIP independently of the microarray, the nine alternative 5′ splice sites with the highest iCLIP cDNA counts at predictive positions were analysed by RT-PCR. RT-PCR detected alternative isoforms for five of these exons, and all of these showed a splicing change in KD cells (Figure S9D). iCLIP correctly predicted the direction of splicing change for all of these exons (Figures 5C). Furthermore, we comprehensively analysed the positions of TIA crosslink events in target RNAs with respect to the observed splicing change (Figure 6A). In agreement with the predictive regulatory logic, there was a significant increase in the number of crosslink events downstream of the enhanced 5′ splice sites, but not at any other positions in the RNA map.
To further verify TIA regulation of an alternative 5′ splice site, we constructed a reporter minigene containing the variable exon 33 and downstream intron and exon from CHD9 pre-mRNA (Figure 6B). The only TIA crosslink cluster present in this region located downstream of the intron-proximal 5′ splice site (Figure 5E). Overexpression of either TIA1 or TIAL1 significantly increased inclusion of the variable portion of the exon, whereas knockdown of the TIA proteins showed the opposite effect, which could again be rescued by TIA overexpression (Figure 6B). Replacing 40 nt of intronic sequence downstream of the intron-proximal 5′ splice site with a sequence from CDC25C pre-mRNA rendered the minigene unresponsive to the changing TIA protein levels (Figure 6B). This confirmed that the TIA crosslink clusters identified the RNA sites that mediated the effect of TIA proteins at the alternative 5′ splice sites.
The microarray also detected a splicing change in 84 alternative 3′ splice sites, four of which were analysed by RT-PCR with a 75% (3/4) validation rate (Figure S9C, Table S4). The RNA map of TIA binding showed no enrichment of TIA crosslink events at the regulated alternative 3′ splice sites (Figure 6C). Instead, there was an increase in TIA crosslink event 10–28 nt downstream of the preceding 5′ splice site if TIA silenced inclusion of the variable portion of the exon, as shown for C3orf23 pre-mRNA (Figures 6C and S8B). Although the enrichment did not reach the significance level, these observations indicated that TIA proteins might silence the intron-proximal alternative 3′ splice sites by binding at the preceding 5′ splice site.
In order to test this hypothesis, we constructed a reporter minigene containing variable-length exon 4 and the upstream intron and exon from C3orf23 pre-mRNA (Figure 6D). The only TIA crosslink cluster present located downstream of the 5′ splice site (Figure S8B). Since the intron has a size of more than 6 kb, we fused its first and last 600 nt to form the intron in the minigene. Overexpression of TIA1 or TIAL1 protein significantly decreased the inclusion of the variable part of the exon, whereas TIA knockdown significantly increased the inclusion (Figure 6D). In order to verify functionality of this binding site, the 83 nt downstream of the 5′ splice site were replaced with the corresponding region from constitutive exon 2 of GAPDH pre-mRNA (Figure 6D). This resulted in a loss of TIA regulation. Surprisingly, the mutation abolished inclusion of the variable portion of the exon, possibly due to the enhanced TIA-independent recognition of the GAPDH 5′ splice site. Taken together, the minigene reporter analysis confirmed that TIA binding downstream of the 5′ splice site can affect the usage of distal alternative 3′ splice sites.
The present study predicts the positive and negative effects of TIA proteins on splicing of cassette and variable-length exons based on experimental analysis of in vivo TIA-RNA interactions. Consistent with previous observations, we found strong binding of TIA1 and TIAL1 downstream of 5′ splice sites where it exerts local and distal effects on splicing regulation. In addition, binding of both proteins was also enriched in ncRNAs and 3′ UTR of mRNAs. The latter is consistent with their role in translational repression via binding AU-rich element (ARE) and with the role in stress granules, where they promote the assembly of translation pre-initiation complexes [18],[19],[39],[40]. Further analyses of iCLIP data could therefore provide novel insights into how TIA binding mediates translational inhibition.
As shown in the example of the 3′ UTR of MYC pre-mRNA, TIA1 and TIAL1 crosslinking is quantitatively reproducible at the level of individual nucleotides (Figure 2F). This allowed us to show that TIA1 and TIAL1 have identical RNA binding specificity and that both proteins bind to the same primary RNA sites. This result is interesting from an evolutionary perspective, raising the question of why two proteins with the same specificity have been maintained in the genome. The redundancy between TIA1 and TIAL1 might increase the robustness of gene expression. Moreover, it is likely that the two proteins differ in their interactions with other proteins and in their post-translational modifications, allowing different signalling pathways to regulate activity of the two proteins. Since the relative expression of TIA1 and TIAL1 varies among human tissues [16], this could allow different signalling pathways to modulate expression of TIA-target RNAs in different tissues.
TIA crosslink clusters were identified downstream of 5% of alternative and constitutive exons, suggesting that TIA proteins play a widespread role in 5′ splice site recognition. However, in spite of this widespread TIA binding, we were able to use consistent rules to predict the effects of TIA binding on splicing of regulated alternative exons. Assessing TIA crosslinking at the 5′ splice sites of both the alternative and the preceding exon enabled us to distinguish between silenced and enhanced exons. Splice-junction microarray analyses and minigene experiments validated the accuracy of these predictions on a genome-wide scale and at the level of individual regulated exons, supporting the finding that the position of TIA binding on pre-mRNA determines its dual splicing effects.
The ability of iCLIP in predicting the dual TIA splicing effects is particularly noteworthy in the case of variable-length exons, since the distance between alternative splice sites is often very short, as shown in the example of CLIP4 pre-mRNA (Figure 1C). The ability of iCLIP to directly identify crosslink sites and thereby resolve TIA binding at such proximal sites was crucial for the accuracy in separating silenced from enhanced variable-length exons.
In the present study, we have evaluated both the local and distal splicing effects of TIA binding at 5′ splice sites. Locally, TIA proteins can regulate alternative 5′ splice sites in a manner consistent with the splice site competition model [41]. TIA proteins enhanced the closest upstream 5′ splice site, leading to a concomitant decrease in usage of the competing 5′ splice site (Figure 7A). Similarly, TIA binding downstream of a cassette exon acts by promoting the usage of its 5′ splice site (Figure 7B). A past study evaluating splicing intermediates showed that in cases of Nova binding downstream of enhanced exons, the downstream intron is removed prior to the upstream one [4]. This result is consistent with a splice site competition model, indicating that Nova and TIA proteins can enhance cassette exons by promoting the splicing pathway that uses the 5′ splice site of the alternative exon, with concomitant decrease in the exon skipping pathway that uses the 5′ splice site of the preceding exon (Figure 7B).
We also found that TIA binding can lead to distal splicing silencing effects. This supports the hypothesis of indirect silencing action, where an RNA-binding protein could cause a distal negative splicing effect by its local enhancing function, initially proposed to explain the same observation in the Nova RNA map [4],[25]. A recent study of PTB-RNA interactions observed an opposite scenario where the local silencing function of PTB causes a distal positive splicing effect [23]. These findings point to the observation that a local splicing function of RNA-binding proteins generally leads to reciprocal distal splicing effects. The study analysing the distal effects of PTB proposed a model where competition between the constitutive and the alternative 5′ splice site was responsible for these distal effects [23].
To gain further insights into the distal regulation of alternative splicing, we analysed the effect of TIA binding on distal alternative splice sites. Surprisingly, we found that TIA proteins regulate usage of alternative 3′ splice sites without binding directly at the 3′ splice sites. This distal effect does not involve a competition between the constitutive and the alternative 5′ splice sites. Instead, TIA proteins regulate the distal alternative 3′ splice sites by modulating recognition of the upstream 5′ splice site (Figure 7C). This result was also supported by the minigene experiment, which showed that a constitutive non-TIA dependent 5′ splice site promotes skipping of the variable portion of the distal exon. This suggests that regulation of a 5′ splice site recognition can affect splicing of downstream alternative 3′ splice sites even if it doesn't compete with another 5′ splice site. This effect could also contribute to the splicing regulation of distal cassette exons (Figure 7D).
Unlike Nova, FOX2, and PTB proteins [4],[23],[24],[42], which bind at positions close to either 3′ or 5′ splice sites, TIA RNA maps did not identify significant binding at 3′ splice sites or within the alternative exons. Instead, TIA binding was enriched only downstream of the exons, where TIA proteins were reported to recruit U1 snRNP to the 5′ splice sites [9]–[11]. It is clear that the uridine-rich motifs downstream of exons are not the only determinant of TIA binding, since these motifs are present also in the polypyrimidine tracts upstream of exons. It is therefore possible that TIA binds to RNA cooperatively in complex with U1 snRNP, which would ensure TIA binding only downstream of exons.
Interestingly, TIA crosslinking was equally common downstream of constitutive and alternative exons. Past studies found that uridine tracts are among the most enriched motifs downstream of constitutive exons, but the function of TIA binding to these motifs was not validated [15],[43]. We found that TIA binding downstream of constitutive exons often promotes efficient splicing of the corresponding intron. Interestingly, this function is shared with the yeast orthologue Nam8p, suggesting that it represents a primary evolutionary function of the TIA proteins [44]–[46].
We find that TIA proteins can cause a distal splicing effect by regulating recognition of a constitutive 5′ splice site, even if this site does not compete with an alternative 5′ splice site (Figure 7C). Several models of splicing regulation could account for this effect. TIA proteins might change the conformation of the U1 snRNP complex in a way that promotes its pairing with the intron-distal 3′ splice site. Alternatively, binding of TIA proteins could lead to a change in the long-range RNA-RNA interactions, or a change in interactions with other RNA-binding proteins that bind at a distal site. Finally, the result could also be explained in light of the splicing kinetics model. Kinetic parameters of splicing were shown previously to influence the splice site choice [41]. By enhancing U1 snRNP recruitment to the 5′ splice site, TIA proteins might allow the splicing reaction to proceed faster, thereby shortening the time available for definition of the downstream alternative exon. In contrast, the slower splicing kinetics in knockdown cells could allow additional time for trans-acting factors, such as SR proteins, to recognise exonic elements that prevent skipping of the alternative exon (Figure 7C, D) [47]. Such effects of splicing kinetics might be related to the transcriptional effects on splicing, which act partly by changing the ability of SR proteins to define the alternative exons [48]. Taken together, this study of TIA proteins highlights the importance of identifying the positions of protein-RNA interactions with high precision in order to reveal the full complexity of splicing regulation.
HeLa cells were irradiated with UV light. Upon cell lysis, RNA was partially fragmented using RNase I. For iCLIP, TIA1 or TIAL1 were immunoprecipitated with protein G Dynabeads (Invitrogen) conjugated to goat-anti TIA1 (Santa Cruz, C-20) or goat-anti TIAL1 (Santa Cruz, C-18) antibody. For iCLAP, the Strep/His-tagged proteins were affinity purified using Streptavidin and Cobalt beads. RNA was ligated at 3′ ends to an RNA adapter and radioactively labelled on beads. After gel electrophoresis and nitrocellulose membrane transfer, protein-RNA complexes were visualised by autoradiogram. RNA was recovered by proteinase K digestion and reverse transcribed using primers with adapter regions separated by a BamHI restriction site and a barcode region at their 5′ end. cDNA was size-purified, circularised, annealed to an oligonucleotide complementary to the restriction site, and digested with BamHI. Linearised cDNA was then PCR-amplified using primers complementary to the adapter regions and subjected to high-throughput sequencing using Illumina GA2. A more detailed description is available in Text S1.
The sequences corresponding to the individual experiment were identified by their defined barcode, the random barcodes were registered, and the barcodes were removed before mapping the sequences to the human genome sequence (version Hg18/NCBI36), allowing one mismatch using Bowtie version 0.10.1 (command line: -a -m 1 -v 1).
The randomisation was done within co-transcribed regions that were expected to have the same expression levels. For instance, a single gene contains abundant exonic and non-abundant intronic RNA, and intronic RNA contains non-coding RNA genes that are usually highly abundant. We have randomised positions within each individual intron, excluding the non-coding RNA genes. Each non-coding RNA gene was randomised separately. Exons were grouped into one single region for randomisation. However, as evident in Figure 2B, the 3′ UTR contains a higher cDNA enrichment than the coding sequence; therefore, we randomised positions in each UTR region separately from the coding sequence. All annotations were based on the version Hg18/NCBI36 of the human genome sequence. Each cDNA was considered independent when randomising the positions.
Only crosslink positions (but not the cDNA count) were compared between the different datasets. Crosslink positions in the first dataset define the position of 0 when analysing the positions in the second dataset.
For sequence analysis of iCLIP crosslink sites, the position of the crosslinking site was extended 10 nt in both directions. The z score for pentamer enrichment at the 21 nt region surrounding the crosslink sites was then calculated relative to randomised genomic positions. A more detailed description is available in Text S1.
This followed the same statistical approach as the analysis of CLIP sequence clusters [42] with a few modifications as described in [28]. Rather than combining crosslink sites into larger clusters, the crosslink sites within the clusters were kept individually in order to preserve the nucleotide resolution of the data. A more detailed description is available in Text S1.
Three different oligonucleotides were used together with a scrambled control (Invitrogen, 12935-112). The following siRNA oligonucleotides were used:
siTIA1-1: 5′-GCAAGUUCCUGCAUAUGGAAUGUAU-3′
siTIA1-2: 5′-AGAAUAUCAGAUGCCCGAUGGUAA-3′
siTIA1-3: 5′-GGCAACAGGAAAGUCUAAGGGAUAU-3′
siTIAL1-1: 5′-CGGAUAUGGUUGGCAAGUUACCAA-3′
siTIAL1-2: 5′-CCGAACCAAUUGGGCCACUCGUAAA-3′
siTIAL1-3: 5′-GCGUCUGGGUUAACAGAUCAGCUUA-3′
5 nM of each siRNA (siTIA1 and siTIAL1) were transfected using Lipofectamine iMax (Invitrogen) according to manufacturer's instructions. Cells were transfected again 2 d after the first transfection and were harvested 2 d after the second transfection for protein and RNA analyses.
The protein concentration was determined using Lowry's Assay (Bio-RAD). Goat anti-TIA1 antibody (1∶1000) (Santa Cruz, C-20) and anti-TIAL1 antibody (1∶1000) (Santa Cruz, C-18) were used to detect TIA1 and TIAL1 protein. Rabbit anti-GAPDH (1∶5000) (Cell signalling) was used for loading control. For overexpression, mouse anti-Strep tag antibody (1∶1000) (Qiagen) was used to detect the tagged protein. Donkey anti-goat HRP, goat anti-rabbit HRP, and goat anti-mouse HRP (Invitrogen) were used as secondary antibodies. The membrane was visualised using ECL kit (Amersham).
A total of six samples were used, three from the siRNA control and three from KD3. The high-resolution splice-junction microarrays were produced by Affymetrix, monitoring 260,488 exon-exon junctions (each with eight probes) and 315,137 exons (each with 10 probes). cDNA samples were prepared using the GeneChip WT cDNA Synthesis and Amplification Kit (Affymetrix). Analysis of microarray data was done using version 3 of ASPIRE (Analysis of SPlicing Isoform Reciprocity). ASPIRE predicts splicing changes from reciprocal sets of microarray probes that recognise either inclusion or skipping of an alternative exon. In version 3 the background detection levels are experimentally determined for each probe, allowing background subtraction in a probe-specific manner [28]. Description of RT-PCR validation is available in Text S1.
RNA maps were produced by assessing the positioning of clustered crosslink sites at the exon/intron boundaries of alternative and flanking exons. For each exon, the positions between 20 nt of exonic and 60 nt of intronic sequence for each of the splice sites were analysed. When introns or exons were shorter than two times the length of the analysed area, analysis was restricted to region up the middle of the intron and exon. To draw the RNA map, the percentage of exons containing a crosslink site at the corresponding position is drawn.
The alternative exons with the flanking constitutive exons were amplified by PCR with genomic DNA from HeLa cells and cloned into pcDNA3 vectors. The 40 nt downstream of the exon-intron boundaries where TIA binding site are present were mutated to sequences with no observed TIA binding sites. In the case of the enhanced cassette exon (OGT1) and the alternative 5′ splice sites (CHD9), the 40 nt were replaced with sequences downstream of a silenced cassette exon (CDC25C). In the case of the alternative 3′ splice sites (C3orf23), the 83 nt were replaced with sequences downstream of GAPDH exon 2. The intron of C3orf23 was too long, so only 600 nt from the exon-intron boundaries on either side were cloned. The constructs were transfected together with GFP, TIA1, or TIAL1-pcDNA3/Step-His vectors using polyfect (Qiagen). The siRNA for TIA1 (KD3) and negative control were transfected at the same time using Lipofectamine iMax (Invitrogen). The cells were collected 2 d later and the splicing effects were assessed by RT-PCR using a T7 forward primer (5′-TAATACGACTCACTATAGGG-3′) and gene-specific reverse primers (Table S4).
|
10.1371/journal.ppat.1001001 | Murine Gamma-Herpesvirus 68 Hijacks MAVS and IKKβ to Initiate Lytic Replication | Upon viral infection, the mitochondrial antiviral signaling (MAVS)-IKKβ pathway is activated to restrict viral replication. Manipulation of immune signaling events by pathogens has been an outstanding theme of host-pathogen interaction. Here we report that the loss of MAVS or IKKβ impaired the lytic replication of gamma-herpesvirus 68 (γHV68), a model herpesvirus for human Kaposi's sarcoma-associated herpesvirus and Epstein-Barr virus. γHV68 infection activated IKKβ in a MAVS-dependent manner; however, IKKβ phosphorylated and promoted the transcriptional activation of the γHV68 replication and transcription activator (RTA). Mutational analyses identified IKKβ phosphorylation sites, through which RTA-mediated transcription was increased by IKKβ, within the transactivation domain of RTA. Moreover, the lytic replication of recombinant γHV68 carrying mutations within the IKKβ phosphorylation sites was greatly impaired. These findings support the conclusion that γHV68 hijacks the antiviral MAVS-IKKβ pathway to promote viral transcription and lytic infection, representing an example whereby viral replication is coupled to host immune activation.
| Innate immunity represents the first line of defense against pathogen infection. Recent studies uncovered an array of sensors that detect pathogen-associated molecular patterns and induce antiviral cytokine production via two closely related kinase complexes, i.e., the IKKα/β/γ and TBK-1/IKKε. To counteract host immune defense, herpesviruses have evolved diverse strategies to evade, manipulate, and exploit host immune responses. Here we report that infection by murine gamma-herpesvirus 68 (γHV68), a model gamma-herpesvirus for human Kaposi's sarcoma-associated herpesvirus and Epstein-Barr virus, activated the IKKβ kinase and IKKβ was usurped to promote viral transcriptional activation. As such, uncoupling IKKβ from transcriptional activation by biochemical and genetic approaches impaired γHV68 lytic replication. Our study represents an example whereby viral lytic replication is coupled to host innate immune activation and sheds light on herpesvirus exploitation of immune responses.
| Host cells activate innate immune signaling pathways to defend against invading pathogens. Pattern recognition receptors, including Toll-like receptors and cytosolic sensors (such as NOD-like receptors and RIG-I-like receptors), recognize pathogen-associated structural components and initiate signal transduction that leads to the biosynthesis and secretion of pro-inflammatory cytokines and interferons, thereby mounting a potent host immune response [1], [2]. To survive within an infected host, viruses have evolved intricate strategies to counteract host immune responses. Herpesviruses and poxviruses have large genomes and therefore have the capacity to encode numerous proteins that modulate host immune responses.
Mitochonrial antiviral signaling (MAVS, also known as IPS-1, VISA, and CARDIF) protein serves as an adaptor to activate both the NFκB and interferon regulatory factor (IRF) pathways [3], [4], [5], [6]. MAVS relays signals from RIG-I and MDA-5, cytosolic sensors that recognize viral dsRNA or ssRNA bearing 5′-triphosphate [7], [8], to the IKKα/β/γ and TBK-1/IKKε (also known as IKKi) kinase complexes [4], [6]. IKKα/β, together with the scaffold protein IKKγ, phosphorylates the inhibitor of NFκB (IκB) and promotes its subsequent ubiquitination and degradation by the proteasome, thereby unleashing NFκB that translocates into the nucleus to activate gene expression of pro-inflammatory cytokines [9], [10]. By contrast, TBK-1 and IKKε directly phosphorylate a serine/threonine-rich sequence within the carboxyl termini of IRF3 and IRF7, leading to the dimerization and nuclear translocation of these transcription factors [11], [12]. Together with NFκB and c-Jun/ATF-2, IRF3 and IRF7 bind to the interferon (IFN)-β enhancer and initiate the transcription of IFN-β [13], [14]. Ultimately, these signaling events promote cytokine and interferon production, establishing an antiviral state in infected cells. Although it is not clear how MAVS activates these immune kinases, recent findings have established the vital roles of MAVS in host antiviral innate immunity [15]. Interestingly, the mitochondrial localization of MAVS is critical for its ability to activate downstream signaling events. As such, various RNA viruses, exemplified by human hepatitis C virus (HCV), encode proteases that cleave MAVS from the outer membrane of the mitochondrion, thereby disarming MAVS-dependent signaling cascades and the host antiviral innate immunity [6], [16], [17], [18].
Murine gamma-herpesvirus 68 (γHV68 or MHV-68) is closely related to human Kaposi's sarcoma-associated herpesvirus (KSHV) and Epstein-Barr virus (EBV) [19]. KSHV and EBV are lymphotropic DNA viruses that are causally linked to malignancies of lymphoid or endothelial/epithelial origin, including lymphoma, nasopharyngeal carcinoma, and Kaposi's sarcoma [20], [21]. Persisting within host immune cells, KSHV and EBV are known to evade, manipulate, and exploit host immune pathways [22], [23]. Emerging studies suggest that γ-herpesviruses may usurp host innate immune responses for their infection [24], [25], [26]. However, it is not known how human KSHV and EBV manipulate innate immune pathways in vivo. Such investigations are greatly hampered by the lack of permissive cell lines and animal models for both KSHV and EBV. By contrast, γHV68 infection in laboratory mice leads to a robust acute infection in the lung and a long-term latent infection in the spleen. For murine γHV68 and human KSHV, the replication and transcription activator (RTA, encoded by ORF50) is necessary and sufficient to initiate lytic replication from latently-infected cells, supporting the notion that RTA integrates diverse signaling pathways to initiate lytic replication [27], [28], [29]. Using γHV68 as a surrogate for human KSHV and EBV, we have unexpectedly discovered that γHV68 activated IKKβ to phosphorylate RTA and promote RTA transcriptional activation, thereby increasing viral gene transcription and lytic replication. As such, RTA phosphorylation by IKKβ couples γHV68 gene expression and lytic replication to host innate immune activation, representing the first example whereby a virus hijacks the antiviral MAVS-IKKβ pathway to promote its lytic replication.
To investigate the roles of MAVS in γHV68 infection, wild-type (MAVS+/+), heterozygous (MAVS+/−), and knockout (MAVS−/−) mice were intranasally (i.n.) infected with 40 plaque-forming unit (PFU) γHV68. γHV68 acute infection in the lung was measured by plaque assays at 4, 7, 10, 13, and 16 days post-infection (d.p.i.). In MAVS+/+ mice, γHV68 titers peaked at 7 d.p.i. with approximately 500 PFU/lung and declined to 100 PFU/lung at 10 d.p.i. Viral load was undetectable by 13 d.p.i., indicating that γHV68 acute infection in the lung had been cleared (Figure 1A). Similar viral loads in the lungs of heterozygous mice (MAVS+/−) were observed (data not shown). Surprisingly, although viral loads at 7 d.p.i. in the lungs of MAVS−/− mice were comparable to those in the lungs of MAVS+/+ mice, γHV68 was nearly undetectable at 10 d.p.i. (Figure 1A). By contrast, γHV68 latent infection as characterized by viral genome frequency, persistent lytic replication, and reactivation was similar in splenocytes of MAVS+/+ and MAVS−/− mice at 16 and 45 d.p.i. (Figure S1). These observations suggest that MAVS plays a specific role(s) in γHV68 acute infection.
To determine whether γHV68 infection altered MAVS expression, we infected BL/6 mice intranasally with a high dose (1×105 PFU) of γHV68, presumably permitting synchronized and maximal infection of lung epithelial cells. MAVS mRNA levels were determined by quantitative real-time PCR (qRT-PCR). The levels of MAVS mRNA were transiently increased at 2.5 and 5 d.p.i. in the lung and spleen, respectively (Figure S2A). Interestingly, the up-regulation of MAVS mRNA preceded that of viral RTA mRNA (Figure S2A and S2B), and that higher viral RTA mRNA levels tightly correlated with higher MAVS mRNA levels at 2.5 and 5 d.p.i., when MAVS mRNA levels peaked in the lung and spleen (Figure S2C). Together with the reduced viral load in the lungs of MAVS−/− mice (Figure 1A), these results suggest that MAVS is necessary for efficient lytic replication in mice and that the transiently induced MAVS expression by γHV68 infection may facilitate viral lytic replication in vivo.
To investigate the roles of MAVS in γHV68 infection, we then assessed the effects of MAVS-deficiency on γHV68 lytic replication ex vivo. Mouse embryonic fibroblasts (MEFs) were infected with a GFP-marked recombinant γHV68 (γHV68 K3/GFP) and viral replication was examined by fluorescence microscopy and plaque assays. Surprisingly, γHV68 displayed delayed replication kinetics in MAVS−/− MEFs compared to MAVS+/+ MEFs at multiplicities of infection (MOI) of 0.01 and 0.1 (Figure 1B, 1C and S3). To quantitatively determine the effect of MAVS on γHV68 lytic infection, we examined γHV68 lytic replication in MAVS+/+ and MAVS−/− MEFs by plaque assays. In fact, γHV68 formed approximately four-fold more plaques in MAVS+/+ MEFs than those in MAVS−/− MEFs, indicative of reduced initiation of lytic replication in MAVS-deficient MEFs (Figure 1D, S4A, and S4B). Interestingly, the plaque size of γHV68 was equivalent in MAVS+/+ and MAVS−/− MEFs (Figure S4C and S4D). To test whether MAVS−/− MEFs are defective in supporting viral lytic replication in general, we examined the lytic replication of vesicular stomatitis virus (VSV), a prototype RNA virus, with a plaque assay. Consistent with an antiviral activity of MAVS against RNA viruses, VSV formed 10-fold more plaques in MAVS−/− MEFs than those in MAVS+/+ MEFs (Figure 1D). The diminished lytic replication of γHV68 in MAVS-deficient MEFs is consistent with the reduced acute infection observed in the lung. To test whether exogenously expressed MAVS is able to restore γHV68 lytic replication, we generated lentivirus in 293T cells and MEFs stably expressing human MAVS (hMAVS) was established with puromycin selection (Figure 1E). As shown in Figure 1F and 1G, exogenous hMAVS restored γHV68 lytic replication by a plaque assay and multi-step growth curves. Nevertheless, these results together support the conclusion that MAVS is necessary for efficient γHV68 lytic replication in vivo and ex vivo.
Two known pathways, the IKKα/β/γ-NFκB and TBK-1/IKKε-IRF3/7 pathway, have been characterized downstream of MAVS (Figure 2A) [3], [4]. We therefore used MEFs deficient in key components of aforementioned pathways to identify downstream effectors of MAVS that are critical for γHV68 lytic infection. Plaque assays and multi-step growth curves of γHV68 lytic infection showed that deficiency in TRAF6, IKKγ, and IKKβ, but not deficiency in the closely related IKKα, recapitulated phenotypes of MAVS deficiency (Figure 2B and 2C). Notably, TRAF6 is necessary for MAVS to activate IKKβ that requires IKKγ, a scaffold protein for both IKKα and IKKβ [5]. By contrast, deficiency in type I IFN receptor (IFNAR) and double deficiency in IRF3 and IRF7 had no discernable effect on the plaque numbers of γHV68 in MEFs, indicating that the IRF-IFN signaling pathway is not critical for the initiation of γHV68 lytic replication (Figure 2B). Furthermore, the exogenous IKKβ expression reconstituted by lentivirus restored the lytic replication of γHV68 as determined by a plaque assay and multi-step growth curves (Figure 2D, 2E, and 2F). Interestingly, the expression of IKKβ in MAVS−/− did not increase γHV68 lytic replication by a plaque assay (Figure 2E), suggesting that the MAVS-dependent activation of IKKβ, rather than the absolute expression level of IKKβ, is crucial for efficient γHV68 lytic replication. Additionally, exogenous IKKβ did not increase γHV68 plaque numbers in MAVS+/+ MEFs (Figure 2E), implying that endogenous IKKβ is sufficient to support efficient γHV68 lytic replication. Of note, lentivirus infection reduces the difference of γHV68 plaque forming capacity in wild-type MEFs and in MEFs deficient in MAVS and IKKβ (Figure 1F and 2D). Collectively, these data indicate that the MAVS-dependent IKKβ activation is critical for efficient γHV68 lytic replication.
To assess whether the kinase activity of IKKβ is important for γHV68 lytic infection, we performed plaque assays with or without the specific IKKβ inhibitor, Bay11-7082 (Bay11). This experiment revealed that Bay11 reduced the plaque number of γHV68 in a dose-dependent manner (Figure 3A). Whereas treatment with 1 µM of Bay11 at 0.5 h before infection reduced γHV68 plaque number by 52%, the same treatment at 7 h post-infection (h.p.i.) reduced the plaque number by 29%, emphasizing the important roles of IKKβ during early γHV68 infection (Figure 3A). We further examined IKKβ activity by an in vitro kinase assay with IKKβ precipitated from MAVS+/+ and MAVS−/− MEFs infected with γHV68. The IKKβ kinase activity was transiently and moderately increased in MAVS+/+ MEFs, however, it was drastically diminished in MAVS−/− MEFs after γHV68 infection (Figure 3B). The activation of IKKβ was further supported by the rapid degradation of IκBα concurrent to IKKβ activation by γHV68 infection in MAVS+/+ MEFs, but not in MAVS−/− MEFs (Figure 3C). To test whether UV-inactivated virus is able to trigger IKKβ activation, we examined the levels of IKKβ kinase activity and IκBα in MAVS+/+ MEFs by in vitro kinase and immunoblot assays, respectively. Interestingly, UV-inactivated γHV68 activated IKKβ and reduced IκBα protein levels, although less efficiently than live γHV68 (Figure 3D and 3E). This observation suggests that γHV68 lytic replication is necessary to activate the MAVS-IKKβ pathway. Alternatively, UV treatment may damage or disrupt viral structural components whose integrity is necessary to activate the MAVS-IKKβ pathway.
MAVS activation by RNA viruses is known to increase the expression of pro-inflammatory cytokines and interferons. However, γHV68 appears to be a poor inducer for these antiviral molecules, suggesting that γHV68 evades signaling events downstream of the MAVS adaptor. Indeed, γHV68 infection failed to up-regulate the expression of IFN-β (Figure S5A). In agreement with this observation, γHV68 RTA, similar to KSHV RTA [30], is sufficient to reduce IRF3 expression (Figure S5B). Meanwhile, it was previously shown that γHV68 infection did not significantly activate NFκB during early infection [31], suggesting that γHV68 uncouples NFκB activation from activated IKKβ. Taken together, these results support the conclusion that γHV68 infection selectively activates IKKβ to promote viral lytic replication.
To discern the molecular mechanisms underlying the requirement of the MAVS-IKKβ pathway in γHV68 lytic infection, levels of γHV68 genomic DNA and mRNA were assessed by PCR or reverse transcription followed by real-time PCR analyses, respectively. At a low MOI (0.01), analyses by PCR (Figure 4A) and real-time PCR (Figure 4B) revealed comparable levels of viral genomes in MAVS+/+ and MAVS−/− MEFs early after de novo infection, suggesting comparable viral entry into MAVS+/+ and MAVS−/− MEFs. Interestingly, levels of viral mRNA transcripts representing immediate early (RTA, ORF73, and ORF57) and early (ORF60 and ORF9) gene products in MAVS+/+ MEFs were higher than those in MAVS−/− MEFs as determined by reverse-transcriptase PCR (Figure 4C). Real-time PCR analyses with cDNA showed approximately 4- to 16-fold higher levels of γHV68 mRNA transcripts in MAVS+/+ MEFs compared to those in MAVS−/− MEFs at 2 and 3 d.p.i. (Figure 4D). It has been shown that TRAF6 is necessary for MAVS to activate IKKβ [5] and exogenous TRAF6 is sufficient to activate IKKβ. To further examine the effects of the MAVS-IKKβ pathway on levels of γHV68 mRNA transcripts, a bacterial artificial chromosome (BAC) containing the γHV68 genome and a plasmid expressing TRAF6 were transfected into 293T cells. The effects of exogenous TRAF6 (that activates IKKβ) on viral transcription were determined by reverse transcription and real-time PCR. At 28 h post-transfection, a time point when immediate early and early genes are transcribed, exogenous TRAF6 efficiently increased the mRNA levels of γHV68 RTA, ORF57, ORF60, and ORF73, without discernable effect on levels of viral genomic DNA (Figure 4E and 4F). These results, obtained under conditions of loss of function (MAVS−/− MEFs) and gain of function (TRAF6 expression), indicate that the activated IKKβ increases the levels of γHV68 mRNA transcripts.
MAVS is an adaptor that activates IKKβ and the MAVS-dependent IKKβ increases γHV68 mRNA levels. We thus postulated that MAVS influences γHV68 transcription via its downstream IKKβ on RTA, because RTA, the master transcription activator, is critical for γHV68 lytic replication. To test this hypothesis, we examined whether IKKβ phosphorylates γHV68 RTA. IKKβ was purified from 293T cells and bacterial GST fusion proteins containing the RTA internal region (RTA-M, aa 335–466) or the RTA C-terminal transactivation domain (RTA-C, aa 457–583) were purified from E.coli (Figure 5A). In the presence of [32P]γATP, IKKβ efficiently transferred the phosphate group to GST-RTA-C. By contrast, GST was not phosphorylated and GST-RTA-M was weakly phosphorylated by IKKβ. Furthermore, the kinase domain deletion variant of IKKβ (IKKβΔKD) failed to phosphorylate GST-RTA-C and GST-RTA-M (Figure 5A), and IKKα had only residual kinase activity toward RTA-C (Figure S6). To confirm the MAVS- and IKKβ-dependent phosphorylation of RTA, RTA phosphorylation in γHV68-infected cells was analyzed by autoradiography and immunoblot. We found that MAVS- and IKKβ deficiency reduced RTA phosphorylation by 50% and 85%, respectively, while reconstituted IKKβ expression restored RTA phosphorylation to that of RTA in MAVS+/+ MEFs (Figure 5B). To assess the roles of phosphorylation of RTA in transcription regulation, luciferase reporter assays were carried out with plasmids containing RTA-responsive promoters of RTA, ORF57, and M3. As shown in Figure 5C, the transcription activity of RTA on all three promoters was significantly increased by exogenous TRAF6 and IKKβ, but not by the kinase dead variant IKKβΔKD, supporting the notion that IKKβ promotes RTA transcription activation via phosphorylation. When expressed to similar levels of IKKβ, IKKβΔKD had no significant effect on RTA transcriptional activation (Figure S7). Given that RTA is a substrate for IKKβ, we sought to examine whether RTA can physically associate with the IKKα/β/γ complex. However, we were unable to detect interaction between RTA and any of the three subunits of IKKα/β/γ by co-immunoprecipitation (data not shown), suggesting that the RTA interaction with the IKKα/β/γ complex is transient or mediated via additional cellular proteins.
To identify IKKβ phosphorylation sites, series of truncations from the C-terminus of RTA were constructed and purified as GST fusion proteins for in vitro kinase assays with IKKβ. These experiments demonstrated that the IKKβ phosphorylation sites were located within the region containing residues 540 through 567 (Figure S8). Given that IKKβ is a serine/threonine kinase, clusters of various serine/threonine residues were changed to alanines and RTA phosphorylation was assessed similarly. Two clusters of mutations, replacement of S550T552S556 (STS/A) and T561T562S564 (TTS/A) by alanines, reduced the phosphorylation levels of RTA-C by approximately 72% and 45%, respectively (Figure 5D and S8). These results indicate that the STS and TTS sequences represent two major IKKβ phosphorylation sites within the transactivation domain of RTA.
To further examine the roles of IKKβ phosphorylation in regulating RTA transcription activity, reporter assays with plasmids containing wild-type RTA, the STS/A and TTS/A variants were carried out with exogenously expressed IKKβ. The STS/A and TTS/A variants had lower basal activity to activate promoters of RTA, ORF57, and M3. Moreover, exogenous IKKβ failed to further stimulate the transcription activities of the STS/A and TTS/A variants to activate promoters of RTA and ORF57 (Figure 5E). Interestingly, the STS/A variant activated M3 promoter to the level of wild-type RTA with or without IKKβ, indicating that the STS site is dispensable for IKKβ to promote RTA transcriptional activity on the M3 promoter (Figure 5E). It is noteworthy that the STS/A and TTS/A variants were expressed at higher levels than wild-type RTA, the transcription activities of the STS/A and TTS/A variants were approximately 50% and 20% of that of wild-type RTA, respectively, when luciferase activity was normalized against protein levels (Figure 5F). Collectively, these results demonstrated that IKKβ promotes RTA transcriptional activation via phosphorylation of the TTS and STS sites within the transactivation domain.
To further investigate the roles of RTA phosphorylation, we assessed the effects of the STS/A and TTS/A mutations on γHV68 lytic replication. Taking advantage of the γHV68-containing BAC with a transposon insertion that inactivates RTA (ORF50 Null) [32], a recombination-based strategy [33] was employed to generate viruses carrying wild-type RTA (Null Rescued, designated NR), the STS/A allele, or the TTS/A allele (Figure 6A). Whereas we easily obtained recombinant γHV68 containing wild-type RTA (γHV68.NR) or the TTS/A allele (γHV68.TTS/A), the STS/A variant failed to support γHV68 recombination in multiple independent experiments. This observation suggests an essential role for the phosphorylated STS sequence in γHV68 lytic replication. To confirm the integrity of viral genomic DNA, we performed restriction digestion with KpnI and EcoRI, and analyzed with agarose gel electrophoresis. As expected, the removal of the Kanamycin cassette within RTA alleles reduced the 9-kb fragment to 7.5-kb counterpart released by KpnI digestion (Figure 6B), and abolished an EcoRI site within the Kanamycin cassette (Figure 6C). To assess the transcriptional activity of RTA derived from BAC DNA, BAC DNA and the M3p luciferase reporter plasmid were transfected into 293T cells and RTA transcriptional activity was assessed by luciferase reporter assay. The activity of wild-type RTA to activate M3 promoter was approximately 6-fold higher than that of the TTS/A mutant (Figure 6D). Using 293T cells transfected with the γHV68 BAC containing the TTS/A allele and a plasmid expressing TRAF6, we assessed the effects of TRAF6 (that activates IKKβ) on γHV68 gene expression. In contrast to what was observed for the γHV68 BAC containing wild-type RTA (Figure 4F), exogenous TRAF6 had marginal effects on the levels of viral mRNAs transcribed from γHV68 BAC containing the TTS/A allele (Figure 6E). These findings are consistent with the observation that IKKβ failed to further promote the transcription of the TTS/A variant (Figure 5E), supporting the conclusion that the TTS residues constitute an IKKβ phosphorylation sequence by which RTA-dependent transcription is positively regulated.
Next, we examined whether recombinant γHV68.TTS/A recapitulates the defects of wild-type γHV68 lytic replication in MEFs deficient in MAVS and IKKβ (plaque assays and multi-step growth curves). To assess the effects of the TTS/A mutation on γHV68 transcription activation, we normalized viral genomes immediately after γHV68 de novo infection of MEFs by qRT-PCR. With equal number of viral genomes, γHV68.NR displayed approximately 32-fold higher of RTA mRNA than recombinant γHV68.TTS/A in MAVS+/+ MEFs at 30 h.p.i. (Figure 7A). This is consistent with the observation that RTA activates its own promoter to facilitate viral lytic replication (Figure 5C and 5E). Furthermore, multi-step growth curves (at an MOI of 0.01) demonstrated that γHV68.TTS/A had delayed replication kinetics and produced >3 orders of magnitude less virion progeny in MAVS+/+ MEFs (Figure 7B). To test whether RTA phosphorylation and the MAVS-IKKβ pathway are functionally redundant, we examined the replication kinetics of recombinant γHV68.NR and γHV68.TTS/A in wild-type, MAVS−/−, and IKKβ−/− MEFs. Consistent with our previous observations (Figure 1C, 2C, and S3B), γHV68.NR showed delayed lytic replication in MAVS−/− and IKKβ−/− MEFs (Figure 7B and 7C). Remarkably, γHV68.TTS/A replicated with similar kinetics in wild-type, MAVS−/−, and IKKβ−/− MEFs, suggesting that the MAVS-IKKβ pathway functions on RTA to promote viral lytic replication (Figure 7B and 7C). However, these replication defects of recombinant γHV68 carrying the TTS/A mutation are much more pronounced than the phenotypes of wild-type γHV68 in MAVS−/− and IKKβ−/− MEFs, implying that additional kinases may influence RTA transcriptional activation via phosphorylation of the TTS site. Taken together, we conclude that the TTS site of RTA is likely phosphorylated by IKKβ and is crucially important for γHV68 lytic replication.
Here we provide evidence that murine γHV68 hijacks the antiviral MAVS-IKKβ pathway to promote its lytic replication. The MAVS adaptor is important for host defense against invading pathogens, including various DNA and RNA viruses. For example, mice lacking MAVS were severely compromised in innate immune defense against VSV infection, leading to an elevated peak viral load and prolonged acute viral infection [34]. The antiviral effects of MAVS have been observed against the infection of a number of RNA and DNA pathogens [35], [36], [37]. To our surprise, γHV68 viral load in the lungs of MAVS−/− mice was significantly lower than that in the lungs of MAVS+/+ mice at 10 d.p.i. The reduced viral load of γHV68 in MAVS−/− mice is counter-intuitive to the presumed antiviral function of the MAVS adaptor in promoting innate immune responses. Although type I interferons in γHV68-infected mice were undetectable [38], mice deficient in type I IFN receptor had higher viral loads and succumbed to γHV68 infection [39]. We surmise that the effects of MAVS deficiency on γHV68 acute infection is likely under-estimated, providing that MAVS is critical for interferon production in response to viral infection. Thus, the viral load of γHV68 acute infection in MAVS−/− mice likely represents a “neutralized” phenotype, in which reduced γHV68 lytic replication is compensated by the lack of type I interferon inhibition. Moreover, the observation that viral RTA mRNA levels correlates tightly with the MAVS mRNA levels during early γHV68 acute infection suggests that MAVS is necessary for γHV68 lytic replication (Figure S2). Although we have not formally excluded the contribution of host immune responses against γHV68 infection to the reduced viral load at 10 d.p.i. in MAVS−/− mice, our experiments with γHV68 replication ex vivo demonstrated critical roles of the MAVS-IKKβ pathway in facilitating γHV68 lytic infection.
During early stages of viral infection, γHV68 activated IKKβ in a MAVS-dependent manner, a signaling event that is likely triggered by a variety of pathogens. The MAVS-dependent activation was supported by elevated IKKβ kinase activity and accelerated IκBα degradation, signature signaling events downstream of the MAVS adaptor. Although the up-regulation of IKKβ kianse activity appears modest, γHV68 may direct IKKβ kinase activity to efficiently modify cellular and viral components that are critical for γHV68 infection, such as RTA. Consequently, γHV68 can harness activated IKKβ without inducing NFκB activation that may be resulted from massive IKKβ activation. Indeed, it was reported that γHV68 infection does not induce NFκB activation during early infection [40], suggesting that modest IKKβ activation is beneficial for γHV68 infection and that γHV68 may uncouple NFκB activation from IKKβ activation. Interestingly, γHV68 appears to block the interferon limb of the MAVS-dependent innate immune pathway. In fact, we found that γHV68 infection failed to induce the expression of IFN-β (Figure S5A). Consistent with this observation, γHV68 RTA, similar to KSHV RTA [30], is sufficient to reduce IRF3 protein (Figure S5B), potentially abrogating the production of interferons that otherwise would potently thwart γHV68 replication. Moreover, ORF36 was reported to deregulate the phosphorylated form of IRF3 and inhibit interferon production [41]. These observations suggest that γHV68 selectively activates the MAVS-IKKβ pathway to promote viral lytic replication.
Within this report, we have identified one requisite role of the MAVS-IKKβ pathway in γHV68 lytic replication with MEFs deficient in key components of this pathway. Phenotypically, γHV68 displayed similar replication defects in MEFs deficient in MAVS, IKKβ, and IKKγ, although the replication defects in IKKβ−/− and IKKγ−/− MEFs were more pronounced than those in MAVS−/− MEFs (Figure 1C, 2B, and 2C). This result supports the corollary that IKKβ, with the scaffold protein IKKγ, functions downstream of MAVS and likely integrates additional signaling emanating from other innate immune pathways including Toll-like receptors. It is worthy to point out that our result does not exclude the antiviral activity of the IRF-IFN pathway in γHV68 lytic replication, although deficiency of IRF3 and IRF7 or IFNAR did not appear to impact the initiation of γHV68 lytic infection as assessed by plaque assays (Figure 2B). It is possible that the IRF-IFN pathway may inhibit molecular events other than the initiation of lytic replication and reduce viral yield during γHV68 infection. Mechanistically, we identified γHV68 RTA, the master viral replication transactivator, as one of the IKKβ kinase substrates. Phosphorylation of RTA by IKKβ increases RTA transcriptional activity and consequently viral mRNA production. Indeed, γHV68 had lower levels of various mRNA transcripts that correlated with reduced lytic replication in MAVS−/− MEFs (Figure 1 and 4). Conversely, exogenous TRAF6 potentiated RTA transcriptional activity and substantially increased the levels of viral mRNA transcripts (Figure 4F and 5C). Additionally, exogenously reconstituted expression of MAVS and IKKβ restored RTA phosphorylation (Figure 5B) and restored γHV68 lytic replication (Figure 1 and 2). Moreover, lytic replication of recombinant γHV68 viruses carrying mutations within the IKKβ phosphorylation sites was greatly impaired, displaying phenotypes that are more pronounced than those of wild-type γHV68 in MEFs deficient in components of the MAVS-IKKβ pathway. Conceivably, other kinases and signaling pathways may converge to modulate RTA transcriptional activation via phosphorylation within these identified IKKβ sites. For example, virus-encoded kinases, such as the functionally conserved ORF36, may amplify the phosphorylation cascade that is initiated by the MAVS-IKKβ pathway [42]. Most importantly, RTA auto-activates its own promoter and increases RTA protein that, in turn, up-regulates the expression of numerous immediate early and early genes during γHV68 infection. Thus, the 50–80% reduction in RTA transcriptional activity of the STS/A and TTS/A variants (Figure 5F) likely translates into, through the aforementioned amplification cascades, the viral yields that are less than 0.1% of the recombinant γHV68.NR (Figure 7B). Finally, it is noteworthy that deficiency in MAVS and IKKβ and mutations within RTA exhibited distinct phenotypes (such as peak viral titers of multi-step growth curves), in addition to the shared reduction of γHV68 lytic replication. These differing effects on γHV68 infection are likely due to their unique hierarchical position within the MAVS-IKKβ-RTA signaling axis. In essence, these experiments identified novel phosphorylation sites within RTA that couples γHV68 lytic replication to the antiviral IKKβ kinase. These findings collectively demonstrate that the MAVS-dependent IKKβ kinase activity is critical for RTA transcriptional activation and γHV68 lytic replication. Interestingly, Gwack et al. reported that phosphorylation of the internal serine/threonine-rich region of KSHV and γHV68 RTA inhibited RTA transcriptional activity and suppressed viral lytic replication [43]. Together with our findings, these results indicate that site-specific phosphorylation determines the transcriptional activity, and likely the promoter-specificity, of gamma-herpesvirus RTA.
Although it is well accepted that the NFκB pathway is crucial for gamma-herpesvirus latent infection [44], the roles of this pathway in gamma-herpesvirus lytic replication appear to be inconsistent. Particularly, Krug et al. reported that the recombinant γHV68 expressing the IκBα super suppressor replicated indistinguishably compared to wild type γHV68 [31]. Thus, the authors concluded that the NFκB pathway is dispensable for γHV68 lytic replication. By contrast, it was shown that RelA, the p65 subunit of an NFκB transcription dimer, inhibits γHV68 lytic replication through suppressing RTA transcription activity in 293T cells [45]. Finally, our current report indicates that the MAVS-IKKβ pathway is necessary for efficient γHV68 lytic replication. However, the seemingly paradox can be explained by the differential effects of three distinct components of the NFκB pathway on γHV68 lytic replication. Although the IκBα super suppressor is commonly employed to inhibit the activation of the NFκB transcription factors, it is important to note that no significant NFκB activation was observed during early γHV68 infection (within the first 6 hours post-infection) [40], temporal phase in which the critical roles of IKKβ was indentified by our genetic and biochemical experiments. Conceivably, the unphosphorylatable IκBα super suppressor may not impact IKKβ kinase activity. By contrast, we have focused on the IKKβ kinase and our study indicated that the ability of IKKβ to promote viral lytic replication largely stems from IKKβ kinase activity to phosphorylate RTA and increase RTA transcriptional activation. Apparently, neither IκBα, nor RelA can do so in replace of IKKβ function. On the other hand, although RelA was shown to suppress γHV68 lytic replication [45], the lack of NFκB activation during early γHV68 infection implies that γHV68 uncouples NFκB activation from IKKβ activation, which are otherwise tightly correlated. As such, γHV68 infection may selectively activate the IKKβ kinase, while sparing the inhibition by preventing NFκB activation. Therefore, a scenario that potentially accommodates all three reports is that nuclear activated RelA is necessary to inhibit γHV68 lytic replication and γHV68 is capable of preventing RelA activation in an IκBα-independent manner. Crucial to this hypothesis is the mechanisms that γHV68 has evolved to thwart NFκB activation and future experiments are necessary to address this possibility.
It was previously reported that γHV68 was impaired for latency establishment and reactivation in MyD88-deficient mice, although the lytic replication of γHV68 appeared to be normal in these mice [26]. Moreover, agonists specific for TLR7/8, which activate downstream signaling events through MyD88, induced KSHV lytic gene expression and reactivated KSHV replication from latently-infected B cells [25]. The specific roles of MAVS in lytic replication and MyD88 in latent infection are consistent with their distinct functions in innate immune responses of epithelial cells and immune cells, respectively. Given that MyD88 also activates the IKKα/β kinase complex, it is possible that IKKβ-dependent activation of RTA may contribute to γHV68 and KSHV latent infection as well. Finally, reduced lytic replication of human KSHV and cytomegalovirus has been observed under experimental conditions in which IKKβ was inhibited by Bay11, implying that human KSHV and cytomegalovirus have evolved similar molecular mechanisms to facilitate lytic replication [46], [47], [48]. Taken together, the mechanism whereby an antiviral innate immune signaling pathway is exploited to promote viral lytic replication may be applied to other herpesviruses and viral reactivation from latency. This study thus has uncovered an intricate interplay between the viral replication transactivator, RTA, and the MAVS-IKKβ pathway. To our best knowledge, this is the first example that illustrates how a virus hijacks an antiviral signaling pathway, downstream of cytosolic sensors, to initiate its lytic replication. Perhaps, co-evolution between the persistent herpesviruses and their hosts has selected viruses that exploited the inevitable innate immune activation by viral infection. Although our current study delineates the key signaling events downstream of MAVS and IKKβ, it remains unknown what viral components and cellular factors activate the MAVS-IKKβ pathway and whether these mechanisms are shared by the oncogenic KSHV and EBV to promote lytic replication or reactivation.
For protein expression in mammalian cells, all genes were cloned into pcDNA5/FRT/TO (Invitrogen) unless specified. For protein expression and purification in E.coli, the internal region (RTA-M, aa 335–466) and C-terminal transactivation domain (RTA-C, aa 457–583) of RTA were cloned into pGEX-4T-1 (Promega) with BamHI and XhoI sites.
NIH3T3 cells, HEK293T (293T) cells and mouse embryonic fibroblasts (MEFs) were maintained in DMEM (Mediatech) with 8% newborn calf serum (NCS) or fetal bovine serum (FBS), respectively. MAVS+/+, MAVS−/−, IKKβ−/−, IKKγ−/− and TRAF6−/− MEFs were described previously [4], [34]. IKKα−/− MEFs were kindly provided by Dr. Amyn A. Habib (Neurology, UT Southwestern). IFNAR+/+ and IFNAR−/− MEFs were kindly provided by Dr. Michael Gale (Immunology, University of Washington). IRF3+/+IRF7+/+ and IRF3−/−IRF7−/− MEFs were kindly provided by Dr. Jae Jung (Microbiology, University of Southern California). γHV68 K3/GFP was kindly provided by Dr. Philip Stevenson (Cambridge University, UK). Wild-type γHV68 and γHV68 K3/GFP were amplified in NIH3T3 cells, and VSV-GFP virus was amplified in BHK-21 cells. Viral titer was determined by a plaque assay with NIH3T3 cells.
All animal experiments were performed in accordance to NIH guidelines, the Animal Welfare Act, and US federal law. The experimental protocol (entitled: Innate immune pathways in γHV68 infection) were approved by the Institutional Animal Care and Use Committee (IACUC). All animals were housed in a centralized research animal facility that is accredited by the Association of Assessment and Accreditation of Laboratory Animal Care International, and that is fully staffed with trained husbandry, technical, and veterinary personnel.
Wild-type (MAVS+/+), heterozygous (MAVS+/−), and knockout (MAVS−/−) mice were described previously [34]. Gender-matched, 6- to 8-week old littermate mice were intranasally (i.n.) inoculated with 40 plaque-forming unit (PFU) wild-type γHV68. To assess MAVS expression in the lung and spleen, BL/6 mice were intranasally infected with 1×105 PFU γHV68. The lungs and spleens were harvested and homogenized in DMEM.
Viral titer of mice tissues or cell lysates was assessed by a plaque assay on NIH3T3 monolayers. After three rounds of freezing and thawing, 10-fold serially-diluted virus supernatants were added onto NIH3T3 cells and incubated for 2 hours at 37°C. Then, DMEM containing 2% NCS and 0.75% methylcellulose (Sigma) was added after removing the supernatant. Plaques were counted at day 6 post-infection. The detection limit for this assay is 5 PFU. To assess the infectivity of γHV68 on various MEFs, a similar plaque assay was carried out with the initial cell density of 5000 cells/cm2. In Bay11-7082 treatment assay, 0.5µM or 1µM Bay11-7082 was added at 0.5 h before infection or 7 h post-infection. Supernatant was removed after 30 min incubation at 37°C, and cells were washed with medium and incubated for plaque formation.
Glutathione-S-transferase (GST) and GST fusion proteins containing the internal region and the transactivation domain of RTA were expressed with IPTG induction and purified with glutathione-sepharose as previously described [33]. Eluted proteins were re-suspended in 25% glycerol and stored at −20°C for kinase assays. To purify IKKβ and IKKβΔKD, 293T cells were transfected with pcDNA3 containing Flag-IKKβ and Flag-IKKβΔKD. At 48 h post-transfection, cells were lysed with kinase purification buffer (150 mM NaCl, 20 mM Tris.HCl pH7.4, 10% Glycerol, 0.5% Triton X-100, 0.5 mM DTT) and subject to one-step affinity purification with anti-Flag M2-conjugated agarose (Sigma). Proteins were eluted with 0.2 mg/ml Flag peptide in kinase buffer (50 mM KCl, 2 mM MgCl2, 2 mM MnCl2, 1 mM DTT, 10 mM NaF, 25 mM HEPES, pH7.5) and stored in 25% glycerol at −80°C.
Commercial antibodies used in this study include: anti-Flag (Sigma), anti-GFP (Covance), anti-IKKβ (H4), anti-IκBα (C20) (Santa Cruz Biotech.), anti-actin (Abcam.). To generate antibody to γHV68 RTA, the mixture of GST fusion proteins containing the RTA-M and RTA-C was used to immunize a rabbit and polyclonal antibodies were tested for the specificity with pre-immune serum as control.
Endogenous IKKβ or exogenously expressed IKKβ and IKKβΔKD were used for in vitro kinase assays. The kinase reaction includes 0.5 µg GST or GST fusion proteins, 100 µCi [32P]γATP, and approximately 250 ng kinase in 20 µl of kinase buffer. Reaction was incubated at room temperature for 25 min and denatured proteins were analyzed by SDS-PAGE and autoradiography.
To determine the relative levels of viral transcripts, total RNA was extracted from MEFs or mice tissues using TRIzol reagent (Invitrogen). To remove genomic DNA, total RNA was treated with RNase-free DNase I (New England Biolab) at 37°C for 1 hour. After heat inactivation, total RNA was re-purified with TRIzol reagent. cDNA was prepared with 1.5 µg total RNA and reverse transcriptase (Invitrogen). RNA was then removed by incubation with RNase H (Epicentre). Abundance of viral transcripts was assessed by qRT-PCR. Mouse β-actin was used as an internal control. Primers used in this study were summarized in Table S1.
Bulk splenocytes were re-suspended in DMEM, and plated onto primary MEF monolayers in 96-well plates in 2-fold serial dilutions (from 105 to 48 cells/well) as previously described [49]. Twelve wells were plated every dilution. Reactivation percentage was scored for cytopathic effects (CPE) positive wells on day 6. In order to measure preformed infectious virus, disrupted cells were plated onto primary MEF monolayers. This procedure destroys over 99% of the cells, but has minimal effect on preformed infectious virus, thus allowing distinction between reactivation from latency and persistent infection.
The frequency of splenocytes harboring wild-type γHV68 genome was assessed by a single-copy-sensitive nested PCR analysis of serial dilutions of splenocytes as previously described [49]. Briefly, mice spleens were homogenized and re-suspended in isotonic buffer and subjected to 3-fold serial dilutions (from 104 to 41 cells/well) in a background of uninfected RAW 264.7 cells, with a total of 104 cells per well. Twelve replicates were plated for each cell dilution. After being plated, cells were subjected to lysis by proteinase K at 56°C for 8 hours. After inactivating the enzyme for 30 minutes at 85°C, samples were subjected to nested PCR using primers specific for γHV68 ORF72. Positive controls of 10, 1, and 0.1 copies of viral DNA and negative controls of uninfected RAW 264.7 cells alone were included on each plate. Reaction products were separated using 2.5% UltraPure agarose (Invitrogen) gels and visualized by ethidium bromide staining.
Reactivation and LDPCR results were analyzed using GraphPad Prism software (GraphPad Software, San Diego, CA). The frequencies of genome-positive cells were statistically analyzed using the paired Student's t-test. The frequencies of viral genome-positive cells were determined from a nonlinear regression analysis of sigmoidal dose-response best-fit curve data. Based on a Poisson distribution, the frequency at which at least one event is present in a given population occurs at the point at which the regression analysis line intersects 63.2%. Pooled data of at least three independent experiments were used to calculate P values with the two-tailed, unpaired Student's t-test.
293T cells (2×105 cells/well) were seeded in 24-well plates 16 hours before transfection. A total of 377 ng of plasmid DNA per well was co-transfected by the calcium phosphate method (Clontech). The plasmid cocktail includes 75 ng of luciferase plasmid (RTAp_luc, ORF57p_Luc or M3p_luc), 200 ng of pCMV-β-galactosidase plasmid, 2 ng of pcDNA5_RTA and 100 ng of pcDNA5 containing TRAF6, IKKβ or IKKβΔKD. At 21 hours post-transfection, whole cell lysates were used to measure the firefly luciferase activity and β-galactosidase activity.
The bacterial artificial chromosome (BAC) system was used to generate recombinant γHV68 similarly to what was described previously [33]. Briefly, wild-type RTA or the STS/A and TTS/A alleles were PCR amplified with overlapping PCR primers. Purified PCR products, along with the BAC clone 5.15 [32] containing a transposon within the transactivation domain of RTA (between nucleotide 69269 and 69270, according to accession number U97553), were transfected into NIH3T3 cells with Lipofectamine 2000 (Invitrogen). Virus in the supernatant was further amplified with NIH3T3 cells. To isolate circular BAC DNA, NIH3T3 cells were infected with recombinant γHV68 and DNA was extracted according to Hirt's protocol [50] and electroporated into ElectroMAX DH10B cells (Invitrogen). BAC DNA containing γHV68 genome was digested with EcoRI and KpnI to rule out chromosome rearrangement. Meanwhile, the RTA alleles were amplified by PCR and sequenced to confirm desired mutations. Selected clones were transfected into NIH3T3 cells and recombinant γHV68 was amplified for subsequent experiments.
RIG-I, 230073; MDA-5, 71586; MAVS, 228607; TBK-1, 56480; IKKε, 56489; IRF3, 54131; IRF7, 54123; c-Jun, 16476; ATF-2, 11909; IFNβ, 15977; IFNAR, 15975; TRAF3, 22031; TRAF6, 22034; IKKγ, 16151; IKKα, 12675; IKKβ, 16150; IκBα, 18035; NFκB1, 18033; RelA, 19697; MyD88, 35956; TLR7, 170743; TLR8, 170744.
|
10.1371/journal.pgen.1002206 | Stress-Induced PARP Activation Mediates Recruitment of Drosophila Mi-2 to Promote Heat Shock Gene Expression | Eukaryotic cells respond to genomic and environmental stresses, such as DNA damage and heat shock (HS), with the synthesis of poly-[ADP-ribose] (PAR) at specific chromatin regions, such as DNA breaks or HS genes, by PAR polymerases (PARP). Little is known about the role of this modification during cellular stress responses. We show here that the nucleosome remodeler dMi-2 is recruited to active HS genes in a PARP–dependent manner. dMi-2 binds PAR suggesting that this physical interaction is important for recruitment. Indeed, a dMi-2 mutant unable to bind PAR does not localise to active HS loci in vivo. We have identified several dMi-2 regions which bind PAR independently in vitro, including the chromodomains and regions near the N-terminus containing motifs rich in K and R residues. Moreover, upon HS gene activation, dMi-2 associates with nascent HS gene transcripts, and its catalytic activity is required for efficient transcription and co-transcriptional RNA processing. RNA and PAR compete for dMi-2 binding in vitro, suggesting a two step process for dMi-2 association with active HS genes: initial recruitment to the locus via PAR interaction, followed by binding to nascent RNA transcripts. We suggest that stress-induced chromatin PARylation serves to rapidly attract factors that are required for an efficient and timely transcriptional response.
| Cells respond to elevated temperatures with the rapid activation of heat shock genes to ensure cellular survival. Heat shock gene activation involves the synthesis of poly-[ADP-ribose] (PAR) at heat shock loci, the opening of chromatin structure, and the coordinated recruitment of transcription factors and chromatin regulators RNA polymerase II and components of the RNA processing machinery. The molecular roles of PAR and and ATP-dependent chromatin remodelers in heat shock gene activation are not clear. We show here that the chromatin remodeler dMi-2 is recruited to Drosophila heat shock genes in a PAR–dependent manner. We provide evidence that recruitment involves direct binding of dMi-2 to PAR polymers and identify novel PAR sensing regions in the dMi-2 protein, including the chromodomains and a series of motifs rich in K and R residues. Upon HS gene activation, dMi-2 associates with nascent transcripts. In addition, we find that dMi-2 and its catalytic activity are important for heat shock gene activation and co-transcriptional RNA processing efficiency. Our study uncovers a novel role of PAR during heat shock gene activation and establishes an unanticipated link between chromatin remodeler activity and RNA processing.
| The activity of eukaryotic genomes is regulated by dynamic changes in chromatin structure. A multitude of nucleosome remodeling enzymes, histone modifying activities and chromatin binding proteins cooperate to establish, maintain and reprogram chromatin structures that determine genome activity.
Drosophila heat shock (HS) genes provide a textbook example of how dramatic changes in the organismal and cellular environment affect chromatin structure in a manner that promotes transcriptional activation of genes coding for molecular chaperones required during the HS response. Upon temperature shift, the HS loci of polytene chromosomes form transcriptionally active “puffs”. This rapid chromatin decondensation correlates with a strong decrease in nucleosome density [1]. Puff formation can be uncoupled from transcription and much of the nucleosome loss at the hsp70 gene occurs prior to the first round of transcription [1], [2]. Recently, heat shock factor (HSF), GAGA factor and poly-[ADP-ribose] polymerase (PARP) have been shown to be required for the rapid removal of nucleosomes upon activation of the hsp70 gene [1]. In addition, HS puffs accumulate PARylated proteins and puff formation depends on PARP activity [3]. The mechanisms underlying PARP action during HS gene activation are not clear. It has been suggested that PARylation may be removing proteins, including histones - which are themselves a good PARP substrate - thereby promoting chromatin opening [1]. The accumulation of PARylated proteins at HS loci has recently been proposed to build up a “transcription compartment” which hinders the diffusion of proteins into and out of the compartment, thus favouring factor recycling [4]. In addition to histone displacement and transcription compartment formation at HS genes, recent evidence suggests that PARylation could also act as a signaling scaffold for the recruitment of PAR-sensing factors during DNA damage. In mammals PARylation at DNA damage sites can mediate the recruitment of several ATP-dependent nucleosome remodeling enzymes [5]–[10]. Here we sought to address whether and how nucleosome remodelers may be recruited to PARP activation sites upon environmental stresses other than DNA damage. We have investigated a paradigm of environmental stress, the activation of HS loci in Drosophila and have analyzed the mechanism through which the nucleosome remodeler dMi-2 is recruited to HS genes.
Mi-2 (CHD3/CHD4) is a conserved ATP-dependent nucleosome remodeler. In both vertebrates and invertebrates, it is a subunit of Nucleosome Remodeling and Deacetylation (NuRD) complexes. NuRD complexes repress cell type specific genes during differentiation [11]–[13]. dMi-2 is also a subunit of the Drosophila-specific Mep-1 complex (dMec) which represses neuron-specific genes during differentiation of the peripheral nervous system [12], [14].
Mi-2 containing complexes lack subunits with sequence-specific DNA binding activity. Two main mechanisms for their recruitment to chromatin have been suggested. First, NuRD complexes contain subunits with methylated DNA binding domains (MBD) which direct NuRD to methylated DNA [15], [16]. This is unlikely to be a major recruitment mechanism for Drosophila Mi-2 complexes, however, given the low and transient levels of DNA methylation in this organism [17]. A second mode of Mi-2 recruitment involves interactions with DNA bound transcription factors [11], [12], [14], [18]–[22]. In addition, SUMOylation of transcription factors can increase their affinity for Mi-2 complexes [21], [22].
Despite its well established role in repression, dMi-2 localises to actively transcribed chromosome regions suggesting an unexpected potential function of dMi-2 in transcription [23]. Here we sought to establish how dMi-2 is recruited to actively transcribed chromatin and to clarify its role in transcriptional activation using genetic, biochemical and pharmacological assays. We show that dMi-2 rapidly associates with activated HS loci, covering the entire transcribed region of the hsp70 gene. dMi-2 recruitment is not affected when transcriptional elongation is blocked but is abrogated when PARP is inhibited. Indeed, we find that dMi-2 specifically binds PARP's oligomeric product PAR in vitro. Significantly, a dMi-2 mutant unable to bind PAR is not recruited to active HS loci in vivo. We have identified several regions of dMi-2 that bind PAR in vitro. These include the chromodomains and a series of K/R-rich motifs near the N-terminus. Further, dMi-2 depletion or expression of an inactive enzyme greatly decreases transcript levels, suggesting that dMi-2 actively supports efficient HS gene expression. Indeed, dMi-2 associates with nascent hsp70 transcripts in vivo and ablation of dMi-2 function results in inefficient RNA processing. RNA and PAR compete for dMi-2 binding suggesting a two step process of dMi-2 association with HS genes: intial recruitment of dMi-2 is effected by its binding to PAR which is produced prior to the onset of transcription, dMi-2 then switches to interacting with the emerging nascent transcripts. Taken together, our results uncover PAR binding as a novel mechanism for the recruitment of the nucleosome remodeler dMi-2 to targeted sites of PARP activitation upon environmental stress and demonstrate that dMi-2 acts as a co-activator for the full transcriptional activation of HS genes. This study provides the first evidence for an in vivo function of PARylation in promoting the recruitment of a nucleosome remodeler to support the transcription of stress induced genes.
As shown previously, dMi-2 colocalised with active RNA polymerase II (Pol II) on polytene chromosomes [23] (Figure 1A). In addition, dMi-2 significantly colocalised with different forms of elongating Pol II (Ser2P and Ser5P) and elongation factors (Spt5). This suggests that dMi-2 may play an unanticipated role in active transcription. Upon HS, dMi-2 associated with the loci 87A and 87C which contain multiple copies of the hsp70 gene (Figure 1B), further strengthening a potential link between dMi-2 and active transcription.
Chromatin immunoprecipitation (ChIP) analysis of dMi-2 binding to the activated hsp70 gene in Kc cells revealed an enrichment of dMi-2 in the transcribed region (Figure 2A and 2B). dMi-2 association was detected as early as 2 min after HS and progressively increased for 20 min (Figure S1).
We considered three recruitment mechanisms:
First, dMi-2 might bind histone modifications enriched in actively transcribed genes, such as H3K4me3 or H3K36me3. However, we did not find a methylation sensitive interaction of recombinant dMi-2 with histone peptides in pulldown assays (data not shown).
Second, dMi-2 might bind and travel with RNA Pol II or elongation factors. This hypothesis predicts that HS-dependent dMi-2 recruitment to the transcribed part of hsp70 is transcription-dependent. To test this hypothesis, we inhibited transcriptional elongation with DRB (Figure 2C). Although this treatment efficiently ablated production of hsp70 transcripts, it did not significantly reduce HS-dependent recruitment of dMi-2. In addition, we failed to detect robust biochemical interactions of dMi-2 with RNA Pol II or elongation factors in co-immunoprecipitation assays (data not shown). We conclude that the HS-dependent recruitment of dMi-2 to the hsp70 gene can be uncoupled from the transcriptional activity of hsp70.
Third, dMi-2 might be recruited by interaction with PAR, a modification that rapidly accumulates over the hsp70 locus upon HS [3]. We therefore treated Kc cells with the small molecule PARP inhibitor PJ34 (Figure 2D). This led to a significant decrease of global PARylation levels, but did not abrogate hsp70 transcription or nucleosome depletion (Figure 2D and Figure S2). Nevertheless, dMi-2 recruitment to hsp70 was severely decreased during HS, suggesting that efficient PARylation of the locus is a requirement for stress-dependent enrichment of dMi-2.
To determine whether dMi-2 binds PAR directly, we auto-PARylated PARP1 in vitro and incubated the reaction with immobilised dMi-2. mH2A1.1 which contains a macrodomain known to interact with PAR was used as a positive control in this assay. Western blot analysis revealed that dMi-2, like mH2A1.1, bound PARylated PARP1 efficiently (Figure 3A). We confirmed that dMi-2 also interacted with radioactively labeled PARylated PARP1 (Figure S3). To ensure that dMi-2 interacted directly with the PAR polymer, we assayed binding to purified PAR using a dot blot assay (Figure S4). This verified the apparent direct interaction between dMi-2 and PAR.
Next, we sought to define the dMi-2 region required for PAR binding. We tested an array of dMi-2 truncation mutants for their ability to interact with PARylated PARP1 in vitro (Figure 3B). This revealed that the N-terminal region had a high affinity for PAR. Within this part of dMi-2, both the PHD finger containing region N-terminal of the chromodomains (aa 1-485) and (to a lesser extent) the chromodomains (aa 484-690) were capable of binding PAR. To verify these results we also tested binding of dMi-2 mutants to PAR in dot blot assays (Figure S4). We conclude that dMi-2 possesses at least two PAR-sensing regions that can function independently of each other.
To assess the functional importance of dMi-2′s PAR binding activity, we compared recruitment of GFP-dMi-2 fusion proteins to the activated hsp70 loci in transgenic flies (Figure 3C). GFP fused to full length dMi-2 and a GFP-dMi-2 fusion lacking the N-terminal PAR-binding regions were expressed to similar levels in 3rd instar larvae and correctly localised to salivary gland nuclei (Figure S5). Full length GFP-dMi-2 was enriched at active HS loci, the PAR binding mutant, however, failed to accumulate. This supports the notion that dMi-2 binding to PAR makes an important contribution to the recruitment of this nucleosome remodeler to the stress-activated hsp70 gene.
The N-terminal PAR binding region of dMi-2 contains two highly conserved domains, a pair of PHD fingers (residues 377 to 484) and a tandem chromodomain (residues 488 to 673). We generated GST fusions containing these domains and tested their ability to bind PAR in dot blot assays (Figure S6). This confirmed that the chromomodomains can bind PAR independently. However, the PHD fingers did not display PAR binding activity.
We next sought to better define the PAR binding region near the N-terminus of dMi-2. The N-terminal 375 residues of dMi-2 are characterised by a high content in charged residues (24% D/E, 21% R/K). This general feature is conserved between dMi-2 and mammalian CHD4 proteins (Figure 4A). In addition, these proteins share a region with high sequence similarity, the CHDNT domain (Pfam family PF08073). The function of this domain is not known. A number of diverse PAR binding motifs have recently been identified [24]–[26]. A common feature of these motifs is that they all contain several R/K residues that are interspersed by hydrophobic residues which often play critical roles in mediating PAR binding [24]–[26]. We subjected different dMi-2 fragments to the PAR binding assay, including four K/R-rich fragments (K/R I to IV in Figure 4A). This analysis revealed strong PAR binding activity for three of the four K/R-rich fragments (K/R I, K/R II and K/R IV; Figure 4B). By contrast, K/R-rich fragment II and a fragment encompassing the CHDNT domain failed to interact with PAR.
Taken together, our results suggest that dMi-2 contains multiple PAR binding regions in its N-terminus: three are characterised by a high content of basic amino acid residues (K/R I, K/R III and K/R IV) and one region containing the tandem chromodomain.
PARylation of the hsp70 locus has been proposed to assist in the opening of chromatin structure and to increase access of factors to DNA and nascent hsp70 transcripts [1]. Given that dMi-2 localises to the entire transcribed region and given that PAR exhibits chemical and structural similarity to RNA, we speculated that dMi-2, once recruited, might interact with nascent hsp70 RNA. We immunoprecipitated dMi-2 from nuclear extracts of heat shocked Kc cells and probed for the co-precipitation of nascent (unprocessed) hsp70 and hsp83 RNA (Figure 5A). Indeed, two independent dMi-2 antibodies precipitated these transcripts arguing for a physical, potentially direct interaction. In agreement with this, dMi-2 bound to single-stranded hsp70 RNA in an electrophoretic mobility shift assay in vitro (Figure 5B).
Next, we performed competition assays to gain insight into the relative affinities of dMi-2 for DNA, RNA and PAR and to determine if dMi-2 can bind to several types of nucleic acid simultaneously or if binding is competitive. First, we tested dMi-2 binding to RNA and DNA, respectively, in the presence of increasing amounts of PAR in electrophoretic mobility shift assays (mass ratios 1∶1, 1∶2 and 1∶4; Figure 5C). In this assay, PAR was able to compete with RNA and DNA for dMi-2 binding. However, whereas dMi-2 no longer bound to DNA at a DNA:PAR mass ratio of 1∶2, residual dMi-2/RNA complexes were still detectable at an RNA:PAR mass ratio of 1∶4. This suggests that dMi-2 has a higher binding affinity for RNA than for DNA. We confirmed this hypothesis by incubating dMi-2 with different mass ratios of RNA and DNA (Figure 5C): At a DNA:RNA mass ratio of 1∶1, dMi-2/RNA complexes formed readily but dMi-2/DNA complexes were not detected. dMi-2/RNA complexes formed even at DNA:RNA mass ratios of 4∶1.
To test if RNA or DNA can compete with dMi-2 for binding to the branched PAR polymer we performed dot blot assays (Figure S7). RNA competed with immobilised PAR for binding to dMi-2 whereas DNA failed to do so.
Taken together, our results suggest that dMi-2 has a higher affinity for binding to RNA and PAR than for binding to DNA. In addition, dMi-2 appears to bind RNA and PAR in a mutually exclusive manner. These results are consistent with the hypothesis that dMi-2 is first recruited to HS loci by interaction with PAR (which is produced prior to and independent of transcription) and, once RNA synthesis has been strongly activated, switches to binding the nascent RNA.
We hypothesised that dMi-2 binding to nascent RNA might influence hsp70 transcription or processing. We used transgenic fly lines to deplete dMi-2 by RNAi (Figure 6A). We subjected transgenic larvae to HS and determined the HS gene transcription by RT-QPCR. Although hsp70, hsp26 and hsp83 genes were all activated by HS, transcript levels were severely reduced in dMi-2 depleted larvae compared to controls. Importantly, transcription of a housekeeping gene was not significantly affected. We conclude that dMi-2 makes a positive contribution to transcription and is essential for full HS gene activation in larvae.
We next determined whether dMi-2 enzymatic activity was required to activate HS genes. We generated transgenic fly lines overexpressing wild type dMi-2 or a dMi-2 mutant carrying a point mutation in the ATP binding site (K761R) predicted to prevent ATP binding (Figure 6B). Indeed, dMi-2K761R could not hydrolyse ATP in vitro (Figure S8). We subjected 3rd instar larvae to HS and determined effects on HS gene transcription as before. Whereas overexpression of wild type dMi-2 had little effect, levels of HS gene transcripts were greatly reduced in larvae overexpressing the enzymatically inactive dMi-2 (Figure 6B). We conclude that the ATPase activity of dMi-2 is essential for full HS gene activation.
Next, we sought to assess whether dMi-2 influences RNA processing. Because dMi-2 depletion and expression of enzymatically inactive dMi-2 resulted in an overall reduction of hsp70 transcript levels we determined the ratio of 3′ unprocessed to total hsp70 RNA as a measure of RNA processing efficiency. We reasoned that a mere reduction in hsp70 activation (e.g. a reduction in the number of initiation events per time) would not change the ratio of unprocessed to total hsp70 RNA. By contrast, processing defects might give rise to a higher relative proportion of unprocessed RNA and, therefore, to a higher unprocessed:total RNA ratio. Depletion of dMi-2 increased the relative proportion of unprocessed hsp70 RNA (Figure 7A). An even more striking effect was observed in larvae overexpressing inactive dMi-2, whereas overexpression of wild type dMi-2 was of little consequence. Similar effects on 3′ RNA processing were observed with the hsp83 gene (data not shown). Hsp83 is one of the few HS genes possessing an intron. Therefore, we determined the ratio of unspliced to total hsp83 transcripts in transgenic larvae (Figure 7B). Again, we observed a significant increase in the relative proportion of unspliced RNA in dMi-2-depleted larvae and in larvae overexpressing inactive enzyme. This suggests that dMi-2 activity is required for the efficient processing of HS gene transcripts and that dMi-2 affects both RNA 3′ end cleavage and splicing.
Mi-2 is strongly linked to transcriptional repression in both vertebrate and invertebrate organisms. Within NuRD and dMec complexes it contributes to the repression of cell type-specific genes [11], [14], [19]–[21]. Therefore, the widespread colocalisation of dMi-2 with active Pol II and elongation factors at many chromosomal sites is surprising and suggests that dMi-2 might play an unappreciated role during active transcription, at least (or specifically) during environmental stresses such as HS. Indeed, dMi-2 is recruited to HS genes within minutes of HS. This property is not shared by other chromatin remodelers: Brahma (BRM) is not enriched at HS puffs and HS gene activation is independent of BRM function ([27] and data not shown). Moreover, although imitation switch (ISWI) containing complexes are important for HS gene transcription, ISWI does not accumulate to high levels at active HS loci ([28], [29] and data not shown). Recruitment to HS puffs has previously been reported for Drosophila CHD1 [30]. Thus, accumulation at active HS genes is shared by at least two members of the CHD family of nucleosome remodelers but not by SWI/SNF and ISWI proteins.
Depletion of dMi-2 or a reduction of dMi-2 recruitment does not significantly perturb hsp70 transcription in Kc cells and, therefore, dMi-2 is dispensable for HS gene activation in this system (Figure 2D and data not shown). By contrast, depletion of dMi-2 in larvae strongly decreases hsp70, hsp26 and hsp83 activation (Figure 6A). It is possible, that the RNAi-mediated depletion of dMi-2 is more efficient in transgenic flies compared to cell lines. In addition, it is believed that several factors contributing to HS gene activation are highly abundant or redundant in Kc cells but more limiting in other contexts. Accordingly, FACT and Spt6 are required for a HS gene activation in flies but are not essential in Kc cells [31], [32].
The strong decrease of HS gene activation in dMi-2 RNAi larvae indicates a positive contribution of dMi-2 to transcription in vivo. Overexpression of inactive dMi-2 also results in reduced HS gene transcription implying that its enzymatic activity is critical (Figure 6B). It is presently unclear whether this reflects a requirement for dMi-2 catalysed nucleosome remodeling or whether its activity is directed towards different substrates.
While dMi-2 could indirectly influence transcription by remodeling nucleosomes within the transcribed part of hsp70, its physical association with nascent HS gene transcripts argues for a more direct effect. Indeed, dMi-2 is not only required for high HS gene mRNA levels, but also affects the efficiency of co-transcriptional 3′ end formation and splicing. A role of chromatin remodelers in splicing has been suggested before: Both CHD1 and BRG1 bind components of the splicing apparatus [33], [34]. CHD1 associates with Pol II and binds nucleosomes containing H3K4me3, which are enriched near the 5′ end of active genes [34], [35]. BRG1 is present at the coding region of genes and influences splice site choice [33], [36]. It has been proposed that CHD1 and BRG1 physically recruit splicing factors but it is unclear if their ATPase activities play a role. Indeed, inactive BRG1 retains the ability to affect exon choice [33], [34]. Inefficient processing of the hsp70 and hsp83 transcripts is not only observed in larvae expressing reduced levels of dMi-2. Importantly, even stronger processing defects are generated by overexpression of inactive dMi-2 (Figure 7). This strongly suggests, for the first time, that the catalytic activity of a chromatin remodeler is required for correct co-transcriptional RNA processing. It remains to be determined whether dMi-2 nucleosome remodeling activity influences RNA processing indirectly, e.g. by altering Pol II elongation rates, or whether it has a more direct role.
A series of complementary results support our hypothesis that dMi-2 interacts with PAR polymers that are rapidly synthesized at activated HS loci. First, the broad distribution of dMi-2 over the entire transcribed region correlates with the distribution of PAR polymer [3]. Second, pharmacological inhibition of PARP greatly decreases dMi-2 binding to activated hsp70. Third, dMi-2 directly binds PAR polymers in vitro. Fourth, an dMi-2 mutant unable to bind PAR also fails to localise to active HS loci. As discussed above, dMi-2 physically associates with nascent HS gene transcripts and binds RNA in vitro. While this interaction is potentially important for the efficiency of transcription and processing, it likely plays a minor role in dMi-2 targeting. Accordingly, inhibition of transcriptional elongation has no significant effect on dMi-2 recruitment (Figure 2C).
It is important to note, that while our results argue for an important role of PAR binding in the recruitment of dMi-2 to HS loci, we cannot exclude that protein-protein interactions with histone or non-histone proteins also play a role.
Our analysis indicates that dMi-2 harbours several PAR binding motifs in its N-terminal region. Polo and colleagues have recently demonstrated that human CHD4 is recruited to double stranded DNA breaks in a PARP-dependent manner [10]. They have mapped PAR binding activity to the region N-terminal of the ATPase domain of CHD4. This agrees well with our data and suggests that the PAR binding function of CHD4/dMi-2 has been conserved in evolution.
Two structural protein modules directly interact with PAR, the macrodomain and the PBZ domain; however, these domains are not present in dMi-2 [5], [7], [37]. In addition, several shorter PAR binding motifs have been identified [5], [26]. These motifs bear little sequence similarity but share the presence of several K/R residues which are interspersed by hydrophobic residues. Our results have uncovered three K/R-rich regions with PAR binding activity near the N-terminus of dMi-2. Two of these three K/R-rich regions (K/R III and K/R IV) consist of interspersed basic and hydrophobic residues and are therefore reminiscent of the previously described PAR binding motifs [24], [25], the third (K/R I) lacks hydrophobic residues completely. None of the three K/R regions matches the consensus PAR binding motifs. It is possible that a consensus motif should generally be chosen less stringently and that a high content of K and R-residues in these regions is sufficient to provide PAR binding activity in vitro. Further characterisation of these regions will be required to resolve this issue. In addition to the K/R regions, the tandem chromodomains of dMi-2 bind PAR in vitro. We have previously shown that the chromodomains are required for interacting with nucleosomal DNA in vitro [38]. Our new data suggests that these domains can interact with different nucleic acids.
Several potential molecular functions of PARylation at HS genes have been suggested. First, PARP activity is required for the rapid loss of nucleosomes at hsp70 within the first two minutes after HS [1]. It has been suggested that PARylation of histones aids rapid nucleosome disassembly [1]. Second, at later stages of the HS response (20–60 minutes after HS), PARP activity is required to establish a compartment which restricts the diffusion of factors such as Pol II and Spt6 and promotes efficient factor recycling [4]. Our results suggest that PARylation carries out a third task, namely, to recruit factors via their direct interaction with PAR. The earliest time point when we can detect dMi-2 binding to hsp70 is between 2 and 5 minutes after HS. This places dMi-2 recruitment between the early PARP-dependent nucleosome removal (0–2 minutes after HS) and effects of the transcription compartment (20–60 minutes after HS).
The ability of dMi-2 to bind both PAR and RNA and the finding that RNA can compete for PAR binding to dMi-2 is consistent with the hypothesis that dMi-2 association with active HS genes is a two step process (Figure 8). We propose that dMi-2 is initially recruited via interaction with PAR polymers. Synthesis of these starts prior to the onset of hsp70 transcription [1]. This results in a rapid local increase of the dMi-2 concentration. In the second step, when hsp70 transcripts are produced by elongating RNA polymerase II at high rates, dMi-2 can switch from binding PAR to interacting with nascent transcripts.
Severe cellular stresses, such as DNA strand breaks and acute HS, must be dealt with quickly and efficiently. In both cases, a multitude of factors are rapidly recruited to orchestrate the repair of DNA and the massive transcriptional activation of HS genes, respectively. We postulate that rapid synthesis of PAR polymers at both DNA damage sites and HS genes affords an efficient mechanism to recruit chromatin remodelers and other factors. It has recently been shown that PARylation of DNA breaks is instrumental in recruiting chromatin remodelers, including mammalian dMi-2 homologs, to damaged sites [8], [9], [10], [39], [40], [41]. Here, we show that dMi-2′s recruitment to activated HS genes requires PARP activity and that dMi-2 binds PAR directly. The high local concentration of PAR polymers at DNA breaks and HS genes might exploit the general affinity of dMi-2 for nucleic acids. Indeed, dMi-2 binds both DNA and RNA as well as PAR in vitro ([38] and this study). In this manner, PAR polymers might act as a scaffold to redirect dMi-2 to chromatin regions where high levels of dMi-2 activity are required, thus acting as a stress-dependent, transient affinity site for chromatin remodeling and possibly RNA processing activities (Figure 8). Our results highlight a signaling and scaffolding function for PARP activity during transient environmental stresses other than DNA damage, suggesting that PARylation carries out important modulatory functions in the stress-dependent reprogramming of nuclear activities.
Kc cell HS treatment and ChIP was performed as decribed using dMi-2C antibody [14], [42]. For primer sequences see Dataset S2. Triplicate mean values of percentage input DNA and standard deviations are plotted. dMi-2 knockdown by RNAi was described previously [23]. For RNAi primer sequences see Dataset S4.
Kc cells were treated with 125 µM DRB (Sigma) to inhibit transcription and with 5 µM PJ34 (Alexis) to inhibit PARP activity for 20 min before subjecting cells to HS.
Chromosomes were prepared as before [23]. The following antibodies were used: Primary antibodies: anti-dMi-2N (rabbit) 1∶200, anti-pol II (mouse H5, Covance) 1∶50, anti-GFP (rabbit, Abcam) 1∶50, anti-Spt5 (guinea pig) 1∶200. Secondary antibodies: Alexa Fluor 488 goat anti-rabbit 1∶200, Alexa Fluor 546 goat anti-mouse or anti-guinea pig 1∶200 (Invitrogen).
Analysis was performed with a Zeiss fluorescence microscope (Axioplan).
For baculovirus production, dMi-2 mutants (aa 1-691) and (aa 1-485) were generated by PCR using appropriate sets of primers and cloned with NotI and XbaI into the pVL1392 transfer vector. Vectors for dMi-2 WT and other mutants were described previously [38]. dMi-2 GST-fusion fragments were generated by PCR using appropriate sets of primers and cloned with NotI and SalI into the pGEX4T1 vector. All constructs were verified by DNA sequencing. For primer sequences see Dataset S1.
Protein extracts from 3rd instar larvae were prepared as described in [14].
Purification of recombinant dMi-2 and ATPase assays are described [23]. Recombinant mH2A1.1 was purified as in [43]. GST-fusion proteins were expressed in E.coli BL21(DE3) and purified with Glutathione Sepharose 4 Fast flow (GE Healthcare) according to the manufacturer's instructions.
A typical DNA or RNA binding reaction (25 µl) was performed in the presence of 0.2 µg of dMi-2F and 80 ng of nucleic acid (DNA or ssRNA) in 40 mM KCl, 20 mM Tris pH 7.6, 1.5 mM MgCl2, 0.5 mM EGTA, 10% glycerol, BSA (200 ng/µl), 1 mM DTT (supplemented with 0.4 units of RNAsin). For competition assays, samples were preincubated for 15 min at 26°C before the different amounts of competitor (PAR or DNA or RNA) were added. Reactions were further incubated at 26°C for 75 min. Products were analyzed on 6% native PAA gel and visualized with ethidium bromide (EtBr) staining. ssRNA was synthesized by in vitro transcription using a fragment of hsp70 DNA as a template. This template (also used for the DNA bandshift assays) was produced by PCR amplification of cDNA derived from heat shocked Kc cells using the following primers: T7-hsp70_f - TAATACGACTCACTATAGGGCCTACGGACTGGACAAGAAC and hsp70_r -AGGGTTGGAGCGCAGATCCTTCTTGTAC.
Total RNA was isolated from 3rd instar larvae using PeqGold total RNA Kit (PeqLab). 10-12 larvae from each cross were pestled in 400 µl of lysis buffer before loading the material on the column. 1 µg of RNA was reverse transcribed by incubation with 0.3 µg of random primers (Invitrogene) and 100 U of M-MLV reverse transcriptase (Invitrogen). cDNA synthesis was performed according to the manufacturer's protocol. cDNA was analyzed by QPCR using Absolute SybrGreen Mix (Thermo Fisher) and the Mx3000P real-time detection system (Agilent). For primer sequences used in RT-QPCR see Dataset S3. All amplifications were performed in triplicate using 0.6 µl of cDNA per reaction. Triplicate mean values were calculated according to the ΔΔCt quantification method using rp49 gene transcription as reference for normalization. Relative mRNA levels in uninduced control larvae were set to 1 and other values were expressed relative to this. The RT-QPCR results were reproduced several times using independent fly crosses and representative data sets are shown.
RNA immunoprecipitation was performed as described previously [44]. Briefly, Kc cells were crosslinked as for ChIP. Cells were washed once with PBS buffer and lysed on ice for 15 min in FA buffer (50 mM Hepes- KOH, pH 7,6, 140 mM NaCl, 1% Triton X-100, 0,1% sodium deoxycholate, proteinase inhibitors, RNAsin (100 u/ml of buffer)). Cells were sonicated, spun down and chromatin was digested with DNAse I. The chromatin containing solution was adjusted to 25 mM MgCl2 and 5 mM CaCl2. 1 ul of DNAse I (Qiagen) was added and reactions were incubated for 10 min at room temperature and then stopped with 20 mM EDTA. Chromatin was spun down for 10 min (13000 rpm) at 4°C. 300 µl of chromatin was used for IP. 2 µl of anti-dMi-2(C) and anti-dMi-2(N) antibodies, 2 µl rabbit IgG, 2 µl rabbit preimmuneserum and beads only (control) and were used for IP. Samples were incubated over night at 4°C. RNA-protein complexes were precipitated with 30 ul of 50% protein G Sepharose beads for 2 hr at 4°C. IPs were washed 5 times in FA buffer, twice with TE buffer and eluted twice with 100 µl of elution buffer (100 mm Tris HCl, pH 8,0, 10 mM EDTA, 1% SDS) – once at room temperature and once at 65°C. All buffers were supplemented with RNAse inhibitor (RNAsin, Promega). All samples were digested with proteinase K for 1 hr at 42°C and decrosslinking was performed at 65°C over night. Immunoprecipitated RNA was purified using PeqGold total RNA Kit (PeqLab), digested with DNAse on the column and eluted with 30 µl of RNAse free dH20. cDNA was synthesized with 10 µl of eluted RNA and 2 µl of input with random hexamers and analysed by Q-PCR with appropriate primer pairs.
Non-radioactive PAR synthesis was performed according to the standard protocol [45]. Briefly, PARP reactions were set up in a final volume of 0.5 ml: 2 µg recombinant Parp1, 100 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 mM MgCl2, 2 µg/ml DNA oligonucleotides, 1 mM NAD+, 1 mM DTT. Reactions were incubated at 37°C for 25 min. PJ34 inhibitor was added before the reaction to control samples to a final concentration of 5 µM. All reactions were stopped with PJ34. Control beads and beads with bound proteins (dMi-2, dMi-2 mutants and mH2A1.1 or GST fusions) were equilibrated in binding buffer (50 mM Tris, pH 8.0, 0,2 mM DTT, 4 mM MgCl2, 200 mM NaCl, 0,1% NP-40). 10 µl of bead-bound proteins were used for each pulldown. Pulldowns were performed with the whole PARP reaction (0.5 ml) and 500 µl of binding buffer (for baculovirus expressed proteins) or in 250 µl of PARP reaction and 250 µl of binding buffer (for GST fusions) for 1 hr at 4°C. After extensive washing (5 times), beads were boiled in SDS loading buffer, loaded on 4–12% gradient SDS-Page gels and analysed by Western blot. For Western blot anti-PAR (10H, 1∶500) antibodies were used. mH2A1.1 was used as a positive control. Radioactive pulldown reactions were prepared in the same way, in the presence of 2 µl of radioactive NAD+ (PerkinElmer). After washing, samples were resuspended in 30 µl of SDS-loading buffer and 10 µl was resolved by SDS PAGE. The gel was dried and and subjected to autoradiography.
Generation of transgenic fly strains and the PAR dot blot assay are described in Text S1.
|
10.1371/journal.ppat.1000878 | Helicobacter pylori VacA Toxin/Subunit p34: Targeting of an Anion Channel to the Inner Mitochondrial Membrane | The vacuolating toxin VacA, released by Helicobacter pylori, is an important virulence factor in the pathogenesis of gastritis and gastroduodenal ulcers. VacA contains two subunits: The p58 subunit mediates entry into target cells, and the p34 subunit mediates targeting to mitochondria and is essential for toxicity. In this study we found that targeting to mitochondria is dependent on a unique signal sequence of 32 uncharged amino acid residues at the p34 N-terminus. Mitochondrial import of p34 is mediated by the import receptor Tom20 and the import channel of the outer membrane TOM complex, leading to insertion of p34 into the mitochondrial inner membrane. p34 assembles in homo-hexamers of extraordinary high stability. CD spectra of the purified protein indicate a content of >40% β-strands, similar to pore-forming β-barrel proteins. p34 forms an anion channel with a conductivity of about 12 pS in 1.5 M KCl buffer. Oligomerization and channel formation are independent both of the 32 uncharged N-terminal residues and of the p58 subunit of the toxin. The conductivity is efficiently blocked by 5-nitro-2-(3-phenylpropylamino)benzoic acid (NPPB), a reagent known to inhibit VacA-mediated apoptosis. We conclude that p34 essentially acts as a small pore-forming toxin, targeted to the mitochondrial inner membrane by a special hydrophobic N-terminal signal.
| VacA is a toxic protein produced by Helicobacter pylori, the bacteria that cause gastritis and ulcer diseases. p34, the toxic component of VacA, is known to damage mitochondria, defined cell organelles in the target cells. However, both the mechanism of mitochondrial targeting and the toxic activity inside the mitochondria are unclear. In this study, we show that p34 carries a unique targeting signal that is different from all targeting signatures that were previously identified in endogenous mitochondrial proteins. Eventually, p34 seems to act as an anion channel in the mitochondrial inner membrane and thus to destroy the balance of salt ions in the organelles.
| Helicobacter pylori is a gram-negative bacterium infecting the human gastric mucosa, causing gastritis and peptic ulcer and, in some cases, gastric cancer [1]–[5]. One of the major virulence factors of the bacteria is the vacuolating toxin VacA, a protein of about 90 kDa [1], [6], [7]. VacA forms hexameric or heptameric flower-shaped oligomers [8]–[11]. These contain a central cavity and are able to form an ion channel [12]–[17]. The VacA toxin, as it is released by the bacteria, is a hetero-dimeric protein, comprising the subunits p58 and p34 that stay associated by non-covalent interactions [7], [18]–[20]. Similar to the subunits of A/B toxins, the two components have different functions in targeting and toxicity: The p58 subunit mediates binding to target cells. Following entry into host cells by endocytosis, the p34 subunit is essential to cause toxic effects [6], [21].
The p34 subunit is a polypeptide of 319 amino acid residues [22]. The first 32 residues are uncharged and required both for VacA insertion into the plasma membrane of host cells and for toxicity [7]. Inside the target cells, p34 can be imported into mitochondria [23], and mitochondria were shown to play an important role in the toxicity of the protein [24]–[29]. However, p34 does not reveal an obvious mitochondrial targeting signal, and nothing is known about the fate and the relevant molecular activities of p34 inside the mitochondria. p34 can dissipate the mitochondrial membrane potential, interfere with the mitochondrial energy metabolism and trigger apoptosis [23], [24], [30], but the mechanism is unclear. In this study we therefore asked: How is p34 imported into mitochondria? What is the mitochondrial targeting sequence? To which subcompartment is p34 targeted, and what is the activity of p34 inside the mitochondria? Combining biochemical and biophysical investigations, we found that the p34 subunit essentially acts as a small pore-forming toxin targeting the mitochondrial inner membrane by a peculiar import signal.
The p34 subunit of VacA is specifically imported into mitochondria, however it does not reveal an obvious mitochondrial targeting signal. To identify the segment that determines mitochondrial import, we designed hybrid proteins containing an EGFP moiety (enhanced green fluorescent protein) and investigated their distribution after expression in HeLa cells. Expression of p34-EGFP, a hybrid protein carrying the EGFP domain at the p34 C-terminus, causes cell death [23], however we confirmed that an EGFP-p34 protein containing the EGFP domain at the N-terminus was targeted to mitochondria keeping the transfected cells intact (Fig. 1A, upper panel). To determine the possible role of the hydrophobic p34 N-terminus in targeting, we tested a truncated construct comprising residues 37–319 of p34 fused to the EGFP domain. The construct stayed in the cytosol (Fig. 1A, middle panel.) Probably due to an affinity of the EGFP moiety [31], the construct partially co-localized with the nuclei of the cells. On the other hand, we found that the 36 N-terminal residues of p34 were sufficient for targeting of the EGFP domain to mitochondria (Fig. 1A, lower panel). Minor differences in the distribution of EGFP-p34(1–319) and p34(1–36)-EGFP suggest that the interactions of the N-terminal residues with the mitochondria may be facilitated by the authentic subunit. However, the N-terminal residues of p34 were essentially sufficient for targeting. The first 32 residues of p34, followed by a lysine in position 33, are non-charged or hydrophobic (Fig. 1B). They appear to represent a novel type of a mitochondrial targeting signal. The N-terminus of p34 is both essential and sufficient for mitochondrial targeting.
To investigate the interactions of p34 with mitochondria in more detail, we used an in vitro assay (Fig. 2). For this purpose, we synthesized 35S-labelled p34 in reticulocyte lysate. Under these conditions, proteins are synthesized in the presence of the cytosolic proteins that determine the import-competence of mitochondrial proteins in mammalian cells [32]. Both hydrophilic and membrane proteins can subsequently be imported into isolated mitochondria. In the absence of mitochondria, 20 µg/ml of proteinase K (PK) were sufficient to completely hydrolyze p34 at 0°C within 10 min (Fig. 2A). We then incubated the 35S-labelled p34 with freshly isolated rat liver mitochondria at 25°C, the mitochondrial outer membrane protein Tom70 [33] was likewise synthesized in reticulocyte lysate and included as a control protein (Fig. 2B). After an incubation of 10 min, the mitochondria were reisolated by centrifugation and incubated in the presence of increasing concentrations of the detergent digitonin and proteinase K (PK, 50 µg/ml) at 0°C. Tom70 and p34 associated with the mitochondria (Fig. 2B, lane 1). In the presence of proteinase K (lane 2), Tom70 was rapidly degraded while about 40% of the p34 was retained, indicating that p34 was transported across the mitochondrial outer membrane. High concentrations of digitonin were required to allow the protease to get access to the p34 (Fig. 2B, lanes 3-6). In most of the experiments, about 20% of the p34 initially added to the samples was imported to a PK-protected location within 10 min.
Previous studies emphasized the relevance of distinct residues within the p34 N-terminus for VacA-induced cellular vacuolation [34], [35]. We separately exchanged three of these residues against alanine (P9A, G14A, K33A), but a substantial influence on the import efficiency was not observed (Fig. 2C). Moreover, the import was not blocked by the uncoupling reagent valinomycin, demonstrating that the mitochondrial membrane potential was not required in this reaction (Fig. 2C, lanes 2 vs. lanes 4).
To investigate the relevance of the N-terminal residues of p34, we used the construct p34(37–319) lacking the N-terminal sequence, or the hybrid protein p34(1–35)-DHFR comprising the N-terminal 35 residues of p34 linked to a complete DHFR (dihydrofolate reductase) domain. The p34 part of this construct was easily degraded by proteinase K, similar to the authentic p34 (Fig. 2D). The DHFR domain, however, was resistant even against high concentrations of the protease (Fig. 2D). Both constructs were synthesized in reticulocyte lysate and incubated with isolated rat liver mitochondria (Fig. 2E). The truncated version of p34 lacking the hydrophobic N-terminus was not imported (Fig. 2E, p34[37–319], lanes 1 and 2), in agreement with the observation that in intact cells, the p34(37–319)EGFP fusion protein did not co-localize with mitochondria (Fig. 1A). In a parallel assay, we tested the construct p34(1–35)-DHFR. The same samples also contained authentic DHFR. Upon incubation with mitochondria, p34(1–35)-DHFR efficiently associated with the mitochondria, while the DHFR was removed (Fig. 2E, lanes 3 vs. lane 4). The associated p34(1–35)-DHFR was partially protected against proteases, similar as the authentic p34 (in Fig. 2C), but a fraction of about 35% was degraded to a fragment of slightly smaller size (Fig. 2E, lane 5).
The observations show that the p34 N-terminus is necessary and sufficient for mitochondrial targeting of p34 both in vitro and in vivo. Since the p34(1–35)-DHFR fusion protein was only partially transported across the outer membrane, the p34 N-terminus might be a weaker import signal as compared to conventional positively charged mitochondrial targeting sequences. However, similar fragments as with p34(1–35)-DHFR were not observed with authentic p34. The p34 N-terminus thus appears to be sufficient for efficient targeting and import of the complete p34 subunit.
To act as a specific targeting signal, p34 should have the capability to interact with specific sites at the mitochondrial surface. These sites could be provided by mitochondrial outer membrane proteins. We pretreated rat liver mitochondria with trypsin at low concentrations, reisolated the mitochondria, and investigated if the import of p34 was affected. The rate of import was clearly reduced (Fig. 2F). We used the same mitochondria to import porin, a mitochondrial outer membrane protein, and found that the rate of import was similarly reduced (data not shown). Similar to endogenous mitochondrial proteins, import of p34 seems to be facilitated by proteins that are exposed at the mitochondrial surface. The effect was also observed with import of p34(1–35)-DHFR (Fig. 2G). In summary, the assays for import of p34 into rat liver mitochondria confirm that the p34 N-terminus can act as a mitochondrial targeting signal. The p34 N-terminus is sufficient to target specific recognition sites at the mitochondrial outer surface. The complete p34 is not only able to target mitochondria but to traverse the outer membrane and to accumulate inside the organelles.
To identify the structures that are targeted by p34, we used isolated yeast mitochondria as a model system (Fig. 3). To improve the import efficiency, we followed the observation that insertion of the VacA holotoxin into the plasma membrane is facilitated by an acid pre-treatment of the toxin [9], [13], [36]. We pre-incubated the lysate at pH 5, however, the import experiments were subsequently carried out at pH 7.2. The acid-pretreated p34 was completely degraded by proteinase K at a concentration of 20 µg/ml (Fig. 3A). After incubation with isolated yeast mitochondria, a fraction of p34 was protected against degradation (Fig. 3B, lanes 1 and 2), indicating that p34 was imported into the organelles. To exclude an unspecific aggregation of p34, we tested samples lacking mitochondria (Fig. 3B, lanes 3 and 4). Similar as with rat liver mitochondria, the import was independent of the mitochondrial membrane potential, as demonstrated by samples containing valinomycin (Fig. 3B, lanes 5 and 6). Import was also independent of mtHsp70 (encoded by SSC1; [37]), the major heat shock protein of 70 kDa in the mitochondrial matrix (Fig. 3B, lanes 7 and 8).
The import rate of p34 was significantly reduced if the yeast mitochondria were pretreated with trypsin, indicating that p34 similarly interacts with proteins at the surface of mitochondria from yeast and from mammalian cells (Fig. 3C). We took advantage of the availability of yeast mutants that show defined defects in the mitochondrial import machinery, and we isolated mitochondria from several of these strains. Tom20, a protein of the outer membrane, is the major receptor for import of endogenous proteins into mitochondria [32], [38]–[40]. Comparing the rate of import into mitochondria of a tom20 deletion strain (tom20Δ; [41]) and the corresponding wildtype strain, we found that the import of p34 was significantly reduced (Fig. 3D).
The delay in import into mitochondria lacking Tom20 was also observed with p34(1–35)-DHFR (Fig. 3E). p34 and p34(1–35)-DHFR were similarly affected, and the results resembled data obtained with porin, which was again included as a mitochondrial standard protein (Fig. 3E). The experiments show that the N-terminal segment of p34 is able to recognize the import receptor Tom20. Interestingly, import of p34 into mitochondria from a tom70 deletion mutant showed only minor effects (Fig. 3E, right column). Tom70 is a second import receptor besides Tom20 and involved in the uptake of a subset of mitochondrial proteins [32], [38]. p34 seems to specifically follow the Tom20 pathway for import.
For import into the inner compartments of mitochondria, proteins have to pass the general import pore that is mainly formed by Tom40 [32], [38], [40], [42], [43]. Using the mutant tom40-4 [42], we found that Tom40 was clearly involved in the import of p34 (Fig. 3F). To enter mitochondria, p34 targets the same import pore as newly synthesized endogenous mitochondrial proteins.
Where is p34 localized after import into mitochondria? EGFP-p34 was previously detected by immuno gold labelling in the interior of HEp-2 cell mitochondria but the precise localization was unclear [23]. We imported radiolabelled p34 into isolated yeast mitochondria, disrupted the membranes by sonication, and separated the membrane vesicles by sucrose density centrifugation (Fig. 3G and H). The peak fractions of the outer membrane protein Tom40 and the inner membrane protein Tim23 were clearly separated, p34 was found to co-fractionate with Tim23. No significant amounts of p34 were detected in the outer membrane. In summary, we conclude that p34 targets the TOM complex in the outer membrane and subsequently accumulates in the mitochondrial inner membrane.
A previous study reported that after insertion of VacA into membranes, most parts of the p34 subunit are protected against proteases [44]. However, structure and function of p34 are still unclear. To obtain purified p34, we expressed p34 in Escherichia coli and isolated the protein by ammonium sulphate precipitation, hydrophobic chromatography and anion exchange chromatography (Fig. 4A). CD spectra of the purified protein indicated a high content of β-strands and a very low probability to form α-helices (Fig. 4B). The algorithm of the spectrapolarimeter calculated a content of 40–45% anti-parallel β-strands. The values are similar to results obtained with typical pore-forming β-barrel proteins [45]. A pore-forming activity of p34 is also indicated by the observation that an expression of p34 in HeLa cells entails a quick loss of the mitochondrial membrane potential [23]. Interestingly, using the program TMB hunt [46], we calculated a probability of p34 to form a β-barrel protein of 99%. Therefore we asked: Is the p34 subunit able to form a channel in the absence of the p58 subunit, and does the purified p34 oligomerize independently of p58?
We incubated purified p34 with the chemical cross-linking reagents DSS (Disucciminidylsuberate) or Sulfo-MBS (Sulfo-m-maleimidobenzoyl-N-hydroxysulfo-succinimide ester), respectively (Fig. 4C, lanes 1–3). With both reagents we observed a series of cross-linking products, indicating that p34 formed homo-oligomers containing several subunits. The same pattern of cross-linking products was observed in the presence and in the absence of Triton X-100 (Fig. 3C, lanes 3 and 4), suggesting that similar complexes assemble in the absence and in the presence of membranes. To determine the size of the complexes we carried out a blue native electrophoresis (BN-PAGE). In this system, the mobility of the proteins depends on their size, but also on their affinity for detergent molecules and for the Coomassie dye that is used to keep protein complexes in solution. In comparison to hydrophilic proteins, most membrane proteins show a reduced mobility [47]. Monomeric p34 showed a mobility similar to hydrophilic marker proteins of about 50 kDa (Fig. 4D, lower panel). In the presence of 500 mM ε-aminocaproic acid, two different complexes of native p34 were resolved, corresponding to an apparent molecular mass of about 330 or 650 kDa, suggesting that native p34 assembles in complexes of 6–7 or 12–14 monomers, respectively (Fig. 4, upper panel). Complexes of similar size were found if derivatives of p34 were tested lacking the N-terminal 36 amino acid residues (Fig. 4, middle panel). The uncharged N-terminus is obviously dispensable for complex formation.
To obtain higher amounts of the subunit, we purified a derivative of p34 containing a (His)10-tag instead of the hydrophobic N-terminus (p34 residues 37–319 connected to 10 histidine residues; Fig. 5A). Complex formation was then tested by size exclusion chromatography. In the presence of 10% glycerol, (His)10-p34 eluted in several fractions, partially corresponding to very high molecular mass, probably due to a tendency to form aggregates (Fig. 5B). From the Kav value of the most prominent peak fraction, we calculated a molecular mass of 193 kDa (Fig. 5C). Since (His)10-p34 is a protein of 32.2 kDa, we infer that the oligomers contained 6 subunits. To prevent aggregation of the protein, the chromatography was repeated in the presence of 2 M urea (Fig. 5D and E). p34 eluted under these conditions in a single peak corresponding to a molecular mass of the hexameric complex. The hexamers showed an impressive stability, even in the presence of 4 M urea most of the complexes were retained, while a smaller fraction started to dissociate (Fig. 5F and G).
Following the observation that the p34 N-terminus was not required for complex formation, we also tested a construct additionally lacking a segment of 28 residues at the p34 C-terminus (p34 residues 37–292 linked to a 10 histidine tag; Fig. 5H and I). In the presence of 4 M urea, most of the protein was stable. Only a small fraction dissociated and eluted in fractions corresponding to the molecular mass of the monomers (Fig. 5I).
The possible pore formation of p34 was tested directly by incorporation of purified p34 in artificial membranes and subsequent electrophysiological characterization (Fig. 6). Current recordings confirmed a low but significant conductivity and a high dynamics in gating behaviour (Fig. 6A). A closer investigation of the current-voltage relationship revealed a conductance state of about 12 pS at 1.5 M KCl, pH 7.5 (Fig. 6B). Under asymmetrical buffer conditions the channels showed a clear preference for anions, with a PCl/PK value of 19 (data not shown). Similar electrophysiological characteristics were observed at pH 4 (not shown). p34 obviously acts as an anion channel of low conductivity. Based on the main conductance state, the diameter of the p34 ion channel can be calculated to be between 0.5 and 1.5 Å (assuming a cylindrical restriction zone of 1 nm with a five-fold higher resistance than the bulk medium). This dimension is sufficient to accommodate single chloride ions, but complex molecules such as organic acids should be excluded.
Interestingly we found similar values for the p34 construct lacking the hydrophobic N-terminus (p3436–311) (Fig. 6C and D). At 1.5 M KCl we determined a conductance state of about 10 pS (Fig. 6D), only slightly less as compared to the authentic p34. Similar data were also obtained with p34 containing exchanges of single residues within the N-terminus (P9A, G14A, K33A) or an N-terminal extension of 12 amino acids (s2 subtype, data not shown). It is therefore unlikely that the p34 N-terminus forms the ion channel.
The conductivity of some pore-forming toxins is modulated by ATP [48]. However, using a fluorescent-labelled ATP derivative, we did not obtain any evidence of ATP binding to p34 (not shown). Previous investigations demonstrated that the ion channel of VacA complexes can efficiently be blocked by NPPB (5-nitro-2-[3-phenylpropylamino] benzoic acid) [14]-[49]. Moreover, it was shown that the proapoptotic effect of VacA can be blocked by a preincubation of the target cells with NPPB [26]. A pretreatment of HeLa cells in the presence of 200 µM NPPB was demonstrated to prevent the cytochrome c release that is induced if purified VacA is added to the cells [30]. We added NPPB at a final concentration of 100 µM to our assay and found that it inhibited the conductance of the p34 channel completely (Fig. 6E and F).
In this study, we investigated targeting, mitochondrial import, the final location, and the function of the VacA p34 subunit of Helicobacter pylori. Because previous investigations on VacA were nearly exclusively carried out on the holotoxin [1], [7], [19], only little was known about the specific features of the toxic p34 subunit. Our data indicate that p34 acts as a pore-forming protein in the mitochondrial inner membrane.
It was previously shown that p34 targets mitochondria [23], but the targeting signal and the import pathway of the subunit remained enigmatic. We found that the 36 N-terminal residues of p34 are both necessary and sufficient for mitochondrial targeting. This observation is surprising because a similar mitochondrial targeting sequence was not described for any endogenous mitochondrial protein [32], [38], [40], [50], [51], or any other bacterial effector protein [52]. The 32 N-terminal residues of p34, shown in Fig. 1B, are uncharged or hydrophobic, the first charged residue of the sequence is the lysine in position 33. We found that the mitochondrial import receptor Tom20 is involved in the uptake of p34, in agreement with studies showing Tom20 to interact with hydrophobic residues of precursor proteins [53]. Subsequent transport of p34 across the mitochondrial outer membrane is mediated by the general import pore formed by Tom40.
In this case, the length and hydrophobicity of the p34 N-terminus seem to determine the specificity for import into mitochondria. Data on protein traffic in plant cells have already shown that the specificity for different membranes can be determined by the length of a hydrophobic segment of the sequence [54], [55]. Interestingly, it was demonstrated that a mitochondrial outer membrane protein can be changed to an inner membrane protein by an artificial increase in the mean hydrophobicity [56]. To our knowledge, the p34 subunit is the first example of an authentic protein that uses a hydrophobic N-terminus for targeting of the mitochondrial inner membrane. Membrane insertion of long hydrophobic peptides is membrane potential-independent [57], and in fact we found that p34 is efficiently imported both in the presence and in the absence of the mitochondrial membrane potential.
The p34 N-terminus is not only required for mitochondrial import but also for interactions of the VacA holotoxin with the plasma membrane during entry into the target cell [7]. The uncharged N-terminal sequence of the p34 subunit seems to confer both a general affinity for lipid bilayers, and specific targeting to mitochondria. The situation is reminiscent of the affinities of the outer membrane porin that similarly has a capability of spontaneous membrane insertion [58] but shows specific import into mitochondria depending on interactions with the mitochondrial TOM complex [42]. Some inner membrane proteins are first imported into the matrix compartment and subsequent insertion starts at the inner side of the membrane [32], however, we did not obtain any evidence of p34 import into the matrix. Because import of p34 is independent of the mitochondrial membrane potential and of matrix mtHsp70, it is more likely that p34 inserts at the outer surface of the inner membrane.
The oligomeric state of the complete VacA toxin was characterized in several studies [8]–[11]. We find that the p34 subunit alone is able to form highly stable complexes of about 200 kDa, corresponding to the molecular mass of 6 subunits. The formation of these hexamers is independent of membrane insertion, raising the possibility that soluble p34 monomers may pass the TOM complex, reassemble in the mitochondrial intermembrane space and insert into the inner membrane after complex formation. Assembly pathways of this type are well documented for bacterial small pore-forming toxins [59], [60]. A fraction of purified p34 formed oligomers of about 400 kDa in BN-PAGE (Fig. 4D), raising the possibility that p34 hexamers may form double donut-like structures under appropriate conditions. This assumption is supported by data on the complete VacA toxin that similarly showed a formation of 12mers [9], [10]. Previous data from a yeast two-hybrid assay had suggested that p58 may be essential for complex formation of p34 [18]. Our data on purified p34 demonstrate that this subunit is able to assemble autonomously, independently of the p58 component. The central part of p34 (residues 37–292) is sufficient for the formation of stable hexamers.
A crystal structure of the VacA p58 subunit was resolved [20], [61], but the structure of p34 has not yet been determined. Our CD spectra indicate that p34 has a content of >40% β-strands, probably in anti-parallel orientation. This value is similar to data reported for β-barrel proteins, although it does not exclude that the p34 β-strands may assemble in a different structure. The endogenous porin of human mitochondria, hVDAC1, shows a content of 32-37% β-strands, depending on the experimental conditions [45]. A high content of β-strands is also a feature of classical small pore-forming toxins, as exemplified by the α-hemolysin of Staphylococcus aureus (Gouaux, 1997). Similar to α-hemolysin, also p34 is able to form an anion channel of low conductance. The p34 complex seems to act essentially as a small pore-forming toxin targeting the mitochondrial inner membrane.
The pore-forming activity of the VacA holotoxin has been known for a long time [12]–[15], [49], [62], but the role of the two subunits had not been clarified. Surprisingly, we found that p34 is not only able to form a channel in the absence of the p58 subunit, but the conductivity of the p34 pore, about 12 pS in 1.5 M KCl buffer, is very similar to the values that were previously reported for the VacA holotoxin [12]. Both VacA [14], [62] and the purified p34 complex (this study) form an anion channel that can be blocked by the reagent NPPB, demonstrating that p34 is the essential pore-forming subunit of the toxin.
Several observations demonstrated that in the VacA holotoxin, the conductivity of the channel is dependent on the p34 N-terminus. For example, a formation of membrane channels was not detected with VacA containing an amino acid exchange P9A or G14A [16]. Following this observation, a model was proposed, suggesting that the p34 N-termini within the VacA toxin adopt an α-helical structure and associate to form the ion-conducting channel [17]. Remarkably, the first X-ray structures of pentameric ligand-gated ion channels recently confirmed that the central ion-conducting channel of these proteins is indeed formed by α-helices while other parts have a high content of β-strands [63]. However, other data already indicated that the p34 N-terminus might not be essential for pore formation. A mutant VacA protein lacking residues 6–27 oligomerized properly and showed a conductivity similar to the wildtype protein [34]. Strikingly, the current was detectable only after a much longer delay than when compared with the wildtype VacA [34]. Similar effects were described for the s2 subtype of the VacA toxin which is produced by some strains of H. pylori. s2 subtype VacA carries an additional peptide of 12 hydrophilic residues at the N-terminus [1], [7]. It forms membrane channels at a significantly reduced rate as compared to the abundant s1 type, but the channels exhibit similar anion selectivities [15]. Our experiments on the isolated p34 subunit indicate that the N-terminus is essential for mitochondrial targeting, but dispensable for assembly and for channel formation. We therefore regard it as unlikely that the p34 N-terminus is the channel-forming domain. Discrepancies between the data on the purified p34 complex and on the VacA holotoxin could be due to direct or indirect interactions between the N-terminus and the p58 subunit. Functional interactions with distant sites were reported for the N-terminus of Staphylococcus aureus α-hemolysin [64]. However, most effects of modifications in the p34 N-terminus, such as a reduced cell vacuolation [34] could be the consequence of an inhibition in membrane insertion rather than in conductivity.
Strains producing the s2 subtype are less virulent as compared to other strains [65]. We found that the s2 subtype p34 was imported into mitochondria with similar efficiency as the common s1 subtype, and it showed the same conductivity (data not shown). The observations confirm that the N-terminus is not relevant in determination of the channel properties. We assume that the reduced virulence of the corresponding H. pylori strains is due to effects in the pathway of the toxin from the plasma membrane to the mitochondria, in line with the observation that the N-terminus is primarily required for targeting in the cell.
It is currently unclear if p34 separates from p58 after uptake of VacA by target cells. However, our data show that both the VacA holotoxin and the p34 subunit carry the same mitochondrial targeting signal. Moreover, they indicate that both the VacA holotoxin and p34 are able to form hexamers that act as anion channels of very similar conductivity. What are the physiological consequences if these anion channels are formed in the mitochondrial inner membrane? It is difficult to asses the ionic equilibria inside mitochondria because the activities of the ions are largely determined by the extremely high protein content inside these organelles. A study on endosomal membranes suggested an interaction of VacA with pyruvate [66], raising the interesting question if the toxic activity of p34 complexes may depend on a depletion of the mitochondria from organic acids and thus from energy substrates. However, the permeability for pyruvate that was reported for the purified toxin was extremely low [49]. Our own data on the purified and reconstituted p34 indicate an inner diameter of the ion channel of only 0.5 to 1.5 Å. These dimensions correspond to the requirements of a single chloride ion and the ion channel should be unable to mediate the diffusion of complex molecules. We conclude that p34 essentially interferes with the homeostasis of inorganic anions inside the mitochondria.
An increasing number of studies already demonstrated a role of the VacA toxin as a mediator of apoptosis [23], [25], [26], [30], [67]–[69], but the relation of the VacA toxin to the mechanism of apoptosis remained unclear. Modifications of the inner mitochondrial membrane potential seem to be the first sign of a VacA activity in relation to cell death [27]. Remarkably, the reagent NPPB that we found to efficiently block the conductance of the p34 ion channel also inhibits VacA-dependent cytochrome c release and apoptosis [26], [30]. In the context of our data we suggest that (i) the anion conductivity described for the VacA holotoxin is due to the autonomous oligomerization and pore-formation of subunit p34, (ii) p34 contains an N-terminal signal for targeting to the mitochondrial inner membrane, deletion of the N-terminus abolishes transport to the mitochondrial target membrane. (iii) The p34 N-terminus is dispensable for pore-formation. (iv) Both pore formation and inner membrane targeting are essential features of p34 toxicity. The p34 subunit of the Helicobacter pylori VacA toxin may serve as a model system in further investigations on the pathological consequences of an anion channel in the mitochondrial inner membrane.
HeLa cells were cultured in DMEM (Bio-Wittaker, Verviers, Belgium) with 5% fetal calf serum (Bio-Media, Boussens, France), penicillin/streptomycin and 2 mM L-glutamine (Gibco-BRL, Paisley, UK) at 37°C in an incubator with 5% CO2. For expression of EGFP-labelled proteins, the corresponding DNA fragments were cloned into the vectors pEGFP-C1 or pEGFP-N1, respectively (Clontech). The cells were transfected using Lipofectamine 2000 reagent (InvitroGen). Immunofluorescence labelling was performed following a conventional protocol. Briefly, 24 h after transfection, cells grown on glass coverslips were rinsed with PBS and fixed for 15 min with a 3.7% paraformaldehyde solution. After a 4 min permeabilization in PBS+0.1% Triton X-100, the coverslips were incubated in a 1∶100 dilution of a rabbit anti-Tom20 antibody (Santa-Cruz). A secondary antibody raised in donkey and coupled to Cy3 (Jackson Immunoresearch) was subsequently used, together with DAPI (Sigma). The coverslips were finally rinsed and mounted in Mowiol mounting medium (Calbiochem) prepared according to the manufacturer's instructions. Pictures were acquired on a TCS SP2 confocal microscope (Leica) equipped with a 63× HCX PL APO, NA = 1.40 objective (Leica), under oil immersion. Following acquisition, the images were combined using Photoshop software (Adobe).
Plasmids for expression in HeLa cells: The constructs encoding EGFP-p34(1–319) or p34(1–319)-EGFP, respectively, were described previously ([23]; the VacA sequence is available using the NCBI acc. no S72494.1 or GI∶619248). To obtain 34(37–319)-EGFP, the plasmid encoding p34(1–319)-EGFP was cut with XhoI and EcoRI within the p34 sequence, and the gap was closed using a synthetic DNA fragment encoding residues 37–58. For construction of p34(1–36)-EGFP, a synthetic DNA fragment encoding the N-terminal 36 amino acids of p34 was introduced in pEGFP-N1 via the XhoI and BamHI site. In both cases, the synthetic DNA fragments were obtained by hybridization of two complementary oligonucleotides, encoding the desired sequence together with an initiating start codon and containing the appropriate restriction site overhangs for subsequent cloning.
Plasmids for synthesis in reticulocyte lysate and for expression in Escherichia coli: For construction of p34(37–319), the plasmid pET28a-p34(1–319) was used as template [23]. This plasmid contains a NcoI site upstream of the p34 coding sequence. An additional NcoI site was introduced upstream of the codon of residue 37. By cutting with NcoI, the DNA segment corresponding to the N-terminal residues 1–36 was subsequently removed and the remaining plasmid was religated. For construction of p34(1–35)-DHFR, a BamHI site was introduced directly downstream of the codon of residue 35 in the plasmid pET28a-p34(1–319). The main part of the p34 coding sequence was subsequently removed by cutting with BamHI and HindIII and substituted by a DNA fragment encoding the entire DHFR (dihydrofolate reductase) of the mouse. (His)10-p34(37–319) was constructed by insertion of the p34 sequence encoding residues 37–319 into the plasmid pET10N (a modified pET19b plasmid; [70]). The DNA segment encoding p34(37–319) was obtained from pET28a-p34(1–319) by PCR, creating a NotI site upstream of the triplet encoding p34 residue 37, and subsequent cutting with NotI and XhoI. The DNA segment was ligated into pET10N downstream of the sequence for the dekahistidine tag. Due to the NotI site, the histidine residues of (His)10-p34(37–319) and the glutamic acid of position 37 are connected by a linker of three alanine residues. Essentially following the same procedure, a DNA insert encoding p34(37–292) was amplified from pET28a-p34(1–319) by PCR, creating a NotI site in front of the codon of p34 residue 37 and a XhoI site after the codon of residue 292.
For import of radiolabelled proteins, rat liver mitochondria were used within 1 h after the isolation. The livers were obtained from animals of 80–120 g weight. Liver pieces were homogenized in buffer A (300 mM sucrose, 2 mM EGTA, 10 mM Tris-HCl pH 7.4) containing BSA (5 mg/ml, fatty acid free) and 1 mM PMSF using a Dounce homogenizer. Cell debris was removed by centrifugation (500 g, 10 min, 4°C). The supernatant was centrifuged at 12.000 g for 6 min to obtain a crude fraction of mitochondria. The mitochondria were subsequently resuspended in 12 ml buffer A, and Percoll (Sigma, P1644) was added to a final concentration of 5% (v/v). The mitochondria were pelleted by centrifugation at 17.000 g for 10 min, washed once in buffer A, and eventually resuspended in buffer A at a final concentration of 10 mg protein/ml. Yeast strains were grown in YPG medium (1% (w/v) yeast extract, 2% (w/v) bacto-peptone, pH 5.0, containing 3% (v/v) glycerol and mitochondria were subsequently isolated following standard procedures [71].
Radiolabelled proteins were imported into yeast and mammalian mitochondria following similar protocols [71]. The proteins were synthesized in rabbit reticulocyte lysate (TNT T7 Coupled Reticulocyte Lysate System, Promega, L4610) in the presence of 35S-labeled methionine (ICN Biomedical Research Products). For import into yeast mitochondria, the reticulocyte lysate containing p34 was preincubated with HCl (final concentration 30 mM) at pH 5–5.5 for 10 min at 25°C. (With most preparations of yeast mitochondria, the efficiency of p34 import without acid pretreatment was very low.) For import into mammalian mitochondria, the acid pretreatment was omitted. For protease-protection assays, the samples contained BSA buffer (3% [w/v] BSA, 80 mM KCl, 10 mM MOPS-KOH, pH 7.2), 2 µl reticulocyte lysate, 2 mM NADH, 1 mM ATP, 20 mM potassium phosphate and 30 µg (yeast) or 40 µg (rat liver) mitochondrial protein in a total volume of 100 µl. The import reactions were carried out at 25°C. The samples were subsequently cooled on ice and proteinase K was added at a final concentration of 25 µg/ml. Following an incubation for 10 min at 0°C, the protease was inactivated by 2 mM PMSF (phenylmethylsulfonyl fluoride) and an additional incubation for 5 min at 0°C. To dissipate the membrane potential, valinomycin (Sigma, V-0627) was used at a final concentration of 1 µM. Digitonin was used as described previously [72].
For preparation of membrane vesicles, mitochondria (2 mg protein in 200 µl SEM) were mixed with 200 µl 0.6 M Sorbitol, 20 mM HEPES-KOH pH 7.4 and incubated for 5 min at 0°C. 2.6 ml 0.5 M EDTA, 20 mM HEPES-KOH pH 7.4 and 100 mM PMSF were added for swelling of the mitochondria. Following an incubation for 30 min at 0°C, a mixture of protease inhibitors was added. Swelling of the mitochondria was stopped by the addition of sucrose to a final concentration of 1.8 M and an additional incubation for 10 min. Membrane vesicles were formed by sonication using a sonifier (Branson 250; duty cycle 70%, Output control 3). Each sample was treated with 3 cycles of each 30 sec. sonification and 15 sec., using a 3 mm Microtip (Heinemann, Schwäbisch Gmünd). The suspension of vesicles obtained by sonification was centrifuged for 10 min at 16.000 g to remove residual mitochondria. The vesicles were collected from the supernatant by centrifugation at 160.000 g (30 min., 4°C). The membranes were carefully resuspended in 400 µl 10 mM KCl, 5 mM HEPES-KOH pH 7.4 and 100 mM PMSF. The suspension was centrifuged for 10 min at 16.000 rpm×g to remove aggregates. The supernatant was applied on a step gradient of 0.85, 1.1, 1.35 and 1.6 M sucrose in 100 mM KCl, 5 mM HEPES-KOH pH 7.4 in a total volume of 11 ml, using Ultra-Clear centrifuge tubes (Beckman, 14×95 mm, No. 344060). The centrifugation was carried out for 16 h at 100.000×g using a SW41 rotor (Beckman, 30.000 rpm, 4°C). 1 ml fractions were collected for TCA precipitation and SDS-PAGE. The proteins were subsequently transferred on nitrocellulose and polyclonal antisera were used for labelling of marker proteins.
p34 was expressed in Escherichia coli, strain C43(DE3), using the vector pET28a (Novagen). After an induction by 1 mM IPTG in a culture of 3000 ml for 3 h at 37°C, the cells were harvested and then opened using a french press. Inclusion bodies containing p34 were recovered by centrifugation and washed once in 100 mM urea, 1% (v/v) Triton X-100, 10 mM Tris/HCl pH 8.0, 0.1% (v/v) mercaptoethanol, and subsequently three times in 1 M urea, 10 mM Tris/HCl pH 8.0, 0.1% (v/v) mercaptoethanol. The inclusion bodies were eventually dissolved in 8 M urea, 100 mM Na2HPO4, 1 mM EDTA, 10 mM Tris/HCl pH 8 (PETurea, 5 ml/g cells), cell debris was removed by centrifugation. The supernatant was incubated for 30 min at 0°C with ammonium sulfate corresponding to a final concentration of 10% saturation. Precipitated proteins were removed and a fraction containing most of the p34 was precipitated from the solution by addition of ammonium sulfate at 30% saturation and 90 min incubation on ice. The precipitate was dissolved in PETurea buffer. The solution was diluted 1∶1 with 4 M urea, 1 M ammonium sulfate, 100 mM Na2HPO4, pH 8 and applied to a column containing 3 ml Phenyl-Sepharose (Amersham Pharmacia). The column was washed with fractions of decreasing concentrations of ammonium sulfate (500 mM, 300 mM) and p34 was eventually eluted with 4 M urea, 100 mM Na2HPO4, pH 8. The solution was dialyzed over night against 2 M urea, 1 mM EDTA, 50 mM Tris/HCl pH 8 (TEurea). The solution was then applied to a 3 ml DEAE-Sephacel column. p34 was easily eluted in subsequent washing steps using the TEurea buffer. Impurities were eventually eluted using TEurea buffer containing 1 M NaCl. Starting with 1 g E. coli cells, about 0.2 mg p34 was isolated. The truncated protein p34(37–319) and derivatives containing single amino acid exchanges were purified following the same protocol.
Derivatives of p34 (comprising residues 37–319 or 37–292) containing a (His)10-tag were isolated by Ni-NTA affinity chromatography. For expression, we used the E. coli strain C43. The expression was induced by addition of 1 mM IPTG following standard conditions and continued for 4 h at 37°C. The cells were opened using the sonifier (Branson 250; 2×1.30 min, 30% amplitude, 1.5 sec impulse, 0.5 pauses) and dissolved in 8 M urea, 1 mM EDTA, 50 mM Tris/HCl pH 8.0. Following a clarifying spin, the solution was applied to a 1 ml Ni-NTA column (Histrap, Amersham Pharmacia) using an Äkta-Prime system. (His)10-p34 eluted in a gradient at concentrations between 50 and 200 mM imidazole. The eluate was dialyzed over night against 4 M urea, 1 mM EDTA, 50 mM Tris/HCl pH 8 or 2 M urea, 1 mM EDTA, 50 mM Tris/HCl pH 8. The eluate was applied to a ‘Superose 6’ 10/300 or ‘Superose 12’ 10/300 column (Amersham Pharmacia). About 20 mg (His)10-p34 were obtained from 1 g E. coli cells. Samples of purified (His)10-p34 were used to immunize two rabbits and to obtain polyclonal antisera. The molecular mass of the eluted oligomers was determined using standard marker proteins (Amersham Pharmacia). The average elution position, Kav, was calculated using the equation Kav = (Ve - V0)/(Vt - V0), with Ve representing the elution volume, V0 the void volume, and Vt the total column volume.
For cross-linking, 30 µg p34 were dissolved in 0.5 ml 2 M urea, 1 mM EDTA, 10 mM MOPS, pH 7.2. DSS (Disucciminidylsuberate, Pierce Biotechnology Inc.) was used at a final concentration of 50 µM, Sulfo-MBS (Sulfo-m-maleimidobenzoyl-N-hydroxysulfo-succinimide ester, Pierce) was added at a final concentration of 0.5 mM as described previously [72]. Blue native electrophoresis (BN-PAGE) was carried out according to published procedures [47], [72]. Complexes of purified p34 (30 µg/lane), dissolved in 0.5% Triton X-100, 10% Glycerol, 50 mM NaCl. 0.1 mM EDTA, PMSF 1 mM, 20 mM Tris-HCl pH 7.0, were separated in gels containing 500 mM ε-aminocaproic acid (EACA).
Electrophysiological characterization of p34 was carried out using the planar lipid bilayer technique as detailed in ref. [73]. Briefly, purified urea solubilised p34 was applied directly below the bilayer in the cis chamber. An acid pretreatment of the protein was omitted. Buffer conditions were symmetrical with 1.5 M KCl, 10 mM Mops-Tris (pH 7.0) in the cis/trans compartment or 1.5 M KCl, 10 mM Na-Acetat (pH 4.0) in the cis/trans compartment. Two Ag/AgCl electrodes covered by 2 M KCl-agar bridges were inserted into each chamber with the trans chamber electrode connected to the headstage (CV-5-1GU) of a Geneclamp 500 current amplifier (Axon Instruments) and thus was the reference for reported membrane potentials. A solution of purified azolectin (60 mg/ml; Sigma type IV-S) in n-decan (purity >99%, Sigma) was used to generate the planar lipid bilayers. Current recordings were carried out using a Digidata 1200 A/D converter. Data analysis was performed by self written Windows-based SCIP (single-channel investigation program) in combination with Origin 7.0 (Microcal Software). Current recordings were performed at a sampling interval of 0.1 ms, filtered with a low-pass-filter at 2 kHz.
Purified p34 was dialyzed against 8 mM N-Decyl-β-D-Maltopyranosid, 10 mM KCl, 20 mM K2HPO4/KH2PO4, pH 7.0. CD-spectroscopy and calculation of the secondary structure of p34 was performed as described in ref. [74]. Briefly, CD spectra were recorded using a Jasco J-810 spectrapolarimeter. All measurements were carried out in a quartz cuvette with an optical path length of 0.01 cm at room temperature. The scans (n = 16) were averaged to improve the signal/noise ratio. Blank buffer spectra were collected and subtracted from the sample spectra.
|
10.1371/journal.pntd.0007266 | Detecting space-time clusters of dengue fever in Panama after adjusting for vector surveillance data | Long term surveillance of vectors and arboviruses is an integral aspect of disease prevention and control systems in countries affected by increasing risk. Yet, little effort has been made to adjust space-time risk estimation by integrating disease case counts with vector surveillance data, which may result in inaccurate risk projection when several vector species are present, and when little is known about their likely role in local transmission. Here, we integrate 13 years of dengue case surveillance and associated Aedes occurrence data across 462 localities in 63 districts to estimate the risk of infection in the Republic of Panama. Our exploratory space-time modelling approach detected the presence of five clusters, which varied by duration, relative risk, and spatial extent after incorporating vector species as covariates. The Ae. aegypti model contained the highest number of districts with more dengue cases than would be expected given baseline population levels, followed by the model accounting for both Ae. aegypti and Ae. albopictus. This implies that arbovirus case surveillance coupled with entomological surveillance can affect cluster detection and risk estimation, potentially improving efforts to understand outbreak dynamics at national scales.
| Dengue cases have increased in tropical regions worldwide owing to urbanization, globalization, and climate change facilitating the spread of Aedes mosquito vectors. National surveillance programs monitor trends in dengue fever and inform the public about epidemiological scenarios where outbreak preventive actions are most needed. Yet, most estimations of dengue risk so far derive only from disease case data, ignoring Aedes occurrence as a key aspect of dengue transmission dynamic. Here we illustrate how incorporating vector presence and absence as a model covariate can considerably alter the characteristics of space-time cluster estimations of dengue cases.
| Dengue disease, a viral infection transmitted to humans by Aedes mosquitoes, is endemic to 128 countries, with 3.9 billion people considered at-risk globally [1]. Dengue disease cases have increased dramatically worldwide throughout the previous several decades [2], likely a result of urbanization [3], globalization [4], climate change [5], the breakdown of regional control programs [6], and the spread of the invasive Aedes albopictus [7]. As a result of both recent and historical risk, many countries employ national surveillance programs to monitor trends in dengue and inform local health authorities to the places and times where preventative practices are most required. However, despite the commonality of these programs and unforeseen cost of cutting them [8], surveillance budgets are often limited [9,10], restricting the scope and quality of the work. This is concerning in developing regions such as Central America, where the burden of disease is high [1] and per capita public health expenditure is among the lowest of any region of the world [11].
Surveillance of both diseases and vectors is an essential component of integrated disease management programs that can be used to determine risk changes in space and time, thus providing the evidence for more targeted prevention and control interventions [12]. Nevertheless, with few exceptions, it is rare for surveillance programs to concurrently monitor both arbovirus cases and vector populations in the same locations and at regular intervals. Most projections of disease risk used to justify public health actions are derived purely from disease case data, ignoring vector population dynamics, which is a key aspect of the vector transmission model. This is particularly concerning when more than one vector species is present, and little is known about their likely role in local transmission, which may result in inaccurate or incomplete risk projection or case clustering models.
The Republic of Panama has been monitoring dengue cases alongside vector presence through the National Department of Epidemiology (NDE) since 1988, making it one of the most long-standing surveillance programs of its kind in Latin America. Of the two known dengue mosquito vectors, Ae. aegypti is considered resident to Latin America and Panama since the 19th century, and the primary source of transmission [12] while Ae. albopictus, considered a secondary vector, has been spreading throughout the region ever since it got introduced in Panama in 2004 [13,14]. Widespread extirpation of Ae. aegypti by a superior ecological competitor like Ae. albopictus has occurred throughout the world in recent decades [15–17], with unknown consequences on arbovirus transmission risk. Encompassing this period of growing interspecific competition among two vector species, Panama’s surveillance system is particularly unique and potentially useful to modelling dengue transmission risk while considering Aedes species interaction. Attaining a better understanding of dengue outbreak dynamics over time may improve the capacity of public health authorities to combat the spread of other arboviruses, such as Zika Virus and Chikungunya Virus.
Our overall aim is to examine the influence that concurrent dengue case surveillance and Aedes species monitoring can have on cluster detection and relative risk estimation. Based on previous studies incorporating covariates into spatiotemporal cancer cluster detection [18,19], we hypothesize that the inclusion of vector surveillance data will alter the identification of dengue clusters in space-time. In so doing, we describe the results of 13 years of dengue and Aedes surveillance data, including two competing vector species plus virus data originating from long-term cooperatively organized surveillance programs. We believe this is the first effort to adjust for vector presence and absence in a vector-borne disease cluster detection model, which we hope sheds light on the characteristics of space-time clusters and relative risk estimation of dengue after Aedes species are used as model covariates.
We utilized dengue prevalence data collected by the National Department of Epidemiology (NDE), housed within the Panamanian Ministry of Health (MINSA) [20]. Systematic national surveillance of dengue cases in Panama have been continuous since 1988. Suspected cases are defined by a patient with a fever and one or more of the following symptoms: headache, retro orbital pain, myalgia, exanthema, rash, vomiting, malaise, leukopenia, and jaundice. A confirmed case is defined as a suspected case with a positive dengue test, conducted using either viral isolation, reverse transcription polymerase chain reaction (RT-PCR), IgM enzyme-linked immunosorbent assay platform (ELISA), or secondary IgG ELISA. RT-PCR was established as the original standard by the National Reference Laboratory at the Gorgas Memorial Institutes for Health Studies (ICGES) in 2003. Yet since 2009, MINSA established national decentralization of serological confirmation of dengue using ELISA tests, which has improved efficiency by allowing district health officials to confirm cases without needing to send samples to a single central facility in Panama City. Data is recorded at the Corregimiento, or neighborhood, scale as the number of confirmed cases in a given year at a given location. This is the lowest scale of data granularity available, and thus, we do not have patient-level detail nor temporal detail at smaller units than year.
We utilized vector data from the Vector Control Department (VCD) at MINSA. Systematic entomological surveillance has occurred in Panama since 2000 in order to establish Aedes infestation rates, and thus, areas of potential dengue transmission risk. Surveys of both Ae. aegypti and Ae. albopictus are performed annually at the Corregimiento-scale and consist of solely larval surveillance. Each year, a random block of houses is chosen and all houses in the block are searched for containers holding Aedes larvae. The larvae are collected and allowed to mature to the fourth instar, at which point they are taxonomically identified to species based on morphological keys [21]. The number of houses positive for Ae. aegypti, Ae. albopictus or both are recorded in the raw datasets. However, because there is no record of the number of houses in each block, we have transformed the data into a presence-absence format in each Corregimiento rather than analyzing the number of positive houses.
We conducted our analyses on dengue and vector data from 2005–2017, encompassing the period in Panama when both Ae. aegypti and Ae. albopictus have been interacting. Overall, data was collapsed from the original Corregimiento scale to the district scale. This is due to unreliable human population estimates at scales smaller than the district. Human population data was gathered from the National Institute of Statistics and Census (INEC), which conducts a national census every 10 years. Because the national census is only conducted every ten years, we used the population levels from 2010 to calculate PR for data from 2005–2017. While this is not ideal, and incurs inherent error in the year to year accuracy of the PR estimate, there is no more frequent population estimate available. This is an unfortunately common situation, especially in Central America, where no country conducts national population assessments more frequently than every 10 years.
For the spatial analyses, we utilize discrete Poisson space-time modelling STSS [22], which systematically moves cylindrical search windows across the geographic and temporal space to detect space-time clusters. Essentially, STSS determines if the observed disease cases in a particular region and time period exceed the expected cases under baseline conditions. In vector-borne disease research, STSS have been used to examine outbreaks of dengue [23–25], chikungunya [26], malaria [27,28], Chagas [29], and West Nile [30,31], for example. STSS have also been used to examine the co-circulation of dengue and chikungunya in Colombia [32]. However, none of these prior studies have utilized vector surveillance data as a covariate.
The cylinders are centered on the centroids of the Panamanian districts while the base of a cylinder is defined as the spatial scan, and the height of a cylinder represents the temporal scan. The number of observed and expected dengue cases are computed for each cylinder. Conceptually, a vast number of cylinders of various space-time dimensions are generated until an upper bound is reached, while each cylinder is a potential cluster. For this study, the maximum spatial scan was set to 25% of the total population in Panama, while the maximum temporal scan was set to 4 years. A Poisson-based likelihood ratio is calculated for each cylinder, which is proportional to (n/μ)n[(N-n)/(N-μ)]N-n[33]. For the parameters, μ is the expected number of dengue cases in a cylinder, and n is the total observed dengue cases in the cylinder. The expected number of dengue cases is computed by multiplying the fraction of population that lives within the cylinder (p) by the total number of cases in Panama (C) divided by the total population (P), that is: E[c] = p*C/P - The cylinder with the highest likelihood ratio is the most likely space-time cluster. To evaluate the statistical significance of the candidate space-time clusters, 999 Monte Carlo simulations are performed under the null hypothesis that there are no significant clusters. Subsequently, we report secondary space-time clusters with a p-value less than 0.05. It must be noted that the cylindrical shape of the clusters does not represent the true shape of the clusters. While it is possible to use irregular search windows [34–36], cluster borders in the model outputs are not perfectly in line with the borders of the risk in reality.
For this study, we ran four STSS models: (1) dengue cases only; (2) dengue cases controlled for the presence and absence of Ae. aegypti and/or Ae. albopictus (i.e. absence of both species, Ae. aegypti presence, Ae. albopictus presence, and presence of both species); (3) dengue cases controlled for Ae. aegypti presence/absence only; and (4) dengue cases controlled for Ae. albopictus presence/absence only. For the covariate adjusted models, the expected number of dengue cases is defined the same way for the non-adjusted model, but includes covariate category i. That is: E[c] = ∑ipi*CiPi. In other words, the adjusted STSS searchers for clusters “above and beyond that which is expected due to these covariates” [37]. For each model, we also report the relative risk of prevalence in each district that belongs to a space-time cluster, which is defined as (c/e)/[(C-c)/(C-e)], where c is the total observed dengue cases in a particular district; e is the expected cases in a district; and C is the total observed dengue cases in the country of Panama. Clusters with a relative risk > 1 indicates that there were more observed dengue cases than expected under baseline conditions. We created all maps in ArcGIS [38].
From 2005–2017, there were a total of 49,910 cases of dengue in Panama (Fig 1), with 2009 and 2014 being the most severe at 6,941 and 7,423 cases respectively. These two years represented 28% of the total dengue cases during the 13-year period. The spatial distribution of the total number of dengue cases and the crude rate per 10,000 people per district is shown in Fig 2A and 2B, respectively. The district with the largest number of dengue cases between 2005–2017 was San Miguelito (13,109 cases), which is situated in Metropolitan Panama City. The highest rate of dengue per 10,000 people during this same time frame in Panama was shared among major urban centers across the country, including again the districts of San Miguelito (416.13), Arraijan (310.0), and Chepo (294.76) in the Province of Panama plus Aguadulce (384.45), Santiago (338.21), Nata (269.69), Guarare (277.42) in central Panama, and Bocas del Toro (758.59) and Changuinola (298.74) in northwestern Panama (Fig 2A and 2B). Additionally, surveillance of each vector species indicated prolonged endemic presence of Ae. aegypti alongside increasing presence of Ae. albopictus from just one district in 2005 to 53 in 2016 (Fig 3).
The results of our space-time modelling detected the presence of five clusters in each of the four models, varying by cluster center and duration (Figs 4–7 and Table 1). Incorporating covariates into the models had considerable effects on the duration, relative risk (RR), and spatial extent of clusters (Table 2). The model adjusting for the presence of Ae. aegypti encompassed the greatest spatial range and highest number of districts with a RR > 1, while the model adjusting for the presence of Ae. albopictus encompassed the smallest spatial range and the lowest number of districts with a RR > 1. The duration of the space-time clusters is notably different when adding the vector surveillance data to the model, however, the one exception is cluster 1 for each model (most likely cluster). For example, the duration of cluster 2 was 2015–2017 for the no covariate and Ae. aegypti model; while the Ae. albopictus and Aedes (both) model reported a duration of only 1 year, which occurred six years earlier (2009). Furthermore, cluster 2 was found in different geographic locations for the Aedes (both) and Ae. albopictus models. This variation in duration of the clusters between the four models is a result of adjusting for the presence of Aedes during the 13-year study period. In other words, the start, end, and duration of the clusters is substantially affected by the presence of one or more Aedes species. The relative risk may be higher if Aedes was found in a district during the entire duration of a space-time cluster. During the 13 years of our study period combined with the 63 districts containing data (13 * 63 = 819), Ae. aegypti was present 690 times, Ae. albopictus was present 245 times, while both Aedes species were found in a district 224 times. As a result, the difference in species presence during the study period partly explains why the clusters for the Ae. albopictus model contained 19 less districts than the Ae. aegypti model, and 10 less districts that the model adjusting for both species.
Overall, our results highlight the extreme heterogeneity of dengue is Panama. We find that clusters and risk vary widely across both time and space, with adjacent districts and years experiencing vastly different rates of transmission. While this may complicate response programs that operate under assumptions of spatially and temporally homogenous risk, it highlights the need for comprehensive risk projection studies which can identify particularly regions that most require attention. Additionally, our results specifically highlight the changes incurred by adding vector data into systematic dengue risk projections. In the model where Ae. aegypti presence and absence was accounted for, more than double the number of districts were contained in clusters than the model where Ae. albopictus presence and absence was accounted for. The Ae. aegypti model also contained the highest number of districts with a relative risk > 1, indicating more dengue cases than would be expected given baseline population levels. As an invasive species that has systematically replaced Ae. aegypti throughout numerous regions in its endemic range [15], Ae. albopictus has been spreading throughout Panama for the previous 13 years [13,14]. Globally, while Ae. albopictus has been implicated in several small outbreaks [39], the majority of dengue serotypes are thought to be transmitted by Ae. aegypti, due to its preference for both urbanized habitat [17,40] and human hosts [41,42].
Overall, based on our findings, we suggest that efforts be undertaken to further understand how the incorporation of vector surveillance data can affect risk projections, especially given the ubiquity of exploratory space-time scan statistics studies in the literature that did not utilize covariate surveillance data [26,43,44]. SATSCAN results can be useful in assessing risk when data granularity is low, as is often the case in developing regions with low-resource surveillance programs. Overall, these results can direct efforts to attain finer scale data in regions where general risk is deemed the greatest, at which point more confirmatory models can be applied, validating the role of certain covariates in space-time risk estimation. Balboa, for example, was identified as a cluster in all four models and had a steady presence of Ae. aegypti throughout the sample period as well as increasing presence of Ae. albopictus since 2006. This district is semi-urban with approximately 2400 people spread across 400km2 area. Another district, San Miguelito in metropolitan Panama City, contained the most observed cases during our study period, despite being only 49.9km2. This district can be characterized by high density housing and residents of relatively low socioeconomic status, which has previously been linked to increased vector-borne disease risk [45–48]. The staggering number of cases should be a cause for concern, yet its small geographic area may facilitate public health interventions such as vector control and community education. Overall, now that the identification of high risk districts at the national scale has been completed and informed by vector presence, the subsequent step of illustrating the comparative characteristics of each district relative to dengue transmission risk can be undertaken. Understanding what caused Balboa and San Miguelito to experience such high relative risk, for example, is the next task necessary for adjusting public health interventions to effectively address the needs and conditions of each district.
Despite the longevity of our data and thoroughness of the surveillance efforts, there are clear considerations and limitations of our work which we would like to see addressed in future studies. First, we recognize that there is no way to statistically compare the outputs of the four models. This exploratory analysis was meant to illustrate that vector surveillance data can impact the location and duration of detected clusters, yet ranking the four models by predictive power is not possible with the statistical methods employed. Second, we recognize the spatial uncertainty by using districts for the analysis and our study is subject to the modifiable areal unit problem (MAUP) [44], however, this is an exploratory study that seeks to guide targeted interventions. Finer-level data can identify the locations (e.g. towns, blocks, homes) within each district that are experiencing the greatest burden of dengue, yet since such fine scale surveys cannot realistically be employed at the national-scale, studies such as ours can indicate where to direct future surveillance efforts. Third, it is possible that the reported cases of dengue in certain districts are travel cases (seeking treatment in a district different than actual residence), and while adjusting for the presence of Aedes can shed light on the districts where an individual is more likely to get infected with dengue, we cannot confirm that every patient was specifically infected in Panama. The lack of population data for more than one year across such a lengthy period is another limitation of this work. While the frequency of a census in Panama is on par with much of Latin America, this greatly impacts our ability to determine accurate prevalence rates year to year. Since linear interpolation is often inaccurate for non-linear trends like population growth rate, we would like to see more frequent population assessments conducted in regions where dengue is an ongoing risk, and while we understand that resources may not easily allow for this, the role of national census efforts in public health is often under-appreciated. An additional limitation originates with the vector surveillance methods employed. Values are reported as the number of houses containing larvae of each respective species. No information is given on the number of houses surveyed, and thus we were forced to transform the data into presence and absence. Had the total number of surveyed houses been reported, we would have been able to compute each district’s infestation rate, which would have provided a scaled and more nuanced covariate in the model. Lastly, no data was available on short-term cross immunity and long-term serotype specific immunity, which may have incurred bias in the model parameter estimations.
Overall, it is key to recognize that adding vector surveillance data as a covariate changes the location, duration, and relative risk of dengue case clusters. As methods such as SaTScan are often used in mapping areas for public health intervention, we suggest that future efforts to utilize these tools in spatial vector epidemiology are conducted with an awareness that outputs can change when vector data is utilized as a covariate. Maps of clusters created without vector surveillance data may be missing key information that alters the distribution of the clusters in space-time, thus creating possible situations where public health efforts may not be planned in the areas which require it most. Although unadjusted cluster analysis is a valuable and commonly utilized tool for public health officials to identify high risk areas of vector-borne disease, our study illustrates the role that incorporating relevant covariates can play in altering the model output. While this has been demonstrated in cancer cluster studies [48,49], this is the first use of covariates in space-time cluster detection modelling of neglected tropical disease. With this comes potential to expand into other classes of covariates. For example, in addition to vector surveillance data, we support the incorporation of additional covariates such as vector genetic background, climate, vegetation, and land cover to dengue cluster models. Host population characteristics, such as housing density, relative isolation, and connectivity may also influence dengue risk in space-time scan statistic models. In general, the dengue transmission model contains numerous variables that we would have been interested in incorporating, had adequate data been available. Still, we demonstrate that vector surveillance clearly provides valuable information in the determination of virus case clusters, and thus should be conducted alongside virus surveillance so that it may be included in modelling efforts.
|
10.1371/journal.pcbi.1004209 | Computational Modeling of Seizure Dynamics Using Coupled Neuronal Networks: Factors Shaping Epileptiform Activity | Epileptic seizure dynamics span multiple scales in space and time. Understanding seizure mechanisms requires identifying the relations between seizure components within and across these scales, together with the analysis of their dynamical repertoire. Mathematical models have been developed to reproduce seizure dynamics across scales ranging from the single neuron to the neural population. In this study, we develop a network model of spiking neurons and systematically investigate the conditions, under which the network displays the emergent dynamic behaviors known from the Epileptor, which is a well-investigated abstract model of epileptic neural activity. This approach allows us to study the biophysical parameters and variables leading to epileptiform discharges at cellular and network levels. Our network model is composed of two neuronal populations, characterized by fast excitatory bursting neurons and regular spiking inhibitory neurons, embedded in a common extracellular environment represented by a slow variable. By systematically analyzing the parameter landscape offered by the simulation framework, we reproduce typical sequences of neural activity observed during status epilepticus. We find that exogenous fluctuations from extracellular environment and electro-tonic couplings play a major role in the progression of the seizure, which supports previous studies and further validates our model. We also investigate the influence of chemical synaptic coupling in the generation of spontaneous seizure-like events. Our results argue towards a temporal shift of typical spike waves with fast discharges as synaptic strengths are varied. We demonstrate that spike waves, including interictal spikes, are generated primarily by inhibitory neurons, whereas fast discharges during the wave part are due to excitatory neurons. Simulated traces are compared with in vivo experimental data from rodents at different stages of the disorder. We draw the conclusion that slow variations of global excitability, due to exogenous fluctuations from extracellular environment, and gap junction communication push the system into paroxysmal regimes. We discuss potential mechanisms underlying such machinery and the relevance of our approach, supporting previous detailed modeling studies and reflecting on the limitations of our methodology.
| Neurons communicate via different types of synapses on very fast time scales. The combination of hundred thousand of such interconnected cells within a fluctuating extracellular environment forms a complex network that gives rise to function and behavior via the formation of dynamical patterns of activity. In the context of epilepsy, the functional properties of the network at the source of a seizure are disrupted by a possibly large set of factors at the cellular and molecular levels. It is therefore needed to sacrifice some biological accuracy to model seizure dynamics in favor of macroscopic realizations. Here, we present a neuronal network model that convenes both neuronal and network representations with the goal to describe brain dynamics involved in the development of epilepsy. We compare our modeling results with animal in vivo recordings to validate our approach in the context of seizures. Such system-level methodology has significant bearing in understanding neuronal network dynamics that entangle multiple synaptic and extracellular modalities.
| Epilepsy is characterized by seizures, a paroxysmal behavior that results from abnormal, excessive or hypersynchronous neuronal activity in the brain [1], with a various set of symptomatic outcomes depending on brain regions involved in its generation and propagation processes. Clinically, epilepsy affects 1% of the population, from whom 30% are drug-resistant. Physiological investigations of neural tissue in the context of human Temporal Lobe Epilepsy and experimental models revealed neuronal loss in the hippocampus, rewiring of excitatory and inhibitory pathways [2], in keeping with the hypothesis on unbalanced excitation/inhibition ratio observed in epilepsy [3,4].
Understanding seizure mechanisms from micro to macro scales is necessary to provide clinicians and basic scientists with a reliable theoretical basis to develop new therapeutic approaches. Computational modeling reproducing brain activity is a genuine approach to investigate such multi-scale paradigms. Neural network models in the context of epilepsy typically use multi-compartment Hodgkin-Huxley type neurons with a collection of ion-channels dynamics and multiple excitatory and inhibitory synaptic combinations. We will here refer to them as biophysically-realistic, see for example [5,6]. Reduced population models (so-called neural masses or mean field models) absorb a significant amount of biophysical details in constant parameter values and are referred to as large-scale or macroscopic [4,7–9], see for a review [10]. A third type of modeling scheme consists in approaching the dynamics of seizures in an abstract manner, and describing them in terms of generic dynamic features [11]. The advantage of this approach is its generality, allowing the identification of invariant seizure classes based on basic dynamical properties. The drawback lies in the difficulty to find biophysical correlates to the state variables used in such an approach. Certain elementary features such as dynamics evolving on different time scales guides the identification of the biophysical correlates. For example, recent emphasis in seizure modeling is directed towards the role of extracellular environmental fluctuations, which evolve on a significantly slower time scale than neuronal discharges. By incorporating slow extracellular potassium or oxygen levels as key parameters, state-of-art studies displayed transitions between pathological brain states observed during paroxysmal activity [12–14]. Such approaches combine dynamical systems theory and large-scale neural network computations to propose key insights into seizure mechanisms. However, extracellular potassium homeostasis provides only a partial answer. Many different biophysical factors can lead to seizure genesis [15,16], in keeping with the concept that different parameter sets can produce the same type of activity at the network level [17]. Introducing all those parameters in a detailed model poses a computational and theoretical challenge, but may be only useful if the arising network behavior can be characterized dynamically.
In the present study, we make the link of the seizure state dynamics across neuron and population levels explicit. Fig 1 illustrates the line of thought deriving the here presented intermediate system architecture (panel B) from the phenomenological model (panel A); future work could comprise spatially structured networks of neurons (panel C). We specifically develop a network of inhibitory spiking and excitatory bursting neurons, driven by a slow environmental variable, reproducing characteristic features of the temporal evolution of human and experimental seizures. Inspired by the phenomenological model of spontaneous seizure generation from [11], the so-called Epileptor, and mean field dimension-reduction techniques [18], we derive the former equations as introduced in [11] from more biophysically-inspired representations of the system, including single neuron dynamics, linear and non-linear synaptic interactions portraying gap junctions and chemical synapses, respectively. Meanwhile adding this new level of complexity, we keep track of the emergent network behavior exhibited in the abstract model by systematically exploring dynamical changes induced by parameter sweeps in our new system. Our slow environmental variable is more abstract than the biophysically explicit extracellular milieu and oxygenation as considered in [12–14] and involves a wider set of physiological factors such as intra/extracellular pH ratio, oxygen availability, extracellular potassium, calcium, and chloride concentrations, for example. These parameters have been studied experimentally and demonstrated to be part of the influencing agents leading to seizures [3]. Introducing these variables into existing detailed biophysical models has been subject to great consideration for the last decade [19–22]. The abstract integration of such local mechanisms into a slow environmental variable, as presented here, aims at describing the slow sub-system from a more conceptual perspective.
Following this direction, large scale simulations of neural systems become possible at reasonable computational cost, but with sufficient accuracy to treat complex dynamical mechanisms, such as multi-clustered synchronization or entrainment, while implying biologically realistic synapses and firing patterns. We systematically identify synchronization regimes and reproduce potential routes through status epilepticus (by definition, a seizure lasting more than 20 minutes) in our parameter space, providing interpretation of its underlying neuronal mechanisms and further validating our model. We accredit the different parametric regimes experimentally against rodent data recorded in vivo and demonstrate that the different routes in parameter space are consistent with the theoretically predicted topology. Then, exploring a regime that generates spontaneous seizures from background activity, we infer new insights from the role of excitatory and inhibitory synapses, as well as the extracellular environment.
The Epileptor [11] is a five-dimensional model and comprises three different time scales accounting for various electrographic patterns: On the fastest time scale, two state variables (ensemble 1) exhibit bistable dynamics between oscillatory activity modeling fast discharges and a stable node representing interictal activity. On the intermediate time scale, two state variables (ensemble 2) model the spike and wave events (SWE) and form the second neuronal ensemble. On the slowest time scale (order of tens of seconds), the evolution of a very slow permittivity variable guides the neural population through the seizures including seizure onset and offset. The first ensemble is linearly inhibited by the second ensemble in order for fast discharges to occur only during the wave part of the SWE, the second ensemble is excited by the first ensemble through a low-pass filter coupling in order to generate SWE and interictal spikes. Both ensembles are coupled through the permittivity variable. A separatrix divides the state space of the first ensemble between ictal and interictal states and acts as a barrier. As the permittivity variable evolves over time, seizure onset occurs through a saddle-node bifurcation showing a direct current (DC) shift at the transition between the interictal and the ictal state. Seizure offset occurs through a homoclinic bifurcation showing the logarithmic scaling of interspike intervals when approaching seizure offset. The time course of the local field potential is related to the total activity of both ensembles. A more detailed description of the model can be found in Jirsa et al [11], with an extended analysis of its embedded dynamics in [23].
We model the dynamics of brain activity across a set of two populations P1 (excitatory) and P2 (inhibitory) of N neurons each, x1,j and x2,j, respectively (j∈[1,N]). The dynamics of the population Pi is determined by x¯i=1N∑j=1Nxi,j as described by the mean field approximation used for neural masses [24], and the dynamics of membrane potentials xi (i∈[1,2]) is defined by Hindmarsh-Rose (Eq (1.1), (1.2) and (1.3)) and Morris-Lecar (Eq (2.1), (2.2), (2.3), (2.4) and (2.5)) neuronal models for P1 and P2, respectively, as follows. For sake of clarity, the jth index is omitted in the notation. Our choice of the Hindmarsh-Rose model is motivated by the fact that its phase flow is isomorphic to the phase flow of the first Epileptor ensemble [11]. This neuron model is a square-wave burster governed by saddle-node and homoclinic bifurcations, and so is the first Epileptor ensemble [25]. Similarly, the Morris-Lecar model, via its saddle-node-on-invariant-circle (SNIC) bifurcation, captures the same excitable properties as the second Epileptor ensemble.
Population 1 neuron (Hindmarsh-Rose):
Population 2 neuron (Morris-Lecar):
CMV˙=I2−gL(V−EL)−gKn(V−EK)−gCam∞(V)(V−ECa)+σ2CE(x¯2−x2)+Isyn(x¯1,x2)+Isyn(x¯2,x2)−σ2×0.3(z¯−3)+σ2W(t)
(2.1)
n˙=ϕ(n∞(V)−n)/τn(V)
(2.2)
with
m∞(V)=12[1+tanh((V−V1)/V2)]
(2.3)
τn(V)=1/cosh((V−V3)/2V4)
(2.4)
n∞(V)=12[1+tanh((V−V3)/V4)]
(2.5)
and
Coupling term CE and function Isyn are detailed in a next section. σi is a scaling ratio between the two pools of neurons in order to have similar membrane potential amplitudes across different neuron types (i∈[1,2]). Parameters are the same as in previously published studies [5,24,25], unless otherwise mentioned. They are in part enumerated in Table 1 together with their biophysical interpretation, when applicable. W(t) is white Gaussian noise of uniformly distributed values in the interval [-Wmax,Wmax] where Wmax ranges from 2.5 to 20 mV as provided in Table 1, 2.5 mV being used for lowest noise simulations and 20 mV for highest noise simulated traces. Here, all simulations were performed with bounding value of Wmax at 6 mV unless stated otherwise. The other parameters are constant and their ranges are included in Table 1. All values are set at the initialization of the simulation and remain constant for the time of the run.
I1, I2 are baseline input currents of neurons for populations 1 and 2, respectively. x0 captures the equilibrium point of the permittivity z (Eq 1.3) and has previously been referred to as degree of epileptogenicity [26] in the context of the Epileptor. As mentioned above, it corresponds to the mutual effect of a set of factors influencing neural excitability including ATP availability, oxygenation, extracellular potassium concentration etc.
We use symmetric linear difference coupling across membrane voltage equations on fast time scales to model gap junctions, also referred to as electrical synapses. This type of communication, being only electrically conductive through the cell's membrane, is considered nearly instantaneous so its transmission delay is negligible. It is described in the equations by the difference CE(x¯i−xi) where x¯i refers to the mean activity of the neural population i, and CE to the coupling strength. Note that when all neurons from a population are synchronized this difference is zero. Only neurons belonging to the same population are connected to each other via such linear coupling as it is observed in neural tissue that gap junctions usually connect neurons from the same class [27].
Chemically mediated synaptic transmissions follow more complex dynamics induced by intermediate biological mechanisms such as pre-synaptic neurotransmitter release, post-synaptic receptor binding, G-protein activation and so on. Models of such dynamics are typically non-linear and invoke physiologically most relevant parameters, including transmitters’ concentration and release time, conductance strength, or re-uptake time. We used the model described in [5,28], as it provides a sufficiently accurate level of biological description for our study.
The equations are:
Isyn(xi,xj)=−GSi,ju(xj−E)
(3.1)
u˙=αT(1−u)−βu
(3.2)
T=Tmax1+e−(xi−Vt)/Kp
(3.3)
where Isyn is the post-synaptic current, xi and xj are the pre- and post-synaptic neuron activity, respectively. u is an auxiliary variable for the computation of the post-synaptic current, E is the reversal potential, Gsi,j the conductance strength of synapses from neurons of population i to neurons from population j, with (i,j) ∈ [1,2]. α, β the forward and backward binding rate constants with transmitter concentration T in the synaptic cleft, of which the maximum is set by the constant Tmax. Kp gives the steepness and Vt sets the value at which the function is half-activated. We note that dendritic spatial summation corresponds to the process, in which the input xi is averaged over the whole population when sent to the synapse’ function Isyn.
EEG time series of epileptic seizures display a large diversity of temporal characteristics such as interictal spikes, ripples, tonic/clonic discharges etc. It is therefore non-trivial to define a precise measure that delineates these features with sufficient accuracy. To analyze the output of our neuronal populations, we consider an index of synchronization between oscillators: the Kuramoto Order Parameter (KOP) [29]. To visualize the dynamics of the phases of each neuronal oscillator, it is convenient to imagine a swarm of points tracing out the unit circle in the complex plane when action potentials occur. The complex order parameter [30] is the norm of the sum of all vectors between the origin of the unit circle and the points around the circle. It is defined as follow:
The norm r of this macroscopic quantity is near zero when action potentials are uniformly distributed over time, and increases as firings get synchronized. An animated version of these motions is presented in S1 Movie. It can be interpreted as the collective rhythm produced by the whole population. Although it is a convenient method to quantify synchrony, this measure is applicable only when oscillations are present. Neurons at rest (i.e. not oscillating) have been considered separately and labeled explicitly at rest in the analysis. Special care must be taken while interpreting our synchronization measure: we do not consider the synchronization between P1 and P2 but rather the synchronization within P1 (KOP1) and within P2 (KOP2). The overall KOP is calculated as the sum of KOP1 and KOP2.
The code is written in Python with an object-oriented architecture, of which an online version is made available together with its documentation at the Github repository (https://[email protected]/sebnaze/epilepton.git). Simulations were performed on a parallel computing cluster using the Euler-Maruyama integration scheme [31] with step size dt = 0.05.
In this section, we first present how a network of spiking and bursting neurons is derived from the phenomenological Epileptor model, using methods from dynamical systems theory. Then, numerical simulations of the model jointly with experimental data in the context of seizures are exposed. 40 excitatory and 40 inhibitory neurons arranged in a fully synaptically connected network with clustered gap junction communications were used to simulate relevant biophysical mechanisms. With this setup, we explored a set of physiologically pertinent parameters. After comparing simulated data with experimental recordings from animal models of epilepsy, we propose sequential stages that may underlie status epilepticus and validate our model. Further investigating the role of synaptic coupling in a regime generating spontaneous seizures, we make predictions of seizure profiles in the context when synaptic communication is not present.
To build a system dynamically isomorph to the Epileptor model [11], but with a link to biophysical properties, we constructed a network of Hindmarsh-Rose bursting and Morris-Lecar spiking model neurons. Fig 2 illustrates the dynamical similarities between the neuron models and the original Epileptor ensembles, and provides a comparison of their phase space topologies for various levels of excitation. In the Epileptor, the first ensemble shows a saddle-node bifurcation at seizure onset and a homoclinic bifurcation at seizure offset. The Hindmarsh-Rose model is a square-wave burster also governed by saddle-node and homoclinic bifurcations [25] and thus isomorph regarding its phase flow with the first Epileptor ensemble [11] (Fig 2, left column). The spike-wave discharge is modeled in the second Epileptor ensemble by a saddle-node on invariant cycle (SNIC) bifurcation, which is also present in the Morris-Lecar model [25] (Fig 2, right column). When the excitatory and inhibitory spiking neurons are electrically coupled within the populations via gap junctions, then synchronization occurs. By construction, the synchronization manifold is identical to the uncoupled neuron models and thus has the same bifurcations and phase flow topology as the Epileptor ensembles. This is valid precisely for full synchronization and approximately for partial synchronization as a function of the coupling strength. Due to the dynamic isomorphism of single neuron models and the Epileptor ensembles, when full synchronization is approached the mean field of the populations expresses the full dynamic range of behaviors known from the Epileptor model. This will be shown in the next section, in which a parameter space analysis maps out the synchronization behaviors to large detail.
Status epilepticus (SE), i.e. uninterrupted seizure lasting at least 20 min, is a traumatic experience that can transform a non-epileptic brain into a brain displaying spontaneous seizures [32,33]. SE can be induced in rodents by injecting convulsant agents, such as kainic acid or pilocarpine [34,35]. It develops through 3 main phases called impending, established and subtle SE (Fig 3, timeseries II, III and IV, respectively), each of which being part of a continuum of electrophysiological fingerprints [36]. In the following, we retrace the route that neuronal networks follow during SE through a selected set of physiological parameters i.e. the extracellular excitability x0, inter- and intra-population synaptic coupling strengths Gsi,j and Gsi,i (glutamatergic and GABAergic for pyramidal cell and interneuron populations, respectively) and gap junction coupling strength CE (Fig 4). We identify regimes according to the synchronization ratio between neurons within populations (color code), and map it to the characteristic phases observed experimentally (Fig 3 and Fig 4, roman numbers). Since a single metric has not yet been identified to discriminate appropriately the different electrophysiological regimes observed during SE, the color code does not separate the SE states but rather offers a coarse map for orientation amongst the observed dynamical regimes. There are no clear demarcation lines between the dynamic regimes, since the transitions are more gradual than discrete.
After SE, animals experience a latent period during which complex network reorganizations take place. During such period, although neuronal networks exhibit interictal-like activity [38], there are no spontaneous seizures. The latter occur during the chronic phase, a few days or weeks after SE. They are difficult to predict; the brain appears to operate “normally” before an abrupt change happens, characterized by 2 to 10-fold larger amplitude oscillations, which is the seizure. Our model reproduces the most important features of such transitions i.e. an abrupt fast firing discharge pattern at seizure onset, and a decrease of spike-wave frequency towards the end of seizure. We predict interictal spikes and spike-wave discharges are generated from synchronized activity of inhibitory neurons, and are affected by synaptic coupling strengths within and between the two populations of neurons. Fig 6 displays a simulation of about a minute of activity in which a seizure takes place, together with its experimental counterpart. The model produces the different states of seizure evolution without any change of parameters; the states include pre-ictal population spikes, abrupt transitions to tonic firing, and seizure offset. Hysteresis effects have been predicted in the Epileptor [11] and are preserved in the coupled neuronal population dynamics relayed by the slow permittivity variable. As permittivity traces out its trajectory, seizure onset and offset occur at different values of permittivity and the two different neuronal spiking patterns of the populations may co-exist for the same permittivity value. These behaviors are characteristic for hysteresis.
Synaptic coupling is also supposed to play a central role in seizure genesis. However, extracellular Ca2+ concentration nearly drops to zero during seizures, including in primates [39]. In the absence of extracellular Ca2+ seizures can occur in neuronal networks [40], following the general rules of seizure dynamics [11]. When glutamatergic and GABAergic synaptic couplings are removed in the model (GSi,j = 0 in Eq 3.1), we still observe seizure-like events but the temporal features of the signal are slightly different (Fig 7), although hysteresis is maintained. In such conditions, inhibitory populations’ spikes and pyramidal fast discharges influence each other’s excitation through the extracellular environment, thus on slower timescale than spiking discharges. We propose a mechanism around seizure onset: the frequency of inhibitory population spikes increases until the excitatory population starts to fire (onset), and from there, inhibitory neurons reduce their activity to finally stop firing for the remainder of the event. Considering the temporal evolution of the slow variable in the deprived synaptic situation, we note that the inhibitory population’s activity is present when the variable is low. As soon as the excitatory lead takes place, the slow variable rises and quickly the inhibitory population becomes quiescent. This prediction would imply that the action of the excitatory activity on the extracellular medium acts as brake on synchronized inhibitory population’s spikes.
Running systematic simulations over coupling strength parameters, we also observe that seizure duration decreases as we increase inter-population coupling (Fig 8, left). This can be understood as follows: as inter-population coupling increases, the synaptic gain of GABA synapses coming from synchronized spiking of inhibitory cells increases, which reduces the activity in the excitatory cells, eventually leading to the destruction of the self-sustained epileptiform discharges. It is notable that the modification of inter-population coupling results in multi-scale effects on fast and slow time scales and, in particular, significantly affects slower time-scales such as seizure duration.
We finally performed simulations with different noise levels, as noise has a strong effect on neuronal network dynamics [41,42]. Our results support the idea that a noisy environment impacts epileptiform activity. In the context of epilepsy, the degree of stochasticity versus determinism in neural systems has been analyzed, but evidence for one or the other mechanism is still lacking [43]. With our model, low noise intensity in the spontaneous seizure configuration leads to long lasting, sharp onset and offset seizures with long periods between seizure-like activities. Increasing noise intensity results in more frequent but shorter seizure-like discharges (Fig 8, right), with less synchronized ensembles of neurons.
The role of extracellular mechanisms in epilepsy has become more prominent over the last decade [44,45]. Previous computational modeling studies incorporated selected features of extracellular environment (oxygenation, [K+]o…) in their mathematical formulation and investigated mechanisms of transitions between brain states [12–14,20]. Here we have adopted the Epileptor modeling approach of Jirsa and colleagues [11], in which multiple intra/extracellular factors are absorbed in a permittivity variable that influences population neural excitability on a time scale of several seconds. Permittivity may include various factors such as [K+]o but also pH [46], calcium concentration [47,48], ATP resources [49], state of blood vessels [45], and oxygen availability [14,50]. Thus, instead of attempting to incorporate them all in a biophysical description resulting in an unmanageable complexity of simulation parameter tuning, they are lumped into the slow permittivity variable affecting neural excitability.
The link between the phenomenological Epileptor and a network model of coupled spiking neurons allows us to gain a deeper understanding of the robustness of the dynamic structures (in particular its phase flow topology) present in the Epileptor, when being subjected to biophysical constraints such as network connectivity or spike averaging. Gold standard Hodgkin Huxley equations [51] have many parameters and degrees of freedom, hence concerns exist with regard to parameter choices, in particular in light of the non-bijective nature of neuron dynamics (many parameter combinations can give rise to the same dynamics, see for instance [17]). Here we chose the reduced neuron models of Hindmarsh-Rose and Morris-Lecar as an intermediate step, which still allow us to pose questions on population averaging and coupling effects, but within a reduced parameter set and well-defined dynamics. These reduced neuron models [52,53] isolate conceptually the essential mathematical properties of excitation and inhibition from the electrochemical properties of sodium and potassium flows and thus provide a simpler mathematical description. In particular, the mathematical properties of the neuron models resemble those present in the Epileptor, which offers a starting point for the modeling process.
As full synchronization is approached with strong electrical coupling within a neuronal population, we demonstrated in Fig 6 and Fig 7 that seizure onset and offset are triggered in a synergistic manner between fast dynamics, typically neuron spikes, and a slower dynamical environment embedding multiple physiological factors. Uncoupled isolated neurons show, by construction, the same dynamic isomorphism, but then noise will drive the synchronization apart, which is not the case in the electrically coupled population. There, the dynamic range of behaviors survives the noise effects, in particular due to gap junction couplings (see Fig 4 and Fig 8). Electrical coupling via gap junction is modeled here by the difference coupling CE(x¯i−xi) as known from electrical circuit theory. This mathematical representation could also be considered a linear approximation of field effects, which are a special case of ephaptic communication [54], but a rigorous mathematical representation thereof still needs to be developed.
Previous computational studies have highlighted the role of gap junctions for synchronization and epilepsy, in particular for gap junctions located on the axons of glutamatergic neurons [55]. Experimentally, there is evidence for gap junctions only between the dendrites of GABAergic interneurons, but despite numerous electron microscopic studies, there is as yet no ultrastructural evidence of gap junctions between somatodendritic domains of hippocampal pyramidal neurons [27]. Our study suggests that such electrical coupling is in particular responsible for the synchronization of inhibitory neurons, which then in turn give rise to the spike-wave part of interictal spikes. In contrast, most of the current literature, both from computational and experimental studies, supports the hypothesis that interictal discharges arise from synchronization of bursts of pyramidal neurons and interneurons [56–58]. Our results (as hypothesized in [11]), support the mechanism by which the large amplitude discharge of the spike-wave component arises from synchronized discharges of inhibitory interneurons only, and the slower wave component with its coincident high-frequency discharges is generated by excitatory neurons. This high frequency firing is consistent with more detailed computational studies [6,55], however the generation of the spike-wave component is different. Supporting our prediction, GABAergic neurons recorded during spike-wave events stopped firing during the fast discharge, and resumed firing when spike-wave events reoccurred during seizures [11]. Current clamp recordings revealed that they stopped firing during the fast discharge because they entered into depolarization block. There is obviously further experimental work to be performed to disambiguate these effects although some evidence is being accumulated [59]. The theoretical framework presented here should be able to aid in sharpening the investigations. Our scheme does not necessarily oppose previous studies, as the way interictal spikes are generated may vary along the course of epileptogenesis [38].
Several physiological mechanisms have been proposed and investigated to explain pathological activity in the context of epilepsy. Paroxysmal depolarization shifts (PDS) became more prominent as a potential mechanism to explain excitatory neural activity during interictal spikes and seizures [60]. PDS are strong neural depolarization blocks that lead to sustained bursting and maintained membrane depolarization with a very long-lasting return to baseline resting membrane potential (see Figure 4 in [61]). Our simulations support the view of such hyper-excited activity in excitatory ensembles during seizure, in which the permittivity in our model regulates excitability. However, the design of our model regarding inhibitory ensembles does not allow testing whether interneurons follow the same track of depolarization. Such behavior can be observed experimentally when GABAergic transmission becomes depolarizing [62–64], a phenomenon known to happen in early development [65].Future studies using more physiological inhibitory neuron models than our two-dimensional regular spiking Morris-Lecar models could possibly investigate this question more adequately. Another mechanism is cell swelling, which in astrocytes and neurons can affect cellular and network function [66], potentially increasing neural excitability and synchronizing populations of neurons [16,67]. Our results support the hypothesis that electrotonic coupling between neurons via gap junctions is increased in epileptic conditions [37,55]. Extracellular space shrinkage, conveying more surface contact between cells and thus favoring electric communication, has also been demonstrated experimentally in previous studies [68] and is compatible with our results concerning the involvement of permittivity and gap junction coupling in seizure genesis and evolution. Cell swelling together with extracellular alkalization or intracellular acidification [56], induced by trans-membrane movement of ions and water associated with bursting activity during paroxysmal depolarization shifts, also increases field interactions through exogenous communication [54]. Permittivity comprises such field effects as part of the slow variable z, but their role and relevance for neural activity is so far unknown. Our simulations in deprived synaptic conditions argue towards the idea that strong pyramidal cell activity would reduce interneuron activation by such coupling via the extracellular space. This “extracellular environmental disinhibition” may be negligible as compared to synaptic strengths in standard conditions, nevertheless leading to spike-waves with embedded fast discharges. It is important to note that the extracellular concentration of calcium drops to the 100 μM range in vivo during seizures in primates [39]. At this concentration, synaptic transmission is largely compromised [69], yet spike waves and fast discharges still occur, supporting the above-mentioned mechanism. Surprisingly, using our model with high noise level, the variance of this slow extracellular fluctuation variable is reduced. This result brings new possible insights about the mechanisms of seizure occurrence frequency shifts on long time scale in some types of epilepsy. For example, time between consecutive seizures often increase in patients enduring frequent seizures in childhood [70]. Following our results, this condition may be due to very slow decrease of background noise level (in the order of months and years) during development [71,72], as empirical evidence states that signal variability decreases with age [73].
Approaches using dynamical system theory may provide guidance when deriving biophysical models of brain function and dysfunction. Here we used the abstract Epileptor model to guide the design of population models of coupled excitatory and inhibitory spiking neurons, which now allows studying the organization of spiking patterns as a function of coupling and excitability. We demonstrated that gap junction coupling plays the dominant role in synchronizing both neuron types, whereas the slow permittivity changes act rather as a slowly changing control parameter aiding in organizing the seizure progression. Our simulations support that the large amplitude discharge of spike-wave components of interictal and ictal population spikes arises from synchronized discharges of inhibitory interneurons only, which is not in line with current thinking in the literature, though finds support by recent empirical studies. Our approach does not rule out other physiological organizations giving rise to the same dynamics as described in the Epileptor; in fact, there is a large range of candidates. Thus although our findings may have validity only within a small range of physiological realizations, they nevertheless can give insight about experimental paradigms via simulations and analyses bridging the gap between neuronal spiking, network and abstract seizure evolution across large temporal scales.
|
10.1371/journal.ppat.1005077 | Transgenic Rabbits Expressing Ovine PrP Are Susceptible to Scrapie | Transmissible spongiform encephalopathies (TSEs) are a group of neurodegenerative diseases affecting a wide range of mammalian species. They are caused by prions, a proteinaceous pathogen essentially composed of PrPSc, an abnormal isoform of the host encoded cellular prion protein PrPC. Constrained steric interactions between PrPSc and PrPC are thought to provide prions with species specificity, and to control cross-species transmission into other host populations, including humans. Transgenetic expression of foreign PrP genes has been successfully and widely used to overcome the recognized resistance of mouse to foreign TSE sources. Rabbit is one of the species that exhibit a pronounced resistance to TSEs. Most attempts to infect experimentally rabbit have failed, except after inoculation with cell-free generated rabbit prions. To gain insights on the molecular determinants of the relative resistance of rabbits to prions, we generated transgenic rabbits expressing the susceptible V136R154Q171 allele of the ovine PRNP gene on a rabbit wild type PRNP New Zealand background and assessed their experimental susceptibility to scrapie prions. All transgenic animals developed a typical TSE 6–8 months after intracerebral inoculation, whereas wild type rabbits remained healthy more than 700 days after inoculation. Despite the endogenous presence of rabbit PrPC, only ovine PrPSc was detectable in the brains of diseased animals. Collectively these data indicate that the low susceptibility of rabbits to prion infection is not enciphered within their non-PrP genetic background.
| Prions are infectious pathogens causing irremediably fatal neurodegenerative diseases in human and in farmed or wild animals. They are formed from abnormally folded assemblies (PrPSc) of the host-encoded prion protein (PrPC). Different PrPSc conformational variants or ‘strains’ can propagate in the same host, giving distinct biological phenotypes. Like other pathogens, prions can transmit from one species to another, representing a zoonotic risk. A barrier, commonly referred to as “species barrier”, limits prion cross-species transmission. This barrier is supposed to reflect the steric incompatibility between invading prion PrPSc and PrPC of the infected host. Rabbit is one of the species that exhibit a pronounced resistance to prions. To gain insights on the molecular determinants of the relative resistance of this species to prions, we generated transgenic rabbits expressing sheep PrPC and assessed their experimental susceptibility to sheep scrapie prions, as routinely done with transgenic mouse models of prion disease. All transgenic rabbits developed a typical prion disease within 200 days, whereas wild type rabbits remained healthy more than 700 days after inoculation. These data indicate that the low susceptibility of rabbits to prion infection is not enciphered within their non-PrP genetic background.
| Scrapie in small ruminants, bovine spongiform encephalopathy (BSE) in cattle, chronic wasting disease (CWD) in cervids, transmissible mink encephalopathy (TME) or Creutzfeldt-Jakob disease (CJD) in humans belong to a group of fatal neurodegenerative diseases referred to as transmissible spongiform encephalopathies (TSEs) or prion diseases [1]. TSEs are characterized by the accumulation in the central nervous system (CNS) of a ß-sheet enriched, protease-resistant and aggregated isoform (PrPSc) of the host encoded cellular prion protein (PrPC). During TSE pathogenesis, PrPSc seeds, acquired through infection or arising from spontaneous conversion of wild type or mutant PrPC, are believed to template the conformational change of host PrPC to nascent PrPSc forms. This autocatalytic polymerization process leads to deposition of injurious deposits into the brain. PrPSc particles are thought to be the major if not the sole component of TSEs infectious agent or prion [2]. Distinct strains of prions are recognized phenotypically in a given host species. They cause TSEs with specific phenotypic traits, including time course to disease, neuropathological features and PrPSc biochemical properties. There is compelling evidence that prion strain diversity reflects stable differences in PrPSc conformations, at the level of the tertiary and/or quaternary structure [3–5].
A wide range of mammals like ruminants, pigs, rodents, carnivores or primates can be naturally and/or experimentally infected with prions. Prions are usually easy to transmit between individuals of the same species. Prions can also transmit between species, as exemplified by the emergence of variant CJD, following dietary exposure of humans to BSE prions. However, such events are restricted by a so-called ‘species’ or ‘transmission’ barrier, the strength of which depends essentially on interactions between host PrPC and the infecting prion strain type(s) [3, 4]. The force of the transmission barrier is classically gauged by the appearance of disease-specific, clinical signs and/or PrPSc in the brain and, sometimes, extraneural tissues of the new host. Their concomitant absence would usually suggest a resistance to infection or a disease incubation time exceeding that of the exposed host life span [6–9]. Rabbits have long been recognized as one of the best examples of a species refractory to TSEs agents as attempts to transmit Kuru, CJD, sheep scrapie, TME isolates and mouse adapted scrapie strains repeatedly failed [6, 10]. In marked contrast, cell-free PrPSc-templated conversion of PrPC by assays such as the protein misfolding cyclic amplification (PMCA, [11]) revealed that rabbit PrPC was fully convertible into rabbit PrPSc by seeds from scrapie infected, BSE and mouse scrapie brain sources ([12–14]). Yet, the amplified agents propagated with limited success in rabbits [12], leading support for the hypothesis that rabbit may barely develop clinical TSEs.
Numerous transgenetic studies in mice have demonstrated that mouse resistance to foreign prions can be abrogated by introducing in the murine genome the corresponding PRNP gene (the gene encoding PrPC). Transgenes can be introduced in animals knocked out for their own Prnp gene or on a murine wild-type background (reviews: [15, 16]). Vidal et al. demonstrates that transgenic mice expressing rabbit PrP are permissive to a broad panel of TSE sources from different species [14], strongly advocating for the full convertibility of rabbit PrPC into disease-associated isoforms. While transgenesis on a murine Prnp knockout background usually abolishes the transmission barrier, co-expression of the transgene with the wild-type Prnp gene can interfere with prion replication [17–19]. In sheep, different PRNP alleles tightly modulate the incidence and pathogenesis of classical scrapie. Three codons (for amino acids at positions 136, 154, and 171) act as major determinants, the V136R154Q171 allele conferring the highest susceptibility [20–22]. Experimental transmission of scrapie prions was markedly improved in transgenic mice overexpressing this allele [3, 23].
In this study, we report that rabbits engineered to express transgenetically the V136R154Q171 ovine PRNP allele develop a TSE syndrome upon experimental inoculation with scrapie. Disease occurred at full attack rate, with no apparent interfering effect of the rabbit wild-type PrPC.
All the experiments involving animals were done in strict accordance with the European Community Council Directive 86/609/EEC. The French committee for GMO,-formerly the governmental "Commission de Génie Génétique" that belonged to the French Ministry for Education, Research and Technology-, has approved the creation and experimental infection with prions of the rabbit transgenic animal model. The agreement number is 2992-II. It was issued the 12th of November 1998.
A fragment of the sheep genome containing the V136R154Q171 allele of the Prnp gene was cloned in a bacterial artificial chromosome (BAC) vector previously used to generate the tg338 mouse line [23, 24]. This 125 kb insert, including 40 kb upstream and 60 kb downstream of the PRNP gene, was purified following a Not1 digestion and microinjected into New Zealand White rabbit embryos. Approximately 700 hundred embryos (from 41 donor females) were microinjected (mostly in the male pronuclei). 600 embryos were implanted into 30 pseudo-pregnant females. There were 28 live births, of which one was transgenic, as identified by PCR. Rabbit DNA was extracted by digesting ear fragments using proteinase K-SDS method. Digestion was followed by a DNA precipitation by ethanol. The two primers used to identify the sheep PRNP gene were TAGGCAGTTGGATCCTGGTT and CCCTATCCTACTATGAGAAA. The primers used to identify the control endogenous αS1-casein gene were CACTCCCTTGTTGAAAACTCTCCTCAG and ATTTTGTGGTTTCAGATCAACCAATAGG.
To estimate the concentration of ovine and rabbit PRNP RNA in ovine PrP transgenic rabbits (TgOv), RT-qPCR analyses were performed on 3 animals per line, as previously described [25]. Two micrograms of total RNA (extracted from brain tissue) were reverse-transcribed with random adapters, following the manufacturer’s instructions [25]. Quantification was achieved by SYBR Green quantitative PCR (Applied Biosystems) using sets of primers specific to the PRNP ovine and rabbit sequences, resulting in the amplification of 100 bp long fragments, respectively. Two sets of primers were used for rabbit PRNP, owing the existence of two variants (with or without exon 2). The sequences of the specific primers were as follows: forward primer: 5’- TCATGGTGAAAAGCCACATAGG-3’, reverse primer: 5’- CCTCCGCCAGGTTTTGGT-3’ for ovine PRNP; forward primer: 5’-TCCTCTCGGCAGCTGTCAT-3’, reverse primer: 5’-GCTTCGGCCGCTTCTTG-3’ for Rabbit PRNP without exon 2; forward primer: 5’-AGAGGCCCCAGTCCAGTGTA-3’; reverse primer: 5’-CACTCCACGTGGCCACAA-3’ for Rabbit PRNP with exon 2. These primers were chosen so that they are located in different exons. The primers for β-actin were as follows: forward: 5’- CCGCATGCAGAAGGAGATCA-3’; reverse: 5’-AGAGCGAGGCCAGGATGGA-3’. For each sample, RNA concentration normalization was achieved using RT-qPCR on β-actin, as previously described [24]. It is thus given by the formula 2(CtPRNP-Ctβ-actin).
Eighteen tgOv rabbits (9 males and 9 females) and 6 wild type (WT) rabbits (3 males and 3 females) aged 103 to 132 days (all bred at INRA Nouzilly) were selected for the experiment. Among them, twelve individually identified tgOv animals and six WT rabbits were injected with 50 μL of a 1% (wt/vol in 5% glucose) brain homogenate from tg338 mice infected with the LA21K fast strain. As controls, six tgOv animals were injected similarly with 50 μL of a 10% brain homogenate from a healthy tg338 mouse.
The inocula were prepared in a class II microbiological cabinet using disposable equipment, with strict safety rules, and immediately inoculated to the animals by the intracerebral route, at the level of the right parietal cortex (depth: 1 cm). Inoculations were performed under general anesthesia by injecting a mixture of xylazine (Rompun, Bayer, France) and ketamine (Imalgene, Merial, France) by the intramuscular route. All inoculations were carried out in compliance with ethics and animal welfare according to regulation requirements. The LA21K fast strain has been obtained through serial transmission and biological cloning by limiting dilutions of the Langlade field scrapie isolate (INRA Toulouse [26]) to tg338 mice. The LA21K fast infectious titer is 109.4 50% lethal doses (LD50)/g of tg338 brain [26].
Rabbits were housed in individual cages, in a dedicated biosafety level-3 facility. They were monitored using a video system and monitored daily for clinical signs and food consumption, by different investigators. Any death arising during the experiment was recorded and animals were necropsied. At clinical stage, rabbits were sacrificed by carbon dioxide suffocation and autopsied for brain and spleen collection. One half of each brain and spleen was stored at –20°C for immunoblotting and biochemical analyses, while the other half of organs was fixed in neutral-buffered 4% formalin for 1 week before paraffin-embedding for immunohistochemistry and histology.
Brains and spleens were analyzed for the presence of either PrPC or PrPSc by Western blotting as previously described [27]. Briefly, homogenates were prepared with a tissue homogenizer (Precellys, Bertin Technologies, France) in a 5% glucose solution. PrP was purified and concentrated using the Bio-Rad TeSeE purification kit. When needed, suspensions were treated with 100 μg/mL of proteinase K (Roche diagnostics, Germany) for 30 min at 37°C and the final pellet was suspended in Laemmli buffer. Samples were denatured at 100°C for 10 minutes, centrifuged at 20 000 g for 5 minutes and the supernatants were run on 12% SDS-PAGE gels. When two-dimensional gel electrophoresis was performed, as previously described [28], the second dimension was run on a 6–16% linear gradient SDS-PAGE. After transfer onto nitrocellulose membranes, samples were probed with Sha31 anti-PrP monoclonal antibody (generous gift of J. Grassi, CEA, Saclay, France, [29]), which binds to the YEDRYYRE amino acid sequence of the PrP protein (amino acid (aa) residues 146 to 153 of the rabbit PrP, Fig 1). This step was followed by the addition of a horseradish peroxidase-conjugate. Peroxidase activity was revealed using a chemiluminescent substrate (SuperSignal West Dura, Pierce, USA), and the signals were captured with a digital imager (Fluorchem 8900, Alpha Innotech, USA) or GeneGnome digital imager (Syngene, Frederick, Maryland, United States). The PrP levels and glycoforms ratios were quantified with the GeneTools software.
Expression levels of PrPC in brains and spleens of rabbits and sheep were compared. Immunoblots revealed with the mAb Sha31 were quantified using the Alpha-Ease software (Alpha-Innotech, USA). Brain samples were diluted and PrP amounts were calculated according to a dilution curve of full-length recombinant ovine PrP (a generous gift of D. Marc, INRA, Tours, France), used as a reference.
Rabbit brain sections (5 μm thick) were treated as described previously [30]. Briefly, after being deparaffinised and rehydrated, tissue sections were incubated in 98% formic acid (MERCK) for 30 min at room temperature, then autoclaved for 15 min at 121°C in 10 mM citrate buffer (pH 6.1) and allowed to cool for 20 min. Sections were then subjected to a 15 min proteolysis at 37°C with 20 μg/mL of proteinase K. Endogenous peroxidase was inhibited using 0.3% hydrogen peroxide in methyl alcohol for 30 min at room temperature. Immunostaining was performed on a DAKO Autostainer according to the manufacturer’s instructions using the mouse monoclonal anti-PrP antibody 2G11 (gift from J. Grosclaude, INRA, Jouy-en-Josas, France, [30]) as primary antibody, followed by the DAKO EnVision+ System labelled polymer-HRP anti-mouse with 3, 3’-diaminobenzidine (DAB) as chromogen. After immunostaining, sections were counterstained with Mayer’s haematoxylin and cover-slipped. The 2G11 antibody was selected for optimal results with TSE-infected-tgOv rabbit brains without any background staining on tissue sections from both uninfected-tgOv and–WT rabbits.
PET blot were performed using a method previously described [31]. Immunodetection was performed with Sha31 monoclonal antibody (4 μg/mL), followed by application of an alkaline phosphatase labeled secondary antibody (Dako, 1/500 final dilution). Enzymatic activity was revealed using NBT/BCIP substrate chromogen.
Paraffin-embedded sections were mounted on glass microscope slides and stained with hematoxylin and eosin.
The equivalent of 6 mL of 20% brain homogenate from tgOv and WT rabbit, treated or not with proteinase K and denatured (see above) were resuspended in 100 mM NaCl, 10 mM EDTA, 10 mM TrisHCl pH 7.8, 0.5% DOC, 0.5% Igepal (Sigma). Monoclonal antibody Sha31 coupled to magnetic beads (Dynabeads M-280 Tosylactivated, Dynal) was added at the rate of 100 μL beads/240 mg tissue equivalent and reacted for 2 hours at 37°C. Beads were collected with a magnet, washed twice with PBS buffer and denatured in Laemmli buffer (10 min at 99°C). Samples were loaded on a 12% acrylamide gel, before electrophoresis and either immunoblotted (60 mg of tissue equivalent) or silver stained (1140 mg of tissue equivalent). The silver stained gels were compared to the western blot and protein bands that were at the same molecular masses than the PrPSc reactive bands were cut, rinsed and then reduced with dithiothreitol and alkylated with iodoacetamide. Samples were incubated overnight at 37°C with 12.5ng/μl trypsin (sequencing grade, Roche, Meylan, France) in 25 mM NH4HCO3 [28]. Tryptic peptides were analyzed by nanoLC-MS/MS with Q-q-TOF and Linear Ion trap.
For CapLC system coupled to Q-TOF Ultima Global (Waters Micromass, Manchester), the digested peptides were loaded on a precolumn (300μm i.d x 5mm, packed with C18 PepMap, LC Packings, Dionex) and desalted. Peptide separations were conducted on a C18 column (Atlantis dC18, 75mm I.D x 150 mm Nano Ease, Waters). Peptides were eluted with a 5–60% linear gradient with water/acetonitrile 98/2 (v/v) containing 0.1% formic acid in buffer A and water/acetonitrile 20/80 (v/v) containing 0.1% formic acid inbuffer B. Mass data were acquired using one MS survey followed by MS/MS scans on the 3 most intense ions detected. Data were processed using ProteinLynx Global server 2.2. The peptide and fragment masses were matched in database (nrNCBI) using MASCOT software (http://www.matrixscience.com). The mass tolerance was 0.2 Da for both precursor and fragment ions.
For Ettan MDLC system (GE Healthcare, Germany) coupled to LTQ Linear Ion Trap Mass Spectrometer (Thermo Electron, US), each sample was desalted using Zorbax 300-SB C18 trap column, 300μm i.d x 5 mm (Agilent Technologies, Germany). Peptide separations were conducted on a Zorbax 300-SB C18 column, 75 μm i.d x 150 mm (Agilent Technologies, Germany). Buffer A consisted of water with 0.1% formic acid while buffer B was 84% acetonitrile with 0.1% formic acid. Separation was performed by applying gradient of 15–55% B for 60 minutes at a flow rate of 400 nL/min. Mass data were acquired using one MS survey (m/z 500–2000) followed by MS/MS scans on the 3 most intense ions detected using Collision Induced Dissociation fragmentation mode. Identification was then performed with Bioworks 3.2 (Thermo Finnigan, San Jose, CA) software. MS/MS spectra were searched against the non-redundant Uniprot database (2006_12) and analysed using TurboSEQUEST (Thermo Finnigan, San Jose, CA). Search parameters included differential amino acid mass shifts for oxidized methionine (+16 Da) and carbamidomethylation on cystein (+57 Da). The output data were evaluated in term of Xcorr magnitude up to 1.7, 2.2 and 3.5 for charge states 1+, 2+ and 3+, respectively.
Confluent Rov cells (P2FJ6 clone, [26, 32]) were grown for 2 days in single wells of 12-well plates. Rov cells were incubated in culture medium containing 10 μL of 20% brain homogenate from tgOv rabbits, WT rabbits and tg338 mice challenged with LA21K fast prions, and from uninfected tgOv rabbits. After 2 days, the medium was removed; the cells were rinsed in phosphate-buffered saline (PBS) and split into 25 cm2 flasks. Each week, one flask was used for subpassaging, whereas another was used to prepare a cell lysate for PrP content analysis (see above). The total protein content was estimated by using a protein assay kit (bicinchoninic acid assay (BCA); Pierce).
The Swiss-Prot accession numbers for the proteins mentioned in the text are sheep (P23907) and rabbit PrP (Q95211).
One transgenic rabbit founder animal was obtained following microinjection of an ovine BAC DNA insert encompassing the entire PRNP transcription unit. This insert has already been used to produce various mouse transgenic lines that express the PrPVRQ allele [23]. For animal production, the tgOv transgenic founder was mated with a WT rabbit. Transmission of the transgene was of about 50% indicating that the rabbit founder was not a mosaic. The transgenic F1 rabbits were mated with WT rabbits giving birth to 39 offspring including 50% of heterozygous transgenic rabbits and 50% of non-transgenic control rabbits. Health of the rabbits did not appear to be affected by the presence of the ovine PRNP transgene (period of observation > 700 days).
The concentration of ovine PRNP RNA relative to that of rabbit PRNP RNA was estimated by RT-qPCR analyses on brain tissue extracts of tgOv rabbits and WT rabbits. In WT and tgOv rabbits, two rabbit PRNP RNA variants were found, as expected, resulting from the splicing (or not) of exon 2. The values were cumulated to obtain the total concentration of rabbit PRNP transcripts. In tgOv rabbits, ovine PRNP RNA levels were 1.5–2 fold higher than those of rabbit PRNP RNA (n = 3 rabbits analyzed). Rabbit PRNP RNA levels were similar between WT and TgOv rabbits, suggesting that expression of ovine PRNP has no interfering effects on the transcription of rabbit PRNP.
In the absence of rabbit-specific anti-PrP antibody, it was not possible to quantify the relative expression levels of ovine versus endogenous rabbit PrPC in tgOv rabbits. Immunoblots analyses indicate that the total levels of PrPC in brain of tgOv rabbits were approx. 1.5–2 fold higher than those found in the brain of their WT counterparts or sheep carrying the VRQ allele (Fig 2A), in agreement with the transcriptional analysis. TgOv rabbits expressed about 50 fold more PrPC in the brain than in the spleen, which compares with the ratio of about 60 found in WT rabbit.
Two-dimensional gel electrophoresis was performed to determine whether any change in the isoforms pattern of PrPC was visible between WT and tgOv rabbits. Equivalent amounts of brain extracts were separated and transferred to nitrocellulose and probed with the Sha31 antibody (Fig 2B). Both extracts gave similar 2D patterns and none of the isoforms appeared specific or quantitatively different between the two types of rabbits.
TgOv and WT rabbits were intracerebrally inoculated with LA21K fast scrapie strain, a fast strain that kills tg338 mice in less than 2 months [26].This agent induced a neurological disease in all inoculated tgOv rabbits (n = 12). The behavioral and clinical signs were invariant from animal to animal. The first behavioral signs were a drop in food consumption and restlessness. Early neurological signs (referred to as moderate in the S1 Video) were characterized by amaurosis and decreased time of random exploration in the cage. With disease progression, more severe neurological and behavioral signs were progressively observed, including loss of balance, disordered gait, paparesis, drop in food consumption and more severe amaurosis, (S2 Video). Animals were euthanized as soon as at least 3 of these signs were recorded. Neither pruritus nor tremor was observed. The clinical phase lasted less than 2 weeks. Animals were euthanized between 161 and 239 days post-inoculation (mean ± SEM incubation time: 192 ± 8 days; Fig 3). None of the mock-infected tgOv rabbits (n = 6) or LA21K fast-challenged WT rabbits (n = 6) presented any clinical signs during the time course of the experiment (Fig 3, S3 Video). They were euthanized healthy at 701 days post-inoculation, except two animals from each control group that were sacrificed, for comparison purpose with scrapie-sick animals.
The brains of diseased tgOv animals were analyzed by immunoblotting and immunohistochemistry for the presence of PrPSc. Proteinase-K resistant PrPSc (PrPres) was readily detected in all infected tgOv rabbits (Figs 4 and 5), consistent with the efficient transmission. All controls remained PrPres negative (Fig 4A and 4B). In the absence of PK-treatment, PrPSc was essentially detected as full-length PrPSc (Fig 4A) as in tg338 brain [33], suggesting absence of endogenous cleavage generating the so-called C2 fragment [33]. Remarkably, LA21K fast electrophoretic pattern was conserved in tgOv rabbits with regards to apparent molecular mass and relative proportions of glycoforms (Fig 4B and 4C). None of the animals of the experiment, including positive transgenic rabbits, showed any detectable PrPres deposits in the spleen (Fig 4B). LA21K fast, as other fast ovine strains is lymphotropic in tg338 mice ([34], Fig 4B). It is likely that disease after intracerebral inoculation has occurred too rapidly in the rabbits to allow centrifugal spreading and replication of LA21K fast scrapie prions in the spleen.
PrPres distribution in the brain was examined by PET blot analyses and immunohistochemistry. The greatest levels of PrPres deposits were observed in the thalamus, hippocampus and frontal cortex of scrapie-sick TgOv rabbits (Fig 5A and 5B). Sparse or moderate PrPSc deposition was seen in the cerebellum, obex and medulla oblongata (Table 1, analysis of 9 animals). Aged, uninfected tgOv rabbits and LA21K fast-challenged WT rabbits remained PrPres negative (Fig 5C and 5D).
Examination of histopathologic lesions in several brain areas of scrapie-sick TgOv rabbits (5 animals analyzed) indicated that spongiosis was prominent in the thalamus (Fig 5E). Mild spongiosis was also observed in the hippocampus (Fig 5F). Sparse spongiosis was observed in the medulla oblongata, obex and cerebellum (Table 1). The cortex was not vacuolated. There was no evidence of vacuolation in aged, uninfected tgOv rabbits and LA21K fast-challenged WT rabbits (Fig 5H–5J).
Collectively, these data indicate that scrapie infected tgOv rabbits exhibited the major clinical, biochemical and neuropathological hallmarks of TSEs.
To determine which of the rabbit or the ovine PrPC had been converted in scrapie-sick tgOv rabbits, immunoprecipitated brain extracts of healthy and infected tgOv rabbits,—treated or not with PK-, were analyzed by two techniques of mass spectrometry. Both analyses allowed detecting PrP fragments in the different gel bands corresponding to the western blot immunoreactive bands. In PK-treated, LA21K fast infected tgOv rabbit brains, five fragments were identified: ESQAYYQR; GENFTETDIK; VVEQMCITQYQR; GENFTETDIKIMER; EHTVTTTTKGENFTETDIK. Only one fragment VVEQMCITQYQR was identified in mock-infected tgOv rabbit, while no fragments were obtained from WT rabbits. All these fragments were located within the C-terminal part of the PrP protein and two could be assigned without ambiguities to the sheep PrP sequence (ESQAYYQR; VVEQMCITQYQR; Fig 1). Thus, PrPres molecules that accumulate in the brain of scrapie-sick tgOv rabbits was essentially of ovine origin, suggesting limited conversion, if any of endogenous rabbit PrPC during disease pathogenesis.
LA21K fast scrapie prions can be efficiently passaged in Rov cells expressing ovine PrP [26, 32, 35]. We examined whether the prions produced in the brain of LA21K fast-sick tgOv rabbits would infect Rov cells with similar efficacy. Rov cells were exposed to similar amounts of brain homogenate from LA21K fast-infected tgOv rabbit and tg338 mice, and grown for up to 4 passages. At each passage, PrPres accumulation was monitored to assess the success of the infection and compare the levels of protein produced. In parallel, cells were exposed to brain extracts from aged, uninfected tgOv rabbits and from LA21K fast-inoculated WT rabbits. While immunoblots analyses of PK-digested cell lysates failed to detect PrPres in these controls, cells exposed to LA21K fast prions from either tgOv rabbit or tg338 mouse origin accumulated similar levels of PrPres at each passage (Fig 6). LA21K fast prions derived from tgOv rabbits and tg338 mice exhibit therefore similar efficacy to infect Rov cells. These data would further sustain the view that the prions produced in the brains of tgOv rabbits are of ovine origin.
The limited number of prion-permissive cell models, the prolonged incubation time in farm species and the low susceptibility of conventional mouse lines to TSEs agents have favored development of mouse transgenesis in the TSE field. Some of the Prnp0/0 mouse lines used to demonstrate the key role of PrPC in susceptibility to prions [36] were eventually engineered to (over)express PRNP genes from a wide range of mammalian species. These models considerably improved our knowledge on prion diversity, and most particularly on the molecular determinants of the transmission barrier during interspecies prion transmission [3]. A side effect of these studies was that transmission barriers that were considered as strong or ‘absolute’ were essentially abrogated [19, 23, 37, 38]. Here, we applied the same strategy to the rabbit species to clarify the respective roles of rabbit PrPC and non-PrP host factors in their pronounced, albeit not absolute [12], resistance to TSEs. We show that transgenic rabbits expressing a scrapie-susceptible ovine PRNP allele develop, at full attack rate, classical hallmarks of TSEs upon inoculation with scrapie prions, including fatal neurological diseases, clinical signs, PrPres deposition and vacuolation in the brain. This demonstrates that rabbits do not bear non-PrP factors that make them intrinsically resistant to prions.
The strategy used to generate the tg338 mouse line, a model highly susceptible to sheep scrapie sources [3, 23, 26, 34], was transposed to rabbit. At variance with the tg338 mouse line, transgenesis was performed on a wild type rabbit background, as Prnp0/0 rabbits are not available, leading to the likely expression of both rabbit and sheep PrP proteins. Analyses of the PRNP transcripts indicated that ovine PRNP transcripts were present in tgOv rabbit brain at levels 1.5–2 fold higher than those of rabbit PRNP. Consistently, expression level of total PrPC in tgOv rabbit brains was 1.5–2 fold higher than that in WT rabbit and sheep. Co-expression of two different prion proteins can have a strong inhibitory effect on the conversion into prions of the transgenic PrPC protein, resulting in either no transmission, or a marked increase of the incubation time or no clinical disease [17–19, 39]. Here, scrapie developed at full attack rate in tgOv rabbits. The molecular, LA21K fast strain-specific [3] signature was conserved upon passage to another species expressing the same transgene in a different genetic background. The potential of LA21K fast prions to infect Rov cells [26, 32] was unaltered by the intermediate passage onto TgOv rabbits. Collectively, these data support the view that transgenic expression of the ovine PrPVRQ allele abrogated the rabbit species barrier to LA21K fast scrapie prions. Demonstrating formally that the species barrier has been fully abrogated and/or that co-expression of rabbit and sheep PrPC had no major interfering effects on scrapie pathogenesis would necessitate further subpassaging on tgOv rabbits to measure a potential reduction in incubation time, if any [3]. This experiment has not been done. However, it can be noticed that the mean incubation time observed at primary passage in tgOv rabbits (<200 days) is within the range of incubation time observed in sheep and transgenic mice expressing physiological levels of ovine PrP upon intracerebral infection, at the same dose, of fast scrapie prions ([23, 30]; 140 ± 5 days (6/6) in tg335 mice [23]).
While scrapie-infected tgOv rabbits showed characteristic signs of prion diseases, their WT counterparts remained healthy for more than 700 days, with neither detectable signs nor lesions or PrPSc deposits. This confirmed with another natural TSEs source the pronounced resistance of the rabbit species to foreign prions [6, 10]. We also showed by mass spectrometry that most if not all of the converted PrPC molecules in the brains of scrapie-sick tgOv rabbits were of ovine origin, indicating that conversion of rabbit PrPC was not favored by the conversion of ovine PrPC at vicinity. Collectively, these data suggest that rabbit PrPC is poorly convertible into LA21K fast PrPSc in vivo. Consistently, cell-free conversion of rabbit PrPC by SSBP/1 prions,-which exhibit a LA21K fast phenotype in tg338 mice (mean survival time of 63 ± 1 days in 6/6 tg338 mice on primary passage)-, necessitated a relatively high number of PMCA rounds as compared to other prion sources [12]. To further address the issue of the presence of rabbit PrPSc in the brain of diseased tgOv rabbits, secondary passage to transgenic mice expressing rabbit PrPC are planned ([12] and the accompanying paper by Vidal et al.).
The common shared view that rabbits were resistant to prion infection was not only attributed to rabbit PrPC sequence but also to its genetic background [12, 40–42]. Vidal and Castilla’s groups demonstrate by using transgenic modeling that rabbit PrPC, as many other mammalian PrPC, is fully convertible into disease-specific isoforms after infection with a broad panel of TSE sources [14]. Thus, taken separately, rabbit genetic background and rabbit PrPC cannot explain the apparently low susceptibility of rabbits to prion infection. What makes the rabbit species comparatively resistant to prion disease remains to be clarified. On the one hand, the diversity of prion sources inoculated to this species has remained too limited [6, 10] to definitely conclude that rabbits are poor acceptor for prions. The difficulty in identifying TSEs agents able to replicate on certain PrP sequence has been recently exemplified by studies on scrapie prions zoonotic potential. Such evidence (contrarily to a common- shared view) was provided because a panel of diverse scrapie sources was inoculated to human PrP transgenic mice [43]. On the other hand, it is possible that prion disease in rabbit would develop too slowly to be observed, because of a low conversion rate of rabbit PrPC. Transgenic modeling with mice expressing PrP at variable levels may help to verify this hypothesis [23, 44].
To summarize, we found that rabbit expressing ovine PrP at near physiological levels can develop a bona fide TSE upon infection with scrapie prions. The low susceptibility of rabbits to prion infection is not enciphered within the rabbit genetic background. Owing to its sensitivity and intermediate size, this model may be a valuable tool for studying TSE pathogenesis, most notably prionemia.
|
10.1371/journal.pgen.1007047 | Alx4 relays sequential FGF signaling to induce lacrimal gland morphogenesis | The sequential use of signaling pathways is essential for the guidance of pluripotent progenitors into diverse cell fates. Here, we show that Shp2 exclusively mediates FGF but not PDGF signaling in the neural crest to control lacrimal gland development. In addition to preventing p53-independent apoptosis and promoting the migration of Sox10-expressing neural crests, Shp2 is also required for expression of the homeodomain transcription factor Alx4, which directly controls Fgf10 expression in the periocular mesenchyme that is necessary for lacrimal gland induction. We show that Alx4 binds an Fgf10 intronic element conserved in terrestrial but not aquatic animals, underlying the evolutionary emergence of the lacrimal gland system in response to an airy environment. Inactivation of ALX4/Alx4 causes lacrimal gland aplasia in both human and mouse. These results reveal a key role of Alx4 in mediating FGF-Shp2-FGF signaling in the neural crest for lacrimal gland development.
| The dry eye disease caused by lacrimal gland dysgenesis is one of the most common ocular ailments. In this study, we show that Shp2 mediates the sequential use of FGF signaling in lacrimal gland development. Our study identifies Alx4 as a novel target of Shp2 signaling and a causal gene for lacrimal gland aplasia in humans. Given this result, there may also be a potential role for Alx4 in guiding pluripotent stem cells to produce lacrimal gland tissue. Finally, our data reveals an Alx4-Fgf10 regulatory unit broadly conserved in the diverse array of terrestrial animals from humans to reptiles, but not in aquatic animals such as amphibians and fish, which sheds light on how the lacrimal gland arose as an evolutionary innovation of terrestrial animals to adapt to their newfound exposure to an airy environment.
| The lacrimal gland plays an essential role in protecting the ocular surface by secreting the aqueous components of the tear film. Defects associated with the lacrimal gland are the main cause of dry eye disease, which is highly prevalent in the geriatric population [1]. Left untreated, dry eye disease may progress from eye irritation and corneal scarring to eventual vision loss. However, lacrimal gland dysfunction is currently incurable and the common treatment option for the resulting dry eye pathology is the application of artificial tears that provides only temporary relief. Recent studies have shown that engraftment of lacrimal gland germ can restore lacrimation in animal models, but the procurement of lacrimal gland cells remains an unresolved challenge [2]. A better fundamental understanding of lacrimal gland development may inform cell-based therapies to repair or regenerate the lacrimal gland, which holds great promise for the treatment of dry eye disease [3].
The neural crest is a multipotent stem cell population that gives rise to many diverse tissues, including craniofacial bones and cartilage, smooth muscle, neurons and ganglia of the peripheral nervous system, adipose cells and melanocytes [4, 5]. Upon induction at the neural plate border, the neural crest undergoes an epithelial-to-mesenchymal transition to delaminate from the dorsal neural tube. These cells then migrate to different regions of the embryo and differentiate into distinct cell types, guided by both their origins along the anterior-posterior axis and the signaling cues they are exposed to in their immediate environment [6]. Once at their destination, neural crest cells closely interact with their host organs, influencing their patterning and morphogenesis [7]. The cranial neural crest cells originating from the midbrain are the source of the periocular mesenchyme, which expresses the chemoattractive signal of Fgf10 to regulate lacrimal gland development [8, 9]. By binding to Fgfr2b and heparan sulphate proteoglycan co-receptors, Fgf10 induces the invasion and branching of the lacrimal gland epithelium [8, 10, 11]. This essential role of Fgf10 in branching morphogenesis is conserved in glandular organs that include the lung, prostate, and pancreas. Nonetheless, the control of Fgf10 expression in the neural crest derived tissues remains unknown.
In this study, we showed that FGF signaling mediated by the protein phosphatase Shp2 is required for the proper patterning and differentiation of the neural crest-derived mesenchyme to produce Fgf10. Genetic evidence further demonstrates that Shp2 is recruited by Frs2 to activate Ras-MAPK signaling downstream to Fgfr1 and Fgfr2 but not to Pdgfrα in the neural crest. By differential gene expression analysis, we identified the homeodomain transcription factor Alx4 as the key effector of Shp2 signaling to control the expression of Fgf10 in the periocular mesenchyme. Importantly, the Alx4 binding sequence in the Fgf10 gene locus is conserved in land species from human to lizard, but not in aquatic animals such as frog and fish, which provides a new genetic insight into how the lacrimal gland arose as an evolutionary innovation of terrestrial animals to adapt to the dry environment. Alx4 conditional knockouts disrupted lacrimal gland development in mouse models and a homozygous ALX4 mutation causes lacrimal gland aplasia in human. Our results reveal a FGF-Shp2-Alx4-Fgf10 axis in regulating neural crest and lacrimal gland development.
FGF signaling is important for development of the neural crest derived craniofacial structures [12–18]. Using the neural crest specific Wnt1-Cre, we observed that conditional knockout of Fgfr1 resulted in significant craniofacial abnormalities, whereas deletion of Fgfr2 did not exhibit any obvious effect (Fig 1A, 1C and 1E, arrows). Lacrimal gland development begins with the invasion of an epithelial bud from the conjunctiva into the periocular mesenchyme at embryonic day 14.5 (E14.5) (Fig 1B, arrowhead). In Fgfr1 and Fgfr2 single mutants, lacrimal gland development was mostly unaffected (Fig 1D and 1F, arrowheads). Combined deletion of both Fgfr1 and Fgfr2, however, abrogated lacrimal gland budding (Fig 1G and 1H, arrows), indicating that Fgfr1 and Fgfr2 can compensate for each other in the neural crest during lacrimal gland development. Fgfr1ΔFrs and Fgfr2LR alleles encode the mutants Fgfr1 and Fgfr2 that lack the docking site for the adaptor protein Frs2 [16, 19]. Although Fgfr2LR homozygous mice were viable and fertile, the craniofacial and lacrimal gland mutant phenotypes were observed in both the Wnt1-Cre;Fgfr1 f/ΔFrs;Fgfr2 f/f and Wnt1-Cre;Fgfr1 f/f;Fgfr2 f/LR mutants (Fig 1I–1L, arrows). The essential role of Frs2 in the neural crest for lacrimal gland development was further demonstrated in Wnt1-Cre;Frs2f/f mutants, which displayed a less severe craniofacial phenotype than Fgfr mutants, but a similar cessation of lacrimal gland budding (Fig 1M and 1N, arrows). Finally, lacrimal gland development was also aborted in Wnt1-Cre;Frs2f/2F animals, which carried mutations in two tyrosine residues of Frs2 (Frs22F) required for the binding of the Shp2 protein phosphatase (S1 Fig, n = 6) [20]. In contrast, although Pdgfrα was expressed in the periocular mesenchyme and required for craniofacial development, its neural crest specific knockout failed to impair lacrimal gland development (Fig 1O–1Q, arrows). These results demonstrated that lacrimal gland development specifically requires FGF-Frs2-Shp2 signaling in the neural crest.
To investigate the potential downstream targets of neural crest FGF signaling occurring during lacrimal gland development, we next generated Wnt1-Cre;Shp2f/f mutants, which failed to develop a lacrimal gland as expected (Fig 2A and 2B, dotted lines, n = 6). Consistent with the idea that the neural crest is the main contributor of the periocular mesenchyme, immunostaining confirmed that Shp2 protein was depleted in the periocular mesenchyme, but preserved in the ectoderm-derived conjunctival epithelium (Fig 2C and 2D, arrows and dotted lines). Although the epithelial cells maintained Pax6 and E-cadherin staining, there was no increase in Col2a1 expression, a hallmark of the nascent lacrimal gland bud (Fig 2E–2H, dotted lines). By contrast, the periocular mesenchyme expression of Col2a1 was preserved, suggesting that the identity of these neural crest-derived cells was unchanged. The Wnt1-Cre transgene was recently reported to cause ectopic expression of Wnt1 in the midbrain-hindbrain boundary [21]. To ensure that this complication did not compromise our results, we used another neural crest-specific deletor, Sox10-Cre, to ablate Shp2, which also resulted in the dysgenesis of the lacrimal gland (S2A and S2B Fig, arrows). Altogether, these results show that Shp2 signaling in the neural crest is required for lacrimal gland budding in a non-cell autonomous manner.
The initial budding of the lacrimal gland requires the inductive signal of Fgf10 that emanates from the periocular mesenchyme. In E13.5 control embryos, Fgf10 was found to exist in a ring-type pattern along the presumptive eyelid surrounding the eye (Fig 2I, arrowheads), with the strongest signal occuring in the mesenchyme adjacent to the future lacrimal gland bud (Fig 2I and 2K, arrows). In both Wnt1-Cre;Shp2f/f and Sox10-Cre;Shp2f/f mutants, however, Fgf10 was absent in the entire periocular mesenchyme (Fig 2J and 2L, arrows and arrowheads, and S2C and S2D Fig). As a result, ERK phosphorylation was maintained in the adjacent retina but abolished in the conjunctival epithelium (Fig 2M and 2N, dotted lines), suggesting a specific loss of FGF signaling in the lacrimal gland primordia. This evidence was further supported by the observed down regulation of FGF signaling response genes, Etv4 and Etv5, in the presumptive lacrimal gland epithelium (Fig 2O–2R, dotted lines). Considering the essential role of Fgf10 signaling in inducing lacrimal gland budding, we concluded that the absence of Fgf10 expression accounted for the lacrimal gland aplasia seen in neural crest Shp2 mutants.
FGF signaling is known to activate the Ras family of small GTPases, which play important roles in cell proliferation, migration and differentiation. Previous studies have identified multiple downstream targets of Ras, including Raf kinases, type I phosphoinositide (PI) 3-kinases, Ral guanine nucleotide exchange factors, the Rac exchange factor Tiam1, and phospholipase C3 [22]. Among these molecules, Raf kinases activate the mitogen-activated protein kinase (MAPK) cascade that culminates with the phosphorylation of Mek kinases (Mek1 and 2) and their direct Erk kinase targets (Erk1 and 2) [23]. At E10.5, ETS transcription factors Etv1, 4 and 5 were strongly expressed in tissues known to have active FGF signaling (Fig 3A, arrows). In both Wnt1-Cre;Shp2f/f and Wnt1-Cre; Mek1f/f;Mek2-/- embryos, these expression patterns were significantly down regulated in the cranial neural crest-derived mesenchyme in the midbrain, branchial arches and nose (Fig 3A, arrowheads), supporting the claim that Shp2 and Mek operate in the same signaling cascade leading to Etv1, 4, and 5 expression. Furthermore, lacrimal gland development was never initiated after the genetic removal of Mek1/2 in the neural crest (Fig 3B, arrowhead, n = 8). Interestingly, however, a small lacrimal gland protrusion was seen in Wnt1-Cre; Erk1-/-;Erk2f/f embryos, suggesting that Mek may have additional key targets other than Erk (Fig 3B, arrowhead, n = 2) that participate in budding morphogenesis. Furthermore, by taking advantage of a conditional allele of oncogenic Kras (LSL-KrasG12D), we showed that constitutively active Ras signaling in the neural crest rescued the Shp2 deficiency during lacrimal gland budding (Fig 3B, arrow, n = 10), supporting the downstream role of Ras-MAPK activation in the FGF-Shp2 signaling cascade in the neural crest [24–27].
The faithful expression of Etv1, 4 and 5 in response to Ras-MAPK activity prompted us to investigate the functional significance of these three transcription factors. Surprisingly, even the combined inactivation of Etv1/4/5 in the neural crest lineage failed to perturb lacrimal gland development (Fig 3C, n = 8), suggesting that these genes may be compensated by other ETS domain transcription factors that share similar binding specificity. To overcome this genetic redundancy, we used a Cre-inducible transgene (R26-EtvEnR) to express Etv4 fused with the Engrailed repressor domain, which acts as a dominant negative ETS transcription factor [28]. Wnt1-Cre; R26-EtvEnR embryos not only exhibited the previously observed craniofacial defect (Fig 3D, arrow), but also showed reduced elongation of the lacrimal gland (Fig 3D, arrowheads, n = 8). This result suggests that ETS domain transcription factors are downstream effectors of FGF-Shp2-Ras-MAPK signaling in neural crest development.
FGF signaling has been implicated in the induction, proliferation, migration and differentiation of neural crest cells [13, 29–32]. The periocular mesenchyme originates from the neural tube in the midbrain, where active FGF signaling indicated by Etv5 expression coincides with Fgf8 expression (Fig 4A, arrows). This suggests that Fgf8 may activate FGF signaling during the induction of cranial neural crest cell progenitors. Considering that Fgf15 is also expressed in the midbrain, we ablated Fgf8 in the midbrain-hindbrain junction using En1-Cre in the Fgf15 null background. As expected, both Fgf8 and Etv5 midbrain expressions were absent in En1-Cre;Fgf8f/f;Fgf15-/- embryos (Fig 4A, arrowheads), demonstrating a loss of FGF signaling. Nevertheless, the lacrimal gland bud still developed normally in these mutants (Fig 4A, asterisks; n = 3), showing that FGF signaling at the induction of cranial neural crest cells is not required for lacrimal gland development.
After induction at the dorsal neural tube, the neural crest progenitors express Sox10 as they begin to migrate toward their final destination. At E10.5, although Sox10-positive neural crest cells were present in the cranial mesenchyme in Wnt1-Cre;Shp2f/f mutants, both their number and extent of migration were slightly reduced as compared to those in the control embryos (Fig 4B, arrows), suggesting that the loss of Shp2 produces subtle defects in neural crest proliferation and migration. This phenotype was reproduced in Wnt1-Cre; Mek1f/f;Mek2-/- embryos, but ameliorated in Wnt1-Cre;Shp2f/f;LSL-KrasG12D embryos (Fig 4B, arrowheads), supporting a role for Shp2-Ras-MAPK signaling in post-inductive neural crest cells.
Previous studies in zebrafish suggested that Shp2 may have a MAPK-independent function in preventing p53-mediated apoptosis in the neural crest [26]. Using lysotracker dye to stain acidic lysosomes in cells undergoing apoptosis, we observed extensive cell death in the first pharyngeal arch in E10.5 Shp2 mutant embryos (Fig 4C, arrows). In sections, cleaved-caspase 3 staining also detected abnormal cell apoptosis in the periocular mesenchyme, although the apoptotic regions were far removed from the conjunctiva (Fig 4C, arrowheads). We reasoned that if the apoptosis induced by the Shp2 deletion was indeed dependent on p53, then the apoptotic events may be avoided by the removal of p53. However, ablation of p53 in Shp2 mutants failed to prevent cell death in the first pharyngeal arch or to rescue any craniofacial phenotype (Fig 4C, arrows and arrowheads). Further, in lacrimal gland development, budding morphogenesis was still aborted in Wnt1-Cre; Shp2f/f;p53f/f embryos (Fig 4C, asterisks, n = 6). Therefore, p53 was not responsible for either the neural crest cell death or the lacrimal gland aplasia observed in Shp2 mutants.
To determine whether these early onset neural crest defects affect periocular mesenchyme development, we crossed Wnt1-Cre mice with those containing the R26R Cre reporter to follow the fate of the neural crest cells. Interestingly, by the time of lacrimal gland budding at E13.5, the periocular mesenchyme adjacent to the conjunctival epithelium was already occupied by the neural crest derived cells in Shp2 mutants (Fig 4D, arrows). Furthermore, the expression of Pitx2 and Foxc1, two markers of the neural crest derived periocular mesenchyme, were similar in wild-type control and Shp2 mutant eyes (Fig 4E, arrows). Therefore, despite causing an initial delay in neural crest migration and abnormal apoptosis, Shp2 ablation did not disrupt the occupancy of the periocular mesenchyme by the neural crest-derived cells at the time of lacrimal gland budding. We thus concluded that the subtle neural crest migration, survival and proliferation defects seen in Shp2 mutants were unlikely to account for the complete failure of lacrimal gland development.
To determine the molecular basis of the lacrimal gland defect observed in Shp2 mutants, we isolated the E14.5 periocular mesenchyme via laser capture micro-dissection and subsequently performed RNAseq analysis (Fig 5A). Among genes that were downregulated at least two folds in Shp2 mutants, the third and eighth most highly expressed transcription factors were Alx4 and Alx1, respectively (Fig 5B). These results were confirmed by a qPCR analysis of micro-dissected tissues, which also showed significant reductions in Shp2 and Fgf10 expressions as expected (Fig 5C).
We next focused on Alx4 and Alx1 as downstream targets of Shp2 signaling. At both E10.5 and E11.5, Alx4 was widely expressed in the cranial mesenchyme surrounding the wild-type eye, but the expression was moderately reduced in Shp2 mutants (Fig 5D, arrows). At E12.5, a more pronounced reduction of Alx4 expression was evident at the temporal side of the mutant eye, where the lacrimal gland bud would have normally emerged. By E13.5, Alx4 expression was absent in all areas of the periocular region except the dorsal side, but recovered in Wnt1-Cre;Shp2f/f;LSL-KrasG12D embryos (Fig 5D, arrowheads). Immunostaining on sections further confirmed that Shp2 deletion led to a progressive down regulation of Alx4 in the periocular mesenchyme, until it was entirely lost by E14.5 (Fig 5E, arrows). Similarly, Alx1 in control wild-type embryos was expressed just anterior to the elongating lacrimal gland bud at E14.5, but this domain of Alx1 expression eventually vanished in Shp2 mutant embryos (Fig 5E, arrowheads). These results demonstrate that the periocular expressions of both Alx1 and Alx4 are regulated by Shp2 signaling.
The results above revealed a close resemblance of Alx1 and Alx4 expressions in the periocular mesenchyme to that of Fgf10 during embryonic development. To evaluate this further, we examined their expression patterns in the neonatal lacrimal gland. At postnatal day 0 (P0), Fgf10 was detectable in the mesenchymal cells, whereas the FGF-inducible gene Etv5 was expressed in the adjacent ducts and acini, suggesting that FGF signaling remained active at this stage (Fig 6A, arrows). As expected, both Alx1 and Alx4 mRNA were also found in the lacrimal gland mesenchyme. Through immunostaining, we further demonstrated that the P3 lacrimal gland expressed the Alx4 protein, which was separated from both the epithelial marker Pax6 and the myoepithelial marker SMA (Fig 6B). Finally, to trace the origin of these Alx4-expressing cells in the lacrimal gland, we crossed Wnt1-Cre with an R26TdT (Ai14) reporter to indelibly label the neural crest-derived cells with tdTomato fluorescence. We then confirmed through immunostaining that Alx4 resided exclusively in the tdTomato-positive cells, demonstrating that Alx4 persisted in the neural crest lineage throughout lacrimal gland development.
Based on the similarities observed between Alx1/4 and Fgf10 expression patterns during lacrimal gland development, we hypothesized that Alx1 and Alx4 were direct regulators of Fgf10 transcription. Because formation of the lacrimal gland was an adaptation of terrestrial animals to an airy environment, we searched the Fgf10 locus for regions that were evolutionarily conserved from human to chicken but not in stickleback fish (Fig 6C). We next overlaid these regions with DNase hypersensitive sites in a 3T3 fibroblast cell line identified by the ENCODE project, because this cell line expressed both Alx4 and Fgf10 at high levels [33]. Finally, we screened these sequences using the Alx1/3/4 binding motif and identified a perfect match within intron 1 of Fgf10 (Fig 6D). Interestingly, sequence alignment showed that this site was evolutionarily conserved among reptiles that have the lacrimal gland, such as the lizard, but not in Xenopus frog, which lacks one (Fig 6C) [34].
To ascertain whether this sequence was a bona fide Alx binding site, we performed chromatin immunoprecipitation in 3T3 cells followed by qPCR using specific primers. Compared to the IgG control, there was a ~3 fold enrichment of this putative Alx binding element in chromatins pulled down by the Alx4 antibody (Fig 6E). This was further validated in vivo by Alx4 chromatin immunoprecipitation using the lacrimal gland mesenchyme isolated from neonatal pups, which resulted in a ~11 fold enrichment. We next knocked down Alx1 and Alx4 using siRNAs in cultured lacrimal gland mesenchymal cells (Fig 6F). Interestingly, Alx1 depletion led to a modest reduction in Fgf10 mRNA levels, but the effect was not statistically significant (Fig 6G). In contrast, the Alx4 knockdown decreased Fgf10 expression by ~50%, which was not further reduced by the combined treatment of both Alx1 and Alx4 siRNAs. This result suggested that Alx4 plays a more dominant role than Alx1 in regulating Fgf10 within the lacrimal gland mesenchyme.
To determine the functional role of Alx4 in lacrimal gland development, we analyzed Alx4lst-J mice, which carried a frameshift mutation that removed both the homeodomain and downstream CAR domain. Homozygous Alx4lst-J animals displayed craniofacial defects, dorsal alopecia and preaxial polydactyly at birth as previously reported in Alx4 knockouts [35, 36]. At E14.5, Alx4lst-J homozygous embryos maintained normal expression levels of Connexin43 and Col2a1 in the periocular mesenchyme, but the domain of Alx1 expression was more restricted (Fig 7A, arrows). Importantly, there was a drastic reduction of Fgf10 adjacent to the lacrimal gland bud, accompanied by a downregulation of FGF-target genes Etv4 and Etv5 in the lacrimal gland epithelium (Fig 7A, dotted lines). At E16.5, histology and immunostaining revealed a complete loss of Alx4 expression in the periocular mesenchyme and a much shorter Pax6-expressing lacrimal gland bud, characterized by reduced phospho-Histone H3 (pHH3) and increasing TUNEL signal (Fig 7B, dotted lines). By P1, no lacrimal gland was detectable by Carmine staining in Alx4lst-J homozygous pups (Fig 7B, black arrows). These results demonstrated that inactivation of Alx4 markedly disrupted Fgf10 expression and downstream FGF signaling, affected cell proliferation and survival, and ultimately caused a failure of lacrimal gland development.
In human, ALX4 loss-of-function mutations underlie autosomal recessive frontonasal dysplasia 2 syndrome, characterized by skull defects, wide nasal bridge, notched nares, depressed nasal tip, hypertelorism and alopecia (OMIM 613451). We reanalyzed one patient carrying a homozygous c.503delC mutation in exon 2 of the ALX4 gene, which resulted in the truncation of the homeobox (HD) and C-terminal OAR domains [37]. MRI imaging in that patient revealed a bilateral absence of lacrimal glands (Fig 7C, arrows). The patient lacked tearing and experienced irritable eyes and multiple episodes of eye infection since birth. This finding is consistent with the role of ALX4 in human lacrimal gland formation.
In this study, we show that FGF signaling in the neural crest is required for Fgf10 production within the periocular mesenchyme, thereby triggering a second round of FGF signaling in the conjunctival epithelium to form the lacrimal gland (Fig 7D). This is mediated by Frs2 and Shp2, which together activate the Ras-MAPK pathway to control the survival, migration and differentiation of the cranial neural crest cells. The downstream effector of Shp2 signaling in the periocular mesenchyme is the homeodomain transcription factor Alx4, which binds a terrestrially conserved element to regulate Fgf10 expression in the periocular mesenchyme, reflecting the evolutionary history of the lacrimal gland. Our results highlight the sequential use of FGF signaling in neural crest development and reveal the etiology of lacrimal insufficiency in an ALX4 patient.
RASopathies represent a spectrum of congenital abnormalities caused by aberrant Ras-MAPK signaling, but the particular RTK signaling pathway mediated by Ras in the normal development of a specific tissue is not always clear [38, 39]. Using mouse genetics, we showed that defective FGF signaling, and not PDGF signaling, in the neural crest reproduced the Shp2 conditional knockout phenotype seen in the lacrimal gland, thereby positioning FGF receptors as the primary regulators of Shp2 function in the neural crest cells that partake in directing the development of the lacrimal gland. Contrary to a previous study in zebrafish, we did not observe that Shp2 acts upstream of p53 to suppress neural crest cell apoptosis [26]. This discrepancy could be due to differences either intrinsic to the species used or to the experimental approaches utilized as we took advantage of conditional knockouts in mice whereas the zebrafish study used a morpholinos knockdown. Instead, our genetic evidence demonstrates a fundamental role for the Shp2-Ras-Mek-Erk signaling cascade in neural crest survival and development. MAPK is known to phosphorylate and induce the ETS domain transcription factors, which act as downstream effectors in gene regulation. In particular, the expressions of Pea3 family genes Etv1/4/5 correlate closely with FGF signaling activities during embryonic development [10]. While deletion of all three Pea3 family genes in the neural crest failed to produce any craniofacial or lacrimal gland defects, the overexpression of a dominant-negative Etv4 lead to stunted lacrimal gland growth. This suggests that other members of the ETS domain transcription factors, which recognize similar binding sites as Etv1/4/5, can play redundant roles in transmitting FGF-MAPK signaling during neural crest development.
Our study demonstrates that Alx genes are the ultimate downstream effectors of Shp2 signaling in the periocular mesenchyme. Alx4 shares both sequence and structural homologies of paired-type homeodomain and C-terminal aristaless domain with two other transcription factors, Alx1 and Alx3. These proteins are present within the craniofacial mesenchyme and limb bud, displaying overlapping expression patterns [40]. Members of this family of transcription factors also exhibit functional redundancies as shown by genetic interactions in specific tissues. Alx3 knockout mice were morphologically normal, but Alx3/4 double mutants displayed more severe defects in the neural crest-derived craniofacial structures than the Alx4 knockout alone [40]. Alx1 null mice showed craniofacial defects distinct from Alx4 mutants and combined deletion of both genes led to developmental abnormalities not found in either of the single mutants, indicating that Alx1 and Alx4 have both unique and redundant roles [36]. The lacrimal gland mesenchyme expresses Alx1 and Alx4, but not Alx3. Although we did not observe a synergistic effect of Alx1 and Alx4 in our in vitro experiments, it remains possible that Alx4/Alx1 double knockout mice will present comparably severe lacrimal gland defects as the neural crest Shp2 mutant did.
The precise level of FGF10/Fgf10 expression in the periocular mesenchyme is critical for lacrimal gland induction. This is clearly shown by aplasia of the lacrimal and salivary glands (ALSG) and Lacrimo-auriculo-dento-digital (LADD) syndromes, in which even heterozygous mutations in human FGF10 can lead to congenital lacrimal gland defects [41, 42]. Our study has demonstrated that neural crest FGF signaling is required for Fgf10 expression in the periocular mesenchyme, but the ligand of the neural crest FGF signaling that leads to lacrimal gland development remains an open question. It is unlikely to be the autocrine signaling of Fgf10, because deletion of Fgfr2, the cognate receptor for Fgf10, in the neural crest only produced minor defects in lacrimal gland development (Fig 1). In limb development, the mesenchyme-derived Fgf10 signals the epithelium to induce Fgf8 and later Fgf4, Fgf9 and Fgf17, which in turn act on Fgfr1 and Fgfr2 in the mesenchyme to maintain Fgf10 expression [43–45]. During lung development, Fgfr1 and Fgfr2 in the mesenchyme respond to Fgf9 expressed by the lung epithelium and mesothelium. This maintains the mesenchymal expression of Fgf10 that signals back to the epithelium [46, 47]. Submandibular salivary gland development is yet another example where the epithelium-mesenchyme interaction plays an important role. In this case, Fgf10 in the mesenchyme originated from the cranial neural crest is modulated by ectodermal-derived Fgf8 [48]. However, neither a systemic knockout of Fgf9 nor deletion of Fgf8 using Cre transgenes specific to the midbrain-hindbrain junctiondisrupt lacrimal gland development (S3 Fig and Fig 4A). Considering the complexity of the FGF family, further work is needed to identify the relevant FGF ligands for the neural crest FGF signaling pathway during lacrimal gland development.
The main and accessory lacrimal glands secrete the aqueous component of the tear film, and thereby play an important role in maintaining the health and transparency of the ocular surface. Because the tear is only necessary for land animals whose eyes are constantly exposed to the air, the lacrimal gland emerged relatively late in the evolution of the vertebrate tetrapod. Even among animals living both on land and in water, the lacrimal gland is only present in reptiles such as the alligator, but not in amphibians such as the frog (S4 Fig). In this study, we show that the Alx4 binding site in the Fgf10 locus lies within a region that’s conserved from humans to alligators, but not in frogs or fish. This suggests that, although both Alx4 and Fgf10 arose in more primitive organisms, these two genes were most likely not functionally linked until the emergence of the lacrimal gland in reptiles. Considering that Fgf10 lies at the top of the genetic cascade for inducing branching morphogenesis in many glandular organs, this represents an example of evolution that coopts an existing genetic circuitry to develop new organs that enable the adaptation to new environments. By showing that the Alx4-Fgf10 axis is conserved from mouse to human, our study contributes to the understanding of the role of Alx4 in human neural crest cell and lacrimal gland development and points in the direction of generating the lacrimal gland from pluripotent stem cells.
The animal experiments were approved by Columbia University Institutional Animal Care and Use Committee (IACUC).
Mice carrying Erk1-/-, Erk2flox, Frs2αflox, Frs2α2F, Mek1flox, Mek2KO, Shp2flox alleles were bred and genotyped as described [20, 49–52]. We obtained Etv1flox mice from Dr. Silvia Arber (University of Basel, Basel, Switzerland), Etv4-/- and Etv5flox mice from Dr. Xin Sun (University of California at San Diego, San Diego, CA), En1-Cre and R26-EtvEnR from Dr. James Li (University of Connecticut Health Center, Farmington, CT), Fgf8flox from Dr. Suzanne Monsour (University of Utah, Salt Lake city, UT), Fgf15-/- from Dr. Steven Kliewer (UT Southwestern Medical Center, Dallas, TX), Fgfr1ΔFrs from Dr. Raj Ladher (RIKEN Kobe Institute-Center for Developmental Biology, Kobe, Japan), Fgfr2LR from Dr. Jacob V.P. Eswarakumara (Yale University School of Medicine, New Haven, CT) and Fgf9-/- and Fgfr2flox from Dr. David Ornitz (Washington University Medical School, St Louis, MO) [16, 19, 28, 53–58]. LSL-KrasG12D mice was obtained from the Mouse Models of Human Cancers Consortium (MMHCC) Repository at National Cancer Institute [59]. Alx4lst-J (Stock No: 000221), Fgfr1flox (Stock No: 007671), p53flox (Stock No: 008462), Pdgfrαflox (Stock No: 006492), R26R (Stock No: 003474), R26RTdT (Ai14, Stock No: 007914), Sox10-Cre (Stock No: 025807) and Wnt1-Cre (Stock No: 009107) mice were obtained from Jackson Laboratory [16, 40, 60–64]. Animals were maintained on mixed genetic background. Wnt1-Cre or Shp2flox only mice did not display any lacrimal gland phenotypes and were used as controls.
Histology, carmine staining, TUNEL assays and immunohistochemistry are performed as previously described [11, 65]. The following primary antibodies were used: Alx4 (sc-33643, Santa Cruz Biotechnology), E-cadherin (U3254, Sigma, St Louis, Missouri), Cleaved-caspase 3 (#9664, Cell signaling Technology), Col2a1 (ab34712, Abcam), Connexin43 (#3512, Cell signaling Technology), pHH-3 (#06–570, Millipore), Pax6 (PRB-278P, Covance, Berkeley, CA, USA), RFP (#600-401-379, Rockland), α-SMA (#C6198, Sigma-Aldrich).
E13.5 embryos were incubated in 4% PFA for 1 hr at 4°C and washed twice in PBS containing 0.02% NP-40, 0.01% sodium deoxycholate and 2 μg/ml MgCl2 for 30 min each, followed by overnight incubation in X-gal staining solution (1 mg/ml X-gal, 10 mM Potassium Ferricyanide, 10nM Potassium Ferrocyanide, 2 μg/ml MgCl2 in PBS) at 4°C. The samples were then cryopreserved in OCT (Sakura Finetek), sectioned at 10 μm thickness and counter-staining with nuclear red.
RNA in situ hybridization was performed as previously described [66]. The following probes were used: Alx1 (from Dr. Terence Capellini, Harvard University, Boston, MA), Alx4 (from Dr. Yang Chai, University of Southern California, Los Angeles, CA), Etv4, Etv5 (from Dr. Bridget Hogan, Duke University Medical Center, Durham, NC, USA), Foxc1 (from Dr Anthony Firulli, Indiana University School of Medicine, Indianapolis, IN, USA), Fgf10 (for whole mount) (from Dr. Suzanne Monsour, University of Utah, Salt Lake city, UT)), Fgf10 (for sections)was generated from a full length cDNA clone (IMAGE: 6313081 Open Biosystems, Huntsville, AL, USA), Pitx2 (from Dr. Valerie Dupé, CNRS, Strasbourg, France).
Freshly harvested embryos were frozen in the OCT, sectioned at 10μm thickness and transferred to PEN slides (Zeiss). Slides were dipped in 95% ethanol for 2 min to fix the samples and stained with crystal violet stain (3% in ethanol) on ice. This was followed by dipping in 70% ethanol for 30–40 sec to remove the OCT and dehydration in 100% ethanol for 2 min. The periocular mesenchymal tissue was micro-dissected using Laser capture microscope (Zeiss AxioObserver.Z1 inverted microscope). 500 pg of RNA was isolated from each sample, converted to cDNA and amplified using Nugen Ovation kit (Nugen) to obtain 2–3 μg cDNA, which was then converted to cDNA library for RNA-sequencing analysis at core facility in Columbia University. The RNAseq data is available at the GEO repository under accession number GSE103402.
Lacrimal glands mesenchymal culture was performed as described previously [67]. Briefly, glands were isolated from P0-P2 pups and trypsinized (Gibco 1:250) at 4°C for 1 hr. After neutralizing trypsin, the mesenchyme was manually separated from the epithelium using fine needle and grown in the complete medium (DMEM+10% FBS with antibiotics) for 3 days before passage. The primary mesenchymal cells were transfected with siRNA using Lipofectamine RNAimax as previously described and harvested after 24–48 hrs [68]. For Alx1 and Alx4, the results were confirmed using two different predesigned Silencer® Select siRNAs from Ambion (Life technologies).
Quantitative-PCR was performed as described [69]. Primer sequences used were, Alx4: 5’-ACACATGGGCAGCCTGTTTG3’, 5-TGCTTGAGGTCTTGCGGTCT-3’, Alx1: 5’ GGAGGAAGTGAGCAGAGGTG-3’, 5’- TTCAAATGCGTGTCCGTTGGT-3’, Fgf10: 5’ CAATGGCAGGCAAATGTATG-3’, 5’- GGAGGAAGTGAGCAGAGGTG-3’, Gapdh: 5’-AGGTCGGTGTGAACGGATTTG-3’, 5’-TGTAGACCATGTAGTTGAGGTCA-3’, Shp2 (exon 4): 5’- CTGACGGAGAAGGGCAAGCA-3’, 5’- CGCACGGAGAGAACGAAGTCT-3’.
The Chromatin Immunoprecipitation (ChIP) assays were performed in 3T3 fibroblasts cells and primary lacrimal gland mesenchymal cells as described [70]. Briefly, the cells grown in DMEM/10% FBS with antibiotics were crosslinked with 1% Formaldehyde for 8–10 min with gentle shaking. This was followed by quenching with 125 mM glycine or 5 min, 3X washing with cold PBS and addition of 1 ml of cold CHIP lysis buffer. After incubation for 10 min at 4°C, the lysed cells were centrifuged at 3000 rpm for 3 min and the pellet were stored at -80°C until later use. The pellet was then resuspended in 1.2 ml of RIPA buffer, sonicated on ice for 8 min using probe sonicator (1 sec “on”, 2 sec “off”, power 3.5) and centrifuged at 13000 rpm for 15 min at 4°C. The supernatant was precleared by adding 45 μl Protein G agarose beads (50% slurry, Millipore) and incubated for 2 hrs at 4°C on rotor. After centrifugation at 5000 rpm for 1 min, the supernatant was transferred to a fresh tube and the protein concentration was measured by Bradford assay. For pull down, 1 μg of antibodies were added per 1mg of protein for overnight incubation at 4°C, followed by addition of 20 μl agarose beads for another 1–2 hours incubation. After brief centrifugation, the beads were washed 1X with RIPA buffer at room temperature, 3X with cold RIPA buffer, 2X with cold Wash buffer A and Wash buffer B, 1X with TE/150mM NaCl. Next, the samples were decrosslinked in Elution buffer containing RNAase (40μg/ml) and Proteinase K (20μg/ml) for 1 hr at room temperature and 50°C overnight. After brief centrifugation, the supernatant was treated with equal vol. of Phenol/Chloroform and the DNA was precipitated with 2.5 vol. of 100% ethanol and Glycoblue for 1 hr at -80°C and dissolved in 20 μl sterile water for qPCR analysis. The antibodies used were IgG as isotype control (sc-2028, Santa Cruz Biotechnology) and anti-Alx4 (sc-22066, Santa Cruz Biotechnology). Buffer recipes: CHIP lysis buffer- 10mM Tris-Cl, pH8, 85mM KCl, 0.5% NP-40, 5nM EDTA, 0.25% Triton; RIPA- 1% Triton, 150mM NaCl, 0.1% SDS, 0.1% Na-Deoxycholate, 10mM Tris-Cl, pH8, 5mM EDTA; Wash buffer A- 50mM HEPES, pH7.9, 500mM NaCl, 1mM EDTA, 1% Triton, 0.1% Na-deoxycholate, 0.1% SDS, Wash buffer B- 20mM Tris-Cl, pH8, 1mM EDTA, 250 mM LiCl, 0.5% NP-40, 0.5% Na-deoxycholate; Elution Buffer- 1% SDS, 30 mM Tris-Cl (pH8), 15mM EDTA, 200mM NaCl. Protease inhibitor cocktail is added prior to use in all the buffers until ready to elute. The primers used for CHIP in intron 1 of Fgf10- F- 5’-GGTTGGAGCTTGTTGTGTGT-3’, R- 5’-GCTCTGCTAATAAAGGTCTCCC-3’.
We retrieved 200 KB upstream and 100 KB downstream of Fgf10 transcription start site from Mouse Genome assembly GRCm38/mm10 and analyzed this sequence for evolutionary conservation using UCSC genome browser. These sequences were also overlaid with the DNase-hypersensitivity data from 3T3 cell line retrieved from ENCODE database and scanned for Alx4 consensus binding sites based on TRANSFAC (release 2013.1) database using MATCH algorithm, with minFP as parameter to identity sites with minimum false positives.
|
10.1371/journal.ppat.1007707 | A specific sequence in the genome of respiratory syncytial virus regulates the generation of copy-back defective viral genomes | Defective viral genomes of the copy-back type (cbDVGs) are the primary initiators of the antiviral immune response during infection with respiratory syncytial virus (RSV) both in vitro and in vivo. However, the mechanism governing cbDVG generation remains unknown, thereby limiting our ability to manipulate cbDVG content in order to modulate the host response to infection. Here we report a specific genomic signal that mediates the generation of a subset of RSV cbDVG species. Using a customized bioinformatics tool, we identified regions in the RSV genome frequently used to generate cbDVGs during infection. We then created a minigenome system to validate the function of one of these sequences and to determine if specific nucleotides were essential for cbDVG generation at that position. Further, we created a recombinant virus unable to produce a subset of cbDVGs due to mutations introduced in this sequence. The identified sequence was also found as a site for cbDVG generation during natural RSV infections, and common cbDVGs originated at this sequence were found among samples from various infected patients. These data demonstrate that sequences encoded in the viral genome determine the location of cbDVG formation and, therefore, the generation of cbDVGs is not a stochastic process. These findings open the possibility of genetically manipulating cbDVG formation to modulate infection outcome.
| Copy-back defective viral genomes (cbDVGs) regulate infection and pathogenesis of Mononegavirales. cbDVGs are believed to arise from random errors that occur during virus replication and the predominant hypothesis is that the viral polymerase is the main driver of cbDVG generation. Here we describe a specific genomic sequence in the RSV genome that is necessary for the generation of a large proportion of the cbDVG population present during infection. We identified specific nucleotides that when modified altered cbDVG generation at this position, and we created a recombinant virus that selectively produced cbDVGs based on mutations in this sequence. These data demonstrate that the generation of RSV cbDVGs is regulated by specific viral sequences and that these sequences can be manipulated to alter the population of cbDVGs generated during infection.
| Defective viral genomes (DVGs), which are generated during the replication of most RNA viruses, potentiate the host innate immune response [1–5] and attenuate the infection in vitro and in vivo [4, 6–9]. Importantly, in naturally infected humans, the presence of DVGs correlates with enhanced antiviral immune responses during RSV infection [6] and reduced disease severity in influenza virus infection [8]. Significant effort is currently invested in harnessing DVGs as antivirals due to their strong immunostimulatory activity and ability to interfere with the replication of the standard virus. However, despite over 50 years of appreciating their critical functions in multiple aspects of viral infections, the molecular mechanisms that drive DVG generation remain largely unknown. This lack of understanding hampers our ability to effectively harness DVGs for therapeutic purposes and limits our capacity to generate tools to elucidate their mechanism of action and impact during specific viral infections.
There are two major types of DVGs: deletion and copy-back (cb) [10]. Both types are unable to complete a full replication cycle without the help of a co-infecting full-length virus [11, 12] and can be packaged to become part of the viral population [13]. Deletion DVGs, common in influenza virus and positive strand RNA viruses, retain the 3’ and 5’ ends of the viral genomes but carry an internal deletion [14–16]. These types of DVGs are believed to arise from recombination events [17, 18] and can strongly interfere with the standard virus [19]. cbDVGs are common products of non-segmented negative sense (nns) RNA virus replication, including Sendai virus (SeV), measles virus, and respiratory syncytial virus (RSV), and are the primary stimulators of the innate immune response during nnsRNA virus infection [6, 7, 20, 21]. cbDVGs arise when the viral polymerase detaches from its template at a “break point” and resumes elongation at a downstream “rejoin point” by copying the 5’ end of the nascent daughter strand [12, 22]. This process results in the formation of a new junction sequence and a truncated genome flanked by reverse complementary ends [23]. cbDVGs have long been thought to result from errors made by the RNA-dependent RNA polymerase (RdRp) during replication due to a combination of lack of proofreading activity and the presence of a polymerase with lower replication fidelity [12]. No pattern or specific sequences for the break and rejoin points of cbDVGs have been reported so far.
Based on our consistent observations of discrete populations of cbDVG generated during RSV infections in vitro and in vivo [6], we set out to test the hypothesis that the generation of cbDVGs is not completely stochastic but instead is regulated by a carefully orchestrated process. Upon identification of cbDVG populations generated during infection, we show that specific viral sequences within the viral genome are preferred sites for cbDVG generation and that these sequences are conserved across viral strains. Utilizing this knowledge, we generated a recombinant virus that produced a restricted set of cbDVGs, serving as strong evidence that specific sequences dictate where cbDVGs are generated. In addition, we demonstrate for the first time that common cbDVGs are generated independently in natural infections in humans, further supporting an orchestrated origin for cbDVGs.
To acquire a comprehensive view of the population of cbDVGs generated during infection, we developed an algorithm to identify cbDVG junction regions within RNA-seq datasets with high sensitivity. The principle of this Viral Opensource DVG Key Algorithm (VODKA) is illustrated in Fig 1A. In brief, the break point (T’) and the rejoin point (W) are far apart in the parental viral genome, but in cbDVGs they become continuous and form the new cbDVG junction sequence when the viral polymerase (Vp) is released from the template and rejoins in the nascent strand. Since cbDVG junction sequences are absent in the full-length viral genome, VODKA identifies the sequence reads that capture junction sequences and then filters out false-positives reads that fully align to the reference genome.
We corroborated VODKA’s performance by testing for the presence of the highly dominant DVG-546 in samples from infections with SeV Cantell (Fig 1B). Briefly, 98,543 out of the 98,626 (99.9%) cbDVG junction reads identified by VODKA from an RNA-seq data set obtained from SeV Cantell-infected cells mapped exactly to the known junction region of DVG-546 (Fig 1C, bottom panel). By aligning cbDVG junction reads to the SeV full-length antisense genome, we determined the location of the major break and rejoin regions (blue peaks in the upper panel of Fig 1C). Each read that aligned to a break or rejoin region contained two portions, one of which fully aligned to the reference genome. The aligned reads in the break region (pink box, Fig 1C) and in the rejoin region (gray box, Fig 1C) corresponded to the DVG-546 sequence before the break point and after the rejoin point, respectively. The breakpoint for DVG-546 predicted by VODKA (14932±1_15292±1) exactly matched the one identified by Sanger sequencing (↑ in Fig 1C), thereby establishing the efficiency and accuracy of VODKA in identifying cbDVG-specific sequences.
We then used VODKA to identify the population of cbDVGs generated during RSV infection. RSV is a virus known to generate immunostimulatory cbDVGs in infected patients [6], and thus a subject of interest in our laboratory. We analyzed pooled RNA-seq datasets from six RSV-infected cell cultures. These cultures were infected with RSV generated from the same parental stock (strain A2) that was first depleted of DVGs and then passaged independently in different cell lines to generate six different stocks enriched for DVGs. The presence of cbDVGs in these infections was confirmed using a specific RT-PCR followed by Sanger sequencing. By aligning VODKA-identified cbDVG junctions to the RSV A2 reference antigenome, we observed 4 major break hotspots spanning over 1300 nucleotides (red down-facing arrows in Fig 1D). In contrast, only 2 major rejoin hotspots were observed within a narrower region of 223 nucleotides in length at the 3’end of the viral genome (black down-facing arrows in Fig 1D). Remarkably, the rejoin area with the highest peak included counts present in all six virus stocks. We then compared these break and rejoin hotspots to those generated in infections with a different stock of RSV enriched in cbDVGs (stock7) from which the major cbDVGs were identified upon Sanger sequencing of PCR amplicons. We observed that the cbDVG rejoin points from stock7 were located within the strongest rejoin hotspot, whereas its four break points were distributed more broadly across the genome (Fig 1D). These results reveal strong hotspots for the polymerase to rejoin during cbDVG formation and suggest a large degree of conservation of the RSV cbDVGs rejoin positions.
To determine if candidate hotspots were involved in cbDVG generation, we selected the region containing the most break points (Break1, dark grey in Fig 1D), or the most rejoin points (Rejoin1+Trailer, light grey in Fig 1D) for further testing using a minigenome system [24]. We constructed an RSV minigenome backbone (BKB) that included the reporter gene mKate2 for flow cytometry quantification of transcripts produced by the viral polymerase. In addition, the minigenome BKB included restriction enzyme sites to insert the selected Break1 and/or Rejoin1 regions (Fig 2A). The goal was to use this system to establish whether sequences in the candidate break and rejoin regions altered the polymerase elongation capacity, eventually leading to the generation of cbDVGs. The strategy used for detection of cbDVGs is illustrated in S1 Fig. As illustrated in Fig 2B, in this system mKate2 expression should only occur if the viral polymerase replicates the entire minigenome sequence from the trailer to the leader. Co-transfection of the minigenome construct with the four helper plasmids expressing the polymerase proteins (L, P, NP, and M2-1), resulted in mKate2 expression in 8–17% of the cells, whereas no mKate2 expression was detected in control transfections that lacked the viral polymerase (Fig 2C; -Vp). Constructs containing only the Rejoin1 sequence led to similar mKate2 expression as the BKB construct, whereas constructs containing Break1 caused a ~30% reduction in mKate2 expression (Fig 2D and 2E). We verified that the difference in mKate2 expression among transfections with different constructs was not due to variable transfection efficiency (S2A–S2C Fig), or cell death (S2D Fig). These results are consistent with the concept that during cbDVG generation, the viral polymerase falls off the template at the break region leading to a reduced amount of newly synthetized template available for mKate2 transcription.
To formally assess whether candidate break and rejoin sequences lead to cbDVG formation, we cloned the designated Pair1 composed of Break1/Rejoin1 into the minigenome system. Upon transfection, we observed that the construct containing Pair1 led to a similar degree of mKate2 expression than the construct bearing Break1 alone (Fig 3A and 3B). We also observed two major amplicons (white arrowheads in Fig 3C), both of which were absent in cells transfected with the construct bearing Break1 alone. These two amplicons contained cbDVGs that were confirmed by conventional Sanger sequencing (S3A Fig). The individual break and rejoin points of these minigenome-generated cbDVGs are indicated in Fig 3D. Interestingly, the rejoin points clustered in close proximity to the rejoin points that we identified from in vitro infected cells. Taken together, these data demonstrate that RSV cbDVG rejoin points fall into a discrete region of the viral genome, which is critical for cbDVG generation.
Since the late Rejoin1 + early Trailer region of the RSV genome was highly enriched with DVG rejoin points relative to other regions in the RSV genome, we then examined which specific features within this region impacted cbDVG generation. This region is within one of most A and U enriched areas of the RSV genome (Fig 4A), suggesting that nucleotide composition might play a role in directing cbDVG formation. To avoid affecting the L “gene end” signal and the genome trailer region, we chose to mutate six nucleotides at the beginning of this rejoin region (nucleotide positions 191–186 from the 3’ end of antigenome) to either all Us (named GC>Us) or all GCs (named AU>GCs). We then used RT-PCR (DI-1/DI-R primer set) to detect cbDVG-like fragments formed in the cells co-transfected with all U’s or all GCs mutant constructs and polymerase-expressing plasmids, as described earlier. Mutant GC>Us generated a dominant amplicon (lane2, Fig 4B) that was absent in cells transfected with mutant AU>GCs (lane4, Fig 4B). From sequencing PCR products within the areas marked by asterisks in Fig 4B, we identified five distinct rejoin points from mutant AU>GCs and three from mutant GC>Us (↑ in Fig 4C). Compared to WT Pair1, mutant GC>Us did not generate rejoin points proximal to the mutated region (grey area in Fig 4C), whereas mutant AU>GCs still produced cbDVG-like fragments at the mutated area.
To rule out bias due to primer location, we designed two additional forward primers to detect cbDVG-like fragments from the same samples. Rejoins detected with DI-F2 primers are identified with red arrows in Fig 4C, while rejoins detected with DI-F3 are indicated with blue arrows. Transfections with mutant GC>Us resulted in one strong amplicon while no predominant amplicon was observed in transfections with mutant AU>GCs (Fig 4D), agreeing with results obtained using the DI-F1 primer set. Sequencing confirmed that the strong amplicons produced by all three different primer sets in transfections with mutant GC>Us were analogous cbDVG-like fragments and shared their break and rejoin points (sequence in S3 Fig, DVG 303bp). To examine if the observed lack of a predominant product resulting from mutant AU>GCs was due to a general reduction of replication ability of the viral polymerase induced by mutations, we introduced the same mutations in the construct with Rejoin1 alone and examined mKate2 expression by flow cytometry. We found no significant differences between Rejoin1 and the two mutants, Rejoin1-GC>Us and Rejoin1-AU>GCs. Neither of these constructs reduced mKate2 expression compared to BKB (S4 Fig), suggesting that the function of the RSV minigenome system remained intact despite of the mutations. Altogether, these data suggest that a minimal content of C nucleotides in the rejoin region determines if cbDVGs are produced at that particular genomic location.
To determine if any of the two Cs within the mutated sequence was critical for cbDVG rejoining at this location, we performed a similar analysis using three new constructs: first C at position 188 or second C at position 186 from the 3’ end of antigenome mutated to U (named C188U or C186U, respectively), or both Cs mutated to Us (named AU). Transfection of the C186U, but not the C188U construct, resulted in one major DVG amplicon (indicated with an asterisk in Fig 4E; sequences in S3 Fig). The C186U construct rejoin points skipped the mutation area and concentrated in the early trailer region, similar to GC>Us. This was confirmed by the two other primer sets. A strong band shown in lane C188U at a high molecular weight (indicated with an arrowhead in Fig 3E) was determined to not correspond to a cbDVG by Sanger sequencing. The construct bearing the double mutation (AU) behaved similar to C186U in terms of the rejoin positions (Fig 4C). Thus, we found the second C at position 15037 (position 186 from 3’ trailer end of antigenome) to be critical for cbDVG generation.
Next, to establish whether Rejoin1 impacts on cbDVG generation during viral infection, we created a mutant virus harboring mutations identical to the GC>Us minigenome construct. This virus is herein identified as gRSV-FR-GC>Us. The backbone of the recombinant RSV (Line 19) included the mKate2 gene and we used mKate2 expression to estimate its replication. As shown in Fig 5A, cells infected with gRSV-FR-GC>Us expressed the same level of mKate2 protein as cells infected with the WT reporter virus (gRSV-FR-WT) at 72 h post infection. Both viruses began to generate cbDVGs at passage 3 (P3) and the pattern of cbDVGs was maintained, and became stronger, by P5 (Fig 5B). We verified that P5 gRSV-FR-GC>Us still carried the mutations we introduced (Fig 5C). Interestingly, gRSV-FR-WT produced 4 major DVGs, whereas gRSV-FR-GC>Us only generated one dominant cbDVG (asterisks in Fig 5B, confirmed sequence in S3G and S3H Fig), which is consistent with results from the minigenome system. The dominant cbDVG generated in cells infected with gRSV-FR-GC>Us rejoined at the early trailer region and skipped the mutation site, similar to what was observed in the minigenome system. Cells infected with gRSV-FR-WT produced one cbDVG that rejoined within the mutation site and three other cbDVGs that rejoined at the same region of the mutant virus (Fig 4D). A population of cbDVGs lacking generation at the mutation site can be repeatedly observed upon independent passages of the mutant virus, albeit the specific species of DVGs varied in different lineages (S5 Fig). These data further support a critical role of Rejoin1 in cbDVG generation.
The majority of rejoin points found in infection with gRSV-FR-WT, which derived from RSV Line 19, located within the early trailer sequence, rather than around the mutation site as found during infection with RSVstocks1-7 derived from RSV line A2 (Fig 5D). Alignment of both sequences revealed one natural mutation in RSV Line 19 that introduced three GCs right at the beginning of the trailer sequence, which are not present in RSV A2 (sequence indicated with a red horizontal line in Fig 5D). The increased GC content in this position in Line 19 likely explains why gRSV-FR-WT generates more cbDVGs at this location than RSV A2 stocks1-7. Regardless of this natural preference for rejoining in the early trailer, gRSV-FR-GC>Us diminished the rejoin signal at the mutation site as no cbDVGs rejoin points were found at this location resulting in less diversity of cbDVG generation compared to the WT virus. Overall, these data confirm that the common rejoin region sequence tested in the minigenome system determines cbDVG rejoining during RSV infection and that the content of C nucleotides, and possibly G nucleotides, in this region critically determines the site of cbDVG rejoin.
To examine whether the Rejoin1 region was utilized during natural infections, we applied VODKA to RNA-seq datasets obtained from RSV-positive pediatric samples. A total of 10 clinical specimens were sequenced; 4 were classified as DVG-low and 6 as DVG-high based on semi-quantification following cbDVG PCR. VODKA outputs were aligned to the reference genome of an RSV strain A isolate (Reference genome NCBIKJ672447, 2012) and showed that, consistent with previous cbDVG-RT-PCR results, samples from DVG-low patients (upper panel in Fig 6A) contained ~8 fold less cbDVG junction reads than DVG-high patients (lower panel in Fig 6A). In addition, coverage mapping showed the presence of multiple break and rejoin regions. Some of them were a mix of both break points and rejoin points (Fig 6A, read and black arrows). The rejoin points were particularly noteworthy because the majority of them clustered within one narrow AU-rich “Rejoin1+ Trailer” region (red ticks in Fig 6B) similar to that identified in in vitro infections (blue ticks in Fig 6B). According to the frequency of different cbDVG junction positions, we illustrated the top 6 major cbDVGs (one of them is a snap-back) in Fig 6C (details summarized in Table 1, Break Rejoin position shown as T’_W). All of them were found in multiple patients (Fig 6C and Table 1). The most abundant cbDVG again had the rejoin point within the “Rejoin1+Trailer” region, despite of a higher diversity of rejoin points compared to in vitro infection. Taken together, these results demonstrate that a conserved rejoin region drives the generation of most cbDVGs during RSV infection in vitro and in vivo and that identical RSV cbDVGs are generated in different naturally infected individuals.
DVGs are critical regulators of viral replication and pathogenesis in multiple RNA virus infections, but the mechanisms modulating their generation are unknown. Historically, DVGs were thought to result from random errors introduced by the viral polymerase during replication. However, mounting evidence indicates that the generation of cbDVGs is not totally stochastic. Specifically, we show that during RSV infection discrete hotspots in the viral genome mark sites for the viral polymerase to release and rejoin during cbDVG formation, both in vitro and during natural RSV infections in humans. Moreover, we show that the content of C nucleotides, and possibly G nucleotides, within the major rejoin hotspot critically impacts the generation of cbDVGs at that position. We also identified specific nucleotides that, when mutated, altered the ability of recombinant viruses to generate diverse species of DVGs. The identification of a specific sequence involved in cbDVG formation opens the unprecedented possibility of genetically manipulating the content of cbDVGs during infection. This possibility may significantly impact our ability to generate tools to further understand the role of these viral products in virus pathogenesis, as well as potentially manipulate the cbDVG content with antiviral and/or therapeutic purposes.
In this study, we utilized a custom-designed algorithm, VODKA, to identify cbDVG in infections in vitro or from children naturally infected with RSV. VODKA outputs were consistent with previous results obtained using classic DVG-RT-PCR and demonstrated a higher sensitivity in the detection of cbDVGs both in vitro and in clinical samples. False-positive DVG junction reads were ruled out by screening all reads aligned to the host (reads from human transcriptome) using VODKA. This test resulted in a minimal number of hits compared to viral samples, adding to the evidence reported throughout this manuscript to support the specificity of cbDVG detection by VODKA. VODKA can successfully identify cbDVGs in a number of viruses, including SeV (Fig 1), offering a powerful tool for cbDVG detection in clinical samples. Furthermore, since cbDVGs, compared to other types DVGs, have the most potent immunostimulatory function, VODKA can be used to identify candidates for development of novel cbDVG-based adjuvants.
Based on our data, we conclude that the rejoin position significantly influences cbDVG generation. One C nucleotide substitution alone can influence the location of the DVG rejoin point implying that a strong rejoin signal likely needs an optimal number of C, and possibly G nucleotides, in specific locations. However, the total amount of cbDVGs produced and their immunostimulatory activity are not necessarily altered by the single C substitution in one rejoin hotspot, suggesting the redundancy of rejoin hotspots in cbDVG generation. More research needs to be done to investigate whether mutations in other rejoin hotspots or in combination will alter the overall amount of cbDVGs and their function. Interestingly, the same differential distribution of cbDVG rejoin points was observed when we compared cbDVG generation from infections with RSV A2 and Line19, which differ in their GC content at the beginning of the trailer region. This observation also implies that the preference of usage among different hotspots as cbDVG rejoin points may vary among different RSV subtypes. In addition, our data suggest that rejoin sequences influence the function of break signals when inserted as pairs in the construct. Our data is in agreement with data from in vitro infections with measles virus lacking the C protein, where break points of cbDVGs were widely distributed along the genome, whereas the rejoin points were clustered in a narrow region close to 5’ end of the genome [25].
Further investigation into the molecular details of how the viral polymerase recognizes these signals may lead to important insights about the mechanism involved in RSV virus replication and the generation of cbDVGs. A lower density of nucleocapsid proteins (NPs) at certain genomic locations has been shown to result in increased cbDVG formation in SeV infection [26]. However, the mutations described to be responsible for low NP density were absent in our SeV stocks, suggesting that alternative mechanisms are likely involved. The usage of C nucleotides as a signal closely resembles the recognition of “gene end” or “gene start” by the viral polymerase when working on transcription [27, 28] and it would be intriguing to evaluate if the mechanisms of cbDVG generation and viral RNA transcription are related. Another factor influencing DVG accumulation is their length, which is tightly related to the spatial structure of the viral RNPs. In paramyxoviruses, although it is thought that “only genomes with hexametric or heptametric lengths are efficiently replicated” [29, 30], some cbDVGs generated in vitro do not obey this rule [25, 31, 32]. For RSV, we observed that a number of cbDVGs do not follow the rule of six or seven. Nonetheless, cbDVGs with certain length may have increased replication efficiency and thus an enhanced fitness advantage. Interestingly, in our minigenome system, although cbDVGs from Pair1 contained the expected rejoin point positions, break points frequently fell into a region further ahead of Break1, suggesting that the distance between the Break and Rejoin points may also play a role in determining where the break position is.
In addition to genomic sequences, other factors, such as viral proteins, likely play an important role in DVG generation. For instance, influenza viruses harboring a high fidelity polymerase generate fewer deletion DVGs [33]. Mutations in non-structural protein 2 of influenza have also been shown to increase the de novo generation of DVGs by altering the fidelity of viral polymerase [34]. Host factors may be essential contributors to DVG generation as well [10]. For example, vesicular stomatitis virus produces a large amount of snap-back DVGs in most cell lines, except human-mouse somatic cell hybrids, and this cellular attribute was mapped to human chromosome 16 [35]. Similarly, infection with measles virus did not show de novo generation of defective interfering particles (DIPs) in human WI-38 cells and SeV did not produce cbDVGs in chicken embryo lung cells [36, 37]. Despite the potential importance of these additional factors on DVG generation, the current work represents a major paradigm shift with the identification of sequences that regulate cbDVG formation.
Remarkably, we found various common cbDVGs present in more than one patient and at least one of those cbDVGs was also present in infections in vitro. These observations support a conserved origin for cbDVGs during infection and challenge the idea that DVGs occur as random product of virus replication. To date, all studies on DVG biology have been correlative in nature. This work opens up new areas of investigation and can ultimately allow us to manipulate the ability of viruses to produce DVGs as a powerful tool to study the role of DVGs in viral pathogenesis.
Studies of human samples were approved by University of Pennsylvania Institutional Review Board. The embryonated chicken eggs used in these studies were 10 days old and were obtained from Charles River.
A549 cells (human type II alveolar cells, ATCC, #CRM-CCL185) and HEp2 cells (HeLa-derived human epithelial cells, ATCC, CCL23) were cultured at 7% CO2 and 37°C with Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (FBS), 1 mM sodium pyruvate, 2 mM L-Glutamine, and 50 mg/ml gentamicin. BSR-T7 cells (Hamster kidney cells, BHK cells constitutively expressing the T7 polymerase, provided by Dr. Christopher Basler’s lab at Icahn School of Medicine) and were maintained in 10% FBS DMEM with 1 mg/ml Geneticin (Invitrogen). All cell lines were treated with mycoplasma removal agent (MP Biomedicals) and routinely tested for mycoplasma before use. Sendai virus Cantell stock (referred to as SeV HD, containing a high DVG particle content) was prepared in embryonated chicken eggs as described previously [7, 38]. The SeV HD stock used in these experiments had a high infectious to total particle ratio of 500:15,000. RSV-HD stocks 1–7 (stock of RSV derived from strain A2, ATCC, #VR-1540 with a high content of cbDVGs) were prepared and characterized as described previously [6, 39] in MAVS KO (3 lineages, stock1-3), STAT1 KO (3 lineages, stock4-6), and WT A549 cells (1 lineage, stock7), respectively. Briefly, RSV was fixed-volume passaged until stocks accumulated a high content of cbDVGs. The cell lines were kindly provided by Dr. Susan Weiss (University of Pennsylvania).
Mammalian expression vectors for RSV N (NR-36462), P (NR-36463), M2-1 (NR-36464), and L (NR-36461) proteins, and the RSV reverse genetic backbone pSynkRSV-line19F (rRSV-FR, NR-36460) were obtained from BEI Resources. Detailed information of the constructs can be found in reference [40]. The backbone plasmid of the RSV minigenome used for testing various DVG junction regions was constructed by cloning two regions of sequences amplified from pSynkRSV-line19F into the pSl1180 vector. The first region included a T7 promoter, a hammerhead ribozyme, RSV leader sequence, and genes encoding monomeric Katushka 2 (mKate2), while the second region included the RSV trailer sequence, a Hepatitis delta virus ribozyme and a T7 terminator. These regions were sequentially cloned into psl1180 vector using restriction enzyme pairs SpeI/SandI and SandI/EcoRI, respectively. The potential cbDVG break and/or rejoin regions (positions in S1 Table) were then inserted between those two regions using restriction enzyme pairs NotI/SandI and SandI/SpaI, respectively. A detailed scheme of the construct can be seen in Fig 2A. Pair1 and Rejoin1 mutations were introduced using the site-directed mutagenesis commercial kit QuickChange II XL (Agilent, CA) according to the manufacture’s protocol. All primers used for cloning are listed in S1 Table. Mutations in reverse genetic backbone pSynkRSV-line19F were generated by fusion PCR using primers in S1 Table as previously described [41].
Nasopharyngeal aspirates from pediatric patients were obtained from the Clinical Virology Laboratory of the Children’s Hospital of Philadelphia. All samples used were banked samples obtained as part of standard testing of patients. Samples were de-identified and sent to our lab for RNA extraction and cbDVG detection as indicated below.
Total RNA was extracted using TRIzol or TRIzol LS (Invitrogen) according to the manufacturer’s specifications. For detection of RSV DVGs in RSV infection, 1–2 μg of isolated total RNA was reverse transcribed with the DI1 primer using the SuperScript III reverse transcriptase (Invitrogen) without RNase H activity to avoid self-priming. Recombinant RNase H (Invitrogen) was later added to the reverse transcribed samples and incubated for 20 min at 37°C. DVGs were partially amplified using both DI1 primer and DI-R primer. The temperature cycle parameters used for the cbDVG-PCR in a BioRad C1000 Thermal Cycler were: 95°C for 10 min and 33–35 cycles of 95°C for 30 sec, 55°C for 30 sec and 72°C for 90 sec followed by a hold at 72°C for 5 min. Ramp rate of all steps was 3 degree/sec. Detailed method can be found in[6]. For detection of cbDVGs in the RSV minigenome system, extracted RNAs were treated with 2 μl TurboDNaseI (Invitrogen) for 15 min at 37°C, followed by reverse transcription. Same procedures as above were utilized, except replacing DI1 primer with DI-F1, DI-F2, and DI-F3 primers. These were then all paired with DI-R reverse primer to amplify the different sizes of PCR products. Sequences of all primers are listed in S1 Table.
Total RNA (1 μg) was reversed transcribed using the high capacity RNA to cDNA kit from Applied Biosystems. cDNA was diluted to a concentration of 10 μg/μl and amplified with specific primers in the presence of SYBR green (Applied Biosystems). qPCR reactions were performed in triplicate using specific primers and the Power SYBR Green PCR Master Mixture (Applied Biosystems) in a Viia7 Applied Biosystems Light-cycler. Gene expression levels of RSV G were normalized to the GAPDH copy number. Sequences of primers used in these studies can be found in S1 Table.
RNA-Seq for SeV Cantell and RSV HD stocks 1–6 were performed as previously described [42]. RNA was extracted using TRIzol reagent and was re-purified using the PicoPure RNA isolation kit (Thermo Fisher Scientific). RNA quality was assessed using the RNA Pico 6000 module on an Agilent Tapestation 2100 (Agilent Technologies) prior to cDNA library preparation. For SeV RNA-Seq dataset, total cDNA libraries were prepared starting from 75 ng (SeV Cantell) and 450 ng (RSV HD stocks) of extracted raw RNA using the Illumina TruSeq Stranded Total RNA LT kit with Ribo-Zero Gold, according to the manufacturer’s instructions. Samples were run on Illumina NextSeq 500 to generate 75 bp, single-end reads, resulting in 21–53 million reads per sample, with an average Q30 score ≥ 96.8%. For sequencing of samples from RSV-positive patients, including 4 DVG low patients and 6 DVG high patients, 100–450 ng of extracted raw RNA was used for preparation of cDNA library using the same kit as above. Samples were run on Illumina NextSeq 500 to generate 150bp, paired-end reads, resulting in 60–170 million reads/sample with average Q30 score ≥ 84.6%. To analyze genomic AU-content relative to DVG break and rejoin points, we calculated the percentage of A or U nucleotides over sliding windows of 40 bases using the Python programming language (Python Software Foundation, https://www.python.org/). We plotted AU-content and cbDVG rejoin points in R using the ggplot2 package [43].
Based on our in vitro RSV experiments, we made the assumption that most cbDVGs are generated from the viral sequence near the 5’ end region of the genome (close to the Trailer sequence). Therefore, starting with the last 3kb of a reference viral genome, we built an index of potential DVG sequences by taking all possible combinations of two non-overlapping segments of L bases, where L is the read length. The segments are linked by reverse complementing the second segment (C-D) and adding the first segment (A-B) to it (S6 Fig). Sequenced reads are aligned to the potential DVGs using bowtie2 [44], and subsequently undergo two filtering steps. First, reads are removed unless they map across a breakpoint (A_C) with at least 15bp of mapped segment on each side. Second, the reads that map cleanly to the reference genome are filtered out. This pipeline gives the output read counts for each breakpoint (A_C). To be consistent with the structure of copy-back DVGs in Fig 1A, A is equivalent to break point T’ and C is equivalent to rejoin point W. VODKA output reads were further aligned to reference viral genomes (RSV A2: NCBI accession number KT992094.1; RSV 2012 clinical isolate: NCBI accession number KJ672447) or known SeV DVG-546 to identify the potential DVG junction regions using the Geneious 7.0 software.
BHK cells constitutively expressing the T7 polymerase (BSR-T7 cells) were transfected with different minigenome constructs, gRSV-FR-WT, gRSV-FR-GC>Us, or gRSV-FR-AU>GCs as well as the sequence–optimized helper plasmids encoding N, P, M2-1, and L, all under T7 control as described previously [40]. Cells were incubated with transfection complex (total plasmid: lipofectamine = 1:3.3) for 2 h at room temperature and then at 37°C for overnight using Opti-MEM as medium. The following morning, the medium was replaced with antibiotic free tissue culture medium containing 2% FBS. For minigenome experiments, cells were harvested at 48 h post-transfection for either RNA extraction or flow cytometry. For mutant virus production, cells were maintained and split every 2–3 days until cytopathic effects (CPEs) were observed. Then viruses were collected and blindly passaged in HEp2 cells three times to obtain P3. P3 was titrated and passaged two more times at MOI of 10 to generate P4 and P5.
Transfected BSR-T7 cells were trypsinized 48 h post transfection and were either directly diluted in FACS buffer (PBS containing 2% FBS and 20 mM EDTA) or stained with aqua LIVE/DEAD. Cells were washed twice in FACS buffer before flow cytometry analysis on an LSRFortessa (Becton Dickinson). Data analysis was performed using Flowjo version Legacy.
All statistical analyses were performed with GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA) and R v3.4.1. A statistically significant difference was defined as a p-value <0.05 by one-way analysis of variance (ANOVA) with a post hoc test to correct for multiple comparisons (based on specific data sets as indicated in each figure legend).
The VODKA algorithm is open-source and available at: https://github.com/itmat/VODKA.
All data are available upon request to the corresponding author. Raw RNA-Sequencing data of FISH-FACS sorted SeV infected cells and RSV infected samples have been deposited on the Gene Expression Omnibus (GEO) database for public access (SeV: GSE96774; RSV: GSE114948).
|
10.1371/journal.pgen.1001372 | Genome-Wide Association Study Using Extreme Truncate Selection
Identifies Novel Genes Affecting Bone Mineral Density and Fracture
Risk | Osteoporotic fracture is a major cause of morbidity and mortality worldwide. Low
bone mineral density (BMD) is a major predisposing factor to fracture and is
known to be highly heritable. Site-, gender-, and age-specific genetic effects
on BMD are thought to be significant, but have largely not been considered in
the design of genome-wide association studies (GWAS) of BMD to date. We report
here a GWAS using a novel study design focusing on women of a specific age
(postmenopausal women, age 55–85 years), with either extreme high or low
hip BMD (age- and gender-adjusted BMD z-scores of +1.5 to +4.0,
n = 1055, or −4.0 to −1.5,
n = 900), with replication in cohorts of women drawn from
the general population (n = 20,898). The study replicates
21 of 26 known BMD–associated genes. Additionally, we report suggestive
association of a further six new genetic associations in or around the genes
CLCN7, GALNT3, IBSP, LTBP3, RSPO3, and
SOX4, with replication in two independent datasets. A novel
mouse model with a loss-of-function mutation in GALNT3 is also
reported, which has high bone mass, supporting the involvement of this gene in
BMD determination. In addition to identifying further genes associated with BMD,
this study confirms the efficiency of extreme-truncate selection designs for
quantitative trait association studies.
| Osteoporotic fracture is a major cause of early mortality and morbidity in the
community. To identify genes associated with osteoporosis, we have performed a
genome-wide association study. In order to improve study power and to address
the demographic group of highest risk from osteoporotic fracture, we have used a
unique study design, studying 1,955 postmenopausal women with either extreme
high or low hip bone mineral density. We then confirmed our findings in 20,898
women from the general population. Our study replicated 21 of 26 known
osteoporosis genes, and it identified a further six novel loci (in or nearby
CLCN7, GALNT3, IBSP, LTBP3, RSPO3, and
SOX4). For one of these loci, GALTN3, we
demonstrate in a mouse model that a loss-of-function genetic mutation in
GALNT3 causes high bone mass. These findings report novel
mechanisms by which osteoporosis can arise, and they significantly add to our
understanding of the aetiology of the disease.
| Osteoporotic fracture is a leading cause of morbidity and mortality in the community,
particularly amongst the elderly. In 2004 ten million Americans were estimated to
have osteoporosis, resulting in 1.5 million fractures per annum [1]. Hip fracture is
associated with a one year mortality rate of 36% in men and 21% in
women [2]; and the
burden of disease of osteoporotic fractures overall is similar to that of colorectal
cancer and greater than that of hypertension and breast cancer [3]. Bone mineral density (BMD) is
strongly correlated with bone strength and fracture risk, and its measurement is
widely used as a diagnostic tool in the assessment of fracture risk [4]–[6]. BMD is known to
be highly heritable, with heritability assessed in both young and elderly twins, and
in families, to be 60–90% [7]–[14]. Although the extent of
covariance between BMD and fracture risk is uncertain, of the 26 genes associated
with BMD at genome-wide significant levels to date, nine have been associated with
fracture risk (reviewed in [15]), supporting the use of BMD as an intermediate phenotype
in the search for genes associated with fracture risk.
There is considerable evidence from genetic studies in humans [12], [16], [17], and in mice [18], indicating
that the genes that influence BMD at different sites, and in the different genders,
overlap but are not identical. Thus far all genome-wide association studies (GWAS)
of BMD have studied cohorts of a wide age range, and with one exception have
included both men and women; when only women have been studied, both pre- and
postmenopausal women have been included. Therefore, to identify genes involved in
osteoporosis in the demographic at highest risk of osteoporotic fracture we have
performed a GWAS in postmenopausal women selected on the basis of their hip BMD, and
replicated the GWAS findings in a large cohort of adult women drawn from the general
population.
Considering markers previously reported as associated with BMD, our discovery dataset
replicates previously associated SNPs in 21 of the 26 genes reported to date to have
genome-wide significant associations (Table S6) (P<0.05, association in the same
direction as initially reported, or, in the case of LRP5 and
GPR177, with the next flanking SNP genotyped) [17], [21], [22], [23], [28], [32], [33], [34]. Replicated
genes include ARHGAP1, CTNNB1, ESR1, FAM3C, FLJ42280, FOXL1, GPR177, HDAC5,
JAG1, LRP5, MARK3, MEF2C, MEPE, OPG, RANK, RANKL, SOST, SOX6, SP7
(Osterix), STARD3NL and ZBTB40. Considering the
combined Anglo-Australasian Osteoporosis Genetics Consortium (AOGC) and
deCODE/TwinsUK/Rotterdam cohorts, 97 SNPs from six loci achieved
P<5×10−8 at the femoral neck (FN), of which four had
previously been reported (FLJ42280, MEF2C, SOX6, ZBTB40). At the
lumbar spine (LS), six SNPs from two known loci (RANKL, OPG)
achieved P<5×10−8. No support was seen for previously
reported associations involving SNPs in ADAMTS18, CRHR1, DCDC5,
MHC, or SBTBN1 (P>0.05).
This study also identifies and replicates two novel loci with confirmed association
with BMD in GALNT3 (MIM: 601756) and at chromosome 6q22 near
RSPO3 (MIM: 610574), and provides strong evidence of a further
four BMD-associated loci (CLCN7 (MIM: 602727),
IBSP (MIM: 147563), LTBP3 (MIM: 602090),
SOX4 (MIM: 184430)) (Table 1). Although these did not achieve
‘genome-wide significance’ in the discovery set alone, they achieved
P-values in the AOGC-discovery cohort of P<10−4, and support in
the AOGC-replication cohort, TwinsUK, Rotterdam and deCODE cohorts; and all have
additional evidence supporting their role in bone. Support was also seen for
TGFBR3 (MIM: 600742), a gene previously reported to have
suggestive association with BMD [33].
SNPs at chromosome 2q24, in and around GALNT3, achieved near
genome-wide significance in our discovery cohort (peak P-value rs1863196, total
hip (TH) P = 2.3×10−5; LS
P = 0.037) (Figure 1A). This SNP was not typed or imputed by either the
Rotterdam or the TwinsUK cohorts, but a nearby SNP showed strong association in
both AOGC and the combined replication cohorts (rs6710518; AOGC discovery, TH
P = 6.9×10−5; combined
replication sets, FN P = 2.7×10−7).
In the combined datasets the finding achieved genome-wide significance at the FN
(P = 1.7×10−10). Strong
association was also seen with this SNP at LS
(P = 7.5×10−5). Another marker
within GALNT3, rs4667492, was also associated with fracture
risk, including vertebral fractures (OR = 0.89;
95%CI = 0.80–0.99;
P = 0.032) and overall low trauma fractures
(OR = 0.92;
95%CI = 0.85–0.99;
P = 0.024).
We have recently identified a mouse with an
N-ethyl-N-nitrosourea induced
loss-of-function GALNT3 mutation (Trp589Arg), that develops
hyperphosphataemia with extraskeletal calcium deposition, and hence represents a
model for FTC [35]. To establish further the association of
GALNT3 and BMD, we determined BMD in these
GALNT3 mutant mice. This revealed that homozygous
(−/−) GALNT3 mutant male and female adult mice had
a higher areal BMD than their wild-type (+/+) litter mates, with
heterozygous (+/−) mice having intermediate BMD (Figure 2). This loss-of-function
GALNT3 mutation is predicted to lead to a reduced
glycosylation of FGF23, which increases its breakdown and leads to reduced serum
FGF23 concentrations [35].
A novel genome-wide significant association was also seen at markers on
chromosome 6q22-23 (Figure
1B). In the combined dataset, marker rs13204965 achieved genome-wide
significance at this locus at the FN
(P = 2.2×10−9), with strong
support in both the AOGC discovery set, and the combined replication sets
(AOGC-discovery, TH P = 2.1×10−4;
combined replication P = 3.5×10−5).
Strong association was also seen with LS BMD (rs13204965
P = 0.00067). The peak of association at this locus lies
within a cDNA fragment, AK127472. The nearest gene, RSPO3
(R-spondin-3), is 275 kb telomeric of the strongest associated SNP, but is
within the associated linkage disequilibrium region (Figure 1B).
Association was observed at chromosome 16p13 with SNPs in and around
CLCN7, which encodes a
Cl−/H+ antiporter expressed primarily in
osteoclasts, and critical to lysosomal acidification, an essential process in
bone resorption. Peak association at this locus was seen with SNP rs13336428 in
the discovery set (TH P = 7.0×10−4;
LS P = 0.028) (Figure S3A), which was confirmed in the
replication set (FN P = 3.6×10−5; LS
P = 0.00012), achieving
P = 1.7×10−6 at the FN and
1.2×10−5 at LS in the overall cohort. Association has
previously been reported between two SNPs in exon 15 of CLCN7
(rs12926089, rs12926669) and FN BMD (P = 0.001–0.003)
[36];
no association was seen with either of those SNPs in the current study (P>0.4
at FN and LS).
Association was observed with SNPs in IBSP (integrin-binding
bone sialoprotein) (Figure S3B), encoded at chromosome 4q22, a
gene which has previously had suggestive association reported with BMD in two
studies (rs1054627, Styrkarrsdottir et al
P = 4.6×10−5
[22];
Koller et al P = 1.5×10−4
[37]). In the
current study, moderate association was observed in the discovery set with the
same SNP as previously reported (rs1054627, AOGC discovery TH,
P = 6.6×10−5), with support in
the replication set and strong association overall (FN combined replication
P = 9.2×10−5; FN overall
association P = 7.6×10−7). Nominal
association was observed at LS (rs1054627, P = 0.019).
Association with BMD was also seen at chromosome 11p13, with SNP rs1152620
achieving P = 4.4×10−5 (TH) in the
discovery set, P = 0.0051 (FN) in the replication set, and
P = 3.6×10−4 overall (Figure
S3C). This SNP was also nominally associated with LS BMD in the discovery
set (P = 0.041). The nearest gene to this locus is
LTBP3 (latent transforming growth factor beta binding
protein 3), which is located 292 kb q-telomeric of rs1152620.
At chromosome 6p22, SNPs in and around SOX4 (Sex determining
region Y box 4) were moderately associated with BMD in our discovery set (most
significant association rs9466056, TH
P = 5.3×10−4; LS
P = 0.0036) (Figure S3D), with support at the hip and LS
in the replication set (FN P = 0.00013, LS
P = 0.013), achieving association overall with
P = 2.6×10−7 (FN) and
P = 0.00081 (LS).
This study demonstrates convincing evidence of association with six genes with BMD
variation, GALNT3, RSPO3, CLCN7, IBSP, LTBP3 and
SOX4. Using a moderate sample size, the use of a novel study
design also led to the confirmation of 21 of 26 known BMD-associations. This study
thus demonstrates the power of extreme-truncate selection designs for association
studies of quantitative traits.
GALNT3 encodes N-acetylgalactosaminyltransferase 3, an enzyme
involved in 0-glycosylation of serine and threonine residues. Mutations of
GALNT3 are known to cause familial tumoral calcinosis (FTC,
OMIM 2111900) [38]
and hyperostosis-hyperphosphataemia syndrome (HOHP, OMIM 610233) [39]. FTC is
characterised by hyperphosphataemia in association with the deposition of calcium
phosphate crystals in extraskeletal tissues; whereas in HOHP, hyperphosphataemia is
associated with recurrent painful long bone swelling and radiographic evidence of
periosteal reaction and cortical hyperostosis. FGF23 mutations
associated with FTC cause hyperphosphataemia through effects on expression of the
sodium-phosphate co-transporter in the kidney and small intestine, and through
increased activation of vitamin D due to increased renal expression of CYP27B1
(25-hydroxyvitamin-D 1 alpha hydroxylase) [40]. It is unclear whether FGF23
has direct effects on the skeleton or if its effects are mediated through its
effects on serum phosphate and vitamin D levels. FGF23 signals via a complex of an
FGF receptor (FGFR1(IIIc)) and Klotho [41]; mice with a loss-of-function
mutation in Klotho develop osteoporosis amongst other
abnormalities, and modest evidence of association of Klotho with
BMD has been reported in several studies [42], [43], [44], [45]. We saw no association with
polymorphisms in Klotho and BMD in the current study (P>0.05 for
all SNPs in and surrounding Klotho). To our knowledge, this finding
is the first demonstration in humans that genetic variants in the FGF23 pathway are
associated with any common human disease.
RSPO3 is one of four members of the R-spondin family (R-spondin-1 to
−4), which are known to activate the Wnt pathway, particularly through effects
on LRP6, itself previously reported to be BMD-associated [46], [47]. LRP6 is inhibited by the
proteins Kremen and DKK1, which combine to induce endocytosis of LRP6, reducing its
cell surface levels. R-spondin family members have been shown to disrupt
DKK1-dependent association of LRP6 and Kremen, thereby releasing LRP6 from this
inhibitory pathway [48]. R-spondin-4 mutations cause anonychia (absence or severe
hypoplasia of all fingernails and toenails, OMIM 206800) [49]. No human disease has been
associated with R-spondin-3, and knockout of R-spondin-3 in mice is embryonically
lethal due to defective placental development [50].
Mutations of CLCN7 cause a family of osteopetroses of differing age
of presentation and severity, including infantile malignant
CLCN7-related recessive osteopetrosis (ARO), intermediate autosomal
osteopetrosis (IAO), and autosomal dominant osteopetrosis type II (ADOII,
Albers-Schoenberg disease). These conditions are characterized by expanded, dense
bones, with markedly reduced bone resorption. Our data support associations of
polymorphisms at this locus with BMD variation in the population.
IBSP is a major non-collagenous bone matrix protein involved in calcium and
hydroxyapatite binding, and is thought to play a role in cell-matrix interactions
through RGD motifs in its amino acid sequence. IBSP is expressed in all major bone
cells including osteoblasts, osteocytes and osteoclasts; and its expression is
upregulated in osteoporotic bone [51]. IBSP knockout mice have low cortical
but high trabecular bone volume, with impaired bone formation, resorption, and
mineralization [52]. IBSP lies within a cluster of genes
including DMP1, MEPE, and SPP1, all of which have
known roles in bone and are strong candidate genes for association with BMD.
MEPE has previously been associated with BMD at genome-wide
significance [17]. In the current study the strongest association was seen
with an SNP in IBSP, rs1054627, as was the case with two previous
studies [22], [37]. Linkage disequilibrium between this SNP, and the
previously reported BMD-associated SNP rs1471403 in MEPE, is modest
(r2 = 0.16). Whilst out study supports the
association of common variants in IBSP in particular with BMD,
further studies will be required to determine if more than one of these genes is
BMD-associated.
Recessive mutations of LTBP3 have been identified as the cause of
dental agenesis in a consanguineous Pakistani family (OMIM 613097) [53]. Affected family
members had base of skull thickening, and elevated axial but not hip BMD.
LTBP3−/− mice develop axial osteosclerosis with
increased trabecular bone thickness, as well as craniosynostosis [54]. LTBP3 is
known to bind TGFβ1, -β2 and -β3, and may influence chondrocyte
maturation and enchondral ossification by effects on their bioavailability [54].
Our study also confirms the previously reported association of another TGF pathway
gene, TGFBR3, encoded at chromosome 1p22, with BMD [33] (Figure S3E). In
that study, association was observed in four independent datasets, but overall the
findings did not achieve genome-wide significance at any individual SNP (most
significant SNP rs17131547,
P = 1.5×10−6). In our discovery set,
peak association was seen at this locus with SNP rs7550034 (TH
P = 1.5×10−4), which lies 154 kb
q-telomeric of rs17131547, but still within TGFBR3 (rs17131547 was
not typed or imputed in our dataset) (Figure S3E). This supports
TGFBR3 as a true BMD-associated gene.
This study also demonstrated that SOX4 polymorphisms are associated
with BMD variation. Both SOX4 and SOX6 are
cartilage-expressed transcription factors known to play essential roles in
chondrocyte differentiation and cartilage formation, and hence endochondral bone
formation. SOX6 has previously been reported to be BMD-associated
at genome-wide significant levels [17]. Whilst
SOX4−/− mice develop severe cardiac abnormalities
and are non-viable, SOX4+/− mice have osteopaenia with
reduced bone formation but normal resorption rates, and diminished cortical and
trabecular bone volume [55]. Our data suggest that SOX4
polymorphisms contribute to the variation in BMD in humans.
This study has a unique design amongst GWAS of BMD reported to date, using an
extreme-truncate ascertainment scheme, focusing on a specific skeletal site (TH),
and with recruitment of a narrow age- and gender-group (post-menopausal women age
55–85 years). Our goal in employing this scheme was to increase the study
power by reducing heterogeneity due to age-, gender- and skeletal site-specific
effects. Whilst osteoporotic fracture can occur at a wide range of skeletal sites,
hip fracture in postmenopausal women is the major cause of morbidity and mortality
due to osteoporosis. To date, with only one exception, all GWAS of BMD have studied
cohorts unselected for BMD [28], and no study has restricted its participants to
postmenopausal women ascertained purely on the basis of hip BMD. Assuming
marker-disease-associated allele linkage disequilibrium of
r2 = 0.9, for
alpha = 5×10−8 our study has
80% power to detect variants contributing 0.3% of the additive genetic
variance of BMD. An equivalent-powered cohort study would require ∼16,000
unselected cases.
Considering the 26 known genes (or genomic areas) associated with BMD, P-values less
than <0.05 were seen in our discovery for 21 of the BMD-associated SNPs. Of the
26 known BMD genes, 16 would have been included in our replication study on the
basis of the strength of their BMD association in our discovery cohort, but were not
further genotyped as they were known already to be BMD-associated. Had these 16
genes replicated, 22 genes would have been identified in this single study,
demonstrating the power of the design of the current study.
A potential criticism of studies of highly selected cohorts, such as the
AOGC-discovery cohort, is that the associations identified may not be relevant in
the general population. However, the confirmation of our findings in replication
cohorts of women unselected for BMD confirms that our findings are of broad
relevance.
In summary, our study design therefore represents a highly efficient model for future
studies of quantitative traits and is one of the first reported studies using an
extreme truncate design in any disease. We have identified two new BMD loci at
genome-wide significance (GALNT3, RSPO3), with
GALNT3 SNPs also associated with fracture. Strong evidence was
also demonstrated for four novel loci (CLCN7, IBSP, LTBP3,
SOX4). Further support was also provided that TGFBR3
is a true BMD-associated locus. Our discovery cohort replicated 21 of 26 previously
identified BMD-associated loci. Our novel findings further advance our understanding
of the aetiopathogenesis of osteoporosis, and highlight new genes and pathways not
previously considered important in BMD variation and fracture risk in the general
population. Our study also provides strong support that the use of extreme truncate
selection is an efficient and powerful approach for the study of quantitative
traits.
All participants gave written, informed consent, and the study was approved by
the relevant research ethics authorities at each participating centre.
The discovery sample population included 1128 Australian, 74 New Zealand and 753
British women, between 55–85 years of age, five or more years
postmenopausal, with either high BMD (age- and gender-adjusted BMD z-scores of
+1.5 to +4.0, n = 1055) or low BMD (age- and
gender-adjusted BMD z-scores of −4.0 to −1.5,
n = 900) (Tables S1 and S2). BMD
z-scores were determined according to the Geelong Osteoporosis Study normative
range [19]. Low
BMD cases were excluded if they had secondary causes of osteoporosis, including
corticosteroid usage at doses equivalent to prednisolone ≥7.5 mg/day for
≥6 months, past or current anticonvulsant usage, previous strontium usage,
premature menopause (<45 years), alcohol excess (>28 units/week), chronic
renal or liver disease, Cushing's syndrome, hyperparathyroidism,
thyrotoxicosis, anorexia nervosa, malabsorption, coeliac disease, rheumatoid
arthritis, ankylosing spondylitis, inflammatory bowel disease, osteomalacia, and
neoplasia (cancer, other than skin cancer). Screening blood tests (including
creatinine (adjusted for weight), alkaline phosphatase, gamma-glutamyl
transferase, 25-hydroxyvitamin D and PTH) were checked in 776 cases, and no
differences were found between the high and low BMD groups. Therefore no further
screening tests were done of the remaining cases.
Fracture data were analysed comparing individuals who had never reported a
fracture after the age of 50 years, with individuals who had had a low or
non-high trauma (low trauma fracture = fracture from a
fall from standing height or less) osteoporotic fracture (excluding skull, nose,
digits, hand, foot, ankle, patella) after the age of 50 years. Vertebral, hip
and non-vertebral fractures were considered both independently and combined.
All participants were of self-reported white European ancestry.
DNA was obtained from peripheral venous blood from all cases except those
recruited from New Zealand, for whom DNA was obtained from salivary samples
using Oragene kits (DNA Genotek, Ontario, Canada). We have previously
demonstrated that DNA from these two sources have equivalent genotyping
characteristics [20].
After quality control checks including assessment of cryptic relatedness,
ethnicity and genotyping quality, 900 individuals with low TH BMD and 1055
individuals with high TH BMD were available for analysis.
The replication cohort consisted of 8928 samples drawn from nine cohort studies,
outlined in Tables S3 and S4 (‘AOGC replication cohort’)
which were directly genotyped, These replication cases were adult women (age
20–95 years), unselected with regard to BMD, and who were not screened for
secondary causes of osteoporosis. Replication was also performed in silico in
11,970 adult women from the TwinsUK and Rotterdam, and deCODE Genetics GWASs
[21],
[22], [23], in which
association data were available at LS and FN.
High and low BMD ascertainment was defined according to the TH score, because
this has better measurement precision than FN BMD [24]. However, neither TwinsUK
nor the Rotterdam Study had TH BMD on the majority of their datasets and
therefore were analysed using the FN measurement for which data were available
on the whole cohort. All replication findings at the hip are reported therefore
for FN BMD. TH and FN BMD are closely correlated (r = 0.882
in the AOGC dataset), with FN BMD one of the components of the TH BMD
measurement.
Genotyping of the discovery cohort (n = 2036) was performed
using Illumina Infinium II HumHap300 (n = 140), 370CNVDuo
(n = 4), 370CNVQuad (n = 1882) and
610Quad (n = 10) chips at the University of Queensland
Diamantina Institute, Brisbane, Australia. Genotype clustering was performed
using Illumina's BeadStudio software; all SNPs with quality scores <0.15
and all individuals with <98% genotyping success were excluded. 289499
SNPs were shared across all chip types. Cluster plots from the 500 most strongly
associated loci, were manually inspected and poorly clustering SNPs excluded
from analysis. Following imputation using the HapMap Phase 2 data, 2,543,887
SNPs were tested for association with TH and LS BMD (Manhattan plot of
association findings, Figure S1). After data cleaning, minimal
evidence of inflation of test statistics was observed, with a genomic inflation
factor (λ) of 1.0282 (qq plot, Figure S2).
A total of 124 SNPs were successfully genotyped in the AOGC replication cohort.
These replication study SNPs were selected from the findings of the discovery
cohort, either based on the strength of association (P-value) or following
analysis with GRAIL (n = 45) [25], using as seed data
all SNPs previously reported to be associated with BMD at GWAS significant
levels (results for all replication SNPs presented in Table S5).
GRAIL is a bioinformatic program that assesses the strength of relationships
between genes in regions surrounding input SNPs (usually derived from genetic
association studies) and other SNPs or genes associated with the trait of
interest, by assessing their co-occurrence in PubMed abstracts. Where genes
surrounding input SNPs occur more frequently in abstracts with known associated
genes, these SNPs are more likely themselves also to be associated, and can thus
be prioritized for inclusion in replication studies.
For the replication study, genotyping was performed either by Applied Biosystems
OpenArray (n = 113) or Taqman technology
(n = 11) (Applied Biosystems, Foster City, CA, USA),
according to the manufacturer's protocol.
Eleven individuals were removed because of abnormal X-chromosome homozygosity
(X-chromosome homozygosity either <−0.14, or >+0.14). Outliers
with regard to autosomal heterozygosity (either <0.34225 or >0.357,
n = 40) and missingness (>3%,
n = 4) were removed. Using an IBS/IBD analysis in PLINK to
detect cryptic relatedness, one individual from 35 pairs of individuals with
pi-hat >0.12 (equivalent to being 3rd degree relatives or closer)
were removed. SNPs with minor allele frequency <1%
(n = 561), and those not in Hardy-Weinberg equilibrium
(P<10−7, n = 170) were then
removed, leaving 288,768 SNPs in total. Nine replication SNPs were removed
because of excess missingness (>10%) or because they failed tests of
Hardy-Weinberg equilibrium (P<0.001).
To detect and correct for population stratification EIGENSTRAT software was used.
We first excluded the 24 regions of long range LD including the MHC identified
in Price et al. before running the principal components analysis, as suggested
by the authors [26]. Sixteen individuals were removed as ethnic outliers,
leaving 1955 individuals in the final discovery dataset.
Imputation analyses were carried out using Markov Chain Haplotyping software
(MaCH; http://www.sph.umich.edu/csg/abecasis/MACH/) using phased data
from CEU individuals from release 22 of the HapMap project as the reference set
of haplotypes. We only analyzed SNPs surrounding disease-associated SNPs that
were either genotyped or could be imputed with relatively high confidence
(R2≥0.3). For TH measurements, a case-control association
analysis of imputed SNPs was performed assuming an underlying additive model and
including four EIGENSTRAT eigenvectors as covariates, using the software package
MACH2DAT [27]
which accounts for uncertainty in prediction of the imputed data by weighting
genotypes by their posterior probabilities. For FN and LS BMD analyses,
Z-transformed residual BMD scores (in g/cm2) were generated for the
entire AOGC cohort after adjusting for the covariates age, age2, and
weight, and for centre of BMD measurement. Because the regression coefficient
for BMD on genotype would be biased by selection for extremes, we adopted the
approach detailed in Kung et al (2009) [28]. Specifically, the regression
coefficient of genotype on BMD was estimated, and subsequently transformed to
the regression coefficient of BMD on genotype through knowledge of the
population variance of the phenotype and the allele frequencies. For fracture
data, analysis was by logistic regression. Only SNPs achieving GWAS significance
were tested for fracture association. The SNPs used for replication from the
Rotterdam Study were analyzed using MACH2QTL implemented in GRIMP [29]. Data from
the discovery and replication cohorts were combined using the inverse variance
approach as implemented in the program METAL [30].
SNPs associated with BMD were also tested for association with fracture in the
AOGC discovery and replication cohorts (hip, vertebral, nonvertebral, and all
low trauma fractures, age ≥50 years, as defined above), by logistic
regression.
Study power was calculated using the ‘Genetic Power Calculator’ [31].
All animal studies were approved by the MRC Harwell Unit Ethical Review Committee
and are licensed under the Animal (Scientific Procedures) Act 1986, issued by
the UK Government Home Office Department. Dual-energy X-ray absorptiometry
(DEXA) was performed using a Lunar Piximus densitometer (GE Medical Systems) and
analysed using the Piximus software.
Data related to this study will be available to research projects approved by a
Data Access Committee including representatives of the University of Queensland
Research Ethics Committee. For enquiries regarding access please contact the
corresponding author, MAB ([email protected]).
|
10.1371/journal.ppat.1003911 | The Major Cellular Sterol Regulatory Pathway Is Required for Andes Virus Infection | The Bunyaviridae comprise a large family of RNA viruses with worldwide distribution and includes the pathogenic New World hantavirus, Andes virus (ANDV). Host factors needed for hantavirus entry remain largely enigmatic and therapeutics are unavailable. To identify cellular requirements for ANDV infection, we performed two parallel genetic screens. Analysis of a large library of insertionally mutagenized human haploid cells and a siRNA genomic screen converged on components (SREBP-2, SCAP, S1P and S2P) of the sterol regulatory pathway as critically important for infection by ANDV. The significance of this pathway was confirmed using functionally deficient cells, TALEN-mediated gene disruption, RNA interference and pharmacologic inhibition. Disruption of sterol regulatory complex function impaired ANDV internalization without affecting virus binding. Pharmacologic manipulation of cholesterol levels demonstrated that ANDV entry is sensitive to changes in cellular cholesterol and raises the possibility that clinically approved regulators of sterol synthesis may prove useful for combating ANDV infection.
| As obligate, intracellular parasites viruses are dependent upon the host cell for numerous factors and processes. However, for many important viruses few of the required host factors have been identified. Hantaviruses are rodent-borne viruses that are associated with severe human disease. Transmission to humans occurs sporadically with a recent notable example in Yosemite National park. In the present study, we utilized two independent genetic strategies to discover cellular factors needed for replication of the highly pathogenic hantavirus Andes virus. We found that four genes, encoding components of a complex involved in regulation of cholesterol synthesis and uptake, were critical for Andes virus infection. Drugs that inhibit an enzyme in this complex or that reduce cellular cholesterol levels effectively blocked Andes virus infection, suggesting new ways for combating this pathogenic virus.
| Hantaviruses are a genera of the Bunyaviridae family that includes a large number of human pathogens. Hantaviruses found in the Americas, the so called New World hantaviruses, including Andes virus (ANDV) from Argentina and Chile, can cause a lethal hemorrhagic fever known as hantavirus pulmonary syndrome (HPS) while the Old World hantaviruses from Europe and Asia are associated with Hemorrhagic Fever with Renal Syndrome (HFRS) [1]–[5]. Unlike other members of the Bunyaviridae family, ANDV and the other hantaviruses are not transmitted by arthropod vectors but instead infect humans directly by aerosolized excreta from infected rodents. Entry into host cells by the membrane enveloped hantaviruses depends upon the viral glycoproteins GN and GC, which form a heterodimeric complex on the virion surface following cleavage of a polyprotein precursor [6]–[8]. Although it is clear that hantaviral infection relies upon transit to an acidic intracellular compartment where the viral glycoproteins mediate membrane fusion [9], [10], the overall entry process is not fully elucidated.
As with other viruses, ANDV must utilize host cell molecules and pathways during the virus life cycle for replication to occur. However relatively little is known about how ANDV, or other hantaviruses, interact with their host cells. High-throughput genetic screens have changed the way viral host co-factors are identified since these approaches have the ability to reveal not only host cell molecules that directly interact with viral components to facilitate virus infection, but also the cellular pathways that orchestrate the expression and activity of these molecules. Identifying pathways rather than individual molecules that are needed for virus replication could lead to the development of multiple therapeutic targets. Moreover, pathways used in common by multiple viruses within a family would represent ideal candidates for therapeutic development.
To identify cellular factors and pathways important for hantavirus replication, we employed two genetic screens: a haploid human cell line that was insertionally mutagenized with a gene-trap vector and a large-scale siRNA screen. A recombinant vesicular stomatitis virus (VSV) recombinant in which the ANDV glycoproteins are expressed on a VSV core (rVSV-ANDV [11]) focused our screening efforts on cellular processes involved in early steps of the ANDV infectious pathway. Key findings were confirmed with replication competent, wild-type ANDV. These independent genetic screens identified members of the major cellular cholesterol regulatory pathway as important for ANDV entry. Inhibiting this pathway using complementary genetic and pharmacologic approaches demonstrated that ANDV is exquisitely sensitive to the cellular levels of cholesterol. Decreased cellular cholesterol blocked ANDV infection at the level of virus entry. Despite normal binding to the cell surface, virus failed to be internalized, resulting in a profound block to infection. Overall these studies provide a framework with which to identify additional cellular components involved in the entry of ANDV, and potentially other hantaviruses, and raise the possibility that approved inhibitors of sterol regulation and synthesis may find clinical application in treating ANDV infection.
As one approach to identify human genes required for ANDV entry we employed an insertional mutagenesis strategy (Figure 1B) in the human haploid cell line (HAP1, [12]). Approximately one billion HAP1 cells were transduced with a gene-trapping vector, LentiET (Lentiviral Exon Trap, Figure 1A), to generate a library of cells with insertionally-inactivated genes. Survival of parental and mutagenized HAP1 cells was selected for in parallel with either replication competent recombinant Vesicular Stomatitis Virus that uses its endogenous glycoprotein (rVSV-G) or a replication competent VSV enveloped with the glycoprotein of ANDV (rVSV-ANDV). Challenge of ∼75 million LentiET mutagenized HAP1 cells with rVSV-ANDV produced hundreds of colonies while parental HAP1 cells yielded no surviving colonies when infected with rVSV-G or rVSV-ANDV. In contrast to the results with rVSV-ANDV, no mutagenized HAP1 cells survived selection with rVSV-G. Since the mutagenized cells specifically survived rVSV-ANDV infection, the infection resistance maps to the ANDV glycoprotein and not to the replication of the VSV core.
Mutagenized HAP1 cells surviving rVSV-ANDV selection were pooled and an aliquot was tested for sensitivity to rVSV-ANDV and rVSV-G infection. Confirming the results of the screen, the selected population displayed significant resistance to rVSV-ANDV infection while remaining sensitive to rVSV-G (Figure S1). Genomic DNA prepared from pooled rVSV-ANDV-resistant cells was used to map pLentiET genetic integration sites. In total, 676 independent integrations sites were mapped in the human genome with 80% being intragenic (Table S1). Of these insertions, 253 (37%) were located within four genes of a sterol regulatory element-binding protein pathway: Sterol Regulatory Element Binding Protein 2 (SREBF2; 59 insertions), Sterol Regulatory Element Binding Protein Cleavage Activating Protein (SCAP; 62 insertions), Site 1 Protease (S1P; 62 insertions), and Site 2 Protease (S2P; 70 insertions), (Figure 1C). These insertional frequencies were compared to the insertional frequency within the HAP1 library pre-virus selection. As the p-values indicate (Figure 1C, all less than 1×10−60), it is highly unlikely for these genes to have been enriched by chance. Furthermore, no other genes contained more than 2 integrations within the selected population (Table S1). In addition, integrations within SCAP and S1P highly favor the orientation where the gene trap vector effectively “captures” the transcript whereas the opposite is true for the SREBF2 gene (Figure 1C, Table S1).
In parallel an RNAi screen was performed using an optimized high throughput luciferase-based assay in a human HEK293T cell line engineered to constitutively express Firefly Luciferase (ffLuc). The Ambion Druggable Genome library (9,102 genes) was employed in a 384-well format with 4 siRNAs per gene and 2 siRNAs per well. Infection was assessed using a non-replicating ANDV (VSV-(ANDV)) viral pseudotype system in which the native glycoprotein (VSV-(G)) was replaced with a Renilla Luciferase (rLuc) reporter and the ANDV glycoprotein was provided in trans [13]. This assay exhibited a linear correlation between rLuc activity and the amount of input virus across a 4-log titration (data not shown). Likewise, the ffLuc signal derived from the HEK293T cell line correlated with cell number across a broad range and so could be used to assess cell viability in each well (Figure S2). The library was reverse-transfected into the HEK293T-ffLuc cells and 72 hours post-transfection cells were infected with VSV-(ANDV). Twenty-four hours post-infection, relative light units (RLU) for cell viability (firefly) and infection (Renilla) were measured. For each plate, robust Z scores were calculated for both cell viability and infection. Genes were identified as hits with a robust Z score for infection of <−1.5 in both siRNA pools (p<0.009). This stringent approach requires that at least 2 unique siRNAs against the gene of interest impacted infection. siRNAs were considered cytotoxic and excluded if they had a robust Z score for viability <−2 in both pools. Based on these criteria, 105 genes were identified as important for infection (Table S2A).
To validate the genes identified in the RNAi screen and to differentiate genes important for ANDV glycoprotein-mediated entry from those related to replication of the VSV core, 3 additional, unique siRNAs targeting 96 of the initial hits were screened using both ANDV and VSV-(G) viral pseudotypes (Table S2A). Genes validated if at least one additional siRNA inhibited infection in at least two biological replicates of the secondary screen, using a cut-off of a robust Z score for infection of <−1.3 (p<0.05) with no cytotoxicity (robust Z score>−2). Thirty-three genes met these criteria, with 9 specific for ANDV glycoprotein-mediated entry (Table 1, Table S2B). Comparison of these results with the haploid cell screen revealed that SREBF2 was the only gene in common, making this a strong candidate since it influenced infection in two different screens.
The results obtained from the siRNA and haploid cell screens indicated that components of the sterol regulatory pathway (SREBF2, SCAP, S1P, and S2P) were required for rVSV-ANDV infection. To probe the importance of this pathway for recombinant ANDV infection, a panel of well-characterized Chinese Hamster Ovary (CHO) cell lines individually null for S1P, S2P, or SCAP (Figure S3) [14]–[16] were challenged with VSV pseudotypes bearing the VSV or ANDV glycoproteins (Figure 2A). Additionally, these cell lines were infected with VSV pseudotypes carrying the glycoproteins from an Old World hantavirus, Hantaan virus (HTNV). rVSV-ANDV infection was severely impaired showing a 1–2 log decrease in each of the mutant CHO cell lines, whereas the infection level by viral pseudotypes bearing the VSV-(G) glycoprotein was similar to the parental CHO cells. Infection with VSV-(HTNV) decreased by roughly 1-log in cells null for S1P and SCAP, but not S2P, suggesting a more modest dependence on this pathway. A stable cell line lacking SREBF2 was not available, therefore SREBP-2 expression was knocked-down with two independent siRNAs in HEK293T cells (Figure 2B). Knockdown of SREBP-2 expression resulted in a significant (p<0.05) decrease in VSV-(ANDV) infection with no significant impact on VSV-G or VSV-(HTNV) infection (Figure 2C).
To explore the importance of the cholesterol regulatory complex for ANDV glycoprotein-dependent infection in human cells we developed a Transcription Activator Like Exonuclease (TALEN) pair that disrupted the coding region of SCAP (TALENSCAP). HEK293T cells transfected with TALENSCAP were expanded and infected with rVSV-ANDV to kill susceptible cells (Figure 2D). SCAP disruption was quantified pre- and post-infection using a quantitative PCR heteroduplex cleavage assay [17]. The heteroduplex assay revealed that following TALENSCAP transfection, ∼3% of the population had evidence of gene disruption. After infection with rVSV-ANDV, the SCAP-disrupted population increased to ∼43%. Sequence analysis confirmed the gene disruption (Figure S4). Enrichment for the disrupted SCAP gene was not observed in cells passaged without virus infection (Figure 2D). Wild-type VSV efficiently infected and killed the SCAP disrupted cells (data not shown). Taken together, these three independent experimental approaches provide strong genetic evidence that members of this sterol regulatory complex are required for efficient ANDV glycoprotein-mediated infection in diverse cell types and hosts.
PF-429242 is a reversible, competitive inhibitor of S1P that blocks cleavage and subsequent of SREBP-2 and has been shown to reduce the rates of cholesterol synthesis in cultured cells and in mice [18]. Human airway epithelia derived A549 cells were pretreated with varying concentrations of PF-429242 for 24 hours to allow turnover of activated SREBP-2 and subsequently infected with VSV-(G), rVSV-ANDV or VSV-(HTNV) (Figure 3A). As expected, treatment with PF-429242 caused a dose-dependent reduction in the levels of total cellular cholesterol (Figure S5). rVSV-ANDV infectivity also decreased in a dose-dependent manner with an approximately 50-fold reduction at 20 µM PF-429242 (dashed line). PF-429242 had an intermediate effect on the infectivity of VSV-(HTNV) (dotted line). In contrast, VSV-(G) infection of A549 cells is only modestly inhibited in the presence of PF-429242 (Figure 3A; continuous line) at this concentration (20 µM), however it is increasingly abrogated at concentrations ≥40 µM (data not shown). Inhibition of rVSV-ANDV by PF-429242 was not due to delayed viral entry kinetics since the vast majority of fusion had occurred within the first 3 hours in the presence or absence of PF-429242 (Figure S6).
To address whether statins, a clinically approved class of cholesterol lowering drugs, could significantly reduce rVSV-ANDV infectivity, we selectively inhibited cholesterol synthesis with the HMG-CoA reductase inhibitor mevastatin 24 hours prior to infection in delipidated growth medium. Pretreatment of human A549 cells with 1.25 µM mevastatin reduced rVSV-ANDV infectivity (>10-fold, Figure 3B). Controls included a VSV-Sindbis virus pseudotype (VSV-(SINV)) known to be sensitive to sterol levels, and vaccinia virus which is not affected by cellular cholesterol [19], [20]. VSV-(SINV) and rVSV-ANDV displayed a dose-dependent effect of mevastatin on infection, with rVSV-ANDV appearing significantly (p<0.05) more sensitive than VSV-(SINV) (Figure 3B). Infection by VSV-(G) also displayed a dose-dependent decrease compared to untreated cells; however the magnitude of this decrease was significantly less pronounced compared to rVSV-ANDV (Figure 3B). As anticipated, infection by a cholesterol-independent virus, vaccinia, was unaffected by mevastatin treatment (Figure 3B). Supplementation of media with mevalonate and sera reversed the inhibitory effect of mevastatin on rVSV-ANDV infection (Figure 3C). This complementation was likely the result of a combination of mevalonate uptake and LDL scavenging from this rich FBS, since it is expected that the inhibition of cholesterol synthesis would enhance expression of the LDL receptor. The increase in rVSV-ANDV infection under these conditions is consistent with the observation that cholesterol levels in statin-treated cells grown in FBS supplemented with mevalonate rebound to near wild-type levels (data not shown). Overall, VSV-G-dependent infectivity is largely unaffected (<2-fold) by 20 µM PF-429242, or in the loss of SCAP, S1P, or S2P, or depleted SREBP-2, yet mevastatin does appear to have a greater impact on VSV-(G) infectivity. This effect may be the result of LDL-R surface expression dynamics as this family of proteins have been shown to act as VSV-(G) receptors [21].
Collectively, the data presented thus far have established that the cholesterol regulatory pathway leading to SREBP-2 cleavage is required for infection by viruses bearing the ANDV glycoproteins. Because this pathway regulates a number of genes required for sterol biosynthesis and internalization as well as cholesterol production within the cell, we wished to investigate whether cellular cholesterol levels are important for ANDV glycoprotein mediated infection. To this end, cells were treated with methyl-β-cyclodextrin (MβCD), which extracts sterols from membranes [22]. Additionally, wild-type ANDV (strain 9717869) was employed for this analysis. MβCD treatment of Vero cells had a modest effect (2-fold) on infection mediated by VSV-(G), whereas ANDV infection was inhibited by more than 10-fold (Figure 4A). Treatment with MβCD reduced total cellular cholesterol levels to ∼80% of untreated samples (data not shown).
To determine if the cholesterol depleting drugs used above to effectively block rVSV-ANDV could also inhibit infection by wild-type ANDV, Vero E6 cells pretreated with PF-429242, mevastatin, or DMSO were infected with ANDV. Three days post infection, cells were fixed, labeled with antibodies against the ANDV nucleoprotein (ANDV-N), and infectivity was determined by flow cytometry (Figures 4B and C). Treatment with either PF-429242 or mevastatin had a significant and dose-dependent effect on ANDV infection, decreasing the percentage of infected cells by approximately 100-fold at 40 µM PF-429242 (Figure 4B) and 6-fold at 10 µM mevastatin (Figure 4C). In contrast, treatment with PF-429242 diminished the infectivity of VSV-(G) with roughly a 3-fold effect at both 20 µM and 40 µM concentrations (Figure 4B). Overall, the MβCD, PF-429242, and statin experiments provide strong evidence that cellular cholesterol levels dictate the permissivity of cells to ANDV infection.
To investigate the mechanism by which cells lacking a functional cholesterol regulatory pathway resist ANDV infection, we compared early stages of the rVSV-ANDV replication cycle in HAP1 cells with an insertional LentiET mutation (not shown) into the S1P gene (HAP1S1P) that abrogates S1P protein expression (Figure S7). HAP1S1P cells were resistant to infection by rVSV-ANDV as compared with rVSV-G (Figure 5A), similar to studies in mutant CHO cells. Next, a qRT-PCR assay was used to monitor binding and internalization of incoming rVSV-ANDV particles. Binding was performed at 4°C with equal amounts of virus added to wild-type or HAP1S1P cells. After extensive washing to remove unbound virus, S1P-deficient cells bound ∼2-fold more rVSV-ANDV virions than wild-type cells (Figure 5B) despite the fact that infection levels were ∼10-fold lower (Figure 5A). As expected for surface bound virus, protease treatment decreased the PCR signal by more than 90% for both cell lines (Figure 5B; background). To measure virus uptake, rVSV-ANDV was bound at 4°C, then cells were transferred to 37°C for one hour to allow virus internalization. We chose this time point because we found that rVSV-ANDV is resistant to the lysosomotropic agent ammonium chloride (NH4Cl) by one hour post infection, indicating that acid-dependent membrane fusion has occurred by this time (Figure S8). After internalization, cells were treated with protease to remove any remaining external virions (Figure 5B; internal). HAP1WT cells internalized nearly 100% of the measured bound virus (Figure 5B; compare HAP1WT bound to internal). While this appears remarkably efficient, there is precedent for high levels of internalization with other viruses, as 90% of bound influenza virions are internalized under similar conditions [23]. In contrast, S1P-deficient cells were unable to internalize rVSV-ANDV virions (Figure 5B; compare HAP1WT to HAP1S1P internal). Indeed, the amount of viral RNA inside these cells was comparable to the background levels of virus detected on the surface of cells stripped with protease prior to endocytosis (Figure 5B; compare HAP1S1P internal to HAP1WT and HAP1S1P background). Given that cells lacking S1P possess a ten-fold defect in both ANDV-glycoprotein mediated internalization (Figure 5B) and infectivity (Figure 5A), we infer that this internalization defect is responsible for the resistance.
To investigate this further, we performed confocal microscopy on Vero E6 cells infected with DiO-labeled viral particles (Figure 6). This allows us to track incoming particles. Furthermore, DiO will stain endocytic compartments following viral fusion so we can monitor binding, uptake and fusion events [24]. Vero E6 cells were pretreated with DMSO (Figure 6A) or the S1P-inhibitor PF-429242 (40 µM, Figure 6B) for 24 hours. Following treatment, cells were chilled to 4°C and incubated for 90 minutes with sucrose-purified viral particles previously stained with the lipophilic dye DiO. As a control for the virion preparation, DiO-labeled concentrated and purified supernatant from mock-infected cells did not produce visible puncta on cells (Figure S9). Following incubation, cells were extensively washed with cold PBS and fixed with paraformaldehyde immediately (‘0 min’), or following an incubation at 37°C (‘20 min’). Cellular membranes where counterstained with Wheat Germ Agglutinan-647 and imaged by confocal microscopy. In the absence of drug, both rVSV-ANDV (left) and VSV-(G) (right) samples displayed similar patterns of distribution- at 0 minutes puncta are found distributed along the cellular membranes, and appear internalized with larger puncta after 20 minutes. Although viral fusion could have occurred by 20 minutes, DiO will stain endocytic compartments zfollowing viral fusion [24]. In agreement with the qRT-PCR data, pretreatment of cells with PF-429242 appears to restrict DiO-labeled rVSV-ANDV to the cell periphery (Figure 6B, left). Uptake of labeled VSV-(G) also appears to be impaired (Figure 6B, right) and is consistent with the observed 3-fold decrease in VSV-(G)-infectivity of these cells at 40 µM PF-429242. The results are representative of two independent experiments. Although not quantitative, the microscopy results in S1P inhibited cells, coupled with the qPCR analysis in cells carrying a genetic lesion in S1P, suggest that a functional cholesterol regulatory pathway is needed for effective internalization and subsequent infection by ANDV.
Viral entry is a complex process often requiring orchestration of protein-protein interactions, cellular signaling, and cellular uptake mechanisms [25]. To begin dissecting this process for ANDV, two independent genetic screens were performed. In the first, insertional mutagenesis was carried out with a gene-trap vector in human haploid cells, a method used previously to identify host cell molecules and pathways used by an array of viral and bacterial pathogens [12], [26]–[33]. This approach allows near saturation of the human genome, although genes required for cell viability in vitro cannot be interrogated. In the second method, a large-scale RNAi screen provided a complementary approach by producing varying degrees of gene suppression and allowing one to potentially query genes required for cell viability. Previously, the low level extent of overlap in genes discovered in various RNAi screens for the same pathogen has hampered identification of specific requirements [34]–[37]. By employing both of these genetic approaches, we sought to identify cellular pathways important for the early stages of ANDV infection comprehensively. The discovery of components of a cholesterol regulatory complex as an ANDV entry requirement by these two independent screens reinforces the significance of this finding. Moreover, the identification of more than 60 independent insertional mutations in each of the 4 genes of this complex attests to the strength of this observation.
Interestingly, integration into the SREBF2 gene appears to be dramatically skewed toward the “antisense” orientation in which the gene trap vector would not efficiently disrupt gene expression. We hypothesize that these integrations likely cripple, but not fully inactive this gene presumably due to the poor fitness of cells completely deficient for SREBP-2.
The requirement for this sterol regulatory complex in ANDV infection was verified by several orthogonal approaches that included analysis of cells deficient in S1P, S2P, or SCAP function, along with generation and study of TALEN-driven deletions in SCAP, and siRNAs to deplete SREBP-2. These experiments confirmed the importance of this pathway for ANDV infection in multiple species and cells types. In addition, tests of a pharmacologic inhibitor of S1P, statins and cholesterol depletion extended these findings to wild-type ANDV. SREBP-2 functions as an initially ER-resident, regulated transcription factor whose activity is controlled by interactions with SCAP and proteolysis by S1P and S2P. After release from the ER, transit to the Golgi, proteolysis and transport to the nucleus, processed SREBP-2 activates transcription via Sterol Responsive Elements (SREs) upstream of genes such as HMG-CoA reductase, squalene synthase and the low density lipoprotein receptor (LDLR) which increases cholesterol production or uptake. Inactivating any of the four genes that were found to be important for ANDV entry blocks SREBP-2 mediated gene activation resulting in lower levels of cholesterol in the absence of an exogenous source. Given the cellular location of the sterol regulatory pathway components in the ER and cis-Golgi, they appear unlikely to play a direct role in ANDV entry. However, reduced expression of one or more genes transcriptionally induced by SREBP-2 or altered cellular cholesterol levels could account for this phenotype. Pharmacologic inhibitors in conjunction with delipidated media made it possible to separate sterol-dependent gene induction from cellular cholesterol levels. When cellular cholesterol levels were reduced by inhibiting cholesterol synthesis with mevastatin or by treating cells with MβCD, sterol-dependent gene expression is induced. Despite this, ANDV infection remained suppressed by more than 10-fold, arguing that it is the lower levels of cellular cholesterol rather than reduced expression of SREBP-2 target genes that is important. Also consistent with this conclusion is the finding that infection could be restored to cholesterol-depleted cells upon addition of exogenous cholesterol. Altogether, these results demonstrate that ANDV entry is particularly sensitive to perturbations of cellular cholesterol levels.
Several viruses have been shown to exhibit some degree of cholesterol dependence for entry into cells [38]–[41]. In some, the effect of cholesterol depletion is direct and highly specific. The clearest example is the alphavirus Semliki Forest virus (SFV), where a mechanistic role has been demonstrated for cholesterol in membrane fusion. In contrast to our findings with ANDV, cholesterol-depleted cells were unaltered in their ability to bind and internalize SFV, but were blocked at the downstream step of membrane fusion, which blocked subsequent virus replication [42], [43]. For other viruses cholesterol depletion inhibits infection indirectly by inhibiting uptake via endocytic pathways. Caveolin-dependent endocytosis is especially sensitive to cholesterol levels and internalization of non-enveloped viruses such as SV-40 is blocked when cholesterol is depleted [44], [45]. In A549 cells at the concentrations of statins we employed to block ANDV entry, SV-40 infection appeared unchanged (data not shown) suggesting that the observed effects are not due to impaired caveolin-mediated entry. Clathrin-mediated endocytosis, a common route of entry into cells by enveloped viruses, is also sensitive to cholesterol depletion using β-cyclodextrin [46]. However, ANDV as well as rVSV-ANDV infection exhibited exquisite sensitivity to cholesterol depletion and disruption of the sterol regulatory pathway under conditions where VSV-(G) mediated infection, a clathrin dependent process, was only marginally affected. Thus it is unlikely that the mechanism underlying reduced ANDV infection is linked to general clathrin-mediated endocytosis. It is possible that a modest reduction of cholesterol levels impacts lipid raft integrity with concomitant effects on localization of proteins needed by ANDV or upon signaling by proteins that partition into these cholesterol-rich domains. For example, signaling by lipid raft localized DAF1 has been found to be critically important for Coxsackie virus entry [47]. Future studies will be required to examine whether lipid rafts or raft-mediated signaling is important for ANDV entry.
Underscoring the importance of cholesterol homeostasis for viral entry is the recent observation that cellular antiviral systems interfere with cholesterol regulation or trafficking (reviewed in [48]). Interferon induced transmembrane proteins (IFITMs) appear to exert their antiviral activity by causing the accumulation of cholesterol in late endosomal compartments thereby blocking infection of a wide variety of viruses that enter through this compartment [49]. Infection by several Bunyavirdae family members, including ANDV, is inhibited by IFITM's [50]. Two other recent studies revealed that interferon induced production of a sterol (oxysterol 25 hydroxycholesterol), which is known to be involved in cholesterol homeostasis, can block infection at the point of viral entry [51], [52]. Although likely mechanistically dissimilar, our results coupled with these findings, highlight the importance of cholesterol homeostasis in viral entry and suggest targeting this process for the development of broadly effective antivirals.
We found that ANDV bound equally well to wild-type and S1P-deficient cells, suggesting that the surface levels of the cellular factors to which ANDV glycoproteins bind are not dependent on the sterol regulatory pathway. Whether ANDV has an absolute requirement for a specific cell surface receptor is not known, though the integrin αvβ3 has been implicated as a binding factor in some cell types [53]–[56]. However, αvβ3 levels were below the limits of detection in the HAP1 cells used for the insertional mutagenesis screen and integrins were not identified in the HEK293-based RNAi screen. Additionally, surface expression levels of αvβ3 on Vero cells were not affected by pharmacologic treatments that blocked ANDV glycoprotein-mediated infection (data not shown). Taken together these observations suggest that αvβ3 integrin is not involved in the cholesterol-dependent phenotype observed in the diverse cells used in these studies. This does not preclude integrin involvement in other cell types.
Although the current screens converged on genes regulating sterol synthesis, it is likely that altering the parameters of the screens by including cholesterol enriched media or constitutively expressing activated SREBP-2 during screening, will uncover additional host factors and/or pathways important for ANDV entry. Finally, the sensitivity of ANDV to safe, effective cholesterol-lowering drugs may suggests new treatments for ANDV infection and pathogenesis.
pLentiET viral pseudotypes were prepared via co-transfection of HEK293T cells with pCAGGS-VSV-(G) (Addgene), pSPAX (Addgene), and pLentiET genome plasmids. Viral supernatants were harvested 48 hours later. Replication competent rVSV-G and rVSV-ANDV were previously described [11], [57]. VSV-(HTNV), VSV-(ANDV), VSV-(SINV) and VSV-(G) pseudovirions were created via coexpression of a VSV-ΔG-reporter genome along with a pCAGGS-viral glycoprotein expression plasmid, as previously described [58]. Wild-type ANDV (Chilean strain 9717869) was provided by Connie Schmaljohn at the U.S. Army Medical Research Institute of Infectious Diseases and used under BLS3 conditions. A recombinant Vaccinia virus expressing GFP was previously described [59].
HAP1 cells were assessed by fluorescent cytometry for ploidy by Hoechst 33342 staining of nuclei. Haploid cells were enriched by size selection to ∼80% haploid immediately before creation of an insertionally-mutagenized library of ∼1×109 cells using three rounds of mutagenesis with pLentiET gene-trap virus. Virus was added such that ∼80% of cells were transduced per round of mutagenesis as determined by a lenti-GFP control virus made in parallel. ∼75 million library cells were selected with either rVSV-G (MOI of 2) or rVSV-ANDV (MOI of 3–5). Cells selected for resistance to rVSV-ANDV were collected and pooled within 3 weeks and saved as DMSO frozen stocks or used for chromosomal DNA preparation. Clonal populations of HAP1 cells were achieved by limiting dilution.
Chromosomal DNA was prepared from either pools or clonal populations of ANDV resistant HAP1 cells. A DNA amplicon preparation protocol that specifically amplifies LTR-host junctions was carried out essentially as previously described [60]. Amplicons derived from pools of ANDV resistant cells or the unselcted library were subjected to deep sequencing analysis using either 454 or illumuna based platforms respectively and aligned to the human genome using the University of California, Santa Cruz BLAST Like Alignment Tool, BLAT (hg18, version 36.1). Enrichment of the sterol regulatory complex genes was calculated by comparing how often that gene was mutated in the screen to how often the gene carries an insertion in the control library data set. For each sterol regulatory complex gene a P-value was calculated using the one-sided Fisher exact test.
siRNAs from the Ambion Druggable genome library representing 9,102 genes were spotted in 54 384-well white bottom plates in a 2×2 format such that each gene was targeted by 2 different pools of 2 siRNAs - 4 unique siRNAs in total. Positive and negative control siRNAs were plated in triplicate on each plate. Using a liquid handler (WellMate, Thermo Fisher) to decrease variability, 0.5 µL of HiPerFect (Qiagen) in 9.5 µL of OptiMem (Gibco) was added to each well and incubated for 15 min at room temperature to allow complex formation. HEK293T/ffLuc cells per well were plated to achieve a 40 nM final siRNA concentration. 72 hours post-transfection, cells were infected with VSV-(ANDV)*rLuc. 24 hrs post-infection, firefly and Renilla luciferase expression were measured. Robust z-scores were calculated for each plate using the median and interquartile ranges of log-transformed RLUs [61]. For the secondary screen, 3 unique siRNAs for each gene, different from those used in the primary screen, were obtained from Ambion and arrayed in 96-well plates.
Sub-confluent cells were spin-infected (45 minutes at 1200 × g, 20°C) with viral pseudotypes and harvested for infectivity assays 8–12 hr post-infection. Infected cells were quantified by FACS using RFP expression or staining with antibodies against the matrix protein of VSV [62], [63] followed by a secondary antibody conjugated to AF-647. At least 104 events, in duplicate, were counted for at least three independent experiments. For PF-429242 and mevastatin studies, cells were pretreated for 24 hr before infection. All infections and overlays were carried out in the continued presence of drug or DMSO, for the length of the infection. For MβCD studies, cells were pretreated for one hour and washed prior to infection. Cells were infected with ANDV at an MOI of 3 by adding 1 mL of inoculum to a 6 well dish for two hours at 37°C. Viral inoculum was removed, cells were overlaid with fresh media containing drug (except for MβCD assays) or DMSO where indicated. Cells were harvested three to four days post-infection, fixed for one hour in 4% formaldehyde then analyzed by flow cytometry using anti-ANDV N.
rVSV-ANDV binding was performed at 4°C in 24 well plates of HAP1 cells in IMDM containing a 1∶8 ratio of 10% FBS to delipidated FBS. After one hour on ice, cells were washed with ice-cold PBS to remove unbound virus, and samples to measure bound virus were collected by scraping cells into PBS, followed by additional washing. Trypsin-EDTA was used to remove surface bound virus. rVSV-ANDV internalization was measured by first binding and washing at 4°C then incubating samples at 37°C for one hour in IMDM containing delipidated media. Cells were then washed with PBS and treated with 0.05% trypsin and washed to remove surface bound virus. Samples were kept on ice after the final washing step, then processed for RNA. Primers specific to the VSV N segment were used for qRT-PCR. Data were analyzed using the ΔΔCT method [64] by calculating the change in gene expression normalized to that of GAPDH as a housekeeping gene.
A TALEN pair targeting exon 3 in SCAP was designed and constructed as previously described [65]. Mutations induced by non-homologous end joining (NHEJ) following expression of the SCAP TALEN were measured as previously described [17]. Band intensities were quantified using ImageJ and utilized to estimate mutation rates as previously described using the formula: % gene modification = 100×(1-(1-fraction cleaved)1/2) [66]. The T7 endonuclease assay has a range of detection from approximately 1% to 50% NHEJ.
More detailed Methods descriptions are given in the Supplemental Information.
|
10.1371/journal.pgen.1004486 | LIN-42, the Caenorhabditis elegans PERIOD homolog, Negatively Regulates MicroRNA Transcription | During C. elegans development, microRNAs (miRNAs) function as molecular switches that define temporal gene expression and cell lineage patterns in a dosage-dependent manner. It is critical, therefore, that the expression of miRNAs be tightly regulated so that target mRNA expression is properly controlled. The molecular mechanisms that function to optimize or control miRNA levels during development are unknown. Here we find that mutations in lin-42, the C. elegans homolog of the circadian-related period gene, suppress multiple dosage-dependent miRNA phenotypes including those involved in developmental timing and neuronal cell fate determination. Analysis of mature miRNA levels in lin-42 mutants indicates that lin-42 functions to attenuate miRNA expression. Through the analysis of transcriptional reporters, we show that the upstream cis-acting regulatory regions of several miRNA genes are sufficient to promote highly dynamic transcription that is coupled to the molting cycles of post-embryonic development. Immunoprecipitation of LIN-42 complexes indicates that LIN-42 binds the putative cis-regulatory regions of both non-coding and protein-coding genes and likely plays a role in regulating their transcription. Consistent with this hypothesis, analysis of miRNA transcriptional reporters in lin-42 mutants indicates that lin-42 regulates miRNA transcription. Surprisingly, strong loss-of-function mutations in lin-42 do not abolish the oscillatory expression patterns of lin-4 and let-7 transcription but lead to increased expression of these genes. We propose that lin-42 functions to negatively regulate the transcriptional output of multiple miRNAs and mRNAs and therefore coordinates the expression levels of genes that dictate temporal cell fate with other regulatory programs that promote rhythmic gene expression.
| MicroRNAs play pervasive roles in controlling gene expression throughout animal development. Given that individual microRNAs are predicted to regulate hundreds of mRNAs and that most mRNA transcripts are microRNA targets, it is essential that the expression levels of microRNAs be tightly regulated. With the goal of unveiling factors that regulate the expression of microRNAs that control developmental timing, we identified lin-42, the C. elegans homolog of the human and Drosophila period gene implicated in circadian gene regulation, as a negative regulator of microRNA expression. By analyzing the transcriptional expression patterns of representative microRNAs, we found that the transcription of many microRNAs is normally highly dynamic and coupled aspects of post-embryonic growth and behavior. We suggest that lin-42 functions to modulate the transcriptional output of temporally-regulated microRNAs and mRNAs in order to maintain optimal expression of these genes throughout development.
| MicroRNAs (miRNAs) are non-coding RNA molecules that post-transcriptionally regulate gene expression [1]. The maturation of miRNAs is a stepwise process that begins with the RNA polymerase II-dependent transcription of long capped and polyadenylated primary miRNAs (pri-miRNAs) [2], [3]. Most pri-miRNAs are then endonucleolytically cleaved by the nuclear Microprocessor complex, composed of Drosha (an RNase III enzyme) and its binding partner Pasha, to yield a ∼70 nt precursor miRNA hairpin (pre-miRNA) [4]. After export to the cytoplasm, the pre-miRNA is cleaved by Dicer (a second Type III RNase) yielding a ∼22 nt duplex that consists of the mature miRNA and its corresponding passenger RNA [5], [6]. The mature single-stranded ∼22 nt miRNA is then loaded into the Argonaute and GW182 to form the miRNA-induced Silencing Complex (miRISC) [7]–[9]. Through partial complementary base-pairing between the miRNA and target mRNA, the miRISC complex negatively regulates gene expression by either translational repression or mRNA degradation [7], [10]. In vivo, target mRNA down-regulation is directly proportional to the amount of miRNA associated with miRISC [1].
Experimental and computational approaches indicate that an individual miRNA can bind to and regulate hundreds of mRNAs and that the majority of protein-coding genes are miRNA targets [11]–[14]. As such, miRNAs have been implicated in a variety of developmental and cellular processes including cell fate specification, proliferation and apoptosis [15]–[19]. In many of these scenarios, the expression of distinct miRNAs is tightly controlled and/or the individual steps of miRNA biogenesis are actively regulated at either the transcriptional or post-transcriptional level by sequence-specific transcription factors or RNA-binding proteins, respectively. For example, some regulatory proteins control miRNA biogenesis by directly binding structural elements within the pri- or pre-miRNA transcript whereas others broadly impact global miRNA biogenesis by inhibiting enzymes required for general miRNA processing and/or activity [20]. Importantly, many of the proteins that regulate miRNA biogenesis are highly conserved and mutations in these genes result in a variety of developmental disorders and diseases [20].
The C. elegans heterochronic pathway has been instrumental to our understanding of the principles of miRNA-mediated gene regulation and for the identification of components that are required to control miRNA expression, metabolism and activity [21]. Post-embryonic development in C. elegans proceeds through a series of four larval stages, punctuated by molts, in which the temporal and spatial patterns of cell division and differentiation are tightly orchestrated and invariant [22]. Heterochronic genes organize temporal patterns of development by controlling stage-specific gene expression. Defects in heterochronic genes cause animals to display temporal cell fate transformations including either the inappropriate skipping or reiteration of stage-specific patterns of cell divisions [23]. An overarching feature of the heterochronic pathway is that many protein-coding genes that are important for controlling temporal patterning are post-transcriptionally regulated by miRNAs [16], [24]–[28]. In this context, miRNAs are expressed at defined times during post-embryonic development and function as molecular switches to inhibit earlier patterns of development and promote the emergence of later gene expression profiles. Throughout post-embryonic development, the expression of heterochronic miRNAs is regulated at both the transcriptional and post-transcriptional levels [20], [29]–[32]. In addition, mutations that alter heterochronic miRNA expression often display strong temporal patterning and behavioral phenotypes [16], [33]–[36].
While the regulatory strategies that dictate patterns of cell fate specification have rapidly emerged through the identification of conserved heterochronic genes, we still lack a deep understanding of how the temporal expression of heterochronic genes are coordinated with aspects of growth and behavior. This coupling is especially important as many post-embryonic cell division and cell fate specification events are intimately tied to the molting cycle [37], [38]. Surprisingly, most of the known genes required for molting do not dramatically alter temporal cell fates and only a few heterochronic genes disrupt the reiterative process of molting [23], [38]–[45]. The molting phenotypes associated with heterochronic mutants usually result from inappropriate temporal cell fate transformations that lead to either a cessation (for precocious heterochronic mutants) or an inappropriate reiteration (for retarded heterochronic mutants) of molting [16], [23]–[28], [31], [33], [34], [42]–[45]. To date, only a single heterochronic gene, lin-42, is known to alter both temporal patterning of cell fate specification and the precise timing of recurrent developmental events [39]. lin-42 is the C. elegans homolog of human and Drosophila PERIOD and was initially identified as a heterochronic mutant that precociously executes adult-specific patterns of development after the third larval molt [46]–[48]. The lin-42 locus is complex and encodes three protein isoforms (LIN-42A, LIN-42B and LIN-42C) that are expressed from two distinct promoters (Figure 1A) [39], [46]–[48]. During post-embryonic development, lin-42 mRNA levels fluctuate over the molting cycles and peak once during each larval stage [39], [46]–[48]. While its precocious developmental phenotypes are similar to other heterochronic mutants, the periodic expression pattern of LIN-42 distinguishes it from other monotonically expressed heterochronic proteins. Therefore, lin-42 has been proposed to play a more iterative role in developmental timing. However, its relationship to and interplay with other heterochronic genes has been difficult to establish at the molecular level. In addition to altering temporal patterns of development, mutations that disrupt the expression of LIN-42A and LIN-42B isoforms display dramatic defects in behavior and molting [39]. Specifically, lin-42a/b mutants alter the normally synchronous molting patterns displayed by wild-type animals and these defects frequently result in lethality [39]. Given that LIN-42 is a nuclear protein, an attractive hypothesis is that LIN-42 coordinates gene expression programs that control the molting cycles with regulatory pathways that mediate stage-specific cell lineage programs [48]. However, this potential role for LIN-42 remains elusive because 1) the molecular nature of LIN-42 activity is yet to be defined and 2) LIN-42 downstream targets that mediate iterative (molting) and sequential (cell fate patterning) gene regulatory programs are unknown.
In this study, we employed multiple forward genetic screens that were collectively geared to identify negative regulators of miRNA expression. As a product of this approach, we identified mutations in lin-42 that suppress multiple stage-specific lineage defects associated with heterochronic miRNAs. Analysis of miRNA expression in lin-42 mutant animals suggests that LIN-42 broadly functions to negatively regulate miRNA expression is therefore is likely to act in a variety of pathways that require miRNAs for proper cell fate specification. Consistent with this hypothesis, we find that lin-42 also plays a role in the miRNA-mediated specification of asymmetric gene expression patterns in gustatory neurons. Analysis of LIN-42 interactions with chromatin suggests that LIN-42 potentially regulates the transcription of both miRNAs and mRNAs. We demonstrate, through the use of transcriptional reporters, that lin-42 mutations alter the transcription of lin-4 and let-7. Surprisingly, mutations that remove LIN-42 isoforms containing the conserved PAS domains (required for circadian gene regulation by human and Drosophila PERIOD) do not uncouple miRNA expression from the molting cycle but, instead, dramatically alter the transcriptional output of miRNA genes. We conclude that a key molecular function of lin-42 is to dynamically inhibit the transcription of post-embryonically expressed miRNAs and mRNAs to ensure the robustness of developmental gene expression.
The inherent dependency of the heterochronic pathway on precisely controlled miRNA activity provides a unique genetic context to identify components that control aspects of miRNA metabolism or expression. To accomplish this, we performed forward genetic screens in either lin-4(ma161), alg-1(ma192) or let-7(n2853) mutant backgrounds to identify novel heterochronic mutations that correct the phenotypes associated with aberrant L1 to L2 (early), L2 to L3 (middle) or L4 to adult (late) cell fate transitions, respectively. These mutants are unique in that they express miRNAs at a much lower level than wild-type animals but do not completely eliminate their expression. lin-4(ma161) and let-7(n2853) mutations alter the conserved seed sequence of the mature miRNA and reduce levels of these miRNAs in vivo [16], [24]. Animals harboring lin-4(ma161) and let-7(n2853) mutations are phenotypically indistinguishable from null mutants and reiterate L1- and L4-specific cell fates, respectively (Table 1) [16], [24]. alg-1(ma192) mutations alter one of the two miRNA-specific Argonautes and disrupt the ability of processed miRNAs to repress downstream target mRNAs [49]. Animals harboring the alg-1(ma192) mutation inappropriately express hbl-1 (the major miRNA target of miR-48, miR-241, and miR-84) in the L3 stage and reiterate L2-specific seam cell division patterns [49]. Consistent with the defects associated with the misregulation of each of these stage-specific transitions, lin-4(ma161), alg-1(ma192) and let-7(n2853) animals display highly penetrant heterochronic phenotypes and fail to express adult-specific gene regulatory programs, including the expression of the adult-specific Pcol-19::GFP transcriptional reporter (Table 1). Suppressors of the retarded heterochronic phenotypes in each of these genetic backgrounds were identified as F2 progeny of mutagenized animals that were able to restore normal development (Figure 1C). Five mutants (ma206, ma208, csh1, csh4 and csh5) were able to suppress multiple retarded heterochronic phenotypes associated with all three mutant backgrounds. Each mutant mapped to a single locus on chromosome II (Figure 1A and Table 1), and subsequent SNP-SNP mapping and sequencing results demonstrated that all five alleles contain mutations that lie within lin-42 and would be predicted to create a premature truncation of the lin-42b or lin-42c open reading frames (Figure 1A and Table S1) [50]. Consistent with previous analyses of lin-42 mutations, animals harboring the ma206, ma208, csh1, csh4 or csh5 allele display highly-penetrant precocious heterochronic phenotypes (Table 1) which were rescued with a fosmid containing the genomic fragment of the wild-type lin-42 gene [46]–[48]. Mutations in lin-42 have been demonstrated to suppress heterochronic phenotypes associated with multiple heterochronic mutants, including lin-4 and let-7 [16], [46], [47], [51]. In these reports, only terminal cell lineage phenotypes, including a correction of the L4-to-adult vulval bursting phenotypes, restoration of adult-specific expression of Pcol-19::GFP, and formation of adult-specific alae were assayed.
We next sought to determine whether the lin-42 mutations we isolated suppressed only terminal heterochronic phenotypes or if they corrected additional stage-specific cell lineage defects associated with lin-4(ma161), let-7(n2853) and alg-1(ma192) mutations. To test if our new lin-42 mutants correct retarded cell lineage phenotypes, we compared multiple hypodermal cell lineages in lin-4(ma161), alg-1(ma192), and let-7(n2853) single mutants to double mutants that also harbored the individual lin-42 candidate suppressor mutations. lin-4 animals lack vulval structures as a consequence of reiterating L1-specific developmental programs in the hypodermis and failing to interpret inductive cues from the anchor cell that initiate vulval morphogenesis at the L3 stage [40], [52]. The vulvaless (Vul) phenotypes of lin-4(ma161) animals are highly penetrant (Figure 1B, F) and are almost completely suppressed by lin-42(ma206), lin-42(ma208), lin-42(csh1), lin-42(csh4) and lin-42(csh5) (Figure 1B, F). These results indicate that lin-42 functions to control cell fate specification in at least the mid-L3 stage, when the vulval precursors are spatially patterned.
The ability of several of these suppressors to alleviate hypodermal cell lineage phenotypes in miRNA hypomorphic mutants was not limited to the vulval cell lineage. The lateral seam cells of lin-4(ma161), alg-1(ma192), and let-7(n2853) animals display altered temporal cell fate specification and also fail to terminally differentiate at the L4 molt. As a consequence, lin-4(ma161), alg-1(ma192), and let-7(n2853) animals lack alae structures as young adults (Table 1, Figure 1D). The alae phenotypes in lin-4(ma161), alg-1(ma192), and let-7(n2853) mutants was strongly suppressed by the lin-42(ma206) allele (Table 1). alg-1(ma192) mutants reiterate L2-specific seam cell division programs due to the inappropriate perdurance of hbl-1 expression at the L3 stage [49]. As a consequence, young adult alg-1(m192) animals harbor supernumerary seam cells (23.5+/−3.78; WT = 11) (Figure 1D, E). lin-42(ma206) mutations strongly suppress the L2-to-L3 heterochronic phenotypes of alg-1(ma192) mutants as lin-42(ma206); alg-1(ma192) animals exhibit a significant reduction in the number of supernumerary seam cells (11.9+/−1.3) and display normal adult alae (Figure 1D and E). Therefore, lin-42 has a role in controlling L2-to-L3 temporal cell fate transitions.
We asked whether our new lin-42 alleles could suppress the heterochronic phenotypes associated with lin-4(e912) and let-7(mn112) null mutants to a level similar to that observed with the hypomorphic alleles used in our initial screens. To test this, we compared aspects of vulval cell proliferation and morphogenesis at the early L4 stage in lin-4(ma161), lin-42(csh5) lin-4(ma161), lin-4(e912) and lin-42(csh5) lin-4(e912) mutants to those of similarly staged wild-type and lin-42(ma205) animals. Lowering lin-42 function in the context of the hypomorphic lin-4(ma161) background results in a strong restoration of vulval development with 85% of animals exhibiting induction/proliferation and invagination of P cells from the larval cuticle (Figure 1B and F). Surprisingly, 42% percent of lin-42(csh5) lin-4(ma161) animals exhibited morphologically normal adult vulva and were competent for egg laying (n = 100). In contrast, reducing lin-42 activity in lin-4(e912) animals has little or no effect on P cell proliferation and vulval morphogenesis (Figure 1F). lin-42 exhibits a similar genetic relationship to let-7 mutations. Both hypomorphic (n2853) and null (mn112) alleles of let-7 display highly penetrant vulval bursting phenotypes at the L4-to-adult transition (Figure 1G) [16], [28]. lin-42 mutations almost completely suppress the lethality associated with larval-to-adult transitions in let-7(n2853) animals but do not statistically improve the viability of let-7(mn112) adults (Figure 1G). These results strongly suggest that lin-42 mutations are not bypass suppressors of lin-4 or let-7 mutant phenotypes but likely require a minimum level of lin-4 or let-7 activity for suppression.
One mechanism by which lin-42 mutations could suppress multiple hypomorphic miRNA mutants would be that lin-42 normally functions to repress some aspect of miRNA metabolism. To directly test this hypothesis, we measured the abundance of several mature miRNAs when lin-42 function is compromised. Northern blot analysis of total RNA extracted from morphologically-staged, young adult animals demonstrates that the total amount of lin-4 and let-7 miRNAs in alg-1(ma192) mutants is 1–1.5 fold lower than the levels found in wild-type animals (Figure 2A and B). In addition to reducing the levels of mature let-7 miRNA, alg-1(ma192) animals display a slight reduction in pre-miRNA processing and accumulate the pre-let-7 hairpin precursor. This under-accumulation phenotype of mature lin-4 and let-7 miRNAs in alg-1(ma192) mutants is suppressed when lin-42 function is compromised (Figure 2A). Consistent with our hypothesis that lin-42 normally inhibits miRNA biogenesis, similarly-staged lin-42(ma206) mutants over-accumulate both lin-4 and let-7 miRNAs (Figure 2A). While the amount of mature let-7 miRNA increases in lin-42(ma206); alg-1(ma192) double mutants, the ratio of pre-let-7 to mature let-7 miRNA is similar to that detected in alg-1(ma192) single mutants (Figure 2A). Therefore, although mature let-7 miRNA over-accumulates in lin-42(ma206) mutants, there is no change in pre-let-7 to mature let-7 processing efficiency as compared to wild type. These data suggest that lin-42 mutations alter aspects of miRNA expression upstream of pre-miRNA processing.
To determine if lin-42 plays a more broad role in modulating miRNA expression, we employed real-time quantitative PCR to measure the expression levels of additional miRNAs in morphologically-staged, young adult wild-type, lin-42(ma206), alg-1(ma192) and lin-42(ma206); alg-1(ma192) animals. We measured a variety of miRNAs that display tissue-specific and temporal expression patterns that are distinct from lin-4 and let-7 miRNAs [35], [53]–[58]. For comparison, we also assayed the expression of two additional small nuclear RNAs (U18 and sn2343) as well as two 21U RNAs that associate with PRG-1, a distinct Argonaute involved in the C. elegans piRNA pathway [59]–[61]. Consistent with the observation that alg-1(ma192) mutations broadly affect miRNA expression, the abundance of all miRNAs tested (lin-4, miR-48, miR-241, miR-84, let-7, miR-1, miR-46, miR-58 and miR-79) was decreased in alg-1(ma192) mutants (Figure 2B). The general miRNA under-accumulation phenotype displayed in alg-1(ma192) mutants was suppressed by removing lin-42 function (Figure 2B). Importantly, the expression levels of the 21U-RNA transcripts were not significantly altered in lin-42(ma206) mutants (Figure 2B). Examination of miRNA expression in lin-42(ma206) mutants indicate that all tested miRNAs were overexpressed from ∼1.8 to ∼3.2 fold when compared to similarly-staged wild-type animals (Figure 2B). miRNA stability is dependent on a variety of factors, including the expression levels of the Argonaute components of miRISC [62]. To determine if the increase in miRNA levels in lin-42 mutant backgrounds was due to the overexpression of the C. elegans miRNA-specific Argonautes (ALG-1 and ALG-2), we quantified the levels of functional ALG-1 and ALG-2 fluorescent reporters in animals with reduced lin-42 activity. The results of this analysis, presented in Figure S1, indicate that ALG-1 and ALG-2 expression is not altered in lin-42(RNAi) animals. Collectively, these results indicate that lin-42 functions to negatively regulate the expression of a wide range of miRNAs.
Because lin-42 regulates the abundance of many miRNAs, we asked if lin-42 functions in other gene regulatory pathways where controlling the expression levels of specific miRNAs is critical for proper cell fate determination. To test this idea, we examined how mutations in lin-42 affected the cell fate specification of two bilaterally symmetric gustatory neurons, ASE left (ASEL) and ASE right (ASER). Normally, a complex gene regulatory network composed of miRNAs and transcription factors form a bi-stable, double-negative feedback loop that ensures mutually exclusive gene expression programs in ASEL and ASER neurons [63], [64]. A major determinant of the exclusive gene expression programs in these two neurons is the ASEL-specific expression of the lsy-6 miRNA and the resulting down-regulation of its target, cog-1. Animals completely lacking lsy-6 fail to down regulate COG-1 in ASEL, and, as a consequence, ASEL neurons in lsy-6(ot71) null mutants adopt an ASER cell fate [63]. These phenotypes can be monitored by a failure to express the Plim-6::GFP transcriptional reporter in ASEL in lsy-6 mutants (Figure 3A). Importantly, lsy-6-mediated repression of cog-1 is dosage-dependent; weak alleles of lsy-6, such as ot150, under-accumulate lsy-6 miRNA as a consequence of reduced lsy-6 transcription and result in a partially penetrant ASEL-to-ASER cell fate transformation phenotype (Fig. 3B) [64]. The ot150 allele of lsy-6 has been used in a variety of contexts as a sensitized genetic background to identify gene products that function in the miRNA pathway [65]–[67]. While 13% of animals harboring only the lsy-6(ot150) allele fail to maintain Plim-6::GFP in ASEL, the penetrance of this phenotype is partially suppressed in lin-42(ma206); lsy-6(ot150) double mutants (Figure 3B), suggesting that lin-42 may play a modulatory role in neuronal cell fate specification. To further explore a potential role for lin-42 in assuring proper neuronal cell fate specification, we developed a more sensitive assay for the lsy-6-mediated repression of cog-1. As previously mentioned, alg-1(ma192) mutants display defects in variety of miRNA-mediated processes, including developmental timing [49]. While the alg-1(ma192) mutation alone does not alter Plim-6::GFP expression in ASEL, combining alg-1(ma192) with lsy-6(ot150) results in a dramatic increase in ASEL to ASER cell fate mis-specification (Figure 3B). As with the suppression of alg-1(ma192) heterochronic phenotypes, reducing lin-42 function significantly restores normal ASEL cell fate specification in lsy-6(ot150); alg-1(ma192) animals (Figure 3B). Because lsy-6-mediated cell fate specification is established during embryonic development, we conclude that lin-42 functions throughout development and is critical for multiple miRNA-mediated developmental processes.
To characterize the spatial and temporal expression patterns of lin-42-regulated miRNAs, we generated a series of engineered transcriptional reporters that contain between 2 and 5 kB of genomic upstream regulatory sequence that drives the expression of GFP fused to an optimized proline-glutamate-serine-threonine-rich (PEST) sequence. PEST domains have been demonstrated, in a variety of heterologous systems, to accelerate the degradation of target proteins via the nuclear and cytoplasmic 26S proteasome [41], [68]–[71]. In contrast to transcriptional reporters that drive the expression of stable GFP, analysis of GFP-pest expression in Plin-4::GFP-pest, Plet-7::GFP-pest or PmiR-1::GFP-pest transgenic animals indicates that the expression of each transcriptional reporter is highly dynamic, with peak GFP-pest expression occurring once each larval stage (n>30 animals per time point)(Figure 4A) [29], [53], [55]. The highly dynamic nature of each expression pattern was then monitored in a population of worms that were transiently arrested at the L1 diapause and then developmentally synchronized by restoring bacterial food. For each of the mir::GFP-pest reporters, post-embryonic GFP-pest expression was first detected at approximately 14 hours (Figure 4B, D and F). Once transcriptionally activated, Plin-4::GFP-pest and Plet-7::GFP-pest reporters peak in expression by 18–20 hours and diminish with similar kinetics (Figure 4B and D). For animals expressing the Plet-7::GFP-pest reporter we monitored GFP-pest expression for longer periods after release from L1 arrest. Consistent with the highly pulsatile nature of this expression pattern, GFP-pest expression was reinitiated at 30 hours, which correlates with the later portions of the L2 stage (Figure S3). While transcriptional activation of the Pmir-1::GFP-pest reporter was also initiated at 14 hours post-L1 arrest, the peak of Pmir-1::GFP-pest expression occurred at a later time point, and diminished with slower kinetics, as compared to Plin-4::GFP-pest and Plet-7::GFP-pest expression (Figure 4F).
We then asked whether the temporal expression pattern of each Pmir::GFP-pest reporter was synchronized with defined stages of the molting cycle, specifically lethargus and ecdysis. To accomplish this, we isolated late-L3-staged transgenic animals and cultured them on separate nematode growth media (NGM) plates at 20°C. Individual animals were then monitored for GFP-pest expression in relation to the induction and termination of both lethargus and ecdysis (Figure 4C, E, and G). We find that the majority of animals which harbor the Plin-4::GFP-pest transgene cease GFP-pest expression by L3 ecdysis and resume expression by the mid-L4 stage. The pulse of Plin-4::GFP-pest expression at the L4 stage extends through the early portion of young adulthood and completely overlaps with the lethargus period in all animals (Figure 4C). Plet-7::GFP-pest expression followed a similar pattern (Figure 4E). However, GFP-pest expression was more variable at the L3-to-L4 transition and L4-specific induction of this transgene was primarily restricted to the lethargus period (Figure 4E). In contrast to the expression profiles of the lin-4 and let-7 reporters, induction of Pmir-1::GFP-pest expression began during, or immediately after, L3 ecdysis and persisted into the L4 stage. A second pulse of Pmir-1::GFP-pest expression completely overlapped with the L4 lethargus period and continued into early adulthood (Figure 4G). Collectively, these results suggest that the expression patterns of lin-4, let-7 and mir-1 are dynamic throughout development and that the cyclical transcription of these miRNAs is mediated by their cognate promoter sequences. Furthermore, these data show that, while each of the Pmir::GFP-pest reporters display pulsatile expression patterns, the transcriptional dynamics for each gene do not display a complete unity of phase in their expression profiles.
To compare the temporal expression patterns of these three miRNAs with that of lin-42, we constructed transgenic strains that expressed either Plin-42a::GFP-pest or Plin-42b::mCherry-pest and subjected these animals to the same time course analyses. It has been previously demonstrated that two independent promoters drive the expression of LIN-42A, LIN-42B and LIN-42C isoforms [39]. Consistent with these findings, Plin-42a::GFP-pest and Plin-42b::mCherry-pest reporters displayed highly pulsatile expression during the L1 stage with initiation and termination of expression at 12 and 28 hours post L1 arrest, respectively (Figure 4H). In addition, we find that Plin-42a::GFP-pest expression peaks at 16 hrs, immediately preceding the expression of the Pmir::GFP-pest reporters, while the peak of Plin-42b::mCherry-pest expression occurs at 20 hrs (Figure 4H). Detailed analysis of individual L3-to-adult animals indicates that Plin-42a::GFP-pest expression displays a temporal expression pattern that is highly similar to Plin-4::GFP-pest and Plet-7::GFP-pest expression (Figure 4I). Specifically, in all three reporters, GFP-pest expression diminishes prior to L3 ecdysis, resumes prior to the L4 lethargus period, and terminates immediately after L4 ecdysis (Figure 4I). In striking contrast to our mir and lin-42 transcriptional reporters, Pcol-12::mCherry-pest expression does not occur during the molting cycle, but rather is exclusively expressed after each ecdysis (Figures 4J and 4K).
Previous analysis of lin-4 and let-7 expression indicates that these miRNAs are expressed in a variety of tissues, including the hypodermis, intestine and muscle [29], [35], [53], [57]. To determine if Plin-4::GFP-pest displays differential temporal expression patterns in a subset of these tissues, we conducted a detailed examination of GFP-pest expression from the early-L3 to the young adult stage. Twenty animals from each of eight morphologically-defined stages were imaged (Figure 4L and Figure S4) and then qualitatively scored for GFP-pest expression in seam cells, hyp7 cells or lateral muscle cells (Figure 4M). Expression of the Plin-4::GFP-pest reporter peaked in hyp7 and seam cells at the mid- and late-L3 stage and then again at the late-L4 stage. In addition, the majority of animals exhibited a cessation of hyp7 and seam cell Plin-4::GFP-pest expression immediately after L4 ecdysis (Figure 4M). In contrast, Plin-4::GFP-pest expression in muscle cells displayed a different transcriptional profile. In the majority of animals, expression of GFP-pest in muscles peaked at L3 ecdysis, gradually diminished throughout the remainder of the L4 stage, and increased again at the young adult stage (Figure 4M). These results suggest that, while lin-4 is dynamically expressed once each larval stage, its promoter activity may be differentially regulated in distinct tissues.
Analysis of the Plet-7::GFP-pest and Plin-4::GFP-pest reporters in lin-42 loss-of-function (lf) animals demonstrated that mutants that alter either lin-42 b/c (lin-42(n0189)) or lin-42a/b (lin-42(ok2385)) isoforms display elevated Pmir-GFP-pest expression in late larval development (Figure 5A, B and Figure S2). These altered temporal expression patterns suggested that lin-42 may normally function to modulate aspects of miRNA transcription. To investigate the potential interactions between LIN-42 and transcriptional regulatory elements, we performed chromatin immunoprecipitation coupled to high throughput sequencing (ChIP-seq) using extracts prepared from animals harboring a functional, GFP-tagged allele of lin-42 (Figure 5C). From two independent biological ChIP-seq replicates derived from separate L4-staged extracts, we obtained 413 high confidence peaks corresponding to chromosomal regions in which LIN-42 is enriched (see Table S2 and Materials and Methods). In agreement with the hypothesis that LIN-42 regulates let-7 transcriptional activity, we find LIN-42 binding sites at conserved let-7 promoter regions that have been previously demonstrated to control let-7 expression (Figure 5C) [31], [35], [57]. Annotation of additional high confidence peaks revealed that 38% (158/413) of LIN-42 peaks fell within the promoters (defined as 2 kb upstream of each gene) of either coding or non-coding genes, 24% (99/413) fell within the introns of coding genes, 8% (34/413) fell within gene bodies and 29% (121/413) fell within other intergenic regions (Figure 5D). Comparison between LIN-42 peak frequency and their distribution relative to the closest annotated transcription start site (TSS) revealed that LIN-42 has two major regions of enrichment: 1) directly at TSSs and 2) at approximately 750 bp upstream of a TSS (Figure 5E). Of these high confidence peaks, 323 were also detected in LIN-42 ChIP-seq samples obtained using an antibody against endogenous LIN-42, suggesting that this list forms a short, but high confidence, group of LIN-42 target genes. A list describing all high-confidence annotated LIN-42 peaks is provided in Table S2. Using the Generic Gene Ontology Term Mapper, we found that numerous genes with high-confidence LIN-42 peaks can be categorized into groups that function in many diverse biological processes, including development, transport, small molecule metabolism, embryogenesis and growth (Table S3). Collectively, these results strongly suggest that LIN-42 plays a role (either directly or indirectly) in a broad range of biological processes and that it predominately interacts with the promoter regions of coding and non-coding genes to regulate their expression.
The genetic and regulatory relationships between lin-42 and lin-4 or let-7, as well as the overlapping temporal expression patterns of these three genes, suggest that lin-42 may play a role in modulating the dynamics of lin-4 and let-7 transcriptional activity. To directly test the idea that lin-42 regulates miRNA levels at the transcriptional level, we quantified the transcriptional profiles of Plin-4::GFP-pest and Plet-7::GFP-pest reporters in wild-type animals and lin-42(n1089) mutants. The n1089 allele of lin-42 deletes genomic sequences that eliminate the coding potential of the lin-42b and c isoforms (Figure 1A) and displays strong heterochronic phenotypes [39], [46], [47]. Importantly, these isoforms contain the domains, PAS-A and PAS-B, that most closely link LIN-42 to PERIOD, a protein involved in controlling the cyclical expression patterns of circadian-regulated genes [39], [47], [48], [72], [73]. We focused on quantifying the GFP intensities of 1) the hypodermal cells in L3-to-adult-staged Plin-4::GFP-pest animals and 2) the seam cells of similarly-staged Plet-7::GFP-pest animals. These tissues and stages were selected for analysis because the majority of well-characterized heterochronic phenotypes are detected in these tissues [16], [24], [39], [42], [46]–[48]. Expression levels for each transcriptional reporter were analyzed throughout eight defined and sequential stages that spanned from early L3 to young adult (Figure 6A–C and Figure S4). In agreement with our previous observations, expression of Plin-4::GFP-pest in hypodermal cells of wild-type animals is dynamic throughout development and displays two main peaks of GFP expression: one at the late-L3 stage and the other at the L4 molt (Figure 6A, B). Similar results are also observed in the seam cells of wild-type animals expressing the Plet-7::GFP-pest reporter (Figure 6B, C). One exception, however, is that that the first peak of Plet-7::GFP-pest expression occurs at the mid-L3 stage (Figure 6B, C). Surprisingly, we find that the cyclical pattern of expression of these reporters is not affected in animals carrying the lin-42(n1089) mutation; both lin-42(n1089) and wild-type animals display nearly identical Plin-4::GFP-pest and Plet-7::GFP-pest temporal expression patterns (Figure 6A, B and C). In contrast, the abundance of GFP-pest expression for each reporter is universally higher in lin-42(n1089) mutants as compared to similarly-staged wild-type animals (Figure 6 A, B and C). In the case of the Plin-4::GFP-pest reporter, higher levels of GFP-pest intensity are observed in hypodermal cells throughout all developmental stages, with the greatest difference occurring between the late-L3 and L3-molt stages (3.1 and 4.3 fold respectively)(Figure 6B). Interestingly, although Plet-7::GFP-pest expression in lin-42(n1089) mutants is also greater in seam cells between the late-L3 and L3-molt stages (2 fold each), Plet-7::GFP-pest expression in lin-42(n1089) and wild-type animals is practically indistinguishable from wild-type during the mid-L4 to the young adult stages (Figure 6C). Taken together, these results suggest that mutations that abolish the expression of PAS domain-containing LIN-42 isoforms do not alter the cyclical expression patterns of miRNA genes during development. Rather, these mutations alter the transcriptional output of miRNAs that display oscillatory expression patterns.
As demonstrated in Figure 5D and Table S2, LIN-42 binds the putative regulatory regions of multiple protein coding genes. This observation raises the possibility that LIN-42 may modulate the transcriptional output of other developmentally regulated genes, including those whose expression, like that of lin-4 and let-7, is also linked to the molting cycle. To determine if lin-42 mutants alter the transcriptional output of other cyclically expressed mRNAs, we observed the expression of two transcriptional reporters for genes involved in the molting process, Pmlt-10::GFP-pest and Pcol-12::mCherry-pest. In wild-type animals, Pmlt-10::GFP-pest transcription begins at the end of each larval period when the new cuticle is being synthesized [39], [41]. We monitored the expression of Pmlt-10::GFP-pest in F1 animals that had been exposed to control RNAi or two RNAi constructs that target all major isoforms of lin-42 and induce precocious expression of Pcol-19::GFP and adult alae [74]. As with the expression of Plin-4::GFP-pest and Plet-7::GFP-pest reporters, the Pmlt-10::GFP-pest reporter maintained its normal, oscillatory pattern of expression in lin-42(RNAi) animals. Quantification of Pmlt-10::GFP-pest reporter expression at the late-L4 stage (where Pmlt-10::GFP-pest normally peaks [39], [41]) indicates that lin-42 depletion does not alter the transcriptional output of the mlt-10 promoter (Figure 6D). In addition, quantification of the Pcol-12::mCherry-pest reporter in young adult lin-42(n1089) animals also indicates that mutations in lin-42 do not alter the temporal expression patterns or levels of the col-12 promoter (Figure 6E). Therefore, while lin-42 mutations alter the transcriptional output of the lin-4 and let-7 genes, lin-42 does not play an essential role in controlling the oscillatory expression patterns or transcriptional output of all genes whose expression is tied to the molting cycle.
Using an unbiased genetic approach, we sought to identify factors that modulate the expression of miRNAs that are critical for controlling temporal patterns of development throughout post-embryonic development. Our strategy was two-fold: 1) we sought to identify suppressors of heterochronic miRNA mutant phenotypes characterized by stage-specific alterations in temporal patterning and 2) we focused on identifying suppressors that preferentially alleviate phenotypes that result from a reduction in, rather than a complete loss of, miRNA expression. These efforts identified lin-42, the C. elegans homolog of the circadian period gene, as a component that not only modulates heterochronic miRNA expression, but also regulates the expression of a wide range of broadly expressed, and functionally distinct, C. elegans miRNAs. Previous genetic analyses implicated lin-42 as a heterochronic gene that normally inhibits the precocious expression of adult characteristics [39], [46]–[48]. The precise placement of lin-42 in the developmental timing pathway has been difficult to incorporate due to the observation that lin-42 mutations alter cell lineage programs that occur exclusively in late development, namely the transition from the L3 to the L4 stage [46]–[48], [51]. In addition, epistasis experiments with other developmental timing mutants suggest that its interaction with other heterochronic genes is complex [46], [47], [51], [75], [76]. Furthermore, unlike other components that control discrete aspects of temporal patterning and display monotonic expression patterns, lin-42 expression is highly dynamic, suggesting a reiterative role for it in the heterochronic pathway.
Results from our screens have identified five new alleles of lin-42 that suppress the adult-specific gene expression defects of hypomorphic alleles of heterochronic miRNAs. We also find that lin-42 corrects stage-specific cell fate specification defects present throughout larval and adult development in these miRNA mutants. These results indicate that lin-42 functions iteratively to control temporal cell fate specification by controlling the transcription of distinct miRNAs. In addition, we demonstrate that our newly-identified lin-42(lf) mutants precociously express adult-specific programs and that these defects are suppressed by mutations in components of the miRNA machinery. Accordingly, these data suggest that lin-42(lf) heterochronic phenotypes are due to an overexpression of specific miRNAs that control temporal patterning. Also, our results demonstrate that lin-42 mutations are not bypass suppressors of the heterochronic phenotypes displayed by lin-4 and let-7 null mutants, suggesting that lin-42 suppresses retarded heterochronic phenotypes by increasing the expression of heterochronic miRNAs or enhancing their effectiveness in regulating miRNA targets.
Multiple lines of evidence described in this manuscript support the conclusion that LIN-42 regulates the transcription of a wide array of miRNAs. First, we investigated how our lin-42 suppressor alleles affected the overall levels of a subset of miRNAs involved in developmental timing. alg-1(ma192) animals display profound defects in temporal cell fate specification and also under-accumulate both lin-4 and let-7 miRNAs. Genetic and molecular experiments indicate that lin-42 suppresses alg-1(ma192)-dependent phenotypes by increasing the available amount of mature miRNAs. Second, lin-42(ma206) mutants over-accumulate multiple miRNAs, including those with no apparent role in developmental timing. Consistent with the hypothesis that lin-42 functions in additional gene regulatory pathways that require miRNA activity, we demonstrated that lin-42(lf) mutants suppress phenotypes associated with the under-accumulation of a miRNA that is essential for proper neuronal cell fate specification. Because lsy-6-mediated regulation of cog-1 expression is dosage-dependent, we speculate that lin-42(lf) mutations suppress neuronal cell fate specification defects by de-repressing lsy-6 transcription in ASEL neurons.
In order to understand how lin-42 may modulate miRNA expression, we pursued two lines of inquiry. First, we constructed a series of reporters that allowed us to measure, in detail, the transcriptional dynamics of multiple miRNAs in developing animals. Using these reporters, we found that several heterochronic miRNAs, such as lin-4 and let-7, exhibit highly dynamic expression patterns that are synchronized with the expression of genes required for each molting cycle. Importantly, the expression of the Plin-4::GFP-pest and Plet-7::GFP-pest reporters coincided with the transcriptional activation of lin-42. Further analysis of the lin-4 and let-7 reporters in a lin-42 mutant background indicated that one function of lin-42 is to negatively regulate the transcriptional output of miRNA promoters. Therefore, LIN-42 functions in a manner similar to the human and Drosophila PERIOD proteins, which inhibit the transcription of circadian regulated genes [77], [78]. Second, it has previously been shown that LIN-42 is a nuclear protein, which suggests that it may play a role in directly regulating the pulsatile expression patterns of its downstream targets [48]. In order to explore potential roles for LIN-42 in directly controlling aspects of miRNA transcription, we performed ChIP-seq experiments to determine if LIN-42 interacts with the putative regulatory regions thought to control the expression of miRNAs and mRNAs. These experiments demonstrated that LIN-42 interacts with the promoters of non-coding genes (including let-7) as well as protein-coding genes, suggesting that lin-42 may regulate the temporal expression of broad class of genes.
Given the role of human and Drosophila period in regulating circadian gene expression, we were surprised to find that animals harboring the lin-42(n1089) allele, which abolishes the expression of PAS-containing lin-42 isoforms, maintained lin-4 and let-7 periodic expression patterns in later larval development. The PAS domains of human and Drosophila PERIOD are absolutely required to maintain the oscillatory expression patterns of circadian-regulated genes [72], [77], [78]. In our experiments, peak expression of the lin-4 and let-7 transcriptional reporters occurred at roughly the same developmental stages in both wild-type and lin-42(n1089) animals. Interestingly, although the temporal expression patterns were similar, the levels of each reporter were elevated (as high as four fold) in lin-42(n1089) mutants as compared to wild-type animals (Figure 7A and B). Notably, lin-42(n1089) mutations do not alter the expression of the lin-42a isoform, which has been implicated in controlling the periodicity of the molting cycle [39]. While the dissection of lin-42 function will require further study, these findings are consistent with the modular nature of LIN-42 activities and suggest a novel role for the PAS domains of LIN-42 in regulating the transcriptional output of periodically expressed genes.
Based on our current observations, we propose a model in which each of the lin-42 isoform functions to sculpt the dynamic transcription of both miRNAs and mRNAs. In out model, cis-regulatory elements within the promoters of specific miRNAs and mRNAs would be sufficient to drive periodic transcription. Regulatory elements within these sequences would be bound by a sequence-specific transcription factor (TF) that would promote the periodic transcription of these genes near the end of each larval stage (Figure 7C). Based on the role of PERIOD in other organisms and our data demonstrating that LIN-42 binds to the putative cis-regulatory elements of several miRNAs and mRNAs, we propose a model in which distinct isoforms of LIN-42 function to regulate the activity of the TF at multiple, genetically separable levels. Our evidence suggests that mutations that specifically disrupt isoforms containing the PAS domain (LIN-42B and LIN-42), fail to properly limit the transcriptional output of genes regulated by the temporal specific TF (Figure 7C). As a consequence, although these mutants display essentially normal temporal patterns of miRNA transcription, the elevated levels of heterochronic miRNAs lead to precocious developmental phenotypes. Importantly, mutations that only alter PAS-domain containing isoforms of LIN-42 retain the expression of LIN-42A (Figure 7B) [39]. Our model would also predict that mutations that disrupt LIN-42 isoforms that contain the conserved SYQ/LT domains (LIN-42A and LIN-42B) would have complex phenotypes with regard to periodic transcription. Indeed, animals harboring the lin-42(ok2385) allele, which disrupts the expression of the LIN-42A isoform (containing the SYQ/LT domains only) and deletes portions of the LIN-42B isoform (containing both the PAS and SYQ/LT domains), precociously execute stage-specific gene expression, fail to maintain periodic molting cycles and overexpress Pmir::GFP-pest transcriptional reporters [39](Figure S2). We interpret the complex phenotypes of lin-42(ok2385) animals as a reduction of the two modular activities of LIN-42 domains. Specifically, a reduction of LIN-42 PAS domain expression alters transcriptional output and deletion of LIN-42 isoforms which contain the SYQ/LT domains results in defects in periodic transcription. Further studies will be needed to define a specific molecular role for LIN-42 isoforms in maintaining normal periodic transcription.
Recent reports suggest that a significant portion of the C. elegans transcriptome is dynamically expressed [79]–[81]. The combined interpretation of these studies suggests that the post-embryonic expression of 5–20% of mRNAs is synchronized with the molting cycles. The conservation of this process implies that these temporal gene expression patterns confer fitness to an organism and raise a number of interesting questions regarding the nature of developmental gene regulation [81]. Because many genes whose oscillatory expression patterns are coupled to the molting cycle control cell fate decisions and cell metabolism in a dosage-dependent manner, it is interesting to speculate how their temporal expression patterns, and levels, are coordinated with their targets. We suggest that lin-42 plays a fundamental role in this process for a wide range of non-coding and protein-coding genes. Because many of the transcriptional targets of lin-42 include miRNAs, each of which may regulate a vast array of genes, the impact on the dynamic nature of the C. elegans transcriptome during development may be immense.
C. elegans strains were grown under standard conditions and mutagenized as previously described [82]. Positional cloning of each suppressor was performed using standard methods [50]. Transformation of animals and integration of extrachromosomal arrays were performed as previously described [83]. See Text S1 for details of transgenic animals used in this manuscript.
Lineage analysis and scoring of adult alae phenotypes were performed by picking staged animals of the indicated genotypes and monitoring seam cells derived from the V lineage as previously described [22]. All images were taken with an Axio Scope.A1 microscope equipped with a monochrome camera (Diagnostic Instruments Inc) and SPOT imaging software (SPOT Imaging Solutions). GFP images of the hypodermal and seam cells were used for further quantification of Pmir::GFP-pest intensity. The average GFP intensity per area (arbitrary units) was quantified using ImageJ64. For each reporter, 20 individual animals were analyzed per developmental stage. For Plin-4::GFP-pest, 10 hypodermal cell nuclei per animal per stage, or a total of 200 nuclei per time point, were used to calculate the average GFP intensity. For Plet-7::GFP-pest, 5 seam cells per animal per stage, or a total of 100 cells per time point, were used to calculate the average GFP intensity.
Total RNA was isolated from staged populations of worms, and northern blots were performed as previously described [27]. Multiplex microRNA TaqMan assays were performed according to the manufacturer's specifications (Life Technologies) and quantified using the ABI 7900HT Fast Real-Time PCR system (Applied Biosystems). For each biological replicate (3 total), the means and standard deviations of the raw Ct values were calculated and the representative heatmap demonstrating the fold change signal was created using R packages (www.r-project.org).
For the characterization of behavioral and GFP/mCherry reporter expression, animals were prepared in one of two ways. For analysis of L1-stage expression, embryos were bleached and staged according to standard protocols and then plated on standard NGM media with OP50 [84]. At indicated times after the release from L1 synchronization, L1-staged animals were imaged with an Axio Scope.A1 microscope. For analysis of the molting cycle and GFP/mCherry-pest reporter expression, individual animals (non-motile, non-pharyngeal pumping) were picked to fresh NGM plates seeded with 20 µL of OP50. Time courses were initiated for each animal after each animal ecdysed. To determine the active and lethargic periods of animals at each stage, the pumping rates of individual animals were observed for 30 s of every hour. GFP/mCherry-pest expression was then monitored using a Zeiss SteREO Discovery V12 microscope with appropriate filters. To prevent photo-bleaching, each animal was exposed to <3 s of UV light.
See Text S1 for details.
|
10.1371/journal.pntd.0003003 | Long-Term and Seasonal Dynamics of Dengue in Iquitos, Peru | Long-term disease surveillance data provide a basis for studying drivers of pathogen transmission dynamics. Dengue is a mosquito-borne disease caused by four distinct, but related, viruses (DENV-1-4) that potentially affect over half the world's population. Dengue incidence varies seasonally and on longer time scales, presumably driven by the interaction of climate and host susceptibility. Precise understanding of dengue dynamics is constrained, however, by the relative paucity of laboratory-confirmed longitudinal data.
We studied 10 years (2000–2010) of laboratory-confirmed, clinic-based surveillance data collected in Iquitos, Peru. We characterized inter and intra-annual patterns of dengue dynamics on a weekly time scale using wavelet analysis. We explored the relationships of case counts to climatic variables with cross-correlation maps on annual and trimester bases.
Transmission was dominated by single serotypes, first DENV-3 (2001–2007) then DENV-4 (2008–2010). After 2003, incidence fluctuated inter-annually with outbreaks usually occurring between October and April. We detected a strong positive autocorrelation in case counts at a lag of ∼70 weeks, indicating a shift in the timing of peak incidence year-to-year. All climatic variables showed modest seasonality and correlated weakly with the number of reported dengue cases across a range of time lags. Cases were reduced after citywide insecticide fumigation if conducted early in the transmission season.
Dengue case counts peaked seasonally despite limited intra-annual variation in climate conditions. Contrary to expectations for this mosquito-borne disease, no climatic variable considered exhibited a strong relationship with transmission. Vector control operations did, however, appear to have a significant impact on transmission some years. Our results indicate that a complicated interplay of factors underlie DENV transmission in contexts such as Iquitos.
| Description of long-term temporal patterns in disease occurrence improves our understanding of pathogen transmission dynamics and facilitates predicting new epidemics. Dengue, the most prevalent mosquito-borne, viral disease of humans, typically varies seasonally and on longer, inter-annual time scales. In most studies of these patterns, however, only a fraction of putative dengue cases are confirmed with laboratory diagnostics. Here we analyzed 10 years of fully confirmed dengue cases reported to a sentinel surveillance system in Iquitos, Peru. We describe the inter and intra-annual patterns of weekly case counts and relate these to climate and local vector control efforts. We show that dengue case counts vary seasonally in Iquitos despite very little variation in key climatic conditions, such as temperature and humidity. Overall, transmission correlated poorly with climate regardless of time lag. In seasons when vector control was conducted early, there was an apparent decline in cases later that season. We speculate that the relationships between climatic conditions and transmission of DENV in Iquitos are complex and non-linear, and that other factors, such as herd immunity, virus diversity, and vector control efforts, play key roles determining the timing and intensity of transmission.
| Dengue is a mosquito-borne disease common throughout the tropics and sub-tropics [1], [2]. It is caused by infection with any of four antigenically-distinct, but related, dengue viruses (DENV-1, 2, 3, and 4) in a human-mosquito transmission cycle. The anthropophilic mosquito, Aedes aegypti, is the predominant vector [3], [4]. The long-term patterns of dengue incidence have been studied at numerous endemic sites, especially in Southeast Asia [5]–[12] and the Americas [8], [13]–[15]. Results highlight intra-annual (seasonal) and inter-annual (across multiple years) signatures in transmission intensity [8], [10], [16], [17], as well as occasional abrupt shifts in the age of people with clinically apparent illness [18]. Conclusions from these studies are mixed, although in aggregate they highlight that dengue occurs across a diverse array of conditions and that the key drivers of transmission similarly vary across those different contexts [8], [10], [16]. Continued, detailed documentation of these temporal dengue patterns in different, endemic populations is useful for improving our understanding of DENV transmission and testing the link of key variables like temperature to components of the virus transmission cycle [10], [19]–[21]. With this goal in mind, here we examined the temporal patterns of laboratory-confirmed dengue cases over a 10-year period encompassing the introductions of two novel serotypes into the Amazonian city of Iquitos, Peru.
Despite their informational value, long-term disease data sets often lack detail because of the costs associated with detection of potential cases and laboratory-based diagnosis [22]. Furthermore, the symptoms associated with dengue fever are non-specific and can lead to misdiagnosis [23], [24]. Nevertheless, many surveillance systems report suspected cases with confirmation of only a small fraction. While severe, hospitalized cases are less prone to misdiagnosis and are usually laboratory confirmed, they typically represent only a small proportion of the total number of people infected [25]. Moreover, severe disease outcomes are influenced by a variety of intrinsic factors (e.g., virus virulence, host exposure history) [26] and not necessarily external drivers, such as climate conditions.
Limitations of many long-term dengue datasets analyzed to date [9], [13], [15], [e.g. 27]–[30], in addition to variation in reporting methods, increase the difficulty and reduce confidence in defining universal properties of dengue transmission dynamics [8]. Johansson et al. [16] concluded that results of these analyses are sometimes biologically implausible and confusing, such as a negative effect of increasing temperatures on transmission [see references in 16]. Because transmission is seasonal, it will correlate with other seasonal patterns even though there is no mechanistic link. Thus, any statistical analysis should be rigorously scrutinized from a biological perspective and, preferably, cross-validated with additional data. A recent study analyzed seasonal dengue in Ecuador using linear mixed models incorporating entomological, epidemiological, and climate data [15]. The investigators found important influences of climate and entomological indices on monthly dengue case counts. Nevertheless, even using new and improved modeling approaches, in aiming to fit a particular statistical model to temporal disease data–which is often aggregated–to predict transmission patterns over time, the analysis potentially obscures other features of the time series that might generate hypotheses about underlying mechanisms.
Here, we examined the seasonal patterns of dengue over a 10-year period in relation to climatic factors and citywide vector control efforts. Our analysis focused on laboratory-confirmed dengue fever cases reported to a surveillance network based in multiple health-care facilities in Iquitos, Peru. During the period of study, two novel DENV serotypes invaded Iquitos, which was already endemic for DENV. In response to the invasions and subsequent epidemics, the local ministry of health conducted citywide house-to-house insecticide fumigation campaigns to kill adult mosquitoes and reduce virus transmission. Our analyses indicate that, although climatic variables correlate weakly with variation in transmission intensity, mosquito control efforts do appear to curtail epidemics when properly applied.
Iquitos is a city of ∼377,000 inhabitants that sits at the confluence of the Nanay, Itaya, and Amazon Rivers in the department of Loreto in northeast Peru. Iquitos has been thoroughly described in previous publications [23], [31]–[35]. In 2000, as part of a collaborative effort between the Peruvian Ministry of Health and the U.S. Naval Medical Research Unit No. 6, a surveillance network was established in public and military hospitals and clinics throughout Iquitos. For most years, 12 or 13 health centers participated, representing predominantly urban and peri-urban areas in and around Iquitos. A core of 3 hospitals and 6 clinics consistently provided samples throughout the study. A few health centers discontinued participation mid-study but were replaced by other health centers from the same geographic area. Additional details are described in Forshey et al. [23].
All data collection was conducted under study protocol NMRCD.2000.0006, approved by the Naval Medical Research Center Institutional Review Board (Bethesda, MD) in compliance with all U.S. Federal regulations governing the protection of human subjects. In addition, the study protocol was reviewed and approved by health authorities in Peru (Dirección General de Epidemiología). Written consent was obtained from participants 18 years of age and older. For participants younger than 18 years, written consent was obtained from a parent or legal guardian. Additionally, written assent was obtained from participants between 8 and 17 years of age. Prior to analysis, all data were de-identified and aggregated into weekly case counts.
Details of the surveillance system, including inclusion criteria and laboratory assays are detailed in Forshey et al. [23]. Briefly, consenting participants (≥5 years old) provided an acute blood sample on the day they visited the health care facility for laboratory confirmation of DENV infection. Laboratory procedures included RT-PCR and virus isolation to identify acute infections and IgM ELISA to detect anti-DENV antibodies consistent with a recent infection. Convalescent samples collected 10 days to 4 weeks later were tested for anti-DENV IgM by ELISA. We identified the infecting serotype when possible (55%); positive diagnosis was generically defined a “DENV” infection when based solely on IgM assay results (Table 1).
In response to dengue outbreaks in Iquitos over the period of study, the Loreto Regional Health Department (LRHD) conducted large-scale vector control interventions (Table 1). In these, they sprayed inside houses with an ultra low volume (ULV), non-residual insecticide (deltamethrin [2002–2006], cypermethrin [2006–2008], or alpha-cypermethrin [2008–2010]) three times over approximately a three-week period. The LRHD attempts to treat all houses within designated sectors of the city, which are chosen based on epidemiological information (Table 1). These citywide efforts usually treated ∼40% of all houses in Iquitos, which total ∼80,000 houses. Data on interventions were provided by the LRHD. For our analyses we identified weeks when fumigation was conducted in the city and examined whether treatments were associated with reductions in dengue incidence within and across years.
Case data were restricted to the period between 1 July 2000 and 30 June 2010. Positive cases were those with evidence of virus (RT-PCR or virus isolation) or immunologic evidence of recent infection (acute or convalescent IgM ELISA titer>1∶100). We combined all DENV+ cases into weekly totals for use in correlation and wavelet analyses (see below).
Generally, people visiting health centers were received for 5–7 hours a day, 5 days a week, although there was some variability in rates across seasons and clinics. A major exception was a 2-week period of 2004 when surveillance in one hospital was extended to 24 hours a day due to the large number of dengue cases they were receiving. To correct for this extended effort, we rescaled the number of cases captured in these 2 weeks by the ratio of the maximum number of negative cases observed in the remainder of the time-series to the number of negative cases observed during those particular weeks (approximately 1∶5). For disaggregated analyses, the data were randomly thinned in these two weeks based on the same scaling factors.
Using the corrected time-series, we conducted autocorrelation analysis to characterize the temporal structure of the case data. Subsequently, we used wavelet analysis to identify temporal variation in the periodicity of dengue case reports. Our analysis was conducted on the square-root transformed and normalized (by standard deviation) time series using the Morlet wavelet transform and implemented in Matlab using the algorithm of Torrence and Compo [36].
Daily climate data for Iquitos was acquired from a US National Oceanic and Atmospheric Administration (NOAA) weather station located at the Iquitos airport. Reported variables include: mean, maximum, and minimum temperatures; precipitation; air pressure; wind speed; and dew point. From these data we generated several derived variables, including: daily temperature range (DTR; max - min), degree-days (DD), relative humidity (RH; 100 - 5*(Temp_mean - dewpoint)), and precipitation events (per week). We calculated degree-days using the triangle method and a 24°C threshold temperature for virus replication [DD 24]; [ 37,38]. We considered the river depth of the Amazon River as a covariate, because this variable changes dramatically over the course of the year as a function of rainfall in the Andes Mountains. At high river levels, fringe areas of Iquitos have occasionally flooded, which could have had an impact on mosquito populations. It is more probable, however, that river depth serves as an indicator of broader scale climate patterns that might correspond with conditions suitable for DENV transmission. River depth data for the Amazon River in meters above sea level was provided by the Servicio Nacional de Metereologia e Hidrologia, Peru.
Seasonal and annual climate patterns were summarized graphically using a loess smoother, which summarizes the data by fitting a local polynomial [39]. The degree of smoothing desired is controlled by the parameter α, where large values indicate more smoothing. We heuristically chose values of α to emphasize short-term and long-term temporal patterns in the data.
Because climate variables are highly collinear, interpretation of the relationship between any single variable and epidemiological patterns could be misleading. Maximum and minimum temperatures, for instance, should correlate in time. To address this issue, we conducted principal components analysis (PCA) on the climate variables. Briefly, PCA reorients a set of n covariates into n principal components (PCs) based on their covariation structure. The first PC (PC1) always captures the largest proportion of the covariance between the covariates, with successive PCs explaining less and less of the remaining variation. With more correlation among covariates, fewer PCs are required to capture most of the variation in the dataset. By definition, the resulting principal components are orthogonal with each other (i.e., they do not correlate) and the set of n PCs exactly encapsulates all covariation among the covariates. Within a PC, loading values describe the relative contribution of each original covariate. Higher loadings indicate greater correlation and high loadings on the first few PCs indicates the overall importance of that covariate in the covariation structure of the dataset.
We examined the relationships between weekly DENV cases and climate variables using cross-correlation maps [CCMs; 40]. For each variable, maps were generated by varying the temporal lag and the period over which the variable was aggregated. Briefly, cases in week t0 were correlated with each covariate aggregated over a range of weeks prior to t0, defined by the interval [t0-a, t0-b]. We evaluated the mean, median, and maximum values of the covariate, and, in some cases, the sum for each period. We present results for the median unless the sum was more appropriate. For example, it is possible that rain influences dengue cases 4–8 weeks later because of effects on mosquito population dynamics. In that case, we set a = 4, b = 8 and looked at the correlation of maximum rainfall over that interval with the number of dengue cases a month in the future. To limit the identification of spurious correlations, we did not investigate lags more than half a year before the cases were observed (27 weeks). We believe, however, that effects most likely to have biological relevance on transmission would occur within a lag of 17 weeks (1 trimester). To investigate both linear and monotonic associations between climatic variables and cases we calculated Pearson and Spearman correlations. We categorized the correlation coefficient, r, as follows: |r|<0.1, no correlation; 0.1≤|r|<0.2, very weak, 0.2≤|r|>0.3, weak; 0.3≤|r|<0.4, weak moderate; 0.4≤|r|<0.5, moderate; |r|≥0.5, moderate strong to strong. Because of the large number of tests conducted (each CCM equates to 338 correlation tests), we did not calculate p-values and rather focused on the relative strength of correlations. Unless otherwise stated, all analyses were conducted with R 2.13.
Over the 10 years of study, 12,602 febrile participants were enrolled, 40% of whom were laboratory diagnosed as having acute or recent DENV infection (Table 1). Although very few dengue cases were detected at the beginning of the study (consistent with serology data [34]), after late 2001, outbreaks occurred on an annual basis (Figs. 1, 2). Overall, weekly case reports fluctuated seasonally (i.e., DENV positive and negative cases together; Fig. 1). Cross-correlation analysis showed that the number of cases diagnosed as something other than dengue (DENV negative cases) mirrored the number of dengue positive reports (i.e., the best lag was 0; Fig. 1c).
Wavelet analysis indicated that the annual periodicity in transmission was particularly strong from the 2004–2005 season forward (Fig. S1). A longer, ∼3 year periodicity was also suggested by the analysis, but the 10 year time-series was too short to place confidence in this result. Over all years, 75% of DENV cases were reported between the 37th week of the preceding year and 13th week of the subsequent year, peaking on average in the last week of December (Fig. 2). We thus define the dengue season in Iquitos as occurring between September and April (between trimester III and trimester I of the subsequent year).
Over the 10 dengue seasons single serotypes accounted for the majority of all cases. DENV-1 was dominant in the first season, followed by the emergence of DENV-3 in 2001, [34], [genotype III;41], and the emergence of DENV-4 in 2008 [genotype II; 42](Fig. 1, Table 1). DENV-2 (lineage I of American/Asian genotype) was only detected in a few study participants in 2001–2002. DENV-1 appeared at low levels in 2002–03 and 2005–06 when DENV-3 was dominant.
Although transmission intensified on an annual basis, the magnitude and timing of the peaks varied across seasons. Temporal autocorrelation of the number of weekly DENV cases indicates a strong positive auto-correlation at a lag around 2 years and a negative correlation around a lag of 1.5 years (Fig. 2). This result is consistent with an apparent shift in the timing of peak transmission from year to year (Fig. 2). In other words, the inter-epidemic period fluctuated between approximately 8 and 16 months.
Climatic variables demonstrated seasonality in Iquitos, although the magnitude of variation was small (Table 2, see SI). Maximum and mean weekly temperatures were warmest in trimesters III and I (between November and April), coinciding with the timing of detection of most dengue cases (Figs. S2, S4). Mean and minimum temperatures showed a gradual increasing trend over the 10 years, culminating in a ∼1°C increase between 2000 and 2010 (Fig. S4, S6). Cumulative weekly DD24 largely mirrored trends in mean and maximum temperatures, peaking in late trimester III (November [28.04°C•days]; Fig. S8) and bottoming in trimester II (June [19.27 C°•days]). The 10-year trend in DD24 was highly non-linear, lowest in early 2008 and increasing rapidly to its highest levels in 2009 and 2010 (Fig. S8). Precipitation occurred throughout the year, but it was usually lower in the later part of trimester II (July [3.77 cm•week−1] and August [4.66 cm•week−1]; See Table 2, Fig. S12). Over all years, rainfall amounts were highest between 2003 and 2008, dropping significantly in later years, although the number of precipitation events remained the same (Fig. S14). Additional climatic variables are shown in the SI.
Because climate variables correlate, we conducted principal components analysis (PCA) to simplify the data and identify subsets of highly collinear drivers. The results of the PCA identified three components that described 79% of the variation among the climate variables (Table 3). The first, PC1, related most strongly to temperature variables and humidity. PC1 increased with increasing humidity and decreased with increasing temperatures. The second, PC2, captured variability in temperatures. PC2 decreased with increasing minimum temperature and river level and increased with larger DTR. The third component, PC3, decreased with precipitation and wind speed and increased with river level (Table 3). All three principal components exhibited seasonal periodicity, although this was attenuated for PC3 in later years. (Fig. S21, S23, S25).
Taken together, conditions in Iquitos can be described by three seasons: In trimester I, temperatures are warm, rainfall is elevated, the level of the Amazon river is increasing and dengue cases subside; in trimester II, conditions are relatively cooler and drier, the river begins to subside, and there are few dengue cases; in trimester III temperatures are their warmest and precipitation increases, the river subsides to its lowest levels, begins to rise again, and dengue transmission picks up.
We related weekly reported dengue cases to climate variables using temporal cross-correlation maps (CCMs; Fig. 3; see Methods and SI). Because pair-wise relationships to individual climate variables can be misleading and conflated by collinearity between climate variables, we first examined CCMs of the three principal components described earlier. We subsequently considered specific individual variables commonly associated with DENV transmission. In all instances, we produced CCMs for the whole year and for trimesters I and III (Fig. 3), when most DENV transmission took place (see above). Overall, CCMs showed that there was a correlation between most climatic variables or their components and reported dengue cases, although the correlations—especially on an annual basis—were often weak (|r|<0.3; Fig. 4). For each CCM, we identified the maximum absolute r and plotted weekly case reports against the climate covariate to characterize the nature of the relationship (linear, non-linear; Fig. 3).
We first examined the relationship between weekly dengue reports and the first three principal components, which consolidate highly collinear variables into orthogonal components (Table 3). The first component, PC1, which associated negatively with temperature variables and positively with RH, correlated weakly and positively with dengue cases when aggregated over a broad period from 17 to 1 week earlier (Figs. 3, 4, S23). This means that a period of relatively lower temperatures and elevated RH preceded high case counts. When we focused only on trimester III, the correlation was weaker (0.18) and the lag was greater ([−26, −21]; Figs. 3, 4). In trimester I, the correlation was stronger (0.29) and the lag was less ([−9, −6]; Figs. 3, 4). The second component, PC2, aggregated over 26 to 12 weeks prior, correlated more strongly (weak moderate) with cases on an annual basis (−0.34, [−26, −12]; Figs. 4, S25). In the principal components analysis, PC2 correlated most strongly with minimum temperatures and DTR, thus when minimum temperature was high and DTR was small 3–6 months previous, case counts were elevated (Table 3). In trimester III, PC2 again correlated negatively (−0.32) with cases, but at a smaller lag ([−6, −2]; Figs. 4, S25). The PC2 correlation and lag for trimester I was similar to the annual pattern (−0.3, [−26, −6]; Figs. 4, S25). Finally, PC3, which correlated most strongly with wind speed and river level, showed a weak correlation with cases on an annual basis (0.12, [−7, −6]; Figs. 4, S27). In trimester III, PC3 correlated weakly and positively at a large lag (0.22, [−26, −24]). In trimester I, the correlation was negative and strongest at a large lag as well (−0.33, [−26, −20]; Figs. 4, S27). PC3 also correlated positively with cases at shorter, biologically relevant lags in this trimester (Fig. S27). Pearson and Spearman correlations for PC1 and PC2 were similar (Fig. S28). PC3, however, differed markedly in annual and trimester III CCMs (Figs. S27, S28).
Within what we considered a biologically relevant window of 17 weeks, PCs 1 and 2 correlated with cases on an annual basis. By trimester, only PC2 correlated significantly in trimester III and both PC1 and PC2 correlated in trimester I, although the relationship with PC2 was distributed over a broad range of lags.
Examination of scatterplots relating components to weekly cases revealed distinct non-linear patterns. The number of cases increased more rapidly with increasing PC1 [−17, −1] than expected of a linear relationship (Fig. 3). There was a considerable increase in variation in the number of cases each week at higher values of PC1 (i.e., at lower maximum/mean temperatures and increasing humidity). Thus, few cases should be expected when PC1 is low 1 to 17 weeks earlier, but it is uncertain how many cases will result when PC1 is elevated over the same period. The patterns by trimester were mostly similar. Conversely, the number of cases decreased more rapidly than expected (linear) in relation to increasing PC2 (Fig. S25). The scatterplot of cases against the best PC2 lag shows a decrease in both the mean and variance of cases as PC2 increases, indicating that the weeks of highest incidence occurred when PC2 was very low (high minimum temperature, low DTR; Table 3) between 26 and 15 weeks before. As with the relationship between PC1 and cases, due to heteroskedasticity, high values of PC2 always correspond to few cases. The patterns were similar by trimester, except the lag was much less in trimester III. Finally, the scatter plot of cases relative to PC3 showed a distinct humped pattern with most transmission occurring when PC3 was between −0.5 and 0.5, suggesting that there is a stronger association between this component and cases than that measured with simple correlation (Fig. S27). Partitioning this analysis by trimester partly resolved this non-linearity: in trimester III the relationship is positive and linear while in trimester I it is negative and linear (Fig. S27).
Mosquito development and virus replication in the mosquito are temperature dependent [43], so ambient temperatures are often thought to play an important role in DENV transmission [37], [44]. Precipitation, too, is often thought to be a key local variable influencing DENV transmission because mosquitoes require aquatic habitats for larval development [8], [43]. Relative humidity combines aspects of temperature and precipitation and is probably directly important for mosquito survival because it influences desiccation rates. All of these variables naturally correlate with each other and for this reason we focused on the analysis of principal components. When considering individual variables, however, we found that correlations on an annual basis were mostly weak (Fig. 4; See the SI for results, figures S8—S27). The number of precipitation events and relative humidity correlated strongest at relatively large lags (Fig. 4). Several individual variables, temperature related variables in particular, correlated with case reports within a 17-week lag (Fig. 4). In trimester III, maximum temperature and DD24 showed moderate negative correlations, but at very large lags. Within our biologically relevant window of 17 weeks, only minimum temperature, RH and river level showed appreciable correlations in this trimester (Fig. 4). In trimester I, precipitation events, RH, wind, and river level were most strongly correlated with weekly case numbers, but at large lags. Only precipitation and wind speed correlated within a lag of 17 weeks (Fig. 4).
There was evidence of non-linear relationships and heteroskedasticity in many instances (see, for example, mean temperature in Fig. S5). These were occasionally resolved when portioning the analysis by trimester. That is, a positive relationship in trimester III changed to a negative relationship in trimester I.
On an annual basis, results for Spearman correlations were largely similar to those for Pearson correlations, although the correlations were stronger and extended over a longer period for temperature covariates (Fig. S28). The one exception was DTR, which correlated positively in Pearson tests, but negatively in Spearman tests at a shorter lag—although in both cases the correlation was very weak and may not be important (Fig. S11). On a trimester basis, several variables correlated well with weekly DENV cases within a 17-week lag. These were, for trimester III, minimum temperature (0.39 [−2, 0]), DTR (−0.44 [−8, 0]), and RH (−0.42 [−15, −4]; Fig. S28).
In addition to climatic variation, city-wide efforts to fumigate households with insecticide to curtail transmission hold large potential for shaping inter and intra-annual patterns of transmission in Iquitos. Using data provided by the LRHD on their vector control efforts, we assessed the potential effect of citywide interventions on the number of reported dengue cases by plotting cases in week t0 with the total number of cases in the subsequent 3 weeks. We split the data by whether an intervention was taking place in week t0 and by trimester (Fig. 5). As indicated above, in trimester III dengue outbreaks were usually beginning and so the relation between cases this week and cases over the following three weeks was approximately 1∶1 or greater (compare black and red lines in Fig. 5). In seasons when an intervention was conducted in trimester III (blue points), however, the relation was less than 1∶1, which indicates a reduction in the rate new cases were captured. Conversely, in trimester I transmission was subsiding and the relationship was usually less than 1∶1 even in the absence of vector interventions. Moreover, there did not appear to be any impact of interventions when they were conducted in trimester I (compare black and blue lines). That is, when interventions were conducted in trimester I any reduction in transmission was masked by the natural decline in the number of new cases reported. Over the full 10 year study period, when transmission and interventions both occurred in trimester III there appeared to be lower transmission in the subsequent trimester I (Fig. 5). We did not observe any seasons with high trimester III transmission without any intervention activities.
Dengue was not reported in Iquitos from the late 1970s—the end of the hemisphere-wide campaign to eradicate Ae. aegypti from the Americas—until a DENV-1 outbreak in 1990 [45]. Continuous DENV transmission has been detected since that time. DENV-2 American was detected in 1995 [46]. Over the period of this study, 2000–2010, DENV-3 [34] and then DENV-4 [42] invaded the city. DENV-3 was dominant over 6 transmission seasons until it was replaced by DENV-4 in 2008 [17]. Virus transmission dynamics in Iquitos have, therefore, been largely due to single serotypes and marked by annual periodicity, suggestive of seasonal forcing. The magnitude and timing of outbreaks were variable from year to year. Because of the obvious seasonality of dengue in Iquitos and elsewhere, we examined the role of climatic drivers in transmission dynamics. Our descriptive analysis of temporal variation in dengue cases in relation to climate did not, however, resolve clear relationships. The magnitude of seasonal climatic variation in Iquitos was quite small and at least low-level transmission was detected year-round. On an annual basis, almost all of the climatic variables we considered correlated weakly (|r|<0.3) with the number of dengue cases reported each week, with a few exceptions that were only slightly better correlated (e.g., relative humidity). Partitioning the analysis by trimester revealed stronger relationships, but most of these were distributed over very long lags (>20 weeks), suggesting that the observed correlation was due to the phase difference between seasonal signals and not a mechanistic link. Principal components analysis facilitated interpretation of the observed patterns, but generally highlighted that the relationship between climate and dengue in a place like Iquitos—where climate conditions may be suitable for transmission year-round—is complex, with no single dominant climate driver. Finally, citywide vector control efforts targeting adult mosquitoes—depending on their timing—appeared to reduce transmission.
In many regions of the world, particularly Southeast Asia, dengue epidemiology is characterized by co-circulation of multiple serotypes [47]. Serotype co-circulation complicates analysis of disease dynamics because the different virus serotypes interact immunologically at the level of the host and may be differentially transmitted by local mosquito vectors [25], [48]–[50]. After examining laboratory-confirmed dengue cases reporting to a network of clinics and hospitals in Iquitos, Peru, we provide a different perspective on the dynamics of this disease from that reported for other contexts. In this isolated population of ∼400,000, transmission has largely been dominated by single serotypes. On a few occasions, a small fraction of cases were due to other serotypes. Because of its population size [13], circulation of a single serotype (and genotype of a serotype) over multiple years and at least one confirmed dengue case in the majority of weeks (81%), we conclude that DENV is endemic and persists in Iquitos year-round. Dengue is not hyper-endemic (i.e., stable, year-to-year, co-circulation of multiple serotypes [47]), probably because of limited connectivity to other dengue endemic areas. Occasionally, new virus strains are amplified in other parts of Peru, Colombia and/or Brazil, from which they are introduced and become established in Iquitos. Indeed, the molecular epidemiology and timing of DENV-3 and DENV-4 emergence [41], [42] suggests that those viruses arrived to Iquitos via the Peruvian cities of Pucallpa to the south (population ∼120,000; DENV-3) and Yurimaguas to the southwest (population ∼48,000; DENV-4) both separated from Iquitos by a short flight or multi-day boat ride. There are no roads connecting those cities to Iquitos.
Our data show that dengue incidence in Iquitos follows a clear, seasonal pattern with the number of dengue cases peaking around December (calendar year trimesters III and I in this analysis). The timing of this peak varied year to year such that a short inter-epidemic period appeared to be followed by long inter-epidemic period. While this pattern is intriguing, our time series was too short to determine whether it is real and not a coincidence. Wavelet analysis suggests a 3-year cycle in incidence similar to that reported for hyper-endemic settings [27], but, again, 10 years is insufficient data to confirm this result statistically.
We find it compelling that transmission was distinctly seasonal, especially after 2004, even though the magnitude of seasonal variation in climate was very small. When looked at on an annual basis, PC2, which aggregated minimum temperature, DTR, and river level, was the best linear covariate. This correlation, however, was distributed over large lags and so may simply be the result of the phase difference between two seasonal signals. PC1, which aggregated temperature variables and RH, showed some correlation with cases and over shorter lags, but it was weak. PC3 showed a very weak linear correlation, but scatterplots indicated that the actual relationship was highly non-linear. PC3 aggregated precipitation and wind speed. When we partitioned the analysis by trimester, we observed that PC2 correlated with cases in a biologically reasonable time frame in trimester III and PC1 did so in trimester I. Neither of these correlations was very strong. Also, the non-linearity in the relationship between PC3 and weekly cases was partly resolved, i.e., the correlation was positive in trimester III and negative in trimester I. Altogether, the analysis of principle components with CCMs suggests, at best, weak climatic forcing of dengue transmission in Iquitos. This is confounded by the impacts of vector control (see below), herd immunity [17], [34], [51], and non-linearities in the relationships—in addition to the caveats associated with our analysis (see below).
Both RH and minimum temperatures have been cited elsewhere as strong correlates of DENV transmission [10], [13], [28], [29]. Precipitation, too, is commonly observed to drive transmission [15]. On an annual basis, temperature-related variables predominantly correlated with dengue cases within a 17-week lag. On a trimester basis, minimum temperature, RH, and river level stand out in trimester III (Fig. 4). Precipitation and wind speed stand out in trimester I. Spearman correlations highlighted minimum temperature and RH, but also DTR. Elevated minimum temperature could accelerate larval development and reduce the DENV extrinsic incubation period. Although RH has been shown to correlate positively with transmission [10], within the range of values we observed, it correlated negatively with cases. This is likely due to the relationship between RH and temperature (see Table 3). River level, which is driven by precipitation in the Andes mountains and not in Iquitos, probably serves as a proxy for some other proximate factors influencing local mosquito populations or transmission because it has limited impact in the areas of the city where dengue is most common. When river levels are high, transport times are significantly reduced (AC Morrison, personal communication) and Ae. aegypti abundances on boats are highest in October (Guagliardo et al. in review). Similarly, although wind speed might affect mosquito behavior, it seems more probably that wind proxies for other environmental conditions.
The range of temperatures experienced each day (DTR) may modify Ae. aegypti life history traits and Ae. aegypti-DENV interactions [44], [52]–[54]. In Thailand, large daily fluctuations corresponded with less transmission. On an annual basis, we found DTR to be weakly correlated with dengue cases. DTR did load heavily on PC2, which was more strongly correlated with transmission. This latter relationship indicates that high DTR over a period 3–6 months earlier correlated with high current case counts (Fig. 4). In trimester III, however, DTR was one of the strongest correlates over short lags in Spearman tests (Figs. S11, S28). While this result was not apparent in Pearson tests, it suggests that DTR may be epidemiologically important for DENV transmission in Iquitos, as suspected for parts of Thailand.
Overall, it appears that climatic conditions in Iquitos always hover near to a critical threshold for transmission. For instance, a small difference in temperatures could allow female mosquitoes to become infectious after only 2 gonotrophic cycles, as opposed to 3 or more, which would be expected to markedly increase vectorial capacity [55]. Clearly, though, other undefined factors are playing important roles in determining the temporal patterns of DENV transmission in Iquitos [17].
Although mosquito abundances must be important, we do not think that dengue seasonality (especially the increase in transmission) is uniquely driven by fluctuations in Ae. aegypti populations. Aedes aegypti is found in Iquitos year-round and, although population size fluctuates, is relatively abundant when dengue transmission is low [Reiner et al. unpublished 32]. This may contrast with other contexts where climatic variables, especially precipitation, vary more than in Iquitos [15]. On the other hand, our results indicate that vector control efforts targeting adult mosquitoes in large portions of the city were effective, accelerating virus fade-out when the intervention was applied early in the dengue season. In addition to truncating lifespan and killing infected and incubating mosquitoes, these control efforts may transiently reduce the vector population below a threshold density necessary for sustaining epidemic transmission.
Health authorities in Iquitos have responded to a number of outbreaks since 2000 with the intent to kill infected and/or infectious adult Ae. aegypti and reduce mosquito abundance (Table 1). Their interventions usually involved three cycles of non-residual, intra-domicile ULV space spraying with an adulticide (deltamethrin, cypermethrin, or alpha-cypermethrin). Spraying was organized by spatial units defined by the ministry of health and was typically guided by epidemiological information in order to prioritize areas with the largest number of cases. A large number of domiciles were usually treated over a period of several weeks to months. Our analyses indicate that these responses were effective at reducing transmission, which is most easily detected when cases peaked in trimester III and an intervention was conducted in this same period, i.e., early in the transmission season (Fig. 5). Later, in trimester I, it was more difficult to detect a reduction in the number of cases caused by fumigation efforts, presumably because transmission intensity was fading for reasons other than vector control. We assume conditions become less suitable for transmission, but cannot say whether this is due to the effects of temperature on virus replication, a natural reduction in the vector population (although vector abundances remain high in trimester I; [Reiner, et al. Unpublished]), increasing herd immunity or some combination of these and other factors.
We deliberately focused on characterizing the temporal patterns of dengue case reporting in Iquitos in relation to commonly studied covariates, namely climate variables. One of our major goals was to inform the development of mechanistic models. In doing so, we made two methodological observations. First, CCMs are a useful tool for describing the nature of a linear correlation between two covariates. In our case we used them to find the ‘best’ periods of correlation, but found also that the maps were often very “flat.” This simply indicates that the correlation was similar across a range of lags and periods. In other instances, there were clearly multiple possible solutions; i.e., there were several different ‘best’ lags. Second, when plotting the scatter plot of cases against covariates at the best lag and period, we found many non-linear patterns. PC3, which had no linear correlation with dengue case counts at any lag on an annual basis, exhibited a distinct humped relationship. Together, these observations bring into question interpretation and use of classical, linear modeling methods for fitting case data without first doing careful exploratory data analysis. Mixed modeling approaches incorporating appropriate lags and confounds might then prove appropriate tools for modeling and predicting transmission [15], [56]. Nevertheless, where the shape of relationships is uncertain, a priori, non-parametric methods such as general additive models would be more useful. New tools are needed for exploratory analysis, however, to search across lags in order to identify the periods when covariates are most strongly associated with the variable of interest, which will guide model development. Regardless, it is critical that we develop an improved understanding of the relationship between virus transmission dynamics [e.g. 17], [57], [58], per se, and disease.
Although the surveillance program that generated our data was largely uniform across the years of study, changes in personnel, protocol modifications, and variation in transmission intensity likely affected the number of cases captured on a daily basis by the system. Moreover, our surveillance only covered approximately 40% of the Iquitos population, participation rates were far short of 100%, and participation was only sought during the day. We specifically addressed one period of a large increase in surveillance effort during a particularly intense dengue outbreak, but otherwise did not attempt to correct for variation in case capture efficiency. We acknowledge this limitation and in our correlation analyses used a non-parametric method (Spearman correlations) that–for the most part–confirmed results from the Pearson correlations. Reporting rates probably varied over the 10 years as a function of disease severity and other factors influencing individual care-seeking behavior. Although each year there was an increase in the number of dengue cases each year, awareness both in the medical community and the general public would be expected to lag actual transmission. We speculate that care-seeking behavior may change during the course of a dengue outbreak. Initially, during the increase in DENV transmission, people may be more likely to report to a clinic or hospital at the first signs of a fever or other symptom. After a period of transmission and the recognition that dengue cannot be cured with a drug, people may self medicate mild disease with an antipyretic and, thus, be less likely to visit a clinic or hospital. We acknowledge that although all of the cases in our data set were laboratory confirmed, factors not associated with transmission per se likely influenced the patterns we observed and so we considered these patterns only indicators of the true transmission dynamic.
In contrast to the seasonal patterns described above (i.e., transmission typically peaking in late December), we note that the DENV-1 outbreak in 1990 peaked in May [45] and the DENV-2 outbreak in 1995 peaked in August [46]. While climatic averages may have changed some since then, the seasonality has not, which begs explanation. In light of the weak, direct relationships between climatic variables and dengue case totals we measured and the observation that conditions in Iquitos may always support some level of transmission [10], we posit that other factors that we did not measure are important for determining the timing of intra-annual fluctuations and seasonal peaks in transmission. Both the 1990 and 1995 outbreaks were associated with novel virus introductions and we found no record of attempts during those times to perform citywide fumigation campaigns such as those begun in 2003. Because Ae. aegypti, is present year-round in Iquitos, herd immunity and the timing of virus introduction emerge as key determinants of when outbreaks occur. Consistent with this idea, DENV-3 transmission remained high in April/May of 2002, later than all other seasons in our analysis (Fig. 2). DENV-4, on the other hand, peaked in October of 2008. After invasion, as herd immunity rises, variation in mosquito abundances and the suitability of environmental conditions for transmission, should play more of a role determining transmission dynamics. We speculate that the timing and intensity of mosquito interventions to control mosquito populations influenced dynamics in subsequent seasons through their effect on herd immunity [59]. Our future work will focus on testing these hypotheses using mechanistic models [e.g.21].
We emphasize that while climate plays a key role in DENV transmission at broad spatial scales [8], [10], there remain significant uncertainties regarding its specific role and importance when weighed against other drivers at local, fine scales. In different geographic contexts, climate could play a greater role in DENV transmission than in Iquitos, highlighting that DENV ecology is complex and context dependent. Nevertheless, the patterns we document here provide valuable material for the development of mechanistic models that can be used to explore alternative hypotheses about transmission drivers in addition to climate. Importantly, our results indicate that vector control efforts, albeit intensive, can reduce transmission if timed and placed properly. This indicates that vector control can be an effective tool for preventing dengue.
|
10.1371/journal.pbio.0050185 | A Dendrite-Autonomous Mechanism for Direction Selectivity in Retinal Starburst Amacrine Cells | Detection of image motion direction begins in the retina, with starburst amacrine cells (SACs) playing a major role. SACs generate larger dendritic Ca2+ signals when motion is from their somata towards their dendritic tips than for motion in the opposite direction. To study the mechanisms underlying the computation of direction selectivity (DS) in SAC dendrites, electrical responses to expanding and contracting circular wave visual stimuli were measured via somatic whole-cell recordings and quantified using Fourier analysis. Fundamental and, especially, harmonic frequency components were larger for expanding stimuli. This DS persists in the presence of GABA and glycine receptor antagonists, suggesting that inhibitory network interactions are not essential. The presence of harmonics indicates nonlinearity, which, as the relationship between harmonic amplitudes and holding potential indicates, is likely due to the activation of voltage-gated channels. [Ca2+] changes in SAC dendrites evoked by voltage steps and monitored by two-photon microscopy suggest that the distal dendrite is tonically depolarized relative to the soma, due in part to resting currents mediated by tonic glutamatergic synaptic input, and that high-voltage–activated Ca2+ channels are active at rest. Supported by compartmental modeling, we conclude that dendritic DS in SACs can be computed by the dendrites themselves, relying on voltage-gated channels and a dendritic voltage gradient, which provides the spatial asymmetry necessary for direction discrimination.
| The visual system dedicates substantial resources to detecting motion and its direction. For more than 40 years, researchers have tried to decipher the underlying computational mechanisms by which retinal neurons compute directed motion. One type of retinal interneuron involved in direction discrimination is the “starburst” amacrine cell. Starburst-cell dendrites are strongly activated by visual motion from their somata towards the dendritic tips, but not by motion in the opposite direction. It has been proposed, for example, that directional selectivity arises from lateral inhibitory interactions in which activated cells inhibit their neighbors. However, despite extensive modeling, the underlying physiological mechanism has remained elusive. Here, by combining whole-cell recordings, two-photon microscopy, and modeling, we show that discrimination of motion direction in starburst-cell dendrites does not require lateral inhibitory interactions in the retina, but can be generated by a “dendrite-autonomous” computation, which relies on intrinsic electrical mechanisms. Blocking inhibitory interactions does not eliminate directional responses, whereas differential activation of voltage-gated membrane conductances and a dendritic voltage gradient can provide the necessary spatial asymmetry to produce directional signals. The computation underlying dendrite-autonomous direction selectivity may represent one of the most intricate examples to date of dendritic information processing.
| The detection of image motion and the computation of its direction and speed is a major function of the visual system. It is needed for the trajectory prediction of moving objects and provides information about motion relative to the environment [1,2]. In addition to being an intensely studied example of retinal signal processing, motion detection represents an important class of computational problems in neuroscience: the detection of spatiotemporal patterns. The retina's ability to detect the direction of image motion was discovered more than 40 y ago when Barlow and his colleagues [3] found direction-selective (DS) ganglion cells (DSGCs). It was later shown that DSGCs receive DS synaptic input [4–6], which suggested that the direction of motion is computed by retinal interneurons. This was confirmed directly by showing that starburst amacrine cells (SACs; Figure 1) [7,8], which provide input to DSGCs [9–11] and had been proposed early on to participate in the DS computation [12–14], generate DS Ca2+ signals in their dendrites [15].
Optical recordings of visual stimulus-evoked Ca2+ signals also showed that dendritic sectors can be activated independently and that the strongest [Ca2+] increases are in the distal portions of the dendrites, near the output synapses [15]. Electrical isolation of sectors from each other, which is necessary for dendritic processing to be local [14,16,17], is provided by the peculiar dendritic morphology of SACs (Figure 1B) [8,18,19] and a low impedance at the soma [20]. In addition, each sector, unlike the whole cell, is functionally polarized in that synaptic inputs from bipolar and amacrine cells are distributed along the entire dendrite, whereas synaptic outputs are restricted to the distal part [21].
At least two fundamentally different mechanisms have been proposed for dendritic DS in SACs: dendrite-intrinsic electrotonics (e.g., [22,23]) and lateral inhibition (e.g., [14,24]). Despite extensive modeling (reviewed by [25,26]), the physiological mechanism underlying intrinsic DS computation in SAC dendrites has remained elusive. In passive models [22,23], the differences that are found between the voltages induced by centrifugal (CF, towards the dendritic tips) and centripetal (CP, towards the soma) stimuli are small and too sensitive to stimulation parameters to explain the robust direction discrimination seen in DSGCs over wide contrast [27] and velocity ranges [28]. Passive models also predict smaller voltage responses at the soma for CF than CP stimuli [23,29], whereas experiments show the opposite [15,30–32]. The inability of passive models to account for these properties of SAC DS suggests that active (voltage-gated) conductances or strong interactions between inhibitory and excitatory synaptic inputs [14,33] are involved.
To investigate this further, we focused on the electrical properties of SACs using whole-cell recording, two-photon imaging, and compartmental modeling. We show that DS in SACs can be generated by a dendrite-autonomous mechanism that relies on active conductances and a somatodendritic voltage gradient.
In axon-bearing neurons, the somatic voltage is always a direct measure of the cell's output signal, i.e., action-potential generation. This is not the case in SACs, in which synaptic output depends on the voltages in the distal dendritic branches, yet the soma is the only place SACs can be recorded electrically. Nonetheless, even weak electrotonic coupling of the dendritic sectors to the soma can provide information about electrical events in one or more dendritic sectors, but only as long as these signals reinforce rather than interfere with each other. In fact, somatic voltage responses are not systematically DS when probed with full-field moving grating or bar stimuli (Figure 1C, upper traces, and [15,30]), probably because signals from different branches arrive at the soma with different, even opposite, phases and thus cancel. This is not surprising since their circular symmetry alone precludes a DS response of SACs to full-field stimuli.
Cancellation can be avoided either by restricting stimulation to part of the dendrite (Figure 1C; middle and lower traces) or by ensuring—with circularly symmetrical stimuli—that all branches respond in the same way. Then signals from different branches are in phase and reinforce at the soma. Here, we used CF and CP circular wave stimuli (drawing in Figure 1D, see Materials and Methods and [15]), which evoke distinctly DS somatic voltage (Figure 1D) and dendritic Ca2+ responses (Figure 1E, see also [15]).
Dendritic Ca2+ responses (Figure 1E) to light stimuli were seen in a minority of cells (nine out of 50) while the patch electrode was attached, whereas electrical responses were present in all cells. In some cells (eight of 16) that remained intact after the patch electrode had been retracted, light-evoked Ca2+ responses developed after about 15 min. When present, dendritic Ca2+ responses (Figure 1E) were similar to those observed in cells loaded via high-resistance microelectrodes [15]. In simultaneous optical and electrical recordings, we found that [Ca2+] and voltage rise sharply and in parallel as the bright phase of the circular wave enters the distal dendritic field (Figure 1E). This suggests a common physiological cause, possibly the opening of voltage-gated Ca2+ channels (VGCCs).
Nonsinusoidal temporal responses (Figure 2A), particularly to CF motion, indicate the presence of a nonlinearity, because a linear system never generates any “new” frequencies and sinusoidal circular-wave stimuli contain no temporal frequencies other than the fundamental. A sensitive way to recognize and quantify nonlinearity in responses evoked by sinusoidal stimuli is Fourier analysis. The response waveforms can often be well approximated by a sum of a direct current (DC) voltage component (V0), a voltage component at the fundamental frequency (V1), and a few low-order (second and third) harmonic voltage components (V2 and V3) (Figure 2A and 2B; see also inset in Figure 2C). Component amplitudes were used to quantify the nonlinearity (Table 1) [34].
Circular wave stimuli generated responses with harmonics that were larger for CF than CP stimuli (Figure 2B); this was most evident for V2. The relative phases (ϕ2 − 2 · ϕ1) cluster near −π/2 for CF motion (Figure 2C, orange), indicating a steep rise (inset in Figure 2C). CP motion typically evoked smaller-amplitude harmonics, i.e., a more sinusoidal waveform; the broad distribution of ϕ2 − 2 · ϕ1 (Figure 2C, blue) indicates a lack of correlation between fundamental and second-harmonic phases for CP motion. (Note that only cells with V2 ≥ 0.3 mV were included in the histogram, since phase measurements become meaningless for amplitudes close to or below the noise level.) There was a small but significant (n = 83, p ≤ 0.01; Figure 2D) difference in V0 between CF (0.7 ± 0.2 mV) and CP (1.1 ± 0.2 mV) stimuli. However, despite a smaller V0, CF motion evoked larger peak depolarizations than CP motion (ΔVPeak [CF − CP] = 1.4 ± 0.2 mV; n = 26). Also note that voltage excursions might well be substantially larger in the dendrite than in the soma (see also Discussion).
To facilitate comparisons, we used directional asymmetry indices (AI1, AI2, and AI3; see Materials and Methods). In almost all cells, CF motion evoked larger fundamental responses (AI1 > 0 in 90% of the cells; Figure 2E) with higher harmonic content (AI2 > 0 in 80%, AI3 > 0 in 73% of cells; Figure 2E). The peak positions of the Gaussian fits to the AI distributions are (n = 83 cells): AI1 = 0.13 ± 0.01, AI2 = 0.25 ± 0.04, and AI3 = 0.22 ± 0.07. In order to check for the involvement of Cs+-sensitive K+ channels, we replaced intracellular K+ by Cs+ and found an increase in resting potential (with K+: n = 58, VRest = −62 ± 1 mV; with Cs+: n = 25, VRest = −48 ± 1 mV) but little effect on DS (with K+: n = 58, AI1 = 0.13 ± 0.01, AI2 = 0.27 ± 0.03, and AI3 = 0.19 ± 0.05; with Cs+: n = 25, AI1 = 0.14 ± 0.02, AI2 = 0.17 ± 0.07, and AI3 = 0.28 ± 0.01).
The response of a nonlinear system can depend on the stimulus amplitude in a much more complicated way than the simple scaling that is characteristic of a linear system. We therefore varied the stimulus contrast and found that V1 grew almost linearly with contrast (Figure 3A and 3B), albeit with different slopes for CP and CF motion. V2 for CP motion was indistinguishable from the (no-motion) background for all but the largest contrast (72%). For CF motion, V2 increased roughly linearly with increasing contrast and ϕ2 − 2 · ϕ1 became more tightly clustered near -π/2 (Figure 3C). V0 increased with contrast but with no consistent DS.
To rule out that the observed response asymmetry is restricted to a small velocity range we measured the dependence of the harmonic amplitudes on stimulus velocity (Figure S1). With a 4-fold change in speed (from 0.5 to 2 mm/s), V0, V1, and V2 varied only little; V3, however, fell by a factor of almost three, presumably due to capacitive filtering at higher frequencies.
SACs receive direct and indirect inhibitory synaptic input [35] from other cells in the circuitry. The direct inputs include inhibitory interactions between SACs [36], which involve GABA acting on GABAA receptors [37], and glycinergic input [38] from other types of amacrine cells. The indirect inhibition results from GABA acting on GABAC receptors at bipolar cell axon terminals [39,40], which provide SACs with glutamatergic excitatory input. Lateral inhibitory inputs are involved in the generation of DS SAC responses when image motion includes the surround [24].
To test whether inhibitory input is required for the generation of DS SAC responses evoked by circular wave stimuli that are spatially confined to the SAC's dendritic arbor, we tested the effect of a mixture of selective antagonists that block GABAA, GABAC, and glycine receptors (i.e., GABAzine [Gbz], TPMPA, and strychnine, respectively). Even with a full cocktail of inhibitors, which eliminates all lateral interactions [24], the electrical response remained strongly DS (Figure 4A) for V2 and peak responses (ΔVPeak); only V1 DS was reduced (Figure 4C; Table 1). The depolarization of the resting potential (VRest) that is also seen is consistent with blocking tonic inhibitory input. Similar results were obtained when applying Gbz and TPMPA together (Figure 4D). No significant reduction in the DS of V1, V2, or ΔVPeak was found with either Gbz (Figure 4B and 4E), which selectively disrupts the SAC-to-SAC GABAergic inhibitory network, or TPMPA (Figure 4F), which selectively abolishes GABA-mediated presynaptic inhibition. AI2 increased, but not significantly, under both Gbz and TPMPA.
We also applied the widely used but nonspecific GABA-receptor antagonist picrotoxin (PTX) at a concentration (≈300 μM) reported to abolish GABA-induced currents in SACs (see Figure 4D and 4E in [38]). At this concentration, PTX also affects glycine receptors [41] and therefore should have an effect similar to a mixture of Gbz, TPMPA, and strychnine. PTX, like the Gbz+TPMPA+strychnine mixture, had no effect on V2 DS, but unlike the mixture, decreased ΔVPeak (Figure 4G). PTX, but not the mixture, also caused an increase in V0 for both CF and CP stimuli—the only blocker to have a consistent effect on V0. The reason for this difference in the actions of these antagonists is not clear. That the effects of PTX, a nonselective antagonist, are different than the effects of a combination of selective blockers, suggests that PTX's actions are not restricted to GABAA, GABAC, and glycine receptors.
In the presence of blocker mixtures (Gbz+TPMPA or Gbz+TPMPA+strychnine), dendritic [Ca2+] is very unstable, with frequent large and spontaneous transients (X. Castell, unpublished data). This makes it impossible to reliably monitor the DS Ca2+ signaling when using our circular wave stimuli that have a physiologically realistic contrast and intensity.
In summary, the results with inhibitory blockers clearly demonstrate that a moving (circular wave) stimulus that is confined to the spatial extent of the cell can produce a DS electrical response that is independent of lateral inhibitory interactions. This shows that under these conditions, DS signaling in SACs is generated by a dendrite-intrinsic mechanism, which is investigated further in the following experiments.
Voltage-gated channels (VGCs) have a limited and channel type–specific activation region. Thus, varying the somatic voltage (VCOM) can help to test whether VGCs are involved in DS and/or response nonlinearity, and may help in the identification of the ion channel(s) involved. We measured light-induced current responses, which in shape and asymmetry resemble the voltage responses measured in current-clamp (Figure 5A and 5B), while stepping the somatic potential from −75 mV to voltages between −105 mV (below which cells became leaky) and −35 mV (Figures 5 and S2). Both fundamental (I1) and harmonics (I2 and I3) were substantially voltage dependent (Figure 5C). I0 is dominated by currents due to the voltage-step protocol and was not analyzed for DS. The direction sensitivities of I1 and I2 behave very differently as a function of step voltage (Figure 5C and 5D). At voltages negative to the resting potential (see Materials and Methods), the DS of I1 decreases, while the DS of I2 increases. The third harmonic (I3) is close to background (no-motion) levels around the resting potential but becomes substantially DS for strong hyperpolarization (Figure 5C).
Of all voltage dependencies analyzed, only that of I1 for CP motion (blue traces in Figures 5C and S2) resembles what would be expected from a passive dendrite, i.e., a linear I–V curve reversing at approximately 0 mV. Most other results (for CF motion I1, I2, and I3; and for CP motion I2) indicate the presence of a nonlinearity that depends on holding voltage, in a manner consistent with the presence of VGCs.
The strong correspondence between motion-induced [Ca2+] transients and nonlinearity in the voltage responses (Figure 1E) suggests a role for VGCCs, which are present in SACs [42]. Furthermore, various Ca2+ channel blockers abolish DS in DSGCs while leaving light responsiveness largely intact [43]. We confirmed this effect of Cd2+ with recordings from DSGCs (unpublished data), albeit at concentrations of 20 μM rather than the 60–110 μM used by Jensen [43].
Because the loss of DS in ganglion cells could result either from the suppression of synaptic inputs from SACs or from the block of DS generation in SAC dendrites, we measured the effect of Cd2+ on SAC responses. Both AI1 and AI2 were reduced by 10 μM Cd2+ (Figure 6; Table 1). At a higher concentration (90 μM), Cd2+ eliminated SAC light responses altogether (six cells, unpublished data), presumably by blocking synaptic pathways. The effects of Cd2+ were variable (Figure 6B–6D), possibly because the concentration range that leaves the synaptic input from bipolar cells intact is narrow. In experiments with other VGCC blockers, such as ω-conotoxin M7C (reviewed in [44]), the inputs to SACs were either strongly reduced or the DS was unaffected (unpublished data).
To test whether distal dendritic VGCCs are active near the resting potential, we measured changes in dendritic [Ca2+] during somatic voltage steps (Figure 7). The Ca2+ signals reflect local VGCC activity, whereas when measuring whole-cell currents (as in [42]), the locations of the activated channels are unclear, and currents due to channels electrotonically distant from the recording can be severely filtered. Substantial voltage-evoked Ca2+ responses (Figure 7A) were observed in a majority of cells (20 out of 27). Some of these cells (seven out of 20) displayed large, spiky [Ca2+] transients riding on top of smooth responses (Figure 7B). Light-induced Ca2+ responses (for which 12 of the 20 cells were tested) were found more often (three of five) in SACs with spiky transients than in cells without them (one of seven). We constructed an activation curve (Figure 7C) using only SACs without spiky transients (compare Figure 7A and 7B) and fitted it with a channel activation curve (Equation 5, Materials and Methods), finding a half-activation voltage V50 = −49 ± 2 mV and a slope voltage VSlope = 8.7 ± 2.0 mV (n = 6 cells; other fit parameters: ACa,0 = −0.17 ± 0.08; g = −0.016 ± 0.002). V50 is much lower and VSlope is much larger than the nominal values for high-voltage–activated (HVA) VGCCs found in SACs [42] (e.g., for P/Q-type: V50 ≈ −11 mV, VSlope ≈ 4.56 mV, from data in [45]). This could, for example, be due to different, i.e., low-voltage–activated (LVA), types of VGCC in the distal dendrites. That T-type or LVA L-type channels [46] contribute significantly is unlikely because SAC Ca2+ currents are not sensitive to Ni2+ [42], which blocks T-type channels more effectively than other Ca2+ channels [47], or nifedipine, which inhibits L-type channels [42]. Another possibility is a dendritic voltage gradient, which is, in fact, plausible since SAC dendritic branches are rather thin [21]. The requisite steady radial current, outward at or near the soma and inward near the tips (Figure 9A), could flow out through K+ channels that are open at the resting potential (and are not completely blocked by intracellular Cs+ [48]) and in through channels with a more positive reversal potential, such as glutamate-gated channels, which are tonically activated on SACs [35]. Since glutamate receptors are preferentially located on branch points and varicosities [49], glutamatergic input should, in fact, be stronger distally, where it is also more effective because the somatic current-sink [20] shunts proximal inputs.
An antagonist of AMPA-type glutamate receptors, CNQX, shifted the [Ca2+] versus V curve towards depolarized potentials by 15 mV (Figure 7D; n = 3 cells; fit parameters: V50 = −34 ± 3 mV; ACa/0 = −0.09 ± 0.06; g = −0.019 ± 0.003; VSlope = 10.2 ± 2.2 mV, see Equation 5) and reduced the holding current (from −96 ± 2 pA to −76 ± 9 pA at VCOM = −75 mV, n = 5; and from −186 ± 7 pA to −141 ± 5 pA at VCOM = −83 mV; n = 3).
This reduction, which confirms the tonic activation of the AMPA-type receptors, may be considerably smaller than the actual tonic current through AMPA-type channels, which is partially compensated due to reduced GABA-receptor–mediated current into SACs because CNQX blocks excitatory input onto inhibitory interneurons (other amacrine cells). The current carried by NMDA-type glutamate receptors in SACs [38] is likely too small to play a major role.
Finally, we examined where along the dendritic branches the moving stimulus has to be presented in order to elicit DS nonlinearities. To do this, different regions of circular wave stimuli were masked by concentric gray rings (Figure 8). Reducing the stimulus diameter (Figure 8B, 8C, and 8F) led to a reduction in DS, mostly due to an increase in the amplitude for CP motion. Responses to CP and CF stimuli became more similar (Figure 8E and 8F) for both fundamental (V1) and harmonic responses (V2 and V3). When the masked region in the center was enlarged (Figure 8D), DS increased for V1, V2, and V3. The DC component (V0) dropped to nearly zero when most of the center region was masked (Figure 8D and 8E), with no clear DS in any except the “standard” configuration (Figure 8A, 8E, and 8F).
Taken together, this suggests that the motion-response asymmetry is computed in the distal section of the dendrite, which is consistent with the radial location of the intensity gradient at the time when voltage and [Ca2+] rise steeply (Figure 1E). The decrease in V1 DS for smaller stimulus diameters is expected as a result of a reduction of distal input (see model), but a reduction in lateral inhibition may be a contributing factor. The finding that V2 DS, which does not depend on inhibition, also decreases with stimulus diameter highlights the importance of the distal region for dendritic DS computation.
Although a number of studies (reviewed in [26,50]) have shown that SACs are necessary for the computation of motion direction in the retina and that SAC dendrites generate direction-dependent signals [15,24], it is still not well understood how the DS signal is actually computed. In this study, we have focused on dendritic electrical mechanisms and found experimental and modeling evidence that they are, by themselves, able to generate DS signals, in the absence of lateral inhibitory interactions, and are thus almost certain to play an important role in generating DS synaptic output from SACs.
Based mainly on two observations, it has been proposed [14] that the computation of SAC DS requires lateral synaptic inhibition. First, blocking GABAergic inhibition abolishes DS responses (but not general responsiveness to light) in DSGCs [51–53], which may, however, simply result from blocking DS input into DSGCs. Second, anatomical ([21,54] but see [10]) and physiological [24,36] evidence indicate GABAergic inhibition between SACs. Recent modeling even suggests that reciprocal inhibitory interactions in the SAC network could be sufficient to create SAC DS [55].
Surround stimulation inhibits SACs [30,35] and depresses dendritic [Ca2+] levels [15,24]. Thus, it was surprising that neither surround inhibition nor DS of dendritic Ca2+ signals were abolished by blocking GABAA receptor–mediated transmission [15]. To completely block surround effects on the dendritic Ca2+ signals, simultaneously blocking GABAA, GABAC, and glycine receptors is required [24]. This indicates that GABA- and glycine-mediated lateral inhibition can contribute to dendritic DS in SACs.
Our pharmacological data show, however, that for stimuli with movement constrained to the SAC dendritic field, DS persists almost unchanged when the direct, GABAA receptor–mediated [24] connections between SACs are blocked. Only with all lateral inhibitory interactions blocked, which also leads to substantial tonic depolarization, is DS affected (Figure 4E and 4F). Consistent with Lee and Zhou [24] the response to CP motion (V1,CP) increases (Figure 4E–4G), leading to a reduced V1 DS. It is, furthermore, inconsistent with a lateral-inhibition–based mechanism that V0,CP (the DC response for CP motion) is larger than V0,CF (Figure 2D), because stronger inhibition for CP motion should to lead to a smaller V0,CP. This and our pharmacological results suggest that an inhibition-independent mechanism is used for DS detection of movement inside the dendritic field (reminiscent of the CF facilitation observed by Lee and Zhou [24]). There may be additional (inhibition-based) mechanisms used for stimuli that extend into the surround [24]. That V1 DS for dendritic field–restricted stimuli is reduced by blocking GABAA, GABAC, and glycine receptors may be caused by an indirect effect of the strong (as much as 8 mV) shift in resting potential, rather than the interruption of lateral inhibition. An absolute requirement of lateral inhibitory network interactions for DS generation is ruled out by the undiminished presence of V2 DS and ΔVPeak DS under these conditions.
That the somatically measured fundamental-frequency voltage and current components (V1 and I1) are larger for CF than for CP stimuli (Figure 1D and 1E, and [15]) contradicts model predictions [23,29]. One possible explanation is that SACs receive DS synaptic input. If excitatory inputs were DS, the difference in the amplitude of the fundamental response for CP and CF stimuli should increase with hyperpolarization, i.e., increased driving force (light-evoked synaptic currents are nonrectifying in SACs [35]). Instead, I1 DS becomes smaller at more-negative potentials (Figures 5C and 5D). For inhibitory inputs, the situation is more complicated. Although the voltage dependence of I1 DS is consistent with that of direct inhibitory synaptic inputs, such as those resulting from SAC–SAC interactions [36], V1 DS persists in the presence of GABAA blockers. Indirect inhibition via presynaptic GABAC receptors has the voltage dependence of glutamatergic currents. The reduction of V1 DS by blocking GABAA, GABAC, and glycine receptors might alternatively be due to changes in the center-surround structure of bipolar cells [56] or due to changes in the active electrical mechanism (discussed next) as a result of the depolarization caused by the elimination of tonic inhibition.
An alternative explanation for the DS at the fundamental frequency is a dendrite-intrinsic direction-dependent amplification of synaptic input currents by activation of VGCs. One way to picture such signal amplification is that the negative slope conductance [57] that occurs in a certain part of the VGCs' activation range compensates some of the native (including synaptic) conductance and thus increases the input resistance. This in turn leads to a larger (amplified) voltage change for a given input current. For significant amplification, the negative slope conductance needs to be comparable to the passive conductance. SACs possess N-, P/Q-, and R-type Ca2+ channels [42], with N- and P/Q-type channels contributing approximately 70% of the Ca2+ current. In addition, SACs may express TTX-resistant voltage-gated Na+ channels (B. J. O'Brien, personal communication). We used P/Q-type channel data from the literature [45,58] and estimated (see Materials and Methods) the size of the slope conductance provided by these VGCCs. For approximately 930 channels, which are needed to account for I2 (see below), the negative slope conductance amounts to approximately 1/(−130 MΩ). This value is comparable to the light-induced synaptic conductance (≈3.7 nS = 1/(270 MΩ) deduced from Figure 4B in [35]). The slope conductance depends strongly on the voltage (location on the activation curve; Figure 7E) and becomes negligible below the activation range. This means that no VGC-based amplification should occur at very hyperpolarized potentials, which is consistent with the observed dependence of fundamental frequency DS on the holding voltage (Figure 5C and 5D).
The presence of a second harmonic component (V2) even for lower-contrast CF motion (Figure 3) indicates that an electrical nonlinearity is involved in SAC DS. The clustering of the relative phases (ϕ2 − 2 · ϕ1) between fundamental and second harmonic at negative values around −π/2 (Figure 2C) shows that, even when it is not obvious from the response trace, nonlinearity occurs mainly during the rising phase of the fundamental. This and the steep rise in the voltage and [Ca2+] for CF motion (Figure 1E) suggest that the distal-dendritic resting potential might hover near the threshold of regenerative events (e.g., Ca2+ spikes). This might also be the reason that action potentials have been seen in some [59], but not all ([15,30,35] and present study), SAC electrical recordings. Presynaptic pathways do not appear to contribute substantially to the response nonlinearity, since only at the highest contrasts do the values of V2 for CP motion rise substantially above background (Figure 3B).
As mentioned above, HVA VGCCs are likely responsible for the nonlinearity. Therefore, it was somewhat surprising that the measured I2 shows only a shallow maximum (Figure 5C) as a function of the somatic holding potential, since we initially expected that hyperpolarization beyond the activation range of these channels would suppress the nonlinear current component completely. That this was not possible, within the range of holding potentials that the cells would tolerate, indicates that there is voltage attenuation between soma and dendrite (see next section). We can, however, estimate whether the number of Ca2+ channels needed to generate the second harmonic signals we see is consistent with Ca2+ currents reported by Cohen [42] by assuming moderate voltage modulation amplitudes and expanding the I–V curve into a Taylor series (see Protocol S1). If we neglect inactivation and kinetic aspects, and assume a modulation amplitude (VA) of 5 mV (Figure 1D), we find that approximately 930 channels are needed to get I2 = 8 pA (Figure 5C). To generate the Ca2+ conductance seen in SACs (≈4 nS, [42]), we need approximately 300 channels. (This takes into account that Cohen used 110 mM Ba2+, which increases the single-channel conductance by a factor of approximately three, see [58].) The discrepancy in channel-number estimates might simply reflect the fact that they are based on measurements from different preparations and are affected differently by dendritic voltage attenuation and current shunting. If, for example, the voltage modulation in the dendrite (VA) is twice that measured in the soma and even if the current from the dendrite (I2) is attenuated by the same factor, which is greater shunting than is likely, the number of channels (Protocol S1, Equation S6) would be half as large (465) and thus in closer agreement with the number estimated using Cohen's data [42]. That voltage excursions in the distal dendrite (where, owing to the branching pattern, one finds most synaptic inputs) are larger than in the soma is likely, because graded potentials decrement as they spread electrotonically towards the soma.
One problem when investigating dendritic processing is that it is not only impossible to record electrically from most of the dendrite, but it is also difficult to control the voltage in parts of the dendrite that are remote from the electrode location (the “space clamp” problem). This is particularly acute in SACs, in which the dendritic connections between the soma and distal branches are very thin [8], and synapses are tonically activated (Figure 7D and [35]). We therefore need to consider explicitly the voltage attenuation and shift due to the dendritic voltage divider (Figure 9A).
Earlier Ca2+ measurements from SAC dendrites [15] showed that at rest (i.e., with constant, uniform illumination), there is a steady-state Ca2+ influx that can be transiently reduced by suitable stimuli. The [Ca2+] versus voltage-step data (Figure 7) confirm this since they show that the dendritic voltage is still within the activation range of VGCCs when clamping the soma at the zero-current potential (VRest). Because LVA Ca2+ channels have not been found in SACs [42], it is conceivable that the Ca2+ influx is through HVA channels and the measured Ca2+ activation curve (Figures 7E) has been shifted by attenuation of the somatic control voltage via a longitudinal dendritic resistance (RPD) [16] in combination with a distal leak conductance (1/Rleak) that has a positive reversal voltage (Figure 5E). Note that the shape of the activation curve for SAC–SAC synaptic currents (Figure 1E in [24]) is consistent with the presence of such a dendritic voltage divider.
We can roughly estimate the properties of the dendritic voltage divider (Figure 9A) by comparing the activation parameters (V50 = −48.5 mV, VSlope = 8.7 mV) obtained from the [Ca2+] versus V data (Figure 7) to those measured under proper voltage clamp. For P/Q-type channels [45], we estimate V50 = −10.9 mV and VSlope = 4.56 mV, and for R-type channels [60], V50 = −27.5 mV and VSlope = 8.57 mV. (Because their activation parameters are sufficiently similar, we lumped N- and P/Q-type channels together.) A combination of 70% P/Q-type and 30% R-type channels (see above) is fit approximately using V50 = −10.4 mV and VSlope = 8.3 mV. The relation between distal voltage (VD) and the somatic voltage (VS) is
where Erev,leak is the reversal potential of the distal leak. Changes in the somatic voltage reach the dendritic tips attenuated by the factor γ = Rleak/(Rleak + RSP); it can be shown that, therefore, the apparent VSlope is increased by 1/γ. For the P/Q-type channel, γ = 4.56/8.7 = 0.524, while for the R-type channel, γ would need to be 0.985, and for the combined current, γ = 8.3/8.7 = 0.95. To get Erev,leak, we can set VS to the apparent half-activation voltage (V50S) and VD to the actual half-activation voltages (V50VC) and use Equation 1 to find Erev,leak = (V50VC − γV50S)/(1 − γ), which, obviously, becomes very error prone as γ approaches one. For the P/Q-type and the combined currents, we find Erev,leak = 30.5 mV and Erev,leak = 780 mV, respectively.
The combined currents lead to rather unlikely values for both γ and Erev,leak, which may mean that R-type channels are not involved in distal Ca2+ influx. For P/Q-type channels, the γ value is reasonable, but the leak reversal potential seems too high. We have not, however, taken into account: (1) persistent currents [61] through VGCCs or voltage-gated Na+ channels (B. J. O'Brien, personal communication), which steepen the activation curve and shift Erev,leak to more positive potentials, because as these VGCs get activated, the ensuing current depolarizes the dendrite, leading in turn to an accelerated activation of VGCs. This could account for the otherwise implausible value of γ ≈ 1 for the R-type and the P/Q-R mixture; (2) The channel activation curves used for comparison (Figure 7E) are measured using Ba2+, which not only shifts the activation range of VGCCs to more positive values (e.g., [60]), but also increases the open-channel conductance (reviewed in [62]), potentially increasing the slope of the activation curve under conditions of imperfect voltage clamp. Please note that we have here lumped together the two dendritic compartments shown in Figure 9A, because, due to the larger number of distal branches, RSP ≫ RPD.
In conclusion, the voltage-divider (voltage gradient) hypothesis reconciles the voltage-step–evoked Ca2+ signals with the pharmacological evidence [42] for the predominance of HVA Ca2+ channels in SACs.
Published models of dendritic DS in SACs can be roughly divided into two classes. One requires lateral inhibition (e.g., [14,31,55]), which, through the interaction with excitatory input, can generate nonlinearity and substantial DS [14,33]. Although lateral inhibition can contribute to dendritic DS [24], even blocking GABAA, GABAC, and glycine receptor-mediated inhibition leaves signaling in SACs clearly DS (Figure 4F). Models in the other class rely on the passive electrical properties of the dendrite and show that moderate DS can be generated by the SAC's morphology alone (e.g., [23]), but predict somatic voltage excursions that are larger for CP than for CF motion, which does not fit the experimental data (Figure 1C and 1D, and [15,30,31]). Passive models (e.g., [23]) also do not generate enough discrimination to explain how the DS system as a whole [27] is able to prefer a substantially lower contrast, preferred-direction stimuli over higher-contrast, null-direction stimuli.
Our data suggest that active (voltage-gated) currents are involved in DS computation, but the question remains as to how. In order to explore this issue, we constructed a two-compartment model (representing, e.g., distal and proximal dendrites; Figure 9B) in which each compartment contains, in addition to a linear leak conductance and a capacitance, a single type of VGC, to which we gave the properties of HVA Ca2+ channels. This we did because HVA Ca2+ channels are present in SACs [42]; other VGC types should work as well, but we have not explored this here. The compartments are connected by a resistor, representing the longitudinal intradendritic current pathway. The moving stimulus is represented by the injection of phase-shifted sinusoidal currents (Figure 9C).
We found that for a certain set of parameters, even if both compartments are otherwise identical, a small difference in the resting potentials (e.g., −19 mV and −16.5 mV) between the compartments allows a strongly DS response, such that in the more depolarized (distal) compartment, the “CF” stimulus causes an approximately 8.5-fold larger (fundamental-frequency) voltage amplitude than the “CP” stimulus (Figure 9D). Consistent with the somatic voltage recordings (e.g., Figure 1C–1E), the more negative (proximal) compartment also prefers the CF stimulus but only moderately so (by factors of ≈1.3 for V1 and ≈1.9 for V2). Thus, the model explains why the somatic voltage and the dendritic [Ca2+] both prefer CF motion (Figure 1E). The model is also consistent with our experimental data in that (1) DS is stronger (Figure 2E) for the second harmonic, (2) DS is voltage dependent (Figure 5C and 5D), (3) DS is largely independent of stimulus contrast (Figure 3D), and (4) in the absence of distal stimulation, V1 for “CP” increases (Figure 8C; Table 2). In the distal model compartment, the responses to preferred-direction stimuli with weak contrasts (corresponding to small Iosc) are larger than responses to null-direction stimuli with as much as 4-fold larger contrasts (Figure 9D; Table 2). These are important features that cannot be reproduced by passive models, and strengthen the case for a central role of active conductances in dendrite-autonomous DS. In this regard, the SAC DS mechanism resembles frequency selection in the rod-photoreceptor network [63] and in auditory hair cells [64]. We would like to emphasize that ours is a proof-of-concept model. Using the current parameters, significant DS for both V1 and V2 is found over a frequency (velocity) range spanning only about a factor of three (unpublished data). More detailed versions of this model (with more radial compartments and multiple types of channels) will be needed to explain the full range of experimental observations.
A central property of our model is that a strong directional preference is imposed on an otherwise symmetrical arrangement by a mild voltage gradient (compare Figure 9D and 9E), as it appears to be present in SAC dendrites (Figure 7, see also Figure 1E in [24]). Only a small fraction of the total voltage drop between dendrite and soma (most of which may occur in the proximal dendrite [16]) would be needed. This is consistent with our experimental finding (Figure 8D) that strong DS results when presenting motion to only the outer part of the dendritic arbor.
The dendritic voltage gradient thus generates the necessary spatial asymmetry for computation of DS, by causing the VGCs in the two compartments to operate in slightly different regimes. Although in the model the proximal compartment is slightly DS, strong DS is limited to the distal compartment, which is where it is needed to drive DS Ca2+ signals [15] and, in turn, DS synaptic release.
Amplification of Ca2+ influx by calcium-induced calcium release (CICR) has been proposed as a mechanism potentially well matched to the timing requirements for the detection of slowly moving stimuli [65]. Although the biochemical machinery necessary for CICR has been found in cultured GABAergic amacrine cells [66,67], these cells were not classified further, and it is currently unknown whether CICR is present in SACs. The abolition of light stimulus–induced [Ca2+] transients during most whole-cell recordings is not due to a loss of VGCC function, because depolarization can still evoke an increase in [Ca2+] (Figure 7A). Also, in retinal ganglion cells recorded under similar conditions, dendritic Ca2+ responses persist [68], suggesting the loss of light-induced Ca2+ responses to be SAC specific. CICR is potentially sensitive to washout of cellular components during whole-cell recording and might be required in SACs to generate detectable Ca2+ responses to light stimuli. The latter is supported by our observation that cells exhibiting spontaneous spiky [Ca2+] transients during depolarization (Figure 7B), which could reflect CICR activity, are also more likely to show light-induced Ca2+ responses.
The persistence of DS electrical responses (linear and nonlinear) in the absence of any light stimulus–induced [Ca2+] transients suggests, however, that DS computation as such is independent of CICR. It remains to be explained why, if Ca2+ fluxes underlie the nonlinear currents (Figure 5), they cannot be detected as changes in Ca2+-dependent fluorescence. During a 100-ms period (the rising phase of the stimulus), about 32 amol (atto-mol = 10−18 mol) of Ca2+ enters the cell (estimated α = 7.4 pS [slope conductance, see Protocol S1, Equation S3], 500 VGCCs, and ΔV = 10 mV). This is small compared to the total amount of Ca2+ indicator; approximately 320 amol, estimated using a volume of the SAC's dendrite of 3.2 pL (based on [16,21]) and [OGB-1] ≈ 100 μM; the change in fluorescence intensity with this amount of indicator would be much less than the light-induced changes actually seen [15].
Since DS Ca2+-current densities are thus expected to be rather small in absolute terms (even though the inward/outward ratios might still be quite large), it is possible that CICR is needed to amplify [Ca2+] changes to levels that are sufficient for driving synaptic output from SACs, which is Ca2+ dependent [36,69]. It may, in fact, be impossible to allow entry of sufficient Ca2+ through channels to drive release without increasing Ca2+ channel density into a regime in which the dendrite becomes electrically unstable. CICR, due to its lack of electrogenicity, would circumvent this problem. Finally, CICR is expected to be highly nonlinear and thus could contribute to the rectification of the DS signal.
Information about image motion is computed in SAC dendrites. The persistence of DS signals in SACs in the presence of GABA and glycine receptor antagonists shows that SAC dendrites by themselves generate DS signals, in the absence of lateral inhibitory interactions. There is strong evidence that VGCs, probably Ca2+ channels, and a gradient in membrane potential between dendritic tips and soma are involved. As we show for a two-compartment model, VGCs, when combined with the voltage gradient, can provide electrogenic timing-, frequency-, and DS amplification of synaptic input currents. VGCCs could, in addition, produce either directly or via CICR the increase in local [Ca2+] necessary for synaptic release and thus for the transmission of DS signals to DSGCs. Dendrite-autonomous direction detection may thus be one of the most convincing examples yet for the power of dendritic computation (reviewed in [70]).
Experiments were done on adult New Zealand White rabbits or pigmented rabbits (Charles River Laboratories, http://www.criver.com, or Harlan Winkelmann, http://www.harlan.com); there were no systematic differences between strains, and the data were pooled. All procedures were approved by the animal care committee. After dark adaptation for two or more hours, rabbits were deeply anesthetized by intramuscular injection of ketamine (50 mg/kg body weight; Curamed Pharma/Pharmaselect, http://www.plasmaselect.com) and xylazine (Rompun, 10 mg/kg body weight; Bayer, http://www.bayer.com) and then sacrificed with intravenous pentobarbital (Narcoren, 160 mg/kg body weight; Merial, http://www.merial.com). The eyes were quickly enucleated, and the retinae were dissected in Ames medium under dim red illumination.
As described previously [35], a piece of mid-peripheral retina was flat-mounted in a the recording chamber [15,71] and superfused (at ≈2–4 ml/min) with warm (≈36 °C), oxygenated (95% O2, 5% CO2) Ames medium. A fluorescent dye (3 to 7 μM Sulforhodamine 101) was added to the bath, which allowed the location, size, and shape of retinal neurons to be visualized using two-photon fluorescence microscopy (2PM, see below and [72]). ON SACs were identified by their characteristic round 10 μm–diameter somata (Figure 1A). The inner limiting membrane (ILM) next to a targeted SAC was opened by scratching a hole in it with a patch pipette and expanding it by “blowing” Ames into the space under the ILM while viewing the retina by infrared video imaging with LED illumination. Cellular debris was cleared away using a suction pipette. All subsequent electrical and optical recordings were conducted in sulforhodamine-free medium.
Antagonists were added to Ames solution without ionic substitution at the following concentrations: 25–50 μM Gbz (GABAzine or SR-95531), 300 μM PTX (picrotoxin), 50–75 μM TPMPA (1,2,5,6-tetrahydro-pyridine-4-yl-methylphosphinic acid), 1 μM strychnine, 5–90 μM CdCl2, 70–100 nM ω-Conotoxin M7C. All chemicals were purchased from Sigma-Aldrich (http://www.sigmaaldrich.com) unless noted otherwise.
SACs were recorded using the whole-cell tight-seal (patch-clamp) technique [73] with 3–12 MΩ patch pipettes pulled from borosilicate glass (O.D.: 1.0 mm, I.D.: 0.58 mm, with filament; Hilgenberg, http://www.hilgenberg-gmbh.de) using a multi-clamp amplifier (Axon Instruments, http://www.axon.com). Data were acquired with a digitizer interface (Digidata 1322A) using pClamp 8 software (both, Axon Instruments) and analyzed off-line with IgorPro (Wavemetrics, http://www.wavemetrics.com). For current-clamp recordings, patch pipettes were filled with solution that contained (in mM): K-aspartate (or K-gluconate) 110–120, KCl, 5–10, NaCl 5, HEPES 5–10, MgCl2 0–1, Mg-ATP 1, Na2-GTP 0.1–1. For voltage-clamp experiments, K-aspartate (or K-gluconate) and KCl in the filling solution were replaced by equal amounts of Cs-methanesulfonate and CsCl, respectively. In some experiments, 0.2–0.5 mM CaCl2, 2–5 mM HEDTA (N-(2-hydroxyethyl)ethylenediamine-N,N′,N′-triacetic acid) and NMG (4-N-methyl-D-glucamine) were added to the intracellular solution without discernable effect on the results. All filling solutions contained the fluorescent Ca2+ indicator (100–200 μM) Oregon Green 488 BAPTA-1 (OGB-1; Invitrogen, http://www.invitrogen.com). For current-clamp recording, the membrane voltage (Vm) was low-pass filtered at 2 kHz and digitized at 5–10 kHz. In voltage-clamp mode, the current (Im) was digitized at 2 to 25 kHz and low-pass filtered at 1 kHz. Voltages (Vm and VCOM) were corrected off-line by subtracting 15 mV for the liquid junction potential (calculated: 14 to 16 mV, measured: 17 to 18 mV; see [74]. Voltage-clamp recordings were, in addition, compensated off-line for series resistance (23 to 75 MΩ) effects. The average resting potentials were −62 ± 1 mV (n = 58; range −26 to −82 mV) and −48 ± 1 mV (n = 25; range −35 to −58 mV) with the K+- and Cs+- based filling solutions, respectively. The average input resistance (only determined for Cs+) was 164 ± 21 MΩ (n = 17; range 83 to 478 MΩ).
Two different custom-built upright 2PMs (both in-house designs), each with two detector channels, were used to image sulforhodamine fluorescence as well as to record Ca2+ indicator signals (Figure 1, and [15,75]). One microscope was equipped with a 20× water immersion lens (0.95 W, XLUMPlanFI, Olympus, http://www.olympus.co.jp/en/), to allow through-the-objective (TTO; see below) visual stimulation over a sufficiently large field. Different parts of the dendrite were imaged by shifting the laser-scanner offset without moving the objective lens, thus, leaving the TTO stimulus pattern stationary. The other microscope was equipped with through-the-condenser (TTC; see below) stimulation and a 60× water immersion lens (0.9NA; Leica, http://www.leica-microsystems.com). In both cases, the two-photon excitation source was a mode-locked Ti/sapphire laser (Mira-900; Coherent, http://www.coherent.com) tuned to approximately 930-nm wavelength. The scanning laser beam caused only moderate and/or quickly adapting electrical responses, which made it possible to record calcium signals in response to visual stimuli [75]. Spectrally nonoverlapping filters in the stimulation light path (see Light stimulation, below) and in front of the two detectors (“green channel,” Ca2+ indicator signal: D 535 BP 50 or 520 BP 30; “red channel,” Sulforhodamine: HQ 622 BP 36, Chroma/AHF, http://www.chroma.com), ensured that the 578- or 600-nm stimulus light did not interfere with the detection of fluorescence.
For Ca2+ imaging, SACs were filled with OGB-1 by diffusion from the patch pipette. The imaging software (CfNT) was written by R. Stepnoski (Bell Labs) and M. Müller (Max-Planck Institute for Medical Research). Small-image series (64 × 8 pixels at 62.5 Hz) were acquired from short dendritic segments and analyzed off-line with IgorPro (Wavemetrics). The observed changes in fluorescence suggest changes in [Ca2+] of hundreds of nM; no attempt was made at absolute quantification.
Light stimuli were generated using custom-written Windows software compiled with Delphi7 (Borland, http://www.borland.com). Two different stimulus projection systems were used: In the TTC setup, the light stimuli were output via a video projector (Astrobeam 530, 80 Hz frame rate; 800 × 600 pixels; A+K, http://www.anders-kern.de), band pass-filtered (600 BP 20), and focused through the substage condenser (0.32 NA; Carl Zeiss, http://www.zeiss.com) onto the photoreceptors (≈4.4 μm/pixel; approximately 130–2,000 photons·s−1·μm−2 for 600 nm). In the TTO setup, stimuli were displayed using a miniature liquid crystal on silicon (LCoS) display (i-visor DH-4400VP; Cybermind Interactive, http://www.cybermindnl.com, or i-glasses VGA; EST, http://www.est-kl.com, 30 or 60 Hz frame rate; 800 × 600 pixels) illuminated by a band pass-filtered (578 BP 10) yellow LED. The output from the stimulator passed through some scaling and focus-correction optics as well as scan and tube lenses to be finally projected by the objective lens (≈1-mm field of view) onto the retina (stimulus resolution on the retina ≈ 2.1 μm/pixel; approx. 10,000–40,000 photons·s−1·μm−2 for 578 nm; measured with an Optical Power Meter, Model 840; Newport, http://www.newport.com). The focus-correction optics provides a shift, by up to approximately 100 μm, between microscope and stimulus focus planes. The stimulus thus remained focused on the photoreceptors while imaging dendrites in the inner plexiform layer.
With either stimulus projection system, stimulus contrast (C) was calculated using
with LMIN and LMAX being the minimal and maximal light intensities, respectively, which were chosen to always be in the linear range of the display device. The velocities of the moving stimuli (gratings, circular waves) ranged from 0.5 to 2 mm/s (temporal frequencies: 2 to 10.5 Hz).
Circular wave stimuli were centered on the cell and consisted of expanding or contracting concentric sinusoidal waves (first term in Equation 3) with one full cycle per stimulus radius (rMax). Pixel values (P) were calculated using
with the distance from the stimulus center (soma), r (in μm; 0 ≤ r ≤ rMax), the stimulus frequency fStim , the display refresh frequency fDisplay, and the frame index i. To avoid sharp edges and to restrict the motion to roughly the outer two-thirds of the SAC dendrites, the resulting sine wave was filtered with a “doughnut-shaped” super-Gauss (second term) with r0,Filter = 0.5·rMax, and wFilter = 0.562·rMax. To get the desired contrast, the values were then scaled by
the background was kept at average intensity ((LMAX − LMIN)/2), such that the stimulus consisted of alternating brighter- and darker-than-average waves (e.g., drawing in Figure 1D).
For most SACs recorded (dendritic field diameters ≈ 280 to 360 μm), we used rMax = 192 μm, which means that the annular area with the circular wave extended between approximately 45 and 147 μm from the center. For the few cells that had smaller or larger dendritic fields, we used an rMax of 128 or 256 μm, respectively. For the data shown in Figure 8A–8D, rMax was 192 μm and the stimulus was presented only inside a broad ring (inner-outer diameter: Figure 8A: 90–360 μm; 8B: 90–270 μm; 8C: 90–210 μm; and 8D: 210–360 μm). The background was set to the mean intensity of the circular wave.
Fluorescence (F) data were spatially averaged over regions of interest (ROI) that covered a 5–15 μm–long section of a distal dendrite and temporally filtered (box-car filter with a 42-ms window) and background corrected using a ROI placed next to the dendrite. For each response, ΔF(t)/F0 was calculated (F0 being the prestimulus F averaged over 200 to 500 ms). For quantification, the area under the trace (
) for a defined time interval (t1 < t < t2) was calculated (for details, see [15]).
For the Ca2+-activation curves (Figure 7), we normalized ACa for each cell to the Ca2+ response at −35 mV taken from the interpolated [Ca2+] versus VCOM relationship of the cell (with spline-interpolated Ca2+ values between voltage steps). The data points pooled from all cells were then fitted using (adapted from Equation 1 in [76]):
with somatic clamp potential VCOM and free parameters ACa,0 (offset), g (max. conductance), V50 (half-activation voltage), and VSlope (the slope voltage).
To quantify the electrical response to circular wave stimuli, we used spectral analysis by Fourier decomposition (Figure 2) of trace segments that contained an integer number of stimulus cycles and excluded the onset of the motion response (Figure 2A, black rectangle). Discrete Fourier decomposition can introduce spurious harmonic frequencies if the signals contain components (such as a stimulus-onset transient) that are not periodic with the stimulus frequency. Therefore, the analysis window was shifted in 50-ms steps along the trace until the residual (difference between data trace and reconstruction, see below) was minimal. Four components at different frequencies were extracted from this analysis: the DC component V0 (at 0 Hz), the fundamental, V1 (at f1), and the second and third harmonics, V2 (at 2·f1) and V3 (at 3·f1) (I0, I1, I2, and I3 for voltage-clamp measurements). These four components (each comprising amplitudes and, except for V0 or I0 phases) are sufficient to capture the response characteristics (Figure 2C, inset). To ensure reliable analysis, we only included cells with a robust light-evoked voltage response (≥2 mV) to CF motion of the standard circular-wave stimulus at the lowest used stimulus contrast ≥45%. To facilitate the comparison between cells, we used normalized asymmetry indices (AI0..3):
with n = 1, 2, or 3. A positive index means that CF motion is preferred. Note that, because only amplitude but no phase information is used, all indices equal zero does not imply identical waveforms. Although the absolute phases depend on the origin of the time window, the quantities
(relative phase) do not and thus can be used to characterize the waveforms (Figures 2C and 3C). The maximal response excursions were determined from a histogram of the voltages within the analysis window (VValley = 5%; VPeak = 95%).
All averages are given as mean ± standard error of the mean (SEM). Statistical significance of differences was evaluated using the Wilcoxon matched-pair signed-ranks test [77] implemented in Matlab (Mathworks, http://www.mathworks.com) with p ≤ 0.05 considered as significant.
The circuit diagram of the two-compartment model is shown in Figure 9B. Both compartments contain the same type of active conductance, which is modeled as a Hodgkin-Huxley–type channel, containing one m and one h element (e.g., [48]). For the m element, the activation probability follows
where m′ denotes the time derivative of m and km the off-rate. Only the on-rate is assumed to be voltage (V) dependent, whereby Vm50 and VmSlope are the half-activation and slope voltages, respectively. The equation for the h element is identical. The voltages in the two compartments (VP and VD) can be calculated by solving the following system of coupled differential equations:
With longitudinal resistor (RPD), membrane capacitor (C), leak resistor (Rleak), injected input currents (IP,zero, ID,zero, IP,osc, and ID,osc), and for the VGCs: maximal conductance (gopen·nch), reversal potential (Erev). For simulation parameters and results, see Figure 9 and Table 2.
|
10.1371/journal.pbio.1001786 | Contribution of Orb2A Stability in Regulated Amyloid-Like Oligomerization of Drosophila Orb2 | How learned experiences persist as memory for a long time is an important question. In Drosophila the persistence of memory is dependent upon amyloid-like oligomers of the Orb2 protein. However, it is not clear how the conversion of Orb2 to the amyloid-like oligomeric state is regulated. The Orb2 has two protein isoforms, and the rare Orb2A isoform is critical for oligomerization of the ubiquitous Orb2B isoform. Here, we report the discovery of a protein network comprised of protein phosphatase 2A (PP2A), Transducer of Erb-B2 (Tob), and Lim Kinase (LimK) that controls the abundance of Orb2A. PP2A maintains Orb2A in an unphosphorylated and unstable state, whereas Tob-LimK phosphorylates and stabilizes Orb2A. Mutation of LimK abolishes activity-dependent Orb2 oligomerization in the adult brain. Moreover, Tob-Orb2 association is modulated by neuronal activity and Tob activity in the mushroom body is required for stable memory formation. These observations suggest that the interplay between PP2A and Tob-LimK activity may dynamically regulate Orb2 amyloid-like oligomer formation and the stabilization of memories.
| The formation of stable long-term memories involves the synthesis of new protein, however the biochemical basis of this process is unclear. A family of RNA binding proteins, Cytoplasmic Polyadenylation Element Binding (CPEB) proteins, are known to regulate synaptic activity and stabilization of memory. The Drosophila CPEB is called Orb2, and its amyloid-like oligomers are critical for the persistence of long-lasting memories. Amyloid formation is often unregulated and stochastic in nature, and the amyloid state is usually dominant and self-sustaining. However, to serve as a substrate for long-lasting memory, the amyloid-like oligomerization of Orb2 must be regulated in a space-, time-, and stimulus-specific manner. Orb2 has two protein isoforms: Orb2A, which is present only in low abundance, and Orb2B, which is the abundant form. Orb2A is important for oligomerization as well as memory persistence. Previous studies suggested that Orb2A may act as a seed to induce oligomerization of the constitutive Orb2B isoform. Therefore, the availability of Orb2A protein would be an important determinant of Orb2 oligomerization. Here we have analyzed how Orb2 conversion to the oligomeric state is regulated. We find that Orb2A is a very unstable protein and that phosphorylation-dephosphorylation of this isoform via canonical neuronal signaling modules can regulate Orb2A stability, and thereby its abundance. We also show that Tob, a known regulator of CPEB-mediated translation, acts as a stabilizer for Orb2A and triggers Orb2 oligomerization. These observations suggest that amyloid formation can be regulated in a dynamic manner by controlling the availability of the seeding Orb2A protein.
| Synthesis of new protein is important for the formation of stable memory [1]. The Cytoplasmic Polyadenylation Element Binding (CPEB) proteins are a family of RNA binding proteins that regulate the translation and subcellular distribution of a specific set of cellular mRNAs in various cell types including neurons [2]. Previous studies found that some CPEB family members play a causal role in long-term change of synaptic activity and in stabilization of memory [3]–[9]. For example, in marine snail Aplysia, in the absence of a neuron-specific ApCPEB, serotonin mediated enhancement of synaptic transmission fails to persist beyond 24 h [7],[10]. Likewise, the Drosophila CPEB, Orb2, is required specifically for long-term memory but not for learning or short-term memory [3]–[5]. In humans, a particular CPEB family member, CPEB3, has been linked to episodic memory formation, suggesting a conserved role of CPEB in synaptic plasticity and memory [11].
Interestingly, ApCPEB and Orb2 form self-sustaining amyloidogenic oligomers (prion-like) in response to the neurotransmitters serotonin in Aplysia and octopamine or tyramine in Drosophila [5],[6],[12],[13]. More importantly, the oligomeric CPEB is required for the persistence of synaptic facilitation in Aplysia [6] and for the stabilization of memory in Drosophila [5]. These observations led us to propose that the persistent form of memory recruits an amyloidogenic oligomeric form of neuronal CPEB to the activated synapse, which in turn maintains memory through the sustained, regulated synthesis of a specific set of synaptic proteins [5]. However, considering the dominant and stable nature of amyloids, a central question is how the conversion of neuronal CPEB to the amyloidogenic state is regulated to confer activity dependence and restrict it to the relevant neuron/synapse.
The Drosophila Orb2 gene has two protein isoforms, Orb2A and Orb2B, and the oligomers are composed of both Orb2A and Orb2B. In the adult brain, in comparison to the Orb2B protein, the Orb2A protein is expressed at an extremely low level [4],[5]. In spite of its low abundance, the Orb2A protein is critical for Orb2 oligomerization, and Orb2A forms oligomers more readily than Orb2B. More importantly, a mutation that impedes Orb2A oligomerization selectively affects persistence of memory [5], and the Orb2A prion-like domain is sufficient for long-term memory formation [4]. These observations suggested a model in which the rare Orb2A protein either acts directly as a seed to induce activity-dependent amyloid-like oligomerization of the constitutive Orb2B protein or Orb2A oligomerization indirectly affects oligomerization of Orb2B [5]. In either case the amount and localization of Orb2A protein would therefore be a key determinant of when and where amyloid-like conversion would occur.
Here we find that Orb2A has a very short half-life and the Orb2 interacting protein Transducer of Erb2 (Tob), a known regulator of cellular growth, stabilizes Orb2A and induces Orb2 oligomerization. Expression of dsRNA against Tob in the mushroom body neurons does not affect learning, but impairs long-term memory formation. Tob recruits the neuronal protein kinase Lim Kinase (LimK) to the Tob-Orb2 complex to induce Orb2 phosphorylation. Phosphorylation regulates Tob-Orb2 association as well as the stability of both proteins, and Protein Phosphatase 2A (PP2A) is a key regulator of the phosphorylation status of Tob and Orb2. Intriguingly, inhibition of PP2A stabilizes Orb2A, but destabilizes Orb2-associated Tob, providing a mechanism for temporal restriction on Orb2A stabilization. Since PP2A and LimK activity can be regulated in a synapse-specific manner [14],[15], the phosphorylation-dephosphorylation of Orb2 and Tob provides a putative mechanism of restricting the Orb2 oligomerization to the activated synapse. Tob is also known to regulate the function of CPEB family members [16]. Therefore, the Tob-Orb2 association-dissociation may also regulate Orb2 function in the nervous system.
A regulator of Orb2 oligomerization could potentially fall into at least two distinct categories: an activator that associates with Orb2 and facilitates conversion to the oligomeric state or a repressor that binds to Orb2 and prevents its oligomerization. To identify both types of regulators we used a proteomics approach to perform a comprehensive search for Orb2 interacting proteins in the adult Drosophila brain. The Orb2 proteins were expressed pan-neuronally as C-terminal HA-tagged proteins (ElavGal4: UAS-Orb2AHA or Orb2BHA), and the Orb2 complex was immunopurified using anti-HA antibodies from RNaseA-treated adult head extract (Figure 1A and B). Previously we observed that the C-terminal tags are inaccessible in the Orb2 oligomeric state [5]; thus, the anti-HA antibody preferentially immunopurified the Orb2 monomers. Therefore, to identify proteins that interact with oligomeric Orb2, we also immunopurified Orb2AHA with an anti-Orb2 antibody (Figure 1B and C). The factors associated with Orb2 were identified using Multidimensional Protein Identification technology (MudPIT) (Table S1) [17].
We found 61 proteins that were significantly enriched (p<0.05) in the Orb2 immunoprecipitates compared to eight independent control immunoprecipitates (Figures 1D and S1A and Table S1). Eleven proteins were overrepresented in Orb2 IP compared to the controls, albeit not to statistical significance (Table S1). To determine the validity of the proteomic approach, we randomly sampled 20 candidate proteins (out of 72 proteins) by pair-wise interaction in S2 cells (Figure 1E and Figure S1B). Approximately 50% (11 out of 20 proteins) of the proteins thus tested formed a complex with at least one of the Orb2 proteins in an RNA-independent manner (Figure 1E and Figure S1B). Therefore, the proteomics approach indeed identified specific components of an Orb2 protein complex in the adult Drosophila brain. The candidate proteins either interact directly with Orb2 or indirectly as part of a larger Orb2 protein complex. A gene ontology (GO) analysis revealed that the Orb2 proteome is significantly enriched for proteins involved in translation initiation, mRNA binding, and synaptic activity (Figure 1F). The enrichment of these protein complexes supports the idea that Orb2 is involved in regulation of synaptic protein synthesis.
The Orb2A protein is undetectable by Western analysis, and a genomic construct expressing Orb2A-EGFP suggests it is ∼100 times less abundant than Orb2B protein in the adult brain [5]. Moreover, monomeric Orb2A has a very short half-life compared to Orb2B (Figure 2A and Table S2). Taken together, these observations suggest that availability of the Orb2A protein could be an important determinant of efficient Orb2A oligomerization and/or function. In the course of our interaction studies in S2 cells, we noticed one of the candidate proteins, Tob, may influence the Orb2A protein level (Figure 1E). To determine Orb2A and Orb2B stability independent of each other, we used Drosophila S2 cells, in which Orb2 is normally not expressed and Tob is expressed at low levels. S2 cells were transfected with only HA-tagged Orb2 or coexpressed with Flag-tagged Tob. To determine half-life, total Orb2 or Tob protein levels were measured at several time points following treatment with cycloheximide (CHX), which blocks new protein synthesis. The coexpression of Tob nearly doubled the half-life of monomeric Orb2A (Figure 2A). However, Tob had no significant effect on Orb2B (Figure 2B), indicating that association with Tob does not automatically enhance half-life. Likewise, incubation with dsRNA against Tob reduced the level of Orb2A protein but not Orb2B (Figure S2A). Earlier studies with Tob family members have suggested that the stability of Tob itself can be regulated [18],[19]. We found a fourfold increase in Tob half-life in the presence of either Orb2A or Orb2B compared to Tob alone (Figure 2C and Table S2). These results suggest that not only does Tob stabilize Orb2A, but Orb2 proteins have stabilizing effects on Tob.
The recombinant Drosophila Tob interacts with in vitro transcribed and translated Orb2 proteins, suggesting direct interaction between these proteins (Figure S2B). In mammals, the Tob family consists of six members, with Drosophila Tob most closely related to the mammalian Tob1 and Tob2 proteins [20]. We found both Aplysia CPEB and mouse CPEB3 interact with the closely related Tob1 and Tob2, and Tob2 increases the steady-state level of ApCPEB and CPEB3 (Figure S2C). Recently, others have reported a direct interaction between mouse CPEB3 and Tob1 [16], suggesting that Tob is an evolutionarily conserved interactor of CPEB proteins. Tob is required for long-term potentiation of hippocampal CA1–CA3 synapses, a cellular correlate of long-term memory in mammals [21], and Tob activity is modulated by bone morphogenetic proteins or BMPs [22]–[24]. These observations suggest Tob could function as an extracellular signal-dependent regulator of Orb2 in the nervous system.
Does Tob influence Orb2 oligomerization in the adult fly brain? To answer this, we increased Tob level in the fly brain using the Gal4-UAS system and assessed Orb2 oligomerization by immunopurification. Overexpression of Tob-TdTomato (Elav-Gal4: UAS-TobTdTom), but not the fluorophore alone (Elav-Gal4: UAS-TdTom), increased the levels of 10% SDS and boiling-resistant oligomeric Orb2 in the fly brain (Figure 2D). The amount of Orb2 oligomers in Tob-expressing flies increased nearly 2-fold compared to control flies (fold increase in oligomers normalized to monomer ± SEM, 1.95±0.27, p<0.05, t test). Tob has been implicated in a number of cellular processes, including transcriptional regulation and RNA metabolism [23],[25]–[28], raising the possibility that the increase in Orb2 oligomerization is a secondary effect of Tob overexpression. We generated a series of deletion mutants of Tob and found that deletion of a conserved 28 amino acid motif, TobΔ28 (Figure S3A), decreased the interaction between Tob and Orb2 in both S2 cells (Figure S3B) and the adult fly brain (Figure S3C). However, it had no effect on the association between Tob and the deadenylase Pop2 (Figure S3D) or with the transcriptional repressor, Mad (Figure S3E). Overexpression of TobΔ28 (Elav-Gal4: UAS-TobΔ28TdTom) in the adult brain did not enhance Orb2 oligomerization (fold increase in oligomers normalized to monomer ± SEM, 0.5±0.14) (Figure 2D).
Does Tob enhance oligomerization of Orb2A, Orb2B, or both? EGFP-tagged Orb2A and Orb2B formed stable oligomers in the adult fly brain and the oligomers associated with Tob (Figure S3F). To determine the effect of Tob on Orb2A and Orb2B, we coexpressed TdTomato-tagged Tob with EGFP-tagged Orb2A or Orb2B in the adult fly brain. To distinguish from the endogenous Orb2, we quantified changes in the number of fluorescent puncta, since the abundance of fluorescence puncta co-relate with extent of oligomerization (Figure 2E) [5]. The number of Orb2A-EGFP puncta increased ∼2-fold in the presence of Tob (number of puncta/100 µm2 ± SEM, Orb2A: 4.41±2.88, N = 12; Orb2A+Tob: 8.06±3.69, N = 15, t test, p = 0.012) but not in the presence of TobΔ28 (3.58±1.62, N = 11, t test, p>0.5) (Figure 2F). Unlike Orb2A, Orb2B∶EGFP by itself remained mostly diffused and Tob overexpression had no significant effect on the rare Orb2B puncta (number of puncta/100 µ∶m2 ± SEM: Orb2B, 2.40±2.11, N = 12, Orb2B+Tob, 1.07±1.30, N = 9, t test, p = 0.155) (Figure 2F and Figure S3G). In addition to being more numerous, the size of Orb2A puncta also increased significantly when Tob was overexpressed (size of puncta ± SEM; Orb2A, 0.39±0.08 µm2, Orb2A+Tob, 0.51±0.07 µm2, p = 0.0003), an effect not seen with the rare Orb2B puncta (Orb2B, 0.43±0.05 µm2, Orb2B+Tob, 0.38±0.06 µm2, p = 0.20) (Figure 2G). Taken together, these observations suggest that Tob-Orb2 association promotes Orb2 oligomer formation either by increasing the Orb2A protein levels and/or enhancing oligomerization.
Is Tob involved in activity-dependent oligomerization of Orb2? Previously we and others have observed that a neurotransmitter such as tyramine or dopamine regulates Orb2 oligomerization [4],[5]. Therefore, we checked whether tyramine modulates Orb2-Tob interaction. To this end, we fed-starved flies 10 mM tyramine and after 4 h immunopurified the Tob-Orb2 complex from tyramine-stimulated or -unstimulated adult fly brain using a Drosophila Tob-specific antibody (Figure S4A). Tyramine stimulation increased the Tob-bound oligomeric Orb2 nearly 4-fold (fold increase in oligomers normalized to monomer ± SEM, 3.82±0.88, n = 5, t test, p<0.05) (Figure 3A), and the oligomers are resistant to boiling in the presence of 10% SDS and 2 M urea, consistent with it being amyloid-like (Figure 3B). The neurotransmitter serotonin (5-HT) had less effect on Tob-Orb2 association (Figure S4B), consistent with our earlier observation that Orb2 oligomerization is influenced by tyramine and not by 5-HT [5]. Use of Orb2B-specific antibody (Figure 3A, right panel) indicated Tob-Orb2B association is enhanced by tyramine stimulation. To determine whether Tob-Orb2A association is also modulated by neuronal activity, we used a genomic construct that encompasses only Orb2A coding region and carries EGFP at the C-terminal end (pCasperOrb2AEGFP) [5]. In Tob immunoprecipitate from tyramine-treated samples, we see EGFP reacting bands that correspond to the size of the monomeric- (∼87 KDa) and oligomeric-Orb2AEGFP (Figure 3C). Since it is difficult to determine which neuronal populations are activated by tyramine feeding, we also directly activated the mushroom body neurons (c747-Gal4, MB247-Gal4) with the temperature-sensitive dTrpA1 channel [29]. The flies were transiently exposed to 30°C (dTrpA1 active) for 25 min and then returned to 22°C (dTrpA1 inactive). Compared to flies carrying only dTrpA1 or Gal4, flies carrying both transgenes (C747Gal4::UAS-dTrpA1 or MB247Gal4:UAS-dTrpA1), there was enhanced Tob-Orb2 association (Figure 3D). Taken together these observations suggest that neuronal activity that enhances Orb2 oligomerization also enhances Tob-Orb2 association.
Because Tob was initially identified as a transcriptional regulator [23],[24], we asked whether Tob is restricted to the cell body or distributed throughout the neuron, including the synaptic region. Immunostaining of the adult fly brain revealed that, as expected, Tob is present mostly in the cell body (Figure S4C). However, at low levels Tob staining was also detected in the synaptic-neuropil regions (Figure 3E, mushroom body lobes). Previously we established a method to purify synaptosomes from adult Drosophila head [5]. In Western blotting of synaptosome fractions (Figure S4D, left panel) Tob was found in the synaptic membrane fraction, similar to Orb2 (Figure S4D). In Δ80QOrb2 flies, which has significantly less Orb2 protein compared to wild-type flies [5], the distribution of Tob was unaffected, suggesting synaptic localization of Tob is independent of Orb2 (Figure S4D). Similar to the fly brain, Tob was also detected in the synaptic membrane fraction prepared from the mouse brain (Figure S4E). Activity-dependent association with Orb2 and presence in the synaptic region suggest that Tob may act to regulate Orb2 function and/or oligomerization in the synapse.
Because Tob is constitutively present in the adult fly brain, we wondered how Tob-mediated oligomerization of Orb2 could be temporally regulated by neuronal activity. Phosphorylation is known to regulate the activity of both Btg/Tob [30]–[32] as well as the CPEB family members [33]–[35]. Consistent with these observations, protein phosphatase 1 (PP1-87B) and protein phosphatase 2A (PP2A) regulatory subunit twins were found in the Orb2 protein complex (Table S1), suggesting that Orb2 may also be regulated via phosphorylation and/or that Orb2 recruits these phosphatases to regulate phosphorylation of other proteins (such as Tob) in the complex. Blotting of Orb2 immunoprecipitates from the adult brain with phospho-tag™ [36], a biotin-tagged dinuclear metal complex that selectively binds to phospho-proteins, detected a small amount of phosphorylated monomeric Orb2B protein (Figure 4A). Similar to the fly brain, when expressed ectopically in S2 cells, both Orb2A and Orb2B are phosphorylated, albeit at very low levels (Figure 4B), suggesting Orb2 proteins are transiently phosphorylated in a regulated manner or kept primarily in an unphosphorylated state by the phosphatase. We observed that Tob is also phosphorylated in the adult fly brain (Figure 4C). To avoid a secondary consequence of prolonged inhibition or activation of phosphatases or kinases in the nervous system, we took advantage of the phosphorylation of Orb2 and Tob in S2 cells to determine the acute role of phosphorylation.
To determine if phosphorylation has any effect on Tob-Orb2 association, we blocked dephosphorylation using calyculin (CY), a cell-permeable serine-phosphatase inhibitor that blocks protein phosphatase 2A (PP2A) at 0.5–1.0 nM concentration and protein phosphatase1 (PP1) at ≥2 nM concentration [37]. We observed that an hour after treatment with 1 nM CY, the amount of Orb2A associated with Tob was reduced (Figure 4D). The reduction in association was not due to reduction in Tob or Orb2A protein level an hour after treatment with CY (Figure 4D). Because phosphatases influence a large number of proteins in the cell, the reduction of Tob-Orb2 association could be a secondary consequence of phosphatase inhibition. To test more directly the effect of phosphorylation, in a reciprocal experiment, we first treated cell lysates expressing Tob and Orb2A with calf intestinal phosphatase (CIP) and then isolated the Tob-Orb2 complex (Figure 4E). We observed that prior dephosphorylation enhanced the association of Tob with Orb2A (Figure 4E). Likewise, when the Orb2-Tob complex was immunopurified with anti-Orb2 antibody and probed with phospho-tag™, only phosphorylated Orb2, but not the hyperphosphorylated Tob, was detected in the immunoprecipitate (Figure 4F). Taken together, these results indicate phosphorylation regulates Tob-Orb2 association. Hypophosphorylation promotes Tob-Orb2A association, and hyperphosphorylation reduces it.
Because the Tob-Orb2 association alters the half-life of both proteins and phosphorylation affects their association, we examined the effect of phosphatase inhibition on the half-life of both proteins. When Tob was expressed by itself there was modest change in stability in the presence of CY (Table S2) compared to the untreated samples (Figure 5A). Interestingly, the increase in Tob stability that occurred when co-expressed with either Orb2A (Figure 5B) or Orb2B (Figure 5C) was ∼50% reduced when the phosphatases were inhibited (Table S2). The destabilization of Tob was observed only in the presence of the PP2A/PP1 inhibitor CY or okadaic acid (1 nM) but not the PP1 selective inhibitor tautomycin (10 nM) (Figure S5A) [37],[38]. Moreover, the extent of Tob phosphorylation appears to be specifically linked to Orb2 complex formation (Figure 5D). The Orb2 proteins, but not the other homologue of CPEB in Drosophila, Orb1, enhance phosphorylation of Tob, although Tob interacts with both Orb2 and Orb1 (Figure S5B and C). These results suggest un- or hypophosphorylated Tob binds Orb2. Association of Tob with Orb2 and PP2A inactivation leads to additional phosphorylation of Tob-Orb2, which results in dissociation and eventual destabilization of Tob.
How does phosphorylation affect Orb2? Treatment of S2 cells with PP2A/PP1 inhibitors CY (1 nM) and okadaic acid but not PP1-specific inhibitor tautomycin (10 nM) enhanced phosphorylation of both Orb2A and Orb2B (Figure 5E and Figure S5D). Treatment with alkaline phosphatase, which removes phosphate from serine/threonine, and λ phosphatase, which removes phosphate from serine/threonine as well as tyrosine residues [39], revealed that upon inhibition of PP2A, Orb2 proteins are phosphorylated at multiple sites (Figure 5F). One of the outcomes of these multiple phosphorylations is a significant increase in Orb2A half-life, from 1 h to >24 h, t(1/2) Orb2A, 1.13±0.08, Orb2A+CY, 35.5±17.5 h; p = 0.010, and doubling of the Orb2B half-life, t(1/2) Orb2B, 4.32±0.53, Orb2B+CY, 8.09±2.95 h, p = 0.05 (Figure 5G). As decreases in PP2A activity increased Orb2 level, likewise increases in PP2A activity by overexpression of PP2A catalytic subunit microtubule star (Mts) that associates with Orb2 (Figure S5E) resulted in a ∼4-fold decrease in Orb2A (0.23±0.01, n = 5) and a ∼2-fold decrease in Orb2B (0.51±0.02, n = 3) protein level (Figure 5H). Increases or decreases in protein phosphatase 1 87B (PP1) activity had no effect on Orb2A or Orb2B abundance (Figure 5E and Figure S5F). These results suggest like Tob, Orb2 phosphorylation is regulated by PP2A. However, unlike Tob, inhibition of PP2A stabilizes Orb2, particularly Orb2A.
How does Tob promote Orb2A stabilization and/or enhanced Orb2 oligomerization? Because phosphorylation enhances Orb2 stability, one possibility is that Tob prevents PP2A from accessing Orb2A. However, the association of PP2A catalytic subunit Mts or regulatory subunit Tws with Orb2 was not affected by increased levels of Tob, and the effect of PP2A on Orb2A half-life was not dependent on Tob level (Orb2A, 25.6±14.7 h, p = 0.02, and Orb2B, 19.5±8.3, p = 0.01) (Figure S6A). However, we found Tob promotes Orb2 phosphorylation by recruiting LimK to Tob-Orb2 complex.
In our effort to identify kinases that phosphorylate Tob, we initially focused on MapK, as in mammals and in C. elegans Tob is phosphorylated by Map Kinase (MapK) [19],[30],[31] and MapK sites are conserved in Drosophila Tob (Figure S6B). However, in an in vitro kinase assay, MapK did not phosphorylate recombinant Drosophila Tob, although as expected mammalian Tob1 and Tob2 were phosphorylated (Figure S6C). We searched for other kinases and focused on the neuronal kinase LimK for several reasons. First, Tob activity is regulated by BMPs, and in the nervous system LimK is a key mediator of BMP signaling [40]–[44]. Second, neuronal activity regulates the synaptic concentration of LimK [15]. Finally, LimK is required for synapse formation [40],[45],[46], which is reminiscent of the function of ApCPEB [10] and Orb2 (our unpublished observation). In an in vitro kinase assay, we found LimK efficiently phosphorylates recombinant Drosophila Tob as well as the mammalian Tob1 and Tob2 (Figure 6A) but weakly phosphorylates maltose binding protein or Tob family member Btg. Tob is a LimK substrate because in the adult fly head (Figure 6B) as wells as in S2 cells (Figure S6D) LimK associates with Tob.
Next we sought to determine whether Tob phosphorylation by LimK is influenced by Orb2. We performed in vitro LimK assays on immunopurified Tob-Orb2 complex or on Tob alone (Figure 6C). To our surprise, we observed that Orb2 is phosphorylated by exogenously added LimK in the presence of Tob (Figure 6C). The Tob-Orb2 immunoprecipitate from cells contains other proteins in addition to Tob and Orb2, and therefore Orb2 may be phosphorylated by other kinases in the presence of LimK. To test such a possibility, we incubated recombinant-soluble Orb2B protein and LimK in the presence or absence of recombinant MBP-tagged Tob. We observed phosphorylation of Orb2B by LimK in the presence of Tob (Figure 6D). Furthermore, LimK copurified with both Orb2A and B only in the presence of Tob. However, in the presence of TobΔ28, which binds efficiently to LimK (Figure S6D) but not to Orb2, there was a marked reduction in the LimK-Orb2 complex (Figure 6E). Together, these data suggest that Tob is a substrate for LimK and that Orb2 proteins become a substrate of LimK when associated with Tob.
Does LimK affect Orb2 oligomerization? To determine whether LimK regulates activity-dependent oligomerization of Orb2 in the adult brain, we examined Orb2 oligomer formation in a LimK hypomorphic mutant LIMK1EY08757 [40]. In the LIMK1EY08757 adult brain, the level of monomeric Orb2B protein level was similar to that of wild-type flies (Figure 7A). We exposed wild-type and LimK mutant flies to 10 mM tyramine and immunopurified either the Orb2 oligomers (Figure 7B) or the Orb2 oligomers associated with Tob (Figure 7C). In the unstimulated brain extract, little or no oligomeric Orb2 was observed in the LimK mutant flies (Figure 7B and C). More importantly, unlike wild-type flies, LimK mutant flies did not undergo a tyramine-dependent increase in Orb2 oligomerization (Figure 7B and C).
To determine whether an increase in LimK activity enhances Orb2 oligomerization, we analyzed Orb2 puncta formation in the larval neuron, where unlike the adult brain, ectopic expression of LimK did not cause any observable developmental problem. We found that Orb2A-EGFP coexpressed with active LimK (ElavGal4::UAS-Orb2A-EGFP; UAS-LimK) has twice the number of puncta (16.10±1.36, N = 24) compared with flies coexpressing a kinase dead version of LimK, LimKKD (ELAV::UAS-Orb2A-EGFP; UAS-LimKKD) (8.71±1.74, N = 6, p<0.05) or flies expressing only Orb2A-EGFP (6.79±1.01, N = 12, p<0.001) (Figure 7D). From these several results, we conclude that Tob serves two functions for Orb2A. First, it binds and stabilizes unphosphorylated Orb2A, and second, it allows Orb2A to be phosphorylated by LimK. Each of these events results in an increase in the effective concentration of Orb2A, which induces Orb2A and/or Orb2A-Orb2B oligomerization.
Because Orb2 oligomerization is important for long-term memory and Tob affects Orb2 oligomerization, we wondered whether Tob activity is important for long-term memory. To this end, we used the male courtship suppression paradigm in which a virgin male fly learns to suppress its courtship behavior upon repeated exposure to an unreceptive female (Figure 7E) [47]. Previously we and others have found male courtship suppression memory is dependent on Orb2 activity [4],[5]. The TobRNAi was expressed under mushroom-body-specific driver 201Y Gal4, which drives expression primarily in the γ-lobe neurons [48]. Expression of Orb2 in γ-lobe in an otherwise orb2 null background is sufficient to rescue the long-term memory defect [3],[4]. We found that male flies expressing TobRNAi (201Y:Gal4-UAS-TobRNAi) in the γ-lobe showed courtship suppression after training in the short term (5 min), but the courtship suppression was lost when measured at 24 h or 48 h after training (Figure 7E). In contrast, flies harboring just the RNAi (UAS-Tob RNAi) or Gal4 (201Y:Gal4) had no impairment in courtship suppression 5 min or 24 to 48 h after training. These results are consistent with the idea that Tob activity is important for long-term courtship suppression memory.
Our previous work suggested that conversion of neuronal CPEB to an amyloid-like oligomeric state provides a molecular mechanism for the persistence of memory [5],[6]. However, it is not known how Orb2 oligomerization is regulated so that it will occur in a neuron/synapse-specific and activity-dependent manner. Here we report that factors that influence Orb2A stability and thereby abundance regulate Orb2 oligomerization.
We find that Tob, a previously known regulator of SMAD-dependent transcription [23],[24] and CPEB-mediated translation [16], associates with both forms of Orb2, but increases the half-life of only Orb2A. Stimulation with tyramine or activation of mushroom body neurons enhances the association of Tob with Orb2, and overexpression of Tob enhances Orb2 oligomerization. Both Orb2 and Tob are phosphoproteins. Phosphorylation destabilizes Orb2-associated Tob, whereas it stabilizes Orb2A. Tob promotes Orb2 phosphorylation by recruiting LimK, and PP2A controls the phosphorylation status of Orb2A and Orb2B.
PP2A, an autocatalytic phosphatase, is known to act as a bidirectional switch in activity-dependent changes in synaptic activity [14],[49]–[51]. PP2A activity is down-regulated upon induction of long-term potentiation of hippocampal CA1 synapses (LTP) and up-regulated during long-term depression (LTD) [14]. Similarly, Lim Kinase, which is synthesized locally at the synapse [15] in response to synaptic activation, is also critical for long-term changes in synaptic activity and synaptic growth [46].
Based on these observations we propose a model for activity-dependent and synapse-specific regulation of amyloid-like oligomerization of Orb2 (Figure 8). We postulate that in the basal state synaptic PP2A keeps the available Orb2A in an unphosphorylated and thereby unstable state. Neuronal stimulation results in synthesis of Orb2A by a yet unknown mechanism. The Tob protein that is constitutively present at the synapse binds to and stabilizes the unphosphorylated Orb2A and recruits the activated LimK to the Tob-Orb2 complex, allowing Orb2 phosphorylation. Concomitant decreases in PP2A activity and phosphorylation by other kinases enhances and increases Orb2A half-life. The increase in Orb2A level as well as phosphorylation may induce conformational change in Orb2A, which allows Orb2A to act as a seed. Alternatively, accumulation and oligomerization of Orb2A may create an environment that is conducive to overall Orb2 oligomerization. In the absence of an in vitro Orb2A-Orb2B oligomerization assay, we could not distinguish between these two possibilities.
For Tob, initial Orb2 association stabilizes Tob. However, association with Orb2 as well as suppression of PP2A activity leads to additional phosphorylation, which results in dissociation of Tob from the Orb2-Tob complex and destabilization. The destabilization of Orb2-associated Tob provides a temporal restriction to the Orb2 oligomerization process. The coincident inactivation of PP2A and activation of LimK may also provide a mechanism for stimulus specificity and synaptic restriction.
We find that Orb2A and Orb2B are phosphorylated at multiple sites, including serine/threonine and presumably tyrosine residues. These phosphorylation events are likely mediated by multiple kinases because overexpression of LimK did not affect Orb2 phosphorylation to the extent observed with the inhibition or activation of PP2A, raising several interesting questions. In what order do these phosphorylations occur? What function do they serve individually and in combination? What kinases are involved? Moreover, similar to mammalian CPEB family members, in addition to changing stability, phosphorylation may also influence the function of Orb2A and Orb2B.
Does Tob regulate Orb2 function? In mammals Tob has been shown to recruit Caf1 to CPEB3 target mRNA, resulting in deadenylation [16], and CPEB3 is known to act as a translation repressor when ectopically expressed. We find Drosophila Tob also interacts with Pop2/Caf1 (Figure S3E) [25] and Orb2A and Orb2B can repress translation of some mRNA [52]. Orb2 has also been identified as a modifier of Fragile-X Mental Retardation Protein (FMRP)–dependent translation, and Fragile-X is believed to act in translation repression [53]. Therefore, the Tob-Orb2 association may contribute to Orb2-dependent translation repression, and the degradation of Orb2-associated Tob may relieve translation repression. Additionally, if the oligomeric Orb2 has an altered affinity for either mRNA or other translation regulators, Tob can affect Orb2 function by inducing oligomerization. However, the relationship between Tob phosphorylation and its function is unclear at this point.
Does involvement of Tob both in transcription and translation serve a specific purpose in the nervous system? Tob inhibits BMP-mediated activation of the Smad-family transcription activators (Smad 1/5/8) by promoting association of inhibitory Smads (Smad 6/7) with the activated receptor [18],[24],[54]–[56]. In Drosophila BMP induces synaptic growth via activation of the Smad-family of transcriptional activators, and subsequent stabilization of these newly formed synapses via activation of LimK [57]–[60]. Our studies suggest Tob and LimK also regulate Orb2-dependent translation, raising the possibility Tob may coordinate transcriptional activation in the cell body to translational regulation in the synapse.
Please see Text S1 for a detail description of the proteomic analysis.
The Orb2 lines have been previously described [5],[52]. The following Drosophila strains were obtained from Bloomington Stock Center: mtsXE-2258 (Stock 5684), Pp2A-29BEP2332 (Stock 17044), P{EPgy2}LIMK1EY08757(Stock 17491), UAS-LimK1HA (Stock 9116), and UAS-LimK1 Kinase dead (Stock 9118). The TobRNAi (Stock 38299) on the second chromosome was obtained from Bloomington TRiP collection. The Gal4 lines were generously provided by Douglas Armstrong (c547-Gal4, c747-Gal4) [61], Troy Zars (MB247, 201Y) [48], and Haig Keshishian (elav-GeneSwitch) [62]. The c547 drives expression primarily in the ellipsoid body, c747, MB247 in all lobes of the mushroom body and 201Y primarily in the γ-lobe of the mushroom body. The elav-GeneSwitch drives expression pan-neuronally in an inducible manner. The UAS-dTrpA1 line was generously provided by Paul Garrity [29]. For expression using the GeneSwitch system, the flies were starved for 16–18 h and then transferred to 2% sucrose containing 200 µM RU486 (mifepristone, SigmaM8046) for 12 h. Various genetic combinations were made by standard genetic crosses.
Orb2AHA and Orb2BHA constructs were previously described [5]. The untagged Orb2 and Orb2-interacting protein constructs were made by cloning the full-length PCR products into TopoDonor vector (Invitrogen) and were subsequently transferred to p AWF using the Gateway cloning system (Invitrogen). The Drosophila Tob cDNA was amplified by RT-PCR and cloned with Topo-TA (Invitrogen). Flag-tagged Tob was created by the subsequent transfer to the mammalian expression vector, pCMV24 (Invitrogen). Standard molecular techniques were then used to subclone into pMT (Invitrogen) for S2 cell expression and pUAST (DGRC) for use as a Drosophila transgene. To create TobΔ28, containing an internal deletion of 28 amino acids (AA235–262), the amino terminal region and C-terminal regions were amplified separately and engineered to contain an internal NotI site. The two fragments (EcoRI/NotI and NotI/SalI) were cloned into pCMV24C. Standard techniques were then used to subclone into pMT and pUAST. For the imaging studies, the tdTomato cDNA was inserted in frame to the C-terminal to create pUAST-TobTdTom and pUAST-TobΔ28TdTom. For antigen production, the cDNA encoding Tob AA 267–564 were amplified by PCR and cloned into pRSETA (Invitrogen) in frame with the 6XHis tag. The mammalian cDNAs for Tob1, Tob2, Ana, and Btg were amplified by RT-PCR from mouse RNA and cloned with Topo-TA, which was subsequently used to produce pCMV24. For production of recombinant proteins in E. coli, Tob, Tob1, Tob2, and Btg were reamplified using primers designed to produce an in-frame 6XHis tag at the C-terminus and then subcloned into pMal-c2X. A full-length cDNA encoding LimK, LD15137 was obtained from DGRC and amplified by PCR for Topo TA cloning. The insert was subsequently transferred to pAcV5 for S2 cell expression. LimKMT was engineered to mutate D500K by site-directed mutagenesis (Stratagene). All sequences were confirmed against the NCBI sequence prior to use.
The pCasperOrb2AEGFP construct is comprised of a genomic fragment 1446 nucleotides 3′ of the last Orb2B-specific exon and 1338 bp 5′ of the exonic sequence of the neighboring gene and therefore does not contain coding region of any of the Orb2 isoforms except Orb2A. The ∼8.27 Kb genomic fragment was cloned into the SpeI/XhoI site of pCasper4, and EGFP was introduced at the C-terminal end by creating an in-frame SgrA1 site.
Mammalian HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS. Transfections were performed using Lipofectin reagent (Invitrogen). Drosophila S2 cells were maintained in Schneider's medium supplemented with 10% FBS with transfections performed using Effectene reagent (Qiagen). The constructs used are as indicated in the figures. When examining quantitative changes, an empty vector was used to ensure equal quantity of DNA in each transfection. Imagequant software was used to determine densiometric changes, which were subsequently analyzed using Graphpad Prizm software.
For immunoprecipitations from cell culture, 3×105 transfected cells were used for each immunoprecipitation. The expression constructs used are as indicated in the figures. Following transfection (36–48 h), the cells were washed in PBS and lysed in 500 µl of 1% Igepal buffer (50 mM Tris-Cl, 7.5, 150 mM NaCl, 1% NP-40 [Igpal], 1 mM DTT, EDTA free protease inhibitor) and clarified by centrifugation at 14,000 rpm for 10 min. For immunoprecipitations from flies, adult heads were collected following flash freezing and vortexing, lysed in 1% Igepal buffer, and clarified by two rounds of centrifugation at 14,000 rpm for 10 min. Protein concentration was determined using a BCA kit (Pierce Biotechnology), and between 1–4 mg of head lysate were used for each immunoprecipitation. The following antibodies were used for immunoprecipitation: anti-HA agarose (Sigma), anti-Flag agarose (Sigma), anti-Tob antibody (raised in guinea pig 2163), and anti-Orb2 (raised in guinea pig-2233 and rabbit-273,402) in conjunction with Protein-A agarose (Repligen). The anti-Tob antibody was raised in guinea pig against the C-terminal end of Tob (Pocono Rabbit Farm), purified using Melon resin (Pierce Biotechnology), and used at 1∶100 dilution. Immunoprecipitations performed using S2 cells were incubated for 2 h at 4°C with continuous rocking, and immunoprecipitations performed using head lysates were incubated for 2 h, and then the ProteinA agarose beads were added with additional 2 h incubation. Following four washes, samples were boiled for 5 min in SDS-PAGE gel loading buffer containing 10% SDS and 2 mM freshly prepared DTT. For immunoprecipitation of Orb2 ∼1 mg of total protein and for Orb2AEGFP ∼3 mg of total protein were used. Western analysis was performed following standard protocols. The following antibodies were used for Western analysis: anti-Flag-HRP (Sigma, 1∶1,000), anti-HA-HRP (Roche, 1∶500), anti-Tob (guinea pig, 1∶1,000), anti-Orb2 (rabbit, 1∶2,000), anti-Orb2 (guinea pig, 1∶1,000), anti-Orb2B (rabbit, 1∶1,000), and anti-EGFP (MBL, 1∶1,000).
To examine endogenous Tob expression in wild-type CantonS flies and c547-Gal4::UAS-Orb2AEGFP and c547-Gal4::UAS-Orb2BEGFP flies, the proboscis was removed and the flies were decapitated. The heads were fixed for 2 h at 4°C in 4% paraformaldehyde (PFA)/PBS, incubated overnight in 20% sucrose/PBS, followed by 2 h in a 30∶70 mixture of 20% sucrose/PBS and OCT embedding media (Tissue-Tek). The heads were then embedded in 100% OCT, and frontal cryosections were made of 12 µm. The sections were permeabilized in 1% TritonX containing PBS for 5 min followed by 10 min in 0.1% TritonX containing PBS (PBST). The slides were blocked in 10% goat serum containing PBST for 1 h, followed by overnight incubation in 1∶50 dilution of melon-purified (Pierce Biotechnology) anti-Tob (2163) antibody. For the CantonS flies, 1∶50 dilution of nc82 (Developmental Studies Hybridoma Bank) was also added to mark the synaptic regions. Anti–guinea pig Alexa-Fluor 633 (Invitrogen) secondary antibody was used for Tob detection, and anti-mouse Alexa Fluor 488 (Invitrogen) was used for nc-82 detection. Images were acquired at 512×512 pixels with a Zeiss LSM 5.0 confocal microscope as 1 µm Z-stacks. Images shown are projections of 10 slices.
To examine changes in aggregate number in the adult Orb2EGFP flies, the whole brain was dissected to remove the exoskeleton and air sacs. The brain was fixed in 4% PFA/PBS for 30 min at room temperature, washed three times with PBST for 10 min, and then the whole brain was mounted. Expression of Orb2EGFP and TobTdTom was driven using the ellipsoid body-specific driver, c547. Images were acquired as above. To quantitate the changes in aggregate number, projections of 20 slices were made for each image centering on the central structure of the ellipsoid body. To examine changes in aggregate number in Lim kinase and Orb2EGFP-expressing animals, third instar larvae were filleted and fixed in 4% PFA/PBS for 10 min at room temperature, washed three times with PBST for 10 min, and mounted. Images of the neurites extending from the ventral ganglia were acquired as described. Projections of 10 slices were made.
Axiovision software (Zeiss, v.4.7.1) was used to quantitate total area, aggregate number, and aggregate size. A commander script was written to identify the region of interest and the puncta within the region. All measurement parameters were kept constant for each image.
pMT∶FlagTob by itself or in conjunction with pMT∶Orb2AHA or pMT∶Orb2BHA was transfected into S2 cells. Expression Tob and Orb2 were induced by adding 700 µM CuSO4. Following 16 h, the cells were washed and incubated with 50 µg/ml cycloheximide. At the indicated times, samples were collected and later analyzed by Western blot using either anti-Flag or anti-HA antibodies. Densitometric measurements were carried out using ImageQuant and plotted (percent remaining of time zero versus time) using Prism Graphpad 5. The decay curve was fitted using first-order kinetics. To determine the half-life of hyperphosphorylated Tob, a similar analysis was performed with the cells being treated with both cycloheximide and calyculin.
To examine Tob phosphorylation, amylose-bound MBP-tagged proteins were incubated with 5 ng of recombinant LimK (Upstate Biotechnology) and 10 µCi of [γ–32P]ATP for 20 min at 30°C with semiconstant shaking. Control reactions were performed identically but in the absence of LimK. Kinase dilution buffer and reaction buffer were prepared according to the manufacturer's specifications. Following phosphorylation, the proteins were washed four times in PBS with 0.1% TritonX and once with PBS prior to loading an 8% SDS/PAGE. Following electrophoresis, the gel was dried and exposed from 4 h to overnight. To examine phosphorylation of recombinant Orb2B, His-tagged Orb2B was expressed in E. coli BL21(DE3) using a slow induction protocol, and a low amount of soluble protein was purified in Ni+2 column. Approximately 10 ng of Orb2B, MBP-tagged Tob was used in the kinase reaction.
To examine phosphorylation of the Orb2-Tob complex, 6×105 S2 cells were transfected with pAct∶Orb2AHA or pAct∶Orb2B individually and in combination with pMT∶Tob. The cells were lysed in 1% Igepal buffer (50 mM Tris-Cl, 7.5, 150 mM NaCl, 1% NP-40 [Igpal], 1 mM DTT, EDTA free protease inhibitor) and incubated for 15 min with 50 U/ml CIP. Following centrifugation at 14,000 rpm for 10 min, the supernatant was incubated for 2 h at 4°C with anti-HA agarose. The immunoprecipitates were washed twice with 1% Igepal buffer and once with a modified RIPA buffer (50 mM Tris, 300 mM NaCl, 0.1% SDS, 1% Igepal). The sample was then split into thirds, with one-third examined by Western blot to ensure equality in protein levels and the other two-thirds used for the in vitro kinase assay described above. For the Tob alone samples, 12×105 S2 cells were transfected with pAcOrb2AHA and pMTTob, and the complex was purified as above and dissociated in 1% Igepal buffer containing 1 M NaCl for 15 min at room temperature. The eluate was then normalized to 150 M NaCl and Tob purified by precipitation with anti-Flag agarose (Sigma). Complete dissociation was ensured by Western analysis.
The protein abundance studies were carried out in 4%–12% Bis-Tris SDS-PAGE (Invitrogen) and run in MES-SDS (50 mM MES, 50 mM Tris Base, 0.1% SDS, 1 mM EDTA, pH 7.3) buffer. In these buffer conditions and in the gradient gel, the phosphorylated bands migrate close to each other, which simplifies the quantification of band intensity. Also, in protein abundance studies, the total cell lysates were prepared, unless mentioned, in the absence of phosphatase inhibitors, again to ensure quantification of the total protein accurately. The 4%–12% gels were also used for the detection oligomeric Orb2 and phospho-tag™ blotting of Orb2- or Tob immunoprecipitate from the adult fly head.
To measure phosphorylation status via mobility shift, we found that an 8% SDS-PAGE run in conventional Tris-Glycine buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, pH 8.6) is more effective, and in 8% gel the different phosphorylated forms of Orb2 and Tob proteins were better separated. For detection of the phosphoproteins via phospho-tag™, both 8% and 4%–12% SDS-PAGE were used.
Flies were maintained using standard fly husbandry methods. For behavioral analysis, flies were maintained on standard cornmeal food at 25°C and 60% relative humidity on a 12 h/12 h light-dark cycle. Virgin males and females were collected at eclosion under CO2 anesthesia. Males were isolated and placed in individual food vials. All flies were aged for 5 d before behavioral training and testing. To increase the efficiency of RNAi, flies were shifted to 30°C for 3 d before training. The control flies were treated similarly. Canton S females (4 d old) were mated the night before they were used in training. Males were assayed for courtship conditioning using a modified version of the spaced training described by McBride et al. (1999) [63]. For spaced training, individual males were placed in individual small food tubes (16×100 mm culture tubes, VWR) with a mated female for 2 h. The female was removed, and the male was left alone for 30 min. A different mated female was placed in the tube with the male for another 2 h. The female was removed and the male again rested for another 30 min. A third mated female was introduced in the tube for 2 h and removed at the end of the trial. Control males were treated exactly the same way, except no mated females were introduced into the tube. Memory was tested 5 min, 24 h, and 48 h after training. All tests were performed in a 1 cm courtship chamber. Fresh mated females were used for all time points. All memory tests were recorded (for 10 min) and analyzed using a customized software. The courtship index of each male was obtained by manual and/or automatic analysis of the movies by an experimenter blind to the genotype and experimental conditions.
|
10.1371/journal.pgen.1000841 | Genome-Wide Association Study in Asian Populations Identifies Variants in ETS1 and WDFY4 Associated with Systemic Lupus Erythematosus | Systemic lupus erythematosus is a complex and potentially fatal autoimmune disease, characterized by autoantibody production and multi-organ damage. By a genome-wide association study (320 patients and 1,500 controls) and subsequent replication altogether involving a total of 3,300 Asian SLE patients from Hong Kong, Mainland China, and Thailand, as well as 4,200 ethnically and geographically matched controls, genetic variants in ETS1 and WDFY4 were found to be associated with SLE (ETS1: rs1128334, P = 2.33×10−11, OR = 1.29; WDFY4: rs7097397, P = 8.15×10−12, OR = 1.30). ETS1 encodes for a transcription factor known to be involved in a wide range of immune functions, including Th17 cell development and terminal differentiation of B lymphocytes. SNP rs1128334 is located in the 3′-UTR of ETS1, and allelic expression analysis from peripheral blood mononuclear cells showed significantly lower expression level from the risk allele. WDFY4 is a conserved protein with unknown function, but is predominantly expressed in primary and secondary immune tissues, and rs7097397 in WDFY4 changes an arginine residue to glutamine (R1816Q) in this protein. Our study also confirmed association of the HLA locus, STAT4, TNFSF4, BLK, BANK1, IRF5, and TNFAIP3 with SLE in Asians. These new genetic findings may help us to gain a better understanding of the disease and the functions of the genes involved.
| In this study, we first conducted a genome-wide association study in a Hong Kong Chinese population, followed by replication in three other cohorts from Mainland China and a cohort from Thailand, which totaled 3,300 Asian patients and 4,200 ethnically and geographically matched controls. We identified novel variants in ETS1 and WDFY4 associated with SLE with genome-wide significance and confirmed the association of HLA locus, STAT4, BLK, IRF5, BANK1, TNFSF, and IRF5 with the disease. ETS1 encodes a critical transcription factor involved in Th17 and B cell development. Allelic expression study showed a significantly lower expression of ETS1 from the risk allele, which provided functional support to the genetic findings. WDFY4 is a huge protein with unknown function but is predominantly expressed in primary and secondary immune tissues, and a nonsynonymous SNP in this gene was found to be highly associated with SLE susceptibility. Our findings shed new light on the function of these genes as well as the mechanism of this devastating disease.
| Systemic lupus erythematosus (SLE) is a prototype autoimmune disease characterized by auto-antibody production and multi-organ damage. Genetic factors are known to play an important role in the disease, with the monozygotic twin concordance rate between 20–59, and the risk for siblings of affected individuals 30 times higher than that for the general population [1]–[3]. There are also population differences for the disease both in terms of genetic susceptibility and disease manifestations. African Americans, Hispanics and Asians all have higher disease prevalence than Caucasians; with Asians known to have more lupus nephritis than patients of European ancestry [4],[5].
Genome-Wide Association studies (GWAS) have dramatically changed the landscape of SLE genetics, with a pace of discovery the field has never seen before. In less than two years time, STAT4 [6], ITGAM [7]–[9], BLK [8], PXK and KIAA1542 [7], BANK1 [10] and TNFAIP3 [11],[12] and several other genes have been identified as associated with SLE [13]–[16]. More susceptibility loci were reported recently in two other GWAS on this disease [17],[18].
Despite varied disease prevalence and severity across different populations, it is noteworthy that most previous studies were conducted on patients of European ancestry with under-representation of other ethnicities. We previously examined some of the GWAS findings from populations of European origin [19]. In addition to confirming the association of STAT4 and BLK with SLE in our population, our data indicated differences between the Asian and Caucasian populations. For example, our study did not detect any significant disease association for PXK, a result that was later confirmed by an independent study on a Korean population [20]. Data from our study indicated that, although the risk alleles in ITGAM are rare in Asians (<2%), they are risk factors in our populations and are closely related to lupus nephritis in particular [21].
In this study, we first genotyped 320 SLE patients collected in Hong Kong by the Illumina 610-Quad Beadchip and analyzed the data against 1500 control individuals genotyped on the same platform. Selected SNPs were then replicated in four independent sample collections from Hong Kong, Shanghai and Anhui, (China), as well as Bangkok (Thailand). Genetic variants in and around two genes, ETS1 and WDFY4, were identified as associated with SLE with genome-wide significance. Functional characterization of the risk alleles also supported potential roles of these genetic variants in disease pathogenesis.
The whole-genome genotyping data was thoroughly examined by quality control measures and by population substructure analysis. Analysis of principal component using Eigenstrat [22] did reveal that the samples collected in Hong Kong clustered together, suggesting that confounding population substructure or admixture is not a major concern if Hong Kong controls were used in association analyses. It did indicate, however, that samples collected in Taiwan (obtained from deCODE Genetics) and Han Chinese in Beijing (HCB, available from HapMap) cluster very differently from Hong Kong samples, suggesting population substructure among Chinese living in different geographical regions and potential pitfalls in association studies when cases and controls are not well-matched (Figure 1).
Genome-wide association analysis confirmed significant association of some established susceptibility genes in our population, including SNPs in the HLA locus, STAT4, TNFSF4, BANK1, TNFAIP3, IRF5 and BLK (Table 1 and Figure 2). Similar to the Caucasian findings [12],[23], the risk allele for rs2230926 in TNFAIP3 is low in frequency but with a relatively large effect size in disease association. SNP rs9271366 located between HLA-DRB1 and HLA-DQA1 is the most significantly associated SNP in the whole genome in our study. Although the HLA locus has been consistently shown to be the locus conferring the largest effect size with SLE association, there is little overlap between previous GWAS findings from populations of European ancestry and our results. The most significant SNPs in the Caucasian data [7],[8] are either monomorphic in our population or are not associated with disease susceptibility based on our GWAS result, something worth further pursuit in future studies. Association of STAT4, BLK, IRF5, BANK1 and TNFSF4 with the disease has been reported in our population previously [19],[21],[24],[25].
To answer the question on whether there are still other genetic variants contributing to disease susceptibility, we reexamined the Q-Q plot comparing expected and observed P values by removing all the SNPs in the known susceptibility loci mentioned above. After removal of these SNPs, we still observed an excess of association signal (Figure 3), suggesting involvement of additional susceptibility loci for this disease. Since our GWAS involved a limited number of patients, and is therefore prone to false positive and false negative findings, we selected SNPs for replication based on both their significance in GWAS results as well as the function and expression pattern of the nearby genes. Selected SNPs were replicated first using Sequenom genotyping on limited number of additional samples (360 cases and 360 controls), and variants with significant association in the Sequenom data were then examined by TaqMan genotyping on a much expanded sample collection from four independent cohorts. SNPs in and around two genes, v-ets erythroblastosis virus E26 oncogene homolog 1 (avian) (ETS1) and the WDFY family member 4 (WDFY4) regions were chosen based both on initial GWAS data as well as their known function (in the case of ETS1) and expression pattern (in the case of WDFY4). Table 2 displays the SNPs in these two loci that showed disease association with a P value<0.01 from our GWAS data. Initial replication by Sequenom showed consistent results with the GWAS trend for these two genes and they were further tested in the remaining samples.
Making use of the remaining samples from Hong Kong not included in GWAS, and sample collections from Shanghai and Anhui, China, and Bangkok, Thailand, we went on to replicate the whole-genome findings on these two loci. SNPs in ETS1 listed in Table 2, rs12223943, rs7932088 and rs10893872, were examined in the expanded samples. SNPs rs10893872 and rs4937333 have absolute LD with each other, so only rs10893872 was chosen for replication. SNP rs1128334 was chosen in the place of rs6590330 for replication due to its high LD with rs6590330 (r2 = 0.97, Figure 4) and its relative position to the gene (3′-UTR) and predicted effect on microRNA binding. For SNPs rs12223943 and rs7932088, the association seen from whole-genome data was inconclusive (data not shown) in the replication stage and were not further pursued. We genotyped rs10893872 and rs1128334 in all the samples from the four cohorts and both were found to be highly associated with SLE (Table 3).
Independence test by logistic regression by Plink pointed to a major contribution from rs1128334. And indeed, the sequence around rs1128334 has high sequence conservation among different species (Figure 4). Haplotype analysis indicated that the TA haplotype formed by the two SNPs (rs10893872 and rs1128334) is the major risk haplotype, whilst the CG haplotype is the major protective haplotype (Table 4) with other haplotypes having low allele frequencies. Subphenotype analysis was also performed for these SNPs, and both SNPs were found to have larger effect sizes for patients with lupus nephritis in all four cohorts, although no statistical significance was reached in any cases. Analysis of other subphenotypes showed insignificant, or inconsistent results among different cohorts.
Since rs1128334 is located at the 3′-UTR region of the gene, it is predicted that it may have an effect on the expression level of ETS1. Therefore, we examined allelic expression of ETS1 gene for the two alleles of rs1128334, “A” and “G”, from healthy individuals heterozygous for this SNP (N = 33). This assay assesses directly whether the two alleles of the SNP correlate with different steady state mRNA levels. Pyrosequencing results from PBMC of healthy individuals heterozygous on the SNP showed a significantly higher expression from the “G” allele than from the risk “A” allele, with a P value<0.0001 (Figure 5).
Two of the SNPs in WDFY4 that showed the most significant association with the disease in our GWAS, SNPs rs10857650 and rs877819, were selected for further replication. In addition, three nonsynonymous SNPs in this gene not genotype by the Illumina 610-Quad Beadchip, rs2170132 (Ser1528Pro), rs7097397 (Arg1816Gln) and rs2292584 (Pro3118Leu), were also selected to test for disease association, aiming at identifying functional variants in this gene. SNPs rs2170132 and rs2292584 did not show significant difference between the cases and the controls in the Hong Kong cohort (Table 5) and Thai samples (data not shown) and were not further tested in other cohorts. Genotyping results on rs10857650 using TaqMan showed significant discordance with results from the Illumina Beadchip, and was thus removed from further analysis.
SNP rs7097397 and rs877819 were confirmed to have significant association with the disease (Table 3). Conditional logistic regression test indicate that the nonsynonymous SNP coded by rs7097397 is probably the functional variant, with P = 1.01×10−5 when controlling the effect of rs877819. Independent contribution from rs877819 is questionable, with a P value of 0.088 considering the effect of rs7097397 in the same test. The two SNPs have intermediate LD (r2 = 0.44, Figure 6A). The genetic result is consistent with the fact that the arginine residue at 1816 in WDFY4 protein is well conserved among orthologs in different mammals (Figure 6B). Preliminary analysis on subphenotype stratification suggests that this nonsynonymous change may be more closely involved in male and early onset patients in a case only analysis (onset age 12 and below vs. above, OR = 1.96, P = 0.0021; male vs. female: OR = 1.52, P = 0.023).
Association of the SNPs in the two genes with disease risk was also corrected by logistic regression using age and sex as covariates, and the associations found in this study all remain highly significant after all the corrections.
Several GWAS on SLE have been conducted on populations of European ancestry [7],[8], but populations of Asian or African ancestry were seriously underrepresented. Only during the process of submission of our current work, a GWAS study on Chinese populations was reported [18]. Considering the population differences in both disease prevalence and clinical manifestations, GWAS on non-Caucasian populations may have novel findings and help to elucidate the differences between populations.
An interesting analysis result from our GWAS data is the difference between Hong Kong samples and samples collected in Taiwan and Beijing, shown by principal component analysis (Figure 1). It suggests population substructure for Chinese living in different regions, which may cause spurious findings in association studies when cases and controls are not well matched. With most of the genetic variants of relatively larger effect sizes already being identified, GWAS becomes more susceptible to effects from mismatches between cases and controls in dealing with SNPs of smaller effect sizes. Our analysis echoed two very recent reports delineating population substructures in Chinese populations living in different geographical regions [26],[27].
Ets-1 is a member of the ETS family of transcription factors that share a unique Ets DNA binding domain. They control a wide variety of cellular processes including cell proliferation and differentiation [28]. Ets-1-deficient mice develop lupus-like disease characterized by high titers of IgM and IgG autoantibodies, immune complex-mediated glomerulonephritis, and local activation of complement [29]. Ets-1 is also involved in many cellular abnormalities that are known to participate in SLE pathogenesis as illustrated in Figure 7.
Ets-1 is a negative regulator of terminal differentiation of B cells and plays critical roles in maintaining B cell identity [30]–[32]. Ets-1-deficient B cells were present in normal numbers but have a large proportion of IgM plasma cells [33]. Ets-1 blocks the function of B-lymphocyte-induced maturation protein 1 (Blimp-1), an essential transcription factor for plasma cells [34]. The number and frequency of plasma cells were known to correlate with disease activity and the titer of anti-dsDNA antibodies in SLE [35],[36].
Ets-1 is also a negative regulator of Th17 cell differentiation, and naïve CD4+ T cells deficient in Ets-1 undergo greatly enhanced differentiation into Th17 cells when cultured in vitro under Th17-skewing conditions [37]. Th17 cells with specificity for self-antigens are known to be highly pathogenic and lead to the development of inflammation and severe autoimmunity [38]. Higher plasma IL-17, IL-23 and higher number of Th17 cells in SLE patients were reported and correlated positively with SLE disease index (SLEDAI) [39]–[41]. IL-17 and IL-21 produced by Th17 cells may also induce B cell terminal differentiation [42]–[44].
In this study, we found that in the PBMC of healthy individuals, expression of ETS1 from the risk “A” allele is reduced comparing to that from the “G” allele. The expression level of ETS1 may be tightly regulated. It was shown that resting T cells express high levels of ETS1 mRNA and protein, which decreased to very low levels upon T cell activation [45]. Lower expression of ETS1 for the risk allele carriers may play a role in disease pathogenesis through increased differentiation and activity of both plasma cells and Th17 cells.
It is likely that the association SNPs identified in this study may affect the response of ETS1 gene to other upstream signals. SNP rs1128334 locates in the 3′-UTR of ETS1, and rs10893872 is in absolute LD with rs4937333, another SNP that is also located in the 3′-UTR of the gene and both SNPs are on putative microRNA (miRNA) binding sites. In a recent study by Du et al, the expression level of a microRNA, miR-326, was found to be related to disease severity in patients with multiple sclerosis and mice with experimental autoimmune encephalomyelitis. ETS1 was shown to be the major target of miR-326, through downregulation of which miR-326 promoted the generation of Th17 cells both in vitro and in vivo [46]. Another microRNA, miRNA-146a, was also found to be involved in SLE pathogenesis [47].
WDFY family member 4 (WDFY4, NCBI GeneID: 57705) codes for a huge protein (3184 amino acid residues) with unknown function. Its closest paralog is WD repeat and FYVE domain containing 3 (WDFY3, NCBI GeneID: 23001). Similar to WDFY3, WDFY4 does contain WD40 domains and a BEACH (Beige and chediak-kigashi) domain, although FYVE zinc finger domain is truncated. Its sequence is well conserved among various species. For example, there is an 84% sequence similarity between the human protein and its orthologs from Bos Taurus and Canis lupus familiaris. And there is a 42% identity between the human protein and protein XP_701288.3 in Danio Rerio (zebra fish). Although protein XP_701288.3 is annotated as WDFY3 in NCBI, it has a higher sequence similarity with human WDFY4 than with WDFY3.
WD40 domain is found in a number of eukaryotic proteins that cover a wide variety of functions including adaptor/regulatory modules in signal transduction, pre-mRNA processing and cytoskeleton assembly, while the BEACH domains are implicated in membrane trafficking. While very little is known about the potential function of this well conserved protein, an interesting phenomenon is that the gene is predominantly expressed in immune tissues such as lymph node, spleen, thymus and tonsil (UniGene Hs.287379, http://www.ncbi.nlm.nih.gov/UniGene/ESTProfileViewer.cgi?uglist=Hs.287379), unlike WDFY3, which is expressed in a wide variety of tissues and organs.
After submission of our work, Han et al reported their GWAS work on SLE on Chinese populations [18] and reported genome-wide significant association signals on ETS1 (rs6590330, downstream of the gene) and WDFY4 (rs1913517, in the intron of both WDFY4 and LRRC18, leucine rich repeat containing 18). Independent identification of these two genes in SLE susceptibility underlined the validity of the findings. Identifying susceptibility genes provided a good start in our effort to elucidate the disease mechanisms and the functions of the genes involved, although many questions remain to be answered. Characterization of the functions of the genes in autoimmunity may eventually help us to translate genetic findings into clinical and therapeutic applications.
This study was conducted according to the principles expressed in the Declaration of Helsinki. The Hong Kong study was approved by the Institutional Review Board of the University of Hong Kong and Hospital Authority, Hong Kong West Cluster, New Territory West Cluster, and Hong Kong East Cluster. The study on Shanghai, Anhui and Thai samples was approved by the Institutional Review Board of Renji Hospital, Research Ethics Committee of Anhui Medical University and the Ethics Committee of the Faculty of Medicine, Chulalongkorn University, respectively. All patients provided written informed consent for the collection of samples and subsequent analysis.
1073 SLE samples collected in Hong Kong were from four hospitals in Hong Kong Island and the New Territories: Queen Mary Hospital, Tuen Mun Hospital, Queen Elizabeth Hospital and Pamela Youde Nethersole Eastern Hospital. The patients were all of self-reported Chinese ethnicity living in Hong Kong. The average onset age was 28 years old and the ratio of female to male patients was 9∶1. About half of the patients had renal involvement, and about 70% tested positive for anti-dsDNA antibodies. The SLE samples collected in Shanghai were patients attending Renji Hospital of Jiaotong University Medical School, a tertiary referral hospital covering Shanghai and the surrounding areas. There is an 8∶1 female to male patient ratio, and about 52% of patients have lupus nephritis. 951 SLE patients collected in Anhui were all self-reported Chinese ethnicity living in Anhui province, central China. They were recruited from the Departments of Rheumatology at Anhui Provincial Hospital and the First Affiliated Hospital of Anhui Medical University, both located in Hefei, Anhui province, about 450 km from Shanghai. The average onset age was 31 years old and the ratio of female to male patients was 17∶1. 314 Thai patients with SLE (female∶male ratio = 14∶1) attending King Chulalongkorn Memorial Hospital, a tertiary referral center in Bangkok were also recruited in this study. Medical records were reviewed to confirm that all subjects met the revised criteria of the American College of Rheumatology for SLE diagnoses [48].
Controls used in the GWAS stage were from both healthy individuals and from other studies conducted in the same institution genotyped with the same platform. For the replication stage, Hong Kong controls were healthy blood donors kindly contributed by the Hong Kong Red Cross and were all of self-reported Chinese ethnicity living in Hong Kong. Controls from Shanghai and Anhui were selected from a pool of healthy blood donors recruited from Renji Hospital (Shanghai) and Hefei City (Anhui), respectively, with an effort to match for the age and sex of corresponding SLE patients. Thai controls were recruited from unrelated voluntary healthy donors from the same ethnic background and geographic area as the Thai SLE patients.
320 (27 males, 293 females) SLE patients were genotyped by Illumina 610-Quad Human Beadchip with a total number of SNPs reaching 620,901. 24 individuals were removed due to low call rate or hidden first degree relationship. A total number of 104,395 SNPs were also removed from the initial analysis on the ground of low genotyping call rate (<90%, 30,133 SNPs) and/or low minor allele frequency (MAF) (<0.005, 102,970 SNPs). 2,285 SNPs were also removed due to violation of Hardy-Weinberg equilibrium in controls (P values<0.0001). After quality control measures, 314 case (27 males, 287 females) and 1484 controls (840 males and 644 females) were analyzed on a total of 514,221 SNPs. The call rate for the remaining SNPs reached 0.999 with a genome-wide inflation factor of 1.03, which is an indication of good match between the cases and controls. To overcome any potential effect from the heterogeneity in the controls, three independent comparisons were initially conducted by separating the controls into three different subsets and only SNPs reaching significance in all subsets in association analysis were selected for further replication.
The SNPs were analyzed for association with the disease by means of comparison of the minor allele frequency in patients and controls (basic allelic test) as well as other tests using Plink [49]. Linkage disequilibrium (LD) patterns were analyzed and displayed by HaploView [50]. Association of the SNPs with disease risk was also corrected by logistic regression using age and sex as covariates. Average odds ratios (OR) and P values jointly analyzed from four sample collections were obtained by Cochran-Mantel-Haenszel (CMH) test of disease association conditional on SNP frequency differences among different populations. Test of independent contributions of a SNP controlling for the effect of other SNPs in the same locus was done by conditional logistic regression as well as haplotype analyses. Subphenotype stratification was performed by comparing cases with and without a given subphenotype.
SNPs rs1128334, rs10893872, rs7097397, rs10857650, and rs877819 were genotyped by TaqMan SNP genotyping method using assay-on-demand probes and primers (Applied Biosystems, Foster City, CA94404, USA). Some of the initial screening was also done using Sequenom MassARRAY iPLEX Gold system. Genotyping accuracy was confirmed by direct sequencing of PCR products for some randomly chosen samples. Genotyping concordance between Illumina Human 610-Quad Beadchip and TaqMan SNP genotyping method was also examined on selected samples and probes: rs877819 has a concordance rate of 99.64% (1 out of 277 differed by the two platforms); rs10893872 has a concordance rate of 99.16% (1 out of 119 samples differed). SNP rs1128334 and rs7097397 were examined by direct sequencing of selected samples and showed complete consistence between TaqMan and sequencing (50 samples each). The results of rs10857650 were discarded due to low concordance between the Illumina Beadchip data and the TaqMan results.
Thirty-three healthy individuals heterozygous for rs1128334 were chosen to assess the relative ETS1 mRNA levels from the two alleles, “A” and “G”, by pyrosequencing [51],[52]. In the meantime, DNA detection ratio was used as a control for amplification efficiency. Briefly, total RNA was extracted from peripheral blood mononuclear cell (PBMC) from each individual. RNA samples were then treated with DNAase to eliminate genomic DNA contamination before being reverse-transcribed into cDNA using oligo-dT primer. cDNA was then amplified by PCR using transcript-specific primers, together with DNA from the same individuals. The cDNA and DNA PCR products were purified using the Qiaquick PCR purification kit, and then subjected to allele quantitative pyrosequencing. The sequencing primer was designed using Pyrosequencing Assay Design Software v.1.0. Reactions were performed on a Biotage PSQ96MA machine, and allele quantification was analyzed using PSQMA 2.1 software. The average G/A cDNA expression ratio of each individual was normalized by the G/A DNA ratio from the same sample. Paired student' t test was used to compare the normalized expression level from the “A” and “G” alleles.
|
10.1371/journal.pgen.1003641 | SLC26A4 Targeted to the Endolymphatic Sac Rescues Hearing and Balance in Slc26a4 Mutant Mice | Mutations of SLC26A4 are a common cause of human hearing loss associated with enlargement of the vestibular aqueduct. SLC26A4 encodes pendrin, an anion exchanger expressed in a variety of epithelial cells in the cochlea, the vestibular labyrinth and the endolymphatic sac. Slc26a4Δ/Δ mice are devoid of pendrin and develop a severe enlargement of the membranous labyrinth, fail to acquire hearing and balance, and thereby provide a model for the human phenotype. Here, we generated a transgenic mouse line that expresses human SLC26A4 controlled by the promoter of ATP6V1B1. Crossing this transgene into the Slc26a4Δ/Δ line restored protein expression of pendrin in the endolymphatic sac without inducing detectable expression in the cochlea or the vestibular sensory organs. The transgene prevented abnormal enlargement of the membranous labyrinth, restored a normal endocochlear potential, normal pH gradients between endolymph and perilymph in the cochlea, normal otoconia formation in the vestibular labyrinth and normal sensory functions of hearing and balance. Our study demonstrates that restoration of pendrin to the endolymphatic sac is sufficient to restore normal inner ear function. This finding in conjunction with our previous report that pendrin expression is required for embryonic development but not for the maintenance of hearing opens the prospect that a spatially and temporally limited therapy will restore normal hearing in human patients carrying a variety of mutations of SLC26A4.
| Mutations of SLC26A4 are the most common cause for hearing loss associated with a swelling of the inner ear. This human disease is largely recapitulated in a mutant mouse model. Mutant mice lack Slc26a4 expression and their inner ears swell during embryonic development, which leads to failure of the cochlea and the vestibular organs resulting in deafness and loss of balance. SLC26A4 is normally found in the cochlea and vestibular organs of the inner ear as well as in the endolymphatic sac, which is a non-sensory part of the inner ear. The multitude of sites where SLC26A4 is located made the goal to restore function through restoration look futile, unless some sites were more important than others. Here, we generated a new mutant mouse that expresses SLC26A4 in the endolymphatic sac but not in the cochlea or the vestibular organs of the inner ear. Fantastically, this mouse did not develop the detrimental swelling of the inner ear and even more exciting, the mouse developed normal hearing and balance. Our study provides the proof-of-concept that a therapy aimed at repairing the endolymphatic sac during embryonic development is sufficient to restore a life-time of normal hearing and balance.
| Enlargement of the vestibular aqueduct (EVA; OMIM #600791) is a malformation of the temporal bone that is commonly observed in children with sensorineural hearing loss [1], [2], [3], [4], [5]. Mutations of SLC26A4 are the most common cause for EVA-associated hearing loss that can either be non-syndromic (DFNB4; OMIM # 600791) or syndromic with enlargement of the thyroid gland (Pendred syndrome; OMIM #274600). SLC26A4 codes for the anion exchanger pendrin that transports anions such as Cl−, I− and HCO3− [6], [7]. Although EVA is a malformation of the temporal bone, it is not the cause for hearing loss since no correlation was found between the degree of EVA and the severity of hearing impairment [8]. EVA, however, is an indication of an enlargement of the endolymphatic duct epithelium that was present during embryonic development. Cartilage cells that form in the periphery of the endolymphatic duct epithelium preserve the diameter of the duct in a ‘fossil-like’ record when they give rise to the bone of the vestibular aqueduct.
The mature inner ear consists of seven interconnected fluid spaces that house six sensory organs (Fig. 1): The cochlea for hearing, the utricle and saccule for sensing linear acceleration including gravity, and three ampullae with semicircular canals for sensing angular acceleration in three spatial axes. The seventh fluid compartment is the endolymphatic duct and sac, which is devoid of sensory cells and which is suspected to play a role in fluid homeostasis [9], [10]. Pendrin is expressed in a variety of epithelial cells that enclose endolymph, which is the luminal fluid of the inner ear (Fig. 1). Pendrin is expressed in outer sulcus, spiral prominence and spindle-shaped cells in the cochlea, transitional cells in the utricle, saccule and ampullae and mitochondria-rich cells (synonym: Forkhead-related or FORE cells) of the endolymphatic sac [11], [12]. Each cell type represents a small domain in the heterogeneous epithelium that encloses endolymph. The many locations and cell types that express pendrin in a normal inner ear made the goal to restore function through restoration of expression look futile unless some sites of expression were more important than others.
The earliest onset of pendrin expression in the murine inner ear occurs in the endolymphatic sac at embryonic day E11.5, which precedes the onset of expression in the cochlea by 3 days, in the saccule and utricle by 4 days, and in the ampullae by 5 days [13]. The expression in the endolymphatic sac surges dramatically at E14.5, a time in development when there is very little pendrin expressed elsewhere in the inner ear [13].
Studies in a mouse model, Slc26a4Δ/Δ, have revealed that loss of pendrin leads to an enlargement of endolymph volume followed by an acidification and a failure to develop normal hearing and balance [13], [14]. The onset of the enlargement in the cochlea and the endolymphatic sac occurs at E14.5 which precedes the onset of the luminal acidification by 1 day in the cochlea and by 3 days in the endolymphatic sac [13]. The enlargement develops in Slc26a4Δ/Δ mice between E14.5 and E18.5, which is the phase of rapid growth of the cochlea [4]. The coincidence of the surge in pendrin expression in the endolymphatic sac at E14.5 and the onset of the enlargement in Slc26a4Δ/Δ mice points to the importance of pendrin expression in the endolymphatic sac for inner ear fluid homeostasis.
We hypothesized that restoration of pendrin expression in the endolymphatic sac would prevent enlargement and permit normal development of the cochlea and the vestibular labyrinth including the acquisition of sensory function. To test this hypothesis, we generated a mouse line that expresses human pendrin SLC26A4 controlled by the promoter of the B1-subunit of the human vacuolar H+ ATPase (ATP6V1B1) and crossed this transgene into the Slc26a4Δ/Δ line to generate mice that lack expression of mouse pendrin but express human pendrin in the endolymphatic sac. No expression of pendrin protein was detected in these mice in the cochlea or the vestibular labyrinth but in mitochondria-rich cells of the endolymphatic sac. Analysis of this mouse model revealed normal hearing and balance function. Our data indicate that the expression of pendrin solely in the endolymphatic sac of the inner ear is sufficient to permit the development of normal hearing and balance.
A transgenic mouse line, referred to as Tg(B1-hPDS)Tg/+; Slc26a4+/+ and abbreviated here to Tg(+);Slc26a4+/+, was created by the laboratory of Dr. Dominique Eladari (Paris, France) [15]. This mouse expresses human SLC26A4 (formerly named hPDS) controlled by the promoter of ATP6V1B1, which codes for the B1-subunit of the vH+ATPase. Transgenic founders were crossed with wild-type C57BL/6× CBA F1 mice and three Tg(+);Slc26a4+/+ mice were shipped to Kansas State University (Manhattan, Kansas, USA). At Kansas State University, Tg(+);Slc26a4+/+ mice were crossed with Slc26a4Δ/Δ mice, which are maintained in an isogenic 129S6SvEv background, to generate the desired Tg(+);Slc26a4Δ/Δ mice in an F2 generation. Expression of SLC26A4 (human pendrin) in this mouse was expected to originate solely from the transgene since Exon 8 in the Slc26a4Δ allele was replaced with a neomycin-cassette that introduced a frame-shift [14]. Littermates with the genotype Tg(−);Slc26a4Δ/Δ served as negative controls. These mice were expected to lack functional pendrin protein expression. Further, littermates with genotypes Tg(−);Slc26a4Δ/+, Tg(+);Slc26a4Δ/+, and Tg(+);Slc26a4+/+ served as positive controls. These mice were expected to express murine pendrin with or without augmentation of human pendrin and have normal hearing and balance.
Expression of Atp1a1, Atp6v1b1, Slc26a4 and SLC26A4 was determined by quantitative RT-PCR and normalized to the expression of 18S rRNA (Fig. 2). The highest levels of Atp6v1b1 and Slc26a4 mRNA among the different inner ear tissues were found in the endolymphatic sac (Fig. 2B and C). Expression of Slc26a4 was reduced by factors between 6 and 16 in Tg(+);Slc26a4Δ/Δ mice compared to Tg(−);Slc26a4Δ/+ mice (Fig. 2C vs G). Expression levels of Atp1a1 and Atp6v1b1 exhibited a similar pattern among inner ear tissues of Tg(−);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/Δ mice (Fig. 2A vs E and Fig. 2B vs F). Most interesting, expression levels of human SLC26A4 in Tg(+);Slc26a4Δ/Δ resembled the pattern of mouse Slc26a4 in Tg(−);Slc26a4Δ/+ mice with the highest levels being expressed in the endolymphatic sac (Fig. 2C vs H). Whether or not expression levels of human SLC26A4 in Tg(+);Slc26a4Δ/Δ exceeded expression levels of mouse Slc26a4 in Tg(−);Slc26a4Δ/+ mice remained undetermined, since the efficiency of the reverse transcription of mRNA into cDNA remains generally unknown in quantitative RT-PCR experiments. Taken together, the data demonstrate that the transgene restored pendrin mRNA expression to the endolymphatic sac, the cochlea and the vestibular labyrinth of the inner ear.
The ability of the ATP6V1B1 promoter to drive protein expression in different tissues including the cochlea, the vestibular labyrinth and the endolymphatic sac was evaluated in a transgenic mouse line, Tg(B1-eGFP) in which the expression of eGFP is controlled by the same 6.9 kb promoter of the human ATP6V1B1 gene that drives the expression of human pendrin in Tg(+);Slc26a4Δ/Δ mice [16]. No expression of eGFP was detected in the cochlea or the vestibular labyrinth of E15.5 Tg(B1-eGFP) mice, although expression was present in the endolymphatic sac and the kidney (Fig. 3). These data suggest that the ATP6V1B1 promoter does not drive protein expression in the cochlea or the vestibular labyrinth.
Soft tissues of the cochlea and the vestibular labyrinth, exclusive of the endolymph sac, were collected from adult mice by microdissection and pooled into an ‘inner ear’ sample. Crude membrane protein preparations were obtained from these inner ear samples and from kidneys and subjected to gel-electrophoresis and Western blotting. Membrane proteins were obtained from Tg(+);Slc26a4Δ/Δ mice as well as from Tg(−);Slc26a4Δ/+ mice, which served as positive controls, and from Tg(−);Slc26a4Δ/Δ mice, which served as negative controls. Pendrin was detected in the inner ear and kidney of Tg(−);Slc26a4Δ/+ mice as a ∼110 kDa band (Fig. 4A). Inner ear from Tg(+);Slc26a4Δ/Δ mice lacked this band. The observation that there was no difference in the pattern of faint bands between inner ears from Tg(+);Slc26a4Δ/Δ mice and Tg(−);Slc26a4Δ/Δ mice, which is the negative control, suggests that pendrin was either not detectable or not present. The pendrin band, however, was found in kidney from Tg(+);Slc26a4Δ/Δ mice (Fig. 4B), which suggests that the antibody recognizes both mouse and human pendrin. The observation that pendrin was detected at similar levels in descending amounts of kidney proteins isolated from Tg(−);Slc26a4Δ/+ mice (Fig. 4A) and Tg(+);Slc26a4Δ/Δ mice (Fig. 4C) suggests that the detection threshold for mouse and human pendrin was similar. Whether the antibody differed in the sensitivity between mouse and human pendrin remains unknown, since the relative abundance of mouse pendrin in kidneys of Tg(−);Slc26a4Δ/+ mice and human pendrin in kidneys of Tg(+);Slc26a4Δ/Δ mice is not known.
The meaning of pendrin being not detectable in the inner ear was evaluated by comparison of the intensity of the pendrin band in inner ear to the intensities in descending amounts of kidney protein (Fig. 4A). This comparison suggests that a ∼5-fold lower amount pendrin should have been detectable in the inner ear. This means that pendrin expression in the inner ear of Tg(+);Slc26a4Δ/Δ mice is either absent or expressed at a level that does not exceed 20% of the expression level in Tg(−);Slc26a4Δ/+ mice.
Temporal bones from Tg(+);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/Δ mice were isolated at E15.5, fixed and injected with white paint (Fig. 5A and B). Most striking is that there was no enlargement of the endolymphatic sac, duct or cochlea in Tg(+);Slc26a4Δ/Δ mice and that the morphology of Tg(+);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/Δ mice was grossly similar. These data demonstrate that the introduction of the transgene rescued the malformation previously described in Slc26a4Δ/Δ mice [10], [14].
Whole-mounted specimens of the endolymphatic sac were prepared for immunocytochemistry from Tg(+);Slc26a4Δ/Δ, Tg(−);Slc26a4Δ/Δ, and Tg(+);Slc26a4Δ/+ mice (Fig. 5C–E). Most striking is the enlargement and lack of pendrin expression in the endolymphatic sac of Tg(−);Slc26a4Δ/Δ mice (Fig. 5E) and the similarity in size and similarity in pendrin expression between the endolymphatic sac of Tg(+);Slc26a4Δ/Δ mice (Fig. 5C) and Tg(+);Slc26a4Δ/+ mice (Fig. 5D). These data demonstrate that the transgene drives pendrin expression in the endolymphatic sac and that the introduction of the transgene rescued the malformation [10], [14].
Gross morphological examination of inner ears revealed greater similarity between Tg(+);Slc26a4Δ/Δ mice and Tg(+);Slc26a4Δ/+ than between Tg(+);Slc26a4Δ/Δ mice and Tg(−);Slc26a4Δ/Δ mice (Fig. 6A–C). Cochlear turns in Tg(+);Slc26a4Δ/Δ mice appeared normal in width and did not show widening of turns or thinning of the otic capsule that was seen in Tg(−);Slc26a4Δ/Δ and that was previously described in Slc26a4Δ/Δ mice [12], [17]. Inspection of the oval window revealed ‘glittering’ otoconia in the saccule in Tg(+);Slc26a4Δ/Δ and Tg(+);Slc26a4Δ/+ mice in contrast to Tg(−);Slc26a4Δ/Δ mice where no ‘glittering’ was visible (Fig. 6D–F).
Midmodiolar sections of cochlear tissues were prepared for immunocytochemistry from Tg(+);Slc26a4Δ/Δ mice and positive controls consisting of Tg(−);Slc26a4Δ/+ or Tg(+);Slc26a4Δ/+ mice. No evidence for cochlear enlargement was found in Tg(+);Slc26a4Δ/Δ mice at E16.5 (Fig. 7B), P1 (Fig. S1A), P16 (Fig. 7A; Fig. S1E) or P18 (Fig. S1C) suggesting that the introduction of the transgene rescued the cochlear malformation previously described in Slc26a4Δ/Δ mice, which includes a ∼10-fold enlargement of the cochlea [12], [14]. No detectable pendrin expression was found in the spiral prominence or outer sulcus epithelium of the cochlea in Tg(+);Slc26a4Δ/Δ mice although prominent expression was observed in these cells in positive controls (Fig. 7 and S1). The absence of pendrin in Tg(+);Slc26a4Δ/Δ mice was observed with two different anti-pendrin antibodies (Pds #1 and Pds #2). The patterns of pendrin expression in the positive controls, Tg(−);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/+ mice, were similar for both antibodies to the pattern previously observed in Slc26a4Δ/+ mice [11], [12], [13]. Expression of pendrin was further examined in whole-mounted specimens that encompassed the spiral limbus, organ of Corti and outer sulcus. No detectable pendrin expression was found at age P35 in the spiral limbus of Tg(+);Slc26a4Δ/Δ mice (Fig. 7C) or Tg(−);Slc26a4Δ/+ mice (Fig. 7H) in contrast to the prominent expression of pendrin in the outer sulcus epithelia of Tg(−);Slc26a4Δ/+ mice (Fig. 7J). For completeness, it needs to be reported that some punctate staining was found in nerve terminals near inner hair cells of Tg(+);Slc26a4Δ/Δ (Fig. 7D) and Tg(−);Slc26a4Δ/+ mice (Fig. 7I).
The endocochlear potential and the difference in pH between endolymph and perilymph was measured with double-barreled ion selective electrodes in Tg(−);Slc26a4Δ/Δ mice, Tg(+);Slc26a4Δ/+ mice, and Tg(+);Slc26a4Δ/Δ mice (Fig. 8). Tg(−);Slc26a4Δ/Δ mice failed to develop a normal endocochlear potential and the pH of endolymph was lower ( = more acidic) than in perilymph, as previously reported [18]. In contrast, Tg(+);Slc26a4Δ/+ mice, as reported for Slc26a4Δ/+ mice [18], developed a normal endocochlear potential and a normal endolymphatic pH that was higher ( = more alkaline) than in perilymph. Similar to Tg(+);Slc26a4Δ/+ mice, Tg(+);Slc26a4Δ/Δ mice developed a normal endocochlear potential and a normal endolymphatic pH even though no detectable pendrin expression was observed in the cochlear epithelium. These data demonstrate that the introduction of the transgene, which rescued the malformation, also rescued the loss of the endocochlear potential and the loss of normal endolymphatic pH homeostasis.
Hearing tests were based on auditory brain stem recordings and thresholds in response to tone bursts of 8 kHz, 16 kHz and 32 kHz. Tests performed in Tg(+);Slc26a4+/+, Tg(−);Slc26a4Δ/Δ and Tg(+);Slc26a4Δ/Δ mice confirmed profound deafness in Tg(−);Slc26a4Δ/Δ mice (Fig. 9B) consistent with previous findings in Slc26a4Δ/Δ mice [14], [18]. Waveforms of auditory brain stem recordings as well as thresholds were similar between Tg(+);Slc26a4Δ/Δ mice (Fig. 9A) and Tg(+);Slc26a4Δ/+ (Fig. 9C). These findings demonstrate that the introduction of the transgene rescued normal hearing although the cochlea did not express detectable levels of pendrin. We next evaluated whether the rescued hearing phenotype in Tg(+);Slc26a4Δ/Δ would be stable through at least 3 months of age. Auditory brain stem recordings were performed in Tg(+);Slc26a4Δ/+ mice (Fig. 9D–F), in Tg(+);Slc26a4Δ/Δ mice (Fig. 9J–L), and in Tg(+);Slc26a4+/+ mice (Fig. 9G–I) at 1, 2 and 3 month of age. Hearing in Tg(+);Slc26a4Δ/Δ mice at 8 kHz and 16 kHz was stable through 3 months. Thresholds were very similar among individuals and did not differ from Tg(+);Slc26a4Δ/+ and Tg(+);Slc26a4+/+ mice. A greater variability in hearing thresholds was observed at 32 kHz (Fig. 9L), with 10 of the 19 Tg(+);Slc26a4Δ/Δ mice maintaining excellent hearing (thresholds ≤30 dB at 32 kHz) and 5 developing a high-frequency hearing loss (thresholds ≥60 dB at 32 kHz). About one half of the Tg(+);Slc26a4Δ/Δ mice (9 of 19) developed progressive threshold elevations at 32 kHz with thresholds increasing by ≥10 dB between the monthly measurements. This variability is reflected in the greater error bars at 32 kHz but did not lead to a statistically significant difference between Tg(+);Slc26a4Δ/Δ and Tg(+);Slc26a4+/+ mice.
Sections and whole-mounted specimens of vestibular tissues were prepared for immunocytochemistry from Tg(+);Slc26a4Δ/Δ mice and positive controls consisting of Tg(−);Slc26a4Δ/+ or Tg(+);Slc26a4Δ/+ mice. No evidence of pendrin expression was found in the three sensory organs in Tg(+);Slc26a4Δ/Δ mice at P14 (Fig. S2A), P16 (Fig. 10A,C,D and Fig. S2C,G,K), P18 (S2E), and P35 (Fig. 10B). The absence of pendrin in Tg(+);Slc26a4Δ/Δ mice was observed with two different anti-pendrin antibodies (Pds #1 and Pds #2). In contrast, pendrin expression was found in transitional cells of the utricle, saccule and ampullae of controls that consisted of Tg(−);Slc26a4Δ/+ or Tg(+);Slc26a4Δ/+ mice. The expression patterns in Tg(−);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/+ mice with both antibodies were similar to the pattern previously described in Slc26a4Δ/+ mice [11], [12].
Vestibular labyrinths were isolated by microdissection from Tg(−);Slc26a4Δ/Δ, Tg(+);Slc26a4Δ/Δ and Tg(+);Slc26a4Δ/+ mice and the roof of the utricle was removed to permit an unobstructed view onto the utricular macula (Fig. 11A–C). Glittering otoconia were observed in Tg(+);Slc26a4Δ/Δ and Tg(+);Slc26a4Δ/+ mice and giant otoconia in Tg(−);Slc26a4Δ/Δ. Otoconia were transferred into glass-bottom dishes and inspected by laser-scanning microscopy using a 405 nm laser. Giant otoconia from Tg(−);Slc26a4Δ/Δ mice were ∼10-fold larger than normal otoconia (Fig. 11D). The shape of the giant otoconia resembled the shape previously observed in Slc26a4Δ/Δ mice [12]. Otoconia in Tg(+);Slc26a4Δ/Δ and Tg(+);Slc26a4Δ/+ mice were similar consisting in both genotypes of larger (∼20 µm, Figs. 11E.1 and F.1) and smaller (∼10 µm; Figs. 11E.2 and F.2) otoconia, some of which revealed a concentric structure (Figs. 11E.3 and F.3). These data suggest that the introduction of the transgene rescued normal otoconia formation.
Balance tests were performed in Tg(+);Slc26a4Δ/Δ mice and Tg(+);Slc26a4Δ/+ mice as well as in Tg(−);Slc26a4Δ/+ and Tg(−);Slc26a4Δ/Δ mice (Fig. 12). Tg(−);Slc26a4Δ/Δ mice failed the test, which confirmed the vestibular phenotype previously described in Slc26a4Δ/Δ mice [14]. There was no apparent difference in the performance of Tg(+);Slc26a4Δ/Δ, Tg(+);Slc26a4Δ/+ and Tg(−);Slc26a4Δ/+ mice. These data demonstrate that the introduction of the transgene rescued normal gross motor vestibular function.
In this study we generated a mouse that expresses human pendrin in the endolymphatic sac but lacks detectable pendrin protein expression in the cochlea or in the vestibular labyrinth. The most biologically interesting and clinically relevant observation of this study is that this mouse develops normal hearing and balance. Our findings support the hypothesis that pendrin expression in the endolymphatic sac is chiefly responsible for the development of normal endolymph volume, that lack of pendrin in the endolymphatic sac is mainly responsible for the development of the membranous labyrinth enlargement in Slc26a4Δ/Δ mice, and that the complex inner ear pathology found in Slc26a4Δ/Δ mice is largely a consequence of the enlargement during embryonic development. This hypothesis was based on the studies in Slc26a4Δ/Δ mice that revealed that the enlargement is a key event on the path toward organ failure resulting in deafness and vestibular dysfunction [12], [13], [18] and on studies in Foxi1−/− mice that lack pendrin expression in the endolymphatic sac, develop an enlargement of the inner ear, but express pendrin in the cochlea and the vestibular labyrinth [19].
To test our hypothesis, we generated a transgenic mouse, Tg(B1-hPDS), which expresses human SLC26A4 (previously named PDS) controlled by the promoter of ATP1V1B1 [15]. The ability of the promoter of ATP1V1B1 to control gene expression had previously been evaluated in a transgenic mouse that expresses eGFP controlled by the promoter of ATP6V1B1 [16], [20]. Expression of eGFP had been found in this mouse in intercalated cells of the renal collecting duct, and in narrow and clear cells of the epididymal epithelium of adult mice [16]. We found expression of eGFP in the embryonic kidney and in mitochondria-rich cells of the endolymphatic sac but not in the cochlea or the vestibular labyrinth (Fig. 3). Expression in mitochondria-rich cells of the endolymphatic sac was expected since these cells are members of the FORE family (forkhead related) of cells. FORE cells express FOXI1, which drives the expression of Atp6v1b1 and Slc26a4 [21], [22], [23]. Consistently, mitochondria-rich cells of the endolymphatic sac express the mRNAs Atp6v1b1 and Slc26a4 [24], [25] and the corresponding proteins, the B1-subunit of the vH+ATPase and pendrin [11], [23]. The onset of expression of Atp6v1b1 in the endolymphatic sac is at E11.5, which is similar to the onset of pendrin [13], [26]. Thus, it was likely that the transgene Tg(B1-hPDS) would drive a timely expression of pendrin in mitochondria-rich cells of the endolymphatic sac.
Although FOXI1 drives the expression of Atp6v1b1 and Slc26a4 in FORE cells such as the mitochondria-rich cells of the endolymphatic sac, the expression of Atp6v1b1 and Slc26a4 is not limited to FORE cells. Indeed, Slc26a4 is expressed in the inner ear in spiral prominence and outer sulcus epithelial cells as well as in spindle-shaped cells of the cochlea and in transitional cells of the vestibular labyrinth, none of which are FORE cells [11], [12]. Further, Atp6v1b1 expression has been found in the spiral limbus of the cochlea, which does not contain FORE cells [25], [26]. The expression of Atp6v1b1 in the cochlea provided the possibility that the transgene would drive an ectopic expression of pendrin in the spiral limbus. Our studies of eGFP expression (Fig. 3) and of pendrin expression by Western blotting (Fig. 4) and immunocytochemistry of whole-mounted specimens and sections using two different anti-pendrin antibodies (Pds #1 and Pds #2, Fig. 7 and S1) revealed no detectable expression in the cochlea or vestibular labyrinth of Tg(+);Slc26a4Δ/Δ mice. The observed absence of pendrin expression in the vestibular labyrinth of Tg(+);Slc26a4Δ/Δ mice (Fig. 10 and S2) is consistent with the reported lack of Atp6v1b1 expression in the vestibular labyrinth based on detection by in situ hybridization [25], [26] and by quantitative RT-PCR (Fig. 2B and F), which is a more sensitive technique. The observation that human SLC26A4 mRNA but no pendrin nor eGFP protein was detected in the cochlea or the vestibular labyrinth suggests the presence of strong translational regulation [27]. Taken together, our data demonstrate that we have generated a mouse that expresses pendrin in the endolymphatic sac but not in the cochlea or the vestibular labyrinth, although we cannot completely rule out that low levels of pendrin protein expression escaped our detection. Such low pendrin expression is unlikely the reason for the restored endolymphatic volume, since pendrin expression in the cochlea and vestibular labyrinth of Foxi−/− mice, which lack pendrin expression in the endolymphatic sac, did not prevent endolymphatic enlargement [19] and since mice that express a mutant pendrin protein that supports anion exchange at a reduced rate are deaf, develop mega-otoconia and are balance impaired [28]. Moreover, hypomorphic mutant alleles of SLC26A4 show no difference in the resulting auditory phenotype from that of functional null alleles in patients with Pendred syndrome [29], indicating that small amounts of pendrin activity are insufficient to rescue hearing in humans.
Measurements of the endocochlear potential and pH revealed that the introduction of the transgene, which rescued normal endolymph volume, also rescued the loss of the normal endocochlear potential and the loss of the normal endolymphatic pH homeostasis (Fig. 8). It appears that the Cl−/HCO3− exchanger pendrin, which is normally expressed in the apical membranes of spiral prominence and outer sulcus epithelial cells, is not the sole mechanism responsible for the alkaline pH of endolymph in a normally developed cochlea. A similar conclusion can be drawn based on measurements in the doxycycline-inducible Slc26a4 mouse model where termination of pendrin expression at P6 led to the development of a nearly normal endocochlear potential and of a nearly normal alkaline pH [30]. We hypothesize that the epithelial barrier enclosing endolymph is permeable to H+, OH− and HCO3− and that the pH of endolymph follows the endocochlear potential.
Hearing and balance tests in Tg(+);Slc26a4Δ/Δ mice revealed normal sensory function (Fig. 9 and 12). The observation that hearing thresholds at 32 kHz had some variability in Tg(+);Slc26a4Δ/Δ mice and that some Tg(+);Slc26a4Δ/Δ mice developed progressive high-frequency hearing loss is most likely a function of the genetic background. Tg(+);Slc26a4Δ/Δ were generated in a F2 generation from Slc26a4Δ/Δ mice that were maintained isogenic in the 129S6 background and Tg(+);Slc26a4Δ/Δ mice that were recently generated in a mixed background of C57BL/6 and CBA. DNA from the three background strains, 129S6, C57BL/6 and CBA, which differ in their hearing thresholds, are expected to comprise variable amounts of the genomes of individual mice. Hearing thresholds for 1 to 3 month-old 129S6, C57BL/6 and CBA mice range between 20–35 dB-SPL at 8 kHz, 10–28 dB-SPL at 16 kHz and 20–50 dB-SPL at 32 Hz [18], [31], [32], [33], [34], [35]. In general, 129S6, C57BL/6 and CBA mice have similar thresholds at 8 kHz, whereas at 16 and 32 kHz CBA mice have lower thresholds than 129S6 and C57BL/6 mice. Thus, the greater variability in hearing thresholds that was observed at 32 kHz particularly in Tg(+);Slc26a4Δ/Δ may be due to a variability in the mixture of these background strains.
Our observation that normal hearing developed in the absence of pendrin expression in the cochlea in combination with the published finding that normal hearing was maintained when pendrin expression was terminated after completed development [30], could suggest that pendrin in the cochlea has no physiologic significance beyond the developmental phase. However, it is also conceivable that pendrin-mediated HCO3− secretion provides a buffer that stabilizes the pH in the lateral wall tissues as well as in endolymph, and that this buffering is important during stress situations associated with normal life. Pendrin expression may indeed be important for the maintenance of hearing into advanced age.
In summary, we demonstrated that restoration of pendrin to the endolymphatic sac is sufficient to restore normal inner ear function. This implies that pendrin in the endolymphatic sac is more important for the development of normal hearing than pendrin expression in the cochlea and more important for the development of normal balance than pendrin expression in the vestibular labyrinth. This finding, in conjunction with our previous report that pendrin expression is required for embryonic development but not for the maintenance of hearing, opens the prospect that a spatially and temporally limited therapy will restore normal hearing in human patients carrying a variety of mutations of SLC26A4.
All animal experiments and procedures at Kansas State University were performed according to protocols approved by the Animal Care and Use Committees at Kansas State University (IACUC#: 2961). All animal procedures at Sorbonne University Paris Cité were performed according to protocols approved by the ethics committee from University Pierre et Marie Curie, and were performed in accordance with the Guide for the Care and Use of Laboratory Animals (NIH publication No. 93-23, revised 1985).
Human SLC26A4 cDNA was ligated into a pBluescript vector that contained 6.9 kbp of the human ATP6V1B1 promoter [16], [20]. An SV40 late region polyadenylation signal was cloned downstream of the SLC26A4 cDNA. The transgene Tg(B1-hPDS) included the 5′-flanking region of the ATP6V1B1 gene extending to but excluding the endogenous translational start codon, the human SLC26A4 cDNA, with its own translational start site, and the SV40 late region polyadenylation signal. The integrity of the transgene was confirmed by restriction digest and bidirectional sequencing of ligation sites. In preparation for injection, the transgene was linearized by SalI and NotI digestion, followed by gel purification using an electroelution method and then concentrated using ElutipD columns (Whatman). The transgene was then further concentrated by ethanol precipitation and resuspended in low EDTA injection buffer (10 mM Tris with 0.1 mM EDTA). Tg(B1-hPDS) transgenic mice were created by the University of Utah transgenic mouse core facility using standard procedures [16], [20]. Genotyping revealed that 63 pups were positive for transgene integration. One founder, which transmitted the transgene in a Mendelian fashion, was crossed with wild-type C57BL/6× CBA F1 mice to establish a colony. Three Tg(B1-hPDS);Slc26a4+/+ transgenic mice were shipped to Kansas State University in Manhattan, Kansas, USA.
At Kansas State University, a colony of Tg(B1-hPDS)Tg/+;Slc26a4Δ/Δ mice was established. Colony management was supported by software (Litter tracker, written in Microsoft Visual Basic and Excel 2010 by P.W.) Tg(B1-hPDS)Tg/+;Slc26a4+/+ mice were crossed with Slc26a4Δ/Δ mice to generate Tg(B1-hPDS)Tg/+;Slc26a4Δ/+ mice. Matings of Tg(B1-hPDS)Tg/+;Slc26a4Δ/+ mice generated 28 Tg(B1-hPDS)Tg/+;Slc26a4+/+, 58 Tg(B1-hPDS)Tg/+;Slc26a4Δ/+ and 39 Tg(B1-hPDS)Tg/+;Slc26a4Δ/Δ mice in a near Mendelian ratio of 1 ∶ 2 ∶ 1 with a 75% rate of transmission for the transgene (based on 169 pups).
Mice were genotyped for Slc26a4+ and Slc26a4Δ alleles by PCR using established primers [14] and for the transgene Tg(B1-hPDS) (Transnetyx, Cordova, TN). Primers for the transgene were designed to amplify a 345 bp PCR-product spanning the hPDS cDNA and the SV40 polyadenylation signal sequence (left primer: 5′-aga ggg tca agg ttc cat ttt ag-3′; right primer: 5′-caa acc aca act aga atg cag tg-3′) [15].
Time-pregnant dams were deeply anesthetized with 4% tri-bromo-ethanol (0.014 ml/g body weight, i.p.) and embryos were harvested by laparotomy. Dams and embryos were sacrificed by decapitation. Gestational age was counted from the day when the vaginal plug was detected. This day was set to embryonic (E) day 0.5. Gestational age, however, was verified by evaluating gross morphological features including limbs, digits and the appearance of the pinna and auditory meatus [36], [37].
The age of mice was counted from the day of birth, which was set to postnatal (P) day 0. Postnatal mice were deeply anesthetized with 4% tri-bromo-ethanol (0.014 ml/g body weight, i.p.) and sacrificed by decapitation or cardiac perfusion with fixative.
Quantitative RT-PCR was performed on total RNA [17]. Total RNA was isolated from tissues obtained by microdissection from Tg(−);Slc26a4Δ/+ and Tg(+);Slc26a4Δ/Δ mice and subjected to quantitative RT-PCR using gene-specific primers for 18S rRNA as well as for mRNA coding for the α-subunit of the mouse Na+/K+ ATPase Atp1a1, the B1-subunit of the mouse vH+ATPase Atp6v1b1, mouse pendrin Slc26a4 and human pendrin SLC26A4, which was introduced via the transgene.
Postnatal mice were genotyped by PCR prior to tissue collection. Embryonic Tg(−);Slc26a4Δ/+ mice were generated by mating Tg(−);Slc26a4Δ/Δ dams and Tg(−);Slc26a4+/+ sires, which yielded 100% of the desired genotype. Embryonic Tg(+);Slc26a4Δ/Δ mice were generated by mating Tg(+);Slc26a4Δ/Δ dams and sires, which yielded Tg(+);Slc26a4Δ/Δ and Tg(−);Slc26a4Δ/Δ mice in a ratio of 3 ∶ 1. Since embryonic mice could not be genotyped prior to tissue collection, the desired Tg(+);Slc26a4Δ/Δ mice among Tg(−);Slc26a4Δ/Δ mice were initially identified by visual inspection of the size of the endolymphatic sac and the presence of ‘glittering’ otoconia, This phenotypic identification was subsequently confirmed by the presence of human pendrin SLC26A4 transgene by RT-PCR.
Tissues were obtained by microdissection. Endolymphatic sacs (8–10 endolymphatic sacs from 4–5 animals per sample) were obtained from mice at age E17.5. Cochlear ducts (4 cochlear ducts from 2 animals per sample and 2 cochlear ducts from 1 animal per sample) were obtained from mice at ages E17.5 and P2, respectively. Vestibular labyrinths (6 vestibular labyrinths from 3 animals per sample) were obtained from mice at age P8. Total RNA was isolated from microdissected tissues (RNeasy micro kit, Qiagen, Valencia, CA, USA), treated with DNAse (RNeasy micro kit), combined with RNA storage solution (Applied Biosystems/Ambion, Austin, TX), adjusted to a concentration of 10 ng/µl, and stored at −80°C.
Quantity and quality of total RNA were evaluated by microfluidic electrophoresis (BioAnalyzer, Agilent, Santa Clara, CA), by microliter absorption photometry (Nanodrop, Wilmington, DE) and by quantitative RT-PCR of 18S rRNA. RNA samples were accepted for quantitative RT-PCR only when they were free of contamination and excellent RNA quality. RNA quality was quantified by the RNA integrity number (RIN) on a scale from 0 (worst) to 10 (best) (BioAnalyzer, Agilent). RIN numbers for total RNA isolated from E17.5 endolymphatic sac and cochlea were 8.2±0.3 (n = 3) and 9.2±0.1 (n = 10).
Chemicals were assembled with the assistance of an automatic pipetting station (Biomek NXp, Beckman Coulter, Fullerton, CA) with hardware modifications and software programming by P.W. Quantitative RT-PCR reactions were carried out in 96-well plates with each well containing ∼10 ng of total RNA, gene specific primers, and an enzyme mix containing reverse transcriptase and DNA polymerase (iScript, BioRad, Hercules, CA) in a total volume of 25 µl. Reverse transcription was performed for 10 min at 50°C and terminated by heating to 95°C for 5 min (OneStepPlus, Applied Biosystems, Foster City, CA). PCR consisted of 40 cycles of 10 s melting at 95°C, 30 s annealing and elongation at 58°C, and 15 s hot-measurement at 78°C (OneStepPlus, Applied Biosystems).
Left and right primers (exon, product size) were for 18S 5′-gag gtt cga aga cga tca ga-3′ and 5′-tcg ctc cac caa cta aga ac-3′ (316 bp), for Atp1a1 5′-tgc ccg cct caa cat tcc-3′ (exon 14) and 5′-gac aca tca gag cca aca atc c-3′ (exon 16, 291 bp), for Atp6v1b1 5′-tga ccc gaa act aca tca cc-3′ (exon 1) and 5′-gcc aga gcc att gaa aat cc-3′ (exon 5, 305 bp), for mouse Slc26a4 5′-tct gat gga ggc aga gat ga-3′ (exon 20) and 5′-ggc cag cct aac aga gac ag-3′ (exon 21, 430 bp), and for human SLC26A4 were 5′-tcc caa agt gcc aat cca ta-3′ and 5′-aca tca agt tct tct tcc gtc ag-3′ (360 bp). Primer pairs for Atp1a1, Atp6v1b1, mouse Slc26a4 spanned introns to prevent amplification of genomic DNA. Primer pairs for mouse Slc26a4 detected the Slc26a4+ allele as well as the Slc26a4Δ allele that is lacking exon 8 [14]. Left and right primers for mouse Slc26a4 differed by 7 and 10 nucleotides from the corresponding human sequence and left and right primers for human SLC26A4 differed in 4 and 6 nucleotides from the corresponding mouse sequence, thereby maximizing species-specific amplification. Since the human transgene did not contain introns, some reactions were carried out without reverse transcriptase to determine whether products of SLC26A4 originated from cDNA rather than from genomic DNA. These experiments revealed no evidence for significant amplification of genomic DNA. Amplification of a single product of the appropriate size was verified by microfluidic electrophoresis (BioAnalyzer, Agilent).
The number of template molecules (cDNATemplate) was estimated according towhere 6.02×1023 molecules/mol represents Avogadro's number, ProductThreshold is the weight of the PCR-product at threshold (0.49×10−9 g) that was obtained from calibration experiments, ProductSize is the size of the product in base pairs (bp), Weightbp is average weight of one bp (660 g/mol), Efficiency is the PCR-efficiency obtained from the slope of the log-linear phase of the growth curve [38] and Ct is the cycle at which the fluorescence of the product molecules reaches a common threshold chosen in the middle of the log-linear part of the growth curve.
Bisected heads of embryos age E15.5 were fixed overnight in Bodian's fixative, contained (vol/vol) 75% ethanol, 5% acetic acid, and 5% formalin in water. Heads were then dehydrated overnight in 100% ethanol and cleared in methyl-salicylate [39], [40]. The membranous labyrinth was injected via the lateral wall of the basal turn of the cochlea and via the endolymphatic sac with diluted paint (Liquid Paper, Newell Rubbermaid, Atlanta, GA, 0.1–0.2% in methyl-salicylate) using a fine glass-electrode, a manipulator (NM-151 Narishige) and a micrometer-driven oil-filled microinjector (CellTram Vario, Eppendorf, Hamburg, Germany). For each genotype, at least three inner ears were injected.
Whole mounts of fresh cochlear ducts, endolymphatic sacs and slices of kidney were prepared from E15.5 Tg(B1-eGFP) mice and visualized with a fluorescence microscope (AxioScope, Carl Zeiss Göttingen).
Mice were anesthetized with 4% tri-bromo-ethanol for in situ measurements of the endocochlear potential and pH with double-barreled microelectrodes. Measurements were made in the basal turn of the cochlea by a round-window approach through the basilar membrane of the first turn of the cochlea [18], [42]. The surgical cavity was covered with liquid Sylgard 184 (Dow Corning) to limit the loss of tissue CO2 into ambient air.
Double-barreled glass microelectrodes were pulled (micropipette puller PD-5; Narishige) from filament-containing glass tubing (1B100F-4; World Precision Instruments) and baked at 180°C for 2 h to ensure dryness. One barrel was silanized by a 30 s exposure to 0.008 ml dimethyldichlorosilane (40136; Fluka). After silanization, microelectrodes were baked again at 180°C for 3 h and tips were broken to a final O.D. of ∼3 µm. The reference barrel was filled with 150 mM KCl and the ion-selective barrel was filled at the tip with liquid ion exchanger (Hydrogen ionophore II - Cocktail A, 95297; Fluka) and back-filled with buffer solution (500 mM KCl, 20 mM HEPES, pH 7.4).
Each barrel of the double-barreled microelectrode was connected via a Ag/AgCl2 electrode to an electrometer (FD223, World Precision Instruments). A flowing KCl electrode (1 M KCl in 0.2% agar) was inserted under the skin of the animal to serve as ground electrode. Data were recorded in analog (BD12E Flatbed recorder, Kipp & Zonen, Delft, The Netherlands) and digital form (DIGIDATA 1322A and AxoScope 10, Axon Instruments, Union City, CA). pH electrodes were calibrated in situ at 37°C using three calibration solutions with different pH values. Calibration solutions contained (in mM): pH 6: 130 NaCl, 20 MES; pH 7: 130 NaCl, 20 HEPES; and pH 8: 130 NaCl, 20 HEPES. pH-sensitive electrodes had a slope of 56.9 ± 0.3 mV/pH unit (n = 11).
Mice were deeply anesthetized with a mixture of dexmedetomidine and ketamine (0.375 mg/Kg body weight dexmedetomidine and 56 mg/kg body weight ketamine; i.p.) and placed on a thermal pad to maintain normal body temperature. The mastoid, vertex and ventral neck region of the animal were connected via sub-dermal platinum needle electrodes (F-E2, Astro Med, Rhode Island, RI) and short (31 cm) leads to the main channel, reference channel and ground of the preamplifier, respectively. Auditory brainstem recordings were performed in a custom constructed, electrically shielded and sound-attenuated chamber (inner dimensions: 23 cm×23 cm×23 cm) using a digital data acquisition system (BioSig32 software, RA4LI Preamplifier, RP2.1 Enhanced Real Time Processor, PA5 Programmable Attenuator, ED1 Electrostatic Speaker Driver, Tucker-Davis Technologies, Alachua, FL). Tone burst stimuli were presented (21 per sec) via a free field electrostatic speaker (SigGen software, ES1 speaker, Tucker Davis). Acoustic stimuli were calibrated using a 1/4 inch condenser microphone (SigCal IRP4.2 software, Tucker Davis, PS9200 microphone, Acoustical Interface, Belmont, CA) placed at the location of the mouse head. Tone bursts (2 ms duration, 0.5 ms gate time; 8, 16 and 32 kHz) were presented with alternating phase (0 and 180°). Responses, recorded over 10 ms, were filtered (300 Hz high pass, 3000 Hz low pass and 60 Hz notch) and 1000 recordings were averaged. Tone burst stimuli were presented at intensities varying between 90 and 0 dB SPL in 5 dB intervals. Auditory thresholds were obtained by a visual comparison of wave forms. After the procedure, mice were rapidly recovered from anesthesia with atipamizole (1.875 mg/kg body weight; i.p.).
Balance testing consisted of determining the time that mice could balance on a rotating 1″ rod with rotations ramping up from 4 to 40 rpm in 60 s (RotaRod, IITC Life Science, Woodland Hills, CA). Test chambers were cushioned with bubble-foil to provide a soft landing for mice falling off the rod.
|
10.1371/journal.pntd.0003176 | Genome of the Avirulent Human-Infective Trypanosome—Trypanosoma rangeli | Trypanosoma rangeli is a hemoflagellate protozoan parasite infecting humans and other wild and domestic mammals across Central and South America. It does not cause human disease, but it can be mistaken for the etiologic agent of Chagas disease, Trypanosoma cruzi. We have sequenced the T. rangeli genome to provide new tools for elucidating the distinct and intriguing biology of this species and the key pathways related to interaction with its arthropod and mammalian hosts.
The T. rangeli haploid genome is ∼24 Mb in length, and is the smallest and least repetitive trypanosomatid genome sequenced thus far. This parasite genome has shorter subtelomeric sequences compared to those of T. cruzi and T. brucei; displays intraspecific karyotype variability and lacks minichromosomes. Of the predicted 7,613 protein coding sequences, functional annotations could be determined for 2,415, while 5,043 are hypothetical proteins, some with evidence of protein expression. 7,101 genes (93%) are shared with other trypanosomatids that infect humans. An ortholog of the dcl2 gene involved in the T. brucei RNAi pathway was found in T. rangeli, but the RNAi machinery is non-functional since the other genes in this pathway are pseudogenized. T. rangeli is highly susceptible to oxidative stress, a phenotype that may be explained by a smaller number of anti-oxidant defense enzymes and heat-shock proteins.
Phylogenetic comparison of nuclear and mitochondrial genes indicates that T. rangeli and T. cruzi are equidistant from T. brucei. In addition to revealing new aspects of trypanosome co-evolution within the vertebrate and invertebrate hosts, comparative genomic analysis with pathogenic trypanosomatids provides valuable new information that can be further explored with the aim of developing better diagnostic tools and/or therapeutic targets.
| Comparative genomics is a powerful tool that affords detailed study of the genetic and evolutionary basis for aspects of lifecycles and pathologies caused by phylogenetically related pathogens. The reference genome sequences of three trypanosomatids, T. brucei, T. cruzi and L. major, and subsequent addition of multiple Leishmania and Trypanosoma genomes has provided data upon which large-scale investigations delineating the complex systems biology of these human parasites has been built. Here, we compare the annotated genome sequence of T. rangeli strain SC-58 to available genomic sequence and annotation data from related species. We provide analysis of gene content, genome architecture and key characteristics associated with the biology of this non-pathogenic trypanosome. Moreover, we report striking new genomic features of T. rangeli compared with its closest relative, T. cruzi, such as (1) considerably less amplification on the gene copy number within multigene virulence factor families such as MASPs, trans-sialidases and mucins; (2) a reduced repertoire of genes encoding anti-oxidant defense enzymes; and (3) the presence of vestigial orthologs of the RNAi machinery, which are insufficient to constitute a functional pathway. Overall, the genome of T. rangeli provides for a much better understanding of the identity, evolution, regulation and function of trypanosome virulence determinants for both mammalian host and insect vector.
| Human trypanosomiases result in high morbidity and mortality, affecting millions of people in developing and underdeveloped countries. In Africa, Trypanosomiasis (sleeping sickness) is tsetse-transmitted and is caused by Trypanosoma brucei gambiense and T. b. rhodesiense; whereas, in the Americas, Trypanosomiasis (Chagas disease) is transmitted by triatomine bugs and is caused by Trypanosoma cruzi. Trypanosoma rangeli (Tejera, 1920) is a third human infective trypanosome species that occurs in sympatry with T. cruzi in Central and South America, infecting a variety of mammalian species, including humans [1]. Natural mixed infections involving T. rangeli and T. cruzi have been reported in a wide geographical area for both mammals and the triatomine insect vectors [2], [3].
Literature on serological cross-reactivity between T. rangeli and T. cruzi has documented an ongoing controversy, probably influenced by the parasite form and/or strain, the host infection time and the serological assay used. While several authors have reported serological cross-reactivity between T. cruzi and T. rangeli in assays of human sera by conventional immunodiagnostic tests [1], [4]–[6], others have reported no cross-reactivity when recombinant antigens or species-specific synthetic peptides are used [7]. Recently, some species-specific proteins were identified in T. rangeli trypomastigotes which may provide for an effective differential in serodiagnosis [8].
In contrast to T. brucei and T. cruzi, T. rangeli is considered non-pathogenic to mammalian hosts but harmful to insect vectors, especially those from the genus Rhodnius. It causes morphological abnormalities and death of triatomine nymphs during molting [9], [10]. T. rangeli is transmitted among mammals through an inoculative route during hematophagy [1]–[3]. The parasite life cycle in the triatomine is initiated by ingestion of trypomastigote forms during a blood meal on an infected mammal. After switching to its epimastigote form, the parasite multiplies and colonizes the insect gut, prior to invading the hemocoel through the intestinal epithelium. Once in the hemolymph, T. rangeli replicates freely and invades the salivary glands, wherein it differentiates into infective metacyclic trypomastigotes [1]. T. rangeli infection via the contaminative route (feces) may also occur, as observed for T. cruzi, given that infective trypomastigotes are also found in the vector gut and rectum.
Although T. rangeli has been found to infect more than 20 mammalian species from five different orders, the parasite's life cycle in these hosts is poorly understood. Between 48 to 72 hours after the inoculation of short metacyclic trypomastigotes (10 µm), a small number of large trypomastigotes (35–40 µm) are found in the bloodstream and appear to persist for 2–3 weeks, after which the infection becomes subpatent. Despite the lack of a visible parasites in the blood, the parasite has been isolated from experimentally infected mammals up to three years after infection [1]. However, neither extracellular nor intracellular multiplication of the parasite in the mammalian host has been clearly demonstrated thus far.
High intra-specific variability has been described between T. rangeli strains, using multiple molecular genetic markers [2], [11]–[16]. A strong association of T. rangeli genetic groups with their local triatomine vector species has been demonstrated, and it has been proposed that the geographic distribution of the parasite' genotypes is associated with a particular evolutionary line of Rhodnius spp., indicating diversification may be tightly linked to host-parasite co-evolution [11], [16]–[18].
The gene expression profiles of distinct forms and strains of T. rangeli representing the major phylogenetic lineages (KP1+ and KP1−) were assessed via sequencing of EST/ORESTES [19]. Despite the non-pathogenic nature of T. rangeli in mammals, comparison of these transcriptomic data with data from T. cruzi and other kinetoplastid species revealed the presence of several genes associated with virulence and pathogenicity in other pathogenic kinetoplastids, such as gp63, sialidases and oligopeptidases.
Although T. rangeli is not particularly pathogenic in mammals, in light of its resemblance, sympatric distribution and serological cross-reactivity with T. cruzi, we decided to sequence and analyze the genome of T. rangeli. Here, we present the T. rangeli genome sequence and a comparative analysis of the predicted protein repertoire to reveal unique biological aspects of this taxon. Our findings may be useful for understanding the virulence and emergence of the human infectivity of Trypanosoma species.
Epimastigotes from the T. rangeli SC-58 (KP1−) and Choachí (KP1+) strains were maintained in liver infusion tryptose (LIT) medium supplemented with 15% FCS at 27°C after cyclic mouse-triatomine-mouse passages. The T. cruzi CL Brener and Y strains were maintained in liver infusion tryptose (LIT) medium supplemented with 10% FCS at 27°C. All samples tested negative for the presence of Mycoplasma sp. by PCR. For DNA sequencing, exponential growth phase epimastigotes from T. rangeli SC-58 strain were washed twice in sterile PBS and genomic DNA was extracted from parasites using the phenol/chloroform method.
Chromosomal DNA was isolated and fractionated via PFGE as described elsewhere [20], [21]. Briefly, 1.1% agarose gels were prepared in 0.5X TBE (45 mM Tris; 45 mM boric acid; 1 mM EDTA, pH 8.3), and agarose plugs containing the samples were loaded into the gels and electrophoresed using the Gene Navigator System (Amersham Pharmacia Biotech) at 13°C for 132 hours. The gels were then stained with ethidium bromide (EtBr) (0.5 mg/mL). The chromosomal bands of T. rangeli (Choachí and SC-58 strains) and T. cruzi (CL Brener clone) were fractioned using a protocol optimized to separate small DNA molecules in the CHEF Mapper system to assess the presence of minichromosomes.
Library generation and sequencing were performed at the Computational Genomics Unit Darcy Fontoura de Almeida (UGCDFA) of the National Laboratory of Scientific Computation (LNCC) (Petrópolis, RJ, Brazil). 454 GS-FLX Titanium sequencing was utilized. Two sequencing libraries were prepared from T. rangeli SC-58 gDNA: one shotgun library (SG) and one 3 kb paired-end library (PE). Each library was constructed from 5 µg of genomic DNA (gDNA) following the GS FLX Titanium series protocols. All titrations, emulsions, PCR, and sequencing steps were carried out according to the manufacturer's protocol. One full PicoTiterPlate (PTP) was used to sequence each library.
In order to estimate the T. rangeli genome size, a pipeline developed at the Karolinska Institutet (KI) generated a genome assembly. Briefly, the 454 SFF (Standard Flowgram Format) files were processed using custom Perl scripts to generate paired-end (PE) FASTQ files. Subsequently, the SFF files were assembled without prior treatment using the Newbler assembler. The resulting assembly was scaffolded using SSPACE 2.1.0 with the generated 454 PE reads, and finally, assembly gaps were improved using GapFiller 1.11.
In order to specifically identify conserved protein coding regions, an alternate, protein-centric procedure was also utilized. A reference-guided assembly of T. rangeli genic regions was carried out using protein sequences from TriTrypDB as formerly described [22], resulting in an overview of the predicted parasite proteome. For this, 73,808 protein sequences were selected from the TriTrypDB (release 3.3 – http://tritrypdb.org/common/downloads/) and used for comparative analysis. All proteins retrieved from TriTrypDB were clustered by BBH (Bidirectional Best Hit), totaling 8,807 clusters. Parasite proteins that were not clustered were also used, for a total of 16,347 protein sequences. Sequences containing start codons different from ATG or containing stop codons in the middle of the sequence were filtered out. For each cluster, one protein was selected based on the following hierarchical criteria: (1) a T. cruzi protein with annotated function, or (2) a protein with annotated function from an organism different than T. cruzi, or (3) a T. cruzi hypothetical protein, or (4) the largest protein. The selected sequences were compared to reads from T. rangeli using tBLASTn, applying an E-value cut-off threshold of 1e–5 to define a set of significant reads to reconstruct each protein sequence. Each protein sequence was reconstructed with the counterpart set of reads selected using the software Newbler 2.5.3 according to the default parameters.
Automatic functional annotation of the T. rangeli genome was performed using the System for Automated Bacterial Integrated Annotation (SABIA) [23], including the previously generated and annotated EST/ORESTES database [19] and proteomic data obtained from surface of T. rangeli trypomastigotes [8].
The assembled nucleotide sequences were translated to aminoacid sequence and annotated according to the following criteria:
Transposable elements were screened in genome assembly (KI) based on similarity using BLASTn, tBLASTn and tBLASTx tools [25]. As queries, the Repbase sequences described for the Euglenozoa group were used [26]. The BLAST results were filtered using the following parameters (e-value≤0.01, identity ≥50%, score≥80), tBLASTx (e-value≤0.01, identity ≥30%, score≥100) and tBLASTn (e-value≤0.01, identity ≥30%, score≥100). The retrieved sequences (protein and nucleotide) were aligned with the reference sequences and were manually curated. For ab initio searches, the software RepeatScout, release 1.0.5 was used [27].
Peptides sequences from nine selected trypanosomatid multigene families (MASP, GP63, Trans-sialidase, Amastin, DGF, KMP-11, Tuzin, RHS and Mucin) were downloaded from TriTrypDB (tritrypdb.org). T. rangeli reads were then aligned against all members of each multigene family using BLASTx algorithm [25] and the reads from the best hits were selected. Those reads were assembled using CAP3 [28] and the resulting contigs were re-aligned against the NR (non-redundant) database from GenBank (https://www.ncbi.nlm.nih.gov/genbank/) and manually inspected to verify that they belong to the aforementioned multigene families. These validated contigs were used to construct a database corresponding to a subset of T. rangeli coding sequences belonging to the selected multigene families, except for the mucin genes. To determine gene copy number, the entire read dataset from the T. rangeli genome and all contigs generated, as described above, were aligned using reciprocal MegaBLAST and all reads corresponding to each contig were selected. After checking, the cut off for minimal identity (with no convergence in reads picking) was set as 95% identity, 10e-15 e-value and at least 80% of read coverage. The best hits were computed and used to calculate the read depth for each nucleotide and the regions covered with the highest rates were selected for the downstream analyses. The selected regions from each contig displaying high coverage values were realigned to NR protein database to verify specific multigene family before the copy numbers for each contig were calculated using the nucleotide by nucleotide coverages obtained with the z-score algorithm. The final coverage for each contig was then calculated after dividing the z-score value by the calculated genome sequencing coverage of 13.78. For all multigene families we added the values obtained as a copy number estimation for each contig to determine the final values displayed as the gene copy number of each family. For mucin genes, because signal peptide sequences are highly conserved in the different members of this family, the read coverage was carried only for the first 75 nucleotide present in the AUPL00006796 gene. To validate our method for copy number estimations and also to verify that the cutoff values were accurate this pipeline was applied to three genes known to be present as single copy genes in most trypanosomatid genomes (msh2, msh6 and gpi8).
A phylogenomic analysis was carried out using all orthologous proteins from distinct species of the Trypanosomatidae family (T. rangeli SC-58, T. cruzi CL Brener Esmeraldo-like, T. cruzi CL Brener non-Esmeraldo-like, T. cruzi Sylvio X10, T. brucei, L. braziliensis, L. infantum and L. major). The multi-FASTA ortholog files containing the best representative of each trypanosomatidae protein sequence were used as inputs for multiple alignments with the default parameters of the CLUSTAL Omega algorithm [29]. All alignments were visually inspected and manually annotated whenever necessary the removal of low quality alignments. Subsequently, protein concatenation of the 1,557 alignment files obtained was carried out using SCaFos software [30].
Phylogenies from the concatenated deduced amino acid sequences of all species were estimated through both protein distance and probabilistic methods, using the PHYLIP package [31] and TREE-PUZZLE [32], respectively. The Seqboot program of the PHYLIP package was used to generate multiple 100-bootstrapped datasets, which were submitted to ProtDist software to compute a distance matrix under the JTT (Jones-Taylor-Thornton) model of amino acid replacement. The neighbor-joining (NJ) method [33] was applied to the resultant multiple datasets, implemented in Neighbor software, which constructed trees via successive clustering of lineages.
The quartet-puzzling [34] search algorithm implemented by TREE-PUZZLE was used to reconstruct phylogenetic trees based on maximum likelihood (ML). The Jones-Taylor-Thornton (JTT) model of amino acid substitution was applied. The quartet-puzzling tree topology was based on 1,000 puzzling steps. The consensus tree was constructed considering a 50% majority rule consensus. The TreeView program [35] and MEGA 5 [36] were used to visualize and edit the resultant phylogenies.
All protein kinase and phosphatidylinositol kinase sequences were selected and manually curated and re-annotated using the following software: Kinomer v. 1.0 web server [37], Kinbase (http://www.kinase.com/kinbase/), SMART (http://smart.embl-heidelberg.de/), Interproscan (http://www.ebi.ac.uk/Tools/pfa/iprscan/) and Motifscan (http://myhits.isb-sib.ch/cgi-bin/motif_scan). The presence of accessory domains and the domain architecture of some proteins, such as those from the AGC group, were decisive in classifying them into a group. PIK and PIK-related kinases were classified according to previous reports [38]–[41].
Analyses were performed using Tandem Repeat Finder (TRF) [42] and Tandem Repeat Assembly Program (TRAP) [43] software. The T. rangeli genome assembly (KI) and transcriptome [19] (2.45 Mb) sequences were submitted to TRF using the default parameters, except for minimum score of 25, as were 32.5 Mb of T. cruzi CL Brener Esmeraldo-like genome sequences from TriTrypDB using the same software parameters. The TRF output files were compiled using TRAP software, and we categorized the repeat sequences into four groups: microsatellites (1 to 6 nucleotides), unclassified (7 to 11 nucleotides), minisatellites (12 to 100 nucleotides) and satellite sequences (up to 100 nucleotides). The abundance, frequency and density of all T. rangeli repeat categories were calculated. Microsatellite classes were also analyzed considering all possible combinations; e.g., the repeat locus AGAT also included GATA, ATAG, TAGA and the reverses complements TCTA, CTAT, TATC and ATCT.
To identify RNAi-related genes in the T. rangeli genome assembly, a set of 39 primers targeting the five genes constituting the RNAi machinery were designed and used to amplify these genes from the parasite genome by PCR. The PCR products were then purified using the Illustra GFX PCR DNA and Gel Band Purification kit (GE Healthcare) and cloned into pGEM-T-Easy vectors (Promega) or directly sequenced. Both strands of the PCR products or inserts were sequenced in a MegaBase automated sequencer, as directed by the manufacturer (GE Healthcare). After quality assessment using the Phred/Phrap/Consed package, sequences showing a Phred>30 were used along with the genome sequences to assemble the RNAi genes. Alignment of the consensus T. rangeli sequences with the T. brucei RNAi genes (TriTrypDB accession numbers Tb927.10.10850, Tb927.8.2370, Tb927.3.1230, Tb10.6k15.1610 and Tb927.10.10730) was carried out using MultiAlin [44].
Functional characterization of the T. rangeli RNAi machinery was performed using parasites transfected with the pTEXeGFP plasmid, kindly donated by Dr. John Kelly (LSHTM, UK). Silencing of eGFP was conducted using the TriFECTa exogenous reporter gene EGFP-S1 DS Positive Control (IDT) or the eGFP antisense siRNA EGFP-AS (5′-UGC AGA UGA ACU UCA GGG UCA-3′). Vero cells transfected with the pEGFP e1 plasmid (Clontech) were used as a positive control. All transfections were carried out in biological triplicates using a Nucleofector II device and the Human T Cell Nucleofector kit (Lonza). eGFP expression and silencing was assessed in both parasites and cells by Western blotting, flow cytometry analysis (FACS), direct fluorescence (FA) and qPCR. In the Western blot assays, an anti-GFP antibody (Santa Cruz Biotechnology) diluted 1∶2,000 was employed, according to standard protocols, and flow cytometry was carried out in a FACSCanto II (BD) apparatus.
Additionally, the functionality of the T. rangeli RNAi machinery was assessed through the transfection of epimastigote forms with the TUBdsRNA-RFP plasmid [45]. The evaluation of cell morphology and detection of RFP fluorescence were carried out at 6, 12, 24, 48 and 72 hours post-transfection using a BX FL 40 microscope (Olympus).
The karyotypes of representative strains from two major T. rangeli lineages [Choachí (KP1+) and SC-58 (KP1−)] were obtained via pulsed-field gel electrophoresis (PFGE). Two chromosomal-band size classes were defined: 1) megabase bands (those ranging from 2.19 to 3.5 Mb) 2) smaller bands, (ranging from 0.40 and 1.48 Mb). This analysis revealed at least 16 chromosomal-bands, whose sizes varied from 0.40 to 3.44 Mb; two megabase bands and 13–14 smaller bands (Figure 1A). We used specific PFGE separation conditions to confirm the absence of minichromosomes (Figure 1B), which are present in T. brucei, [46], but not in T. cruzi. The fluorescence intensity varied between these chromosomal bands, suggesting that co-migrating chromosomes are not necessarily homologous and that ploidy differences exist. The occurrence of aneuploidy has been demonstrated in different T. cruzi strains [21], [47] and in various species and isolates of Leishmania spp. [48], [49]. Of the 16 chromosomal bands identified, only seven were of a similar molecular size in the two T. rangeli isolates, confirming the existence of chromosomal size polymorphism, as demonstrated previously [50]–[52]. Therefore, analogously to T. cruzi, these 16 chromosomal bands may not reflect the actual number of chromosomes. Rather, this number is most likely higher than 16, as a single band may contain co-migrating heterologous chromosomes of similar sizes. Further studies will be needed to define the exact number of chromosomes and ploidy in T. rangeli.
Based on ssu rDNA and gapdh gene sequences, T. rangeli was phylogenetically positioned relatively closer to T. cruzi than to T. brucei [12]. This evolutionary proximity may also be reflected in the chromosomal organization of these species. It has been suggested that the common ancestor of trypanosomes exhibited smaller and more fragmented chromosomes and that fusion events occurred in the T. brucei lineage, leading to the smaller number of chromosomes currently observed [53]. Consistent with this idea, the chromosomal organization of T. rangeli also shows smaller and possibly more fragmented chromosomes, similar to those of T. cruzi [21].
The general characteristics of the T. rangeli genome sequence are shown in Table 1 (GenBank accession AUPL00000000). The applied 454-based approach allowed the generation of 2,206,288 reads, which after reference-guided assembly to representative kinetoplastid gene sequences available at TriTrypDB, resulted in identification of a total of 7,613 coding sequences (CDS) from the T. rangeli reads. These CDSs include tRNAs encoding all 20 amino acids. In addition, we identify 33 genes corresponding to the typical trypanosomatid rRNAs (5.8S, 18S and 28S) (GenBank accession KJ742907). As has been observed for numerous other pathogenic and non-pathogenic trypanosomatids [54], a high percentage of T. rangeli genes (∼65.6%) encode hypothetical proteins. Among these genes, 44 show evidence of expression as revealed by BLASTx similarity to proteins detected via mass spectrometry on the surface of T. rangeli trypomastigotes [8]. Comparative sequence analysis revealed that 7,101 CDS (93%) of the identified T. rangeli genes are shared with other human pathogenic trypanosomes (Figure 2). T. rangeli shares 403 gene clusters exclusively with T. cruzi, thus reinforcing the phylogenetic relationship of these species. The conserved genome core of the 5,178 gene clusters present in all species (T. rangeli, T. cruzi, T. brucei and L. major) are mainly involved in fundamental biological processes and to host-parasite interactions (Figure 2), representing ∼84% of the TriTryp (T. cruzi, T. brucei and L. major) genome core [55].
In addition to reference-based gene assembly, a relatively high-quality de novo genome assembly was generated from paired-end reads utilizing the Karolinska Institutet pipeline. The final genome assembly contains 259 scaffolds with 4.42% gaps. Given the NG50 (statistic of scaffold lengths) of (202,734 bp) and the low repeat content of this genome, it is clear that most of the genome has been reconstructed. The assembly obtained by using the pipeline corroborates our draft reference-guided assembly data, suggesting a size of the T. rangeli genome of ∼24 Mb. Thus, the T. rangeli genome is the smallest and least repetitive trypanosomatid genome obtained to date including T. cruzi CL Brener and Sylvio X-10, T. cruzi marinkellei, T. brucei and Leishmania sp. [56]–[61].
Based on a total of 1,557 orthologous sequences representing different CDSs encoded by 8 different trypanosomatid genomes, an alignment of 964,591 concatenated amino acid residues was obtained and used to create NJ and ML tree topologies that were robust and revealed that South American trypanosomes (T. rangeli and T. cruzi) are equidistant from the African trypanosome (T. brucei) (Figures 3A and 3B). Despite the well-established genomic variability among T. cruzi strains, sequences derived from all strains CL Brener - Esmeraldo and non-Esmeraldo-like haplotypes - and Sylvio X10, clustered closer to T. rangeli than to T. brucei with high bootstrap values. The use of a phylogenomic approach to assess the evolutionary history of trypanosomatids clearly positioned T. rangeli closer to T. cruzi than T. brucei at the genomic level, corroborating former studies using single or a few genes [2], [3], [11], . T. rangeli and T. cruzi share conserved gene sequences with remarkably few genes or paralog groups that are unique to each one of the two species. Nevertheless, the divergence between T. rangeli and any T. cruzi strain is much greater than the differences among T. cruzi strains. As expected, all Leishmania species (L. braziliensis, L. infantum, and L. major) were clustered to a distinct branch.
The abundance, frequency and density of non-coding tandem repeat sequences found in the T. rangeli genome and transcriptome sequences; as well as a comparison of satellite DNA sequences to the T. cruzi haploid genome; are presented in Table S1. Approximately 1.27 Mb (6%) of the current T. rangeli genome assembly (∼24 Mb) is composed of tandem repeat sequences. Microsatellites are the most abundant repeats in both the T. rangeli (0.78 Mb, or 3.9%) and T. cruzi CL Brener (1.01 Mb, or 2.8%) genomes. We were able to identify 42,279 microsatellite loci, distributed in 400 non-redundant classes, in the T. rangeli genome sequence (Table S2). Approximately 4.7% (1,997) of these loci were found in the T. rangeli transcriptome [19] (Table S2). The microsatellite density and relative abundance in the T. rangeli genome assembly were estimated to be 38,678 bp/Mb and 3.87%, respectively. Interestingly, despite the relative abundance and the variation in the copy number of the 125 bp of satellite DNA observed in T. cruzi strains [62], these repeats were not found in the T. rangeli genome.
Transposable elements (TEs) represent a significant source of genetic diversity, and the fraction of particular genomes that correspond to TEs is highly variable [63]. Furthermore, TEs have been widely used as tools for genome manipulation as transgenic vectors or for gene tagging in organisms ranging from different microbes to mammals [64], [65], including the protozoan parasites Leishmania sp., Trypanosoma sp. and Plasmodium sp. [66]–[68]. In the genomes of the kinetoplastid protozoa analyzed thus far, only retrotransposon elements have been found. Trypanosomes retain long autonomous non-LTR retrotransposons ∼ ingi (T. brucei) and L1Tc (T. cruzi); site-specific retroposons SLACS (T. brucei) and CZAR (T. cruzi); and short nonautonomous truncated versions (RIME, NARTc), in addition to degenerate ingi-related retroposons with no coding capacity (DIREs) as also observed for L. major [60], L. infantum and L. braziliensis [61]. A long autonomous LTR retrotransposon, designated VIPER, has also been described in T. cruzi [56], [57]. L. braziliensis contains SLACS/CZAR-related elements and the Telomeric Associated Transposable Elements (TATEs) [61].
Intact copies and putative autonomous TEs were not found in the T. rangeli genome. However, we identified 96 remnants of retrotransposons, which are most closely related to those of T. cruzi. The LTR retrotransposon VIPER was present as 39 copies, the non-LTR retroposons ingi/RHS as 51 copies; L1TC, five copies; and a single copy of CZAR. In contrast to T. cruzi and T. brucei, which maintain autonomous elements, and L. braziliensis with intact TATE elements at chromosome ends, T. rangeli, L. major and L. infantum harbors only degenerate elements, suggesting that TEs have been selectively lost during the course of recent evolution.
Typically, a significant proportion of a trypanosomatid genome contains large families that encode surface proteins. Many of these proteins function as host cell adhesion molecules involved in cell invasion, as components of immune evasion mechanisms or as signaling proteins. We selected nine gene families that encode surface proteins present in T. cruzi, T. brucei and Leishmania spp. to search for orthologous sequences in the T. rangeli genome. Because the draft assemblies of the T. cruzi and T. rangeli genomes are still fragmented, we applied a read-based analysis to estimate the copy numbers of members of these families. Three single-copy genes that are known to have two distinct alleles in the T. cruzi CL Brener genome were also included in this analysis to validate our estimations. We found that the T. rangeli genome contains a smaller number of copies of three gene families, the MASPs, Mucins and Trans-sialidases, which are known to be present in far greater numbers in T. cruzi. Conversely, high copy numbers of amastin and kmp-11 are present in the T. rangeli genome compared to T. cruzi (Table 2).
T. cruzi amastins are small surface glycoproteins containing approximately 180 amino acids encoded by a gene family that has been subdivided into α-, β-, γ-, and δ-amastins and which are differentially expressed during the parasite life cycle [69], [70]. δ-amastins are mainly expressed by T. cruzi and Leishmania sp. intracellular amastigotes, a developmental stage that has not been observed during T. rangeli life cycle. Surprisingly, whereas T. cruzi has 27 copies of amastin genes, we estimate that 72 copies belonging to α-, β- and δ- amastin subfamilies are present in T. rangeli. Since the function of these proteins are still unknown, the study of their expression pattern and the significance of the expansion of this gene family in T. rangeli may shed new light into the role of these trypanosomatid specific surface glycoproteins.
Also in contrast to T. cruzi CL Brener strain, where forty alleles of genes encoding KPM-11 are present, there are 148 members in the KMP-11 in the T. rangeli genome. KMP-11 is a 92-amino acid antigen present in a wide range of trypanosomatids and is a target of the host humoral immune response against Leishmania spp. and T. cruzi infections, which, in the T. cruzi infection, induces an immunoprotective response [71]. The T. rangeli KMP-11 antigen shares 97% amino acid identity with its T. cruzi homologue [72]. These proteins are distributed in the cytoplasm, membrane, flagellum and flagellar pocket, most likely associated with the cytoskeleton of this protozoan [73]. The expansion of this family could have provided a selective growth advantage to T. rangeli in its insect vector. However, as a target for the immune response in mammals, it might have contributed to the poor pathogenicity of this organism.
The copy numbers of mucin glycoprotein-encoding genes, which are one of the largest and most heterogeneous gene families found in T. cruzi (TcMUC), are considerably reduced in T. rangeli. In T. cruzi, these surface glycoproteins cover the cell surface of several parasite stages and form a glycocalyx barrier [74]. Read coverage analysis of the region encoding the N-terminal conserved domain of the TcMUC family suggests the presence of only 15 copies in T. rangeli compared to 992 copies in T. cruzi. This finding is in agreement with the fact that only a few mucins were identified in the T. rangeli transcriptome [19], and only one TrMUC peptide was found through proteomic analysis [8]. In contrast to T. cruzi, T. rangeli lacks trans-sialidase activity, retaining only sialidase activity [75]. T. cruzi trans-sialidases (TS) are encoded by the largest gene family present in its genome. This enzyme catalyzes the transfer of sialic acid from sialylated donors present in host cells to the terminal galactose of mucin-glycoconjugates present at the parasite cell surface [76]. As a consequence of TS activity, in T. cruzi, large quanitities of multiple sialylated mucins form a protective coat when the parasite is exposed to the blood and tissues of the mammalian host. The relative paucity of the TrMUC repertoire correlates with the lower parasite load of T. rangeli in mammalian hosts and may in turn reflect the increased susceptibility to host immune mediators of T. rangeli compared with T. cruzi.
T. cruzi TS (TcTS) is a virulence factor integral to T. cruzi infection of the mammalian host [76], [77]. TcTS contains 12-amino acid repeats at the C-terminus, corresponding to the shed acute antigen (SAPA) [78], which is unnecessary for its activity but required for enzyme oligomerization and stability in the host [79]. This repeat is not present in T. rangeli sialidase sequences, and no T. rangeli proteins were detected in western blot assays using an anti-SAPA monoclonal antibody (unpublished results). In T. cruzi, TSs containing SAPA repeats are present only in infective trypomastigotes [80], while the TSs purified from epimastigotes lack the SAPA domain [81]. In addition to genes encoding the catalytic TS (subgroup Tc I), the trans-sialidase/sialidase superfamily in T. cruzi comprises eight subgroups, designated TcS I to VIII [82]. TcS group II encompasses proteins involved in host cell adhesion and invasion, and members of TcS group III display complement regulatory properties. The functions of the other groups are unknown, but all exhibit the conserved VTVxNVxLYNR motif, which is shared by all known TcS members [82], [83]. Sialidases/sialidase-like proteins similar to TcS groups I, II and III have been reported in T. rangeli [19], [84]–[86]. Here, we confirmed the presence of all TS subgroups in T. rangeli (Figure S1), although this parasite exhibits fewer members of the trans-sialidase/sialidase superfamily compared with T. cruzi (Table 2). It is therefore likely that all TS subgroups originated prior to the last common ancestor of the two species and that there was selective pressure in favor of the expansion and diversification of copies in T. cruzi. These observations also imply that the acquisition of SAPA repeats might have occurred after the appearance of the multiple gene family, when the T. cruzi ancestor gained mammalian infectivity, as proposed previously [81]. It has been suggested that the extensive sequence copy number expansion of the T. cruzi TS family could represent an immune evasion strategy driving the immune system to a series of spurious and non-neutralizing antibody responses [87]. It is tempting to speculate that the smaller number of copies of this large gene family found in T. rangeli could be related to the reduced virulence of this parasite in vertebrate hosts. Although, the expression of TS by both T. rangeli and T. brucei suggests a role for this enzyme during infections of the insect vector.
We identified 50 sequences in the T. rangeli genome encoding conserved domains of mucin-associated surface proteins (MASPs), which is fewer than that found in T. cruzi, in which the MASPs constitute the second largest gene family [57], [88]. Because MASPs are expressed at the surface of trypomastigotes and are highly polymorphic, the vast repertoire of MASP sequences present in the genome may contribute to the ability of T. cruzi to infect several host cell types and/or participate in host immune evasion mechanisms [89]. Changes in T. cruzi MASP family antigenic profiles during acute experimental infection have been established [89] and recent data has proposed a direct role for T. cruzi MASPs in host cell invasion (Najib El-Sayed, personal communication). Since T. rangeli lacks discernable ability to invade and multiply within the mammalian cells, the reduced repertoires of MASPs and of trans-sialidases in T. rangeli correlates may imply concerted action between these two groups of surface proteins during cell invasion and intracellular parasitism in T. cruzi.
African trypanosomes (T. brucei, T. congolense and T. vivax) are blood-living, extracellular parasites, having variable surface glycoproteins (VSG) as key elements required for immune evasion in these species [90]. As with T. cruzi, sequences related to the (VSG) could not be discerned through rigorous searches of the T. rangeli genome.
In some strains of T. rangeli, the epimastigotes are highly resistant to complement-mediated lysis [91]. In this context, genes showing similarity to gp160, a member of the large super-family of trans-sialidases identified as complement regulatory protein (CRP, or GP160) in T. cruzi [92], are found in the T. rangeli genome. However, their sizes are smaller than the corresponding T. cruzi genes, and considering the domain conservation observed in this family, their function as complement regulatory proteins remains unproven. Other T. cruzi molecules have been shown to confer resistance to complement-mediated lysis, such as calreticulin, GP58/68 and the complement C2 receptor inhibitor trispanning (CRIT) [93]. Our data showed that CRIT protein is absent in T. rangeli.
The mitochondrial genome of trypanosomes is a structure composed of concatenated large (maxi-) and small (mini-) circular DNAs. Minicircles are more abundant, comprising several thousand copies per genome, and are 1.6 to 1.8 kb long in T. rangeli. Minicircles encode gRNAs that are utilized in the editing of mitochondrial transcripts derived from maxicircle DNA, which are present at about 20 copies per genome. Minicircles exhibit heterogeneous and highly conserved regions [94]. Probes generated against conserved regions have been previously used as sensitive tools for discriminating T. rangeli and T. cruzi lineages [15].
We assembled the maxicircle of T. rangeli as a single contig of 25,288 bp. The length of this sequence is >10 kb longer than those sequenced from T. cruzi (Sylvio 15,185 bp, CL Brener 15,167 bp, Esmeraldo 14,935 bp). The maxicircle of T. cruzi marinkellei was found to be slightly longer (20,037 bp) than those of other T. cruzi strains. These length differences were attributed to variability of the repetitive region [59], [95]. Similarly, the T. rangeli maxicircle exhibits repetitive regions of ∼6 Kb that, along with non-coding regions, have increased the overall size by ∼15 Kb. The coding region of the T. rangeli maxicircle has maintained a high degree of synteny with that of T. cruzi (Figure S2). We found no in silico evidence of additional coding sequences outside this region. Transcripts from rRNA, cyb, coII and nadh were identified in the T. rangeli EST database [19].
Three chromosome ends were identified in the genome assembled in this study (Figure S3) corresponding to telomere ends. These sequences contain previously described structures found in the terminal region of T. rangeli telomeres, which is characterized by a specific telomeric junction sequence in T. rangeli (SubTr) separating the hexameric repeats from interstitial gene sequences [96], [97]. Although T. rangeli (SubTr) and T. cruzi (Tc189) telomeric junctions share very low sequence identity, related sequences have been identified in several intergenic regions in both protozoa (mainly between gp85 genes of the trans-sialidase superfamily), suggesting that the two structures could have a common origin. According to our analysis of the sequence immediately upstream of SubTr, two types of chromosome ends could be identified (Figure 4). In the first type, SubTr is preceded by a gp85/trans-sialidase gene/pseudogene, while the second exhibits a copy of the mercaptopyruvate sulfurtransferase gene. The presence of this single copy gene so close to the telomeric end of a chromosome in T. rangeli is interesting because it is absent at this location in T. cruzi telomeres where only pseudogenes belonging to multigene families have been found. Notwithstanding, the chromosome ends of T. rangeli differ from those of T. brucei and T. cruzi in that they exhibit a simpler homogeneous organization, with short subtelomeric regions [57]. The subtelomeric region extending between SubTr and the first internal (interstitial) chromosome-specific gene in the scaffolds analyzed here is quite short (∼5 kb) (Figure 4). Two of the analyzed scaffolds exhibit a high level of gene synteny with T. cruzi chromosome ends (CL Brener). However, this synteny is lost in subtelomeric regions due to the absence of interspersed “islands” of trans-sialidase, dgf-1 and rhs genes/pseudogenes in the chromosomes of T. rangeli (Figure 5) [55], [98]. Therefore, the differences in subtelomeric structure observed between T. rangeli and T. cruzi are consistent with the reduced number of repeated sequences found in the genome of the former and with the expansion of these sequences in the latter.
Although telomerase activity has not been reported in T. rangeli, a putative telomerase reverse transcriptase (tert) gene, along with an ortholog of a telomerase-associated protein (TEP1) gene were identified in the genome of this parasite. Taken together, the presence of the tert and tep1 genes and the lack of transposable elements or blocks of non-hexameric tandem repeat sequences at chromosome ends suggest that the maintenance of telomere length in T. rangeli is primarily due to telomerase activity.
Among the telomere-binding proteins, a putative TTAGGG binding factor (TRF2) homolog was identified in the T. rangeli genome. In T. brucei, TRF2 interacts with double-stranded telomeric DNA as a homodimer and is essential for maintaining the telomeric G-rich overhang [99]. Moreover, homologs of the RBP38/Tc38 and RPA-1 proteins, which are single-stranded DNA-binding factors involved in telomere maintenance mechanisms, and two other putative proteins (JBP1 and JBP2) participating in base J biosynthesis [100]–[102] were also detected in T. rangeli. Base J is a hypermodified DNA base localized primarily at telomeric regions of the genome of T. brucei, T. cruzi and Leishmania with elusive function. However, J in chromosome-internal positions has been associated with regulation of Pol II transcription initiation in T. cruzi [103], whereas in Leishmania sp. when present at the ends of long polycistronic transcripts, it was shown to be involved in transcription termination [104].
Most of the major components of the translation machinery found in other trypanosome and leishmania genomes are also found in T. rangeli (Table S3). In general, one copy of the genes encoding the aminoacyl-tRNA synthetases is present, except for glutaminyl-tRNA synthetase and aspartyl-tRNA synthetase, which display three copies each, and leucyl-tRNA synthetase, lysyl-tRNA synthetase, valyl-tRNA synthetase, tryptophanyl-tRNA synthetase, and seryl-tRNA synthetase, which exhibit two copies each. N-terminal mitochondrial targeting signals were also predicted in some of the deduced amino acid sequences of tRNA-synthetases from T. rangeli.
Compared to the other trypanosome genomes, similar numbers of genes encoding ribosomal proteins and other factors involved in translation were found in T. rangeli with some minor variation. For example, three copies of genes encoding eukaryotic initiation factor 5A were detected in T. rangeli, compared to two in T. cruzi and one in T. brucei. Only one copy of elongation factor 1-beta was identified in T. rangeli, compared to three in T. cruzi and T. brucei and there are eight paralogs of Elongation factor 1-alpha in T. rangeli that are similar to the paralogous expansion observed in T. cruzi, with eleven copies.
In many eukaryotes, RNA interference (RNAi) is a cellular mechanism for controlling gene expression in a sequence-specific fashion. This phenomenon has been described in a large number of organisms, including T. brucei, T. congolense, L. braziliensis and Giardia lamblia. It is, however, absent in many other trypanosomes, such as T. cruzi, L. major and L. donovani, and other protozoa, such as Plasmodium falciparum [45], [105]–[107]. Since the discovery of RNAi in T. brucei [108], a total of five major components of the RNAi machinery have been identified, including cytosolic (TbDCL1) and nuclear (TbDCL2) dicers, the Argonaute 1 (TbAGO1) protein, and two additional RNA Interference Factors, designated TbRIF4 and TbRIF5. It has been proposed that TbRIF4 acts in the conversion of double-stranded siRNAs into single-stranded form, and TbRIF5 functions as an essential co-factor for the TbDCL1 protein [109]–[112].
By searching for orthologs of components of the RNAi machinery in the T. rangeli genome using the T. brucei protein sequences as queries in tBLASTn analyses, we found that four of the five components of the T. brucei RNAi machinery are present in the T. rangeli genome as pseudogenes, as they exhibit one or more stop codons or frame shifts. To further evaluate whether these defective genes were a strain-specific phenomena restricted to the SC-58 strain, another strain representative of the northernmost distribution of the parasite was also assayed via PCR amplification and sequenced using Sanger sequencing chemistry. In addition to punctual differences among the strains, large deletions in T. rangeli ago1 and dcl1 were found (Figure S4). Among these five RNAi components, only Dicer-like 2 can be functional, since it contains insertions and deletions that do not cause frame-shifts or a premature translational stop. The T. rangeli Dicer-like 2 protein is 54 amino acids shorter in its N-terminal portion, exhibiting approximately 30% identity with T. congolense and T. brucei DCL2, with higher conservation in the RNaseIII domain (C-terminus) (Figure S5). The explanation for why only dcl2 was retained in the T. rangeli genome is unclear. However, it has been shown in T. brucei, that the dcl2 knockout cell line shows reduced levels of CIR147 (Chromosomal Internal Repeats – 147 bp long) and SLACS siRNAs (Spliced Leader Associated Conserved Sequence) and accumulation of long transcripts derived from retrotransposons (ingi and SLACS) [110]. This TbDCL2 knockout cell line also showed an increasing in the RNAi response to exogenous dsRNA. It is, however, difficult to speculate whether TrDCL2 plays a similar role in T. rangeli because the TbAGO1 ortholog is defective in this organism, and TbAGO1 knockout cells shows phenotype overlap compared to TbDCL2 -/- parasites [110].
Furthermore, a gene encoding a member of the AGO/PIWI family without the PAZ domain (conferring small RNA binding activity) was found in the T. rangeli genome (AUPL00000858). It encodes a protein of 1,083 amino acids that shares highest identity with T. cruzi (71% identical), followed by T. brucei (58%) and T. congolense (52%) throughout its entire sequence. This gene is present in the genome of all trypanosomatids, including RNAi-negative parasites, but its function is still unknown [113]. It may be that the protein encoded can work together with the TrDCL2 as part of an RNA metabolism pathway, but further work is needed to test this hypothesis.
In addition to re-sequencing PCR products corresponding to RNAi factors, the presence of a functional RNAi mechanism was investigated through transient transfections of a siRNA targeting eGFP, or a plasmid that can drive the expression of a long dsRNA targeting endogenous β-tubulin and a fluorescent marker (red fluorescent protein). In agreement with the in silico analysis, the transfection of eGFP-expressing cells or wild type parasites with the siRNA (Figure 6) or a plasmid encoding tubulin dsRNA, respectively, failed to inhibit eGFP expression or alter the parasite's morphology, which suggests an absence of a functional RNAi machinery in T. rangeli.
The T. rangeli genome encodes 151 eukaryotic protein kinases (ePKs), which corresponds to 1.94% of the total coding sequences in the genome. Like other trypanosomatids, T. rangeli lacks members of the protein tyrosine kinase (PTK), tyrosine kinase-like (TKL) and receptor guanylate cyclases (RGC) groups. T. rangeli displays some ePKs with predicted transmembrane domains, including nine genes, in addition to five with a signal peptide (Table S4).
The protein kinases of eukaryotes are subdivided into 8 groups according to the nomenclature of Miranda-Saavedra and Barton (2007) [114] and KinBase (http://www.kinase.com/kinbase/). In the T. rangeli genome, the largest group is “Other” (kinases that could not be assigned to a specific group), with 40 members, followed by the CMGC (cyclin-dependent kinases, mitogen-activated protein kinases, glycogen synthase kinase 3 and CK2-related kinases) group, with 30 members, two of which are catalytically inactive. The least represented group is the casein kinases (CK1), with only two members. The other groups display 26 members in AGC (Protein kinase A, G and C families), 22 members in CAMK (Calcium and Calmodulin-regulated kinases) and 31 members in STE (Kinases related to MAPKs activation).
The phosphatidylinositol kinases (PIK) and PIK-related proteins of T. rangeli are described in Table S5. These are lipid kinases that play a key role in a wide range of cellular processes, such as cell growth and survival, vesicle trafficking, cytoskeletal reorganization and chemotaxis, cell adhesion, superoxide production and glucose transport [115]. Like T. cruzi [38], T. rangeli lacks a tor-like 2 gene, although a truncated version of this gene without the catalytic domain has been identified. The accessory domains of the PIK-related families of both T. cruzi and T. rangeli can be seen in Table S6.
In addition, T. rangeli possesses four phosphatidylinositol phosphate kinases (PIPK), which have not been evaluated in other trypanosomatids as yet, including in T. cruzi. These kinases phosphorylate already-phosphorylated phosphatidyl inositols to form phosphatidylinositol bisphosphates. The PIPK functions have been mainly established for mice and humans, which include vesicular trafficking, membrane translocation, cell adhesion, chemotaxis, the cell cycle and DNA synthesis [116].
Genes that encode most of the proteins responsible for DNA repair and recombination mechanisms in other trypanosomatids were also found in T. rangeli, suggesting that this protozoan displays all of the known functional DNA repair pathways. In other organisms it has been demonstrated that errors generated during DNA replication can be corrected via DNA mismatch repair, involving the recruitment of heterodimers of MSH2 and MSH3 or MSH6, which signalize MLH1 and PMS1 binding [117]. Homologs of these proteins are present in T. rangeli, but in common with other trypanosomatids, no homolog of PMS2 was found [56], [57], [60]. Different DNA base modifications can be corrected via base excision repair [118]. Sequences encoding the OGG1, UNG and MUTY DNA glycosylases were identified. However, whether the long and short pathways are functional is a question that remains to be answered because important homologs, such as LIG3, XRCC1 and PARP, are missing. Lesions that alter DNA conformation can be repaired through nucleotide excision repair (NER) [119], and as with other trypanosomatids, T. rangeli contains sequences encoding most of components of the NER pathway, including proteins constituting the TFIIH complex. It has been shown that in T. brucei, two trypanosomatid-specific subunits of TFIIH (TSP1 and TSP2) are important for parasite viability because they participate in the transcription of the splice-leader gene [120]. Both proteins are also present in T. rangeli, as well in T. cruzi and L. major.
DNA recombination is an essential process involved in DNA repair and in the generation of genetic variability in these parasites. No major differences in genes encoding components of DNA recombination machinery were observed between T. rangeli and other trypanosomatids [121]. They all exhibit genes encoding MRE11, RAD50, KU70 and 80, BRCA2 and RAD51, which play important roles in homologous recombination (HR) and non-homologous end joining (NHEJ). However, T. rangeli lacks homologs of DNA Ligase IV and XRCC4, like other trypanosomatids, indicating that it does not exhibit a functional NHEJ [122].
Several antioxidant enzymes work sequentially in different sub-cellular compartments to promote hydroperoxide detoxification (Table S7) [123]. During its life cycle, T. rangeli is exposed to reactive oxygen species (ROS) in its triatomine vectors and possibly in its mammalian host. ROS are generated through oxidative metabolism and oxidative bursts in the host immune system [124]. Interestingly, epimastigotes of T. rangeli (SC-58 strain) are 5-fold more sensitive to hydrogen peroxide (H2O2) than T. cruzi (Y strain) forms, with IC50 values of 60 µM±2 and 300 µM±5, respectively (Figure 7). It has been reported that the membrane-bound phosphatases of T. rangeli are more sensitive to the addition of sublethal doses of H2O2 than T. cruzi phosphatases [125].
In trypanosomatids, the major antioxidant molecule is a low molecular weight thiol trypanothione, which maintains the intracellular environment in a reduced state, essentially through the action of trypanothione reductase [126]. Trypanothione is a conjugate formed in two-steps via the bifunctional enzyme trypanothione synthetase (TRS) using two glutathione molecules and one spermidine. Two genes coding to trypanothione synthetase, and one to trypanothione synthetase-like are present in T. rangeli. Considering the substrates, glutathione synthesis is observed in T. rangeli, as in all trypanosomatids, despite the absence of de novo cysteine biosynthesis [127]. However, while in T. brucei, Angomonas fasciculata and Leishmania spp., the spermidine is synthesized from ornithine and methionine; in T. cruzi, the key enzyme ornithine decarboxylase (ODC) is absent, and the parasite solely depends on polyamine uptake by transporters to synthesize trypanothione. The odc gene is not present in T. rangeli, suggesting that this parasite also requires exogenous polyamines [128].
Trypanothione reductase (TR), a key enzyme involved in antioxidant defense in trypanosomatids, is present in T. rangeli and shares 84% identity with the T. cruzi enzyme at the amino acid level. Trypanothione is maintained in its reduced form (T-SH2) by the action of trypanothione reductase and the cofactor NADPH [126]. The reactions of the trypanothione cycle are catalyzed by tryparedoxin peroxidase (TXNPx) and ascorbate peroxidase (APX), which are responsible for the subsequent detoxification of H2O2 to water [126]. These enzymes use tryparedoxin and ascorbate as electron donors, respectively, which are in turn, reduced by dihydrotrypanothione.
As with other trypanosomatids, T. rangeli produces superoxide dismutase (SOD), an enzyme that removes excess superoxide radicals by converting them to oxygen and H2O2 [129]. Three Fe-sod genes were found in T. rangeli: Fe-sod-a, Fe-sod-b and a putative Fe-sod, sharing 90%, 88% and 84% identity with T. cruzi Fe-sod genes, respectively. Additionally, as with to T. cruzi, T. rangeli exhibits genes encoding distinct TXNPx proteins, including one cytosolic, one mitochondrial and one putative TXNPx sequence. Both enzymes possess two domains that are common to subgroup 2-Cys, and is present in antioxidant enzymes from the peroxiredoxin family [130]. The T. rangeli genome also contains two glutathione peroxidases (gpx), which act as antioxidants by reducing H2O2 or hydroperoxides with a high catalytic efficiency in different cellular locations [128]. In addition, enzymes related to sensitivity of nifurtimox or benzonidazol were identified in T. rangeli, including nitroreductase and prostaglandin F2 synthetase.
An ortholog of the ascorbate peroxidase gene from T. cruzi (apx) is present as a pseudogene in T. rangeli, as it exhibits a premature stop codon or frame shifts. Interestingly, this enzyme, which is a class I heme-containing enzyme, is present in photosynthetic microorganisms, plants and some trypanosomatids, such as Leishmania spp. and T. cruzi, but is absent in T. brucei [131]–[133]. In T. cruzi, ascorbate peroxidase and glutathione-dependent peroxidase II metabolize H2O2 and lipid hydroperoxides in the endoplasmic reticulum. It can be speculated that the higher sensibility of T. rangeli to H2O2 compared to T. cruzi could be related to the absence of ascorbate peroxidase activity. Proteomic analyses conducted in T. cruzi have demonstrated upregulation of components of the parasite antioxidant network during metacyclogenesis, including TcAPX, reinforcing the importance of the antioxidant enzymes for successful infection [134], [135]. Wilkinson et al. [136] suggested that T. brucei may not require ascorbate-based antioxidant protection because, as an extracellular parasite, it is not exposed to the oxidative challenge from host immune cells produced in response to intracellular infection of T. cruzi or Leishmania spp. Thus, the limited capability of T. rangeli to respond to oxidative stress could be related to the inability of this parasite to infect and multiply inside vertebrate host cells. This observation may suggest a distinct replication site for this parasite in the mammalian host, similar to the extracellular cycle of T. brucei.
In Table S8, the genes encoding the stress response proteins of T. rangeli are presented. A large set of heat shock protein genes is found in the genome of this parasite, occasionally displaying a reduced copy number compared with T. cruzi. Similarly to T. cruzi, the T. rangeli genome contains 17 hsp70 genes, 13 of which are cytosolic, while 3 are mitochondrial, and one localized to the endoplasmic reticulum. On the other hand, only one hsp85 and hsp20 genes were found in the T. rangeli genome, compared to 6 and 11 copies in T. cruzi, respectively. The large number of hsp40 genes observed in kinetoplastids (68 copies in T. cruzi) [137] is also reduced in T. rangeli (24 copies).
Thus, where the reduced repertoire of transialidases and MASPs may correlate with diminished ability to enter mammalian cells, it can be speculated that the reduced number of genes related to different cellular stress responses provides for a more limited capability of T. rangeli to respond to oxidative stress and that this in turn corresponds with an apparent inability to survive and multiply within mammalian cells.
At 24 Mb (haploid), the T. rangeli genome is the shortest and least variable genome from the mammalian-infective trypanosomatids to date. Our elucidation of its sequence both answers and poses a variety of intriguing questions about the biology of a trypanosome which is infectious but non-pathogenic to humans and which is carried by triatomine bugs and sympatrically distributed with T. cruzi, but which shows a salivarian rather than a stercorian route for infection. Based on phylogenomic analysis, T. rangeli is undoubtedly positioned as a stercorarian parasite, chromosome structure and progressive loss of RNAi machinery in this lineage lend support to this interpretation and the results presented here corroborate previous results based on distinct nuclear and mitochondrial markers. The different evolutionary path of this trypanosome species is, though, writ large on its genome by a differential in the preponderance of gene duplication and divergence, particularly at the telomeres, with reduced diversity in genes known to be associated with infection of the mammalian host such as transsialidases, MASPs and oxidative stress and rather more diversity in other non-telomeric gene families such as KMP-11s and amastins which may imply roles for these families in vector interactions. It is interesting to consider to what extent the T. rangeli-Rhodnius vector species co-evolution of salivary gland colonization (and anterior transmission) is an example of parallel or convergent evolution with the colonization of the tsetse salivary gland by African trypanosomes, and to what extent the apparatus for this phenotype was already present in a progenitor. Our release of the T. rangeli genome casts further light on the evolutionary origins and relationships of trypanosomes, and provides a resource for better understanding the function of genes and factors related to the virulence and pathogenesis of trypanosomiasis and with which to address unknown aspects of the T. rangeli life cycle in mammalian hosts.
|
10.1371/journal.pbio.0060224 | A Genome-Wide RNAi Screen to Dissect Centriole Duplication and Centrosome Maturation in Drosophila | Centrosomes comprise a pair of centrioles surrounded by an amorphous pericentriolar material (PCM). Here, we have performed a microscopy-based genome-wide RNA interference (RNAi) screen in Drosophila cells to identify proteins required for centriole duplication and mitotic PCM recruitment. We analysed 92% of the Drosophila genome (13,059 genes) and identified 32 genes involved in centrosome function. An extensive series of secondary screens classified these genes into four categories: (1) nine are required for centriole duplication, (2) 11 are required for centrosome maturation, (3) nine are required for both functions, and (4) three genes regulate centrosome separation. These 32 hits include several new centrosomal components, some of which have human homologs. In addition, we find that the individual depletion of only two proteins, Polo and Centrosomin (Cnn) can completely block centrosome maturation. Cnn is phosphorylated during mitosis in a Polo-dependent manner, suggesting that the Polo-dependent phosphorylation of Cnn initiates centrosome maturation in flies.
| A major goal of the cell cycle is to accurately separate the duplicated chromosomes between two daughter cells. To achieve this, a pair of centrosomes organise a bipolar spindle made of microtubules; the chromosomes line up on the spindle and are then separated to the two spindle poles. Centrosomes are also required for the formation of cilia and flagella, which are present in many eukaryotic cells; centrosome dysfunction is a common feature of many human cancers and several neurological disorders, whereas mutations in genes that affect cilia function give rise to several human diseases. Here, we perform a genome-wide screen using RNA interference to try to identify all of the proteins required for centrosome function in the model organism Drosophila melanogaster (a fruitfly). We identified all 16 of the centrosomal proteins that were already known to be required for centrosome function in Drosophila, as well as 16 new centrosomal components or regulators. We confirmed the centrosomal location of several of the components and performed some analysis of their functions. We believe we are approaching a complete inventory of the proteins required for centrosome function in flies.
| In most cells, the centrosome functions as the major microtubule (MT) organising centre (MTOC), and, as such, it has been implicated in organising many cellular processes, including vesicle transport, cell polarity, cell migration, and cell division [1,2]. There is also evidence that centrosomes have essential roles within the cell that are independent of their ability to organise MTs [3,4]. Indeed, many key regulators of cellular physiology, such as those required for cell cycle progression, cell signalling, and DNA damage response pathways, are concentrated at centrosomes, suggesting that the centrosome functions as a scaffold where many regulators meet and coordinate their response to various events in the life of the cell [5].
Centrosomes consist of a centriole pair surrounded by pericentriolar material (PCM). At the end of mitosis, the two centrioles disengage to allow duplication in the next cell cycle [6]. Subsequently, new centrioles are formed perpendicular to the mother centrioles in S-phase. As cells enter mitosis, the centrioles recruit PCM (a process termed centrosome maturation), and many MT nucleation and anchoring factors concentrate at the centrosomes as they form the poles of the mitotic spindle [5]. In addition to their function in organising the centrosome, centrioles also form the basal bodies present at the base of cilia and flagella, and cilia have been shown to have a variety of essential functions in development [7].
Centrosome amplification is a common feature of many cancers, and this has been linked to genetic instability, which is widely believed to be an important driver of tumourigenesis [8–12]. Furthermore, mutations in several human centrosomal proteins cause primary autosomal microcephaly, in which patients are born with small brains [13,14]. The reason for this phenotype is unclear, but it is postulated that centrosomes play a particularly important role during the asymmetric cell division of neural stem cells [15], and this is certainly the case in flies [16]. Finally, defects in cilia function have been identified as the cause of several human syndromes such as Bardet-Biedl syndrome (BBS) and Kartagener's syndrome, which lead to relatively pleiotropic defects during the development of affected individuals [17,18].
Although more than one hundred proteins are concentrated at centrosomes [5,19], it is unclear how these proteins are assembled into a functional unit, and how many of these proteins are actually required for centrosome function. Traditional genetic screens and genome-wide RNA interference (RNAi) screens in the early Caenorhabditis elegans embryo have identified just four proteins that are essential for centriole duplication (ZYG-1, SAS-4, SAS-5, and SAS-6), three that are essential for the recruitment of the PCM to the centrioles during mitosis (SPD-5, Protein Phosphatase-4 [PP-4], and the Aurora A kinase [AIR1]), and one that appears to have a role in both processes (SPD-2) [20–26]. Thus, a surprisingly small number of proteins appear to be essential for these “core” centrosomal functions in worms. Experiments in other systems, however, have identified many additional proteins that appear to have a role in centrosome maturation and/or centriole duplication ([5] and references therein; [27–35]). As the initial genome-wide screens in worms were not specifically designed to identify proteins required for centrosome function, it remains unclear how many proteins are required for the key functions of centriole duplication and centrosome maturation.
Here, we have performed a genome-wide RNAi screen in Drosophila tissue culture cells (S2R+) designed to identify proteins required for centriole duplication and centrosome maturation. After an extensive series of localisation studies and secondary screens, we have identified just 32 proteins that are required for these core centrosomal processes. Importantly, this screen successfully identified every Drosophila protein that had previously been implicated in centriole duplication and/or centrosome maturation, as well as several new factors, some of which have been implicated in centrosome function in other systems, and some of which are novel proteins that we confirm are components of the centrosome. Thus, we believe we are approaching a near-complete inventory of proteins required for these processes in flies. Finally, we noticed that only the depletion of either Polo kinase or Centrosomin (Cnn) could completely suppress centrosome maturation, indicating that they are major players in this process. We show that Cnn is phosphorylated exclusively during mitosis in a manner that is dependent on Polo kinase, and that these two proteins are codependent for their localisation at centrosomes. This suggests that the Polo-dependent phosphorylation of Cnn plays a crucial part in initiating centrosome maturation in flies.
We devised a microscopy-based screen to search for proteins required for centriole duplication and centrosome maturation (Figure 1A and 1B). We used a library of double-stranded RNAs (dsRNAs) targeted against 13,059 individual genes (approximately 92% of all predicted protein-coding genes in Drosophila melanogaster) to deplete individual proteins in S2R+ cells. Treated cells were grown for 4 d in 384-well plates, then incubated with colchicine to depolymerise the MTs and arrest cells in mitosis for 8 h prior to fixation. The colchicine treatment increased the number of mitotic cells to facilitate the analysis, but did not interfere with centrosome maturation, which occurs robustly even in the absence of centrosomal MTs (Figure 1C). Cells were then fixed and processed for immunofluorescence microscopy with antibodies raised against phospho-histone H3 (p-H3) to identify mitotic cells and Cnn to label the PCM. The colchicine arrest often prevented proper centrosome separation and resulted in a mix of mitotic cells with one or two centrosomes (1.2–1.5 centrosomes per mitotic cell on average—see Materials and Methods).
We used anti-Cnn antibodies in our screen because Cnn is a very robust PCM marker, but also because Cnn appears to be a very general centrosome maturation factor: in its absence, the centrosomal recruitment of every other PCM component that has been tested is severely compromised during mitosis [36–38]. Thus, proteins that cause defects in the mitotic recruitment of Cnn to centrosomes are also likely to be general recruitment factors that are required for the proper recruitment of many other PCM components. Moreover, we reasoned that this screen would also identify proteins that are required for centriole duplication, as the PCM only assembles on the centriole scaffold in flies (Figure 1C) [16]. Thus, a reduction in centriole numbers would lead to fewer Cnn dots being observed and would therefore be detected in our screen.
In S2R+ cells, anti-Cnn antibodies only label centrosomes during mitosis (Figure 1C), as is true in many Drosophila cells in vivo [39,40]. The number of Cnn dots per mitotic cell was used as our readout in the primary screen. We quantified the number of centrosomes per mitotic cell after the depletion of individual proteins in three different ways (Figure 1A). First, each well of RNAi-treated cells was examined manually on a fluorescence microscope. Second, digital images of four fields of cells (typically containing more than 50 mitotic cells/field) from each well were acquired automatically and analysed manually. Third, these digital images were used to automatically count the number of centrosomes in each mitotic cell using CellProfiler [41] (see Materials and Methods). All of these analyses were performed “blind.” In this way, we identified 119 genes whose depletion significantly decreased or increased the average number of centrosomes per mitotic cell (Figure 1C and Table S1).
We performed an extensive series of secondary screens with 79 of these initial 119 hits. We used several criteria to exclude 40 genes that we thought less likely to be of interest for further analysis (see Table S1 and Materials and Methods), although we cannot exclude the possibility that some of these genes play a role in centriole duplication and/or centrosome maturation. We synthesised new, nonoverlapping, dsRNAs against these 79 genes (Table S3), and repeated the screen in both 384-well and 96-well formats with a 20× objective, but this time we examined the centrosomal localisation of Cnn, γ-tubulin, and DSpd-2 in both colchicine-, and noncolchicine-treated cells. All experiments were performed in triplicate to ensure the robustness of our screening procedures. Only 39 of the 79 genes tested were confirmed as positive hits after this analysis (Table S1). These 39 genes were then further tested in a set of functional assays that were specifically designed to distinguish whether individual proteins were required for centriole duplication, centrosome maturation, or both. We analysed the depletion of these 39 proteins in 24-well plates with a 63× objective using markers to detect centrioles (DSas-4), PCM (Cnn, DSpd-2, and γ-tubulin), and mitotic spindles (α-tubulin).
This analysis gave a final list of 32 genes whose depletion gave highly reproducible centrosome defects (Tables 1–4). For simplicity, we named any of these genes that had not previously been named, or that did not have homologs in other systems that had been assigned a function, Rcd proteins for “Reduction in Cnn Dots.” The 32 proteins were classified into four groups (Figure 1B; Tables 1–4). Nine proteins appeared to be required primarily for efficient centriole duplication (Class I, Figure 1B). The depletion of these proteins led to a reduction in the number of centrioles and centrosomes per cell, but in those cells that retained centrioles, the recruitment of the PCM was largely unperturbed (Figures 2 and S1). Nine proteins appeared to be required for both efficient centriole duplication and efficient PCM recruitment (Class II, Figure 1B). The depletion of these proteins led to a reduction in the average number of centrioles per cell, and in those cells that retained centrioles, the recruitment of the PCM to the centrioles was also reduced (Figures 3 and S1). Eleven proteins appeared to be required primarily for the efficient recruitment of the PCM to the centrioles (Class III, Figure 1B). The depletion of these proteins had only a minor effect on the average number of centrioles per cell, but significantly reduced the amount of PCM that was recruited to the centrioles (Figures 4 and S1). Finally, three proteins appeared to be required for centrosome separation (Class IV, Figure 1B). The depletion of these proteins led to an apparent reduction in the average number of centrosomes per cell, but staining with the centriole marker revealed that this was due to the clustering of several centrioles (Figure 5).
To quantitate the defect in Cnn recruitment in cells depleted of each of these 32 proteins, we took optical sections through the entire cell volume and measured total centrosomal Cnn intensity. The average centrosomal intensity was measured in three independent depletion experiments (Figure S1—note that we typically analysed a total of ∼100 centrosomes in total, but in cases where centriole numbers were dramatically reduced, we could analyse only 20–40 centrosomes in total). Virtually all of the proteins classified as being required exclusively for PCM recruitment (Class III) showed a statistically significant decrease in the recruitment of Cnn to centrioles, but this was not true for any of the proteins classified as having a defect in only centriole duplication (Class I), strongly supporting the robustness of our scoring procedures. The proteins classified as being required for both PCM recruitment and centriole duplication (Class II), however, showed an intermediate phenotype: in eight of nine cases, the recruitment of Cnn was less than that seen in controls, but in only three cases was this difference statistically significant (Figure S1). As we consistently scored these proteins as having a defect in PCM recruitment in multiple experiments with multiple PCM markers, we suspect that this reflects the fact that the defect in PCM recruitment is more subtle in this class, and we would need to assay larger numbers of centrosomes to show statistical significance (see Discussion).
The nine proteins we identified as being required for centriole duplication included the three proteins already known to be essential for this process in flies (DSas-4, DSas-6, and Sak/Plk4) as well as three proteins implicated in centriole duplication on the basis of their anastral spindle phenotype when depleted from S2 cells (Ana1–3) [42]. We created stable S2 cell lines expressing green fluorescent protein (GFP) fusions to Ana1 and Ana2 (we had difficulty in cloning full-length Ana3) under the control of either the metallothionein or ubiquitin promoter and found that they both localised to centrioles when expressed at low levels, as described previously [42] (Protocol S1, pages 5 and 11; and Table 1). When expressed at higher levels, Ana1 and 2 formed extra dots (usually 5–10) in the cytoplasm, a feature shared with the overexpression of GFP fusions to DSas-4, DSas-6, and Sak [43,44] (Figure S4). This suggests that, like these core centriole duplication proteins, Ana1 and Ana2 are structural components of the centriole required for efficient centriole duplication.
The three remaining proteins in this class have not previously been implicated in centriole duplication. Rcd1 (CG8233), Rcd2 (CG4786), and emb (CG13387) all have human homologs that have been implicated in various processes (Table S5), but none of these proteins were detectable at centrioles in stable cell lines expressing GFP fusions to any one of these proteins; instead these fusions localised to the nucleus, the plasma membrane, and nuclear membrane, respectively (Protocol S1, pages 6, 7, and 9; and Table 1). Thus, although it is possible that GFP-tagging disrupts the centriolar localisation of one or more of these proteins, it seems likely that they influence centriole duplication indirectly.
The nine proteins identified as being required for both centriole duplication and PCM recruitment include four that have previously been implicated in centriole/centrosome function either in flies or in other systems (Table 2). Asterless (Asl; CG2919) is a centrosomal protein previously shown to be required for efficient PCM recruitment in flies, and it is related to the human centrosomal protein Cep152 [45] (Table S5). Asl-GFP localised to both centrioles and the PCM, as shown previously [45]; as with the overexpression of DSas-4, DSas-6, Sak, Ana1, and Ana2, its overexpression led to the formation of extra dots in cells (Figure S4A and S4B). Thus, we conclude that Asl is required for both centrosome maturation and centriole duplication in flies.
CG17081 is the fly homolog of human Cep135, CG14617 is the fly homolog of human CP110, and CG3980 is the fly homolog of Cep97; all of these proteins have been implicated in centriole duplication and PCM recruitment in humans [19,27,29,46]. We found that when expressed at low levels, GFP fusions to Drosophila Cep135 (DCep135) and Drosophila CP110 (DCP110) were concentrated at centrioles; interestingly, however, high-level overexpression of either protein led to the formation of fibre-like structures in the cytoplasm, most prominently in the case of DCep135 (Figures S4A; Protocol S1, pages 14 and 18). In contrast, a GFP fusion to Drosophila Cep97 (DCep97) localised to centrosomes specifically during mitosis (Figures 6D; Protocol S1, page 13; and Table 2). Together, these findings indicate that these four proteins are very likely to play a direct role in centriole duplication and/or centrosome maturation (see Discussion).
Two of these nine hits, Myb and Rcd5 (CG1135), were recently found in a screen to identify proteins involved in mitotic spindle function, but their exact defects were not characterized [42]. Myb is a transcription factor that has a variety of cell cycle–related functions [47], but GFP-Myb fusions did not detectably localise to centrioles or centrosomes, suggesting Myb's role at centrosomes may be indirect (Protocol S1, page 20; and Table 2). Interestingly, it has recently been shown that perturbing Myb function leads to a reduced Polo levels, perhaps explaining its influence on PCM recruitment [48].
Interestingly, Rcd5 (CG1135) was unique amongst all of the proteins we analysed in that it had only a slight effect on the amount of Cnn recruited to centrioles during mitosis (and it was picked up in our original screen primarily because of the reduction in the number of centrioles in depleted cells), but the amount of γ-tubulin and DSpd-2 recruited to centrosomes was more dramatically decreased, hence the inclusion of Rcd5 in Class II (Protocol S1, page 19; and Table 2). Thus, Rcd5 may act downstream of Cnn in the pathway that leads to DSpd-2 and γ-tubulin recruitment. GFP fusions to this protein were not, however, detectably concentrated at centrosomes (Protocol S1, page 19; and Table 2).
None of the three remaining proteins in this class have previously been implicated in centriole duplication or centrosome maturation. Calmodulin, however, has been implicated in targeting several proteins to centrioles and centrosomes, including CP110 [28], and a GFP-calmodulin fusion protein localised to centrosomes and spindles specifically during mitosis (Protocol S1, page 21). Rcd4 (CG17295) is not obviously related to any protein outside of insects, but GFP fusions to Rcd4 strongly localised to centrioles (Figure 6; Protocol S1, page 16). Thus, these two proteins are likely to have direct roles in centriole function. Rcd3 (CG8231) is homologous to the human T-complex protein 1 subunit zeta, which is needed for proper tubulin folding [49], so the observed defects are probably indirect.
The 11 proteins required for centrosome maturation (Table 3) include five of the six proteins that have previously been implicated in this process in flies: Cnn [36–38], Polo [50], DSpd-2 [51,52], D-PLP [39], and γ-tubulin [53]. The only protein of this type that we did not identify in our screen was Aurora A [54], which we found to be required for centrosome separation, but which is probably also required for PCM recruitment (see below).
Of the six remaining proteins in this class, Grip71WD is a centrosomal protein that is homologous to GCP-WD/NEDD1 in humans. Although it was thought not to be required for PCM recruitment in flies [55], Grip71WD has been implicated in PCM recruitment and centriole duplication in humans [31,56]. Our data suggest that this protein has some function in centrosome maturation flies. The MT-binding protein Map205 is localised to centrosomes and MTs [57] (Figure 6; Protocol S1, page 26), but null mutants in this gene are viable and fertile [57], demonstrating that its function is not essential in flies. Rcd6 (CG11175) is predicted to encode a transmembrane protein, and GFP fusions were predominantly localised to the plasma membrane, suggesting that any role in centrosome maturation is indirect (Protocol S1, page 29).
Surprisingly, the three remaining proteins in this class encode the catalytic subunit (mts), a regulatory subunit (tws), and a structural subunit (PP2A-29B) of the protein phosphatase PP2A, thus providing compelling evidence that this enzyme is essential for efficient PCM recruitment in flies. Components of PP2A are associated with centrosomes in human cells [19], and with the centrosome equivalents in fission yeast and Dictyostelium [58,59], but GFP fusions to any of these fly proteins were not detectably concentrated at centrosomes in our hands (Protocol S1, page 30; and Table 3). Although PP2A activity is required for many cell processes, this form of PP2A (PP2Atws) seems to be the only one that is essential for centrosome maturation; we tested the effect of depleting the three other PP2A regulatory subunits either individually, or in all combinations, and found that none of these were required for efficient centrosome maturation (J. Dobbelaere, unpublished data).
To our surprise, Aurora A, together with the ubiquitin E2 ligase UbcD6 and the protein of unknown function Rcd7 (CG14098), were recovered in our screen as being required for centrosome separation (Table 4). These proteins were picked up in our primary screen because they were originally scored as having too few centrosomes per cell (Table S1). Our secondary screening revealed, however, that cells depleted of these proteins appeared to have too few centrosomes because they had not separated properly (Figure 5). Although Aurora A has previously been implicated in PCM recruitment in worms and flies [54,60], a centrosome clustering phenotype has been described previously in flies [61], and we confirmed that this is the dominant phenotype we observed in aurora A mutant larval brain cells (Figure S2). It seems likely, however, that this centrosome separation defect masks a role for Aurora A and UbcD6 in PCM recruitment, as the single centrioles that we observed in these cells were found to recruit less PCM than normal (Figures 5 and S1).
From our analysis of all the proteins we identified as being required for efficient centrosome maturation, it was clear that the depletion of Cnn or Polo had a significantly stronger effect on this process than the depletion of any other protein (Figures 7A and S1, and Protocol S1, pages 23 and 24—note that for Cnn, this was judged by the strength of its effect on the centrosomal recruitment of γ-tubulin and DSpd-2). This suggests that these two proteins have a particularly important role in centrosome maturation in flies. Since Polo is known to localise to centrosomes in mitosis [50,62] (Figure 7), we tested whether Polo might initiate centrosome maturation by phosphorylating Cnn. Western blotting experiments revealed that Cnn was indeed phosphorylated specifically during mitosis, and that this phosphorylation was dependent on Polo, but not on the centrosomal kinases Aurora A or Sak/Plk4 (Figure 8). Moreover, Cnn and Polo exhibited a reciprocal dependency for their localisation at centrosomes: Cnn was essentially undetectable at centrosomes in cells depleted of Polo, whereas Polo and activated Polo—detected with antibodies raised against Polo phosphorylated on the activating T210/T182 (in humans and flies, respectively)—were undetectable at centrosomes in cells depleted of Cnn (Figure 7B and 7C). These observations raise the intriguing possibility that it is the Polo-dependent phosphorylation of Cnn that initiates centrosome maturation in flies.
In this study, we set out to identify proteins required for centriole duplication and centrosome maturation in Drosophila S2R+ cells. As well as recovering all known Drosophila proteins that had previously been implicated in these processes, we identified several fly homologs of centrosomal proteins previously identified in other systems, and several new proteins that had not previously been implicated in centrosome function, some of which have homologs in other systems. We show that several of these new proteins are centrosomal components, indicating that they probably have a direct role in centrosome function.
One surprising aspect of our results was the identification of a relatively large number of proteins (nine) that appear to be required for both centriole duplication and centrosome maturation (Table 2). It is unclear, however, whether these proteins have separate functions in these processes. Previous studies in worms and human cells have revealed that although centrosome maturation is not essential for centriole duplication, the recruitment of at least some PCM components to the centrioles is required for this process to occur efficiently [63–65]. Thus, although the proteins we identify in this class do not have a particularly strong defect in centrosome maturation (compared to Cnn and Polo, for example, which have stronger defects in centrosome maturation, but no defects in centriole duplication), it may be that these proteins play a particularly important part in recruiting a small amount of PCM to the centrioles during S-phase, and that this is required for efficient centriole duplication. Alternatively, some or all of these proteins may only be required for efficient centriole duplication, but their partial depletion may lead to the formation of defective centrioles that no longer efficiently recruit PCM. Further investigation will reveal how these proteins regulate these two processes, but it is clear that Asl/Cep152, DCep135 (CG17081), DCP110 (CG14617), DCep97 (CG3980), Rcd4 (CG17259), (which so far has no homolog outside of insects), and calmodulin are all centrosomal components that are required for efficient centriole duplication and/or efficient PCM recruitment in fly cells.
Studies in worm embryos have identified just five proteins that are required for centriole duplication, and these have been ordered into a functional pathway: SPD-2 recruits the kinase ZYG-1, which recruits SAS-5 and SAS-6, which in turn recruit SAS-4 [20–26]. Proteins related to ZYG-1, SAS-6, and SAS-4 are required for centriole duplication in several other systems, and it has been postulated that these five proteins constitute a conserved “core” centriole duplication machinery [66]. Previous studies in fly cells suggested that three additional proteins (Ana1–3) may also be required for centriole duplication (inferred from a lack of astral MTs in spindles and absence of γ-tubulin at the poles when these proteins were depleted), and Ana1 and Ana2 were shown to localise to centrioles [42]. We have confirmed these results and extended them by directly showing that centriole numbers decrease in cells depleted of Ana1–3. Further experiments will be required, however, to determine whether these proteins are part of the conserved “core” centriole duplication machinery.
It is worth noting that whereas SPD-2 is a key initiator of centriole duplication in worm embryos [25,26], DSpd-2 was only picked up in our screen as being required for PCM recruitment (see below), consistent with previous analyses of DSpd-2 mutant flies [51,52]. Whether human Spd-2/Cep192 has a role in centriole duplication that is independent of its role in PCM recruitment remains controversial [67,68]. Thus, the exact role of this family of proteins in centriole duplication and PCM maturation remains to be clarified.
We believe we have now identified most, if not all, of the major structural components required for general PCM assembly during mitosis (see below). Cnn, DSpd-2, D-PLP, γ-tubulin, and Grip71WD are all components of the PCM, whereas Map205 is a MT-associated protein that is present in the PCM. Polo and a specific form of PP2A appear likely to play regulatory roles in this process. Moreover, although the depletion of Aurora A and UbcD6 causes primarily a centriole-clustering phenotype, the recruitment of PCM to individual centrioles is reduced when either of these proteins is depleted, indicating that they too play a part in centrosome maturation. Although it remains unclear how these proteins work together to drive centrosome maturation, the individual depletion of two of these proteins, Cnn and Polo, consistently perturbed centrosome maturation to a greater extent than the depletion of any of the other proteins. This indicates that these two proteins may initiate the centrosome maturation pathway in flies. In support of this possibility, we found that Cnn is specifically phosphorylated during mitosis in a Polo-dependent manner. More experiments are required, however, to determine whether Polo phosphorylates Cnn directly, and whether this phosphorylation event really initiates centrosome maturation, or is simply correlated with it.
Interestingly, it has previously been postulated that Cnn functions primarily to “strengthen” the structure of the PCM, thus preventing the PCM from dissipating away from the centrosomes soon after it is recruited [38]. An attractive model is that the Polo-dependent phosphorylation of Cnn may initiate centrosome maturation by allowing Cnn to strengthen the PCM. In such a scenario, the centrioles would actively recruit PCM at all stages of the cell cycle, but in the absence of phosphorylated Cnn, the PCM is structurally weak, and it cannot accumulate to any extent around the centrioles. As cells enter mitosis, Polo phosphorylates Cnn (either directly or indirectly), thus allowing it to strengthen the PCM, which can then accumulate around the centrioles.
An important question is whether the proteins we identify here represent a complete list of those required for centriole duplication and centrosome maturation in flies. Clearly, we may have missed some proteins. Our screen probed only approximately 92% of protein-coding genes, and 108 proteins could not be tested because there were not enough mitotic cells to be scored after their depletion. In addition, some proteins may not have been detected because they are poorly depleted by RNAi, or because their depletion produced such pleiotropic defects that centrosome defects could not be scored properly. On the other hand, all 13 of the known fly proteins previously implicated in centrosome maturation (Polo, Aurora A, Cnn, DSpd-2, D-PLP, Asl, and γ-tubulin) or centriole duplication (DSas-4, DSas-6, Sak, Ana1, Ana2, and Ana3) were successfully identified in our screen. This is despite the fact that many centriolar proteins are known to be difficult to deplete by RNAi [42,69] (J. Dobbelaere, unpublished data). Moreover, the depletion of proteins such as Polo and Aurora A clearly produces pleiotropic mitotic defects, yet both proteins were successfully identified in our screen.
Taken together, these observations suggest that it is unlikely we are missing large numbers of proteins from this list, and that we are at least approaching a near-complete inventory of the proteins required for centriole duplication and centrosome maturation in flies. Although this list is significantly larger than the list that has emerged from studies in worm embryos, it is still surprisingly small, and we conclude that only a relatively small subset of the many proteins concentrated at centrosomes is actually essential for the key centrosomal functions of duplication and maturation. Clearly, this extensive dataset provides an important framework with which to delineate the events that drive the centrosome cycle.
An RNAi library covering nearly the whole Drosophila genome was purchased from Ambion (Silencer(R) Drosophila RNAi Library, AM85000). This library comprises dsRNAs designed against 13,059 Drosphila genes, or approximately 92% of all currently known protein-coding genes (Flybase). The original library, in 96-well plates, was replated onto clear bottom 384-well plates (Corning #3712) to a final concentration of 0.22 μg of dsRNA/well in 5 μl (1× PBS) using a Beckman Biomek FX. Controls were added in the upper left and lower right corner of each plate. dsRNA against DsRed was used as a negative control. dsRNA against Scar, String, and Thread were added as controls for cell morphology, division, and cell death. Finally, dsRNA against Polo and Cnn were added as positive controls to every 384-well plate for this specific screen.
For the primary screen, S2R+ cells were cultured in Shields and Sang medium (Sigma S3652) with 10% FBS (Sigma F9665) and 1% penicillin/streptomycin (Gibco 15070–063). After trypsinising the cells, they were diluted to 7 × 105 cells/ml in serum-free Shields and Sang medium. A total of 15 μl of cells were added to the dsRNA-containing 384-well plates using a Thermo Wellmate (giving a final concentration of ∼10,500 cells per well). Plates were gently spun, and cells were incubated for 30–45 min, and 35 μl of serum-containing medium was added. Plates were sealed and incubated for 4 d at 25 °C. Eight hours prior to fixation, we exchanged the medium for medium containing 25 μM colchicine (Sigma #C3915), a microtubule depolymerising drug that arrests cells in mitosis (this typically resulted in 20%–35% of the cells in a well being in mitosis at the time of fixation). Cells were washed once with PBS, fixed with 4% formaldehyde (in PBS) (Sigma #F8775) for 12 min, and permeabilised with 0.5% SDS in PBS for 10 min. Cells were blocked with 5% goat serum (Sigma G9023) in PBS-T (0.1% Triton) for 20 min and stained overnight at 4 °C with anti-Cnn antibodies (1:1,000, rabbit) to stain centrosomes [38] and anti-pH3 Ser10 antibodies to label mitotic cells (1:2,000, mouse; Abcam 14955). Antibodies were diluted in PBS-T with 5% goat serum. The next day, cells were washed three times with PBS-T for 5 min. Secondary antibodies, anti-rabbit Alexa 488 (1:1,500; Molecular Probes A21206) and anti-mouse Alexa 567 (1:1,500; Molecular Probes A11004), in 5% goat serum in PBS-T were added for 2 h at room temperature. Cells were washed once with PBS-T, incubated with Hoechst 33258 (final concentration of 0.2 μg/ml; Sigma #861405) in PBS for 10 min, and then washed once more with PBS-T. Finally, 20 μl of PBS was added to each well, and plates were sealed with aluminium sealing tape (Corning #6569).
Specimens were imaged on a Nikon TE2000E microscope, with an automated Prior stage controlled with Metamorph software (Molecular Devices) using a 20×, 0.45NA, Plan Fluor air objective. After automated focusing, we took pictures of the three channels (Hoechst, Alexa 488, and Alexa 567) at four different sites per well (an average total of 500–2,000 cells, approximately 150–400 of which were usually in mitosis). All primary pictures and annotation are available on the Flight database (http://flight.licr.org/)
For the secondary screening assays, RNAi was performed as above (using 0.22 μg, 0.6 μg, 2 μg, or 10 μg of dsRNA per well for 384-well, 96-well, 24-well, or 6-well plates, respectively). Detailed immunofluorescence analysis of centrioles and PCM was performed by adding a glass cover slip before seeding the cells in 24-well plates and analysing the cells on a Perkin Elmer Ultraview ERS spinning disk system on a Zeiss Axioskop II microscope using a 63×, 1.4NA, Plan Apo oil objective. Antibodies used in the secondary screen were rabbit anti-DSpd2 (1:500; [51]), rabbit anti-DSas4 (1:500; [16]), mouse anti–γ-tubulin (1:500; GTU-88 Sigma), and mouse anti–α-tubulin (1:1;000; DM1α Sigma). Twenty images at 0.25-μm separation in the Z-axis were taken in each channel, and a maximum-intensity Z-projection was made using the Ultraview ERS software. Note that the anti–DSas-4 antibodies usually cannot distinguish between a single centriole and a centriole pair (as centriole pairs usually stain as a single dot in these cells with this antibody).
To identify proteins that give centrosome defects after depletion with dsRNA, we scored each well by three different methods. First, each well was inspected manually on the widefield microscope system described above, and given a numerical score (from −3 to 3) for the severity of any defect in cell number, mitotic index, centrosome number, and centrosome size. Second, the pictures taken with the automated microscope were manually scored using the same criteria. All of these analyses were performed “blind,” so that we did not know which genes were being analysed. Finally, the pictures were analysed with CellProfiler (http://www.cellprofiler.org) [41] using a self-made pipeline (See Text S1). This resulted in a numerical value for the number of Cnn dots per mitotic cell. The inverse of this numerical dataset was normalised (plate average was set to zero) and corrected for plate-by-plate variations and possible edge effects using the CellHTS software ([70], using the B-score method) (See Figure S3). The Z′-score was calculated using Cnn and Polo as positive controls, and all empty and DsRed wells as negative controls. This analysis enabled us to give a statistical significance to each potential hit. A total of 108 genes were excluded from both the manual and the automated analysis because of the lack of cells or lack of mitotic cells in the well (Table S2); 119 genes were selected for secondary analysis as they were scored as hits with at least two of these three methods. From these 119 genes, only 79 were selected for a more detailed secondary analysis, as we eliminated genes that were commonly identified in previous screens (indicating they are likely false positives), were known components of the ribosome or transcription machinery, or were the result of clear off-target effects (Table S1).
For the secondary analysis, centriole (DSas-4) and centrosome (Cnn) number (shown in the graphs associated with each gene in Protocol S1) were quantified as follows. Maximum intensity z-projections from two independent experiments (at least 30 mitotic cells per experiment) were analysed, and the number of centrioles per mitotic cell were counted. The amount of PCM accumulated around each centriole was scored by eye as either normal or small/absent. For the quantification of PCM recruitment shown in Figures 2E, 3E, 4E, 5E, and S1, PCM size was quantified by measuring the background-corrected mean intensity of the Cnn dots in the z-projected image. Average intensities (normalised against control RNAi set to 100%) are represented from three independent experiments (typically 20–40 Cnn dots were counted per experiment, but occasionally only 10–15 Cnn dots were counted for proteins whose depletion meant there were very few centrosomes that could be counted). The statistical significance was measured using a dual-tailed t-test. p ≤ 0.05 are marked on the graphs by a single asterisk (*), and p ≤ 0.01 are marked with double asterisks (**).
Vectors allowing the expression of GFP-tagged proteins were made using the Gateway system (Invitrogen). A list of the primers used is shown in Table S4. Constructs for all genes, unless otherwise stated (Table S4), were made for both N- and C-terminal (NT and CT, respectively) tagging. Forward primers for NT- and CT-tagging were the same (including ATG), but the NT reverse primer included the STOP codon, whereas the CT-primers lacked the STOP codon. All genes were cloned from cDNA unless stated otherwise (Table S4). Once cloned in the pZEO-Entry vector, inserts were checked by restriction digest and most of them also by sequencing (Table S4).
The genes were then recombined into the expression vectors pMT (Invitrogen) and pwUbq (gift from R. Basto), placing the genes under the control of the metallothionein and ubiquitin promoters, respectively. Transfection of the expression vectors in S2 cells was performed as described previously [71]. Approximately 350,000 S2 cells were plated in 24-well plates for 2 h. At 30 min before transfection, 0.6 μg of vector DNA was mixed with 0.06 μg of pCoBlast (Invitrogen), 5 μl of Cellfectin (Invitrogen), and 50 μl of serum-free Schneider medium (SFM) (Sigma). A total of 450 μl of SFM was added to the transfection mix. The medium of the plated S2 cells was removed, and the transfection mix was added. After 3–4 h, 1 ml of serum-containing Schneider medium was added. Cells were incubated for 4 d before adding 25 μg/ml blasticidin. After 3–4 wk, stable cultures were obtained. GFP expression was analysed by western blotting and immunofluorescence (IM). Cells containing the pMT vector were induced 24 h prior to analysis with 100 μM CuSO4. When S2 cells were to be analysed by immunofluorescence, cells were plated on glass slides coated with 0.05 μg/ml ConcavalinA (Sigma #C5275) and fixed with 4% paraformaldehyde (freshly prepared in PBS). Cells were costained with anti–α-tubulin (1:1,000 DM1α) and anti-DSas4 antibodies (1:500) and Hoechst. Pictures were taken and analysed as described above. Maximum z-projections are shown in all figures.
S2R+ cells incubated with or without dsRNA (as described above) in 24-well plates were washed once with PBS and then suspended in 200 μl of loading buffer. Samples were boiled for 10 min, and 10 μl was loaded on a 3%–8% gels (Nupage; Invitrogen). The samples were blotted on nitrocellulose membranes and probed with anti-Cnn antibodies (1:1,000), as described previously [38]. An anti-actin antibody (MP Biomedicals #08691001) was used as a loading control (1:1,000). For the phosphatase treatment of S2R+ cell extracts, cells were diluted in lysis buffer (PBS, 5 mM EDTA, 1× PMSF, 1× protease inhibitor [Roche Complete]) plus or minus phosphatase inhibitors (25 mM NaF, 1 mM Na3VO4, 20 mM beta-glycerol phosphate, 1× phosphatase inhibitor cocktail [Sigma #P2850]) and syringed through a G24 needle approximately 60 times on ice. Lysates were incubated for 30 min at 30 °C with 10 units/100 μl of lambda phosphatase (Sigma # P9614). The reaction was quenched by the addition of 4× loading buffer. Samples were analysed by western blotting. For the 2-D analysis, samples were suspended in 2-D buffer (10 mM Tris [pH 8–8.5], 5 mM magnesium acetate, 8 M urea, and 4% CHAPS). The protein concentration was measured and 50 μg of proteins analysed using pH 4–10 strips and 12% acrylamide gels, and processed for western blotting.
Third instar larval brains were dissected from wild-type (w67) and aurora-A mutant flies (transheterozygotes between the two hypomorphic alleles aure200 and aure209), and fixed and stained as described previously [16]. Brains were stained with Cnn (1:1,000), α-tubulin (1:1,000), γ-tubulin (1:500), and DSas-4 (1:500) antibodies. More than 50 mitotic cells were analysed for three different brains. For the statistical analysis of the centriole number in these mitotic cells, centrioles were only counted if they were DSas-4 and γ-tubulin positive.
|
10.1371/journal.pntd.0006521 | Chikungunya virus vector competency of Brazilian and Florida mosquito vectors | Chikungunya virus is a vector-borne alphavirus transmitted by the bites of infected female Ae. aegypti and Ae. albopictus. In Brazil between 2014 and 2016 almost 320 thousand autochthonous human cases were reported and in Florida numerous imported CHIKV viremic cases (> 3,800) demonstrate the potential high risk to establishment of local transmission. In the present study, we carried out a series of experiments to determine the viral dissemination and transmission rates of different Brazilian and Florida populations of Ae. aegypti and Ae. albopictus at 2, 5, and 13 days post-infection for the emergent Asian genotype of CHIKV. Our results show that all tested populations of Ae. aegypti and Ae. albopictus have a high proportion (> 0.80) of individuals with disseminated infection as early as 2 days-post exposure. We found no significant treatment effects of mosquito population origin effects on viral dissemination rates. Transmission rates had a heterogeneous pattern, with US Ae. aegypti and Brazilian Ae. albopictus having the highest proportion of individuals with successful infection (respectively 0.50 and 0.82 as early as 2 days-post infection). Model results found significant effects of population origin, population origin x species, population origin x days post-infection and population origin x species x days post infection.
| Chikungunya is considered a serious mosquito-borne disease in many tropical and subtropical countries throughout the world. It is already an epidemic disease in Brazil and poses as a potential risk in Florida. It is mainly transmitted by mosquitoes Aedes aegypti and Aedes albopictus. These mosquito species are common and abundant throughout much of the year in Brazil and Florida. In this study, we determined two components of vector competence from Brazilian and Florida populations of both mosquitoes to the emergent Asian genotype of chikungunya virus: viral dissemination and transmission rates. Both Aedes populations exhibited a high proportion of disseminated infection as early as two days after ingestion of chikungunya virus infected blood. Transmission efficiency was higher in Ae. aegypti from Florida and Ae. albopictus from Brazil. Our findings suggest that mosquito-virus interactions of both Ae. aegypti and Ae. albopictus may vary by geographic population, which may impact public health measures and should be considered during outbreaks of this arboviral disease.
| Chikungunya fever is a vector-borne viral disease that originated in Africa and is caused by a virus (CHIKV; family Togaviridae, genus Alphavirus) transmitted by the bites of infected female Aedes mosquitoes, mainly Ae. aegypti and Ae. albopictus [1]. There are three genotypes of CHIKV, which apparently evolved independently in distinct geographic regions: Asian, West African, and East/Central/South African (ECSA) [2]. CHIKV is widespread worldwide and poses as a major public health problem in tropical and subtropical regions [3–6]. In the Americas, autochthonous transmission of CHIKV was first detected in St. Martin Island in October 2013 and quickly spread throughout the Americas in the following months [7–9]. The initial spread of autochthonous cases in the Americas was due to the Asian genotype, but the ECSA genotype was also detected circulating in Brazil in 2014 [10]. To date, local transmission of CHIKV has been documented in over 43 countries with more than 1,000,000 confirmed cases, where Brazil reported 314,834 until the 15th epidemiological week of 2017 [11–12].
Aedes aegypti and Ae. albopictus are the main vectors of CHIKV, and both are highly invasive species and closely associated with the human peridomestic environment [13, 14, 6]. Aedes aegypti is highly anthropophilic and exhibits endophilic behavior and is mostly associated with high human density. In contrast, Ae. albopictus shows an eclectic feeding behavior, preferentially feeding and resting in the peridomicile and is more common in vegetated and urban/urban forest transition habitats, especially where it is sympatric with Ae. aegypti [15–19]. In Africa, CHIKV is maintained via an enzootic cycle involving several species of arboreal mosquitoes, including Ae. africanus and Ae. furcifer, and non-human primates [20]. Epidemic transmission is maintained mainly by Ae. aegypti in urban environments, but a single-base mutation in a strain of the ECSA genotype during the outbreak in La Réunion Island enhanced vector competence of Ae. albopictus [21, 22]. A second mutation is associated with enhanced vector competence of Ae. albopictus during an outbreak in Kerala, India [23]. In fact, the acquisition of second-step Ae. albopictus-adaptive mutations by CHIKV strains might indicate even more efficient transmission by this invasive vector [24].
Vector competence studies are important to determine the potential of resident mosquito populations to transmit CHIKV. Vector competence is a phenotypic parameter that describes the ability of the vector to become infected, replicate and transmit a pathogen [25, 26]. Moreover, vector competence depends on vector and viral genetic characteristics [27] and environmental factors such as ambient temperature and diurnal temperature range [28–32]. It has been shown that vector competence of Ae. aegypti and Ae. albopictus for CHIKV is a complex interaction dependent on vector population, virus strain and temperature [33, 34]. The vector competence of Ae. aegypti for dengue virus (DENV) has been shown to have high variability and heterogeneity whether it is analyzed at city [35], country [36] or continental level [37].
Previous studies of CHIKV have characterized variation in vector competence among CHIKV genotypes, extrinsic incubation temperature, and geographic populations of Ae. aegypti and Ae. albopictus, and species-specific differences. In Florida, Ae. aegypti and Ae. albopictus were highly susceptible to infection and viral dissemination to ECSA and Asian genotypes of CHIKV, with some variation between strains [38, 39]. Pesko et al. (2009) [40] evaluated vector competence of Ae. aegypti and Ae. albopictus from Florida for infection with a La Réunion island ECSA isolate of CHIKV. Although both species were susceptible to high CHIKV doses, Ae albopictus was more susceptible to infection than Ae. aegypti. Richards et al. (2010) [33] assessed the effect of extrinsic incubation temperature on vector competence of Florida mosquitoes for CHIKV isolates from La Réunion and found highest infection, dissemination, and transmission rates in Ae. albopictus than in Ae. aegypti and Culex quinquefasciatus, but no effect on the extrinsic incubation period. Vega-Rúa et al. (2014) [31] working with three CHIKV genotypes and 35 populations of Ae. aegypti and Ae. albopictus mosquitoes from 10 American countries showed that all Aedes populations tested were susceptible to CHIKV infection by all three genotypes. However, CHIKV transmission was heterogeneous in American Ae. aegypti and Ae. albopictus populations, ranging from 11.1% to 96.7%. In this study, the Aedes populations from Rio de Janeiro showed high transmission rates, and Ae. albopictus from Florida were more competent vectors than Ae. aegypti.
Although Ae. aegypti is considered the primary epidemic vector of CHIKV and Ae. albopictus a potential vector in some areas [2, 21, 31], heterogeneous vector competence of both species may alter risk of disease transmission, as evidenced by the participation of Ae. albopictus in the outbreak in La Réunion Island [21]. Studies comparing vector competence in American populations of both species are necessary in a scenario where travel and global trade in endemic regions have increased the risk for spread of CHIKV, as evidenced by its introduction in the Americas [41]. Also, there is a real risk for the introduction of CHIKV strains with adaptive mutations to enhance vector competence of Ae. albopictus, an invasive species which is widespread in the Americas [24. With the aim to shed light on the causes and consequences of geographical variations in the transmission of arboviruses of public health concern, we carried out an experiment to determine the dissemination and transmission rates of Brazilian and Florida populations of Ae. aegypti and Ae. albopictus for the emergent Asian genotype of CHIKV.
Chikungunya virus (Asian lineage, GenBank accession: KJ451624) used was isolated from the serum of an infected human in the British Virgin Islands in 2013 by other investigators. Subsequently, this isolate was archived with the Centers for Disease Control and Prevention. We requested an isolate of this virus for use in this study and so the sample was already present in an already-existing collection (Centers for Disease Control and Prevention, Arboviral Diseases Branch). The virus sample was anonymized and Institutional Review Board approval was not needed for receipt and use of the sample in this study. No entomological gathering was done on private land or in private residence for this study.
The Ae. aegypti and Ae. albopictus populations used in this experiment were collected in Rio de Janeiro (RJ) and Macapá (MC)—Brazil, Key West (KW) and Okeechobee (OK), Florida—United States (Fig 1, Table 1). All gathering of entomological samples were done on public land. We chose collection sites based on allopatric Ae. aegypti to Ae. albopictus (MC and KW) and sympatric populations (RJ and OK). Some of these areas report local transmission of chikungunya cases (RJ and MC) while others are located near regions in Florida where local transmission has occurred (Miami-Dade, Palm Beach, St. Lucie, and Broward Counties) (KW and OK) [12, 42]. In Brazil, eggs of both species were obtained from oviposition traps during a routine entomological survey. Aedes albopictus from a sympatric population (RJ) were obtained at the Oswaldo Cruz Foundation campus, in the Manguinhos neighborhood in March 2015 from 50 oviposition traps using methods described elsewhere [16]. Aedes aegypti eggs from an allopatric population (MC) were collected with oviposition traps by personnel from the Amapá State Health Secretary in May 2015. In United States, eggs of allopatric Ae. aegypti (KW) were collected in March 2015 with oviposition traps by personnel of Florida Keys Mosquito Control District. Immatures of sympatric Ae. albopictus (OK) were collected from tires in October 2015.
Field-collected mosquitoes (eggs or larvae) were reared in pans containing 1 L of tap water (100 larvae per pan) to adulthood on a diet with 0.6 g of equal amounts of brewer’s yeast and lactalbumin. Mosquitoes were held in a climate controlled room at 26–28°C and a photoperiod of 14:10 hours light:dark. Upon pupation, pupae were collected daily and placed in vials with a cotton seal until eclosion after which adult mosquitoes were identified to species. Adults were transferred to 0.3m3 cages and provided with 10% sucrose solution and water from cotton wicks and allowed to feed on bovine blood once per week using an artificial feeding system with hog intestine membranes. Females and males were held together for eleven days after which females were transferred to cylindrical cages (ht. by dia., 10 cm by 10 cm, 50 females/cage) with mesh screening one day before being fed CHIKV infected blood. The F2 (Okeechobee) and F3 (Rio de Janeiro, Macapá and Key West) generations progeny of field-collected Ae. aegypti and Ae. albopictus were used for the CHIKV infection study.
The strain of CHIKV (Asian lineage, GenBank accession: KJ451624) used was isolated from the serum of an infected human in the British Virgin Islands in 2013. The Centers for Disease Control and Prevention was the source of the virus strain used in this study. The CHIKV isolate was passaged twice in culture using African green monkey (Vero) cells and viral titer was determined in 6-well plates seeded with Vero cells (American Type Culture Collection, ATCC) by plaque assay using a modified procedure by Kaur et al. (2016) [43].
For preparation of the virus suspension, monolayers of Vero cells were inoculated with dilute stock CHIKV at a multiplicity of infection of 0.1 followed by a one-hour incubation at 37°C and 5% carbon dioxide atmosphere. The American Type Culture Collection was the source of Vero cells used in this study. After the inoculation procedure, each flask received 24 ml media (M199 medium supplemented with 10% fetal bovine serum, penicillin/streptomycin and mycostatin) and was left to incubate for an additional 47-hours. Adult females aged 10–11 days were offered CHIKV infected defibrinated bovine blood (Hemostat, Dixon, CA) using an artificial feeding system with hog intestine membranes (Hemotek, Lancashire, United Kingdom). Samples of blood were taken of the virus-blood suspension at the time of feeding to determine the concentration of CHIKV ingested by the adult mosquitoes. Blood meal titers ranged from log10 7.3 to 8.3 plaque forming unit equivalents (pfue)/mL. Fully engorged females were held in cylindrical cages along with an oviposition substrate and maintained at a 14:10 hour light:dark photoperiod and 28°C.
Virus transmission potential using saliva assays was determined at 2, 5, and 13 days after feeding on infected blood. Mosquitoes were deprived of sucrose for 1-day and then individually transferred to plastic tubes fitted with a removable screen lid (37-mL 8 by 3 cm). Honey was dyed with blue food coloring (McCormick) and impregnated on filter paper (1 cm diameter) and fastened to the inside lid of the tube. Mosquitoes that fed on the honey deposited saliva and the blue food coloring was visualized in the crop with aid of an incandescent flashlight. Mosquitoes were examined for blue in the crop after 24 and 48-hours during the transmission assay. Only mosquitoes that fed on honey were used to assess transmission potential. Additionally, saliva was collected from another subset of mosquitoes in capillary tubes with immersion oil as described previously [44,32, 39]. Mosquitoes were stored at -80°C after the transmission assay and later dissected to test the legs and saliva for the presence of CHIKV RNA by qRT-PCR [32]. The sequence of primers targeting a nonstructural polyprotein gene was as follows: forward, 5'-GTACGGAAGGTAAACTGGTATGG-3': reverse, 5'-TCCACCTCCCACTCCTTAAT-3'. The probe sequence was: 5'-/56-FAM/TGCAGAACCCACCGAAAGGAAACT/3BHQ_1/-3' (Integrated DNA Technologies, Coralville, IA). Detection of CHIKV RNA in the legs of a mosquito is considered a proof that the virus infection has disseminated from the midgut, and we use the number of mosquitoes with a disseminated infection over the number of mosquitoes fully engorged on a viraemic blood-meal, as the virus dissemination rate. Detection of CHIKV RNA in mosquito saliva is considered a proof that the mosquito can transmit virus when feeding, and we use the proportion of mosquitoes with virus in saliva among all mosquitoes with a disseminated infection as our expression of transmission rate.
For each mosquito, legs were triturated in 1.0 mL of media (GIBCO Media 199). Saliva from mosquitoes was combined with 300 μL of media. RNA isolation on a 140 μL sample of mosquito legs and saliva homogenate was achieved using the QIAamp viral RNA mini kit (Qiagen, Valencia, CA) and eluted in 50 μL of buffer according to the manufacturer’s protocol. Viral RNA was detected using the Superscript III One-Step qRT-PCR with Platinum Taq kit by Invitrogen (Invitrogen, Carlsbad, CA) using methods described elsewhere [32, 39]. Quantitative RT-PCR was performed with the CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA) with the following program: 50°C for 30 minutes, 94°C for 2 minutes, 39 cycles at 94°C for 10 seconds and 60°C for 1 minute, and 50°C for 30 seconds. The expression of viral titer in mosquito-derived samples used a standard curve method comparing cDNA synthesis for a range of serial dilutions of CHIKV in parallel with plaque assays of the same dilutions of virus, expressed as plaque forming unit equivalents (pfue)/ml [45].
We were interested in analyzing the relationship between the presence or absence of CHIKV in the legs and saliva (dependent variables) and the following independent variables: mosquito species (Ae. aegypti and Ae. albopictus), population origin (Brazil and USA), days post-infection (dpi, 2, 5 and 13), and a three-way interaction of species by population origin by days post-infection. Exploratory analyses were done using chi-square tests to verify possible relationships between both dependent variables (presence or absence of CHIKV in the legs and saliva) and each of the independent variables. We modeled this relationship using two separate binomial generalized linear models: one focused on the viral dissemination to the legs, and the other focused on the viral infection of saliva. To account for numerical problems in the viral dissemination binomial model, we used a Firth's Bias-Reduced Logistic Model [46]. We also analyzed the relationship between the viral titer of legs and saliva and the aforementioned main effects using a Gaussian generalized linear model. All analyses were done using R [47] and RStudio [48], with the libraries ggplot2 [49], logistf [46] and lsmeans [50].
Chikungunya virus (Asian lineage, GenBank accession: KJ451624, repository: Centers for Disease Control and Prevention).
Chikungunya virus dissemination rates were measured by the proportion of mosquitoes that had infected legs from the total that fully engorged on infected blood. A total of 358 Aedes mosquitos were tested for disseminated infection (172 Ae. aegypti and 186 Ae. albopictus). Overall, our results showed the proportion of individuals of both species with disseminated infection significantly increased with each of the days post-infection analyzed (2-dpi, 0.847 ± 0.034; 5-dpi, 0.977 ± 0.013; and 13 dpi, 0.984 ± 0.011) (χ2 = 24.35, df = 2, p<0.0001). Aedes aegypti had higher dissemination rates than Ae. albopictus (mean ± SE, 0.960 ± 0.014 and 0.919 ± 0.020, respectively), although not significant (χ2 = 2.09, df = 1, p = 0.148). Both US and Brazilian populations of Ae. aegypti (0.976 ± 0.016 and 0.946 ± 0.023, respectively) had higher dissemination rates when compared to Ae. albopictus (0.915 ± 0.028 and 0.922 ± 0.028, respectively), but this difference was also not significant (χ2 = 0.03, df = 1, p = 0.857).
When analyzing the dissemination rates per species, population origin and days post-infection interaction, Ae. aegypti reached 100% of individuals at the 5th and 13th days, but the US population had higher dissemination rates at the 2nd day when compared to the Brazilian population (0.913 ± 0.06 and 0.814 ± 0.07, respectively) (Fig 2). These differences, however, were not significant (χ2 = 0.07, df = 2, p = 0.961). For Ae. albopictus, both US and Brazilian populations had similar dissemination rates at the 2nd day (0.843 ± 0.065 and 0.827 ± 0.071). At the 5th day, the US population had a lower dissemination rate when compared to the Brazilian population (0.906 ± 0.052 and 1.0, respectively). At the 13th day, the A. albopictus US population had a higher dissemination rate (1.0) than the Brazilian population (0.933 ± 0.046). The dissemination rate did not significantly differ between population origins (χ2 = 0.36, df = 2, p = 0.834) (Fig 2).
The three-way interaction Firth's bias-reduced logistic model results show that none of the main effects or the interactions were significant for disseminated infection rates (Table 2).
When analyzing the viral titers in the mosquito legs, Gaussian model results show that days post-infection had a significant positive effect, and the interaction of species and population origin had a significant negative effect (Table 2). Overall, both populations of Ae. aegypti had lower levels of viral titer (expressed in log10 pfue/mL) in their legs at 2nd day post-infection, which increased and peaked at the 5th and 13th days (US; 2nd day = 2.884 ± 0.453, 5th day = 4.289 ± 0.179 and 13th day = 4.131 ± 0.053; and Brazilian 2nd day = 2.668 ± 0.411, 5th day = 3.610 ± 0.277 and 13th day = 4.119 ± 0.110). The same pattern was observed for Ae. albopictus for both US (2nd day = 2.060 ± 0.290, 5th day = 4.074 ± 0.263 and 13th day = 3.676 ± 0.244) and Brazilian populations (2nd day = 3.086 ± 0.362, 5th day = 3.971 ± 0.241 and 13th day = 3.988 ± 0.183) (S1 Fig).
Chikungunya virus infection rates were measured by the proportion of mosquitoes that had infected saliva from the total that presented viral dissemination. A total of 224 Aedes mosquitoes that had positive leg infections were tested for saliva infection (107 Ae. aegypti and 117 Ae. albopictus). Overall, we found a significant effect of days post-infection and infection rates when analyzing both species (χ2 = 8.88, df = 2, p<0.05) (Fig 3). The infection rates reached a peak at the 5th day post-infection and decreased at the 13th day (2-dpi, 0.415 ± 0.068; 5-dpi, 0.500 ± 0.050; and 13-dpi, 0.274 ± 0.053). We also found a significant relationship between infection rates per species and population origin (χ2 = 11.55, df = 1, p<0.0001); US Ae. aegypti had higher infection rates when compared to the Brazilian (Ae. aegypti, 0.5 ± 0.068; Ae. albopictus, 0.264 ± 0.061). For Ae. albopictus, the US population had lower infection rates when compared to Brazilian conspecifics (0.245 ± 0.057 and 0.6 ± 0.063, respectively). The analysis of infection rates per species, population origin and days post-infection for Ae. aegypti showed that the US population had similar rates in all days (2-dpi, 0.5 ± 0.166; 5-dpi, 0.52 ± 0.101; 13-dpi, 0.473 ± 0.117). The Brazilian population had a lower infection rate when compared with the US population at all day’s post-infection (0.1 ± 0.1, 0.391 ± 0.2 and 0.104 ± 0.091, respectively), although this difference was not significant (χ2 = 1.32, df = 2, p = 0.67). For Ae. albopictus, the US population had a lower infection rate at the 2nd and 13th days (0.125 ± 0.085 and 0.176 ± 0.095, respectively) and higher infection rates at the 5th day (0.375 ± 0.1). The Brazilian population however had high infection rate at the 2nd day (0.823 ± 0.095), decreasing at the 5th day (0.692 ± 0.092) and finally decreasing further at the 13thday (0.235 ± 0.106). The Brazilian population had a higher infection rate at all day’s post-infection when compared to the US population, but this difference was not significant (Fig 2, χ2 = 3.05, df = 2, p = 0.238) (Fig 3).
The three-way interaction logistic model results showed a significant effect of population origin, and the interactions between population origin x species, population origin x days post-infection and the three-way interaction of population origin x species x days post infection were significant for saliva infection rates (Table 3).
The Gaussian model to analyze the viral titer in the saliva of the tested mosquitoes did not detect significant main effects or interactions of the treatment factors (Table 3). The US population of Ae. aegypti had similar levels of viral titer in the saliva at all three time-points tested (respectively 1.794 ± 0.593, 1.516 ± 0.247 and 1.351 ± 0.171 pfue/mL), while the Brazilian population had a peak at the 5th day and decreasing at the 13th day (respectively 1.659 ± 0.376 and 1.300 ± 0.429 pfue/mL). The US population of Ae. albopictus had higher viral titer in their saliva at the 2nd day, decreasing with each passing time point (2.290 ± 0.730, 1.036 ± 0.247 and 0.810 ± 0.228 pfue/mL). For the Brazilian population of this species, viral titer peaked at 5th days, decreasing at the 13th (respectively 1.095 ± 0.140, 1.501 ± 0.245 and 1.058 ± 0.458 pfue/mL) (S2 Fig).
This study tested the vector competence of two populations of Ae. aegypti and Ae. albopictus from Brazil and Florida for an emergent Asian lineage of CHIKV. We carried out a series of experiments to determine two fundamental characteristics of this phenotypic trait: viral dissemination into the haemocoel of the tested mosquitos and saliva infection. These measurements characterize midgut and salivary gland barriers and are determinants of the vector competence of a mosquito population [26]. While viral dissemination indicates its propagation in the midgut and subsequent spread of the infection to other tissues, saliva infection is needed for the mosquito to successfully transmit the arbovirus by bite to a vertebrate host. Our results shed light on important questions regarding vector competence of Aedes mosquito populations of the Americas. The lack of statistical significance when comparing species and populations shows that viral dissemination occurs equally in these treatment conditions. In fact, more than 90% of all individuals have successful viral dissemination in their bodies, despite heterogeneity in species and population origin. This conclusion is further supported by the model results, which shows that none of the tested effects and interactions were statistically significant. Because high rates of disseminated infection were observed under these conditions, we had greater potential to detect treatment-dependent reductions in disseminated infection and less ability to identify treatment enhanced disseminated infection.
In our study, viral dissemination occurred rapidly, with around 85% of all individuals with positive legs at the 2nd day post-infection, and more than 98% of mosquitoes tested positive at the 13th day-post infection. Rapid viral dissemination together with a short extrinsic incubation period, as observed by saliva infection assays, may have important consequences for CHIKV epidemiology, especially given that both these Aedes species exhibit gonotrophic discordance [51, 52]. For instance, females will remain infectious for longer periods during the adult stage after ingesting CHIKV than pathogens with longer EIPs. Moreover, mosquito adult survival, EIP and host feeding strongly contribute to vectorial capacity which describes the number of infective bites received daily by a single host [53, 6]. A more thorough analysis showed that both populations of Ae. aegypti had similar levels of viral dissemination, reaching 100% of all tested individuals at the 5th day post-infection. For Ae. albopictus, we found a similar pattern with an increasing proportion of individuals with disseminated infection with each passing day post-infection. However, only the US population reached 100% of individuals with disseminated infection. This high number of individuals of both species and populations with disseminated infection might suggest a lack of substantial midgut escape barriers for the CHIKV strain used [31].
It is unclear whether differences in disseminated infection rates may be observed among these invasive Aedes mosquitoes if lower titer CHIKV infected blood were ingested. Studies have shown differences in susceptibility of Aedes vectors to CHIKV depending the dose of virus ingested [54, 55, 39]. Differences in susceptibility of Ae. aegypti and Ae. albopictus from Florida to infection and transmission of two lineages of CHIKV (Indian Ocean and Asian genotype) were tested [39]. In this study, Ae. aegypti tested with a lower dose of CHIKV Asian genotype in two different temperatures (25°C and 30°C) did not have significant differences in viral dissemination and transmission (100% to 40% and 33.3% to 0%, respectively). The low infection rates were attributed to a relatively low dose of CHIKV in blood meals (5.8 log10 pfue/ml). On the other hand, all populations of Ae. aegypti and Ae. albopictus presented higher susceptibility to infection and transmission for these two tested lineages of CHIKV at high titers [39, 54] determined the relative susceptibility of selected strains of Ae. aegypti and Ae. albopictus fed on a viremic monkey to infection with Southeast Asian strain of CHIKV. The results showed that strains of Ae. albopictus, regardless of their geographical origin, were more susceptible to infection (range, 72–97%) and dissemination (36–80%) with CHIKV than Ae. aegypti (infection rate, 12–25% and dissemination 8–25%) even though some strains presented lower infection rates in mosquitoes that ingested the lower dose (104.2–4.6 pfu/ml). Coffey et al. (2014) [55] summarizes numerous chikungunya virus infection with Ae. aegypti and Ae. albopictus, stating lower and higher doses used in infected blood meals. In this review, the authors showed that infection, dissemination, and transmission rates of both Aedes vectors can vary according to the geographic sources of mosquitoes and the titer of the ingested bloodmeal. For instance, using bloodmeal titers of > 7 log10 pfu/ml (high dose) presented 80% of Ae. aegypti from all locations develop disseminated infection. For Ae. albopictus, more than half became infected or develop disseminated infection. The infection and dissemination rates for US Ae. albopictus are dose-dependent and seem to increase with the titer of the ingested bloodmeal [21, 40, 55]. Vega-Rúa et al. (2014) [31] assessed 35 American Ae. aegypti and Ae. albopictus for three CHIKV genotypes with the titer of 107.5 pfu/ml, including mosquitoes populations from Brazil and Florida. Their study demonstrated that all 35 populations of both Aedes vectors were susceptible to CHIKV infection by all genotypes tested and that CHIKV transmission efficiency was highly heterogeneous in American mosquitoes ranging from 11.1% to 96.7%. Indeed, Ae. albopictus from Rio de Janeiro showed high transmission efficiencies even between geographically close populations, i.e., with some populations being able to transmit infectious viral particles as early as 2 days post-infection. However, the vector competence of Ae. aegypti and Ae. albopictus from Vero Beach was not tested for the Asian lineage of CHIKV, but for Indian Ocean and ancestral ECSA genotypes showed that transmission efficiencies were low (<30%).
The proportion of individuals with saliva infection was substantially lower than those with viral dissemination, suggesting salivary gland barrier(s) [31, 39]. Interestingly, US Ae. aegypti had almost twice as many infected individuals when comparing with the Brazilian population. A contrasting relation was observed for Ae. albopictus, with the Brazilian population reaching 60% of infected individuals against 24.5% from the US population. Thus, observed inherent differences in mosquito-virus interactions for both Ae. aegypti and Ae. albopictus might depend on geographic origin, which might impact disease transmission and contribute to its establishment in areas endemic for DENV and/or ZIKV. It is not clear whether heterogeneity exists in other traits that compose vector capacity, such as adult survival and biting rates, adult density, feeding behavior, and others, which would further influence CHIKV transmission and epidemiology in such areas [6]. Also, we observed that saliva infection declined with length of infection suggesting impaired transmission efficiency among older mosquitoes, most likely attributable to virus modulation of the infection as observed in other studies [56, 57]. Further studies on vector competence of Ae. aegypti and Ae. albopictus should be done to analyze the heterogeneity of dissemination and transmission of CHIKV among different populations of endemic or receptive areas for this arbovirus using a range of viral titers.
|
10.1371/journal.pntd.0001292 | Diagnosis, Clinical Features, and Self-Reported Morbidity of Strongyloides stercoralis and Hookworm Infection in a Co-Endemic Setting | Infections with Strongyloides stercoralis and other helminths represent important, yet often neglected issues in developing countries. Indeed, strongyloidiasis can be fatal, but only a few studies provide information regarding its health relevance in Africa. Moreover, clinical data on symptomatology and typical recognition patterns mainly originate from Western travel clinics.
A cross-sectional epidemiological survey was carried out in a rural part of south-central Côte d'Ivoire. Stool samples from 292 randomly selected individuals were examined for intestinal helminths, using a suite of diagnostic techniques (i.e., Kato-Katz, Baermann funnel, and Koga agar plate). Participants were interviewed with a pre-tested questionnaire and clinically examined. Multivariate logistic regression analysis was done to relate perceived morbidity and clinical findings to helminth infection status.
The prevalence of hookworm and S. stercoralis was 51.0% and 12.7%, respectively. Both infections were strongly associated with each other (adjusted odds ratio, 6.73; P<0.001) and higher prevalences were observed with age. S. stercoralis-infected individuals expressed self-reported morbidity considerably more often than those with hookworm infection. Clinical examination identified high prevalences of various pathologies and detected tendencies to worse health conditions in helminth-infected subjects.
The use of multiple diagnostic tools showed that S. stercoralis and hookworm are co-endemic in rural Côte d'Ivoire and that each infection causes clinical symptoms and sequelae. Our findings are important for (re-)estimating the burden of helminth infections, and highlight the need for integrating epidemiological surveys, rigorous diagnostic approaches, and clinical assessments in the developing world.
| Infections with parasitic worms such as hookworm and threadworm (Strongyloides stercoralis) are widespread throughout the developing world. However, the symptoms caused by parasitic worms are unspecific and little is known about clinical presentations in endemic countries, and hence doctors' awareness of these diseases is usually low. Many infections therefore remain undetected and untreated over long time periods. As a consequence, parasitic worms can impair the well-being of infected individuals and cause harmful, sometimes even fatal health outcomes. To improve the knowledge about clinical signs and symptoms caused by parasitic worms, we administered a questionnaire to 292 children and adults in rural Côte d'Ivoire, examined them clinically, and looked at their stool for parasitic worm infections. We found that people with parasitic worms reported some symptoms, especially disorders of the gastrointestinal tract, more often than those without worms. Our clinical examination showed a trend toward worse health conditions in worm-infected people, particularly in those with S. stercoralis. Our results are important to improve patient management and control measures, and to better estimate the true health impact of parasitic worm infections.
| Infections with the nematodes Strongyloides stercoralis and hookworm (Ancylostoma duodenale and Necator americanus) are widespread in tropical and subtropical areas, accounting for a remarkable, yet underappreciated burden [1]. For example, hookworm infection is a leading cause of anemia [2], whereas S. stercoralis is able to maintain itself for decades within its host and may cause a lethal hyperinfection syndrome among immunosuppressed patients [3], [4]. Both infections belong to the heterogeneous group of the neglected tropical diseases (NTDs), which are notoriously under-researched and control efforts not adequately funded. Lately, more attention has been addressed to the NTDs [5], [6], which is justified by their cumulative global burden (>50 million disability-adjusted life years (DALYs) lost annually) [7]. The ‘true’ burden of the NTDs might even be higher [8], but considerable uncertainty remains. For example, estimates regarding the global burden due to hookworm infection range between 60,000 DALYs and 22.1 million DALYs, whereas no burden estimates are available for strongyloidiasis [9]. The lack of accurate diagnostic tools is an important consideration, which explains the often unsatisfactory epidemiological mapping of NTDs [10], [11]. With regard to strongyloidiasis, most information regarding signs, symptoms, and recognition patterns originate from examinations of returning travelers in Western travel clinics [12], and refugee populations in the North [13].
Surprisingly, biomedical research in endemic settings has mainly focused on chronic effects (e.g., stunting and nutritional deficiencies) of helminth infections, while little is known on acute impairment, despite the fact that clinical manifestations may profoundly differ between travelers and individuals from endemic areas. This has been shown for S. stercoralis [14], which may cause a severe and fatal hyperinfection syndrome, particularly in patients with HTLV-1 infection, on steroid medication and other immunosuppressive drugs, such as organ transplant recipients. While a higher suspicion index of strongyloidiasis is likely to improve early diagnosis and adequate treatment, epidemiological data are scarce in tropical countries [15]. It is currently estimated that between 30 and 100 million people are infected with S. stercoralis [1], but little is known about arising consequences in endemic areas. What symptoms does a S. stercoralis infection cause in tropical settings? When should a clinician suspect strongyloidiasis? Are individuals affected by hookworm or S. stercoralis prone to co-infections and co-morbidities? Answers to these questions will deepen our understanding of the true impact of nematode infections and result in improved patient management.
Here, we show a multi-pronged approach to elucidate clinical features of S. stercoralis and hookworm infections in a co-endemic area of Côte d'Ivoire. Our assessment is based on a cross-sectional epidemiological survey and in-depth laboratory investigations, complemented by a morbidity-centered questionnaire survey and standardized clinical examination.
The study protocol was approved by the institutional research commissions of the Swiss Tropical and Public Health Institute (Swiss TPH; Basel, Switzerland) and the Centre Suisse de Recherches Scientifiques en Côte d'Ivoire (CSRS; Abidjan, Côte d'Ivoire). Ethical approval was granted by the ethics committee of Basel (EKBB; reference no. 316/08), and the Ministry of Higher Education and Scientific Research in Côte d'Ivoire (reference no. 124/MESRS/DGRSIT/YKS/sac).
Political and health authorities were informed about the purpose and procedures of the study. Participants were informed about their involvement, including possible concerns and the right to withdraw at any moment without further obligations. Adult participants and the parents/legal guardians of participating children (aged<16 years) signed a written informed consent. At the end of the study, all individuals were invited to learn about their parasitological results and free treatment was offered to participants infected with helminths (i.e., single 400 mg dose of albendazole against hookworm and other soil-transmitted helminths, single 200 µg/kg dose of ivermectin against S. stercoralis, and single 40 mg/kg dose of praziquantel against Schistosoma haematobium and Schistosoma mansoni). Due to the high observed prevalence of soil-transmitted helminths, annual treatment with albendazole has been initiated in this study area in early 2010. Ivermectin is also being administered as part of the lymphatic filariasis control program.
The study was carried out in Léléblé and five surrounding hamlets (‘campements’) in May and June 2009. Léléblé is the second largest village that forms part of the recently established Taabo health demographic surveillance system (Taabo HDSS) in south-central Côte d'Ivoire, located approximately 160 km north of Abidjan. A census done at the end of 2008 revealed that the total population in Léléblé was 5,235 people. The tropical climate follows a seasonal pattern with a rainy season from May to October and a dry season from November to April. Taabo HDSS covers the area around Lake Taabo, a man-made reservoir inundated in the late 1970s [16]. Approximately 38,000 people live within the Taabo HDSS, consisting of a small district town (Taabo-Cité), 14 villages, including Léléblé, and over 100 small hamlets.
In May 2009, a population-representative epidemiological survey was conducted in Taabo HDSS. Approximately 7% of the households were randomly selected; for Léléblé and surrounding hamlets this resulted in a sample size of 351 individuals. Allowing for a drop-out rate of 20–25%, we assumed that complete data records from at least 260 individuals would be available for appraisal of the population prevalence of S. stercoralis in Léléblé and surrounding hamlets with reasonable accuracy [17], [18].
Collection containers were distributed and study participants invited to submit a urine sample and a lemon-sized portion of their morning stool the following day. Samples were collected at a public spot between 08:00 and 12:00 hours and then transferred to a laboratory in Taabo-Cité, 28 km east of Léléblé.
Stool samples were examined for the presence of helminth eggs or larvae, using a suite of quality-controlled diagnostic methods: Kato-Katz [19], Baermann funnel (BM) [20], and Koga agar plate (KAP) [21]. Standard protocols of these techniques have been described elsewhere [22]. Additionally, for sufficiently large samples, ∼2 g of stool was put in 15 ml Falcon tubes filled with 10 ml of 5% formalin and, within 4 months, examined using the formalin-ether concentration technique (FECT), and the Flotac-400 dual technique. The comparison of these two methods for diagnosis of intestinal protozoon infections has been presented elsewhere [23].
From each stool sample, duplicate 41.7 mg Kato-Katz thick smears were prepared with slides read after a clearing time of 30–45 min. The number of helminth eggs was counted and recorded for each species separately. Infection intensity was derived by multiplying egg counts by a factor 24 and expressed as eggs per gram of stool (EPG). The exact number of observed hookworm and S. stercoralis larvae was recorded in the BM and the KAP tests.
Urine samples were analyzed for the presence of S. haematobium eggs. In brief, samples were vigorously shaken and 10 ml was filtered through a 13 µl nylon filter with a syringe. Filters were put on microscope slides, a drop of Lugol added, and S. haematobium eggs counted under a microscope by experienced laboratory technicians. For quality control, a random sample of ∼5% of all slides was re-examined by a senior technician.
We searched the literature for common signs, symptoms, and complaints of hookworm and S. stercoralis infection. These symptoms were included in a questionnaire, which consisted of four parts: (i) demographic and anthropometric measures (e.g., sex, age, occupation, educational attainment, individual's height and weight); (ii) personal hygiene behavior (e.g., source of drinking water, frequency of contact with Lake Taabo, availability of toilets, regular use of soap, and wearing shoes); (iii) nine specific questions pertaining to past and recent medical history (e.g., chronic or infectious diseases such as asthma, diabetes, tuberculosis; hospitalization during the last 2 months; antimalarial and anthelmintic treatment in the last 2 months); and (iv) perceived state of health (recall period: 2 weeks), including questions concerning any disorders in the gastrointestinal tract (i.e., abdominal pain, abdominal distension, diarrhea, constipation, weight loss, abdominal cramping, nausea, vomiting, and blood in stool), the lungs (i.e., dyspnea, cough for more than 2 weeks, fatigue, uncommon expectoration), and the skin (i.e., cutaneous rash, migrating eruptions, pruritus). The questionnaire was pre-tested in a neighboring village. Study participants were interviewed by experienced field enumerators employed by Taabo HDSS.
A standardized clinical examination was performed by the first two authors (SLB and BS). The examination comprised an evaluation of a participant's general habitus, conjunctival inspection, palpation of the abdomen and the thorax, cardiac and pulmonary auscultation, presence and grade of hepatomegaly and splenomegaly, and presence of jaundice. Moreover, the skin was carefully examined for any signs compatible with helminth infection (larva currens, larva migrans, rash, and pruritus).
Data were double-entered, cross-checked in Excel version 10.0 (edition 2002, Microsoft Corporation) and analyzed using STATA version 10.0 (StataCorp.; College Station, TX, USA). Stool or urine samples found positive for a specific helminth infection by any of the employed diagnostic techniques were considered as true-positive, leading to the prevalence results of each method. The combined results of the different techniques served as diagnostic ‘gold’ standard and were used as an estimate of the ‘true’ prevalence. Prevalence, sensitivity (proportion of true-positives identified as positive), and negative predictive value (NPV) were calculated, including 95% confidence intervals (CI) to quantify statistical uncertainty. Infection intensities were classified according to WHO thresholds [24] on the basis of the mean EPG. Distributional differences were assessed by Pearson's χ2 and Fisher's exact test, as appropriate. For statistical significance P<0.05 was used throughout.
Self-reported morbidity results from the questionnaire and findings from the clinical examination were analyzed for associations with helminth infections by univariate logistic regression and odds ratios (OR) were computed. Multivariate logistic regression modeling was used to estimate significant associations between S. stercoralis or hookworm infection and findings from the questionnaire survey and the clinical examination. Outcomes were defined as species-specific helminth infection. Possible explanatory variables were included in the final model if they (i) were biologically plausible, (ii) were present in at least 5% of the study participants, and (iii) showed an association (P<0.2) with infection status in a univariate logistic regression analysis. Adjusted ORs were calculated in order to reveal associations between a specific helminth infection and morbidity indicators.
Complete parasitological and clinical data were available from 292 out of 351 randomly selected individuals, owing to a compliance of 83.2%. There were more females (n = 155, 53.1%) than males (n = 137, 46.9%). The median age of our cohort was 13 years (mean, 20.1 years; range, 2 months to 75 years), thus constituting a representative sample of the village (46% of the population aged<15 years, according to the 2008 census).
Eight different helminth species were detected (Table 1). Hookworm was the predominant species; the overall prevalence was 51.0% and infection intensities were mainly light (<2,000 EPG, 95.1%). S. stercoralis larvae were found in fecal samples of 37 individuals (12.7%). S. haematobium and Ascaris lumbricoides showed prevalences of 8.5% and 5.1%, respectively. The remaining helminth species were found in less than 5% of the participants.
Table 2 shows the comparison of the BM and KAP methods for the diagnosis of S. stercoralis. While the BM technique identified 26 out of 37 infections, 17 infections were detected by the KAP test, and hence the BM technique showed a considerably higher sensitivity than KAP (70.3% vs. 46.0%). NPVs were similar for the two methods (95.9% vs. 92.7%). With regard to hookworm diagnosis, duplicate Kato-Katz thick smears showed a sensitivity of 69.1%, which was slightly higher than a single KAP test (61.8%). Compared to our diagnostic ‘gold’ standard, a combination of duplicate Kato-Katz thick smears and a single KAP test resulted in a prevalence of 46.2% (95% CI, 40.4–52.1%). While S. stercoralis infection rates did not differ significantly by sex, hookworm was more frequently observed in males than females (57.7% vs. 45.2%; χ2 = 4.55, P = 0.033; Table 3). The prevalence of both helminth infections increased with age. While only 5.2% of all children aged<5 years were diagnosed with S. stercoralis infection, 17.9% of all adolescents and young adults (aged 15–29 years), and every fifth adult and elderly (>45 years) were found positive for this helminth species. A similar relationship between age and infection prevalence was detected for hookworm; the prevalence rose from 36.8% in early childhood to 51.4% in school-aged children before reaching a plateau around 60% in adults. The infection prevalence in people working as farmers in agriculture or as merchants in trade was disproportionately higher than in individuals with other professions. A multivariate logistic regression analysis revealed a highly significant association between S. stercoralis and hookworm infection (adjusted OR, 6.73; 95% CI, 2.51–18.08, P<0.001).
Table 4 summarizes risk factors and health-related behaviors, in relation to an infection with either S. stercoralis or hookworm. Non-infected individuals stated more often to have any grade of school education (46%; P = 0.116), to have access to sanitation (36%; P = 0.108), and a recent history of anthelmintic treatment (9%; P = 0.008) compared to their infected counterparts. Among infected individuals, their stomach was more often considered as a major health problem (S. stercoralis-infected, 35%; hookworm-infected, 23%) than among individuals who had neither a hookworm nor a S. stercoralis infection (15%; P = 0.039). Different gastrointestinal symptoms such as severe abdominal pain, blood in stool, and diarrhea were more often reported among helminth-infected individuals.
Multivariate logistic regression analysis revealed that some risk factors and symptoms were associated with S. stercoralis, hookworm, or both infections concurrently (Table 5). Individuals complaining about frequent stomach ache (adjusted OR, 2.35; 95% CI, 0.98–5.62, P = 0.056) were more likely to be infected with S. stercoralis, while self-reported diarrhea (adjusted OR, 1.89; 95% CI, 0.98–3.65, P = 0.057) and “working as a farmer in agriculture” (adjusted OR, 2.62; 95% CI, 1.27–5.40, P = 0.009) were risk factors for a hookworm infection.
Hepatomegaly and splenomegaly were frequently observed in the study population with prevalences of 31.8% and 31.2%, respectively. Abnormal findings on pulmonary auscultation (i.e., wheezing, rhonchi, crackles and rales, bronchial breathing over the chest) were noted in one out of five participants (20.5%).
We only found numeric, non-significant differences between S. stercoralis-infected, hookworm-infected, and participants with neither of these infections (Figure 1). A general poor state of health characterized by inadequate personal hygiene and squalidness was more frequent in S. stercoralis-infected individuals (13.5% vs. 4.7%), as well as the uncommon sign “wheezing on pulmonary auscultation” (5.4% vs. 2.7%). However, a combination of such signs was seldom present, thus hampering an exact differentiation of helminth infections by a combination of specific signs. More than half (54.1%) of all S. stercoralis-infected individuals were found to suffer from either pulmonary wheezing, abdominal pain (as determined by the questionnaire survey or found on examination), or a general malaise, while one of these three signs was present in only 26.9% of their hookworm-infected counterparts. Anthropometrics revealed no statistically significant differences between helminth-infected and non-infected individuals (data not shown).
The study reported here from a rural part of south-central Côte d'Ivoire confirms that hookworm and S. stercoralis are co-endemic [25]. While every second participant was infected with hookworm, S. stercoralis infections were found in 12.7% of our cohort. Interestingly, S. stercoralis was exclusively found in the main study village (Léléblé), but not in the surrounding hamlets, indicating that this nematode species requires distinct environmental and biologic characteristics to cause human infection. Moreover, we found that both hookworm and S. stercoralis disproportionately affect the poorest population segments, i.e., those with no or limited education, lack of personal hygiene, and the worst self-reported and clinically-examined health conditions. While the prevalence of hookworm increases with age, other helminths mainly affect school-aged children [2]. In the present work, the S. stercoralis age-prevalence curve showed an almost parallel shape to the hookworm age-prevalence curve, reaching an infection rate of approximately 20% in older adults.
The highly significant association between hookworm and S. stercoralis (OR>6), and the series of self-reported morbidity markers associated with these two helminth species are particularly noteworthy. Interestingly, the perceived health impact of strongyloidiasis appeared to be greater than morbidity due to a hookworm infection, especially regarding gastrointestinal symptoms, skin rash, and abnormal wheezing on pulmonary auscultation. These findings are coherent with different disease stages of strongyloidiasis, thus indicating an entitative and reproducibly measurable morbidity in infected subjects who live in endemic areas. This observation opposes the widely expressed opinion that most S. stercoralis-infected individuals residing in tropical endemic areas remain asymptomatic [3]. The assessment of acute, impaired well-being is a relevant consideration for the development of accurate disease burden estimates for strongyloidiasis and hookworm disease, particularly in view of the often contradictory data on the long-term effects due to soil-transmitted helminthiasis, e.g., cognitive performance and physical development of infected school-aged children [26]–[29]. Factual data derived from clinical examinations in endemic areas may be more appropriate than questionnaire surveys investigating self-reported morbidity, because many complaints, which are commonly imputed to helminth infections, are also frequently caused by other infective agents and/or nutritional deficiencies. Moreover, medical examinations in zones of rural poverty may allow identification and differentiation of certain helminth infections by a distinct cluster of clinical signs, thus superseding the need for detailed laboratory investigations, which are seldom available in resource-constrained areas.
Our study suffers from several limitations that are offered for consideration. First, the prevalence of hookworm and S. stercoralis infections is likely to be underestimated due to (i) the low sensitivity of stool microscopy [10], [22], and (ii) the examination of only one stool sample per person, as the output of helminth eggs and larvae in fecal material shows considerable day-to-day variation [22], [30]. However, we used a suite of diagnostic techniques to enhance sensitivity, e.g., KAP test and BM technique for the diagnosis of S. stercoralis. The BM funnel, which is considered the method of choice for S. stercoralis diagnosis when analyzing stool samples [31], was indeed more sensitive than the KAP test, confirming recent studies carried out in the People's Republic of China [32], Zanzibar [22], [33], and Uganda [34]. However, accurate diagnosis remains a critical issue, even when techniques other than stool microscopy are employed [35]. The usefulness of serology in endemic areas has been discussed controversially [36], and a recently developed polymerase chain reaction (PCR) assay [37], which may become a new diagnostic ‘gold’ standard, needs to be validated in different epidemiological settings prior to larger scale use. Molecular techniques still have a long way to go until they might become viable alternatives for diagnosis of NTDs that are intimately connected to poverty. Second, a larger sample size may have revealed a more distinct clinical recognition pattern of the helminth infections investigated here, particularly for S. stercoralis, which we expected to be even more prevalent in the study region, based on preliminary results obtained from five villages in the Taabo HDSS study area [25].
In conclusion, S. stercoralis and hookworm infections represent important health threats in tropical regions and both parasites cause morbid sequelae. While especially strongyloidiasis is increasingly recognized as an important cause of morbidity and persistent mortality in some parts of the world [4], its diagnosis remains a challenge and clinical features in resource-constrained settings are still neglected [35], [38]. Our study revealed the difficulty to properly assess helminth infections and proposes clinical in-depth studies to factually substantiate the estimated helminth disease burden in order to obtain a more accurate picture of their relevance in Africa. In view of changing health care patterns in tropical countries, including broader access to steroid treatment, an increasing number of the potentially fatal hyperinfection syndrome has to be expected in Africa and elsewhere. Hence, there is a need to raise physicians' awareness of strongyloidiasis to finally improve the patient management and outcome of this important, but widely underrecognized infection.
|
10.1371/journal.pbio.1001703 | Translational Profiling of Clock Cells Reveals Circadianly Synchronized Protein Synthesis | Genome-wide studies of circadian transcription or mRNA translation have been hindered by the presence of heterogeneous cell populations in complex tissues such as the nervous system. We describe here the use of a Drosophila cell-specific translational profiling approach to document the rhythmic “translatome” of neural clock cells for the first time in any organism. Unexpectedly, translation of most clock-regulated transcripts—as assayed by mRNA ribosome association—occurs at one of two predominant circadian phases, midday or mid-night, times of behavioral quiescence; mRNAs encoding similar cellular functions are translated at the same time of day. Our analysis also indicates that fundamental cellular processes—metabolism, energy production, redox state (e.g., the thioredoxin system), cell growth, signaling and others—are rhythmically modulated within clock cells via synchronized protein synthesis. Our approach is validated by the identification of mRNAs known to exhibit circadian changes in abundance and the discovery of hundreds of novel mRNAs that show translational rhythms. This includes Tdc2, encoding a neurotransmitter synthetic enzyme, which we demonstrate is required within clock neurons for normal circadian locomotor activity.
| The circadian clock controls daily rhythms in physiology and behavior via mechanisms that regulate gene expression. While numerous studies have examined the clock regulation of gene transcription and documented rhythms in mRNA abundance, less is known about how circadian changes in protein synthesis contribute to the orchestration of physiological and behavioral programs. Here we have monitored mRNA ribosomal association (as a proxy for translation) to globally examine the circadian timing of protein synthesis specifically within clock cells of Drosophila. The results reveal, for the first time in any organism, the complete circadian program of protein synthesis (the “circadian translatome”) within these cells. A novel finding is that most mRNAs within clock cells are translated at one of two predominant circadian phases—midday or mid-night—times of low energy expenditure. Our work also finds that many clock cell processes, including metabolism, redox state, signaling, neurotransmission, and even protein synthesis itself, are coordinately regulated such that mRNAs required for similar cellular functions are translated in synchrony at the same time of day.
| Genetic studies carried out in several model systems have provided seminal knowledge about the biochemistry of the circadian molecular oscillator and the neural circuitry regulating circadian behavior. The best characterized circadian oscillators consist of transcriptional/translational feedback loops (TTFLs) [1], although nontranscriptional oscillators (NTOs) exist in organisms ranging from unicellulars to Drosophila and humans [2]–[4]. In Drosophila and mammals, a well-characterized TTFL oscillator consisting of several canonical clock genes regulates circadian behavioral rhythms (reviewed in [1]). Similarly, transcription of many (perhaps most) genes is orchestrated by the circadian clock, based on gene profiling studies carried out in Drosophila, mammals and plants. Only a few studies, however, have documented cell-type–specific transcriptional rhythms [5]–[7], due to methodological limitations. Most of those studies utilized Fluorescence-Activated Cell Sorting (FACS), the manual isolation of identified cells, or cell-specific transcriptional profiling techniques, but such methods are either not applicable to all cell populations or lack the sensitivity to detect the entire transcriptome; nor do they distinguish between ribosome-bound (i.e., translating) and soluble mRNA without the use of polyribosome isolation.
Drosophila is an excellent model for cell-type–specific profiling of clock cells because of its outstanding genetics and well-characterized circadian system. Studies have described the fly circadian molecular oscillator [8] and the circadian neuronal circuitry [9], revealing molecular and functional differences among groups of pacemaker neurons that mediate morning and evening bouts of activity or responses of the clock to environmental cues [5],[10]–[18]. To date, no study has documented genome-wide expression profiles for all clock cells of the fly head or the complete translatome of such cells. In this study, we describe use of the Translating Ribosome Affinity Purification (TRAP) method [19] to define the rhythmic translatome of circadian clock cells. Our results reveal a daily synchronization of protein synthesis and identify novel cycling mRNAs within clock cells that are required for diverse physiological processes.
Previous studies have shown that TRAP reflects the translational status of mRNAs in a manner similar to that of conventional polyribosomal analysis [19]. In addition, a recent study in Drosophila indicates that an EGFP-L10a fusion incorporates into polysomes and can be employed for cell-specific translational profiling [20]. To employ TRAP in our studies, we generated Drosophila strains carrying a UAS-EGFP-L10a transgene insertion (see Materials and Methods). Using a pan-neuronal driver (elav-Gal4), we found that the EGFP-L10a fusion has a cytoplasmic/nucleolar pattern of localization in neurons (Figure 1A–C), consistent with incorporation into ribosomes. Indeed, the ring shape pattern in nucleoli (seen in the nucleus of Figure 1A) likely results from expression in the Granular Component (GC, Figure 1D), a structure within which ribosomal proteins assemble into functional ribosomes. As expected, EGFP-L10a was localized in all neurons of the adult nervous system (Figure 1E). In contrast, the tim-uas-Gal4 driver results in expression within the cytoplasm of clock neurons and glia of the nervous system (Figure 1F) or only clock neurons when combined with repo-Gal80 (Figure 1G), which inhibits expression in all glial cells. A different GFP–Drosophila ribosomal protein fusion (L11) has an identical intracellular localization pattern [21]. In addition, it has recently been shown that our EGFP-L10a fusion localizes to branch points of neuronal dendrites, consistent with incorporation into ribosomes that mediate local protein synthesis [22]. Collectively, these pieces of evidence indicate that EGFP-L10a incorporates into functional ribosomes.
We examined circadian locomotor activity of flies overexpressing the UAS-EGFP-L10a transgene in clock cells (Pigment Dispersing Factor, PDF, or Timeless, TIM) to determine whether there were adverse effects on behavior. As shown in Figure 1H, these files have normal behavioral rhythmicity, indicating that EGFP-L10a does not act in a dominant negative manner even at high levels [Average periods (P) and Rhythmicity Indices (RI) were 23.7±0.08/0.57±0.02, 24.0±0.03/0.55±0.01, and 24.3±0.14/0.50±0.03 for control, pdf-Gal4>UAS-EGFP-L10a, and tim-uas-Gal4>UAS-EGFP-L10a flies; n = 17–30]. Thus, the presence of GFP-tagged ribosomes in clock cells does not affect their function.
We optimized TRAP methods for use with Drosophila and demonstrated that significant amounts of RNA could be immunoprecipitated from head tissues of flies expressing UAS-EGFP-L10a under control of the pan-neural elav-Gal4 or clock cell tim-uas-Gal4 driver (see Materials and Methods). Prior to pursuing genome-wide studies, we wished to determine if our Drosophila TRAP methods could detect bona fide changes in translational status. To ask this question, we employed overexpression of Iron Regulatory Protein (IRP), which is known to repress translation of an unspliced form of ferritin (fer) mRNA by inhibiting binding of the small ribosomal subunit to the message. We generated act5C-Gal4/tub-Gal80ts, UAS-EGFP-L10a/UAS-IRP flies in order to be able to activate expression of the TRAP and IRP transgenes conditionally during larval development (by raising temperature to inactivate Gal80ts, an inhibitor of Gal4). Larvae of this genotype and controls (act5C-Gal4/tub-Gal80ts; UAS-EGFP-L10a) were exposed to 30°C to activate expression of UAS-EGFP-L10a in both genotypes and additionally UAS-IRP in the experimental class. Early pupae were collected for both genotypes and subjected to TRAP coupled with Q-RT-PCR to quantify ribosome-associated fer mRNA (relative to control Rp49 mRNA). Similar to previous studies in Drosophila that employed polysome gradient analysis [23], we observed IRP-induced translational repression of an unspliced but not a spliced form of fer (Figure 1I). Indeed, translation of spliced fer was enhanced slightly by IRP overexpression, similar to that observed from the analysis of a high molecular weight polysome fraction in the previous study [23]. This result shows feasibility for the use of TRAP in Drosophila to detect changes in translational status.
To determine if our methods were able to detect rhythmic changes in the ribosomal association of cycling transcripts, we examined the period (per) and timeless (tim) clock mRNAs. TRAP methods were employed to immunopurify RNA from head tissues of elav-Gal4/UAS-EGFP-L10a flies two times of day (ZT11 and ZT23, the times of high and low per/tim RNA abundance, respectively). Extracted RNA was then subjected to Q-RT-PCR, using gene-specific primers, to detect the clock mRNAs. Figure 1J shows that the abundances of ribosome-bound tim and per clock mRNAs are significantly higher at ZT11 than at ZT23. This result is consistent with the known rhythmic profile of tim and per RNA abundances at the two time points (higher at ZT11) and the expected translational status of the mRNA at the two times of day. We emphasize that Figure 1J shows differences in ribosome association of the clock RNAs, not simply the previously documented RNA abundance for per and tim. In addition, we note that the temporal resolution of our measurements does not exclude translational regulation of per mRNA, which has been suggested in certain studies [24]–[26]. Nonetheless, these results demonstrate that TRAP methods are capable of detecting diurnal changes in the translational status of specific mRNAs.
Using the newly developed methods, we performed TRAP on head tissue lysates of tim-uas-Gal4; UAS-EGFP-L10a flies that were collected at 4-h intervals during the first two days of constant darkness (DD) following entrainment to LD 12∶12. Such flies express the EGFP-L10a fusion in all clock cells of the head, including the ∼150 pacemaker neurons, photoreceptors, and glia. RNA was extracted from affinity-purified samples and used to generate libraries representing all ribosome-associated transcripts (see Materials and Methods). TRAP libraries corresponding to six different times of the circadian cycle (CT0, 4, 8, 12, 16, and 20) were independently constructed for DD1 and DD2 (see details in Materials and Methods). Libraries were sequenced, using a multiplexing strategy, to produce single end, 100 base sequencing reads; these were mapped to the Drosophila reference genome and analyzed as described in Materials and Methods.
We employed two recently developed programs, JTK_CYCLE and ARSER [27],[28], to compare their usefulness for detecting circadian rhythms in the ribosome association of mRNAs. Using criteria and statistical cutoffs described in the Materials and Methods section, 1,195 and 263 translationally cycling mRNAs were detected by the ARSER and JTK_CYCLE programs, respectively. Interestingly, the majority of the cycling mRNAs (203 out of 263) detected by JTK_CYCLE were also detected by the ARSER program (Figure 2A), indicating consistency of the two analyses. Figure S1 shows robust cycling for eight mRNAs out of the 60 identified by JTK_CYCLE but not ARSER. Thus, JTK_CYCLE may identify cycling mRNAs not detected by ARSER. Table S4 lists the 1,255 mRNAs that were identified as exhibiting significant translational cycling by either program (mRNAs identified by both programs are indicated in bold). The False Discovery Rate (FDR) calculated by the ARSER program at the relevant p value was 0.148, indicating that approximately 186 mRNAs are false positives. This FDR is quite low relative to other recent genome-wide studies of cycling mRNAs [29]–[31]. We did not compute an FDR for the JTK_CYCLE program, because 203/263 mRNAs identified by JTK_CYCLE are included in the ARSER dataset, and therefore the latter dataset represents a good approximation of FDR for our analyses. Based on the ARSER analysis, we estimate that approximately 1,069 of these mRNAs show circadian changes in translation in clock cells of the adult head, representing about 10% of all analyzed genes in the genome. This large number of cycling mRNAs is consistent with recent studies utilizing manual dissection approaches to perform cell-specific transcriptional profiling of the Drosophila PDF clock neurons [5],[10]. Cell-specific profiling methods may identify a larger number of cycling Drosophila mRNAs, relative to previous studies, due to a more homogeneous starting cell population (i.e., clock cells).
We examined a number of mRNAs in our datasets that had previously been shown to exhibit abundance rhythms to assess the quality of our datasets. These include both clock and clock-regulated mRNAs (per, tim, vri, clk, to, fer2, slob, ugt35b, 5-HT1A, bw, Ir, and WupA). All showed translational rhythmicity (Figure 2B) with an expected phase, although the tim rhythm damped on DD2. Figure 2C, for example, shows robust rhythmicity in the sequence reads for per and lack of rhythmicity for a nearby gene. Our analysis also revealed translational cycling for many other genes that express rhythmic mRNAs. For example, our list of mRNAs includes 49 of 420 mRNAs showing circadian abundance rhythms identified in five previous microarray-based studies (see Introduction). This comparison does not include a recent study that identified 2,751 cycling mRNAs in hand-dissected PDF neurons [10]; our results include 172 of those mRNAs (see Table S4). Interestingly, Ugt35b mRNA, one of several encoding fly glucuronosyltransferase activity, was previously shown to exhibit transcriptional cycling in head tissues but not in PDF neurons [10]. Given that we employed a clock cell tim-Gal4 driver in our TRAP studies, we suggest that Ugt35b cycles in other clock cells of the head.
We conducted TRAP combined with quantitative PCR for Ugt35b, tim, and 18 novel cycling mRNAs (not previously found to show abundance rhythms in head tissues) to verify results obtained by RNA-seq. As expected, Ugt35b and tim exhibited rhythmicity, presumably a consequence of their mRNA abundance rhythms. Of the novel mRNAs, 15/18 showed rhythmic changes in translation, with a profile very similar to that observed with RNA-seq analysis (Figure S2). We further analyzed cycling of a number of these mRNAs in the per0 mutant, which lacks a functional clock, during the first day of constant darkness (DD1). We found that rhythmic expression of these mRNAs was abolished in the per0 mutant, confirming their circadian clock regulation (Figure S3).
Previous genome-wide studies showed that peaks of mRNA abundance occur at many different circadian phases (see Figure S4). In contrast, our profiling of the clock cell translatome revealed a striking feature of circadianly regulated protein synthesis. We found that peak translation for most of the 1,255 mRNAs identified in our study occurs predominantly during two circadian phases: midday or mid-night (Figure 3A–C; Figure S4). These are times of relative behavioral quiescence and just prior to initiation of locomotor activity bouts (Figure 3A, lower panel). Thus, protein synthesis may be confined to times of day that require reduced metabolic expenditure and/or are just prior to initiation of behavioral activities. A further analysis revealed surprisingly synchronized translation of mRNAs required for the same cellular process: translation is predominantly unimodal (with a peak during the day or night) or biomodal, depending on the process (Figure 3C). This bias in the timing of translation was true of many other cellular processes (Figure 3D). For example, of the 10 enzymes involved in glucose metabolism in our list of cycling RNAs, nine are translated during the day. In contrast, all 10 GPCRs in our list are translated during the night (Figure 3E).
Of note, mRNAs encoding a number of translational initiation factors (eIF4E isoforms) exhibit cycling with a phase that corresponds to the daytime peak of circadian translation (Figure S5). Thus, circadian translation of these eIFs may contribute to a broad clock regulation of protein synthesis. In contrast, the major initiation factor, eIF4E-1, does not exhibit translational cycling, suggesting that it does not participate in circadian regulation (Figure S5). Consistent with previous results indicating that ribosome biogenesis is regulated in a circadian manner [32], 20 mRNAs encoding ribosomal proteins, translation initiation factors, or other translational regulatory components show translational rhythmicity (Table S4).
The synchronized rhythmic expression profiles identified by our cell-specific profiling approach may result from a clock regulation of translation or mRNA abundance. To ask whether changes in translational status contribute to the synchronization of gene expression in clock cells, we carried out additional studies, using TRAP/RNA-seq methods.
We reasoned that total RNA isolated from whole heads contains mRNAs from both clock and nonclock cells. Thus, if a gene is robustly expressed in nonclock cells, the abundance profile obtained from whole head total RNA will not represent its expression profile in clock cells. However, for mRNAs predominantly expressed in clock cells (such as per, tim, and others), assays of total head RNA will reflect clock cell expression. Such an mRNA ought to show enrichment in a TRAP sample from tim-uas-gal4>EGFP-L10a heads relative to total RNA from the starting lysate, and the circadian expression profile, when assayed from total RNA, should approximate the profile in clock cells. Thus, if such an mRNA shows a rhythm by TRAP but not in total RNA, then it is likely to be regulated at the translational level.
To identify mRNAs enriched in clock cells, we created new genome-wide libraries for TRAP and total RNA samples from head tissues of tim-uas-gal4>EGFP-L10a–expressing flies. These were sequenced to identify mRNAs that are substantially enriched by TRAP relative to total RNA—that is, enriched in clock cells. We identified many that show an enrichment within clock cells similar to or greater than that observed for tim mRNA. Forty-nine of them are present in our previous list of cycling mRNAs. We chose 12 cycling mRNAs from the enriched list and examined their expression profiles in total RNA versus TRAP RNA samples using Q-PCR methods. Of the 12 mRNAs tested, three did not show cycling similar to that detected by RNA-seq analysis (25%, and the same as we reported for another set of RNAs; Figure S2); thus, these three were not examined further. Of the remaining nine mRNAs, which showed cycling by Q-PCR similar to that detected by RNA-seq, three of them exhibited constant abundance in total RNA but showed circadian cycling in ribosome association, indicating that they are likely regulated at the translational level. Figure 4 shows cycling profiles for these three mRNAs and a fourth mRNA showing both abundance and ribosome-association rhythms (Figure 4D). Thus, for certain mRNAs, there is good evidence for a clock regulation of translation.
We manually annotated the proteins encoded by the 1,255 cycling RNAs using information obtained from Flybase and classified them by biological process (Figure 5A). Of the annotated genes, the most represented functional class is metabolism/energy production, including NAD-dependent processes and oxidation-reduction reactions. This class includes 85 genes involved in intermediary metabolism, 14 genes with mitochondrial functions, and 46 genes that regulate oxidation-reduction processes. These results are consistent with Drosophila and mouse circadian transcriptional profiling studies that identified a large subset of metabolic genes [33],[34]. Another overrepresented group is signaling (including both intracellular pathways and intercellular signaling mechanisms). Interestingly, 44 members of the signaling class belong to the G Protein signaling family, represented by many G Protein Coupled Receptors (GPCRs) and GTPases.
Several particularly interesting cycling mRNAs encode proteins that potentially modulate the NADP+/NADPH ratio or are known components of the cellular redox (thioredoxin) system. Examples include the CG3483 and CG7755 genes, both predicted to encode isocitrate dehydrogenase-like proteins. While at least one isocitrate dehydrogenase (IDH) is a component of the mitochondrial citric acid cycle, others have a cytoplasmic localization, producing αketoglutarate with a conversion of NADP+ to NADPH [35]. We also found that the mRNA encoding Glutathione Transferase E10 (GstE10), which utilizes the redox regulator glutathione in detoxification reactions, exhibits a translational rhythm (Table S4). Interestingly, it was recently shown that glutathione and a different Gst mRNA (GstD1) show circadian changes in abundance in Drosophila head tissues [36], suggesting a complex regulation of redox state.
Components of the thioredoxin (TRX) system, a general regulator of cellular redox state, are also under circadian control. Thioredoxin T (TrxT) and Thioredoxin reductase (Trxr-2) mRNAs show robust circadian changes in translation, with peaks in the late subjective day (Figure 5B). This circadian translation may reflect an underlying transcriptional control as both TrxT and Trxr-2 show mRNA abundance rhythms in Drosophila head tissues (Figure S6). Of interest, it was previously suggested that TrxT showed an mRNA abundance rhythm within the Drosophila PDF clock neurons, but this was based only on examination of two circadian times in a screen for cycling mRNAs [10]. Thioredoxin reductases are known to catalyze reduction of thioredoxin, in the process converting NADPH to NADP+ [37], an important regulator of cellular redox. In addition to these TRX system genes, Grx-1, a glutaredoxin also involved with cell redox state homeostasis, shows circadian translational cycling (Table S4). Rhythmicity in cellular redox state is significant as it regulates many biochemical processes including circadian transcription factors (see Discussion).
Previous studies have indicated that synaptic vesicle cycling mechanisms are important within clock neurons [38] and glial cells [39] for circadian oscillator or output functions. Similarly, there are reciprocal interactions between the oscillator and neuronal membrane events, including ion channel activity, that are critical for timekeeping in Drosophila and mammals [6],[40],[41]. It is of interest that we identified mRNAs encoding at least 20 ion channels or channel regulatory proteins that exhibit rhythms in ribosome association. These include cac (Ca2+ channel), Ir (K+ channel), SK (K+ channel), Slob (K+ channel regulator), and inaF-B (Trp channel regulator), although Ir showed significant rhythmicity only during day 1 of DD. Interestingly, however, Ir was identified in a previous study as a rhythmic mRNA within PDF neurons that is important for oscillator function [6]. Likewise, a number of mRNAs encoding vesicle trafficking or release proteins, including exo70, syn, and unc-104, exhibited rhythmicity in our experiments.
We note that at least two potential brain glial mRNAs were revealed in our study: CG9977 and CG6218. CG9977 encodes adenosylhomocysteinase activity, whereas CG6218 encodes an ATPase. Both were identified in a previous microarray-based screen for Drosophila mRNAs enriched in glial cells [42], and are known to be expressed in the adult brain according to FlyAtlas [43]. The CG9977 enzymatic activity converts S-adenosyl-L-homocysteine to L-homocysteine and adenosine, the latter a known mammalian gliotransmitter [44]. As the tim-uas-gal4 driver is expressed in neurons and glia (including astrocytes) with PER-based oscillators, CG9977 and CG6218 may be expressed in the latter cell type.
Tdc2 encodes the neurally expressed isoform of tyrosine decarboxylase, which converts tyrosine to tyramine, the latter compound acting as a substrate for octopamine synthesis. In Drosophila, both tyramine and octopamine serve as neurotransmitters, regulating diverse functions including adult locomotion, male aggression, male courtship, drug sensitivity, ovulation, circadian activity rhythms, and appetitive memory formation [45]–[49]. Therefore, it is of interest that Tdc2 mRNA exhibits circadian translational rhythms in clock cells (Figure 6A; Table S4). We verified the circadian translation of Tdc2 using the TRAP technique coupled with Q-RT-PCR (Figure 6B), and showed that this rhythm is abolished in per0 flies (Figure S3). Using an anti-TDC2 antibody, we further verified that expression of the TDC2 protein exhibits the predicted circadian changes in two major groups of clock neurons, the l-LNvs and the LNds (Figure 6C and D).
We used two different strategies to characterize the expression pattern of Tdc2 in the adult brain, in particular in various groups of clock cells. First, we characterized the expression pattern of a Tdc2-Gal4 transgene [50] and its co-localization with PERIOD protein. We found that a UAS-GFP reporter, driven by Tdc2-Gal4, was expressed in multiple regions of the fly brain, as expected. However, the only clock cells showing GFP fluorescence (identifiable by PER expression) were the ventral lateral (LNv) PDF neurons of the brain (Figure S7A), which are critical for circadian behavior [9]. Next, using anti-TDC2 antibody, we localized the TDC2 protein in flies expressing a membrane-bound GFP (mCD8-GFP) in all clock cells (driven by tim-uas-gal4). As expected, we found that there is a strong immunoreactive signal for TDC2 in many nonclock neurons. Within the clock neuronal population, we detected TDC2 immunoreactivity in all l-LNvs (Figure S7B, a–d), s-LNvs (a–d), and LNds (i–k), as well as a few cells in the DN1 region (l–n). Finally, a comparison of TDC2 immunoreactivity and Tdc2-gal4 driven mCD8-GFP expression found that Tdc2-gal4 does not express in all TDC2 immunoreactive cells (unpublished data), indicating that the Tdc2-gal4 transgene does not reflect the complete expression pattern of the Tdc2 gene. These results suggest that rhythmic production of TDC2 in various clock neurons, and a consequent rhythm in release of tyramine and/or octopamine from these cells, may be required for normal circadian behavior.
To assess the role of Tdc2 in circadian behavior, we analyzed locomotor activity of the Tdc2RO54 mutant, which carries a point mutation that abolishes the enzymatic activity of TDC2 [50]. Consistent with previous reports [45], we found that Tdc2RO54 mutants displayed decreased activity (Figure 7A). In addition, however, the mutant population exhibited decreased rhythmicity. The average Rhythmicity Index (RI) for Tdc2RO54 was 0.18±0.02 compared to 0.56±0.02 for control flies, and indeed only 20±3% of the mutant population displayed significant free running rhythms, whereas the control population was 100% rhythmic (Figure 7A). We note that decreased activity does not result in arrhythmicity, as there was no correlation between activity level and rhythmic locomotor activity (Figure 7B). These observations indicate that octopamine and/or tyramine are required for normal circadian behavior (see Discussion).
To ask whether the observed arrythmicity of the Tdc2 null mutant results from loss of Tdc2 function in clock cells, we examine circadian behavior in flies with a Tdc2 knockdown specifically in clock cells. We found that populations of flies expressing Tdc2 RNAi, driven by either pdf-gal4 or tim-uas-gal4, were 75% arrhythmic and had low average Rhythmicity Indices (Figure 8A,B,F), whereas control flies were normally rhythmic (Figure 8C–F) Thus, tdc2 is required within clock neurons for normal locomotor activity rhythms.
This study is the first to profile the circadian translatome of a defined cell population in a complex tissue. In contrast to previous studies showing that mRNA abundance rhythms peak at multiple circadian phases (Figure S4), our results indicate that translation of most rhythmic transcripts within clock cells is restricted to two major phases—midday and mid-night. Furthermore, we provide evidence that circadian regulation of either mRNA abundance or protein synthesis (depending on the mRNA) contributes to this synchronization. We speculate that protein synthesis may occur predominantly at circadian phases that are associated with reduced metabolic expenditure. In Drosophila, such times coincide with behavioral quiescence, just prior to initiation of locomotor activity bouts (Figure 3). The synchronized translation of functionally related mRNAs (Figure 3B–D) suggests a clock-orchestrated sequential activation of biological processes; these results reinforce the concept that fundamental cellular processes are under circadian control within clock cells.
Two significant technical improvements enabled the discovery of synchronized translation. First, our analysis was restricted to circadian clock cells, circumventing the problem of profiling a mixed population, in which some cells express a rhythmic mRNA, whereas others express the same mRNA constitutively (thus masking rhythmicity). In addition, different cell types may express out-of-phase rhythmic mRNAs, also masking a rhythm in a mixed cell population. Second, our technique analyzes ribosome association rather than steady-state mRNA abundance, representing a more direct assessment of protein expression. Although it is not currently technically feasible to directly compare transcriptional and translational rhythms in the same cell types, our results indicate that translational regulatory mechanisms contribute to synchronized protein synthesis. Consistent with this idea, we have shown that mRNAs encoding relevant translational regulatory factors are rhythmically expressed. These include translation initiation factors, ribosomal proteins, and enzymes involved in rRNA and tRNA synthesis (Table S4). In mammals, ribosome biogenesis is known to be regulated by the circadian clock [32]. Thus, it is possible that the circadian clock regulates translation of many mRNAs, including those relevant for clock function [25],[26] by controlling availability of the translational apparatus.
We document rhythmic translation of mRNAs within clock cells that is relevant for diverse biochemical and behavioral functions. A particularly interesting class includes factors important for cell redox homeostasis (CG3483, CG7755, TrxT, and Trxr-2), as it has been demonstrated that a clock control of redox state drives rhythms in the excitability of suprachiasmatic nuclei (SCN) neurons [51]. Furthermore, there is redox control of many cellular factors, including enzymes, receptors, cytokines, growth factors, and transcription factors. Thioredoxin, for example, regulates NFkB activity [52], which is known to be under circadian control [53]. NADP(H) and NAD(H), the reduced forms of these cofactors, stimulate DNA binding of the CLOCK/BMAL1 and NPAS2/BMAL1 transcriptional heterodimers, which are critical components of the mammalian circadian clock [54]. Circadian translational regulation of cellular redox may be important for rhythmicity of clock components and clock outputs as well as metabolic feedback to the clock [33],[55].
The rhythm in TrxT translation may also function in another important circadian output. It has recently been demonstrated that there is circadian control of peroxiredoxin (PRX) protein oxidation state in organisms ranging from unicellulars to humans, and that this rhythm is regulated by an uncharacterized NTO (reviewed in [1]). Significantly, oxidized PRX multimers serve as cellular chaperones and cell cycle modulators. Thioredoxin (TRX) mediates reduction of oxidized PRX molecules to complete the PRX catalytic cycle [1], and thus rhythmic TrxT may contribute to circadian changes in PRX oxidation state. Of relevance, mRNAs encoding other chaperones are also rhythmically translated (Table S4).
Rhythmic factors important for neurotransmission were also identified by our analysis. Among them, Tdc2—encoding the synthetic enzyme for tyramine and octopamine—is rhythmically expressed in clock neurons and localized to the PDF cell population. Rhythmic release of these transmitters from PDF or other clock neurons may contribute to the temporal coordination of the clock cell circuitry, similar to the role of PDF [56]. Alternatively, rhythmic release of octopamine and/or tyramine may regulate downstream neurons that drive locomotor activity rhythms.
Of note, previous studies have suggested a clock control of tyramine synthesis, showing that there was decreased tyrosine decarboxylase activity in the brains of per clock mutants [57]. In addition, mutants of several clock genes, including per, clock, cycle, and doubletime, were found to be required for normal cocaine sensitization, a process depending on induction of tyrosine decarboxylase activity and production of tyramine [58]. Expression of Tdc2 in clock neurons is consistent with a role for tyramine, and perhaps octopamine, in this process.
Tdc2 was not described as a cycling mRNA in several previous genome-wide circadian expression studies [59]–[63] that utilized whole fly heads as a starting material. Based on the expression pattern of Tdc2—that is, broad and strong expression in a large number of neurons including only certain clock neurons —it seems likely that the cycling of Tdc2 eluded detection in previous studies because of the presence of other TDC2-containing neurons in which the transcript does not exhibit rhythms in abundance. Indeed, Tdc2 was included in a long list of mRNAs (2,751) showing enrichment in the large LNv clock neurons at one time of day (ZT12) in a recent study that utilized manual dissection procedures to profile PDF neurons [10]. Our detection of rhythmic Tdc2 translation and confirmation of its role in maintaining circadian locomotor activity rhythms clearly demonstrates the advantage of a cell-type–specific approach in genome-wide studies of gene expression.
For translational profiling of flies with a normal circadian clock, males of a homozygous w1118; tim-uas-gal4 stock were crossed to virgin females of a homozygous w1118; UAS-EGFP-mL10a stock. F1 progeny carrying both the UAS and Gal4 transgenes (and expressing EGFP-tagged ribosomes in all clock cells) were collected and used for TRAP experiments. To profile a clock mutant, females from a homozygous per0 w1118; UAS-EGFP-L10a stock were crossed to a Gal4 strain and only male flies from the F1 progeny were used in the TRAP experiments. All flies were reared in a lighting schedule consisting of 12 h of light and 12 h of dark (LD 12∶12) at 25°C and 60% humidity. The tdc2RO54 strain and its isogenic parental strain were gifts from Dr. Jay Hirsh of University of Virginia. UAS-Tdc2 RNAi flies were obtained from the VDRC stock center (stock numbers 10687R-1 and 10687R-3).
The UAS-EGFP-L10a transgene was generated by cloning the coding sequence of the EGFP-L10a fusion protein (provided by Nat Heintz) into the pUAST vector. We chose to use the mouse L10a ribosomal protein (mL10a) because it is virtually the same as fly L10a (identical in size and ∼90% similar)—not surprising for a ribosomal subunit—and it has been shown to work well for the TRAP method. The cloning service was provided by Entelechon (Regensburg, Germany), and the resulting UAS-EGFP-L10a plasmid was verified by sequencing. The UAS-EGFP-L10a plasmid was purified using a Qiagen Maxi-prep kit and then used to generate transgenic flies (Genetic Services, Cambridge, MA). Genomic insertions were mapped to chromosomes using standard segregation analysis procedures.
Adult flies were collected in 50 ml conical tubes at desired time points and flash frozen in liquid nitrogen. Fly heads were collected by vigorously shaking frozen flies and passing them through geological sieves according to standard procedures. Approximately 200 heads were employed for each affinity purification experiment. Frozen heads were homogenized in a buffer containing 20 mM HEPES-KOH (pH 7.4), 150 mM KCl, 5 mM MgCl2, 0.5 mM DTT, 100 µg/ml Cycloheximide, and 2 U/ml SUPERase (Life Technologies) and centrifuged at 20,000× g for 15 min to obtain cleared lysate. After adding DHPC and Igepal CA-630 to a final concentration of 30 mM and 1%, respectively, the lysates were incubated on ice for 5 min and centrifuged at 20,000× g again for 15 min. After centrifugation, the supernatant was applied to magnetic beads coated with a purified high-affinity anti-EGFP antibody (prepared using the Dyabeads Antibody Couple Kit from Invitrogen) and incubated at 4°C with end-to-end rotation for 1 h to allow binding of EGFP-tagged ribosome to the antibodies. Following incubation, samples were washed with a buffer containing 20 mM HEPES-KOH (pH 7.4), 150 mM KCl, 5 mM MgCl2, 0.5 mM DTT, 100 µg/ml Cycloheximide, and 1% Igepal CA-630 for five times at room temperature. RNA was extracted from the beads using the TRIzol reagent (Life Technologies). Quality and quantity of the isolated RNAs were analyzed using a Bioanalyzer (Agilent).
Using these methods, we affinity purified RNA-containing ribosomes from head tissues of adult flies expressing UAS-EGFP-L10a in all neurons or clock cells. Similar to published studies [19], we optimized homogenization procedures for Drosophila head tissues, included magnesium and cycloheximide in the lysis buffer to preserve polysomes, inhibited RNAase activity, and employed a purified, high-affinity anti-GFP antibody for ribosome precipitation. In those experiments, the UAS-EGFP-L10a transgene was expressed in all neurons or all clock cells using, respectively, the elav-Gal4 or tim-uas-Gal4 drivers. In three pilot experiments—two using elav-Gal4 and one using tim-uas-Gal4—we obtained a total of 305–544 ng RNA from head tissues of 200 UAS-EGFP-L10a–expressing flies, whereas there were negligible amounts (50–100-fold less) of precipitated RNA in control samples (elav-Gal4 or tim-uas-Gal4 alone) (Figure S8). Nearly as much RNA was precipitated using the tim-uas-Gal4 driver as with elav-Gal, and we attribute this result to the strength of the tim-uas-Gal4 driver and the observation that it is expressed in all clock neurons including photoreceptors and thousands of glial cells. With expression of UAS-EGFP-L10a in only the clock neuron population (∼150 neurons, some of which can be seen in Figure 1G), we were able to immunopurify 44 ng of RNA from 200 fly heads—10-fold more than control precipitations—indicating good sensitivity for our methods. Expression of a different ribosomal protein fusion, GFP-Drosophila L11 [21], can also be employed for TRAP analysis; we immunopurified 118 ng of ribosome-bound RNA from elav-Gal4/UAS-GFP-L11 head tissues starting with 150 flies (unpublished data).
We employed standard Illumina protocols and reagents (the TruSeq RNA sample preparation kit) for RNA-seq library construction. RNAs extracted from the immunoprecipitation contain a mixture of mRNAs, ribosomal RNAs, and other small RNAs that are involved in translation, such as tRNAs. Using the TruSeq RNA kit, mRNAs were isolated using poly-dT coupled magnetic beads and fragmented by addition of divalent cations at 94°C. Cleaved mRNAs were then reverse transcribed into cDNA using random primers, and cDNA was subjected to second strand synthesis using DNA polymerase I and RNaseH. DNAs were end repaired, “A” tailed, and then ligated to Illumina sequencing adaptors prior to enrichment by PCR to create a library. Sequencing of libraries was accomplished using an Illumina HiSeq 2000 in the Tufts Medical School Molecular Core Facility. Sequence reads were obtained and their quality analyzed using the quality control metrics provided by the FastQC pipeline (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). We obtained, on average, 21 million high-quality 100-b reads for each of the 24 samples (after removing low-quality reads), and an average of 82% of the high-quality reads could be mapped to the Drosophila 5.22 reference genome (Table S1) using Tophat (v 2.0.0) and Bowtie2 (v 2.0.0.5) [64],[65]. The reads represent approximately ∼12,000 genes that are expressed in clock cells of the Drosophila head.
After mapping with Tophat and Bowtie, we counted the number of reads aligning to individual annotated genes in the Drosophila genome using HTseq-count (EMBL). Using these methods, there was good agreement between the two biological replicates for each time point in the DD1 and DD2 datasets. The correlation coefficient (r) of the two replicates was greater than 0.9 for all time points (Table S2, representative scatter plots of two replicates are shown in Figure S9). Next, we conducted a preliminary assessment of each of the four individual datasets (DD1, replicate 1 and 2 and DD2, replicate 1 and 2) by calculating the “best cosine correlation” for all genes including 10 genes in our datasets that are known to show transcriptional cycling from previous studies (Table S3). The “best cosine correlation” is obtained by calculating the correlation coefficient (r) between read counts of the six time points and corresponding six values on one of 48 cosine curves each with 0.5 h difference in phase, and selecting the highest r value from the 48 comparisons. We found that 10 “control” RNAs had high r values in the two DD1 datasets and DD2 dataset 1. However, poor correlation coefficients were observed for DD2 dataset 2, and thus this dataset was not used in our subsequent analyses. Given the good correlation between DD1 datasets 1 and 2, we pooled reads from these two replicates to generate one set of combined expression values. For DD2, we employed only dataset 1 in the analysis. As a consequence of improved sequencing technology, samples of the DD2 dataset 2 contained roughly the same number of reads as the combined DD1 datasets 1 and 2. Thus, the total number of reads analyzed for each sample was similar across all time points of DD1 and DD2—on average ∼32 million reads per sample. The resulting datasets (six time points for both DD1 and DD2) were quantile normalized to control for variation among experiments.
Relative sequence read coverage at different circadian time points, quantified using HTseq-count and quantile normalized, were used to construct a time-lapse expression series and analyzed using two different programs, ARSER and JTK_CYCLE [27],[28], to identify the presence of circadian periodicity. ARSER was developed by Yang and Su [27], and it analyzes circadian expression data by harmonic regression based on autoregressive spectral estimation; JTK_CYCLE was developed by Hughes et al. [28], and uses a nonparametric algorithm to detect rhythmic components in genome scale datasets. Results obtained from the two different analyses were filtered in several ways to obtain the final set of cycling genes: (1) we required the average raw read counts across the 12 time points to be at least 20; (2) we required a “cycling amplitude,” defined as ½ (maximum expression value – minimum expression value)/median expression value, of at least 0.5; and (3) for results with the ARSER program, p<0.021 was considered statistically significant, whereas for the JTK_CYCLE program, p<0.015 was used as a cutoff. As the two programs appear to have different sensitivities in detecting circadian genes, different cutoff p values were chosen for them in order to include the majority of known clock genes. We think the use of this biological criterion to determine a statistical cutoff is reasonable for this type of analysis.
One-week-old adult flies expressing EGFP-mL10a in all clock cells—that is, carrying one copy each of tim-uas-gal4 and UAS-EGFP-L10a—were entrained to a LD 12∶12 cycle for 3 d at 25°C and flash frozen in liquid nitrogen at ZT8 on the 4th day. Three sets of samples, each containing about 200 flies, were collected. Head collection, homogenization, TRAP, and RNA isolation were carried out as described above in “Affinity Purification of Ribosomes and Isolation of Ribosome-Bound mRNAs.” Before the immunoprecipitation step, 1/10 of the tissue lysate was set aside for extraction of total RNA. RNAs were isolated from the TRAP immunoprecipitates (referred to as “TRAP RNA”) as well as from the input whole head lysates (referred to as “total RNA”). Equal amounts (300 ng) of TRAP RNA and total RNA were used to construct RNA-seq libraries. For each of the three sets of fly heads, one TRAP RNA library and one total RNA library were constructed. RNA-seq library construction, sequencing, and mapping were conducted as described above. Sequence read counts were obtained using HTSeq (EMBL) with BDGP5, Ensembl release 68 for gene coordinates. Normalized sequence read counts were used to test for differential expression between the TRAP RNA samples and whole head total RNA samples. Differential expression was determined using the DESeq package for R [66]. Genes that showed significantly increased abundance in the TRAP RNA samples were considered to be enriched in clock cells.
Adult or larval brain and ventral ganglion were dissected in PBS (137 mM NaCl, 2.7 mM KCl, 8 mM Na2HPO4 • 2 H2O, 2 mM KH2PO4, pH 7.4) under a dissecting microscope and fixed in 4% paraformaldehyde. After fixation, tissues were washed three times with PBST (PBS with 0.1% Triton-X-100), blocked with 5% Normal Goat Serum (NGS) in PBST for 3 h, and incubated with primary antibody solution in PBST with 2% NGS at 4°C overnight. Anti-PER, anti-LARK, anti-TDC2, and anti-PDF primary antibodies were diluted 1∶15,000, 1∶2,000, 1∶300, and 1∶20, respectively. The primary antibody solution was removed the next day and tissues were washed five times in PBST and incubated with fluorescence-conjugated secondary antibody for 3 h (Cy3 conjugated goat anti-rabbit secondary antibody from Jackson ImmunoResearch for PER, LARK, and TDC2; Alexa Fluor 488 or Alexa Fluor 647 conjugated goat anti-mouse secondary antibody from Invitrogen for PDF). Following incubation with secondary antibody, tissues was washed five times in PBST and mounted on slides in VECTASHIELD mounting media (Vector Lab).
Florescence microscopy of brains was conducted using either a Leica SP2 confocal microscope at the Tufts Center for Neuroscience Research (CNR) Imaging Core or a Leica SP8 confocal microscope at the Enhanced Neuroimaging Core of the Harvard NeuroDiscovery Center. GFP, Cy3, and Alexa Fluor 647 were excited using laser light of 488 nm, 561 nm, and 647 nm, respectively. Fluorescence excitation and image acquisition in the three different channels were performed in a sequential manner to avoid signal bleed-through between channels. One-micron optical sections were acquired in the vicinity of the LNv and LNd neurons using a 63× oil objective. Brain specimens collected from ZT1 and ZT9 were imaged in an alternating order so that every ZT1 image was paired with a ZT9 image and paired t tests were used in the final statistical analyses of image quantification. Such analyses minimize random variation due to fluctuation in laser power. To quantify TDC2 immunoreactivity in l-LNv and LNd, Regions of Interest (ROIs) were manually selected to include all l-LNv cells or all LNd cells based on PDF immunoreactivity (for l-LNv) or expression of tim-uas-gal4–driven mCD8-GFP in the appropriate region (for LNd). A custom Image J program was used to calculate the average pixel intensity across the entire stack within the ROI for all pixels that had an intensity value greater than that of a manually selected background region.
Quantitative real-time PCR was conducted on a Stratagene Mx3000P or Mx4000 QPCR system using SYBR Green Real-time PCR Master Mix (Applied Biosystems). Primer sequences are listed in Table S5. Primers were tested to be sure a single product was amplified with the expected melting temperature. A primer pair for an abundant noncycling gene, Rp49, was used in all samples to serve as an internal control for the amount of starting material. The relative abundance of a gene of interest was calculated based on the difference between the Ct value of the specific primer pair and that of the Rp49 primer pair.
|
10.1371/journal.ppat.1007304 | Crystal structure of a soluble fragment of poliovirus 2CATPase | Poliovirus (PV) 2CATPase is the most studied 2C protein in the Picornaviridae family. It is involved in RNA replication, encapsidation and uncoating and many inhibitors have been found that target PV 2CATPase. Despite numerous investigations to characterize its functions, a high-resolution structure of PV 2C has not yet been determined. We report here the crystal structure of a soluble fragment of PV 2CATPase to 2.55Å, containing an ATPase domain, a zinc finger and a C-terminal helical domain but missing the N-terminal domain. The ATPase domain shares the common structural features with EV71 2C and other Superfamily 3 helicases. The C-terminal cysteine-rich motif folds into a CCCC type zinc finger in which four cysteine ligands and several auxiliary residues assist in zinc binding. By comparing with the known zinc finger fold groups, we found the zinc finger of 2C proteins belong to a new fold group, which we denote the “Enterovirus 2C-like” group. The C-terminus of PV 2CATPase forms an amphipathic helix that occupies a hydrophobic pocket located on an adjacent PV 2CATPase in the crystal lattice. The C-terminus mediated PV 2C-2C interaction promotes self-oligomerization, most likely hexamerization, which is fundamental to the ATPase activity of 2C. The zinc finger is the most structurally diverse feature in 2C proteins. Available structural and virological data suggest that the zinc finger of 2C might confer the specificity of interaction with other proteins. We built a hexameric ring model of PV 2CATPase and visualized the previously identified functional motifs and drug-resistant sites, thus providing a structure framework for antiviral drug development.
| Since the launch of the Global Polio Eradication Initiative, the number of poliomyelitis cases has significantly reduced but obstacles to disease eradication remain. In the endgame phase, anti-poliovirus drugs will be critical in controlling transmission of vaccine-derived polioviruses and in treating patients with chronic infection. However, no effective anti-poliovirus drugs are yet available. The 2CATPase encoded by poliovirus is one of the most important drug targets, and many inhibitors of 2C have been found. We report here the crystal structure of a soluble portion of PV 2CATPase, containing an ATPase domain, a zinc finger and a C-terminal helical domain. Our findings not only revealed common and individual structural features in the picornaviral 2C protein family, but also allowed us to visualize a large collection of functional motifs and drug-resistant sites identified from the decades-long investigations of poliovirus 2CATPase. Our findings are invaluable for understanding the function of picornavirus 2C proteins and the development of antiviral drugs.
| Poliovirus (PV) is the pathogen of poliomyelitis. PV infection can directly result in damage of motor neurons and cause neural lesions [1]. Since the launch of Global Polio Eradication Initiative (http://polioeradication.org/) by the World Health Assembly, the number of poliomyelitis cases have been significantly reduced. The incidence of paralytic polio in 1988 was 1,000 children per day, and this number decreased to 400 per day in 2013[2]. Only three endemic countries remain today. Nevertheless, obstacles to global polio eradication remain. To overcome the last hurdles in the endgame phase, effective anti-PV drugs are critical in controlling transmission of vaccine-derived polioviruses (VDPVs) and in treating patients with chronic infection or personnel casually exposed to PV [3]. Further, to minimize poliomyelitis risk in the “post-polio” era, the National Research Council of the Unite States concluded that the development of antiviral drugs would be important, and possibly essential [4].
PV belongs to Enterovirus C species in the Enterovirus genus, Picornaviridae family [5]. It has an icosahedral non-enveloped capsid containing a single-stranded +RNA genome of ~ 7,500 nucleotides. PV has three serotypes, PV1 PV2 and PV3, whose capsids exhibit distinct antigenicity [6], but the replicative enzymes of PV are highly conserved not only among different serotypes, but also in other virus species. Therefore, the virally encoded replicative enzymes are considered as good targets for broad-spectrum antiviral drugs [7].
Among EVs replicative enzymes, 2C protein is arguably the most attractive target for direct-acting antivirals (DAA) development [8]. At least ten inhibitors have been found targeting 2C, including guanidine hydrochloride (GuHCl) [9,10], HBB [11], TBZE-029 [12], MRL-1237 [13,14], pirlindole, dibucaine, zuclopenthixol [15], hydantoin [16], fluoxetine [17,18] and brefeldin A [19]. Genotype analyses of drug-resistant virus clones had revealed a large set of residues of 2C that are likely to be involved in binding. Structural characterization of 2C proteins is therefore essential to visualize these potential drug binding sites and to assist antiviral drug development.
2C protein of enteroviruses has typically ~330 residues. It is composed of a N-terminal membrane binding domain, a central ATPase domain, a cysteine-rich domain and a C-terminal helical domain [20]. We recently determined the crystal structure of a soluble portion of 2C helicase from human enterovirus 71 (EV71) [21], the first high-resolution 2C structure in Picornaviridae. The ATPase domain of EV71 2C exhibits the structural characteristics of superfamily 3 (SF3) helicases. EV71 2C has an unusual zinc finger with three cysteine ligands. The C-terminus of EV71 2C forms an amphipathic helix that mediates self-oligomerization through a specific interaction between 2C-2C. The hexameric ring model of EV71 2C revealed that the central pore is negatively charged and a putative RNA binding motif is located on the rim of the ring, which suggests the mode of RNA binding might be the winding around the hexameric ring.
Poliovirus (PV) 2CATPase has been intensively studied for decades. Numerous genetic and biochemical studies have shown that PV 2CATPase is implicated in series of events in virus life-cycle, from uncoating [22], cellular membrane rearrangement and assembly of membranous replication complexes [23–27], viral RNA synthesis [28–30] to morphogenesis [31–34]. PV 2CATPase has only 329 residues but is very rich in features. It can be divided into an N-terminal domain, a central ATPase domain and a C-terminal domain. The N-terminal domain harbors a membrane binding motif [35], an amphipathic helix [30], an oligomerization motif [36] and an RNA binding motif [37]. The extreme N-terminal sequence of 2C is able to interact both with virally encoded proteins, 2B, 2BC, 3A, 3AB, 3C, VP3 [20,32,38,39] and cellular protein RTN3 [40]. Structural investigation of PV 2CATPase has been challenging because the N-terminal domain confers protein insolubility. Our structural characterization of EV71 2C was possible only when the N-terminal domain was removed. We characterized the roles of Walker A + B motifs, Motif C and arginine finger (R-finger) of EV71 2C in ATP hydrolysis and virus production. All of these conserved ATPase motifs are present in PV 2CATPase, and should exert similar functions [41,42]. It was predicted that the C-terminal cysteine rich motif of PV 2CATPase forms a CCCC type zinc finger [43], but the second potential zinc coordination sites (PCS) of the putative zinc finger is not present in many other enterovirus 2C proteins including EV71 2C. The crystallographic data showed that EV71 2C indeed lacks the PCS2 and it possesses a rare zinc finger with only three cysteine ligands, but it does not seem to affect its zinc coordinating capability [21]. Whether PV 2CATPase has a distinct zinc finger requires structural investigation. PV 2CATPase shares 63.7% amino acid sequence identity with EV71 2C helicase, however, no conclusive evidence has shown that PV 2CATPase has helicase activity to date. To support RNA synthesis, Cho et al found that the 3D polymerase of poliovirus has the activity of unwinding long stretch of RNA duplex, and the nascent RNA strand can be displaced from the template during chain elongation reaction[44]. Thus, the unwindase activity of PV 3D satisfies the necessity of using a helicase during PV replication.
Although many inhibitors have been found targeting PV 2CATPase over decades, the high-resolution structure of PV 2CATPase was still missing to date, which hindered the understanding of the mechanisms of inhibitor’s efficacy and drug-resistance and the development of anti-PV drugs. GuHCl for example inhibits ATPase activity of PV 2CATPase in vitro and in vivo [42]. Mutations that confer resistance to GuHCl were mostly mapped in segments outside the catalytically important ATPase motifs [9,10], whose precise functions were unknown.
We report here the crystal structure of a soluble fragment of PV 2CATPase containing the complete ATPase and the C-terminal domains to a resolution limit of 2.55Å. We found that the C-terminal helix mediates self-oligomerization of PV 2CATPase via a specific interaction between 2C-2C, which is essential to ATPase activity. Comparing the structures of PV 2C and EV71 2C revealed both common and distinct structural characteristics. We built a hexameric ring model of PV 2CATPase to visualize regions important to PV 2CATPase function and drug resistance.
To gain atomic insight into poliovirus 2CATPase, we carried out a crystallographic study. We first expressed a N-terminal maltose-binding protein (MBP) tagged full-length 2CATPase from human poliovirus 1 (strain Mahoney, GenBank: KU866422.1); however, the MBP tagged protein did not yield crystals. Removal of MBP tag by proteinase cleavage only led to precipitation of PV 2CATPase. Therefore, we employed a similar strategy as used in expressing the soluble fragment of EV71 2C [21]. Firstly, we removed the N-terminal 115 residues of PV 2CATPase, (PV-2C-ΔN). This fragment lacks the N-terminal membrane binding domain but retains a complete ATPase domain, a zinc binding site and a C-terminal helical domain. PV-2C-ΔN was soluble, but still failed to crystallize. Considering that surface-entropy reduction may favor crystallization [45], we predicted a set of surface residues charges based on EV71 2C structure [21]. We then introduced alanine substitutions systematically to these residues with an aim of minimizing surface charging. We finally obtained crystals of PV-2C-ΔN bearing mutations E207A, K209A and R149A (PV-2C-ΔN-3Mut). None of these mutations is located within conserved ATPase motifs, zinc binding site or the C-terminal helical domain.
We grew PV-2C-ΔN-3Mut crystals in a hanging drop vapor diffusion system at 20°C. The crystals diffracted the X ray to a 2.55Å. It belonged to the space group of P21 and contained 6 copies of 2C per asymmetric unit (ASU). We solved the crystal structure of PV-2C-ΔN-3Mut by molecular replacement using EV71 2C structure (PDB code: 5GQ1, B chain) as the search model. Several regions of the initial atomic model were built manually, especially at the zinc binding site that accounted for the largest structural discrepancy. In the finally refined model, we were able to located most of residues in chains A, B, C, D and E. But chain H was largely disordered, only a fraction of residues was visible in the electron density. Chain A and B were the most intact 2C copies and they were associated with relatively low temperature factors, 67.82 Å2 and 72.44 Å2, respectively. By contrast, the temperature factors of the other chains C, D, E and H are much higher; the values are 108.38 Å2, 76.55 Å2, 105.38 Å2 and 170.56 Å2, respectively. Therefore, only chain A and B were used for further structural analysis. Because the average temperature factor of this structure is high, we generated a simulated-annealing 2Fo-Fc composite omit map to validate its quality. As illustrated in Fig 1, the simulated-annealing composite omit map has a good fit with the final model PV-2C-ΔN-3Mut, which supported that the model was correctly built. We performed a structural comparison of the chains within ASU (S1 Fig). The r.m.s.d. values among different chains ranges from 0.4Å to 1.3Å. The largest structural deviation is contributed by a small loop region 180-184aa of B chain between β3 and α2, which is apparently resulted from the crystal packing artifacts. The C-terminal helical domain of PV 2C exhibits only slightly different orientation, but this is not comparable with the dramatical conformational changes observed at the C-terminal helix in EV71 2C structure[21] (PDB entries: 5GQ1 and 5GRB). The statistics of data collection, structure refinement and structure validation are summarized in Table 1.
The overall fold of PV-2C-ΔN-3Mut is similar to EV71 2C. Structural comparison between PV-2C-ΔN-3Mut and EV71 2C (PDB code: 5GRB, chain C) gave a Dali Z-score of 30.9 with 197 aligned Cα atoms and the r.m.s.d. value was 1.3Å. PV-2C-ΔN-3Mut is comprised of an ATPase domain with canonical α/β Rossmann fold, a CCCC type zinc finger followed and a long helical C-terminus (Fig 2A). The ATPase domain contains a five-stranded parallel β-sheet (β1-β5) surrounded by three α-helices (α1-α3). We located three conserved ATPase motifs. Walker A motif is located on the loop connecting β1 and α1, forming the phosphate binding loop (P-loop). A phosphate group was found occupying the P-loop. The Walker B motif is located on the loop between β3 and α2. The SF3 helicase specific motif C is located on the loop between β4 and α3 (Fig 2A and 2E).
The cysteine-rich motif of PV 2CATPase (residue 267–289), located between α4 and β6, connects the ATPase domain and the C-terminal helical domain. The cysteine-rich motif folds into a CCCC type zinc finger (Fig 2C). The zinc ion is coordinated by four cysteine ligands, C269, C272, C281 and C286. Three of these cysteine residues are from the long loop between α4 and α5, and one is from the C-terminus of α5. The distances from each Sγ atom to the zinc varies from 2.2Å to 2.4Å, which are typical distances between sulfur and zinc as the median distance of Zn-S is 2.28Å [46]. In addition, two conserved residues C282 and K288 assist the zinc coordination. The carbonyl oxygen of C282 buttress the zinc ion from the bottom (distance, 3.3 Å), whereas the side chain Nζ atom of K288 covers the zinc on top (distance 4.2Å) (Fig 2B and 2C). There is an unusually long loop region between the second and third zinc ligands C272 and C281, on which N277 and F278 dock their side chains into a hydrophobic groove formed between α1 and α6 helices. This hydrophobic interaction between the zinc finger and the ATPase domain may serve to further stabilize the folding of the zinc finger (Fig 2C).
Pfister and colleagues predicted four potential zinc coordination sites (PCS1-4) in the cysteine-rich motif of PV 2CATPase [43]. We therefore adopted this nomenclature in our structural analysis of zinc finger structure for the convenience of comparing with the results reported previously. We found that while the conformation of C269 (PCS1), C281 (PCS3) and C286 (PCS4) are well conserved in both structures (Fig 2D), residue C272 acts as PCS2 for zinc binding in PV 2CATPase, but its structural counterpart E272 in EV71 2C is not a zinc ligand. Instead, E272 forms a salt bridge with K288 stabilizing the overall folding of the zinc finger of EV71 2C. The loop harboring PCS2 (residues 270–276) is among the least conserved regions in picornavirus 2C proteins (Fig 2E), which coincides with the pronounced structural difference of this region between PV 2CATPase and EV71 2C (Fig 2D).
We performed crystal packing analysis of PV-2C-ΔN-3Mut structure and found that all 2C copies in the crystal polymerize via a C-terminus helix mediated 2C-2C interaction (Fig 3A). The mode of PV 2C-2C interaction resembles EV71 2C-2C interaction we characterized previously [21]. Residues C323, M324, L327 and F328 of the C-terminus α6 helix of a PV-2C-ΔN-3Mut monomer formed a pocket-binding domain (denoted: PBD) that docks inside a hydrophobic pocket (denoted: pocket) between the zinc finger and the ATPase helicase domain of an adjacent PV-2C-ΔN-3Mut monomer (Fig 3B). The pocket is formed by 13 residues, most of which are hydrophobic. A salt bridge between E325 from the PBD and R144 from the pocket further stabilizes 2C-2C interaction. To validate PV-2C-ΔN-3Mut oligomerization observed in crystalline state, we investigated self-oligomerization of PV-2C-ΔN using size-exclusion chromatography (Fig 4). The molecular mass of PV-2C-ΔN was observed to be 94kDa, close to the theoretical molecular mass of tetramers 95.2kDa, suggesting PV-2C-ΔN exists as tetrameric in solution. Removing PBD of PV-2C-ΔN (PV 2C 116–319) reduced its molecular mass to 23kDa, nearly matching the theoretical molecular mass for monomers 23.8kDa. Therefore, the tetramerization of PV-2C-ΔN was dependent on the PBD mediated 2C-2C interaction. Our previous results demonstrated that mutation of either L327 or F238 could abolish self-oligomerization of EV71 2C and these two residues are highly conserved among 2C helicases [21] (Fig 2A). We then measured the molecular mass of PV-2C-ΔN bearing mutation L327A and F238A respectively. The molecular mass of these mutants reduced to monomeric size, 26kDa, indicating that both residues play the crucial role in PV 2CATPase self-oligomerization (Fig 4).
Our previous crystallographic study of EV71 2C identified a hinge region on C-terminus α6 helix, which could mediate rotation between 2C-2C, therefore resulting multiple 2C-2C conformations in crystal structure [21]. By contrast, we did not find the similar 2C-2C rotations in the crystal structure of PV-2C-ΔN-3Mut, the C-terminus α6 helix of PV 2CATPase was kept essentially straight in all 2C copies. We compared the PV 2C-2C conformation with multiple EV71 2C-2C conformations, which demonstrated that PV 2C-2C conformation resembles EV71 2C-2C conformation-2 [21], a catalytically nonproductive conformation. We measured the distance from the phosphate group occupying the P-loop of a PV-2C-ΔN-3Mut monomer to the side chain of R241 (arginine-finger) on an adjacent PV-2C-ΔN-3Mut. The distance was 21Å, suggesting this cannot be a catalytically active conformation. In order to explore the biologically relevant conformation of PV 2CATPase, we built a hexameric ring model. We first searched structural homologues of PV-2C-ΔN-3Mut using Dali server, and found that beside EV71 2C, the best hit was the structure of a SV40 Large T Antigen (PDB code: 2H1L). We generated a hexameric model of PV 2CATPase by superimposing six copies of PV-2C-ΔN-3Mut to each subunit within SV40 Large T Antigen hexamer. In the hexameric model of PV-2C-ΔN, the C-terminal PBD of a protomer is located very close to the hydrophobic pocket on the adjacent protomer. We therefore slightly bent the α6 helix so that the PBD could readily occupy the hydrophobic pocket (Fig 5A).
The hexameric model of PV-2C-ΔN has a ring-like shape with the diameter of 117Å and the height of 40Å (Fig 5B). Its size is similar to EV71 2C hexameric model[21]. Adams and colleagues have reported that particles of the MBP tagged PV 2CATPase were composed of 5–8 protomers, and they exhibited increasing size ranging from 150 to 200Å [36]. The larger size of PV 2CATPase homo-oligomers visualized in the electron microcopy was possibly contributed by the N-terminal domain of PV 2CATPase and the presence of the MBP tag. While the unliganded foot-and-mouth disease virus 2C protein lacking the N-terminal domain self-oligomerizes in a concentration-dependent manner, this truncation containing a Motif C mutation (N207A) specifically forms hexamers in the presence of ATP and RNA[47]. The negative stain electron microscopy study further revealed FMDV 2C-ATP-RNA is a hexameric ring with 6-fold symmetry. The hexameric ring model of PV-2C-ΔN shows that all conserved ATPase motifs line up at the gaps between protomers, constituting the active sites (Fig 5A). While Walker A, Walker B motifs and Motif C are located on one side of the active site, the R finger (R241) is located on the other side. The distance from the phosphate group at the P-loop to R241 side chain has now reduced to 5.1Å, suggesting the improved active site conformation. The central pore of the hexameric ring has a funnel-like shape. While the opening on cytoplasm side is wider (diameter = 27Å), the opening on membrane proximal side is narrower (diameter = 14Å). It was previously reported that PV 2CATPase harbors two discrete segments, residues 21–45 and 312–319, both involved in RNA binding [10]. Although the N-terminal RNA binding segment is missing in our structure, the C-terminal RNA binding segment is located on the rim of the hexameric ring, suggesting that the RNA might bind the rim of PV 2CATPase hexameric ring.
To validate our structural findings, we characterized the ATPase activity of the MBP tagged full-length PV 2CATPase and a selection of mutants. We first characterized the ATPase activity of the wild-type enzyme (Fig 6A). Our analysis showed that the ATPase activity of PV 2CATPase obeys the Hill equation. The nonlinear curve fitting using the Hill equation gave an R2 = 0.999. The Hill coefficient n is approximal 1.9, suggesting a positive cooperativity in ATP binding and hydrolysis. We calculated the enzyme kinetic parameters Vmax = 137.7±2.1 μM/min and Km = 641.7±25.1 μM. The turnover rate kcat of the enzyme is 27.5 min-1. This value was calculated by dividing Vmax with the concentration of the monomeric MBP-PV 2C. The turnover rate of the MBP-tagged PV 2C is lower than that measured for the bilayer nanodiscs bound PV 2C[48]. The difference in the ATPase activity is likely due to the different strategy and affinity tag used in protein preparation.
Next, we measured and compared the ATPase activity of the PV 2CATPase mutants (Fig 6B and 6C). When we kept the substrate ATP concentration constant at 500μM, the wild-type PV 2CATPase exhibited an ATPase activity of 8.9±0.8 μmol/μmol/min-1, whereas the catalytic inactive mutants K135A (Walker A), D177A (Walker B), N223A (Motif C) and R241A (R finger) all exhibited the background activities at least 20 folds less than the wild-type activity, similar to MOCK and MBP controls. In most of cases, mutations on either the PBD or the pocket led to severe losses or abrogation of ATPase activity. The loss of ATPase activity was possibly caused by the disruption of PBD-pocket interaction and in turn 2C self-oligomerization, therefore the active site between PV 2C-2C could not form. These results are consistent with size-exclusion chromatography experiment that the disruption of PBD-pocket interaction by mutation F328A or deletion of PBD abolished 2C tetramerization in solution (Fig 4). The ATPase activity of mutant C323A was measured as 12.1±1.2 μmol/μmol/min, similar to WT enzyme. This suggests that substituting of C323 by an alanine could not undermine the interaction between PBD-pocket. This hypothesis is supported by our size-exclusion chromatography results (Fig 4). PV-2C-ΔN bearing mutation C323A at the C-terminal PBD also eluted as tetramers, suggesting the homo-oligomerization remained unaffected (Fig 4).
Among alanine substitutions of four zinc coordinating cysteine, C272A retained more than 70% of WT ATPase activities, 6.5±0.4 μmol/μmol/min, whereas C269A, C281A and C286A lost the majority of their activities; the percentage of the activities remained were 22%, 13% and 10%, respectively. Pfister et al. showed that while eliminating PCS1 (C269) of PV 2CATPase caused the failure in recovering virus from plasmid, but eliminating PCS2 (substitute C272 with serine or glutamine) did not affected PV translation activity in vivo. Nevertheless, substitution of PCS2 could induce temperature sensitive phenotype, encapsidation defects and impairment of RNA replication at high temperature [43]. PCS2 is naturally missing in EV71 2C and many other enterovirus 2C proteins. To further investigate the significance of PCS2, we introduced a set of mutations to an infectious clone of EV71. Structural alignments of PV 2CATPase and EV71 2C (Fig 2D) revealed that residue E272 of EV71 2C is the structural counterpart of C272 of PV 2CATPase. We therefore substituted E272 with cysteine and histidine respectively, with an aim of adding PCS2 to the EV71 2C zinc finger. To cover all possibilities, we also substituted three other nearby residues S271, N273 and N274 with cysteine respectively, so that at least one of these mutants may have a PCS2 for the zinc finger. To our surprises, the infectious clone bearing E272C or E272H did not showed the improved EV71 infectivity, instead they caused >75% losses of activities. Mutations N273C and N274C were also detrimental to virus production. Only S271C retained >75% of WT activity, but there was no improvement in EV71 production (Fig 7A and 7B).
Comparing the available high-resolution structures of enterovirus 2C proteins (PV 2C, EV-C species and EV71 2C, EV-A species), we identified both common and individual structural features. (i) The zinc finger is the most structurally distinct site in PV 2CATPase and EV71 2C. While PC 2CATPase has a CCCC type zinc finger, EV71 2C has only three cysteine (lacking the PCS2) for zinc coordination. Eliminating the PCS2 of PV 2CATPase resulted in temperature-sensitive phenotypes and encapsidation defects [11,49]. We added the PCS2 to the zinc finger of EV71 2C in order to convert it to a CCCC type zinc finger similar to PV 2CATPase zinc finger, but it failed to improve EV71 infectivity. Zinc fingers are ubiquitous small motifs that function as binding module for nucleic acids or proteins, etc.[50]. It was demonstrated that the C-terminal cysteine-rich site of PV 2CATPase, the zinc finger in our structure, is required for morphogenesis[49]. Therefore, the significant difference in sequence and structure of the zinc finger we observed here may underlie the specificity of 2C protein and determine what process it may involve. The cysteine-rich motif is one of the least conserved regions in Picornaviridae 2C (Fig 2E). In fact, the 2C of the foot-and-mouth disease virus (FMDV) does not even have a cysteine-rich motif between ATPase and C-terminal helical domains. So, this region of FMDV 2C might fold into a structure completely different from the zinc finger but it may still function as protein binding module with the distinct specificity. (ii) The hexameric ring models of PV 2CATPase and EV71 2C show that the ATPase active site formed between 2C subunits has nearly identical geometry and the catalytic residues (Walker A & B, Motif C and R finger) identified by structural and biochemical characterizations are invariant. (iii) The C-terminus amphipathic helix mediated self-oligomerization is common in enterovirus 2C. Analogous to EV71 2C, PV 2CATPase undergoes self-oligomerization both in crystalline state and in solution via PBD-pocket interaction. Although residues constituting the PBD and the pocket are not strictly conserved in PV 2CATPase and EV71 2C, the “knob-into-hole” interaction between PBD-pocket is identical. The PBD residues of PV 2CATPase include C323, M324, L327 and F328, whereas the PBD of EV71 2C contains residues T323, I324, L327 and F328. All PBD residues are hydrophobic, in which L327 and F328 are invariant. The strict conservation of L327 and F328 highlights their essential role in 2C activity. We showed that mutations L327 and F328 abrogated the ATPase activity and homo-oligomerization of both PV 2CATPase and EV71 2C helicases (Figs 4 and 5), and these mutations could halt EV71 infection [21]. The dimension of the hydrophobic pocket of PV 2CATPase is 17Å long, 12 Å wide and 7Å deep, which is similar to the pocket size of EV71 2C. We calculated the solvent accessible surface area (SASA) of the pocket of PV 2CATPase as 918 Å2 containing 13 residues, the SASA of the pocket of EV71 2C is 846 Å2 containing 14 residues.
Eight zinc finger fold groups have been classified previously [50]. Comparison of the geometry of the zinc fingers of EV71 2C and PV 2CATPase with eight known groups of zinc finger demonstrates that it belongs to none of them (S3 Fig). Therefore, the 2C zinc finger represents a new fold group, which we denote the “Enterovirus 2C-like” fold group. This group is composed of a N-terminal long loop followed by a short helix and a short loop. Three zinc ligands are contributed from the long loop, one ligand is contributed from the short helix. The short helix has a “PhhC” consensus sequence (h represents hydrophobic residues) in enteroviruses. The PhhC sequence is followed by a “GKA” motif that is invariant among enteroviruses. The conserved lysine of GKA motif plays an auxiliary role in stabilizing the zinc binding. The second zinc ligand is nonessential in zinc coordination. It can be substituted by a solvent water molecule, such in case of EV71 2C [21].
Previous studies have identified a large set of residues important to 2C functions. Genotype analyses of drug resistant virus clones have suggested many residues that were targeted by drugs. We summarized these findings and analyzed the compatibility of the published phenotypes with our structural characterization (S1 and S2 Tables and Fig 8). Based on their locations on 2C structure, we divided these residues into four categories. (i) Residues from Walker A motif, Walker B motif, Motif C and R finger are essential for ATPase/helicase activities. They are gathered in the gaps between subunits in the hexameric ring model of 2C protein. These residues are in general not involved in drug resistance, probably due to high genetic barrier. Only one GuHCl-resistant mutation was mapped to Walker A (A133T) of echovirus-9 2C [11]. (ii) Residues buried deeply into the hydrophobic core of 2C are important to overall folding. These include L125, V218, I142 A143, M246 and I248, most of which are buried between the parallel β-strand plane and the surrounding helices in ATPase domain. These residues were found important to encapsidation[31], morphogenesis [49] or temperature-sensitive virus phenotype [51]. Mutations of these residues account for resistant to GuHCl, MRL-1237 and Hydantoin. (iii) Residues exposed to the molecular surface of 2C hexamer model may directly interact with protein binding partners, RNA or drugs. Most of these residues are mapped on the cytoplasmic side and the rim of the ring, however a few are on the membrane proximal side (Fig 8). Therefore, in vivo the membrane proximal side of the hexameric ring model is unlikely to be the accessible molecular surface of 2C. During infection, this side should be attached to N-terminal portion of 2C, whose structure is yet to be determined. We further divided the surface exposed residues into different regions (Fig 8C). The majority of published drug resistant mutations were mapped into regions around the pore of the hexameric ring. These regions are potentially targeted by GuHCl, MRL-1237, HBB, TBZE-029, Fluoxetine, Dibucaine, Zuclopenthixol and Pirlindole. (iv) Residues located on N-terminal membrane binding domain are still missing in the available 2C structures.
GuHCl is among the earliest identified compounds effective on PV and other EVs. We mapped all published GuHCl-resistant or dependent mutations [9,10] on the structure of PV-2C-ΔN, revealing two distinct sites, site-1 and site-2 (Fig 8A and 8B). Mutations I142V and A143G cluster on site-1, where they are deeply buried inside the hydrophobic core of ATPase domain. Substitute at here may affect the interaction with GuHCl or other ligands via an indirect way; the mechanism require further investigation. The other mutations, N179G on the loop between β3 and α2, M187L on the α2 helix and S225T, I227M and A233T/S on the loop between β4 and α3 (β4-α3-loop), cluster on site-2. The β4-α3-loop is highly flexible in PV 2CATPase, and a segment of this loop residues 227–235 were missing in the electron density map. When displayed these residues on the hexameric ring model of PV-2C-ΔN, we found that six copies of site-2 from each subunit form a belt region surrounding the central pore (Fig 8A and 8B), which may act as the direct binding site of GuHCl. Site-2 might act an essential interface for the binding with other viral or cellular proteins during replication. Hence, a plausible explanation of the inhibitory mechanism of GuHCl is blocking other proteins from binding to 2C by occupying site-2. Mutations responsible for resistance to other drugs were also mapped on the β4-α3-loop. Echovirus 9 bearing 2C mutations I227L and A229V is resistant to HBB. Coxsackievirus B3 containing 2C mutation A224V, I227V and A229V is resistant to TBZE-029, Fluoxetine and Dibucaine. Therefore, the β4-α3-loop of 2C is likely a “hot spot” for drug resistance.
The investigation of PV morphogenesis suggested that the cysteine-rich motif on the C-terminal domain of 2C helicase is involved in encapsidation, possibly via an interaction with a region between Walker A and B motifs of the ATPase domain [34,49]. Our crystal structure of PV 2CATPase revealed that the α1 helix (residues 135–150) located between the Walker A and B motifs (Fig 2E and 2C) directly interacts with residues N277 and F278 from the loop between PCS2 and PCS3 of the zinc finger. Wang et al showed that PV with K279A/R280A mutations showed defects in replication and encapsidation[49]. The same group later demonstrated that the PCSs of the cysteine-rich motif are also involved in encapsidation. According to the crystal structure of PV 2CATPase and the hexameric ring model, the cysteine rich motif folds into a zinc finger and it is located on six vertices of the ring (Fig 8C). K279 and R280 are located immediately downstream of N277 and F278, where they are fully accessible on the surface of the hexameric ring. Therefore, it is suggestive that the interaction between the zinc finger and the ATPase domains are important to stabilizing the folding of the zinc finger, whereas K279 and R280 may be involved in an interaction with other protein partners during encapsidation. Among four PCSs of the zinc finger, C269 (PCS1) and C281 (PCS3) are almost completely buried in the hydrophobic core; by contrast, C272 (PCS2) and C286 (PCS4) are relatively more exposed to the molecular surface. This supports the results reported by Wang and colleagues [49], that while alanine substitution of PCS1 or PCS3 was lethal, substitutions of PCS4 and PCS2 exhibited temperature-sensitive and quasi infectious phenotypes respectively. It suggests that PCS1 and PCS3 are absolutely required in maintaining the folding of the zinc finger and probably the entire hexamer, whereas PCS2 and PCS4 might have an additional role of interacting with other proteins during encapsidation.
Vero and RD cells (American Type Culture Collection) were grown in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) with 10% fetal bovine serum added. EV71 infectious clone was a gift from S. Cen (Institute of Medicinal Biotechnology, Chinese Academy of Medical Sciences and Peking Union Medical College) which contained the full-length cDNA of WT EV71 (strain Fuyang, GenBank: EU703814.1). The mouse anti-EV71 monoclonal antibody was purchased from Millipore. Donkey anti-mouse immunoglobulin G secondary antibody was purchased from LI-COR Biosciences.
Because full-length PV 2CATPaseis unstable for structural and biochemical studies, we prepared the maltose-binding protein (MBP) tagged 2CATPase as described previously [36]. In brief, the cDNA encoding 2CATPase of human poliovirus 1 strain Mahoney (GenBank ID: KU866422.1) was synthesized (GENEWIZ), amplified using PCR and inserted into pMAL-c5X vector (New England BioLabs) between EcoRV and SalI sites. The resulting plasmid expressed a N-terminal tagged full-length PV 2CATPase (MBP-PV 2C). The plasmid was transformed to competent E. coli cells BL21 (DE3). The bacteria were inoculated in a medium containing 100 mg/L ampicillin and grown to a density OD600 = 0.8 at 37°C. The cultures were cooled and induced with 0.3mM IPTG. The bacteria were then cultured at 18°C overnight before harvesting. The cells were pelleted by centrifugation (5,000g, 10min) and resuspended in a cold lysis buffer (50mM HEPES pH 7.5, 500mM NaCl, 1% Triton X-100, 1mM DTT and 1mM EDTA) on ice. The suspended cells were disrupted by ultrasonication on ice and clarified by centrifugation at 25,000g for 30min. The supernatant was loaded onto the amylose resin (New England BioLabs) and the resin was washed thoroughly by using a buffer containing 50mM HEPES pH 7.5, 500mM NaCl, 1% Triton X-100 and 1mM DTT. The MBP-PV 2C protein was eluted with 5 column volume of elution buffer (50mM HEPES, pH 7.5, 100mM NaCl, 1mM DTT and 20mM maltose). The final step of the purification was size exclusion chromatography using Superdex 200 HR10/300 GL column (GE Healthcare) pre-equilibrated with a buffer containing 20mM HEPES pH 7.5 and 100mM NaCl and 5mM DTT.
The MBP tagged full-length PV 2CATPase were used for biochemical characterization, however, this protein did not yield crystals. To crystallized PV 2CATPase, we expressed a fragment of 2C lacking the N-terminal membrane binding domain. The cDNA encoding residues 116–329 of PV 2CATPase was amplified PCR and inserted into pET28-SUMO expression plasmid [21](modified from pET28a) between BglII and XhoI sites. The resulting plasmid expressed a truncation of PV 2CATPase, denoted PV-2C-ΔN. The plasmid was transformed into E. coli competent cells BL21 (DE3) (Novagen). A single colony was picked and grew in 1L of LB media containing 50 mg/L kanamycin at 37°C to a density OD600 = 0.8. The culture was then cooled to 18°C and induced by 0.5mM IPTG (final concentration). The bacteria culture continued at 18°C overnight. The cells were harvested by centrifugation (5,000g, 10min) and resuspended in cold lysis buffer (50mM Tris-HCl pH 7.5 and 100mM NaCl) and disrupted by ultrasonication. The cell debris was removed by centrifugation at 25,000g for 30 min. The supernatant was loaded to Ni-NTA resin (Qiagen). The column was washed thoroughly with 20 column volumes of wash buffer containing 50mM Tris pH 7.5, 100mM NaCl, 20mM imidazole to eliminate nonspecifically bound proteins. Subsequently, Ulp1 peptidase was added to cleave SUMO tag at 4°C overnight. The flow through containing the non-tagged PV-2C-ΔN was collected and subjected to Superdex 200 HR10/300 GL column (GE Healthcare) pre-equilibrated with a buffer containing 20mM HEPES pH 7.5 and 100mM NaCl and 5mM DTT.
Mutations of the full-length MBP-PV 2CATPase and PV-2C-ΔN were introduced by site-directed mutagenesis. The mutants were expressed and purified using the same protocols described above.
PV-2C-ΔN-3Mut was concentrated to ~5 mg/ml before crystallization trials. The protein was crystallized by mixing 1μl of protein sample (5mg/ml) with 1.3μl of buffer containing 0.2M MgCl2, 0.1M MES pH 6.5, 3% (v/v) PEG4000, 9.2% (v/v) polypropylene glycol P 400, and 5mM TECP were freshly added to the buffer before use. The crystals were grown in a hanging-drop vapor diffusion system at 20°C. The crystals were flash frozen in Liquid nitrogen. Glycerol (v/v 25%) was used as the cryoprotectant. Complete datasets were collected at BL18U1 beamline of Shanghai Synchrotron Radiation Facility (SSRF). The crystal diffracted the X ray to 2.55Å. It belonged to a space group of P21, contained six copies of PV 2C-ΔN in the asymmetric unit (ASU). The structure was solved by molecular replacement using EV71 2C structure (PDB code: 5GQ1, B chain) as the searching model. The atomic model of PV-2C-ΔN-3Mut was completed by manual building using the software Coot v0.8.2 [52]. The structure was refined using the software PHENIX v1.10.1[53]. Most residues of PV-2C-ΔN-3Mut were located in chain A, B, C, D and E, however only a fraction of residues was visible in chain H. The loop between β4 and α3 was highly flexible in all chains. The composite omit map of PV-2C-ΔN-3Mut was calculated using the software phenix.composite_omit_map from PHENIX v1.10.1[53]. The software was run in a simulated annealing mode to aggressively remove phase bias. The annealing method used was cartesian and the starting temperature was 5,000 K. The statistics of data collection, structure refinement and structure validation were summarized in Table 1.
ATPase assays were performed as previously described [21,36,54]. The concentration of the enzyme was kept constant at 5μM in all reactions. The volume of the reaction mixtures was 50μl, contained 20mM HEPES pH 7.5, 4mM magnesium acetate, 5mM DTT, 500μM ATP and trace amount of [γ-32P] labelled ATP. The mixtures were incubated at 30°C and the reactions were initiated by adding enzyme. At the given time point, 10μL of the reaction mixture was removed and mixed with EDTA (final concentration = 0.1M) to quench the reaction. At least three time points were recorded for each reaction. The mixtures were resolved by thin-layer chromatography using PEI (Polyethylenimine) Cellulose Plates (Sigma-Aldrich) with a buffer containing 0.8M acetic acid and 0.8M Lithium chloride [21]. The PEI plates were visualized and quantified using Typhoon TrioVariable Mode Imager (GE Healthcare).
Vero and RD cells were grown with 5% CO2 in DMEM, supplemented with 10% Fetal Bovine serum. EV71 infectious clone contained the full-length cDNA of WT virus. The mutations were introduced by the site-directed mutagenesis. The plasmids were linearized by HindIII digestion and were in vitro transcribed to RNAs using the MEGA script T7 Kit (Ambion). Subsequently, the RNAs were transfected in Vero cells with Lipofectamine 2000 (Invitrogen) according to manufacturer’s instructions. The Vero cells were cultured at 37°C. 72 hours post-transfection, the supernatants of cell culture were collected to infect RD cells seeded in a 24-well plate at the given temperatures. The RD cells were fixed 24 hours post-infection. Immunofluorescence assays were performed to probe the production of EV71 virus using mouse anti-EV71 monoclonal antibody and Donkey anti-mouse immunoglobulin G secondary antibody. At least eight different visual fields from each well were photographed randomly under the microscope. The positive dots were counted using the software ImageJ v1.2.0 [55].
The Superdex200 10/300 GL column (GE Healthcare) was equilibrated with a buffer containing 20mM HEPES pH 7.5 and 100mM NaCl and 5mM DTT. The column was calibrated with Gel Filtration Standard (BIO-RAD) containing thyroglobulin (670kDa), γ-globulin (158kDa), ovalbumin (45kDa), myoglobin (17kDa), and vitamin B12 (1.35kDa). The purified proteins were loaded and eluted with a flow rate of 0.5ml/min. The Log of the molecular mass (kDa) of the standards was plotted as the function of the elution volume. Four standard proteins: thyroglobulin, γ-globulin, ovalbumin and myoglobin were used for a linear-fitting to generate a standard curve. The molecular mass of PV-2C-ΔN variants was then calculated using the standard curve.
|
10.1371/journal.pntd.0004012 | Experimental Transmission of Karshi (Mammalian Tick-Borne Flavivirus Group) Virus by Ornithodoros Ticks >2,900 Days after Initial Virus Exposure Supports the Role of Soft Ticks as a Long-Term Maintenance Mechanism for Certain Flaviviruses | Members of the mammalian tick-borne flavivirus group, including tick-borne encephalitis virus, are responsible for at least 10,000 clinical cases of tick-borne encephalitis each year. To attempt to explain the long-term maintenance of members of this group, we followed Ornithodoros parkeri, O. sonrai, and O. tartakovskyi for >2,900 days after they had been exposed to Karshi virus, a member of the mammalian tick-borne flavivirus group.
Ticks were exposed to Karshi virus either by allowing them to feed on viremic suckling mice or by intracoelomic inoculation. The ticks were then allowed to feed individually on suckling mice after various periods of extrinsic incubation to determine their ability to transmit virus by bite and to determine how long the ticks would remain infectious. The ticks remained efficient vectors of Karshi virus, even when tested >2,900 d after their initial exposure to virus, including those ticks exposed to Karshi virus either orally or by inoculation.
Ornithodoros spp. ticks were able to transmit Karshi virus for >2,900 days (nearly 8 years) after a single exposure to a viremic mouse. Therefore, these ticks may serve as a long-term maintenance mechanism for Karshi virus and potentially other members of the mammalian tick-borne flavivirus group.
| Members of the mammalian tick-borne flavivirus group, including tick-borne encephalitis virus, remain a significant cause of human disease and are responsible for at least 10,000 clinical cases of tick-borne encephalitis each year. One of the principal questions in their epidemiology is how they persist from year to year in a given area. To attempt to explain the long-term maintenance of members of this group, we exposed Ornithodoros parkeri, O. sonrai, and O. tartakovskyi ticks to Karshi virus, a member of the mammalian tick-borne flavivirus group. Ticks were exposed to Karshi virus either by allowing them to feed on viremic suckling mice or by intracoelomic inoculation. To determine their ability to maintain the virus for an extended period of time and to transmit Karshi virus, ticks were allowed to feed individually on suckling mice after various periods of extrinsic incubation. Ticks exposed to Karshi virus, either orally or by inoculation, remained efficient vectors of Karshi virus, even when tested >2,900 days (approximately 8 years) after their initial exposure to virus. Therefore, these ticks may serve as a long-term maintenance mechanism for Karshi virus and potentially other members of the mammalian tick-borne flavivirus group.
| Karshi virus is a member of the mammalian tick-borne flavivirus group (genus Flavivirus, family Flaviviridae) [1]. Members of this group include tick-borne encephalitis virus (including subtypes Central European encephalitis virus (CEEV) and Russian spring-summer encephalitis virus (RSSEV), Omsk hemorrhagic fever virus, Langat virus (LGTV), Alkhurma hemorrhagic fever virus, Kyasanur Forest disease virus (KFDV), Powassan virus (POWV), Royal Farm virus, Karshi virus, Gadgets Gully virus, and Louping ill virus [1,2]. This group of viruses, also known as the TBEV serocomplex [1,3], are responsible for at least 10,000 clinical cases of tick-borne encephalitis each year [4]. A second group of tick-borne flaviviruses is known as the seabird tick-borne flavivirus group [5]. Although a member of the mammalian tick-borne flavivirus group, Karshi virus is not known to cause disease in humans [5]. However, its close relationship to both POWV and KFDV indicated that it should be capable of causing disease in humans [1,2].
The natural transmission cycle of the mammalian tick-borne flavivirus group involves ixodid ticks and rodents, with Ixodes ricinus and I. persulcatus being the principal vectors of CEEV and RSSEV viruses, respectively [6,7]. This cycle is essentially identical to that for the Lyme disease spirochete, Borrelia burgdorferi, in I. scapularis. In the Lyme disease cycle, the mouse, Peromyscus leucopus, remains infectious for several months [8]. Therefore, once a mouse becomes infected by being fed upon by an infectious nymphal tick, it would continue to expose larval and nymphal ticks to the spirochete for months. However, because viremias in rodents exposed to members of the mammalian tick-borne flavivirus group are transient, often lasting only a few days [9,10], the timing of nymphal and larval attachment becomes critical. If infectious nymphal ticks attach too early in the season, the viremia in the rodent will have ended prior to the attachment of the larval ticks. Unlike these Ixodid (hard) ticks that normally attach for 2–13 days to complete a blood meal and only feed once during the larval, nymphal, and adult stages [11], members of the genus Ornithodoros attach and complete feeding usually within 10–30 min and most complete feeding within an hour [12]. Also, these ticks will feed multiple times both as nymphs and as adults, often live in rodent borrows, and can live about 20 years [12,13]. Previous studies indicate that Ornithodoros spp. ticks are able to become infected and transmit members of the mammalian tick-borne flavivirus group [14–16] as well as other pathogens [12]. Because of their long life span and repeated feedings, they can remain infectious for an extended period of time. Ornithodoros tholozani were shown to transmit Borrelia persica (a causative agent of relapsing fever) for at least 13 years after a single exposure [13] and field-collected O. turicata were able to transmit B. recurrentis (repsorted as Spirochaeta recurrentis) for at least 6.5 years [17]. In addition, many Ornithodoros spp. ticks are considered to be nidicolous, i.e., living in close association with their vertebrate hosts such as living in rodent burrows [18]. Onithodoros sonrai are found in burrows of many rodent genera in Senegal and western Africa [19]; O. tartakovskyi, which is widely distributed in central Asia from Iran to the Xinjiang Province in western China are found in burrows of various rodent species, but primarily the great gerbil, Rhombomys opimus, [20,21]; and O. parkeri found in the western portions of the United States and Canada, is associated with numerous rodent species, but primarily prairie dogs [22,23]. To determine the potential for these ticks to serve as a long-term maintenance mechanism for these viruses, we evaluated the potential for O. sonrai, O. parkeri, and O. tartakovskyi ticks to transmit Karshi virus over an extended period of time.
We used three species of Ornithodoros ticks. These included a laboratory colony of O. sonrai derived from wild-caught specimens excavated from mammal burrows in the Bandia Forest of Senegal in 1989 [15]. No virus was detected upon examination of parental ticks from this colony. Georgia Southern University provided a colony of O. parkeri derived from specimens captured in Spicer City, CA, in 1965. The National Institute of Allergy and Infectious Diseases provided a laboratory colony of O. tartakovskyi. All three colonies were maintained as described by Durden et al. [24].
We used the U2-2247 strain of Karshi virus. It had been passaged once in Vero cells and once in suckling mice before use in these experiments. Serial dilutions of blood, brain, and tick samples were tested for virus by plaque assay on confluent monolayers of 2- to 3-d-old primary chicken embryo cells or by subcutaneous inoculation into 2- to 4-d-old suckling mice. The identity of the original virus, and virus recovered from ticks and mice, was confirmed by a Karshi-specific quantitative real-time Real Time- polymerase chain reaction (PCR) assay and by direct sequencing of the PCR products [16,25].
One-day-old suckling mice (BALB/c strain) were inoculated intraperitoneally with 106.3 suckling mouse lethal dose50 (SMLD50) units of Karshi virus. Two or 3 days after inoculation, a Karshi virus-inoculated mouse was placed in a cage containing ~50 O. sonrai, O. parkeri, or O. tartakovskyi ticks at various stages of development (larvae through adult, but predominately early nymphs). After the ticks had been allowed to attach to the mouse for about 5 min, the mouse was removed and a second virus-inoculated mouse was added to the cage. This was repeated for up to three mice for each species of tick used in this study. The ticks were allowed to feed on the virus-inoculated mouse for about 2 h. At that time, those ticks that had attached and did not feed were removed and discarded. Each mouse was then euthanized with CO2 and blood was collected by cardiac puncture. Blood was mixed 1:10 in diluent (Medium 199 with Earle’s salts containing 10% heat-inactivated fetal bovine serum and 5 μg of amphotericin B, 50 μg of gentamicin, 100 units of penicillin, and 100 μg of streptomycin per ml and 0.075% NaHCO3) and frozen at -70°C until tested to determine the viremia at the time of tick feeding. The engorged ticks were placed in a cage maintained at room temperature (~20°C) until tested for either infection or for the ability to transmit virus by bite. For each species, some of the ticks that had not attached to a virus-inoculated mouse were inoculated intracoelomically with 104 SMLD50 (107.5 SMLD50/ml) of the same virus strain that had been used to infect the mice [26]. These inoculated ticks were treated in the same manner as the engorged ticks, except that the inoculated O. parkeri were maintained in an incubator maintained at 26°C rather than at ambient air temperature.
To determine transmission rates, virus-exposed ticks were allowed to feed for up to 2 hours on naive suckling mice (either BALBc or Swiss Webster) individually, i.e., one tick per mouse. These suckling mice were marked by subcutaneous inoculation of India ink, returned to their dam, and then monitored daily over the next 21 d for signs of viral infection. Each litter contained one or two suckling mice that were either unexposed to ticks or were fed upon by a tick from the uninfected colony to serve as negative controls. Moribund mice were euthanized with CO2, and brain samples were obtained from a subset of them and then triturated (1:10) in diluent and frozen at -70°C until tested for virus. In most of the tick transmission trials, ticks were caged individually in plastic vials (12 ml, about half filled with washed sea sand) after feeding on the mice. Many of these same ticks were tested multiple times over the following 8 years for their ability to transmit virus by bite.
Research was conducted under an IACUC approved protocol in compliance with the Animal Welfare Act, PHS Policy, and other Federal statutes and regulations relating to animals and experiments involving animals. The facility where this research was conducted is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, International and adheres to principles stated in the Guide for the Care and Use of Laboratory Animals, National Research Council, 2011. The USAMRIID IACUC approved these studies.
Viremias in the suckling mice at the time of the tick feedings ranged from 106.5 to 106.7 SMLD50/ml. When allowed to feed on a susceptible mouse ≤94 days after the initial blood meal, transmission was very inefficient, with none of 17 ticks transmitting virus to the mice (Table 1). However, when tested ≥105 days after the initial feeding, at least 60% of the ticks that had fed on a mouse with a viremia about 106.5 SMLD50/ml transmitted virus, regardless of tick species, including several ticks that failed to transmit virus when allowed to feed at days 59–94 after virus exposure. When ticks that had transmitted virus on one occasion were allowed to feed on a second mouse at some point in the future, nearly all of them (86%, n = 14) transmitted each time they were allowed to feed. Each of the species transmitted virus the last time it was tested, and all species transmitted virus for at least 2,000 days (Table 1). Data for each transmission attempt is provided in S1 Table.
For both O. sonrai and O. tartakovskyi, five of six ticks transmitted virus by bite when tested 43 days after inoculation with Karshi virus (Table 2). However, all 34 ticks (eight O. parkeri, 11 O. sonrai, and 15 O. tartakovskyi) tested at ≥64 days after inoculation transmitted Karshi virus by bite. Data for each transmission attempt is provided in S2 Table.
These 34 ticks took a total of 43 blood meals from susceptible mice and transmitted virus in each case (Table 2). Individuals in each species transmitted virus the last time that species was tested, with the final transmission occurring >2,100 days after the tick had initially been inoculated with Karshi virus.
Ornithodoros spp. ticks were able to transmit Karshi virus for >2,900 days (nearly 8 years) after a single exposure to a viremic mouse. Therefore, these ticks may serve as a long-term maintenance mechanism for Karshi virus and potentially other members of the mammalian tick-borne flavivirus group. This study was a continuation of a study [16] that examined the potential for these Ornithodoros ticks to transmit Karshi virus, but that original study only followed the tick for 3 years.
Traditionally, viruses in the mammalian tick-borne flavivirus group have been associated with ixodid ticks, I. ricinis and I. persulcatus in Europe and Asia, respectively [6,7] and with I. cookei and I. scapularis in the Americas [27]. Larval and nymphal ticks are exposed to virus when overwintering infected nymphal ticks feed on naïve rodents in the spring. These ticks can also be infected by co-feeding with an infected tick [28], regardless of the immune status of the rodent [29]. However, transmission by co-feeding on an immune rodent was only about 10% as efficient as co-feeding on an immunologically naïve rodent when the two ticks were not immediately collocated [29]. Given the relatively short period of viremia for these viruses in their rodent hosts [9,10], one could hypothesize that this cycle would be too inefficient to maintain these viruses for many years in the same location. However, if a rodent became infected after being fed upon by an infectious tick and then went back to its burrow, it could potentially expose many of the Ornithodoros ticks living in that burrow. When that rodent died, or was killed by a predator, the burrow would remain vacant until discovered by a new rodent. Individual Ornithodoros ticks can remain viable for up to 4 years between feedings [13,30,31] and can survive for 10–20 years [13,32–34]. In addition, this study observed transmission of Karshi virus for up to 8 years post infection. Thus, ticks present in the vacant rodent burrow could remain a source of virus for many years. When a new rodent entered that burrow and was fed upon by the infected Ornithodoros ticks, the rodent would become infected and all the ixodid ticks present on that rodent exposed to virus. These ixodid ticks could then spread the virus to other rodents and to larger mammals including humans.
Ornithodoros ticks have a wide distribution, with species found in much of the range of the mammalian tick-borne flaviviruses [35,36]. However, there are regions where members of this virus complex are found, but for which members of the genus Ornithodoros have not been described, i.e., the northeastern US for Powasson virus and deer tick virus, and parts of the northern range of the mammalian tick-borne flaviviruses in Eurasia. Therefore, other methods must exist for the perpetuation of these viruses in those areas.
Experimental studies on members of the mammalian tick-borne flavivirus group have focused on ixodid ticks. However, several members of this and the closely related seabird tick-borne flaviviruses group have been isolated from naturally occurring Ornithodoros ticks. These include Karshi virus [37], KFDV [38], Alkhurma hemorrhagic fever virus [39], Meaban virus [40], and Saumarez Reef virus [41].
Therefore, the susceptibility of O. parkeri, O. sonrai, and O. tartakovskyi to infection with Karshi virus; their ability to transmit this virus for extended periods (at least 2,905 days); their long life span; and the isolation of several members of both the mammalian and seabird tick-borne flavivirus groups from Ornithodoros ticks indicate that Ornithodoros species should be studied as potential long-term reservoir hosts for members of the tick-borne flavivirus groups.
|
10.1371/journal.pbio.1000059 | Oncogenic Kras Initiates Leukemia in Hematopoietic Stem Cells | How oncogenes modulate the self-renewal properties of cancer-initiating cells is incompletely understood. Activating KRAS and NRAS mutations are among the most common oncogenic lesions detected in human cancer, and occur in myeloproliferative disorders (MPDs) and leukemias. We investigated the effects of expressing oncogenic KrasG12D from its endogenous locus on the proliferation and tumor-initiating properties of murine hematopoietic stem and progenitor cells. MPD could be initiated by KrasG12D expression in a highly restricted population enriched for hematopoietic stem cells (HSCs), but not in common myeloid progenitors. KrasG12D HSCs demonstrated a marked in vivo competitive advantage over wild-type cells. KrasG12D expression also increased the fraction of proliferating HSCs and reduced the overall size of this compartment. Transplanted KrasG12D HSCs efficiently initiated acute T-lineage leukemia/lymphoma, which was associated with secondary Notch1 mutations in thymocytes. We conclude that MPD-initiating activity is restricted to the HSC compartment in KrasG12D mice, and that distinct self-renewing populations with cooperating mutations emerge during cancer progression.
| Ras proteins act as molecular switches that relay growth signals from outside the cell. This mechanism is often subverted in cancer, and Ras proteins are activated directly by RAS gene mutations in approximately one-third of human malignancies. We have modeled this in mice engineered to have a Ras mutation. These mice develop a disease similar to chronic leukemias in humans called myeloproliferative disorders. It is marked by a fatal accumulation of mature and immature cells in the blood and bone marrow. We investigated whether some or all of these neoplastic cells were immortal. In agreement with the “cancer stem cell” hypothesis, we found that immortal cells were extremely rare in the bone marrow of diseased mice. They were found only in the same cell populations that contain normal bone marrow stem cells. However, these cells had high rates of replication and produced large numbers of daughter cells. Furthermore, many mice went on to develop acute lymphoid leukemia after acquiring additional mutations in maturing lymphoid cells. These studies exemplify the evolution of malignant stem cells during cancer progression. They also highlight the importance of rare, long-lived cells in the genesis and, potentially, therapy of high-risk chronic leukemias caused by abnormal Ras proteins.
| Self-renewal is integral to the malignant phenotype [1]. In principle, the ability of cancer cells to self-renew may be intrinsic to the compartment in which the tumor-initiating mutation occurs, or may be acquired as a consequence of mutations in more differentiated cells. The hematopoietic system has proven highly informative for addressing how cancer-associated mutations and cell of origin interact to establish malignant self-renewing populations. Accumulating evidence supports the idea that many hematopoietic malignancies exist in a hierarchy of differentiation with only a minor population capable of propagating and maintaining the disease in vivo [2]. These cells are termed leukemia-initiating cells or leukemia stem cells (LSCs), and manifest some biologic properties of normal hematopoietic stem cells (HSCs). However, the precise relationship between these populations is uncertain and appears to depend, in part, on both the leukemia subtype and on the effects of specific mutations. For example, overexpressing MLL fusion proteins found in human acute myeloid leukemia transforms both murine HSCs and more differentiated progenitors [3,4]. By contrast, inactivation of the JunB transcription factor must occur in the HSC compartment for initiation of myeloid malignancies [5]. These proof-of-concept experiments underscore the importance of understanding how oncogenes and tumor suppressors that are commonly mutated in human cancers perturb self-renewal and growth control. Importantly, the functional characteristics of LSCs that distinguish them from HSCs and how these properties are modulated by oncogenes are poorly understood.
RAS gene mutations are highly prevalent in pancreatic (>80%), colorectal (40%–50%), endometrial (40%), lung (30%), and cervical cancers (20%–30%), as well as in myeloid malignancies (20%–40%) [6]. Of the genes in the canonical RAS family, KRAS accounts for ∼90% of cancer-associated mutations, whereas HRAS mutations are rare. In hematologic cancers, NRAS is mutated 2–3 times more often than KRAS [6]. Cancer-associated RAS mutations, which introduce amino acid substitutions at codons 12, 13, or 61, result in oncogenic Ras proteins that accumulate in the active, GTP-bound conformation because of defective guanine nucleotide hydrolysis [7]. Elevated levels of GTP-bound Ras, in turn, deregulate signaling in cancer cells by altering the activation of effector cascades that include the Raf/MEK/ERK, phosphatidylinositol 3-kinase (PI3K)/Akt, and Ral-GDS pathways [8].
Chronic and juvenile myelomonocytic leukemias (CMML and JMML) are aggressive myeloid malignancies that are classified as myeloproliferative disorders (MPDs) [9]. Both diseases are characterized by leukocytosis with excess monocytes in blood and bone marrow, and by significant infiltration of malignant myeloid cells into the liver, spleen, and other organs. Hyperactive Ras is strongly implicated in the pathogenesis of JMML and CMML. Somatic NRAS and KRAS mutations are found in ∼40% of CMML specimens [10,11], and ∼85% of JMML patients have mutations in KRAS, NRAS, NF1, or PTPN11 (reviewed in [12]). The latter two genes encode proteins that regulate Ras-GTP levels. Importantly, children with germline mutations in NF1 or PTPN11 are at markedly increased risk of developing JMML, which argues strongly that deregulated Ras signaling can initiate this MPD [12]. This hypothesis is further supported by studies of Nf1, Kras, and Ptpn11 mutant mice, all of which develop MPDs that resemble JMML and CMML [13–16]. Mutations that alter other signaling molecules also cause MPD. For example, the BCR-ABL fusion gene is the hallmark of chronic myeloid leukemia, and JAK2 mutations are found in nearly all cases of polycythemia vera (reviewed in [17]). While mutations affecting signaling molecules are found in nearly all types of MPD, additional cooperating mutations are rare. MPDs are therefore genetically straightforward and tractable malignancies for understanding how aberrant signal transduction contributes to cancer by perturbation of stem and progenitor cell fates.
Mice expressing oncogenic KrasG12D in hematopoietic cells develop a fatal MPD with 100% penetrance that is characterized by leukocytosis, splenomegaly, and anemia, with death at approximately 3 mo of age [13,14]. In this system, injecting Mx1-Cre, KrasLSL-G12D mice with polyinosinic-polycytidylic acid (pIpC) induces expression of Cre recombinase, which removes an inhibitory loxP-STOP-loxP (LSL) element and activates the KrasG12D allele (we hereafter refer to Mx1-Cre, KrasLSL-G12D mice that have been treated with pIpC as KrasG12D mice). Mx1-Cre, KrasLSL-G12D mice provide a robust experimental system for investigating how expressing oncogenic Kras from the endogenous promoter affects HSCs and their progeny in vivo. Here, we identify tumor initiating cells in the KrasG12D model of MPD and find that, unlike some myeloid oncogenes, KrasG12D does not confer aberrant self-renewal properties to committed progenitor cells. However, small numbers of primitive KrasG12D cells can initiate MPD. This population shows excessive proliferation and rapidly dominates multilineage hematopoiesis in vivo. These data indicate that hyperactive Ras signaling is sufficient for the competitive advantage demonstrated by mutant HSCs and further implicate the HSC compartment as critical for therapy of JMML and CMML. Transplanted KrasG12D HSCs efficiently initiate T lineage acute lymphoblastic leukemia/lymphoma (T-ALL), which is associated with Notch1 mutations and with acquisition of LSC activity in differentiating thymocytes. These results further demonstrate that distinct self-renewing populations can arise through cooperating oncogenic mutations during cancer progression.
Although basal Mx1-Cre activity is low [18,19], many Mx1-Cre, KrasLSL-G12D mice that are not injected with pIpC ultimately succumb with MPD ([14] and unpublished data). This observation suggests that HSCs and/or progenitor cells that activate KrasG12D expression have a substantial proliferative advantage in vivo. To assess the kinetics of this process, we analyzed recombination in the myeloid progenitors of young Mx1-Cre, KrasLSL-G12D mice that were not treated with pIpC. Bone marrow cells from 3–5-wk-old animals were plated in methylcellulose medium to enumerate granulocyte-macrophage colony forming unit progenitors (CFU-GM), and individual myeloid colonies were genotyped by PCR. Surprisingly, nearly all CFU-GM in untreated animals were recombined as early as 3 wk of age (Figure 1A). We then examined more primitive populations in Mx1-Cre, KrasLSL-G12D mice crossed to a ROSA26 yellow fluorescent protein (YFP) reporter strain (Figure 1B) [20]. In these mice, cells expressing Cre are identified by YFP expression. The frequency of YFP+ cells in wild-type (WT) mice was within the expected range of background Cre expression [18,19]. However, we found a much higher incidence of YFP expression in bone marrow cells of Mx1-Cre, KrasLSL-G12D, ROSA26-YFP mice. This result was consistent among all populations analyzed, including the primitive Flk2− Lin−/lo Sca1+ c-kit+ (Flk2− LSK) compartment, which is highly enriched for HSCs [21]. To confirm that YFP expression correlated with KrasG12D expression, we also directly genotyped colonies formed by single Flk2− LSK cells. This method again revealed a predominance of KrasG12D-expressing cells in naïve Mx1-Cre, KrasLSL-G12D mice (Figure 1C). Together, these data are consistent with an advantage for Cre-expressing cells in untreated Mx1-Cre, KrasLSL-G12D mice.
The apparent outgrowth of KrasG12D Flk2− LSK cells in mice that were not injected with pIpC suggested that KrasG12D expression might increase proliferation in this compartment. To test this hypothesis, we stained bone marrow cells collected from KrasG12D mice and WT littermates 2 wk after pIpC treatment with antibodies to cell surface proteins and with dyes that stain DNA and RNA (7-aminoactinomycin D [7-AAD] and pyronin Y, respectively). These studies revealed a significant reduction in the number of quiescent Flk2− LSK cells in KrasG12D animals, which are identified by having a 2n DNA content and low pyronin Y staining (Figure 2A and 2B). Whereas the Flk2− subset of WT LSK contains roughly 80% cells in the G0 phase of the cell cycle, KrasG12D Flk2− LSK are only 50% quiescent. As an initial exploration of mechanisms regulating cell cycle progression, we analyzed D- and E-type cyclin expression by quantitative PCR in sorted Flk2− LSK cells from KrasG12D and WT mice. KrasG12D Flk2− LSK cells had significantly higher expression of cyclin D1 (Figure 2C).
Bone marrow mononuclear cells from KrasG12D mice display a hypersensitive pattern of CFU-GM progenitor growth, which is a cellular hallmark of JMML [12–14]. We found that this abnormal CFU-GM activity resides primarily in the common myeloid progenitor (CMP) compartment, and that these cells demonstrate enhanced proliferation in vivo ([22] and unpublished data). To investigate if this population could initiate and maintain MPD, we collected Lin−/lo Sca1− c-kit+ CD34+ FcγRII/III− CMPs from 5-wk-old KrasG12D mice and WT littermates by FACS, and transferred 10,000 of these cells into lethally irradiated recipients with 106 WT marrow cells for radioprotective support. Transplanted tester cells, recipient cells, and support cells were marked by expression of different isoforms of CD45, allowing them to be distinguished by flow cytometry. Transplanted KrasG12D or WT CMPs demonstrated robust day 8 spleen colony-forming unit (CFU-S8) activity, with colony size somewhat larger for KrasG12D input CMPs (Figure 3A). However, we detected less than 0.1% of circulating myeloid cells derived from transplanted KrasG12D or WT CMPs 1 mo after transplantation (Figure 3B and 3C). As expected from previous studies [23,24], KrasG12D and WT CMPs made minor contributions to the circulating B cell compartment with a statistically insignificant trend towards greater B cell production from KrasG12D cells. Taken together, these data indicate that KrasG12D CMPs do not initiate a hematologic disease. By contrast, transferring 500 Flk2− LSK cells from KrasG12D animals into lethally irradiated recipients rapidly resulted in durable multilineage reconstitution. KrasG12D cells dominated the T cell and B cell compartments sooner and to a higher degree than the progeny of WT Flk2− LSK cells in control animals. The myeloid series demonstrated a more variable time course, but KrasG12D derived cells also eventually out-competed WT cells (Figure 4).
Recipients of KrasG12D Flk2− LSK cells that were euthanized 3 mo after transplantation had mild to moderate MPD, manifested as leukocytosis and splenomegaly with myeloid and erythroid infiltration (Figure 5). As discussed below, we also found that all recipients of KrasG12D Flk2− LSK cells developed T-ALL 2–4 mo after adoptive transfer. Although early mortality from T-ALL precluded analyzing recipient mice beyond 3 mo, KrasG12D Flk2− LSK cells recapitulate the essential features of MPD seen in the original Mx1-Cre, KrasLSL-G12D model [13,14].
We next asked how KrasG12D expression affects HSC function. Immunophenotypic analysis revealed a 2-fold reduction in the number of marrow Flk2− LSK cells in KrasG12D mice 2 wk after pIpC injection (Figure 6A). This was mostly offset by an increased number of splenic Flk2− LSK cells. The reduction in marrow Flk2− LSK cells persisted in older animals (Figure S2). We also performed limit dilution studies to assess functional HSC activity. In these experiments, lethally irradiated recipients received decreasing numbers of whole bone marrow cells from either KrasG12D mice or WT littermates that had been injected with pIpC 2 wk earlier. Recipients were bled monthly, and flow cytometry was performed to assess whether CD45.1+ donor cells were able to provide durable (>2 mo) multilineage (myeloid, B cell, and T cell) engraftment. These studies demonstrated a striking 10-fold decrease in the number of long-term repopulating stem cells in KrasG12D animals compared to the WT littermate controls (Figure 6B; Table S1).
Progression of MPD despite a reduction in the size of the HSC compartment suggests an increased production of mature cells by each KrasG12D HSC. To address this possibility, we examined the patterns of reconstitution from either KrasG12D or WT HSCs in mice that received the limit dilution dose (1 × 105 KrasG12D and 1 × 104 WT bone marrow cells). As ∼50% of these recipients were engrafted with donor cells, the Poisson distribution predicts that approximately 70% of the engrafting mice received a single HSC and ∼24% received two HSC. By 2 mo after transplantation, donor KrasG12D marrow cells made a markedly greater contribution to recipient hematopoiesis than WT cells (Figure 6C).
We were able to compare repopulation of the stem cell compartment by WT and KrasG12D HSCs in a few lethally irradiated recipient mice before the onset of T-ALL. In these experiments, lethally irradiated recipients (CD45.2) were reconstituted with equal numbers of KrasG12D (CD45.1) and WT (CD45.1/CD45.2) Flk2− LSK cells, as well as 106 CD45.2 whole bone marrow cells (CD45.2) for radioprotection. Whereas only one of six recipients euthanized 2 mo after transplantation showed a clear bias towards KrasG12D-derived Flk2− LSK cells, mice that survived for 3 mo without evidence of diffuse T-ALL demonstrated an overwhelming bias towards KrasG12D-derived Flk2− LSK cells, myeloid progenitors, and mature myeloid cells (Figure S3).
Recipients that were injected with KrasG12D Flk2− LSK cells, either alone or in a 1:1 ratio with WT Flk2− LSK cells, uniformly became moribund 8–14 wk after transplantation. Examination of euthanized mice revealed massive thymic enlargement with an arrest in T cell development at the CD4/CD8 double positive stage, and variable infiltration of blast cells within the liver, spleen, and bone marrow (Figure 7A). Importantly, two of three animals from the bone marrow limit dilution transplantation assay that were engrafted with a single KrasG12D repopulating unit developed T-ALL that was identical to that seen in animals repopulated with 500 KrasG12D Flk2− LSK cells (unpublished data).
Next, we asked if T-ALL arose within the bone marrow or the thymus of mice transplanted with KrasG12D HSCs. To do this, we took advantage of our prior observation that recipients conditioned with sublethal irradiation fail to engraft with KrasG12D bone marrow [13]. Therefore, sublethally irradiated recipients can exclusively select for hematologic malignancies with more aggressive biologic behavior. We isolated bone marrow cells and thymocytes from primary recipients of KrasG12D Flk2− LSK cells that developed T-ALL, and injected each population into sublethally irradiated secondary recipients (Figure 7B). As expected, no animals demonstrated multilineage engraftment or MPD. Animals transplanted with thymocytes quickly succumbed with an identical T-ALL as primary recipients; however, none of the mice transplanted with bone marrow developed leukemia. Thus, whereas bone marrow-derived KrasG12D HSCs efficiently give rise to T-ALL, the T-ALL LSC population is initially restricted to the thymus in primary recipient mice.
These results suggested that one or more secondary mutations might have developed in a novel T-lymphoid clone. Somatic NOTCH1 mutations are common in human and murine T-ALL [25–28]. To determine if a similar mechanism might contribute to the evolution of KrasG12D HSCs to T-ALL LSCs, we performed Western blot analysis to detect cleaved (activated) Notch1 protein. Cleaved Notch1 was observed in thymocytes from diseased primary recipients, but not in bone marrow cells (Figure 7C), a finding that is consistent with the secondary transplant data. Direct sequencing around the PEST domain uncovered frameshift mutations in exon 34 of Notch1 in thymocytes from five of six animals transplanted with KrasG12D Flk2− LSK cells that developed T-ALL, but no mutations in thymocytes from control mice that received WT HSCs alone and remained well. The presence of a somatically acquired Notch1 mutation in a large fraction of the tumor provides a molecular indication of clonality.
The propensity of KrasG12D HSCs to generate T-ALL led us to investigate the effects of KrasG12D expression in early T-lineage cells. Flow cytometry of the bone marrow revealed no expansion of Lin− Flk2+ IL-7Rα+ c-kitint Sca1int common lymphoid progenitors (Figure 8A) [29]. However, KrasG12D mice demonstrated consistent thymic enlargement compared to age-matched littermate controls, even prior to the onset of T-ALL (Figure 8B). Immunophenotyping of primary thymocytes demonstrated an essentially normal distribution of CD4+ and CD8+ expression, with a slight trend toward an increased number of CD4/CD8 double negative (DN) cells (after excluding Mac1+ and Gr1+ infiltrating myeloid cells) (Figure 8B). By contrast, further examination of the DN compartment using the cell surface markers CD44 and CD25 uncovered skewed development in Kras mutant mice (Figure 8C). Together, these data demonstrate that oncogenic KrasG12D perturbs thymic homeostasis, particularly in early stages of thymocyte maturation.
We find that oncogenic KrasG12D expression in HSCs confers a strong in vivo growth advantage, increases proliferation, and results in MPD and T-ALL. In MPD, as in normal marrow, stem cell activity is restricted to the Flk2− LSK population, which represents less than 0.1% of nucleated marrow cells. While pathologic behaviors of more mature cells may contribute to tissue infiltration, anemia, and organomegaly, self-renewal is confined to this very primitive population. Therefore, hyperactive Ras signaling promotes excess proliferation in multiple hematopoietic compartments without immortalizing non–self-renewing cells. Similar data have been described in murine models of MPD based on BCR-ABL overexpression or loss of JunB, both of which also deregulate cytoplasmic signaling networks [5,30]. By contrast, recent experiments have provided direct evidence that some oncogenic transcription factors allow committed myeloid progenitors to acquire self-renewal ability [3,4]. Taken together, studies of myeloid oncogenes performed to date support the general idea that mutations that predominately alter cytoplasmic signaling networks and those that affect transcription factors controlling cell fate decisions comprise discrete complementation groups for the fully transformed phenotype [31,32].
Our limit dilution transplantation data demonstrate that oncogenic Kras confers a dramatic growth advantage in the HSC compartment. Under stringent conditions in which the contribution of a single WT HSC can barely be detected, the progeny of one (or at most three) KrasG12D HSC comprise a substantial fraction of the hematopoietic compartment. These studies provide direct experimental evidence that the outgrowth of malignant cells in MPD can be attributed to hyperactive Ras signaling in HSC. Our data suggest a pathogenic model in which JMML or CMML is initiated by a somatic mutation that deregulates Ras signaling in a single HSC. This idea is consistent with limited data from human patients and xenograft studies that implicate the HSC as the cell of origin for JMML [33–38].
To begin to address the mechanism by which mutant HSCs outgrow their WT counterparts, we analyzed the cell cycle in KrasG12D Flk2− LSK cells and found they are preferentially in cycle. The overexpression of cyclin D1 in KrasG12D cells we observed is consistent with many prior studies in cultured cell lines engineered to overexpress oncogenic Ras [39,40]. Intriguingly, HSCs in mice lacking D-type cyclins demonstrate severe proliferative defects and accumulate in the G0 and/or G1 phases of the cell cycle [41]. If increased cyclin D1 levels conversely result in excessive proliferation of HSC, then MPD in KrasG12D mice may be mediated in part by cyclin D1, similar to the requirement for cyclin D1 in a model of Ras-mediated breast cancer [42].
We demonstrate a substantial early growth advantage of HSCs that express KrasG12D; however, it is also possible that oncogenic Ras expression has a negative long-term impact on HSC function. Increased proliferation or oncogenic stress may ultimately detract from self-renewal capacity. Reduced HSC fitness was observed in Pten−/− mice, in which phosphatidylinositol 3-kinase signaling is hyperactive [43,44]. Similar effects of KrasG12D are suggested by the reduced numbers of HSCs in KrasG12D mice, although this finding could also reflect changes in the composition of the Flk2− LSK population or cell-extrinsic effects related to alteration of the marrow microenvironment. The rapid demise of primary KrasG12D mice from MPD, and of transplant recipients from T-ALL, precluded serial transplantation experiments to test the long term fitness of KrasG12D HSC. In addition, the 5-fold discrepancy between HSC numbers that were measured by flow cytometry versus limit dilution transplantation suggests a defect in engraftment of KrasG12D HSC. This idea is consistent with a prior report in which retroviral transduction of mutant NRAS appeared to reduce engraftment potential [45], and with extensive data showing that proliferating HSCs fail to engraft efficiently [46–50].
Despite the reduced number of HSC in the bone marrow of KrasG12D mice, our data are not entirely consistent with the idea that KrasG12D is a cell-intrinsic negative regulator of HSC self-renewal. If it were, we would expect specific loss of KrasG12D cells and outgrowth of WT or KrasLSL-G12D cells, because conditional models using Mx1-Cre typically retain a small pool of cells with the unrearranged locus [51,52]. However, we observed preferential retention of KrasG12D cells, even within the diminishing Flk2− LSK compartment. This observation suggests that residual WT HSCs are unable to compensate for the reduced HSC number. Therefore, we favor the hypothesis that the reduction in HSC number in KrasG12D mice reflects a disordered bone marrow microenvironment with reduced supportive capacity rather than a purely cell-intrinsic effect of KrasG12D in HSC.
The natural history of hematologic disease was different in primary KrasG12D mice than in recipients of transplanted KrasG12D HSC. A subtle but important finding is that MPD is established earlier in primary KrasG12D mice than in transplanted recipients. (Figure 5 and [13]; also see [16,53]). There are several possible explanations for this observation. The hematopoietic microenvironment in young mice may be more permissive for MPD than the irradiated bone marrow of an adult. Additionally, pIpC administration in Mx1-Cre mice may quickly create a field of KrasG12D myeloid progenitors that contributes to the rapid evolution of MPD in primary mice through cytokine-mediated autocrine and/or paracrine mechanisms [54]. It is also possible that nonhematopoietic stromal cells in Mx1-Cre, KrasLSL-G12D mice express K-RasG12D and contribute to the rapid onset of MPD in primary KrasG12D mice.
The kinetics of MPD development relate directly to the high frequency of T-ALL we observed in transplanted recipients as compared to primary KrasG12D mice. The apparent incidence of T-ALL is highly subject to selection bias, because animals that die from MPD cannot be evaluated for subsequent emergence of T-ALL. For example, we have observed ∼10%–15% of KrasG12D mice develop T-ALL on an inbred C57BL/6 strain background (unpublished data), but median time to death from MPD is shorter than the typical latency of T-ALL, and lymphoid tumors exclusively appear in mice with a relatively late onset of MPD. Similarly, we have not observed spontaneous T-ALL in F1 (C57BL/6 × 129Sv/Jae) mice, which die from MPD at a younger age than the C57BL/6 strain described here ([13] and unpublished data). In transplant recipients, aggressive T-ALL arose in mice that also had evidence of underlying MPD that was not yet severe enough to kill the animal. Together, these observations suggest that the attenuation of MPD in transplant recipients was central to the apparent increase in the incidence of T-ALL in the transplant setting.
Cooperation of hyperactive Ras and deregulated Notch signaling in T-ALL has recently been shown [55–57]. Our data extend these studies by demonstrating the remarkable efficiency with which KrasG12D HSCs can initiate T-ALL, and delineating how multiple cell types may participate in the stepwise acquisition of oncogenic mutations in hematologic cancers. In the Mx1-Cre, KrasLSL-G12D model, T-ALL is initiated by oncogenic Kras expression in HSC, but full transformation occurs when cooperating Notch1 mutations arise in a T-lineage cell. In this sense, both KrasG12D HSCs and the fully transformed thymocytes can be considered different types of malignant stem cells with distinct leukemogenic potentials.
One potential implication of these results is that the initiating Kras mutation creates conditions favorable for acquisition of cooperating mutations by increasing the size of susceptible lymphoid progenitor pools and/or conferring resistance to apoptotic signals during thymic selection. KrasG12D appears to most greatly affect the DN population that is characteristically undergoing TCR rearrangement, selection, and proliferation [58,59]. Interestingly, K-RasG12D protein expression may substitute for the pre-T cell receptor rearrangement at this critical checkpoint, thereby allowing propagation of thymocytes that would normally be edited [60]. Consistent with this idea, a patient with impaired lymphoid homeostasis and multiple lymphoid malignancies was recently reported to have a germline NRASG13D mutation, and oncogenic NRAS suppressed apoptosis of lymphocytes after cytokine withdrawal [61]. By contrast, Kindler et al. recently reported reduced thymic cellularity in Mx1-Cre, KrasG12D mice [56]. We speculate that the proliferative effects of K-RasG12D in the T cell compartment were obscured in those studies by the short interval between pIpC injection, which induces systemic interferon production, and histologic analysis.
The idea that patients may harbor a variety of genetically distinct LSC is consistent with studies of patients with chronic myeloid leukemia in blast transformation [62], and has important therapeutic implications. The need to eliminate partially transformed but self-renewing cells, like KrasG12D HSCs, will depend on their propensity to initiate a life-threatening disease. Targeted therapies that are directed against onco-proteins such as K-RasG12D will effectively eliminate premalignant clones only if the targeted lesions are initiating rather than secondary mutations. For example, inhibition of Notch signaling is an attractive therapeutic strategy for T-ALL that is being investigated in the clinic. However, if these cancers arise from aberrant HSCs that do not contain a NOTCH1 mutation and are not eradicated by treatment, relapse could occur through the acquisition of distinct cooperating mutations in a self-renewing preleukemic population. Consistent with this idea, studies of human T-ALL suggest that NOTCH1 mutation occurs as a secondary mutation in at least some cases, with some patients developing recurrent disease having distinct NOTCH1 alleles [63].
Finally, our data have implications for understanding the nature of cancer stem cell populations in nonhematopoietic malignancies. KRAS is the most frequent target of dominant oncogenic mutations in human cancer, and it is particularly important in carcinomas of the lung, pancreas, and colon. Analogous cancers arise in strains of mice expressing conditional oncogenic Kras alleles in these tissues [64–68]. Importantly, whereas oncogenic Kras expression efficiently initiates tumorigenesis in murine lung and pancreas, colon cancer is observed only when the tumor suppressor Apc is inactivated as well [68]. These data are consistent with studies of human patients, which imply that KRAS mutation occurs early in pancreatic cancer but typically after APC mutation in colon carcinoma [69–71]. Lung cancer in KrasG12D mice appears to be initiated in a distinct bronchio-alveolar stem cell population [72]. On the basis of these observations and our data, we speculate that, like HSCs, cells initiating pancreatic and lung cancer will possess inherent self-renewal potential, and that KRAS mutations only contribute to colon tumorigenesis in cells that have already acquired a mutation that enhances self-renewal. Uncovering specific proteins and pathways that are essential for the self-renewal and survival of Kras mutant cancer stem cell populations may reveal novel targets for therapeutic intervention in a variety of human cancers.
All animals were handled in strict accordance with good animal practice as defined by the relevant national and local animal welfare bodies, and all animal work was approved by the institutional animal research committee (University of California, San Francisco IACUC).
Animals were housed in a barrier facility at the University of California, San Francisco. All mice were of the C57BL/6 strain. As described [13], 21-d-old Mx1-Cre, KrasLSL-G12D were injected IP with 250 μg pIpC (Sigma). ROSA26 LSL-YFP reporter mice were a gift from C. Lowell. Ptprc (CD45) congenic mice were from Jackson Labs.
To assay CFU-GM, nucleated bone marrow cells (105) or splenocytes (2 × 105) were suspended in 1 ml methylcellulose medium (M3231, StemCell Technologies) with 10 ng/ml murine GM-CSF (PeproTech). Colonies were counted after 8 d. To assay Flk2− LSK cells, 100 sorted cells were plated in 1 ml methylcellulose medium (M3434, StemCell Technologies), which is provided containing IL-3, IL-6, stem cell factor, and erythropoietin, and incubated at 37 °C for 14 d. For Kras genotyping, individual colonies were picked with a 10 μl pipette and frozen overnight in 10 μl ddH2O prior to PCR analysis [13].
Flow cytometry was performed as described [21,24]. Staining was carried out in FACS Staining Buffer (FSB; HBSS with 2% heat inactivated FCS) at 0 °C unless otherwise stated. HSCs were defined as Lin−/lo Sca1+ c-kit+ Flk2−, and CMP as Lin−/lo Sca1− c-kit+ CD34+ FcγR−. Antibodies were from eBioscience except as specified. To identify HSCs, cells were first stained for 1 h with unconjugated lineage antibodies: CD3 (17A2, BioLegend); CD4 (RM4–5); CD5 (53–7.3); CD8 (53–6.7); B220 (RA3-6B2); Ter119 (TER-119); Mac1 (M1/70); Gr1 (RB6-8C5). Cells were washed then incubated for 30 min with TriColor goat-anti-rat F(ab′)2 (Invitrogen) and murine IgG (Sigma). Cells were washed again, stained for 30 min with Pacific Blue anti-Sca1 (D7, BioLegend); APC anti-c-kit (2B8, BioLegend); PE anti-Flk2 (A2F10, BioLegend); and 7-AAD (Sigma) 5 μg/ml for dead cell exclusion. IL-7Rα was detected using a PE conjugate (eBioscience). In this case, or when analyzing YFP+ cells, Flk2 was detected with biotinylated anti-Flk2 and APC-Alexa Fluor 750 streptavidin (Invitrogen). We found no difference in Flk2− LSK staining when Mac1+ cells were identified an on independent channel but not excluded. For myeloid progenitor analysis, PE anti-Flk2 was replaced with PE anti-FcγRII/III (93) and FITC anti-CD34 (RAM34), and cells were stained for an additional 30 min prior to analysis. Mature cells were assigned lineage using Pacific Blue anti-Mac1 and anti-Gr1 (myeloid), FITC anti-CD3 and anti-CD5 (T), and PE anti-B220 (B). T cell subsets in mice with T-ALL were distinguished with Pacific Orange anti-CD4 (Invitrogen), APC anti-CD8, FITC anti-CD3, and PE anti-CD5. To discern origin of cells in chimeric mice, antibodies to CD45.1 (A20; PE-Cy7) and CD45.2 (104; Alexa Fluor 700) were added. Primary thymocytes were analyzed with PE-Cy7 anti-CD4, Alexa 647 anti-CD8, PE anti-CD25, Pacific Blue anti-CD44, and FITC anti-Mac1 and anti-Gr1.
Prior to sorting, c-kit+ cells were enriched with anti-CD117 microbeads and an AutoMACS (Miltenyi). Cell sorting was performed on a FACSAria and flow cytometry on an LSRII, both using FACSDiva software (BD). Data were analyzed using FlowJo software (TreeStar).
Recipients received a single fraction of 950 rads for lethal irradiation, or 450 rads for sublethal irradiation, from a cesium source. Donor cells were prepared in 100 to 200 μl of FSB, and injected retro-orbitally into anesthetized mice. Recipients received water with neomycin and polymyxin for 2 wk. Blood counts were monitored monthly using a Hemavet 950FS (Drew Scientific).
Limit dilution was performed as described [73]. Lethally irradiated recipients were transplanted with 1 × 104, 3 × 104, or 1 × 105 nucleated bone marrow cells along with 106 unfractionated WT bone marrow cells for radioprotection. Peripheral blood was analyzed monthly by flow cytometry; mice with detectable engraftment in myeloid, B and T lineages 2 mo after transplantation were scored as positive and L-Calc software (StemCell Technologies) was used for statistical analysis.
For the CFU-S8 assay, lethally irradiated mice received 105 CMPs without support cells. Spleen colonies were observed 8 d later by gross examination at harvest and after 48 h fixation in 10% formalin. Sections were stained with hematoxylin and eosin.
Cells were lysed in 1% NP-40 with 30 mM NaF, 30 mM β-glycerophosphate, 20 mM Na4P2O7, 1 mM Na3VO4, and Complete (Roche) and analyzed after SDS-PAGE using cleaved Notch1 and β-actin antibodies (Cell Signaling). Genomic DNA was PCR-amplified sequenced bidirectionally using 5′-ATAGCATGATGGGGCCACTA-3′ and 5′-GCCTCTGGAATGTGGGTGAT-3′.
Staining with 7-AAD and pyronin Y was performed as described [74] in nucleic acid staining solution (NASS; 0.1M phosphate-citrate buffer [pH 6.0] [Sigma], 5mM EDTA, 0.15M NaCl, 0.5% BSA), with 0.02% saponin (Sigma). Nucleated bone marrow cells were stained for 30 min with FITC lineage antibodies (CD3, CD4, CD5, CD8, B220, Ter119, and Gr1), Pacific Blue anti-Sca1, unconjugated anti-CD16/32 (2.4G2, UCSF hybridoma core), and biotinylated anti-Flk2. Cells were washed and then stained with APC-Alexa Fluor 750 streptavidin. Cells were washed again and resuspended in 500 μl of NASS with 1 μg/ml 7-AAD (Sigma), and incubated at room temperature for 30 min, then on ice for 5 min. Pyronin Y (Sigma) was then added to 1 μg/ml, and cells were incubated for an additional 10 min before being washed and resuspended in 200 μL of FSB. Cells were finally stained with APC anti-c-kit and FITC anti-Mac1. Differences in the quiescent fraction of HSCs were analyzed using an unpaired t-test.
The assay was performed as described [50]. Flk2− KLS cells from animals pooled by genotype were double-sorted directly into RNA binding/lysis buffer from the RNEasy kit (Qiagen), and total RNA was extracted per instructions. First strand cDNA synthesis was performed using a SuperScript III kit (Invitrogen) per manufacturer's instructions. Reactions were performed in an ABI-7900 sequence detection system using SYBR green according to manufacturer's instructions (Applied Biosystems). Each amplification was performed in 10 μl with a template cDNA equivalent of 100 sorted HSCs. Each sample was tested in triplicate with each primer pair, and normalized to β-actin expression. Due to limiting numbers of doubly sorted cells, final cell purity was not analyzed; however the staining characteristics of c-kit-enriched and singly sorted cells are presented in Figure S1.
To analyze repopulation by a single HSC, we identified a cohort of mice receiving a cell dose yielding engraftment in only 50% of recipients. The Poisson distribution indicates that the probability of a mouse receiving k HSCs is given by
where n is the average HSC number per mouse. The average that yields k = 0 HSC at a rate of 0.5 is given by 0.5 = e−n, which is solved to give n = 0.693. Using this value for the average number of HSC per mouse, the Poisson distribution can be used to estimate the likelihood of any mouse receiving a given number of HSC: f(0) = e−ln 2 = 0.5; f(1) = (ln 2)e−ln 2 = 0.346; f(2) = 1/2 (ln 2)2e−ln 2 = 0.120; f(3) = 1/6 (ln 2)3e−ln 2 = 0.028. The proportion of mice expected to receive four or more HSCs is the remainder, 1 − (0.5 + 0.346 + 0.12 + 0.028) = 0.006, or 0.6%.
|
10.1371/journal.pntd.0004506 | A Four-Point Screening Method for Assessing Molecular Mechanism of Action (MMOA) Identifies Tideglusib as a Time-Dependent Inhibitor of Trypanosoma brucei GSK3β | New therapeutics are needed for neglected tropical diseases including Human African trypanosomiasis (HAT), a progressive and fatal disease caused by the protozoan parasites Trypanosoma brucei gambiense and T. b. rhodesiense. There is a need for simple, efficient, cost effective methods to identify new molecules with unique molecular mechanisms of action (MMOAs). The mechanistic features of a binding mode, such as competition with endogenous substrates and time-dependence can affect the observed inhibitory IC50, and differentiate molecules and their therapeutic usefulness. Simple screening methods to determine time-dependence and competition can be used to differentiate compounds with different MMOAs in order to identify new therapeutic opportunities.
In this work we report a four point screening methodology to evaluate the time-dependence and competition for inhibition of GSK3β protein kinase isolated from T. brucei. Using this method, we identified tideglusib as a time-dependent inhibitor whose mechanism of action is time-dependent, ATP competitive upon initial binding, which transitions to ATP non-competitive with time. The enzyme activity was not recovered following 100-fold dilution of the buffer consistent with an irreversible mechanism of action. This is in contrast to the T. brucei GSK3β inhibitor GW8510, whose inhibition was competitive with ATP, not time-dependent at all measured time points and reversible in dilution experiments. The activity of tideglusib against T. brucei parasites was confirmed by inhibition of parasite proliferation (GI50 of 2.3 μM).
Altogether this work demonstrates a straightforward method for determining molecular mechanisms of action and its application for mechanistic differentiation of two potent TbGSK3β inhibitors. The four point MMOA method identified tideglusib as a mechanistically differentiated TbGSK3β inhibitor. Tideglusib was shown to inhibit parasite growth in this work, and has been reported to be well tolerated in one year of dosing in human clinical studies. Consequently, further supportive studies on the potential therapeutic usefulness of tideglusib for HAT are justified.
| Drug discovery for neglected tropical diseases must use efficient methods due to limited resources. One preferred drug discovery strategy is target-based drug discovery. In this strategy it is assumed that drug action begins with binding of a drug to its target. However, while binding is required, it is not sufficient to describe all the molecular interactions that translate binding to a therapeutically useful response. The contribution of aspects of the molecular mechanism of action (MMOA) such as time-dependence and substrate competition can influence concentration response relationships. To address this, a four point MMOA methodology was developed to evaluate time-dependence and substrate competition. We used this method to evaluate the MMOA for T.brucei GSK3β inhibitors, and observed tideglusib to have a time-dependent, ATP-competitive mechanism that differentiated it from rapidly reversible inhibitors, such as GW8510. Adjusting the enzyme assays to account for these mechanisms showed that GW8510 and tideglusib had similar activities for TbGSK3β. However, this similarity did not translate to cellular activity, where GW-8510 was more active than tideglusib (0.12 μM to 2.3 μM, respectively). These data suggest that factors other than TbGSK3β MMOA differentiate the effect of these molecules against T. brucei.
| New medicines are needed for neglected tropical diseases (NTDs). Drug discovery and development in all disease areas is inefficient due to high failure rates that increase the costs. The impact of the high failure rates on NTDs is very significant as a result of limited resources afforded to these diseases because of a lack of commercial markets for the medicines [1]. There is an urgent need for cost effective methods and strategies that expedite opportunities to identify new medicines for NTDs [1].
As part of our on-going efforts to identify new treatment molecules for human African trypanosomiasis (HAT; also known as sleeping sickness) we have been investigating approaches to identify new molecular mechanisms of action (MMOAs) early in the discovery process. HAT is a deadly infection affecting mostly impoverished areas in rural sub-Saharan Africa, caused by the protozoan Trypanosoma brucei. There are currently an estimated 30000 cases of HAT, with approximately 70 million people in 36 African countries at risk [2]. At this time, five medicines are available for HAT, all of which must be administered intravenously or intramuscularly, and three of these treat late-stage infections, when the parasite has crossed the blood-brain barrier. Notable limitations of these treatments include ineffectiveness, long periods of drug administration, cost, toxicity, and other severe medical side effects [3]. Consequently, there is a critical need to discover and develop safer, more effective and less expensive therapies for HAT. There is hope that this unmet medical need will be addressed with the promising candidates currently in development, fexinidazole and SCYX-7158 [4]. However, there is still a need for back-up strategies for these compounds in case they do not meet the promise and to also address potential resistance.
Drug discovery and development is facilitated by an understanding of how drugs work (i.e. their mechanism of action). How a drug works is a key feature of the pharmacological response, the therapeutic index, and the drug’s therapeutic usefulness [5–7]. Knowledge of a drug’s mechanism will also facilitate optimization and clinical development. It has been long recognized that pharmacological action begins with an interaction between two molecules. Ehlrich noted in 1913 that a substance will not work unless it is bound “corpora non agunt nisi fixata” [8]. However, binding alone is not sufficient for a medicine to produce an effective and safe response that is therapeutically useful. The molecular mechanism of action (MMOA) through which a medicine connects binding to the response is also important [9, 10]. For example, two similarly structured molecules can bind to an enzyme with similar affinity; however, one will bind as a substrate in a manner suitable for the catalytic reaction while the other will inhibit the reaction. For receptors, two similarly structured molecules can bind to a receptor with similar affinity; however, one will initiate the response (agonist) whereas the other will block the response (antagonist).
Understanding drug action at the molecular level can facilitate the rational design of new medicines as well as provide opportunities to identify new therapies. Understanding the MMOA can provide opportunities to identify new molecules that are differentiated from previous molecules, molecules that can be repurposed for new indications or molecules that have been previously overlooked. A detailed understanding of the MMOA is also important for interpretation of in vitro/ in vivo correlations in target validation studies and understanding pharmacokinetic/pharmacodynamics (PK/PD) relationships.
Two important features of MMOA that have been shown to differentiate medicines are binding kinetics and binding competition. The binding kinetics are the rate at which a molecule binds (association rate) and debinds (dissociation rate). A reaction with a slow dissociation rate can be functionally irreversible when the dissociation rate is sufficiently slow or covalent. Competition occurs when two molecules compete for the same binding site and will result in decreased fractional occupancy of the drug bound to the target. The decrease in fractional occupancy due to competition can be overcome by increasing the concentration of the drug. The decrease in fractional occupancy due to competition can also be overcome with slow dissociation kinetics and irreversibility. This pharmacological behavior is described as insurmountable drug action.
Many examples demonstrate the important role of binding kinetics in effective drug action [9, 11, 12]. Aspirin is an irreversible inhibitor of prostaglandin H2 synthases (also known as cyclooxygenase, COX), whereas ibuprofen is a rapidly reversible inhibitor of these enzymes with a fast dissociation rate [13, 14]. The irreversibility of aspirin contributes to its usefulness for prevention of atherothrombotic disease [15, 16] and differentiates aspirin from ibuprofen. Irreversibility can be achieved by covalent binding as well as long residence times in a system not at equilibrium to provide insurmountable pharmacological behavior [17]. Slow dissociation kinetics in a system not at equilibrium contributes to the use-dependence behavior of channel blockers [18] and the insurmountable behavior of many receptor blockers, including the well-documented, angiotensin receptor blockers [19, 20]. These examples illustrate some of the advantages to time-dependent behavior including a greater inhibition of activity and longer lasting pharmacodynamic behavior and target occupancy enabling administration of lower doses and in some cases greater durability. These mechanistic behaviors contribute to the effectiveness and utility of many anti-infectives including the irreversible inhibitor, penicillin [21], and isoniazid [22, 23]. This behavior also contributes to the effectiveness of many other medicines including lapatinib, tiotropium, and candesartan to name a few [9, 11]. For completeness it must be noted that long-residence time/irreversibility is not suited for all system. When there are liabilities due to mechanism-based toxicity (on-target toxicity), long residence time/irreversible behavior is not appropriate [5, 9].
Competition of a drug with endogenous substrate for binding will reduce the fractional occupancy and may result in a loss of effectiveness. This will require higher concentrations to achieve the same effect and thereby decrease the selectivity, increase the potential for toxicity as well as provide a challenge for pharmaceutical development to administer the drug at a sufficient dose and concentration at the site of action to achieve efficacy.
Competition with endogenous substrates is particularly relevant for protein kinases where ATP competitive inhibitors must compete with high concentrations of endogenous ATP for binding and inhibition of the kinase activity. The physiological concentrations of ATP are estimated to be in excess of 1 mM, which will result in a ≈ 100 fold shift in IC50 for inhibitors with a Km for ATP of 10 μM (see theoretical explanation below). Importantly, mechanisms that avoid competition with ATP have been identified, including slow dissociation kinetics (long residence times) and non-competitive mechanisms. Wilson and coworkers recently demonstrated that the MMOA of the first approved kinase inhibitor, Gleevec (imatinib mesylate), involves time-dependent binding that is important to its action and selectivity [24]. This time-dependent, mechanistic behavior was not identified until over a decade after the drug had been discovered. In general, the many molecular mechanistic features important to the action of first in class drugs are identified long after a drug is discovered, and consequently, not used to inform optimization and development [6]. Accordingly, a simple, cost-effective method to identify MMOAs for lead compounds will be valuable to R&D, notably in resource limited diseases, such as NTDs.
Towards this goal, we have used a simple four point method to characterize the MMOA of compounds for time-dependence and competition as part of our efforts to identify novel and therapeutically useful protein kinase inhibitors of T. brucei. The method involves measuring activity at one inhibitor concentration with and without a preincubation (30 min in this case) to determine time-dependence, and at two substrate concentrations (0.5 Km and 5x Km in this work) to determine competition. A shift in the time dependence indicates that the inhibitor binding does not reach equilibrium in the time frame of the experiment. The inability to rapidly reach equilibrium can be due to a multi-step binding mechanism (Fig 1). A loss of activity with higher substrate concentrations is consistent with a substrate competitive mechanism of inhibition. A lack of shift in inhibition can be considered as noncompetitive, such as inhibitor binding in a different site. However, a combination of noncompetitive behavior and time-dependence can be diagnostic of a molecule with insurmountable pharmacological behavior due to slow binding kinetics (including irreversibility) that prevents competition. For example, an irreversible inhibitor will appear noncompetitive even when it is bound in the substrate binding site.
The utility of the four point MMOA method was demonstrated with GSK3β isolated from T. brucei (TbGSK3β) and the identification of tideglusib (NP-12, NP031112) as a time-dependent, competitive inhibitor of TbGSK3β. Further evaluation of tideglusib showed irreversible behavior against the enzyme and inhibition of parasite proliferation at low micromolar concentrations. Tideglusib, previously evaluated in the clinic, has a good safety profile in humans [25, 26] and therefore, warrants further studies to evaluate its potential for HAT. Altogether this work demonstrates a method for rapidly evaluating MMOA that can help identify opportunities for new NTDs.
Reagents, unless noted, were purchased from Sigma-Aldrich (Saint Louis, Missouri). GSM peptide (sequence = RRRPASVPPSPSLS RHS(pS)HQRR, where pS is a phosphorylated serine residue) [27] was purchased from EMD Millipore Corporation (Temecula, California). ADP-Glo kit was purchased from Promega Corporation (Madison, Wisconsin). Tideglusib and GW8510 were purchased from Sigma-Aldrich (Saint Louis, Missouri).
The short-form of GSK3β from T. brucei was expressed in E. coli with a C-terminal His tag and purified to homogeneity as described in Materials and Methods. Conditions for the enzyme assay were optimized for linearity with time and enzyme concentration using GSM as a phosphoryl acceptor and measuring ADP production using Promega ADP-Glo reagents. The substrate versus rate plots showed saturable kinetics for the two substrates GSM and ATP, with Kms of 23 μM and 21 μM, respectively and Vmax of 31 and 34 s-1 in the two studies (Fig 2).
IC50 for inhibition TbGSK3β was determined as a function of preincubation time (Fig 3). Preincubated reactions were preincubated with inhibitor, TbGSK3β, GSM peptide, and buffer at room temperature; after 30 minutes reactions were initiated with the addition of ATP. Reactions with 0 min preincubation were initiated with the addition of TbGSK3β. Reactions were run for 5 min at room temperature, and stopped at 80°C. ADP product formation was measured by ADP-Glo kit. As shown in Fig 3 a shift in IC50 with preincubation time was observed for tideglusib, but not for GW8510. The IC50 values for tideglusib were 43 nM and 173 nM for preincubated reactions and non-preincubated reactions, with standard errors of 6.5 nM (N = 2) and 22.6 nM (N = 5), respectively. The IC50 values for GW-8510 were 14.5 nM and 15.1 nM for preincubated reactions and non-preincubated reactions, respectively.
The four point methodology was demonstrated for tideglusib and GW8510 (Fig 4). Tideglusib decreased the enzyme activity following preincubation (gray bars no preincubation; black bars 30 min preincubation) at both 10 μM and 100 μM ATP (Fig 4A and 4C, respectively). On the other hand there was no effect of GW8510 preincubation on the TbGSK3β activity at 10 μM and 100 μM ATP (Fig 4B and 4D, respectively). These results can be interpreted that GW-8510 binds to TbGSK3β corresponding to a one-step model, to from an EI complex (Fig 1) while tideglusib follows a two-step model, forming an initial EI complex that rearranges with time to a more stable E’I complex (Fig 1).
In ATP competition studies the percent inhibition of activity by GW8510 was less at higher ATP concentrations (100 μM) than lower ATP (10 μM), 28% vs 67%, respectively irrespective of preincubation time (Fig 4B and 4D). Tideglusib’s activity was also decreased at when there was no preincubation but not when there was a preincubation (100 μM) (Fig 4A and 4C). This suggested that 1) tideglusib’s initial binding to form EI was competitive with ATP, and 2) that there is a time-dependent transition from the EI state to a more stable E’I state and is not sensitive to ATP competition. Due to this transition, tideglusib is more potent than GW8510 at high ATP concentrations following preincubations. We did not observe a shift in tideglusib activity at higher concentrations [80 μM] of the peptide substrate, GSM (supplementary, S1 Fig) indicating that the tideglusib does not interact with the peptide substrate binding site.
Further evaluation of time dependence under more physiological conditions of no preincubation and higher ATP concentration was evaluated using enzyme progress curves. Progress curves measure the progress of the enzyme reaction with time [28]. In these reactions the time-dependent inhibition of TbGSK3β by tideglusib and GW8510 was evaluated under conditions in which inhibitor, 100 μM ATP and 80 μM GSM were combined in buffer and the reactions started with TbGSK3β (Fig 5). The reactions were stopped by heating at the specific times (5, 10, 20, 30 and 60 min). The progress of the reaction, as measured by product formation (ADP) with no inhibitor (DMSO control), was linear with time. When tideglusib was added, the percent inhibition was greater at 30 min as compared to 5 min, consistent with time-dependent loss of activity. In contrast, the effect of GW8510 on the rate of product formation was similar at all time points, consistent with rapid equilibrium reversible inhibition. Also of interest with tideglusib is that the slope of the progress curves approached zero at the later incubation times in a concentration dependent manner, suggesting complete inhibition of the enzyme at the later time points at all concentrations, behavior consistent with irreversible inhibition (Fig 5A).
Tideglusib and GW8510 were evaluated for irreversibility by measuring the TbGSK3β activity following a 100-fold dilution with reaction buffer. In these experiments tideglusib and GW8510 [100 nM] were preincubated with 50 nM TbGSK3β for 30 min at which time the reaction was diluted 100 fold into a solution containing 100 μM ATP and 80 μM GSM. The activity was then measured for up to two hours (Fig 6). The activity was compared to a DMSO control (also preincubated for 30 min and activity measure up to 2 hr). The final concentration of tideglusib and GW8510 following dilution was 1 nM. Inhibition of enzyme activity was retained following dilution of tideglusib whereas enzyme activity was recovered in the reactions with GW8510. The results are consistent with reversible behavior of GW8510 and irreversible behavior of tideglusib. This experiment does not distinguish between a very slow dissociation rate or covalent binding. As discussed below, tideglusib was previously demonstrated to have irreversible inhibition against human GSK3β with an IC50 of 5 nM following preincubation, with no evidence for covalent binding [29]. These investigators concluded in that work that tideglusib was a functionally irreversible inhibitor [29]. The results shown in Fig 6 are consistent with similar behavior for TbGSK3β.
The four point MMOA assay was used to characterize molecules that had previously been identified as TbGSK3β inhibitors. The concentrations chosen for the MMOA assays were determined from the IC50 in the HTS screens. A good starting concentration for the time-dependent studies is the IC25 (concentration at 25% inhibition), since time-dependent inhibition will increase with preincubation time. For the competition analysis, use of an inhibitor concentration equal to the IC50 at substrate Km (theoretically 2x Ki) will allow observation of the decrease in inhibition due to higher substrate concentrations (Fig 5). In practice, concentrations must be chosen that will allow the change to be observed and will be specific to each inhibitor. The activities of these molecules were not time dependent since they were unchanged after 0 and 30 minutes incubation (Table 1) and all were found to be competitive with ATP since the percent inhibition was less using 100 μM versus 10 μM (Fig 7).
Tideglusib was evaluated in T. brucei parasite assays as previously described in Materials and Methods. The concentration of inhibitor at which trypanosome proliferation was inhibited by 50% (GI50) was 2.3 μM (Fig 8). In comparison, Ojo and coworkers reported the activity of GW-8510 in a similar 48 hr parasite growth assay with T. brucei brucei strain 427 to have an EC50 of 119 nM [30].
The effect of tideglusib on trypanosome viability after short term exposure was evaluated to determine if the molecule is cidal or cytostatic. During the short 6 hour treatment at high trypanosome density (5 x 105/mL), tideglusib (at 5 or 10 μM) markedly slowed trypanosome growth (Fig 9B). In fact at 10 μM tideglusib, proliferation was entirely halted. However, dilution of tideglusib (or DMSO) from the media no longer arrested trypanosome proliferation, with a slight lag in recovery after 10 μM tideglusib exposure. After the initial lag in recovered growth, the cells grew with a normal rate of approximately 11 doublings in 48 hours (Fig 9C). This is in contrast to the control, pentamidine, which killed all observed trypanosomes within 24 hours of the 6 hour exposure (Fig 9C).
These results show that tideglusib was cytostatic at high trypanosome cell densities. Trypanosomes treated with 10 μM tideglusib did demonstrate a slight lag in normal proliferation rates after the drug pressure was removed (Fig 9C). We conclude that tideglusib is better at controlling proliferation during short exposure, but pentamidine wins over the long run.
As part of our on-going efforts to identify new treatment molecules for NTDs we have been investigating approaches to identify new MMOAs early in the discovery process. Currently the identification of MMOA occurs after a molecule has been identified as a lead or clinical candidate using rigorous biochemical and pharmacological methods. Previous analyses have suggested that an optimal MMOA can provide an improved therapeutic index, however the challenge is to identify compounds with an optimal MMOA [6, 31]. One approach to accomplish early identification of MMOA is to identify clusters of compounds with different MMOAs. MMOA identification would be used in addition to other factors to cluster molecules, such as activity, chemical structure and physical properties. Representatives from these compound clusters will then be evaluated empirically in phenotypic screens to identify actives and new opportunities. Accordingly, simple, efficient methods are required to easily identify molecules with different mechanistic features. We report here a simple, efficient four point MMOA screening method for evaluating compounds for time-dependent activity and competitive inhibition.
In practice, this four point MMOA method can be applied at any stage in the drug discovery process. In early discovery, hits are identified in screens at one or a few concentrations and confirmed with IC50 determinations. The IC50 concentration will inform the concentration to use for the four point MMOA method. It should be noted that this method can be employed with any enzyme, regardless of the number of substrates by using two concentrations of the substrate of interest (one around Km and the other above Km).
It is worth mentioning that this four point MMOA method is a simplification of more rigorous methods that are common practice in biochemistry and pharmacology labs. Typically studies to look for time-dependence and competition are accomplished by investigating complete dose-response curves whereby the results inform a detailed understanding of the molecular interaction. In preparation for these definitive studies, smaller experiments are used to establish conditions. We have found that the four point MMOA studies are suitable for screening purposes and can be sufficiently reliable to differentiate compounds. In previous work, we applied a similar strategy to differentiate kinetics for antagonists binding to CCR5 using IC50 measures in receptor competition assays with and without preincubation [32]. Guo et al have recently described a dual point competition association assay for assessing ligand-receptor binding kinetics. Their approach was found to be useful for ranking molecules [33]. The four point MMOA method described here is a simpler method as compared to both of these, requiring less data points, thereby resulting in more efficient use of time and resources. Interesting findings from the four point MMOA method can be evaluated in more detailed assays as warranted.
Using the four point MMOA method, tideglusib was determined to be a time-dependent inhibitor of TbGSK3β. Its effectiveness increased with time until its inhibition activity was similar to GW8510, a well-documented potent TbGSK3β inhibitor [30](Figs 3 and 4). Both tideglusib and GW8510 competed with ATP for binding to TbGSK3β without preincubation. Interestingly, tideglusib’s inhibition increased with preincubation time (Figs 3 and 4), incubation time (Fig 5) and showed non-competitive behavior following preincubation (Fig 4). In comparison, the MMOA of GW8510 was competitive following preincubation (Fig 4). These results show tideglusib to have a different mechanistic profile for TbGSK3β than GW8510. Tideglusib shows a time dependent switch from competitive to non-competitive behavior. This type of behavior is seen with other slow-dissociating molecules, which has been termed insurmountable and is a property of successful medicines [5, 17, 19, 20, 34]. We also used the four point MMOA method to characterize other kinase inhibitors that were previously identified in screens as TbGSK3β inhibitors (Fig 7, Table 1). All were found to behave mechanistically as non-time dependent, competitive inhibitors.
Adjusting the enzyme assays to account for these mechanisms showed that GW8510 and tideglusib had similar activity for TbGSK3β under more physiological conditions of preincubation and 100 μM ATP (Fig 4). However, this similarity did not translate to cellular activity, where GW-8510 was more active than tideglusib (0.12 μM to 2.3 μM, respectively [29]). These data suggests that factors other than the TbGSK3β MMOA differentiate the effect of these molecules against T. brucei, and demonstrates the value of early MMOA determination to help understand target specific hypotheses. Many factors may contribute to the difference in activity including poor solubility of tideglusib that limits cellular activity and/or that the cellular activity of GW-8510 maybe due in part to a non-TbGSK3β mechanism. While further studies are needed to evaluate these issues, the understanding of their different TbGSK3β MMOAs will assist in designing and correctly interpreting the data.
The MMOA for tideglusib inhibition of TbGSK3β identified with the four point MMOA assay is consistent with that observed for inhibition of human GSK3β as reported by Dominquez and coworkers [29]. Using a more sophisticated analysis they also reported a binding mechanism which had an ATP-competitive component [29]. Furthermore, their studies on time-dependence led to the conclusion that tideglusib is a functionally irreversible inhibitor of human GSK3β and that the duration of the pharmacological effect caused by this behavior may be exploited to maximize its therapeutic potential. The work described here for TbGSK3β inhibition by tideglusib shows that the mechanism of inhibition is similar against the enzymes from human and T. brucei.
In humans, GSK3β is a regulatory kinase for over 40 different proteins in a variety of pathways, and has been implicated in a number of diseases including: Type II diabetes (Diabetes mellitus type 2), Alzheimer's Disease, inflammation, cancer, and bipolar disorder [26]. Furthermore, studies have shown that one of the T. brucei GSK3 homologs (TbGSK3 short), is necessary for cell growth and viability and thus may serve as a potential drug target for the treatment of HAT [30]. In a recent clinical trial designed to investigate safety and efficacy, tideglusib administered in escalating doses of up to 1000 mg/day to 30 Alzheimer’s patients for 6 weeks was generally well tolerated [25]. Tideglusib was also well-tolerated in phase II studies of Alzheimer’s disease and progressive supranuclear palsy [35, 36]. Blood levels for tideglusib were reported in mice of approximately 2.25±1.55 μg/ml (6.7 μM) after oral administration [37].
In summary, we used the four point MMOA method to evaluate the MMOA for T.brucei GSK3β inhibitors and observed tideglusib to have a time-dependent mechanism which differentiated it from rapidly reversible inhibitors, such as GW8510. Tideglusib was shown to inhibit parasite growth in this work and has been reported to be well tolerated in one year of dosing in human clinical studies [35–37]. Consequently, further supportive studies on the potential therapeutic usefulness of tideglusib for HAT are justified.
|
10.1371/journal.ppat.1006166 | Replication-Coupled Recruitment of Viral and Cellular Factors to Herpes Simplex Virus Type 1 Replication Forks for the Maintenance and Expression of Viral Genomes | Herpes simplex virus type 1 (HSV-1) infects over half the human population. Much of the infectious cycle occurs in the nucleus of cells where the virus has evolved mechanisms to manipulate host processes for the production of virus. The genome of HSV-1 is coordinately expressed, maintained, and replicated such that progeny virions are produced within 4–6 hours post infection. In this study, we selectively purify HSV-1 replication forks and associated proteins from virus-infected cells and identify select viral and cellular replication, repair, and transcription factors that associate with viral replication forks. Pulse chase analyses and imaging studies reveal temporal and spatial dynamics between viral replication forks and associated proteins and demonstrate that several DNA repair complexes and key transcription factors are recruited to or near replication forks. Consistent with these observations we show that the initiation of viral DNA replication is sufficient to license late gene transcription. These data provide insight into mechanisms that couple HSV-1 DNA replication with transcription and repair for the coordinated expression and maintenance of the viral genome.
| HSV-1 is a ubiquitous human pathogen that causes persistent infections for the lifetime of the infected host. Of major interest are the mechanisms underlying how the virus utilizes cellular resources to rapidly replicate with high fidelity. We show that DNA repair and late transcription are coupled to genome replication by identifying the viral and cellular factors that associate with replicating viral DNA. In addition to transcription and repair, the results also describe how RNA processing and virion packaging are temporally coordinated relative to genome replication.
| Herpesviruses belong to a large family of enveloped double-stranded DNA viruses that cause persistent infections in a broad range of metazoan hosts. Much of the infectious cycle occurs in the nucleus of cells, including events underlying transitions between latent and productive states, viral gene expression, DNA replication, and packaging of nascent genomes into capsids. Herpesviruses share strong evolutionary relationships with their hosts and have evolved mechanisms to subvert host antiviral restriction pathways and DNA damage responses, while adapting host mechanisms for viral genome maintenance and gene expression [1,2]. Although virus-host interactions largely determine the outcome of infection, there is limited understanding of how the dynamic interplay between host factors and viral DNA contribute to processes that occur on viral genomes.
Herpes simplex virus type 1 (HSV-1) is a ubiquitous human pathogen that causes recurrent contagious oral and genital sores and serious infections of the eye and central nervous system. HSV-1 consists of a 152 kilobasepair linear genome that codes for over 80 proteins [3,4]. Like all herpesviruses, during productive infection, HSV-1 largely utilizes its own DNA synthetic machinery for genome replication but depends on host RNA polymerase II (Pol II) for the transcription of viral genes [5].
HSV-1 DNA replication requires at least seven viral DNA replication proteins [6] and initiates at least one of three origins of replication [7]. These sites are bound by the origin binding protein (UL9) [8], which in combination with the single-stranded DNA binding protein (ICP8) distorts the origins [9]. This is followed by unwinding and priming by the helicase/primase complex (UL5/UL8/UL52) for the activation of DNA synthesis by the two subunit viral DNA polymerase (catalytic subunit UL30, processivity subunit UL42) [10]. While this set of viral proteins includes a DNA-dependent DNA polymerase and functional analogs of other cellular replication proteins, it is not sufficient to drive origin-dependent replication in vitro. This implies the requirement for additional viral or cellular proteins. Furthermore, the existence of branched, concatemeric, and isomeric viral DNA structures suggests that alternative modes of replication involving recombination must also exist [11–15].
Although the viral DNA polymerase has a high probability to incorporate incorrect bases, the HSV-1 genome is maintained with high fidelity [16]. Furthermore, the genome is known to contain several nicks and gaps [17–19], which would likely impede replication and transcription. Some cellular DNA repair proteins have been implicated in productive stages of infection including those involved in mismatch repair (MMR) and double strand break (DSB) repair [1]. However, HSV-1 inhibits classic cellular DSB repair pathways including nonhomologous end joining and homologous recombination. It is therefore likely that the virus adapts cellular repair proteins to mediate processes for virus-specific genome maintenance.
HSV-1 transcription occurs through a tightly ordered cascade of immediate early (IE)(α), early (β), and late genes (γ) [20,21]. IE genes are expressed without de novo viral protein synthesis and include regulatory factors that alter host cell metabolism and control viral gene expression. The IE gene product ICP4 is a transcriptional repressor of IE genes and activator of early and late genes [22–24]. Early gene products include viral DNA replication factors and late gene products include virion assembly and structural proteins. Late genes are further classified into leaky late or true late, depending on the extent to which they depend on viral DNA replication for expression. True late gene expression strictly depends on viral DNA synthesis, although mechanisms by which transcription and replication are coupled are not understood.
Although multiple aspects of the HSV-1 infectious cycle have been carefully investigated, the role viral and cellular factors play in nuclear events that coordinate viral DNA replication with recombination, repair, and late gene expression are not well understood. We recently adapted the iPOND (isolation of proteins on nascent DNA) method [25] to purify viral genomes from infected cells for the comprehensive analysis of viral genome associated proteins by mass spectrometry [26]. This study provided insight into cellular factors and processes that act on viral genomes at various times during infection. In the current study, we describe the purification and imaging of HSV-1 replication forks for the elucidation of proteins functioning on replicating genomes. Viral replication factors were identified at sites of active DNA synthesis, as well as groups of proteins that have known functions in specific cellular processes, including DNA repair and transcriptional regulation. Pulse chase and imaging studies reveal the temporal and spatial relationships of these proteins with viral DNA with respect to replication and provide insight into mechanisms that coordinate the repair and expression of viral genomes with genome replication. To our knowledge this is the most comprehensive analysis of cellular factors associated with viral replication forks to date.
We previously established methods based on iPOND to label HSV-1 DNA with clickable nucleotide analogs (EdU or EdC) for the covalent attachment of biotin or a fluorophore for purification or imaging of viral genomes [26]. We sought to adapt these methods to selectively label, purify, and image viral replication forks to investigate replication-coupled events. In order to examine the timing of protein interactions with viral replication forks, we first determined the rates of viral DNA replication throughout productive infection (Fig 1A). Infections, quantification of viral DNA, and calculation of replication rates were carried out as described in the Materials and Methods. The rates of HSV-1 DNA replication, calculated as the number of base pairs (bp) synthesized per minute (min), consistently fluctuate from 4–8, 8–12, and 12–24 hours post infection (hpi) (dashed vertical lines), with replication occurring more rapidly earlier during infection. The rates of DNA synthesis range from 1384–2184 bp/min from 4–8 hpi and 1020–1448 bp/min from 8–12 hpi. These rates may represent the use of multiple origins of replication on a single viral genome and therefore DNA synthesis by an individual polymerase is likely slower than the calculated rate.
To determine the resolution of pulse chase assays for the purification of viral replication fork associated proteins at 4 hpi, the number of base pairs labeled during the time intervals used in subsequent experiments were calculated (Fig 1B). These values do not take into account the amount of time it takes for EdC to be incorporated into cells and phosphorylated. Therefore, these values represent the maximum number of base pairs labeled and chased by a single polymerase. A 5–20 min pulse with EdC should provide accurate representation of viral replication forks and a 120 min pulse would result in labeling of entire viral genomes.
To identify factors and complexes enriched on HSV-1 replication forks, we pulse labeled viral DNA with EdC for 5 or 20 min, tagged and isolated labeled DNA, and identified associated proteins by mass spectrometry (Fig 1C). MRC5 cells were used in this study because they enter a quiescent state in which cellular DNA replication is arrested once grown to confluency ensuring that only viral DNA is labeled and purified in our assays (S1 Fig) [26]. Cells were infected with the virus UL2/UL50, which is defective for the viral uracil glycosylase (UL2) and dUTPase (UL50), resulting in greater incorporation of EdU and EdC into viral DNA [26]. For maximum protein yield, we adapted the accelerated native iPOND (aniPOND) method [27], which involves purification of DNA-protein complexes under native conditions and results in higher protein yield compared to the standard iPOND method. We improved this method by switching to magnetic streptavidin-coated beads for purifications, which resulted in reproducibly higher protein yield and decreased background binding.
Proteins enriched relative to the unlabeled control after a 5 min pulse with EdC were mapped using the STRING functional protein interaction database [28] (Fig 2A). A total of 52 human proteins with known functions in DNA repair, chromatin modification, transcription, and transcription-coupled processes were identified as associated with replication forks. Output from the STRING database indicated that there were more interactions among the identified proteins than would be expected for a random set of proteins. In addition, the viral replication proteins UL9, UL52, UL5, UL42, UL30, and ICP8 were enriched on pulse labeled DNA by at least 5 fold. The helicase-primase complex associated protein UL8 was not identified, perhaps because it is more transiently associated with viral DNA. Importantly, cellular DNA polymerases were not identified, consistent with the selective purification of viral and not cellular replication forks in our assays.
Proteins enriched after a 20 min pulse are shown in Fig 2B. A total of 90 human proteins enriched by at least 8 fold compared to the unlabeled negative control were mapped using STRING, revealing striking similarities between the 5 and 20 min pulse datasets. In fact, most human proteins enriched on replication forks pulse labeled for 5 min were found after 20 min. More members of the same complexes were identified in the 20 min pulse compared to the 5 min pulse. These data implicate specific repair processes at replication forks including MMR and single strand break (SSB) and DSB repair, as well as the chromatin remodeling complexes INO80, NuRD, and FACT. Core complexes involved in Pol II mediated transcription were also present including Mediator, Integrator, TFIID, and Pol II. These data provide evidence for the coupling of repair, chromatin remodeling, and transcription to viral DNA replication.
To compare the relative abundance of proteins on viral replication forks to whole viral genomes, infected cells were pulse labeled with EdC for 20 min to label replication forks, or for 120 min to label entire viral genomes (Fig 1B) followed by the isolation of DNA for the identification of associated proteins by mass spectrometry (C). Spectral count (SpC) values for identified proteins were normalized to the total abundance of proteins in that sample as described in the Materials and Methods to account for differences in the amount of total DNA isolated in each condition. Ratios of the normalized values (pulse 20 min/120 min) were compared for individual proteins, which were then graphed to highlight trends associated with factors that function in the same biological processes (Fig 3A, 3C, 3E and 3G and S2A Fig). Factors that were greatly enriched near replication forks include the HSV-1 replication helicase (UL5) and primase (UL52), the DSB repair protein RAD50, the MED12 and MED13 subunits of the Mediator complex, Integrator complex members, and the INO80 complex. Proteins that were more enriched on whole genomes after a 120 min pulse include factors involved in transcription-coupled RNA processing and virion assembly, as well as components of the cellular cytoskeleton and select histones.
To further investigate the dynamics of protein interactions with respect to viral replication forks, pulse chase analysis was carried out. Infected cells were pulse labeled with EdC for 20 min to label replication forks or pulsed for 20 min and then chased for 40 min with deoxyC (Fig 1B) followed by the isolation of DNA and identification of associated proteins (C). Ratios of the normalized SpC values (pulse/chase) were graphed for individual proteins (Fig 3B, 3D, 3F and 3H and S2B Fig). Factors that were greatly enriched in the pulse compared to chase include UL52, several DNA repair proteins, MED12 and MED13, the Integrator complex, the NuRD complex, and the viral portal protein UL6. Proteins that were more enriched in the chase include RNA processing factors and cytoskeletal proteins.
Western blot analysis was used to compare the interaction dynamics of the viral replication proteins ICP8 and UL42 with respect to replication forks (S2C Fig). ICP8 was enriched in the pulse and not in chases and relative levels of UL42 did not change significantly under these conditions. ICP8 binds to single stranded DNA near replication forks and UL42 binds to double stranded DNA nonspecifically. These data suggest that UL42 remains associated with DNA after it is deposited at sites of viral replication and validate trends observed for these two proteins by mass spectrometry (Fig 3B).
After the onset of viral DNA replication many nuclear processes in the viral life cycle occur in viral replication compartments [29]. To visualize viral replication forks relative to replication compartments, Vero cells were infected with HSV-1 strain KOS for four hours and actively replicating viral DNA was pulse labeled with EdC for 20 min followed by fixation, click chemistry to tag labeled DNA, and immunofluorescence for ICP8 (Fig 4, top panel). Distinct foci containing labeled viral DNA were visualized within viral replication compartments within the nuclei of infected cells. These foci colocalized with ICP8 (see red/green trace at right), and could be chased away from ICP8 containing foci (S2D Fig). These data demonstrate that distinct sites of viral DNA replication exist within viral replication compartments and provide a tool to visualize and validate the associations of proteins identified by mass spectrometry with sites of active viral DNA synthesis.
Cellular MMR proteins and PCNA were among the proteins identified to be enriched on viral replication forks after a 5 or 20 min pulse (Fig 2) and were relatively more abundant on pulse labeled compared to chased DNA (Fig 3B). We therefore predicted that these factors should colocalize with sites of active viral DNA synthesis within the nuclei of infected cells. We carried out imaging of viral replication forks as described above coupled with immunofluorescence of MMR proteins (MLH1, MSH2, and MSH6) and PCNA and demonstrated that these proteins do in fact colocalize with sites of active viral DNA synthesis (Fig 4). Compared to mock infected cells (S3 Fig), these factors undergo robust redistribution to viral replication forks within infected cells. In contrast, UL42 localizes to sites of active DNA synthesis, as well as to adjacent sites within replication compartments, consistent with observations from pulse chase mass spectrometry (Fig 3B) and western blot studies (S2C Fig). Taken together, imaging studies corroborate results from viral replication fork pull down assays.
To better understand the relationships of MMR proteins and PCNA with viral DNA, we investigated the recruitment of these factors to viral DNA early during infection before the onset of replication (S3 Fig and Table 1) and to viral genomes inhibited for replication by the addition of the viral DNA polymerase inhibitor acyclovir (S4 Fig and Table 1). In all cases, these factors depend on ongoing viral DNA synthesis to associate with HSV-1 DNA. It can therefore be concluded that PCNA and MMR proteins are recruited to and likely act at viral replication forks during productive HSV-1 infection.
Cellular Pol II and the Mediator complex were among proteins enriched on viral replication forks after a 5 or 20 min pulse (Fig 2) and transcription elongation factors SUPT5H (Spt5) and SUPT6H (Spt6) were abundant after a 20 min pulse. These factors were also relatively more abundant on pulse labeled DNA compared to chased DNA (Fig 3D), although the relative level of enrichment varies greatly between these proteins. We therefore predicted that these factors should colocalize with sites of active viral DNA synthesis. We carried out imaging of viral replication forks coupled with immunofluorescence and demonstrated that Pol II (POLR2A), MED23, MED1, SUPT5H, and SUPT6H do in fact colocalize with sites of active viral DNA synthesis, however, not exclusively (Fig 4). POLR2A, MED23, and MED1 also associate with adjacent sites within replication compartments that are not in the act of viral DNA synthesis and SUPT5H and SUPT6H appear to localize adjacent to sites of active DNA synthesis. POLR2A and MED23 have similar distributions within replication compartments as ICP4, consistent with more general roles in the regulation of replication-dependent transcription of late viral genes, as well as replication-independent transcription of IE and early viral genes. Compared to mock infected cells (S3 Fig), these factors were redistributed to viral replication compartments during infection, consistent with the overall reduction in Pol II occupancy on cellular promoters [30]. We observed recruitment of transcription factors to viral genomes early during infection before the onset of replication (S3 Fig and Table 1) and to viral genomes inhibited by acyclovir (S4 Fig and Table 1). Taken together, these data are consistent with the involvement of cellular transcription factors in replication-dependent and replication-independent transcription of viral genes.
To investigate the coupling of viral transcription to replication, we carried out RNA-Seq to quantify the dependence of viral gene expression on DNA replication. Acyclovir was added to cultures of KOS infected Vero cells at 0, 2, 3, 4, or 6 hpi and DNA or RNA was harvested at 12 hpi to compare the relative abundance of viral genomes and transcripts as a function of time of inhibition of viral DNA synthesis. The control was harvested at 12 hpi without the addition of acyclovir. Acyclovir inhibits viral DNA replication by ~27 fold when added at 4 hpi compared to the control (Fig 5A), but only inhibits transcription of the true late gene UL44 (glycoprotein C, gC) by 1.5 fold when added at this time (B). The addition of acyclovir at 4 hpi also resulted in a 2.5 fold increase in expression of the early gene UL23 (thymidine kinase, tk) compared to the control (C), whose expression typically peaks between 3 and 4 hpi [31]. Additional late genes follow the same kinetics for activation as gC (D, E, shaded) and the early gene UL29 (ICP8) follows the same kinetics as UL23 (E, dashed box). Taken together, these data indicate that initial rounds of viral DNA replication are sufficient to alter the transcriptional landscape of HSV-1 genomes enough to activate late gene and alter early gene expression.
The processes of genome replication, transcription, repair, and maintenance are often coupled in various ways in cells to allow for growth, differentiation, and development. The coupling of these processes is also fundamental to the life cycle of complex DNA viruses such as HSV-1. The expression of approximately 80 viral genes is coordinated with the entry of the viral genome into the nucleus and replication of viral DNA. In addition, the genome undergoes prolific recombination, however it remains relatively stable from a genetic standpoint. These processes, along with genome maturation and packaging, are coordinated by the interactions of viral and cellular factors with the viral genome such that the first progeny virus is assembled within the first 4–6 hours post infection. A pivotal coordinating event is viral DNA replication.
Here we developed methods to selectively purify and image HSV-1 replication forks for the temporal and spatial analysis of protein interactions with sites of viral DNA synthesis to ascertain how these fundamental processes are coupled. We demonstrate that select cellular factors copurify with nascent viral DNA, which include factors that function in SSB and DSB repair, MMR, transcription, and chromatin remodeling (Fig 2). We further demonstrate that identified factors colocalize with distinct sites of ongoing DNA synthesis within replication compartments (Fig 4), confirming the coupling of repair and transcription to viral DNA replication. We test for the dependence of these interactions on initial rounds of viral DNA synthesis and demonstrate that recruitment of PCNA and MMR proteins to viral DNA is strictly replication-dependent. On the other hand, Pol II and Mediator interactions are both replication-coupled and independent (S3 Fig and S4 Fig), consistent with replication-dependent and independent modes of gene expression. Furthermore, we track nascent viral DNA out of replication compartments (S2D Fig), where the newly replicated DNA makes more contact with factors involved in transport and packaging of viral genomes (Fig 3G and 3H) or transcription-coupled RNA processing (E, F). The ability to recognize and track distinct populations of viral DNA with respect to DNA synthesis reveals the existence of functional subdomains within HSV-1 replication compartments (Fig 6A).
We demonstrate that at 4 hpi viral replication proteins are abundantly associated with sites of viral DNA synthesis (Fig 2). After longer pulses with EdC (Fig 3A) or in pulse chase experiments (B), the levels of the origin binding protein UL9 do not change significantly, implicating that UL9 binds to origins during initial rounds of replication and that the relative levels on individual genomes remains constant from 4–6 hpi. Pulse chase studies reveal close associations between UL5/UL52 and sites of active viral DNA synthesis (Fig 3A and 3B). Levels decrease significantly in chases, consistent with close associations with replication forks and indicating that these proteins are not deposited onto newly replicated DNA. In contrast, UL42 remains associated in chases (Fig 3B and S2C Fig) and does not exclusively colocalize with replication foci within replication compartments (Fig 4). It is possible that UL42 remains associated with nascent viral DNA through sequence nonspecific interactions with double-stranded DNA. This is in stark contrast to the pulse chase and binding properties of ICP8 (S2C Fig and Fig 3B), which selectively associates with single stranded DNA at replication forks (Fig 4).
Viral transcription factors ICP4 and ICP22 also associate with sites of viral DNA synthesis (Fig 2), consistent with functions in replication-dependent late gene expression. Levels of these factors remain unchanged in pulse experiments (Fig 3C) implicating ongoing functions in viral gene expression from 4–6 hpi. Furthermore, levels remain unchanged in pulse chase experiments (D), consistent with these proteins being deposited on and remaining associated with newly replicated DNA. ICP4 associates with sites of active viral DNA synthesis, as well as with adjacent sites within replication compartments (Fig 4), consistent with the ability to associate with double stranded DNA in a sequence nonspecific manner [32], although site specific binding across the genome also occurs [33,34].
An interesting observation from these studies is that the viral portal protein UL6 is abundant on viral replication forks (Fig 2). In fact, it is one of the most abundant proteins associated with pulse labeled viral DNA. This protein is also more closely associated with sites of active viral DNA synthesis because the relative levels of UL6 decrease as you move away from replication forks (Fig 3H). This is in contrast to other viral DNA processing and packaging factors, which are only enriched on viral genomes after longer pulses with EdC (G). It is possible that UL6 plays a yet to be identified role in viral DNA replication in addition to its function as the portal for entry into the HSV-1 capsid.
Among proteins enriched on viral replication forks are factors that function in the repair of damaged cellular DNA (Fig 2). These include cellular proteins with known roles in MMR, SSB repair, and DSB repair. In general these proteins may function in the repair of damaged DNA during replication or may mediate DNA recombination.
The MMR proteins, MSH2, MSH3, and MSH6 are among the proteins most abundantly enriched on viral replication forks. After a 20 min pulse, MSH2 and MSH6 are enriched on EdC labeled DNA by ~100 fold and MSH3 by 30 fold compared to the unlabeled negative control (Fig 2). MLH1 was less abundant but still enriched by ~9 fold. The MSH2-MSH6 heterodimer accounts for 80–90% of cellular MSH2 and preferentially recognizes base pair mismatches and small insertion/deletion mutations [35]. These proteins have previously been shown to copurify with viral genomes [26] and with ICP8 from virus infected cells [36]. MSH2, MSH6, and MLH1 have also been shown to colocalize with viral replication compartments [36,37] and knockdown of MSH2, MSH6, and MLH1 results in reduced viral yields and defects associated with replication compartment formation [37]. Here we demonstrate that these proteins colocalize with sites of active viral DNA synthesis within replication compartments (Fig 4), depend on DNA replication for colocalization with viral DNA (S3 Fig and S4 Fig), and are more closely associated with pulse labeled replication forks rather than chased DNA (Fig 3B). Our studies place the cellular MMR machinery at or near viral replication forks and demonstrates an intimate relationship between viral DNA replication and MMR. Because UL30 incorporates incorrect nucleotides with a frequency of 1 in 300 [16], which would result in >500 mismatches per genome, HSV-1 may have adapted cellular MMR to prevent the maintenance of mutations generated during DNA replication.
Cellular MMR has been closely linked to DNA replication [38–40] and the cellular sliding clamp PCNA recruits MMR proteins to replicating cellular DNA [41]. Consistent with the potential involvement of PCNA in MMR of viral DNA, we observe high levels of PCNA and the clamp loader complex (RFC1-5) at replication forks (Fig 2) and observe similar pulse chase kinetics (Fig 3B) and localization with sites of DNA synthesis (Fig 4) for PCNA as compared to MMR proteins. Furthermore, PCNA association with viral DNA depends on active viral DNA synthesis (S3 Fig and S4 Fig) and PCNA knockdown results in reduced virus yield [42]. Taken together, these data strongly implicate cellular PCNA and MMR in the maintenance of viral genome integrity during viral DNA replication (Fig 6B).
The HSV-1 genome contains several nicks and gaps [17–19], which likely cause DSBs during DNA replication. Another source of DSBs on viral DNA is through the actions of type II topoisomerases, which cause transient DSBs to relax supercoiling that occurs during transcription and replication or to allow for the decatenation of linked DNA molecules [43]. In support of this, type II topoisomerases TOP2A and TOP2B are associated with viral replication forks (Fig 2). DSBs are required for HSV-1 genome isomerization [44] but would likely impede transcription and replication of viral DNA.
Here we identified MRE11 and RAD50 components of the MRN (MRE11-RAD50-NBS1) complex at viral replication forks (Fig 2). Furthermore, XRCC5 and 6 were identified just below the enrichment threshold (S1 Table). These proteins play a role in processing DSBs before repair. Although HSV-1 inhibits cellular DSB repair by nonhomologous end joining and homologous recombination [1], alternative modes of recombination have been suggested through the actions of the viral alkaline nuclease (UL12) and ICP8 [45]. UL12 was not identified to be associated with viral replication forks in our assays but has previously been shown to coprecipitate with ICP8 [36]. MRN complex members colocalize with ICP8 in virus infected cells throughout infection and infections in MRE11 defective cells result in reduced viral yields [46,47]. Therefore, this complex likely plays a role in repair or stabilization of DSBs on viral DNA during replication or may act upstream of virus specific recombination events (Fig 6B).
SSBs can form on viral DNA during lagging strand synthesis or by the actions of topoisomerase I (TOP1), which was enriched on viral replication forks (Fig 2). TOP1 cleaves a single strand of DNA to create a transient break to allow for topological changes during replication or transcription. During the repair of TOP1 associated SSBs, PARP1 and RECQL inhibit replication restart, DNA ends are repaired by PNKP and the nick is ligated through the actions of XRCC1 and LIG3 [43]. In this study, all of these factors were enriched on viral replication forks (Fig 2). Furthermore, pulse chase kinetics suggest that some of these factors are most closely associated with sites of active DNA synthesis (Fig 3B). PARP1, XRCC1, and LIG3 are also involved in the repair of nicks left behind after base excision repair, which is inhibited in our assays. It is therefore likely that SSB repair of TOP1 intermediates occurs near viral replication forks (Fig 6B). Some of these proteins may also play a role in the repair nicks left behind during lagging strand synthesis of Okazaki fragments.
INO80 (INO80, RUVBL2, RUVBL1), NuRD (HDAC1, HDAC2, RBBP4, CHD3, CHD4, MTA1, MTA2), and FACT (SSRP1, SUPT16H) chromatin remodeling complex members were selectively enriched on viral replication forks (Fig 2 and S2A and S2B Fig). The NuRD and INO80 complexes catalyzes ATP-dependent chromatin remodeling and play roles in cellular transcription repression and activation, DNA replication, recombination, and repair [48–50]. The FACT complex remodels nucleosomes and tethers histones to prevent their removal during transcription elongation [51]. Although histones in the form of nucleosomes do not appear to be associated with replicated viral DNA, histone H1 and H2A variants are associated with newly replicated viral DNA (S2A and S2B Fig). The functions of chromatin remodeling complexes and histones in the regulation of processes that occur on viral DNA, including replication-coupled events, is of great interest for future studies.
Many DNA viruses undergo replication-coupled late gene expression, although the mechanisms that either repress transcription before or activate transcription after replication are not well understood. Interestingly, we identified cellular transcription factors associated with viral replication forks including Mediator, TFIID, Pol II, ICP4, Integrator, SUPT5H, and SUPT6H (Fig 2). These factors have known functions in transcription initiation and elongation of cellular genes. The promoters of late genes are relatively simple and contain only initiator and TATA containing elements [52,53] and lack sites for Sp1, which are found within promoters of IE and early genes. We did not identify Sp1 in our assays, suggesting that proteins associated with late and not IE or early gene promoters are enriched in our assays. Late gene promoters are bound by ICP4, TBP, and Pol II prior to the onset of DNA replication [54], suggesting a model by which preinitiation complexes (PICs) begin to form before DNA replication. It is therefore likely that genome replication enhances recruitment and some concurrent event promotes promoter conformations that are conducive to initiation and promoter escape (Fig 6C).
The dynamics of transcriptional events with respect to replication forks suggests the involvement of the Mediator complex in the regulation of late gene transcription. In this study, we reveal a strong association of Mediator with viral replication forks (Fig 2). Pulse chase studies reveal a preference for binding at replication forks, with Mediator levels decreasing in chases (Fig 3D). This observation is especially robust for the MED12 and MED13 subunits of the complex. MED12 and MED13 are components of the kinase domain, which sterically blocks interactions with between Mediator and the C-terminal domain (CTD) of Pol II to repress initiation and reinitiation events [55]. Mediator compositional changes have recently been shown to occur at the promoters of cellular genes, which potentially coordinate sequential events in transcription activation [56]. Mediator containing the kinase module first associates with activators bound to enhancers, upon PIC formation the kinase domain leaves allowing Mediator to associate with the CTD of Pol II, which is followed by CTD phosphorylation and promoter escape. Taken together, we propose a testable model whereby Mediator containing MED12 and MED13 associates near the promoters of late viral genes during DNA replication and that conformational changes that occur within Mediator promote transcriptional activation (Fig 6C).
It is likely that ICP4 is involved in the recruitment of Mediator to replication forks. ICP4 interacts with Mediator and is required for the recruitment of Mediator to the promoters of all classes of viral genes [57–59]. Furthermore, the form of Mediator that associates with ICP4 contains MED12 and MED13 and lacks MED26, which is consistent with the form we found associated with viral replication forks.
Mediator was more enriched on genomes pulse labeled for 20 min at 4 hpi as compared to genomes labeled from 4–6 hpi (Fig 3C), suggesting that initial rounds of replication are sufficient to drive conformation changes within the promoters of late viral genes. To test this hypothesis, we carried out RNA-seq to assay for late viral gene expression as a function of time of inhibition of viral DNA replication (Fig 5). Our data indicate that inhibition of replication after initial rounds (at 3 or 4 hpi) results in a significant reduction of the number of viral genomes present at 12 hpi (A), but has little effect on the number of late gene transcripts that are expressed at that time (B,D,E). These data indicate that initial rounds of viral DNA replication are sufficient to drive robust late gene expression, further supporting a role of replication in promoting conformational changes within promoters of late viral genes.
Taken together, our results provide novel insight into the coupling of viral DNA replication with several fundamental processes that occur on the HSV-1 genome during productive infection. Imaging studies reveal functional subdomains within viral replication compartments where DNA synthesis and factors that participate in replication-coupled events are concentrated (Fig 6A). Results also support models whereby dynamic associations of cellular factors with viral replication forks coordinate viral DNA replication with the repair and maintenance of viral DNA (B) and the expression of late viral genes (C).
Experiments were performed using MRC5 (human fetal lung) or Vero (African green monkey kidney) cells obtained from and propagated as recommended by ATCC. The viruses used in this study include the HSV-1 wild type strain KOS and UL2/UL50 [26]. EdC-prelabeled virus was prepared as described previously [26].
To measure DNA quantity for the calculation of viral replication rates (Fig 1A), 5x105 Vero or MRC5 cells were infected with KOS or UL2/UL50 at a multiplicity of infection (MOI) of 10 plaque forming units (PFU)/cell. DNA was collected at 0, 2, 4, 6, 8, 10, 12, and 24 hpi by aspirating growth medium and collecting infected cells in 0.2 ml DNA extraction buffer (0.5% SDS, 400 μg/ml proteinase K, 100 mM NaCl). Samples were incubated at 37°C for 4 hours followed by heat inactivation at 65°C for 15 min and phenol:chloroform extraction. DNA was diluted 1:500 in water and qPCR was carried out to determine the number of viral genomes using primers specific for the HSV-1 tk gene [31]. Standard curves were generated using purified KOS DNA. The rates of viral DNA replication were calculated as bp synthesized per min using the following formula: 152,000bp/genomedurationlog2(finalnumberofgenomes)−log2(initialnumbergenomes)
To measure genome quantity for comparison with RNA-Seq data (Fig 5A), 2x106 Vero cells were infected with KOS at an MOI of 10 PFU/cell and incubated at 37°C for 12 hours in the presence or absence of 100 μM acyclovir. DNA extraction and qPCR were carried out as above using primers specific for the HSV-1 gC gene [31].
Viral DNA isolation was carried out using the aniPOND technique as described previously [27] with the following modifications. A confluent monolayer of ~7x107 MRC5 cells was infected with the UL2/UL50 virus at an MOI of 10 PFU/cell for one hour at room temperature. After adsorption, the inoculum was removed and cells were rinsed with room temperature tris-buffered saline (TBS) before growth medium was replaced. Cells were incubated at 37°C for four hours before adding 25 μM EdC (Sigma-Aldrich) for 5, 20, or 120 min. Chases were carried out after a 20 min pulse by quickly rinsing cells three times with chase medium containing 150 μM 2´deoxycytidine (deoxyC) followed by incubation in chase medium for an additional 40 min. Negative control samples were not labeled with EdC. Harvesting nuclei, cell washes, and biotin conjugation by click chemistry were carried out as described. For cell lysis and DNA fragmentation, resuspended cell pellets were incubated for 45 min in lysis buffer and sonicated 6 times for 30 sec each at 40% amplitude using a Sonics Vibra Cell Ultra Sonic Processer equipped with a 3 mm microtip probe. DNA fragments ranged in size from 100–500 bp. Dynabeads MyOne Streptavidin T1 (ThermoFisher) were used to purify biotinylated DNA-protein complexes. Proteins were eluted from streptavidin-coated beads by boiling in 2x SDS Laemmli sample buffer. DNA isolation from input and bound samples was carried out as described [26]. Experiments yielded ~6 μg total DNA and 100–200 ng bead bound DNA.
Mass spectrometry was carried out by MSBioworks as described [26]. Proteins were considered enriched by aniPOND based on the following criteria: 1) protein had at least 5 SpCs in one experimental condition per data set, 2) protein was not detected in the negative control or was enriched over the negative control by at least four-fold based on dividing SpC values, and 3) protein was detected in duplicate experiments. The normalized spectral abundance factor (NSAF: SpC/MWΣ(SpC/MW)) was calculated for each sample to account for differences in protein size and total protein yield. Average NSAF from two independent data sets were used for all analyses. Raw and normalized SpC data for enriched proteins are provided in S1 Table.
Click chemistry and immunofluorescence experiments were carried out after infection of Vero cells with KOS or EdC-labeled KOS as described [26]. To pulse label viral replication forks, 25 μM EdC was added to KOS infected cells at 4 hpi for 20 min before fixation with 3.7% paraformaldehyde. The Click-iT Alexa Fluor 488 Imaging Kit (ThermoFisher) was used to tag EdC labeled viral genomes and immunofluorescence was carried out using the following primary antibodies and dilutions: mouse anti-ICP8: ab20194 (Abcam), 1:200; mouse anti-PCNA: sc-056 (Santa Cruz), 1:200; mouse anti-MLH1: 550838 (BD Biosciences), 1:200; mouse anti-MSH2: Ab-2 NA27 (Calbiochem), 1:100; rabbit anti-MSH6: A300-023A (Bethyl Laboratories), 1:100; mouse anti-UL42: 2H4 ab19311 (Abcam), 1:200; mouse anti-ICP4: 58S, 1:500; mouse anti-Pol II (POLR2A): 4H8 (Abcam), 1:500; mouse anti-Sur2 (MED23): 550429 (BD Biosciences), 1:500; rabbit anti-Trap220 (MED1): sc-8998 (Santa Cruz), 1:250; rabbit anti-Spt5 (SUPT5H): A300-869A (Bethyl Laboratories), 1:200; and rabbit anti-Spt6 (SUPT6H): ab32820 (Abcam), 1:200. Goat anti-mouse and anti-rabbit alexa fluor 594-conjugated secondary antibodies (Santa Cruz) were diluted 1:500.
Monolayers of 2x106 Vero cells were infected with KOS at an MOI 10 PFU/cell and incubated at 37°C for 12 hours. Acyclovir (100 μM) was added at the indicated times post infection. RNA was isolated using the Ambion RNAqueous-4PCR Kit and quantified using the Agilent RNA 6000 Nano Kit. RNA-Seq was carried out as described [31] and sent to the Tufts University Core Facility for sequencing. Illumina reads were processed and aligned to the KOS genome using CLC Workbench V8. To account for variations in sample reads, each sample was normalized to the total number of reads.
|
10.1371/journal.ppat.1006402 | Seminal plasma induces inflammation and enhances HIV-1 replication in human cervical tissue explants | The most immediate and evident effect of mucosal exposure to semen in vivo is a local release of proinflammatory mediators accompanied by an influx of leukocytes into the female genital mucosa (FGM). The implication of such response in HIV-1 transmission has never been addressed due to limitations of currently available experimental models. Using human tissue explants from the uterine cervix, we developed a system of mucosal exposure to seminal plasma (SP) that supports HIV-1 replication. Treatment of ectocervical explants with SP resulted in the upregulation of inflammatory and growth factors, including IL-6, TNF, CCL5, CCL20, CXCL1, and CXCL8, and IL1A, CSF2, IL7, PTGS2, as evaluated by measuring protein levels in explant conditioned medium (ECM) and gene expression in tissue. SP treatment was also associated with increased recruitment of monocytes and neutrophils, as observed upon incubation of peripheral blood leukocytes with ECM in a transwell system. To evaluate the impact of the SP-mediated response on local susceptibility to HIV-1, we infected ectocervical explants with the CCR5-tropic variant HIV-1BaL either in the presence of SP, or after explant pre-incubation with SP. In both experimental settings SP enhanced virus replication as evaluated by HIV-1 p24gag released in explant culture medium over time, as well as by HIV-1 DNA quantification in explants infected in the presence of SP. These results suggest that a sustained inflammatory response elicited by SP soon after coitus may promote HIV-1 transmission to the FGM. Nevertheless, ectocervical tissue explants did not support the replication of transmitted/founder HIV-1 molecular clones, regardless of SP treatment. Our system offers experimental and analytical advantages over traditional models of HIV-1 transmission for the study of SP immunoregulatory effect on the FGM, and may provide a useful platform to ultimately identify new determinants of HIV-1 infection at this site.
| The majority of HIV-1 transmissions worldwide occur via unprotected vaginal intercourse, with semen and the female genital mucosa (FGM) being the main vector and recipient of virus acquisition in women. It is well known that mucosal exposure to semen via vaginal intercourse or artificial fertilization results in transient inflammation of the FGM. This response is thought to be important to defend the host against eventual pathogens present in semen and clear the mucosa from residual sperm cells, as well as to promote fertilization. However its implication in HIV-1 transmission has never been addressed. Using human tissue explants from the uterine cervix, we developed a model of mucosal exposure to seminal plasma (SP) to study HIV-1 transmission to the FGM. Our model recapitulated the main features of the in vivo response to semen, namely upregulation of proinflammatory factors and leukocyte recruitment. SP treatment of explants enhanced HIV-1 replication, suggesting that a sustained inflammatory response elicited by SP soon after coitus may promote virus infection. Our experimental system may contribute to improve the efficacy of currently available prevention measures by understanding the mechanisms regulating immunity in the FGM, and leading to the identification of new determinants of HIV-1 transmission therein.
| Mucosal exposure to infected semen accounts for the majority of human immunodeficiency virus (HIV) 1 transmission events worldwide, with vaginal intercourse being the most common acquisition route in women [1]. In spite of this evidence, a poor understanding of the mechanisms regulating immunity at mucosal sites, exacerbated by discrepancies in the biology of infection between humans and experimental models, has hampered the development of effective prevention measures against HIV-1 [2–4].
We among others believe that semen plays an active role in modulating the efficiency of HIV-1 transmission to the female genital mucosa (FGM) [5]. Semen is a complex mixture of cells and molecules with immunoregulatory function [6]. An inflammatory response characterized by leukocyte infiltration and upregulation of proinflammatory and growth factors occurs in the FGM soon after exposure to semen, in a highly conserved fashion among mammals [7]. This response, also called leukocytic reaction, was originally described in humans by analyzing cervical smears collected after artificial insemination, showing a significant amount of neutrophils that egressed the mucosa as early as 20 min after treatment [8,9]. These events have been proposed to contribute to fertilization, nevertheless their impact on the FGM susceptibility to sexually transmitted pathogens, including HIV-1, remains elusive.
In the last few years, a growing number of reports investigated the role of semen in HIV-1 infection leading to controversial results [5,10]. Most of these findings rely on isolated cells incubated with highly diluted seminal plasma (SP) and artificial infection settings involving exogenously activated or reporter cells, which may not adequately mimic the biology of mucosal transmission. The few vaginal infection studies conducted to date in non-human primates employed either purified seminal factors or human SP, without assessing the presence of an underlying immune response induced by SP treatment [11,12].
To develop a new, relevant model of HIV-1 sexual transmission we used human mucosal tissue from the lower female reproductive tract (FRT) and devised an experimental system that could reproduce the local immunologic events associated with exposure to semen as observed in vivo. We here demonstrate that treatment of ectocervical tissue explants with SP from HIV-seronegative men recapitulated the hallmarks of the early mucosal response to semen, including proinflammatory and growth factor upregulation, as well as leukocyte recruitment as evaluated in a transwell system. This inflammatory-like response resulted in enhanced HIV-1 replication, as shown in explants infected with the laboratory-adapted CCR5-tropic HIV-1 variant HIV-1BaL both after exposure to SP and in the presence of SP. However, our system did not support the replication of transmitted/founder HIV-1 molecular clones, regardless of SP treatment. Exposure to SP did not affect cell viability as evaluated on individual cells upon isolation from tissue explants, that were mostly viable after a 24 h-culture in agarose. Tissue architecture was partially preserved after 18 days from infection, allowing for the detection and visualization of cells harboring HIV-1 within explants at the end of culture. These aspects highlight the relevance and advantages offered by tissue explants over other in vitro models of mucosal response to semen and HIV-1 transmission. In particular, our experimental system can be implemented to gain a better understanding of the cellular and molecular features of SP early immunoregulatory effect on the FGM. Such knowledge may contribute to improve the efficiency of currently available prevention strategies by enabling the targeting of other determinants of transmission in addition to HIV-1.
Specimens from 54 uterine tissue donors were included. Median donor age (range) was 45 (35–54) years. All women were pre-menopausal. Fourteen women were using progesterone-based medications, namely levonorgestrel intrauterine device (IUD) or the selective progesterone receptor modulator ulipristal acetate or both, at the time of surgery: specimens from one donor were used for cytokine analysis, three for cell viability assay, and ten for HIV-1 infection experiments, of which six were infected with HIV-1BaL. Seventeen donors provided information on sexual habits: all donors reported one sexual partner in the past year, and two reported regular condom use.
Local exposure to semen is associated with an influx of leukocytes into the FGM and rapid neutrophil egress from the mucosa into the vaginal lumen [9,21]. To investigate these events in vitro, we exposed peripheral blood leukocytes (PBL) from healthy donors to ECM for 2 h in order to assess its chemotactic power using a transwell assay. Input cells were obtained by reconstitution of the mononuclear cell and granulocyte fractions of blood. Neutrophils (CD16+CD66b+) were the predominant population among input cells, followed by T cells (CD16-CD3+) and monocytes (CD16-CD14+), reflecting the natural composition of circulating leukocytes (CD45+) (S5 Fig). Compared to ECM of untreated control explants (CM), incubation with ECM from SP-treated explants resulted in a median 2.0-fold increase in the number of transmigrated total leukocytes, and in particular a 2.4- and 4.4-fold increase in neutrophil and monocyte transmigration respectively (n = 6, p = 0.03) (Fig 3A & 3B). The relative contribution of the cellular and plasma fractions of semen to mucosal leukocyte recruitment in humans is not known. However, studies in animal models point at soluble factors in semen such as TGF-β1, prostaglandin-E, and TLR4 ligands as potential inflammatory mediators [7,20,23]. Although we did not investigate the role of those molecules, except for preliminary data on explant cytokine response to LPS (S3 Fig), we asked whether SP-mediated chemokine upregulation measured in ECM could explain the observed PBL migration. To this end, we measured the expression levels of the chemokine receptors CCR5, binding CCL5 among other β-chemokines, and CXCR1 and CXCR2, binding CXCL1 and CXCL8, on transmigrated cells as indirect evidence of cognate chemokine binding to the receptor. The mean fluorescence intensity (MFI) of CCR5 on transmigrated monocytes significantly decreased (median) following incubation with ECM from both CM- (0.26-fold) and SP-treated (0.27-fold) explants compared to untreated cultured PBL (n = 6, p = 0.035) (Fig 3C). There was no difference between n-fold change values of CCR5 MFI of the two treatments (p>0.999). These data suggest that increased local production of CCL5, as measured in SP-ECM, may account for semen-induced recruitment of monocytes to the FGM.
The expression of CXCR1 and CXCR2 on transmigrated granulocytes, identified among CD45+ cells based on their light-scattering properties (S5 Fig) did not decrease as we expected to. On the contrary, CXCR2 MFI on cells incubated with ECM from both CM- and SP-treated explants showed an increase compared to untreated cultured control cells, whereas CXCR1 levels were unchanged (S6 Fig). Of note, our analysis of transmigrated PBL was restricted to a 2 h-incubation with ECM, although the expression of chemokine receptors may vary in time, also under regulation of other factors than receptor ligands. Therefore, chemokine-neutralization experiments are required to substantiate our observations.
A primary limitation of the experimental study of semen is its toxicity to isolated cells incubated with low concentrations of whole semen or SP (e.g., 10%) for 3 to 6 h [13,24,25]. We wanted to verify whether SP treatment affected cell viability in our system. Loss of membrane integrity was characterized by staining cells isolated from explants with an amine-reactive dye (Live/dead, L/d), as indicator of necrotic or late-apoptotic cells, and annexin V (AV), which binds to phosphatidylserine exposed on the outer plasma membrane early during apoptosis. Compared to donor-matched explants incubated with CM, a 12 h-treatment with SP50% did not affect the fraction of cells positive to either one or both cell death markers. This was demonstrated in both immune and non-immune cells, as defined by CD45 expression (Fig 4).
As expected, treatment with the proapoptotic drug camptothecin significantly increased the fraction of double positive (AV+ L/d+) and live/dead-single positive (AV- L/d+) populations, among both CD45+ and CD45- cells, with a significant decrease in the fraction of the double negative (AV- L/d-) population among CD45- cells (p<0.05) (Fig 4D). Also considering that the cell isolation procedure likely contributed to cell death, our analysis showed that the majority of cells retained within explants after a 24 h-culture in agarose were viable, with fractions (mean ± s.e.m., n = 6) of cells negative for both cell death markers in CM- vs. SP-treated explants of 79.8 ± 3.8% vs. 80.7 ± 4.6% of CD45+ cells, and 92.7 ± 1.8% and 92.8 ± 1.3% of CD45- cells (Fig 4C).
In situ detection of chromatin fragmentation revealed a few apoptotic cells that were located in the upper layers of the epithelium of both explants incubated with SP50% and CM for 12 h, and to a lesser extent in donor-matched uncultured explants (S7 Fig). Also, shedding of the upper epithelial layers was visible in cultured explants for both treatments. Although we could not establish a measurable indicator of toxicity using this method, the presence of apoptotic cells and epithelial disruption appeared to vary with tissue donors and manipulation of individual explants rather than being associated with SP treatment.
It has been speculated that the inflammatory nature of the response elicited by semen can affect the FGM susceptibility to HIV-1 [10]. To test this hypothesis, we infected ectocervical tissue explants with HIV-1 either following exposure to SP or in the presence of SP, as a model of HIV-1 transmission to the FGM (S1B Fig). Productive infection with the CCR5-tropic laboratory-adapted virus HIV-1BaL was achieved for explants from all included donors (n = 17) regardless of the infection set-up, as evaluated by measuring the amount of HIV-1 core protein p24gag released into culture supernatant over time (Fig 5A). Exposure of explants to SP25% for 4 h before infection (post-SP) resulted in a median cumulative p24gag production of 2.5 ng/ml compared to 1.5 ng/ml for untreated explants (CM) (Fig 5B). Compared to donor-matched untreated explants, SP treatment resulted in a median 1.6-fold increase in cumulative p24gag production (n = 9, p = 0.003) (Fig 5C). Similarly, infection of tissue explants with a mixture of HIV-1BaL and SP25% for 12 h (SP-mix) resulted in a median cumulative p24gag production of 12.9 ng/ml compared to 7.7 ng/ml for untreated explants (Fig 5B), that was equal to a median 1.5-fold increase as measured in donor-matched tissue explants (n = 8, p = 0.007) (Fig 5C).
To confirm the enhancing effect of SP treatment on HIV-1 replication we measured the amount of HIV-1 DNA in explants infected in the presence of SP (SP-mix) harvested at the end of culture (day 18). In agreement with the results on HIV-1 p24gag levels, SP-treated explants showed a median number of HIV-1 DNA copies of 1319 compared to 1024 for untreated explants (CM) (Fig 5D), resulting in a median 1.4-fold increase as measured in donor-matched tissue explants (n = 8, p = 0.023) (Fig 5E). The data on HIV-1 p24gag levels were consistent with HIV-1 DNA detection also for HIV-infected explants treated with the anti-retroviral drug lamivudine (3TC) throughout culture as a negative control (n = 3), showing a constant decline in HIV-1 replication kinetics associated with sporadic detection of HIV-1 DNA below 25 copies in explants harvested at the end of culture (day 18) (Fig 5A & 5D).
Of note, the use of progesterone-based drugs (n = 6), namely a levonorgestrel-containing IUD (n = 3) and per oral ulipristal acetate (n = 3), was associated with a lower SP-mediated increase in HIV-1 cumulative p24gag production compared to explants from donors who did not use any hormonal drugs at the time of surgery (n = 11) (1.4 vs. 1.8 median n-fold, p = 0.056) (S8 Fig). Nevertheless, SP treatment significantly enhanced cumulative HIV-1 replication compared to donor-matched untreated explants in both groups (p<0.05). The effect of sex hormones on the local defenses of the FRT has been object of a number of investigations over the years, in particular with respect to the use of the injectable progestin medroxyprogesterone acetate in HIV endemic areas [26]. Although in vivo and in vitro experimental data point at a protective role of estrogen as opposed to progesterone [27,28], the mechanisms underlying hormonal regulation of HIV-1 transmission remain to be elucidated in humans. In addition, levonorgestrel and ulipristal acetate differ not only in the route of administration, but also in their affinity to the progesterone and glucocorticoid receptors among others. Therefore, the analysis of their effect on the local immune response in the lower FRT and on HIV-1 transmission therein requires a proper validation and study design.
Histologic analysis of HIV-infected ectocervical tissue explants harvested at the end of culture revealed the presence of HIV RNA in the stroma (S9 Fig), supporting the evidence that productive virus replication occurs within explants until the final days of culture, as already suggested by HIV-1 DNA results. In agreement with previous reports [29], the pluristratified epithelium characteristic of the lower FGM was not retained on ectocervical explants, as indicated by shedding of the upper epithelial layers already after 12 h of culture in agarose medium (S7 Fig). Nevertheless, the overall structure of the stroma, along with resident cells, was preserved upon infection. Morphology retention, although partial, accounts for a critical advantage offered by tissue explants over traditional monotypic and two-dimensional culture systems, which allows characterizing the early events regulating the founder pool of infected cells within their natural environment.
To further validate our findings using viruses representative of mucosally acquired variants, we tested full length infectious molecular clones (IMCs) of viruses derived from circulating virions or infected cells of patients with acute HIV-1 clade B infection, also known as transmitted/founder (T/F) HIV-1 [30,31]. In particular, we selected the viruses pCH077.t/2627 (male host), pRHPA.c/2635 (female), and pTHRO.c/2626 (male) based on the limited data available on ectocervical tissue explant infection with T/F HIV-1 [32,33]. As we expected T/F HIV-1 to replicate at lower levels than HIV-1BaL, we decided to infect explants after SP treatment (post-SP) to avoid diluting the virus inoculum and maximize the time of infection (18 h). Nevertheless, we did not obtain a productive infection with any of the three selected T/F HIV-1, regardless of SP treatment, as showed by p24gag levels in explant culture supernatant measured over time (Fig 6A–6C). Of note pRHPA.c/2635, the only virus isolated from a female host among those tested, exhibited a distinctive replication kinetics reaching a steady p24gag production of about 500pg/ml during the last six days of culture (Fig 6B). The other two tested T/F HIV-1 generally showed lower p24gag levels and undetectable levels at the end of culture in 1 out of 3, and 2 out 4 experiments for pCH077.t/2627 and pTHRO.c/2626 respectively (Fig 6A & 6C). HIV-1 DNA was detected in low copy numbers, comparable to those occasionally observed for HIV-1BaL-infected explants treated with 3TC (Fig 5D), in 1 out of 3 (CM only), and 1 out 4 (SP only) experiments for pCH077.t/2627 and pTHRO.c/2626 respectively, whereas HIV-1 DNA was undetectable in 3 out 3 experiments for pRHPA.c/2635 (Fig 6D). The complete absence of HIV-1 DNA in explants infected with pRHPA.c/2635 seems in contrast with p24gag levels in culture medium of the same explants, although we cannot exclude that infected cells migrated from the explants into gelatin sponges. In addition, the inoculum of pRHPA.c/2635 was higher than the other two T/F HIV-1, thus possibly resulting in higher absorption and unspecific release of virus. In selected experiments (n = 3), we infected donor-matched tissue explants with HIV-1BaL in parallel with one or more T/F HIV-1. Tissue explants from all three included tissue donors were susceptible to HIV-1BaL but none of the T/F viruses, as evaluated by HIV-1 DNA quantification in explants harvested at the end of culture (Fig 6E). In addition, T/F viruses produced in 293T cells productively infected exogenously activated PBMCs as verified by p24gag levels in culture supernatant and HIV-1 DNA copy numbers in cells harvested at the end of culture (day 8–10 post-infection) (Fig 6F).
This is the first work to attempt to extensively reproduce the immunoregulatory effect of SP on the FGM as described in humans, and combine it with HIV-1 infection ex vivo using human mucosal tissue from the lower FGT. Compared to previous studies employing human tissue explants to study specific aspects of semen-mediated inflammation, the goal of the present work is, by providing accurate tissue-donor and specimen inclusion criteria as well as multiple analytical readouts, to develop an experimental platform open to modifications for the study of the early molecular events regulating HIV-1 transmission to the FGM.
To the best of our knowledge, data on in vivo cytokine production in the human FRT following exposure to semen are limited to one recent study, which revealed an overall lack of change in the levels of most of the cytokines analyzed here, except for a reduction in CXCL8 and increase in TGF-β1, in cervicovaginal lavage collected 2 to 6 h after unprotected sex [34]. In mice, semen affects cytokine production at distal sites from the uterus [35], indicating that the bulk of genital secretions collected in the vagina, in addition to the dilution factor introduced with the collection method, may not accurately reflect local changes occurring at the site of semen exposure. In vitro experiments revealed that both epithelial and stromal cells isolated from the upper and lower FRT can produce and upregulate in response to SP the proinflammatory cytokines and chemokines measured here [13,36,37]. Also in agreement with our findings, enhanced production of CXCL8 and no change in IL-10 levels were previously reported in SP-treated human ectocervical tissue explants [38]. The levels of cytokine gene expression in explants generally recapitulated those measured in ECM, showing an induction of proinflammatory cytokines upon SP treatment. Our results are in agreement with those of Sharkey et al., who compared mRNA levels in ectocervical biopsies collected from healthy fertile women before and after unprotected vaginal coitus [21]. In their study, mucosal exposure to semen was associated with enhanced transcription of IL1A, IL6, CXCL8, and CSF2. In the same samples, PTGS2, TNFA, CCL20, and CXCL1 were also upregulated by semen, although not significantly. Increased levels of both PTGS2 and cyclooxygenase 2 were previously reported in SP-treated human ectocervical tissue explants [39], confirming our findings.
In contrast to early artificial fertilization studies implicating whole semen as the mediator of leukocyte recruitment to the FGM [8,9], our data indicate that SP alone can induce such response. As observed in those same reports and other studies in mammals [7], neutrophils and monocytes are the predominant migrating populations early after mucosal exposure to semen. Tissue infiltration by these cell types is indeed characteristic of acute inflammatory responses, and their interplay is key to the progress and eventual resolution of inflammation [40,41]. ECM from both CM- and SP-treated explants attracted a modest fraction of T cells, suggesting that their recruitment to the FGM may occur later on during the mucosal response to semen. This may be due to antigen presentation requirement in secondary lymphoid organs, as observed for post-coital mucosal infiltration of T cells specific to male antigens in mice [42,43]. Phenotypic analysis of transmigrated cells upon incubation with ECM, together with chemokine neutralization experiments, may provide useful information on the molecular mechanisms regulating immune cell recruitment to the FGM, as shown by CCR5 reduction on transmigrated monocytes in our system. In addition, phenotypic and functional analysis upon incubation with SP-ECM of cellular subsets of interest, isolated either from blood or the FRT, may contribute to the study of cell maturation and differentiation processes towards a local tolerogenic response mediated by SP, as observed for the T regulatory response in mice [43,44], as well as in human dendritic cells in vitro [45].
An important aspect of our model is that SP treatment did not affect cell viability as measured on individual cells isolated from explants after 24 h of culture in agarose. This sets an important difference with other in vitro systems based on primary cells or cell lines for the study of SP [13,24,25]. In those settings incubation with low dilutions of SP for short time lengths was associated with elevated toxicity, which is enhanced by the presence of serum [46]. In the present work we used charcoal stripped FBS at the final concentration of 2.5% to prepare agarose medium, and serum-free medium to dilute SP, treat explants as a control (CM) and incubate explants for 12 h after SP or CM treatment. Although the nature and low concentration of FBS could have contributed to limit cytotoxicity, the structural and functional integrity of tissue explants, even if partially preserved, may explain the absence of cell death observed upon treatment with highly concentrated SP (50%) for a relatively long time (12 h). This aspect offers a significant advantage over classical in vitro systems which results rely on possibly over-diluted semen or SP.
Our results suggest that the acute mucosal response to SP creates a favorable environment for virus replication early during transmission to the FRT, and are in line with reports implicating SP as an enhancer of HIV-1 infection [5,10]. Of note, we challenged explants with HIV-1BaL after SP treatment at first (post-SP). This experimental setting provided a longer period for infection than the incubation with SP, 18 and 4 h respectively, in order to maximize the probability of obtaining a productive infection with the selected virus inoculum. In addition, we wanted to primarily address the effect of SP-induced mucosal changes on HIV-1 replication, avoiding any physical interactions between virions and seminal components, as a result of mixing SP and the virus stock, that could have eventually enhanced the infection [24,47]. Nevertheless, with this experimental setting we cannot exclude the presence during HIV-1 infection of seminal factors retained on the explant surface or within the surrounding agarose upon SP wash-out, as observed for seminal TGF-β1. However, to more closely mimic the biology of HIV-1 sexual transmission to the FGM, we also performed infection with a mixture of HIV-1BaL and SP (SP-mix). In order not to extend explant exposure to SP over 12 h, as originally designed for cytokine experiments, and maximize the probability of obtaining a productive infection at the same time, we doubled the HIV-1BaL inoculum used for the previous infection setting (post-SP) due to the shorter incubation time with the virus. This technical difference in the experimental set-up may explain the higher levels of HIV-1 replication achieved with HIV-SP mixing than SP treatment prior to infection, although a biological effect linked to the incubation and direct interaction between virus and seminal factors could have played a role as well [24,47]. As a future perspective, we would like to interfere with candidate seminal factors that signal to the mucosa, such as prostaglandins and TLR4 ligands [17,20,48], in order to harness the induced cytokine response and potentially modulate its effect on HIV-1 replication.
As opposed to infection with HIV-1BaL, the use of T/F HIV-1 full length IMCs did not result in detectable infection and replication in ectocervical tissue explants. These results evidence differences between HIV-1 isolates, as well as the experimental systems used to test them. In fact, most of the data available on T/F viruses have been produced using cell lines or exogenously activated primary cells [49,50]. Based on the limited data available to date, there is no robust experimental evidence that the T/F IMCs tested in our study can establish a productive infection in mucosal tissues as evaluated in explants from the FRT [32,33]. Nevertheless, those same studies failed to show any difference in transmission efficiency between T/F viruses and HIV-1BaL among other reference chronic viruses [32,33], as also suggested by some in vitro studies, although the topic remains debated [51]. On the other hand, we do not exclude the possibility that T/F HIV-infected cells might have exited explants and accumulated into gelatin sponges thus explaining the sustained production of p24gag observed in our system for pRHPA.c/2635-infected explants at the end of culture. Vaginal myeloid dendritic cells were shown to preferentially uptake T/F HIV-1 to trans-infect lymphocytes in vitro upon migration from human vaginal tissue inoculated with full length IMCs ex vivo [52].
Further analysis of the local cytokine milieu, as well as CD4 and CCR5 levels on virus target cells isolated from explants, may help to rule out the nature and the cause of the observed T/F HIV-1 replication kinetics. Explant activation might be required to boost the replication of less infectious HIV-1 variants, as observed for primary isolates ex vivo [53], as well as in women with pre-existing genital inflammation in vivo [54]. Finally, the comparison with primary isolates from acutely infected individuals may also reveal important differences in the biology of infection between isolates and molecular clones in a more complex system than isolated cells such as tissue explants.
One important limitation of our transmission model is that it does not accurately mimic the natural route and dynamics of virus entry into the mucosa due to the lack of explant polarization. However, it is a valid tool for studying the founder pool of HIV-1 infected cells, which fate is believed to ultimately determine the outcome of a transmission event [55]. Based on our results, it appears that the fraction of tissue-resident HIV-1 target cells present within explants are sufficient to support and modulate the magnitude of virus replication in response to SP, at least in the early stage of transmission. This hypothesis is supported by in vivo studies conducted in non-human primates, in which CD4+ T cell infiltrates became apparent only at day 4 post-infection with SIV, or later on during infection [56,57]. Of importance, the presence of infected cells within explants was confirmed by HIV-1 DNA quantification and in situ HIV RNA hybridization at the end of culture (day 18 post-infection), suggesting that HIV target cells residing within the FGM are either long-lived or undergo local replication. Based on previous studies of human cervicovaginal explants infected with HIV-1 and treated with the following factors, we speculate that SP promotes viral replication by upregulating proinflammatory cytokines, such as IL-6 [58] and CXCL8 [59], as well as lymphotropic growth factors such as IL-7 [22]. These factors can directly upregulate HIV-1 gene transcription, and increase the activation and life-span of HIV-infected and bystander target cells. However, the recruitment of immune cells to the transmission site is likely required for amplification and systemic spread of the infection in vivo. For instance, in our system SP upregulated the production of CCL20, a chemokine that was previously shown to be involved in promoting HIV-1 infection of the FGM through recruitment of Langerhans cells in vitro [60], and plasmacytoid dendritic cells in non-human primates already at day 1 post-infection with SIV [56,57]. Moreover, T helper 17 cells expressing the CCL20-binding receptor CCR6 were identified as preferential targets of SIV during vaginal transmission [61].
Of importance, intravaginal challenge with SIV is usually performed in the absence of SP, and this may affect the nature of the local response in the FGM as well as the composition of infiltrating cells. As mentioned above, mucosal exposure to semen in mice resulted in the establishment of local tolerance to paternal antigens in the FRT [43]. The response triggered by the combination of HIV-1 infection and SP treatment may have peculiar features distinct from the two separate treatments. This point remains unaddressed in the present study, as leukocyte transmigration experiments were performed using ECM from explants treated with SP from HIV-uninfected individuals or CM in the absence of HIV-1, in order to specifically investigate the mucosal response to SP. Therefore it is challenging to extrapolate implications for HIV-1 transmission from our results on ECM cytokines and cell transmigration in response to ECM. The use of SP from HIV-infected individuals to treat explants and related ECM, would be more relevant to this end. It is indeed well know that HIV-1 infection, as well as the stage of infection, can significantly affect the immunologic and microbiologic profiles of semen, with potential implications for virus transmission [62–64]. Cell-associated HIV is also an important source of transmitted virus [65–67] that was not addressed in the present study and may be object of future investigations.
The evidence that HIV-1 transmission to the FGM is inefficient [1] may not reflect the likelihood of a locally productive infection being established, the propagation of which would require the maintenance of an inflammatory environment as modeled in our system, but rather the ability of the mucosa to rapidly counteract semen-induced inflammation, among other inflammatory conditions. The response to semen is transient in vivo and may resolve passively once semen is washed out, or through activation of endogenous or seminal factors at a later stage [7]. For instance, TGF-β is highly concentrated in SP as inactive precursor and is reportedly activated in the FGM by numerous mechanisms [36]. The magnitude of the TGF-β1 response, as well as that of other seminal and mucosal factors in vivo, may be affected by modification of the tissue microenvironment by infiltrating leukocytes. Therefore, our system cannot completely reproduce the events following the acute response to semen and particularly the local tolerogenic or low inflammatory environment observed in animal models days post-exposure [43], and in women experiencing frequent unprotected sexual intercourse as sex workers [68,69].
In conclusion, the experimental system presented here reproduces the early events surrounding the interaction between SP and the FGM. Our results suggest that semen and the FGM do not act as a passive carrier and recipient of HIV-1, and understanding their crosstalk may reveal important clues about local susceptibility to infections, as well as other inflammatory disorders of the FRT, such as infertility and cancer. As evidenced by others [2,70], our study supports the notion that including semen or its components in the design of experimental and preclinical models of HIV-1 transmission may increase the relevance of their findings accelerating the discovery of new prevention measures.
Approval for the collection of human uterine samples was obtained from the Regional Ethical Review Board of Stockholm. Written informed consent was obtained from all tissue donors. Medical history information was collected from tissue donors via a questionnaire at the time of surgery.
Semen was obtained via masturbation after ≥48 h of abstinence from 15 donors (aged 18 to 65 years) recruited at the Venhälsan Clinic of the Södersjukhuset, Stockholm. Exclusion criteria were: HIV seropositivity, clinical symptoms of sexually transmitted disease within 3 months prior to donation, systemic immunosuppressive therapy, and infertility (if known). The HIV test was conducted using an Alere Determine HIV-1/2 antibody test (Alere, Waltham MA, USA) at the Venhälsan Clinic.
Semen specimens were collected in a sterile container, allowed to liquefy at room temperature (RT) and processed within 60 min from collection. Specimens were supplemented with penicillin-streptomycin 100U/ml, gentamicin 50μg/ml, Fungizone 2.5μg/ml (all from ThermoFisher Scientific, Waltham MA, USA), and bovine serum albumin (BSA) 0.1% (Sigma-Aldrich, St. Louis MO, USA). SP was separated from cells by centrifugation at 700×g for 15 min, clarified by centrifugation at 17000×g for 5 min in a new tube, aliquoted and stored at -80°C. For treatment of ectocervical tissue explants, aliquots of SP from five randomly selected donors were thawed and pooled in equal proportion before each experiment. Cytokine concentrations in the SP pools used for cytokine measurements in explants and related ECM are reported in S1 Table and S4 Fig.
Uterine specimens were obtained from patients undergoing hysterectomy for benign conditions, such as heavy/irregular menstrual bleeding or benign myoma, at the St. Göran Hospital, Stockholm. Exclusion criteria were: clinical symptoms of sexually transmitted disease within 3 months prior to surgery, systemic immunosuppressive therapy, and human papilloma virus (HPV) positivity. HPV genotyping was conducted on cervical swabs using a PapilloCheck HPV test (Greiner Bio-One GmbH, Kremsmünster, Austria) at the accredited microbiology laboratory of the Karolinska University Hospital.
Culture medium (CM) was prepared by supplementing RPMI 1640 with gentamicin 50μg/ml, Fungizone 2.5μg/ml, non-essential amino acids, and sodium pyruvate 1mM (all from ThermoFisher Scientific). A 4-9cm2 sample of mucosa was dissected from the uterine ectocervix by a pathologist immediately after hysterectomy and maintained in CM at 4°C for transportation. Ectocervical tissue was processed within 6–8 h from surgery. The mucosa was dissected into tissue blocks (i.e., explants) of approximately 8 mm3. One explant was snap-frozen immediately after dissection and cryopreserved at -80°C for in situ terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay (S1A Fig).
Dissected ectocervical explants were placed at the bottom of a 24-well plate (1 explant per well) with the epithelial side facing upwards (S1A Fig). Due to the limited amount of tissue available, only one technical replicate was performed for each experimental condition. Agarose gel 4% (ThermoFisher Scientific) was melted at 70°C for 10 min and equilibrated at 37°C for 2 min. Agarose was mixed 1:1 with CM supplemented with charcoal stripped FBS 5% (ThermoFisher Scientific) pre-warmed at 37°C. A volume of 300–500μl of agarose medium was used to embed ectocervical tissue explants in the 24-well plate to cover all cut surfaces, leaving the epithelial side exposed to the air. Agarose was allowed to solidify at RT for 2 min, before adding 400μl of serum-free CM into the well. The explants were incubated at 37°C, CO2 5%, and humidity 95% for 1 h. The supernatant was removed, and the explants incubated with serum-free CM or SP. Pooled SP was diluted 1:1 (50%, v/v) or 1:3 (25%) in serum-free CM. Explants were treated by adding 400μl of CM (untreated control) or SP to the well and incubated at 37°C, CO2 5%, and humidity 95% for 2, 4, or 12 h. The supernatant was then removed, and the explants were harvested and snap-frozen for in situ TUNEL, or washed and incubated with 500μl of fresh serum-free CM at 37°C (S1A Fig). After 12 h, the supernatant (i.e., ECM) was collected, clarified by centrifugation at 17000×g for 1 min, and stored at -80°C. Explants were harvested and soaked in RNAlater RNA Stabilization Reagent (Qiagen, Hilden, Germany) before storage at -80°C.
To evaluate SP cytokine carry-over, after dissection explants were soaked in a solution of phosphate-buffered saline (PBS) and PFA 4% (Sigma-Aldrich) at RT for 2 h and subsequently at 4°C for 12h. Donor-matched fresh and PFA-inactivated explants were treated with CM and SP50% for 12 h, as described above. The CM and SP solutions used to treat explants were harvested before washing the explants, clarified by centrifugation at 17000×g for 1 min, and stored at -80°C along with a sample of fresh CM collected immediately after washing. Explants were incubated with 500μl of fresh CM at 37°C for 12 h. Thereafter ECM was collected, clarified by centrifugation at 17000×g for 1 min, and stored at -80°C.
The cyclooxygenase-inhibitor indomethacin (Sigma-Aldrich) was reconstituted in absolute ethanol at the concentration of 10mM and stored at -20°C for maximum 2 months. For indomethacin treatment, indomethacin was added to CM and SP50% at the concentration of 10μM, and 30 μM in selected experiments, and to fresh CM for the following 12h-incubation. In selected experiments, explants were also treated with lipopolysaccharides (LPS) from Escherichia coli O11:B4 (Sigma-Aldrich) at 2μg/ml for 12 h, in the presence or the absence of indomethacin 10μM and 30μM.
To evaluate toxicity of SP treatment, donor-matched explants were treated with CM and SP50% for 12 h and subsequently incubated with fresh CM for 12 h, as described above. As a positive control, explants were treated with a solution of medium and the topoisomerase-inhibitor camptothecin (Sigma-Aldrich) at the concentration of 100μM for 24 h to induce apoptosis. At the end of culture, explants were harvested and immediately processed for cell isolation. A total of 9 explants (3 explants per well) were used for each experimental condition.
Cytokine concentration in ECM was measured using a multiplex bead-array immunoassay, developed as previously described [71]. All monoclonal capture antibodies, biotinylated polyclonal detection antibodies, and human recombinant cytokines were purchased from R&D Systems (Minneapolis MN, USA), except for IL-10 capture and detection antibodies that were purchased from BioLegend (San Diego CA, USA). Individual magnetic carboxylated bead sets (Luminex, Austin TX, USA) were coupled to the capture antibodies according to the manufacturer’s recommendations. Human recombinant cytokines were resuspended at concentrations ranging from 5 to 10ng/ml, and diluted serially 1:3 to generate standard curves. All assay procedures were performed in a buffer containing PBS supplemented with BSA 0.1%. The assay was run using 2000 beads per bead set in a total volume of 50μL per well. Samples of ECM were run in duplicates at 2 dilutions 1:1 and 1:9. 50μl of sample were added to the bead mixture and incubated overnight at 4°C in a Bio-Plex Pro flat bottom 96-well plate (Bio-Rad). Plates were washed twice with PBS containing Tween-20 0.05%. The beads were incubated with the detection antibody mixture for 60 min at RT. Biotinylated detection antibodies were used at twice the concentrations for a classic enzyme-linked immunosorbent assay recommended by the manufacturer. Plates were washed twice, and the beads incubated with a solution containing phycoerythrin-conjugated streptavidin 6μg/ml (ThermoFisher Scientific) for 30 min at RT. Beads were acquired with a Bio-Plex100 system (Bio-Rad). The median fluorescence intensity of a minimum of 100 beads per each bead set was recorded in each sample, and analyzed with the Bio-Plex Manager software (Bio-Rad) using a 5P regression algorithm. Concentration values were normalized to explant weight. Concentration values that were below the lower limit of quantification (LLOQ) were reported as the midpoint between zero and the LLOQ for statistical analysis. The LLOQ (pg/ml) was 2.0 for IL-1α, 4.6 for IL-6, 11.0 for TNF, 11.0 for CCL5, 20.6 for CCL20, 2.0 for CXCL1, 3.6 for CXCL8, 6.8 for TGF-β1, and 2.0 for IL-10.
Ectocervical explants preserved in RNAlater were weighed, washed with PBS, and processed to purify RNA using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Eluted RNA was treated with DNase for 30 min at 37°C using a TURBO DNA-free kit (ThermoFisher Scientific). RNA was reversed transcribed into first-strand cDNA using a SuperScriptII reverse transcriptase kit, random hexamers 2.5μM, and dNTP mix 0.5mM (all from ThermoFisher Scientific).
The sequences of the primers used for quantitative real-time PCR are listed in S3 Table. Some primer pairs were designed using the software Primer-BLAST [72,73]. All primers were purchased from ThermoFisher Scientific. Serial dilutions of cDNA were used to evaluate the amplification efficiency of each pair of primers as determined by regression analysis of amplicon abundance vs. threshold cycle (Ct). All used primers had efficiencies within 90 and 110%. Primer specificity was evaluated from the dissociation curve profile of PCR products. PCR amplification was performed in a 20μl-reaction containing HOT FIREPol EvaGreen qPCR Supermix (Solis BioDyne, Tartu, Estonia), forward and reverse primers 0.25μM each, and 3μl of cDNA, using an ABI Prism 7500 real-time PCR system (Applied Biosystems, Foster City CA, USA). The thermal cycler profile was as follows: 95°C for 12 min, 40 cycles of 95°C for 15 sec, 60°C for 30 sec, and 72°C for 1 min. PCR products were analyzed by dissociation curve profile. Samples were run in duplicate.
The relative quantity (RQ) of genes of interest (GOI) was calculated using the equation 2^ -ΔCt, where ΔCt is the difference between the Ct of the GOI and the Ct of an endogenous reference gene. N-fold values were calculated using the equation 2^ -ΔΔCt, where ΔΔCt is the ratio between the GOI RQ in SP and/or indomethacin-treated explants, and the GOI RQ in donor-matched untreated explants (CM). The genes encoding for glyceraldehyde 3-phosphate dehydrogenase (GAPDH), beta-actin (ACTB), and ubiquitin C (UBC) were used as endogenous references. In each independent experiment, GOI RQ and n-fold values were calculated individually for the 3 endogenous reference genes, and individual n-fold values were averaged to perform statistics.
Peripheral blood was collected from healthy volunteers at the Karolinska University Hospital. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation at 400×g for 20 min using Ficoll-Hypaque (GE Healthcare, Wilmington MA, USA). Granulocytes were recovered from the resulting pellet upon resuspension in a solution containing EDTA 0.1mM, NH4Cl 180.0mM, and KCO3 10.0mM to lyse erythrocytes. The PBMC and granulocyte fractions were counted, resuspended at 1×106 cells/ml in medium, and mixed 1:1 (i.e. PBL). A total of 4×105 PBL (400μl) were added into a 24-well plate insert with a 5μm-pore filter (Millipore, Billerica MA, USA) and incubated with 800μl of ECM from donor-matched CM- or SP-treated explants diluted 1:3 in medium added into the well, for 2 h at 37°C, CO2 5%, and humidity 95%. PBL were incubated with medium only or medium supplemented with FBS 10% as a negative and positive control respectively. Due to the limited volume of ECM available, only one technical replicate was performed for each experimental condition. The cell suspension in the wells was harvested, and 50μl of CountBright beads (ThermoFisher Scientific) was added to each sample. Cells were washed with PBS EDTA 1mM, and stained with the following antibodies: CD45-V450 (HI30), CD3-V500 (UCHT1), CD14-FITC (MφP9), CD66b-PE (G10F5), CD195 (CCR5)-APC (2D7), CD16-APC-Cy7 (3G8), CD181 (CXCR1)-PE (5A12), and CD182 (CXCR2)-APC (6C6) (all from BD Biosciences, San Jose, CA, USA). For each sample a minimum of 1x104 events were recorded in the gate corresponding to the beads using a Gallios analyzer (Beckman Coulter, Brea CA, USA) (S5 Fig). Data were analyzed using FlowJo v10.0.7 (FlowJo, Ashland OR, USA). To compare the number of acquired events between samples within the same experiment, the number of cells was normalized to the number of beads.
Ectocervical explants were digested using a solution of Hank’s balanced salt solution with calcium and magnesium (ThermoFisher Scientific), DNase I grade II at 100μg/ml and Liberase Dispase Low Research Grade at 16μg/ml (0.08 Wünsch units/ml) at 37°C for 1 h rocking at 1500 revolution per minutes using a thermomixer (1ml for 9 explants in a 1.5ml microtube). Both enzymes were purchased from Roche Diagnostics (Mannheim, Germany) and reconstituted in distilled cell culture grade water at the concentration of 1mg/ml for storage at -20°C. After tissue digestion, the supernatant was filtered on a 100μm-cell strainer (BD Biosciences) and explants were mechanically disrupted using a pestle, rinsed with PBS EDTA 1mM and the supernatant transferred onto the filter. Cells were washed twice with PBS EDTA 1mM and spun at 250xg for 5 min. Cells were stained separately with the LIVE/DEAD far red dead cell staining kit (ThermoFisher Scientific), CD45-Pacific Blue (HI30) (BioLegend), and Annexin V-PE (BD Biosciences) according to the manufacturers’ instructions. Events were acquired using a CyAn ADP analyzer (Beckman Coulter) and data were analyzed using FlowJo v10.0.7 (FlowJo).
Cryopreserved samples of ectocervical tissue explants were mounted in optimal cutting temperature (OCT) embedding medium (Histolab Products, Askim, Sweden) and cut into 8μm-thick sections. Sections were fixed in PFA 2% (Sigma-Aldrich) at RT for 10 min, and processed to visualize chromatin fragmentation using the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) kit TACS-XL according to manufacturer's instructions (Trevigen Bio-Techne). Briefly, sections were treated with proteinase K 1:60 for 15 min, hydrogen peroxide 3% for 10 min, and incubated with a mixture of terminal deoxynucleotidyl transferase (TdT) and brominated nucleotides (BrdU) for 60 min at 37°C. The TdT enzyme was omitted to generate a negative control staining. Sections were incubated with a biotinylated anti-BrdU antibody and subsequently with streptavidin-conjugated horseradish peroxidase, and staining reactions were developed using diaminobenzidine tetrahydrochloride according to the manufacturer's instructions. Nuclear counterstaining was performed with methyl green provided in the kit. Stained tissue sections were converted to digital images using a NanoZoomer Slide Scanner (Hamamatsu Photonics, Hamamatsu City, Shizuoka, Japan).
HIV-1BaL was obtained through the NIH AIDS Reagent Program (Division of AIDS, NIAID, NIH) from Drs. Suzanne Gartner, Mikulas Popovic, and Robert Gallo [74,75]. A cell-free virus stock was produced in PBMCs. Culture supernatant was harvested at day 10 post-infection, passed through a 0.45μm filter, aliquoted and stored at -80°C.
Plasmids encoding the molecular clones pCH077.t/2627, pRHPA.c/2635, and pTHRO.c/2626 were obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH: Panel of full-length transmitted/founder (T/F) HIV-1 Infectious Molecular Clones (Cat #11919) from Dr. John Kappes and Dr. Christina Ochsenbauer [30,31,76–78]. The plasmids were amplified in One Shot Stbl3 Chemically Competent E. coli (ThermoFisher Scientific) and purified using the EndoFree Plasmid Maxi Kit (Qiagen) according to the manufacturer's instructions. Sequence homology between original and amplified plasmids was confirmed by digestion with restriction enzymes and gel electrophoresis. 293T cells were a kind gift of Mr. Vlad Radoi. Transfection of 293T cells with plasmids was carried out using FuGENE 6 Transfection Reagent (Promega, Madison WI, USA) according to the manufacturer's instructions. Culture supernatant was harvested at 48 and 72 h post-transfection, filtered, aliquoted and stored at -80°C. A cell-free viral stock was also produced by infecting PBMCs with the virus produced in 293T cells. Culture supernatant was harvested at day 8–10 post-infection, filtered, aliquoted and stored at -80°C.
Donor-matched ectocervical explants mounted in agarose (3 explants per well) were incubated with CM or SP25% for 4 h at 37°C, CO2 5%, and humidity 95% (S1B Fig). The supernatant was removed, and the explants washed with CM and incubated for 18 h at 37°C with a final volume of 400μl of HIV-1 stock. For T/F HIV-1, viral stocks produced in 293T cells and PBMCs were mixed 1:1 immediately before explant infection. The p24gag concentration (ng/ml) of virus inoculum was 48.8 ± 5.3 for HIV-1BaL, 153.8 ± 6.7 for pCH077.t/2627, 179.0 ± 6.6 for pRHPA.c/2635, and 127.1 ± 2.6 for pTHRO.c/2626 (mean ± s.e.m., n = 3). Alternatively, donor-matched explants mounted in agarose were infected with 400μl of a mixture of HIV-1BaL and SP25% (final concentration) or CM for 12 h at 37°C. The p24gag concentration (ng/ml) of virus inoculum was 104.9 ± 14.9 (mean ± s.e.m, n = 3). Depending on the amount of tissue available, 9–12 explants were used for each experimental condition to generate two technical replicates.
After infection, the supernatant was removed and the explants were transferred onto gelatin sponges (Aegis Lifesciences, Ahmedabad, India) in a 12-well plate containing CM supplemented with FBS 15% (4–6 explants per well). The explants were maintained at the liquid-air interface for 18 days, as previously described [79,80]. Culture supernatant was sampled every 3 days before replacing it with fresh medium, and stored at -80°C. Viral replication was quantified by measuring the amount of HIV-1 p24gag in culture medium using a bead-based immunoassay [81]. Cumulative virus production was calculated by summing the p24gag concentrations measured over time. At the end of culture, infected tissue explants were cryopreserved at -80°C for DNA extraction or in situ HIV RNA hybridization.
DNA was extracted from infected tissue explants using the DNeasy Blood and Tissue Kit (Qiagen) according to the manufacturer’s instructions. At least five tissue explants for each experimental condition (e.g. CM, SP) were processed to generate a DNA preparation. RNA digestion with RNase A was performed to avoid co-purification of viral RNA. Eluted DNA was diluted 1:4 in water. HIV-1 pol copy numbers were quantified by real-time qPCR as described by Gibellini et al [82]. The gene hemoglobin subunit beta (HBB) was used to normalize HIV-1 DNA copy numbers to the amount of input DNA in order to compare samples. The sequences of HIV-1 pol and HBB primers are listed in S3 Table [83,84]. Serial dilutions of the pTHRO.c/2626 plasmid and human DNA were used to generate standard curves to quantify HIV-1 pol and HBB copy numbers respectively, as determined by regression analysis of amplicon abundance vs. Ct values. Amplification efficiency was within 90 and 110% for all included assays. Amplification was performed in a 20μl-reaction containing HOT FIREPol EvaGreen qPCR Supermix (Solis BioDyne), forward and reverse primers 0.25μM each, and 10μl of DNA, using a QuantStudio 5 real-time PCR system (Applied Biosystems). The thermal cycler profile was as follows: 95°C for 12 min, 40 cycles of 95°C for 15 sec, 60°C for 30 sec, 72°C for 30 sec, 75°C for 5 sec (HIV-1 pol); 95°C for 12 min, 40 cycles of 95°C for 15 sec, 58°C for 30 sec, 72°C for 30 sec (HBB). PCR products were analyzed by dissociation curve profile. Samples were run in duplicate.
Cryopreserved samples of ectocervical tissue explants infected with HIV-1 were mounted in OCT (Histolab Products) and cut into 8μm-thick sections. Sections were fixed overnight in chilled PFA 4% (Sigma-Aldrich) at 4°C and processed to detect HIV RNA using the RNAscope technology [85], according to the RNAscope 2.0 HD Detection Kit (RED) (Histolab Products). The following probes (all from Histolab Products) were used: Probe Hs-HIV (311921), Positive Control Probe Hs-UBC (310041), Positive Control Probe Hs-PPIB (313901), and Negative Control Probe DapB (310043). Nuclear counterstaining was performed with hematoxylin (Histolab Products). Stained tissue sections were converted to digital images using a Panoramic 250 Flash Slide Scanner (3DHistech, Budapest, Hungary).
All experiments were conducted independently using donor-matched tissue or ECM. To minimize the effects of biological variability between tissue donors and SP pools, the results of explant treatment with SP and/or indomethacin were normalized to those of donor-matched untreated control tissue (CM), and the ratio was defined as n-fold. N-fold values were tested against 1 using the Wilcoxon signed rank test. Differences in n-fold change between two groups were evaluated using the Mann-Whitney test for groups with mixed paired and unpaired data, and the Wilcoxon matched-pairs signed rank test for groups containing only paired data. All tests were two-tailed. Differences in ECM cytokine n-fold change between multiple groups were evaluated using the Kruskal-Wallis test for groups with mixed paired and unpaired data, and the Friedman test for groups containing only paired data, with Dunn's multiple comparisons test. p<0.05 indicated statistical significance.
Accession numbers (Entrez Gene) for the genes and transcripts analyzed here are reported in S3 Table.
Proteins (Swiss-Prot): IL-1α (P01583.1), IL-6 (Q75MH2), TNF (P01375.1), CCL5 (P13501.3), CCL20 (P78556.1), CXCL1 (P09341.1), CXCL8 (P10145.1), TGF-β1 (P01137.2), IL-10 (P22301.1).
Viruses (Entrez Nucleotide): HIV-1BaL (AY713409.1), pCH077.t/2627 (JN944941), pRHPA.c/2635 (JN944944), pTHRO.c/2626 (JN944946).
|
10.1371/journal.ppat.1006098 | Receptor Activation of HIV-1 Env Leads to Asymmetric Exposure of the gp41 Trimer | Structural rearrangements of HIV-1 glycoprotein Env promote viral entry through membrane fusion. Env is a symmetric homotrimer with each protomer composed of surface subunit gp120 and transmembrane subunit gp41. Cellular CD4- and chemokine receptor-binding to gp120 coordinate conformational changes in gp41, first to an extended prehairpin intermediate (PHI) and, ultimately, into a fusogenic trimer-of-hairpins (TOH). HIV-1 fusion inhibitors target gp41 in the PHI and block TOH formation. To characterize structural transformations into and through the PHI, we employed asymmetric Env trimers containing both high and low affinity binding sites for individual fusion inhibitors. Asymmetry was achieved using engineered Env heterotrimers composed of protomers deficient in either CD4- or chemokine receptor-binding. Linking receptor engagement to inhibitor affinity allowed us to assess conformational changes of individual Env protomers in the context of a functioning trimer. We found that the transition into the PHI could occur symmetrically or asymmetrically depending on the stoichiometry of CD4 binding. Sequential engagement of multiple CD4s promoted progressive exposure of individual fusion inhibitor binding sites in a CD4-dependent fashion. By contrast, engagement of only a single CD4 molecule led to a delayed, but symmetric, exposure of the gp41 trimer. This complex coupling between Env-CD4 interaction and gp41 exposure explained the multiphasic fusion-inhibitor titration observed for a mutant Env homotrimer with a naturally asymmetric gp41. Our results suggest that the spatial and temporal exposure of gp41 can proceed in a nonconcerted, asymmetric manner depending on the number of CD4s that engage the Env trimer. The findings have important implications for the mechanism of viral membrane fusion and the development of vaccine candidates designed to elicit neutralizing antibodies targeting gp41 in the PHI.
| For HIV, cellular invasion requires merging viral and cellular membranes, an event achieved through the activity of the viral fusion protein Env. Env consists of three gp120 and three gp41 subunits symmetrically arranged on the viral surface. The gp120 subunits bind cellular receptors, which, in turn, coordinate gp41 conformational changes that promote membrane fusion. Understanding these structural rearrangements illuminates the mechanism of viral membrane fusion, and also spurs development of targeted inhibitors of viral entry and vaccine candidates that elicit antiviral immune responses. In this study, we employed a novel strategy to investigate individual subunits in the context of functioning Env complexes. The strategy links distinct gp120-receptor interactions to conformational changes that expose specific gp41 subunits. We found that, despite the initial symmetric arrangement of its subunits, Env conformational changes most often proceed quite asymmetrically, leading to exposure of only one-third of the gp41 trimer for much of the fusion event. This finding might explain why attempts to elicit potent anti-HIV antibodies to a fully exposed gp41 trimer have been largely unsuccessful. The study gives us a glimpse of the early structural transitions leading to Env-mediated membrane fusion and provides a framework for interrogating the fusion proteins of other membrane-encapsulated viruses.
| Entry of human immunodeficiency virus type 1 (HIV-1) into target cells involves fusion of viral and cellular membranes, a process mediated by the viral surface protein Env (gp160) [1]. This heavily glycosylated, type 1 transmembrane protein assembles as a homotrimer following synthesis in the endoplasmic reticulum of virus-producing cells. In the Golgi apparatus, each protomer is cleaved into two subunits that remain noncovalently associated: an N-terminal surface protein (denoted SU or gp120) and a C-terminal transmembrane protein (denoted TM or gp41). Cryo-EM studies on HIV-1 particles revealed that the Env trimer assumes a lobed, mushroom-like appearance, with the gp120 subunits forming a canopy that surrounds a stalk formed by gp41 [2–5]. In high-resolution structures of the Env ectodomain, the N-terminal portion of each gp41, including a 3,4-hydrophobic heptad repeat sequence denoted the N-HR, is cradled by the conserved interior region of a single gp120 subunit (Fig 1) [6–11]. Parts of the N-HR segments adopt a homotrimeric coiled-coil conformation that stabilizes the trimeric interface. The C-terminal portion of the gp41 ectodomain, including a second heptad repeat (denoted C-HR), interacts with and extends beyond the membrane-proximal face of the gp120 trimer (Fig 1). At the other end of the complex, variable loops (V1/V2 and V3) from each gp120 partake in intersubunit interactions to effectively cap the gp120 canopy [6, 7, 10].
In mediating HIV-1 entry, Env undergoes a series of coordinated structural transformations initiated when gp120 binds cellular CD4 [13–15]. This event substantially alters gp120 structure, releasing constraints on the V3 loop and enabling formation of a new antigenic surface called the bridging sheet [16–18]. The released V3 loop and bridging sheet interact with the extracellular loops and N-terminus of chemokine receptors such as CXCR4 or CCR5 (denoted as the coreceptor) [19–22]. CD4-binding also modifies the interaction between gp120 and gp41, often resulting in irreversible shedding of gp120 [23, 24]. Critical to fusion, the gp41 ectodomain can now extend and insert its N-terminal fusion peptide segment into the target cell membrane, thereby spanning the gap between viral and cellular membranes (Fig 1A) [25–27]. Sometime during this extended intermediate state (denoted the prehairpin intermediate state or PHI), the entire gp41 N-HR adopts a coiled-coil conformation, and both the N-HR and C-HR regions become exposed to the extracellular environment (Fig 1A and 1C) [28, 29]. Ultimately, the extended state collapses into its most stable form, a trimer-of-hairpins (TOH), in which the C-HR regions pack in an antiparallel manner into hydrophobic grooves on the outside of the N-HR coiled coil (Fig 1A) [12, 30–32]. The structure brings the fusion peptides and transmembrane domains of the gp41 trimer, and their associated membranes, into the close proximity required for stable fusion [28].
The molecular mechanisms that guide global structural changes of Env are largely unknown. For instance, it remains unclear how the conformational transitions of the gp41 subunits are allosterically coupled to CD4 and coreceptor binding to gp120. As a first step in unraveling these mechanisms, this study focuses on the nature of the native-to-PHI transition that exposes the N-HR and C-HR regions of the gp41 ectodomain. In the current model, the transition is envisioned as a concerted process, with the three gp120 subunits simultaneously disengaging from the gp41 trimer leading to full (symmetric) exposure of the N-HR coiled coil and three C-HR regions [14, 33]. In support of this model, the structure of the Env trimer in its native conformation suggests that all three gp120 subunits must fully separate from the three gp41 N-terminal regions in order for the fusion peptides to extend and N-HR coiled coil to form (Fig 1B) [6, 7]. Furthermore, there is evidence of substantial cooperativity between protomers, which enables the Env trimer to function without engagement of all receptor-binding sites [34–36]. However, biological process involving homomeric proteins rarely proceed in a concerted manner [37, 38]. It is a formal possibility that the Env trimer transitions through one or more asymmetric intermediates, with partial separation of gp120 from the gp41 trimer leading to piecemeal exposure of the gp41 ectodomain.
Probing initial receptor-induced conformational transitions of Env is complicated by the homotrimeric oligomerization of the glycoprotein. The three identical CD4- and coreceptor-binding sites, the three-fold symmetry of the N-HR coiled coil, and the three identical C-HR regions make it difficult to assess the structural transformations of individual gp120 and gp41 subunits. Here, we circumvent this obstacle by studying two types of asymmetric Env trimers. The first type is a heterotrimer of functionally complemented Env protomers, one deficient in CD4 binding and the other deficient in coreceptor binding [34, 35]. The second type is an Env homotrimer containing a gp41 substitution that disrupts three-fold symmetry in the N-HR coiled coil. We explored the exposure of the gp41 ectodomain in these asymmetric Env species using N-HR- and C-HR-targeted fusion inhibitors that block TOH formation. Our results demonstrate that receptor binding does not trigger three-fold symmetric exposure of fusion inhibitor-binding sites as envisioned in the current model. Instead, the findings point to stepwise disengagement of gp120 subunits from gp41 and a gradual uncovering of the trimeric core.
Fusion inhibitors bind the gp41 N-HR and C-HR regions exposed in a kinetic window between receptor activation of Env and formation of the gp41 TOH (Fig 1A and 1C) [39–54]. Once bound, they block N-HR/C-HR association and promote irreversible deactivation of the extended gp41 trimer [55]. Due to the transient exposure of these binding sites, fusion inhibitor potency reflects both equilibrium and kinetic factors that describe inhibitor association/dissociation and Env conformational transitions [50, 55]. For a fusion inhibitor that readily dissociates and re-associates during the intermediate-state lifetime, potency is indeed proportional to binding affinity. However, for extremely tight binding fusion inhibitors that rarely dissociate before gp41 deactivates, kinetic factors are the predominant determinant of inhibitory activity [50, 55]. The potencies of such kinetically restricted inhibitors depend on the rate of inhibitor association and lifetime of the sensitive state, but not strictly on affinity. Finally, a kinetically restricted inhibitor can be converted to an affinity-dependent inhibitor in the setting of an extremely destabilizing escape mutation. S1 Table provides an overview of these properties for the fusion inhibitors and Env variants used in this study.
The kinetic properties of fusion inhibition make these inhibitors particularly useful in monitoring late conformational changes in gp41, the duration of the PHI, and the steric environment around the N-HR and C-HR regions during membrane fusion [21, 49, 50, 52, 53, 55–60]. Fusion inhibitors are less useful for resolving early Env structural transitions that lead to N-HR and C-HR exposure. Typically, Env homotrimers present three identical binding sites for a given fusion inhibitor, and deciphering whether these sites are exposed simultaneously or sequentially in the PHI is nearly impossible from inhibitory titrations. We reasoned that this structural information might be gleaned from asymmetric Env trimers in which the three fusion-inhibitor binding sites possessed unequal affinities.
As a first step in generating asymmetry in the gp41 trimer, we utilized a functional complementation strategy to produce heterotrimeric Env species. We generated HIV-1 from cells expressing two Env (HXB2 strain) variants, denoted in this paper A and B. Env A was competent in binding CD4 but deficient in binding CXCR4, while Env B was deficient in binding CD4 but competent in binding CXCR4 (Fig 2A, top). In Env A, interaction with CXCR4 was disrupted by swapping the V3 loop from the HXB2 strain with that from the SF162 strain (CCR5-tropic). In Env B, interaction with CD4 was ablated by substituting Asp368 with Arg in the CD4-binding site of gp120 [61, 62]. These alterations did not impact Env expression, processing or incorporation into viruses (Fig 2B). Consistent with previous studies [34, 35], HIV-1 expressing only A3 or B3 homotrimers were poorly infectious on CD4+CXCR4+CCR5- target cells (Fig 2B, Env B expression fractions of 0 and 1, respectively). However, HIV-1 generated from cells coexpressing Envs A and B showed substantially higher infectivity, indicating that the functional disruptions in individual protomers could be complemented to generate fusogenic Env heterotrimers A2B and AB2.
To characterize the fusogenic activities of these Env heterotrimers, we adjusted the relative protomer expression while keeping total Env level in virus producing cells constant (Fig 2B). HIV-1 infectivity levels peaked at a relative Env B expression fraction of 0.5 (an A:B expression ratio of 1:1) and were roughly symmetric around this value. The variance in infectivity closely matched the predicted combined population of both heterotrimers on viruses (Fig 2B, black line), suggesting that these two Env species promoted viral entry with comparable efficiencies. The results implied that EnvHXB2 trimers with 2 CD4/1 CXCR4 binding sites (A2B) and EnvHXB2 trimers with 1 CD4/2 CXCR4 binding sites (AB2) have similar fusogenic activities.
The functional complementation strategy provided a platform to interrogate CD4- or CXCR4-triggered conformational transitions of individual protomers in the context of functional Env trimers. We applied the technique in the next two sections in order to study the transition into the PHI. To assess exposure of the gp41 ectodomain, we introduced escape mutations into the N-HR or C-HR regions and monitored the impact on fusion inhibitor potency (Fig 3). When the coexpressed Envs A and B both had the same gp41 mutation, the viral population (designated Mm) contained functional Env heterotrimers with no high affinity fusion inhibitor binding sites. Similarly, when neither Env protomer had a gp41 mutation, the viral population (designated Ww) contained Env heterotrimers with three high-affinity fusion inhibitor binding sites. A2B and AB2 heterotrimers on Mm and Ww viruses mimicked mutant and wild type Env homotrimers in terms of symmetry in their gp41 ectodomains. However, when gp41 mutations were incorporated into either Env A or Env B, the viral populations (designated Mw or Wm, respectively) contained functional Env heterotrimers with asymmetric gp41 trimers (Fig 3). These heterotrimers contained either one high affinity / two low affinity binding sites or two high affinity / one low affinity binding site. Importantly for these functional Env trimers, escape mutations were introduced into protomers restricted to either CD4- or CXCR4-binding. In Mw viruses, the escape mutation was introduced into Env A, the protomer that binds CD4. In Wm viruses, the escape mutation was introduced into Env B, the protomer that binds CXCR4. We reasoned that this linkage might allow us to ascertain whether specific fusion-inhibitor binding sites were exposed upon CD4 or CXCR4 binding.
We probed exposure of the N-HR coiled coil using peptide fusion inhibitors T20, C37-KYI and di-C37 (Fig 1C) [43, 55]. With sequences derived from the gp41 C-HR region, these so called C-peptides bind the hydrophobic groove formed at the interface of two N-HR helices in their coiled-coil conformation [12, 30, 31]. T20 interacts relatively weakly with gp41 and is sensitive to affinity disrupting escape mutations in the N-HR region [63]. C37-KYI and di-C37 bind extremely tightly to gp41 and have kinetically restricted inhibitory potencies that are not sensitive to small affinity disruptions [55]. We employed resistance mutations L544S and V549A that reside within the T20 binding site and decrease C-peptide interaction affinity 50–100 fold (Fig 1C) [50]. Neither point mutation altered Env expression, processing or viral incorporation (S1A Fig and S2A Fig). The V549A mutation had no impact on viral infectivity (S2A Fig), while the L544S mutation had a modest impact on homotrimer infectivity that was not recapitulated in the heterotrimer setting (S1A Fig). In EnvHXB2 homotrimers, the L544S and V549A substitutions conferred 30- and 50-fold resistance to T20, respectively, but had no impact on the potencies of C37-KYI and di-C37 (S1B and S2B Figs). These inhibitory characteristics were qualitatively preserved in A2B and AB2 heterotrimers on Ww and Mm viruses, which contain either three high affinity or three low affinity-fusion inhibitor binding sites like wild type and mutant Env homotrimers (Fig 4).
We first tested the inhibitor sensitivities of Env heterotrimers using viral samples generated from cells expressing equal levels of Env A and Env B. Under these conditions, the A2B and AB2 trimeric species were equally represented and contributed comparably to viral entry. Thus, for the Wm and Mw viral populations, the average fusogenic trimer contained approximately 1.5 high affinity and 1.5 low affinity C-peptide binding sites (Fig 3). We reasoned that if receptor-induced structural changes led to symmetric exposure of the N-HR coiled coil, then Wm and Mw viruses would be similarly sensitive to T20. Instead, we observed that the Wm viruses were 10-fold more sensitive to T20 than Mw viruses (Fig 4). This difference in T20 sensitivity was not due to alterations in Env fusion kinetics as Wm and Mw viruses were equally sensitive to kinetically restricted inhibitors (Fig 4B and 4F and S1 and S2 Figs). Rather, the results suggested that high affinity-binding sites of the N-HR coiled coil were more exposed in Env heterotrimers on Wm viruses than in Env heterotrimers on Mw viruses.
We next asked how the two Env heterotrimeric species, A2B and AB2, individually contributed to the difference in T20 sensitivity between Wm and Mw viruses. HIV-1 was generated from cells in which the relative expression of Env A to Env B was biased either 1:9 or 9:1. At the 1:9 expression ratio, where the AB2 heterotrimer accounted for approximately 90% of the fusogenic Env species, Wm and Mw viruses showed nearly identical sensitivity to T20 (Fig 4C and 4G and S1 and S2 Figs). By contrast, for the 9:1 expression ratio where the A2B heterotrimer predominated, Wm viruses were 20-fold more sensitive to T20 than Mw viruses. Thus, the data indicated that the difference in T20 sensitivities of Mw and Wm viruses was primarily due to the A2B heterotrimeric species (summarized in Table 1). Again, these differences in T20 sensitivity could not be attributed to alterations in Env fusion kinetics since all viral populations (Ww, Wm, Mw and Mm) were similarly inhibited by kinetically restricted fusion inhibitors (Fig 4D and 4H). Rather, A2B heterotrimers on Wm viruses appear to preferentially expose a high affinity site and are potently inhibited by T20 (nearly as well Ww viruses), while the same trimers on Mw viruses appear to preferentially expose a low affinity site and have significantly poorer T20 sensitivity. Hence, A2B trimers (2 CD4 / 1 CXCR4 binding site) appear to uncover N-HR regions from CD4-binding Env A protomers before the entire N-HR coiled coil is uncovered. Curiously, AB2 trimers (1 CD4 / 2 CXCR4 binding sites) from Wm and Mw viruses had similar T20 sensitivities, as expected if all three T20 binding sites were equally exposed. The results suggested that single CD4 binding to Env leads to symmetric gp41 N-HR exposure (see Discussion).
Using the same functional complementation strategy, we examined exposure of the gp41 C-HR using the protein fusion inhibitor 5-Helix (Fig 1C) [44, 50]. The wild type version (denoted 5HWT) binds the C-HR region with extremely high affinity and displays kinetically restricted inhibitory potency. A mutant version (denoted 5HLAVA) binds less tightly and displays affinity-dependent inhibitory potency. We employed resistance mutation N656D in the C-HR region that reduces 5-Helix binding affinity by more than three orders of magnitude (Fig 1C). Incorporation of the N656D escape mutation did not alter viral infectivity, Env expression or processing (S3A Fig). In EnvHXB2 homotrimers, N656D conferred more than 100-fold resistance to 5HLAVA but had no impact on the potency of 5HWT (S3B Fig). The same characteristics were observed for Ww and Mm viruses containing A2B and AB2 heterotrimers (Fig 5).
Unlike our interrogation of T20 inhibition with N-HR escape mutations, no large difference in 5HLAVA inhibition was observed between Wm and Mw viruses (Fig 5, Table 1). The IC50 values were within 2-fold of one another at all Env protomer expression ratios. We did observe a small but persistent trend in Wm viruses toward higher sensitivity as the relative fraction of the Env A protomer increased (Fig 5C). That trend was absent in Mw viruses. All viral populations showed similar sensitivity to 5HWT (Fig 5D), indicating that there were no alterations in fusion rates to confound the interpretation of 5HLAVA sensitivities. The data suggested that the high affinity 5-Helix binding sites in heterotrimers on Wm and Mw viruses were similarly exposed, regardless of the stoichiometry of CD4- and CXCR4-binding sites on the Env trimer. Thus, the three gp41 C-HR regions appeared to be uncovered nearly simultaneously, or, at a minimum, much more symmetrically than the T20 binding sites on the N-HR coiled coil.
The previous experiments took advantage of engineered asymmetry in A2B and AB2 Env heterotrimers to unmask properties of the native-to-PHI transition. However, it was formally possible that the combination of Env A and B protomers unnaturally produced asymmetric exposure of the gp41 trimer, and that Env homotrimers with fully intact CD4 and coreceptor binding sites would undergo a more symmetric conformational transition. Addressing this concern required an Env homotrimer with a naturally asymmetric gp41 containing high and low affinity-fusion inhibitor binding sites. Towards this goal, we were assisted by the serendipitous discovery that the Q552R substitution introduced asymmetry into the N-HR coiled coil.
The thermostable core of the gp41 TOH consists of the three-stranded N-HR coiled coil surrounded by three helical C-HR segments (Fig 6A) [12, 30, 31]. Each C-HR segment packs into a hydrophobic groove formed at the interface of two different N-HR helices. The wild type structure displays an overall three-fold symmetry parallel to the longitudinal axis of the coiled coil (Fig 6B). Notably, the Gln552 residues of each N-HR helix (in the a-position of the hydrophobic heptad repeat, see Fig 6A) point toward the hydrophobic interior of the coiled coil, with their side chain amides forming hydrogen bonds with backbone carbonyls on neighboring helices (Fig 6B).
Propagating HIV-1 (NL4-3 strain) in the presence of fusion inhibitor C37-KYI led to the selection of the escape mutation Q552R. With its additional bulk and positive charge, we suspected that the Arg552 side chains would not adopt the same configuration in the gp41 TOH as observed for Gln552. To determine the structural impact of this mutation, we crystallized a construct composed of 40 residues from the N-HR sequence (N40) and 37 residues from the C-HR sequence (C37) and solved its structure (Fig 6C, S2 Table). Although the mutant six-helix bundle crystallized in a hexagonal space group (P31), three-fold crystallographic symmetry did not run through the center of the N-HR coiled coil as it does in almost all previous gp41 TOH structures. Instead, the structure was markedly asymmetric in that only two of the three N40 segments were folded into α-helices throughout their entire length. The N-terminal region (residues 545–554) of the third N40 segment assumed a distorted, largely extended conformation (Fig 6C, S4A Fig). However, by residue 555 (three residues C-terminal to the Q552R mutation), this N40 segment resumed a helical fold, restoring three-fold symmetry to the coiled coil. This symmetry was maintained throughout the remainder of the molecule (S4C and S4D Fig).
As a consequence of the asymmetric distortion of the N-HR coiled coil, the three Arg552 residues occupied distinctly different chemical environments. One Arg552 side chain extended from an intact N-HR helix into the coiled-coil interior, where it formed a hydrogen bond with the backbone of the other intact N-HR helix (S4B and S5 Figs). The Arg552 side chain from the second intact N-HR helix reached into an N40/C37 interface, replacing the Gln551 side chain from an adjacent N-HR segment. The third Arg552 side chain extended from the distorted N-HR segment (next to the displaced Gln551) radially away from the center of the TOH. We speculated that the spatial arrangement of the three Arg552 side chains minimized their electrostatic repulsion, which would be particularly destabilizing if all three were confined within the interior of the N-HR coiled coil.
Unlike the N40 coiled coil, all three C37 peptides in the Q552R-mutant TOH structure folded into their normal helical structures, even in the asymmetric zone of the N40 segment (Fig 6C). The backbone conformation of these C-HR segments closely mimicked that for the wild type six-helix bundle (RMSD = 0.34 Å, S6 Fig). The lack of distortion in the C37 main chains was surprising, but might have resulted from constraints imposed by crystal contacts with neighboring six-helix bundles. Despite their similar structures, the three C37 helices were unlikely to have equal affinities for the Q552R-mutant coiled coil. One C37 helix bound a groove formed by two intact N40 helices and was stabilized by the same N-HR/C-HR contacts observed in the wild type TOH (S5 Fig). By contrast, the other two C37 helices bound into grooves formed by one intact N40 helix and the distorted N40 helix, resulting in the loss of a number of stabilizing interactions (S5 Fig). Thus, in contrast to the wild type N-HR coiled coil, the Q552R-mutant coiled coil possessed dissimilar C37 binding sites, with the distorted N-HR segment substantially disrupting two out of the three sites.
To determine if asymmetry in TOH structure is a definitive feature of HIV-1 resistance to C37, we also crystallized an N40/C37 construct containing the V549E substitution (Fig 6D, S1 Table). Residue 549 (in an e-position of the N-HR hydrophobic heptad repeat, see Fig 6A) extends directly into the N-HR/C-HR interface where it fills a small hydrophobic pocket bounded by side chains of Asn656, Glu657 and Leu660 and the backbone of the C-HR helix (S7A and S7B Fig). The six-helix bundle containing the V594E mutation crystallized in a rhombohedral space group (R3) with a defined three-fold crystallographic symmetry down the axis of the N40 coiled coil. The V549E substitution did not disrupt the continuity of the N40 coiled coil nor C37 docking on its surface (Fig 6D, S7C and S7D Fig). Overall, the mutant structure closely resembled the wild type six-helix bundle (RMSD = 0.94 Å, S6 Fig). Symmetry dictated that the three C37 peptides bind the V549E-mutant coiled coil with equal affinity, which will be lower than that for the wild type coiled coil due to disrupted hydrophobic packing. Hence, asymmetry is not a general feature of escape from C37-like fusion inhibitors, but rather a special property of the Q552R substitution.
As with the previous functional complementation experiments, we used fusion inhibitors to probe gp41 exposure in wild type and mutant Env homotrimers. We reasoned that the presence of both high and low affinity binding sites on the Q552R N-HR coiled coil would manifest in the shape of the inhibitor titration if the sites were sequentially uncovered, but not if the entire coiled coil was symmetrically exposed. HIV-1 was pseudotyped with wild type Env or one of two resistant variants, EnvQ552R/N637K and EnvV549E/N637K. The additional N637K substitution in the C-HR region stabilized the N-HR/C-HR interaction and supported the fusogenic activity of Envs containing affinity-disrupting N-HR mutations (especially the Q552R substitution) [64, 65]. For these experiments, we employed fusion inhibitors C37-KYI, di-C37, 5HWT and PIE12 (Fig 1C, S1 Table). Inhibition by C37-KYI is kinetically restricted against EnvWT, but affinity-dependent against the two resistant Envs [55]. Inhibition by di-C37 and 5HWT is kinetically restricted against all three Envs [55]. PIE12 is an affinity dependent D-peptide inhibitor that targets the C-terminal region of the N-HR coiled coil [66]. The inhibitor combination allowed us to monitor both binding site exposure and Env conformational dynamics around the site of escape mutations and in regions distal to them.
The most striking feature of these inhibition experiments was the nonparallel dose-response curves for the three C37-KYI titrations (Fig 7A). The titration for EnvV549E/N637K was steeper than that for EnvWT, while the titration for EnvQ552R/N637K was substantially less steep. This variance in slope caused the inhibition curves of the two resistant Envs to intersect near their IC50 values. This observation is unique to the C37-KYI titrations, as inhibition by di-C37, 5-Helix and PIE12 yielded parallel dose-response curves for all three Env species (Fig 7B–7D). These results implied that the slope difference observed for C37-KYI was not due to alterations in Env transition kinetics, nor was it a general property of affinity-dependent fusion inhibitors targeting these three Env species. Rather, the effect was specific for an affinity-dependent inhibitor that binds the N-terminal region of the N-HR coiled coil.
To quantify the slope difference, we initially fit C37-KYI inhibition data to the Hill equation and extracted Hill coefficients nH. The titration curve for EnvWT yielded an nH of 1 (Fig 7A, solid black line), implying a single C37-KYI blocked viral entry. The data were consistent with one C37 peptide being sufficient to disrupt the function of an Env trimer (KH and MJR, submitted). Since C37-KYI is kinetically restricted against EnvWT and, therefore, does not dissociate before gp41 deactivates, the binding of a second or third C37-KYI would not alter inhibition probability. Indeed, inhibition data for kinetically restricted fusion inhibitors di-C37 and 5-Helix similarly fit with an nH of 1 (Fig 7B and 7D, all lines). The C37-KYI titration curve for EnvV549E/N637K yielded an nH of 1.45 (Fig 7A, solid green line), indicating that a significant fraction of inhibition events involved more than one bound peptide. This observation likely reflected the reduced binding affinity for C37-KYI causing the inhibitor to dissociate more rapidly than EnvV549E/N637K could deactivate; in this situation, inhibition probability would be enhanced with more than one peptide bound per trimer. The titration curve for EnvQ552R/N637K yielded an nH of 0.67, substantially less than the Hill coefficients from both the kinetically restricted inhibition of EnvWT and affinity-dependent inhibition of EnvV549E/N637K. The low nH value suggested that C37-KYI inhibited EnvQ552R/N637K by interacting with binding sites of different affinity. Therefore, we reanalyzed the data assuming that C37-KYI inhibited a portion of EnvQ552R/N637K trimers with high potency and the remainder with low potency. In this case, infection probability (PI) can be quantified in terms of two IC50 values (IC50.1 and IC50.2) and the C37-KYI concentration ([Inh]):
PI=f(IC50.1[Inh]+IC50.1)+(1−f)(IC50.2[Inh]+IC50.2)
(1)
where f represents the fraction of inhibition events that titrate with IC50.1. The data fit best with f = 0.31, IC50.1 = 8 nM, and IC50.2 = 180 nM (Fig 7A, solid red line). The results suggested that one third of inhibition events involved binding to a high affinity site, while two thirds involved binding to low affinity sites.
The multiphasic titration curve for EnvQ552R/N637K suggested that its N-HR coiled coil adopts the asymmetric conformation observed in the crystal structure of the mutant TOH. More significantly, the data also pointed to sequential, stochastic exposure of C37-binding sites during membrane fusion. If the N-HR coiled coil were exposed symmetrically following a concerted structural change of the Env trimer, then C37-KYI would always be able to bind its high affinity site, and the titration would be monophasic as observed for EnvWT and EnvV549E/N637K. Rather, the data suggested that this high affinity site was only fully exposed for one third of Envs involved in viral membrane fusion. The remainder of the Envs fully exposed a low affinity site. Such a result is expected if receptor-induced structural changes in Env randomly uncovered a single C37-binding site, with subsequent structural changes required to expose the remainder of the N-HR coiled coil in preparation for gp41 collapse into the TOH.
CD4 and chemokine receptor (CoR) binding guide the structural transformation of HIV-1 Env from a constrained, metastable native state through an extended PHI configuration and ultimately into a compact, low energy TOH conformation [15, 28]. Structures of the native Env trimer reveal how the gp120 canopy effectively restrains gp41 subunits from premature triggering, while structures of the TOH divulge its role in juxtaposing the viral and cellular membranes for fusion [6, 7, 12, 30, 31]. Unfortunately, the structure of Env in the PHI and the disposition of gp120 subunits during this state remain unknown. Evidence from temperature and lipid arrest states of Env-mediated membrane fusion and from fusion inhibitor synergy measurements suggest the PHI might itself be a combination of multiple evolving states, with the N-HR region becoming progressively exposed and the C-HR region becoming progressively occluded as fusion proceeds [52, 55]. Our experiments here asked whether the gp41 trimer was revealed in a three-fold symmetric manner upon entry into the PHI, or whether the N-HR coiled coil and three C-HR regions were uncovered in piecemeal fashion.
To interrogate these early conformational transitions, we employed functionally complemented Env heterotrimers consisting of protomers deficient in either CD4- or CXCR4-binding. Fusion inhibitor-escape mutations were then selectively introduced into either protomer, and the impact on inhibitor potency was measured. For N-HR-targeted T20, potency depended on whether the escape mutation was incorporated into the CD4- or CXCR4-binding protomer. By contrast, for C-HR-targeted 5HLAVA, potency was largely unaffected by which protomer contained the mutation. The results suggested that the C-peptide binding sites on the gp41 N-HR coiled coil were exposed in piecemeal fashion, while the gp41 C-HR regions were exposed symmetrically. Our interpretations were based on experiments complicated by multiple levels of Env manipulation (e.g., functional complementation, fusion inhibitor resistance and biased protomer expression). Such alterations, and perhaps even fusion inhibitor binding itself, could have shifted the conformational landscape of the trimer and played a more active role in promoting what appeared to be an asymmetric transition. It was also possible that our observations were a consequence of the fact that CD4 and CoR could not bind the same protomer in the Env heterotrimer. To partially address these concerns, we analyzed the fusion inhibitor sensitivity of a naturally asymmetric Env homotrimer (EnvQ552R/N637K) that has three intact CD4 and CXCR4 binding sites. Inhibitor titrations revealed that C37-KYI blocked viral entry with multiple potencies, as if the C-peptide bound to sites of different affinities and these sites were asymmetrically exposed. These measurements of viral entry inhibition remain an indirect reporter of Env structural changes. More direct methods, such as cryo-EM structural determination combined with inhibitor binding/precipitation assays, will be required to confirm our results.
Our data suggest that entry into the PHI leads to near simultaneous exposure of the three gp41 C-HR regions but only uncovers a portion of the gp41 N-HR coiled coil. The most likely structure occluding C-peptide binding sites is gp120, but how this protein remains bound to gp41 in the PHI is unclear. In native Env, the interior of each gp120 subunit cradles a gp41 N-terminal region that adopts a conformation vastly different from the one envisioned for the PHI state (Fig 1B) [6, 8–11]. The N-HR domain is segmented into two helical regions (residues 548–568, 571–593), an extended loop preceding the first helix (534–547) and a short linker connecting the two helices (569–570) [10]. The two helical segments run roughly antiparallel to one another, with the C-terminal helix from each N-HR domain assuming a coiled-coil conformation. While the C-terminal helix partially interfaces with gp120, the N-terminal helix and extended loop make extensive contacts with the surface subunit (especially its C1 domain). For the entire N-HR region to adopt the coiled-coil structure and form fusion inhibitor binding sites, each gp120 subunit would need to disengage fully from the N-terminal helix, extended loop and gp41 N-terminus (fusion peptide). This structural rearrangement potentially destabilizes the gp120-gp41 interaction. Indeed, addition of soluble CD4 to EnvHXB2 (the viral strain used in these experiments) is sufficient to promote both N-HR coiled-coil formation and rapid shedding of gp120 [39, 48, 54, 67, 68]. However, if receptor triggered entry into the PHI caused all three gp120 subunits to disengage completely, then the three fusion inhibitor binding sites would be exposed simultaneously. Our results suggest that Env trimers actively involved in viral membrane fusion retain at least some of their gp120 subunits well after the N-HR coiled coil has formed.
The T20 sensitivities of biased heterotrimers (Fig 4C and 4G) suggest that receptor binding stoichiometry plays a role in determining how the gp41 N-HR coiled coil becomes exposed. In AB2 heterotrimers (1 CD4 / 2 CXCR4 binding sites), the three C-peptide binding sites on the N-HR coiled coil appear to be exposed equally (reflected by the similar T20 potency against Wm and Mw viruses, Table 1). By contrast, in A2B heterotrimers (2 CD4 / 1 CXCR4 binding site), the C-peptide binding sites appear to be uncovered asymmetrically (reflected by the different T20 potency against Wm and Mw viruses, Table 1). Formally, either multiple CD4 binding or single CXCR4 binding or both could be responsible for the asymmetric exposure of the N-HR coiled coil in A2B heterotrimers. However, the multiphasic C37-KYI titration of EnvQ552R/N637K suggests that trimers with 3 CD4 binding sites and 3 CXCR4 binding sites also reveal their N-HR coiled coil in piecemeal fashion. Although our experiments do not directly report on receptor interaction stoichiometry during viral entry, the findings point to multiple CD4 binding as the primary driver of an asymmetric transition into the PHI.
Although asymmetric exposure of gp41 appears to require at least two CD4 interactions with Env, our data suggest that these binding events uncover only a single C-peptide binding site. This conclusion explains how C37-KYI seems to bind a high affinity site only one-third of the time when inhibiting EnvQ552R/N637K homotrimers. If, instead, asymmetric exposure of the N-HR coiled coil revealed two C-peptide binding sites, then the high affinity site would be exposed two-thirds of the time and two-thirds of inhibition events would be of high potency. Likewise, the similar impact of the L544S and V549A substitutions in our functional complementation experiments can be best understood if only a single site is exposed. Residues 544 and 549 point into different C-peptide binding grooves on the N-HR coiled coil (S8 Fig), and, therefore, mutant heterotrimers display distinct organizations of low and high affinity C-peptide binding sites. As shown in S8 Fig for A2B heterotrimers of Mw viruses, simultaneous exposure of two C-peptide binding sites would reveal two low affinity sites for one mutation but would uncover one low affinity and one high affinity site for the other mutation. However, for both L544S and V549A substitutions, A2B heterotrimers on Mw viruses are insensitive to T20. Thus, our data are most consistent with exposure of a single site on the N-HR coiled coil of A2B heterotrimers, specifically the one common to both L544S and V549A substitutions and formed by the two Env A protomers (S8 Fig).
In Fig 8, we present a working model of early Env structural changes that describes how multiple CD4 binding might uncover a single C-peptide binding site on the N-HR coiled coil. We speculate that an unliganded gp120 subunit remains closely associated with its cognate N-HR helix following coiled-coil formation. This association spatially restricts access to the two C-peptide binding sites on either side of the N-HR helix. CD4 binding leads gp120 to detach from the gp41 N-terminal region and sterically unveil the two previously occluded half-sites on the coiled coil. The first CD4 interaction with the Env trimer might trigger N-HR coiled-coil formation, but no C-peptide binding site would be completely exposed since the two unliganded gp120 subunits remain attached. Alternatively, formation of the N-HR coiled coil and, by extension, the three C-peptide binding sites, might not even occur until the second CD4 binds [9]. In either case, interaction of this second CD4 would uncover a single site, the one formed by the N-HR helices of the two CD4-bound protomers. The other two sites would continue to be partially occluded by the remaining unliganded gp120 subunit. Interaction of a third CD4 would lead to a fully exposed N-HR coiled coil, a thermodynamically unstable conformation that likely resolves rapidly into the TOH (Fig 8A). The short lifetime of this fully exposed state limits C-peptide binding to the two previously occluded sites, reducing their impact on inhibitory potency. Thus, the properties of the first exposed site will have the greatest impact on C-peptide inhibitory potency. This model explains how C37-KYI inhibits EnvQ552R/N637K homotrimers with varying potencies despite all trimers having the same complement of C-peptide binding sites. Although the model ties receptor binding to g41 exposure, it is important to emphasize that our experiments did not directly measure CD4 interaction with Env and that all gp120 subunits may not engage CD4 prior to viral membrane fusion. The model does not address the role of chemokine receptor binding in coordinating structural changes, nor does it precisely define when the N-HR coiled coil forms during early Env activation. Current research in our lab is attempting to tackle these shortcomings.
Although the model portrays three CD4 interactions, Env trimers can function binding fewer CD4 molecules (and chemokine receptors) (Fig 2) [34–36, 69]. Despite having only one or two CD4 binding sites (and 2 or 1 CXCR4 binding sites, respectively), A2B and AB2 heterotrimers have robust fusogenic activities compared to EnvHXB2 homotrimers (S1A and S2A Figs). The proposed model explains how Wm and Mw viruses have such different sensitivities to T20 when A2B heterotrimers, with 2 CD4 binding sites, predominate (Fig 8B). Following the second CD4-gp120 interaction, the C-peptide binding site formed by N-HR helices from both Env A protomers is uncovered. When Env A protomers have wild type gp41 subunits, as in Wm viruses, the exposed site has a high affinity for C-peptides and the trimer is sensitive. Conversely, when the Env A protomers contain an escape mutation, as in Mw virus, the exposed site has a low affinity and the trimer is resistant. Full exposure of the remaining two C-peptide binding sites requires spontaneous, CD4-independent detachment of the final gp120 subunit, but their impact on potency is limited by the short lifetime of this state.
The proposed model may also explain how Wm and Mw viruses show similar sensitivities to T20 when AB2 heterotrimers, with only 1 CD4 binding site, predominate. Exposure of any C-peptide binding site on these trimers requires spontaneous detachment of at least one Env B gp120 subunit from the N-HR coiled coil (Fig 8C). The similar sensitivities of Wm and Mw viruses imply that wild type and mutant C-peptide binding sites are equally accessible during the PHI regardless of whether the escape mutation is introduced into Env A or Env B. Equal accessibility would be achieved if both Env B gp120 subunits detached nearly simultaneously (Fig 8C). The instability of the resulting fully exposed gp41 trimer would then limit T20 binding, rendering AB2 trimers on Wm and Mw viruses similarly resistant. Alternatively, spontaneous detachment may not be as unidirectional as illustrated in the model, and unliganded gp120 subunits might repeatedly detach and reattach during the lifetime of the PHI. In this case, exposure of high and low affinity C-peptide binding sites would be equal from a time-averaged perspective (as opposed to the spatial perspective as proposed above). Additional experiments will be required to distinguish these possibilities.
Interestingly, N-HR regions from other class 1 viral fusion glycoproteins are segmented in their native, prefusion conformation similar to HIV-1 Env. In the hemagglutinin (HA) protein of influenza virus, each N-HR region is broken into two antiparallel alpha-helices, with the C-terminal helix participating in a coiled coil and the N-terminal helix encapsulated by the surface subunit to maintain the metastable conformation (S9A Fig) [70]. In glycoprotein GP from Ebola virus, the N-HR region is multiply segmented, forming a C-terminal helix that participates in a short coiled coil while a number of preceding helical and extended fragments snake around the surface subunit (S9B Fig) [71]. Like gp120, the surface subunits of HA and GP would need to disengage from the N-terminal regions of the transmembrane subunits in order for the N-HR coiled coil to form. Since a disulfide bond crosslinks the surface and transmembrane subunits of HA and GP, disengagement would not lead to surface subunit dissociation and shedding upon entry into the PHI. It remains to be seen if covalent intersubunit association helps to guide these fusion glycoproteins through symmetric conformational transformations, or if the three-fold symmetry of these structures is lost at least for a short time early in the fusion process, as suggested for HIV-1 Env.
In developing an HIV-1 vaccine, the gp41 N-HR coiled coil has always been imagined as a viable target for neutralizing antibodies based on how well C-peptides inhibit viral entry [33]. Toward this goal, vaccination and selection experiments employing engineered forms of the N-HR coiled coil have been successful at generating antibodies that strongly interact with the antigens [72–77]. However, these studies have largely failed to identify an antibody with broad and potent neutralizing activity. Sterically restricted access to the N-HR coiled coil has been implicated as one potential reason for poor neutralization activity [57]. Our results here suggest a more fundamental problem with the antigen itself. The symmetrically exposed N-HR coiled coil that mimics the antigenic structure appears to be available for only a short period during the PHI. During the state primarily targeted by C-peptides, the N-HR coiled coil seems to be asymmetrically exposed, with one binding site uncovered and two binding sites partially occluded. Antibodies recognizing this conformation should have a better chance of binding gp41 during the PHI. Generating such antibodies might require engineering an asymmetric antigen containing an N-HR coiled coil and, perhaps, one or two attached, unliganded gp120 molecules.
The following cell lines and reagents were obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH: U87.CD4.CXCR4 from Drs. HongKui Deng and Dr. Dan R. Littman [79]; HOS-CXCR4+ from Dr. Nathaniel Landau [78]; pCAGGS_SF162gp160 (expression vector for EnvSF162) from Drs. L. Stamatatos and C. Cheng-Mayer [80]; Chessie 8 from Dr. George Lewis [81]; HIV-Ig (Catalog number 3957) from NABI and NHLBI. HEK 293T cells and hybridomas for anti-gp120 antibodies 55–83 and 46–2 were obtained from the ATCC. Entry inhibitor PIE12 was a kind gift of Dr. Michael Kay (University of Utah) [66].
Variants of the HIV-1HXB2 env gene were generated in the expression plasmid pEBB_EnvHXB2 [82]. To make the Env A protomer for functional complementation experiments, a DNA cassette of the V3-loop sequence from EnvSF162 was produced by PCR (Pfu Polymerase, Stratagene) using pCAGGS_SF162gp160. The cassette was then utilized in a QuikChange reaction (Stratagene) to swap the V3-loop sequence from EnvHXB2 with that from EnvSF162. The aligned protein sequences of the two gp120 V3 loops are listed below with amino acid differences underlined:
HXB2 (residues 296–331): CTRPNNNTRKRIRIQRGPGRAFVTIGK-IGNMRQAHC
SF162 (residues 294–327): CTRPNNNTRKSITI--GPGRAFYATGDIIGDIRQAHC
Additional point mutations were made through QuikChange PCR using standard overlapping oligonucleotides (Integrated DNA Technologies):
All designed Env constructs were confirmed by DNA sequencing the entire open-reading frame (Kimmel Cancer Center Cancer Genomics Facility, Thomas Jefferson University). Plasmid DNA was prepared for transfection using Promega purification kits and DNA concentrations were quantified by absorbance at 260 nm.
To ensure the accuracy of Env expression ratios for functional complementation experiments, we assessed the impact of sequence modifications on the cellular expression of Env A, Env B and Env variants containing T20 or 5HLAVA escape mutations. Env expression was assessed in 293T cells cotransfected (Lipofectamine, Life Sciences) with the Env-deficient HIV-1NL4-3 genome (pNL4-3R-E-Luc+) [82] and either one or two Env-expressing plasmids. Virus-containing supernatants were harvested 48 hours post-transfection and characterized as described below. Cells were washed extensively with phosphate buffered saline (PBS, pH 7.4) to remove uncollected virus and cellular debris before being lysed with 1% Triton in PBS supplemented with protease inhibitors (Complete tablets, Roche). Lysates were clarified by centrifugation (10,000 x g, 10 minutes) and the HIV-1 p24 antigen content was determined by ELISA (Aalto, Ireland). Samples of equal p24 protein content were separated by SDS-PAGE, transferred to nitrocellulose paper (Novex, Life Sciences; Hybond C, GE), and probed with 1) monoclonal antibody Chessie 8, which recognizes an epitope in the gp41 cytoplasmic tail; 2) monoclonal antibodies 55–83 and 46–2, which recognize the C1 and C5 constant regions of gp120, respectively; and 3) primary polyclonal HIV-Ig, which detects, among other viral proteins, p24 capsid. Western blots were developed with an HRP-conjugated secondary antibody (Jackson ImmunoResearch) and ECL reagent kit (Pierce). Env A, Env B and Env variants containing L544S, V549A or N656D mutations were found to express at the same levels as wild type EnvHXB2 (S1, S2 and S3 Figs).
HIV-1 particles were prepared from 293T cells (5x106) transfected with a constant amount of pNL4-3R-E-Luc+ DNA (2.5 μg) and one or two Env expressing plasmids (2.5μg total). For functional complementation experiments, plasmids encoding Env A and Env B constructs were premixed at molar ratios that equaled the indicated expression ratios. Supernatants containing viral particles were harvested 48 hours post-transfection and centrifuged (2500 x g, 10 minutes) to remove cellular debris. Virus was further purified through a 20% sucrose cushion [PBS, pH 7.4, with 2% w/v bovine serum albumin (BSA)] by ultracentrifugation (Beckman L-80, SW 50.1 rotor, 150,000 x g, 70 minutes, 4°C). Pellets were resuspended in PBS with 2% BSA, and viral titers were quantified by p24 antigen ELISA. To assess Env incorporation into virus, samples containing equal p24 content were separated by SDS-PAGE and analyzed by Western Blot as described above. The Env modifications used in this study did not alter the incorporation of Env into the HIV-1 particles (Fig 2B). To assess the fusogenic activities of different Env trimer populations, viral samples were applied to U87.CD4.CXCR4 target cells and infectivity was quantified 48 hours later by expression of a luciferase reporter construct (Promega, Pierce). Luciferase activity was normalized by the p24 content of the input viral sample in order to determine absolute fusogenicity.
C-peptide T20 was synthesized using standard Fmoc chemistry by the peptide synthesis facility at the Kimmel Cancer Center, Thomas Jefferson University. Cleaved, desalted peptides were purified to homogeneity by reverse phase high pressure liquid chromatography (rp-HPLC) using a Vydac C-18 column and a linear gradient of acetonitrile in water containing 0.1% trifluoroacetic acid. The identity of T20 was confirmed using MALDI-TOF mass spectrometry.
C-peptides C37, C37-KYI and di-C37 were prepared from a recombinant gp41 TOH construct NC1 as previously described [44, 55]. NC1 contains a 40 residue N-HR segment (amino acids 543–582, EnvHXB2 sequence), a trypsin-cleavable linker (GGRGG), a 37-residue C-HR segment (625–661) and a C-terminal hexahistidine tag (GHHHHHH). Briefly, the protein was expressed in E. coli RP3098 and purified from lysates using metal chelate chromatography (Ni-NTA agarose, Qiagen). The trimeric protein was subject to trypsin digestion (Sigma, 1:250 mass ratio, overnight at 4°C), and C37 was purified to homogeneity using rp-HPLC as described above. C37-KYI contains the affinity enhancing mutations N637K/T639I. Di-C37 is a dimerized form of C37 connected via an engineered disulfide bond at the C-terminus of each monomer.
5-Helix proteins were recombinantly expressed and purified from E. coli RP3098 as previously described [44, 50]. 5-Helix contains three N40 segments and two C37 segments alternately connected into a single polypeptide. The proteins were solubilized from bacterial inclusion bodies using 8 M guanidine HCl (GdnHCl) in tris-buffered saline (TBS, pH 8), loaded onto Ni-NTA agarose beads, and renatured in 4 M GdnHCl by a reverse thermal gradient (90°C to room temperature over 4 hours). Eluted proteins were subject to size exclusion chromatography (Superdex 75, GE) to separate monomers from aggregates. 5HLAVA contains the affinity-disrupting L556A/V570A mutations in all three N-HR segments.
The concentrations of T20, PIE12, C37 constructs and 5-Helix proteins were determined by absorbance at 280 nm by the method of Edelhoch [83]. The interaction and inhibition properties of these inhibitors have been extensively characterized [50, 55, 66]. Inhibitory titrations were performed using pseudotyped HIV-1 (see above) and either U87.CD4.CXCR4 or HOS-CXCR4+ target cells. Unless otherwise noted, the dependence of infectivity (measured by luciferase activity in target cells) on inhibitor concentration was fit to a Langmuir equation to extract IC50 values.
Crystal structures were obtained using NC1 proteins (see above) containing either the double mutation Q552R/L555M or single mutation V549E. These mutations were detected in HIV-1NL4-3 propagated in the presence of C37 and C37-KYI. The Q552R and V549E mutations disrupt C-peptide binding affinity approximately 5000-fold and confer resistance to C37 (150-fold) and the higher affinity C37-KYI (6-fold) [55]. The L555M mutation impacts neither the affinity nor potency of C37 or C37-KYI either alone or in combination with Q552R. The mutations were incorporated into the NC1 expression plasmid by Quikchange PCR (Stratagene). Proteins were recombinantly expressed and purified by metal chelate chromatography followed by rp-HPLC (described above). Lyophilized proteins were resuspended in water at a concentration of 12–13 mg/mL and ultracentrifuged at 150,000 x g (30 min, 4°C) to remove particulates.
Crystallization conditions were screened (Index HT and Crystal Screen HT, Hampton Research) by sitting-drop vapor diffusion method using robotically dispensed 100 nL sample volumes (HydraIIPlusOne, Thermo Scientific). Following optimization, diffraction quality crystals with hexagonal prism shape of 150 μm were obtained for NC1Q552R/L555M using 20% w/v PEG 8000, 0.3 M Calcium Acetate, 0.1 M Na Cacodylate, (pH 6.5) and 6.35mg/mL of protein. NC1V549E crystallized into a hexagonal prism of 50 μm in 3 M Sodium Chloride, 0.1 M Sodium Acetate (pH 4.5) and 6 mg/mL of protein. Crystals were frozen in mother liquor supplemented with 27% v/v ethylene glycol and stored in liquid nitrogen until exposed.
Diffraction data sets were collected on beamline's X6A at the National Synchrotron Light Source (NSLS, USA) and 23ID-B at the Advanced Photon Source, Argonne National Laboratory. The images were indexed and integrated using DIALS [84] and scaled using AIMLESS [85]. The structures were solved by molecular replacement using Phaser [86] with the HIV gp41 core structure (PDB ID: 1AIK, [12]) as a search model. The manual rebuilding of the model was performed in Coot [87] and refined in PHENIX ver. 1–10.1–2155 [88]. All crystallographic data are summarized in S2 Table. The structures were validated using ADIT Validation Server as implemented on the RSCB PDB website. Structural illustrations were prepared with PyMOL [89] and Coot.
|
10.1371/journal.pgen.1003712 | A Critical Function of Mad2l2 in Primordial Germ Cell Development of Mice | The development of primordial germ cells (PGCs) involves several waves of epigenetic reprogramming. A major step is following specification and involves the transition from the stably suppressive histone modification H3K9me2 to the more flexible, still repressive H3K27me3, while PGCs are arrested in G2 phase of their cycle. The significance and underlying molecular mechanism of this transition were so far unknown. Here, we generated mutant mice for the Mad2l2 (Mad2B, Rev7) gene product, and found that they are infertile in both males and females. We demonstrated that Mad2l2 is essential for PGC, but not somatic development. PGCs were specified normally in Mad2l2−/− embryos, but became eliminated by apoptosis during the subsequent phase of epigenetic reprogramming. A majority of knockout PGCs failed to arrest in the G2 phase, and did not switch from a H3K9me2 to a H3K27me3 configuration. By the analysis of transfected fibroblasts we found that the interaction of Mad2l2 with the histone methyltransferases G9a and GLP lead to a downregulation of H3K9me2. The inhibitory binding of Mad2l2 to Cyclin dependent kinase 1 (Cdk1) could arrest the cell cycle in the G2 phase, and also allowed another histone methyltransferase, Ezh2, to upregulate H3K27me3. Together, these results demonstrate the potential of Mad2l2 in the regulation of both cell cycle and the epigenetic status. The function of Mad2l2 is essential in PGCs, and thus of high relevance for fertility.
| Primordial germ cells (PGCs) are the origin of sperm and oocytes, and are responsible for transferring genetic information to the next generation faithfully. PGCs are first specified from pluripotent epiblast cells early in embryonic development. Second, they reprogram their epigenetic signature by changing histone modifications. This developmental event is specific to germ cells but not somatic cells. Although many players in the specification of PGCs are identified, only little is known about the genes essential for the regulation of the second phase. Here, we report that the Mad2l2 gene product plays an important role in the epigenetic reprogramming of PGCs. In wild type PGCs the cell cycle is arrested, and the methylation of histone 3 on residue K9 is replaced by methylation on K27. Our findings indicate that Mad2l2 is involved in this coordination of cell cycle and epigenetic reprogramming. The elucidation of this mechanism would help to identify the genetic basis of infertility.
| In mice, PGCs are induced by BMP signaling at the onset of gastrulation at day 7.25 of embryonic development (E7.25) in the posterior epiblast. They enter the extraembryonic mesoderm and the hindgut endoderm, and then migrate through the dorsal mesentery, until they accumulate in the genital ridges to participate in the generation of the future gonads [1]. Once specified, PGCs undergo various changes of their transcriptional profile and epigenetic status, which together establish the unique germ cell fate separate from surrounding somatic cells [2], [3]. Two PR-domain containing proteins, Prdm1 (Blimp1) and Prdm14, initiate the PGC-specific program [4], [5]. The reactivation of the pluripotency-associated gene Sox2 that had been silenced in the epiblast of the egg cylinder is an immediate early change upon PGC specification [6], [7]. It leads to the acquisition of a potential to become pluripotent under specific culture conditions [8]–[10]. Around E7.5 the transcription of somatic genes like Hox, Snail or Brachyury become repressed as a result of Prdm1 function, and the characteristic PGC gene Dppa3 becomes upregulated. Together, the typical transcriptional signature of PGCs has developed by E9.0 [11].
The chromatin of PGCs undergoes extensive remodeling, affecting both DNA and histone configurations [3], [12]. De novo DNA methylation is suppressed as the result of the downregulation of the DNA methyltransferases Dnmt3b and Uhrf1 [7]. Consequently, a passive DNA demethylation is initiated at around E8.0, and by E9.5, PGCs become hypomethylated [3]. At E7.75, PGCs harbor a high, genome-wide level of the repressive histone modification H3K9me2, similar to the surrounding somatic cells. This modification is gradually lost, and by E9.25 suppressed in most PGCs. The corresponding histone methyltransferases GLP and G9a, which methylate lysine residue 9 of histone 3, are downregulated by E7.5 or E9.0, respectively [11], [13]. In parallel to H3K9me2 downregulation, H3K27me3, a repressive histone modification providing more plasticity, accumulates in PGCs and finally replaces the H3K9me2 completely at E9.25 [2], [3], [11]. H3K27 trimethylation is catalyzed by Ezh2, a subunit of the polycomb repressive complex 2 (PRC2), and downregulates the expression of typical somatic or differentiation related genes [14], [15]. Ezh2 is subject to phosphorylation at different motifs by the cyclin dependent kinases Cdk1 or Cdk2, which modulate the activity or stability of Ezh2, and thus affect the level of H3K27me3 [16]–[18]. Cdk1/Cyclin B1-mediated phosphorylation of Ezh2 at threonin 487 (pEzh2-T487) disrupts its binding to the other components of PRC2 complex, leading to its inactivation, and therefore to H3K27me3 attenuation [18]. It was previously shown that murine and porcine PGCs, and also PGCs derived in vitro from mouse embryonic stem cells arrest their cell cycle in a G2 phase briefly after their specification [11], [19]–[21]. This phase, which is accompanied by transcriptional silence, may provide time for epigenetic reprogramming. So far, the molecular mechanism coordinating the epigenetic reprogramming and cell cycle prolongation in early PGCs is not clear.
Mad2l2 is a chromatin binding protein involved in both cell cycle control and DNA repair [22]–[24]. Mad2l2 was previously described as an accessory, non-catalytic subunit of the translesion DNA polymerase zeta, and its knockdown led to hypersensitivity towards DNA damage [25], [26]. Mad2l2 appears to function by binding to a diverse spectrum of proteins via its conserved HORMA domain. Several, but not all of these partners bind via the conserved sequence motif PXXXPP [27]. Reported binding partners include Cdh1 and Cdc20, the substrate binding proteins of the APC/C complex, the two translesion polymerases Rev1 and Rev3, the transcription factors Elk-1 and TCF4, the clathrin light chain A, and others [23], [24], [28]–[32]. Accordingly, functions for Mad2l2 were previously claimed in such diverse processes as DNA repair, cell cycle control, and the regulation of gene expression. However, the biological significance of the reported interactions and activities remained unclear due to the lack of appropriate mouse mutants.
In this work we describe a mouse mutant lacking the Mad2l2 gene. Embryos lose PGCs briefly after their specification, and do not proceed in epigenetic reprogramming. We investigated the function of Mad2l2 also by gain- and loss-of-function analysis in fibroblasts, and in biochemical assays. We suggest new functions of Mad2l2 as a regulator of epigenetic reprogramming, which is particularly relevant for primordial germ cells, and therefore required for fertility of males and females.
Low levels of Mad2l2 mRNA are widely expressed in adult and E14.5 embryonic cells, with a particularly high level in testis (Figure 1A). High levels of Mad2l2 protein were detected in pachytene spermatocytes by immunohistochemistry (Figure 1E), while the antibody did not lead to specific signals above background in other tissues, including PGCs. Significant amounts of Mad2l2 RNA were previously detected in E9.5 PGCs by microarray analysis (NCBI database Gene Expression Omnibus GEO; Hayashi et al., 2011).
A conditional knockout of the Mad2l2 gene in embryonic stem cells was generated and ubiquitously active Cre recombinase was introduced through breeding (Figures S1A, B). Heterozygous Mad2l2 mutants were viable, healthy and fertile. Homozygous embryos and postnatal mice were significantly smaller than their littermates, but no morphological abnormalities were observed (Figures S1C–F). Offspring before and after birth appeared in sub-Mendelian ratios, indicating a loss of embryos in midgestation (Table S1). Homozygous males and females were infertile, and gonads were significantly underdeveloped. Ovaries were not formed at all or were small organ rudiments that did not contain ovarian follicles or germ cells (Table S2 and Figure 1B). Such structures may be indicative that some germ cells were present in the gonad during granulosa cell differentiation (Figure 1B). Mutant testes were drastically smaller than control organs of the same age, and seminiferous tubules were devoid of spermatogonial cells (detected by Plzf), pre-meiotic (identified by Stra8) and meiotic cells (detected by γH2AX; Figure 1C,D,F–H) [33]–[36]. Leydig cells appeared hyperplastic, and Sertoli cells, identified by Wt1, were mislocalized and highly vacuolated (Figure 1I) [37], [38]. In summary, finding these deficiencies in both males and females suggested that developmental problems arose earlier during embryogenesis.
For the determination of PGC numbers, embryos were collected at different time points during their early development, were staged as outlined under experimental procedures, and PGCs were identified by the presence of alkaline phosphatase (AP) or Oct4 (Figure 2A) [39]. At the early head fold (EHF) stage, the numbers of PGCs at the base of the allantois were similar in wild type, heterozygous and homozygous embryos. However, while the number of normal PGCs increased at the late head fold (LHF) stage, the number of Mad2l2−/− PGCs fell behind (Figure 2B). It decreased drastically from E8.5 onward, and at E9.0 only few instead of normally ca. 120 PGCs were found in the hindgut endoderm. At E9.5 and E10.5 Oct4-positive PGCs were no longer detected (Figure 2B). At E8.25, both wild type and remaining mutant PGCs co-expressed Oct4 together with Prdm1, Tcfap2c, and Dppa3, indicating a normal specification of mutant PGCs (Figure S2A,B,D). Oct4 and Sox2 were co-expressed in all wild type PGCs with no exception. In contrast, above 40% of Oct4-positive Mad2l2−/− PGCs did not express Sox2 at E9.0, and thus had either failed to reactivate, or at least to maintain its expression (Figure S2C). Emigration to the dorsal mesentery did not occur, and as a result, gonad primordia at E13.5 were devoid of germ cells (Figure 2A). All E9.0 Mad2l2−/− PGCs had accumulated active, acetylated p53 protein, reflecting an activated stress response and impending apoptosis (Figure S3A) [40]. As judged by the TUNEL assay (See Text S1), some SSEA1-positive PGCs undergoing cell death were detected in E9.0 hindgut endoderm (Figure 2C). In addition, the same territory contained accumulations of SSEA1-negative, apoptotic cells. Based on their size we suspected them to be germ cells having lost already expression of their typical marker, although we could not exclude that they represented mutant somatic cells. In summary, Mad2l2−/− PGCs were specified normally, but their numbers decreased progressively, and no PGCs could be detected in Mad2l2−/− embryos beyond E9.5. This time window correlates with an epigenetic transition of PGCs and cell cycle arrest between E7.5-E9.5 [3], [11].
Proper development of PGCs relies on their endogenous program as well as on exogenous signals emanating from surrounding somatic cells that support their induction, migration or survival in various organisms [41]–[44]. To address the cause of early PGC loss in Mad2l2 deficient embryos, we employed a Prdm1-Cre mouse line, which would be expected to delete the Mad2l2 gene specifically in nascent PGCs [4]. The TUNEL assay demonstrated apoptosis in SSEA1-positive PGCs of Prdm1-Cre+, Mad2l2fl/fl embryos at E8.75 (Figure 3). In addition, TUNEL-positive, SSEA1-negative cells with a high nuclear to cytoplasmic ratio were observed in the hindgut. Also some TUNEL-negative, SSEA1-positive PGCs were found, which is explainable by the incomplete efficiency of Prdm1-Cre mediated deletion, although the actual recombination could not be confirmed here for the few available cells [4]. In contrast, no apoptosis was observed in Prdm1-Cre+, Mad2l2fl/+ PGCs of the same age, excluding toxic effect of Cre recombinase on PGCs [45]. Together, these findings demonstrate that Mad2l2 deficient PGCs did not survive even in a wild type somatic environment. Since Mad2l2 is the subunit of a repair DNA polymerase, we asked if Mad2l2 deficient PGCs are affected by DNA damage. We applied an antibody detecting phosphorylated ATM/ATR substrates (pATM/ATR-S) including Chk1, Chk2, and MDM2, as well as specific antibodies against pChk1 and pChk2, respectively. No double-positive PGCs were detected in either wild type or knockout embryos in such staining (Figure S3B–D). Together, these observations indicate that Mad2l2 deficient PGCs are not lost due to DNA damage.
Immediately after their induction in the epiblast, PGCs begin to undergo massive epigenetic reprogramming with regard to both DNA and histone modifications. The genome-wide demethylation of the DNA in PGCs is partially due to a downregulation DNA methyltransferases, which is accompanied by loss of cytidine methylation. To address the epigenetic reprogramming in Mad2l2−/− PGCs, first we performed whole mount staining (See Text S1) against Dnmt3b DNA methyltransferase. Both wild type and Mad2l2 deficient PGCs suppressed Dnmt3b expression (Figure 4A). Immunohistochemistry analysis of DNA methylation showed loss of the 5-methylcytosine (5 mC) at E9.0 in both wild type and knockout sections (Figure 4B). These observations seem to indicate that DNA hypomethylation had been properly initiated and progressed in the absence of Mad2l2.
In PGCs, the repressive histone H3K9me2 should become downregulated during the cell cycle arrest between E7.5 and E9.5. A comparison of stage-matched E9.0 embryos revealed that the majority of mutant, Oct4-positive PGCs had not downregulated H3K9me2, while wild type PGCs mostly had lost this histone modification (Figure 5A). Correspondingly, also G9a and GLP, two H3K9 methyltransferases, were still found in mutant, but not in wild type PGCs (Figure 5B,C; S4A,B).
Addressing the cell cycle profile of PGCs, we confirmed a cytoplasmic localization of Cyclin B1 in the majority of wild type PGCs on E9.0, indicating that they were in the G2 phase of the cell cycle (Figure 6) [11]. In Oct4-positive Mad2l2−/− PGCs, on the other hand, the Cyclin B1 protein was either localized in the nucleus, in the cytoplasm or not present at all (Figure 6). Thus, it appeared that mutant PGCs did not arrest in G2 phase of their cell cycle.
A highly elevated, global H3K27me3 modification could be confirmed for the majority of wild type PGCs, while levels in Mad2l2−/− PGCs were mostly indistinguishable from surrounding somatic cells (Figure 7A). Ezh2, the relevant methyltransferase for residue K27 of histone 3, is expressed in PGCs at a similar level to that of neighboring somatic cells, at least during their specification period [46]. However, we observed that the inactivation of Ezh2 was completely suppressed in the majority of wild type PGCs at E8.5, while above 60% of knockout PGCs contained high or low levels of such inactive Ezh2 protein (Figure 7B). Thus, a significant portion of the Mad2l2−/− PGCs failed to acquire an epigenetic status dominated by H3K27me3, probably due to presence of inactive phosphorylated Ezh2.
The number of early PGCs is too small for biochemical and transfection approaches. Therefore, we performed a set of experiments in fibroblasts with the intention to provide evidence for a function of Mad2l2 in epigenetic and cell cycle regulation. Since the Mad2l2 protein contains a protein-binding HORMA domain Co-immunoprecipitation was applied to identify Mad2l2 interacting partners related to histone modifications (See Text S1). First, to explore a physical interaction between Mad2l2 and G9a or GLP, NIH3T3 fibroblasts were transfected with a plasmid encoding HA-Mad2l2 (See Text S1). Co-immunoprecipitation of NIH3T3 protein extract with anti-G9a, anti-GLP or anti-HA antibodies demonstrated that Mad2l2 interacts with both methyltransferases (Figure 8A, B). Transfection of NIH3T3 cells with a vector encoding a GFP-fused Mad2l2 protein showed that G9a mRNA levels were specifically downregulated in the presence of GFP-Mad2l2 (Figures S5A). G9a protein levels were always low in Mad2l2-GFP transfected cells, while untransfected cells had either high or low levels (Figures 8C). Correspondingly, the level of H3K9me2 became completely suppressed in transfected cells (Figure 8C), while levels of H3K4me2, an unrelated histone modification, remained unaffected (Figure S5B). For the analysis of loss-of-function conditions Mad2l2 deficient MEFs were prepared, and elevated levels of G9a and H3K9me2 were observed (Figure 8D). Together, these findings indicate a negative correlation between the presence of Mad2l2 and the expression and activity of the methyltransferase G9a.
To test whether ectopic expression of Mad2l2 can arrest the cell cycle, NIH3T3 cells were transfected with a HA-Mad2l2 encoding vector. Expressing cells did not enter mitosis, as evident by the complete absence of pH 3 or Cyclin B1 from nuclei, as well as the presence of unseparated centrosomes (Figure 8E) [47], [48]. Several pathways regulating the entry into mitosis converge at the cyclin dependent kinase 1 (Cdk1), which needs to be dephosphorylated and associated with phosporylated Cyclin B1 to be active [49], [50]. We hypothesized that Mad2l2 might interact physically with Cdk1 or Cyclin B1 to regulate the G2/M transition. Protein lysate from HA-Mad2l2 transfected NIH3T3 cells was precipitated with antibodies against Cdk1, pCdk1 (phosphorylated Cdk1), Cyclin B1, and the HA-tag. Co-precipitate analysis revealed a physical interaction of Mad2l2 with Cdk1, but not pCdk1 or Cyclin B1 (Figure 8F–H). We then looked for a regulatory effect of Mad2l2 on the kinase activity of Cdk1/Cyclin B1 in an in vitro assay (See Text S1), containing recombinant GST-Mad2l2, Cyclin B1 and Cdk1, as well as the specific substrate Cdc7 [51]. GST-Mad2l2, but not GST alone could specifically attenuate the kinase activity of Cdk1-Cyclin B1 in a concentration-dependent manner (Figure 8I). Together, our experiments suggest that the ectopic presence of Mad2l2 prolongs the cell cycle.
To address whether Mad2l2 can principally be involved in H3K27me3 upregulation, gain-of-function experiments with a GFP-Mad2l2 fusion protein were performed in NIH3T3 cells. Immunocytochemistry showed a very high level of H3K27me3 in all GFP-positive cells, while surrounding untransfected cells had mostly low levels, with some exceptions possibly dependent on the state of their cell cycle (Figure 8J). Given the inhibitory function of Mad2l2 on the kinase activity of Cdk1, we asked if it might attenuate the inhibitory phosphorylation of Ezh2 (Figure 8K, L). The highest level of pEzh2 was observed in mitotic cells correlating with the highest activity of Cdk1/Cyclin B1 (Figure 8K) [18]. In contrast, Mad2l2 over-expressing cells showed the lowest level of pEzh2, even less than that in untransfected interphase cells (Figure 8K). Consistently, western blot analysis confirmed the drastic suppression of pEzh2 in Mad2l2 over-expressing FACS-sorted fibroblasts, while the overall level of Ezh2 itself remained unchanged (Figure 8K). The loss-of-function situation was analyzed in Mad2l2 deficient MEFs, which showed an increased level of pEzh2, while the amount of H3K27me3 was decreased (Figure 8L). Apparently, here the Cdk1/Cyclin B1 was active, and could phosphorylate and thereby inactivate Ezh2. Our analysis of fibroblasts and of a cell free system demonstrate the capacity of Mad2l2 to suppress the kinase activity of Cdk1/Cyclin B1, and thus to support the activity of Ezh2 and by that promote the tri-methylation of histone 3 on K27.
Several mutations are known to affect or terminate the development of PGCs (for review see [44]). In principal, every step proved to be sensible, particularly the primary induction by BMP signaling, the early specification, the migration to the developing gonad, and the pre- or postnatal oogenesis or spermatogenesis. The early BMP response genes, Prdm1 and Prdm14, are crucial for PGC specification directly after induction, where numbers of mutant PGCs are drastically reduced already on E8.0, and only few mutant PGCs survive to E9.5 [4], [5]. Similar kinetics for PGC loss were observed in mice lacking the transcription factor Tcfap2c, which mostly phenocopy the Prdm1−/− mice [52]. A slightly later timing, shifted by about one day, was found for the Mad2l2 mutants in our study. Although embryos at EHF stage were relatively small, they harbored stage-adequate numbers of PGCs expressing Prdm1 and the commitment markers Dppa3 and Tcfap2c arguing for a normal specification in the epiblast. A reduction of PGC numbers was observed in the LHF stage, and there was no survival beyond E9.5. At this point of development, PGCs would normally have undergone a major epigenetic reprogramming, would recover from their cell cycle arrest, and resume transcription. This timing suggests a failure of epigenetic reprogramming and cell cycle arrest in Mad2l2−/− PGCs. In principle, it is conceivable that wrongly developed PGCs might either revert to a somatic fate, or undergo apoptosis. PGCs are lost without evidence for apoptosis in mutants of the Prdm1, the Prdm14, and the Tcfap2c gene, whereas mutations in the Oct4, the Kit and the Mad2l2 genes remove wrongly programmed PGCs by apoptosis [4], [5], [52]–[54]. Somatic Mad2l2−/− cells apparently do not rely on a specific epigenetic reprogramming and cell cycle arrest, and at least some Mad2l2-deficient mice develop normally and live until adulthood. Still, mutants are born in sub-Mendelian ratio and adults are usually smaller, as is the case in many mutant mice. Together, this points to a highly specialized function of Mad2l2 in the unique development of germ cells, but does not exclude lower penetrance effects in somatic cells.
H3K9 methylation is critical for formation of heterochromatin and transcriptional silencing. At the onset of PGC development, H3K9me2 is the dominant epigenetic mark in the genome of embryonic cells [3], [11]. This modification requires the activity of the two methyltransferases G9a and GLP [55]. G9a, the major mammalian H3K9 methyltransferase, plays a critical role in germ cell development, particularly in gametogenesis. The specific deletion of G9a in PGCs after E9.5 leads to germ cell loss during the meiotic prophase, and thus to sterility of both males and females [56]. During the S phase of the cell cycle, G9a binds to DNA methyltransferase DNMT1 and loads on to the DNA at replication foci, ensuring a coordination of DNA methylation and H3K9 methylation in heterochromatin regions [57]. Nascent PGCs leave asynchronously the S phase of their cycle and enter G2 at around E8.0. At this time, the de novo methylation of the daughter chromatin is suppressed, and both Prdm1 and Prdm14 were suggested to be involved [58], [59]. In parallel, the maintained activity of histone demethylases like Jmjd1a erases further the remaining H3K9me2 [60]. Our results indicate that similar to Prdm14 deficient PGCs, the majority of Mad2l2−/− PGCs fail to suppress H3K9me2. The maintenance of a high H3K9me2 level in Prdm14 mutant PGCs was attributed to a failure in downregulation of GLP. Released from repression by genome-wide H3K9me2, PGCs repress RNA Pol-II dependent de novo transcription until they acquire the alternative repressive histone mark, H3K27me3. This probably ensures the maintenance of separate PGC and somatic programs, established previously via combinational functions of Prdm1, Prdm14, and Tcfap2c [61]. A significant portion, but not all, of the Mad2l2−/− PGCs failed to proceed with their epigenetic reprogramming, as it is the case in Prdm14 mutant PGCs. Obviously, shortly before their elimination around E9.0, the Mad2l2−/− PGCs represent a heterogeneous population with respect to their transcriptional and epigenetic status. Thus, Mad2l2 is absolutely essential for the development of PGCs.
We observed that Mad2l2 suppresses G9a on the level of gene expression, which could be related to its ability to interact with transcription factors [29], [32]. The binding of Mad2l2 to the two histone methyltransferases G9a and GLP was previously identified in a systematic analysis of human protein complexes, and represented a first hint for an involvement of Mad2l2 in the generation of epigenetic modifications [62]. We confirmed this evidence by co-immunoprecipitation of both G9a and GLP with HA-Mad2l2 from transfected fibroblasts, where the level of H3K9me2 was significantly downregulated. Noteworthy, both G9a (PXXXPP) and GLP (PXXXyP) have the sequence motif suggested to be responsible for Mad2l2 binding [27]. G9a and GLP form homo- and heteromeric complexes in vitro, which are necessary for histone methyltransferase activity [13], [55]. Indeed, several proteins, bind to G9a or GLP, and alter their activities [63], [64]. Among those is Prdm1, which binds to G9a and recruits it to assemble silent chromatin [65]. Similarly, the direct interaction between Mad2l2 and G9a or GLP may disrupt formation of the G9a-GLP active heterodimer complex, and thus suppress the methylation of histone 3. Supportive evidence for such an inhibitory binding comes from the negative correlation between Mad2l2 and H3K9me2 levels in PGCs (Fig. 5A) and fibroblasts (Fig. 8D). However, the actual significance of the observed protein-protein interactions needs further investigation.
Cdk1 is a regulatory kinase of central importance for several processes, in particular also in cell cycle control and in epigenetic reprogramming [66], [67]. Our study in transfected fibroblasts and in a cell-free system suggests that Mad2l2 can bind directly to dephosporylated Cdk1, and thus inhibit its kinase activity. Possibly this interaction involves the Cdk1 sequence PXXXPy, which is related to the previously identified Mad2l2 binding motif PXXXPP [27]. The entry into mitosis is mediated by a complex network of proteins that finally activate the Cdk1-Cyclin B1 complex [50]. One of the first functions of Cdk1-Cyclin B1 is the phosphorylation and therefore disruption of Eg5, a protein involved in centrosome adhesion [68]. Overexpression of Mad2l2 abrogated centrosome separation, and caused a cell cycle arrest at the G2 phase. Dephosphorylated Cdk1 in association with phosphorylated Cyclin B1 translocate to the nucleus and initiates prophase by the phosphorylation of a variety of substrates [50]. Thus, via direct binding to Cdk1, Mad2l2 would have the capacity to inhibit Cdk1-Cyclin B1 complex formation, and thus to block the entry into mitosis. Inhibition and/or disruption of the Cdk1-Cyclin B1 complex through direct interaction were previously also observed for Gadd45 proteins, stress factors implicated in the activation of the G2/M DNA damage checkpoint [51], [69], [70]. Previous analyses of Mad2l2 had indicated inhibitory interactions with Cdh1, and possibly also with Cdc20 [23], [24]. These proteins would normally exert their function only after the onset of mitosis, either as part of the spindle assembly checkpoint, or as the substrate recognizing protein of the APC/C protein ubiquitination complex, respectively. However, early knockout PGCs divide relatively normal and only fail to arrest in the G2 phase. Therefore, it is less likely that Mad2l2 functions in mitosis of PGCs via binding to Cdh1, or Cdc20. Overexpression in fibroblasts indicated the possibility that Mad2l2 can be involved in a G2 arrest. This might correlate with the G2 arrest, which coincides with the epigenetic transition of PGCs from a H3K9me2 to a H3K27me3 configuration, and with the timing of PGC loss in Mad2l2 mutants. Among the many functions of the widely distributed kinase Cdk1 is the inhibition of the histone 3 methyltransferase Ezh2 by phosphorylation [66], [67]. Our analysis in fibroblasts indicates that Mad2l2 can interfere with this inactivation, and thus in effect, promote the activation of Ezh2. Consequently, we observed an increase of H3K27me3 levels upon overexpression of Mad2l2. Our data do not allow at present to decide if the primary defect in knockout PGCs lies in the regulation of the cell cycle, if the epigenetic failure precedes misregulation of the cycle, or if the two tightly coupled processes are not separable. Nevertheless, the outcome is that Mad2l2 mutated PGCs are not able to make the developmental transition from E7.5 to E9.5, and are quickly eliminated from the embryo (Figure 9). Thus, Mad2l2 is absolutely required for the development of PGCs, and thus for fertility.
While this manuscript was under revision, a related set of data was published demonstrating the necessity of Mad2l2 for PGC maintenance [71]. However, detailed characterization of knockout PGCs and the mechanism by which Mad2l2 may function were not studied.
All animal works have been conducted according to relevant national and international guidelines.
Genomic sequences were amplified from a 129 strain mouse PAC clone. The vector was assembled using the recombineering protocol and materials as described (see Figure S1; [72]. The loxP sites were introduced 113 bp upstream of the first coding exon, and 20 bp dowstream of the last exon, deleting finally a region of 5330 bp. The vector was introduced into MPI-II ES cells, which were selected with G418 and Ganciclovir. Cells with homologous recombination were aggregated with morula-stage embryos. The Mad2l2 gene was inactivated by crossing of heterozygote mice with CMV-Cre mice [73], and then breeding to homozygocity. Genotyping was performed using the primers
#1 (GCTCTTATTGCCTTGACATGTGGCTGC),
#2 (GGACACTCAGTTCTGGAAAGGCTGG), and
#3 (CTGCAGCCCAATTCCGATCATATTCAATAAC).
The day of the vaginal plug was taken as E0.5, and embryos were dissected accordingly. Embryos were staged [11] by corresponding time and morphology as follows: before E8.0 (EHF), E8.0 (LHF), E8.25 (less than 5 somites), E8.5 (before turning, 6 to 8 somites), E8.75 (turning embryos, 10 to 12 somites), E9.0, (after turning, 14 to 18 somites, with only the first branchial arch obvious, and with open otic vesicles, E9.5 (two branchial arches, closed otic vesicles, 20–24 somites).
The following antibodies were used. Rabbit anti-Cyclin B1 (Sigma-Aldrich), 1∶100; mouse anti-phospho-Histone H3 (ser10; Cell Signaling), 1∶200; rat anti-HA (Roche), 1∶100; mouse anti-γTubulin (Abcam), 1∶200; mouse anti-Cdk1 (Santa Cruz), 1∶50; rabbit anti-pCdk1 (Cell Signaling), 1∶50; mouse anti-Oct4 (BD), 1∶100; rabbit anti-Oct4 (Abcam), 1∶100; mouse anti-SSEA1 (Santa Cruz), 1∶100; rabbit anti-Nanog (abcam), 1∶100; rabbit anti-Sox2 (Millipore), 1∶200; rabbit anti-H3K9me2 (Upstate) 1∶100; and (Millipore), 1∶100; rabbit anti-G9a (Cell Signaling), 1∶25; mouse anti-GLP (Abcam), 1∶50; rabbit anti-Mad2l2 (Abcam), 1∶100; mouse anti-γH2AX (Millipore), 1∶200; rabbit anti-pChk2 (Cell Signaling), 1∶200; mouse anti-Vimentin (gift of M. Osborn), 1∶100; rabbit anti-WT1 (Abcam), 1∶1000; rabbit anti-Ezh2 (Cell Signaling), 1∶2000; rabbit anti-pEzh2 T487 (Epitomics), 1∶1000; rabbit anti-H3K4me2 (Active Motif), 1∶100; rabbit anti-H3K27me3 (Active Motif), 1∶100; rabbit anti-Dppa3 (abcam), 1∶500; rabbit anti-Stra8 (abcam), 1∶2000; rabbit anti-Plzf (abcam), 1∶100; rabbit anti-Dnmt3b (abcam), 1∶100; rabbit anti-Tcfap2c (Santa Cruz), 1∶100; mouse anti-5mC (abcam), 1∶200.
GST-fused Mad2l2 protein was expressed in and purified from E. coli. Full length Mad2l2 cDNA was cloned in frame with the N-terminal GST-tag into the pGEX-KT vector. Expression was induced by the addition of 1 mM IPTG (isopropyl-β-D-thiogalactopyranoside, Sigma). Bacterial cells were harvested; proteins were lysed on ice in 50 mM Tris, pH 7.5, 500 mM NaCl, 2 mM EDTA, 5 mM DTT, 10% glycerol, freshly added 1 mM PMSF and Complete EDTA-free protease inhibitor cocktail tablet (Roche). Glutathione Sepharose 4B (Amersham Biosciences) was used to purify the GST-fused protein. The elution was done twice, each time with 2 ml elution buffer (500 mM Tris, pH 8.0, 100 mM Glutathione supplemented with protease inhibitor). The protein was dialyzed in dialysis buffer (20 mM Tris-Cl pH 7.5) using a dialysis cassettes (Pierce) at 4°C overnight. The protein concentrations were measured and determined according to the standard curve.
Kinase activity of Cdk1-cyclin B1 was analyzed using purified, recombinant proteins (CycLex), and a human Cdc7 peptide as substrate, applying an assay system from CycLex [51]. To test effect of Mad2l2 on kinase activity of Cdk1-Cyclin B1, dilutions of GST-Mad2l2 or GST alone protein were incubated for 15 min at 37°C with 12.5 mUnits of recombinant kinase. These protein mixes were individually given into substrate-coated wells, and incubated for 45 min at 37°C. For detection of phospho-Cdc7 a specific monoclonal antibody (TK-3H7) and HRP-conjugated anti-mouse IgG was applied, and the absorbance at 450 nm was measured.
|
10.1371/journal.ppat.1004413 | The pH-Responsive PacC Transcription Factor of Aspergillus fumigatus Governs Epithelial Entry and Tissue Invasion during Pulmonary Aspergillosis | Destruction of the pulmonary epithelium is a major feature of lung diseases caused by the mould pathogen Aspergillus fumigatus. Although it is widely postulated that tissue invasion is governed by fungal proteases, A. fumigatus mutants lacking individual or multiple enzymes remain fully invasive, suggesting a concomitant requirement for other pathogenic activities during host invasion. In this study we discovered, and exploited, a novel, tissue non-invasive, phenotype in A. fumigatus mutants lacking the pH-responsive transcription factor PacC. Our study revealed a novel mode of epithelial entry, occurring in a cell wall-dependent manner prior to protease production, and via the Dectin-1 β-glucan receptor. ΔpacC mutants are defective in both contact-mediated epithelial entry and protease expression, and significantly attenuated for pathogenicity in leukopenic mice. We combined murine infection modelling, in vivo transcriptomics, and in vitro infections of human alveolar epithelia, to delineate two major, and sequentially acting, PacC-dependent processes impacting epithelial integrity in vitro and tissue invasion in the whole animal. We demonstrate that A. fumigatus spores and germlings are internalised by epithelial cells in a contact-, actin-, cell wall- and Dectin-1 dependent manner and ΔpacC mutants, which aberrantly remodel the cell wall during germinative growth, are unable to gain entry into epithelial cells, both in vitro and in vivo. We further show that PacC acts as a global transcriptional regulator of secreted molecules during growth in the leukopenic mammalian lung, and profile the full cohort of secreted gene products expressed during invasive infection. Our study reveals a combinatorial mode of tissue entry dependent upon sequential, and mechanistically distinct, perturbations of the pulmonary epithelium and demonstrates, for the first time a protective role for Dectin-1 blockade in epithelial defences. Infecting ΔpacC mutants are hypersensitive to cell wall-active antifungal agents highlighting the value of PacC signalling as a target for antifungal therapy.
| Inhaled spores of the pathogenic mould Aspergillus fumigatus cause fungal lung infections in humans having immune defects. A. fumigatus spores germinate within the immunocompromised lung, producing invasively growing, elongated cells called hyphae. Hyphae degrade the surrounding pulmonary tissue, a process thought to be caused by secreted fungal enzymes; however, A. fumigatus mutants lacking one or more protease activities retain fully invasive phenotypes in mouse models of disease. Here we report the first discovery of a non-invasive A. fumigatus mutant, which lacks a pH-responsive transcription factor PacC. Using global transcriptional profiling of wild type and mutant isolates, and in vitro pulmonary invasion assays, we established that loss of PacC leads to a compound non-invasive phenotype characterised by deficits in both contact-mediated epithelial entry and protease expression. Consistent with an important role for epithelial entry in promoting invasive disease in mammalian tissues, PacC mutants remain surface-localised on mammalian epithelia, both in vitro and in vivo. Our study sets a new precedent for involvement of both host and pathogen activities in promoting epithelial invasion by A. fumigatus and supports a model wherein fungal protease activity acting subsequently to, or in parallel with, host-mediated epithelial entry provides the mechanistic basis for tissue invasion.
| Spores of the mould pathogen Aspergillus fumigatus are agents of multiple human diseases, most of which initiate with inhalation of fungal spores and, dependent upon host immune status, compromise pulmonary integrity. Amongst the resultant diseases, invasive aspergillosis (IA) exerts the highest fatal toll, resulting globally in an estimated 200,000 deaths per annum [1]. Recipients of allogenic hematopoietic stem cell- or solid organ transplants, are particularly susceptible to IA which accounts for 43% and 19% of all invasive fungal infections in these cohorts and causes 58% and 34% mortality, respectively, at 12 weeks post-transplant [2]–[4]. Structural or immunological lung defects also lead to chronic, semi-invasive, pulmonary aspergillosis (CPA) having a 5 year mortality of 50% [5]. Amongst more than 200 Aspergillus species, Aspergillus fumigatus accounts for the majority of these diseases [6].
In diseases caused by A. fumigatus the initiating host-pathogen interaction occurs at pulmonary epithelia where inhaled spores can exit from dormancy, swell and generate invasive cells called hyphae, which traverse the lung epithelium. A key pathological feature of invasive- and semi-invasive aspergilloses is the destruction of the lung parenchyma, hypothesised to be governed by proteolytic enzymes secreted by the invading pathogen. Exposure of in vitro-cultured bronchial, or alveolar, epithelial cells (ECs) to fungal culture supernatants has revealed a role for fungal proteases in destruction of the mammalian F-actin cytoskeleton and loss of focal adhesion [7]–[9]. However, in whole animal studies of disease, it has not been possible to attribute lung damage solely to the activity of fungal proteases since A. fumigatus mutants lacking individual, or multiple, enzymes retain the ability to cause fatal invasive infections in immunocompromised hosts [9]–[15].
The interaction of A. fumigatus spores with alveolar epithelia can result in the internalisation of spores [16]–[18] but the role of this process in disease outcome remains unknown. Cells of the A549 pneumocyte cell line [18] and 16HBE14o- transformed human bronchial epithelial cells [19] internalise 30–50% of encountered spores, via an actin-dependent mechanism. Whilst the vast majority of internalised spores are killed, a small proportion (∼3%) survives and germinates inside acidic organelles [20]. This has prompted hypotheses of latent occupation of host epithelia by A. fumigatus spores, which might thereby evade host immunity and initiate disseminated infections (as reviewed by Osherov, 2011 [21]). Conversely, a curative role for epithelial activities is strongly supported by the observations of Chaudhary et al., 2012 who reported that bronchial ECs harbouring a CFTR mutation (ΔF508) demonstrate impaired uptake and killing of conidia [22]. The impact of EC-mediated activities upon disease outcomes in whole animal models of infection is presently unclear.
In fungi a conserved regulatory pathway governs the pH-dependent expression of secreted proteins and adaptation to alkaline stress [23]–[27]. Acting via PacC/Rim101 transcription factors, this environmental adaptation mechanism promotes the energy-efficient production of exported enzymes and metabolites, and has a demonstrated role in the pathogenicity of Candida albicans [25], [28] and Aspergillus nidulans [29]. Analysis of the transcriptomic response of invasive A. fumigatus hyphae to the mammalian pulmonary niche, identified 102 alkaline-responsive gene functions as being upregulated in leukopenic mice [30]. We therefore hypothesised that in A. fumigatus, PacC would be important for colonisation of the mildly alkaline murine lung. In this study we describe a functional genomics analysis of PacC-mediated activities which govern pathogenicity in mice. Unexpectedly, we discovered that PacC null mutants exhibit an unprecedented non-invasive phenotype, which is not an artefact of defective fungal growth, and can be recapitulated in vitro using cultured epithelia. The capacity to invade host tissues is therefore a genetically regulated trait, requiring PacC regulatory control.
In this study we exploited the differentially invasive properties of wild-type and ΔpacC isolates, to address the cellular and molecular basis of pathogen-mediated epithelial damage during Aspergillus infections. This revealed a novel mode of epithelial entry, occurring in a cell wall-dependent manner prior to protease production, and via the Dectin-1 β-glucan receptor. Our findings reveal novel mechanistic insights, having direct relevance to infection of whole animals, which will focus the onward study of A. fumigatus-mediated lung diseases upon dissecting the synergistic and/or additive impacts of temporally distinct aspects of the host-pathogen interaction at pulmonary epithelia. The multiple deficits in pathogenic activities and heightened sensitivity to echinocandin drugs, observed in ΔpacC isolates, highlight the potential of this receptor-mediated fungal signalling mechanism as a target for antifungal therapies.
To characterise the role of the A. fumigatus PacC transcription factor (AFUA_3G11970) in pathogenicity we constructed null and complemented alleles in two distinct A. fumigatus clinical isolates CEA10 and ATCC46645 (Figure S1). Relative to non-mutated isolates, PacC null mutants assumed a compact colonial phenotype on supplemented solid DMEM medium pH 7.4 (Figure 1A) which, in contrast to colonies of wild type isolates, lacked peripheral invasive hyphae, and were composed of a denser hyphal network indicative of a hyperbranching morphology (Figure 1B). This compact morphology was pH-independent, being also observed in colonies grown on Aspergillus complete medium pH 6.5 (Figure S2A), where sporulation and pigmentation of ΔpacC mutants was equivalent to that of the wild type. To assess pH sensitivity of the ΔpacC mutants we examined, on pH-buffered minimal media, the extent of radial growth at pHs 8.0 and 7.2, relative to growth at pH 6.5 (Figures S2B and S2C respectively). Consistent with a role for PacC in alkaline adaptation, ΔpacC isolates suffered growth impairment at pH 8.0, achieving approximately 10–20% of the radial growth attained at pH 6.5 (Figure S2B), compared with 40% achieved by wild-type and reconstituted strains. However, sensitivity of ΔpacC isolates to growth at pH 7.2, which approximates the pH of the mammalian lung, did not differ from that of the wild type isolates (Figure S2C).
To orchestrate alkaline adaptation, cytoplasmically localised PacC/Rim101 transcription factors must undergo pH-dependent proteolytic cleavage and nuclear entry [31]. Concordant with this model of transcription factor activation, we detected, by electrophoretic mobility shift assay (EMSA), several PacC retardation complexes, the relative quantities of which were altered under acidic and alkaline growth conditions (Figures S3A and S3B). Relative to growth at acidic pH, processed, activated forms of PacC increased in abundance upon shifting to alkaline conditions (Figures S3A and S3B). These findings are consistent with a pH-responsive mode of PacC activation, and with a conserved role for the A. fumigatus transcription factor in alkaline adaptation.
The mammalian pulmonary niche exerts multiple physiological stresses upon invading pathogens, including mildly alkaline pH, hypoxia, and iron-, zinc- and nutrient limitation [30], [32], [33]. To compare the growth rates of wild type and ΔpacC hyphae in a physiologically relevant setting, we devised an in vitro epithelial infection assay (Figure 1C), comprising monolayers of A549 alveolar epithelial cells cultured in supplemented DMEM medium (pH 7.4). We used this assay to assess the growth rates of wild type and ΔpacC isolates, under 5% CO2 (Figure 1C). To promote the visualisation of, and distinction between, fungal and host cells we stained host cells with FITC-labelled concavalin A and fungal cells with calcofluor white. We then performed a quantitative analysis of fungal growth rate (Figure 1D) and hyphal branching (Figure 1E) via immunofluorescence microscopy. Inspection of the infected monolayers revealed significant hyphal growth of both wild type and ΔpacC isolates (Figure 1C), but quantitation of hyphal branching frequency revealed an approximately doubled frequency of hyphal branching amongst ΔpacC isolates relative to wild type. To normalise for heightened branching during quantitation of hyphal growth rates we performed a comparative assessment of individual cell sizes by enumerating the number of pixels per fungal particle. This analysis, revealed that the cross-sectional area of wild type and ΔpacC cells approximates 540 versus 387 µm2 respectively (Figure 1D) which, assuming a uniform hyphal radius for wild type and mutant cells of approximately 3 µm, equates to a maximum deviance of 20 µm in length after 16 hours of in vitro co-culture with mammalian epithelia. On this mathematical basis, the median length of an unbranched ΔpacC hypha would be ∼ 30 µm. Given that ∼ 50% of ΔpacC hyphae remain unbranched, and the thickness of the alveolar epithelium is comparatively tiny, the observed differences in hyphal length between wild type and mutant hyphae are insufficient to explain the non-invasive phenotype of the ΔpacC mutants.
Leukopenia is an important risk factor for IA in humans, and cyclophosphamide-induced leukocyte depletion renders mice highly susceptible to pulmonary infection [34]. To assess the role of A. fumigatus PacC in pathogenicity we assessed the survival of leukopenic mice following infection, via the intranasal route, with spores of wild-type, ΔpacC or reconstituted isolates (Figure 2A). Relative to wild-type strains, ΔpacCATCC and ΔpacCCEA10 mutants were significantly attenuated for virulence (Figure 2A). At day 6 post-infection 100% of mice infected with the ΔpacC mutants remained alive while 78% and 92% of mice infected with wild type (ATCC46645 and CEA10 respectively) isolates were dead.
Histological analysis of infected lung tissues revealed significant differences between mutant and wild-type isolates by 20 hours post-infection (Figures 2B and C). While ΔpacC spores appeared to swell and form primary germ tubes by 12 hours post-infection (Figure 2B), penetration of the lung epithelium was not evident and, despite being competent in hyphal production in the murine airway, ΔpacC germlings remain contained within the epithelial boundary of the airspace at 20 hours post-infection (Figure 2C). Thus, ΔpacC germlings in leukopenic hosts exhibited a marked tissue non-invasive phenotype (Figures 2B and 2C).
To further characterise the non-invasive phenotype of ΔpacC isolates we used our in vitro co-culture assay (Figure 1C) to examine integrity of A549 alveolar epithelial monolayers following 16 hours co-culture with wild-type or ΔpacC isolates by enumeration of detaching epithelial cells. In monolayers infected with wild-type strains, extensive rounding and detachment of up to 40% of host cells was demonstrable, resulting in observable destruction of the epithelial monolayer (Figures 1C and 3A). However, similar to PBS-challenged monolayers, infection with ΔpacC mutants led to detachment of less than 5% of monolayer cells (Figures 1C and 3A). Although ΔpacC hyphae achieved similar hyphal growth rates (Figure 1D) and epithelial coverage (Figure 1C) to isogenic wild type isolates, ΔpacC hyphae were found to navigate the surface of the epithelial monolayer without effecting cellular detachment (Figure 1C).
To further characterise this deficit in epithelial damage, a modified 51Cr release assay [35], adopting a similar time course of infection, was utilised to quantitatively assess epithelial cell lysis upon infection (Figure 3B). Concordant with cell detachment assays, ΔpacC mutants reproducibly failed to fully elicit epithelial damage. Taken together these findings demonstrate that ΔpacC hyphae fail to elicit disintegration of alveolar epithelia during in vitro culture at pH 7.4 despite achieving similar growth rates as wild-type isolates (Figures 1C and D and 3A and B).
Several observations argue strongly against a trivial pH-dependent growth defect as the basis for the ΔpacC virulence defect. First, in A. fumigatus, a compact colonial morphology is not a robust correlate of reduced virulence. For example, despite compact colonial morphology and a highly branching mode of growth, null mutants lacking the ChsG chitin synthase remain fully virulent in an inhalational model of infection [36]. Second, in A. fumigatus, in vitro alkaline sensitivity is not a robust correlate of reduced virulence. For example, mutants lacking the mitogen-activated kinase MpkA, demonstrate more severe radial growth defects than ΔpacC mutants in Aspergillus minimal medium [37] and exhibit highly alkaline sensitive phenotypes [38], but retain full virulence in a low dose inhalational model of aspergillosis. Third, during in vitro cell culture with A549 cells (pH 7.4), hyphae of ΔpacC mutants achieve similar growth rates to that of isogenic wild type isolates (Figures 1C and D). Fourth, despite achieving similar growth rates to wild type hyphae during in vitro cell culture with A549 cells (pH 7.4), hyphae of ΔpacC mutants fail to elicit epithelial decay (Figures 1C, 3A and B). Finally, hyphae of ΔpacC mutants fail to traverse the murine epithelium and are strictly confined to the airway (Figure 2C). ΔpacC mutants are the first-reported non-invasive A. fumigatus mutants, revealing that tissue invasion is a genetically regulated trait under PacC regulatory control. We therefore exploited the differentially invasive properties of wild type and ΔpacC isolates to seek a more detailed mechanistic understanding of tissue invasive growth in this pathogen.
Previously, we devised a strategy for analysing A. fumigatus gene expression during initiation of murine aspergillosis [30]. Here we applied a similar approach, this time performing time-series analyses (4, 8, 12 and 16 hours) of A. fumigatus gene expression. This permitted the capture of stage-specific gene expression during invasive colonisation of the leukopenic murine lung, the first reported longitudinal study of gene expression during mammalian pulmonary infection. We adopted a comprehensive experimental design (Figure S4A), incorporating 12 competitive hybridisations and 12 flip-dye experiments. This permitted the analysis of stage-specific gene expression in both infecting wild-type ATCC46645 (Figure S4B) and ΔpacCATCC isolates (Figure S4C), as well as the directly comparative analysis of gene expression, by time-point, for the wild type and ΔpacCATCC mutants (Figure S4D).
Relative to ungerminated spores, transcript profiling of wild-type gene expression revealed a total of 3733 genes, (log2≥ +/−1.5), which were differentially expressed at a minimum of one time-point during invasive infection. The differentially regulated genes were assigned to three cohorts, corresponding to (i) genes consistently up- or down-regulated across the time series; or differentially regulated during (ii) early (4, 8 and 12 hours) or (iii) late (12 and 16 hours) phases (Dataset S1). This revealed respiration, metabolism and amino acid biosynthesis as being prioritised during early infection of the leukopenic host, while cation transport, secondary metabolism and iron metabolism were subsequently emphasised during commencement of invasive growth (Dataset S1). Throughout the time series of growth in the host, upregulated expression of secreted gene products remained highly significant. A comprehensive functional, and statistical, analysis of differentially regulated gene products is provided in Dataset S1.
Directly comparative analysis of ΔpacCATCC and ATCC46645 activities (Dataset S2) revealed 1116 genes to be differentially expressed. Of these, 577 were up-regulated and 539 were down-regulated in the ΔpacCATCC isolate, relative to the wild type isolate, in at least one time point of the analysis. Scrutiny of the datasets revealed dysregulated expression of secreted protein gene products, defined as having predicted signal peptide motifs (Figure S5), cell wall biosynthetic enzymes (Figure S6), and gliotoxin biosynthetic genes (Figure S7) during infections caused by the ΔpacCATCC isolate. A comprehensive functional, and statistical, analysis of differentially regulated gene products is provided in Dataset S2. The differential regulation of 5 genes was independently validated by quantitative PCR (Figure S8).
A. fumigatus spores adhere rapidly (within 30 minutes) to lung pneumocytes and become quickly internalised and killed [16]–[18], [20]. In response to challenge with A. fumigatus conidia, host injury can be observed as cell rounding and detachment from monolayers [16]–[18], [20], and cytoskeletal fibres of lung pneumocytes suffer major reorganisation, an effect which can be blocked by antipain-mediated protease inhibition [8]. We found secreted factors to be the major functional cohort amongst those aberrantly regulated during ΔpacC infections (Figure S5, and Dataset S2). To assess the role of secreted factors in epithelial disintegration we exposed A549 monolayers to A. fumigatus culture filtrates derived from young (16 hours), or mature (48 hours), wild-type or ΔpacC cultures grown in supplemented DMEM culture medium. Filtrates obtained from mature cultures of wild-type or reconstituted A. fumigatus isolates, prompted significant reductions (∼30–40%) in the numbers of adherent cells after 20 hours of co-incubation with alveolar epithelia (Figure 4A). In contrast, filtrates derived from mature ΔpacC cultures led to detachment of less than 10% of monolayer cells (Figure 4A). Concordant with a protease-mediated basis for epithelial destruction, pre-treatment of wild-type A. fumigatus culture filtrates with the protease inhibitor antipain [8] reduced cellular detachment by up to 50% (Figure 4A). To further probe protease production by wild type and mutant isolates, we used a qualitative assay based upon the clearance of gelatin from the surface of unprocessed X-ray film [39]. In agreement with our assays of epithelial degradation (Figure 4A) we detected gelatin-degrading activity in filtrates of mature wild type and reconstituted A. fumigatus cultures, which was absent in cultures from ΔpacC isolates (Figure S9).
Together, these findings are consistent with the release of a damaging proteolytic entity by mature A. fumigatus hyphae, the production of which is dependent upon PacC-mediated signalling. Notably, epithelia challenged with ΔpacC culture supernatants were somewhat protected by pretreatment of fungal extracts with antipain (Figure 4A). This might indicate the production of a protective, host-derived enzyme which is degraded or inactivated by culture filtrates of wild type A. fumigatus isolates, or the existence of a protective, host-derived enzyme whose action, in this assay, is masked by the high degree of epithelial detachment imposed by wild type isolates.
Critically, our analysis of filtrates obtained from younger fungal cultures (16 hours) revealed a novel finding. Regardless of the fungal strain tested, exposure to filtrates from young cultures did not impact monolayer integrity (Figure 4B). This finding suggested that epithelial disintegration occurring at 16 hours of spore and EC co-incubation (Figure 4B) requires direct interaction between host and pathogen cells and represents a genetically regulated fungal assault upon epithelial integrity which is temporally, and mechanistically distinct from protease-mediated damage. Further, that this damaging interaction between host and pathogen cells occurs during immediate proximity between host and pathogen cells, most likely in a contact-dependent manner. To substantiate this view we reiterated the detachment analysis, this time omitting monolayer washing to analyse only host cells directly contacting the pathogen (Figure 4C). A549 cells in contact with ΔpacC hyphae underwent significantly less rounding and detachment than those contacting hyphae of wild-type or reconstituted isolates (Figure 4C). Taken together, these data reveal that A. fumigatus elicits host damage in a biphasic manner and, that the pH-responsive transcription factor PacC governs functions required for epithelial disintegration during both early- and late-phases of the host-pathogen interaction.
Amongst the functional cohorts aberrantly regulated during ΔpacCATCC infections (Figures S5–S7 and Dataset S2) cell wall biosynthesis offered a plausible mechanism for contact-dependent host damage. To analyse cell wall compositions of mutant and wild-type isolates, strains were stained with the chitin-binding agent calcofluor white (CFW). Microscopic examination revealed intensified CFW-staining of ΔpacC germ tube tips relative to those of the parental isolates (Figure 5A) and quantitative analysis of fluorescence intensities revealed significantly higher CFW in ΔpacC germ tubes (Figure 5B). In addition, electron microscopy showed a thickened cell wall in the ΔpacCATCC mutant, which was highly evident after 16 hours of growth (Figure S10). Hyphal cell wall composition (Figures 5C, D and E) was assessed by high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD). After 16 hours of growth, cell wall chitin content was found to be 20% higher in extracts from ΔpacC isolates, relative to wild-type cells (Figure 5C). The quantity of cell wall glucan and mannan was measured as equivalent between mutant and wild type cells (Figures 5D and 5E).
To probe the contribution of cell wall components to epithelial cell detachment, we exposed A549 monolayers to cell wall extracts. This revealed that, relative to the vehicle control, the percentage of adherent cells was drastically reduced upon exposure to cell wall extracts from wild-type isolates, an effect which was not elicited by ΔpacC cell wall extracts (Figure 5F).
If detachment of alveolar epithelial cells occurs via a cell wall-mediated mechanism, infection with dead hyphae would be predicted also to cause cellular detachment. To test this, A549 monolayers were incubated with thimerosal-killed hyphae and cell detachment, per unit of hyphal length, was measured from unwashed monolayers. Killed hyphae from parental isolates caused damage to the epithelial monolayer independently of fungal viability, an effect which was significantly impaired in monolayers incubated with killed ΔpacC hyphae (Figure 5G). In conclusion, epithelial detachment elicited early in the interaction between A. fumigatus and host cells requires contact between host and pathogen and occurs independently of fungal viability. Crucially, the inability of killed ΔpacC hyphae to injure epithelial monolayers eliminated the trivial possibility that mere collision between host and fungal cells can account for loss of monolayer integrity.
The mammalian C-type lectin receptor Dectin-1, is predominantly expressed by myeloid cells and recognizes a variety of fungal β-1,3-linked and β-1,6-linked glucans [40]–[42]. Recent transcriptional and immunohistochemical analyses have revealed Dectin-1 gene expression in bronchiolar epithelia, and alveolar type II cells (ATIIs) of murine lungs [43] and Han et al., showed that A549 cells internalise germinated A. fumigatus spores in a phospholipase D-dependent manner, a process inhibited by an anti-Dectin-1 antibody [44]. Given these observations and aberrant cell wall remodelling in (Figure 5), and epithelial damage by (Figure 4), ΔpacC mutants we hypothesised non- or reduced involvement of Dectin-1 in promoting A. fumigatus ΔpacC spore recognition and internalisation.
Epithelial cells have been shown to internalise and kill up to 50% of the A. fumigatus spores they come into contact with [16]–[19], [22]. We therefore hypothesised that contact-dependent perturbations of epithelial integrity might result from internalisation of fungal spores. To assess this we first assessed the numbers of internalised wild type and ΔpacC spores using a nystatin protection assay (Figure 6A). This revealed that the proportion of wild-type spores internalised by A549 cells ranges from 16 to 23% (Figure 6A). However, epithelial cells were found to internalise ΔpacC mutants significantly less avidly than the respective parental isolates (Figure 6A), whereby only ∼10–12% of the initial inoculum had become internalised after 4 hours of co-incubation. At the concentrations used in this assay nystatin exposure was 100% efficient in killing A. fumigatus spores (Figure S11).
Given the altered cell wall morphology of ΔpacC isolates a plausible explanation for their defective internalisation might include an adhesion defect. We therefore tested the ability of the isolates to adhere to plastic surfaces and epithelial monolayers after 30 minutes of incubation. However, for neither substrate was a difference in adherence observed, either for the ΔpacC mutants or wild type isolates (Figure S12). Taken together, these data suggest that internalisation of A. fumigatus spores, which is hampered by pacC deletion, contributes to contact-dependent monolayer decay.
To assess the role of Dectin-1 in promoting epithelial decay in vitro we pre-incubated epithelial cells with the monoclonal anti-Dectin-1 antibody (Mab1859) prior to performing nystatin protection assays. Concordant with a role for Dectin-1 in spore internalisation, epithelial monolayers pre-treated with Mab1859 exhibited a ∼20–30% reduction in internalised spores, relative to untreated epithelia (Figure 6B). As a control we used 0.2 µM cytochalasin D (CD), an inhibitor of actin polymerization, which prevents spore internalisation into A549 cells [18]. The impact of CD treatment (50% reduction) was consistently greater than that of Mab1859, possibly indicating the contribution of additional spore-detecting PRRs driving spore internalisation and/or opsonic phagocytosis and/or incomplete Mab1859-mediated inhibition of Dectin-1 activity.
To assess the impact of spore internalisation upon epithelial integrity, A549 monolayers were pre-treated for 1 hour with CD and detachment after 16 hours was evaluated using our in vitro assay. For wild type infections, cellular detachment from epithelial monolayers was significantly reduced from ∼45% to ∼25% in the presence of CD (Figure 6C) but pre-incubation with CD did not alter integrity of ΔpacC-challenged monolayers (Figure 6C). These results suggested that actin-mediated internalisation of wild-type spores contributes to epithelial detachment during initial interactions with fungal spores.
To further probe the molecular basis of epithelial decay, monolayers were preincubated with the monoclonal anti-Dectin-1 antibody (Mab1859) prior to co-incubation with A. fumigatus spores. Concordant with a damaging role for Dectin-1 mediated internalisation of A. fumigatus spores and hyphae in A549 alveolar monolayers, Mab1859 pre-treatment conveyed an almost complete protection of monolayer integrity during co-incubation with wild type A. fumigatus isolates (Figure 6C).
To characterise the cell wall defect inhibiting Dectin-1-mediated uptake of ΔpacC we examined β-glucan content in the cell walls of wild type and ΔpacC spores and hyphae, by incubating fungal cells with a soluble chimeric Dectin-1 Fc protein [45], [46] and quantifying immunofluorescence. This revealed highly anomolous organisation of β-glucan content in ΔpacC spores (Figure 6D). Relative to wild type spores which deposit β-glucan at a highly localised, and singular focus of the spore cell wall prior to germination, β-glucan in ΔpacC spores adopts a highly diffuse distribution. This phenotype is not a consequence of slowed spore swelling as wild type and ΔpacC spores demonstrate equivalent sizes at 4 hours of culture in supplemented DMEM (Figure 6E).
In Histoplasma capsulatum α-1,3-glucan promotes virulence by blocking innate immune recognition of β-glucan by Dectin-1. To examine α-1,3-glucan content in A. fumigatus spores and hyphae we used immunofluorescence microscopy and anti-α-1,3-glucan antibody, which has previously been used for analysis of the H. capsulatum cell wall [47]. Immunofluorescence-mediated detection of this antibody revealed similar distributions of α-glucan in wild type and ΔpacC hyphae (Figure S13).
Taken together these data support an important role for internalisation of A. fumigatus spores during invasion of the pulmonary epithelium which, in a cell wall-, actin- and Dectin-1 dependent manner permits endocytosis of fungal particles contacting alveolar epithelia.
The extent of epithelial protection afforded by in vitro delivery of Mab1859 (Figures 6B and C) was suggestive of a detrimental role for Dectin-1 engagement during A. fumigatus-epithelial interactions. To decipher between protective and exacerbatory roles for Dectin-1 in maintenance of epithelial integrity in whole animals, we assessed pulmonary damage after 24 hours of A. fumigatus infection in Dectin-1+/+ and Dectin-1−/− mice. To study epithelial activities in the absence of confounding leukocyte responses, mice were depleted of leukocytes using a cyclophosphamide and hydrocortisone protocol and lung injury was scored via histological, biochemical and immunoblot assays. In the lungs of Dectin-1+/+ and Dectin-1−/− mice, fungal lesions were equivalent in size and invasiveness (Figure 7A) although frequency of fungal lesions was increased in Dectin-1−/− animals (not shown). Epithelial damage was surveyed via quantitation of lactate dehydrogenase (LDH) in BALs (Figure 7B) and analysis of expression of the Dectin-1-independent damage associated molecular pattern (DAMP) protein S100B (Figure 7C), whose major source during A. fumigatus infection is epithelial cells [48]. Both assays revealed heightened epithelial damage in the lungs of Dectin-1−/− animals relative to wild type counterparts. Our results indicate that, despite a highly protective role for the anti Dectin-1 antibody Mab1859 during in vitro epithelial infections, integrity of Dectin-1 (Figures 7A–C) is essential for limitation of epithelial damage in vivo. As neutrophil-depletion and macrophage dysfunction were chemotherapeutically implemented in our murine model we conclude that Dectin-1 activity is essential for protecting the lung epithelium from the damage inflicted by germinating A. fumigatus spores.
PacC homologues in the fungal pathogens C. albicans and Cryptococcus neoformans indirectly modulate interactions with the host interface via governance of fungal cell wall architecture. For both organisms, cell surface defects in Rim101 null mutants appear to be a critical component of altered pathogenicity. In C. albicans, oropharyngeal pathogenicity, estimated from in vitro assessment of damage in the FaDu cell line, can be partially restored to attenuated Rim101 null mutants via overexpression of the Rim101 target genes ALS3, CHT2, PGA7/RBT6, SKN1 or ZRT1 [49]. In C. neoformans Rim101 null mutants, via cell wall defects, prompt aberrant inflammatory responses, resulting in mild hypervirulence [50]. Thus, in the case of both organisms, host immune responses to altered cell wall composition play a functional role in disease outcome. To test the immunostimulatory capacity of wild-type and ΔpacC spores, immunocompetent CD1 mice were infected with 106 spores and the recruitment of macrophages (F4/80+) and neutrophils (Ly-6G+) to the pulmonary niche was quantified. Relative to mice infected with a wild-type isolate, no significant difference in the number of recruited neutrophils was recorded for mice infected with the ΔpacC mutant (Figure 7D). Further, and in stark contrast to findings in C. neoformans, we did not observe (in immunocompetent hosts) a hyperinflammatory response to infection with ΔpacC mutants. We therefore conclude that anomalous innate immune responses are unlikely to contribute to the altered pathogenicity of A. fumigatus ΔpacC mutants. Therefore, in stark contrast to oral and pulmonary infections, respectively with C. albicans [49] and C. neoformans [50]–[52], modulation of host innate immunity is unlikely to contribute to A. fumigatus disease outcome.
Taken together our results indicate that the predominant host-mediated mechanism promoting the non-invasive phenotype of A. fumigatus ΔpacC mutants is a failure to engage the Dectin-1 receptor. It is therefore highly likely that A. fumigatus exploits this innate immune mechanism to gain entry to the pulmonary epithelium. The important role for Dectin-1 in epithelial protection in vivo implies that the full extent of systemic Dectin-1 depletion upon epithelial defences likely extends well beyond defective internalisation of inhaled fungal spores. However, given the propensity of A. fumigatus to exploit this mode of tissue entry, it remains possible that the targeted depletion of epithelial Dectin-1 activity would afford protection against invasive, and other, A. fumigatus diseases of the lung.
The fungal cell wall is a premier, pathogen-specific target for antifungal drugs. Given the significant cell wall defect observed in ΔpacC mutants we predicted altered echinocandin sensitivity relative to wild type isolates. A standard EUCAST assay [53] was used to calculate the susceptibility of isolates, revealing increased susceptibility (Figure 8A) of ΔpacC mutants (minimum effective concentration, MEC, of 0.11 µg/ml) compared to that of ATCC46645 (∼ 0.58 µg/ml) and CEA10 (∼ 0.75 µg/ml). A. fumigatus strains grown in the presence of 16 µg/ml caspofungin displayed aberrant morphology, elevated branching and shortening of hyphae. Heightened severity of these phenotypes was observed for ΔpacC mutants which demonstrated extensive ballooning of hyphal tips (Figure 8B). Given the tendency for chitin increase to promote echinocandin tolerance the heightened susceptibility of ΔpacC mutants is surprising; however, an obvious explanation for this effect would be increased porosity due to altered cell wall architecture.
To test the potency of echinocandin agents against pH non-sensing mutants in vivo we examined, via viable counts, the effect of caspofungin (5 mg/kg) treatment at 48 hr post-infection. Relative to animals infected with a wild type isolate, fungal burden was significantly decreased in caspofungin-treated mice infected with ΔpacCCEA10 (Figure 8C). Histological analysis of lung tissues recovered from mice infected with CEA10 or ΔpacCCEA10 in the presence or absence of caspofungin confirmed this observation (Figure 8D). Thus, ΔpacC mutants are hypersensitive to the cell wall-active drug caspofungin, a phenotype which extrapolates to mammalian infections (Figures 8C and 8D). Given that the A. fumigatus cell wall is essential for viability, agents which selectively inhibit the pH-dependent activation of PacC signalling might provide useful adjuncts to existing antifungal therapies.
Amongst an annual global caseload of 1.5 million fatal mycoses, more than 75% of infections are initiated by inhalation of fungal particles [1]. Despite this, our understanding of the interactions between inhaled fungal pathogens and the respiratory epithelium remains in its infancy. This study addresses disease caused by the major mould pathogen of humans, A. fumigatus, and assigns an essential role for the transcription factor PacC in epithelial invasion and pathogenicity. A critical discovery made during this study is the inability of ΔpacC mutants to invade the mammalian respiratory epithelium, a hallmark of invasive diseases caused by A. fumigatus. Our study confirms that epithelial invasion by this pathogen is a genetically-regulated trait, under PacC regulatory control, and identifies the cellular basis of the deficit to lie with at least two temporally and mechanistically distinct processes, namely protease-mediated monolayer decay and epithelial entry.
Compared to our previous studies of A. nidulans pH regulation [29], loss of A. fumigatus PacC has a milder impact on alkaline tolerance and in vivo germination. At least one mode of essential micronutrient acquisition is different between the two species. In A. nidulans, siderophore-mediated iron acquisition is PacC-dependent [54], in A. fumigatus it is not (Hubertus Haas personal communication). Certainly this could explain the differences between these two species in germination/growth rates in the mammalian lung.
The A. fumigatus genome is predicted to encode more than 100 hydrolytic enzymes [55] some of which are assumed as crucial for liberation of proteinaceous nutrients from host tissues [56], [57]. Early studies found a correlation between high elastinolytic potency of A. fumigatus isolates and pathogenicity in mice [58] but a subsequent survey of 73 isolates revealed discordant production of extracellular elastase, acid proteinase and phospholipase amongst strains causing human disease. The elastinolytic neutral metalloprotease Mep (AFUA_8G07080), secreted in A. fumigatus culture filtrates and leukopenic murine hosts [59], is dispensable for pathogenicity.
A. fumigatus culture filtrates cause epithelial desquamation and destroy F-actin cytoskeletal fibres of in vitro-cultured pneumocytes [8], [60]. Kogan et al., 2004 found deletion of the alp1 gene (AFUA_4G11800) encoding a secreted alkaline protease, or antipain treatment of wild-type culture supernatants to be equally ablative of secreted protease activity in vitro. In addition, immunolabelling of the F-actin cytoskeletal fibres of A549 cells revealed that F-actin disruption requires Alp1 integrity. However, an A. fumigatus mutant lacking Alp1 retained full virulence in both cortisone-treated [61] and leukopenic mice [15]. Attempts to implement wholesale depletion of A. fumigatus protease expression have also failed to unearth avirulent mutants. A doubly protease-deficient mutant lacking Alp1 and Mep, which is completely deficient in collagenic proteolytic activity at neutral pH in vitro, is fully virulent in a cortisone-treated murine model [13]. Furthermore, an A. fumigatus ΔprtT mutant lacking a conserved positive regulator of secreted proteases suffers a 70% reduction in casein proteolytic activity, but is also fully virulent in leukopenic mice. These findings cast doubt upon the true relevance of protease production to A. fumigatus pathogenicity [10]. Comparison of the PacC and PrtT [9] regulons revealed 83 and 31 secreted gene products as being down-regulated by PacC and PrtT (AFUA_4G10120) respectively, amongst which only 8 are commonly down-regulated (Dataset S3). Given the fully virulent phenotype of the ΔprtT, mutant we can confidently surmise that none of these gene products, acting alone or in combination with each other, can support tissue-invasive growth in the mammalian host. Thus their aberrant expression in the ΔpacC mutant is unlikely to explain its non-invasive phenotype. Concordant with this conclusion, and with the fully virulent phenotype of a Δmep;Δalp1 mutant [13], the inclusion of genes encoding both Alp1 and Mep1 (Dataset S3) amongst those down-regulated in both PacC and PrtT null mutants cannot, alone, explain the non-invasive phenotype of PacC null mutants. Loss of PacC negatively impacts a further 75 uncharacterised secreted gene products. If the tissue invasive growth of A. fumigatus is solely proteolytically mediated, we predict that important enzymatic functions will be amongst them. Investigation of this hypothesis lies beyond the scope of this study but is an ongoing component of our further study, as is the co-operative activity of fungal protease- and toxin-mediated assaults upon the mammalian pulmonary epithelium.
The finding that epithelial monolayers are resistant to culture filtrates obtained from earlier (16 hour) time points of fungal growth highlights the existence of at least one additional, and earlier acting, perturbation of host tissue. Soluble effectors of epithelial detachment are not immediately secreted by metabolically active A. fumigatus spores and a phased mode of A. fumigatus assault, commencing with contact-dependent perturbation, is likely responsible for monolayer perturbation. In agreement with this hypothesis, we found cytochalasin D-mediated inhibition of actin polymerisation to be partially protective of epithelial monolayer integrity, and ΔpacC spores to be far less avidly internalised than wild-type counterparts. Amongst the repertoire of invasion tactics employed by fungal pathogens at host epithelia, induced endocytosis, active penetration and participation of host factors have been implicated [52]. The results of our study implicate all three activities as having relevance to the host-A. fumigatus interaction, but with several critical differences relative to studies of other fungal pathogens. Firstly, the occurrence of induced endocytosis, as evidenced by sensitivity of the internalisation process to cytochalasin D-mediated inhibition (Figure 6C), promotes epithelial disintegration by both live and dead fungal elements. This finding stands in stark contrast to epithelial invasion by C. albicans where internalisation of live germinated cells, but not killed cells, leads to host damage [62]. Second, the existence of A. fumigatus invasins remains thus far unproven. Certainly, from bioinformatics analyses, evidence for highly conserved homologues for the C. albicans invasin Ssa1 can be gleaned. However the expression of this homologue is not impacted by pacC gene deletion and A. fumigatus lacks any homologue of the Als3 invasin altogether [63], [64]. Our finding that cell wall extracts can impact epithelial integrity, and the relevance of Dectin-1 to this process, implies the existence of an invasin-independent mode of epithelial entry for A. fumigatus. In our studies cytochalasin D imposed an incomplete block upon spore internalisation suggesting that at least a subset of fungal elements can access the internal environment of epithelial cells via an ‘active penetration’ mechanism, as recently documented for C. albicans [65]. These important differences in the way in which different fungal pathogens interact with physiologically distinct epithelia highlight the current paucity of information on fungal interactions with pulmonary epithelia and, given that the vast majority of invasive mycoses are initiated via inhalation of fungal particles [1], should prompt renewed scrutiny of fungal interactions with mammalian lung tissues.
The mechanism by which epithelial cells recognise and internalise A. fumigatus conidia remains poorly characterised, as does the relevance of such activity to disease outcome. Our study demonstrates that uptake of A. fumigatus spores, by type II pneumocytes, is dependent upon a) actin polymerisation b) fungal cell wall/surface composition c) integrity of PacC and d) the β-1,3-glucan receptor Dectin-1, and moreover, that such interactions can negatively impact epithelial integrity. Dectin-1 expression in non-myeloid cells is increasingly frequently reported and has demonstrated relevance in β-1,3-glucan-(curdlan) exposed bronchiolar and alveolar type II cells [43], poly IC challenge of human bronchial epithelial cells [66], Mycobacterium ulcerans infection of epidermal keratinocytes and Mycobacterium tuberculosis infection of A549 epithelia [67], [68]. The A549 cell surface constitutively expresses Dectin-1, regardless of infection by A. fumigatus [44]. In humans, a Y238X Dectin-1 polymorphism is a risk factor for invasive aspergillosis in haematopoietic stem cell recipients. Cunha et al. (2010) found, in experimental HSCT, that transplant of stem cells from Dectin-1- donors to wild-type recipients resulted in lessened susceptibility to invasive aspergillosis. Thus by restricting the Dectin-1 deficiency to cells of myeloid origin, susceptibility to invasive aspergillosis was not altered. We found Dectin-1 deficiency, in the absence of leukocytes, to heighten epithelial damage in the whole animal host relative to leukopenic wild-type animals. This observation and several of our findings support a curative role for Dectin-1 mediated internalisation of A. fumigatus spores. According to Han et al., 2011, internalisation of A. fumigatus by A549 epithelial cells can be correlated with membrane phosphatidylcholine cleavage, a process which is closely linked to alteration of cytoskeletal actin dynamics, and prompted by exposure to β-1,3-glucan. This finding is consistent with our observation that cell wall extracts and killed A. fumigatus hyphae can perturb epithelial integrity. Our data, and those of Chaudhary et al., 2012, who found that bronchial ECs from cystic fibrosis sufferers demonstrate impaired uptake and killing of conidia, are highly suggestive of a curative role for EC activities during exposure to A. fumigatus spores. It does, however, remain feasible that the pulmonary epithelium provides a reservoir for A. fumigatus spores, and the attenuated phenotype of the ΔpacC mutant is consistent with such a theory. It also remains feasible that contact-dependent perturbation of epithelia facilitates subsequent protease-mediated damage by exposing subepithelial structures and facilitating fungal adhesion. Studies of C. albicans have revealed several means of fungal entry into epithelial cells, including self-induced endocytosis, and protease-mediated decay, the latter impacting, via Rim101-mediated E-cadherin degradation, the disintegration of oral epithelia [69]. It is most likely that multiple mechanisms contribute also to the pathogenicity of A. fumigatus. What is unique about PacC in A. fumigatus, is the critical role played by PacC-dependent factors in all of these processes and the profound requirement for PacC to orchestrate epithelial entry, protease-mediated epithelial decay, invasive growth and pathogenicity. These findings not only identify PacC as a critical master regulator of pathogenicity determinants in A. fumigatus, but also heighten its relevance as an antifungal target.
A. fumigatus strains used in this study are listed in Table S1. A. fumigatus strains were cultured at 37°C in Aspergillus Minimal Media (AMM) or Aspergillus Complete Media (ACM) [70]. For preparation of A. fumigatus culture supernatants, 106 spores/ml were grown in Dulbecco's Modified Eagle Medium (DMEM, Sigma) supplemented with 10% foetal bovine serum (FBS, Sigma) and 10% penicillin and streptomycin (Sigma) at 37°C, 5% CO2 for either 16, 48, 60 or 72 hr. Supernatants were doubly filtrated through Miracloth and centrifuged for 10 min at 4000 rpm to remove any hyphal fragments. To inhibit protease activity in culture supernatants, filtrates were treated with the serine and cysteine protease inhibitor antipain (10 µg/ml) for 1 hr. For preparation of A. fumigatus cell wall extracts, see Text S1. For analyses of cell wall composition, A. fumigatus strains were harvested from an ACM plate (0 hr) as previously described or from 105 spores/ml cultures grown in AMM broth for 4, 8 and 16 hr at 37°C with shaking at 200 rpm. For in vitro challenge of epithelial monolayers with killed hyphae 104 spores/ml were incubated in supplemented DMEM at 37°C, 5% CO2 for 18 hr (parental isolates and reconstituted strains) or 36 hr (ΔpacC mutants). A. fumigatus hyphae were killed by incubating them in PBS supplemented with 0.02% thimerosal (Sigma) overnight at 4°C [71], a treatment which preserves cellular integrity. Before incubation with monolayers, killed hyphae were washed twice in PBS. Killing was verified by plating a 1∶100 dilution of killed hyphae in ACM plates.
ΔpacC mutants in the genetic backgrounds CEA10 (ΔpacCCEA10) and ATCC46645 (ΔpacCATCC) were constructed by gene replacement (Figure S1A) using a split-marker strategy [72]. For details, see Text S1. Initial phenotypic analyses, including survival analyses were performed using both ΔpacCATCC and ΔpacCCEA10 mutants, while subsequent analyses of gene expression were limited to the ΔpacCATCC mutant. For analyses of ΔpacC fungal burden in mice we opted to use the lesser attenuated ΔpacCCEA10 to prolong fungal occupancy of the murine lung.
Dilutions of 103 conidia were inoculated onto ACM or supplemented AMM (as shown on Table S3) and incubated for 48 hr and 72 hr respectively. Images were captured using a Nikon Coolpix 990 digital camera.
Spores were inoculated at a density of 1×106 to 2×106 spores/ml, in 100 ml of liquid ACM, and grown for 16 hr at 37°C. Next, media was buffered to pH 5.0 with 100 mM glycolic acid pH 5.0 or to pH 8.0 with 100 mM Tris-HCl pH 8.0, pH shifts were performed for 1 hour. 10 mg of protein was extracted from washed mycelia as described previously [73]. Protein concentrations were determined using the Bradford assay [74]. The ipnA2 probe was synthesised and labelled as described previously [27], [73]. Densitometry data were obtained by measuring pixel intensity/mm2 for the relevant bands using a Phosphorimager FLA-3000 (FujiFilm) and Multi-Gauge V3.0 software.
Murine infections were performed under UK Home Office Project Licence PPL/70/6487 in dedicated facilities at Imperial College London. For all experiments A. fumigatus spores were harvested and prepared as previously described [75] and viable counts from administered inocula were determined, following serial dilution, by growth for 24–48 hr on ACM. Mice were housed in individually vented cages and anaesthetized by halothane inhalation and infected by intranasal instillation of spore suspensions. Mice were rendered leukopenic by administration of cyclophosphamide (150 mg/kg, intraperitoneal) on days −3, −1, +2 and every subsequent third day, and a single subcutaneous dose of hydrocortisone acetate (112.5 mg/kg) administered on day −1.
Survival analysis. Leukopenic male CD1 mice (18–22 g) were infected by intranasal instillation of 5.0 × 104 or 6.0 × 105 conidia in 40 µl of saline solution. Mice were weighed every 24 hr from day of infection and visual inspection made twice daily. In the majority of cases the end-point for survival experimentation was a 20% reduction in body weight measured from day of infection, at which point mice were sacrificed.
Histological and transcriptomic analyses. Leukopenic male CD1 mice (n = 8) were infected with 108 conidia in 40 µl of saline solution. At the relevant time-points post-infection, mice were sacrificed and lungs were partitioned, using surgical sutures, into lobes destined for transcriptomic or histological analysis. Bronchoalveolar lavages (BALs) were performed using three 0.5 ml aliquots of pre-warmed sterile saline. BALs were snap frozen immediately following harvesting using liquid nitrogen. Lobes for histological sectioning were removed and immediately fixed in 4% (v/v) formaldehyde (Sigma). Lungs were embedded in paraffin prior to sectioning and stained with haematoxylin and eosin or light green and Grocott's Methenamine Silver. Images were taken using a Reichert-Jung (Slough, UK) Polyvar microscope using brightfield illumination at 40X magnification.
Fungal burden analyses. Immunosuppressed male CD1 mice (n = 5) were infected with 3.75 × 105 spores. Mice were culled and whole lungs were collected after 24 and 48 hr of infection.
BAL samples were centrifuged at 14000 rpm for 5 min and the pellet was washed with 500 µl ice cold H2O to lyse host cells. Seven BALs were pooled, resuspended in 450 µl ME-RLC buffer (QIAGEN) and ground in liquid nitrogen with a pestle and mortar. RNA was then extracted using RNeasy Kit (QIAGEN). A reference RNA sample was extracted from A. fumigatus ATCC46645 conidia harvested from an ACM plate. Conidia were washed thoroughly with sterile water, quickly frozen in liquid nitrogen, and disrupted by grinding. Total RNA was extracted using RNeasy Kit (QIAGEN). The quality of RNA used for microarray analysis was checked using a Nanodrop ND-1000 Spectrophotometer (Nanodrop, Wilmington, USA). Only RNA with an A260/280 and an A260/230 ratio> 1.9 was used for the experiments. Labelled cDNA samples were synthesised as described previously [30]. Protocols for direct labelling and hybridisation of cDNA probes can be found on the JCVI website (http://pfgrc.jcvi.org/index.php/microarray/protocols.html). The A. fumigatus oligonucleotide slides version 3 was used for microarray hybridization (http://pfgrc.jcvi.org/index.php/microarray/array_description/aspergillus_fumigatus/version3.html).The phase- or strain-specific comparative analysis of gene expression datasets was conducted in Genespring GX 11.02 (Agilent). Normalised log2 expression ratios were filtered on expression level and differentially regulated transcripts were defined as having log2 (Cy5/Cy3) greater than the arbitrary thresholds of ± 1.5. Raw data have been deposited in the Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE54810. Functional analysis of differentially-expressed gene cohorts was implemented by DAVID (http://david.abcc.ncifcrf.gov/) [77], [78]. Microarray data was validated by qPCR as described in Text S1.
Human pulmonary carcinoma epithelial cell line A549 (American type culture collection, CCL-185) was used throughout this study. For all experiments, cells were maintained at 37°C, 5% CO2 in supplemented DMEM. Epithelial cells were used after the second or third passage. For all experiments, 105 A549 cells were seeded in 6-well tissue culture plates and incubated to ≥ 90% confluence. Monolayers were challenged with 105 spores/ml, 200 µl of supernatant or 200 µl of cell wall extract. Following co-incubation with A. fumigatus spores, cell wall extracts or supernatants, monolayers were washed 3 times with PBS and adherent A549 cells were counted in 3 fields of view at magnifications of 20 or 40 (Nikon Eclipse TS100). Washing was omitted for analyses of contact-dependent damage.
Damage to A549 epithelial cells by the various strains of A. fumigatus was determined using a previously described method at 16, 20 and 24 hr of co-incubation [35]. The 51Cr content of the medium and lysates was measured and the degree of epithelial cell damage was calculated and corrected for spontaneous chromium release by uninfected epithelial cells.
To analyse the localisation of α-1,3-glucan or β-1,3-glucan on A. fumigatus cell walls, isolates (105 or 104 spores/ml) were grown in 8-well slide culture chambers (Nalge Nunc International, Rochester, NY) in supplemented DMEM for 4, 8, 12 or 16 hr. α-1,3-glucan was visualized using 0.1 mg/ml mouse IgMγ MOPC-104E (Sigma, in PBS buffer) as primary antibody and 0.1 mg/ml Alexa Fluor 488 goat anti-mouse IgM (μ chain) antibody (Life technologies, in PBS buffer) [47], [82], [83]. β-1,3-glucan was visualized 5 µg/ml Fc-dectin-1 fusion (kind gift from Dr G.D. Brown, University of Aberdeen) coupled with 15 µg/ml goat anti-human IgG (H+L) Fluorescein conjugated antibody [45], [46]. Briefly, samples were incubated with primary antibodies for 30 minutes, before incubation with the secondary antibodies for 30 minutes in the dark.
To visualise epithelial monolayers co-incubated with A. fumigatus strains, A549 cells were seeded in 2-well slide culture chambers (Nalge Nunc International, Rochester, NY) in supplemented DMEM. At 90% confluence, epithelial monolayers were incubated with A. fumigatus isolates (105 spores/ml) for 16 hr. After washing the monolayers three times with PBS, samples were incubated in supplemented DMEM with 10 µg/ml FITC-labelled concanavalin A (Molecular Probes) and 0.4 mg/ml calcofluor white (Sigma), to visualize respectively epithelial cells and hyphae. Labelling was performed for 30 minutes at 37°C, 5% CO2.
After rinsing with PBS, samples were imaged using a Nikon Eclipse TE2000E microscope with DIC optics, a 20× plan fluor objective or 60× (1.3 NA) plan fluor objective, and equipped with an ORCA-ER CCD camera (Hamamatsu, Welwyn Garden City, UK) driven by the MetaMorph NX1.1 software for image acquisition. For Alexa Fluor 488, FITC and fluorescein, a Nikon B-2A filter cube (excitation filter 470/20 nm BP, dichroic mirror 500 nm LP, emission filter 515 nm LP) was used. For calcofluor white, a Nikon UV-2A filter cube (excitation filter 355/15 nm BP, dichroic mirror 400 nm LP, emission filter 420 nm LP) was used. Images were processed and analysed using the software Image J version 1.47.
Adhesion was tested using a modification of the protocol in Gravelat et al., 2012 [84] as described in Text S1.
The release of lactate dehydrogenase (LDH) was assessed in BALs using the Cytox 96 Non-Radioactive Cytotoxicity Assay kit (Promega) according to manufacturer's instructions. BAL samples were assessed in triplicate and averaged values were normalised to the total amount of protein as measured in triplicate using a bicinchoninic acid assay (BCA) assay (Sigma) according to manufacturer's instructions.
Lungs were homogenised in 1 ml of PBS (pH 7.4) containing protease inhibitor cocktail (Roche) and protein concentration was measured by BCA, using a BSA as standard (Sigma). 9 µg of protein was analysed by western blotting [85]. A 1∶1000 dilution of a α-S100B antibody (Abcam) was used, in parallel with an α-actin antibody (Cell Signaling) for normalisation of loading.
BALs were collected using 3 ml of PBS and a further 5 ml of PBS were added at the time of preparation of the samples for FACS analysis. Cell pellets were resuspended in 1 ml red blood lysis buffer (Sigma). Blocking of the Fc receptor to remove unspecific signal was achieved by incubating the samples with 0.5 µg of an anti-Mouse CD16/CD32 antibody (E-bioscience) in 100 µl of 0.1% BSA PBS. 14 µl of antibody mix was added for labelling of macrophages (α-F4/80-APC-Cy7, 5 µl, Biolegend), leukocytes (α-CD45-PE, 2 µl, E-bioscience) and neutrophils (α-Ly-6G-BV421, 2 µl, Biolegend). Samples was analysed using a BD Fortessa cell analyser. Data acquisition and analysis were performed using respectively the software Diva and FlowJo. For each sample (n = 4, plus 2 controls), cell population size for macrophages (F4/80+) and neutrophils (Ly-6G+) were expressed as cells/ml.
In vitro susceptibility testing of A. fumigatus strains was performed according to the European Committee for Antimicrobial Susceptibility testing (EUCAST) standard method [53]. Caspofungin was tested on A. fumigatus strains in biological and technical triplicate. 1.25 × 105 strains were grown with RPMI1640, 0.165 mol/L MOPS, pH 7.0 and incubated at 37°C for 48 hr. The final concentration of caspofungin tested ranged from 0.03 to 16 ug/ml.
GraphPad Prism was used to interpret data and p values were calculated through Log Rank analysis (for comparative survival), unpaired t tests or 1-way ANOVA tests as indicated. Error bars show the Standard Error of the Mean (SEM). ***p<0.001, 0.001 <**p<0.01, and 0.01 <*p<0.05
Figure S1 shows the strategies for construction and validation of A. fumigatus ΔpacC mutants. Figure S2 shows growth of A. fumigatus isolates on laboratory culture media, as images and as radial growth rates normalised to pH 6.5. Figure S3 shows pH-dependency of PacC processing, as measured using EMSA analyses. Figure S4 shows the experimental set-up and outputs of in-host transcriptomic analysis of A. fumigatus wild-type and ΔpacC activities. Figure S5 shows a heat map of differentially expressed A. fumigatus gene products having predicted signal peptides. Figure S6 shows a heat map of differentially expressed A. fumigatus gene products having putative or demonstrated roles in cell wall biosynthesis. Fig. S7 shows a heat map of differentially expressed A. fumigatus gene products involved in gliotoxin biosynthesis (AFUA_6G09570-AFUA_6G09740). Figure S8 shows qPCR validation of microarray data. Figure S9 shows analysis of A. fumigatus protease activity using a qualitative gelatine degradation assay. Figure S10 shows the electron microscopy of A. fumigatus ΔpacCATCC mutant and the respective parental isolate at 0, 4, 8 and 16 hr of growth. Figure S11 shows equivalent nystatin-mediated killing of wild type and mutant isolates used in this study. Figure S12 shows equivalent adhesion of mutant and wild types isolates to plastic, and to epithelia in vitro. Figure S13 shows immunofluorescence analysis of α-glucan distribution in A. fumigatus germlings. Table S1 and S2 list the A. fumigatus strains and oligonucleotides used in this study respectively. Table S3 lists the A. fumigatus phenotypic testing conditions tested. Dataset S1 shows the temporal analysis of A. fumigatus gene expression during intiation of murine pulmonary aspergillosis. Dataset S2 shows the genes differentially regulated, relative to wild type, in a host-infecting A. fumigatus ΔpacCATCC mutant. Dataset S3 shows the expression of genes encoding secreted gene products which are regulated by either PrtT [9] or PacC, or both transcription factors. Text S1 contains the supplementary material and methods.
The genes and gene products (with accession numbers at http://www.cadre-genomes.org.uk/) studied in this work is pacC (AFUA_3G11970). Also mentioned in the text are prtT (AFUA_4G10120), mep (AFUA_8G07080), alp1 (AFUA_4G11800), β-tubulin (AFUA_7G00250) and the gliotoxin cluster (AFUA_6G09570-AFUA_6G09740).
|
10.1371/journal.ppat.1005127 | Leptomonas seymouri: Adaptations to the Dixenous Life Cycle Analyzed by Genome Sequencing, Transcriptome Profiling and Co-infection with Leishmania donovani | The co-infection cases involving dixenous Leishmania spp. (mostly of the L. donovani complex) and presumably monoxenous trypanosomatids in immunocompromised mammalian hosts including humans are well documented. The main opportunistic parasite has been identified as Leptomonas seymouri of the sub-family Leishmaniinae. The molecular mechanisms allowing a parasite of insects to withstand elevated temperature and substantially different conditions of vertebrate tissues are not understood. Here we demonstrate that L. seymouri is well adapted for the environment of the warm-blooded host. We sequenced the genome and compared the whole transcriptome profiles of this species cultivated at low and high temperatures (mimicking the vector and the vertebrate host, respectively) and identified genes and pathways differentially expressed under these experimental conditions. Moreover, Leptomonas seymouri was found to persist for several days in two species of Phlebotomus spp. implicated in Leishmania donovani transmission. Despite of all these adaptations, L. seymouri remains a predominantly monoxenous species not capable of infecting vertebrate cells under normal conditions.
| In this work we performed a comprehensive evaluation of the infective potential of Leptomonas seymouri, repeatedly isolated from kala-azar patients infected by Leishmania donovani in India and neighboring countries, and have tested the capacity of this monoxenous trypanosomatid to utilize the sand fly vectors permissive for Leishmania donovani. We concluded that despite several genetic adaptations it has developed, Leptomonas seymouri remains a predominantly monoxenous species not able to infect mammalian macrophages either alone or in co-infection with Leishmania. Under certain circumstances it is able to infect mammals, but probably only when the host is immunocompromised by infection with another pathogen, such as Leishmania donovani or HIV.
| Flagellates of the family Trypanosomatidae are single-celled obligatory parasites. They can be either dixenous (i.e. those with two hosts in their life cycle—Trypanosoma, Leishmania, and Phytomonas spp.) or monoxenous (i.e. those having only one host). For decades, monoxenous trypanosomatids of insects were effectively neglected. However, this situation is rapidly changing, as a remarkable diversity of these flagellates is being revealed within insects—a group which is known to be extraordinarily species rich [1,2]. In addition, the study of these parasites is expected to shed light on the origin of the dixenous life cycle (alternation of an insect vector and a vertebrate or plant host). It is generally accepted that the dixenous species have evolved from their monoxenous kins and that this transition has happened independently at least three times during the evolution of Trypanosomatidae, as the dixenous genera Trypanosoma, Leishmania, and Phytomonas are interspersed by the monoxenous genera Angomonas, Blastocrithidia, Blechomonas, Crithidia, Herpetomonas, Kentomonas, Leptomonas, Paratrypanosoma, Sergeia, Strigomonas, and Wallacemonas (S1 Fig) [3,4]. This suggests that some (presumably) monoxenous species may occasionally try switching to dixeny. Indeed, the presence of the monoxenous trypanosomatids in vertebrates has been noted already about 100 years ago [5]. More recently, several monoxenous flagellates belonging to the genera Herpetomonas, Crithidia, Leptomonas, and Blechomonas have been identified from human clinical isolates [6–8]. Importantly, most of them involved immuno-compromised individuals, leading to a hypothesis that these usually non-infectious species may explore new ecological niches in vertebrates that have their immune system suppressed [9,10]. Within this paradigm, about two dozen cases of monoxenous trypanosomatids co-infecting humans along with various Leishmania spp. have been reported almost exclusively from the Indian subcontinent. Most of them implicated causative agents of visceral leishmaniasis (kala-azar) of the L. donovani complex [11]. It was also demonstrated that both dixenous and monoxenous flagellates may be transmitted by the same Phlebotomus vector, yet the evidence is not very strong [12,13]. The cytochrome b and 18S rRNA-based PCR analyses were confined to the isolates from a small geographical area and the identity of non-Leishmania parasites could not be elucidated to the species level.
The species most often recovered from co-infections in leishmaniasis patients is Leptomonas seymouri Wallace, 1959 [14]. Together with all Leishmania spp. it belongs to the subfamily Leishmaniinae (S1 Fig) [15] and was originally isolated from a cotton stainer Dysdercus suturellus (Hemiptera: Pyrrhocoridae) [16]. Nonetheless, when a broad-scale survey of trypanosomatids parasitizing pyrrhocorids throughout the world was undertaken, none of the samples proved to contain L. seymouri [17]. So the question remains whether the original isolate was obtained from a specific host (e.g. species that is evolutionary adapted for parasite's life cycle). L. seymouri can even multiply in plants under experimental conditions [18] proving it to be non-fastidious and able to adapt to quite different environments.
Recent whole-genome analysis of kala-azar clinical isolates from splenic aspirates demonstrated heavy "contamination" with unidentified Leptomonas sp. [19]. This result is not so surprising provided that both parasites are almost indistinguishable by morphology and that Leptomonas outgrows Leishmania in culture [20].
We speculate that several species of monoxenous trypanosomatids are capable of surviving in the hostile environment of the vertebrate body. Molecular details of such adaptation are not worked out, yet it is clear that some monoxenous trypanosomatids must be able to tolerate heat shock up to the temperatures they might experience in warm-blooded vertebrates. Indeed, a number of representatives of the genera Crithidia and Herpetomonas can withstand elevated temperature reaching 37°C [21–23].
In this study we addressed the issue of Leishmania–Leptomonas co-infection from the point of view of the monoxenous partner. To understand molecular mechanisms and biochemical pathways responsible for survival within warm-blooded vertebrates, we have demonstrated that Leptomonas seymouri can withstand elevated temperatures in vitro, sequenced its genome, and assessed transcriptional profiles of cells cultivated in different conditions. Furthermore, we tested L. seymouri ability to survive in Phlebotomus argentipes and P. orientalis, two sand fly species implicated in Leishmania donovani transmission.
Whole genome sequencing of two clinical Indian kala-azar field isolates, a strain resistant to sodium antimony gluconate therapy (Ld 39, May 2000, Muzaffarpur, Bihar) and a strain sensitive to treatment (Ld 2001, February 2000, Balia, Uttar Pradesh), revealed numerous (over 95%) sequences apparently derived from Leptomonas sp. in addition to those of L. donovani [19]. These isolates were cultivated from splenic aspirates in frame of a large screen aimed to understand molecular differences between confirmed kala-azar cases. For precise identification of the co-infecting species we applied an arsenal of molecular tools developed over the years [24–27]. Three genetic loci, namely 18S rRNA, glycosomal glyceraldehyde-3-phosphate dehydrogenase (gGAPDH), and ITS regions were amplified, sequenced and compared with other representatives of the subfamily Leishmaniinae [15]. 18S rRNA sequences of the isolates Ld 39 and Ld 2001 (GenBank accession numbers KP717894 and KP717895, respectively) were identical and indistinguishable from the corresponding sequence of L. seymouri (GenBank accession number AF153040). gGAPDH sequences (GenBank accession numbers KP717896 and KP717897 for isolates Ld 39 and Ld 2001, respectively) were nearly identical with only 1 nt substitution in the coding sequence. They both were very similar (except for the degenerative primer sequences) to the gGAPDH sequence of L. seymouri (GenBank accession number AF047495). 18S rRNA and gGAPDH sequences are informative for higher level taxonomy, and are usually adequate for the genus (and up) level ranking [4,28]. For proper species identification we used other well-established markers, ITS1 and ITS2 [14,20,29]. Their sequences were identical with the exception of a 2 nt-long indel (GenBank accession numbers KP717898 and KP717899 for isolates Ld 39 and Ld 2001, respectively). BLAST search revealed 100% identity with the ITS1-5.8S rRNA region of L. seymouri (GenBank accession number JN848802).
The data presented above allowed us to conclude that the monoxenous co-infectant of the clinical kala-azar isolates Ld 39 and Ld 2001 is L. seymouri. We also would like to note that the cases of co-infections of Leishmania and Leptomonas are likely underreported in the literature, as several sequences attributed to L. donovani in GenBank do in fact belong to L. seymouri. Our analysis of the ITS-containing region, SL, gGAPDH, HSP70, HSP83, RNA polymerase II, α-tubulin and some mitochondrial genes (A6, cytb, COI, COII, COIII, NADH) revealed that 38 out of 170 (22%) and 3 out of 217 (1.4%) ITS sequences of L. seymouri were misidentified as Leishmania donovani and L. tropica, respectively (see S1 Table for GenBank accession numbers).
The presence of monoxenous L. seymouri in co-infections with dixenous L. donovani implies several adaptations to the environment of the human body. One of the important factors to be considered is temperature. Typical monoxenous trypanosomatids of the insect gut are temperature-sensitive and cannot withstand conditions of the warm-blooded vertebrates [6]. In order to investigate temperature resistance of several trypanosomatid species in vitro, we compared growth kinetics of two different Leptomonas species, L. seymouri ATCC 30220 (hereafter used as a proxy of filed isolated Ld 39 and Ld 2001, which were not available) and L. pyrrhocoris H10, under different experimental conditions. Parasites were incubated at temperatures 23°C, 29°C, and 35°C for up to 7 days. The highest temperature (35°C) approximately corresponds to that faced by the flagellates upon transfer from a sand fly into a vertebrate. To imitate the conditions of insect gut and vertebrate blood, SDM and two-phased blood-agar were used, respectively. No considerable difference was observed in growth kinetics of two trypanosomatid species incubated at 23°C in both media. Interestingly, increasing the cultivation temperature to 29°C and 35°C inhibited growth of L. pyrrhocoris, while growth of L. seymouri was not significantly affected (Fig 1). We concluded that L. seymouri is capable of withstanding the elevated temperature reaching that of the human body. In contrast, L. pyrrhocoris is temperature-sensitive and halts its cell division in non-optimal conditions. In all cases, cultivation on blood-agar medium resulted in higher cells density.
Light microscopy of Giemsa stained smears of L. seymouri cultivated under different experimental conditions revealed statistically significant morphological changes (Fig 2). The most noticeable one was shortening of the free portion of the flagellum observed in cells cultivated at high temperature. This phenomenon was observed for both media used but it was more pronounced in blood-agar. Also elevated temperature resulted in more diverse body sizes and shapes with the most conspicuous feature being elongated and tapered posterior end of some cells.
The genome of L. seymouri ATCC 30220 was assembled into 1,222 scaffolds (maximum length 326,845 bp) with N50 of 70,646 bp and a total assembly length of approximately 27.3 Mbp. This is a substantial improvement over the previously reported assembly of the unidentified Leptomonas sp. (14,518 contigs with maximum length of 26,366 and N50 of 3,370 bp) [19]. Both assemblies had almost the same total genome length (27.3 and 27.4 Mb). Importantly, over 85% of the reads could be cross-mapped (length fraction = 0.9; similarity fraction = 0.9) confirming identity of the L. seymouri isolates. The number of annotated protein-coding genes, 8,488, was also within the range of previously reported genomes (6,451 for Phytomonas sp. HART1; 8,309 for Leishmania major; 10,109 for Trypanosoma brucei) [30–32]. Consistent with other trypanosomatids, the protein-coding genes lack conventional introns. The only exceptions reported so far in Trypanosoma spp. and Leishmania spp. are poly(A) polymerase and DEAD/H RNA helicase [32,33]. Indeed, their L. seymouri orthologs also contain introns and thus require cis-splicing for proper expression.
A typical aspect of the L. seymouri genome is that it contains a relatively small number of genes that have undergone tandemly linked duplications. Using a cutoff value of 10−50, the number of genes present in two or more homologous copies has been estimated at about 9.9% in L. seymouri. Same numbers for Phytomonas sp., L. major, T. brucei, and C. fasciculata are 9.6%, 18.3%, 26.0%, and 40.2%, respectively. This is one of the major components determining differences in genome size among these species.
Genomic information was used to predict the metabolic pathways in L. pyrrhocoris and L. seymouri, two phylogenetic kins with different sensitivity to temperature and ability to co-infect vertebrate hosts (S2 Fig). In essence, the metabolism in these two species is very similar, with important features and differences highlighted below. A classical glycolytic pathway, partly inside glycosomes (as inferred from the presence of peroxisome targeting signals), is responsible for the metabolism of various exogenous sugars (S2 Table). Carbohydrate metabolism is characterized by an incomplete aerobic oxidation because one of the classical mitochondrial tricarboxylic acids (TCA) cycle enzymes (NAD-linked isocitrate dehydrogenase) is absent. However, the other TCA cycle enzymes can be used for the inter-conversion of metabolic building blocks required for gluconeogenesis and other biosynthetic purposes (S3 Table). While both L. pyrrhocoris and L. seymouri are able to synthesize their own pyrimidines, they depend on a supply of external purines. They lack the capacity to oxidize aromatic amino acids and require an external supply of most of the essential amino acids, cofactors and vitamins for growth (S4 Table). Both Leptomonas spp. have a fully developed mitochondrion with 9 of the 10 TCA cycle enzymes present, a complete respiratory chain with the respiratory complexes I—IV, and a fully functional mitochondrial F1-ATPase (S5 Table).
Although lacking the alternative oxidase found in many other trypanosomatids, L. seymouri possesses an alternative NADH dehydrogenase gene. Our analysis predicts that it is able to feed on a large variety of polysaccharides, carbohydrates, both hexoses and pentoses, with the anticipated end products of carbohydrate metabolism being acetate, succinate, carbon dioxide, ethanol, alanine, and D-lactate. L. seymouri has a complete set of β-oxidation enzymes, which are associated with the mitochondrion. A few additional lipid-metabolizing enzymes are present in the glycosomes. It appears that the analyzed flagellate does not possess a type-I system of fatty acid synthesis, but makes its fatty acids in the cytosol by the action of a series of elongases (S6 Table). It is able to oxidize 16 of the 20 amino acids, but the necessary enzymes for the metabolism of lysine and the three aromatic amino acids (phenylalanine, tyrosine and tryptophan) are lacking. The urea cycle is not functional since two mitochondrial enzymes of the cycle are missing (S7 Table). The remaining three cytosolic enzymes have all been acquired by lateral gene transfer and allow arginine to be utilized in polyamine biosynthesis. Surface proteins, previously identified in Trypanosoma, Leishmania and Crithidia spp., have also been found in Leptomonas (S8 Table). Homologues of GP63, amastin, 3’-nucleotidase, integral membrane protein, prohibitin, membrane-bound acid phosphatases MBPA1 and MBPA2 and tartrate-sensitive acid phosphatase, but not oligosaccharyl transferase, are present. Protection against oxidative stress in monoxenous trypanosomatids differs from their dixenous kins. In addition to the trypanothione system and the presence of many homologues of tryparedoxins and peroxiredoxins, all monoxenous species analyzed thus far have a bacterial-type catalase acquired by lateral gene transfer (S9 Table).
Enzymes of the RNA interference pathway, namely the homologs of the Argonaute (AGO1) and the two dicer proteins (DCL1 and DCL2) were not detected in L. seymouri (S10 Table). Importantly, they were found in the genome of L. pyrrhocoris arguing that these two closely related species differ in their ability to regulate gene expression by RNA interference.
In the evolution of Trypanosomatidae many events of lateral gene transfer (LGT) have taken place, since genes of bacterial origin are frequently encountered in all trypanosomatid lineages [34]. This suggests that an ancestral flagellate had already acquired such genes, which include a number of enzymes of glycolysis, pentose-phosphate shunt and pyrimidine biosynthesis, as well as trypanothione reductase and pterin transporters [35–37]. Some LGT events including genes involved in sucrose and pentose sugar metabolism, haem synthesis and urea cycle seem to be more recent and specific to the Leishmaniinae clade that comprises Leishmania, Crithidia and Leptomonas spp. [38–40]. Even more recent acquisitions, shared only among Crithidia spp. and Leptomonas spp. include catalase, the diaminopimelate-metabolizing enzymes and those of β-glucosidase, nitroalkane oxidase, phenolic acid dehydrogenase and glycerol dehydrogenase families (S11 Table). In total, 70 out of 586, or 12% of all the metabolic genes analyzed, have resulted from the events of lateral transfer.
For this analysis full proteomes for 23 trypanosomatid species were downloaded from TriTrypDB v. 7.0 and combined with newly annotated proteins from L. seymouri, L. pyrrhocoris, B. ayalai and Paratrypanosoma confusum (S12 Table). Comprehensive characterization of L. seymouri gene family repertoire and its comparison to that of other trypanosomatids may help to shed light on possible adaptations of this species to the dixenous lifestyle. Recently, a comparative genomics approach was used to define a "gene kit" implicated in cell invasion and intracellular parasitism in Leishmania spp. and Trypanosoma cruzi [41]. Authors have found that despite substantial differences in mechanisms of host cell invasion and survival within the host cell, 3,340 orthologous gene clusters are exclusively shared between intracellular parasites when compared to extracellular T. brucei. Many proteins within these clusters were already proven to play a pivotal role in Leishmania and Trypanosoma virulence (e.g. GP63, amastin, ascorbate peroxidase), while functions of other proteins require further detailed investigation.
In our study we were aiming to identify candidate proteins in L. seymouri that may define its ability to occasionally infect warm-blooded organisms. For that purpose Orthologous Groups (OG) presence/absence patterns in L. seymouri were analyzed and compared to those of other trypanosomatids. In the reference dataset for OrthoMCL analysis several Leishmania spp. (medically and veterinary important dixenous species), along with C. fasciculata and L. pyrrhocoris (both never encountered in vertebrates) are of primary interest for comparison with L. seymouri. According to a widely accepted view of trypanosomatid phylogeny, Leptomonas spp. are most closely related to Crithidia spp., and together they form a clade that clusters as a sister group to the genus Leishmania [2,15] (S2 Fig). Firstly, OG content was compared in L. seymouri, C. fasciculata, and L. pyrrhocoris in order to exclude from the analysis OGs that are present in typical monoxenous trypanosomatids. Leptomonas pyrrhocoris has a typical promastigote morphology and dwells in insect species of the family Pyrrhocoridae [17,42], while C. fasciculata uses various culicids as hosts [43–45]. Notably, some representatives of the genus Crithidia (C. hutneri, C. luciliae thermophila) can survive at temperatures of the mammalian and avian bodies [21,22]. Therefore C. fasciculata may possess genes involved in survival at elevated temperatures, and in order to exclude possible biases caused by the presence of C. fasciculata genes in our OrthoMCL analysis, OG repertoire comparisons were performed twice: with and without C. fasciculata in the datasets being compared.
Out of 7,935 L. seymouri OGs, 79 OGs were absent in L. pyrrhocoris, and 26 OGs were absent from both L. pyrrhocoris and C. fasciculata (S13 Table and Fig 3). Our assumption is that among the genes belonging to the above-mentioned groups there are at least several that predispose L. seymouri metabolism to dixeny. Fifty five out of 79 OGs absent in L. pyrrhocoris do not have any functional annotation assigned and thus represent a broad field for further studies (S13 Table). Nevertheless, several genes identified by comparative genomics approach in our study were already proven to play a pivotal role in parasite survival and virulence (see below).
In order to further narrow down the set of such genes we introduced one more condition into the comparison: gene family present in L. seymouri must be also present in all Leishmania species considered in the analysis (L. braziliensis MHOM/BR/75/M2903, L. braziliensis MHOM/BR/75/M2904, L. donovani BPK282A1, L. infantum MCAN/ES/98/JPCM5, L. major MHOM/IL/80/Friedlin, and L. mexicana MHOM/GT/2001/U1103). A reptile parasite L. tarentolae ParrotTarII was not included in the analysis due to its inability to infect warm-blooded organisms [46]. Additional BLASTP search (E-value ≤ 10−10) for proteins belonging to OGs and meeting the criteria stated above was performed in order to determine whether these OGs have related OGs with homologous proteins clustered separately by the sensitive OrthoMCL algorithm. Cases when related OGs have a presence/absence pattern which violates the abovementioned criteria are not discussed here since unambiguous conclusion cannot be made concerning the role of such proteins in L. seymouri thermotolerance.
Sixteen OGs absent from L. pyrrhocoris are shared by L. seymouri and Leishmania spp. Importantly, only 2 of them are absent from both L. pyrrhocoris and C. fasciculata (Fig 3). These two OGs represent a kinase-like protein and a ubiquinol-cytochrome c reductase-like protein. According to the results of additional BLASTP search, the latter protein OG does not have any related OGs and all of its orthologs in the TriTrypDB are annotated as ubiquinol-cytochrome c reductase-like proteins. Aiming to identify homologs of this protein in other species beyond the TriTrypDB, we conducted a BLAST search against the NCBI nr database and found a close homolog only in Strigomonas culicis (ubiquinol-cytochrome c reductase subunit 6, E-value ≤ 10−30, protein accession number: EPY16273.1). The kinase-like protein mentioned above has weak hits with E-value over 10−30 to several other OGs containing protein kinases. Due to the relatively high E-values of the BLAST hits and quite unspecific annotations of kinases within related OGs, this protein was not excluded from our analysis (although several related OGs have absence/presence patterns that differ from the required ones), and its possible role in L. seymouri thermotolerance cannot be ruled out.
Having excluded the requirement of OG being absent from C. fasciculata, the overlap mentioned above extends to 16 OGs (Fig 3), which include one more group with putative protein kinases as well as putative anaphase-promoting complex subunit, putative epsin and several hypothetical proteins with unknown functions. In order to obtain a global picture of corresponding OG distribution for subunits of the anaphase-promoting complex and putative epsin, we extended analysis of OG presence/absence patterns to the whole dataset of 27 trypanosomatid species. As expected, OGs containing these proteins have shown nearly omnipresent distribution (being absent from L. pyrrhocoris as required in our analysis and additionally missing in several Trypanosoma spp.). Additional BLAST search (with more relaxed parameters) for these proteins against proteins belonging to other OGs also did not return any hits. Such results can be explained assuming considerable sequence diversity in these proteins families. For epsins, a group of eukaryotic proteins broadly implicated in clathrin-mediated endocytosis, there is evidence for substantial sequence dissimilarities and lineage-specific protein architecture [47]. Anaphase-promoting complex is a multi-subunit E3 ubiquitin ligase that is necessary for proteolytic degradation of crucial cell cycle regulators, which causes segregation of sister chromatids [48]. Taking into account a universal role of the proteins mentioned above and their phyletic patterns (especially their presence in several monoxenous species) we conclude that they are unlikely to be involved in L. seymouri thermotolerance. Interestingly, 3 OGs containing hypothetical proteins (OG_09193, OG_10013, and OG_10042) within the group of 16 OGs fully satisfy the conditions applied in the study, including the absence of closely related OGs. Moreover, these groups of homologous proteins do not occur in any Trypanosoma spp. and in monoxenous trypanosomatids for which genome sequences are available (except for C. fasciculata). Proteins within these groups represent primary targets for additional studies aiming to reveal mechanisms contributing to L. seymouri thermotolerance.
Prompted by our observation that L. seymouri lacks RNAi machinery (see above) and by patterns of RNAi retention in Trypanosomatidae [49], we also examined L. seymouri for the presence of dsRNA viruses. Two complementary methods, the nuclease digestion assay and immunofluorescence microscopy, were used [50]. Indeed, the anti-dsRNA antibodies detected small sharp dots, which are reminiscent of those found in the virus-positive isolate of Leishmania guyanensis [51]. Importantly, these putative viral particles did not co-localize with the mitochondrion (Fig 4A). The nuclease digestion assay of L. seymouri RNA was performed in parallel with the virus-free Blechomonas pulexsimulantis used as a negative control [52]. It detected dsRNA bands resistant to DNase I and S1 nuclease, which were present in RNA preparations from L. seymouri (Fig 4B). Interestingly, this dsRNAs differ in size from that of the previously characterized LRV1 virus of Leishmania guyanensis (1.5 + 2.9 kb versus 5.3 kb, respectively) [53,54]. It remains to be investigated whether this reflects critical differences in genomic organization of viruses, such as segmented versus whole dsRNA genomes.
To identify genes and/or pathways responsible for thermoresistance of L. seymouri, we profiled whole transcriptomes of the parasites cultivated at low (23°C) and high (35°C) temperature. We presumed that in addition to genetic factors (e.g. chromosome ploidy) regulation of gene expression may also be involved in adaptation to dixeny. Reads passing the filtering step (61.4; 52.5; 39.1 million reads for replicates at 23°C and 61.1; 58.9; 37.9 million reads for replicates at 35°C) were used in subsequent analyses. Out of 8,488 genes identified in the L. seymouri genome 8,482 genes were recovered in our analysis. Results of the FDR test are shown in S3 Fig. In total, 340 genes (4% of the total number) were shown to be differentially expressed at the elevated temperature (S14 Table). Expression of 139 genes (1.6% of the total number) was found to be down-regulated at 35°C, whilst 201 genes (2.4% of the total number) were upregulated at least 1.5 fold (p-value ≤ 0.05). Several interesting cases are discussed in detail below.
Experimental infections of the two proven vectors of Leishmania donovani, Phlebotomus orientalis and P. argentipes, were compared side-by-side. Insects were fed on either blood or sugar meals to mimic the range of conditions which may favor infection (Fig 5A and 5B). On day 2 after infective sugar meal all females of P. orientalis were infected, while the infection rate of P. argentipes females was lower (59%) (Fig 5A). Intensity of infection was generally weak in both species tested. On day 6 p. i. percentages of infected sand flies decreased to 77% and 46% for P. orientalis and P. argentipes females, respectively. On day 9 every third female remained infected, yet most of them harbored only few flagellates.
Infection via the blood meal was less efficient when compared to the sugar meal. On day 2 less than half of blood-fed females of P. orientalis and P. argentipes were infected (47% and 32%, respectively). Freely moving promastigotes were found enclosed in the ingested blood. On days 6 and 9 L. seymouri promastigotes persisted only in a few females (Fig 5B).
The experimental co-infection of sand fly females of P. argentipes were performed by blood meals containing either mCherry- (L. seymouri, ATCC-30220) and/or GFP-expressing (L. donovani, strain GR-374) flagellates. In the control dissection of five sand flies performed just a few hours p. i., both mCherry- and GFP-labeled cells were encountered at about 100 cells of each species per sand fly gut. On day 2 p. i. the infection rate of L. seymouri was lower than that of L. donovani (82.6% versus 95.7%, respectively), and also the intensity of infection with the former species was significantly weaker (Fig 5C). The differences between both parasite species became even more pronounced on day 5 p. i., when the percentage of infected sand flies remained unaltered for L. donovani (86.7%), while it markedly dropped for L. seymouri (13.3%). Moreover, the intensity of infection with L. donovani was high, whereas the few insects still infected with L. seymouri harbored only negligible number of free swimming mCherry-expressing cells (Fig 5C).
Survival of parasites inside mammalian host cells J774 or BMMɸ was evaluated 3, 4, 5 and 6 days p. i. No viable L. seymouri cells were found in macrophages by fluorescent microscopy or in Giemsa-stained smears. In contrast, the control represented by L. donovani survived inside both J774 and BMMɸ cells. Similar results were obtained using peritoneal macrophages from BALB/c mice. The transformation assay has confirmed microscopic observations, as no L. seymouri cells were found after the lysis of macrophages. On the contrary, L. donovani propagated very well under the same conditions. Similar results were obtained when either J774 or BMMɸ macrophages were simultaneously co-infected with both parasites.
Here we performed a multifarious evaluation of the infective potential of L. seymouri, repeatedly isolated from kala-azar patients infected by L. donovani in India and neighboring countries, and have tested the capacity of this monoxenous trypanosomatid to utilize the sand fly vectors permissive for L. donovani. Moreover, we attempted to find genetic and corresponding metabolic adaptations responsible for its survival at 35°C.
Firstly, we have sequenced the whole genome of L. seymouri and compared it with L. pyrrhocoris and C. fasciculata, the only monoxenous species for which high-quality assemblies are available. Twenty six OGs carried by the thermotolerant L. seymouri and absent in these closely related thermosensitive flagellates may potentially be associated with this adaptation. Including dixenous species into the comparative analysis narrowed down our search to just two OGs shared by L. seymouri and five Leishmania spp. and absent from L. pyrrhocoris and C. fasciculata, namely a kinase-like protein and a ubiquinol-cytochrome c reductase-like protein. It was shown previously that protein kinases are involved in amastigote differentiation in Leishmania spp. [58], a process in which temperature switch plays a decisive role [59]. Moreover, our search has identified a number of proteins with specific distribution among trypanosomatid lineages (e.g. absent in Trypanosoma spp. and/or Leishmania spp. but present in monoxenous flagellates) that are prime targets for functional analysis. In any case, the fact that the vast majority of genes within OGs with this phyletic distribution are annotated as hypothetical proteins with unknown function indicates our scarce knowledge of trypanosomatid metabolism.
A number of metabolic changes observed in L. seymouri exposed to elevated temperature are evocative of those in Leishmania amastigotes or T. brucei bloodstream forms in glucose-poor environment [60,61]. For example, inhibition of the de novo synthesis of sterols in L. seymouri resembles Leishmania amastigotes in which the relative abundance of C24-alkyl sterols was significantly decreased upon their differentiation from procyclics [62,63]. Similarly to their dixenous cousins, L. seymouri cells at high temperature reduce the uptake of glucose and shift their acetyl-CoA production in mitochondria from mainly pyruvate-based to the fatty acids-derived [64–66].
The detection of double-stranded viruses in L. seymouri is particularly relevant in the light of recent findings that their presence in Leishmania guyanensis correlates with its virulence and metastatic potential [51,67]. While molecular mechanisms of this phenomenon are just becoming to be understood, it is already clear that the host immune response is rewired [68,69]. We and others have detected dsRNA-containing viruses in several other monoxenous trypanosomatids parasitizing dipteran and heteropteran insects [27,51,70,71], but their relationships to the characterized viruses of Leishmania still remain a mystery. Two analyzed Leptomonas spp. differ in their acceptability for dsRNA viruses. This indicates fundamentally different mechanisms they may utilize to regulate their gene expression.
In summary, we conclude that although L. seymouri has developed several adaptations that allow it to grow well at 35°C, it remains a predominantly monoxenous species not able to infect mammalian macrophages either alone or in co-infection with Leishmania. This agrees with a recent report on selective elimination of Leptomonas from co-cultures with Leishmania [72]. Under certain circumstances it is able to infect mammals, but probably only when the host is immunocompromised by infection with another pathogen, such as L. donovani or HIV [14,73]. However, it is quite likely that such co-infections are much more frequent than the available literature suggests. This conclusion is further supported by our finding that L. seymouri can survive up to 9 days in the same sand fly species that is responsible for the transmission of pernicious Leishmania spp. Therefore, it will be important to analyze samples from patients suffering from visceral and other leishmaniases with primers specific for L. seymouri and related (presumably) monoxenous trypanosomatids to address the possibility that we see only the tip of the iceberg. In addition to the capacity to withstand elevated temperature, other factors, such as its ability to escape the host immune response, may likely play an important role in establishment of the Leptomonas infection in mammals. We cannot exclude the possibility that some isolates of L. seymouri may be exclusively transmitted by sandflies and spend part of their life cycle in vertebrates similar to their Leishmania spp. relatives.
Animals were maintained and handled in the animal facility of Charles University in Prague in accordance with institutional guidelines and Czech legislation (Act Number 246/1992 and 359/2012 coll. on Protection of Animals against Cruelty in present statutes at large), which complies with all relevant European Union and international guidelines for experimental animals. The experiments were approved by the Committee on the Ethics of Animal Experiments of the Charles University in Prague (Permit Number 24/773/08-10001) and were performed under the Certificate of Competency (Registration Number CZU945/05 ext. CZ02573) and the Permission Number 31114/2013-MSMT-13 ext. 24115/2014-MZE-17214 of the Ministry of the Environment of the Czech Republic.
Leptomonas seymouri isolate ATCC 30220 was obtained from the American Type Culture Collection (ATCC, Manassas, USA). It was isolated from the cotton stainer Dysdercus suturellus in the United States in 1959. Leptomonas pyrrhocoris isolate H10 [17], Blechomonas ayalai isolate B08-376 [52] and Leishmania donovani isolate MHOM/ET/2010/GR374 have originated from the research collections at Charles University in Prague, Institute of Parasitology in České Budějovice, and Life Science Research Centre in Ostrava.
Cultures of the monoxenous trypanosomatids were routinely maintained in the Schneider's Drosophila medium (SDM) (Thermo Fisher Scientific, Waltham, USA) supplemented with 10% Fetal Bovine Serum (FBS) (Thermo Fisher Scientific), 50 units/ml of penicillin, 50 μg/ml of streptomycin (both from Sigma-Aldrich, St. Louis, USA), and 10 μg/ml of hemin (Jena Bioscience GmbH, Jena, Germany) at 23°C. All isolates used in this work can also be cultivated in the Brain Heart Infusion (BHI) medium (Sigma-Aldrich) supplemented with 10% FBS and antibiotics as above, or in the two-phased blood-agar medium [74].
To estimate the dynamics of growth, 5 x 104 parasites were seeded into the SDM or the blood-agar medium. Cultures were incubated at 23°C, 29°C, and 35°C for 7 days. Cell numbers were counted using a hemocytometer and plotted in log scale. Morphology of the cells cultivated at low (23°C) and high (35°C) temperature, either in SDM or blood-agar media, was analyzed at day 4 (exponential phase) after staining cells with Giemsa as described previously [75,76]. One hundred cells per sample were measured and analyzed using ANOVA statistical models [77].
Leishmania donovani (MHOM/ET/2010/GR374) transfected with Green Fluorescent Protein (GFP) was cultured in M199 medium (Sigma) containing 20% heat-inactivated FBS (Thermo Fisher Scientific) supplemented with 1% BME vitamins (Sigma), 2% sterile urine, 50 units/ml penicillin, 250μg/ml amikacin (Bristol-Myers Squibb, New York, USA), and 150 μg/ml of geneticin, G418 (Sigma).
The internal transcribed spacer, ITS region of the rRNA locus was amplified using primers IAMWE and Tc5.8-rev and conditions described elsewhere [78]. Total genomic DNA samples of clinical Indian kala-azar field isolates Ld_39 and Ld_2001 were used as templates [19]. The 18S rRNA and gGAPDH genes were PCR-amplified, cloned into the pCR2.1 vector system (Thermo Fisher Scientific), sequenced and analyzed as described previously [79,80]. The obtained sequences were deposited to GenBank with the following accession numbers: KP717894, KP717895 (18S rRNA); KP717896, KP717897 (gGAPDH); KP717898, KP717899 (ITS1 + ITS2 regions).
The Leptomonas seymouri ATCC 30220 genome was sequenced with 100 nt paired-end reads using the Illumina HiSeq 2000 platform (Macrogen, Seoul, South Korea). Prior to assembly, reads were subjected to trimming and filtering using CLC Genomics Workbench v. 7.0 (CLC Inc, Aarhus, Denmark): regions with Phred quality < 20 were trimmed, no more than one N was allowed in the remaining sequence, then TruSeq adapter trimming and a minimum length threshold of 75 nt were applied.
Draft genome of L. seymouri was assembled with the CLC Genomics Workbench v. 7.0 employing a De Bruijn graph-based algorithm with the average coverage of 180 x. Augustus v. 2.5.5 was used to annotate the draft genome of L. seymouri [81]. Prediction accuracy of Augustus was improved by retraining using a training set of L. seymouri conserved proteins. In brief, de novo assembled contigs were searched against proteins in the TriTrypDB v. 7.0 database [82] (BlastX E-value ≤ 10−5) and best BLAST hits were chosen based on the following criteria: a) E-value ≤ 10−30, b) hit length longer than 80 amino acids (aa), c) percent identity higher than 40. Subsequently, a non-redundant training set of 727 high-confidence gene models with unambiguous start site positions was created based on best BLAST hits to annotated proteins from TriTrypDB and RNA-seq coverage data. Non-redundancy of the training set was achieved by excluding genes with more than 70% identity at the amino acid level. Further analysis of the Augustus annotation included manual curation of predicted genes based on transcriptome sequencing data, e.g. removing start sites predicted in regions with no transcriptomic coverage and adding transcribed ORFs >200 aa in length not predicted by Augustus. For tRNA gene prediction tRNAscan-SE Search Server [83] was used with default parameters. For annotating other non-coding RNAs BlastN algorithm (E-value ≤ 10−10) was employed with subsequent manual inspection of BLAST results. As a result, 8,488 genes were annotated in the L. seymouri genome, which has been submitted to the NCBI (BioProject accession number PRJNA285179) and the TriTryp database, a part of the EuPathDB [84].
Orthologous groups are the set of genes descended from a single common ancestral gene, containing both paralogs and orthologs. OGs for L. seymouri proteins were inferred using the OrthoMCL v.2.0 software [85]. Full proteomes for 23 trypanosomatid species were downloaded from the TriTrypDB v. 7.0 and combined with newly annotated proteins from L. seymouri and 3 other trypanosomatid species (Leptomonas pyrrhocoris, Blechomonas ayalai and Paratrypanosoma confusum). The reference protein dataset was subjected to removal of poor quality proteins (based on sequence length and percent of in-frame stop codons), all vs. all BLAST (E-value 10−10) and a clustering procedure implemented in the OrthoMCL algorithm. This resulted in 19,866 OGs, 7,935 of which contained proteins of L. seymouri.
L. seymouri was cultivated at 23°C and 35°C for 75 hrs. Total RNA was isolated from 2.5 x 107 cells using RNeasy Mini Kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer’s instruction. The mRNA-derived libraries were sequenced with 100 nt paired-end reads on the Illumina HiSeq 2000 platform (Macrogen). Total of 3 independent biological replicates were analyzed. The whole transcriptome data from this study have been submitted to TriTrypDB database [82].
Differential gene expression analysis was done using the RNA-Seq tool in CLC Genomics Workbench. Raw reads were subjected to quality-based trimming (regions with Phred quality < 20 were trimmed, no more than one N was allowed in the remaining sequence), adapter trimming, and a minimum length threshold of 30 bp. Processed reads were then mapped to the annotated L. seymouri genome with the following parameters: maximum number of mismatches, 2; minimum fraction of read length mapped, 0.8; minimum identity within the mapped sequence, 0.8; maximum number of best-scoring hits for a read, 30. All libraries were mapped as paired-end, and expression values (RPKM) for each gene were calculated. To identify gene sets that are differentially expressed between the two conditions, the FDR test was employed [86]. Genes with expression fold change ≥ 1.5 and FDR p-value ≤ 0.05 were chosen for further analyses.
Gene ontology (GO) terms for genes up- and down-regulated at high temperature were generated using the Blast2GO plugin in CLC Genomics Workbench [87]. Initially, BlastP search against the NCBI nr database was performed, GO terms associated with all the hits were retrieved, and most appropriate GO terms were selected according to the standard Blast2GO procedure. GO term enrichment was assessed using Fisher's exact test.
For detection of dsRNA viruses, two complementary protocols were used. Cells were stained with 4′,6-diamidino-2-phenylindole (DAPI), mitotracker Red CMXRos (both from Thermo Fisher Scientific) and mouse monoclonal anti-dsRNA antibody (Scicons, Szirák, Hungary), followed by goat anti-mouse IgG–Alexa Fluor 488 (Thermo Fisher Scientific) antibody as described previously [50]. In addition, 50 μg of total RNA isolated using TRI reagent (Sigma-Aldrich) was treated with 1 unit of DNase I (New England Biolabs, Ipswich, USA) at 37°C for 1 hr, followed by digestion with 35 units of S1 nuclease (Sigma-Aldrich) for 45 min at the same temperature. Samples were analyzed on 0.8% native agarose in 1xTAE buffer [88].
A fragment encoding mCherry fluorescent protein was amplified with primers 5´-TTATCCATGGTTAGTAAAGGAGAA-3´ and 5´-TGTTAGCGGCCGCTTATGCGGTACCAGAACC-3´ using plasmid p2686 as a template [89]. The resulting 745 bp fragment was cloned into the pF4T7polNLS1.4sat vector digested with NcoI and NotI replacing the T7 polymerase ORF [90]. Log-phase L. seymouri cells (4 x 107) were transfected with 15 μg of SwaI-linearized pF4mCherry1.4sat as described before [91]. Recombinant clones were selected on agar—BHI growth medium supplemented with 10% FBS, 40mM HEPES, pH 7.4 and nourseothricin (Jena Bioscience) at final concentration of 250 μg/ml. Expression of mCherry was confirmed by fluorescence microscopy.
Colonies of two sand fly species, Phlebotomus orientalis and P. argentipes, both representing major proven vectors for L. donovani, were maintained under standard conditions as described elsewhere [92]. Females of both colonies were fed either through a chick-skin membrane on suspension of heat-inactivated rabbit blood containing exponentially growing 1 x 107 promastigotes per ml of blood or on 20% sucrose solution containing 5 x 107 promastigotes per ml. In order to recognize sugar-fed females, the sucrose solution was stained by indigo carmin. Blood- and sugar-fed females were kept at 26°C with free access to 50% sucrose solution by day 1 post infection (p. i.).
Sand fly females were dissected at different intervals p. i. (1–2, 5–6 and 7–9 days). Numbers and location of flagellates in the sand fly gut were checked microscopically. Parasite loads were graded as previously described, i.e.: light (< 100 parasites/gut), moderate (100–1,000 parasites/gut) and heavy (> 1,000 parasites/gut) [93].
Females of P. argentipes were fed through a chick-skin membrane on suspension of heat-inactivated rabbit blood containing 1 x 106 per ml promastigotes of Leptomonas seymouri mCherry (passage 4) and 1 x 106 per ml promastigotes of Leishmania donovani (MHOM/ET/2010/GR374) GFP (passage 10) originating from exponentially growing cultures. Assorted blood-fed females were kept at 26°C with free access to 20% sucrose solution. Sand fly females were dissected on days 2 and 5 p. i., and the presence of parasites as well as other characteristics were analyzed as described previously [93].
Macrophage cell line J774 was cultured in complete RPMI-1640 medium (Sigma) containing 10% FBS, 100 U/ml of penicillin, 100 μg/ml of streptomycin, 2mM of L-glutamine, and 0.05 mM of β-mercapto-ethanol (all from Sigma) at 37°C with 5% CO2. Bone marrow was obtained by flushing of tibias and femurs of BALB/c mice and flagellates were cultured in complete RPMI-1640 medium (Sigma) supplemented as above along with 20% of L929 fibroblast cell culture supernatant serving as a source of macrophage colony-stimulating factor at 37°C with 5% CO2. The differentiation from bone marrow precursor cells to bone marrow-derived macrophages proceeded for 7 to 8 days in sterile polystyrene Petri dishes. The bone marrow derived macrophages (BMMɸ) were washed and seeded into plates at density of 5 x 105 cells per ml. Consequently, stationary cultures of Leishmania donovani (GFP), Leptomonas seymouri (mCherry), alone or in combination were added in ratio of 8:1 (parasites: BMMɸ). Three days p. i. BMMɸ were extensively washed with pre-warmed RPMI-1640 to remove excess of parasites and the viability of trypanosomatids was monitored by fluorescence microscope Olympus CX-31 (Olympus, Tokyo, Japan) up to 6 day p. i. In addition, Giemsa staining was used to analyze intracellular forms in macrophages by light microscopy. All experiments were performed in two independent biological replicates.
To analyze survival of parasites, the transformation growth assay was used [94]. In brief, macrophages infected with Leishmania donovani, Leptomonas seymouri, alone or in combination for 96 hrs were extensively washed with RPMI-1640 and lysed with 0.016% SDS in RPMI-1640 for 7 min at room temperature to release their intracellular forms. The lysis reactions were neutralized by RPMI-1640 supplemented with 17% heat-inactivated FBS. Parasites were spun down at 3,200 rpm for 10 min at 4°C, washed in RPMI-1640, and re-suspended in a relevant promastigote medium (BHI or M199) supplemented with an appropriate selective antibiotic at 23°C. For macrophages co-infected with both parasites, two types of media and antibiotics were assessed. The status of viable parasites was checked for 6 consecutive days.
|
10.1371/journal.pntd.0001487 | Evaluation of the Human IgG Antibody Response to Aedes albopictus Saliva as a New Specific Biomarker of Exposure to Vector Bites | The spread of Aedes albopictus, a vector for re-emergent arbovirus diseases like chikungunya and dengue, points up the need for better control strategies and new tools to evaluate transmission risk. Human antibody (Ab) responses to mosquito salivary proteins could represent a reliable biomarker for evaluating human-vector contact and the efficacy of control programs.
We used ELISA tests to evaluate specific immunoglobulin G (IgG) responses to salivary gland extracts (SGE) in adults exposed to Aedes albopictus in Reunion Island. The percentage of immune responders (88%) and levels of anti-SGE IgG Abs were high in exposed individuals. At an individual level, our results indicate heterogeneity of the exposure to Aedes albopictus bites. In addition, low-level immune cross-reactivity between Aedes albopictus and Aedes aegypti SGEs was observed, mainly in the highest responders.
Ab responses to saliva could be used as an immuno-epidemiological tool for evaluating exposure to Aedes albopictus bites. Combined with entomological and epidemiological methods, a “salivary” biomarker of exposure to Aedes albopictus could enhance surveillance of its spread and the risk of arbovirus transmission, and could be used as a direct tool for the evaluation of Aedes albopictus control strategies.
| Aedes-borne viruses like dengue and chikungunya are a major problem in Reunion Island. Assessing exposure to Aedes bites is crucial to estimating the risk of pathogen transmission. Currently, the exposure of populations to Aedes albopictus bites is mainly evaluated by entomological methods which are indirect and difficult to apply on a large scale. Recent findings suggest that evaluation of human antibody responses against arthropod salivary proteins could be useful in assessing exposure to mosquito bites. The results indicate that 88% of the studied population produce IgG to Ae. albopictus saliva antigens in Reunion Island and show that this biomarker can detect different levels of individual exposure. In addition, little cross-reactivity is observed with Aedes aegypti saliva, suggesting that this could be a specific marker for exposure to Aedes albopictus bites. Taken together, these results suggest that antibody responses to saliva could constitute a powerful immuno-epidemiological tool for evaluating exposure to Aedes albopictus and therefore the risk of arbovirus infection.
| The incidence of arthropod-borne disease is on the rise and mosquito-borne diseases in particular constitute a world-wide threat [1]. In Asia, Africa and South America, arbovirus diseases are re-emerging, notably dengue and chikungunya. According to the World Health Organization, there are 50 million cases of dengue fever every year and the number of countries declaring cases is increasing [2] Chikungunya is an emerging arbovirus [3] and several outbreaks have been recorded, such as the 2006 epidemic on Reunion Island in the Indian Ocean [4]. The threat of these diseases in the developed world is real with, in addition to the chikungunya outbreak in Italy in 2007 [5], sporadic autochthonous cases of dengue and chikungunya recently reported in Southern France [6]. Therefore, epidemiological tools for evaluating such risks are urgently needed in both developing and developed countries. Aedes aegypti and Aedes albopictus are both vectors of the dengue and chikungunya viruses, and Ae. albopictus populations are dramatically expanding worldwide. Epidemiological evaluation of Aedes-borne diseases is currently based on pathogen detection in human populations and entomological methods. The exposure of human populations to Aedes is currently evaluated by mapping breeding sites and using mosquito-capture strategies. But these methods have substantial limitations when it comes to large-scale studies in the field, e.g. vector density and transmission risk are estimated by counting immature Aedes in breeding sites to derive House and Breteau Indices, a process which is too demanding for regular implementation in the field [7], especially in the urban setting. In addition, current methods for evaluating Aedes exposure are mainly applicable at the population level and cannot be used to gauge the heterogeneity of individual exposure. In order to improve vector control and follow the risk of arbovirus transmission, much effort is being devoted to developing new, simple, rapid and sensitive indicators to evaluate human exposure to Aedes bites and thus the risk of arbovirus transmission in exposed populations. One promising approach is based on the idea that exposure could be directly assessed by measuring human-vector contact as reflected by the human antibody (Ab) response to arthropod salivary proteins [8]. At the time of biting, the female mosquito injects saliva containing biologically active molecules to favour feeding and some of these are highly immunogenic [9]. Human Ab responses to the saliva of a number of vectors, including Triatoma (Chagas disease) [10] and Phlebotomus (Leishmaniasis) [11], have been identified as promising biomarkers for vector exposure. Ab responses to the saliva of Glossina (the vector of Human African Trypanosomiasis) have been shown to have high diagnostic value [12]. For mosquitoes, Ab responses to whole saliva have been correlated to human exposure to Culex mosquitoes [13], and Anopheles gambiae [14], Anopheles dirus [15] and Anopheles darlingi [16], vectors of Plasmodium. Recently, it has been shown that the IgG response to whole An. gambiae saliva could be a useful biomarker for evaluating the efficacy of malaria vector control [17]. Studies on Ab responses to Aedes saliva have tended to focus on human allergic reactions [18] and the identification of the immunogenic proteins [19] although they have shown that quantitative evaluation of anti-saliva Ab responses (IgG and specific isotypes) could give a measure of human exposure to biting Aedes [20], [21]. It was recently demonstrated that IgM and IgG responses to Ae. aegypti saliva could be used to estimate exposure in transiently exposed populations [22]. Finally, recent data showed that IgE and IgG4 responses to Ae. aegypti saliva could be detected in young Senegalese children during the exposure season [23]. The present study addresses one important application of this salivary biomarker as a tool to evaluate the specific exposure of individuals to Ae. albopictus bites. Human IgG responses to Ae. albopictus saliva (salivary gland extracts; SGE) were measured in adults living on Reunion Island. In this area, Ae. albopictus represents the only Aedes species which is known to bite humans and is the unique vector of chikungunya. Ae. aegypti, which is non anthropophilic in Reunion Island is totally absent from the study area [24]. To check the specificity of this biomarker for Ae. albopictus, cross-reactivity was tested by comparing IgG Ab levels i) to Ae. aegypti SGE in individuals from Reunion Island and ii) in sera from Bolivian subjects who had only ever been exposed to Ae. aegypti species.
All studies followed ethical principles as stipulated in the Edinburgh revision of the Helsinki Declaration. The studies in La Reunion and the North of France were approved by a French Ethics Committee (the Sud Ouest, Outre Mer Ethics Committee, 25/02/2009) and authorized by the French Drug Agency (AFFSAPS, Ministry of Health; 12/01/2009). The study in Bolivia was approved by the Bolivian Committee of Bioethics (September 2006) and the Institut de Recherche pour le Dévelopement (IRD) “Comité Consultatif de Déontologie et d'Ethique” (July 2006). Written informed consent was obtained from every subject.
The study populations were from two different areas, namely Reunion Island and Bolivia, for specific exposure to Ae. albopictus or Ae. aegypti, respectively.
In the south of Reunion Island (Le Tampon), Ae. albopictus is abundant and is found up to 1200 meters in winter. Chikungunya transmission was high during the 2006 epidemic [25]. Blood samples were collected in May–June 2009 during the seasonal peak of Ae. albopictus exposure, from adults of between 18 and 30 years of age (n = 110). Subjects exposed only to Ae. aegypti were randomly selected from a large study conducted in the city of Santa Cruz de la Sierra, Bolivia (n = 104) and pair-matched for age with the Reunion Island subjects. Sera from unexposed individuals (n = 18) in a region free of either Ae. albopictus or Ae. aegypti (North of France) were used as a negative control.
SGE were obtained from 10 day-old uninfected females reared in insectaries. Ae. albopictus was bred from larvae collected in the field in Reunion Island (Direction Regionale des Affaires Sanitaires et Sociales, Saint Denis, Reunion Island) and the Bora-Bora strain of Ae. aegypti was used (IRD, Montpellier, France). Briefly, two days after a blood meal, the mosquitoes were sedated with CO2 and then their salivary glands were dissected out and transferred into a tube containing 30 µl of phosphate buffered saline (PBS). The dissected glands were then pooled in 30 or 60 pairs per batch and frozen at −80°C before extraction. A simple technique consisting of 3 successive freeze-thaw cycles in liquid nitrogen was used to disrupt the membranes. The soluble SGE fraction was then separated by centrifugation for 20 minutes at 30,000 g at +4°C. The concentration of protein was evaluated by the Bradford method (OZ Biosciences) after pooling of the different batches to generate a homogenous SGE for immunological assessment. SGEs were then stored at −80°C before use.
An enzyme-linked immunosorbent assay (ELISA) was carried out on Maxisorp plates (Nunc, Roskilde, Denmark) coated with Ae. albopictus or Ae. aegypti SGE, (0.8 µg/ml PBS) at 37°C for 150 min. Plates were blocked using 250 µl of protein free Blocking-Buffer (Pierce, Thermo Fisher, France) for 60 minutes at room temperature. Individual sera were incubated in duplicate at a 1/100 dilution in PBS-Tween 1%, 4°C overnight. Monoclonal mouse biotinylated Ab against human IgG (BD Pharmingen, San Diego, CA) was incubated at a 1/1000 dilution for 90 minutes at 37°C. Peroxidase-conjugated streptavidin (GE healthcare, Orsay, France) was added at 1/1000 for 60 minutes at 37°C. Colorimetric development was carried out using ABTS (2,2′-azino-bis (3-ethylbenzthiazoline 6-sulfonic acid) diammonium, Pierce) in 50 mM citrate buffer (pH 4) containing 0.003% H2O2, and absorbance was measured at 405 nm. Each test sample was assessed in duplicate wells and in a blank well containing no antigen (ODn) to measure non-specific reactions, as previously described [17], [23], [26]. Individual results were expressed as the ΔOD value calculated using the equation ΔOD = ODx-ODn, where ODx represents the mean of the OD readings in the two antigen wells. A subject was considered as an “immune responder” if the ΔOD result was higher than the mean ΔOD+(3 SD) for unexposed individuals (negative control). The threshold of positivity was 0.271 for IgG against Ae. albopictus and 0.161 for IgG against Ae. aegypti.
Graph Pad Prism Software (San Diego, CA USA) was used to analyse the data. After confirmation of non-normal distribution, a non-parametric Mann-Whitney test was used to compare Ab levels between two independent groups, and a non-parametric Kruskal-Wallis test was used for comparisons between more than two groups. A Spearman test was used to assess the correlation between IgG levels against Ae. albopictus and Ae. aegypti SGEs. All differences were considered significant at p<0.05.
IgG responses to Ae. albopictus SGE were evaluated in individuals from Reunion Island and North of France (Figure 1). In the unexposed group, one individual IgG response was slightly above the cut-off value (ΔOD = 0.297) (Figure 1). In contrast, a high percentage (88%) of the exposed group from La Reunion responded positively to the anti-SGE IgG Ab. Although considerable differences in specific Ab level were observed between exposed individuals (ΔOD from 0.034 to 3.308), a significant difference in specific IgG level was observed between the exposed group (median = 1.067) and the unexposed group (median = 0.015) (p<0.0001, Mann-Whitney test).
Cross-reactivity between Ae. albopictus and Ae. aegypti SGEs was evaluated by two complementary approaches. First, the specific IgG response to both SGEs (Figure 2) was evaluated in individuals only exposed to Ae. albopictus (from Reunion Island); in parallel, IgG responses to both SGEs were assessed in individuals only exposed to Ae. aegypti (from Bolivia). In the Bolivian group, 16% showed a positive IgG response against Ae. albopictus SGE with only one strong response (ΔOD>1). The median (0.107) level of IgG against Ae. albopictus SGE was significantly lower in the Bolivians than in the Reunion Island group (1.067) (P<0.0001, Mann-Whitney test). The IgG response to Ae. aegypti SGE was also evaluated in both groups of subjects. As expected, the Bolivian group (exposed only to Ae. aegypti) presented high levels of IgG against Ae. aegypti SGE with 76% of immune responders. In contrast, only 19% of the subjects from Reunion Island were responders to Ae. aegypti SGE and the median (0.068) IgG level was very low compared with the Bolivian group (0.991) (p<0.0001, Mann Whitney test). Only one individual from Reunion presented very high level of IgG to Ae. aegytpi SGE (ΔOD>3). In addition, in 58% of double immune responders from Reunion Island, the level of IgG against Ae. albopictus SGE was above the 75% percentile value (ΔOD = 2.426; data not shown). In the Bolivian double immune responders, the corresponding figure for Ae. aegypti SGE was 50% (ΔOD value of 75% percentile = 2.341). These results indicate that IgG to Ae. albopictus and Ae. aegypti SGE are cross reactive, particularly in individuals presenting a very high level of IgG.
Secondly, the level of specific IgG Ab against both SGEs was compared by a statistical correlation analysis (Figure 3). High positive correlation between IgG against Ae. albopictus SGE and Ae. aegypti SGE was observed for both the Reunion Island group (r = +0.445; P<0.0001, Spearman test) and the Bolivian group (r = +0.617; P<0.0001, Spearman test). For each population, IgG cross-reactivity was low with few individuals responding to both Ae. albopictus and Ae. aegypti SGEs. For the Reunion Island group, only 5 individuals showed a strong IgG response to Ae. aegypti SGE. These correlations indicate that cross-reactivity between the two Aedes species may depend on IgG level. In the Bolivian group, the same trend is observed with only 8 individuals showing a strong IgG response to Ae. albopictus SGE.
In this study, we investigated IgG responses to Ae. albopictus SGE in exposed adults from Reunion Island where Ae. albopictus—the only anthopophilic Aedes species—transmits chikungunya. First, specific IgG responses were high in the exposed group and significantly different to those observed in an unexposed population from Europe: 88% of exposed individuals developed IgG against Ae. albopictus SGE. In addition, specific IgG Ab levels showed considerable inter-individual variations. Since the intensity of exposure in a given population living in the same area can obviously vary between individuals, these results suggest that the anti-SGE IgG response may be a reliable biomarker for exposure to Ae. albopictus bites. Furthermore, the significant difference in Ab levels between unexposed and exposed individuals shows that this biomarker could distinguish individuals exposed to Ae. albopictus bites. A useful biomarker for Ae. albopictus bites needs to be highly specific and devoid of immune cross-reactivity with other Aedes species. We evaluated the cross-reactivity between two species using complementary approaches, i.e. in individuals only ever exposed to Ae. aegypti and by analysing IgG levels against both SGEs. In the Bolivian group only exposed to Ae. aegypti, 16% of individuals responded to Ae. albopictus SGE (heterologous ELISA). The ΔOD values for these “cross-reactive” individuals are characterised by very low IgG levels whereas a high percentage (88%) and high IgG levels (homologous ELISA) were observed in subjects from Reunion Island. In addition, IgG responses against Ae. aegypti SGE are significantly different: 76% of immune responders in Bolivia compared with 19% in Reunion Island. Interestingly, we observed that IgG cross-reactivity was mainly detected in high immune responders. In Reunion Island, in 58% of double immune responders, the level of IgG against Ae. albopictus SGE was above the third quartile. In parallel, in Bolivia, the level of IgG against Ae. aegypti SGE of 50% of double immune responders was above the third quartile. Taken together, these results suggest that there is cross-reactivity between Ae. albopictus SGE and Ae. aegypti SGE, especially in high immune responders. Further investigations would be required to identify species-specific salivary antigens. IgG response to Ae. albopictus SGE has been detected in individuals exposed to the bites of this mosquito [27]. It can be hypothesised that, in Reunion Island, the observed specific IgG responses were elicited as a result of antigenic stimulation following biting by Ae. albopictus. In the urban area of Reunion Island, Ae. albopictus is highly antropophilic [28] and characterised by numerous “artificial” breeding sites [25] which could explain the high percentage (88%) of specific responders to Ae. albopictus SGE. These results point up the relevance of this approach to developing a specific biomarker for exposure to Ae. albopictus. However epidemiological factors—history of exposure, genetic background, immune tolerance, etc.—have to be taken into account when explaining variations in responsiveness. Further longitudinal studies could focus on this. To our knowledge, measuring IgGs against Ae. albopictus SGE represents the first direct method for evaluating human exposure to Ae. albopictus and this parameter probably represents the first genuine biomarker for man-vector contact. This method could help overcome the shortcomings of the current methods which only give indirect measurements of exposure to Ae. albopictus and therefore have considerable limitations for evaluating the risk of arbovirus transmission [7]. The current standard methods—immature stage counting and trapping techniques—are both “static” and mainly target “household exposure”, ignoring all the anthropogenic factors that can affect exposure (e.g. water storage practices). Ae. albopictus and Ae. aegypti are diurnal mosquitoes, biting both indoors and outdoors, and these characteristics may complicate the assessment of exposure using conventional methods. Using anti-SGE IgG responses to evaluate exposure to Ae. albopictus, highly heterogeneous Ab levels were observed between exposed individuals, as previously reported for another vector [26]. It could be hypothesized that different levels could reflect the intensity of exposure to biting vectors, e.g. the experimental results indicate that a high Ab response is the result of high exposure and the ame association is also observed for low Ab levels [29]. This has also been observed in human populations in the field where arthropods vectors are endemic. In exposed individuals from endemic areas, it has been shown that the level of anti-saliva Ab was closely associated with the intensity of exposure to vector bites [12], [23], [26], [30]. Therefore, this could be a useful tool for comparing the exposure to Ae. albopictus at different sites in a given study area or between different areas, and could be useful for evaluating the efficacy of vector control. In addition, recent findings in the malaria field have shown that Ab responses to saliva antigens are useful in the assessment of low-level exposure to Anopheles bites [31]. Detection of low level exposure in newly colonized areas is of particular interest for Ae. albopictus due to its ongoing worldwide spread, especially in urban contexts. Moreover, a clear correlation between larval indices and pathogen transmission is difficult to establish when the level of exposure is low [32], [33]. In both cases, evaluation of the anti-saliva IgG response could complement entomological methods. However, several validation steps (e.g. seasonal variation and correlation with entomological measurements) will have to be checked.
In this study, we measured the Ab response to whole SGE. As long as any salivary proteins are shared with other species or genera, the degree of cross-reactivity with major other Aedes vectors will have to be assessed. Thus, cross-reactivity was investigated between Ae. albopictus and Ae. aegypti, closely related species with shared salivary proteins [34], [35]. Only weak cross-reactivity was detected between Ae. albopictus and Ae. aegypti SGEs, mainly observed in high immune responders. This may suggest that species-specific proteins are more immunogenic than genus-shared proteins. Species-specificity has been already reported with several Aedes species [20], [23], [36] and for a broad range of vectors [12], [29], [37]. In contrast, western-blot analysis reveals extensive cross-reactivity and shows that some antigens are common to all Aedes species [21], [38], [39]. This low level of cross-reactivity also raises the roles of intensity and history of exposure in determining the acquired IgG response against SGE. Individuals from Reunion Island are unlikely to have been exposed to Ae. aegypti because this mosquito is not found in town and its breeding sites are restricted to natural habitats [24]. Cross-reactivity between salivary proteins common to all members of the Aedes genus seems therefore to be the most likely explanation of the observed IgG responsiveness to Ae. aegypti SGE in Reunion Island. Travelling could also lead to contact with Ae. aegypti which is present in most of the islands of the Indian Ocean.
To enhance the usefulness of this biomarker for large-scale applications and to exclude cross-reactivity, an Ae. albopictus-specific salivary antigen needs to be identified. In malaria, only one peptide in whole Anopheles salivary antigen is an efficient biomarker for exposure to Anopheles bites [30]. To improve the sensitivity and the specificity of detection, an immuno-proteomic study is currently underway to identify Ae. albopictus-specific proteins and peptides.
The study described here represents the first step for estimating human exposure to Ae. albopictus by quantifying the IgG response to vector salivary antigens. In an area of chikungunya transmission, it was shown that the level of Ab against Ae. albopictus SGE can be used to identify individuals who have been exposed to the bites of this important vector. Low level cross-reactivity was observed with Ae. aegypti SGE suggesting that it will be possible to develop a specific biomarker for human exposure to biting Ae. albopictus. By combining the use of such a biomarker with classical entomological and epidemiological methods, it could enhance the assessment of human exposure to Ae. albopictus and therefore contribute to both accurate prediction of the risk of arbovirus transmission and evaluation of the efficacy of vector control.
|
10.1371/journal.pcbi.1003784 | Kinetic Memory Based on the Enzyme-Limited Competition | Cellular memory, which allows cells to retain information from their environment, is important for a variety of cellular functions, such as adaptation to external stimuli, cell differentiation, and synaptic plasticity. Although posttranslational modifications have received much attention as a source of cellular memory, the mechanisms directing such alterations have not been fully uncovered. It may be possible to embed memory in multiple stable states in dynamical systems governing modifications. However, several experiments on modifications of proteins suggest long-term relaxation depending on experienced external conditions, without explicit switches over multi-stable states. As an alternative to a multistability memory scheme, we propose “kinetic memory” for epigenetic cellular memory, in which memory is stored as a slow-relaxation process far from a stable fixed state. Information from previous environmental exposure is retained as the long-term maintenance of a cellular state, rather than switches over fixed states. To demonstrate this kinetic memory, we study several models in which multimeric proteins undergo catalytic modifications (e.g., phosphorylation and methylation), and find that a slow relaxation process of the modification state, logarithmic in time, appears when the concentration of a catalyst (enzyme) involved in the modification reactions is lower than that of the substrates. Sharp transitions from a normal fast-relaxation phase into this slow-relaxation phase are revealed, and explained by enzyme-limited competition among modification reactions. The slow-relaxation process is confirmed by simulations of several models of catalytic reactions of protein modifications, and it enables the memorization of external stimuli, as its time course depends crucially on the history of the stimuli. This kinetic memory provides novel insight into a broad class of cellular memory and functions. In particular, applications for long-term potentiation are discussed, including dynamic modifications of calcium-calmodulin kinase II and cAMP-response element-binding protein essential for synaptic plasticity.
| Cellular memory exists in a wide range of organisms from prokaryotes to eukaryotes and can persist, in some cases, for days at a time. Mounting evidence supports the notion that cells indeed can retain information associated with previous environmental exposures via posttranslational modifications. Molecules with multiple modification sites, such as those regulated by changes in methylation or phosphorylation, are expected to play an important role in this process. However, how such epigenetic cellular memory is preserved is not yet fully understood, and theoretical models must be developed. Here we demonstrate that long-term maintenance of the modification state occurs as a result of enzyme-limited competition in a wide class of multimeric protein systems consisting of modification reactions that share enzymes. This maintenance is explained by very slow relaxation resulting from the limited amount of enzymes available. Therefore, memory is easily affected by continuous changes in external stimuli. The proposed mechanism for kinetic memory does not require any fine-tuning in system parameters and is applied to a broad class of cellular memory including long-term potentiation of synapses.
| The importance of cellular memory, in which information from experienced environmental exposures is preserved within cellular states, has received a great deal of attention in recent years. The capability of cells to translate environmental exposures into cellular memory has been reported in various organisms, ranging from bacteria to unicellular protozoa and multicellular vertebrates [1]–[5]. Such examples of cellular memory are thought to result from stored epigenetic changes that are not restricted to histone modifications but rather include long-term modifications (e.g., phosphorylation, methylation, and acetylation) of proteins and DNAs that regulate gene expression and thereby affect cellular states [6]–[9]. Generally, cellular epigenetic memory is regarded to occur more slowly than elementally biochemical reactions without affecting the genome sequence, and is considered to be important for various cellular functions, such as adaptation to external stimuli, cell differentiation, and synaptic plasticity [1], [4], [5], [9], [10].
One of the most prominent examples of cellular memory is long-term potentiation (LTP) for synaptic plasticity, which is defined by an increase in the synaptic strength over a long time span. Persistent phosphorylation of /calmodulin-dependent protein kinase II (CaMKII) is known to be particularly important in early LTP [6], [7]. CaMKII phosphorylation is elevated by transient increases in the concentration of and sustained even after decreases in the concentration of . Similarly, in late LTP, the persistent phosphorylation of transcription factors such as cyclic AMP-response element-binding protein (CREB) is important [6], [8]. Another important example of cellular memory is found in the determination of cell fates. For example, when nerve growth factor is administrated to PC12 cells, extracellular-signal-regulated kinase (ERK) is persistently phosphorylated and transmits information to downstream molecules, eventually leading to cell differentiation [9].
Such cellular memory consists of three events: induction, maintenance, and expression. Signaling input can modify the state of the molecule in question, and the modification can be maintained over a time span longer than the time scale of elemental biochemical reactions (e.g., minutes to days). Such modifications can result in changes in expression or in the activation of other molecules. In this study, we focus on the question of how modified states are maintained over a long duration.
One possible avenue for addressing this question is the use of the attractor concept, which includes the “memories-as-attractors” viewpoint whereby dynamical systems governing the modification states have multiple attractors (steady states), in each of which memory is stored as a stable modification state of the substrates [11]–[14] (see Fig. S1A). For example, models for the persistent phosphorylation of CaMKII have been proposed in which the phosphorylation can take on two stable states, such that hysteresis appears against the change in [15]–[17]. In an in vitro experiment, however, CaMKII did not show bistability; it only showed ultrasensitivity against the change in [18]. There have been reports that modification levels are shifted continuously upon stimulation, rather than taking only a few discrete states [18] (see also [19]); this continuity of modification states cannot be explained by the multistability model. Furthermore, the relaxation time: 20 minutes after inhibition of CaMKII in ref [20] suggests long-term dynamics without resorting to the bistability discussed therein. In addition, the inclusion of positive feedback processes in gene expression dynamics, which are necessary for the maintenance of multiple states, requires energy, and as a result, memory retention incurs housekeeping costs. In summary, the stability of time-invariant attractors is important in some cases, but in other cases it cannot be explained by the “memories-as-attractors” viewpoint.
Thus, it is important to identify other forms of memory that allow very slow changes of cellular states following transient stimuli. In this respect, we propose “a kinetic memory hypothesis” for epigenetic cellular memory. In this scheme, memory is stored as a slow relaxation process far from an attractor, in which slowness enables long-term maintenance of an embedded state (see Fig. S1B). When cells are stimulated, their cellular states are shifted continuously and are thus kept apart from attractors, and relaxation occurs more slowly than with elemental chemical reactions. Based on this slowness in relaxation, stimulus-induced excited states are memorized over long time spans. This is in strong contrast to the memory-as-attractor scheme, in which new stable states are generated or in which different attractors are selected to store the memory of the stimulus. In kinetic memory, states can be changed continuously depending on the magnitude of the stimulus, and thus epigenetic kinetic memory is more flexible than attractor memory. However, to date, an explicit biochemical model that realizes kinetic memory has not been proposed. In this study, we apply the kinetic memory scheme to model processes underlying cellular memory.
At first, we introduce a chained modification model showing very slow kinetics (Fig. 1). This model consists of a substrate with modification sites and a catalyst, where and denote the substrates and a substrate-catalyst complex with -th modified sites, respectively, and denotes the catalyst. The total abundance of each is defined as conservation quantities given as and , respectively. The use of ‘modification’ here refers to changes such as phosphorylation and methylation. In the first model we study, the catalyst can only facilitate each demodification reaction of the substrate, which is achieved by assuming that enzymes used in modification reactions are always maintained at sufficiently high levels so as not to be rate-limiting. Or, the first-order kinetics of the modification reaction may be considered to be derived from the autophosphorylation reaction (e.g., CaMKII phosphorylation reaction). However, the same results can be obtained, even though both the modification and demodification are catalytic reactions as described below in the kinase-phosphatase model.
The whole reactions in the chained modification model are described below.where and denote the substrates and a substrate-catalyst complex with modified sites, respectively, and denotes the catalyst.
Here, the formation and dissociation of substrate-catalyst complexes occur at much faster rates than other reactions () and these reactions are therefore eliminated adiabatically. (We have also confirmed the validity of the approximation numerically.)
By denoting the concentration of free catalyst that does not form a complex as and the total concentration of modified substrate, i.e, summation of the concentration of free substrate and substrate-catalyst complex, as , the model can be described as:(1)(2)where are the dissociation constants between and . For simplicity, the rate is chosen to be independent of , as .
Generally, the binding energy of complexes depends on the modification sites. For simplicity, we assume that binding energy is reduced linearly per single modification [21], but as long as the binding energy distribution is not narrow (this condition is described below), a deceleration of the relaxation is obtained. The change in affinity is natural, as the modification generally changes the function and shape of the protein [22]. With this simplification, the dissociation constant between and increases exponentially as , where is . The input is given as changes in the speed of the modification reactions, expressed as for all (see Table 1 in Text S1). This is a simplified model for reactions with several modification sites, as discussed in the context of, e.g., phosphorylation and methylation of proteins, whereas several extensions will be discussed later.
To analyze the relaxation process from a highly modified state to a minimally modified state, we set the initial condition as and for all . Then, only demodification reactions progress. Under this condition, a modification level () relaxes from to in a single direction. When the concentration of the catalyst is sufficiently high (i.e., in the limit ), the behavior of this model is same as that of the first-order reactions, such that fast exponential relaxation occurs.
At low catalyst concentrations, however, the relaxation process is quite different. In this case, as shown in Fig. 2, the modification level relaxes more slowly; it cannot be fitted by the exponential form (see Fig. S2A) and is better fitted by the form of in a certain range (see Fig. S2B), which is termed as the logarithmic relaxation in time. Moreover, when the concentration of the catalyst is within a particular range, the relaxation process shows a plateau. Indeed, such logarithmic relaxation processes and an emergence of a plateau have been shown to exist in glasses [23]. In (statistical) physics, glasses are known to exhibit a very slow relaxation dynamics with a plateau before reaching the final (equilibrium) state, while their long-term relaxation course is fitted by [23], [24]. The origin of such slow relaxation is attributed to kinetic constraint [24], [25]. In this sense, the present relaxation is also referred to “glassy” relaxation.
As shown in Fig. 2, the relaxation process shows a transition from a fast exponential relaxation phase to a slow logarithmic relaxation phase as the catalyst concentration is decreased. To examine this transition quantitatively, we analyzed the dependence of the relaxation time on the concentration of the catalyst. In Fig. 3, the dependence of upon the total concentration of catalyst is plotted, and a sudden increase of with the decrease of below the concentration of total substrates can be observed. Although nonlinear dependence on the catalyst concentration is expected based on Michaelis-Menten kinetics, this sharp increase in near the critical point results from limitation of the catalyst.
Here, by slightly decreasing the concentration of the catalyst below that of the substrate or by increasing the concentration of the substrate beyond that of the catalyst, the relaxation time suddenly increases by several orders of magnitude. In other words, the relaxation time becomes much longer than the elemental chemical time scale, such that the modification state remains almost at the original level. Hence, the modification state is “memorized” over a long time scale, and storing or erasing memory is achieved by slightly changing the ratio of catalyst to substrate below and beyond the critical value. In summary, our model exhibits a transition from fast exponential relaxation to slow logarithmic relaxation, thereby providing kinetic memory.
In this section, we discuss the conditions for the existence of a sharp transition to memorized states.
The relaxation of the modification level is accompanied by changes in the modifications of substrate sites. Initially, was set larger than other values for , and the relaxation process to the stationary state with was investigated. At high catalyst concentrations, relaxed in descending order (Fig. 4A); decreases in resulted in increases of , whose decrease resulted in an increase of and so forth. This ordered relaxation agrees with that of first-order kinetics, which is expected given that a substrate with modified sites is demodified to a substrate with sites.
At low catalyst concentrations, however, the relaxation process is not ordered in such a monotonic manner. Indeed, highly modified substrates cannot relax readily (Fig. 4B), whereas less-modified substrates are able to take on modified states more easily. This relaxation of reversed order results from the limitation of catalysts and differences in catalyst/substrate affinity according to the number of modified sites , both of which underlie slow, logarithmic relaxation.
Now we theoretically estimate the slow relaxation dynamics. Here, the relaxation dynamics, even after elimination of the variable by assuming fast equilibration of binding and unbinding dynamics, are complicated and nonlinear to be solved analytically.
Note that slow relaxation requires competition for a catalyst, which is present at low abundance, and heterogeneity of affinity between substrates and the catalyst. Because of these two conditions, the time scales of relaxation for each modified substrate are separated, which slows the relaxation of modification dynamics. Here, we roughly estimate the long-term relaxation dynamics asymptotically, by approximating it by superposition of the eigenmode relaxation dynamics, which are approximated by relaxation of each modification level . In our case, the affinities between substrates and a catalyst are distributed exponentially, and thus the time scales of demodifications, accordingly the eigenvalues, are also distributed exponentially. As for the estimate by superposition, we follow the scheme adopted in the theory for slow relaxation dynamics in glass (see e.g., [25]). Then, in the limit of large and , with small , the total relaxation dynamics, given as a summation of such demodification dynamics of abundant substrates, are estimated to be logarithmic.
To estimate the relaxation dynamics, we first focus on the dynamics of substrates consisting of only a single demodification reaction, i.e., . When and is sufficiently small, the abundance of free catalyst is larger than ; therefore, the dynamics of substrate abundance are estimated by(3)
Thus, the relaxation of substrates is exponential and independent of . In contrast, when , almost all catalysts are bound to substrates to form a complex form; therefore, that the abundance of free catalyst molecules approaches zero. Hence, is smaller than . In this case, is estimated as . Here, in the large limit, for is negligible; thus, . When is sufficiently small, ; therefore, is estimated by(4)
Thus, the dynamics of for are given by(5)
Here, the total modification level is given by .
Although the dynamics of the total modification level depend on the initial condition and both the influx and efflux of each , the eigenvalue of each 's relaxation dynamics is mainly governed by the efflux, because a contribution to the eigenvalue of the influx is given as and of the efflux is given as , where . Hence, a contribute of the influx is negligible for large , thus the eigenvalue of each dynamics is governed by . Then, the time evolution of the total modification level is given by , where is the fraction of each eigenmode. Therefore, in the low catalyst concentration regime, the relaxation time depends on the inverse of and is determined by for maximal ; therefore, when the number of modification sites, , increases, the relaxation time increases in proportion to . In the large limit, the summation of can be estimated by integration as . When there is no singular dependence of on (i.e., the initial condition should not have singular dependence that is quite exceptional as in for and for ), by setting , the integral is calculated as(6)
Here, the divergence of and as and is considerably slower than the exponential. When is sufficiently large, such that(7)
Therefore, the dependence is obtained asymptotically for large when . The above estimation suggests that a small is needed for switching between fast and slow relaxation, and a large is needed for slower relaxation that allows separation of the time scale of each substrate, and a large is needed for logarithmic relaxation. Thus, when and are large and is small, the time evolution of protein modifications is expected to asymptotically follow a logarithmic pattern. Our simulation results show that the relaxation is much slower than exponential and does not follow the logarithmic form perfectly, because is small but finite (see Fig. S2B).
The slowing of relaxation caused by competition for the catalyst is also intuitively understood. When the modified molecules are demodified, the amount of modified molecules will increase. Because the modified molecule has stronger affinity for the catalyst than the modified molecule, the binding of the modified molecules to the catalyst hinders the binding of the modified molecules. Thus, the demodification process is slowed depending on the modification level with an increase of the timescale as for the -modified substrate. Therefore, the order of relaxation becomes reversed in the logarithmic regime (see Fig. 4B).
The above slowing mechanism resulting from the summation of distributed exponential relaxation dynamics is studied as the slowing of the equilibration process of glass. In particular, the mechanism we describe here is identical to that previously proposed for a chemical glass in catalytic networks [25], where a negative correlation between the abundance of a substrate and that of its catalyst suppresses the relaxation. In contrast to the abstract catalytic network model, our study adopts a realistic protein model with modification sites; therefore, input signals are easily administrated, storing information, and erasure of memory is achieved with applicability to the present cells.
When kinetic memory is formed via logarithmic relaxation, the relaxation time is not constant; rather it is instead further increased with the magnitude and the duration of stimuli given as for and for , as shown in Fig. 5A. Moreover the maximal modification level also depends on the magnitude and duration of stimuli (Fig. 5B and Fig. S3). Thus, information regarding the input stimulus (i.e., magnitude and duration) is “memorized” as the difference between the relaxation time and the modification level. This continuous memory is in contrast to the on-off type of attractor memory.
Both the relaxation time and the modification level are candidate mediators of memory storage, as cells can access both of these mechanisms depending on the output pathway from the modification level. If cells use threshold dynamics as an output pathway and if the temporal integration of such output product contains information, then the relaxation time will be important for cells to decide their fate. In contrast, if cells use the modification level itself as an output pathway, the modification level will be more important. Depending on the output pathway, either the relaxation time or the modification level provides a candidate mechanism for useful information to be stored.
To further demonstrate the utility of kinetic memory, we also investigated the kinase-phosphatase (K-P) model (Fig. 6A). This model contains three components, i.e., kinase , phosphatase , and substrate , with multiple modification sites [27]–[29]. These modification states are characterized by , which denotes the number of phosphorylated residues . Kinases mediate an increase in the number of phosphorylated residues, whereas phosphatases facilitate inverse reactions. Generally, substrate modifications lead to changes in the affinities of substrates and catalysts. The dissociation constant between and and that between and increase exponentially as and , respectively, identical to the effect observed in the chained modification model.
We studied the relaxation process of after the amount of the total kinase () was varied, which functions as the input stimulus. We analyzed the relaxation process following the input of stimuli with a higher concentration of active kinases for a sufficient length of time. Here, again, the dephosphorylation processes in the K-P model show slow logarithmic relaxations when is positive. As shown in Fig. 6B, when the total amount of phosphatase is lower than that of the substrate, the dephosphorylation process shows very slow relaxation. The transition between fast and slow relaxation occurs at the point where the amount of phosphatase and that of the substrate is balanced.
As another example, we studied an extended version of the Asakura-Honda (A-H) model. The original A-H model was introduced to explain processes of adaptation to changes in the concentration of external signal molecules (attractant and repellant) in chemotactic behavior [30]. This model represents a two-state receptor with multiple modification sites. Receptors in different states are recognized by distinct enzymes that facilitate an increase or decrease in the number of modified sites. In the model, the enzymes are always maintained at sufficient levels so as to not be rate-limiting. This A-H model consists only of first-order reactions, without Michaelis-Menten type reactions. Here, to discuss kinetic memory, we explicitly took the dynamics of the co-factor as a catalyst that catalyzes each modification reaction into account (Fig. 7A).
In the present paper, we introduced three models, i.e., a chained modification model having a single-state substrate and one catalyst, a kinase-phosphatase model having a single-state substrate and two catalysts, and a modified Asakura-Honda model having a two-state substrate and one catalyst. To demonstrate that the kinetic memory functions for all of these cases, we studied the A-H model here.
We analyzed the adaptation process after the input stimulus was applied to change the fraction of two states, and we found that in the extended A-H model, slow logarithmic relaxation occurs when the abundance of catalyst is limited, as observed in the chained modification model (Fig. 7B). The modified state memorizes the input amplitude and duration during the process of adaptation, and the conditions for this kinetic memory are essentially identical to those of the chained modification model (see Supporting Information and Fig. S5–S7). It is noted that a slow process exists only in the relaxation in the adaptation; the response remains fast independently of the parameters and the abundance of catalysts. The time scales for the response and relaxation are separated, and they are independently controlled by tuning the dissociation constants between the substrates and catalyst.
In the present study, we evaluated three models, i.e., the chained modification model, the kinase-phosphatase model and the extended A-H model, which consist of a substrate with multi-modification sites and a catalyst that facilitates modification of the substrate's sites. As shown in Fig. 3 and Fig. 6B and Fig. S6, all of these models reveal a transition from fast exponential relaxation to slow logarithmic relaxation at the point where the concentration of the catalyst falls below that of the substrate.
The conditions for this transition are summarized as follows: There must be heterogeneity in the binding affinity of the catalysts that depends on the number of modified sites, such that the affinity for highly modified substrates should be sufficiently small. This leads to two requirements in our model in which the dissociation constant is set at for the modification sites .
(i) low value: this supports a high affinity for substrates, such that the competition for catalysts among substrates is induced. To satisfy this requirement, should be smaller than the substrate concentration for all ; thus, the upper limit of is restricted as .
(ii) : affinity depends on the number of modified sites on the substrates. When the above conditions are not satisfied, the relaxation time follows that of the Michaelis-Menten equation. In contrast, when is small and , the relaxation changes drastically at the point where the concentrations of the substrate and the catalyst coincide. This change is much sharper than that expected from the Michaelis-Menten equation and is an example of ultrasensitivity [26].
When the concentration of the substrate () is lower than that of the catalyst (), the concentration of a catalyst-substrate complex () becomes approximately identical to that of the substrate (), whereas when , approaches . When the concentration of the catalyst decreases to levels below that of the substrate, various modified forms of the substrate compete for catalyst molecules. As a result, a transition to the slow relaxation occurs, induced by this enzyme-limited competition.
The kinetic memory described here is expected to be advantageous over attractor memory with regards to the housekeeping cost. To maintain attractor memory, continuous invertible reactions are necessary, which consumes housekeeping energy [31]. Kinetic memory mediated by protein modifications also incurs housekeeping costs because of its irreversible modification reactions. However, during slow relaxation in kinetic memory, reversible association and dissociation reactions progress, but irreversible reactions are suppressed. Hence, the housekeeping costs in kinetic memory are expected to be lower than those for attractor memory.
Moreover, the memory erasure mechanism is different from that involved in multistable memory. The memory of a stable attractor is almost always constant except for the short time span needed for a switch to a different attractor. Erasure in kinetic memory, in contrast, is achieved by simply increasing the abundance of the enzyme, and thus could require a lower cost.
Attractor memory, however, may have some advantage with regards to the stability of the memorized state against noise and the constancy of the memorized state.
The kinetic memory we studied here is generated by a substrate with a few modification sites and a catalyst (enzyme) shared by each of the different modification states, resulting in enzyme-limited competition (ELC) [32], [33]. Hence, multimeric proteins, for example, can provide a molecular basis for kinetic memory. A candidate for a multimeric protein bearing kinetic memory is CaMKII, which forms a dodecameric structure with phosphorylation sites in each monomer [7]. It is known that phosphorylation of sites on CaMKII plays a critical role in the maintenance of early LTP. CaMKII is phosphorylated with increases in concentration and is dephosphorylated by protein phosphatase 1 (PP1). After the concentration decreases, phosphorylation levels remain high; therefore, CaMKII stores memory in its phosphorylation state. Indeed, if PP1 is limited, ELC among CaMKII molecules is expected to lead to kinetic memory, according to our argument presented here. In fact, it has been reported that the concentration of PP1 is lower than that of CaMKII in the postsynaptic density, as suggested by the condition for ELC [17]. As already described in the introduction, the phosphorylation state of CaMKII may not have multistability [18].
To confirm our “kinetic memory” hypothesis, it is important to analyze the time evolution of dephosphorylation of CaMKII over a long time course. If the dynamics of CaMKII dephosphorylation are slowed and distinguishable from a simple exponential, our hypothesis may be supported. Moreover, analysis of mutants that mimic phosphorylated and unphosphorylated states may also be effective. By using phosphorylated and unphosphorylated mutants, the difference in binding energy between phosphorylated and unphosphorylated states may be determined biochemically. Such results are helpful to determine the actual of CaMKII and how the relaxation dynamics depend on .
Another candidate for kinetic memory may be CREB in brain synapses, which is known to mediate potentiation through altered phosphorylation levels. In late LTP, CREB is phosphorylated by CaM-dependent kinase and is gradually dephosphorylated by calcineurin, whereas phosphorylation of CREB leads to activation of gene expression [8], [34]. Here, it is reported that with increased duration of input, the relaxation time for dephosphorylation of CREB is prolonged, which is consistent with our kinetic memory [34].
It could also be expected that several other proteins with multi-modification sites may provide kinetic memory in our scheme. For example, the phosphorylation level of ERK, which has multiple phosphorylation sites, is elevated over a long time span after a transient increase in the level of nerve growth factor [9]. Such long-term phosphorylation may be a result of logarithmic relaxation in kinetic memory.
In a multistability (attractor) model, the memorized states are discrete and few in number, as the number of attractors typically increases only linearly with the number of modification sites. In contrast, kinetic memory can store continuous information regarding inputs, as we have discussed above. Indeed, cellular memory refers to the process by which organisms integrate information from continuous external conditions and retain it in their cells. Unicellular protozoa P.caudatum, when placed on a temperature gradient, accumulate in a region maintained at the previous cultivation temperature; this memory of the cultivation temperature is stored over approximately 40 min [3], [35]. Similar temperature memory in nematodes C.elegans over several hours has also been observed and is suggested to be stored at the level of a single thermosensory neuron [36]. It was also reported that C.elegans can memorize the NaCl concentration at the level of a single neuron [37], [38]. The kinetic memory scheme may shed light on such “continuous” cellular memory.
An important condition for kinetic memory is competition for the catalyst. The relevance of ELC to cellular functions has recently been discussed. For example, our previous study suggested the importance of ELC for temperature compensation of the period of biological clocks [32], [33]. Both in that study and in the present study, we found that distributed affinity leads to slow dynamics, and non-linear dependence of the reaction rate on substrate and catalyst concentrations is essential. Further applications of ELC to other cellular functions will be revealed in the near future.
Ultimately, it will be important to experimentally verify the kinetic memory hypothesis that we have proposed and tested here. Steps toward this aim should include measurements of relaxation processes for protein modifications under various catalyst concentrations. In addition, the development and use of mutant proteins that are able to mimic substrates with several modification states would enable the testing of varying affinities between modification sites. Such experiments would not only validate our model but also facilitate future investigations seeking to uncover mechanisms underlying primitive forms of cellular memory.
The whole reactions are described below.
By assuming that association and dissociation reactions are faster than the other reactions, we obtained the following equations:(8)
The total concentrations of the kinase and the phosphatase are conserved quantities, and thus(9)(10)
For parameter values, see Table 2 in Text S1.
The complete set of reactions in the extended Asakura-Honda model is given as follows:
Assuming that catalyst association and dissociation reactions, in addition to flip-flop reactions between S and T, are much faster than modification reactions, they can be eliminated adiabatically; therefore, the model can be described as:(11)
Where and . Here, is the free catalyst that is not bound to or , which satisfies(12)where and are dissociation constants between and or , respectively. We assumed that the affinities between the catalyst and substrate decrease exponentially with the number of modified sites in the substrate, that is, the dissociation constants and increase exponentially. Exponential increases of and are required for perfect adaptation when the amount of the catalyst is sufficiently low (see Supporting information). In addition, we assumed that the dissociation constants are not different between the form and the form; therefore, the dissociation constants are described as . However, this is only for simplicity and is not essential. Stimuli are given as changes in , identical to the original A-H model.
For parameter values, see Table 3 in Text S1.
|
10.1371/journal.ppat.1004012 | DHX36 Enhances RIG-I Signaling by Facilitating PKR-Mediated Antiviral Stress Granule Formation | RIG-I is a DExD/H-box RNA helicase and functions as a critical cytoplasmic sensor for RNA viruses to initiate antiviral interferon (IFN) responses. Here we demonstrate that another DExD/H-box RNA helicase DHX36 is a key molecule for RIG-I signaling by regulating double-stranded RNA (dsRNA)-dependent protein kinase (PKR) activation, which has been shown to be essential for the formation of antiviral stress granule (avSG). We found that DHX36 and PKR form a complex in a dsRNA-dependent manner. By forming this complex, DHX36 facilitates dsRNA binding and phosphorylation of PKR through its ATPase/helicase activity. Using DHX36 KO-inducible MEF cells, we demonstrated that DHX36 deficient cells showed defect in IFN production and higher susceptibility in RNA virus infection, indicating the physiological importance of this complex in host defense. In summary, we identify a novel function of DHX36 as a critical regulator of PKR-dependent avSG to facilitate viral RNA recognition by RIG-I-like receptor (RLR).
| Cellular responses to environmental stress are critical for maintaining homeostasis in living organisms. In one type of response, eukaryotic cells exhibit rapid formation of aggregates with RNA and multiple RNA-binding proteins in the cytoplasm termed stress granules (SGs). Over the past decade, SGs have been suggested to be important compartments and play essential roles in cellular stress responses. We have previously reported that virus infection induced SG-like aggregates and are crucial for antiviral response, therefore termed them as antiviral (av) SGs. In this report, we describe a novel function of DExD/H box RNA helicase 36 (DHX36), as a critical activator of double-stranded RNA (dsRNA)-dependent protein kinase (PKR), which directly trigger avSG formation in virus-infected cells. Our results reveal a novel link between DHX36 and avSG which functions as a platform to facilitate sensing of viral invasion and triggering antiviral responses.
| Cellular responses against various stresses are crucial for maintaining homeostasis. Virus infection is one class of cellular stress that induces a number of host responses through the activation of pattern recognition receptors (PRRs), which detect pathogen-associated molecular patterns (PAMPs) with high specificity for protection against microbial invasion. RIG-I, a DExD/H-box RNA helicase family member, functions as a cytosolic viral RNA sensor, recognizes specific structures of non-self RNAs, including double-stranded RNA (dsRNA) and in vitro transcripts containing 5′ppp- and partial double-stranded structure due to copy back (5′ppp-cbRNA) [1], [2]. Upon sensing non-self RNA, RIG-I promptly induces type I interferon (IFN) and other cytokines to provide a primary defense program against viruses [3].
In addition to RIG-I, interferon-inducible dsRNA-dependent protein kinase (PKR) represents another cytosolic foreign RNA sensor and plays a pivotal role in regulating innate immune responses [4], [5], [6], [7], [8], [9]. Both RIG-I and PKR exert an antiviral effect by detecting particular RNA species such as dsRNA or secondary-structured RNA, which can be generated during replication of both RNA and DNA viruses [2], [10]. Although accumulating evidence has suggested that PKR has an important role in antiviral host defense, the exact mechanisms underlying the regulation of innate immunity remain to be elucidated.
Recently, we and others have shown that PKR is a key factor for the induction of cytoplasmic bodies called antiviral stress granules (avSGs) by viral infection, and we further clarified that SGs provide a critical platform for interactions between antiviral proteins and non-self RNA ligands [11], [12], [13], [14]. Increasing evidence has clearly shown a tight link between SGs and antiviral innate immunity and, in fact, most viruses appear to antagonize SG formation through multiple strategies for the evasion of host antiviral responses [15], [16].
Previous studies have suggested the involvement of various DExD/H-box RNA helicases in innate immunity by either direct sensing of various PAMPs or functioning as an adaptor molecule in the signaling pathway [17]. DExD/H-box helicase 36 (DHX36), also termed RNA helicase associated with AU-rich RNA element (ARE) (RHAU) is implicated as a factor in ARE-mediated mRNA decay [18]. DHX36 can directly bind and unwind the specific structure of guanine-tetramolecular quadruplex-DNA or -RNA (G4-DNA or G4-RNA) through its amino-terminal RHAU-specific motif together with helicase activity [19], [20]. Recently, DHX36 has also been characterized as a critical factor for host immune responses. For instance, DHX36 was shown to function as a foreign DNA sensor in plasmacytoid dendritic cells (pDC) [21]. More recently DHX36 has been suggested to play a pivotal role in the IFN signaling pathway by forming a complex together with other RNA helicases for dsRNA sensing [22]. Furthermore, it has also been demonstrated that DHX36 localizes in stress granules under various cellular stress conditions [23]. These results prompted us to investigate whether DHX36 has the potential to regulate innate immunity by interacting with antiviral proteins in virus-induced stress granules.
In this report, we examined the role of DHX36 in the host innate immune response by virus infection and non-self RNA ligand transfection in fibroblast cells. It was found that DHX36 regulates innate immunity by facilitating virus-induced stress granule formation and contributes to PKR activation by viral RNA.
To further evaluate the function of DHX36 in the RLR-mediated pathway, we employed the conditional dhx36 KO system [24]. First, the efficiency of tamoxifen-induced deletion of the dhx36 gene in MEFs was examined (Figure 1A). After 72 hours of tamoxifen treatment, DHX36 protein became undetectable. Next, the effect of dhx36 KO on IFN-ß gene activation was examined. IFN-ß induction by infection with influenza A virus (IAV) ΔNS1, Newcastle disease virus (NDV) or by transfection of poly rI: poly rC (pIC) was significantly reduced in the absence of DHX36 as revealed by quantification of IFN-ß protein (Figure 1B–D) and mRNA (Figure 1E and Figure 1F). Moreover, DHX36 is required for efficient IRF-3 dimerization in NDV-infected cells (Figure 1G). In summary, these results reveal the involvement of DHX36 in virus-induced signaling for IFN-ß gene activation.
It was reported that 5′ppp-cbRNA activates RIG-I [25], [26]. Interestingly, IFN-ß production, as well as IRF3 dimerization, induced by transfection of 5′ppp-cbRNA did not change in the presence or absence of DHX36 (Figure 1G and Figure 1H). Whereas IFN-ß mRNA level was increased in the absence of DHX36 (Figure 1I). Because these IFN-ß mRNAs are likely polyadenylated (Figure S1), a possible involvement of DHX36 in posttranscriptional regulation is suggested [18]. However we focus on the function of DHX36 on the virus-induced signaling in this report.
In NDV-infected HEK293T cells, knockdown of DHX36 profoundly diminished active IRF-3 dimers (Figure 1J). Since NDV replication was comparable up to 12 hr (Figure 1E and Figure 1J), in the presence or absence of DHX36, the data suggest the involvement of DHX36 in antiviral signaling through IRF-3. Altogether, these results indicate that DHX36 is involved in IFN-ß production in a stimulus-dependent manner.
We recently reported that antiviral stress granules (avSGs) are crucial for both sensing virus infection and type I IFN signaling by providing a critical platform for interaction between antiviral proteins and non-self RNA ligands [11], [13]. DHX36 was reported to localize in SGs induced by various stresses [23]. We used HeLa cells for DHX36 localization by immunostaining because this cell line was used in our previous studies. We confirmed that DHX36 upregulates virus- or pIC-induced signaling in HeLa cells (Figure S2). Infection of HeLa cells with IAVΔNS1 induced speckles of SG marker T-cell intracellular antigen-1-related protein (TIAR) and these speckles also co-localized with RIG-I and IAV nucleocapsid protein (NP) (Figure 2A). Quantification also confirmed that viral infection specifically induced co-localization of these proteins (bar graph in Figure 2A). DHX36 is normally distributed diffusely in both the nucleus and cytoplasm; however, IAVΔNS1 infection induced its re-localization to speckles, which co-localized with RIG-I and TIAR (Figure 2B). Co-localization of DHX36/TIAR and DHX36/RIG-I was highly inducible and was observed in the majority of speckles (∼90%, bar graph in Figure 2B). We also examined co-localization of PKR with RIG-I in avSGs induced by infection with IAVΔNS1 (Figure 2C). These results confirmed that DHX36, RIG-I, PKR and TIAR localizes in avSGs.
Because DHX36 co-localized with RIG-I and PKR, we examined their physical interaction by co-immunoprecipitation. Cells were transfected with expression vector for full-length RIG-I or RIG-IΔCTD, which lacks the C-terminal domain (CTD). The cells were mock treated or infected with IAVΔNS1 for 12 h and physical interaction was monitored by co-immunoprecipitation. DHX36 interacted with RIG-I regardless of virus infection in its CTD-dependent manner (Figure 2D, Figure S3). However, interaction between PKR and RIG was dependent on virus infection. Moreover, RIG-I CTD was partly responsible for this interaction. To further confirm the interaction of both DHX36 and antiviral proteins with dsRNA, we performed pIC pull-down analysis using an extract from uninfected cells (Figure 2E). The result clearly showed that DHX36 has binding affinity to dsRNA in addition to AU-rich RNA, as reported previously [18]. Purified recombinant DHX36 exhibited clear binding with pIC (Figure S4), demonstrating direct interaction between DHX36 and dsRNA. RIG-I and PKR also bound to the dsRNA, presumably as a complex (Figure 2E). However, TIAR, a SG component and a known RNA binding protein, did not bind to the dsRNA.
To further analyze whether DHX36 associates with viral RNA, we performed RNP-IP analysis using anti-DHX36 monoclonal antibody to pull down endogenous DHX36-RNA complex. Direct interaction between DHX36 and RIG-I was confirmed again by this experiment as shown in Figure 3A. We then isolated RNA from RNP complex and performed real-time qPCR to evaluate binding of DHX36 and IAV viral RNAs. We found that DHX36 associates with viral RNAs (Figure 3B). Among the viral RNAs examined, DHX36 showed the highest binding affinity to IAV segment 8 (Figure 3C). Next, we checked the localization of IAV RNA. Since IAV produces viral RNAs containing partial double-stranded panhandle-structure [1], we used a propidium iodide (PI), which specifically binds to both dsDNA and dsRNA, for staining of cytoplasmic dsRNA. IAVΔNS1 infected cells showed the cytoplasmic granules stained by PI and these foci coincided with RIG-I and DHX36 (Figure 3D). Therefore, these results suggest that DHX36 directly recognizes viral RNAs together with antiviral proteins in the avSG.
To further prove the functional relevance of avSG to host antiviral responses, we investigated the localization of TRIM25, which was shown to be critical for RLR signaling [27]. We used HeLa cells stably expressing GFP-tagged RasGAP SH3-domain-binding protein 1 (G3BP1), a SG marker (HeLa/G-G3BP) [11] for this purpose. TRIM25 is normally distributed diffusely in the cytoplasm. Interestingly, virus-infected or pIC-transfected cells exhibited translocation of TRIM25 to avSG together with DHX36 and G3BP1 (Figure 4A). In contrast, TRIM25 was not localized in SG induced by arsenite treatment (Figure 4B), which does not induce IFN signaling (Figure 4C), strongly suggesting that interaction between TRIM25 and antiviral proteins in the avSG is critical for the efficient antiviral signaling.
To explore the underlying mechanism of DHX36 requirement in NDV-induced IFN-ß gene activation, we knocked down DHX36 in HeLa/G-G3BP cells and examined them for avSG formation by NDV infection. NDV-induced avSG was markedly diminished by DHX36 knockdown (Figure 5A and Figure 5B); however, DHX36 deletion did not affect viral RNA production up to 12 h (Figure 1J). We previously reported that RIG-I associates with G3BP in an IAVΔNS1 infection-dependent manner [13]. Therefore, we examined the association of RIG-I and G3BP in NDV-infected cells and confirmed that these proteins physically associated in NDV-infected cells. Interestingly, this interaction was significantly attenuated by DHX36 knockdown (Figure 5C). Consistently, we also observed that IAVΔNS1-induced avSGs and antiviral signaling was inhibited by DHX36 depletion (Figure S5), confirming a critical role for DHX36 in the regulation of IFN-ß gene activation through avSGs.
It has been hypothesized that different stresses activate one of the four distinct eIF2α kinases, including PKR, and phosphorylation of eIF2α at Ser 51 triggers the assembly of SG [28]. Upon binding of dsRNA with PKR, autophosphorylation of PKR takes place and the resultant hyperphosphorylated PKR catalyzes the phosphorylation of heterologous substrates [29], [30]. Because our data suggest that DHX36 is required for avSG formation, we further examined the effect of DHX36 deletion on PKR phosphorylation induced by viral infection. Intriguingly, virus infection induced hyperphosphorylation of PKR and this phosphorylation was markedly attenuated by DHX36 deletion (Figure 6A). pIC transfection similarly induced the phosphorylation of PKR in mouse (Figure 6B) and human cells (Figure 6C). However, depletion of DHX36 diminished the PKR phosphorylation. These results clearly suggest that DHX36 is required for efficient activation of PKR induced by dsRNA.
To examine the role of DHX36 in the activation of PKR, we transiently overexpressed HA-tagged DHX36 in HEK293T cells and analyzed the phosphorylation of PKR in control and NDV-infected cells. While overexpression of DHX36 alone did not induce PKR phosphorylation, NDV-induced PKR phosphorylation was augmented by ectopic expression of DHX36 (Figure 7A, lane 2 and 5 and Figure 7B). Because DHX36 is a putative RNA ATPase/helicase, we next asked whether these catalytic activities are involved in PKR activation by overexpression of an ATPase-defective DHX36 mutant, E335A [18]. Interestingly, ectopic expression of DHX36 E335A inhibited NDV-induced PKR phosphorylation, suggesting that the ATPase/Helicase activity is involved in PKR activation (Figure 7A, lane 4–6 and Figure 7B). We further asked if the physical interaction between PKR and dsRNA is affected by the presence of DHX36. To evaluate this, we pulled down RNA-protein complex from normal and DHX36-knocked down cell lysates by pIC-agarose (Figure 7C). Although the cells were not infected, PKR is activated during incubation with pIC-agarose, presumably by ATP present in the extract. In the absence of DHX36, the association of PKR, but not RIG-I, with pIC was moderately reduced with a concomitant reduction of phospho-PKR (Figure 7C and Figure 7D). Since total PKR amount was not affected by the absence of DHX36, these data suggest that DHX36 is capable of facilitating PKR association with dsRNA, hence its activation.
Although our data clearly showed that DHX36 augments PKR phosphorylation in a RNA ligand-dependent manner, it is still possible that other cellular factors may also participate and contribute to PKR activation. In fact, DHX36 forms a complex with other RNA helicases for innate immune responses induced by dsRNA [22]. To clarify this issue, we further asked whether DHX36 directly facilitates PKR phosphorylation or requires other cofactors for PKR activation. To examine this, we performed an in vitro PKR phosphorylation assay using purified recombinant PKR, DHX36 WT and DHX36 E335A (Figure 8A). We confirmed that purified recombinant DHX36 WT, but not E335A, retained the ATPase activity (Figure 8B). We also confirmed that the recombinant PKR was functional (Figure 8C). Based on the data from Figure 8C, we chose 1 ng pIC for further analysis. As shown in Figure 8D, PKR phosphorylation was detected by pIC treatment (Figure 8D, lane 1 and lane 2). Interestingly, this PKR phosphorylation was markedly increased in the presence of DHX36 in a dose-dependent manner (Figure 8D, lane 3–5). In contrast, DHX36 E335A lost its function on facilitating the PKR phosphorylation (Figure 8D, lane 6–8) confirming the involvement of ATPase activity in vitro. Additionally, BSA was also tested as a non-related control for this analysis and the result did not show any significant effect (Figure 8D, lane 9–11, Figure S6). Taken together, these data indicate that DHX36 solely functions for efficient PKR activation in the presence of dsRNA through its ATPase activity.
Type I IFN plays a central role in defense against virus infection by suppressing viral yield and promoting cell survival. First, we examined the impact of deleting DHX36 on the viral cytopathic effect. WT or DHX36 KO MEFs were infected with NDV. At early time points (up to 12 h), viral RNA replication was not impaired by DHX36 knockdown (Figure 1J). However, at 24 h post infection, augmented cell death was observed in DHX36 KO cells compared to WT cells (Figure 9A and Figure 9B). Consistent with the enhanced cytopathic effect, DHX36 KO cells produced enhanced viral RNA (Figure 9C) and viral titer (Figure 9D), suggesting that DHX36 exerts its antiviral effect through the regulation of IFN production.
The cytoplasmic viral RNA sensor, RLR, plays a major role in detecting viral infection and triggering antiviral responses. In this report, we describe the involvement of DHX36 in sensing viral infection and subsequent activation of RIG-I. Kim et al. reported that DHX36 functions as DNA sensor in plasmacytoid dendritic cells (pDC) [21]. The reported function of DHX36 is apparently distinct from our discovery. First, we used fibroblast and epithelial cell lines, quite distinct cell types from pDC. Second, the stimuli used did not contain or produce DNA. There is another report describing DHX36 as a dsRNA sensor in myeloid dendritic cells (mDC) [22]. DHX36 appears to function upstream of TIR domain-containing adaptor inducing IFN-ß (TRIF also termed TICAM1) to activate NF-κB and IRF-3/7. Again, their observation is distinct from ours because in our system, RIG-I was essentially involved in the signal induced by dsRNA, IAVΔNS1, and NDV in the fibroblast cells [31]. It remains to be shown whether dsRNA-stimulated mDC promotes PKR activation and SG formation through DHX36.
We observed that DHX36 augments signaling by RIG-I in a stimulus-dependent manner: signaling induced by infection with IAVΔNS1, NDV and transfection of pIC is enhanced by DHX36 (Figure 1). These stimuli commonly induce cells to form SG. We and others have shown that SG formation is required to facilitate antiviral responses of host cells [11], [12], [13], [14]. Interestingly, the majority of viruses reported, including Sindbis, encephalomyocarditis, polio, adeno-, and measles viruses induce SG upon infection [11], [12], [13], [32].
We discovered that DHX36 facilitates the activation of PKR in the presence of dsRNA, resulting in subsequent SG formation (Figure 5–8 and also see a hypothetical model in Figure 10). DHX36 was found to physically interact with RIG-I regardless of virus infection. Under the condition of viral infection, DHX36:RIG-I complex recognizes dsRNA and interacts with PKR. Although small increase of binding of PKR to dsRNA was observed by DHX36, this does not explain the marked activation of PKR by DHX36. Interestingly, ATPase-deficient DHX36 inhibited PKR activation, suggesting that ATP hydrolysis and/or dsRNA unwinding are mechanistically involved in this regulation. In fact, DHX36 is capable of resolving several types of nucleotides containing unusual structures, such as G4-DNA and G4-RNA, and promotes further biological events [33], [34]. It is tempting to speculate that DHX36 changes conformation upon binding with dsRNA in the presence of ATP in a similar manner as another DExD/H box RNA helicase, RIG-I [35], to accelerate PKR activation with dsRNA (Figure 2D and Figure 7C). Finally, PKR activation subsequently induces avSG assembly and provides a platform of antiviral signal transduction by recruitment of TRIM25, a critical signaling molecule.
Generally, replication of RNA viruses does not take place in soluble compartments. Viruses hijack host cell structures or create a de novo compartment to replicate, resulting in evasion from host sensing of foreign nucleic acids. Induction of avSGs may be one of the means to force viral nucleic acids to be exposed to host immune sensors and facilitate antiviral responses. This idea is consistent with the observations that avSGs facilitate immune sensing but is not an absolute requirement.
In contrast to the viral infections examined, DHX36 did not affect IFN-ß induction by 5′ppp-cbRNA (Figure 1), even though this stimulus activates RIG-I [3]. This is partly because 5′ppp-cbRNA inadequately activates PKR and subsequent avSGs (data not shown). It remains to be explored if conferring PKR activation function to 5′ppp-cbRNA results in enhanced signaling. Alternatively, 5′ppp-cbRNA may bypass the requirement of avSG formation through an unknown mechanism. In this regard, it is tempting to hypothesize the existence of an adaptor molecule(s) that facilitates the activation of RIG-I with short 5′ppp-dsRNA to signal without avSGs. This may be relevant to the previous observation that the 5′ppp moiety is essential for RIG-I to detect short dsRNA (<50 bp) but is dispensable for long dsRNA sensing (100–500 bp) [36], [37].
In this report, we describe a novel function of DHX36 in virus-induced SG formation through PKR activation. We have been proposing a model in which SG functions as a platform to facilitate viral dsRNA recognition by RLR as well as to execute antiviral reactions [13]. The results described here further support our model and emphasize the critical involvement of the virus-induced stress response in antiviral innate immune responses. IAV and picornaviruses attenuate antiviral signaling by the activity of viral non-structural protein NS1 and 3C protease, respectively [11], [13].
Our current study was mainly analyzed by using DHX36 WT and KO MEF to confirm its phenotype of IFN signaling and antiviral activity. Additionally, HEK293 cells were used for the experiments of transient over-expression (Figure 2D and Figure 7A) and measuring the IFN signaling by knocking down of DHX36 (Figure 1J, Figure 6C and Figure S5 D) to support our data from KO system. HeLa cells are useful for analysis of cellular imaging and we have an excellent cellular imaging system with GFP-fused G3BP1 stably expressing HeLa cells. Thus, HeLa and GFP-G3BP1 HeLa cells were mainly utilized for monitoring of localization of antiviral proteins and avSG (Figure 2A–C, Figure 3, Figure 4 and Figure 5).
HeLa and HEK293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (100 U/ml and 100 µg/ml, respectively). DHX36 WT or KO-inducible MEFs (GFP or CRE-GFP), and DHX36 knockdown-inducible HeLa cells were kindly provided by Dr. Nagamine (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland) [24] [38]. Briefly, DHX36 KO was induced by treatment with 1 µM tamoxifen (Sigma-Aldrich, St Louis, MO) in the culture medium for 72 h and the DHX36 KO was confirmed by Western blot analysis. For shRNA-derived DHX36 knockdown, HeLa cells were treated with doxycycline (SIGMA) at a final concentration of 1 µg/ml for 72 h. HeLa cell stably expressing EGFP-G3BP1 (HeLa/G-G3BP) was previously reported by our group [11]. The pEF-Bos-HA-DHX36 was constructed by inserting the PCR-amplified full-length DHX36 coding sequences from pEGFP-DHX36 provided by Dr. Nagamine [23] between BglII and BamHI sites of pEF-Bos vector. The pEF-Bos-HA-DHX36 E335A was constructed with a KOD-Plus-Mutagenesis kit (Toyobo, Osaka, Japan) using the pEF-Bos-HA-DHX36 as a template. pBos-FLAG-RIG-I WT (RIG-I 1-925) and pBos-FLAG-RIG-IΔCTD (RIG-I 1-801) were described previously [39]. All constructs were transfected into HEK293T cells with Lipofectamine 2000 (Invitrogen, Carlsbad, CA) for the experiments.
Poly I∶C was purchased from Amersham Biosciences (Arlington Heights, IL). 5′ppp-RNA (5′- pppGGAAACUGAAAGGGAGAAGUGAAAGUG -3′) [35] was synthesized by in vitro transcription using the AmpliScribe T7-Flash Transcription kit (Epicentre, Madison, WI). RNAs were delivered into the cells with Lipofectamine 2000 according to the manufacturer's instructions. Influenza A virus ΔNS1 strain (A/PR/8/34, ΔNS1) was originally produced by Dr. A. Garcia-Sastre (Mount Sinai School of Medicine, USA), and provided by Dr. S. Akira (Osaka University, Japan). Newcastle disease virus (Miyadera strain) was provided by Dr. Taniguchi (University of Tokyo, Japan). For virus infection, cells were washed with PBS and treated with the culture medium (‘mock-treated’) or infected with IAVΔNS1 or NDV in serum-free and antibiotic-free medium. After adsorption at 37°C for 1 h, the medium was changed and infection was continued for various periods in the presence of serum-containing DMEM.
Culture supernatants were collected and subjected to ELISA with mouse IFN-ß kit (PBL Interferon Source, Piscataway, NJ) according to the manufacturer's instructions.
Poly I∶C pull down assay was performed as described previously [40]. Stable transformants of FLAG-RIG-I in rig-i null MEFs (RIG-I null/FLAG-RIG-I) were established by transfection of a linearized plasmid (pBos-FLAG-RIG-I WT) [39], and selected with Puromycin. Immunoprecipitation was performed using whole-cell extracts from HeLa, HEK293T, HeLa/G-G3BP, or RIG-I null/FLAG-RIG-I cells (200 µg), together with 1 µg anti-FLAG (Sigma-Aldrich, St. Luis, MO, F3165), or anti-GFP (Wako, Osaka, Japan, mFX73) antibodies. After overnight incubation at 4°C, immune complexes were precipitated with protein A-Sepharose beads (Amersham Biosciences) and analyzed by SDS-PAGE and Western blotting. Anti-human and -mouse IRF-3 polyclonal antibody, and anti-RIG-I polyclonal antibody were previously described [13], [41]. Anti-human DHX36 monoclonal antibody was kindly provided by Dr. Nagamine [38]. The monoclonal antibody against Influenza NP (mAb61A5) was generated by Dr. Y. Kikuch (Iwaki Meisei University, Japan), and provided by Dr. F. Momose (Kitasato University, Japan) [42]. The anti-NDV NP monoclonal antibody was produced by Dr. Y. Nagai, and provided by Dr. T. Sakaguchi (Hiroshima University, Japan). Other antibodies used in this study were as follows: anti-phospho PKR T451 (Abcam, ab4818), anti-G3BP (Santa Cruz, sc-70283), anti-TIAR (Santa Cruz, sc-1749), anti-Actin (Sigma-Aldrich, A-1978), and anti-TRIM25 (Santa Cruz, sc-22832) antibodies. For immunofluorescence analysis, Alexa 488-, 594-, and 633- conjugated anti-mouse, anti-rabbit, or anti-goat IgG antibodies purchased from Invitrogen were used as secondary antibodies. Propidium Iodide (PI) (1∶2,000 in PBST) (Miltenyi Biotec) was used for cytoplasmic dsRNA staining.
RNA binding assay was previously reported [35]. Briefly, recombinant DHX36 (1.5 µg) was mixed with 1 µg of pIC in total 10 µl of DHX36 buffer [50 mM Tris pH 7.5, 50 mM KCl, 5 mM DTT, 20% (v/v) glycerol] and incubated at 25°C for 15 min. Then, the mixture was applied to 1% agarose gel and stained with ethidium bromide (EtBr).
RIP assay was performed using cell extracts from mock- or IAVΔNS1- infected HeLa cells with anti-human DHX36 monoclonal antibody by RiboCluster Profiler RIP-Assay Kit (MBL, Japan, RN1001) according to the manufacturer's recommendations. Briefly, RNA-protein complex was pulled-down with 5 µg of mouse normal IgG (Santa Cruz, sc-2025) or anti-DHX36 monoclonal antibody. Then, bound RNAs were recovered from the RNA-protein complex and used for cDNA synthesis. Real-time qPCR was further performed to evaluate the RNA level bound to DHX36 with the specific primer sets targeting IAV gene. As an internal control, human glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene was targeted. Sequence information of primers is as follows: IAV segment 5, 5′-GATGGAGACTGATGGAGAACGCCAG-3′ (Sense), 5′-AGCTGTTTTGGATCAACCGTCCCTC-3′ (antisense); IAV segment 6, 5′-GGACAGAGACTGATAGTAAG-3′ (sense), 5′-GTTAGCTCAGGATGTTGAAC-3′ (antisense); IAV segment 8, 5′-GATAACACAGTTCGAGTCTC-3′ (sense), 5′- TTTCTCGTTTCTGTTTTGGA-3′ (antisense); human GAPDH, 5′-ACTGCCAACGTGTCAGTGGT-3′ (sense), 5′-TTACTCCTTGGAGGCCATGT-3′ (antisense).
Quantitative reverse transcription-PCR for IFN-ß was performed as described previously [13]. For the evaluation of viral RNA, quantitative reverse-transcription PCR was performed using SYBR green reagent (Applied Biosystems, CA, USA, 4385612) with the specific primer sets targeting NDV (F and N) or IAV (segment 5, 6 and 8) genes. As an internal control, human and mouse GAPDH gene was targeted and amplified. Sequence information of primers is as follows: NDV F, 5′-GCAGCTGCAGGGATTGTGGT-3′ (sense), 5′-TCTTTGAGCAGGAGGATGTTG-3′ (antisense); NDV N, 5′-CTCAAGAGAGGCCGCAATAC-3′ (sense), 5′-AGTGCAAGGGCTGATGTCTT-3′ (antisense); mouse GAPDH, 5′-ATCTTCTTGTGCAGTGCCAGCCTCGTCCCG-3′ (sense), 5′-AGTTGAGGTCAATGAAGGGGTCGTTGATGG-3′ (antisense). Sequence information of human GAPDH is described above.
Virus-infected HeLa or HeLa/G-G3BP cells were fixed with 4% paraformaldehyde (PFA) for 20 min at 4°C, permeabilized with 0.05% Triton X-100 in PBS for 5 min at room temperature (RT), blocked with 5 mg/ml BSA in PBST (0.04% Tween20 in PBS) for 30 min, and incubated at 4°C overnight with the relevant primary antibodies diluted in blocking buffer. The cells were then incubated with secondary antibodies at room temperature for 1 h. Nuclei were stained with 4.6-dimaidino-2-phenylinodole (DAPI) and analyzed with a confocal laser microscope, TCS-SP (Leica).
For fluorescence live imaging, HeLa/G-G3BP cells were stimulated by either NDV infection or RNA ligand transfection as described above. After 12 h stimulation, GFP fluorescence images were taken and analyzed with a fluorescence microscope system, AF6500 (Leica). The percentages of avSG-containing cells were calculated in more than five randomly chosen fields for each slide.
The siRNA negative control (Invitrogen, Cat. No. 1293-112) and siRNAs targeting human PKR (sense: 5′-UUUACUUCACGCUCCGCCUUCUCGU-3′, antisense: 5′-ACGAGAAGGCGGAGCGUGAAGUAAA-3′) and DHX36 (sense: 5′-UUCUACUGCUUACAAAUCCAGCUCC-3′, antisense: 5′-GGAGCUGGAUUUGUAAGCAGUAGAA-3′) were purchased from Invitrogen, and the siRNAs targeting human RIG-I (sense: 5′-CGGAUUAGCGACAAAUUUAUU-3′, antisense: 5′-UAAAUUUGUCGCUAAUCCGUU-3′) and mouse DHX36 (sense: 5′-CUACAACUGGCUUAUCUAUUU-3′, antisense: 5′-AUAGAUAAGCCAGUUGUAGUU-3′) were purchased from Genolution Pharmaceutical (Seoul, Korea). For knock down of target genes, siRNAs were transfected into the cells with RNAi MAX (Invitrogen) according to the manufacturer's recommendations. At 48 h or 72 h post-transfection, cells were harvested or infected with viruses for further experiments.
pEGST-PKR/λPP plasmid that encodes GST-fused PKR and λ protein phosphatase was kindly gifted from Dr. Takayasu Date (Kanazawa Medical University). The vector was transformed into E. Coli BL21(DE3)pLysS strain. The Bacteria was first grown at 37°C in LB medium containing 100 µg/ml ampicillin. Protein expression was induced by addition of 1 mM IPTG when the absorbance at 600 nm was approximately 0.4. The cells were then grown at 25°C for 16 h. After incubation, the cells were harvested by centrifugation and resuspended in PBS supplemented with protease inhibitor cocktail (EDTA free) (nacalai tesuque) and was lysed via sonication. The supernatant was collected by centrifugation and mixed with Glutathione Sepharose 4B (GE Healthcare) for 3 hr at 4°C. The protein bound to Glutathione Sepharose 4B was washed with PBS and further washed with PBS supplemented with 500 mM NaCl to remove the E. Coli derived nucleic acid from GST-PKR. Then the protein was eluted by elution buffer containing 50 mM Tris pH 7.5, 50 mM KCl, 5 mM DTT, 20% (v/v) glycerol, 10 mM Glutathione and was concentrated. The purity of PKR was estimated by the Gelanalyzer program (http://www.gelanalyzer.com/) and approximate purity was 80%. No contamination of E. Coli derived nucleic acid was confirmed by UV spectrometer.
To obtain the purified recombinant DHX36 WT and its ATPase-dead mutant, E335A, the intact human DHX36 and E335A were amplified by PCR and inserted into a pEt22b(+) (Novagen) to produce a C-terminally hexa-histidine tagged protein. The vector was transformed into E. Coli BL21(DE3) strain. The Bacteria was similarly grown at 37°C in LB medium containing 100 µg/ml ampicillin. The protein expression was induced by the addition of 0.01 mM IPTG when the absorbance at 600 nm was approximately 0.4. Then the cells were grown at 16°C for 16 h. The cells were harvested by centrifugation and suspended in lysis buffer containing 50 mM Tris pH 8 500 mM NaCl, 20 mM Imidazole supplemented with protease inhibitor cocktail and were lysed via sonication and centrifuged. The supernatant was suspended with a Ni-NTA (Qiagen) affinity column, then the resin was washed with lysis buffer. The protein was eluted with a gradient of 20–500 mM Imidazole dissolved in lysis buffer. The buffer containing the protein was exchanged to 50 mM Tris pH 7.5, 50 mM KCl, 5 mM DTT, 20% (v/v) glycerol and concentrated. The purity of proteins was further confirmed and approximate purity was 80%. No contamination of E. Coli derived nucleic acid was confirmed by UV spectrometer.
In vitro PKR phosphorylation was determined in a total volume of 10 µl containing 50 mM Tris–HCl (pH 7.5), 2 mM MgCl2, 50 mM KCl, 1 µg of purified GST-PKR [43], and the indicated amounts of WT or E335A hexa-histidine-DHX36 and pIC. After incubation at 30°C for 5 min, 1 µl of 10 mM ATP was supplemented and the reaction mixtures were further incubated at 30°C for 15 min. The reaction was stopped by adding 10 µl of 2× SDS sample buffer and boiled for 5 min. Samples were subjected to SDS-PAGE and the level of phosphorylated PKR was evaluated by Western blot analysis using anti-phospho PKR specific antibody. Total amount of input protein level was also confirmed by silver staining.
0.35 mg of recombinant DHX36 WT or E335A was mock-treated or mixed with 0.25 µg pIC in a total volume of 20 µl buffer containing 20 mM Tris (pH 8.0), 1.5 mM MgCl2, and 1.5 mM DTT and the mixture was incubated at room temperature for 15 min. After incubation, 5 µl of 5 mM ATP was added and the mixture was further incubated at 37°C for 30 min. Finally, 5 µl of the mixture from the incubated samples was taken and diluted with 45 µl water, and 100 µl BIOMOL Green (Enzo life sciences, Farmingdale, NY) solution was added to the mixture. ATPase activity was examined by measuring the absorbance of 630 nm of the samples using a Microplate Reader 680 (Bio Rad).
DHX36 WT and KO MEF were infected with NDV for 24 h and culture supernatant was collected. Then, virus yield in culture supernatant was determined using Hep2 cells as previously described [40].
1. Human DHX36: NM_001114397
2. Mouse DHX36: NM_028136
3. Human RIG-I: NM_014314
4. Human PKR: NM_001135651
5. Mouse PKR: NM_011163
6. Human TIAR: NM_001033925
7. Human G3BP: NM_005754
8. Human IRF3: NM_001197122
9. Human Actin: NM_001100
10. Mouse Actin: NM_007392 NM_183274
11. Influenza A virus Segment 5 (NP): NC_002019.1
12. Influenza A virus Segment 6 (NA): NC_002018.1
13. Influenza A virus Segment 8 (NS): NC_002020.1
14. Newcastle disease virus N: NP_071466
15. Newcastle disease virus F: NP_071469
|
10.1371/journal.pbio.1002093 | A Novel Bipartite Centrosome Coordinates the Apicomplexan Cell Cycle | Apicomplexan parasites can change fundamental features of cell division during their life cycles, suspending cytokinesis when needed and changing proliferative scale in different hosts and tissues. The structural and molecular basis for this remarkable cell cycle flexibility is not fully understood, although the centrosome serves a key role in determining when and how much replication will occur. Here we describe the discovery of multiple replicating core complexes with distinct protein composition and function in the centrosome of Toxoplasma gondii. An outer core complex distal from the nucleus contains the TgCentrin1/TgSfi1 protein pair, along with the cartwheel protein TgSas-6 and a novel Aurora-related kinase, while an inner core closely aligned with the unique spindle pole (centrocone) holds distant orthologs of the CEP250/C-Nap protein family. This outer/inner spatial relationship of centrosome cores is maintained throughout the cell cycle. When in metaphase, the duplicated cores align to opposite sides of the kinetochores in a linear array. As parasites transition into S phase, the cores sequentially duplicate, outer core first and inner core second, ensuring that each daughter parasite inherits one copy of each type of centrosome core. A key serine/threonine kinase distantly related to the MAPK family is localized to the centrosome, where it restricts core duplication to once per cycle and ensures the proper formation of new daughter parasites. Genetic analysis of the outer core in a temperature-sensitive mutant demonstrated this core functions primarily in cytokinesis. An inhibition of ts-TgSfi1 function at high temperature caused the loss of outer cores and a severe block to budding, while at the same time the inner core amplified along with the unique spindle pole, indicating the inner core and spindle pole are independent and co-regulated. The discovery of a novel bipartite organization in the parasite centrosome that segregates the functions of karyokinesis and cytokinesis provides an explanation for how cell cycle flexibility is achieved in apicomplexan life cycles.
| Apicomplexan parasites infect many different hosts and tissues, causing numerous human diseases, including malaria. These important pathogens have a peculiar cell cycle in which chromosomes sometimes amplify to remarkable levels, followed by concerted cell division—providing an unusual proliferative capacity. This capacity for proliferation, combined with an ability to change the scale of replication when needed, are hallmarks of the cell cycles of these parasites. Yet the molecular mechanism responsible for these peculiar cell cycles remains one of the unsolved mysteries of Apicomplexa biology. Here we show that the centrosome—an organelle that orchestrates several aspects of the cell cycle—of the apicomplexan parasite Toxoplasma gondii contains specialized structures that coordinate parasite cell division. Our findings demonstrate that a two-part centrosomal architecture, comprising an inner and an outer core with distinct protein compositions, segregates the processes of mitosis from the assembly of new daughter parasites. The modular organization of the centrosome offers an explanation for how cell division can be suspended while the parasites amplify their genome to the biotic scale required for their life cycles. It is unknown whether these distinct centrosome core complexes evolved independently in Apicompexa. Another possibility is that the foundations for these mechanisms were present in the original eukaryote, which could explain how the distinct extranuclear centrosome of animal cells and the novel yeast spindle pole body of the nuclear envelope may have evolved from a common ancestor.
| Infection with apicomplexan parasites is the cause of numerous important human diseases, including malaria, cryptosporidiosis, and toxoplasmosis. Pathogenesis of these diseases is closely tied to parasite replication [1] and the destruction of host cells, leading to tissue and organ damage. This fundamental relationship between parasite growth and disease is evident by the action of drugs used to combat these infections since the best treatments all reduce or block parasite proliferation. Existing therapies, in particular for malaria, are under constant pressure from acquired parasite drug resistance, a situation that requires a broad portfolio of antiparasitic compounds with different parasite-specific targets. The peculiar proliferative cycles of Apicomplexa parasites differ substantially from the hosts they inhabit and should offer fertile ground to supply an active pipeline of new treatments. To fulfill this promise, we need a better understanding of the unique structural and molecular features of parasite proliferation.
Modern Apicomplexa are the result of millions of years of evolution [2], involving successful encounters with many invertebrate and vertebrate hosts that have led to an extraordinary worldwide distribution. The development of specialized invasion and replication strategies [3–5] has permitted these parasites to surmount a variety of host-defensive barriers and achieve sufficient expansion in many different host tissues. Apicomplexan replication has adapted to different host cells, most commonly using a sequence of two chromosome replication cycles uniquely regulated in different parasite genetic lineages [4]. A single G1 phase that varies in length with the scale of parasite production precedes a first chromosome cycle (S/Mn), the biosynthetic focus of which is genome replication (nuclear cycle), followed by a unconventional chromosome cycle (S/Mn+1) that produces infectious parasites (budding cycle). The budding cycle is restricted to a single round of chromosome replication, and therefore, the amplification of the genome in the nuclear cycle determines the scale of biotic expansion. That scale can range depending on the species from a few to thousands of parasites produced from a single infected cell. Through simple variation in the nuclear to budding cycle sequence, apicomplexan parasites have solved the problem of adjusting proliferation to a wide variety of host cells. What is not understood are the mechanistic details that afford this tremendous cell division flexibility, while also preserving the fidelity of chromosome replication. Viewed from the restricted principles of model eukaryotic cell cycles, successful Apicomplexa replication often appears chaotic and in violation of some basic cell cycle restrictions (e.g. “copy once only once” in the nuclear cycle). This paradox is one of the major mysteries of the phylum Apicomplexa.
During their life cycle, Toxoplasma gondii parasites switch between multi- (merozoite stage) and binate-nuclear replication (tachyzoite stage) [6], with the binary division cycle of the tachyzoite (called endodyogeny, i.e. “inside two are borne”) now a major experimental model for understanding basic principles of apicomplexan proliferation. The cell cycle periods of the tachyzoite [4,7–10] are reasonably defined and have provided evidence for major checkpoint control that was exploited to synchronize parasite growth [9,11,12]. The most unusual feature of Apicomplexa cell division is budding, which occurs by the assembly of daughter cells within or from the mother cell using a highly ordered process [3] that is accompanied by the de novo synthesis and packaging of invasion organelles [13]. In T. gondii tachyzoites, as in Plasmodium falciparum merozoites, assembly of new parasites (budding) is guided by a unique cell cycle transcriptome that delivers proteins in a "just-in-time" order [7,14]. A major conclusion from these studies is that the tachyzoite centrosome has a vital role in coordinating budding and mitotic events. The centrosome is one of two sites of microtubule nucleation in the tachyzoite with direct responsibility for assembly of the intranuclear spindle. A second microtubule-organizing center (MTOC) is the polar ring of the apical complex. This MTOC organizes the subpellicular microtubules (MTs) scaffold supporting the pellicle that gives the parasite its shape [3]. Importantly, the centrosome also governs the position and activity of this second MTOC through a physical tether, the striated fiber [15]. During mitosis in tachyzoites, a polar striated fiber emerges from each centrosome and at its other end gives rise to the MTOC that defines the daughter cells. Both centers are thus aligned and physically connected [15].
Across the Apicomplexa phylum, similar structural and mitotic principles are observed to govern parasite division. The centrosome (also called a centriolar plaque) has a central role in regulating the Apicomplexa cell cycle [4]. Ultrastructural studies of different coccidian parasites have demonstrated that the duplication of the centrosome occurs prior to budding, and this involves an unusual parallel configuration of the internal centriole structures [4,16–18]. Assembly of new daughters initiates in close proximity to these structures [8,19,20], and in Toxoplasma mutants that fail to duplicate the centrosome parasite, budding is inhibited [21–23]. Perinuclear centrosome structures also duplicate prior to each nuclear division during the intraerythocytic cycle of P. falciparum merozoites, with coordination of these structures in the last nuclear division that precedes budding [1]. A fixed spatial orientation of the centrosome to a unique protein complex embedded in the nuclear envelope, called the centrocone, is also a common structural feature of the apicomplexan mitosis [24]. Importantly, these mitotic structures are positioned near where centromeres tether chromosomes during interphase [3,4,25], and it is through the centrocone structure that spindle fibers pass into the nucleus to segregate chromosomes during mitosis [4]. The centrosome also has a master function in limiting chromosome replication in the tachyzoite that appears to require physical contact. Evidence from tachyzoite growth mutants indicates that "copy once" restrictions in the budding cycle require a connection between the centrosome and new daughter cytoskeleton [26,27]. Breaking this contact by drug treatment [28] or by genetic ablation of centrosome factors [15,22,26] leads to unregulated nuclear re-duplication and abnormal budding. To help explain the master functions associated with this mitotic organelle, we describe here the discovery of a complex internal structure of the T. gondii centrosome that is comprised of two independent and replicating core structures. The centrosome cores have unique structural and regulatory protein composition, and each core has a fixed spatial orientation to the budding and mitotic machineries. Genetic analysis of specific centrosome proteins indicates that each type of core serves distinct roles in regulating either cytokinesis or karyokinesis. Together, these results support a model whereby differential modulation of the centrosome cores could provide the flexibility required to achieve different modes of apicomplexan parasite replication.
To proliferate, T. gondii tachyzoites build two daughter parasites, internally enclosing an intact nucleus and a full complement of metabolic and invasion organelles. The daughter buds then consume the residual mother cell to become infectious parasites. The centrosome must be duplicated for this process to unfold, indicating a central organizing function (shown as red dots in diagram, Fig. 1A) [3,4,8]. Budding structures emerge close to the centrosome, and the centrosome is oriented to the unique centrocone compartment embedded in the nuclear membrane that is required for chromosome segregation [25,29]. The molecular basis for the complex coordinating functions of the centrosome in these parasites is not well understood. To decipher the structural features of the T. gondii centrosome, we mined Apicomplexa genome sequences for conserved centrosomal proteins (see Table 1 and S1 Fig. for gene lists) and examined how these factors are expressed during tachyzoite division. One of the key centrosome structural proteins present in T. gondii is an ortholog of the cartwheel protein Sas-6 [30]; note that this protein is distinct from the recently described TgSas-6L, which is associated with the apical MTOC [31]. Other canonical centrosomal proteins identified in T. gondii are centriole elongation factor Sas-4 [32,33], centrin-binding protein Sfi1 [34], gamma-tubulin, and a large protein that contains a single Aurora-like kinase domain (Table 1). The T. gondii Sas-6 ortholog retains the three-domain structure of the human protein known to be responsible for the 9-fold symmetry of the centriolar barrel (S1A Fig.) [35,36]. The organization of the PISA (Present in Sas-6) motif and Sas-6 conserved domain [30,37] relative to the central coiled-coil domain is also preserved. The major difference between these proteins is the extended and unstructured N- and C-tails of TgSas-6 (S1A Fig.). Epitope tagging the TgSas-6 protein in T. gondii tachyzoites by genetic knock-in (TgSas-6HA, C-terminal 3xHA fusion) demonstrated co-localization with centriolar TgCentrin1myc that was also epitope tagged in the endogenous gene locus in this clone (Fig. 1B and see S2 Table for a full list of transgenic strains used in this study). DAPI (4’, 6–diamidino-2-phenylindole) co-staining was included in this analysis in order to follow the distribution of genomic DNA and to localize the nucleus in the parasites studied. This immunofluorescent microscopy analysis (IFA) validates the centrosome assignment of the TgSas-6 ortholog. We identified an ortholog of the centrin-binding protein Sfi1 (Table 1 and Fig. 1C and S1B and S2 Figs.) first discovered in budding yeast and also present in human cells [34,38]. The TgSfi1 ortholog is a highly disordered protein comprised of multiple short domains and a large disordered C-terminus (1,145 residues, http://www.disprot.org) [39]. A total of 31 divergent centrin binding sites were identified that are evenly distributed over two-thirds of the protein length with none in the disordered C-terminal tail (S1B Fig. and S2 Fig.). Consistent with its known role in pairing with centrin, we found endogenously tagged TgSfi1myc closely co-localized with TgCentrin1HA in the tachyzoite centrosome (Fig. 1C). As expected, T. gondii γ-Tubulin (Table 1) localized to the centrosome and duplicated along with TgCentrin1-associated structures (Fig. 1D).
Studies of the centrosome in higher eukaryotes identified a number of large coiled-coil proteins with important functions associated with centrosome and are, therefore, named the CEP (CEntrosomal Proteins) family of proteins (for review see [40]). Using a combination of BLAST and protein structural analysis, we found several T. gondii genes encoding proteins similar to the CEP protein family of higher eukaryotes. Apicomplexan orthologs for CEP76, CEP97, CEP120, and CEP250 are listed in Table 1 (structural features shown in S1D Fig. and Fig. 2A). The CEP250/c-Nap1 factor is known to be required for centrosome separation in animal cells [41], and it was of particular interest to determine if this factor was present in tachyzoite centrosomes. The T. gondii gene encoding a protein with highest similarity to human CEP250/C-Nap is TGME49_212880 (Table 1 and S1 Table). Human CEP250 has coiled-coil domains running across the entirety of the protein length (Fig. 2A, indicated by the red plot line). The T. gondii TGME49_212880 protein also has numerous coiled-coil domains and appears to preserve an overall architecture of the two central extended helical domains surrounded by short coiled-coil regions (Fig. 2A, blue line and conservation block) as well as having long extended N- and C-terminal tails. To verify centrosomal localization, we tagged TGME49_212880 by genetic knock-in with a 3xmyc epitope. TGME49_212880myc was associated with a discrete perinuclear structure consistent with the position of the centrosome (Fig. 2B). We have designated TGME49_212880 (TgCEP250) as the ortholog of hCEP250.
Curiously, the T. gondii genome also encodes an abundant complement of coiled-coil factors with significant homology to human CEP250. The ten CEP250-related T. gondii proteins with highest similarity scores predicted to have extended coiled-coil domains (S1 Table) comprise a group of novel proteins with variable length (from 1,232 to 6,668 amino acid residues). Two of the TgCEP250-like factors on this list were recently described as proteins containing the CRMP domain, were localized to the apical cone (TGME49_244470) or conoid (TGME49_252880) [42], and were not studied here. Two other CEP250-like proteins were endogenously tagged, and immunofluorescence microscopy analysis revealed differential subcellular localizations. Protein encoded by TGME49_242790 gene localized to the peripheral annuli of the parasite cytoskeleton (S3A Fig.), a novel compartment that was previously shown to house TgCentrin2 [43], and therefore, was given the name TgPAP1 (Peripheral Annuli Protein 1). A non-periodic CEP250-related protein encoded by gene TGME49_265840 formed a peculiar fibrous net enclosing nucleus (S3B Fig.) and was named TgNMP1 (Nuclear Mesh Protein 1).
The only other factor in the group of CEP250-like proteins that showed centrosomal localization similar to TgCEP250 was a novel member of the family encoded by the TGME49_290620 gene (Fig. 2C, co-localization with TgCentrin1). The distinguishing structural feature of this CEP250-related factor was a single coiled-coil domain that spanned a central 800 residues within the large 2663 amino acid protein (Fig. 2D). We designated this protein TgCEP250-L1 (TgCEP250-like protein 1). To determine co-localization of this novel centrosomal protein with the related factor TgCEP250, we created a dual-tagged transgenic strain expressing TgCEP250-L1HA and TgCEP250myc (S2 Table). Surprisingly, in tachyzoites undergoing mitosis we observed four TgCEP250-positive perinuclear foci, while TgCEP250-L1HA protein co-localized with only two of the TgCEP250myc positive foci that were always proximal to the nucleus (Fig. 2E; TgCEP250-L1HA, red; TgCEP250myc, green).
The discovery of multiple structures in the perinuclear region occupied by centrosomal factors TgCEP250 and TgCEP250L-1 indicated the centrosome of tachyzoites has a complex organization. To investigate this further, we produced several single or dual epitope-tagged transgenic strains (see S2 Table) and in combination with antibodies to TgCentrin1, TgMORN1 (centrocone), or TgCenH3 (centromere/inner kinetochore) examined in detail the structural organization of the parasite centrosome and associated mitotic nuclear structures. An unexpected result of these studies was finding that the TgCentrin1 marker commonly used to identify centrosomes in these parasites was contained in only one of two internal protein core structures of the tachyzoite centrosome. The TgCentrin1-associated protein core was named the outer core because it was always distal to the tachyzoite nucleus (Fig. 3A), while an inner core (protein complex with immediate perinuclear orientation) lacking TgCentrin1 harbored proteins TgCEP250myc and TgCEP250-L1HA (Fig. 3A and B, and Fig. 2E). Paired with TgCentrin1HA, the TgSfi1myc protein was exclusively localized to the outer core (Fig. 3C) as was the centriole cartwheel protein TgSas-6 (dual tagged strain, TgSas-6HA/TgCentrin1myc, Fig. 1B). The TgCEP250-L1HA-associated inner core structure was closely aligned, but resolved from the centrocone in late mitosis (Fig. 3D, anti-TgMORN1 co-staining) [29]. Co-staining with putative nuclear envelope factor TgNMP1myc placed the inner core on the outside of the nucleus (Fig. 3E, TgCEP250-L1HA/TgNMPmyc). TgCEP250myc protein was associated with both types of core structures in mitotic tachyzoites (Fig. 3A), while TgCEP250-L1HA showed more exclusive association to the inner core (Fig. 3B) with transient translocation to the outer core providing possible linkage between two structures (Fig. 3D). T. gondii γ-Tubulinmyc showed preferential localization to the TgCentrin1-containing outer core and not to the inner core (Fig. 3F, CEP250-L1HA/γ-Tubulinmyc). Finally, the visualization of the two cores (TgSas-6HA and TgCEP250-L1myc) with the centromere/inner kinetochore complex (TgCenH3) by structured illumination microscopy (Fig. 3G, Sas-6HA/CEP250-L1myc/anti-CenH3) confirmed that the parasite centrosome contained two novel cores that aligned to the kinetochore constituting a multi-layered mitotic machinery spanning the nuclear membrane of the dividing tachyzoite. A diagram (Fig. 3H) summarizes the protein composition and alignment of the two centrosomal cores in one-half of a dividing nucleus.
To understand the duplication of the unusual tachyzoite centrosome, we followed the development of the dual centrosome cores and the centrocone in all phases of the parasite cell cycle (Fig. 4A: TgSas-6HA, blue; TgCEP250-L1myc, red; anti-TgMORN1, green). Newly formed G1 parasites inherit a single, condensed centrosome as demonstrated in the first series of images (Fig. 4A, vertical series 1). The compact G1-centrosome was tightly associated with the nuclear envelope and enlarged during G1 progression. In late G1, the outer core containing TgSas-6HA expanded and duplicated (Fig. 4A, series 2). Simultaneously, the inner core containing TgCEP250-L1myc separated from the outer core, although it retained close association with the nuclear centrocone (Fig. 4A, series 2 green stain). The expanded inner core duplicated immediately after the outer core had replicated, also retaining the distinctive spatial orientation with respect to the nucleus, inner core proximal, and the outer core distal (Fig. 4A, series 3). The duplication of the cores prepared the tachyzoite for mitosis and analysis by super-resolution microscopy demonstrated that the cores aligned with the TgCenH3 containing inner kinetochore [25] in a typical metaphase relationship (Fig. 4B). The two outer cores (TgSas-6HA, red stain), two inner cores (TgCEP250-L1myc, green stain), and centromeres (anti-TgCenH3, blue stain) were aligned in a linear array extended 1,300 nm from one outer core to the other. At this mitotic stage the distance between each pair of outer and inner centrosome cores measured 300 nm, reaching a maximum of 400 nm in anaphase and telophase (Fig. 4B). The nuclear centrocone containing TgMORN1 protein split in two last in the observed mitotic sequence (Fig. 4A, series 4). The distinct centrosome core structures segregate into each daughter parasite (Fig. 4A, series 5) indicating there is a physical mechanism that ensures each new parasite receives one copy of each type of centrosome core.
The discovery of two distinct centrosome core complexes and the association of TgCentrin1 to only one core [44] indicates the Toxoplasma centrosome is more complex than previously appreciated. We turned to the large collection of conditional cell cycle mutants generated by earlier studies [21,27] for insight into their interdependence and regulation. Among the group of mutants with defects in mitosis, the mutant 9–86E4 possessed a temperature-sensitive allele of the outer core protein TgSfi1 (see characterization of the protein in Figs. 1C, 3C, S1B, S1C, and S2). Mutant 9–86E4 parasites grew normally at the permissive temperature of 34°C, while at 40°C they quickly growth arrested (Fig. 5A). At 40°C mutant 9–86E4 parasites were typically larger than parasites grown at 34°C, although this was not systematically quantified. At high temperature, there was absence of budding. Parasites were able to duplicate the nucleus (see circled parasite in Fig. 5B, 40°C panel), but with the primary budding defect this led to unequal DNA and nuclei distribution with rare DNA-free zoites present. Genetic complementation, followed by marker rescue (Fig. 5C) [23,26] and whole genome sequencing of mutant 9–86E4 (see “Identification of Temperature-Sensitive Mutations” in Materials and Methods) [45,46], pinpointed a E1759K mutation in the TgSfi1 gene as responsible for these growth defects (Fig. 5D). To monitor the ts-allele of TgSfi1 protein, we introduced 3xHA-epitope into the ts-TgSfi1 locus using modified CRISPR technology [47]. Western blot analysis of the mutant parasites expressing ts-TgSfi1HA showed high instability of the tagged protein after 24 h incubation at 40°C (Fig. 5E), indicating that parasites grown at 40°C were phenotypically null for this factor. Consistent with a centrosomal function, and as a known partner with Centrin [48], the loss of ts-TgSfi1 led to dramatic reduction in the TgCentrin1-associated cores in the 9–86E4 mutant at 40°C (Fig. 5B and F).
To determine how the inner core containing TgCEP250-L1 was affected by the loss of ts-TgSfi1 at high temperature, we introduced into the 9–86E4 mutant a C-terminal tagged version of the TgCEP250-L1 protein regulated by its own promoter (see fosmid construction in “Generation of Transgenic Tachyzoite Strains” of Materials and Methods). The resulting transgenic strain carrying the ts-TgSfi1(E1759K) mutation was used to visualize both the outer (anti-Centrin, green) and inner core (TgCEP250-L1HA, red). At the permissive temperature, parasites maintained the expected 1:1 ratio of outer/inner cores (Fig. 5F, average 1.37 at 34°C in a population average from an asynchronous culture comprised of 60% G1 and 40% S and M/C parasites). However, a shift to 40°C reduced the number of the TgCentrin1-outer cores below the minimal one per nucleus (Fig. 5F, dot plot: 0.8 at 40°C versus 1.4 at 34°C), while the number of TgCEP250-L1HA-inner cores dramatically increased beyond the proper 1:1 nuclear stoichiometry (Fig. 5F, red). While the TgCentrin1 outer cores were reduced, the few outer cores that remained were maintained in the normal distal perinuclear position observed in normal replicating parasites. By contrast, over-amplified inner TgCEP250-L1 cores decorated the nucleus in a pattern that was independent of the remaining outer TgCentrin1 cores, indicating the position and replication of the inner core structures was uncoupled from the outer core. Intriguingly, the amplification of inner TgCEP250-L1 cores was always paired with and orientated to replicated TgMORN1-associated centrocone structures, suggesting a common mechanism was directing the amplification of both structures (Fig. 5G). Therefore, the loss of ts-TgSfi1 limited function of the outer core at the restricted temperature, which caused blocking of the outer core duplication while loosening control of the single duplication of the inner core. The other consequences of the defect were physical separation of two cores and severe restriction of daughter budding.
The uncoupling of the replication of outer and inner cores in the ts-TgSfi1 mutant above indicates that complex regulatory mechanisms operate in the Toxoplasma centrosome to control duplication of the centrosome cores. To investigate the replication of these cores further, we studied a serine/threonine protein kinase related to mammalian ERK1, which previous studies determined is required for tachyzoite growth, although the underlying mechanism responsible was not reported [45]. The gene TgME49_312570 is one of three protein kinase genes in T. gondii possessing a MAPK-like kinase domain [49]. The similarity of TGME49_312570 to eukaryotic MAPK factors lies almost exclusively in the ATP binding pocket, which corresponds to approximately 270 amino acids (aa) of an otherwise 1,298 aa novel protein. Given the lack of the MEKK-MEK-MAPK signal transduction module in the Apicomplexa [50], a similarity limited to part of the kinase domain in TGME49_312570 and no established mechanistic function for this protein, we have designated this gene as TgMAPK-like 1 (TgMAPK-L1). Epitope tagging of this protein by genetic knock-in demonstrated a prominent pericentrosomal pattern surrounding the TgCentrin1 outer core in the tachyzoite S-phase and early in mitosis (TgMAPK-L1HA; Fig. 6A and S4A Fig.). During budding, expression of the TgMAPK-L1HA rapidly decreased and dropped below detection level in the newly emerged G1 parasites, consistent with the cyclical pattern of the encoded mRNA (protein cell cycle properties; S4A Fig.; for mRNA profile see Toxodb). Similar staining is often seen for proteins localized in the pericentriolar matrix (PCM) [51], which, because of low conservation of PCM markers, have not been identified in T. gondii. The pericentrosomal localization of TgMAPK-L1 makes this protein a first candidate for the PCM compartment.
There are few reports in animal cells of MAPKs exclusively localized to the centrosome as we observed for TgMAPK-L1HA in tachyzoites, nor is direct mitotic control the mechanism typically associated with MAPKs of higher eukaryotes, in which signal transduction in response to external growth factors is the more common function [52]. Insight into the function of TgMAPK-L1 was provided by a temperature-sensitive mutant, 11–31G12, recently identified in the large collection of tachyzoite cell cycle mutants [21]. Mutant 11–31G12 parasites carrying a L534Q mutation immediately C-terminal of the TgMAPK-L1 kinase domain (S4 Fig.) rapidly growth arrest at 40°C (Fig. 6B) with defects in the coordination of daughter budding with mitosis leading to abnormal numbers of internal daughters and nuclei (Fig. 6C, 40°C panel). The unlinking of parasite budding and nuclear duplication in the mutant led to an elevated ratio of centrosome to bud numbers, as evident from anti-Centrin staining (Fig. 6D, dot plot: 1.53 at 34°C versus 4.6 at 40°C). Genetic complementation followed by marker rescue and ts-allele sequencing identified ts-TgMAPK-L1 as responsible for the high temperature defects (S4B Fig. and S4C Fig.). To verify that the mutation in ts-TgMAPK-L1 was solely responsible for the cell cycle defects, we transferred the L534Q mutation into parent RHΔku80 parasites and simultaneously tagged the ts-TgMAPK-L1 protein with three copies of the HA epitope (S4D Fig.). The introduction of the L534Q mutation recapitulated the temperature and cell cycle defects (S4E Fig.) of the 11–31G12 mutant. After 20 h at 40°C, tachyzoites carrying the ts-TgMAPK-L1 mutation were yet to produce a mature daughter parasite with multiple buds forming in a single mother cell (S4E Fig., anti-IMC1 panel), indicating there is a loss of the critical controls that restrict binary division in these parasites. Western blot and IFA analysis of the ts-TgMAPK-L1HA protein (S4E,F Fig.) indicated that instability of this protein at high temperature creating a null phenotype is the major cause of conditional growth arrest in ts-TgMAPK-L1 mutant parasites.
To further investigate the mechanism of irreversible and lethal growth arrest caused by ts-TgMAPK-L1, we introduced 3xHA-tagged TgCEP250-L1 into the original ts-TgMAPK-L1 mutant, 11–31G12, and with the use of anti-Centrin antibody analyzed the outer and inner centrosome cores in this mutant. Upon shift to high temperature, we observed rapid amplification of both centrosomal cores that roughly maintained the internal alignment (Fig. 6E; anti-Centrin, green; TgCEP250-L1HA, red). Replication of the inner core was often delayed in the mutant, leading to accumulation of the distinctive “dumbbell” forms (Fig. 6E, inset 40°C panel). We next examined the relationship between the centrosome and developing daughter structures and how the loss of ts-TgMAPK-L1 in the centrosome affected critical features of daughter budding in mutant parasites (Fig. 7). In these experiments we monitored TgMORN1, which is present in the distinctive spindle-associated nuclear centrocone and early (daughter) basal ring [29]. Both of the basal ring and centrocone structures are tightly associated in a single complex in the parasite S phase and early mitosis (Fig. 7A and 7B, 34°C) and are in close proximity to the PCM localized TgMAPK-L1HA in wild-type tachyzoites (Fig. 7A, magnified merge images on the right). Although deficiency in ts-TgMAPK-L1 in mutant parasites did not affect alignment of the nuclear centrocone with centrosome (Fig. 7B; 40°C; lower panel; anti-TgMORN1, green; inner core visualized with TgCEP250-L1HA, red), it severed stable associations between the centrocone and the early forming daughter basal rings (Fig. 7B, 40°C panel, loose basal rings). Disconnection of the basal ring and centrocone compartments was observed in 50% of the ts-TgMAPK-L1 parasites at high temperature (Fig. 7C). Further analysis of the ts-TgMAPK-L1 mutant revealed that a delay in basal ring development accompanied by subsequent karyokinetic counting defects was the prevalent phenotype of ts-TgMAPK-L1 parasites at 40°C. The images of co-markers TgMORN1 and TgCEP250-L1HA in the Fig. 7D illustrate three consecutive stages of the centrocone/basal ring uncoupling defect in ts-TgMAPK-L1 parasites. Note that under normal conditions, the cell cycle length of T. gondii tachyzoite used in this study is 8 h at 37°C [7,10], and by 20 h post-infection, replicating parasites typically complete three cell cycles in a single infected host (four to eight parasites per vacuole). By contrast, mutant 11–31G12 parasites at 40°C do not divide following invasion, forming large cells that retain the original mother cell IMC1 and basal complex (Fig. 7D; panels 1 and 2; anti-IMC1, red; anti-TgMORN1, green). The majority of ts-TgMAPK-L1 deficient parasites had over-amplified centrocones, with a distinct subpopulation also failing to form daughter basal rings (Fig. 7D, panel 1). The lack of basal rings appeared to induce another round of nuclear duplication in the absence of budding. In other parasites, we observed the formation of daughter basal rings that retained the proximity to the centrocone (Fig. 7D, panel 2) leading to the assembly of multiple abnormal daughter buds (Fig. 7D, panel 3). At high temperature, all daughter formation in these vacuoles was nonviable due to numerous mitotic defects (Fig. 6C, retention of the mother DNA in the lower panel). Together, these results indicate TgMAPK-L1 has a specific role in restricting tachyzoite nuclear replication that likely involves the preservation of the physical connection between the karyokinetic and cytokinetic centers.
It had been shown that several mitotic kinases, including cyclin-dependent, Polo-like, Aurora, and NIMA-related kinases, coordinate timing and fidelity of the centrosome function in higher eukaryotes [53]. Similarly, a recently identified T. gondii ortholog of the NIMA-related kinase, TgNek1–2, showed dynamic association with the centrosome in which it positively regulated the TgCentrin1-core complex duplication [54]. Another group of the conserved eukaryotic kinases is Aurora family, which plays a central role in the establishment of the bipolar spindle [55]. A single gene in the Toxoplasma genome encodes a protein possessing a kinase domain similar to Aurora kinases of higher eukaryotes [56,57]. The single exon gene (Table 1) encodes a large, 2,812 amino acid protein that is dynamically regulated over the tachyzoite cell cycle at the mRNA and protein level (Fig. 8A and B). We epitope tagged TgAurora-related kinase 1 (TgArk1) with 3xHA and discovered by IFA analysis that this regulatory factor provides another example of the heterologous composition of the centrosome core structures in this parasite. As predicted by the mRNA profile, TgArk1HA was not detected in G1 parasites but quickly reached maximum expression in S phase and gradually disappeared from parasites containing mature daughter buds (Fig. 8B, red). In S phase and early mitosis, the TgArk1HA factor was localized exclusively to the TgCentrin1-associated outer core of the parasite centrosome (Fig. 8B, TgArk1-head morphology) and not in the centrocone compartment that is visualized using anti-TgMORN1 antibody (Fig. 8C) [24,29,58]. Later in mitosis, TgArk1HA became associated with a linear structure (Fig. 8B and C, TgArk1-tail morphology) lying along one side of the growing daughter bud. To determine whether TgArk1-tail associated with nuclear or microtubule structures, we treated TgArk1HA transgenic parasites with microtubule disrupting agent oryzalin. In the absence of the assembled subpellicular microtubules, the TgArk1-tail associated with the inner membrane disrupted material (Fig. 8D, anti-IMC1 co-staining) and not with the parasite nucleus (Fig. 8D, DAPI).
The T. gondii tachyzoite stage studied here divides using a budding cycle that is limited to one round of genome duplication (endodyogeny), which is a restriction lifted in the merozoite stage of the cat life cycle that replicates by a sequence of nuclear and budding cycles (Fig. 9). This ability of Toxoplasma and other parasites of this family to adapt their cell cycles to different hosts and tissues is one of the unsolved mysteries of Apicomplexa molecular biology. The key event that initiates the tachyzoite budding cycle is the duplication of the centrosome at the G1/S boundary [10,23]. Once duplicated, a complex process unfolds in and around these central structures, including assembly of the striated fiber required for budding initiation [15], formation of the spindle needed for mitosis, and organelle segregation [4]. How the centrosome performs the myriad of coordinating functions may be explained in part by our discovery of a unique binary internal organization. In this study, we show that the tachyzoite centrosome has two replicating core complexes. These cores have a distinct protein composition and a stereotypical geometry that defines their orientation with respect to the nucleus and forming daughter cells (Fig. 9, budding cycle). Interestingly, protein kinases, including TgArk1, TgMAPK-L1, and TgNek1–2 [54] appear to decorate specific structures in and around the centrosome indicating a complex and specialized regulatory machinery likely operates from the centrosome. The inner core complex (closest to the nucleus) of the tachyzoite centrosome contains factors TgCEP250 and TgCEP250-L1 and is aligned with the centrocone (Fig. 9), which in turn is oriented to the intranuclear kinetochore/centromeres throughout the cell cycle [25]. This arrangement suggests the inner core and centrocone may work in concert to control the adjacent nuclear environment. The unusual co-amplification of the centrocone and the inner centrosome core in the ts-TgSfi1 mutant shown here is consistent with a shared regulatory relationship. By contrast, the principle role for the centrosome outer core complex, which contains factors TgCentrin1/TgSfi1, TgSas-6, and the large TgArk1 protein kinase, is in regulating the initiation and assembly of daughter buds (Fig. 9, budding versus nuclear cycles). Genetic experiments strongly support this spatial segregation of function. The loss of the outer core in mutant 9–86E4, which we show in this study to be defective in ts-TgSfi1, leads to a primary block in budding (Fig. 6). Conversely, the Gubbels laboratory has recently produced a mutant of TgCEP250 leading to a loss of the inner core and a primary disruption of mitosis, while duplication of the outer core and budding initiation was not lost (Chen and Gubbels, personal communication). These consequences are reciprocal to the loss of budding and outer cores in the ts-TgSfi1 mutant; here the TgCEP250 inner core amplifies and the nucleus divides (Fig. 6). Altogether, these results indicate that the principle roles of the outer and inner centrosome cores in cytokinesis and karyokinesis, respectively, can be uncoupled. Notwithstanding the distinct protein composition and unusual independent function of the tachyzoite centrosome cores, there are other mechanisms that ensure each type of core is inherited by daughter parasites. Both cores duplicate at the G1/S phase transition, segregate by pairs into each daughter, and are co-regulated by factors residing in the PCM, which surrounds the internal cores. The loss of ts-TgMAPK-L1 in the temperature-sensitive mutant 11–31G12 leads to abnormal amplification of both centrosome cores and disrupts the normal linkage between the new daughter cytoskeleton and the centrocone structure (Fig. 7 and Fig. 8). Consistent with previous studies in which spindle and subpellicular microtubules were disrupted with oryzalin [28], breaking the physical connection between mitotic and budding machinery in the ts-TgMAPK-L1 mutant also causes a loss of normal restrictions over nuclear and daughter bud duplication. These results indicate TgMAPK-L1 has a key role in determining the scale of parasite counting in Toxoplasma replication (Fig. 9).
Proteomic analysis of human centrosomes reveals a more extensive asymmetrical protein distribution in centrosome structures than previously considered [59], and examples of this core diversity are elegantly demonstrated by recent high-resolution microscopy of human centriole structures [51]. Structural heterogeneity of the centrosome is thought to pave the way for specialized development [60] and also represents stages of procentriole maturation, although the diversity of novel protein complexes in centrosomes from different eukaryotes [60] suggests we are only beginning to understand the functions involved. The family history of the centrosome cores in the T. gondii tachyzoite is not fully understood; however, the spatiotemporal behavior of these structures is very distinct from animal cells. Coccidian parasites, which include T. gondii, possess recognizable centrioles in their centrosomes that are arranged not orthogonal, but in distinct parallel configuration [17,61]. They are also shaped differently (200 nm x 200 nm) and have a nine plus one all singlet organization [17,61]. There does not appear to be an equivalent procentriole maturation process as seen in animal cells; rather, we have shown here that the outer TgCentrin1/TgSfi-1 core duplicates first in early S phase, followed in minutes by the duplication of the TgCEP250/TgCEP250-L1 inner core. These core structures progressively resolve from each other over the next hour, reaching > 400 nm of separation in the post-metaphase of tachyzoite mitosis. Our results also demonstrate the TgCEP250/TgCEP250-L1 inner core never acquires outer core proteins TgCentrin1/TgSfi1, TgSas-6, or TgArk1 kinase. This would be expected as cells progress through successive cell cycles, if the inner core was a procentriole destined to become a mature mother centriole. These findings raise questions about the biogenesis of the tachyzoite centrosome core structures. The association of TgSas-6 cartwheel protein only with the outer TgCentrin1/TgSfi1 core provides a known template capable of seeding the duplication of this structure, as this factor templates the 9-fold symmetry of MT assembly of centrioles in other eukaryotes [30,35]. How the inner core replicates is more of a mystery because we did not detect TgSas-6 in this structure. It is possible we missed a transient association of TgSas-6 with the inner core, or de novo synthesis of the inner core centriole is responsible for duplication, which is known to occur in rare instances in eukaryotic cells [60]. Higher resolution studies of centrosome biogenesis in these parasites should help resolve these questions and test if both outer and inner centrosome cores in Toxoplasma tachyzoites have centrioles, although it remains a possibility that the inner core is part of the unique spindle pole complex, and therefore lacks a centriole.
Comprehensive phylogenetic analysis of several centriolar factors defined the inheritance of an ancestral module that regulates the 9-fold symmetry and the assembly of the centriole microtubules that precedes bikont and unikont divergence [37]. We have shown here that several of these core centrosome proteins are conserved in T. gondii, and they are also found encoded in the genomes of most other apicomplexans. It is therefore surprising that experimental evidence of the MT centriole barrels in the centrosome exists only for the coccidian branch, while the cartwheel protein Sas-6 that templates this structure is present in all apicomplexan branches (Table 1) [30,37]. Centrioles of Toxoplasma are small (200 x 200 nm) compared to animal cells (700 x 250 nm) and experimentally challenging to recognize in ultrastructure preparations [4]. Intriguingly, P. falciparum orthologs of several T. gondii outer core proteins have peak mRNA expression quite late in the intraerythrocytic cycle. Importantly, the mRNA encoding PfSas-6 cartwheel protein (PF3D7_0607600) is maximum at >35 h in growth-synchronized merozoites. This profile indicates that peak PfSas6 expression occurs in late schizogony, which is several hours after the initiation of nuclear replication in nuclear cycle [14]. PfSas-4 (PF3D7_1458500) and the ortholog of the large T. gondii aurora-related kinase (PF3D7_0309200) as well as one of two centrins (PfCEN2) that associate with the P. falciparum centrosome are also exclusively expressed in the late schizont (PfCEN2 and 3) [62]. The late expression of PfSas-6, PfCEN2, and the aurora-related kinase may indicate they function in the final cell cycle as part of the global control machinery that coordinates budding. Such a switch from local to global control is required to allow for synchronous budding and to complete P. falciparum merozoite replication [1]. Many of the molecular details of centrosome architecture and function in the Apicomplexa remain to be explored. However, the critical problem of how cytokinesis (budding) might be suspended during nuclear reduplication in the Apicomplexa may be solved by the fundamental independence of a two-compartment centrosome, in which one compartment controls budding while the other rules mitosis that we describe here for the T. gondii tachyzoite. This model provides a new framework to understand how multi-nuclear schizogony replication of P. falciparum and other important apicomplexan parasites is achieved. As noted previously [1], the unusually complex mitotic structures in these parasites appears to be a mix of strategic elements reminiscent of the mammalian extranuclear centriolar centrosome and the nuclear embedded yeast spindle pole body. This complex architecture may have evolved in these parasites to achieve regulatory diversity. Importantly, these distinct structural elements are connected and often remain physically tethered in a fixed linear array during the tachyzoite budding cycle. Spindle microtubules from the chromosomal centromeres hitch the genome to the inner core of the centrosome (note that centromeres remain sequestered in this region through interphase). Emanating from the centriolar region of the centrosome, the striated fiber connects to the new apical microtubule-organizing center of the daughter bud [15]. This fiber extends during the growth of the bud pellicle and then disappears as the daughter cytoskeleton reaches maturity. Given the elegant demonstration of active cyclin-CDK protein complexes tethered to the mitotic spindle in Hela cells [63], it is not a stretch to suggest that cell cycle checkpoints in the Apicomplexa likely exploit the remarkable physical connections from chromosome to daughter bud in order to coordinate cytokinesis and karyokinesis in the budding cycle. There are few examples of eukaryotic cells with specialized centrosome cores, and we would propose that this arrangement could achieve the regulatory flexibility required for these parasites to adapt to multiple host life cycles (Fig. 9). The implication of this model is that differential regulation of centrosome core composition and/or activity could provide the switch between the local control of chromosome replication in the nuclear cycle to the globally controlled "copy once" regimen of the budding cycle (Fig. 9). It is conceivable that post-translational regulation of the outer core could regulate activation or suspension of the budding cycle (Fig. 9, nuclear cycle), and the presence of TgArk1 exclusively in this core structure highlights a possible candidate regulator. Further studies to define how multiple organizational hubs (i.e., centrosome and centrocone) segregate responsibilities within the apicomplexan cell cycle will be important to understand this critical and truly fascinating aspect of the parasites’ life cycles.
Parasites were grown in human foreskin fibroblasts (HFF) as described [64]. All transgenic and mutant parasite lines are derivatives of the RHΔhxgprt parasite strain [65]. Temperature-sensitive clones 9–86E4 and 11–31G12 were obtained by chemical mutagenesis of the RHΔhxgprt strain [21]. Growth measurements were performed using parasites pre-synchronized by limited invasion, as previously described [12,26]. Parasite vacuoles in the infected cultures were evaluated over various time periods with average vacuole sizes determined at each time point from 50–100 randomly selected vacuoles.
Endogenous tagging by genetic knock-in technique. Selected T. gondii proteins were tagged with a triple copy of the HA or myc tag by genetic knock-in (See S2 Table for full list of genes, primers and transgenic strains created in the current study). PCR DNA fragments encompassing the 3′-end of the gene of interest (GOI) were used to construct the plasmids pLIC-GOI-HA3X/dhfr-DHFR-TS, pLIC-GOI-HA3X/dhfr-HXGPRT or pLIC-GOI-myc3X/dhfr-DHFR-TS and the constructs were electroporated into RHΔku80 strain deficient in non-homologous recombination [66]. The double-tagged transgenic lines were established by sequential selection under alternative selection markers with cloning. Expression of the epitope tagged fusion proteins was verified by IFA.
A new strain expressing the ts-TgMAPK-L1 mutation was generated in the RHΔku80 strain. To introduce the L534Q mutation into a new genomic background, we PCR amplified a 3,354 bp DNA fragment from mutant 11–31G12 that includes the 3′ end of the TGGT1_312570 (ts-TgMAPK-L1) using primers LIC-TgMAPK-L1_FOR and LIC-TgMAPK-L1_REV (S2 Table). In parallel, we also amplified a genomic fragment from the wild-type TGGT1_312570 locus in the RHΔku80 to generate 3xHA tagged native TgMAPK-L1HA. The PCR products were cloned into pLIC-HA3X/dhfr-HXGPRT vector, and the resulting construct was introduced in RHΔku80 strain [66]. Strains were tested for growth at 40°C and analyzed by IFA and western blot analysis.
Endogenous tagging using CRISPR/Cas9 technology. To introduce 3xHA-epitope to the C-terminus of ts-TgSfi1, we constructed gsTgSfi1 CRISPR/Cas9 plasmid by modifying sgUPRT-CRISPR/Cas9 plasmid generously provided by Dr. David Sibley (Washington University, MO, United States), as previously described [47]. Replacement was driven by Q5 DNA polymerase mutagenesis (New England Biolabs, Ipswich, MA, US) using primers specific for TgSfi1 gsRNA (S2 Table). To obtain the insertion cassette that includes C-terminus of TgSfi1 gene fused to 3xHA-epitope and HXGPRT selection marker we introduced synonymous mutation into the PAM site of the corresponding gsRNA in the pLIC-tgSfi1-HA3X/dhfr-HXGPRT plasmid using Q5 DNA polymerase mutagenesis (see primers design in the S2 Table). The amplified insertion cassette and sgTgSfi1 CRISPR/Cas9 plasmid were mixed in 1:1 molar ratio and electoporated into the mutant 9–86E4 parasites. Selection for growth in the mycophenolic acid/xanthine media was performed at 34°C.
Ectopic expression of epitope-tagged TgCEP250-L1. A large insert fosmid clone containing a fragment of chromosome IX (3757055–3790435) that includes the TgCEP250-L1 gene (RHfos10J10) [67] was modified by recombination with a cassette containing a 3xHA epitope tag, a chloramphenicol selection cassette (T. gondii selection), and a gentamycin selection cassette (for selection in bacteria) downstream of the TgCEP250-L1 gene. Primers contained appropriate overhangs to provide homology for recombination into the TgCEP250-L1 gene 3′end in the fosmid were used (S2 Table). Recombined fosmid was introduced in the mutants 9–86E4 and 11–31G12 and selected in the medium with 20 μM chloramphenicol. Established clones were analyzed by IFA and tested for growth at the permissive (34°C) and non-permissive (40°C) temperatures.
Ectopic expression of epitope-tagged TgCentrin2. The coding sequence of TgCentrin2 (TGME49_250340) was amplified from RHΔhxgprt cDNA library (see primers design in the S2 Table) and cloned into the pDEST_tub-YFP_CAT plasmid by recombination (Gateway, Life Technologies), which resulted in the C-terminal fusion of TgCentrin2 with YFP-protein. Recombinant plasmid was introduced into RHΔku80 parasites expressing endogenously tagged TgPAPHA protein.
Temperature-sensitive mutants 9–86E4 (ts-TgSfi1) and 11–31G12 (ts-TgMAPK-L1) were complemented using the ToxoSuperCos cosmid genomic library as previously described [21,23,26]. Mutant parasites were transfected with cosmid library DNA (50 μg DNA/5 x 107 parasites/transfection) in 20 independent electroporations. After two consecutive selections at 40°C, parasites were selected by the combination of high temperature and 1 μM pyrimethamine. Double-resistant (temperature and drug) populations were passed four times before genomic DNA was isolated for marker-rescue [21]. To identify the complementing locus in T. gondii chromosomes, rescued genomic inserts were sequenced using a T3 primer and the sequences mapped to the T. gondii genome (Toxodb). To resolve the contribution of individual genes in the recovered locus, we transformed the mutants with individual cosmids from a cosmid collection mapped to the T. gondii genome (toxomap.wustl.edu/cosmid.html). For direct complementation of the mutant 11–31G12 with DNA fragments, the TGGT1_312570 gene locus (including 550 bp 5′UTR and 367 bp of 3′UTR) was amplified from genomic DNA isolated from the parental strain RHΔhxgprt or the mutant (S2 Table). Specific cosmids or PCR fragments were transfected into 1 x 107 parasites using 6–10 μg of purified DNA. To quantify genetic rescue, established drug-resistant populations were tested for growth at the high temperature by standard plaque assay performed in triplicate [21,23,26]
To validate the cosmid genetic complementation we used next generation sequencing to verify the mutation in mutant 9–86E4 (ts-TgSfi1). Whole genome DNA libraries were prepared, sequenced, and analyzed for single nucleotide variation (SNV) according to published methods [68]. In brief, genomic DNA from the mutant and parent strains was fragmented and then, following end-repair, ligated with Illumina paired-end adaptors (Illumina, CA, US). Purified library fragments were enriched via PCR amplification using Illumina paired-end PCR primers (Illumina, CA, US), the fragments normalized to 2 nM and denatured using 0.1 N NaOH. Denatured libraries were cluster amplified on V2 flowcells using V2 chemistry according to manufacturer’s protocol (Illumina, CA, US). Flowcells were sequenced on Genome Analyzer II’s, using V3 Sequencing-by-Synthesis kits and analyzed with the Illumina’s v1.3.4 pipeline following manufacturer’s protocol (Illumina, CA, US). The resulting FASTQ sequence traces were aligned to T. gondii GT1 genome reference v7.3 and the Human genome reference build 37. MOSAIK was used to perform the alignments using the standard parameters described in the documentation V1.0 (available at https://github.com/wanpinglee/MOSAIK/wiki/QuickStart). SNVs were called using the SNV caller FreeBayes, using standard parameters as described in the documentation, software version 0.7.2. SNVs were filtered to remove SNVs whose calls had less than 5x coverage in the mutant and 3x in the parent, a p-value less than 0.8, and did not have a single allele that comprised 70% or more of the sequence reads.
Purified parasites were washed in PBS and collected by centrifugation. Total lysates were obtained by resuspending the parasite pellets with Leammli loading dye, heated at 95°C for 10 min, and briefly sonicated. After separation on the SDS-PAGE gels, proteins were transferred onto nitrocellulose membrane and probed with monoclonal antibodies against HA- (rat 3F10, Roche Applied Sciences), myc-epitope (mouse, Cell Signaling Technology), and α-Tubulin (mouse 12G10, kindly provided by Dr. Jacek Gaertig, University of Georgia, GA, US). After incubation with secondary HRP-conjugated anti-mouse or anti-rat antibodies, proteins were visualized by enhanced chemiluminescence detection (PerkinElmer).
Confluent HFF cultures on glass coverslips were infected with parasites for the indicated times. Infected monolayers were fixed, permeabilized, and incubated with antibody as previously described [26]. The following primary antibodies were used: mouse monoclonal αMyc (Santa Cruz Biotechnology, Santa Cruz, CA, US), αTgCenH3 [25], rat monoclonal αHA (Roche Applied Sciences), rabbit polyclonal αMyc (Cell Signaling Technology), αHuman Centrin 2 [23], αMORN1 (centrocone and basal complex stains, kindly provided by Dr. Marc-Jan Gubbels, Boston College, MA, US), and αIMC1 (parasite shape and internal daughter bud stains, kindly provided by Dr. Gary Ward, University of Vermont, VT, US). All Alexa-conjugated secondary antibodies (Molecular Probes, Life Technologies) were used at dilution 1:500. Coverslips were mounted with Aquamount (Thermo Scientific), dried overnight at 4°C, and viewed on Zeiss Axiovert Microscope equipped with 100x objective. Images were processed in Adobe Photoshop CS v4.0 using linear adjustment when needed. Super-Resolution images were acquired using the Zeiss ELYRA S1 (SR-SIM) microscope using a 63x lens. Images were collected and processed using Zeiss Zen software.
|
10.1371/journal.pgen.1004966 | Genomics of Divergence along a Continuum of Parapatric Population Differentiation | The patterns of genomic divergence during ecological speciation are shaped by a combination of evolutionary forces. Processes such as genetic drift, local reduction of gene flow around genes causing reproductive isolation, hitchhiking around selected variants, variation in recombination and mutation rates are all factors that can contribute to the heterogeneity of genomic divergence. On the basis of 60 fully sequenced three-spined stickleback genomes, we explore these different mechanisms explaining the heterogeneity of genomic divergence across five parapatric lake and river population pairs varying in their degree of genetic differentiation. We find that divergent regions of the genome are mostly specific for each population pair, while their size and abundance are not correlated with the extent of genome-wide population differentiation. In each pair-wise comparison, an analysis of allele frequency spectra reveals that 25–55% of the divergent regions are consistent with a local restriction of gene flow. Another large proportion of divergent regions (38–75%) appears to be mainly shaped by hitchhiking effects around positively selected variants. We provide empirical evidence that alternative mechanisms determining the evolution of genomic patterns of divergence are not mutually exclusive, but rather act in concert to shape the genome during population differentiation, a first necessary step towards ecological speciation.
| A variety of evolutionary forces influence the genomic landscape of divergence during ecological speciation. Here we characterize the evolution of genomic divergence patterns based on 60 fully sequenced three-spined stickleback genomes, contrasting lake and river populations that differ in parasite abundance. Our comparison of the size and abundance of divergent regions in the genomes across a continuum of population differentiation suggests that selection and the hitchhiking effect on neutral sites mainly contributes to the observed heterogeneous patterns of genomic divergence. Additional divergent regions of the genome can be explained by a local reduction of gene flow. Our description of genomic divergence patterns across a continuum of population differentiation combined with an analysis of molecular signatures of evolution highlights how adaptation shapes the differentiation of sticklebacks in freshwater habitats.
| During ecological speciation, divergence along the genome has been observed to be heterogeneous in numerous taxonomic groups [e.g., [1–4]]. Typically, the average genome-wide divergence is low, interspersed with regions of exceptional differentiation. However, studies describing divergence patterns across the genome have found regions of exceptional differentiation to be either numerous and small [4] or few and large [5, 6], the latter sometimes referred to as ‘genomic islands’. A variety of explanations have been proposed for the observed heterogeneity in genomic divergence, including stochastic processes such as genetic drift, but also deterministic mechanisms such as locus-specific reduction of gene flow in the vicinity of genes causing reproductive isolation, hitchhiking around selected variants, or variation in recombination and mutation rates [7]. Generally, genetic drift, population expansion, migration, and other demographic events affect the whole genome, whereas natural selection modified by local environmental differences impact only those regions of the genome that affect the respective phenotypes and fitness.
It is not known whether or not genomic patterns such as the variation of divergence and recombination along the genome tend to follow a predictable evolutionary trajectory as populations proceed along a speciation continuum [7]. We investigated the early phase of divergence using lake-river stickleback population pairs varying in their degree of genetic differentiation. If divergence patterns are driven by locus-specific effects of gene flow and divergent selection, the extent of divergence is expected to be more localized than widespread, in line with the “island view” [6]. These regions might hold “speciation genes” maintaining reproductive isolation between species including genes underlying a fitness reduction in hybrids [8]. Furthermore, “divergence hitchhiking”, the accumulative effect of selectively advantageous loci, predicts a positive correlation between genomic divergence and island size progression [9]. An alternative explanation posits that the lack of differentiation across most of the genome is due to shared ancestral polymorphism rather than ongoing gene flow [10, 11], whereas regions of high differentiation represent regions influenced by selection at linked sites [12]. Such a hitchhiking pattern may be caused by both advantageous (positive selection) and deleterious alleles (background selection). Therefore, if adaptation alone (assuming some degree of geographic separation) shapes the genomic landscape, population genetic processes unrelated to the extent of overall genomic differentiation govern divergence patterns. Disentangling such alternative scenarios is a crucial yet challenging step in understanding the genomics of divergence, especially in parapatry where the current and historic extent of migration and gene flow contribute to the overall genomic patterns.
We tested predictions inherent to the different scenarios explaining genomic patterns of divergence using whole-genome sequencing data of replicated population pairs of three-spined sticklebacks varying in their degree of genetic differentiation. Five population pairs were sampled from connected lakes and rivers from the United States (Us), Canada (Ca), Norway (No), and from two sites in Germany (G1 and G2; Fig. 1 and S1 Table). As ice sheets covered these regions during the last glaciation, these populations represent recent colonization events (~12 000 years ago). Both lake and river populations are derived from marine ancestors that became landlocked during de-glaciation, and in which ecotype differentiation between watersheds has occurred repeatedly. Some phenotypic traits such as feeding morphology [13], brain development [7], and parasite resistance [14] seemingly differentiated in parallel with habitat (i.e. lake and river) suggestive of local adaptation. Furthermore, experimental studies have shown evidence for local adaptation to lake and river habitats mediated by parasites [15]. Hence, contrasting the differentiation between populations from distinct ecosystems permits us to study the onset of divergence, which might eventually lead to complete reproductive isolation (i.e. speciation). Here, we scan genomic divergence patterns and evaluate differences and commonalities across a wide geographic sampling of parapatric population pairs to uncover the relative importance and interaction of evolutionary factors like drift, selection, and recombination during adaptive divergence.
One consistent difference between lake and river habitats is that lake fish posses a higher parasites diversity than parapatric river fish. From previous work on three-spined sticklebacks, lakes and rivers in Northern Germany are known to harbour distinct parasite communities [14, 16]. Despite the relatively low sample size for individual locations in this study (n = 12–17), this ecological difference between lakes and rivers is here confirmed on a broader geographic scale (Fig. 1). From each of the ten sampled populations, six stickleback genomes were sequenced using a combination of paired-end and mate-pair libraries on the Illumina HiSeq platform to an average genomic coverage of 26-fold (S2 Table). Instead of sequencing many individuals with low coverage, a small number of genomes per population was chosen to be sequenced to high coverage. This approach takes advantage of the greater resolution of single nucleotide polymorphisms (SNPs) and copy number variations (CNVs; evaluated in greater detail in a companion paper [17]) plus increased genotype accuracy within each individual to decipher the divergence mechanisms acting towards an apparent repeated differentiation between lake and river fish. Besides evaluating allele frequencies, the high individual sequence coverage permits us to infer haplotypes and examine recombination patterns. After stringent quality filtering, we accessed 297,437,667 bp from the 20 autosomes (380,547,835 bp). SNP density varied from 3 to 10 SNPs per kilobase (kb) within each population (S3–S4 Tables). For each of the five parapatric comparisons, pairwise genome-wide averages of divergence (FST) ranged from 0.10 to 0.28, disclosing a varying degree of differentiation in the ascending order of Us, G2, No, G1, and Ca (Table 1). The parapatric pairs emerge as repeated independent differentiation events (neighbor joining tree, Fig. 1A) except for the German populations, despite belonging to different draining systems (North Sea versus Baltic Sea). Due to low land levels and historically varying water levels, water bodies and connections across Northern Germany have most likely fluctuated over time. Thus the two lake and river population pairs in Germany (G1 and G2) might have been originally connected. Because of this, G1 and G2 share some postglacial history, common ancestral variation, and divergence while currently the two water systems are physically separated. Specifically, studies on the German system have proposed parasite communities as a promising candidate mediating divergent selection, pointing out their role in local adaptation [15, 18]. As a further global perspective of this hypothesis, we find a signal of isolation-by-adaptation (partial mantel test: r = 0.622, P = 0.0007) shown by a significant association of genome-wide FST and parasite community (jaccard distance of parasite sums across individuals, counts were 4th square root transformed) while correcting for geographic distance (geodetic distance between GPS coordinates of each sampling location). As we detected isolation-by-adaptation at a spatial scale beyond which gene flow occurs, this signal might be most likely caused by a loose linkage between locally adapted loci and the genome-wide neutral regions [19]. This result suggests a role of parasites for the local adaptation of freshwater stickleback populations.
Spatial heterogeneity along the genome was analyzed between parapatric populations by applying a genome scan approach, which averaged genetic divergence (FST) in 10 kb and 100 kb non-overlapping windows across the 20 autosomes (Fig. 2). The shape of the distribution of FST values across the genome qualitatively matches a skewed Poisson distribution, suggestive of divergence with gene flow (S1 Fig.) [9]. The pronounced right tail of the distributions aided the identification of outlier windows, which are significantly different from the genome-wide average. Outlier windows were detected for each population pair as the top 1% of the empirical distribution in addition to being significantly differentiated compared to a random permutation of markers across windows, applying a false discovery rate (FDR) of 0.01. Using the exact same approach comparing marine and freshwater populations, regions known to be under strong divergent selection such as Eda and Atp1a1 were detected as outliers demonstrating the robustness and reliability of the applied methods (details see Methods). Across all five parapatric lake-river comparisons, we identified a total of 1,530 extreme 10 kb outlier windows, in which 47 are shared between at least two of the five population pairs, a proportion that is slightly more than expected by chance (10,000 permutations of random sampling gave on average 28 overlaps, one-tailed P = 0.0006), but none of the windows were shared across all five population pairs. Although we found a weak positive correlation of FST along the genome between the five lake and river ecotype pairs (Fig. 2 and S2 Fig.), there is a negative correlation of FST among the 1,530 outlier windows (Pearson correlation ranging from r = -0.2531 to -0.1064, all P<10-4). These results indicate that outlier windows in one population pair are often windows of low FST in the other population pairs. Hence, outlier windows are not the same across the different population pairs. Annotations for all genes overlapping common outlier windows can be found in S5 Table. None of these outlier windows overlapped with those detected in a previous lake and river comparison of different stickleback populations on the Haida Gwaii archipelago [20]. Thus outliers of exceptional differentiation appear to be locally specific for lake-river ecotypes on a wide geographic scale as well as on a narrow scale [20, 21]. This is in contrast to earlier comparisons between marine and freshwater sticklebacks where few loci are repeatedly found under divergent selection on a global scale [22, 23]. Our results are in line with the notion that the repeated differentiation between derived freshwater stickleback populations occurs as a response to different ecological pressures specific to their local environment [24]. This might reflect locally specific parasite communities, aside from the general trend of an increase in parasite diversity in lakes compared to rivers. However, genomic diversification seem to be an inevitable consequence following the dispersal across habitats, reinforcing the concept that local adaptation is a major contributor to the evolution of species.
In order to further understand processes shaping the heterogeneity of genomic divergence, we evaluated if divergence is widespread or localized along the genome. Divergence hitchhiking predicts a trend towards an increase in size of divergent regions with overall population differentiation [8, 19]. Conversely, if size was largely determined by the strength and duration of selection, the size of divergent regions would be independent of overall population differentiation. To test these predictions in our dataset, we exploited our comprehensive sequencing resolution to identify precise borders and dimensions of regions of exceptional differentiation. Amongst the 1,530 outlier windows, adjacent outlier windows were combined into 794 continuous outlier “regions” of exceptional differentiation estimated to the nearest 1 kb (S6 Table). The size of a region of exceptional differentiation was determined utilizing barrier strength (b, ref [25]) to contrast local divergence to the genome-wide average. We found a high degree of size heterogeneity among divergent regions within and across population pairs, with no evidence that the size of these regions increases with higher levels of genome-wide differentiation (Table 1, S3 Fig.). This also holds true when recombination rates are taken into account (see below). Therefore, the genomic pattern of divergence observed across a continuum of population differentiation suggests that selection at linked sites drives the observed pattern rather than the interplay of gene flow and divergent selection, consistent with the perspective of geographically specific local adaptation. However, additional factors such as soft sweeps resulting from adaptation based on standing genetic variation might also contribute to the observed patterns, further complicating interpretations.
To further explore if the observed divergence patterns are indeed facilitated by selection and not induced by drift alone, we investigated fine-scale linkage patterns and their effects on genomic heterogeneity across a populations. For each population, we estimated the realized population-scaled recombination rates (ρ/Θ) along the genome. Both a local reduction of gene flow mediated by divergent selection and selection with the hitchhiking of linked neutral sites are predicted to produce a negative correlation between FST and recombination rate [12, 26], however this association would be unlikely mediated by drift alone. In addition, divergence hitchhiking predicts that over time, linkage will extend along the genome and eventually encompass large tracts of the genome [27]. In our study, realized recombination rates in regions of exceptional differentiation were often significantly reduced compared to genome-wide estimates (Fig. 3). We found that genome-wide recombination rates tended to decrease with increasing overall differentiation (Fig. 3). However, realized recombination rates in divergent regions are not significantly correlated with genome-wide differentiation, adding to the growing lack of empirical evidence for divergence hitchhiking [28]. These results suggest that either actual recombination rates coincide regions of the genome, which become divergent, or selection drives local reductions in realized recombination rates. The coalescent-based population recombination rates (4Ner) estimated in this study are simultaneously affected by the variation in genomic structure within and across populations, which may influence actual recombination rates, as well as by selection. Hence, selection might have locally reduced realized recombination rates in certain genomic regions or actual recombination has been reduced due to the intrinsic genomic structural variations thereby promoting genomic divergence. Previous studies evaluating large-scale map-based recombination patterns in sticklebacks have also found a correlation between recombination and divergence, suggesting that genome structure, via its influence on recombination, is important in understanding patterns of genomic differentiation [29, 30]. Here, the low correlation in divergence (FST) between different population pairs (Fig. 2) suggests that local factors specific to each population pair drive genomic differentiation, and that population specific selection reduces realized recombination, particularly if genomic structure is conserved across populations. However, it is possible that genome structure is not so strongly conserved across these geographically distant pairs. Structural variations such as inversions and CNVs have been shown to be abundant within stickleback populations [31]. A companion paper [17] highlights the prevalence of CNVs among and between the populations studied here, in which CNVs tend to also be population specific. These findings indicate that genome structure might be more variable than expected, and therefore might hold potential for promoting genomic differentiation in a population specific manner. We cannot here distinguish between selection-induced influences on realized recombination rates, and actual variation in recombination rates due to differences in genome structure and resultant effects on patterns of genomic differentiation. Further understanding of genome structure’s influence on recombination rates, and its variability within and across populations, will be crucial for disentangling the combined influences of selection and recombination on patterns of genomic variation.
Relative divergence (FST) in regions with low levels of recombination might be misleadingly interpreted as conclusive evidence for a local reduction of gene flow. For this reason, measurements of absolute divergence such as Dxy have been suggested as a complement to more reliably identified regions of locally reduced gene flow [10, 12, 32]. However, absolute divergence measurements are unreliable statistics for nascent populations and in non-equilibrium situations during population differentiation. Hence, we aim to disentangle different mechanisms shaping regions of exceptional differentiation by assessing selective sweep signatures in one or both populations of each parapatric pair. Utilizing the base pair resolution of our whole genome sequence data, we evaluated allele frequency spectra to differentiate between molecular signatures of selection among individual regions of exceptional differentiation. In divergent regions differentiated due to a local restriction of gene flow mediated by selection, the spectrum is not expected to be affected locally and should reveal a signature of neutral evolution [12]. The opposite is true for regions resulting from selection with hitchhiking at linked sites, which causes a characteristic skew of the spectrum. An excess of rare alleles is expected in a population experiencing a selective sweep [33], or in both populations in the case of background selection [34]. Distortions in the allele frequency spectrum were calculated for each population as Tajima’s D (TD) across the genome in 100 kb windows and in each region of exceptional differentiation. Genome-wide averages of TD varied from 0.0385 to 0.5936 suggesting predominantly neutral evolution across the genome with no indication for an excess of low frequency polymorphism in any of the populations. TD values within regions of exceptional differentiation were shifted towards negative values except for the Alaskan river (Us_R, Fig. 4A). These negative shifts of TD are consistent with selection as a major mechanism responsible for localized divergent regions along the genome.
In order to quantify the relative contribution of different mechanisms shaping the genomics of speciation, we partitioned individual regions of exceptional differentiation into four mutually exclusive categories with different molecular signatures of evolution based on contrasting local TD values to the genome-wide average (Table 1 and Fig. 4B–F). The minority of divergent regions is consistent with background selection (12%, TD reduced in both populations, Fig. 4B), whereas adaptation seems to shape most of the divergent regions (48%), consistent with the influential role of selection. Divergent regions with signals of positive selection (TD reduced in one of the two populations) should harbor those genes responsible for local adaptation. Genes in divergent regions with a signature of positive selection in lakes (Fig. 4C) were overrepresented with functions involved in structural molecule activity (18 out of 260 annotated genes, P = 0.0018), while genes in divergent regions with signals of positive selection in rivers (Fig. 4D) were overrepresented with functions involved in G-protein coupled receptor activity (15 out of 105, P = 0.0038), antiporter activity (6 out of 36, P = 0.0280), and drug transmembrane transporter activity (4 out of 8, P = 0.0367), suggesting functions in environmental response. Divergent regions with neutral TD patterns (TD in both populations similar to genome-wide average, Fig. 4E) potentially harbor genes restricting gene flow. Despite the prominent occurrence of neutral TD patterns among divergent regions (35%), we found no functional overrepresentation of genes within those regions (S6 Table). This indicates that a variety of different genes and functions might be involved in reproductive isolation, but the current state of gene annotations does not allow drawing compelling conclusions. Overall, the variety of molecular signatures of selection found in divergent regions suggests that different evolutionary processes shape regions of exceptional differentiation. We acknowledge that our approach of strictly categorizing regions based on thresholds simplifies a complex situation, in which various factors most likely interact to shape genomic divergence. However, our analysis suggests that different processes have different impacts across the genome, with selection being a probably major contributor. Therefore, the effects of a local reduction of gene flow and local adaptation are mutually compatible and probably act in concert to shape the genomic landscape of divergence between differentiating parapatric stickleback populations.
We presented multiple lines of evidence for the role of adaptation shaping the genomic divergence patterns between lake-river populations of three-spined sticklebacks. Aside from adaptive processes, stochastic variation in coalescent times and variable mutation rates could further contribute to the observed heterogeneity of genomic divergence [35]. In particular, demographic history such as colonization events (population range expansions) might lead to a substantial variation in allele frequencies across the genome, possibly mimicking the patterns of adaptive hitchhiking [36]. Here, we have chosen the genome-wide average as proxy of the underlying demographic history and the effect of random drift on these populations, as detailed demographic information is scarce. Today, fish migration from the sampled rivers flowing into lake habitats is possible while migration in the opposite direction is likely constrained by physical barriers (S1 Table). However, as freshwater systems have been subject to recurrent water-level changes during de-glaciation, the spatial context at different stages of population divergence might have fluctuated over the years affecting demographic history of the populations. Due to pronounced local differences and variable genomic patterns across the sampled continuum of genetic population differentiation we conclude that the main mode of contemporary divergence between parapatric three-spined sticklebacks is associated with population-specific local adaptation. This is potentially partially mediated by differences in the parasite, as we also found a corresponding signature of isolation by adaptation. Furthermore, our fine-scale examinations of molecular evolution suggest that some heterogeneity of genomic divergence is also the result of locus-specific differences in gene flow mediated by divergent selection. Our study has taken an important step towards deciphering the underlying mechanisms responsible for the genomic patterns during speciation, one of the fundamental enigmas in evolutionary biology.
Three-spined stickleback fish were caught from five pairs of lakes and rivers in North America and Northern Europe (S1 Table and Fig. 1). Between 12 and 17 fish were screened for macroparasites following established procedures [14]. Both Shannon diversity indices for each population and jaccard distance between populations were estimated on the basis of 4th square root transformed parasite counts. Muscle tissue from six sampled individuals from each location was used for DNA extraction (using a Qiagen DNA Midi Kit following the manufacturer’s protocol for high molecular weight DNA) and Illumina sequencing following previous methods [31]. To capture natural variation present in the wild, we randomly picked individual fish for sequencing (albeit targeting equal sex ratio per population and similar fish sizes across populations), thus without pre-selection of any particular morphological or parasitological characteristics. For each individual, two paired-end libraries (100bp reads, average insert size of 140bp and 300bp) and a mate-pair library (50bp reads, average insert gap of 3kb) were produced, achieving an average depth of coverage of 26x (S2 Table). Data is deposited in the European Nucleotide Archive (PRJEB5198). Raw sequence data was processed and filtered following previous procedures [31] and mapped against the three-spined stickleback reference genome [22] from Ensembl version 68 [37] with BWA (Burrows-Wheeler Aligner) software [38].
Mapped reads were further filtered and processed utilizing the Picard toolkit following previous procedures [31]. SNPs and indels were called with GATKv1.6 [39, 40] using concordant SNP calls from SAMtools v0.1.18 [41] for variant recalibration. Phasing and imputation was performed with BEAGLE v3.1 [42]. VCFtools [43] was utilized for processing genotypes. Positions overlapping with ‘N’s and repeat-masked regions from the Ensembl annotations (version 68) were removed from the final genotype file. Furthermore, variants within 10bp of an indel or indicating copy number variation were also excluded. Copy number variable (CNV) regions were identified by deviations in expected read depth with the software CNVnator [44]. More details on the CNV analysis are given in a companion paper submitted by Chain et al. The following analyses were performed on the 20 autosomes, spanning 380,547,835 sites in the reference genome. After removing masked sites and CNV region and imputing genotypes across 60 individuals, 297,437,667 sites were reliably genotyped and used for estimating population genetics parameters.
We used Illumina’s Golden Gate platform for cross checking genotypes from SNP sites distributed across the genome. Each chromosome held on average 9 (range 2–21) markers and the total of 183 loci were mostly interspersed by at least 50 kb. We found a high overall concordance (98% in 12,041 comparable sites) between genotype calls from the Golden Gate assay and our sequencing pipeline.
The population genetics estimators of nucleotide diversity (π and Θ) and Tajima’s D (TD) were calculated with VCFtools v0.1.11 [43] for each of the 10 populations (S3 Table), in addition to the relative divergence (Weir and Cockerham FST) and absolute divergence (Dxy [45]) estimated for each of the 5 parapatric lake-river pairs (S4 Table). Numbers of polymorphic sites per population and per population pair are reported in S3–S4 Tables. To illustrate the relationship amongst all sampled populations, we utilized a set of 1,074,467 intergenic autosomal polymorphic loci to estimate pairwise divergence (Weir and Cockerham FST) and built a neighbor joining tree. To gain support for the tree topology we randomly down sampled this dataset 100 times to 100,000 loci. For the genome scan, FST was calculated on the full dataset that was further filtered for minor allele frequencies below 25% across each pairwise comparison excluding uninformative polymorphism [46]. This way we evaluated the divergence between parapatric population pairs on the basis of 691,957 to 1,227,732 sites across the 20 autosomes. Population genetics estimators were averaged across the genome (20 autosomes) in non-overlapping windows to ensure statistical independence of windows. We used window sizes of 10 kb and 100 kb and confirmed that results are qualitatively the same. Diversity estimates have been corrected for the number of sites for which genotypes are available.
Outlier windows were determined by combining an empirical approach with a permutation approach. First, windows above the top 1% of the empirical distribution were identified as putative outlier windows. Second, we applied a permutation approach in which loci across the genome were permuted 1,000,000 times and window estimates of FST were tested against permutations holding the same amount of variable sites. Putative outlier windows from this permutation approach were identified after adjusting for a FDR of 0.01. Our final set of outlier windows consisted of those windows that were significant outliers in both approaches. All statistical procedures and visualizations were implemented in R [47]. Outlier window positions were compared across the five replicated lake-river comparisons. To evaluate how many overlapping outlier windows were expected by chance, windows were permutated 10,000 times utilizing bedtools [48].
To approximate the size of regions of exceptional differentiation more in detail, adjacent outlier windows were combined to form larger contiguous divergent regions of extreme differentiation. In each resulting candidate region, the locus of maximal divergence was determined as a starting point, in which outward steps of 1 kb windows were binned to estimate barrier strength (b, ref [25]). Margins of divergent regions showing extreme differentiation were determined when b dropped below 1 (genome-wide average) in two consecutive 1 kb bins. This resulted in divergent regions of exceptional differentiation with distinct sizes estimated to the nearest 1 kb. Divergent regions with sequence coverage (sequence information accessible, see details above) spanning less than 50% of their length were excluded from subsequent analyses. Average sizes of about 50 kb are independent of the initial window size used but specific values reported here are based on the 10 kb window size approach (Table 1).
We acknowledge that estimates of FST based on allele frequencies can vary depending on samples size [49]. To reduce variation of estimates between populations we kept the samples size constant at 12 alleles per populations. Additionally, our analysis did not rely on per site estimates but instead on averages of FST over larger regions (see above). We evaluated the effect of sample size on our power to describe genomic patterns, detect outlier windows, and define divergent regions in the three following ways. (i) We tested the accuracy of our FST estimates at individual loci by comparing them to estimates based on a larger sample size. The 183 loci used for validating the genotypes (see above) were also used to genotype a larger population sample (n = 26–59 per population) to validate allele frequencies and resulting FST estimates. For all population pairs, the FST estimates based on the sequencing approach with 6 individuals per population (12 alleles) had a significant positive correlation with the FST estimates from the Golden Gate assay using at least 26 individuals (Pearson correlation, r = 0.85, P< 10-16, df = 241, S4 Fig.). (ii) We tested the consistency of window FST estimates across the whole range of potential FST values by jack-knifing samples (S5 Fig.). On average, jack-knifed values (comparing 10 alleles per population) had 95% confidence intervals of 0.039 up to a maximum of 0.175. Windows with high FST values (>0.75) had even narrower confidence intervals (average of 0.027 and maximum of 0.088). These results support the notion that pronounced differences (“near-” and “post-fixation”) can be more reliably detected using our sample sizes than more settled differences (“pre-fixation” regime). (iii) We tested our ability to detect known candidate genes, which highly differentiate between marine and freshwater populations. For this we utilized previous sequencing data available for a marine population from Denmark [31]. Our genome scan based on FST estimates averaged across 10 kb windows reliably detected windows overlapping ATP1a1 [50], a well known candidate gene for physiological adaptation to osmotic differences on linkage group I, in all 6 pairwise European marine-freshwater comparisons (S6 Fig.). Eda, the major gene (linkage group IV) underlying the reduction of lateral plate number frequently observed in freshwater populations [51], was detected in 5 out of the 6 pairwise European marine-freshwater comparisons (S7 Fig.). As expected G1_L, a lake population showing phenotypic variation at this trait did not show significant differentiation in the Eda region, in which two of the six sequenced individuals were fully plated and carried the same haplotype as the fully plated marine fish. This is in line with a simulation demonstrating that sampling 12 haplotypes yields between 67–95% power compared to a gold standard, while notably, sampling fewer individuals has the greatest impact in the “pre-fixation” regime (a beneficial allele is starting to rise in one population) compared to “near-fixation” and “post-fixation” regimes (a beneficial allele is nearly or completely fixed in one population) [52].
To assess the molecular signature of selection in regions of exceptional differentiation, shifts in the allele frequency spectrum were evaluated utilizing TD. TD in these regions was compared to the genome-wide average of each respective population. A 5% threshold was applied to classify divergent regions into four mutually exclusive categories: background selection if TD dropped below the threshold in both parapatric populations, adaptation in lake or river if TD dropped below the threshold only in the respective population, and reduced gene flow if TD appeared neutral (not below the threshold). Comparing the utilization of population specific thresholds for each pairwise comparison with the utilization of the same overall averaged threshold for all populations resulted in minor differences in absolute numbers of regions in different categories. Furthermore, these differences did not affect qualitative changes with respect to the functional annotation of different categories, nor the proportion of different categories across the five parapatric population pairs.
Direct measures of fine-scale population recombination rates (ρ = 4Ner) were obtained with LDhat [53, 54] from patterns of genetic variation for each population separately. We filtered highly localized breakdowns of linkage disequilibrium (values of ρ above 100 between adjacent SNPs), as those are most likely artifacts, possibly due to local misassembly of the reference genome or clusters of erroneous SNPs [55]. Resulting recombination rate estimates were averaged over each 10 kb window and over each divergent region with exceptional differentiation, and corrected by the population specific mutation rate (Θ = 4Neµ) estimated as an average across all autosomes.
Regions overlapping with gene annotations from version 68 of Ensembl were identified using intersectBed of bedtools [48]. Annotations for shared outlier windows and divergent regions are reported in S5–S6 Tables. To determine enrichment of functional classes of genes among regions, topGO [56] was used with a universe of autosomal genes, and significance was determined at the 0.05 level using FDR adjusted p values to correct for multiple testing.
This study was performed according to the requirements of the German Protection of Animals Act (Tierschutzgesetz) and was approved by the ‘Ministry of Energy, Agriculture, the Environment and Rural Areas’ of the state of Schleswig-Holstein, Germany (reference number: V 312–72241.123–34). Wild sticklebacks were caught using minnow traps or hand nets. Before dissection, the fish were anesthetized with MS222 and sacrificed by an incision into the brain followed by immediate decapitation, and every effort was made to minimize suffering. No further animal ethics committee approval was needed. The species used in this study are not endangered or protected in any of the populations studied.
|
10.1371/journal.ppat.1005533 | Structure-Guided Mutations in the Terminal Organelle Protein MG491 Cause Major Motility and Morphologic Alterations on Mycoplasma genitalium | The emergent human pathogen Mycoplasma genitalium, with one of the smallest genomes among cells capable of growing in axenic cultures, presents a flask-shaped morphology due to a protrusion of the cell membrane, known as the terminal organelle, that is involved in cell adhesion and motility and is an important virulence factor of this microorganism. The terminal organelle is supported by a cytoskeleton complex of about 300 nm in length that includes three substructures: the terminal button, the rod and the wheel complex. The crystal structure of the MG491 protein, a proposed component of the wheel complex, has been determined at ~3 Å resolution. MG491 subunits are composed of a 60-residue N-terminus, a central three-helix-bundle spanning about 150 residues and a C-terminal region that appears to be quite flexible and contains the region that interacts with MG200, another key protein of the terminal organelle. The MG491 molecule is a tetramer presenting a unique organization as a dimer of asymmetric pairs of subunits. The asymmetric arrangement results in two very different intersubunit interfaces between the central three-helix-bundle domains, which correlates with the formation of only ~50% of the intersubunit disulfide bridges of the single cysteine residue found in MG491 (Cys87). Moreover, M. genitalium cells with a point mutation in the MG491 gene causing the change of Cys87 to Ser present a drastic reduction in motility (as determined by microcinematography) and important alterations in morphology (as determined by electron microscopy), while preserving normal levels of the terminal organelle proteins. Other variants of MG491, designed also according to the structural information, altered significantly the motility and/or the cell morphology. Together, these results indicate that MG491 plays a key role in the functioning, organization and stabilization of the terminal organelle.
| Mycoplasma genitalium is one of the smallest bacteria known and a common human pathogen. M. genitalium cells present a flask-shaped morphology due to the presence of a characteristic protrusion, known as the terminal organelle, that has several key biological roles. The terminal organelle allows mycoplasmas to move on solid surfaces by a distinctive type of cellular motility that is also related to pathogenicity. In the present study we have determined the first crystal structure of a terminal organelle protein (MG491). Using the structural information, we have designed and prepared several variants of this protein. M. genitalium cells expressing the variant proteins showed striking differences with respect to the unmodified mycoplasma cells. In one of the variants, we replaced a cysteine residue by a serine, which implies the exchange of just one sulfur atom by oxygen, resulting in a drastic reduction of motility and important alterations in cell morphology. Other variants changed the speed or the frequency of the movements. These results demonstrate that MG491 plays a key role in the functioning, organization and stabilization of the terminal organelle and are also a clear example of the interplay between atomic resolution details and the highest levels of cellular organization.
| Mycoplasmas are microorganisms belonging to the class of Mollicutes (‘soft skin’) that evolved from Gram-positive bacteria by genome reduction and are characterized by the absence of a cell wall, by their small cell sizes and by their reduced biosynthetic machinery. Consequently, these microorganisms live in nature as obligate parasites depending on the uptake of essential nutrients from their hosts. Mycoplasma species vary in form and many are able to move by gliding motility [1,2]. In particular, Mycoplasma genitalium, one of the smallest autoreplicative microorganisms known, is a motile species belonging to the pneumoniae cluster of mycoplasmas. Due to its small genome, of only ~480 protein-coding genes, M. genitalium has been used as a model of minimal cell [3,4] and is the subject of intense work in systems biology research [5–7]. M. genitalium is an emergent and prevalent sexually transmitted pathogen involved in urogenital infections in humans, including non-gonococcal and non-chlamydial urethritis and inflammatory reproductive tract diseases in women. A review article on this issue described the need for an early diagnostic of the infection, which increases the risk of HIV transmission when persistent [8,9]. In addition, an intense search for novel therapeutic agents against M. genitalium has been launched as several studies revealed the existence of isolates resistant to treatments with azithromycin [10–12], indicating that there is a continuous need to search for potential drug and vaccine targets in this microorganism [11,13].
M. genitalium has a flask-shaped morphology that consists of a cell body with a protrusion of the cell membrane, called the terminal organelle, which is the scaffold for cell adhesion, division and motility, processes deeply related to infectivity. The terminal organelle is supported by a complex cytoskeleton that is formed by three main substructures: the terminal button, the rod and the wheel complex located, respectively, at the tip, the center and the rear with respect to the cell body [2,14–16]. Moreover, adhesins P110 and P140, which are very abundant at the surface of the terminal organelle, are essential for attachment to host cells, together with the accessory proteins MG218, MG312, MG317 and MG491 [17–20] A model for gliding motility in mycoplasmas from the pneumoniae cluster proposed a cyclic process where the rod, anchored to the wheel complex, has a central role [14,15]. According to this model in a first step, the tip of the terminal organelle binds to the substrate with the rod fully extended and, in a second step, the rod contracts, dragging the cell forward. However, we have very recently demonstrated that this model is no longer valid since M. genitalium cells remain motile in the absence of the rod element [20]. The same work also highlighted the role of P110 and P140 adhesins and of the P32 protein on promoting cell movement as previously proposed [1,21]. The wheel complex might also be involved in chromosome segregation by attaching the mycoplasma chromosome to the terminal organelle [22]. M. genitalium MG200 and MG491 proteins have been proposed as components of the terminal organelle wheel complex ultrastructure [23,24] and in agreement with this, it has recently been found that MG200 and MG491 interact with each other specifically influencing cell motility [25].
In this work, the crystal structure of MG491 was determined and found to present a unique tetrameric organization as a dimer of asymmetric pairs of subunits. The structural information guided the design of MG491 variants, which presented striking alterations in cell motility and in cell morphology demonstrating the key role played by MG491 in the organization and functioning of the terminal organelle of M. genitalium.
Crystals of the same type were obtained from both the full length MG491 protein (residues 1 to 346) and from a construct of the protein N-terminal region, MG491-Nt (residues 1 to 308), though times required for crystallization changed from a few months to a few weeks, respectively. Initial phases were derived from Single-wavelength Anomalous Diffraction data [3] collected at the selenium absorption edge of a MG491-Nt variant where three isoleucine residues (Ile36, Ile168, Ile205) had been replaced by seleno methionines (see Material and Methods) (Fig 1A). Four selenium sites were located within the crystal asymmetric unit, with two sites related to the other two by a Non-Crystallographic Symmetry (NCS) two-fold axis. Structure determination was then achieved by density modification, averaging between both the less isomorphous crystals (see Materials and Methods) and using the NCS two-fold axis (S1 Fig). The final refined structure, with four subunits in the crystal asymmetric unit (residues 65–204, 67–203, 66–203 and 62–205 for subunits A, B, C and D, respectively; Figs 1B, 2A and 2B), has agreement Rwork and Rfree factors of 22.17% and 24.93%, for a seleno methionine MG491-Nt data set at 3.0 Å resolution (Table 1, S2 Fig, PDB entry code 4XNG). The unexpected presence of four subunits in the crystal asymmetric unit had two major implications: i) extensive proteolysis had to have happened in the C-terminal region of the protein (not visible in the determined structures) during crystallization. Four subunits each with 308 residues would give an unacceptably low crystal solvent content of 4%. ii) Four subunits cannot be symmetrically related by the only two-fold symmetry found.
The structure determined for MG491 subunits consists of an antiparallel three-helix-bundle, with helices α1 (residues 70–102), α2 (111–145) and α3 (167–203) connected by loops L1 (103–110) and L2 (146–166), respectively (Fig 1A and 1B). Helix α1 is kinked in its central part, due to the insertion of a π-helix turn starting in residue Cys87 (Fig 1B). Only a small and variable number of residues, from three in subunit B to eight in subunit D, could be traced preceding helix α1. Therefore, about sixty residues in the N-terminal region of MG491 appear to be flexible with respect to the subunit three-helix-bundle domain. Differences between the four subunits present an averaged root mean square deviation (r.m.s.d.) for Cα atoms of only 0.32 Å, which increases to 0.70 Å between subunits not related by the NCS two-fold axis, with the largest deviations corresponding to loop L1 and to the central part of loop L2 (residues 152–160).
The four MG491 subunits found in the crystal asymmetric unit present three different types of intersubunit interfaces that were named symmetric, tight and loose (Figs 2B and 3A–3C). The symmetric interface, with a total buried area of ~650 Å2, corresponds to interactions across the two-fold symmetry axis and involves only subunits A and C, while subunits B and D do not contact with each other (Fig 3A). The tight interfaces, with a total buried area of ~2*1200 Å2, correspond to the two interfaces between subunits in the pairs A/D and C/B (within each pair subunits are related by ~72° rotation) (Fig 3B). The loose interfaces, with a total buried area of ~2*450 Å2, correspond to the two interfaces between subunits in the pairs A/B and C/D (within each pair subunits are related by ~108° rotation) (Fig 3C). Therefore, the four subunits found in the crystal asymmetric unit present a network of (extensive) intersubunit interactions strongly suggesting that the MG491 molecule can form tetramers, in agreement with studies by gel filtration, crosslinking with glutaraldehyde and nano-ElectroSpray Ionization Mass Spectrometry (S3 Fig). Despite the fact that the four subunits in the tetramer are structurally similar, as reflected by the low r.m.s.d. values, they are placed in two different environments. Therefore, two types of subunits can be identified according to the residues that participate in the tight and loose interfaces of each subunit. The organization of the MG491 tetramer, with only a two-fold symmetry, can be defined as a dimer (C2 molecular symmetry) of asymmetric pairs of subunits. Quantification of the deviation from an accurate four-fold molecular symmetry gives an average for the Cα atoms of all the residues of 7.3 Å (Fig 3D). Attempts to form a regular (symmetric) oligomer using only the interactions corresponding to the tight interface would result in a helical aggregate with four subunits at most due to steric clashes (S4A Fig). In turn, oligomerization using only the loose interface would result in helical aggregates with three subunits at most (S4B Fig). Loop L2 mediates interactions in both the tight and the loose interfaces presenting a different conformation in each interface. In the loose interface the backbone in the central part of loop L2 moves towards the neighbor subunit by ~2 Å, with the side chain of Phe158 flipping to a different conformation about 7.5 Å away (Fig 4A and 4B). Interestingly, the conformational changes observed for loop L2 result in similar intersubunit interactions for the two interfaces when analyzed with LigPlot+ [28]. In particular, interaction of Phe157 with Gly91 in the tight interface is mirrored as interaction of Phe157 with Gly80 in the loose interface (S5 Fig).
In the MG491 tetramer, the four Cys87 residues (Cys87 is the only cysteine in the whole MG491 sequence) are located close to each other (Fig 4C), in particular across the symmetric interface, suggesting the possible formation of disulfide bonds between subunits. However, in the structure determined disulfide bonds are absent, which might be due to the presence of reducing agents required for crystallization. Following an established oxidation protocol to form disulfide bonds in vitro [29], the protein was diluted in 1x PBS (pH 7.5) to a final concentration of 1 mg/ml and incubated for 2–4 h in 1% (v/v) DMSO, resulting in the formation of intersubunit disulfide bridges but only between ~50% of the subunits (Fig 4C). This result supports a departure of symmetry in the molecular organization of MG491 that would agree with disulfide bridges being formed only between the two subunits at the symmetric interface of the MG491 tetramer.
The sequence alignment (http://espript.ibcp.fr) [30] between MG491 and its M. pneumoniae homolog, P41, gives an overall identity of 53% mainly due to the high identity found for the proteins N-terminal regions (until about residue 203 in MG491, Fig 1A). In particular, the most conserved regions are for MG491 helices α1 and α3 as well as for loops L1 and L2. Accordingly, the MG491 structure is expected to be well preserved in the M. pneumoniae protein P41.
To study the biological relevance of the unique molecular organization of MG491 and to further investigate the function of this protein, three structure-guided M. genitalium mutant strains were engineered with mutations in residues directly involved in the interactions between subunits. These three protein variants were: i) Cys87 replaced by a serine; ii) Phe157 and Phe158 substituted both by alanines and iii) the peptide from Asn155 to Lys160, corresponding to a large fragment of loop L2, deleted (Fig 1A). An additional mutant strain lacking the N-terminal region of MG491 (residues 1–61) was also engineered to gain insight into the function of this region. Mutant alleles mg491C87S, mg491F157A-F158A, mg491ΔloopL2 and mg491ΔNt were introduced in pMTncat plasmids [23] under the control of the MG438 promoter (Fig 5A). These mini-transposons were electroporated into cells from M. genitalium Δmg491 null mutant strain lacking MG_491 [20]. One colony from each transformation experiment was selected for the different alleles and named mg491-C87S, mg491-F157A-F158A, mg491-ΔloopL2 and mg491-ΔNt, respectively. Transposon insertion sites were investigated by direct genome sequencing and all the selected transformants showed transposon insertion sites in genes other than those involved in the terminal organelle architecture and/or gliding motility functioning (Table 2).
Upon introduction of the wild type allele in Δmg491 cells, steady-state levels of the MG491 protein were restored in the Δmg491-mg491cat strain (Fig 5B). Normal levels of MG491 were also observed in the mg491-F157A-F158A, mg491-ΔloopL2 and mg491-ΔNt mutant strains (Fig 5B). However, a lower amount of MG491-Cys87Ser was detected in Δmg491-C87S cells, suggesting that Cys87 might play an important role in protein stability. The apparent molecular weight of the deletion variant protein MG491ΔNt was ~40 kDa, in agreement with the expected value. Lower levels of MG491 have already been shown to correlate well with the existence of several downstream events in terminal organelle related proteins [20]. Therefore, it was not surprising to observe in Δmg491 cells, a drastic decrease in the amount of adhesion proteins P110 and P140, and of most of the cytadherence accessory proteins (Fig 6A and 6B). However, these cells exhibited normal amounts of the cytadherence accessory proteins MG200 and MG219. The adhesin and cytadherence accessory proteins levels were also restored upon reintroduction of the wild type MG_491 allele in the M. genitalium Δmg491 strain and a similar effect was observed in the transformants containing the mutant alleles mg491-C87S, mg491-F157A-F158A and mg491-ΔloopL2. In contrast, the levels of adhesins and cytadherence accessory proteins were not restored after the introduction of the mutant allele in mg491-ΔNt, indicating that the N-terminal region of MG491 has an important role in the formation and stabilization of the terminal organelle (Fig 6A and 6B).
Cells from the Δmg491 strain showed a filamentous morphology when observed by scanning electron microscopy (Fig 7A). The characteristic flask-shaped morphology typically observed in wild type cells was restored in the Δmg491-mg491cat strain (Fig 7B). The gliding properties were also restored in these cells, showing no significant differences when compared to those exhibited by G37 wild type cells (Table 3, S1 and S2 Movies, S6 Fig). Cells from the mg491-C87S strain showed normal terminal organelles (Fig 7C) but this strain also presented a high frequency of cells bearing multiple terminal organelles, which correlated with a reduced number of motile cells and a slower mean velocity as measured by time lapse microcinematography (Table 3, S3 Movie). When examining microcinematographies of G37 wild type cells, 18% of the motile cells show one or more resting periods. These resting periods are short and seem not to be related to cell division. Remarkably, the frequency of motile cells showing resting periods in mg491-C87S strain was 49%, indicating that the high frequency of non-motile cells might be a consequence of these resting periods. Likewise, a large amount of cells bearing multiple terminal organelles was also observed when examining the mg491-ΔloopL2 strain (Fig 7E) but these cells showed, in addition, a drastic decrease in different gliding motility parameters (Table 3, S4 Movie) and a low hemadsorption activity (S6 Fig). Moreover, both strains showed normal levels of all known proteins involved in gliding motility (Fig 6A and 6B) and exhibited no significant changes in the overall terminal organelle architecture (Fig 8). These data suggest that gliding motility impairments detected in these strains are a direct consequence of the mutations introduced in MG_491. In contrast, the gliding properties and the frequency and architecture of terminal organelles in mg491-F157A-F158A cells were similar to those of wild type cells (S5 Movie and Figs 7D and 8C). However, this variant shows a lower hemadsorption activity than G37 wild type (Table 3, S6 Fig) and a large amount of minute cells smaller than 0.35 μm in size as revealed by electron microscopy (Fig 7D and Table 3). Cells from this strain were stained with Hoechst 33342, examined by time lapse microcinematography and finally visualized by epifluorescence microscopy. Most of the minute cells analyzed (93.3%) showed no detectable fluorescence after staining with Hoechst indicating that these cells did not contain detectable amounts of DNA. Among these non-fluorescent cells, 53 of them (54.1%) were found motile during the examination period (S7 Fig), indicating that these minute cells were consequence of terminal organelle detachments. Such minute cells are rarely observed in M. genitalium G37 wild type strain. In contrast, cell detachments are frequently observed when the terminal organelle is not properly anchored to the cell body. Minute cells were previously described to be the result of terminal organelle detachments from the main cell body in M. genitalium cells lacking the C-terminal region of MG491 [22] and also in M. pneumoniae cells with a disrupted MPN311 gene, which codes for the P41 protein (S1 Table) [30]. Thus, the presence of minute cells in the mg491-F157A-F158A strain suggests that the intersubunits interactions promoted by Phe157 and Phe158 are required for the proper assembly of MG491, possibly playing an important role in the stabilization of the protein quaternary structure. However, oligomerization of protein variant Phe157Ala-Phe158Ala appears similar to the wild type protein presenting, surprisingly, even a slightly increased stability (S8 Fig). Finally, electron microscopy analysis of the mg491-ΔNt strain revealed the presence of a large amount of cells with filamentous morphology and the absence of rods inside these filaments (Fig 8E), similar to what was observed when examining the parental Δmg491 strain [20]. Moreover, no motile cells were observed for this strain (S6 Movie), suggesting that MG491 is involved in the assembly of the terminal organelle and in its stabilization through the protein N-terminal region.
Human pathogen M. genitalium, from the pneumoniae cluster of mycoplasmas, presents a flask-shaped morphology conferred by a polar structure, known as terminal organelle, which neither structurally nor functionally is yet well understood. MG491 protein from M. genitalium shares a high sequence identity with the M. pneumoniae protein P41, which is known to be an important component of the terminal organelle in M. pneumoniae and has been located at the base of the electron-dense core [31]. The location of MG491 in the terminal organelle of M. genitalium was also supported by the finding that a 25-residue region interacts specifically with MG200 [25], a protein that had been shown to be involved in gliding motility [23,24]. The structural characterization of MG491 in this work, indicates that MG491 subunits are composed of three distinct regions with a 60-residue N-terminus, a central three-helix-bundle spanning about 150 residues and a C-terminal region that contains the residues that interact with MG200 and appears to be mostly unstructured (Fig 1A). Only the central helix-bundle is well defined in the electron density maps of crystals from several constructs of MG491 (Fig 1B). All the solved crystal structures contain four crystallographically independent subunits, which are interwoven by a network of interactions with each other (Fig 2A and 2B). Surprisingly, each one of this tetrameric ensembles is organized with only one two-fold symmetry axis that relates pairs of subunits (Fig 3A). These pairs can be defined in two alternative ways referred as loose or tight according to the extension of the interacting interface between the two subunits in the pair (Fig 3B and 3C). Steric clashes make it impossible to model regular oligomers containing only one kind of these interacting surfaces (S4 Fig). The biological relevance of this unique organization has been assessed by the characterization of M. genitalium mutant strains with alterations in residues involved in intersubunit interactions. The MG491 variant Cys87Ser preserves normal levels of all the other terminal organelle proteins but presents a very significant reduction in motility (Table 3), comparable to the effects observed in deletion mutants of whole proteins involved in gliding motility [24,32]. Terminal organelle development is synchronized with cell division and cytokinesis appears to be highly coordinated with gliding motility, which is also essential for segregation of the terminal organelles to the opposite cell poles. In this way, alterations in motility often result in the presence of cells bearing multiple terminal organelles [32–34]. In contrast, cells from mg491-C87S mutant strain show only a very modest increase (7.6%) in the frequency of cells with multiple terminal organelles. Moreover, the gliding velocity of these cells is not significantly lower than that exhibited by wild type cells (Table 3). The large number of non-motile cells in the mg491-C87S mutant is strongly correlated with an increased frequency of cells showing resting periods, rather than with the presence of cells stalled in the cytokinesis process, as observed in other gliding mutants [24,32]. The higher frequency of resting periods in the mg491-C87S mutant strain suggests that the Cys87 residue of MG491 might have an important role in the regulation of gliding motility. Interestingly, the frequency of resting periods was also found increased in M. genitalium cells lacking the EAGR box from MG200 [16], reinforcing the relevance of the interplay between MG200 and MG491 in the regulation of gliding motility [22]. The strikingly severe effects of a single point mutation on the only cysteine residue suggests a major role for this cysteine that is likely related with the asymmetric formation of intersubunit disulfide bridges observed in vitro (Fig 4D). Only the two subunits in the tetramer of MG491 that interact across the molecular two-fold symmetry axis (subunits A and C in Figs 3 and 4) are expected to participate in this interaction, while the cysteine residues of the other two subunits would remain reduced or available for different interactions. MG491 variants designed to alter the tight and loose interfaces by deleting the central part of loop L2 (ΔloopL2) or replacing two of the loop residues (Phe157Ala-Phe158Ala) also resulted in significant changes in cell motility and cell morphology (Table 3 and Figs 7E and 7F and 8C and 8D). As expected, alterations in the deletion variant ΔloopL2, which also includes residues Phe157 and Phe158, are stronger than those observed for the Phe157Ala-Phe158Ala variant and, accordingly, the frequency of cells with multiple terminal organelles is higher in ΔloopL2 cells (Table 3). In contrast, the MG491 double mutation variant Phe157Ala-Phe158Ala showed a significant increase in the amount of minute cells or terminal organelles detached from the main cell body despite the fact that no clear changes were observed in vitro for the oligomerization of the variant (S8 Fig). The increased frequency of minute cells strongly supports that Phe157 and Phe158 residues have a main role in the stability of the wheel complex or in the interactions of the wheel complex with the rod.
Despite the complexity of the terminal organelle of mycoplasmas, here we show that this structure can be a reachable target for a thorough characterization zooming out in resolution from atomic to cellular levels. In this work, the structural information obtained from the crystal structure of MG491 has guided the preparation of several M. genitalium mutant strains of this protein. As a result, the motility and morphology of M. genitalium cells have been importantly affected, providing, for the first time, information on how the structure of a protein relates with the organization, stabilization and functioning of the terminal organelle. Motile mycoplasmas with spreading deficiencies are associated to a reduced infectivity [17,35], which emphasizes the relevance of MG491 in the virulence of M. genitalium.
The E. coli XL1-Blue strain was used to amplify the plasmids used in this study and was grown on LB agar plates or liquid LB media overnight. Ampicillin was added at 0.1 mg/ml. M. genitalium G37 wild type and mutant strains were grown in SP-4 broth at 37°C under 5% (v/v) CO2 in tissue culture flasks (from TPP, Switzerland) until mid-log phase of growth. Transformant colonies were isolated on SP-4 agar plates supplemented with 2 μg/ml tetracyclin and 34 μg/ml chloramphenicol.
The coding sequence of the MG_491 gene was amplified from M. genitalium G37 wild type genomic DNA with oligonucleotides 5MG491 and 3MG491 and ligated into a pBE plasmid [36]. The triplet coding for Trp232 from the MG491 protein was changed from TGA to TGG by amplification of this plasmid with oligonucleotides MutMG491PA and MutMG491PB and circularization of the amplicon with T4 DNA ligase. Afterwards, the sequence coding for this mutated version of the full length MG491 protein was cloned between NdeI and XhoI restriction sites of a pET21d expression vector (Novagen, Madison, WI, USA), which also codes for a C-terminus hexa-histidine tag. The resulting vector was transformed into E. coli BL21 (DE3) cells and the transformant cells were plated on LB/agar plates supplemented with ampicillin. After checking the correctness of the DNA sequence, the transformant cells were cultivated in 1 l LB medium containing 0.1 mg/ml ampicillin and induced overnight with 1 mM IPTG at 20°C with constant shaking after reaching an OD600 of ~0.6. Subsequently, the cells were harvested by centrifugation at 4500 xg for 15 min at 4°C. The pellet was resuspended in lysis buffer (0.02 M Tris-HCl (pH 8.0), 0.5 M NaCl, 0.02 M imidazole, complete EDTA free protease inhibitor (Roche Diagnostics, Mannheim, Germany)) and the cells disrupted by sonication. The total lysate was then centrifuged twice for 20 min at 45000 xg to remove cells debris and filtered through a 0.22 μm filter. The his-tagged MG_491 gene product present in the resulting supernatant was firstly purified through a 5 ml HisTrap HP column (GE Healthcare Life Sciences, Uppsala, Sweden) previously equilibrated in 0.05 M Tris-HCl (pH 8.0) buffer containing 0.5 M NaCl and 0.02 M imidazole, concentrated to a suitable volume and then loaded on a Superdex 200 16/60 gel filtration column (GE Healthcare Life Sciences, Uppsala, Sweden) equilibrated in 0.05 M Tris-HCl (pH 8.0) containing 0.15 M NaCl.
To obtain the phases for the X-ray structure determination several methionine residues (absent in the MG491 sequence) were introduced based on secondary structure element predictions, in positions corresponding to Ile36, Ile168, Ile205 and Ile313. A new expression vector was prepared (pET21d-MG491-B) using pET21d-MG491 as template and the oligonucleotide primers containing the appropriate target substitutions (see S1 Table). Limited proteolysis experiments performed with Trypsin on a MG491 sample generated a fragment of about 30–35 kDa with an intact N-terminal (revealed by Edman sequencing), which suggested that the C-terminal region is more accessible and thus more susceptible to proteolysis. Given this, and using the pET21dMG491-B vector as template and the appropriate primers (see S1 Table), a shorter variant of the protein was designed spanning MG491 residues 1 to 308 (MG491Δ308). The resulting PCR fragment was finally cloned into a pOPINE expression vector [37], which encodes for an extra lysine and a hexa-histidine tag at the C-terminal end of the construct. This new vector was then transformed into E. coli BL21 (DE3) cells and the MG491-Nt protein was expressed and purified following the same protocol used to prepare the full length protein. Additionally, the seleno methionine-labeled MG491-Nt protein was produced by growing a 0.1 l pre-culture overnight at 37°C in presence of 400 μl L-methionine at 10 mg/ml, 2 ml of 50% (w/v) glucose (freshly prepared and filtered through a 0.22 μm filter) and the appropriate antibiotic. Cells were then recovered by centrifugation at 4500 xg for 15 min, washed three times with 1x PBS, to remove the L-methionine that has not been incorporated by the cells, and finally resuspended in 2 ml 1x PBS. This cell pellet was then used to inoculate 1 l of SelenoMet media (Molecular Dimensions Ltd., Newmarket, UK) in presence of 9 ml L-seleno methionine at 10 mg/ml and supplemented with OnEx solutions 1, 2 and 3 from the Overnight Express Autoinduction Systems 1 (Novagen, Madison, WI, USA). Cells were grown for 6 h at 37°C, then the temperature was lowered to 25°C and growth was continued for 20 h with constant shaking before harvesting. The seleno methionine-labeled MG491-Nt protein was finally purified following the same protocol used for the full length MG491 and MG491-Nt proteins. Under these conditions, the proteins eluted as single peaks consistent with tetramers of ~200 kDa, respectively. The propensity of MG491 to form tetramers was also assessed and confirmed by crosslinking with glutaraldehyde [38] and by nano-ElectroSpray Ionization Mass Spectrometry (S3 Fig).
Crystals of the full length MG491, MG491-Nt and seleno methionine-labeled MG491-Nt (respectively at concentrations of 10 mg/ml, 8 mg/ml and 15 mg/ml), were grown at 20°C by the vapour-diffusion method over a reservoir containing 0.2 M lithium sulphate monohydrate, 25% (w/v) PEG 3350 and 0.1 M Bis-Tris (pH 6.5) or 0.1 M HEPES (pH 7.5) or 0.1 M Tris-HCl (pH 8.5). Before data collection crystals were transferred to a drop of reservoir solution containing 15% (v/v) propylene glycol as cryoprotectant and flash-cooled in liquid nitrogen. Crystals of MG491-Nt, soaked for 10 to 60 sec in a drop of mother liquor containing 12.5–100 mM of 5-amino-2,4,6-triiodoisophthalic acid (I3C, Sigma), were then rapidly back-soaked [39] in a drop of mother liquor containing 15% (v/v) propylene glycol as cryoprotectant and flash-cooled in liquid nitrogen.
X-ray diffraction data was collected at 100 K on beamlines ID23-1 [40] and ID29 [41] (ESRF, Grenoble, France) for crystals of the full length MG491 and seleno methionine-labeled MG491-Nt proteins and on beamline PROXIMA1 (SOLEIL, Gif-sur-Yvette, France) for crystals derivatized with the I3C compound. All beamlines used were equipped with PILATUS 6M-F detectors [42]. For an optimal measurement of the anomalous differences on the seleno methionine-labeled MG491-Nt crystals, a MiniKappa goniometer mounted on beamline ID29 (ESRF, Grenoble, France) was used to re-orient the investigated crystal before data collection, aligning a crystallographic axis along the rotation axis such that Bijvoet mates were on the same image [43]. Data were integrated with XDS [44,45], the output unmerged XDS ASCII file reflection.HKL was then converted to MTZ format by COMBAT and a list of free reflections generated (CCP4 Program Suite v6.4.0). The resulting reflection files were finally scaled with SCALA [46,47]. Phasing statistics for each data set containing anomalous differences were assessed with the processing software XDS, SCALA, XPREP (Bruker AXS Inc., Madison, Wisconsin, USA.) or SHELXC from the SHELX suite [48,49]. All crystals from the different protein constructs belonged to the orthorhombic space group P21212, with unit cell parameters in the range of a = 96–98 Å, b = 107–112 Å and c = 62–70 Å, indicating an important non-isomorphism not only between native and derivative crystals but also between different derivative crystals (Table 1 and S1 Fig). The HKL2Map GUI interface [50] was used to run the SHELX triad. Initial maps, obtained from the seleno methionine-labeled MG491-Nt data set with the highest anomalous signal, were improved by extensive density modification procedures including averaging between the less isomorphous crystals with programs DM and DMMULTI [51,52]. The command-line utility phenix.get_cc_mtz_mtz, from Phenix suite [53], which uses RESOLVE [54], was used to facilitate comparisons between density maps with origin shifts compatible with the space group symmetry. The model was completed and refined in rounds of manual rebuilding and restrained refinement with REFMAC [55], using TLS and isotropic B-factors only in the final stages of refinement. The quality of the final model was validated using MolProbity [56] and PROCHECK [57] (Table 1). Interacting surfaces were analyzed with Pymol (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC) and the electrostatic representation was generated with the APBS plug-in.
Deviations from perfect C2 and C4 cyclic symmetry were calculated for the Cα atoms as the interatomic distances differences (null when the symmetry is perfect) between pairs of subunits [58].
The pMTnMG491cat plasmid containing a mini-transposon bearing the coding sequence of MG_491 under the control of the MG438 promoter was amplified using the phosphorylated oligonucleotide P-C87SMG491/5 and the oligonucleotide C87SMG491/3. The resulting PCR fragment was circularized by ligation of the blunt ends to obtain the pMTnMG491C87S plasmid. Similarly, pMTnMG491cat plasmid was amplified using the oligonucleotide FFAAMG491/5 and the phosphorylated oligonucleotide P-loop/3. The amplicon was circularized by ligation to obtain the pMTnMG491FA plasmid. The pMTnMG491cat plasmid was also amplified using the oligonucleotide loop/5 and the phosphorylated oligonucleotide loop/3. The PCR product was circularized by ligation to obtain pMTnMG491loop plasmid. Finally, the 855 bp 3’ coding sequence of MG_491 was amplified using oligonucleotides MG491pr438ct/5 and MG491/3. The PCR fragment was excised with ApaI and XhoI restriction enzymes and ligated into a pMTncat plasmid [23] to obtain the pMTnMG491ΔNt plasmid. The four constructed plasmids were electroporated into Δmg491 cells and the transformants were isolated in SP-4 agar plates supplemented with tetracyclin and chloramphenicol. Transposon insertions were considered to disrupt a gene sequence when they fell within the 5'-most 80% of the ORF and were located after at least three codons from the start of the protein-coding region [3].
Total protein extracts of mycoplasma strains were electrophoresed in standard SDS-PAGE gels and stained with Coomassie Brilliant Blue or transferred electrophoretically to PVDF membranes following standard procedures [59]. PVDF membranes were probed with anti-MG217 at 1:500 dilution [60], anti-HMW3 at 1:5 000 dilution [61], anti-P41 at 1:1 000 dilution [62], anti-P32 at 1:2 000 dilution, anti-MG200 at 1:5 000 dilution [24] and anti-MG219 at 1:1 000 dilution.
The hemadsorption activity of M. genitalium G37 wild type and MG491 mutant strains were quantitatively determined by flow cytometry as previously described [63] using a FACSCalibur (Becton Dickinson). The fraction of non-attached mycoplasma cells was plotted vs the concentration of red blood cells and fitted to inverse Langmuir Isotherm curves by iteration using the KaleidaGraph software (Synergy). The Wald test was used to find statistically significant differences in the dissociation constant (KD) of the different strains with the G37 wild type strain.
Samples of mid-log phase cultures of G37 wild type strain and Δmg491-mg491cat, Δmg491-mg491C87S, Δmg491-mg491F157A-F158A and Δmg491-mg491loopL2 were 200x diluted and grown overnight on 8-well μ-slides ibiTreat (IBIDI). A Δmg491-mg491ΔNt undiluted sample was also grown overnight on 8-well μ-slides ibiTreat. Culture medium was replaced with fresh pre-warmed SP-4 before observations. Cell motility was examined at 37°C and 5% (v/v) CO2 using a Nikon Eclipse TE 2000-E inverted microscope. Images were captured at 2 sec intervals for 2 min. The percentage of motile cells in each strain was measured from 200 single cells and the differences were considered significant when the P value <0.05 using a standard χ² test. The mean velocity was measured from 25 motile cells of each strain and a significant difference was considered to be a P value <0.05 using a standard T-test.
A sample of mid-log phase culture of mg491-F157A-F158A strain was diluted 200x in SP-4 and grown overnight on 8-well μ-slides ibiTreat (IBIDI). Just before visualizing cells, culture medium was replaced with fresh pre-warmed SP-4 containing Hoechst 33342 0.01 mg/ml. Cells were observed by phase contrast and epifluorescence in a Nikon Eclipse TE 2000-E inverted microscope. Phase contrast and DAPI (excitation 387/11 nm, emission 447/60 nm) epifluorescence pictures were captured with a Digital Sight DS-SMC Nikon camera controlled by NIS-Elements BR software.
Samples of M. genitalium G37 wild type and mutant strains were diluted as previously described and grown overnight in SP-4 medium over coverslips at 37°C and 5% (v/v) CO2. Then, coverslips were dehydrated and metalized as previously described [17] and were visualized in a Merlin scanning electron microscope (Zeiss). The percentage of single cells with more than one terminal organelle and the percentage of cells with a size smaller than 0.35 μm were measured from 200 single cells. A significant difference was considered to be a P value <0.05 using a χ² test.
Samples of M. genitalium G37 wild type and mutant strains were diluted as previously described and grown overnight in SP-4 medium over holey carbon-coated grids at 37°C and 5% (v/v) CO2. Each grid was washed with 1x PBS supplemented with 0.9 mM CaCl2 and 0.49 mM MgCl2 (PBSCM, Sigma), blotted to remove the liquid excess and immediately plunged into liquid ethane in a Leica EM CPC cryo-workstation (Leica Microsystems). The grids were transferred to liquid nitrogen and kept at -179°C during image capturing in a 626 Gatan cryoholder (Gatan). The grids were examined on a JEOL 2011 transmission electron microscope operating at an accelerating voltage of 200 kV. Micrographs were recorded using a Gatan USC1000 camera under low electron dose conditions to minimize damage by electron beam radiation. A moderate underfocus between -30 μm and -15 μm was used to increase the contrast of the micrographs.
|
10.1371/journal.pcbi.1002288 | Deciphering the Arginine-Binding Preferences at the Substrate-Binding Groove of Ser/Thr Kinases by Computational Surface Mapping | Protein kinases are key signaling enzymes that catalyze the transfer of γ-phosphate from an ATP molecule to a phospho-accepting residue in the substrate. Unraveling the molecular features that govern the preference of kinases for particular residues flanking the phosphoacceptor is important for understanding kinase specificities toward their substrates and for designing substrate-like peptidic inhibitors. We applied ANCHORSmap, a new fragment-based computational approach for mapping amino acid side chains on protein surfaces, to predict and characterize the preference of kinases toward Arginine binding. We focus on positions P−2 and P−5, commonly occupied by Arginine (Arg) in substrates of basophilic Ser/Thr kinases. The method accurately identified all the P−2/P−5 Arg binding sites previously determined by X-ray crystallography and produced Arg preferences that corresponded to those experimentally found by peptide arrays. The predicted Arg-binding positions and their associated pockets were analyzed in terms of shape, physicochemical properties, amino acid composition, and in-silico mutagenesis, providing structural rationalization for previously unexplained trends in kinase preferences toward Arg moieties. This methodology sheds light on several kinases that were described in the literature as having non-trivial preferences for Arg, and provides some surprising departures from the prevailing views regarding residues that determine kinase specificity toward Arg. In particular, we found that the preference for a P−5 Arg is not necessarily governed by the 170/230 acidic pair, as was previously assumed, but by several different pairs of acidic residues, selected from positions 133, 169, and 230 (PKA numbering). The acidic residue at position 230 serves as a pivotal element in recognizing Arg from both the P−2 and P−5 positions.
| Protein kinases are key signaling enzymes and major drug targets that catalyze the transfer of phosphate group to a phospho-accepting residue in the substrate. Unraveling molecular features that govern the preference of kinases for particular residues flanking the phosphoacceptor (substrate consensus sequence, SCS) is important for understanding kinase-substrates specificities and for designing peptidic inhibitors. Current methods used to predict this set of essential residues usually rely on linking between experimentally determined SCSs to kinase sequences. As such, these methods are less sensitive when specificity is dictated by subtle or kinase-unique sequence/structural features. In this study, we took a different approach for studying kinases specificities, by applying a new fragment-based method for mapping amino acid side chains on protein surfaces. We predicted and characterized the preference of Ser/Thr kinases toward Arginine binding, using the unbound kinase structures. The method produced high quality predictions and was able to provide novel insights and interesting departures from the prevailing views regarding the specificity-determining elements governing specificity toward Arginine. This work paves the way for studying the kinase binding preferences for other amino acids, for predicting protein-peptide structures, for facilitating the design of novel inhibitors, and for re-engineering of kinase specificities.
| Protein phosphorylation is one of the most abundant posttranslational modifications. It is catalyzed by protein kinases, a large group of enzymes that account for approximately 2% of the human genome [1]. Phosphorylation involves the regulation of almost every process in the cell, and numerous diseases, such as diabetes, Alzheimer's disease and cancer, are tightly related to abnormal levels of protein phosphorylation. Thus kinases are considered one of the major drug targets of the 21st century [2], with over a hundred kinase inhibitors in various stages of clinical trials and several drugs already in the clinic [3].
Most kinase inhibitors target the ATP-binding site [4], providing different, but usually low levels of kinase selectivity [5]. In pursuit of additional (non-ATP site) ways of inhibiting kinases, which in some cases may provide kinase-selective inhibition [6], kinase-substrate and other kinase-protein interactions are being actively targeted by various research groups using small molecules [7], [8] and peptidomimetics [6], [9], [10], [11], [12]. Structural information and computational approaches have greatly contributed to the design of low-molecular-weight kinase-targeting drugs [13]. The need for computational tools for peptide design is on the rise, due to increasing interest in protein-protein interactions and their inhibition in general [11], [14], [15], [16] and for protein kinases in particular [6], [9], [17], providing part of the motivation for the current work. While peptides are usually considered poor drug candidates because of low cell permeability and high tendency to be rapidly metabolized, recent improvements in synthetic peptide chemistry [18], successful usage of modulations that enable cell-penetration of proteins and peptides [6], [12], [19], [20], [21], [22] and of different administration routes, open up new avenues in the field of peptidic and peptidomimetic drug discovery [23].
Members of the protein kinase family share a common structure, consisting of a small N-terminal lobe and a larger C-terminal lobe [24]. The ATP-binding site and the main substrate-recognition site lie within the major groove formed between the two lobes. In eukaryotes, most kinases transfer the ATP γ-phosphate to either serine or threonine residues (Ser/Thr kinases), while others phosphorylate tyrosine residues (Tyr kinases) [25]. The Ser/Thr kinases can be further classified into various families and subfamilies based on sequence similarity, such as ACG, CAMK, etc. [1]. Another common classification of Ser/Thr kinases is into three main groups, basophilic, acidophilic and proline-directed. This classification is based on basic, acidic or proline substrate residues that govern kinase-substrate recognition [26], [27], and is assumed to confer global specificity between the three groups of kinases [28].
The above-mentioned residues are part of a set of amino acid residues immediately flanking the substrate phosphorylation site (which is referred to as P0) that play an important role in the tendency of the substrate to be recognized and phosphorylated by a particular kinase. The term substrate consensus sequence (SCS) refers to the essential sequence elements surrounding the phosphorylated site [29]. The flanking residues are referred to as P−n or P+n according to their location along the substrate sequence, n residues N-terminal or C-terminal to the P0 position, respectively.
Early studies of the prototypical basophilic protein kinase A (PKA) showed a pronounced preference for Arginine (Arg) at positions P−2 and P−3 of the substrate [30]. Later, the strong preference for Arg at P−3 was shown to be a general feature of many basophilic kinases. In a recent work that tested the consensus phosphorylation motifs of 61 out of the 122 kinases in Saccharomyces cerevisiae, 57% were detected as basophilic, with 87% of the basophilic kinases showing a primary preference for Arg at the P−3 position [31]. In fact, peptide phosphorylation screening approaches often fix Arg at this position, and concentrate on exploring preferences at other positions [32].
Aside from the P−3 position, P−2 and P−5 are the only two positions for which the frequency of Arg is greater than its average occurrence in the human proteome [33] and these are the focus of the current study. While position Glu127 (PKA catalytic domain numbering used throughout the paper) of the kinase, located at the hinge that connects the two lobes, has been shown to be the source for Arg specificity at position P−3 [31], [34], [35], [36], [37], the identities and roles of kinase residues defining Arg specificities at P−2 and P−5 are more intricate. Mutational analysis [33], [38], as well as the crystal structures of several kinase-peptide complexes, confirm that different basophilic kinases use the same surface site to accommodate Arg at either the P−2 or P−5 substrate positions [39], . Positions 129, 133, 169 and a dominant acidic pair (170/230) have been implicated, but are not always sufficient for explaining the experimental P−2/P−5 Arg preferences [33], [40], [42], [43]. It appears that the P−2/P−5 Arg specificity and the interaction strength is not conferred by a readily observable sequence or structural feature, but rather by a combination of a few subtle attributes which need to be uncovered by particularly sensitive methods.
Prediction of kinase-specific phosphorylation sites is commonly based on sequence-based computational methods [44], but structure-based approaches have begun to emerge as well [45], [46], [47]. Notably, the PREDIKIN method combines structural information on specificity-determining residues with sequence information obtained from known kinase substrates [46], [47]. The currently available methods are trained on kinase and substrate sequences and rely on analogy to known complex structures. Well-suited for the detection of conserved specificity determinants between kinase subfamilies, these methods are less sensitive when specificity is dictated by kinase-unique features, and they are not aimed at supplying information on amino acid binding preferences outside the known spatial organization of the substrate/peptide complexes. Yet, such information is valuable for de novo design of protein-protein interactions inhibitors.
Computational mapping methodologies have the potential of addressing the challenge of kinase-unique binding and to specificity analyses further away from the phosphoacceptor binding region. These approaches identify the favorable binding position of a molecular probe using solely the molecular interaction field embedded in the three-dimensional structure of the protein. Consequently, a sensitive energetic description of independent functional moieties within the investigated binding environment is supplied. A variety of computational mapping methodologies have been developed, including grid based methods [48] combined with fft correlation techniques [49] and methods that employ simultaneous minimization of all probes [50]. Computational surface-mapping have been successfully used as an initial step in fragment-based drug discovery procedures [51], [52], in comparing the binding sites of different related receptors [53], and in classifying protein kinases based on their ATP-binding sites [54]. Nevertheless, since these methods are mostly designed and used for small molecules docking, they are less adequate for detecting binding positions in the context of protein-protein interactions. In the latter case, the amino acid probe is a part of a much larger molecule (protein/peptide) whose presence can modify the local dielectric environment at the probe binding site. This electrostatic shielding effect requires an appropriate treatment in order to obtain reliable scores, specific for protein-protein and protein-peptide interactions.
A specific scoring function for protein-peptide interactions is implemented in the PepSite method, which uses spatial position scoring matrices derived from a large set of protein-peptide complexes and is aimed to identify preferred amino acid binding positions on a given protein surface. By combining the predictions of single amino acid binding sites with the sequence order of the peptide, the method was shown to correctly locate the binding position of many peptides [55]. Yet, the single amino acid predictions were not tested explicitly by the authors.
Here we use ANCHORSmap [56], a recently developed computational mapping procedure specifically designed to identify binding positions of single amino acid side chains, in the context of protein-protein interactions, to study the Arg-binding preferences of representative basophilic and non-basophilic Ser/Thr kinases. ANCHORSmap consists of a specialized scoring function which was calibrated and tested for the ability to accurately position residues at protein-protein interface and to reproduce experimental ΔΔG values that were measured for alanine mutations [56].
We show that ANCHORSmap successfully discriminates between basophilic and acidophilic kinases and accurately identifies and top-ranks all P−2 and P−5 Arg-binding sites previously determined by X-ray crystallography. Furthermore, the Arg-binding positions detected for all 10 examined kinases are in line with their SCSs. A detailed examination of the Arg-binding maps produced for the different kinases, together with in-silico mutagenesis, sequence alignments and available crystal structures of kinase-peptide complexes, indicates important roles for several previously unappreciated positions and structural features in kinase catalytic domain that govern the P−2/P−5 Arg specificity.
The ANCHORSmap algorithm produces detailed binding maps of amino acid side chains on protein surfaces. The predicted binding positions (anchoring spots) are ranked by their calculated ΔG values, and adjacent anchoring spots can also be clustered into a single position to produce a sparser map of mean anchoring spots without significantly lowering the accuracy of the results [56]. In this work, the mean anchoring spots are reported unless otherwise stated, and in order to imitate a real prediction scenario, all of the calculations were performed on the unbound structures of the proteins.
Previous findings indicate that amino acids that have high propensity to form hot spots, such as Arg, Glu/Asp, Tyr, Trp and His, are also highly selective in binding to the entire protein surface [56]. As a preliminary test for the prediction sensitivity of ANCHORSmap for protein kinase surfaces, we tested the method for its ability to distinguish between basophilic and acidophilic kinases. Using both acidic (Glu) and basic (Arg) probes, the method produced a clear differential binding pattern between few representative basophilic and acidophilic kinases, indicating that it is sensitive enough for categorizing the basophilic/acidophilic nature of a given kinase without prior knowledge of its SCS (See Text S1 and Figure S1).
X-ray crystallography studies have shown that different kinases use the same surface site to bind Arg from either the P−2 or P−5 substrate positions. We will refer to this surface site as the −2/5 site. To the best of our knowledge, structures of kinase-peptide complexes in which an Arg residue has been shown to anchor at the −2/5 site are currently available for only four different basophilic kinases from three kinase families, defined in [26],[57]: PKA [39] and PKB [40](AGC family), PIM1 (CAMK family) and PAK4 (STE family) [58].
Using the unbound structures of the proteins listed in Table 1, we tested the ability of ANCHORSmap to correctly reproduce their Arg-binding positions. Both detailed and mean Arg-binding maps were produced and an example from the top 20 mean Arg-binding positions detected on the entire surface of PKB can be seen in Figure 1A.
Remarkably, although the search for Arg-binding positions started from thousands (∼7500) of probes initially scattered over the entire surface of each kinase, the computed positions corresponding to the experimental Arg-binding positions were ranked extremely high. For three out of four cases, the rank of the most accurate position in the detailed maps was lower than 4, and the top-ranking mean solutions coincided with the experimental Arg-binding positions in every case (Table 1). The solutions were also geometrically accurate: the average RMSD from the experimental positions was 1.6±0.3 Å, and for three out of four kinases, the top ranking mean Arg position was less than 1.7 Å from the experimental bound position (measured between the experimental and computed Arg Cζ atoms, which represent the centers of the guanidino groups). The only exception was PIM1, for which a larger distance of 2.6 Å was obtained. Examination of the entire cluster of anchoring spots that contribute to the mean position of PIM1 showed a clear tendency of the lowest-energy binding positions to accumulate in close proximity to the experimental Arg position (Table 1 and Figure 1B).
The acidic residue at position 170 of the kinases has been implicated in imposing a preference for Arg at position P−2 or P−5 of the substrate [33]. A comparison of the unbound structures of different kinases (listed in Table 2) showed that the acidic 170 residue may adopt different conformations. For the peptide-unbound PKB structure (3D0E), the conformation of Glu170 uniquely and significantly deviated from the peptide-complex structure (1O6K) and from the consensus unbound conformation observed for the other kinases (Figure 2). Thus, for PKB, both a bound-like rotamer of Glu170 (reported in Table 1) and the unbound conformation were used in the calculations. The unbound conformation of Glu170 reduced the binding affinity (by 2.3 kcal/mol) and worsened the ranking (from 1 to 8) of the correct mean solution. Nevertheless, the location of the top solution remained the same, in line with Ben-Shimon and Eisenstein's finding that mean anchoring spots are particularly useful for unbound predictions [56].
These results are particularly striking in view of the challenging computational task. For example, we analyzed the 10 top ranking Arg positions produced by the PepSite server (http://pepsite.russelllab.org/) for each of the four kinases listed in table 1. The best result was achieved for PKB, for which PepSite solution ranked 6 was located in a distance of 4.8 Å (measured between the Arg Cζ atoms) from the experimental position. This solution corresponds to the second top ranking solution produced by ANCHORSmap for PKB, see Figure 1. For the rest of the cases tested, no reported solution was found closer than 8 Å from the experimental position.
Six additional kinases from the AGC, CAMK and STE kinase families, for which peptide-bound structures have not been determined experimentally but unbound structures as well as experimental SCSs were available, were analyzed next: PASK, CAMK-II (CAMK family), ASK1 (STE family), p70S6K, PDK1 and PKC (AGC family). This completed the test set to 10 kinases.
Substrate-specificity studies for PKC isozymes have resulted in several, sometimes inconsistent SCS definitions [29], [38], [59], [60]. Therefore, the most frequent SCS of all PKC isozymes (RXXS/TXRX) [61], [62], [63] was compared to the average results obtained for the four PKC isoforms (alpha, betaII, iota and theta) for which unbound crystal structures are available.
Eight out of ten SCSs in the set contained the robust basophilic signature of Arg at P−3, but were diverse in terms of Arg-binding preferences for the −2/5 site: the set included kinases with no clear preference for Arg in either the P−2 or P−5 positions of the SCS (PKC, ASK1, CAMK-II, PDK), kinases with clear and exclusive P−2 (PKA, PAK4) or P−5 (PKB, p70S6K, PIM1) Arg preferences, and a kinase with dual P−2 and P−5 preference (PASK). Our goal was to predict the preference for Arg at the −2/5 site in general and to identify potential reasons for the preferences for −2 vs. −5 substrate positions.
By combining sequence alignment of key kinase residues and information extracted from SCSs, studies have suggested different residues at several sequence positions of the kinase catalytic domain (positions 129, 133, 169, 170 and 230, based on PKA numbering) as crucial for controlling the P−2/P−5 Arg specificity [31], [33], [40], [42]. The most dominant condition for attaining P−2/P−5 Arg specificity was identified by Zhu et al. [33], who showed that among the AGC, CAMK and STE kinases, P−2/P−5 Arg specificity is highly correlated with the existence of a single pair of acidic residues in positions 170 and 230. The importance of this acidic pair is illustrated in the crystal structure of the PKA/PKI complex [66], in which Arg19 (P−2) of PKI, which is anchored at the −2 subsite, is hydrogen-bonded to Glu170 and also forms a tight salt bridge with Glu230 (Figure 6A). However, it was not established whether Arg binding at the −5 subsite would also use the 170/230 acidic pair for binding. Indeed, in the crystal structure of the PIM1-pimtide complex [58], position 170 do not seems to be directly involved in the binding of the P−5 Arg. Moreover, for several kinases, the acidic pair is neither a sufficient nor an obligatory condition for attaining a strong Arg-binding preference [33], [67], and none of the currently suggested residues is able to explain the P−2/P−5 Arg preference of all kinases. We propose that this preference is most likely dictated by a delicate balance between the entire residue composition at the −2/5 site and other essential structural elements which cannot be simply detected by sequence-based methods.
To support this idea we combine the information obtained from the predicted Arg-binding positions of selected kinases analyzed in this study (PKA, ASK1, P70S6K, PASK and PIM1) with structural and sequence comparisons of key positions making up the −2/5 site (some previously suggested in the literature and others suggested here for the first time), to examine in detail the Arg-affinity-determining components at the −2/5 site. We also provide an estimation of the binding-energy contribution of selected key residues to Arg binding, by performing in-silico mutagenesis and recalculating the Arg-binding energies with ANCHORSmap. Except for one case in which Phe was replaced by Ser, the rest of the mutations were chosen such that they minimally affect the shape of the −2/5 site pocket, yet enable to determine the contribution of the residue's functional group to Arg binding. This was done by replacing charged/polar residues with hydrophobic residues of comparable size, and selecting the appropriate side-chain rotamer which would optimally mimic the native surface shape of the −2/5 site pocket. Thus Thr, Asp/Asn and Glu were replaced by Val, Leu and Met, respectively.
We used ANCHORSmap, a novel computational mapping approach specifically designed for the detection of favorable binding positions of amino acid probes on the surfaces of proteins, to investigate the Arg preference determining elements in Ser/Thr protein kinase substrate-binding grooves. We focused mainly on the P−2/P−5 Arg-binding preferences that typically characterize the SCS of a particular surface region, defined by us as the −2/5 site.
Initially, we demonstrated that the ANCHORSmap method produces high-quality predictions by detecting differential binding patterns on the surfaces of representative basophilic and acidophilic kinases. This enabled successful discrimination between the two types of kinases without any prior knowledge of their SCSs. It also suggested that kinases might employ an either/or strategy in which their substrate-binding groove is optimized for binding either acids or bases, but not both. A kinome-wide analysis is needed to investigate this idea and is planned for further studies.
Importantly, using the unbound kinase structures, the method accurately reproduced and top-ranked the X-ray crystallography-determined Arg-binding positions at the −2/5 site of four different kinases; it also showed excellent correspondence between the calculated ΔG values obtained for Arg at the −2/5 site of all 10 kinases tested in this work and the preference for Arg in the experimentally determined SCSs.
In phosphorylatable peptide libraries, the SCSs of basophilic kinases emerge with a dominant signature preference for Arg at P−3 [27], [31]. However, for the group of basophilic kinases tested in this study, carrying an Arg preference at both the P−3 and P−2/P−5 positions, the Arg positions detected at the −2/5 site were almost exclusively top-ranked, pointing to the −2/5 site as the most preferred binding environment for Arg on the entire kinase surface. ANCHORSmap detected an adequate binding position for accommodation of P−3 Arg in all cases, yet both its ranking and calculated energies were significantly weaker (data not shown). However, since the kinase P−3 Arg interaction is known to involve, in some cases, the ATP molecule as well [66], [71], [72], the absence of ATP during the calculations does not allow for accurate ΔG calculation, making the comparison between the two sites difficult. It must also be taken into account that the experimentally observed dominant preference for Arg at P−3 is measured using phosphorylation activity, whereas we estimate the contribution of Arg to binding. Catalytic activity and binding affinity do not necessarily have to be correlated. A series of experiments [73] showed that the intrinsic affinities of several protein substrates to their respective kinases are weak compared to their apparent affinities measured in traditional steady-state kinetic-activity assays. Experimental studies with PKA and the protein kinase inhibitory peptide PKI support the hypothesis that the P−3 position may be important for catalysis but less important for the binding itself. It was shown that replacement of Arg19 of the inhibitory peptide (at position P−2, experimentally shown to bind at the −2 subsite (PDB code 1ATP)) by Gly reduces the inhibition by 520-fold, while similar Arg replacement at the equivalent P−3 position (Arg18) reduces the inhibition by only 90-fold [74]. Yet in substrate peptides of PKA selected for catalytic activity, Arg at position P−3 is the most dominant one [31], [33]. Moreover, the PKA-PKI complex is similarly formed with [66] (PDB code 1ATP) or without [75] (PDB code 1APM) the ATP molecule, even though in the former case, the guanidino group of the P−3 Arg is hydrogen-bonded to the ATP ribose. This indicates that the P−3 Arg is not the most crucial residue for inhibitor binding. Note that while the −2/5 site is located on the C-lobe, the P−3 Arg interacts with the N-lobe, requiring an optimal geometry of the two lobes for contact formation. Such contact may stabilize the two lobes together, enabling the accurate geometry for an appropriate catalytic activity, but might be less important for binding of an inhibitor.
The binding information obtained by ANCHORSmap is not identical to the information that can be gained from phosphorylation peptide arrays. Thus it is particularly useful in the design of kinase peptide inhibitors, as opposed to the design of optimal kinase substrates. Indeed, ANCHORSmap results were recently used to rationalize the structure-activity relations of a peptidomimetic library of novel PKB kinase inhibitors [76].
The four currently available crystal structures containing a P−2/P−5-bound Arg at the −2/5 site imply that Arg can potentially bind in two distinct subsites within the −2/5 site. However, the limited number of available complexes has made it difficult to draw a clear conclusion regarding the subsite separation inside the −2/5 site and its relation to the Arg position along the substrate sequence. The search for potential Arg-binding positions within the −2/5 site helped resolve this issue: it supplied a clear dichotomization of the predicted Arg positions between two separate subsites, −2 and −5, each with a different set of predicted Arg positions that usually corresponded to the location of Arg in the P−2 and P−5 positions of the SCS, respectively.
Finally, we used the predicted Arg-binding positions together with structural, sequence and in-silico mutagenesis analysis of key residues, to explain known and novel structural and sequence specificity-determining features that govern the Arg interaction at the −2/5 site. The analysis showed that in many cases, the interaction strength is underlined by a delicate balance between several attributes of the binding site, architectural and chemical, which cannot be simply obtained from sequence alignment comparisons, but emerge automatically from the ANCHORSmap-predicted positions and accompanying ΔG values. This is due to the fact that ANCHORSmap utilizes the information embedded in the structure of the protein, capturing subtle changes in the physicochemical properties of the binding site. The predicted positions were then used to trace back the role of individual key residues and structural features that determine preference toward Arg in particular substrate positions. This methodology provided several new hypotheses regarding Arg specificity-determining elements at the −2/5 site, which can be further tested experimentally. We suggest that the inability of ASK1 to strongly bind Arg, in spite of the existence of the 170/230 acidic pair, can be explained by a unique sequence composition that affects the architectural arrangement at the −2/5 site of ASK1. We could therefore hypothesize that the preference for a P−5 Arg is not governed by the 170/230 acidic pair, as it is for P−2 Arg and as was previously assumed, but can be governed by several different pairs of acidic residues, selected from positions 133, 169 and 230, whereas position 170 affects only the binding of Arg at the −2 subsite. Acidic residue at position 230 on the other hand, serves as a pivot element in the recognition of Arg at both subsites.
Computational mapping of amino acid-anchoring spots on kinase surfaces can provide testable hypotheses regarding kinase specificity and peptidomimetics affinity [77] and may be used as input for anchor-driven peptide-docking [14] to predict 3D structures of kinase-peptide complexes. It is therefore expected to promote our understanding of kinase regulation and expand the possibilities for the design of kinase-specific signaling modulators.
A compendium of peptide-anchoring sites obtained in this work is available upon request, providing a basis for the development of novel kinase modulators for biochemical research.
Superposition of the −2/5 site was used for probe RMSD calculations and for the comparison between the −2/5 sites of different kinases. Superposition was performed over the Cα atoms of the following list of residues surrounding the −2/5 site: 128–136, 168–170, 201–204, 230, 234–236 (PKA numbering).
The random set of proteins consisted of 20 soluble medium-sized (255–280 amino acids), structurally and functionally unrelated structures, sharing sequence identity of less than 25%, which were taken from the PDB-REPRDB database [82]: http://mbs.cbrc.jp/pdbreprdb-cgi/reprdb_menu.pl. The list of structures are: 1BKC, 1G6H, 7YAS, 1QGI, 1P1X, 1MOO, 1QH5, 1UWC, 1JFR, 2A14, 1MML, 1ARB, 1J1M, 1AKO, 1VIN, 2A3U, 1JOV, 1NPY, 2HVM, 1IC6.
Kinase sequences were retrieved from KinBase (http://kinase.com). Only sequences of the catalytic domain of human kinases were included in the analysis. Sequence alignment was performed using the T-coffee server [83], [84].
Apart from the Phe129Ser replacement in PKA, in which the rotamer of the corresponding Ser in ASK1 was used, for the rest of the mutations, the rotamer which optimally mimicked the native surface shape of the −2/5 site pocket was selected by visual inspection. The selected rotamers were then subjected to a short (20-step) steepest descent energy minimization to remove steric clashes with the protein. Side-chain replacements, rotamer selections and energy minimizations were performed with built-in tools of the Discovery Studio package V2.5.
|
10.1371/journal.pntd.0005145 | Wolbachia-Based Dengue Virus Inhibition Is Not Tissue-Specific in Aedes aegypti | Dengue fever, caused by the dengue virus (DENV), is now the most common arbovirus transmitted disease globally. One novel approach to control DENV is to use the endosymbiotic bacterium, Wolbachia pipientis, to limit DENV replication inside the primary mosquito vector, Aedes aegypti. Wolbachia that is naturally present in a range of insects reduces the capacity for viruses, bacteria, parasites and fungi to replicate inside insects. Wolbachia’s mode of action is not well understood but may involve components of immune activation or competition with pathogens for limited host resources. The strength of Wolbachia-based anti DENV effects appear to correlate with bacterial density in the whole insect and in cell culture. Here we aimed to determine whether particular tissues, especially those with high Wolbachia densities or immune activity, play a greater role in mediating the anti DENV effect.
Ae. aegypti mosquito lines with and without Wolbachia (Wildtype) were orally fed DENV 3 and their viral loads subsequently measured over two time points post infection in the midgut, head, salivary glands, Malpighian tubules, fat body and carcass. We did not find correlations between Wolbachia densities and DENV loads in any tissue, nor with DENV loads in salivary glands, the endpoint of infection. This is in contrast with strong positive correlations between DENV loads in a range of tissues and salivary gland loads for Wildtype mosquitoes. Lastly, there was no evidence of a heightened role for tissues with known immune function including the fat body and the Malpighian tubules in Wolbachia’s limitation of DENV.
We conclude that the efficacy of DENV blocking in Wolbachia infected mosquitoes is not reliant on any particular tissue. This work therefore suggests that the mechanism of Wolbachia-based antiviral effects is either systemic or acts locally via processes that are fundamental to diverse cell types. We further conclude that the relationship between DENV blocking and Wolbachia density is not linear in mosquito tissues
| Dengue fever caused by the dengue virus (DENV) is transmitted by the mosquito, Aedes aegypti. To control the disease, an intracellular bacterium called Wolbachia has been introduced into Ae. aegypti where it blocks/limits success of infection of DENV. The mechanistic basis of blocking is not well understood but may involve Wolbachia activating the host immune system or competing with DENV for host resources. The strength of blocking appears to correlate with Wolbachia density. Here, we aimed to determine if any particular tissues inside the mosquito play a greater role in blocking. Tissues were chosen based on their Wolbachia density and their roles in infection and immunity. Wolbachia infected and uninfected mosquitoes were orally infected with DENV and Wolbachia density and DENV load were assessed in midgut, salivary gland, head, Malpighian tubules, fat body and carcass. Wolbachia density did not correlate with DENV loads in the same tissues nor with DENV loads in the salivary glands. We also showed that no one tissue appeared to play a greater role in blocking. In summary, these finding suggest that in the mosquito a threshold Wolbachia density may be required for DENV blocking. Our findings also suggest that blocking may involve mechanisms that are fundamental to all cells.
| Dengue fever, caused by the dengue virus (DENV), is the most prevalent arthropod transmitted virus, endemic in over 100 countries [1,2].The virus is comprised of four antigenically distinct serotypes (1–4) [3,4]. DENV is transmitted by Aedes aegypti and Ae. albopictus with the former being the principal vector [5]. With no specific antiviral drugs, management of the disease has mainly relied on relieving the associated symptoms of fever, headache and rash [6]. As the current tetravalent dengue vaccine offers incomplete protection [7], vector control remains the primary means of reducing disease prevalence.
One example of an emerging vector control strategy involves the use of a bacterial endosymbiont, Wolbachia pipentis that is naturally present in 40% of arthropods [8] and 28% of mosquito species, including Ae. albopictus, and Ae. notoscriptus. Interestingly, Ae. aegypti is not naturally infected with the symbiont [9]. Over the last decade, three different Wolbachia strains have been transinfected into Ae. aegypti where they form stable, inherited infections including; wMelPop-CLA and wMel, both from Drosophila melanogaster, wAlbB from Ae. Albopictus and wMelwAlbB, which is a superinfection from both host donors [10–13]. In these mosquito vectors, Wolbachia demonstrates an ability to limit or “block” the success of infection by viruses, nematodes and parasites [14–16]. This effect forms the basis of Wolbachia-based biocontrol trials to interrupt disease transmission in the human population via the vector [17]. The most advanced of such trials are focused on DENV control where the wMel strain has been released into wild Ae. aegypti mosquitoes and successfully spread [18].
Despite widespread field-testing, the mechanistic basis of Wolbachia-DENV blocking is poorly understood. Pathogen blocking has been partly attributed to the ability of the bacterium to increase the basal immune activity of the host thereby enabling it to resist subsequent DENV infection in a process known as ‘immune priming’ [19–21]. Wolbachia-DENV inhibition may also be as a result of competition between the symbiont and viruses for vital host nutrients such as cholesterol, as demonstrated in Drosophila [22]. Such competition may be expected given that the Wolbachia genome lacks a range of key genes in lipid biosynthesis pathways [23] and because viruses are heavily reliant on host cholesterol for replication [24,25]. Neither immune priming nor cholesterol competition however, can completely explain Wolbachia-DENV blocking.
The strength of blocking appears to correlate with Wolbachia density, whereby higher densities of the symbiont are associated with greater viral inhibition [11,26–28]. In mosquito cell lines, only highly infected cells show almost complete DENV inhibition [27,28]. The same relationship has been documented in other insects. In Drosophila simulans, the wMel, wAu and wRi strains grow to high densities and provide protection against Drosophila C virus (DCV). In contrast, the wHa and wNo strains that grow to very low densities show little blocking [29]. The fact that wAlbB is unable to block DENV in its natural host Ae. albopictus has also been attributed to low symbiont numbers. In Ae. aegypti where wAlbB has been introduced and hence grows to higher densities, DENV blocking is much stronger [28]. The correlation is further shown in Ae. aegypti by the disparity in blocking between the virulent wMelPop-CLA strain, which grows to very high densities compared to the wMel strain which grows to moderate densities [11].
Several studies have reported that Wolbachia is found at different densities in various tissues of the mosquito body, with the ovaries and Malpighian tubules tending to have high densities [14,21,28,30,31]. Osborne et al., [32] have suggested that Wolbachia density within the head, gut and Malpighian tubules correlated with the ability to mediate protection against DCV in D. simulans. These different tissues may be of varying importance for pathogen blocking as predicted by their Wolbachia densities or if they play a particular functional role in Wolbachia-based pathogen blocking. For example, the fat body is mainly involved in pathogen defence [33,34] and the Malpighian tubules, that happen to have very high Wolbachia densities now appear to have immune function [35]. It is unknown if there is a correlation between the Wolbachia encountered by DENV in these tissues and the subsequent progression of infection to the salivary glands as the endpoint of transmission.
When a mosquito takes a viremic blood meal, the virus first infects the midgut and then it disseminates to other tissues such as the Malpighian tubules, fat body, trachea and the salivary glands, where it can be transmitted to a human via the saliva on a subsequent bite [5]. The rate of DENV transmission correlates with the titre of virus in the salivary glands when studied in animal models [36] and mosquito infection rate is also known to correlate with virus infectious dose [37]. Even though several studies have suggested that intermediate mosquito tissues are infected by the virus differentially over time [38–41], it is not clear if there is a correlation between DENV infection in these tissues and that in the salivary glands.
Here we have examined the infectivity and viral load of a DENV serotype 3 strain in the tissues of Wildtype and wMel-infected Ae. aegypti. Specifically we have assessed whether Wolbachia densities predict DENV load in the same tissue and if densities in intermediate tissues predict subsequent DENV loads in the salivary glands. We found that there was a positive correlation between DENV loads in intermediate tissues and salivary glands in Wildtype but not Wolbachia-infected mosquitoes. There was also no correlation between Wolbachia densities and DENV loads in any particular tissue. Together, these findings suggest that no one tissue is particularly important for Wolbachia-based blocking and that Wolbachia may simply be limiting virus at the level of each individual cell, by fundamental processes shared by diverse cell types.
Two mosquito lines were used for this experiment; Wolbachia infected [11] and Wolbachia uninfected Ae. aegypti mosquitoes designated wMel.F and Wildtype [31,42], respectively. The wMel.F mosquito line was collected in 2012 from field release sites in Cairns, Australia [18] while the Wildtype line was collected in 2014 from Babinda, Australia. The Wildtype mosquito line was used within four generations of field collection to limit inbreeding. At every generation, the wMel.F mosquito line was outcrossed with 20% Wildtype males to prevent genetic drift between the two lines. Adult mosquitoes were maintained on 10% sucrose while the larvae were fed TetraMin® fish food (Melle, Germany) ad libitum. Mosquitoes were reared under standard conditions of 25°C temperature, 65% relative humidity and photoperiod 12 hours light: dark.
The DENV 3 strain used for this experiment was sampled from a patient during an outbreak in Cairns, Australia in 2008/2009 [43]. This strain was selected because it caused one of the largest dengue outbreaks in Australia [43] and because it has been demonstrated to infect both wMel and Wildtype mosquitoes at a high rate [44]. Passage 6 of DENV 3 (PFU 106) was propagated using the protocol by Ye et al., [45] and stored in single use aliquots of 1mL at -80°C. The virus was mixed with defibrinated sheep’s blood in the ratio 1:1 and fed through a membrane feeder to three to five day old mosquitoes. The mosquitoes were starved for 24 hours prior to oral infection. The wMel.F and Wildtype mosquitoes were both fed simultaneously over a period of three hours [44]. Mosquitoes were then anesthetized on ice and females that did not feed were sorted out and discarded. Engorged mosquitoes were maintained on 10% sucrose at 25°C until they were dissected.
The midguts, salivary glands, head, fat body, Malpighian tubules and carcass were dissected from each individual mosquito. These tissues were chosen mainly based on their functional role in DENV infection, dissemination and transmission in the mosquito. DENV first infects and replicates in the midgut before being disseminated to other tissues [5]. The end point of disseminated DENV is the salivary glands from where it is transmitted to the human host through the saliva when the mosquito takes a blood meal [5]. Assessment of DENV dissemination in mosquitoes is commonly done using the head tissue given ease of dissection [21,42]. Tissues were dissected on 8 and 14 days post infection (dpi). These time points were chosen to reflect the early stage of infection where DENV would have disseminated from the midgut to other tissues and the late stage of infection where infection would have been well established [44]. Dissections were done in 1X phosphate buffered saline (PBS). Tissues of each individual mosquito were placed in 96-well PCR plates (VWR LabAdvantage, Australia) containing 200ul of extraction buffer (0.01M Trizma base, 0.001M EDTA, 0.05M NaCl and 2.5ul proteinase K) and 2-mm-diameter glass beads (Merck KGaA, Darmstadt, Germany). Ovaries were separated from the carcass and discarded to ensure that Wolbachia density was not unduly influenced by gravid females. To minimize contamination within mosquito lines the dissecting pins were immersed in 80% ethanol for ~10 seconds between individual mosquitoes and discarded after every 20 individuals. New dissecting pins were used for each line to avoid cross contamination between Wildtype and wMel.F mosquitoes. All tissues were stored at -80°C prior to RNA/DNA co-extraction. The entire experiment was replicated three times.
Plates containing dissected tissues were homogenized for 1 min 30 seconds in a mini-Beadbeater (BioSpec Products, Bartlesville, OK). They were then incubated in a thermo cycler (C1000Tm Thermal cycler, Bio-Rad, California USA) at 56°C for 5 min, then 98°C for 5 min for the simultaneous extraction of RNA and DNA. The extracted RNA/DNA was stored at -80°C and subsequently used for the quantification of DENV 3 RNA copies and Wolbachia density.
Taqman qPCR was used to quantify DENV 3 RNA copies in LightCycler480 (Roche, Applied Science, Switzerland). The RealTime Ready RNA Virus Master (©1996–2016 Roche Diagnostics) was used for concurrent cDNA synthesis and DENV 3 RNA copies quantification following manufacturer’s protocol. Primers for DENV were designed from the 3’UTR region with HEX labelled probes [46]. The following qPCR cycling conditions were used: reverse transcription at 50°C for 10 min, initial denaturation at 95°C for 30s, 45cycles of amplification at 95°C for 5s and 60°C for 30s and a final cooling step at 40°C for 10s. Absolute quantification of DENV 3 RNA copies for individual tissues was extrapolated from a standard curve as previously reported [14].
Taqman multiplex qPCR was used for the quantification of the WD0513 Wolbachia gene [47] in LightCycler480 (Roche, Applied Science, Switzerland). The WD0513 gene was normalised to the mosquito housekeeping gene RPS17 [48,49] to account for different tissue sizes. The qPCR cycling conditions used are as follows: An initial incubation at 90°C for 5min followed by 45 cycles of amplification at 95°C for 10s, 60°C for 15s and 72°C for 1s and a final cooling step of 40°C for 10s. Relative quantification of Wolbachia was done using the inbuilt algorithm of LightCycler480.
Tissue infectivity (proportion infected) of DENV 3 was analysed using the binary logistic function in a generalized linear model with presence or absence of DENV 3 infection as the response variable and tissue type and time as predicting factors. DENV 3 RNA copies (DENV load) in tissues was analysed using the tweedie distribution with log link function in a generalized linear model with DENV load as the response variable and tissue type and time as predicting factors. Wolbachia density in tissues was analysed using the tweedie distribution with log link function in a generalized linear model with Wolbachia density as the response variable and time and tissue as the predictive factors. Models were run separately for the Wildtype and wMel.F mosquito lines. Non-Parametric Spearman correlation co-efficient was used to test for correlation between the following: (1) DENV loads in intermediate tissues and salivary glands, (2) Wolbachia density in tissues and DENV load in salivary glands and (3) DENV load and Wolbachia density in the same tissue. All statistical analyses were performed in SPSS® (IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY)
To determine if time post infection and tissue type had an effect on DENV 3 infectivity, we examined head, salivary glands, midgut, Malpighian tubules, fat body and carcass at 8 and 14 dpi. There was a significant effect of tissue for both Wildtype (Wald = 139.60; df = 5; p< 0.0001) and wMel.F (Wald = 40; df = 5; p< 0.0001) mosquitoes (Fig 1). The head was the least infected tissue in both Wildtype and wMel.F mosquitoes, failing to recapitulate patterns of infection in other disseminated tissues including the salivary glands. In a previous study [42] where DENV 3 infection rates in the mosquito head and body were examined, head infection rates were significantly lower than that of the body at 7 dpi in wMel.F mosquitoes. However by 14 dpi in the same study there was no difference between head and body infections in wMel.F mosquitoes. Furthermore, in the Wildtype mosquitoes, head infection rates were lower than that of the body at both 7 and 14 dpi but these differences were not significant [42]. The disparity observed in head infection rates between the present and previous study could possibly be due to the comparatively small sample size used by the previous study. Midgut and carcass were the most highly infected tissues in both Wildtype and wMel.F mosquitoes, respectively. In Wildtype mosquitoes (Fig 1A) there was a significant interaction between tissue and time (Wald = 15; df = 5; p = 0.011,). Midgut infections decline with time, becoming less of a source of infection beyond 8 days. Conversely, salivary glands are still becoming increasingly infected post 8 days. Interestingly, the pattern of infection across hemocoel-associated tissues indicates early dissemination and a plateau of infection rates as well as a similarity in the capacity for these tissues to support DENV replication. There was a clear effect of time (Wald = 10; df = 5; p = 0.002) in wMel.F mosquitoes with infectivity increasing from 8 to 14 dpi across all tissues (Fig 1B). Across the board, tissue infection rates are reduced in wMel.F mosquitoes as expected [14,42] but unlike in Wildtype mosquitoes, more tissues show rising infection rates with time, suggesting the power of blocking is strongest early in infection.
To determine if time post infection and tissue type had an effect on DENV load in tissues, the DENV loads of all the tissues were compared at 8 and 14 dpi. There was a significant variation in DENV loads in tissues of both Wildtype (Wald = 1497; df = 5; p< 0.0001) and wMel.F (Wald = 50; df = 5; p<0.0001) mosquitoes. Midguts had the highest DENV load in the Wildtype mosquitoes while head had the lowest load in the wMel.F mosquitoes (Fig 2). Even though time did not have a significant effect on DENV loads in tissues of both Wildtype (Wald = 2.1; df = 1; p = 0.144) and wMel.F (Wald = 1.9; df = 1; p = 0.166) mosquitoes, there was a significant interaction between time and tissue for both Wildtype (Wald = 131; df = 5; p< 0.0001) and wMel.F mosquitoes (Wald = 180; df = 5; p<0.0001). For instance while DENV load in the carcass decreased over time that of the salivary glands increased in Wildtype mosquitoes (Fig 2A). On the other hand, DENV loads in the carcass increase over time while that of the salivary glands decreased in the wMel.F mosquitoes (Fig 2B). For the most part, however, DENV appears to infect tissues early, reach a peak DENV load and remain relatively stable in Wildtype mosquitoes. In general, wMel.F mosquitoes, exhibited greater variation in DENV load across time and tissues and between individual mosquitoes than is seen for Wildtype possibly demonstrating variation in the efficacy of blocking.
We examined if DENV load in a range of tissues early in the infection process was predictive of loads in the salivary glands by testing for correlations. In the wMel.F mosquito line, the efficacy of blocking effect rendered many mosquitoes uninfected. As such sufficient numbers of DENV positive heads were not obtained for either time point and all tissues at 8 dpi had to be excluded. For Wildtype mosquitoes head DENV loads were not predictive of salivary gland DENV loads at either time point, 8dpi (r = 0.250; p = 0.516) or 14dpi (r = -0.071; p = 0.867) (Fig 3C). There was a significant correlation between midgut and salivary gland DENV loads only at 14 dpi (r = 0.701; p<0.0001) (Fig 3A) in the Wildtype but not in the wMel.F mosquitoes (r = 0.460; p = 0.550) at 14 dpi (Fig 3B). In the Wildtype, salivary glands DENV loads were positively correlated to that of the Malpighian tubules at both 8 (r = 0.684; p<0.0001) and 14 (r = 0.783; p<0.0001) dpi (Fig 4A). A positive correlation was also found between fat body and salivary gland DENV loads at both 8 (r = 0.594; p = 0.002) and 14 (r = 0.684; p<0.0001) dpi (Fig 4C). Carcass DENV loads were positively correlated to salivary gland DENV loads only at 14 dpi (r = 0.701; p<0.0001) (Fig 4E). Malpighian tubule (r = -0.260; p = 0.917), fat body (r = 0.299; p = 0.188) and carcass (r = 0.127; p = 0.545) DENV loads were not predictive of DENV loads in wMel.F mosquito salivary gland (Fig 4B, Fig 4D and Fig 4F). In summary, these findings show that DENV load in upstream tissues may predict salivary gland loads in Wildtype but not wMel.F infected Ae. aegypti.
To determine if time and tissue type affect Wolbachia density in wMel.F mosquitoes, we compared Wolbachia density in the head, salivary glands, midgut, Malpighian tubules, fat body and carcass over two time points (8 and 14 dpi) (Fig 5). We observed that time had no effect (Wald = 0.18; df = 1; p = 0.671) on Wolbachia density in tissues. There was a significant tissue effect (Wald = 3423; df = 5; p<0.0001) demonstrating that Wolbachia density varied across tissue types. For instance, Wolbachia was most abundant in the Malpighian tubules with the head having the lowest bacterial density. There was an interaction between time and tissue type (WALD = 28; df = 5; p<0.0001). For example in the Malpighian tubules, Wolbachia density decreased from 8 dpi to 14 dpi while that of the carcass increased from 8 to 14 dpi (Fig 5). In summary, Wolbachia density varied across different tissue types with the Malpighian tubules and head harbouring the highest and the least number of Wolbachia respectively.
We first examined if the Wolbachia density in particular tissues was predictive of DENV load in that same tissue. In the carcass Wolbachia density was negatively correlated (r = -0.580; p = 0.005) with DENV load at 8 dpi, but this was not the case at 14 dpi (r = 0.160; p = 0.898) (Fig 6E). At 8 dpi Wolbachia density in the midgut (r = -0.125; p = 0.601), salivary glands (r = 0.453; p = 0.0680), Malpighian tubules (r = -0.095; p = 0.700) and fat body (r = 0.070; p = 0.765) were not correlated with DENV load (Fig 6A–6D). Neither was there a significant correlation between Wolbachia density in the midgut (r = -0.060; p = 0.702), salivary glands (r = -0.063; p = 0.626), Malpighian tubules (r = -0.026; p = 0.873) and fat body (r = 0.0044; p = 0.784) and DENV load at 14 dpi (Fig 6A–6D). Infection rates for both DENV and Wolbachia were too low to be statistically analysed in the head for both 8 and 14 dpi. These findings demonstrate that Wolbachia density is not predictive of DENV load within any of tissue types tested.
To determine if Wolbachia density in any particular tissue has an effect on DENV load in the salivary gland, we compared Wolbachia density in all the five dissected tissues to DENV load in the salivary glands. At 8 dpi there was no correlation between Wolbachia density in midgut (r = 0.21; p = 0.940), Malpighian tubules (r = 0.412; p = 0.101), fat body (r = 0.221; p = 0.395) and carcass (r = 0.051; p = 0.844) and DENV load in the salivary glands (Fig 7A–7D). Neither was there a significant correlation at 14 dp between the Wolbachia density in midgut (r = 0.155; p = 0.232), Malpighian tubules (r = 0.046; p = 0.732), fat body (r = 0.060; p = 0.642) and carcass (r = 0.32; p = 0.801) (Fig 7A–7D). Infection rates for both Wolbachia and DENV were too low to be statistically analysed in the head at both 8 and 14 dpi. These results show that Wolbachia density in intermediate tissues does not predict salivary gland DENV load.
This study investigated whether particular tissues, especially those with high Wolbachia densities or immune function, are important in Wolbachia-mediated DENV blocking. Specifically, we assessed whether DENV loads or Wolbachia densities encountered in intermediate tissues predict subsequent infection in the salivary glands. All tissues examined were susceptible to DENV 3 infection. As expected, the tissues of Wolbachia infected mosquitoes had considerably lower DENV infection rates compared to the Wildtype, due to pathogen inhibition [11,14,42]. Interestingly, we observed that the strength of blocking was strongest early in the infection process but then declined with time. This is consistent with observations in Wolbachia infected flies [29] where DCV numbers are initially low but progressively climb from 2 to 30 days post infection. Moreira et al., and Ye et al., [14,44] also observed slight increases in infection rates over time in wMelPop-CLA and wMel infected mosquitoes respectively, but only when DENV titres in the blood meal were high. The differences in blocking between early and late stages of infection raise the question whether the midgut may be playing a particular role in Wolbachia-DENV inhibition.
The midgut, not surprisingly had a high initial infection rate in Wildtype mosquitoes but as observed in other studies [41], the infection rate declined as the salivary glands became increasingly infected. This increasing infection suggests that salivary glands may be a site of DENV replication as well as accumulation. The very low infection rates in the head for both Wildtype and wMel.F mosquitoes were unexpected due to the fact that the head is generally used as a proxy for DENV dissemination. Our finding is contrary to previous studies that found higher head infection using immunofluorescence assays and RNA estimates of DENV 1, 2 and 3 [39,41,50]. However infection rates may be affected by the specific DENV and mosquito genotypes studied [51] or environmental effects that often vary across studies [52,53]. Lastly, head infection rates could also be inflated if head tissues become contaminated with the salivary gland during dissections.
While it is known that infection of the midgut is dependent on the amount of DENV the mosquito ingests [54], whether this affects virus dissemination from the midgut to other tissues including the salivary glands is not well understood. Salazar et al., [41] have reported that the trachea may facilitate DENV 2 dissemination from the midgut. However other studies suggest that virus disseminates to mosquito tissues through the hemolymph [55,56]. DENV loads in almost all hemocoel-associated tissues of Wildtype mosquitoes were similar to one another and predictive of salivary gland loads at 8 dpi. This pattern suggests that DENV infects these tissues in parallel and then seeps into the hemolymph and then makes its way to the salivary glands. Interestingly, midgut DENV loads in Wildtype mosquitoes were only predictive of salivary gland DENV loads late in the infection process. This is likely explained by incomplete dissemination of DENV out of the midgut in the early stages of infection.
Our findings demonstrate that the infection dynamics of the fat body is not different from other hemocoel-associated tissues in Wildtype mosquitoes. This finding does not support a special role of this immune active tissue [34,57] in DENV inhibition. Innate immune genes, particularly in the TOLL pathway, have been shown to decline in the fat body after 3 days when challenged with DENV [33]. Therefore the capacity for DENV to successfully infect this tissue may reflect declines in transcription of immunity genes over time. Another tissue that has been reported to be involved in insect immunity is the Malpighian tubules. In Drosophila Malpighian tubules exhibit basal expression of antimicrobial peptides (AMPs) that then increase in response to immune challenge [58]. Malpighian tubules have further been shown to fight infection independent of the fat body in Drosophila [59] and melanise larvae of the dog heartworm, Dirofilaria immitis in Ae.sollicitans [60]. Regardless, the Malpighian tubules had similar DENV loads to other tissues in Wildtype mosquitoes. This suggests that the Malpighian tubules are unlikely to be playing an immune role in modulating DENV replication and transmission in Ae. aegypti.
Unlike in Wildtype mosquitoes, DENV loads in the wMel.F mosquito midgut, Malpighian tubules, fat body and carcass did not significantly influence that of the salivary glands. There was no correlation between DENV load in the salivary glands and Wolbachia density in any of the tissues studied. In almost all cases, Wolbachia densities in particular tissues were also not predictive of DENV loads in those same tissues. This is true even for the Malpighian tubules that harbour extremely high densities of Wolbachia. This unique tropism may relate to access to nitrogen given Wolbachia’s reliance on host amino acids for nutrition [61]. Our findings are contrary to those for Drosophila [32] where Wolbachia density in the head, gut, and Malpighian tubules correlated with DCV inhibition in the whole fly. Drosophila, however, is naturally infected with Wolbachia unlike Ae. aegypti [11]. Native hosts for Wolbachia appear to have more restricted tissue distributions and reduced bacterial densities compared to novel hosts that may be the result of coadaptation [62]. Therefore the relationships between tissue densities and blocking in flies may not be the same as in novelly infected mosquitoes.
Across a range of studies Wolbachia density appears to correlate with the strength of DENV blocking/viral inhibition [11,26–29]. There are several possible models for this relationship. Firstly, in the simplest case there is a negative linear relationship between the two. In whole mosquitoes this hypothesis is supported based on a comparison of DENV blocking between the wMelPop strain, which grows to very high densities, and the wMel strain, which grows to moderate densities [11]. Secondly, blocking may become apparent only after particular thresholds of Wolbachia densities are reached. In an Ae. albopictus cell line, a minimum density of ~960 Wolbachia per host cell (wsp/actin) was required for complete blocking of DENV [28] with no obvious correlations at lower Wolbachia densities. We observed the highest density of ~30 Wolbachia per host cell (WD0513/RPS17) in the Malpighian tubules demonstrating that Wolbachia does not normally grow to such high densities in wMel.F Ae aegypti tissues. Our work therefore confirms the observations in cell lines [28] that at lower densities, there is no correlation between Wolbachia densities and blocking. Given different functional roles of particular tissues, especially with regards to immunity, we also hypothesized there may be tissue specific contributions to DENV blocking. Our work suggests, however, that none of the tissues we examined played a greater role in the expression of blocking. Instead, efficacy of blocking may be determined at the level of the individual cell. Our work does not rule out the involvement of immunity[19–21] or nutrient competition for key resources [22] in the mechanism of inhibition, but suggests Wolbachia must act through aspects of host cell biology that are either systemic or fundamental to diverse cell types.
Our findings in Wildtype mosquitoes demonstrate that DENV disseminates from the midgut and infects mosquito hemocoel-associated tissues equally through time. They also suggest that infection of the mosquito head is not an accurate proxy for the assessment of dissemination. In terms of Wolbachia-based blocking of DENV this study reports two main findings. Firstly, the Wolbachia tissue densities in the mosquito are not linear predictors of DENV load as has been reported in cell lines where densities are usually very high. This may be related to the much lower densities naturally present in insect tissues. Secondly, DENV inhibition is unlikely to be explained by tissue specific mechanisms. Future studies seeking to dissect the involvement of either immunity, resource competition or other unknown contributors to mechanism, should focus on aspects of host cell biology that are fundamental across tissues. Generalisations from cell line based-studies are likely to be more biologically meaningful when Wolbachia densities are lower and more reflective of those found in insect tissues.
|
10.1371/journal.pntd.0006643 | Snakebite incidence in two townships in Mandalay Division, Myanmar | The global incidence of snakebite is estimated at more than 2.5 million cases annually, with greater than 100,000 deaths. Historically, Myanmar has one of the highest incidences of venomous snakebites. In order to improve the health outcomes of snakebite patients in Myanmar, access to accurate snakebite incidence data is crucial. The last population-based study in Myanmar was conducted more than a decade ago. In 2014, the Ministry of Health and Sports data from health facilities indicated an incidence of about 29.5 bites/ 100,000 population/year (a total of 15,079 bites). Since data from health facilities lack information about those who do not seek health care from government health services, a new population-based survey was conducted in 2 rural areas of Mandalay region. The survey data were compared to those obtained from healthcare services.
4,276 rural respondents in Kyaukse and Madaya townships in Mandalay Division were recruited using cluster sampling that involved random selection of 150 villages and random sampling of 30 households from each village. One adult member of each household was interviewed using a structured questionnaire.
One respondent from each of 4,276 households represented 19,877 residents from 144 villages. 24 people in these households had suffered snakebite during the last one year giving an annual incidence of 116/100,000. During the last ten years, 252 people suffered snakebites. 44.1% of the victims were women. 14% of the villages reported 4 or more bites during the last ten years, whereas 27% villages reported no snakebites. 92.4% of the victims recovered fully, 5.4% died, and 2% suffered long term health issues. One victim was reported to have died from causes unrelated to the snakebite. While there was no statistically significant difference between outcomes for children and adults, 4 of 38 of those under 18 years of age died compared to 7 of 133 adults between 19 to 40 years of age.
This incidence reported by the community members points to substantially more snakebites than the number of snakebite patients attending health facilities. This higher incidence points to the need for a nation-wide population-based survey, community education about gaining access to care where antivenom is available, and to the potential need for a larger supply of antivenom and expansion of medical care in rural areas.
| Snakebite is a major health issue affecting large numbers of people, particularly in tropical developing countries. Myanmar has one of the highest incidences of venomous snakebite in the world. Considering changes in demography, development, agricultural practices, knowledge about prevention and preventive practices, regular and accurate assessment of incidence is needed in order to improve public health programs and health services provision to improve health outcomes for snakebite patients. For that purpose, we conducted a large population based survey in Mandalay, which is one of the seven high incidence regions in Myanmar. The survey indicated a substantially higher incidence of snakebites than that suggested by the number of snakebite victims attending health care centres and hospitals. This higher incidence of snakebite has implications for community health service planning, scale of production of antivenom, and the need to improve access to health care centres or hospitals where antivenom is available, and suggests a need for community health education regarding appropriate 1st aid.
| Snakebite is a major global health issue. The estimated global annual incidence is about 5 million snakebites resulting in between 81,000 and 138,000 deaths [1]. In 1998, Chippaux [2] estimated the incidence as high as 5.4 million bites and in 2000, White [3] suggested more than 150,000 deaths. In India alone, a well-designed community-based study documented more than 45,000 deaths in 2005 [4]. Many venomous snakebites lead to severe and persisting morbidity resulting from amputation, infection, scarring, stigmatisation and psychological trauma, all resulting in substantial physical, psychological and economic disabilities and hardship. The global burden of snakebite is predominantly in tropical developing countries where manual farming practices predominate. With inadequate access to appropriate health care and the high indirect cost associated with receiving care in hospitals in major urban centres, snakebites victims and their families often face significant economic loss [5, 6].
Many snakebite victims do not seek or receive care in the formal health care sector. Epidemiological studies dependent on data from the health care system will therefore fail to capture information about these victims. This results in serious underestimation of the true snakebite incidence. This problem is exacerbated by the fact that snakebites mainly affect people in countries where health systems are evolving and the health management information systems are insufficiently robust to capture information even about those presenting to health facilities.
Global burden estimates indicate that South and Southeast Asia and sub-Saharan Africa have the highest snakebite burden in the world [1]. Historically, within Southeast Asia, Myanmar has one of the highest incidences of venomous snakebite. In 2014, Ministry of Health and Sports (MOHS) data from health facilities indicated about 29.5 bites / 100,000 population / year (a total of 15,079 bites). The most recent community level data available on snakebite in Myanmar were obtained more than a decade ago [7]. For planning an effective response to this important public health issue and to facilitate effective clinical care for all snakebite victims, it is important to re-assess the population-based snakebite incidence in addition to having information about the patients who receive care within the health system.
With demographic changes, changes to agricultural practices, education, economic development and associated changes in prevention and access to health care, it is important periodically to re-assess the incidence, morbidity, mortality, community knowledge and practices in relation to prevention and service utilisation. We conducted a survey in two townships near Mandalay to describe snakebite incidence and mortality in rural areas, as an overwhelming majority of snakebites occur in these areas. A secondary aim is to compare these survey data with data from health centre and hospital presentations. The research was conducted with ethics approval from the Ethics Committee, Department of Medical Research, Ministry of Health, Myanmar, and the Human Research Ethics Committee at the University of Adelaide.
A survey to assess the population-based snakebite incidence in the Mandalay region was conducted as part of a large health system and community development collaborative project involving the Myanmar Ministries of Health and Industry in conjunction with the University of Adelaide and other Australian and international institutions. This collaborative project aims to improve outcomes of snakebite patients, with a focus on improving the quantity and quality of nationally produced antivenom, its distribution in quantities sufficient for the needs of the various regions of the country, health services development through in-service training and resource improvement, and community education targeting training for prevention, first aid and appropriate use of health services. As part of the larger project, this survey was conducted to measure incidence of snakebite in the previous year, snakebite mortality in the previous 10 years, and snakebite-related knowledge and practices among community members.
After consultations with Myanmar colleagues and visits to numerous candidate townships and hospitals, we determined that the appropriate combination of a high incidence of snakebite, accessibility and local support from senior health care workers could be found in the Kyaukse and Madaya Townships of Mandalay Division. The project was in collaboration with the Myanmar MOHS and had approval for these two townships only. MOHS data based on reports from all facilities across the country indicated that Mandalay region and these two townships were representative of the high incidence regions. Incidences based on MOHS data in 2014 were 29.4/100,000 across the whole country, 38.3/100,000 across the 7 highest incidence regions (where about 90% of all snakebites in the country occur), and 44.7/100,000 in Mandalay region.
In Myanmar, a ‘township’ comprises not just an urban centre, but also the surrounding rural hinterland that may cover a significant area with many rural villages. All residents of the rural areas of Kyaukse and Madaya townships in Mandalay Division were eligible for inclusion in the survey. We excluded residents who lived in urban areas of Kyaukse and Madaya Townships (16% and 9% respectively). There is a clear distinction in the Myanmar health system between urban and rural areas. The regional health management authority had separate lists for rural and urban areas and that was the basis of distinction. We excluded urban areas as we considered that snakebite would be relatively infrequent in these areas and also that the issues related to accessing treatment would be different. Data from Mandalay General Hospital (MGH) and Mandalay Teaching Hospital (MTH) provide evidence of the relative paucity of snakebite cases in urban areas compared to rural areas. MGH takes patients from the rural regions, whereas the MTH takes patients from the urban areas. In the 12 months from April 2017 till March 2018, MGH admitted a total of 856 patients, and MTH admitted only 153 patients. Also note that many rural bite victims were cared for in Township Hospitals and did not require transfer of care to MGH at any stage. Our estimate, based on the observations across the hospitals is that only about 1 out of 10 rural patients treated at the Township Hospitals required transfer to MGH. The rest were managed at their respective Township Hospitals without onward referral.
For each selected household, an adult member was interviewed. History of snakebites and snakebite-related deaths among all household residents covering a period of 10 years prior to the survey was collected.
The household was the sampling unit. The sample size was defined as 4,500 households with about 20,000 household members. This is similar to most other recent studies in South and Southeast Asia. The sample size of those studies ranged from about 2,000 to 11,000 households [8, 9, 10,11] with the exception of one large study in Sri Lanka where the sample size was 44,136 households [12].
Cluster sampling was used, in line with the method recommended in the World Health Organisation Vaccination Coverage Cluster Surveys [13]. Three stages of sampling involved: (i) stratification by township, (ii) random selection of 75 villages from each of the two townships, (iii) random sampling of 30 households from each village selected from lists of households provided by the local government health departments. Census data on the population of each village was used in stage (ii) so that random sampling of villages was done with probability proportional to size. We used a 150x30 sampling strategy, determined after considering both statistical (precision) and practical (feasibility) issues. Based on a previous survey in Myanmar (7), the anticipated snakebite incidence was 100/100,000. The design effect for snakebite incidence in Myanmar is unknown. Design effects for most variables in UNICEF’s 2009–2010 Myanmar Multiple Indicator Cluster Survey [14] (thirty households per village, as in our survey) were between 1.5 and 2 and so we used a design effect of 2 in the design of our survey. We felt that it was feasible to conduct interviews in 30 households in a single day and that 150 villages would give good geographical coverage of the study area. The resulting sample size (4,276 households, 19,877 individuals), with correction for cluster sampling methodology, was considered adequate to enable some estimation of snakebite incidence. This sample size is similar to several previous studies of snakebite incidence in Asia (8,9,10).
Data were collected using an interviewer-administered questionnaire that took about 20 minutes to complete. The structured questionnaire (Table 1) included questions about the types of snakes in the area as reported by the respondents. Photos were not shown for the respondents to identify the type of snakes sighted in the area as this method has proved misleading and unreliable. However, patients’ descriptions and use of Burmese names for the different species contributed to our tentative identifications. Snakebites during the last 10 years, first aid and health care use, snakebite outcomes, as well as knowledge and practices for prevention, first aid and treatment.
The questionnaires were administered by primary health care workers, midwives, health assistants and public health staff employed at government health centres providing outreach care. The data collectors were trained by the research team and local Project team members, all of whom were native speakers of the Myanmar language. The completed questionnaires were reviewed daily for completeness by the trained data collector supervisors and were then reviewed by the Project team field supervisors for the survey. The questionnaires were double-entered for quality assurance. The statistical analysis was conducted using SPSS (IBM SPSS Statistics Version 24) after the survey data were weighted for respondent sampling probability. Population rates and incidence were adjusted for the design effect of the survey using the SPSS Complex Samples Function. All numerical findings, including 95% confidence intervals, are adjusted for the three stage sampling methodology.
The survey was conducted in 74 villages in Kyaukse and 70 villages in Madaya townships. One respondent from each of the 4,276 households was interviewed (participation rate of 94%). The total number of people in these 4,276 households was 19,877. 49.9% of the respondents were women. 29.1% were 21–40 years of age, 43.8% were 41–60 years of age, and 20.5% were more than 60 years of age.
In the 4,276 study households, representing 19,877 residents, 24 people had suffered a snakebite during the previous one year, an unadjusted incidence of 123/100,000 and adjusted incidence of 116/100,000 people (95% CI 74/100000–182 /100,000).
252 respondents informed us that at least one person in their households had been bitten by a snake during the previous 10 years (5.9% of households, 95% CI 4.8%–7.3%). Among the victims, 44.1% were women. Relatively fewer households in Madaya Township had experienced a snakebite (4.7% of households, 95% CI 3.3%–6.7%) during the last 10 years than in Kyaukse Township (7.3%, 95% CI 6.7–9.2%). 111 of 252 (44%) snakebites were reported from 21 villages (14% of the villages) were there had been 4 or more snakebites during the last 10 years. From 40 villages, there were no reports of snakebites, and only 1 snakebite was reported from 38 villages during the previous 10 years, 2 snakebites in 25 villages, and 3 snakebites in 19 villages.
Outcomes were reported for 230 victims in the previous 10 years. The respondents informed us that 209 (92.4%, 95% CI 87.6%–95.5%) of these had fully recovered, 14 (5.4%, 95% CI 2.9%–10.0%) died due to that snakebite, and 6 (2%, 95% CI 0.9–4.7%) did not recover fully and had long term health issues as a result of the snakebite. It was reported that one victim died from causes unrelated to the snakebite. While there was no statistically significant difference between poor outcomes for children and adults, 4 of 38 victims under 18 years of age died compared to 7 of 133 adults 19 to 40 years of age. A higher proportion of women died (7 out of 87) than of men (7 out of 141). This difference was not statistically significant.
The species of snakes reported are listed in Table 2.
The respondents informed us that most snakebites occur in and around fields and in the forests (Table 3).
Snakebites were reported to occur during all four seasons but peak incidence was believed to during hot periods before the monsoon (Apr-June–Tu Gu—Na Yone season) and during the monsoon period (Table 4). (Note that Tu Gu—Na Yone is shorter than the other three seasons.) The activities most associated with people being bitten were farming and collecting crops (Table 5).
Considering the likely better recall about symptoms and treatment for those 24 snakebite victims who were bitten during the one year preceding the survey, the household members were asked questions about symptoms and treatment received. 6 (25%) of those 24 that reported snakebite to one of their family members reported that it was a venomous snake, whereas 3 (12.5%) reported it was not a venomous snake. Of the 24 respondents 19 could not recall the type of treatment received. Three reported that the victim received anti-venom. There were 2 reports of intravenous fluid being used with one reported to have received oxygen as well. The respondents informed that 8 (33.3%) of these 24 patients fully recovered. Others did not report if their family member fully recovered or had long term consequences.
Our 2015 survey discovered an incidence of snakebite 116/100,000 in two areas chosen for their high incidence of snakebite. This is much higher than the 2014 and 2016 national incidence as reported in the public health statistics [15] and an incidence of 68.5/100,000 rural populations in Mandalay region calculated using the Ministry of Health and Sports health facilities statistics and Myanmar Information Management Unit Census [15, 16, 17].
Considering the relatively small number of snakebites as reported by the respondents pointing to an estimated incidence of 116/100,00 with wider confidence intervals, our extrapolations must be interpreted with caution. The incidence identified through our survey suggests that the number of snakebites in the rural areas of seven regions of Myanmar, where 90% of snakebites occur and where about 23 million people live, may be significantly higher than the number of snakebites reported in the national health facilities data. Nationally, in 2014, 15,079 snakebite patients sought medical care from government facilities, out of which 13,382 were from the seven high incidence regions (Mandalay, Sagaing, Bago, Magwe, Yangon, Nay Pyi Taw and Ayeryarwady). According to health facility data, the snakebite in all these high incidence regions (except Nay Pyi Taw, where incidence in about 19/100,000), range between 38/100,000 and 52/100,000, with an incidence in Mandalay of 44/100,000. This suggests that information about incidence in one of these regions might be applied to the others.
The case fatality rate of snakebite was about 5% in the rural areas of Kyaukse and Madaya where this survey was conducted. This suggests about 1,250 deaths (0.05 x 25,000) due to snakebite in the 23 million people living in the rural areas of 7 high incidence regions. The total number of deaths caused by snakebites across the whole of Myanmar as reported in the national health services data was 630 in 2014 [16]. This paper reports the community based incidence based on a community survey. As the overall incidence per 100,000 population is relatively low and for that reason despite a relatively larger sample size it was not possible to explore the health services use and outcomes in the community survey. Therefore, to further understand the health services use and health outcomes for a large number of patients, a hospital record database was established at Mandalay General Hospital (MGH). MGH admits patients from across the whole region, mainly from the rural areas. In 2016, 965 patients were admitted to that hospital. A minority, 11.5%, sought care from traditional healers before visiting a formal health care facility, while 87.7% sought care from the formal health system as their first point of contact. A large majority, 85.4%, of patients were bitten by Russell’s vipers (RV), and all 9.8% deaths of patients admitted to this hospital were caused by RV. The complications included Acute Kidney Injury, (59.8%), requiring dialysis in 23.9% cases. Green pit viper bites were the next most common cause of bites (7.6%).
The reasons why relatively fewer snakebites and snakebite deaths were recorded in the health services data may include failure of some victims to seek health care, or seeking care from traditional healers only, and deaths of some victims before they could seek medical care. A proportion of snakebites cause no symptoms of envenoming: because the venomous snake injects no venom (“dry bites”), because some bites are by non-venomous snakes or other animals, or because the “bite” was just a sharp injury from a thorn. These victims may decide not to seek medical care because they develop no symptoms.
Among other factors, characteristics of a disease under investigation influence recall bias [18]. It could be assumed that there is a good recall for snakebite compared to other diseases, because snakebite is so memorable and is greatly feared as it is life threatening and causes serious complications including amputations and deaths in many patients. The concordance of our survey’s one year (24 bites) and ten year (252 bites) incidences suggests that the information provided by the respondents is affected little by recall bias. Hence, we are confident that the incidence identified by our survey is close to the true incidence in our study area. A study in 2005 in two other townships of the high incidence regions of Myanmar reported similarly high annual incidence of as high as 100/100000 in Kyaukapadaung and 115/100000 in Taungdwingyi Township [7]. Myanmar is in a region having one of the highest snakebite incidence in the world. Kasturiratne reported that South and South East Asia suffer about 232,000 envenomings every year [19]. A survey in Bangladesh reported an incidence of 623/100,000 [8] and a survey in Laos reported an incidence of 355/100,000 in one area and 1162/100000 in another [9].
The high snakebite incidence and case fatality rate identified through this survey has important implications. There is an imperative to work with the communities to raise awareness about the need for early access to health care facilities by all of the victims regardless of whether envenoming has already occurred or not. While it is true that a proportion of bites by venomous snakes do not cause envenoming, to avoid delays in treatment leading to severe envenoming and systemic complications it is important that the victims are taken, as soon as possible after the bite, to health care facilities where antivenom is available. There, they should be observed and tested for symptoms and signs of envenoming so that treatment with relevant antivenom can be provided if needed. The long term complications of snakebites from Russell’s Vipers, which are responsible for most of the bites in this region, have not been studied formally. However, long term complications seen among the snakebite patients across the project hospitals including Mandalay General hospital include chronic renal failure, chronically infected bite site wound, hypopituitarism and loss of a limb from amputation. This aspect of snakebite sequelae requires further research in the future.
Our findings suggest a higher incidence of snakebite than recorded in health services data. This has implications for service planning, including the potential need for production of more antivenom.
The survey also revealed the large variety of snakes known to be present in this area. Symptoms of envenoming and envenoming syndromes, i.e. haemotoxic, neurotoxic, nephrotoxic, help make a more precise determination of snake species. However, as it would have been difficult for the community members to remember the symptoms correctly, this survey, which focussed on incidence, did not include questions about signs and symptoms affecting those who were reported bitten in the last ten years. Other aspects of the project, particularly examining of dead snakes brought by the patients to Mandalay General Hosital, confirmed that most of the snakebites in this area are by Russell’s Viper, Rat Snake, Cobra, Checkered Keelback and Green Snakes. A number of the snakes reported by the respondents are non-venomous. Some of the victims may not seek care if they consider that have been bitten by a non-venomous snake if envenoming symptoms do not appear early. There is a need for a herpetological survey in high snakebite incidence regions of the country.
Another important finding is that even within an overall high incidence region most of the snakebites may be clustered in some areas, with other areas having relatively few or no snakebites. Out of 144 villages, respondents from 40 villages reported no snakebite during the previous ten years. Likely reasons, that require further exploration, include the type of crops grown, proximity to rivers and marshy areas, sanitation in and around houses, and use of personal protective equipment such as long boots. This heterogeneous incidence of bites demands that the trends and situation within an area be carefully assessed when planning targeted health education, antivenom distribution, and in-service training for the health care providers and managers who work in the high incidence areas.
This survey was one of the many aspects of the health system development project. Other initiatives included auditing of hospital patient data, participatory research with the communities, examination of dead snakes brought by the patients, and discussions with the primary care staff about the snakebite problem in the area. These initiatives helped generate sufficient information to be able to assist the government health services in strengthening the algorithms for diagnosis and treatment, training a large number of staff at primary care facilities and hospitals, greatly increasing the quantity of antivenom, and lyophilisation of antivenom to avoid difficulties associated with cold chain in areas lacking electricity networks.
We acknowledge that the sample size was too small to provide a precise estimate of snakebite incidence. With only 24 snakebites in the past year, the point estimate of 116 snakebites per 100,000 population had wide confidence intervals (74/100,000 to 182/100000). To achieve a precise estimate of snakebite incidence in Myanmar would require a sample size of at least 100,000 people.
The lack of detailed information about symptoms, treatment and outcomes including long term consequences is another limitation. Considering difficulty in recalling clinical information by the community members it is challenging to capture such clinical information through a community survey. Research that involves reviewing the data of patients admitted to the township, district and referral hospitals will help capture such clinical information which when used in conjunction with the information about incidence would help refine and strengthen in-service trainings and better define which anti venoms are necessary.
|
10.1371/journal.ppat.1002474 | Trypanosoma cruzi trans-Sialidase in Complex with a Neutralizing Antibody: Structure/Function Studies towards the Rational Design of Inhibitors | Trans-sialidase (TS), a virulence factor from Trypanosoma cruzi, is an enzyme playing key roles in the biology of this protozoan parasite. Absent from the mammalian host, it constitutes a potential target for the development of novel chemotherapeutic drugs, an urgent need to combat Chagas' disease. TS is involved in host cell invasion and parasite survival in the bloodstream. However, TS is also actively shed by the parasite to the bloodstream, inducing systemic effects readily detected during the acute phase of the disease, in particular, hematological alterations and triggering of immune cells apoptosis, until specific neutralizing antibodies are elicited. These antibodies constitute the only known submicromolar inhibitor of TS's catalytic activity. We now report the identification and detailed characterization of a neutralizing mouse monoclonal antibody (mAb 13G9), recognizing T. cruzi TS with high specificity and subnanomolar affinity. This mAb displays undetectable association with the T. cruzi superfamily of TS-like proteins or yet with the TS-related enzymes from Trypanosoma brucei or Trypanosoma rangeli. In immunofluorescence assays, mAb 13G9 labeled 100% of the parasites from the infective trypomastigote stage. This mAb also reduces parasite invasion of cultured cells and strongly inhibits parasite surface sialylation. The crystal structure of the mAb 13G9 antigen-binding fragment in complex with the globular region of T. cruzi TS was determined, revealing detailed molecular insights of the inhibition mechanism. Not occluding the enzyme's catalytic site, the antibody performs a subtle action by inhibiting the movement of an assisting tyrosine (Y119), whose mobility is known to play a key role in the trans-glycosidase mechanism. As an example of enzymatic inhibition involving non-catalytic residues that occupy sites distal from the substrate-binding pocket, this first near atomic characterization of a high affinity inhibitory molecule for TS provides a rational framework for novel strategies in the design of chemotherapeutic compounds.
| Chagas' disease, or American trypanosomiasis, is an endemic illness that affects approximately 8 million people in Latin America. The etiologic agent is the protozoan parasite Trypanosoma cruzi. To survive in the mammalian host and invade its cells, leading to the chronic infection, the parasite incorporates a charged carbohydrate (sialic acid). However, the parasite is unable to synthesize sialic acid, having to scavenge it from the host's sialo-glycoconjugates, through a transglycosylation reaction catalyzed by the enzyme trans-sialidase, which is unique to these organisms. We have obtained a monoclonal antibody that fully inhibits T. cruzi trans-sialidase actually being, at the best of our knowledge, the most potent inhibitor available. We now report a complete characterization of this neutralizing monoclonal antibody, at the functional and molecular levels. The antibody displays very high affinity and specificity for the T. cruzi enzyme, labels the parasites' surface and effectively blocks its sialylation and host cell invasion capacities. The determination of the 3D structure of the enzyme-antibody immunocomplex by X ray diffraction, allowed us to unveil the inhibition mechanism, providing clues for rational drug design. Given that sialidases are virulence factors in several pathogenic microorganisms, the reported data shall help to expand informative knowledge in this area.
| Chagas' disease, the American trypanosomiasis, is a chronic disabling parasitic disease caused by the flagellate protozoon Trypanosoma cruzi. With an estimated global burden of 100 million people at risk, 8 million already infected, and approximately 40,000 new cases/year, Chagas' disease represents a major health and economic problem in Latin America [1]. The infection is naturally transmitted by triatomine vectors (“kissing bugs”), from the south of the USA to the southern region of South America, although chagasic patients are in fact dispersed worldwide due to migrations. Patients can also transmit the disease either by in utero infection leading to the congenitally acquired disease or by accidental transmission through contaminated blood. The acute infection is characterized by patent parasite burden. During this initial stage, T. cruzi induces several alterations in the infected mammal including intense polyclonal activation of lymphocytes [2], transient thymic aplasia [3], [4] and other clinical hematological findings [5], [6]. The majority of the patients control the parasitemia, survive the acute phase, and enter into an indeterminate form of the disease that may last for many years or even indefinitely [1]. Up to 20 years after the infection, ∼35% of patients develop different pathologies, such as cardiomyopathy, peripheral nervous system damage, and/or dysfunction of the digestive tract [1].
Sialic acids have proven to be crucial during the parasite's life cycle and survival in the mammalian host [7]–[10]. However, T. cruzi is unable to perform de novo synthesis of sialic acids [11]. This family of nine-carbon carbohydrates, is actually scavenged from the host's glycoconjugates, through a glycosyl-transfer reaction mediated by trans-sialidase (TS), a modified sialidase expressed by the parasite. In this way, the surface of the parasite becomes rapidly sialylated, with mucins being the main sialyl acceptors, in a process that allows the parasite to evade its destruction by serum factors [9], [10]. TS activity is also involved in host cell attachment and invasion [7], [8], as well as in parasite escape from the parasitophorous vacuole into the cytoplasm, where the parasite replicates [12].
In the trypomastigote stage, TS is a glycosylphosphatidylinositol-anchored non-integral membrane protein [13], actively released to the extracellular milieu, leading to a systemic distribution of the enzyme through the bloodstream. Its half-life in blood is significantly extended due to the presence of a C-terminal repetitive domain named SAPA [14]. TS activity is detectable in the bloodstream of infected humans and mice, until antibodies able to neutralize its catalytic activity are elicited [15]. The systemic distribution of TS is associated with several pathologies observed during the early steps of infection including depletion of thymocytes [16], absence of germinal centers in secondary organs [17] and thrombocytopenia and erythropenia [5], [6], all alterations that can be prevented by the passive transfer of TS-neutralizing antibodies [17], [18]. In fact, administration of the enzyme in mice before T. cruzi challenge, leads to more severe evolution of the infection [19]. These finding are also consistent with the fact that increased shedding of the enzyme correlates with increased virulence of the corresponding parasite strains [20].
TS has thus been identified as a potential target for drug discovery and design. Added to its key roles in host response evasion, cell invasion and pathogenesis, TS is not present in the mammalian host. The development of suitable drugs to treat/prevent Chagas' disease is urgently needed [21]. Only two compounds, benznidazol and nifurtimox, are currently available for treating both acute and chronic infections. These drugs are far from being optimal: fairly toxic, they trigger serious side effects, while also showing suboptimal efficacy in a high proportion of patients. The emergence of resistant parasite strains adds a concerning issue [22]. Several attempts to obtain suitable TS inhibitors have been made, especially once its 3D structure became available [23], [24]. However, only low affinity molecules have been obtained so far [25], [26], some of them toxic in in vivo assays [27], ultimately suggesting that further and more active efforts must be pursued.
We have obtained a TS-neutralizing mouse monoclonal antibody (mAb 13G9) that displays very high affinity and specificity towards the T. cruzi enzyme. This mAb is able to prevent immune system and hematological abnormalities, even when assaying highly virulent parasites under lethal infection conditions [5], [17]. We now report an extensive functional characterization of mAb 13G9, as well as the crystal structure of the 13G9-TS binary complex. The molecular features of the inhibitory mechanism are unveiled, providing novel insight for the development of TS inhibitors, which might also be relevant for related neuraminidases in other pathogens.
Mice were immunized with a TS recombinant protein (Δ1443TS), identical to the wt except it includes a deletion of a non-neutralizing epitope. Δ1443TS retains full enzymatic activity, while avoiding the otherwise typical delay in eliciting TS-neutralizing antibodies [28], [29]. Hybridomas were screened by TS-inhibition assay [30] and the 13G9 clone secreting a TS-neutralizing mAb (IgG2aκ) was obtained. The specificity of this mAb was confirmed by the absence of reactivity against the closely related sialidase from Trypanosoma rangeli and the TS from Trypanosoma brucei (data not shown). As depicted in Figure 1A, this mAb showed high affinity for the T. cruzi TS (KD ∼7.2×10−10 M) as calculated from the kinetic constants determined by surface plasmon resonance. In agreement, isothermal titration calorimetry assays indicated an equilibrium dissociation constant lower than 10−9 M (raw data not shown).
The mAb was purified by Protein A-affinity chromatography from filtered hybridoma supernatants. This purified material was further subjected to anionic chromatography (Figure S1). The mAb eluted as a single peak as evaluated both by TS-neutralizing activity (not shown) as well as by TS recognition in dot-blot assays (Figure S1). The same sequence was found in several mRNAs encoding for the antibody (not shown), in support of a clonal nature of the hybridoma. Purified mAb was proteolized with papain to generate the Fab fragment. Inhibitory activity of the fragment was determined and compared with that from the whole IgG protein (Figure 1, panels B and C). Although the full-length mAb appears to have a higher inhibitory activity (half maximal inhibitory concentration IC50 5.6×10−11 M), its Fab fragment still retains a nanomolar IC50 (1.6×10−9 M), clearly conserving its antigen-binding mechanism. These high inhibitory potencies are consistent with the apparent dissociation constant determined by surface plasmon resonance (see above), even though IC50 figures cannot be compared with affinity constants in absolute terms at this point (allosteric effects, or yet mixed inhibition mechanisms, may flaw a linear relationship). The purified Fab proved to be fairly unstable when non-complexed to TS, requiring immediate use for biochemical characterizations. This may be one of the main reasons for the observed inhibitory potency decrease compared to the entire immunoglobulin molecule. The Fab's instability precluded its use for further in vivo and in vitro biologic assays.
T. cruzi TS belongs in fact to a huge superfamily of genes, among which at least four families can be discriminated [31]. TSs are only included in one of these families, which encodes for a number of enzymatically active and inactive members [32]. These two forms of TS can be distinguished by the single Tyr342His mutation [33]: only the active TSs have the Tyr342 residue acting as the enzyme's nucleophile during the ping-pong reaction [34]. TS-mAb competition assays performed with the inactive TS showed that both proteins reacted similarly with the mAb. An equimolar mixture of inactive and active TSs, displayed ∼50% reduction of the neutralizing reactivity (Figure 1D). In a separate set of assays, heat-inactivated TS was not recognized by the mAb 13G9 (Figure S1), consistent with the hypothesis that the neutralizing epitope is conformational [35]. In the infective trypomastigote stages, all TSs include the SAPA C-terminal extension [31], which is absent in all the other TS-related families allowing for clear-cut discrimination. To address whether the mAb 13G9 was specific only for TS proteins, extracts from biotinylated trypomastigotes were reacted with the antibody (Figure 1E). Pulled-down material was subjected to Western blot and developed in parallel with anti-SAPA (for TS) and streptavidin for all the biotinylated parasite surface components. Strong signals were readily observed in both lanes, matching the TS expected protein sizes. No differential pattern was detected whatsoever, confirming the very high specificity of 13G9 antibody only towards proteins belonging to the TS family.
The reactivity of mAb 13G9 with whole parasites was assayed by immunofluorescence showing surface labeling consistent with the expected cellular membrane localization of TS (Figure 2A). The ability of the mAb to inhibit TS-mediated transfer of sialic acid from the surrounding environment to the parasite's surface molecules was then tested. To reduce the basal sialylation of parasites, sialyl residue donors were largely depleted replacing fetal bovine serum (FBS) by bovine serum albumin (BSA) in the infected tissue cultures; only host cells remained as the unique source of the sugar. Trypomastigotes were then collected and incubated with α(2,3)sialyllactose as sialic acid donor and TS, in the presence of mAb 13G9. The amount of transferred sialic acid was determined by the thiobarbituric acid method [36]. As shown in Figure 2B, mAb 13G9 very efficiently inhibited the parasites' sialylation, demonstrating its biologic relevance as a TS-inhibitory molecule. The sialylation observed in the treated parasites corresponds to the sugar acquired before the addition of the mAb. These quantitative results are in agreement with the Western blot assays we have recently reported for sialyl-transfer inhibition by mAb 13G9 using azido-modified sialic acids [37].
TS is involved in cell invasion [8], [12] given that sialic acid is required for competent interplay with the host cells. The ability of mAb 13G9 to interfere with the invasion process was therefore studied. The addition of the mAb (Figure 2C) strongly reduced the number of infected cells, highlighting its biologic activity and contributing direct evidence that TS is a valid target for drug discovery.
To gain atomic insight into the antigen-antibody interactions allowing mAb 13G9 to neutralize the TS catalytic activity with extremely high efficiency, we solved the structure of the immunocomplex by X ray crystallography.
Crystallogenesis screenings were performed under a sitting-drop vapor diffusion setup with a Honeybee963 robotic station, using standard 96-well plates. Several initial hits were obtained. Further manual optimization eventually allowed to grow crystals (0.7×0.05×0.05 mm) in polyethylene glycol (PEG) 20,000 plus dioxane, suitable for X ray diffraction data to be collected (Table 1). Limiting resolution was 3.4Å on a Cu rotating anode generator, and indexing was straightforward, indicating a primitive cell in the trigonal/hexagonal system. Cell parameters (a = b = 178.1Å, c = 140.7Å) suggested the presence of as many as 3 binary complexes per asymmetric unit, raising as well the hypothesis that its weak diffraction could respond to limiting X ray beam intensity in the context of a fairly large unit cell (low number of scattering cells per crystal unit volume). To rule out this possibility, several crystals were tested at the ALS (Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, CA) beamline 5.0.2 (8×1011 photons/s with 1.5 mrad divergence at 12.4 keV), with no detectable improvement in resolution as judged by standard quantitative statistics, strongly suggesting that crystal disorder linked to high solvent content (66% as determined after full refinement) is the major cause for maximum resolution sphere limitation.
No 6-fold peaks were found in self-rotation function maps, and the κ = 180° section revealed significantly weaker signals than the 3-fold axis (data not shown) consistent with point group 3. Systematic extinctions were observed in the reciprocal 00l axis, strongly suggesting space groups P31 or P32. The structure was solved by molecular replacement confirming SG P31. Two search probes were used to calculate rotation and translation functions: Protein Data Base (PDB) 3CLF (mouse IgG Fab fragment, chosen according to sequence similarity to mAb 13G9) and 2AH2 (high resolution T. cruzi TS model). Iterative cycles of maximum likelihood refinement [38] were interspersed with manual rebuilding [39]. The high resolution of the molecular replacement search models resulted in excellent maps and straightforward rebuilding, mostly adding missing side chains on the immunoglobulin heavy and light chains. Tight non-crystallographic symmetry restraints were kept only in the first refinement cycles, thereafter allowing for automatic local NCS detection, with variable weights according to evolving rms deviations, as implemented in the program Buster/TNT [40]. Model refinement statistics are summarized in Table 1. Interestingly, the PISA server (European Bioinformatics Institute, Hinxton) predicts that the TS-Fab 13G9 complex would not be stable in solution, contradicting our experimental results. This discrepancy reveals the still challenging task of predicting energetic and thermodynamic properties of protein/protein associations, based on the analysis of crystal structures of partners and derived complexes, despite the fact that prediction algorithms are complex and attempt integrating enthalpic and entropic effects, as well as solvent accessible surface burial and geometric complementarity [41].
Indeed, three binary Fab-TS complexes are located in the asymmetric unit, all very similar at the level of precision of our data. Refined models of immunocomplex 2 (IC2, composed by TS chain B, and chains I and M of the Fab molecule) and IC3 (TS chain C, complexed to Fab J and N) were superposed sequentially onto complex IC1 (TS chain A with H "heavy" and L "light" chains from the Fab molecule) minimizing root mean squared deviations (rmsd) of atomic coordinates. Such structural alignments resulted in 0.84Å rmsd between IC1 and IC2, and 0.82Å between IC1 and IC3. Regions of highest variation correspond to intrinsically mobile segments, as reflected by detailed analysis of atomic displacement parameters (isotropic B factors). The mean B factor for all atoms is relatively high (59.9 Å2), consistent with the low resolution to which these crystals diffract X rays. Crystal packing is indeed loose, leading to high bulk solvent content and corresponding protein flexibility. TS molecules display lower B factors then the Fab dimers to which they are bound. A global tendency is also maintained among the independent complexes, IC3 showing greater mobility than IC2, which in turn is more flexible than complex IC1 (59>53>48 Å2), probably due to the different packing environments. In the case of the immunoglobulin heterodimers, chains also display a clear difference among variable domains, more rigid, compared to the constant domains, which show a reproducible flexibility on the distal half, away from the interdomain hinge.
Given the overall structural similarity among the three complexes and the fact that complex IC1 resulted in a model with lower B factors, subsequent analyses will be referred only to this complex. Figure 3 shows the immunocomplex IC1 highlighting that the variable regions of the Fab light chain are interacting with TS loops located closer to the entrance of the enzyme's catalytic pocket, while the heavy chain associates to an adjacent, more distal patch.
The solvent accessible surface that becomes buried due to the enzyme-antibody interaction corresponds to 1810.2 Å2 (916.5 Å2 on the TS and 893.7 Å2 on the Fab, adding 506 Å2 from the heavy chain, and 387.7 Å2 from the light chain), within the typical range of antibodies reacting with protein antigens. On this interface, 15 hydrogen bonds and one salt bridge can be distinguished, as well as a number of residues that establish contact interactions (van der Waals forces), as listed on Table 2. The resolution limit of the diffraction data allowed for the identification of very few water molecules, none of which are directly involved in the accessible nor the buried surfaces engaged in interaction. The shape complementarity statistics [42] correspond to 0.673 and 0.645, after analysis of the interface areas with the light and the heavy chains, respectively. These figures are within the typical range (0.64–0.74) of specific protein:protein interfaces. The epitope (Figure 4) consists of residues H171, Y248, R311–W312, and loops 199–201 (KKK) and 116–128 (SRSYWTSHGDARD - W120 and A126 do not interact directly).
The structural bases of the catalytic inhibitory effect that this mAb elicits, can start to be elucidated by modeling the entrance of the sialylated substrate into the TS reactional center in the context of the TS-Fab complex (Figure 5). Superimposing TS PDB models 1S0I and 1S0J, onto our structure, allowed to define the positions of the substrates N-acetyl-neuraminyl-lactose (α(2,3)sialyllactose) and 4-methylumbelliferyl-N-acetyl-neuraminic acid (MU-NANA), respectively (Figure 5). The most readily observable feature is the steric hindrance that TS residue Y119 imposes, blocking the entrance of the sialyl residue in the reactional pocket.
The free mobility of the phenolic side chain of Y119 is limited by the juxtaposed residue S30 from the Fab's light chain (Figure 5). This restraint seems to play a central role in precluding the entrance of sialylated substrates into the catalytic pocket, entrance that absolutely requires the movement of Y119 [23]. A second effect could not be excluded, namely the spatial constraint exerted by the overall architecture of the associated complex. Residues S26–S28 (within the light chain complementarily determining region CDRL1) and S66–G67 on the same Fab chain, establish direct contact with TS residues R311 and W312. This interaction is located just on top of the catalytic pocket entrance, functioning as a ‘roof’ (SG/RW roof), where the catalytic center itself would be the floor. As shown in Figure 5B, when sialyllactose is located in position, the substrate pocket appears to be too small, predicting direct clashes of the glucosyl residue with the SG/RW roof (particularly residues Ser66–Gly67 of the Fab light chain). This scenario of course implies that Y119 could eventually be forced to move out of the sialic acid binding site, an unlikely event. The light chain loop 29–31 is also prone to interfere with the saccharide, if rearrangements are to be considered during its accommodation (data not shown). In order to obtain further experimental data evaluating the relative effects of Y119-mobility hindrance and/or the spatial constraints exerted by the SG/RW roof onto the catalytic pocket cavity volume, MU-NANA was assayed in TS-catalyzed sialidase reactions. MU-NANA is an artificial substrate that allows for TS-catalyzed hydrolytic and trans-glycosidase activities [43], and given its smaller volume, could better accommodate, avoiding steric clashes with the SG/RW roof structure (Figure 5B). TS-mediated MU-NANA hydrolysis was efficiently inhibited by mAb 13G9 (Figure 6), suggesting that the immobilization of Y119 does play a central role. The spatial confinement in the pocket, partly due to the SG/RW roof structure, might impose secondary constraints precluding torsional accommodation, even in the case of smaller compounds.
This report describes an extensive biochemical and structural characterization of the mouse mAb 13G9, which is herein demonstrated to act as a powerful inhibitor of the T. cruzi TS catalytic activity, displaying high specificity and affinity for the enzyme. T. cruzi TS is a virulence factor required for the survival of the parasite in the mammalian host. Several different biologic activities of the enzyme can be discriminated. The parasite uses TS activity to sialylate its own surface molecules, allowing it to evade lysis by serum factors [9], [10]. In this context, it should be noted that the addition of mAb 13G9 inhibited this sialylation process (Figure 1) in agreement with our previous findings with azido-modified sugars [37]. As well, TS is not only directly involved in the parasite/host cell interaction through the generation of a required sialylated epitope [7], [8] but also in escaping from the parasitophorous vacuole to the cytoplasm [12]. In concert with these findings, here we report that mAb 13G9 significantly reduces parasite infection of cell cultures (Figure 1). Passive transfer of neutralizing mAb 13G9 to heavily infected mice, protects them against TS-induced deleterious effects on the immune system and platelets [5], [17]. In this sense, it is well known that antibodies against neuraminidases are also effective in preventing other diseases such as Influenza [44]. These protective effects are very much promising to delineate a therapeutic tool. The high molecular weight of antibodies constitutes a main drawback in their use, due to eventual hindrance for effective diffusion into infected tissues, where high concentrations of locally produced TS are expected to be found. On the other hand, Fab fragments, small recombinant antibody-derived molecules (e.g. scFv), or yet antibody-mimetic engineered molecules [45], can be cleared exceedingly fast from the bloodstream [46], resulting in poor pharmacokinetic figures. PEGylation, and other modifications to improve bioavailability of these smaller protein scaffolds, constitute interesting approaches to be tested using mAb 13G9 as starting lead [47].
As a second interesting avenue to explore for therapeutic derivatives, the high affinity and specificity of this mAb, prompted us to elucidate its neutralizing mechanism, as an attempt to thereafter conceive low molecular weight inhibitors, suitable as chemotherapy leads. Some information can be gathered in this respect from previous studies of the neuraminidase from Influenza virus, a protein orthologous to TS. The overall geometry of the antibody/TS association that we are now reporting, is reminiscent of the one described for a Fab/Influenza-N2 neuraminidase complex (PDB 2AEP; [48]), which shows interaction with enzyme's loops on the same side of the reaction pocket, opposite to the patch where most other anti-neuraminidase antibodies have been reported (such as the ones involving avian N9 neuraminidase with antibodies NC41 and NC10, PDBs 1NCA and 1NMB, respectively; among others) [49]–[51]. The interaction surfaces of TS-13G9 mAb (this report) and N2NA-Mem5Fab (2AEP) are largely overlapping, although the antibodies are bound in inverted configurations with respect to the location of the heavy and light chains. Well defined escape mutations in Influenza (loops including positions 198–199 and 220–221, following N2 Influenza numbering scheme) identify epidemiologically important antigenic sites of neuraminidase, revealing antigenic drift in human viruses seemingly under natural antibody selection of enzyme variants [52]. These loops, connecting β2–β3 within the second blade of the six-bladed β-propeller domain, and β4 of this blade with β1 of the next one, are not structurally conserved between T. cruzi and Influenza enzymes, being longer in the former. Nevertheless, it is clear that the equivalent loops in T. cruzi TS do play a critical role in the 13G9 Fab association that we are now reporting.
One of the specific mAb loops that interact in a proximal position to the catalytic pocket of the enzyme, was observed precluding the displacement of Y119, a critical residue that has already been shown to be flexible in TS [24], [53]. Indeed, the mobility of Y119 plays a key role in the trans-glycosidase mechanism of TS. The determination of the three-dimensional coordinates of the paratope, including these features that lead to spatial constraints, uncovers relevant information. This is to be used as a precise guide, not only to undertake peptidomimetic syntheses, but most importantly, to use as a working template for the synthesis of non-peptidic molecules including critical pharmacophores [54].
The protocol of this study was approved by the Committee on the Ethics of Animal Experiments of the Universidad Nacional de San Martín, which also approved protocol development under the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health.
Recombinant T. cruzi TSs (constructs 1N1, 2Vo, Δ1443TS and 3.2) [24], [28], [33], T. rangeli sialidase [23] and T. brucei TS [55] were used. The 1N1 and 2Vo clones correspond to the full-length (including the SAPA repeats [33]) wild type genes that encode for enzymatically active and inactive molecules, respectively. The Δ1443TS recombinant TS was used for immunization procedures. Δ1443TS is an engineered variant where the deletion of a non-neutralizing epitope in the globular domain was done [28]. The TS 3.2 construct [24] is engineered to express the enzymatically competent globular domain only, containing seven mutations of surface-located residues that allow for protein crystallization. All TSs were expressed in Escherichia coli BL21 and immediately used after purification, avoiding >3 weeks storage at 4°C. Recombinant proteins were purified to homogeneity as described elsewhere [56], briefly, TS was subjected to immobilized metal affinity chromatography (Ni++-charged, Hi-Trap Chelating HP) followed by MonoQ anionic exchange chromatography (both from GE-Healthcare).
C3H/HeJ male animals (60 day old) were used. Mice received three intramuscular doses of Δ1443TS recombinant enzyme [28], 10 µg each with 100 µg of thiophosphodiester backbone CpG-ODN 1826 oligonucleotide (5′-TCCATGACGTTCCTGACGTT-3′, CpG motifs underlined) (Sigma-Genosys) as adjuvant [57]. TS-inhibition assay was performed as previously described [30], preincubating sera with TS and then testing for remnant activity using α(2,3)sialyllactose (Sigma) and [D-glucose-1-14C]-lactose (GE-Healthcare) as donor and acceptor substrates, respectively. Best responders were selected for cell fusion procedures.
Splenocyte suspensions were mixed with Sp2/0-Ag14 cells (ATCC) and fusions performed with polyethylene glycol (GIBCO) following standard procedures [58]. Cells were seeded on 96-well flat-bottom plates at a density of 1×105 cells/well in RPMI 1640 with 2 mM Na Piruvate, 10% FBS, 1X hypoxanthine-aminopterin-thymidine (HAT) solution (all from Invitrogen) and supplemented with 2% supernatant of Sp2/0–Ag14 cultures. One-week later, plates were observed under microscopy and the supernatant of those wells containing hybridomas were taken and refilled with fresh medium. ELISA was performed with these samples in search for TS-specific antibody production. To preserve discontinuos epitopes, the recombinant TS 1N1 containing the C-terminus repetitive extension (SAPA) was linked to the plate (MaxiSorb, NUNC) by Protein A-Sepharose (HiTrap, GE-Healthcare)-purified rabbit IgG anti-SAPA, a procedure that safely retained the enzymatic activity (not shown). Those culture wells where anti-TS antibodies were detected were further assayed by TS-inhibition assay [30]. Hybridomas secreting neutralizing antibodies were cloned twice by cell dilution. From four inhibitory antibody-secreting hybridomas detected, only one (named 13G9) was successfully recloned twice by the dilution method and then expanded. The mAb 13G9 was typed as IgG2aκ using the Mouse Antibody Isotyping Kit (GIBCO).
The 13G9 hybridoma was cultured in RPMI 1640 plus 2 mM Na Piruvate and 10% FBS. Supernatants were clarified and subjected to Protein A-Sepharose (GE-Healthcare) affinity chromatography. The mAb was eluted with 150 mM NaCl, 0.1 M Glycine-HCl pH 3.5 and aliquots were received on 0.1 M Tris-ClH pH 7.6 and dialyzed against 50 mM NaCl, 20 mM Tris-HCl, pH 7.6. Fractions were then loaded into an ion-exchange column (MonoQ, GE-Healthcare) and eluted with a 50–500 mM NaCl gradient in the same buffer (Figure S1). Purified 13G9 mAb was tested by TS-inhibition assay [30] and by reactivity to native and denatured TS-SAPA molecules spotted on nitrocellulose (Figure S1).
cDNA was obtained from 13G9 hybridoma cultures from total RNA using the SuperScript II retrotranscriptase (Invitrogen). cDNA quality control was performed by GAPDH amplification. To amplify the immunoglobulin Fab chains, oligonucleotide primer sets Fwh1 (5′-GTCAGGAGTTGAGCTGGTAAG-3′), Fwh2 (5′-CCTGGGACTTCAGTGAAGATG-3′) and Rvh (5′-TGGAGGACAGGGCTTGATTG-3′) were used for the heavy chain, and Fwl1 (5′-AACAATCATGTGTGCATCTATA-3′), Fwl2 (5′-GAGGAGATCACCCTAACCTG-3′) and Rvl (5′-TCAGGATGTGGTTGCAACAC-3′), for the light chain. Pfu DNA polymerase (Promega) was used and amplicons cloned and sequenced.
The association/dissociation kinetic constants (kon/koff) were determined with a BIAcore 2000 (BIAcore AB, Uppsala, Sweden). Purified mAb was dialyzed against 20 mM sodium acetate pH 5.6 and immobilized to sensor chips CM5 by using the amine-coupling kit (BIAcore AB). Chips were quenched with 1 M ethanolamine/HCl. After equilibration with 150 mM NaCl, 0.05% P20 surfactant, 10 mM HEPES pH 7.4 (HBS-EP), different concentrations of TS (from 1 nM to 10 µM) were injected at 50 µl/min. After each recording cycle, chips were regenerated with an injection of 2 mM HCl for 30 sec. A free surface of the chip was used as control throughout the experiments. Kinetic constants were evaluated using the program BIAevaluation 3.01 (BIAcore AB). Isothermal titration calorimetry assays were performed in the laboratory of Dr. Alan Cooper (Department of Chemistry, Joseph Black Building University of Glasgow, UK).
Inhibition constants of TS activity were determined for mAb 13G9 and its derived Fab fragment (see below for digestion details) by testing increasing amounts of inhibiting antibody with 2 ng of TS in 30 µl of 150 mM NaCl, Tris-HCl pH 7.6. After 5 min at room temperature (RT), 1 mM sialyllactose and 0.4 nmol (about 40,000 cpm) of [D-glucose-1-14C]-lactose (54.3 mCi/mmol, GE-Healthcare) were added. Remnant TS activity was evaluated [30] after 30 min incubation at RT.
Trypomastigotes (120×106) were purified from supernatants of infected Vero cell cultures, biotinylated (Sulfo-NHS-LC-Biotin kit form Pierce, Rockford, IL) washed and lysed in the presence of protease inhibitors and centrifuged at 16,000 g. Supernatant was precleared with Protein A-Sepharose (GE-Healthcare) and then reacted with 50 µl of mAb 13G9 hybridoma supernatant for 30 min. Then, Protein A-Sepharose was added and beads extensively washed before SDS-PAGE sample buffer addition and boiling. SDS-PAGE was performed with two parallel aliquots that were then transferred to polyvinylidene fluoride (PVDF) membrane (GE-Healthcare) and developed with either rabbit IgG anti-SAPA followed by horseradish peroxidase (HRP)-labeled secondary antibody or HRP-streptavidin and Super Signal West Pico Chemiluminescent substrate (Pierce).
T. cruzi trypomastigotes (CL-Brenner strain) obtained from Vero cell cultures (Minimum Essential Medium (Invitrogen) supplemented with 0.2% BSA instead of FBS to reduce sialic acid donors) were exhaustively washed with PBS. Parasites were tested by infection of Vero and HeLa cell cultures in the same medium at a multiplicity of infection of 30 in the presence of 0.1 mg/ml of mAb 13G9. After 3 h, cells were washed and medium plus 10% FBS was added. Cells were fixed and stained 24 h later for counting infected cells under microscopy. IgG purified from naïve mouse was used as control.
Parasites obtained under low sialic acid conditions as above were incubated with 1 mM sialyllactose (Sigma) as sialyl residue donor substrate and TS (2 µg/ml) with or without mAb 13G9 (0.1 mg/ml). After washings with PBS, sialyl residue content was determined by the thiobarbituric HPLC assay after hydrolysis in 0.1 M HCl for 1 h at 80°C [36]. IgG purified from naïve mouse was used as control.
Cell culture-derived trypomastigotes were washed with PBS and incubated with mAb 13G9 (0.05 mg/ml) for 15 min, washed, fixed with 1% paraformaldehyde for 10 min on ice, washed again and blocked for 1 h with 2% BSA plus 5% swine serum in PBS. After that, the parasites were adhered to glass slides via Poly-L-Lysine (Sigma), blocked again, developed with a FITC-conjugated secondary antibody (DAKO, Denmark) and observed by epifluorescence microscopy.
The sialidase activity of TS was determined by measuring the fluorescence of 4-methylumbelliferone released by the hydrolysis of 0.2 mM MU-NANA (Sigma). To 50 ng of TS, different amounts of hybridoma culture supernatant (0–10 µl) or RPMI plus 10% FBS (control) were added. The assay was performed in 50 µl of 150 mM NaCl, 20 mM Tris-ClH pH 6.8. After 10 min at RT, 200 µM of MU-NANA was added and incubation continued for 30 min. The reaction was stopped by dilution in 0.2 M NaHCO3 pH 10, and fluorescence was measured with a DYNA Quant TM 200 fluorometer (GE-Healthcare). Fluorescence values were referred to each RPMI control.
Purified mAb was dialyzed against 2 mM EDTA, 0.1 M Tris-HCl pH 7.6. Before papain digestion 1 mM dithiothreitol (DTT) was added. Papain-agarose beads (Sigma) were washed with the same buffer and activated by addition of 1 mM DTT for 15 min at 37°C. The Fab fragment was generated by digestion for 5 h at 37°C with papain-agarose beads (3U papain/mg mAb; 30 mg of beads for 14 mg of mAb) with gentle end-over-end agitation [58]. After centrifugation at 3,000 rpm, 10 µM trans-epoxysuccinyl-L-leucylamido(4-guanidino)butane (E-64) was added. Undigested antibody and Fc fragment were depleted by Protein A-Sepharose (GE-Healthcare) chromatography and Fab digestion and purity was assayed by SDS-PAGE.
To generate the immunocomplex, pure TS (3.2 clone) was immediately added after the depletion of papain-beads and E-64 addition step before subjecting the mixture to Protein A-Sepharose chromatography as above (Figure S1). The immunocomplex was brought to 25 mM NaCl and concentrated on a BIOMAX 30 K (Millipore) to 14 mg/ml and the buffer changed to 25 mM NaCl, 20 mM Tris-HCl pH 7.6. The purified immunocomplex was essentially free from contaminating proteins and only traces of TS activity remained (see Figure S1). Before crystallization trials, the immunocomplex was repurified by size exclusion chromatography (Superdex200 10/300, GE Healthcare) in an AKTA Purifier, (GE Healthcare) with isocractic elution in 100 mM NaCl, 20 mM Tris-HCl pH 7.6. The resulting single symmetric peak was pooled and concentrated to 7.5 mg/ml by ultrafiltration (Vivaspin, Sartorius-Stedim Biotech; 30 kDa-cutoff membrane) in buffer 25 mM NaCl, 20 mM Tris-HCl pH 7.6.
Crystallogenesis conditions were screened with a HoneyBee 963 robot (Digilab), using the vapor diffusion method in sitting-drops and reservoirs filled with 150 µl mother liquors (kits JCSG Core Suites I, II, III and IV, Qiagen), rendering 396 different conditions in 96-well plates (3-drop round bottom, Greiner). Protein drops were dispensed mixing equal parts of protein and reservoir solutions (300 nl + 300 nl). Plates were immediately sealed and incubated at 20°C. Hits were obtained in several conditions, one of them was chosen for manual optimization in 24-well plates (VDX, Hampton Research). Final optimized conditions consisted in 2+2 µl hanging-drops, 0.1 M bicine pH 8.5, 10% PEG 20,000, 4% 1,4-dioxane as mother liquor. To obtain larger crystals suitable for single crystal X ray diffraction experiments, repeated macroseeding cycles proved to be essential. Each cycle included selection of best crystal seeds that were transferred to protein-free drops of mother liquor and crystals etched for 30 sec (this washing procedure was repeated three times). Finally, the seed was added to a fresh hanging-drop containing 2 µl protein + 2 µl mother liquor, over 1 ml pure mother liquor. Single needles grew in 5–10 days, cryoprotected with mother liquor containing 12% PEG 20,000 and 30% glycerol and flash frozen in liquid nitrogen until data collection.
Single crystal X ray diffraction experiments were performed with a rotating copper anode (Micromax007-HF, Rigaku), multilayer mirrors (Varimax HF, Rigaku) and an image plate detector (Mar345 dtb, Mar Research). Crystals were mounted to collect data under cryogenic temperature (108°K, Cryostream Series 700, Oxford Cryosystems). To attempt improving diffraction resolution, similar crystals were subjected to X ray diffraction using synchrotron radiation at beamline 5.0.2 ALS, equipped with a wiggler inserted device. All data sets were processed with MOSFLM [59], SCALA and TRUNCATE [60].
The structure was solved by molecular replacement with the program Phaser [61], using the models 3CLF (mouse IgG Fab) and 2AH2 (T. cruzi TS in complex with 3-flourosialic acid) as search probes. The Fab probe was previously modified using Chainsaw [60], keeping only the conserved side chains, the rest pruned to alanine or glycine.
The model was refined to the highest collected resolution (3.4 Å) with the program Buster/TNT [38], using a maximum likelihood target function and non-crystallographic restraints throughout the entire process. A TLS model was used to refine correlated anisotropic atomic displacement parameters in large rigid-body domains. Reciprocal space refinement cycles were iterated with manual model rebuilding [39]. Validation tools within Coot were inspected regularly during the refinement process. Last validation steps were done with MolProbity [62].
The atomic coordinates and structure factors of the Fab-TS immunocomplex that we have solved in this report are accessible in the PDB with accession code 3OPZ. The models used to solve the phase problem have PDB accession codes 3CLF (mouse IgG Fab fragment) and 2AH2 (T. cruzi TS). A certain number of sialidase and trans-sialidase structures solved previously by us or by other groups, are mentioned in the Discussion section and can be accessed in the PDB with codes: 2AEP (Fab/Influenza-N2 neuraminidase complex); 1NCA (avian N9 neuraminidase complexed with antibody NC41); 1NMB (avian N9 neuraminidase complexed with antibody NC10); 2AEP (N2NA-Mem5Fab); 1S0I (T. cruzi TS in complex with sialyllactose) and 1S0J (T. cruzi TS in complex with MUNANA). Sequence of T. cruzi trans-sialidase can be accessed from the GenBank with the code L26499.
|
10.1371/journal.pntd.0005679 | Development of ELISAs for diagnosis of acute typhoid fever in Nigerian children | Improved serodiagnostic tests for typhoid fever (TF) are needed for surveillance, to facilitate patient management, curb antibiotic resistance, and inform public health programs. To address this need, IgA, IgM and IgG ELISAs using Salmonella enterica serovar Typhi (S. Typhi) lipopolysaccharide (LPS) and hemolysin E (t1477) protein were conducted on 86 Nigerian pediatric TF and 29 non-typhoidal Salmonella (NTS) cases, 178 culture-negative febrile cases, 28 “other” (i.e., non-Salmonella) pediatric infections, and 48 healthy Nigerian children. The best discrimination was achieved between TF and healthy children. LPS-specific IgA and IgM provided receiver operator characteristic areas under the curve (ROC AUC) values of 0.963 and 0.968, respectively, and 0.978 for IgA+M combined. Similar performance was achieved with t1477-specific IgA and IgM (0.968 and 0.968, respectively; 0.976 combined). IgG against LPS and t1477 was less accurate for discriminating these groups, possibly as a consequence of previous exposure, although ROC AUC values were still high (0.928 and 0.932, respectively). Importantly, discrimination between TF and children with other infections was maintained by LPS-specific IgA and IgM (AUC = 0.903 and 0.934, respectively; 0.938 combined), and slightly reduced for IgG (0.909), while t1477-specific IgG performed best (0.914). A similar pattern was seen when comparing TF with other infections from outside Nigeria. The t1477 may be recognized by cross-reactive antibodies from other acute infections, although a robust IgG response may provide some diagnostic utility in populations where incidence of other infections is low, such as in children. The data are consistent with IgA and IgM against S. Typhi LPS being specific markers of acute TF.
| In many African countries, clinical management of children that present with symptoms of bacterial sepsis, such as typhoid fever (TF) caused by Salmonella Typhi, consists of empiric broad spectrum antibiotics. Blood culture remains the gold-standard for diagnosis, but is slow, suffers from poor sensitivity, and often unavailable. Consequently multi-drug resistant bacteria have emerged that are difficult to manage with antibiotics. There is an urgent need to develop rapid, sensitive and affordable tests for patient diagnosis, help curb antibiotic resistance, and inform public health preventive strategies such as the deployment of vaccines. Here, we have assessed antibodies to S. Typhi lipopolysaccharide (LPS) and hemolysin E (HylE, t1477) in the sera of Nigerian children with acute TF and compared them with heathy children, children with other febrile infections, and adults from around the world with a variety of other bacterial infections. The key finding concerns LPS. This is a common cell-wall component present in many bacterial species. Yet despite this, S. Typhi LPS-specific IgA and IgM are excellent markers of acute TF in Nigerian children, and insensitive to other non-salmonelloses. This surprising finding suggests a rapid point-of-care test for TF can be developed based on detection of LPS-specific IgA+IgM.
| Salmonelloses are a group of potentially fatal bacteremias caused by different serovars of Salmonella enterica. Typhoid fever (TF), caused by the human-specific serovar S. Typhi, is a global health problem, especially in developing countries [1, 2]. In 2010 there were an estimated 26.9 million TF episodes worldwide, with a case-fatality rate of ~ 1% [3]. Salmonellosis caused by nontyphoid Salmonella (NTS) serovars are caused predominantly by the zoonotic serovars, S. Typhimurium and S. Enteritidis [4–7]. These are emerging in sub-Saharan Africa as an important cause of bacteremia in young children, typically when associated with malnutrition, malaria, severe anemia, and/or HIV co-infection [6, 8–11]. Case-fatality rates for blood-borne, or invasive, NTS (iNTS) infection is higher than that for typhoid, typically ~20% [4, 6, 12], although the antibiotic treatment regimen is the same. Over-use of empiric broad-spectrum antibiotic treatment for undifferentiated febrile disease has led to an increase in antibiotic resistance in both typhoidal and nontyphoidal serovars, and the potential for new antibiotics is not encouraging [1, 13, 14]. The development of effective vaccines to prevent invasive salmonellosis is therefore an important global health priority [1, 15]
Accurate diagnosis of salmonellosis remains a challenge in endemic settings. Clinically, initial presentation with typhoid or NTS disease is usually with non-differentiating fever alone, and often without symptoms of gastroenteritis that would indicate a Salmonella infection [7]. Bacterial culture is the gold standard for diagnosis of both typhoid and iNTS disease. However, culture suffers from poor sensitivity, and culture facilities are very limited in resource poor settings such as Nigeria and other countries in Africa. Even when such facilities are available, the time to a laboratory diagnosis is around 48 hours, and is often unaffordable for most patients. An inactivated-Salmonella agglutination test, developed by Widal >100 years ago, is a rapid and affordable single-step test. It remains the mainstay of diagnosis in many developing countries, even when culture facilities are available. However, the Widal’s test has poor specificity, thought to be caused by antigens shared between S. enterica serovars, and between other species of bacteria, such as Brucella melitnesis [16]. The Widal’s test also fails to discriminate between current and previous exposure, thus requiring two samples to be taken 7–10 days apart to monitor for an increase in titer. In practice, the decision to treat with antibiotics has to be made on the basis of the first test, and confirmatory convalescent testing is often not practicable or irrelevant for immediate patient management. It is also less sensitive in the acute stage of infection when IgG titers are lower.
The lack of accurate tests for surveillance also has resulted in only limited understanding of epidemiology of salmonellosis in Africa. The high mortality, particularly in children with iNTS infections, and the recent emergence of drug resistance, emphasize the need for a better understanding of the epidemiology before the rational design and implementation of control measures, including vaccines, can be effectively deployed.
In this study we have addressed the development of improved serodiagnostics with well-defined serum samples collected from febrile children in Nigeria. Based on proteome microarray screening data published recently [17], we hypothesized that LPS and/or the hemolysin E (HylE, t1477) antigen may have diagnostic utility for TF. However, it was unknown from the original study if these antigens were cross-reactive for other bacteremias. Here we have evaluated IgG, IgM and IgA ELISAs using purified S. Typhi LPS and HlyE using culture-confirmed pediatric bacteremias, including typhoid, iNTS disease, and ‘other’ febrile diseases, as well as healthy Nigerian children, and healthy adults from the U.S. We find LPS-specific IgA and IgA+M ELISA, in particular, are sensitive in diagnosing acute typhoid in these children, and descriminate well between typhoid and healthy, and other febrile bacteremias commonly encountered in Nigeria.
In a previous study [17] we confirmed the potential utility of S. Typhi LPS-specific IgA for the diagnosis of acute typhoid in sera from Nigerian children. Before pursuing this further, we wished to investigate potential cross-recognition of LPS by antibodies from other acute infections that might lead to a false-positive diagnosis of typhoid. For this we conducted a pilot study using a microarray displaying LPS from 7 different bacterial pathogens (S. Typhi, S. Typhimurium, F. tularensis, B. pseudomallei, B. melitensis, V. cholerae and E. coli), and probed it with sera from Nigerian pediatric samples and adult samples available from other febrile diseases, and controls. The data for IgA and IgG reactivity are summarized in the box plots in Figs 1 and 2, respectively. Panels A-D show Nigerian pediatric samples. As reported previously, IgA reactivity for S. Typhi LPS was strongest in typhoid cases (N = 16; Fig 1A), largely absent from ‘No Growth’ (N = 16; Fig 1C) and healthy control (N = 16; Fig 1D) samples, while present in a few individuals with culture-confirmed NTS (N = 16; Fig 1B) presumably owing to the antigenic similarities between LPS from related Salmonella serovars. Although there is a range of signals from the typhoid cases, only one sample was negative. We then examined the reactivity of sera from other bacteremias for other locations outside Nigeria, as follows: tularemia from Spain (N = 12; Fig 1E), melioidosis from Thailand (N = 7 acute, and N = 7 convalescent; Fig 1F), brucellosis from Peru (N = 12 acute, and N = 16 convalescent; Fig 1G), cholera from Bangladesh (N = 7 acute, and N = 7 convalescent; Fig 1H), and C. difficile infections (CDI) from the UK (N = 16; Fig 1I). Also probed were malaria samples from Mali, PNG and Kenya (N = 16; Fig 1J) and healthy controls from the U.S. (N = 20; Fig 1K). With the exception of two melioidosis cases and two malaria cases, IgA from these other infections did not cross-react with S. Typhi or S. Typhimurium LPS in this study.
Of note, IgA from other gram negative bacteremias did recognize the LPS appropriate to the infecting organism. Thus, IgA in individuals with acute tularemia specifically recognized F. tularensis LPS (Fig 1E), melioidosis IgA specifically recognized B. pseudomallei LPS (Fig 1F), brucellosis IgA specifically recognized LPS from B. melitensis (Fig 1G) and cholera IgA specifically recognized V. cholerae LPS (Fig 1H). This shows the lack of cross-reactivity for S. Typhi LPS was not due to a lack of anti-LPS antibodies in these other infections. In several cases, the signal intensity of the LPS-specific IgA response correlated with stage of infection. For example, of the 12 tularemia samples, the six 2nd time-point samples after MA seroconversion (late acute stage) gave maximal signals against F. tularensis LPS, while the remaining samples were the 1st time point prior to seroconversion (early acute).
For IgG, the most robust signals against S. Typhi and S. Typhimurium LPS were from the Nigerian acute typhoid cases. However, IgG was not a reliable marker of acute typhoid in Nigerian children in this array study. IgG against these antigens were particularly common among all the samples tested, but particularly in the Thai melioidosis, Peruvian brucellosis, malaria samples from various locations (Fig 2F, 2G and 2J, respectively), and the US negative controls (Fig 2K), consistent with the exposure to Salmonella species being widespread globally. These data are consistent with IgG being associated with both acute and convalescent (previous) exposure. Indeed, most individuals tested had IgG to multiple LPS species. For example, many of the Nigerian individuals in panels A-D also have LPS-specific IgG to E. coli and B. pseudomallei, which may indicate a previous exposure to these organisms and/or cross-reactivity from other infections. Antibodies against F. tularenisis are also quite common among different populations where tularemia is non-endemic (e.g., U.S), and may reflect cross-recognition of antibodies to other non-pathogenic Francisella species [18].
The data from the pilot study described above indicated that LPS-specific IgA may have utility for discriminating between acute and convalescent typhoid or other acute infections. While a more deployable array format is currently under development [19], in parallel we decided to develop an ELISA test for typhoid. The ELISA is inexpensive, robust and provides results more quickly than blood culture. Two batches of HlyE were used in the course of this study which, when compared by ELISA using all 349 Nigerian samples correlated with an r2 = 0.922 and a slope = 1.05 using Spearmann’s rank correlation (S1 Fig). The optimal conditions for ELISAs were initially determined using individual serum samples from three Nigerian typhoid patients and a healthy control. The optimized concentrations of the coating antigens were determined by titration to be 1.25μg/ml for LPS, and 2.5μg/ml for HlyE (t1477). Two serum dilutions, 1/100 and 1/200, were evaluated for the highest ratio when comparing heathy controls and culture-confirmed typhoid. For t1477 ELISA, 1/100 was selected, while for LPS ELISA, 1/200 was found to give the higher ratio and selected for subsequent studies. Two secondary antibody dilutions recommended by the manufacturer, 1/12,500 and 1/25,000, were evaluated for optimal signal to background ratio. For IgA and IgM ELISA, 1/12500 dilution was selected, while for IgG ELISA 1/25000 dilution was used. Once established, a standard operating procedure was used throughout the study. Batches of ELISA plates were prepared by pre-coating plates with antigen, blocking, and then storing dried at 4°C in desiccated pouches until required for use.
A total of 495 serum samples were used (Table 1) and tested for LPS-specific IgG, IgA, IgM and IgA+IgM in separate ELISAs. The samples comprised 369 Nigerian pediatric samples, consisting of culture-confirmed typhoid (“S. Typhi”, n = 86), non-typhoid Salmonella (“NTS”) disease (n = 29) or other bacteremias (“Other”, n = 28; listed in Table 2), as well as febrile cases that were blood culture-negative for any bacteria (“No Growth”, n = 178), and healthy Nigerian control children (“Healthy”, n = 48). Also tested by ELISA were well-defined sera from tularemia, brucellosis, and malaria cases, as well as U.S. controls.
Results of all ELISAs are summarized as box plots in Fig 3; the same data are also presented as bar charts in Supporting Information S2 Fig. In IgA ELISAs, the median OD value of the typhoid group was statistically different from all other groups when tested by the Wilcoxon method. Of these other groups, the NTS disease group showed the highest reactivity, due presumably to antibodies to LPS from NTS serovars cross-reacting with LPS from S. Typhi. While the difference between the medians of the S. Typhi and “No Growth” groups were highly significant (P < 0.0001), the latter group contained a large number of outliers that may correspond to blood culture false-negatives. Significantly, none of the 48 healthy Nigerian controls or 28 Nigerian “other” infections were seropositive for LPS IgA. Reactivity by non-Nigerian other infection groups was generally very low, although some Peruvian brucellosis cases had reactivity or cross-reactivity to S. Typhi LPS. There were two outliers in the malaria group (N = 48) with an LPS-IgA response. Although it is not known whether these individuals had a co-infection with Salmonella, association of malaria with salmonellosis is well known to the medical community. The IgG response to LPS (Fig 3B) was elevated in all groups, consistent with widespread previous exposure to Salmonella sp. The P-values for NTS, tularemia and brucellosis were larger than for IgA, with brucellosis failing to reach significance. Overall, the pattern of reactivity by IgM to LPS (Fig 3C) was similar to that of IgA, with the notable exception of NTS which was not significantly different to typhoid cases. As for IgA, IgM reactivity in healthy Nigerian children was very low, whereas IgM reactivity by the ‘other’ infections from Nigeria and elsewhere were overall higher than for IgA. Overall, the data indicate that the LPS-specific IgA has the best potential of the three isotypes for the diagnosis of acute typhoid from other febrile diseases.
Previous experiments using a S. Typhi full proteome array [17] revealed very few protein antigens with utility for diagnosing typhoid fever in Nigerian children. However, the hemolysin E protein (HylE, t1477) did emerge as a potential candidate, and is examined further here and in the following section for sensitivity and specificity using the full serum collection (N = 495 as described for LPS above) by ELISA for IgA, IgG and IgM (Fig 4). The same data are also presented as bar charts in Supporting Information S3 Fig. Overall, IgA reactivity was low among all the groups. Nevertheless, the ‘S. Typhi’ and ‘No Growth’ groups had the largest number of seropositive individuals (Fig 4A). IgA-responses to t1477 provided better discrimination between ‘S. Typhi’ and ‘NTS’ groups, although sensitivity of detection in both groups was low (detailed in the next section). By comparison, the IgG response to t1477 was elevated in all groups (Fig 4B). The highest median IgG signal was seen in the pediatric typhoid group, with the ‘No growth’ and ‘NTS’ groups having the next highest signals overall. Interestingly the Nigerian healthy children were the lowest of the Nigerian groups, although there were a number of outliers with IgG signals. IgG alone does not allow discrimination between ongoing or previous episodes of typhoid, although the negligible reactivity by this group to LPS by Ig of any isotype tested (seen in Fig 3) would support the notion the outliers with IgG responses to t1477 are convalescent cases. The IgM reactivity against t1477 reflected the IgA response. A notable exception was broadly similar levels of IgM reactivity by ‘S. Typhi’, and ‘No Growth’ groups. This contrasts with the highly significant difference seen when LPS-specific IgM was measured (Fig 3C). This appears to be caused by the reduced sensitivity for detection of typhoid by t1477-specific IgM, rather than any increase in sensitivity for detection of potential typhoid cases among the ‘No Growth’ group. As with LPS-specific IgM, t1477-specific IgM did not discriminate well between typhoid and ‘NTS’ groups, although again, this appears to be caused by the reduced sensitivity for detection of typhoid.
The accuracy of LPS and t1477 ELISAs to discriminate between Nigerian pediatric S. Typhi patients and controls were determined by ROC analysis. Plots of true positive rate (sensitivity) and false positive rate (1-specificity) for discriminating between typhoid cases and healthy children are shown for LPS and t1477 in Fig 5A and 5B, respectively. Table 3 shows corresponding percent specificity and sensitivity with either set at 90%, and areas under the curve (AUC). With LPS, IgA and IgM both gave 94% sensitivity (at fixed specificity) when used alone, which was increased slightly (to 95%) by combining the detection of IgA and IgM in the assay. Combining IgA and IgM could also be achieved in silico by summing the OD450nm data for IgA and IgM ELISAs performed individually (S4 Fig). LPS-specific IgA and IgM also give identical specificity when used alone (98% at fixed sensitivity) which was unchanged by combining both isotypes. The AUC of IgA and IgM were very similar (0.963 and 0.968, respectively) and increased slightly (0.978) after combining. Despite similar performance of IgA and IgM in the ROC analysis, the IgA ELISAs were characterized by lower backgrounds in the control groups compared to IgM, as can be seen from the raw data in S2 Fig. In contrast, LPS-specific IgG provided the lowest accuracy for distinguishing typhoid cases from healthy controls.
In the t1477 ELISAs, although AUC values of IgA and IgM were identical (0.968), IgA provided superior sensitivity than IgM (94% and 86%, respectively, at fixed specificity) and specificity (96% and 88%, respectively, at fixed sensitivity). Multiplexing IgA and IgM did not increase sensitivity or specificity over IgA alone, although there was a modest increase in AUC (to 0.976). As with LPS, t1477-specific IgG also gave lower accuracy than IgA or IgM for diagnosing acute typhoid. These data indicate both LPS and t1477-specific IgA and IgM provide good discrimination between healthy Nigerian children and those with acute typhoid fever, which is improved by detection of both IgA and IgM isotypes together.
We then compared acute typhoid with 28 Nigerian ‘other’ (non-Salmonella) infections (listed in Table 2), since this is more relevant to the diagnosis of typhoid in the clinical setting. ROC Plots are shown in Fig 5C and 5D, with corresponding AUC, and percent sensitivity and specificity given in Table 4. Here, LPS-specific IgA and IgM give comparable sensitivity when used alone (86% and 87%, respectively, at fixed specificity), which is increased to 90% when IgA and IgM are combined. LPS-specific IgM provided considerably greater specificity than IgA when used alone (82% and 75%, respectively, at fixed sensitivity), which is dramatically increased (to 96%) when combined. The AUC is also increased slightly by combining IgA and IgM to 0.938. LPS-specific IgG provides the lowest sensitivity and specificity of all three isotypes. By comparison, the relative accuracy of t1477 in ELISAs for diagnosing acute typhoid was lower for all three Ig isotypes compared to LPS, and also reduced relative to discrimination of typhoid vs. healthy controls. Unexpectedly IgG emerged as the isotype with the highest sensitivity and specificity of t1477-specific Igs. It is possible this is restricted to childhood, where there is relatively less lifetime exposure to Salmonella than older children and adults, combined with a robust IgG response during typhoid fever.
We also compared the ability of LPS and t1477 to discriminate between Nigerian pediatric typhoid and additional samples from ‘other’ non-Salmonella infections obtained from locations outside Nigeria, namely tularemia (Spain, N = 12), brucellosis (Peru, N = 16) and malaria (various sources, N = 48). ROC plots are shown in Fig 6A (LPS) and 6B (t1477), with corresponding AUC and percent sensitivity and specificity given in Table 5. Data were broadly similar to that seen with Nigerian ‘other’ infections, with combined IgA+IgM providing the most accurate test when using LPS, and IgG providing the most accurate test when using t1477. As noted earlier, brucellosis samples were prominent among the ‘other’ infections for cross-reactivity to S. Typhi LPS. If these samples were removed from the analysis (Fig 6C and 6D) there was a slight increase in sensitivity and specificity in almost all situations, with the exception of t1477-specific IgG (Table 5, values in parenthesis).
Finally we explored the effect of multiplexing LPS and t1477 antigens on the accuracy of the test for typhoid compared to each antigen alone (Fig 7 and Table 6). Multiplexing LPS IgA with t1477IgG in silico increased accuracy compared to either alone, while multiplexing LPS IgA+IgM (as mixed secondary antibodies) with t1477 IgG in silico increased the accuracy further.
In countries in Sub-Saharan Africa, where typhoid and non-typhoidal salmonellosis are major causes of bacterial sepsis in children, accurate and rapid point-of-care tests are urgently needed to replace existing diagnostic methods. Culture of S. Typhi organisms from bone marrow is the gold standard, but because it is invasive, blood culture is often a more practical, albeit less sensitive, alternative. Blood or bone marrow culture is also slow (2–3 days to arrive at a diagnosis), and empiric broad-spectrum antibiotic treatment is often initiated without a diagnosis being made.
The traditional Widal’s test, which is based on the agglutination of inactivated Salmonella Typhi and Paratyphi A organisms by antibodies to flagellin and LPS (H and O antigens, respectively) is rapid, inexpensive and requires no instrumentation. However, interpretation of the results must be made with caution. Sensitivity of the Widal’s test is lower in the early stage of infection when antibody titers are low. The test also fails to discriminate between acute from convalescent infection, leading to reduced sensitivity in endemic settings [20]. Although sensitivity can be improved if a follow-up sample is tested [21], this is not an option for rapid diagnosis. The test also lacks specificity owing to cross-reactivity with antibodies against closely-related NTS serovars [22] and other bacteria, notably Brucella [16]. Misuse of the Widal’s test has contributed to over-diagnosis of Salmonella infection, inappropriate antibiotic use, and the emergence of drug resistance [23].
Recent alternatives for serodiagnosis of typhoid include the Tubex test for LPS-specific IgM and the Typhidot test for IgG or IgM against a 50kDa outer membrane protein [24]. The Tubex test format is based on the interference by patient serum antibodies with the agglutination of latex beads coated with O9-specific monoclonal antibody and S. Typhi LPS-coated magnetic beads. The Typhidot test is a pre-dotted antigen strip. Neither test is currently configured for detection of IgA. Both have been evaluated in several Asian and African study sites; Tubex and Typhidot show comparable performance and were more specific although less sensitive than the Widal test (http://www.who.int/bulletin/volumes/89/9/11-087627/en/).
In this study we have focused on the use of LPS and t1477 (hemolysin E) as antigens to discriminate between S. Typhi infection and other bacterial infections, including commonly-encountered bacteremia seen in Nigeria. LPS has long been recognized as dominant in the response to Salmonella, while the identity of t1477 has come from studies using proteome-wide serological screens using microrrays [17, 25, 26]. Although the microarray has the potential to diagnose multiple infectious diseases on a single chip, it is currently unsuitable as a point-of-care test for many clinics in its current format. An accurate, more deployable test, particularly if configured into a format able to provide a result in <30 minutes, could help curb the inappropriate use of antibiotics and stem the rise in antibiotic resistance in Nigeria.
The data presented here indicate LPS-specific IgA (or IgA+M combined) discriminate well between Nigerian children with typhoid and healthy Nigerian children (AUC = 0.963 and 0.978, respectively; Table 3). More importantly for the clinical setting, LPS-specific IgA (or IgA+M combined) also discriminates between Nigerian children with typhoid and children with ‘other’ (non-Salmonella) infections (AUC = 0.903 and 0.938, respectively; Table 4). Similarly, discrimination between typhoid cases and healthy children using t1477-specific IgA (Table 3) was comparable to that obtained with LPS-specific IgA, although discrimination between typhoid and ‘other’ cases using t1477-specific IgA (Table 4) was far less accurate than for LPS-specific IgA. One possibility is proteins antigenically related to S. Typhi t1477 hemolysin E are found in one or more of the other bacterial infections represented in the collection (see Table 2). Such potential cross-reactivity would reduce the diagnostic utility of the antigen for typhoid.
LPS-specific IgG provided less accuracy for discriminating between Nigerian pediatric typhoid and healthy Nigerian children (Table 3), which was reduced further when discriminating typhoid with ‘other’ infections (Table 4). It is possible that IgG titers remain elevated for longer than IgA, thereby making it more difficult to discriminate between acute and previous or convalescent infections using IgG. This may be less of an issue in children where lifetime exposure to Salmonella species will likely be less than in adults. Although we have not examined Nigerian adults in this study, the expectation is they will have higher and more durable IgG titers to both LPS and t1477 than in children. This notion is supported for LPS by the pilot LPS array (Fig 2) in which all the non-Nigerian samples (i.e., panels E through K) were from adults. Thus, the median IgG signal of the healthy Nigerian children was lowest among all the groups tested, including adults from two non-endemic sites, the US (Fig 2K) and UK (Fig 2I). It remains to be determined whether LPS- and/or t1477-specific IgA has any utility for diagnosing typhoid in adults.
Unexpectedly, t1477-specific IgG performed better than LPS-specific IgG for discriminating between Nigerian pediatric typhoid and healthy Nigerian children (Table 3), and between Nigerian pediatric typhoid with ‘other’ infections (Table 4). Indeed, t1477-specific IgG also performed better than t1477-specific IgA and IgM for discriminating between typhoid and ‘other’ infections in Nigerian children. It is possible this diagnostic performance occurs only in children, where there is less lifetime exposure to Salmonella. It is anticipated that t1477-specific IgG will have less utility for diagnosing acute typhoid in older children and adults where the IgG titers from convalescent infections are likely to be much higher.
Finally we also compared the ability of LPS and t1477 to discriminate between typhoid and non-Nigerian ‘other’ infections from other locations around the world. In the ELISA, LPS-specific IgA+M provided excellent sensitivity and specificity, although we did notice detection of some Peruvian brucellosis cases using S. Typhi LPS (Fig 3 and Table 5). There are accounts in the literature of antigenic cross-reactivity between Brucella sp. and S. enterica serotype Urbana [16, 27, 28] which raises the possibility of cross-reactivity between antibodies generated during human brucellosis and Salmonella antigens. Although brucellosis is rare in Nigeria, if discrimination between acute typhoid and brucellosis is necessary, one option might be to utilize serodiagnostic B. melitensis antigens discovered previously [29–32] or B. melitensis LPS (Fig 1G) to assist in positive identification of brucellosis cases. The accuracy of t1477-specific IgA (or IgA+M) was lower than for LPS, consistent with its performance in discriminating between typhoid and Nigerian “other” infections.
LPS has received considerable interest as a potential diagnostic antigen for typhoid and for the basis of alternative assays to the Widal’s test. In one longitudinal study [33], anti-LPS IgA and IgM titers were seen to peak around d11-21 and decline thereafter, whereas IgG titers remained elevated and did not decline as rapidly. Other studies have also shown the transient nature of anti-LPS IgA in typhoid in saliva samples [34, 35] as well as in sera of gastroenteritis caused by non-typhoidal Salmonella serovars [36]. Thus, LPS-specific IgA appears to be a useful marker of acute Salmonellosis owing to its transient appearance after infection. The transient nature of IgA appears to be a peculiarity of LPS, and possibly other T-independent antigens, since serum and mucosal IgA responses to bacteria are generally long-lived [37–39].
In the present study, the LPS molecule did not discriminate well between typhoid and NTS, presumably because of the presence of shared epitopes present in the conserved lipid A and core oligosaccharide regions [40]. However, the more variant O-polysaccarides where serovar-specific epitopes of the O-antigen are located may discriminate between antibodies engendered by typhoid and NTS serovars. Salmonella O-polysaccharides have been produced from bacterial extracts and conjugated to protein carriers for use as subunit vaccines [41–43], although their utility as specific diagnostics is less well explored. In one such study, the S. Typhi O-polysaccharide O:1,9,12 performs well in IgG dot blots as a discriminator between typhoid and other acute infections or healthy controls, although IgA and the ability to discriminate between typhoid and NTS were not examined [44]. Neither the use of LPS nor measurement of IgA for diagnosis of typhoid is novel, but when used together appear to represent a good marker for acute infection in Nigerian children.
The t1477/hemolysin E (HylE, also known as cytolysin A or CylA) protein is a known dominant antigen in the antibody response to S. Typhi infection [17, 25, 26, 45]. Its utility as a potential serodiagnostic for typhoid has been demonstrated independently in a study of different Ig isotypes in 50 culture-confirmed typhoid cases [46]. In that study, IgA was the most sensitive, detecting 28/50 cases using a cut-off defined by the isotype-matched responses by healthy controls and other febrile infections. IgG was second most sensitive (19/50), and IgM least sensitive (3/50). A subsequent pilot study has demonstrated the utility of anti-HylE IgA in saliva as a biomarker for acute typhoid fever [47]. S. Typhi HylE is a 302 amino-acid long transmembrane protein with a helix hydrophobic segment located between residues 179 and 199. Along with homologs in other bacteria, such as the prototypic ClyA in E. coli, S. Typhi HylE belongs to a family of important pore-forming virulence factors of bacterial pathogens that assemble in cell membranes [48]. The HylE gene (t1477) is present in human-specific typhoid serovars (Typhi and Paratyphi) but absent from others (e.g., S. Typhimurium). In Fig 4, IgA reactivity by the 29 Nigerian children with invasive NTS (iNTS) is negligible with the exception of two outliers with low reactivity. However, sensitivity of t1477-specific IgA for detection of typhoid is also low, indicating this antigen is unlikely to have utility for discriminating iNTS and typhoid.
This study was conducted with informed consent and approved by the Ethics Committees of the Federal Capital Territory of Nigeria, Federal Medical Center Keffi, Aminu Kano Teaching Hospital and University of Nebraska Medical Center (UNMC), Omaha Institutional Review Board (IRB). We used written consent provided by parent or guardian of each child. The process was approved by both local IRB and UNMC IRB.
Sera from Spanish tularemia cases were provided by Drs. Raquel Escudero and Pedro Anda, Instituto de Salud Carlos III, Madrid, Spain. Human subjects approval from Comité de Bioética y Bienestar Animal, Instituto de Salud Carlos III (approval no. PI 33). Sera from Thai melioidosis cases were provided by Direk Limmathurotsakul and Narisara Chantratita, Mahidol University, Thailand. Ethical approval for the study was from the Ministry of Public Health, Royal Government of Thailand, and the Oxford Tropical Research Ethics Committee. Sera from Peruvian brucellosis cases were collected with human subjects approval from the Human Research Protection Committee of the University of California San Diego, the Comite de Ética of Universidad Peruana Cayetano Heredia, Lima, Peru, and the Comite de Ética of Asociación Benéfica PRISMA, Lima, Peru. Sera from Bangladeshi cholera cases were provided by Drs. Edward Ryan, Richelle Charles and Firdausi Qadri, Massachusetts General Hospital, Boston, MA. Human subjects approval by IRB protocol # 1999P009116 and International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDRB) #PR-11041. Sera from Clostridium difficile infections were collected with ethical approval from the University of Liverpool Research Ethics Committee (#08/H1005/32), and each patient provided written informed consent prior to recruitment. Malaria sera were collected with human subjects approvals from Institutional Review Boards at University Hospitals Case Medical Center and the Kenya Medical Research Institute Ethical Review Committee [49], the Medical Research Advisory Council, PNG [50], and the Ethics Committee of the Faculty of Medicine, Pharmacy, and Odonto-Stomatology and the Institutional Review Board at the National Institute of Allergy and Infectious Diseases, National Institutes of Health [51]. Sera from healthy US adults were collected under UCI IRB protocol #2007–5896. Sera were provided to the University of California Irvine (UCI) for assay without patient identifiers and were classified as exempt status by the UCI Institutional Review Board.
A retrospective study was designed using a convenience series of sera samples from Nigerian pediatric febrile cases and healthy controls, as well as other infectious diseases from other locations outside Nigeria, which were assayed by ELISA and/or LPS microarray (Table 1). The Nigerian samples were collected between 2009 and 2014 from children aged 8 months—13 years (median approximately 4 years) who presented to primary or secondary health centers in central and northwest Nigeria with an acute febrile illness and other symptoms that were suggestive of bacteremia. The duration of symptoms ranged from about 3–10 days with a median of 5 days, as documented in the clinical data captured during enrollment. S. Typhi is the leading cause of childhood bacteremia in this area [52]. Baseline demographics of this population have been described previously [52, 53]. Following informed consent from the parent or guardian, blood was obtained aseptically from a peripheral vein for blood culture and simultaneously an aliquot for serum separation was saved. Blood sampling and processing were as previously described [52, 53]. Briefly, only aerobic blood culture bottles were used and held in a Bactec 9050 incubator (Becton Dickinson, Temse, Belgium) for a maximum of 5 days. Bacteria were identified by morphology, and for Enterobacteriacae, by use of an API 20 E rapid identification system (BioMerieux, Marcy-l'Étoile, France). Bacterial isolates were stored in skimmed milk at -70°C, and further characterized at the Clinical Microbiology laboratory, University of Nebraska Medical Center. Bacteremia was defined as the isolation of at least 1 noncontaminant bacteria from the admission blood culture. These samples comprised children with typhoid (N = 86), non-typhoid Salmonella (NTS) infections (N = 29), other bacteremias (N = 28), and febrile cases that were culture negative (‘No Growth’, N = 178). Samples sizes were determined by availability during the collection period. No samples with missing or indeterminate culture test results were used in this study. In addition, we also obtained sera from healthy Nigerian children enrolled from immunization clinics in the same facilities as controls (N = 48). These children present for routine immunizations and typically are in a stable state of health. Only children who were asymptomatic and did not have a history of a febrile illness in the past month, or had taken any antibiotic during the same period, were eligible. No blood cultures were performed on the healthy controls.
For the pilot LPS array (detailed below), an expanded collection of samples from “other” control infections from other countries were tested in addition to Nigerian samples discussed above, as follows. 1) Tularemia sera (N = 12) from a 2007 Spanish outbreak of Francisella tularensis subsp. holarctica. These consisted of paired samples from 6 acute cases that were seronegative by microagglutination (MA) test at the 1st time point at presentation and which seroconverted by the 2nd time point approximately 2 weeks later. These samples were found previously to be seropositive for F. tularensis subsp. tularensis (FTT) strain Schu S4 antigens at both time points using a proteome microarray [18]. 2) Melioidosis sera from Thailand (N = 14). Samples were collected in 2004 from patients presenting with symptoms of melioidosis, and were diagnosed by indirect hemagglutination assay (IHA) and blood and throat swab culture, as described previously [54]. 3) Brucellosis sera collected prior to 2008 from an endemic region of Peru (N = 28), previously shown to be seropositive using a Brucella melitensis proteome microarray [31, 32]. Samples probed were culture positive/Rose-Bengal positive (N = 12) and culture negative/Rose Bengal positive (N = 16). These correspond to samples taken on the first day (acute infection) and within 6 weeks after obtaining the first sample (convalescent infection).4) Cholera sera from Bangladesh collected between 2008 and 2010 presenting to the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDRB) hospital with acute watery and stool culture confirmed V. cholerae O1 infection. Following informed consent, venous blood was collected from adults (age 18–55 years) at the acute phase of infection (N = 7) after clinical stabilization (day 2), and again at convalescent phases of infection (d7 and 30; N = 7). 5) Sera from Clostridium difficile infections (CDI) from diagnosed acute cases in the UK collected between 2008 and 2012 (N = 16) [55]. Each patient was followed-up for minimum period of 30 days initially and then for 1 year from notes for the collection of additional demographics clinical outcome information. 6) Symptomatic malaria cases from Kenya, Papua New Guinea and Mali (N = 48). These were diagnosed with Plasmodium falciparum parasitemia, and all defined as seropositive using different iterations of P. falciparum protein arrays derived from strain 3D7 [56]. 7) Healthy US adults from a non-endemic area (Orange County, CA), Adherence to Standards for Reporting of Diagnostic Accuracy Studies (STARD) is shown by the flowchart (S5 Fig) and checklist (S1 Text) in the Supporting Information.
Lipopolysaccharides (LPS) were obtained as follows: 1) LPS from S. Typhosa (= S. Typhi) was purchased from Sigma-Aldrich (Cat. #L2387); 2) LPS from S. Typhimurium was purchased from Sigma-Aldrich (Cat. #L6511); 3) LPS from Francisella tularensis Subsp. novicida was purified from the live vaccine strain (LVS) (DSTL batch #B07/3564), as described [57]; 4) LPS from Burkholderia pseudomallei was purified from strain K96243 (DSTL batch #B07/3558), as described [58]; 5) LPS from Brucella melitensis was purified from strain 16M, as described [31]; 6) LPS from V. cholerae O1 was purified from Ogawa (strain X-25049) and Inaba (strain T19479) serotypes, as described [59]; 7) Escherichia coli 055:B5 LPS was purchased from Sigma-Aldrich (Cat. #L2880). Each LPS species was diluted in PBS buffer, pH 7.3–7.5 (EMD Millipore Corp., Billerica, MA; Cat. #6506-OP) and printed on nitrocellulose-coated glass slides (Oncyte Avid from Grace Bio-Labs, Bend, OR) using a GeneMachines Omnigrid 100 array printer, and printed at a concentration of 0.1 μg/ml. This concentration was determined previously in titration experiments to be the lowest concentration able to provide near maximal signals. Performers of the LPS microarray assays were blinded to the identity of the samples until after the assays were completed. LPS arrays were probed for 18h at 4°C with sera diluted 1/100 in protein microarray blocking buffer (Maine Manufacturing, GVS North America, Sanford, ME) supplemented with E. coli lysate (Antigen Discovery Inc, Irvine, CA). Bound IgG and IgA were then detected using secondary antibodies conjugated to biotin followed by streptavidin conjugated to quantum dots, and then visualized in an ArrayCAMarray imager, as described previously [19].
Hemolysin E protein (HylE, gene t1477 from S. Typhi Ty2 strain) was expressed in E. coli and purified as described previously [17]. LPS from S. Typhi was as described above for microarrays. ELISAs were performed as described [60]. Briefly, antigens were coated onto microtiter plates (ThermoScientific, Walham, MA) at concentrations 1.25 μg/ml (LPS) and 2.5 μg/ml (HylE) in TBS (100μl/well) overnight at 4°C. The coating concentrations were determined previously for each antigen by serial dilution experiments. The following day, plates were washed 4 times in 1x TBS containing 0.05% Tween20 (T-TBS; ThermoScientific) and blocked with casein/TBS blocking buffer (ThermoScientific) for 1-2h (300 μl/well). Blocking buffer was then decanted, and the plates air-dried and stored in desiccated foil pouches at 4°C until required for use. Performers of the ELISAs were blinded to the identity of the samples until after the assays were completed. For ELISA assay, sera were diluted to 1/200 (LPS) and 1/100 (HylE) in casein/TBS blocking buffer containing E. coli lysate (GenScript, Piscataway, NJ) at 1.5 mg/ml final concentration, and incubated for 30 min prior to placing into the plates. Plates were incubated for 45 min with gentle rocking at room temperature (RT). After washing with T-TBS goat anti-human IgG-, IgA- or IgM-HRP conjugates (Bethyl Laboratories, Inc., Montgomery, TX) diluted 1/25,000 (IgG) or 1/12,500 (IgA, IgM) in Guardian Stabilizer (ThermoScientific) were added to wells (100 μl /well) and incubated for 45 min at RT (100 μl/well). After washing with T-TBS, plates were developed by adding 100 μl/well SureBlueReserve TMB developer (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, MD) for 10 min in the dark. Development was stopped by addition of 100 μl/well of 0.2M H2SO4 and OD read at 450 nm in a Multiskan FC plate reader.
ELISA data were collected at OD450nm and data were corrected by the positive control between runs. Dot plots and comparisons between medians of different groups using the Wilcoxon method, were produced in JMP (SAS Institute, Inc., Cary, NC, USA). Receiver operator characteristic (ROC) analyses were performed between patient groups for each antigen with a varying threshold cut off in the R statistical environment using ROCR. Plots of false positive vs. true positive plots were made, from which areas under the curve (AUC) and sensitivity and 1-specificity values were calculated for each antigen(s).
|
10.1371/journal.pntd.0005564 | Health beliefs of school-age rural children in podoconiosis-affected families: A qualitative study in Southern Ethiopia | Several studies have suggested investigation of health beliefs in children to be an important pre-condition for primary prevention of disease. However, little effort has been made to understand these in the context of podoconiosis. This study therefore aimed to explore the health beliefs of school-age rural children in podoconiosis-affected families.
A cross sectional qualitative study was conducted in March 2016 in Wolaita Zone, Southern Ethiopia. Data were collected through in-depth individual interviews (IDIs) and focus group discussions (FGDs), with a total of one hundred seventeen 9 to15-year-old children recruited from podoconiosis affected families. The study revealed various misconceptions regarding risk factors for podoconiosis. Most children believed barefoot exposure to dew, worms, snake bite, frog urine, other forms of poison, and contact with affected people to be major causes of the disease. Their knowledge about the role of heredity and that of long term barefoot exposure to irritant mineral particles was also weak. Though most participants correctly appraised their susceptibility to podoconiosis in relation to regular use of footwear and foot hygiene, others based their risk perceptions on factors they think beyond their control. They described several barriers to preventive behaviour, including uncomfortable footwear, shortage and poor adaptability of footwear for farm activities and sports, and shortage of soap for washing. Children also perceived low self-efficacy to practice preventive behaviour in spite of the barriers.
Health education interventions may enhance school-age children’s health literacy and be translated to preventive action. Overcoming practical challenges such as shortage of footwear and other hygiene facilities requires other forms of interventions such as livelihood strengthening activities. Linking podoconiosis-affected families with local governmental or non-governmental organizations providing socio-economic support for households may assist school-age children in those families to sustainably engage in preventive behaviours.
| Podoconiosis is an example of a lifestyle-related disease that develops after childhood and affects millions of people with little experience of preventive behaviour. It is caused by prolonged barefoot exposure to irritant mineral particles, accompanied by inherited susceptibility. Children with a family history of podoconiosis are highly vulnerable to the disease. Several studies have underscored the importance of studying health beliefs in children for the early control of diseases that arise from risky behaviour and habits established in childhood, which continue into adulthood. This study attempted to explore the health beliefs of school-age children in podoconiosis affected families. The forms of health beliefs addressed in the present study were knowledge of podoconiosis risk factors and perceptions of severity of and susceptibility to the disease, benefits of and barriers to engaging in preventive actions such as footwear use and feet hygiene, and self-efficacy to perform preventive actions regardless of perceived barriers. Various forms of misconceptions and inaccurate risk perceptions were recorded. Participant children also reported several barriers that limited their engagement in preventive actions and low confidence to overcome them. These findings imply that school-age children at high risk of podoconiosis might benefit from an intervention that improves their knowledge about podoconiosis and enhances their self-efficacy to sustainably perform preventive behaviours.
| Podoconiosis is an example of a lifestyle-related disease that develops later in life and affects millions of people with little experience of preventive behaviour. It is non-infectious (and thus also termed ‘non-filarial elephantiasis’) and is characterized by bilateral swelling of the lower legs, commonly affecting people in the economically productive age groups [1,2]. In Ethiopia, over 1.5 million people are believed to live with podoconiosis [3]. Evidence to date indicates that the combination of inherited genetic susceptibility and barefoot exposure to soil rich in irritant mineral particles contributes to the cause of podoconiosis [2,4]. An estimate of heritability of podoconiosis is 63% while the risk ratio of siblings in affected families is 5 times higher than their counterparts in the general population [2]. Luckily, genetically susceptible individuals can entirely prevent the disease if they consistently protect their feet from exposure to irritant particles by wearing shoes starting at young age [5]. However, few children in podoconiosis-affected families engage in preventive behaviours such as regular use of footwear and foot hygiene in spite of their higher susceptibility to the disease. In the most recent study in an endemic setting in Ethiopia, the proportion of preschool children reported to have “all day, every day” use of footwear was only 31% [6]. Another study also reported poor hygiene among children [7].
Previous studies among adults in communities endemic for podoconiosis have reported higher level of misconceptions regarding the cause and prevention of podoconiosis [6–11], and discussed the implications of the misconceptions to disease prevention behaviour and interpersonal interactions [6]. The beliefs that podoconiosis is contagious, caused by worms in the soil, indiscriminately inherited among relatives, caused by evil eye, curse, witch, or cold weather [8,9] were found to have negative consequences on preventive behavioural choices and interpersonal interactions [10]. The perceptions of adults regarding their own and children’s susceptibility to the disease were also reported to be inaccurate [6,11]. The perceptions that footwear does not permit farm activities and other duties, is uncomfortable for walking in the mud, smells bad in the hot season, wears out too quickly, softens the feet, and should be preserved for special events have all been identified as factors discouraging optimum use of footwear among people at high risk for the disease [11,12]. However, most of these studies focused only on adults. The studies that have investigated preventive behaviour among children [6,7,13] have explained it based on the parents’ health beliefs. Children are perceived as “active, purposeful beings who make sense of their world and contribute substantially to their own development” [14], and whose cognitive developments occur intensively within the age of 7–15 years [15–17]. Researchers have acknowledged increasing levels of social autonomy of school-age children as they spend more time away from home with less parental supervision. This gives them the chance to develop independent beliefs about health [18].
Several studies have underscored the importance of investigating the dimensions of health beliefs in school-age children, particularly for control and prevention of diseases that arise from behaviour and habits established in childhood and continue to adult life [18, 19–23]. The formation of values and behaviour in early childhood necessitates understanding of health beliefs of children [19]. This is supported by other studies which argue for the establishment of accurate beliefs about health in early childhood as habits in childhood are predictive of habits in adulthood [20–22]. Investigating health beliefs in children is also thought to enable better understanding of the impact of health education on the modification of health beliefs and encouragement of preventive behaviour [18,19]. Knowledge of the health beliefs of school-age children can be used to engage them as health messengers to their families and peers. A growing body of thought supports the belief that school-age children are not just passive recipients of health information. Rather, they can act as change agents who positively influence the behaviour of others in their communities through communicating health messages [24, 25]. Investigating the health beliefs of school-age children may not only help promotion of footwear use for preventing podoconiosis, but also prevention of other neglected tropical diseases contracted through the feet. To our knowledge, no studies have actively involved children of this age group in the study of their health beliefs in the context of podoconiosis. The main purpose of the present study is therefore to explore various forms of health beliefs in school-age children.
Ethical approval was obtained from the ethics committee of the Armauer Hansen Research Institute (AHRI) (Project reg. No. P035/15) and College of Health Sciences, Addis Ababa University (Protocol number 047/15/Ext). The Wolaita Zone Administrative Bureau gave written permission to work in the community. The Mossy Foot International also allowed their outreach clinic site staff to help in the identification of study participants. Individual participation of children and group interviews was guided by the international guidelines for involving children in research [26–28]. A developmental psychologist helped during the recruitment and interviews to make sure that the content and format of questions was appropriate to the age and cognitive level of the child. Informed consent was obtained from caregivers and children also gave their assent to participate in the study. Taking into account likely difficulties understanding written consent forms due to low literacy [29], a conversational style oral presentation of consent information was made in the local language to caregivers and children. Caregivers confirmed their permission for a child to participate in the study by signing or thumb-printing on the consent forms. Children expressed their assent verbally in the presence of their caregivers as a witness, to ensure the assent process is without any coercion. The use of verbal assent to children was approved by the ethics committee.
The study was conducted in Wolaita Zone, one of the thirteen zones in Southern Nation Nationalities and Peoples Regional State (SNNPRS). Wolaita Zone is located in the south-west of Ethiopia roughly between 6.30–7.10 N and 37.10–38.10 E, latitude and longitude respectively (Fig 1). According to the 2007 census report, the total population of the area was around 1.7 million; of whom 83.2% resided in rural areas. The dominant means of living is subsistence agriculture [30]. In this Zone, the point prevalence of podoconiosis has been calculated at 5.46% [31]. Since 1998, Mossy Foot International (MFI) (formerly, the Mossy Foot Treatment and Prevention Association), an international non-governmental organization, has been offering community-based prevention and control activities against podoconiosis in 15 outreach sites located at 15 to 65 km from the head office in Wolaita Sodo. Clinics in all outreach sites are run by community podoconiosis agents (CPAs) who are themselves patients, and social workers recruited from the local community. The MFI reaches the wider community through a group of network members who provide voluntary services of awareness and demand creation in collaboration with site workers. The organization has been serving over 30,000 registered patients for over a decade [32].
A cross sectional qualitative study was conducted in March, 2016 using in-depth individual interviews (IDIs) and focus group discussion (FGDs) methods. A purposive sampling technique was used to select three study sites with large numbers of registered patients and a relatively long history of establishment. The selected sites were Damot Pulasa Woreda (district), Boloso Sore Woreda and Ofa Woreda. Study site staff members helped identify affected families and children eligible for interviews in those families. A theoretical sampling technique was used to determine the number of participants in the study, i.e. the process of sampling that continues until theoretical saturation is reached [33]. The major inclusion criteria for children were being part of a podoconiosis-affected family and age 9–15 years. Podoconiosis-affected families having at least one child between the age of 9–15 years were identified through the MFI site workers. One child per household was selected either for IDI or for FGD. None of the selected child participated in both activities. Children affected by podoconiosis or other forms of physical impairment were excluded from the study. Focus group discussions were disaggregated by gender: three with boys and three with girls. Twelve children participated in each FGD, giving a total of 72 participants. A total of 45 IDIs were held with children: 15 in each of the three sites. Together, 117 children participated in individual and group interviews.
Semi-structured questions were developed in English and translated into the local language, Wolaitatto (AT is the native speaker). Children were interviewed individually during home visits while FGDs were held in a place proximate for children coming from surrounding villages. All interviews were digitally recorded. Before assessing the beliefs of children about podoconiosis in both IDIs and FGDs, the mental image of children regarding their recognition of podoconiosis as a disease and its manifestations was assessed through drawings and their verbal description of the pictures they drew. Drawing exercises have been suggested as a tool to understand children’s imagery of disease and to assess their knowledge and conceptions [34,35]. To prompt recalling, the interviewer used the local term “na’u gediya kitisiya hargiya”, literally “disease that causes bilateral swelling of feet”. This term was used instead of common local terms such as “Kita”, “Inchricha” [8], which were reported as derogatory and entailing demeaning connotation against affected persons [29]. Participant children were asked about their thoughts regarding the cause of podoconiosis, their perceptions of severity of disease, their appraisal of susceptibility to the disease, the advantages and disadvantages of various types of footwear, their perceptions of barriers to regular use of footwear and their self-efficacy beliefs to use footwear regularly in spite of these barriers. In the middle or at the end of FGDs, role plays were performed by children, which boosted their confidence to talk openly what goes on within the family and in the community. Whenever children found a given question difficult to understand, guidance and repeated clarifications were used to facilitate responses. While individual interviews lasted a maximum of 30 minutes, it took 1 hour to complete each FGD. At the end of the interviews, participants were provided brief information about the causes of podoconiosis and the role of consistent use of footwear in preventing the disease. After every interview, children received a piece of soap, a pen and a note book as compensation for their time, and this was suggested in a previous study as ethically appropriate if made in consultation with community members in the study setting [29].
Data were transcribed and translated into English and imported to NVivo software version 11 for analysis. Both deductive and inductive approaches were used to analyse the data. Deductive coding of themes in the data was based on Health Belief model constructs such as knowledge, perceived severity, perceived susceptibility, perceived benefits and barriers, and perceived self-efficacy. The Health Belief Model (HBM) provides a useful framework to understand and explain health beliefs in association with disease preventive behaviour [36–40]. A number of studies have used the health belief model to explain health beliefs of young and adolescent children [18,19,41,42].A grounded theory approach was used to inductively identify newly emergent themes and subthemes as coding process proceeded. Grounded theory refers to a set of integrated and inductively generated concepts, categories, and themes that are formulated into a logical, systematic and explanatory scheme [33].
Children were asked if they thought they might or might not be at risk. Some provided logical reasons for considering themselves at risk considering exposure to risk factors in the environment and lack of preventive actions taken to reduce their exposure to these risk factors. They believed that they would be at risk if they went barefoot and were aware that their exposure could be reversed through preventive actions. If I walk barefoot and simply step on harmful things, I think my feet may also swell like that of my father. As a result, whatever they complain about poverty, I urge them to buy shoes for me. (FGD participant, boy, 13 years, grade 4)
On the other hand, misperceptions around personal susceptibility to podoconiosis were also common. Some children thought that they were at risk based on exposure to factors that are not recognized to cause podoconiosis such as exposure to snake bite and sharing of contaminated water.
Some children never thought about the risk of getting podoconiosis. This is mainly because they related occurrence of disease to ‘God’s plan’.
Some children thought that they were not at risk and not worried of getting the disease because their feet were healthy at the time of interview. When a boy was asked whether he had ever thought about getting podoconiosis, he replied “no, I have never thought about getting foot swelling as my feet are healthy now. (IDI participant, boy, 12 years, grade 2)
Participants in all individual and group interviews were aware of the negative consequences of the disease on affected individuals and their families. They noted that podoconiosis “causes illness and results in unexpected medical expenses”, “limits capacity to work”, “embarrassing”, “exposes family of affected person to starvation”, “causes illness and makes them bedridden due to swelling in groins”, “limits their ability to walk or run fast”, and “dissociates them from other people because of the bad smell and flies collected on wounds of their feet”. More importantly, they also revealed the impact of their parents’ condition on their own life, citing stress and starvation because of their parents’ illness.
The participants held a positive outlook towards using footwear. They stated that using footwear protects the feet from dust, snakes, poison, dew, chilly weather, injuries by sharp things.
Children also clearly stated the importance of foot hygiene in preventing podoconiosis. As a FGD participant states, “if people wash feet regularly, before and after wearing shoes, they cannot get the disease” (FGD participant, boy, 12 years, grade 4).
The perceived discomfort of footwear, particularly in hot weather, was commonly thought to be a barrier to regular use. Children associated closed rather than open shoes with discomfort in hot weather. They repeatedly stated that using closed footwear in hot weather caused nail dystrophy and ‘mich’.
On the other hand, girls tended to perceive the disadvantages of both open and closed footwear. They thought that the regular use of sandals caused heel fissures while the regular use of closed plastic boots damaged the toenails. They recommended changing between different types of footwear for better foot health.
Most children also said that they felt uncomfortable using any type of footwear when they engaged in farming activities.
Children also said it was difficult to do sport wearing shoes. While some were worried about shortening of the shoes’ life, others were more concerned about injuring themselves.
Another problem was the limited number of pairs of footwear they owned. Those who had a single pair of shoes kept them for special occasions such as school or church. Some children were concerned about the economic situation of their parents and refrained from wearing shoes regularly so as to preserve the ones they had. The poor quality of shoes purchased for children was also reported to negatively influence the motivation of children to use them regularly.
Regarding foot washing, most children indicated that they washed their feet before putting on shoes, before going to bed in the evening and after performing certain activities barefoot. Almost all of the children reported that they washed their feet at least once a day. They did not perceive barriers to washing their feet regularly, except discipline or consistency. The availability of water nearby and the encouragement of parents were reported to be enabling factors. However, when observed, the participant children’s feet were not as clean as expected since they rarely used soap.
As discussed earlier, because of the discomfort of wearing closed shoes for farm activities and hot weather, children tended to instead wear either open (less protective) types of footwear or to walk barefoot. They were asked if they thought they could use closed footwear for all conditions regardless of the perceived challenges. Most said they could not wear closed shoes while farming because they were too heavy and cause bad smell.
On the other hand, for some participants, self-efficacy depended on the type of closed footwear. For instance, most children thought that they could use canvas or ‘sieve’ plastic boots (with ventilation holes) in all seasons or times of day. They thought that these boots let air circulate around the feet in hot weather and are comfortable for cold weather since they cover the whole foot. Canvas shoes were also thought to be good under all conditions.
Using the Health Belief Model, this study attempted to explore children’s knowledge regarding the cause and prevention of podoconiosis, their perceptions about susceptibility to and severity of the disease, the benefits of and barriers to engaging in preventive behaviour and their self-efficacy to perform preventive behaviour in the face of perceived challenges.
Participant children were found to have intimate knowledge of podoconiosis symptoms, which they expressed best in self-drawn pictures and verbal descriptions. The use of a simple term in the local language that is equivalent to the scientific name of the disease also helped children distinguish the attributes of the disease from other related diseases. However, some younger children, and children with fully-recovered family members or those at the early stages of the disease struggled to name podoconiosis in the local language and consequently faced difficulties drawing pictures. There were also some participants who were too shy to draw pictures and thought they lacked the skill to do so. This implies that using illustrative pictures or photos of the disease in addition to local language is necessary before proceeding to other activities such as interviews with children.
Various understandings regarding the cause and prevention of podoconiosis were apparent in interviews. The most important domain in which children’s knowledge was assessed was their understanding of behavioral and environmental risk factors. Children mentioned exposure to several environmental factors such as snakes, frogs, worms, germs, cold weather, injuries by sharp objects and rusty metals, etc. as causes of the disease. Other kinds of behaviour such as sharing shoes and household equipment, insulting affected people or jumping over a podoconiosis-affected leg were also mentioned as causes of the disease. Barefoot exposure to soil was also raised by children as a cause of podoconiosis though the role of mineral particles in the soil was mentioned infrequently. This is congruent with previous studies which reported poor knowledge of adult community members about mineral particles in the soil as a causal agent of podoconiosis [8–10,12,43,44]. Regarding preventive measures, children were well aware of the importance of footwear and foot hygiene. A previous study in Wolaita also reported large numbers of community members believing that podoconiosis was preventable through good personal hygiene and wearing shoes [8]. However, as children also endorsed contagion and other behaviour as causes, they may not intend to use footwear and keep feet hygiene and prioritize these over other actions such as avoiding contact with affected persons.
The other important domain in which children’s knowledge of podoconiosis cause and prevention were assessed was their beliefs about heredity. Unlike adults in the same study setting who immediately mentioned heredity in relation to podoconiosis in interviews [8,10], children under the age of 15 were less likely to have ideas about the issue of heredity in their short memory. It was only through further probing questions that children were able to discuss heredity in the present study. The use of questions about family clustering of the disease and its disproportionate prevalence in the community helped identify the children’s intuitive understanding of heredity in association with podoconiosis. These questions revealed misinterpretations of heredity among children. Though some children believed that heritability could be mitigated through regular use of footwear and foot hygiene, they had no clear conception of what actually was inherited, mentioning attributes like ‘tumor in the blood line’. Others confused heredity with contagion. Similar misconceptions about heredity have been reported in a previous study [10].
Studies have underlined the importance of accurate understanding of heredity and its link with mineral particles in the soil for self-motivated and self-regulated preventive actions among people at high risk for podoconiosis [10–13]. Inaccurate understandings of the concept of heredity and its role in the cause of the disease contribute to the perceived inevitability of podoconiosis, which is reflected in endorsement of risky behaviors [10]. The other form of inaccurate understanding of the role of heredity is an absolute denial of its role, which is reflected in the endorsement of environmental determinism [10,44]. However, a health education intervention with culturally and linguistically appropriate genetics information for adults suggests the possibility of enhancing genetic literacy in low income rural settings [44]. School age children may also benefit from similar health education interventions tailored to their cognitive scope.
In the study setting, there are limited circumstances in which children get information about podoconiosis in schools. They construct knowledge of podoconiosis through communication with their parents or observation of illness experiences of affected family members. As most adults in podoconiosis-endemic communities were reported to hold high levels of misconceptions about the disease [10], children may learn these misconceptions from their parents. If not addressed, this may contribute to the intergenerational perpetuation of misconceptions about the disease in communities highly endemic for the disease.
The misconceptions children held about the risk factors of podoconiosis were also reflected in their appraisal of their susceptibility to the disease. Some children believed that they were not at risk, because their feet were healthy at the time of interview, which shows lack of understanding of the onset of the disease. Others associated the risk of getting podoconiosis with bad luck or predisposition, which may also limit their motivation to engage in or maintain preventive action. Such a perception was also reported as common among adults in podoconiosis-endemic communities and acted as a barrier to preventive actions against podoconiosis [11,12]. As children associate their perception of susceptibility to podoconiosis with exposure to poison in the soil, and contact with and insults against affected people, they may consider their exposure to these risk factors beyond their control. This may limit their commitment to engage sustainably in preventive behaviours. On the other hand, though not realistic in the case of podoconiosis, the children’s worries related to exposure to other harmful things such as snakes in the environment may be used in framing health messages. This may motivate them to use protective footwear not only to prevent podoconiosis, but also several barefoot related neglected tropical diseases affecting children of the school age.
Children perceived various benefits of using footwear and practicing foot hygiene, including prevention of podoconiosis. However, they perceived several barriers including that footwear was uncomfortable, particularly under circumstances like hot weather, farming and sports activities. Similar barriers have also been reported in previous studies on adults [11,12]. Children considered closed shoes to be cumbersome, potentially smelly and likely to cause nail dystrophy. They related problems concerning footwear not only with the number of pairs of footwear they owned, but also the adaptability and acceptability of footwear they own. Children with multiple pairs of footwear tended to report stronger self-efficacy beliefs to use footwear in all conditions of daily life. Financial constraints were reported to be the major obstacles limiting parents’ capacity to provide acceptable and adaptable shoes [12]. The effect of poverty was also reflected in the children’s use of soap for washing their feet, forcing them to prioritize soap for washing the face and clothes. Hence, these practical challenges should be considered together with their misconceptions about the disease during health promotion activities.
To our knowledge, this study is the first to address health beliefs of school-age rural children in podoconiosis-affected families. We have explored dimensions of health beliefs that are important to public health intervention in relation to this condition and other Neglected Tropical Diseases. However, the findings of this study may not be generalizable to other settings as the participants belong to a homogenous cultural background. In addition, the present study emphasized exploration of the school-age children’s health beliefs. Further study is required to examine the interplay between school-age children’s health beliefs and the socioeconomic circumstances of their families.
In conclusion, children held misconceptions regarding behavioral, environmental and genetic risk factors. They also perceived obstacles that threaten their ability to engage in preventive behaviors. Health education interventions may enhance school-age children’s health literacy and be translated into preventive action. Overcoming practical challenges such as shortage of footwear and other hygiene facilities requires other forms of interventions such as livelihood-strengthening activities. Linking podoconiosis-affected families with local governmental or non-governmental organizations providing socio-economic support for households may assist school-age children in those families to sustainably engage in preventive behaviors.
|
10.1371/journal.pgen.1004995 | Fatty Acid Synthase Cooperates with Glyoxalase 1 to Protect against Sugar Toxicity | Fatty acid (FA) metabolism is deregulated in several human diseases including metabolic syndrome, type 2 diabetes and cancers. Therefore, FA-metabolic enzymes are potential targets for drug therapy, although the consequence of these treatments must be precisely evaluated at the organismal and cellular levels. In healthy organism, synthesis of triacylglycerols (TAGs)—composed of three FA units esterified to a glycerol backbone—is increased in response to dietary sugar. Saturation in the storage and synthesis capacity of TAGs is associated with type 2 diabetes progression. Sugar toxicity likely depends on advanced-glycation-end-products (AGEs) that form through covalent bounding between amine groups and carbonyl groups of sugar or their derivatives α-oxoaldehydes. Methylglyoxal (MG) is a highly reactive α-oxoaldehyde that is derived from glycolysis through a non-enzymatic reaction. Glyoxalase 1 (Glo1) works to neutralize MG, reducing its deleterious effects. Here, we have used the power of Drosophila genetics to generate Fatty acid synthase (FASN) mutants, allowing us to investigate the consequence of this deficiency upon sugar-supplemented diets. We found that FASN mutants are lethal but can be rescued by an appropriate lipid diet. Rescued animals do not exhibit insulin resistance, are dramatically sensitive to dietary sugar and accumulate AGEs. We show that FASN and Glo1 cooperate at systemic and cell-autonomous levels to protect against sugar toxicity. We observed that the size of FASN mutant cells decreases as dietary sucrose increases. Genetic interactions at the cell-autonomous level, where glycolytic enzymes or Glo1 were manipulated in FASN mutant cells, revealed that this sugar-dependent size reduction is a direct consequence of MG-derived-AGE accumulation. In summary, our findings indicate that FASN is dispensable for cell growth if extracellular lipids are available. In contrast, FA-synthesis appears to be required to limit a cell-autonomous accumulation of MG-derived-AGEs, supporting the notion that MG is the most deleterious α-oxoaldehyde at the intracellular level.
| Consumption of sugar and lipid (fat) enriched food increases the risk of developing metabolic diseases and cancers. However, lipids are essential molecules for life, as they are the major components of cell membranes. Metabolism refers to biochemical reactions that transform nutrients into molecules required by an organism, although toxic by-products can also formed. Sugars or their derivatives are likely to induce toxic effects by forming stable conjugates with proteins. To neutralize their toxic potential, sugars are metabolized and stored as fat. Here, we have used the fruitfly model to investigate the consequences of lipogenesis deficiency upon ingestion of sugar-enriched diets. We show that lipogenesis deficient animals are dramatically sensitive to dietary sugar. Further, we have identified the sugar by-product responsible for intracellular toxicity, in the context of lipogenesis inhibition. Our study reveals that inhibiting lipogenesis does not disrupt cellular growth if extracellular lipids are available. In contrast lipogenesis inhibition may have deleterious consequences due to accumulation of toxic by-products. The efficacy of lipogenic inhibitors in fighting cancers and metabolic diseases is currently under investigation. Therefore, to evaluate the clinical benefit of these inhibitors, accumulation of the toxic molecules should be monitored in both sick and healthy cells.
| Deregulation of metabolism occurs in several pandemic human diseases whose incidence has dramatically increased due to changes in lifestyle and extended lifespan. These disorders include metabolic syndrome and type 2 diabetes (T2D) that are typified by insulin resistance and elevated levels of glucose and triacylglycerols (TAGs) in the plasma [1,2]. However, insulin resistance does not directly depend on an increase in TAG levels, but is rather a consequence of diacylglycerol and/or ceramides accumulation [1,3,4], whose levels increase as adipose tissue reaches a saturating point [5,6]. Cancer cells also exhibit metabolic perturbations characterized in part by a dramatic increase in glycolysis and fatty acid (FA) synthesis [7,8]. These changes emphasize direct links between sugar catabolism and FA synthesis.
Recent studies support the notion that glycation of proteins, DNA and/or phospholipids is likely to be responsible for the toxic effects induced by excess sugar [9,10]. The resulting compounds, advanced-glycation-end-products (AGEs), maybe responsible for vascular complication, nephropathy and retinal degeneration in T2D patients [11,12]. Glycation is a spontaneous reaction that occurs between an amine group and a carbonyl group of sugars or α-oxoaldehydes [13]. The latter include methylglyoxal (MG) that largely derives from spontaneous oxidation of the glycolytic intermediates dihydroxyacetone-phosphate (DHAP) and glyceraldhehyde-3-phosphate (G3P) [14]. The glyoxalase system [15], an enzymatic system composed of glyoxalase 1 (Glo1) and glyoxalase 2, maintains tolerable levels of MG.
In healthy organisms, circulating glucose is taken up by cells and is used to produce energy through glycolysis and the citric acid cycle. In postprandial condition, dietary glucose is used to synthesize glycogen in the liver and muscles. Excess glucose is also used for FA synthesis in hepatocytes and adipocytes. Synthesis of FA first requires carboxylation of acetyl-CoA to malonyl-CoA by the enzyme ACC (Acetyl-CoA carboxylase) [16]. Next, the Fatty acid synthase (FASN according to the current mammalian nomenclature) sequentially incorporates several malonyl-CoA molecules onto an acetyl-CoA primer to form a long chain FA (LCFA) [17].
Drosophila genetics has proven a powerful model system to investigate metabolic regulation at the level of the organism [18,19,20]. We previously demonstrated that in larvae, ACC is cell-autonomously required for the synthesis and storage of TAGs in the fat body (FB) [21], an insect organ with hepatic and adipose functions. We also provided evidence that within the oenocytes—abdominal cells with a hepatic-like function [22]—ACC is required to maintain the watertightness of the tracheal system [21].
Here, we have focused on the Drosophila FASN orthologs, of which only one (FASNCG3523) is ubiquitously expressed. By directing inducible RNA-interfering (RNAi) to FASNCG3523 and glycogen synthase (GlyS), we observed that the larval FB synthesizes both TAGs and glycogen. Next, we observed that expression of FASNCG3523 is induced by dietary sugar and that FASNCG3523 deficient animals are extremely sensitive to moderate increases in dietary sugar. Furthermore, we provide evidence that the activity of FASN and Glo1 cooperate both at the organismal and cellular level to protect against sugar toxicity.
To investigate the physiological consequences of FA synthesis defect in Drosophila, we focused on the ortholog of the anabolic enzyme FASN, encoded by three distinct genes (FASNCG3523, FASNCG3524, FASNCG17374) [21]. Previous reports show that in larval tissues, FASNCG3523 is ubiquitously expressed, while FASNCG3524 and FASNCG17374 are mostly expressed in the carcass, which is comprised of epidermal cells, oenocytes and skeletal muscles [23]. To corroborate these findings, transcript levels of the three FASN genes were monitored using quantitative-PCR (RT-Q-PCR) in third stage larvae (L3) separated in two fractions, the internal organs, which can be easily removed and the leftover carcass. Consistently, FASNCG3523 transcripts were detected at high levels in both fractions, whereas FASNCG3524 and FASNCG17374 transcripts were detected at high levels in the carcass, but minimally in the internal organs (S1A Fig.).
To determine whether these enzymes are essential, we made use of the binary Gal4/UAS system to direct specific RNAi to each FASN gene and to the ACC orthologue [21]. Ubiquitous knockdown of these genes caused lethality at late embryogenesis for ACC, at L1 stage for FASNCG3523, at the L2 stage for FASNCG17374, while no phenotype was observed for FASNCG3524 (Table 1). The lethality at the L2 stage observed in FASNCG17374-RNAi knockdown resembled the phenotype previously observed when inducing this RNAi using an oenocyte specific driver [21], typified by a defect in the watertightness of the tracheal system (S1B–S1C Fig.). Therefore, we have used a svp-gal80 transgene to inhibit Gal4 in the oenocytes [22]. Driving FASNCG17374-RNAi in the entire animal, except in the oenocytes, resulted in a total rescue of the lethal phenotype (Table 1 and S1D Fig.), indicating that FASNCG17374 does not serve an essential function in other tissues. To get further insights into the organ-specific function of these enzymes, we used the Cg-gal4 and Mef2-gal4 drivers, which are specific to the FB and the muscles, respectively. When induced in either tissue, knockdown to any of either gene did not affect viability (Table 1). Nonetheless, muscle knockdown of ACC or FASNCG3523, but not of FASNCG3524 or FASNCG17374, led to a motility defect in adult flies (Table 1). Taken together, these findings indicate that the synthesis of LCFA is not essential in either the FB or the muscles. However, consistent with previous studies reporting that muscle-specific knockdown of ACC affects body homeostasis and motility of adult flies [24,25], our findings indicate that FA synthesis plays an important role in muscle development and/or activity.
We previously reported that FB-knockdown of ACC results in a decrease in total TAG levels [21]. To determine which of the three FASN members is necessary for LCFA synthesis in the FB, RNAi to each FASN gene was induced using the FB-specific driver. Consistent with the finding that FASNCG3523 is the only FASN gene expressed in internal organs (S1A Fig.), total TAG levels were dramatically reduced in FASNCG3523 but not in FASNCG17374 and FASNCG3524 knockdowns (S2A Fig.).
We previously observed that in Cg>ACC-RNAi (Cg-gal4 directing ACC-RNAi) animals the drop in whole larvae TAG levels was accompanied by an increase in glycogen storage [21]. Thus, to investigate the physiological relationship between TAG and glycogen storage, RNAi transgenes to either ACC or FASNCG3523 was combined with an RNAi transgene to the gene encoding the unique Drosophila GlyS. Single or dual knockdowns were induced in either the FB, the muscles or in both tissues. Total amounts of TAG, glycogen, trehalose, glucose and protein were measured in 0–5h prepupae, as this is a convenient phase to stage the animals after the feeding period. Prepupal weighing revealed that animals expressing GlyS-RNAi in combination with either ACC-RNAi or FASNCG3523-RNAi in both the muscles and the FB exhibited the most prominent reduction in body weight (S2B Fig. and S1 Table). Total TAG levels decreased dramatically when either ACC-RNAi or FASNCG3523-RNAi were induced in the FB but not in the muscles (S2C Fig.); this decrease was roughly similar when either RNAi were induced in both tissues (S2C Fig.). Furthermore, TAG levels measured in control, ACC-RNAi- or FASNCG3523-RNAi-expressing animals were not markedly modified by the expression of the GlyS-RNAi (S2C Fig.). Total glycogen levels decreased when GlyS-RNAi was expressed in either the FB or the muscles, and decreased further when GlyS-RNAi was expressed in both tissues (S2D Fig.), indicating that in prepupae both organs contribute to glycogen storage. This finding contrasts with a previous study reporting that in larvae, glycogen can be detected in skeletal muscles only [26]. Therefore, since glycogen is unlikely to be transported between organs, it is conceivable that FB glycogen synthesis mostly occurs at late larval stage. Unexpectedly, driving either ACC-RNAi or FASNCG3523-RNAi in the muscles provoked a moderate decrease in glycogen levels (S2D Fig.). This may be a consequence of muscle dysfunction linked to the above mentioned motility defect (Table 1). Importantly, FB-knockdown to ACC or FASNCG3523 provoked a very strong increase in total glycogen levels that is not observed when co-expressing GlyS-RNAi (S2D Fig.), indicating that this extra-glycogen is synthesized inside the FB. Furthermore, we observed that trehalose levels (S2E Fig.) in part mirrored the variations observed with in glycogen (S2D Fig.), as shown by a strong correlation (S2F Fig.). Since energy stores are mobilized during metamorphosis, the decrease in trehalose levels might be a direct consequence of reduced glycogen breakdown. However, when ACC-RNAi or FASNCG3523-RNAi was expressed in the FB, glycogen but not trehalose levels increased dramatically (S2D–S2F Fig.), suggesting that trehalose levels cannot be increased in the prepupae. Finally, neither glucose (S2G Fig.) nor protein (S2H Fig.) levels exhibited severe perturbation in any of the tested genotypes. Taken together these results indicate that at the end of larval life, glycogen accumulates in both the muscles and the FB, whereas TAGs accumulate mainly in the FB. Further, a reduction in TAG storage can, in part, be compensated for by an increase in glycogen storage.
Considering that the synthesis of glycogen and TAG constitutes a metabolic mechanism to safely store high quantities of glucose, we hypothesized that the anabolic enzymes FASN, ACC and GlyS, are induced by dietary sugar. Therefore larvae were fed a low carbohydrate diet (LCD) or a sucrose-supplemented diet (SSD). Using RT-Q-PCR, the expression of ACC, GlyS and the three FASN genes was monitored in larvae fed on 0% (LCD), 5%-, 10%- and 20%-SSDs (S2 Table). FASNCG17374 expression was insensitive to increases in dietary sugar, while the expression of all the other genes was enhanced by sucrose (Fig. 1A). This response was observed following a 5%-SSD but was not further enhanced by 10%- and 20%-SSD, indicating that a moderate increase in dietary sugar elicits an adaptive metabolic response.
Next, we wondered whether the synthesis of FA may protect against excess dietary sugar. Considering that the FB is the main storage organ, RNAi to ACC, FASNCG3523, or GlyS were induced with the Cg-gal4 driver and the duration of larval development was monitored by following the onset of metamorphosis. When fed LCD, no developmental delay was observed in control, Cg>ACC-RNAi, Cg>FASNCG3523-RNAi or Cg>GlyS-RNAi larvae (Fig. 1B). In contrast, when fed a 10%-SSD, the onset of metamorphosis was delayed by roughly two days in Cg>ACC-RNAi and Cg>FASNCG3523-RNAi larvae, while Cg>GlyS-RNAi larvae were only slightly delayed (Fig. 1C and lines 1–3, S3 Table). The effect was enhanced for larvae fed a 20%-SSD. Control larvae exhibited a 3-day delay, Cg>GlyS-RNAi larvae exhibited a 4-day delay, whereas Cg>ACC-RNAi and Cg>FASNCG3523-RNAi larvae exhibited approximately an 8-day developmental delay (Fig. 1D and lines 4–6, S3 Table). Together, these findings indicate that FA synthesis is a crucial metabolic pathway, which buffers the developmental defects induced by excess dietary sugar.
To gain further insights into the physiological requirements of LCFA synthesis we generated FASN mutants. As shown above, FASNCG3523 is an essential gene ubiquitously expressed, whereas FASNCG17374 sustains the synthesis of an essential FA only within the oenocytes (Table 1 and S1C–S1D Fig.). FASNCG3524 is not essential (Table 1) and may be redundant with FASNCG3523, as these two genes are in tandem on the second chromosome (Fig. 2A) and both are induced by dietary sugar (Fig. 1A). Therefore, we took advantage of two FRT-containing P-elements, located within the FASNCG3524 and the FASNCG3523 genes (Fig. 2A). Flipase recombination between the FRT sequences of these two P-elements resulted in a complete deletion of FASNCG3524, hereafter referred to as FASNΔ24. The resulting chimeric P-element links the 5’ region of FASNCG3524 to most of the FASNCG3523 genomic sequences (Fig. 2A). To generate a null FASNCG3523 mutant, we performed a remobilization of the chimeric P-element and looked for imprecise excisions that remove part of the FASNCG3523 gene. 22 excisions were recovered, one of them (hereafter referred to as FASNΔ24-23) removed 1200-bp of the FASNCG3523 gene (Fig. 2A), including the first methionine codon and the sequence coding half of the β-ketoacyl synthase (KS) domain [17].
Both FASNΔ24 and FASNΔ24-23 are lethal at the L1 larval stage. RT-Q-PCR analysis showed that FASNCG3524 expression could not be detected in either mutants fed a lipid-supplemented diet (Fig. 2B and see below). In addition, FASNCG3523 transcript levels were severely reduced in homozygous FASNΔ24 larvae and barely detectable in homozygous FASNΔ24-23 larvae (Fig. 2B). Therefore, both mutations delete FASNCG3524, however, FASNΔ24 appears to be a hypomorphic mutant and FASNΔ24-23 a null mutant for FASNCG3523.
To ascertain that the L1 lethality observed in both mutants was solely due to FASN deficiency, rescue experiments were performed, using UAS lines expressing either FASNCG3524 or FASNCG3523 cDNA. Ubiquitous overexpression revealed that FASNCG3524 cDNA could partially rescue the lethality of FASNΔ24-23 mutants to the pupal stage, although none emerged as adults (S4 Table). In contrast, ubiquitous overexpression of FASNCG3523 cDNA did not rescue the lethal phenotype in either FASN mutants and induced embryonic lethality when driven with any of the ubiquitous gal4-lines tested (S4 Table). However, one of the UAS-FASNCG3523 lines was able to partially rescue the lethal phenotype to pupal or adult stages in both mutants in the absence of gal4 drivers (S4 Table). Consistently, RT-Q-PCR analysis revealed that FASNCG3523 but not FASNCG3524 transcripts were detected at high levels in both FASN mutant rescued animals (Fig. 2B), indicating that an endogenous promoter could drive the expression of this UAS-FASNCG3523 transgene. These findings show that both FASNΔ24 and FASNΔ24-23 are bona fide mutants and suggest that FASNCG3523 protein levels should be maintained within a precise window of expression.
To determine whether the lethal phenotype could be rescued by dietary lipids, a LCD was supplemented with lipids (S2 and S5 Tables). Interestingly, supplementing a LCD with soy lipids could in part rescue the lethality of the hypomorph FASNΔ24 mutant to pupal or adult stages (S5 and see below) but not the lethality of the null FASNΔ24-23 mutant (S5 Table). We therefore, supplemented a LCD with various dietary lipids, including oils, margarine, butter and egg yolk alone or in combination. In isolation, none of the dietary lipids could rescue lethality of FASNΔ24-23 mutants, although a few larvae grew and developed to the L2 or L3 stages (S5 Table). In contrast, a LCD supplemented with butter and egg yolk (beySD) (S2 Table) could partially rescue lethality of both FASNΔ24 and FASNΔ24-23 mutants (S5 Table). To evaluate the metabolic consequences of the FASN deletion, TAG, glycogen, trehalose and glucose levels were measured in the FASNΔ24-23 mutant and control prepupae fed a beySD. FASNΔ24-23 prepupae exhibited a net decrease in TAG levels (Fig. 2C) associated with a moderate increase in glycogen and trehalose levels (Fig. 2D-E), whereas glucose levels were not significantly modified (Fig. 2F). Then, we performed a detailed analysis of FA composition of the TAGs, the sterol esters, and the various phospholipid classes. This analysis revealed that the relative FA content of the various phospholipids was not significantly modified (S3A–S3F Fig.). In each phospholipid class, palmitic acid (16:00) was always the most abundant FA component, although palmitoleic (16:01), stearic (18:00) oleic (18:01) and linoleic (18:02) acids were also highly represented. In contrast, the relative FA content in the sterol ester and TAG classes significantly varied in FASNΔ24-23 mutants versus controls (Fig. 2G-H). For the sterol ester class, oleic acid was less abundant in the mutants than in the control; however, this deficit was compensated for with higher levels of myristic (14:00), myristoleic (14:01), palmitoleic, linoleic (18:02) and arachidonic (20:04) acids (Fig. 2G). For the TAG class, control prepupae contained a higher proportion of saturated lauric (12:00), myrictic and palmitic acids, whereas mutants contained a higher proportion of unsaturated myristoleic, palmitoleic, oleic, linoleic and arachidonic acids (Fig. 2H). Together, these findings suggest that dietary lipids provide phospholipid precursors in sufficient amounts to compensate for the loss of FASN. Further, the difference in the FA composition of the TAG class in mutant versus control animals suggests that the structure of the TAGs is not critical.
Next, we investigated sucrose sensitivity in FASN mutant. Importantly, about 40% of the FASNΔ24 mutants fed a soy-lipid supplemented diet and of the FASNΔ24-23 mutant fed a beySD underwent metamorphosis onset (S4A Fig. and Fig. 3A). As shown by standard deviation values the percentages of rescue was highly variable. Nonetheless, addition of 10% sucrose to either lipid supplemented diet, resulted in a total lethality at L1 stage for both FASN mutants (S4A Fig. and Fig. 3A). These findings indicate that individuals that are unable to synthesized FAs are extremely sensitive to moderate increases in dietary sucrose. Moreover, less than half of the control larvae were able to pupariate when fed a lipid supplemented diet (S4A Fig. and Fig. 3A). The lipotoxicity was markedly suppressed when beySD was supplemented with 10% sucrose (Fig. 3A), possibly due to a reduction in the feeding rate (see below).
Since metabolic analysis is easier to perform on late rather than early larvae—which are very small—, a diet-shift protocol was established. FASNΔ24-23 mutant and control larvae were fed a beySD until the L2/L3 transition, transferred onto the same feeding media with or without 10% sucrose supplementation and left to develop 24h or 40h. First, to evaluate the feeding rate, larvae were transferred onto fresh media stained with brilliant blue FCF dye, and absorption of stained food was evaluated from whole larval extracts after one hour. Colorimetric measurement revealed that FASNΔ24-23 mutants contained much less food in their gut than control animals (Fig. 3B), and that sucrose supplementation also reduced the stained food content in both FASNΔ24-23 and control larvae (Fig. 3B). The lower gut content suggests that food uptake was reduced, although we could not exclude an increase in stool elimination. Next, levels of circulating sugars in larval hemolymph were measured. Interestingly, neither glucose nor trehalose levels increased in control larvae fed a 10%-sucrose supplemented beySD (Fig. 3C-D), suggesting that this feeding protocol does not induce a diabetic-like phenotype. Nonetheless, FASNΔ24-23 mutants fed a beySD exhibited a moderate increase in trehalose levels (Fig. 3D), while glucose levels remained unchanged (Fig. 3C). In contrast, after 24h of feeding on a 10%-sucrose supplemented beySD, both glucose and trehalose levels were strongly increased (Fig. 3C-D). Considering that increases in levels of circulating sugar is a hallmark of diabetes [27], the insulin response was evaluated in the FB of larvae expressing a tGPH reporter [28]. FBs were dissected from larvae fed a beySD with or without a 10%-sucrose supplement, and membrane translocation of tGPH was analyzed after 20 mn incubation with or without insulin. When grown on either feeding media, both control and mutant FBs were highly responsive to insulin stimulation (Fig. 3E-L and S4B Fig.) indicating that neither the FASNΔ24-23 mutant nor control larvae exhibit a T2D-like phenotype when fed a 10%-sucrose supplemented beySD. Importantly, the membrane-GFP fluorescence induced by insulin stimulation was much higher in FASNΔ24-23 mutant than in control larvae (Fig. 3F,H,J,L and S4B Fig.), suggesting that the former were hypersensitive to insulin. Together, our findings indicate that FASNΔ24-23 mutant animals are highly sensitive to dietary sugar but do not exhibit a T2D-like phenotype.
Since an increase in AGEs is linked to high levels of circulating sugar in T2D patients [12], we compared the amounts of AGEs in whole control or FASN mutant larvae. In L3 larvae transferred onto fresh beySD for 24h, the amounts of AGEs were higher in FASNΔ24-23 mutants than in controls. This was the case regardless or whether the beySD was supplemented with 10% sucrose (Fig. 4A). In older L3 larvae transferred on fresh beySD for 40h, the amounts of AGEs were strongly increased in FASNΔ24-23 mutants compare to controls (Fig. 4B). In addition, exposure to 10%-sucrose supplemented beySD further increased AGE levels in FASNΔ24-23 mutants (Fig. 4B), suggesting that FA synthesis constitutes a metabolic pathway to restrict AGE accumulation.
To further investigate the effects of dietary sucrose, we performed RNAi knockdown to two glycolytic enzymes encoded by single genes, Phosphofructokinase 1 (PFK1) and Pyruvate kinase (PK) that catalyze an early and the last glycolytic steps, respectively (Fig. 4C). FB-targeted knockdown to either PFK1 or PK did not result in a phenotypic defect in larvae fed a LCD, as developmental times did not differ markedly from controls (S5A Fig.). However these larvae were very sensitive to sucrose. When fed a 10%-SSD, both RNAi-knockdown larvae exhibited a significant developmental delay (S5B Fig. and lines 7–8, S3 Table). Moreover, when fed a 20%-SSD, the developmental delay was further increased for Cg>PFK1-RNAi larvae, whereas most of the Cg>PK-RNAi animals died during larval life (S5C Fig. and lines 9–10, S3 Table). The difference in sucrose sensitivity suggests either that PK-RNAi induces a more efficient knockdown than PFK1-RNAi, or that some glycolytic intermediates produced downstream of the enzymatic step catalyzed by PFK1 are extremely toxic.
Following the glycolytic step catalyzed by PFK1, an Aldolase cleaves fructose 1,6 bisphophate (Fru-1,6-BP) in the trioses phosphate, DHAP or G3P. Either metabolite leads to pyruvate, or to the highly reactive glycating α-oxoaldehyde MG via a non enzymatic reaction (Fig. 4C). We therefore used UAS-RNAi to the single glo1 ortholog that encodes an MG neutralizing enzyme. FASNCG3523-RNAi and glo1-RNAi were induced independently or together in the FB and the duration of larval development was monitored. When fed a LCD, glo1-RNAi larvae exhibited a moderate developmental delay (Fig. 4D and line 11, S3 Table). This developmental delay was slightly prolonged when fed a SSD, although not to the same extent as FASNCG3523-RNAi larvae, which were much more sensitive to dietary sucrose (Fig. 4D-D” and lines 11,14,17, S3 Table). Furthermore, animals dually expressing FASNCG3523-RNAi and glo1-RNAi in their FB exhibited a high rate of larval lethality and a developmental delay that dramatically increased concurrently with sucrose concentration (Fig. 4D-D” and lines 12–13,15–16,18–19, S3 Table). Conversely, FB-overexpression of Glo1 was able to partially compensate for the developmental delay induced by an increase dietary sugar (Fig. 4E-E” and lines 20,22, S3 Table). FB-overexpression of Glo1 was also able to partially suppress the strong developmental delay of FASNCG3523-RNAi larvae grown on SSD (Fig. 4E-E” and lines 21,23, S3 Table). In each assay, the percentage of pupae is relative to the number of their SM5-TM6B siblings (see material and methods). Intriguingly, when fed a 20%-SSD, the ratio of UAS-glo1 larvae relative to the number of their SM5-TM6B siblings was higher than the control ratio, reaching a maximum at roughly 120% (Fig. 4E”). Furthermore, we also observed that when testing homozygous w- control flies, the rate of larval lethality was significantly higher in 20%-SSD than in LCD or in 10%-SSD (S5D Fig.). This observation suggests that in the Glo1-overexpressing assay, a significant number of the SM5-TM6B siblings underwent lethality when fed a 20%-SSD and that Glo1 overexpression suppresses this lethality. In contrast in the control assay all the larvae underwent the same rate of lethality irrespective of the SM5-TM6B balancers. Together, these findings indicate that sucrose toxicity can be alleviated by overexpression of Glo1 and conversely, the deleterious effects are exacerbated when both FA synthesis and Glyoxalase activity are simultaneously dampened.
To determine whether a lack of FA synthesis induces cell-autonomous defects, we generated flip-out recombination during embryogenesis and analyzed the resulting clones in the FB of feeding larvae at the end of the L3 stage. Interestingly, the size of FASNCG3523-RNAi cells was almost normal in larvae fed a LCD, but drastically reduced in larvae fed a 20%-SSD (S6A–S6B,S6M Fig.). A similar phenotype was observed for PK-RNAi flip-out cells (S6C–S6D,S6M Fig.), although the size reduction observed in larvae fed a 20%-SSD, varied a lot depending on the experiment, possibly because of variability in RNAi efficiency. In contrast, PFK1-RNAi flip-out cells were insensitive to dietary sucrose since cell size remained unchanged irrespective of sucrose supplementation (S6E–S6F,S6M Fig.). To perform genetic interactions at the cellular level, we generated MARCM clones either mutant (FASNΔ24-23) or wild-type (FASN+). Firstly, the sucrose sensitivity of FASNΔ24-23 cells was evaluated in the FB of larvae raised on media containing increasing quantities of sucrose. For larvae fed a LCD, the size of FASNΔ24-23 cells was slightly reduced compare to neighboring control cells (Fig. 5A,M). However, as the sucrose content in the diet increased, a concomitant reduction in the size of FASNΔ24-23 cells was observed (Fig. 5B-D,M). This cell size reduction was not correlated with lipid content, as Nile red staining revealed that FASNΔ24-23 cells were severely depleted in LDs, irrespective of sugar supplementation (S6G–S6I Fig.). Next, we generated FASNΔ24-23 MARCM clones expressing PFK1-RNAi or PK-RNAi. Under these conditions, the size of FASNΔ24-23 cells, expressing either RNAi was hardly reduced in larvae fed a LCD (Fig. 5E,G,N). However, in larvae fed a 20%-SSD, the size of FASNΔ24-23 cells remained unaffected when expressing PFK1-RNAi (Fig. 5F,N), but were dramatically reduced when expressing PK-RNAi (Fig. 5H,N). The phenotypic suppression produced by PFK1-RNAi, suggests that an intermediate metabolite, downstream of PFK1 (Fig. 4C), is responsible for the size reduction of FASNΔ24-23 cells observed in SSD-fed larvae. Therefore, MARCM clones, expressing glo1-RNAi were analyzed. Interestingly, MARCM FASN+ clones expressing only the glo1-RNAi were insensitive to sucrose supplementation (Fig. 5I-J,O). Nonetheless, FASNΔ24-23 cells expressing glo1-RNAi exhibited an extreme size reduction in larvae fed LCD (Fig. 5K,O) accompanied by a severe decrease in nucleus size (see below). Furthermore, these clonal cells could not be observed when larvae were fed 20%-SSD, suggesting that these cells were eliminated during development. Conversely, FASNΔ24-23 MARCM cells overexpressing glo1 were of normal size in larvae fed a 20%-SSD (Fig. 5L,P). However, neither FASNΔ24-23 MARCM cells in LCD-fed larvae, nor FASN+ MARCM cells were affected in size by Glo1 overexpression (S6J–S6L,S6M Fig.). Together, these findings indicate that Glo1 can compensate for cell size reduction due to a sugar-dependent FA-synthesis defect, but is unlikely to promote cellular growth.
Finally, an antibody to MG-derived AGEs (MG-AGEs) was used for immunostaining. In FASNΔ24-23 clonal cells the amounts of MG-AGEs were barely detectable in larvae fed a LCD (Fig. 6A-C) but were dramatically increased in larvae fed a 20%-SSD (Fig. 6D-F). Importantly, increased MG-AGE levels induced by 20%-SSD were abolished in FASNΔ24-23 MARCM clones expressing either PFK1-RNAi (Fig. 6G-I) or UAS-glo1 (Fig. 6J-L). Furthermore, in larvae fed a LCD, FASNΔ24-23 clones expressing glo1-RNAi exhibited a strong accumulation of MG-AGEs (Fig. 6M-O). Nucleus size in these clones, was also dramatically reduced (Fig. 6O,O’). Taken together, these findings indicate that FA synthesis and Glyoxalase activity cooperate in a cell-autonomous manner to neutralize the toxicity of dietary sugar, which may result in cellular growth defects or putative cell elimination.
In this study, we investigated the role of FA synthesis in regulating homeostasis in response to dietary sugar. To maintain tolerable levels of circulating sugars, organisms synthesize and store macromolecules in appropriate organs. In contrast to previous studies in insects, which report that the majority of TAGs stored in the FB are of dietary origin [29,30], we observed that in Drosophila, the larval FB is a lipogenic organ. However, in FASNΔ24-23 mutant fed a beySD, TAG levels were decreased but not abolished. This indicates that as in mammalian hepatocytes and adipocytes [2,31,32], TAGs stored in the Drosophila larval FB originate from either food assimilation or de novo synthesis. Together, our findings confirm that metabolic pathways act within an integrative network to maintain homeostasis and support the notion that in term of post-feeding macromolecules storage (TAGs and glycogen), the Drosophila larval FB constitutes an alternative model for mammalian liver and adipose tissue (Fig. 7A).
Our FASN mutants are lethal at L1 stage, and this lethality can be recued by a beySD. Rescue of FASNΔ24 but not of FASNΔ24-23 mutants by soy lipid extracts likely reflects the strength of the mutation since FASNCG3523 is still weakly expressed in the hypomorphic mutant. Consistently a SREBP mutant that down-regulates but does not abolished the expression of several FA anabolic enzymes including FASNCG3523, could also be rescued by soy lipid extracts [33]. Rescue of the lethal phenotype by dietary lipids, as well as the minor phenotype observed in FASNΔ24-23 clonal cells, suggests that neighboring cells or organs can provide FAs to those that are deficient. This may be achieved through lipophorin activity [34]. Intriguingly, we found that in contrast to other lipid-supplemented media, a mix of butter and egg yolk could rescue the FASNΔ24-23 lethal phenotype. TAGs cannot be directly assimilated by enterocytes; first they require digestive lipases to cleave TAGs to di-acyl-glycerol (DAG), mono-acyl-2-glycerol (MAG) and free FAs (FFAs) [35,36]. In several mammalian species, lipids interact with bile acids to form micelles prior to enzyme cleavage and enterocyte absorption. However, it has been reported that in rats and human infants, FFAs may interact with calcium or magnesium ions to form soaps that are hardly assimilated [37,38,39]. In insects, lipid emulsifiers are poorly characterized although glycolipid or amino acid complexes are likely to be involved in lipid assimilation [40,41]. As egg yolk lipoproteins are highly efficient emulsifiers [42], they may help solubilize lipids, thereby favoring their absorption. The composition in FAs and their positions on the glycerol backbone vary depending on the origin of the TAGs. Regarding FA synthesis, FASNΔ24-23 mutants are expected to lack palmitic acid. Analysis of various oils and fats, revealed that TAGs found in butter contain high quantities of palmitic acid in position sn-2 of the glycerol [36]. Thus, assimilation of MAGs resulting from butter digestion, are high in palmitic acid. Hence, it is possible that MAGs are better assimilated than FFAs in our FASNΔ24-23 mutants. In order to fully understand the process of lipid absorption in FASNΔ24-23 mutants, extensive analysis, including the precise measurement of ingested and excreted FAs, will be required.
Rescue of lethality of FASN mutant by a lipid-supplemented diet indicates that FA synthesis deficiency can be compensated for by an appropriate lipid diet. Previous studies in Drosophila have reported that the FA composition of the various lipid classes varies depending on the diet [43,44]. Here, we show that the relative FA composition of phospholipids is not significantly different in FASNΔ24-23 rescued mutants and control animals fed a beySD. These findings not only confirm that diet contributes to phospholipid composition, but reveal that in the presence of an exogenous lipid supply, the essential FASN enzyme becomes dispensable for phospholipid synthesis. In contrast, sterol esters and TAGs exhibit variation in their FA composition. Compared to controls, TAGs from mutants contain less saturated FAs and more long chain unsaturated FA, suggesting that expression of desaturases and elongases [45] may be increased in FASN mutants. The high variability in TAG composition suggests that TAG structure is not a crucial parameter, which strengthens the notion that TAG synthesis constitutes a metabolic strategy to neutralize the potential toxicity of nutrients. Previous studies suggested that the fat tissue fulfills a protective role against excess sugar. In agreement with this, it has been shown that fat transplantation in lipoatrophic mice reverses T2D [46,47] and that in genetically induced obese mice, a decrease in adipose FASN expression is linked to T2D progression [48]. In addition, mice and flies with defects in ChREBP—a transcriptional activator of lipogenic enzyme expression—do not survive increases in dietary sugar levels [44,49,50]. However, it was unknown that FASN activity also protects against sugar toxicity. This finding is in contrast to a previous report which showed that in flies, lethality induced by ubiquitous expression of FASNCG3523-RNAi can be partially rescued by dietary sugar [50]. Here, we demonstrate that FA synthesis protects against dietary sugar at both a systemic and cell-autonomous level. The media used in our study contained low concentrations of digestible sugar, 64, 164 and 264 mg/ml for the LCD, 10%-SSD and 20%-SSD, respectively. In other studies, which used Drosophila larvae as a model for sugar tolerance, the concentration of digestible sugar was 86, 140 or 80 mg/ml for the low carbohydrate media and 377, 380 or 230 mg/ml for the sugar enriched media [27,50,51]. Importantly, while circulating sugar levels increase in FASNΔ24-23 animals, these mutants do not exhibit a T2D-like phenotype and become insulin hypersensitive. Therefore, as previously suggested [44,49,50], disrupting FA synthesis provides a convenient model to investigate the effect of glucotoxicity independent of lipotoxicity.
Here, we provide evidence to propose that FASN and Glo1 cooperate both in a systemic and in a cell-autonomous manner to protect against the deleterious effect of dietary sucrose. Our findings indicate that when FA synthesis is very active, as in the FB of Drosophila larvae, Glo1 activity is dispensable in term of neutralizing the few toxic metabolites produced through sugar catabolism (Fig. 7A). Conversely, the detoxifying activity of Glo1 becomes critical when FASN activity is disrupted in the larval FB (Fig. 7B). Thus, the observed decrease in lipogenic enzyme expression in the adipose tissue of a diabetic mouse model [48], may require an increase in Glo1 activity. If lipogenic enzyme expression is also decreased in T2D patients, the increase in glycating agents [52] may result not only from an increase in circulating sugar but also from a decrease in FA synthesis. For a few decades, pathological damage induced by excess sugar was thought to be a consequence of AGE formation [53], a paradigm substantiated by recent studies on experimental diabetic nephropathy [9,10,54]. Consistent with a study in Caenorhabditis elegans, reporting that Glo1 overexpression protects against glucose toxicity [55], we show that manipulating Glo1 levels in the larval FB modulate a sugar-induced developmental delay. Studies in diabetic models and patients mostly focused on AGE levels in body fluids [56,57,58,59], although alterations to intracellular products have also been reported [60,61,62]. At the cellular level, glo1 knockdown in FB cells induces a cell-autonomous phenotype, only when clones are also FASN deficient. This phenotype results in either an extreme reduction in cell size or elimination of cells, when larvae are fed LCD or SSD, respectively. The number of FB cells is determined during a proliferative phase at embryogenesis. During larval life, FB cells do not divide, but undergo a rapid cell growth phase [63,64]. The lack of a visible phenotype in glo1-deficient cells indicates that even when larvae are fed SSD, Glo1 does not affect the growth process of FB cells. In contrast, increasing quantities of sucrose in the food, even to moderate levels, induces a size reduction of FASN mutant cells. This phenotype is unlikely to depend directly on sugar since addition of moderate amounts of sucrose to food media does not markedly increase circulating sugar levels. In contrast, it is likely to directly depend on an increase of intracellular MG, since the cell size reduction is suppressed if FASN mutant cells are either deficient in PFK1 or overexpressing glo1 cDNA. In summary, our findings suggest that FASN activity is dispensable in sustaining cell growth but plays a key role in protecting against the potentially toxicity of MG produced through glycolysis.
In conclusion, we have demonstrated that FA synthesis constitutes a metabolic strategy to restrict the production of intermediate toxic molecules, suggesting that obesity is not a harmful process, as long as storage capacity is not overwhelmed. Furthermore, our study highlights the need for caution when using FA synthesis inhibitors to treat cancers and metabolic diseases, as they might provoke negative side effects.
Fly strains: P[tGPH] [28], daughterless(da)-gal4, Mef2-gal4, actin5C>CD2>gal4,UAS-GFP, P[w[+mC] = tubP-GAL80]LL10,P[ry[+t7.2] = neoFRT]40A, UAS-Dcr-2 (Bloomington Stock Center); Inducible RNA-interfering (UAS-RNAi) lines to ACC (VDRC 108631), FASNCG3523 (VDRC 29349), FASNCG3524 (VDRC 4290), GlyS (VDRC 35136), glo1 (remobilized on chromosome III from VDRC 26832), PFK1 (VDRC 3017), PK (remobilized on chromosome III from VDRC 49533); FASCG17374-RNAi, svp-gal80, Cg-gal4 [21]. The P-element insertions (Exelixis collection) P[XP]v(2)k05816d04154 and P[XP]CG3523d06961 were used to generate deficiency as described [65]. All the fly lines were isogenized from single males in a white1118 mutant (w-) background. For clonal analysis FASNΔ24-23 was over SM5-TM6B,Tb- balancers [21]; For survival and metabolic analyses, FASN mutants were balanced by a CyO GFP-labelled chromosome.
The results presented for ubiquitous or tissue-targeted UAS-RNAi lines—including the corresponding controls—were obtained with a UAS-Dcr-2 that strengthens the RNAi effect. Developmental delays were evaluated from overnight egg collection and the number of prepupae formed was counted every morning. For each assay, several tubes were collected, overcrowded tubes were discarded and the numbers of prepupae were pooled. As some of the transgenes used in the genetic combinations were homozygous lethal, all the lines (driver, RNAi, w- control) were balanced with co-segregating SM5-TM6B,Tb- balancers that lead to non-mendelian offspring distribution. Therefore, for each assay, the number of RNAi-expressing Tb+ larvae was divided by the final number of Tb- larvae and all assays were normalized to the control ratio. For controls, a similar calculation was done from the offspring of driver females crossed with w-;SM5-TM6B males and this control ratio was adjusted to reach 100%.
To generate the overexpressing lines, the locus of FASCG3523, and FASCG3524 were recovered by gap repair and the endogenous promoter replaced by a UAST [66]. glo1 cDNA was amplified from GH24818 (DGRC) and cloned into the pUAST vector. Plasmid constructs were injected by BestGene.
RT-Q-PCR were performed as previously described [21] using the following primers:
FASNCG3523 (5’-F CTTCTTCATTTCCCCGA-3’ and 5’-CGAAGGAGTATCCGGC-3’)
FASNCG3524 (5’-CTTTGACAATATGCTCTAC-3’ and 5’-AAGTCCGGAGTGTCCAG-3’)
FASNCG17374 (5’-F ATCAGCTCCAACCTCTAC-3’ and 5’-GGGCTACATGCAAGTCT-3’)
ACC (5’-TTGGGAAACTCATTCGTG-3’ and 5’-CCAGGACCTTGGCATTA-3’)
GlyS (5’-CCCCTCATACTACGAGC-3’ and 5’-CGATATAGCGGCGATCC-3’)
Flip-out clones were performed as described [21]. MARCM clones [67] were heat-shock induced at 4–6h embryogenesis and the larvae were allowed to grow on various sucrose-supplemented media until mid/late L3 stage. FB were dissected as described [21] but fixed with 3.7% formaldehyde in PBT (PBS 0.1% Tween20). FB were stained with Phalloidin-Rhodamine B (sigma) at 625 ηg/ml for 2h at RT, extensively washed and mounted in DABCO (sigma). Relative cell size was expressed as a ratio, clonal:neighboring control cells, which was estimated using the image-j software. The insulin responsive assay was performed as described [27]. tGPH quantification was measured in squares (10X10 pixels) positioned either at the membrane or at the nuclei. Measurements for each assay were recorded from 8 cells taken from 2 different FBs. For each cell, the maximum fluorescence intensity at the membrane was divided by the maximum fluorescence intensity at the nucleus and the mean ratio was plotted (S4B Fig.). Nile Red staining was performed as previously described [21]. For MG-AGE Immunostaining, dissected FB were fixed as described above, but blocked for 20 mn in PBS containing 0.1% Triton X100 and 2% bovine serum albumin. Samples were incubated overnight at 4°C with diluted (1:400) MG-AGE antibody (Cell Biolabs), extensively washed and incubated for 2h at room temperature with secondary antibody and DAPI in the blocking solution. Samples were finally washed in PBT and mounted in DABCO. Image acquisitions were obtained using a Nikon TE2000-U or a Leica SP8 confocal laser-scanning microscopes.
TAGs, protein, glucose, trehalose and glycogen measurements were performed as previously described [21]. To measure circulating sugar, 6μl samples of hemolymph were collected from 20 to 30 bled L3 larvae. For AGE measurement, 5 samples of 10 L3 larvae were washed in PBS and crushed at 4°C in a Precellys 24; extracts were cleared 10 mn at 4° C in a microfuge at maximum speed. Extracts were diluted 100X in PBS and 100 μL of this diluted extract were treated with an ELISA kit (Cell Biolabs STA-317). AGEs estimation evaluated from spectrophotometric dosage at 450nm, was normalized to the protein concentration of each sample. To measure feeding rates, food media was tinted with 0,1% brilliant blue FCF. 3 samples of 10 L3 larvae were collected, frozen, extracted in 200μl water and centrifuged for 7min at maximum speed. The final volume was adjusted to 800μl and measured at 629ηm. Lipidomics were performed in triplicates of 100 mg 0–5h prepupae. Lipids were extracted and analyzed by GC-MS as described [68].
Statistical analyses were performed with R version 3.0.2 [69]. Error bars in figures stand for empirical standard deviations measured independently from the replicates in each category. Significance for the statistical tests was coded in the following way based on the p-values: ***: 0 < p < 0.001; **: 0.001 < p < 0.01; *: 0.01 < p < 0.05. In all the graphs, the error bars represent the standard deviations.
For S3 Table (corresponding to Fig. 1C, 1D, 4D, 4E, and S5A–S5C Fig.), the effect of the genotype was tested with one-way ANOVAs on developmental rates. Developmental rates (in units of days-1) were computed as the inverse of developmental duration to pupation. Lethality (evaluated for each developmental curve and corrected with the lethality rates for control measured in S5D Fig.) was accounted for by including a corresponding number of (unobserved) lethal events (individuals with a developmental rate of 0). Since all nine ANOVAs detected a significant effect of the genotype, pairwise comparisons between genotypes were tested with a post-hoc Tukey "Honest significant difference" test [70] for each sub-figure, and the biologically-relevant comparisons are reported.
|
10.1371/journal.pbio.1001089 | The Neural Basis of Following Advice | Learning by following explicit advice is fundamental for human cultural evolution, yet the neurobiology of adaptive social learning is largely unknown. Here, we used simulations to analyze the adaptive value of social learning mechanisms, computational modeling of behavioral data to describe cognitive mechanisms involved in social learning, and model-based functional magnetic resonance imaging (fMRI) to identify the neurobiological basis of following advice. One-time advice received before learning had a sustained influence on people's learning processes. This was best explained by social learning mechanisms implementing a more positive evaluation of the outcomes from recommended options. Computer simulations showed that this “outcome-bonus” accumulates more rewards than an alternative mechanism implementing higher initial reward expectation for recommended options. fMRI results revealed a neural outcome-bonus signal in the septal area and the left caudate. This neural signal coded rewards in the absence of advice, and crucially, it signaled greater positive rewards for positive and negative feedback after recommended rather than after non-recommended choices. Hence, our results indicate that following advice is intrinsically rewarding. A positive correlation between the model's outcome-bonus parameter and amygdala activity after positive feedback directly relates the computational model to brain activity. These results advance the understanding of social learning by providing a neurobiological account for adaptive learning from advice.
| Learning by following advice is fundamental for human cultural evolution. Yet it is largely unknown how the brain implements advice-taking in order to maximize rewards. Here, we used functional magnetic resonance imaging (fMRI) and behavioral experiments to study how people use one-off advice. We find that advice had a sustained effect on choices and modulated learning in two ways. First, participants initially assumed that the recommended option was most beneficial. Second, and more importantly, gains and losses obtained after following advice received an “outcome-bonus,” in which they were evaluated more positively than after not following advice. In other words, following advice was in general intrinsically rewarding. Computer simulations showed that the outcome-bonus is adaptive, because it benefits from good advice and limits the effect of bad advice. The fMRI analysis revealed a neural outcome-bonus signal in the septal area and left caudate head, structures previously implicated in trust and reward based learning. Participants with greater outcome-bonuses showed a greater gain-signal increase after following advice in the amygdala, a structure implicated in processing emotions and social information. In sum, these results suggest that decision makers adaptively combine advice and individual learning with a social learning mechanism in which advice modulates the neural reward response.
| The nature and level of social learning in human societies is unmatched in the animal world. Especially when decisions are difficult, people rely on advice or recommendations regarding a decision or course of action [1]. Accumulating knowledge through social learning (particularly advice taking) is uniquely human and fundamental to the evolution of human culture [2]–[4], and it is plausible that genetic adaptations to social learning evolved in humans [5]. Cumulative social learning strongly relies on advice taking, which transmits social information more reliably than imitation or observational learning. For the individual, heeding advice can be especially useful when mistakes are costly and social information is accurate [4],[6]. Accordingly, advice taking affects many domains of learning and decision making, such as cooperation [6],[7], financial decisions [8], or consumer behavior [9]. For instance, people do not discover a healthy diet by trial and error but combine recommendations from others with their own experiences to choose their meals.
The influence of advice and social learning in general does not require direct personal interaction but can be observed in situations where social information is transmitted by observation or by written or spoken advice [10]–[12]. Recent fMRI experiments provided the first insights into the neurobiological mechanisms underlying social learning. Social prediction error signals are used to learn about the probability of good advice from advisors with sometimes cooperative and sometimes uncooperative motives [13] and determine to what extent initial judgments are adjusted based on social information [14]. However, these results do not provide a mechanistic explanation for the often-observed sustained influence of advice or, more generally, the human propensity for social learning. In particular, it remains unclear if and how the brain implements an adaptive social learning mechanism to combine supportive advice with individual information gained through personal experience.
Reinforcement learning models [15] can provide hypotheses about the influence of advice on decision making, especially when decisions are based on past experiences. These models specify distinct sub-mechanisms of learning, such as the initial evaluation of choice options or the repeated evaluation of choice outcomes, which have different behavioral [16] and neuronal signatures [17] that may be separately influenced by advice. Behavioral studies have shown that the human propensity for following advice could be explained by its influence on the evaluation of outcomes rather than on initial reward expectations or choice processes [12],[18]. This influence is described best by an outcome-bonus model [12], which postulates a learning mechanism in which a reward bonus is added to both good and bad outcomes of recommended options. More specifically, rewards from recommended options lead to stronger positive reinforcements than rewards with the same objective value from alternative, non-recommended options. Correspondingly, punishments from recommended options inhibit the choice of that option less than punishments with the same objective disutility from non-recommended options. In fact, when the punishment from a recommended option is smaller than the size of the outcome-bonus, the punishment may still lead to a positive reinforcement. The behavioral evidence in favor of the outcome-bonus model suggests that the neurobiological implementation of advice-following relies on the brain's reward system. Neurophysiological experiments in monkeys and fMRI experiments in humans [15],[19] consistently report reward representation in targets of dopaminergic midbrain projection neurons. Positive outcomes (rewards) elicit an increase in blood-oxygen level-dependent (BOLD) responses in the ventromedial prefrontal cortex (VMPFC) [20],[21], the medial prefrontal cortex (MPFC) [22],[23], the amygdala [24], and the ventral striatum (VST) [25]–[27]. Of particular interest in the context of social learning is the septal area, because it signals reward [28] and triggers release of oxytocin [29], which in turn is known to enhance trusting behavior [30]. Hence, we predicted that positive outcomes from a recommended decision would lead to greater positive BOLD responses than positive outcomes from non-recommended decisions in these reward sensitive regions. Furthermore, whereas negative outcomes should lead to a negative BOLD response after choosing a non-recommended option, negative outcomes after choosing a recommended option should lead to an attenuated BOLD response decrease or even to a positive BOLD response.
Based on these predictions, we investigated if and how the outcome-bonus is implemented in the brain. In addition, we compared computational models and used simulations to test whether the outcome-bonus model provides the best explanation of behavior and if it is an adaptive social learning mechanism. We show that, compared to alternative social-learning mechanisms, the outcome-bonus is more adaptive and can better account for the observed behavior. Moreover, we identified a neural outcome-bonus signal in the septal area and the left caudate.
Participants in the experiment learned that advice (i.e., a form on which the advisor marked which option the advice receiver should choose) was given from a second group of participants, who had previous experience with the task and were motivated to give good advice (see Figure 1 and Text S1 for details). Of the 21 participants, 16 received good advice. Regardless of good or bad advice, participants chose the recommended deck (41.5% of all choices) twice as often as they chose the non-recommended deck with the same payoff distribution (21.5% of all choices; p<.0001). Notably, this effect of advice was not limited to the beginning of the experiment, but rather was sustained; Figure 2A shows that participants robustly preferred the recommended deck to the non-recommended deck with the identical payoff distribution throughout the entire experiment. This result is consistent with the outcome-bonus but not with the assumption that advice influences only the evaluation of choice options prior to individual learning.
In the first half and, to a lesser extent, the second half of the trials, recipients of good advice chose the good decks more frequently than recipients of bad advice. The fact that this effect is greater in the first half (p = .039, effect size r = .39) than in the second half of the experiment (p = .137, effect size r = .25) indicates that bad advice harmed learning more during the first half of the experiment (c.f. Figure S2). The relatively weaker influence of bad advice in the second half of the experiment shows that decisions were made based on a combination of advice and individual learning, because only individual learning by the participants receiving bad advice can explain why they performed nearly as well as receivers of good advice in the second half of the experiment.
We evaluated the outcome-bonus model quantitatively by comparing it with alternative models based on standard model selection criteria. The results provide strong empirical support for the outcome-bonus mechanism as essential to explain social learning. In particular, we derived the Bayes factor from the Bayesian information criterion (BIC) [31]. With this model selection criterion, we found strong evidence in favor of the outcome-bonus model and a combined model implementing an outcome-bonus and higher initial reward expectations for the recommended deck (henceforth prior+outcome-bonus model). Figure 2B illustrates that, when considering the models' Bayes factors, these models predict the observed behavior equally well and much better than alternative models. Additionally, we compared the models by their Akaike information criterion (AIC) as an additional model selection criterion. Here, the prior+outcome-bonus model was the best model. Moreover, comparing the outcome-bonus, the prior, and the prior+outcome-bonus model against each other illustrates that removing the assumption of an outcome-bonus hurts the model fit more than removing the prior. When comparing the AIC (or BIC) values of the prior, the outcome-bonus, and the prior+outcome-bonus models with eight alternative models on a participant-by-participant level, the prior model is on average better for 59.7% (or 57.6%) of participants, the outcome-bonus model for 62.3% (or 64.4%), and the prior+outcome-bonus model for 67.1% (or 47.5%) (see Text S1 for details). Like the previous model comparison results, these comparisons underline the relevance of the outcome-bonus mechanism.
Simulated learning paths of the models illustrate that the outcome-bonus model, but not the prior model, predicts our key behavioral result, namely the sustained effect of advice on participants' behavior (c.f. Figures 2A and 3B). It might appear counterintuitive that the outcome-bonus is assumed to stay constant throughout the learning process. However, dynamic versions of the outcome-bonus model and the prior+outcome-bonus model, in which the outcome-bonus increases with time after good advice and decreases with time after bad advice, describe learning more poorly than the models using a constant outcome-bonus. Alternatively, advice could only influence the evaluation of gains or losses. These alternative models, however, again fitted the observed learning processes less well than the original outcome-bonus model (see Text S1 for details). In sum, regardless of the model selection criterion, the change of prior evaluations of options and more importantly the outcome-bonus mechanism are crucial constituents of descriptive social learning models for the influence of advice on learning. Because the prior+outcome-bonus model explained the data altogether best, we used predictions and parameters of this model in the fMRI analysis.
The outcome-bonus mechanism may be crucial to explain learning processes because it helps people to solve the learning task successfully. Indeed, when advice is more likely to be good than bad and the task is difficult and long (as was the case in our experiment), the outcome-bonus model is generally more adaptive (i.e., leading to higher average rewards) than the prior model because it ensures a lasting influence of good advice.
Beyond this basic insight, the simulation results depicted in Figure 3A show that, when good and bad advice are equally likely, the outcome-bonus model performs worse than the prior model only in a situation where at the same time (a) learning is difficult, and (b) the outcome-bonus is so large that recommended bad options are evaluated more positively than the objectively good options. Crucially, however, Figure 3A also shows that, in most situations, the outcome-bonus model outperforms the prior model when good and bad advice is equally likely. When advice is bad, the outcome-bonus model performs better because the prior model learns only late—after the wrong initial expectation for the recommended deck has been unlearned—which options are best (c.f. inset in Figure 3B). The advantage of the outcome-bonus model after bad advice is particularly strong for easier tasks where individual learning is relatively successful, because it does not interfere strongly with individual learning, whereas the prior does (see also Text S1). When advice is good, the outcome-bonus model performs better because it leads to a preference for a good option long after the effect of higher initial expectations has decayed (c.f. Figure 3B).
The superiority of the outcome-bonus model is notable, as the prior model resembles more a Bayesian approach, in which advice as prior information should influence the initial evaluation of choice options. However, whereas the prior model learns the expected values more accurately in the long run, cumulative rewards do depend on the choices made based on the learned values. The sustained overestimation of the rewards from a good option implemented by the outcome-bonus model (after good advice) helps to make the choice of that option occur more frequently and ultimately helps to accumulate more rewards.
We used fMRI to test the prediction that rewards from recommended versus non-recommended options would lead to greater BOLD responses in reward-sensitive brain regions. Regions implementing the outcome-bonus (outcome-bonus regions) should fulfill two conditions. First, when advice is followed, gains should lead to a greater increase in BOLD signals and losses should lead to a smaller decrease in BOLD signals (compared to when not following advice). Second, when advice is not followed, outcome-bonus regions should be gain-preferring; that is, these regions should show a regular reward signal with an increased positive BOLD signal after gains and a reduced BOLD signal after losses [32]. Only one cluster comprising the septal area and the left caudate head showed the predicted effect of advice (max. z-score = 3.49; Montreal Neurological Institute [MNI] coordinates: x = 4, y = 2, z = 4; see Figure 4A and 4B) and was also gain-preferring, suggesting that this region implements the outcome-bonus. This outcome-bonus implementation cannot be explained in terms of different payoff distributions of the good and bad options because the experiment was designed such that good and bad decks were equally likely to lead to gains or losses (although the magnitudes of gains and losses differed). Moreover, because our fMRI analysis controlled for both different gain and loss magnitudes of good and bad options as well as different prediction error magnitudes of the advice and no-advice condition, the result can neither be ascribed to the fact that advisors recommended good options more frequently than bad options nor to differences in prediction errors elicited by feedbacks from the different choice options.
For more detailed insights into how the brain evaluates outcomes that are dependent on advice, we contrasted feedback-related BOLD responses separately for gains and losses in reward signaling regions after following and not following advice (see Figure 4C and 4D, and Text S1). For losses, we found greater BOLD responses after following advice in two gain-preferring regions: the VMPFC (max. z-score = 3.35; x = −10, y = 52, z = −18) and the left caudate (max. z-score = 3.23; x = −16, y = 20, z = −6). However, these regions did not provide a complete outcome-bonus signal because the BOLD response to positive feedback was not greater after following advice. One cluster in the orbitofrontal cortex also showed a weaker BOLD signal reduction for losses after following advice (max. z-score = 3.35; x = 16, y = 28, z = −12), but voxels in this region were not gain-preferring. For gains, we found that the difference between activity in the left amygdala after following or not following advice correlates with the outcome-bonus parameter of the prior+outcome-bonus model (max. z-score = 3.02; x = −26, y = −4, z = −14), suggesting that the amygdala also implements the outcome-bonus.
To investigate how advice modulates standard brain responses to rewards, we investigated advice-dependent changes in brain regions that showed greater activity after not following advice for gains compared to losses. Such reward signals were identified in the VMPFC, the ventral striatum (VST), and the right insula. The parameter estimates of these regions for gains and losses after following and not following advice show that advice led to an attenuation of the BOLD response in the VMPFC and VST, such that gain and loss signals are closer to the baseline BOLD response after advice was followed (see Figure 5).
To check the robustness of the neural outcome-bonus signal resulting from the effect of advice, we performed supplementary fMRI analyses. First, the above described analysis did not reveal a correlation between BOLD responses and prediction errors, likely because it included separate regressors for positive versus negative payoffs, which captured the variance associated with positive versus negative prediction errors. Indeed, a supplementary fMRI analysis tailored to reveal a prediction error signal identified correlations with prediction errors in the VST (Figure S6A). Importantly, this analysis also revealed the above reported effect of advice on reward signals in the septal area and the left caudate head (Figure S6B). Second, to further investigate the existence of a sustained effect of advice on learning and the neural correlates underlying this effect, we performed another fMRI analysis that tested whether the outcome-bonus changed from the first to the second half of the trials in which advice was followed. Consistent with our modeling results showing that models with a dynamic outcome-bonus do not explain behavior substantially better than models with a constant outcome-bonus, we did not find a change in the neural outcome-bonus signal in the septal area over time. However, we found reduced BOLD responses for feedback after following advice in the paracingulate gyrus and the superior temporal sulcus (see Figures S7 and S8 for details), which are commonly associated with theory-of-mind processes and trusting behavior [33],[34].
Taken together, behavioral, modeling, simulation, and neuroimaging data provide strong convergent evidence for a sustained effect of well-intentioned advice on decision making, which can be explained by an outcome-bonus model for following advice. Behavioral data showed that advice had a long-lasting influence on decision making and learning. Simulations suggest that the outcome-bonus is an adaptive social learning mechanism in a broad range of social learning environments. The model comparison showed that the outcome-bonus is necessary to explain the behavioral effect of advice. fMRI data supported this conclusion, as advice modulated reward-related brain activity so that the gain-sensitive septal area and the left caudate head showed a greater reward signal after following rather than not following advice; even negative outcomes elicited a positive reward response when advice was followed.
One feature of the experiment was that participants controlled when to follow advice, so that advice-following trials were not randomized across the experiment. Hence, additional factors might have influenced the observed differences between following and not following advice. Future experiments that randomly interleave trials of tasks with and without advice should further investigate this issue.
Still, the current experiment allowed us to rule out a number of alternative mechanisms that could a priori explain advice following. Among these, the brain could provide a greater expected reward signal for the recommended option. Alternatively, choosing the non-recommended option could be associated with anticipated regret, or negative feedback for the recommended option could lead to particularly strong regret. Moreover, outcomes from the recommended option could be processed with greater attention. Our behavioral and fMRI results do not support these alternative hypotheses. First, the decision phase was not characterized by a greater reward anticipation signal in the VST or the VMPFC when participants chose the recommended deck. Rather, the change in BOLD signal in a number of brain regions was smaller when choices were made and advice was followed (see Figure S9). This replicates the findings of an earlier study, which examined the effect of advice on investment decisions [35] and reported reduced activity in decision-related regions during advice trials. We did not find a greater change in the BOLD signal in regret-associated regions like the anterior cingulate cortex or the orbitofrontal cortex [36] during the choice or during the processing of negative feedback from non-recommended decks. Our fMRI results are also inconsistent with an attention account as we found that the reward signal in the VMPFC, as identified by contrasting gains and losses after not following advice, had a greater magnitude after not following compared to following advice. Similar results were reported for a study that compared orbitofrontal reward signals of self-determined and instructed choices [37]. Finally, simple attention effects cannot be reconciled readily neither with our behavioral finding that participants still learn which of the non-recommended decks is better nor with the notion that they prefer the recommended to the non-recommended of two options with the same expected value. Instead, our results suggest that advice modulates reward processing in two ways. First, the standard reward signal in the VMPFC and VST is attenuated. Second, the septal area and the left caudate head implement an outcome-bonus for recommended options. Importantly, the outcome-bonus signal does not replace the standard reward signal but seems to influence learning in addition to an attenuated standard reward signal.
Prior neuroimaging research on decision making in social contexts addressed the differences between social and nonsocial cognition [38] and the computational processes underlying decision making and learning in a social context [39]. Notably, recent studies showed that a social prediction error signal predicts future conformity with humans and computers [14] and that, when advice is given on a trial-by-trial level during strategic interaction, the brain tracks the quality of advice through social reinforcement learning signals [13].
We discovered that, on a neurobiological level, the human propensity for following trustworthy advice could be explained by the modulation of the neural reward response. Importantly, the outcome-bonus does not replace the standard reward signal. Instead, it supplements a still present, though attenuated, learning signal in the VMPFC and the VST (where a partial outcome-bonus is implemented). More specifically, only the septal area and the left caudate implement the full outcome-bonus signal. These regions signaled a more positive evaluation of outcomes after following advice and were also sensitive to rewards after not following advice. Notably, the septal area is ideally suited as the neural substrate of the outcome-bonus, because it contains neurons that mediate reinforcement [28] and project to nuclei in the hypothalamus that release oxytocin [29], a neurotransmitter known to facilitate trust [30]. Accordingly, a recent study showed greater activity in the septal area during trusting behavior [34]. Hence, our findings suggest that an intrinsic reward signal in the septal area facilitates trust, which, in turn, would facilitate future advice-following. The correlation of the outcome-bonus estimated for individual participants and the difference of positive reward signals in the amygdala after following versus not following advice suggests that this structure is also involved in maintaining the influence of advice. This result is plausible, as the amygdala is known to be involved in the detection of trust from faces [40] during social interaction [41] and in the generation of reward prediction errors during learning [39],[42].
The notion of intrinsic reward for following advice may seem counterintuitive, particularly because a Bayesian approach would suggest that advice influences expectations prior to individual experience. From an adaptive perspective, the relevant criterion to choose a social learning mechanism is the amount of reward that can be accumulated using a specific mechanism. Bayesian models are optimal in the sense that they accurately learn expected payoffs. This does not imply, however, that these models also accumulate most rewards because the obtained rewards depend also on how choices are derived from expected payoffs. Hence, when advice is predominantly good, the outcome-bonus model performs well as it biases choices persistently in the direction of the recommended option, whereas the prior model affects choices only initially.
Another interesting result is that the models implementing a dynamic outcome-bonus did not explain participants' behavior as well as the models implementing a stable outcome-bonus. We ascribe this to characteristics of our task designed to mimic everyday situations of advice following, in which the task at hand is often difficult, and the recommendation comes from a competent and motivated advisor. Hence, future research is needed to show whether the influence of advice is stable or dynamic when the task is relatively easy or the competence of the advisor is less uncertain.
The notion of intrinsic reward for following advice is consistent with both a learning and an evolutionary perspective. The effect of advice on reward representation suggests that following advice acts similarly to a secondary reinforcer. Following advice alone, which is usually followed by positive outcomes, elicits a reward response. Likewise, it has been proposed that imitation—another form of social learning—has the quality of a secondary reinforcer for children, who frequently experience that imitation leads to positive outcomes [43]. Mathematical analysis shows that the propensity for social learning can evolve on the population level in the environmental conditions that characterized the era of human evolutionary adaptation [2],[4]. Moreover, social learning can solve problems that individual learning cannot, such as cooperation in social dilemmas or the accumulation of knowledge across generations [2],[6],[7],[44]. Therefore, it seems plausible that humans have evolved mechanisms for social learning [5],[45]. We complement these explanations by providing a neurobiological account of an adaptive social learning mechanism, which can also explain the human propensity for social learning. Importantly, insights into the neurobiological mechanisms underlying social learning can pave the way for a targeted search of genetic adaptations to social learning. Based on our results, one could speculate that genetic adaptation to social learning involves genes that modulate reward processing.
In conclusion, we present evidence that the brain's reward system implements an adaptive social learning mechanism by generating a greater reward signal for outcomes received after following trustworthy advice. This outcome-bonus could also explain maladaptive social learning, which should occur particularly when the difference between choice options is hard to detect or when social influence is strong. Indeed, others have reported that decision makers will follow advice that implies sub-optimal decisions when decisions are difficult or contain a dilemma [7] and that social influence determines preferential choice beyond the quality of consumed goods [46]. Fundamentally, our results advance the understanding of social learning by providing a neurobiological account of the human propensity for social learning and of the sustained influence of social information on learning and decision making.
Twenty-one right-handed healthy participants performed a four-armed bandit task with 168 trials while being scanned in an MRI scanner. All participants were free of neurological and psychiatric history and gave written informed consent in accord with local ethics. An additional 10 participants were recruited to function as advisors for participants in the fMRI experiment.
Participants in the fMRI experiment received advice from a randomly selected advisor before entering the MRI scanner. To establish incentives for following advice, we truthfully informed participants that the advisor had performed the same task before and that the advisor's payment partially depended on the receiver's earnings. This design comes close to natural settings of advice-giving and -taking, where the advisor is motivated to give good advice, but the advice-receiver still cannot be entirely sure whether she receives the best advice.
In the learning task (performed in the MRI scanner), participants repeatedly chose from four card decks and received feedback after each trial (Figure 1 and Text S1). The four decks were comprised of two identical “good decks” with a high positive expected value and two identical “bad decks” with a low positive expected value (see Figure S1). Therefore, preference for the recommended deck over the corresponding deck with the same payoff distribution would be a clear indicator of the influence of advice. To examine the effect of advice on rewards and punishments, each card deck generated 50% positive and 50% negative payoffs across all trials. The bad decks had slightly higher gains but much larger losses than the good decks.
To investigate the influence of advice on learning, we first compared how a standard reinforcement learning model, an “outcome-bonus” model, a “prior” model, and a combined “prior+outcome-bonus” model described participants' choices.
The standard reinforcement learning model assigns each option i an expected reward qi(t). On the basis of the expected rewards, choices are made according to the softmax choice rule [47], which determines the probability pi(t) of choosing option i of the J options in round t as follows:(1)where τ is a sensitivity parameter determining how likely the option with the largest expected reward will be chosen. Note that this choice function holds for all trials except the first, for which we assumed that the decision maker chooses the recommended option. This assumption was implemented in all tested models.
After a choice is made, the expected rewards are updated on the basis of the prediction error. That is, the deviation between the expected and actually received reward:(2)where ri(t) is the reward obtained from choosing deck i in trial t and α is the learning rate that determines the impact of the prediction error in the updating process.
The outcome-bonus model differs from the standard reinforcement-learning model by changing the reinforcement of outcomes from recommended options. Accordingly, the updating rule (Equation 2) was modified such that when the recommended option was chosen, a constant bonus was added to the objective reward:(3)where g(i) is an indicator function that takes the value 1 if option i is recommended and the value 0 if option i is not recommended, βb is a free outcome-bonus parameter capturing the level of social influence, and μ is the expected payoff from choosing randomly among all options and serves as a normalization constant to allow for comparison across tasks with different payoff magnitudes.
The prior model assumes a higher initial reward expectation for the recommended choice option. Hence, the initial reward expectation in the prior model is defined as , where βp captures the social influence on the prior expectations and N is the number of trials in the learning experiment, which we chose as a simple scaling factor, allowing for the comparison of the weight of the prior compared to the payoff that can be obtained in the experiment. For the combined prior+outcome-bonus model, both the initial reward expectation and the outcome-bonus were used to modify the evaluation of the choice options.
Additionally to the aforementioned models, we examined (a) a dynamic version of the outcome-bonus that becomes increasingly large after good advice and increasingly small and negative after bad advice. We also tested various other modifications of the outcome-bonus model, which (b) combined dynamic outcome-bonus and higher prior reward expectation for the recommended option, restricted the outcome-bonus to only (c) gains or (d) losses, (e) assumed that losses after following advice are processed as zero payoffs (see Text S1 for details).
For all models, we estimated the parameter values that maximized the log likelihood of trial-by-trial choice predictions for each participant separately (see Text S1). Model comparison was performed based on AIC and BIC values, which are derived from the log likelihood but additionally penalize models with a greater number of free parameters.
The functional analysis was based on 12 regressors (plus six motion-parameter regressors): Two regressors modeled the choice of the recommended and the non-recommended option(s), respectively. Four binary regressors modeled (a) positive and (b) negative feedback after choosing the recommended option and (c) positive and (d) negative feedback after non-recommended options, respectively. An additional set of four corresponding parametric regressors controlled for feedback magnitude. One regressor modeled prediction errors estimated with the combined prior+outcome-bonus model. One error regressor modeled feedback after missed trials, in which participants made no decisions. For group-level results, individual-level contrasts were averaged using the FMRIB Local Analysis of Mixed Effects module in FSL (see Text S1), and one-sample t tests were performed at each voxel for each contrast of interest.
To identify regions implementing advice and reward-sensitive feedback signals, we used the four regressors (a) through (d), described above. Advice-sensitive regions were identified by the contrast [1 1 −1 −1] for these regressors. To test if the resulting functional regions of interest (ROIs) were also reward-sensitive, we tested these voxels with the contrast [0 0 1 −1], based on the assumption that, after not following advice, feedback allows for the identification of reward responses that are uncontaminated by advice. An additional whole brain contrast, comparing gains and losses after not following advice [0 0 1 −1], revealed commonly reported reward signals in the VST and the VMPFC.
Following our predictions, we investigated representations of reward in a set of anatomically defined regions comprising the major reward-representing areas of the brain. We defined the reward ROIs based on the Harvard-Oxford subcortical structural atlas and included the following anatomical regions: nucleus accumbens, caudate, putamen, thalamus, medial frontal cortex, and amygdala. For the amygdala, Z (Gaussianized T) statistic images were thresholded with a small volume correction determined by z>2.576, and a minimum cluster size of 36 voxels determined with the AFNI AlphaSim tool (see Text S1). For the ROI comprising all other regions, Z statistic images were thresholded with a small volume correction determined by z>2.576 and a minimum cluster size of 92 voxels, also determined with the AFNI AlphaSim tool.
|
10.1371/journal.pntd.0000809 | Roles of Small GTPase Rac1 in the Regulation of Actin Cytoskeleton during Dengue Virus Infection | Increased vascular permeability is a hallmark feature in severe dengue virus (DV) infection, and dysfunction of endothelial cells has been speculated to contribute in the pathogenesis of dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS). Rho-family GTPase Rac1 is a significant element of endothelial barrier function regulation and has been implicated in the regulation of actin remodeling and intercellular junction formation. Yet there is little evidence linking Rac1 GTPase to alteration in endothelial cell function induced by DV infection.
Here, we showed that actin is essential for DV serotype 2 (DV2) entry into and release from ECV304 cells, and Rac1 signaling is involved these processes. At early infection, actin cytoskeleton rearranged significantly during 1 hour post infection, and disrupting actin filament dynamics with jasplakinolide or cytochalasin D reduced DV2 entry. DV2 entry induced reduction of Rac1 activity within 1 hour post infection. The expression of dominant-negative forms of Rac1 established that DV2 entry is negatively regulated by Rac1. At late infection, actin drugs also inhibited the DV2 release and induced accumulation of viral proteins in the cytoplasm. Meanwhile, the activity of Rac1 increased significantly with the progression of DV2 infection and was up-regulated in transfected cells expressing E protein. Confocal microscopy showed that DV2 E protein was closely associated with either actin or Rac1 in DV2-infected cells. The interaction between E protein and actin was further confirmed by co-immunoprecipitation assay.
These results defined roles for actin integrity in DV2 entry and release, and indicated evidence for the participation of Rac1 signaling pathways in DV2-induced actin reorganizations and E-actin interaction. Our results may provide further insight into the pathogenesis of DHF/DSS.
| An important clinical characteristic of dengue hemorrhagic fever/dengue shock syndrome is increased vascular permeability. Actin cytoskeleton is a significant element of endothelial barrier function regulation. In vitro study showed that dengue virus infection could induce redistributions of actin cytoskeleton. It is not precisely clear the roles of actin and the mechanisms of its reorganization during the infection. Using immunochemical assays, drug inhibition assays and protein interaction profiling methods, we aimed to identify the ways in which dengue virus serotype 2 interacts with actin cytoskeleton. The study showed that dynamic treadmilling of actin is necessary for dengue virus entry, production and release, while small GTPase Rac1 also plays multiple roles during these processes. In addition, we demonstrated the association of viral E protein with actin, indicating a direct effect of viral protein on the structural modifications of actin cytoskeleton. Our results provide evidence for the participation of Rac1 signaling pathways in viral protein-induced actin reorganizations, which may be a mechanism involved in the etiology of dengue hemorrhagic fever.
| Dengue virus (DV) is an enveloped, single-stranded RNA virus belonging to the family Flaviviridae. The DV genome has one open reading frame encoding three structural proteins - capsid, membrane and envelope (E)- that constitute the virus particle, and seven nonstructural proteins. DV infection causes a wide range of symptoms from a mild disease (dengue fever, DF) to severe, life-threatening complications (dengue hemorrhagic fever/dengue shock syndrome, DHF/DSS). The characteristics of DHF/DSS are abnormalities in hemostasis and increased vascular permeability. Sudden and extensive plasma leakage in tissue spaces and various serous cavities of the body in patients with DHF may result in profound shock – DSS – that can be fatal if not clinically managed in time [1]. However, the mechanism of the increased vascular permeability induced by DV infection is not clear yet. Autopsy studies showed rare apoptotic endothelial cells and no severely damaged capillaries vessels, though capillaries in several organs showed endothelial swelling [2]. It seemed that increased vascular permeability without morphological destruction of capillary endothelium is the cardinal feature of DHF/DSS [3].
Dynamics of cytoskeletal and cytoskeleton-associated proteins is a significant element of endothelial barrier function regulation. Actin cytoskeleton, linking to the cytoplasmic tail of junctional adhesive proteins as well as extracellular matrix protein, is relevant in the stabilization of inter-cellular junctions and the maintenance of endothelium integrity. In our previous study, increased permeability of monolayer of ECV304 cells without obvious morphological destruction was observed in DV2-infected cell culture model [4], and β3 integrin, which is an extracellular matrix protein and plays central roles in maintaining capillary integrity, showed an up-regulating expression in human dermal microvascular endothelial cells after DV2 infection [5]. Additionally, several groups also reported that DV infection induce alterations in actin cytoskeletal assembly and junctional protein complexes in human vascular endothelial cells in vitro [6]–[8]. Therefore, it was inferred that actin rearrangement induced by DV infection may contribute to the dysfunction of endothelial barrier, which in turn cause increase of vascular permeability.
Actin and the associated vesicle fission machinery act in concert to liberate nascent vesicles from both the plasma membrane and trans-Golgi network [9]. Recent work revealed that some viruses that enter via receptor-mediated endocytosis and bud at plasma or endosomal membrane recruit cellular actin cytoskeleton for these purposes [10], [11]. After virus infection, actin cytoskeleton may also interact with viral structural proteins, such as envelope protein of another flavivirus (West Nile virus), and the nucleocapsid protein of a hantavirus (Black Creek Canal virus) [12], [13]. As observed by ourselves and others, DV also enter host cells via clathrin-dependent endocytosis, and viral nucleocapsids are released into the extracellular by budding [14], [15]. It has been implicated that actin rearrangement is required for DV entry into the mosquito C6/36 cells and mammalian HepG2 cells [16], [17]. The components involved in regulating actin rearrangements of the infected cultural system remain to be identified.
Rac1 GTPase, a member of Rho GTPases family, has been implicated in the negative regulation of clathrin-mediated endocytosis [18]. Rac1 also has a central role in regulating both actin cytoskeletal remodeling and the integrity of intercellular junctions. In the absence of vasoactive stimuli, dominant negative Rac1 increases endothelial permeability and affects adherens and tight junctions. Recent studies showed Rac1 GTPase is essential for many virus infections. HIV Env-mediated syncytium formation relies on Rac1 activation in target cells [19]. Early herpes simplex virus type 1 infection is dependent on regulated Rac1 signalling in epithelial cells [20]. Rac1 activation is also required for hepatitis B virus replication and Coxsackievirus movement [21], [22]. A very recent study reported that Rac1 and Cdc42 regulates formation of filopodia required for DV2 entry into HMEC-1 cells [23]. However, current knowledge about the mechanism of actin organization induced by DV infection is still limited.
Here we sought to investigate the roles of Rac1 GTPase and the participation of viral E protein in actin rearrangements of ECV304 cells induced by DV2 infection. Our data showed that actin cytoskeleton is required for DV2 infection, during which Rac1 GTPase may play different roles at various stages. In the early stage, DV2 entry reduces activity of Rac1 GTPase and DV2 entry is negatively regulated by Rac1 GTPase. In contrast, Rac1 is activated by DV2 infection in the late stage and also by the expression of viral E protein. Confocal study indicated active Rac1 may be involved in the interaction between actin and E protein, and hence actin reorganizations. Taken together, these results suggested multiple roles of Rac1 GTPase in DV2 infection.
Aedes albopictus mosquito (C6/36) cells and endothelial-like ECV 304 cells (European Collection of Cell Cultures) were grown in Dulbecco's Eagle's minimum essential medium (DMEM, GIBCO) containing 10% fetal bovine serum (FBS). Vero cells were grown in Eagle's minimum essential medium (MEM, GIBCO) with 5% FBS.
DV2 (strain Tr1751) virus was isolated from a patient with dengue fever and propagated in C6/36 cells and stored at −70°C until used. Viral titers were detected by plaque assay, using a Vero cell monolayer culture under 1% methylcellulose overlay medium.
Rabbit anti-E protein of DV2 polyclonal antibodies (PAb) and mouse anti DV2 PAb were produced in our laboratory. Mouse monoclonal antibodies (MAb) 301 and 504 against Japanese encephalitis virus [24],which cross-reacted with E protein of DV2, were kindly provided by Dr. Yasui K (Tokyo Metropolitan Institute for Neuroscience, Japan). FITC-conjugated goat anti-mouse immunoglubin (IgG), rabbit anti-actin PAb, anti-Rac1 MAb, phalloidin–TRITC, and protease inhibitor cocktail were from Sigma. Cytochalasin D (Cyt D), jasplakinolide (Jas), and 3-(4,5-dimethyl thiazol-2yl)-2,55-diphenyltetrazolium bromide (MTT) were from Merck. Lipofectamine reagents were from Invitrogen. pReceiver-M01α vector was from Stratgene. Glutathione sepharose 4B and anti GST antibody were from Amersham.
To attain the vector expressing DV2 E protein (GenBank accession number: L10053.1), the coding sequences of E genes (from 877 to 2421 bp, containing the gene coding signal peptides) were amplified and cloned into the Nsp5 and XhoΙ sites of pReceiver-M01α,named pRec-E. The sequences of PCR generated fragment was verified by DNA sequencing. Then ECV304 cells were transfected with plasmids pRec-E by using Lipofectamine 2000 according to the manufacturer's instruction. The distributions of recombinant DV2 E proteins with actin filaments were analyzed by immunofluorescence assay at 48 h post-transfection. To study the effect of E proteins on Rac1 activity, cells stably expressing pRec-E or pReceiver-M01α were obtained after selection under G418 (400 µg/ml) for 10 days and named as ECV/pRec-E and ECV/pRec respectively.
To attain the vectors expressing Rac1 mutants, the wild-type (WT) forms and constitutively-inactivated mutants of Rac1 were amplified from Rac1 cDNAs, pEXV3-Rac1V12N17 and pEXV3-Rac1N17 respectively [25], with sequence specific primers and cloned in the sense orientation into the Nsp5 and XhoΙ sites of pReceiver-M01α. The sequences of all PCR generated fragments were verified by DNA sequencing. ECV304 cells stably transfected with the His-tagged WT, V12N17, and N17 mutants of Rac1 plasmids were established by using Lipofectamine 2000 according to the manufacturer's instruction. And each cell line carrying either a WT or mutant Rac1 plasmid was first characterized by a GTP-loading assay to confirm its phenotype. These cell lines were used for studying effect of Rac1 on DV2 infection.
Cells were inoculated with DV2 (MOI = 10) and incubated at 37°C for 1 h. Then the inoculum was removed and the cells were treated with acid glycine (pH 3.0) solution for 2 min to inactivate extracellular virus. Afterwards, cell samples were collected and the titer of virus in each sample was determined by standard plaque assays on Vero cells. The titer of mock treatment (DMSO control) was considered as 100%. Experiments were performed in duplicate for at least three independent experiments and the titers were averaged.
Two drugs, Jas and Cyt D, were stored at −20°C as 1 mM or 100 µM stock solutions in dimethyl sulfoxide (DMSO) respectively. The cytotoxicity of each drug to ECV304 cells was determined, using the 3-(4,5-dimethylthiazol-2-yl)-diphenyl tetrazolium bromide (MTT, Sigma) method. From this, 100 nM of Jas and 4 µM of Cyt D were used as the highest working concentrations, and DMSO was added at the same concentration to the control cells (mock treatment).
To study the effects of actin inhibitor on DV2 entry, ECV304 cells were grown to confluence in 24-well culture plates and pretreated with DMEM containing Jas (100, 50, or 10 nM), Cyt D (4, 2, or 0.2 µM), or 0.1% of DMSO (mock-treatment) for 5, 3, or 1 h at 37°C, respectively. Then the infection was performed and the entry efficiency of DV2 was determined as described above.
To study the effects of actin inhibitors on virus yield, Cyt D or Jas was added to the infected cells at different time points after viral entry. ECV304 cells were grown to confluence in 6-well culture plates, and inoculated with DV2 (MOI = 1) at 37°C for 1 h. Medium containing drugs (100 nM of Jas or 2 µM of Cyt D) was added to the cells at the initiation of infection, then the cells were fixed at 1 d or 3 d post infection (p.i.) and virus antigens were detected with immunofluorescence assay as described below. Medium containing drugs was also added at the indicated time points post infection (0 h p.i., 3 h p.i., or 6 h p.i. respectively) and left on for the entire duration of the 9-h incubation. Control cells were incubated with medium containing 0.1% of DMSO (mock-treatment). At 9 h p.i., 2 ml of supernatant was saved and cells were frozen and thawed three times in 2 ml of DMEM. Viral titer of each sample was determined in duplicate by plaque assay and plotted as percentages of the titer in the samples of control cells.
Infected or mock-infected ECV304 cells (approximately 5×106) were rinsed in cold phosphate-buffered saline (PBS) and lysed in 500 µl of lysis buffer (10 mM Tris HCl [pH 7.5], 1 mM EDTA, 100 mM NaCl, 1% NP-40, 1% protease inhibitor cocktail) for 30 min on ice. After centrifugation at 12,000 rpm for 10 min, the clarified cell lysate was mixed with MAbs (mixture of 301 and 504) and rocked end-over-end at 4°C overnight. Subsequently, 20 µl of protein A-Sepharose beads (50% slurry) was added and incubated for 1 h at 4°C. The beads were collected and washed three times in the lysis buffer. The bound proteins were dissociated by boiling for 5 min in the Laemmli sample buffer and separated on 12% polyacrylamide gels.
CRIB was expressed in E. coli strain BL21 as a GST fusion protein and immobilized by binding to glutathione-Sepharose beads. Briefly, the cDNA sequence of CRIB domain was amplified from mouse brain and cloned into the pGEX-6p-1. Then, E. coli transformed with recombinant plasmids was grown at 37°C overnight in 2xYT medium. The next morning, the culture was diluted 1∶20 into fresh 2xYT medium and cells were induced with 0.5 mM ipTG (isopropyl-beta-D-thiogalactopyranoside), and growth was continued for 24 h at 25°C. The cells were harvested with centrifugation at 5000 g for 10 min, and resuspended in PBS (50 µl per ml of culture). Then the cells were lysed by sonication on ice, incubated on ice for 20 min in the presence of 1%Triton X-100. After centrifugation at 8000 g for 10 min, the soluble fraction was incubated with glutathione sepharose 4B at 4°C, with gentle agitation overnight. The beads were then washed three times with 10 ml PBS-1% Triton X-100 and two times with 10 ml PBS. The washed GST-CRIB beads were kept at 4°C in the presence of sodium azide and a cocktail of protease inhibitors.
Activated Rac1 was identified by binding specifically to the GST-fused p21-binding domain of human Pak1. Briefly, ECV304 cells (approximately 5×106) were seeded and serum starved for about 24 h. After infection with DV2, cells were washed twice in ice-cold PBS and lysed in 500 µl of lysis buffer (10 mM Tris HCl [pH 7.5], 1 mM EDTA, 100 mM NaCl, 1% NP-40, 1% protease inhibitor cocktail) at various time points p.i. The lysates were clarified, normalized to equal amounts of total proteins, and incubated with glutathione beads containing bound GST-CRIB for 90 min at 4°C. Bound Rac1 was resolved by sodium dodecyl sulfate-12% polyacrylamide gel electrophoresis (SDS-12% PAGE) and immunoblotted with monoclonal antibodies against Rac1. Immunoreactive bands were visualized, and band intensities were assessed. The bands were scanned, and their intensities were assessed and quantified, with GTP bound Rac1 in mock-infected cells being considered one fold activation for comparison to infected cells.
A novel affinity binding assay has been developed for in situ detection of active form of small GTPases in mammalian cultured cells [26]. Briefly, bound GST-CRIB was eluted off the beads with elution buffer (50 mM Tris-HCl, 10 mM reduced glutathione, pH 8.0) and stored at −80°C before use. ECV304 cells were infected with DV2 ( MOI = 1) for 24 h and then cells were fixed in 2% paraformaldehyde buffer on ice for 1 min and then incubated with GST-CRIB (100 µg/ml) on ice for 10 min. After washing, cells were fixed again in paraformaldehyde buffer on ice for 10 min. Subsequently, GST-CRIB was detected with an anti-GST antibody and a secondary antibody conjugated with horseradish peroxidase.
Cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.2% Triton-X 100 for 5 min. After washing and blocking with 1% bovine serum albumin (BSA) in PBS, polymerized actin was detected by incubating the cells with phalloidin-TRITC (1∶100, 1 h at 37°C). For detections of actin filaments and DV2 antigens or E protein, cells were incubated with mouse anti-DV2 PAb (1∶100, overnight at 4°C) or with a mixture of MAb 301 and 504 (1∶100, overnight at 4°C), and subsequently with a mixture of FITC-conjugated goat anti-mouse IgG (1∶100) and phalloidin-TRITC (1∶100, 1 h at 37°C). For detections of Rac1 and E protein, cells were incubated with primary mouse anti-Rac1 MAb (1∶100, overnight at 4°C) and FITC-conjugated secondary antibody (1∶100, 1 h at 37°C). Then cells were incubated with rabbit anti-E PAb (1∶100, overnight at 4°C) and TRITC-conjugated secondary antibody (1∶100, 1 h at 37°C). Following washing with PBS, slides were mounted with mounting medium and examined under a fluorescent microscope (Olympus, BX-51) or a confocal laser microscope (Leika TCS-NT). The relative virus positive area was measured by Image Proplus 5.0 program and the area of control cells was considered as 100%.
The statistical significance was accessed by two-tailed paired student's t-test.
Actin cytoskeletal assembly/disassembly dynamics are critical for many aspects of clathrin-coated structure dynamics including assembly, constriction, internalization, and lateral motility [27]. Recent studies have shown that WNV and DV enter the mammalian cells in clathrin-mediated endocytosis, but direct functional evidence for a role of actin during DV2 entry is lacking. In this study, the addition of DV2 caused rapid reorganizations of F-actin network in ECV304 cells within 10 min p.i. (Fig. 1A). Actin fibers disassembled and dispersed drastically throughout the cytoplasm at 30 min p.i., then repolymerized into stress fibers along the cell edge at 1 h p.i., indicating that DV2 infection could induce reorganization of actin cytoskeleton.
To investigate the effect of the interrupted actin cytoskeleton on virus entry, ECV304 cells were pretreated with Jas or Cyt D, and then infected with DV2 at 37°C for 1 h. As shown in Fig. 1B, the virus entry into drug-treated cells was inhibited in a time and dose dependent manner when compared with that in mock treated cells. At 1 h of treatment, about 40%, 19%, and 15% reduction at 4, 2, 0.2 µM of Cyt D were induced, and there was no significant inhibition at 50 or 10 nM of Jas and a little reduction with about 13% at 100 nM of Jas. About 50%, 39%, and 27% reduction at 4, 2, 0.2 µM of Cyt D, and about 25%, 20%, and 17% reduction at 100, 50, 10 nM of Jas were observed at 5 h of treatment, respectively. These results showed that actin dynamics is essential for the DV2 entry.
Rac1 is one of the best characterized GTPases and has been implicated in negative regulation of clathrin-mediated endocytosis. To determine the Rac1 GTPase activity during DV2 entry into ECV304 cells, cell lysates from early time points p.i. were used in GST-CRIB pull-down assay. There is no obvious change in amount of total Rac1 during 1 h p.i. In contrast, Rac1 GTPase activity decreased as early as 15 min p.i. (0.3-fold) and the lowest value (0.2-fold) was seen at 30 min p.i. Thereafter, it showed a little recovery tendency with about 0.3-fold at 60 min p.i. (Fig. 2A).
In the dominant negative Rac1-N17 and Rac1-V12N17 mutants, substitution of threonine 17 for asparagines behaved as a dominant inhibitor of endogenous Rac1 function. To study the effect of dominant-negative Rac1 on virus entry, we established ECV304 cell lines expressing WT, V12N17, and N17 forms of Rac1. The cells were infected with DV2 for 1 hour and intracellular viral titers were measured. As shown in Fig. 2B, as compared with WT cells, entry of DV2 increased about 62% in cells expressing Rac1-N17 and 105% in cells expressing Rac1-V12N17, respectively. This confirms that Rac1- GTPase functions as a negative regulator for DV2 entry.
DV2 has previously been shown to egress by budding at the plasma membrane of infected cells, but relatively little is known about the mechanism involved in this mode of release. To access the effect of interrupted actin cytoskeleton on virus release, 2 µM Cyt D or 100 nM Jas was added at the initiation of DV2 infection and distributions of viral proteins were detected at 24 and 72 h p.i. by immunostaining. In mock-treated cells, viral antigens were generally localized at perinuclear area. While in drug-treated cells, more viral antigens accumulated throughout the cytoplasm at 24 h p.i. and the accumulation became more significantly at 72 h p.i. (Fig. 3A and 3B). The observation suggested that disturbing of actin network could inhibit the release of DV2.
Additionally, cells were treated with the two drugs (2 µM Cyt D or 100 nM Jas) from different time points post infection, and the virus titers in the medium and cell fraction were determined 9 h p.i. and were plotted as percentages of the titer in the supernatant or cell fraction of mock-treatment cells. As shown in Fig. 3C and 3D, both Cyt D and Jas could induce the reduction of supernatant viral titer, and more obvious reduction was seen in case of treatment with Cyt D (Fig. 3C). Interestingly, there was a drastic decrease in intracellular virus titers at treatment with both drugs from 0 h p.i., but only a little decrease at treatment from 3 and 6 h p.i. (Fig. 3D).
To confirm that the reduction was due to an inhibition of virus release, the ratios of extra- to intracellular infectious particles (Fig. 3E) were determined. Treatments with Cyt D at various time periods all reduced ratios significantly. In Cyt D-treated cells, the ratio was reduced to 81% when treated from 0 h p.i.; then the ratio was down to about 50% when treated from both 3 h p.i. and 6 h p.i. In contrast, treatments with Jas induce the reduction of ratio in a time-dependent manner. In Jas-treated cells, the ratio was reduced to 61% when treated from 0 h p.i.; the ratio gradually recovered to 68.6% when treated from 3 h p.i., and to about 100% when treated from 6 h p.i. As the decreased ratios indicated that the drugs have more inhibitory effects on extracellular viral titer, in combination with the result of immunostaining, the data showed that disturbing actin cytoskeleton partially blocked DV2 release. Furthermore, Cyt D showed stronger effects than Jas, suggesting that DV2 release might depend more on actin treadmilling than stabled actin filaments.
To determine the Rac1 GTPase activity during DV2 infection, cell lysates from 1, 12, 24 h p.i. were used in GST-CRIB pull-down assay. As shown in Fig. 4A, Rac1-GTPase activity increased gradually with the progression of infection, from 2.3 fold at 12 h p.i. to 3.1 fold at 24 h p.i. Meanwhile, there was no obvious change in the total cellular Rac1 level after DV2 infection. This indicated the steady-state level of endogenous Rac1 in these cells and thus demonstrated DV2 infection induced the activation of preexisting endogenous Rac1.
To investigate the effect of dominant-negative Rac1 on DV2 infection, the cells expressing WT, V12N17, N17 forms of Rac1 were infected with DV2 for 1 hour, then supernatant and cellular samples were collected at 24 h p.i. for viral titration. As compared with Rac1-WT cells, viral titers in supernatant decreased to about 60% and 45% in cells expressing Rac1-V12N17 and Rac1-N17, respectively (Fig. 4B). Meanwhile, the intracellular virus titers were decreased to 70% around in cells expressing Rac1-V12N17 and to about 50% in cells expressing Rac1-N17 (Fig. 4C). Generally, the results indicated that suppression of Rac1 inhibited both extra- and intracellular titers of DV2 significantly. Meanwhile, we found that the dominant forms of Rac1 only led to slight reduction in ratios of extra- to intracellular infectious particles. The ratio was about 90% and 85% in cells expressing Rac1-V12N17 and Rac1-N17, respectively (Fig. 4D). Together, the results indicated Rac1 GTPase has some influence on the virus production or assembly, but little effect on DV2 release.
To gain insights into the mechanism by which activity of Rac1 GTPase is modulated in DV2 infection, we analyzed the distributions of Rac1 and viral envelope proteins first. Rac1 showed a diffused distribution throughout the cytoplasm in mock-infected cells. Compared with that, an enrichment of Rac1 was observed in perinuclear region of infected cells at 24 h p.i., and highly colocalized with E proteins at the same area (Fig. 5A). Accordingly, in-situ detection assay for Rac1 GTP showed that Rac1 is activated in the perinuclear region of infected cells at 24 h p.i. (Fig. 5B), whereas the activated Rac1 was rarely observed in mock-infected cells.
The activation of Rac1 by E protein was further confirmed by pull-down assays in cell lines stable-transfected with pRec or pRec-E. Cells were grown for 48 h to confluence and then cellular samples were collected for pull-down assays. As compared with ECV/pRec cells, expression of E protein in ECV/pRec-E cells induced an increase of about 5-fold in Rac1 GTPase activity (Fig. 5C). In combination with the close colocalization of E protein and Rac1, these data indicated that DV2 E protein the may trigger the activation of endogenous Rac1, which may involve in the rearrangements of actin filaments induced by DV2 infection.
To clarify whether E proteins induced actin rearrangements during DV2 infection, actin rearrangements and distribution of DV2 E proteins were detected with phalloidin–TRITC and anti-E MAb, and further analyzed with confocal immunofluorescent assay. At 1 day p.i., actin rearrangement was seen in viral antigen-positive cells. With infection progress, the actin disorganizations became more evident at 3 day p.i., when most stress fibers were disrupted and F-actin was present significantly in the periphery ruffles of cells (Fig. 6A). Actin filaments were colocalized with DV2 E protein in the perinuclear region of infected ECV304 cells. Accordingly, reorganized actin filaments also highly colocalized with transiently expressed E proteins in transfected ECV304 cells with pRec-E (Fig. 6B). This indicated the possible interaction between E protein and actin microfilaments during DV2 infection.
To further confirm the interaction between them, co-IPs were carried out to examine lysates of DV2-infected ECV304 cells. Cell lysates were immunoprecipitated with anti-E antibodies followed by SDS-PAGE and immunoblot analysis with anti-actin antibody. An actin-E protein complex could be co-immunoprecipitated using anti-E antibodies (Fig. 6C). Co-IP was also performed with DV2- infected suckling mouse brain, and similar result was obtained (data not shown). The reverse immunoprecipitation was performed using anti-actin antibodies followed by immunoblot analysis with anti-E antibodies. However, the association of E with actin could not be observed. This may be due to the fact that the cytoplasma is very rich in actin filaments, so that the loading quantity of 10 µg per lane is below the detection level of E protein. In the subsequent co-IPs, we used anti-E antibodies to first carry out the immunoprecipitation. With the progression of infection, the amount of actin associated with viral E proteins was increased and peaked at 3 d p.i. (Fig. 6D). The results strongly suggested the close association of E proteins with the actin filaments. Taken together, these data indicated E is an important viral component in the signal pathway triggering actin rearrangements during the infection.
Flaviviruses enter the host cells by receptor-mediated endocytosis. DC-SIGN (dendritic-cell-specific ICAM-grabbing non-integrin), GRP78/BiP (glucose-regulating protein 78), and CD14-associated molecules have been suggested as primary receptors for DV [28]–[31]. We and other groups showed that β3 integrin, a prominent endothelial cell receptor, is identified as the functional receptor necessary for WNV and DV2 entry into vertebrate cells [5]. Subsequent to the initial interaction of ligand/virus with the cellular receptor, an internalization signal is usually triggered to facilitate the endocytosis of the virus into cells.
DV, as well as WNV, particles enter mammalian and mosquito cells by clathrin-dependent endocytosis [14], [15], [17], [32], Actin cytoskeleton is critical for clathrin-coated structure dynamics and macropinocytosis [27]. In this study, significant repolymerizations of actin cytoskeleton at early infection were observed, and disruption of actin filaments with Jas or Cyt D inhibited significantly the entry process of DV2 in a time and dose dependent manner. Our results are consistent with recent reports in which, it was found that inhibition of macropinocytosis with Cyt D caused a reduction in DV2 entry, and the inhibition effects of Cyt D depend upon the serotype of the DV to some extent [16], [17]. All together, these results suggested that DV2 may manipulates actin cytoskeleton and the associated signaling cascades to facilitate its entry.
Cytoskeleton is also believed to be play active roles in viral intracellular transport, maturation and release. Previously, we observed that both vimentin and microtubule were reorganized in DV2 infection [33], and role for actin in the release of WNV has been reported by others [11]. Here, we also observed significant repolymerizations of actin cytoskeleton during DV2 infection and accumulation of large amount of DV2 antigens in cells treated by two actin inhibitors, indicating that actin also involved in DV2 release. However, there was a significant difference in inhibition potency between these two drugs. As drug incubation time decreased, the ratio of extra- to intracellular virus in cells treated with Jas gradually recover to the control level, whereas the ratios of cells treated with Cyt D kept down to 50%. The findings are very interesting because these two inhibitors destabilize actin dynamics through different mechanisms: Cyt D inhibits the polymerization of subunits by binding to the plus-ends of the actin filaments, while Jas, a potent inducer of actin polymerization, enhances actin stabilization by inhibiting the depolymerization of actin filaments. The decrease of extracellular DV2 yields could be resulted from the inhibition of the growing actin barbed ends, which are essential for generating the vectorial force for membrane bending and virus budding. Our findings indicated that the availability of monomeric actin to form newly filamentous actin is essential for DV2 release. In other words, DV2 release might depend more on actin treadmilling than stabled actin. In combination with WNV results reported by others, it is speculated that utilizing actin may be a common mechanism for flaviviruses release.
Next, we are interested whether E protein, which is an important viral structural protein and serves as viral adsorption protein, involve in the interaction of DV2 and actin. In this study, the high co-localization of DV2 E protein with actin was observed by confocal microscopy analysis, and their interaction was further confirmed by co-IP assay. Moreover, both the amount of actin associating with E proteins (Fig. 6A) and the level of reorganization of actin cytoskeleton (Fig. 6D) are increased with the progression of infection. These indicated the close association of DV2 E protein with actin. Our results are consistent with a recent study reported by Chu [12], who found that actin is associated with protein C from 6 h p.i. and then with protein E from 10 h p.i. in WNV infected cells, suggesting that the association between actin and viral structural proteins may be common in flaviviruses. Moreover, previous studies have demonstrated that rearrangements of actin filaments and overexpression of integrinβ3 induced by DV2 in endothelial cells may contribute to the increase in vascular permeability observed in DHF/DSS [5]–[7]. Gelatinolytic matrix metalloproteinase-9 and matrix metalloproteinase-2 overproduced by DV2-infected immature dendritic cells down-regulate the expression levels of cell adhesion molecules, induce the redistribution of the actin filaments, and enhance endothelial permeability [34]. All together, it was speculated that a direct effect of viral protein may also contribute to the structural modifications of actin cytoskeleton in infected cells, and to the increased permeability of the endothelial cells during DV infection, especially when large amounts of virons are produced and released.
The negative regulation of clathrin-mediated endocytosis of cell surface receptors mediated by Rac1 has been implicated in both polarized and non-polarized cells. In MDCK cells, Rac1V12 expression decreased the rates of apical and basolateral endocytosis [35]. In Hela cells, overexpression of constitutively active forms of Rac1 causes inhibition of transferrin and EGF receptor internalization [18], [26]. One mechanism by which Rac1 might participate in clathrin-mediated endocytosis is the modulation of phosphatidylinositol (PI) lipid metabolism, which plays a key role in this process [36]. Rac1 has been shown to interact with a number of enzymes that regulate PI metabolism, including type I PI 3 -kinase and synaptojanin 2 [37], [38]. Recently, a guanine nucleotide exchange factor splice variant designated Ost-III is identified as a regulator of Rac1 involved in the inhibition of receptor endocytosis [26].
In this study, we observed that Rac1 GTPase activity decreased at the early stage of DV infection. The role of Rac1 GTPase in regulating the DV2 infection was clearly demonstrated by the increase of virus entry in cells expressing Rac1-V12N17 or Rac1-N17. The results suggested that the initial mediator of transient Rac1 inactivation may be viral receptors in response to DV2 binding and/or internalization. This observation was supported by an earlier study showing that GTPase Rab 5, which regulates transport Rac1 to early endosomes, was essential for cellular entry of both WNV and DV [14]. Since the Rac1 dominant negative mutants show a reduced affinity for nucleotides, the increased DV2 entry observed may be due to the abrogation of downstream signaling of active Rac1.
It is noteworthy that Rac1 activity is up-regulated in the late phase of DV2 infection, and dominant negative Rac1 inhibited DV2 production and release. Further study showed that Rac1 is activated and colocalized with E protein at perinuclear area in DV2- infected cells, and Rac1 activity increased in cells expressing E protein. These results raise the possibility that Rac1 activated by viral protein is involved in the reorganization of actin cytoskeleton after DV2 infection. Rac1 has been implicated in actin nucleation and membrane ruffling. Rac1 and the adapter protein Nck activate actin nucleation through WAVE1 (WASP -family verprolin homologous protein), PAK and Arp2/3 complex [39], [40]. A recent study revealed that there is also a critical function for Rac1 in tight junction assembly. Tight junction assembly was disrupted profoundly in mammalian epithelial cells depleted of Par-3 while Rac1 was constitutively activated, and the assembly of tight junctions is efficiently restored by a dominant-negative Rac1 mutant [41]. These studies support the hypothesis that activated endogenous Rac1 contributes to the rearrangements of actin filaments and intercellular junctions at later phase of DV infection.
The dysfunction of vascular endothelium and increased vascular permeability is the hallmark of severe DV infection in humans. Previous studies showed that the phenomenon of antibody-dependent enhancement plays an important role in the immunopathogenic aspect of DHF. In secondary heterotypic infection, pre-existing, non-neutralizing antibody induces virus-antibody complex binding to Fc-receptor bearing cells, such as dentritic cells and monocytes/macrophages, and results in increased T-cell activation as well as higher serum cytokine levels. Then the virus escapes and infects greater numbers of cells which can lead to a greater viral load. High dengue viremia during dengue fever or viremia detected after defervescence have been shown to associate with development of DHF [42], [43]. Vascular endothelium, a barrier between blood and tissue, will directly interact with the viral particles and virus-antibody complexes during the viremia. Dengue antigens have been demonstrated in biopsies of lung, liver and brain vascular endothelium of patients [44]–[46]. In this study, our results showed a possible correlation between the accumulation of virus envelope proteins and the endothelium dysfunction, which could be a mechanism involved in the etiology of DHF/DSS.
Taken together, the results demonstrated that dynamic treadmilling of actin is necessary for DV2 entry, production and release, and Rac1 GTPase also plays critical roles in those processes. We also confirmed the interaction between E protein and actin, which indicated a direct effect of viral protein on the structural modifications of actin cytoskeleton. Further understanding of the various components of signal cascades manipulated by DV would lead to a better understanding of DV interactions with host cells and their outcomes, all of which would eventually lead to the development of better control measures against DV infection.
|
10.1371/journal.pntd.0003872 | Plasmodium vivax Diversity and Population Structure across Four Continents | Plasmodium vivax is the geographically most widespread human malaria parasite. To analyze patterns of microsatellite diversity and population structure across countries of different transmission intensity, genotyping data from 11 microsatellite markers was either generated or compiled from 841 isolates from four continents collected in 1999–2008. Diversity was highest in South-East Asia (mean allelic richness 10.0–12.8), intermediate in the South Pacific (8.1–9.9) Madagascar and Sudan (7.9–8.4), and lowest in South America and Central Asia (5.5–7.2). A reduced panel of only 3 markers was sufficient to identify approx. 90% of all haplotypes in South Pacific, African and SE-Asian populations, but only 60–80% in Latin American populations, suggesting that typing of 2–6 markers, depending on the level of endemicity, is sufficient for epidemiological studies. Clustering analysis showed distinct clusters in Peru and Brazil, but little sub-structuring was observed within Africa, SE-Asia or the South Pacific. Isolates from Uzbekistan were exceptional, as a near-clonal parasite population was observed that was clearly separated from all other populations (FST>0.2). Outside Central Asia FST values were highest (0.11–0.16) between South American and all other populations, and lowest (0.04–0.07) between populations from South-East Asia and the South Pacific. These comparisons between P. vivax populations from four continents indicated that not only transmission intensity, but also geographical isolation affect diversity and population structure. However, the high effective population size results in slow changes of these parameters. This persistency must be taken into account when assessing the impact of control programs on the genetic structure of parasite populations.
| Plasmodium vivax is the predominant malaria parasite in Latin America, Asia and the South Pacific. Different factors are expected to shape diversity and population structure across continents, e.g. transmission intensity which is much lower in South America as compared to Southeast-Asia and the South Pacific, or geographical isolation of P. vivax populations in the South Pacific. We have compiled data from 841 isolates from South and Central America, Africa, Central Asia, Southeast-Asia and the South Pacific typed with a panel of 11 microsatellite markers. Diversity was highest in Southeast-Asia, where transmission is intermediate-high and migration of infected hosts is high, and lowest in South America and Central Asia where malaria transmission is low and focal. Reducing the panel of microsatellites showed that 2–6 markers are sufficient for genotyping for most drug trials and epidemiological studies, as these markers can identify >90% of all haplotypes. Parasites clustered according to continental origin, with high population differentiation between South American and Central Asian populations and the other populations, and lowest differences between Southeast-Asia and the South Pacific. Current attempts to reduce malaria transmission might change this pattern, but only after transmission is reduced for an extended period of time.
| Plasmodium vivax is the human malaria parasite with the largest geographical expansion, and the predominant malaria parasite outside of Africa [1]. Transmission intensity (according to annual parasite incidence as a surrogate measure) ranges from very low and seasonal in temperate zones and in countries approaching malaria elimination to very high mainly in Asian and South Pacific countries [1]. Prior to malaria control starting early in the 20th century, P. vivax transmission even occurred in large parts of Europe, Russia and the US [2]. P. vivax is difficult to control due to relapsing liver stages, fast and constant formation of gametocytes and a large proportion of asymptomatic carriers contributing to transmission [3, 4]. As a consequence, P. vivax has become the predominant malaria parasite in several countries where P. falciparum transmission has been successfully reduced [5, 6].
Along with other parameters, parasite diversity can be used to assess the effect of interventions, as reduced transmission is expected to result in reduced diversity. This relationship was observed for P. falciparum [7, 8]. Moreover, knowledge on parasite diversity is the basis to study gene flow between populations, or to track the source of imported infections [9]. Thus, global comparisons of population genetic data help to develop and validate molecular tools for surveillance of antimalarial interventions.
Typing of highly polymorphic microsatellites has proven useful to describe the diversity and structure of parasite populations [10–14], to study patterns of relapses [15–17], multiplicity and molecular force of infection [18, 19], and for distinguishing reinfection from recrudescence in drug trials [20–22]. Several studies reported extensive P. vivax microsatellite diversity even in regions of moderate endemicity, and high multiplicity of infection was frequently observed. While some P. vivax populations showed pronounced structuring on small geographical scale [10, 13], this was not the case for other populations [11, 23].
Differences in method of sampling with respect to geographical space as well as different panels of markers used for genotyping make direct comparison of results difficult [24]. Lacking so far was a comprehensive comparison of P. vivax diversity based on samples collected across many different sites and typed with the same set of markers. Therefore we compiled data from published studies that included samples from Peru [10], Brazil [14, 21], Sudan [25], Cambodia [26], Vietnam [27], Papua New Guinea (PNG) and Solomon Islands [11], and complemented this dataset with previously unpublished typing results from Central America and Mexico, Madagascar and Central Asia (Armenia, Azerbaijan and Uzbekistan). All samples were typed with 11 published microsatellite markers [28].
This global data set included 841 isolates from regions of different levels of transmission intensity (Fig 1A and Table 1) and representing 6 out of 9 malaria transmission zones recently shown to differ in relapse patterns [29]. The highest P. vivax prevalence ever has been recorded in the lowlands of PNG (e.g. reaching >50% by PCR in children in East Sepik Province [19]). These South Pacific parasite populations are relatively isolated due to limited migration of human hosts. In Southeast (SE)-Asia transmission is also high, yet often focal [30]. Migration of hosts is high SE-Asia and a major complication of eradication efforts. In Latin America transmission is lower, but increased since the 1960s when the number of P. vivax cases was very low due to successful spraying campaigns [5]. Transmission in Central America is low, and parasite populations are separated from those in South America by the Isthmus of Panama, with no road connecting South and Central America. Also within Central America sub-structuring is likely; e.g. different P. vivax subpopulations were observed in Mexico, following vector distribution [31]. In Africa, P. vivax transmission occurs mostly in Madagascar and East Africa, i.e. Ethiopia and Sudan. In other parts of Sub-Saharan Africa P. vivax transmission is very low as most individuals carry the Duffy-negative blood type, which largely prevents P. vivax infection [32]. Thus, Madagascan parasites are isolated from other P. vivax populations in northern Africa, and transmission in Madagascar is relatively low. In Central Asia, transmission is very low; in Uzbekistan malaria had been eradicated in 1961, but was reintroduced later, and is characterized by small outbreaks in border areas [33].
The wide distribution of parasite populations included in this study permit for the first time assessing P. vivax microsatellite population structure on a global scale. Previous studies on P. vivax population structure across different continents mostly genotyped polymorphic antigens [34, 35] or mitochondrial DNA [36, 37]. These markers differ from microsatellites because antigens are under immune selection and mitochondria are maternally inherited, thus excluding recombination. It remains unclear whether similar results would be obtained from analyses of mtDNA, antigen-coding genes or of putatively neutral microsatellites. The present study allowed a global comparison of parasite microsatellite diversity and population structure made possible by harmonization of methods, and the definition of a minimal subset of markers for epidemiological studies.
Informed written consent was obtained from all individuals or their parents or guardians prior to the study. Details on ethical approval are published for samples from Peru [10], Brazil [14, 21], Sudan [25], Vietnam [27], Cambodia [26] and the South Pacific [11]. For samples collected from returning travelers to the US (samples from Central America, Mexico, Africa, India and Indonesia) the study protocol was approved by the Ethical Committee for Research with Human Subjects of the Institute of Biomedical Sciences, University of São Paulo (960/CEP). In Madagascar, the study protocol was approved by the Ethics Committee of the Ministry of Health of Madagascar (007/SANPF/2007). The study was approved by the WEHI Human Research Ethics Committee.
Details of samples included in this study are given in Table 1. In case of cohort studies, only 1 sample per individual was included. From Africa, India and Indonesia samples from returning travelers were utilized [38]. While the number of travelers’ samples was too small to assess intra-population diversity, linkage disequilibrium, or FST compared to other populations, they were still useful for clustering analysis and principal component analysis (PCA).
Samples from Central America and Mexico had been collected from travelers returning from several countries spanning from Panama to Mexico. Due to the limited number of isolates and lack of precise information on sample origin (some samples derived from travelers visiting several countries), these isolates were combined as ‘Central American’ population, despite possible sub-structuring. From Armenia and Azerbaijan 14 isolates were available; these were combined because migration between both countries is frequent and it was not always clear in which of the two countries the infection had been acquired. From Uzbekistan 20 isolates were available. For calculation of allelic richness all Central Asian countries were pooled to reach the required number of 25 samples.
All samples were typed with the same set of 11 published microsatellite markers [28]. Three additional markers of that panel were excluded. MS3 and MS16 had not been typed in all populations, and MS8 showed signals of positive selection and variation in allele sizing between labs (details below). Microsatellites are considered neutral markers; however, some of them might lie within coding regions or be linked to genes under selection. Selection had been tested using Lositan software [39]. Only marker MS8 showed a weak signal for positive selection. A relationship between P. vivax microsatellite alleles and clinical disease or acquired immunity has never been reported, thus it was not to be expected that different age groups sampled across populations or different proportions of febrile and asymptomatic individuals would influence population genetic parameters.
Minor differences in typing protocols did not affect the results; i.e. samples from PNG, Solomon Islands, Madagascar Sudan and Central Asia were amplified by nested PCR, while a single round of PCR was performed on all other samples. As the nested PCR primers were identical to the single-round primers, allele sizes can be compared directly. From each lab a subset of samples was typed again starting from DNA to ensure comparability of results. Allele lengths were very consistent for all markers, except for MS8, where variation of up to 1.5 base pairs was observed. Due to this variation and possible positive selection MS8 was excluded from analysis.
In case of multi-clone infections the predominant peak only was included into the analysis. Occasionally this can result in incorrect haplotype assembly. This affects analysis such as linkage disequilibrium or clustering, where individual haplotypes are needed. Diversity and FST values are not affected because allelic frequencies are assessed at population level. Clustering analysis and calculation of LD were repeated with only those samples that harbored a single allele at each marker. To obtain sufficient samples from Madagascar and Cambodia, we permitted in the analysis also isolates with >1 alleles at one of the markers. It should be noted that isolates of low multiplicity were selected for genotyping for several parasite populations. Thus the difference between the number of samples of the full data set (including predominant peak haplotypes) and the number of only single clone infections does not reflect the proportion of multiple clone infections.
Alleles were binned using TANDEM software [40] and formatted using PGDspider [41]. Expected heterozygosity (HE) of markers and allelic richness were calculated using FSTAT [42]. HE is the chance that two unrelated parasites carry a different allele of a given marker, and allelic richness is a measure of alleles per each marker adjusted for the different numbers of isolates per site. When the effective population size is reduced, e.g. as consequence of intensified malaria control, rare alleles are expected to disappear first. As a result, allelic richness changes more rapidly than HE, as the influence of low-frequency alleles on HE is small. The number of unique haplotypes was calculated by Dropout [43] and linkage disequilibrium (LD) by LIAN 3.5 with 100,000-fold re-sampling [44]. LIAN compares the observed association of markers to the values expected for random association based on the population diversity. Only unique haplotypes with no missing data were included for calculating LD, resulting in 633 samples in this analysis. Three of the markers used (MS2, MS4, MS5) localize to chromosome 6 and two markers (MS12, MS15) to chromosome 5, thus these markers are physically linked. To assess linkage disequilibrium irrespective of physical linkage of markers, LD was also calculated excluding markers MS2, MS4 and MS12, i.e. with 8 markers located on 8 different chromosomes. As compared to the 11-marker panel, fewer isolates had to be excluded due to missing data, but identical 8-marker haplotypes occurred more often, resulting in 637 samples for this analysis.
Relatedness between haplotypes was assessed by pairwise comparison of all samples within a population and calculating the proportion of shared alleles. Only samples with at least 7 markers available for comparison were included.
Effective population size Ne (i.e. the estimated number of unique haplotypes circulating in each site) was calculated using step-wise mutations models (SMM) as well as infinite allele models (IAM), using mutation rates observed in P. falciparum studies of 1.59*10−4 (95% confidence interval = 3.7*10−4, 6.98*10−5) [45]. While some of the markers harbor simple tri-nucleotide repeats, and SMM are likely applicable, other markers contain more complex repeat structures (e.g. MS2, MS6, MS10, MS20) and IAM are more appropriate [46], thus both values are given.
The software STRUCTURE was used to assess clustering of isolates [47]. This method detects clusters without prior information on the origin of samples. Twenty iterations for K = 1 to K = 12 (K being the number of clusters) were run, each with a burn-in period of 10’000 steps and then 100’000 MCMC iterations. A method developed through simulation studies [48] was applied to estimate the most likely number of clusters. In addition, the optimal number of clusters was assessed using the program STRUCTURAMA [49]. FST values among populations were calculated using FSTAT [42]. To compute Principal Components Analysis (PCA) the smartPCA application of EIGENSOFT was used [50]. In contrast to STRUCTURE analysis, PCA attempts to maximize variance between populations based on the known origin of samples. While smartPCA was designed for SNPs it can be used with microsatellites; each microsatellite allele was treated as a SNP. Preliminary analysis had shown no population substructuring between samples collected in Brazil in 2004 and 2006 [21], in the lowlands of PNG [11] and in Madagascar, thus samples were combined for calculation of FST values and PCA.
In the absence of the same measures of transmission intensity for all populations, such as entomological inoculation rate, force of infection or parasite prevalence, populations were broadly classified as low, medium and high transmission (Fig 2).
A total of 841 P. vivax isolates were available for this study. Fig 1A shows the origin of isolates on a published map of P. vivax endemicity in 2010, represented as parasite rate [1]. All isolates were typed for the same set of 11 microsatellite markers. For all 841 isolates genotyping results were obtained from ≥7 markers. For marker MS4, no amplification product was obtained from 70 isolates (i.e. data was obtained from 91.7% of all isolates), for all other markers data was obtained from 823–834 (97.9–99.2%) out of 841 isolates.
Pronounced differences in population diversity were observed. While expected heterozygosity (HE) was generally high, it was strongly reduced in isolates from Uzbekistan. With the exception of Uzbekistan, HE of all except 2 markers (MS5 and MS7) was >0.5 in all populations (Table 2, supplementary S1 File). Mean HE of 11 markers was lowest in Uzbekistan (0.52), followed by Azerbaijan (0.67) and South America (0.68–0.71), intermediate in Africa and the South Pacific (0.77–0.83), and highest in South-East Asia (0.84–0.87). The differences between continents were slightly higher when only those markers with lower overall diversity were assessed (MS1, MS4, MS5, MS7, MS12, mean HE<0.75). Mean HE ranged from 0.53 in Uzbekistan and 0.6 in Peru to 0.85 in Cambodia. Mean allelic richness showed a similar pattern (Table 2 and Fig 2) with lowest values in South America (5.5–7.0 alleles/locus) and Central Asia (7.1 alleles/locus), intermediate values in Africa and the South Pacific (7.9–9.9 alleles/locus), and highest in Cambodia (13.4 alleles/locus).
When all isolates and all markers were analyzed, linkage disequilibrium (LD) was strong and significant in South American, Madagascan, Central Asian and SE-Asian populations (Table 3). No or limited LD was detected in Central America, Sudan and the South Pacific. Trends were similar when only single-clone infections were analyzed, with exception of Peru and Madagascar, where LD observed in the analysis of all data was no longer detected. Overall levels of LD were lower in the single-clone data set. This can be explained by the reduction in sample size, since similarly low levels of LD were observed in an equally low number of randomly selected multi-clone infections. Most representative results were obtained when only 1 marker per chromosome was included, thus excluding physical linkage of markers (total of 8 markers). All populations in the South Pacific, Peru, Central America and Sudan were in full linkage equilibrium (Table 3 and Fig 2).
Effective population Ne size was about twice as high in Cambodia and Vietnam (IAM: 6378 and 5553) as compared to South America (2423–2829, Table 4). Values for Madagascar, Sudan and the South Pacific were intermediate, and low for Azerbaijan (2422) and Uzbekistan (1606). Estimates based on SMM were 2–3 fold higher. Samples from Central America and Mexico were highly diverse and showed high Ne (5159), likely because they originated from different countries and thus represent different subpopulations.
Across all populations a total of 759 individual haplotypes were found in 818 isolates. 11-loci haplotypes were shared only within the same country. Again isolates from Uzbekistan were unusual, as 10 out of 20 isolates were identical and another two shared the same haplotype. Ten haplotypes were observed more than once in Peru, 5 and 2 in Brazil in 2004 and 2006 (plus 1 shared between the 2004 and 2006 data set), 5 in Sudan, 2 in Azerbaijan, 2 in Vietnam (1 observed 5 times), and 1 in PNG. 730/818 haplotypes were singletons.
Relatedness among haplotypes was calculated for each population (Fig 2). In the pairwise comparisons isolates from Cambodia and Vietnam shared the same allele on average in 12.6% and 15.0% of markers (i.e. a mean of 1.4 and 1.65 out of 11 microsatellites carried the same allele). Samples from Central American, African and South Pacific populations shared the same allele in 16.5–22.5% of markers. Relatedness was 29.3–32.5% in South American populations, and in Azerbaijan 33.3% and in Uzbekistan 48.5% of markers shared the same allele.
To assess discrimination power of a smaller panel of markers compared to the full set of markers, microsatellites were removed successively and haplotype counts recorded from each individual population, as well as for the full dataset (Fig 3). MS4 was removed first, as no data for this marker could be obtained from 70 isolates. The remaining 10 markers were sequentially removed according to their diversity across all populations, starting with the least diverse one.
Only taking into account haplotypes shared within populations, a set of 4 markers (MS2, MS9, MS15, MS20) identified 710/760 (93.4%) haplotypes and 3 markers (MS2, MS15, MS20) still identified 665/760 (87.5%) haplotypes. By reducing to 2 markers (i.e. MS15 and MS10), the number of haplotypes dropped considerably to 513/760 (67.5%). The proportion of haplotypes lost by omitting markers was highest in South American populations. When using only 3 markers, 13–38% of haplotypes were lost in South American populations, but <10% in populations from SE-Asia and the South Pacific (Fig 3).
Taking into account haplotypes observed in several populations, a single haplotype was shared between Peru and Vietnam when the panel was reduced to 5 markers. With four markers 15 additional haplotypes were shared. While some were shared between populations from the same continent (5 identical 4-loci haplotypes in 251 isolates from PNG), 11 haplotypes occurred on different continents (e.g. Peru/Azerbaijan, Madagascar/Vietnam, Brazil/Cambodia/Vietnam, Sudan/Solomon Islands). As a consequence, when haplotypes shared between populations were taken into account, 4 markers identified 687/759 (91.5%) haplotypes, and 3 markers 595/759 (78.4%) haplotypes (black line in Fig 3, panel 1).
Clustering analysis indicated clearly distinct clusters, mainly following geographical lines (Fig 1B). First South American samples (but not those from Central America) and samples from Azerbaijan were separated from all other populations, with admixture in samples from Central America and Africa. When the number of clusters (K) was 3, the South Pacific populations formed a separate cluster, as well as the Brazilian samples, while samples from Peru, Central America, Madagascar and Asia clustered together. Peruvian samples formed a separate cluster when K was set to 4, K = 5 led to the separation of Madagascar, Sudan and Azerbaijan from SE-Asia. Central American samples formed an individual cluster when K = 6. The clonal population in Uzbekistan clustered with different populations in individual STRUCTURE runs for a given value of K, indicating that no clear relationship to any other population was observed. When the optimal number of clusters was calculated as described [48], high values were observed for K = 2, K = 5 and K = 7 (Fig 4). In addition the program STRUCTURAMA was used to assess the best number of clusters. The 99% CI for estimates of cluster number showed a wide range (79–115 clusters) and was thus not informative
As the clear separation of Latin American and South Pacific isolates from those from Africa and Asia might interfere with more subtle population structure within Africa and Asia, clustering analysis by STRUCTURE was repeated including the latter isolates only (Fig 5A). For K = 2 African and Azerbaijani isolates clustered separately from SE-Asian ones. Indian isolates clustered with Africa. As with the full set of isolates, the clonal samples from Uzbekistan clustered with different populations in individual runs for the same number of clusters (e.g. with Africa or with SE-Asia if K = 2). For K = 3 these isolates formed a separate cluster. Both within Africa and SE-Asia admixture was high, and when the number of clusters was set to 4 or 5, the separation between countries was limited. Analysis with STRUCTURE was repeated for single clone infections only. Results were similar as for the full data set (Fig 5B).
Results of clustering analysis were confirmed by FST values (Table 5). Between Brazil and Peru FST values were high (0.16), as well as when South American populations were compared to non-American populations (0.11–0.29). Differences between Cambodia, Vietnam and South Pacific populations were lowest (0.026–0.066). The highest FST value was observed between Azerbaijan and Uzbekistan (0.37), and all values for comparisons between Uzbekistan and other populations were >0.2. FST between Madagascar and Sudan was low (0.07).
FST values were also compared between continents or sub-continents, i.e. South America, Africa, Central Asia, SE-Asia and the South Pacific (Table 6). FST was highest between South America and the South Pacific as well as between Central Asia and all other populations (0.11), and lowest between SE-Asia and the South Pacific (0.042) and SE-Asia and Africa (0.058).
In principal component analysis (PCA), PC1 differentiated isolates from Brazil and Peru on the one hand and the South Pacific on the other hand from a cluster containing African and Asian isolates (Fig 6). PC2 separated isolates from Peru und Brazil.
In summary, clustering analysis, FST values and PCA showed similar results. South American populations were separated from all others, and populations from Brazil and Peru formed clearly separated groups with little admixture. African, SE-Asian and South Pacific populations each formed a large cluster with little sub-structuring. Samples from Central America, Indonesia and India grouped with SE-Asia.
Analysis of 841 P. vivax isolates of global origin, typed with the same markers, allowed direct comparison between populations differing in transmission intensity, geographical isolation and history of malaria control. Determination of the optimal set of microsatellite markers required for differentiation of individual parasite infections will improve strategies for genotyping. Knowledge on population structure can be used to assess the effect of interventions and help to track imported infections.
Diversity of a marker, expressed as expected heterozygosity (HE), is a key criterion for choosing a subset of microsatellite markers for a variety of genotyping questions [51]. Markers of a lower diversity are more suitable to assess population structure, as with less diverse markers, a smaller number of isolates is needed to detect genetic differences among populations. In contrast, many applied genotyping studies aim at distinguishing between “identical” and “different” clones. Samples from treatment failures in a drug trial for example require distinction between new infection versus recrudescence (reappearance of a pre-treatment clone) [18–22]. In this scenario the phylogenetic relationship of genotypes observed is not of interest. Such applied typing tasks often include large numbers of isolates, and thus a minimal set of markers is desirable that is able to reduce the genotyping workload and price without impairing the discrimination power for distinguishing clones. For phylogeographical studies a larger panel of markers is needed. The same is required for tracking the origin of imported malaria cases.
The step-wise removal of markers showed that as little as 3 highly diverse markers were sufficient to detect around 90% of all haplotypes in most populations, only in South American populations up to 40% of haplotypes were missed. Thus, typing only 2–3 markers in SE-Asia, the South Pacific and Africa, and 4–6 markers in South America would lead to only a small underestimation of multiplicity of infection or of treatment efficiency (when a clone during follow-up could not be distinguished from the clone at baseline and a new infection was taken for a recrudescence).
Other limitations affect drug trial results, such as imperfect detectability of minority clones [52] and as a consequence substantial day-to-day variation in the alleles detected [53]. Longitudinal P. vivax studies involving genotyping are also complicated by relapsing hypnozoites, which can be homologous or heterologous to the clone at baseline [16, 54]. Longitudinal studies can also be affected by re-infection with a genotype observed earlier in the same individual. This is a particular threat when the overall diversity in the parasite population is low, as observed in Uzbekistan, or in case of clonal expansion during an outbreak [46]. Thus, when the population diversity is intermediate or high, using a reduced panel of markers is acceptable as this reduces the ability to differentiate clones only in a minor way and is justified in view of substantial savings in time and costs.
Overall parasite diversity measured as HE or allelic richness was high and reflected transmission levels. It was lowest in South America and Central Asia, where transmission is low, and highest in SE-Asia. In the South Pacific diversity was intermediate despite highest levels of transmission (prevalence >50% in Ilaita [19]). This could be due to the relative geographical isolation of South Pacific populations, and thus limited introduction of parasites from other regions. In contrast migration of infected hosts is high among SE-Asian countries, thus high parasite diversity can be maintained, even when transmission is reduced.
Linkage disequilibrium can be the result of selfing in the mosquito of male and female gametes from the same parasite clone [55], as opposed to recombination of different clones resulting in a break up of linkage. Recombination can occur if a mosquito feeds on a host infected with different parasite clones. The level of LD is expected to decrease with increasing transmission intensity, while diversity is expected to increase. In line with this expectation, LD was detected in Brazil, Central Asia and Madagascar, where transmission is low [1]. It is, however, noteworthy that levels of diversity did not fully correlate with transmission intensity. Highest diversity and strong LD was observed in SE-Asia. In the South Pacific an intermediate diversity but no LD was found. Different processes influence diversity and LD; likely the local ecology add to the high parasite diversity in SE-Asia, while the high proportion of individuals carrying multi-clone infections in the South Pacific (up to 75% of all infected [19]) lead to high levels of parasite recombination. High LD is also expected when closely related parasites are sampled. Yet no LD was observed in Ilaita in PNG, despite most dense sampling of all populations with 132 isolates collected across hamlets approx. 5 km from each other [56].
Incorrect assembly of multi-locus haplotypes in multi-clone infections within a host would be expected to lead to incorrect low levels of LD, but the opposite trend was observed: lower LD was found when only single-clone infections were included in the analysis. This unexpected finding was attributed to the smaller sample size after removing multi-clone infections.
Analysis of population structure revealed significant FST values between all populations. While clustering analysis and PCA differentiated among those populations separated by high FST values, no within-continent subdivision was observed in the South Pacific, Africa and SE-Asia. The difference in clustering between populations from Latin American (clearly separated clusters in Peru, Brazil and Central America and Mexico) on the one hand and from the South Pacific (no subdivision between different provinces in PNG and Solomon Islands) or Africa (no subdivision between Madagascar and Sudan) on the other hand is striking, given comparable distances between sites. While high levels of human migration in SE-Asia could explain parasite gene flow, no clusters were found in the South Pacific despite limited human movement (only air and sea transport between East Sepik and Madang provinces, and between PNG and Solomon Islands). Likewise limited sub-structuring was evident between Madagascar and Sudan, despite open sea and countries with very low P. vivax transmission separating sampling locations. In Central Asia, clear subdivision was observed between parasites east and west of the Caspian Sea. Parasites from Azerbaijan and Armenia clustered with those from Africa, while parasites from Uzbekistan were highly clonal and formed a separate cluster.
Distinct parasite subpopulations might be the result of expansion of parasite clones after the near elimination of malaria in the second half of the 20th century in some countries. In Latin America they could also reflect different independent introductions of P. vivax, as it has been shown for P. falciparum [57]. In Uzbekistan P. vivax had been reintroduced after its elimination, most likely from other Central Asian countries. In contrast, in SE-Asia and the South Pacific P. vivax was present much longer than in Latin America, and even during the peak of the eradications campaigns in the 1960s prevalence remained high [58]. Central American and Mexican samples were exceptional as they were highly diverse and LD was low despite low transmission intensity. This is most likely caused by the fact that these samples represent different isolated subpopulations over a large geographical range. Further studies involving additional parasite populations and different molecular markers are needed to establish differences between South and Central American samples, to understand why Central America clusters with SE-Asia, and to identify potential routes for P. vivax colonization of South and Central America.
Beside microsatellites, other molecular markers have been used to assess P. vivax population structure, most importantly mtDNA and polymorphic antigens. In agreement with the present study, sequencing of mtDNA identified a separate subgroup in Latin America (highest support of all subgroups in Bayesian tree analysis) [36], as well as highly diverse populations in Asia and the South Pacific [37]. However, no pronounced separation between South-Pacific, SE-Asian and some South American isolates was observed using mtDNA [36]. Isolates from South and Central America had been found in the same subgroup, yet, only 3 haplotypes from Central America were sequenced [36]. The same study found different subgroups in East Asia (China and Korea) and SE-Asia (Cambodia, Thailand and Indonesia), plus a third subgroup including isolates from locations across Asia and PNG. The present study includes no isolates from China or Korea, and only few from Indonesia, thus no such structure within East Asia could be found.
In concordance with microsatellite results, mtDNA diversity was highest in SE-Asia and high in the South Pacific [36]. In contrast to microsatellite-derived measurements of diversity, mtDNA diversity was lower for Madagascar, Central America and Africa, whereby results of the latter two populations are likely affected by a very small sample size [36]. The same study indicated that overall P. vivax diversity in Latin America was as high as in SE-Asia, despite locally reduced diversity, a result confirmed by analysis of 3 whole-genome sequences from Latin America [59]. The present study found microsatellite diversity across Central America and Mexico to be similarly high as that of SE-Asia, but when isolates from Peru and Brazil were combined diversity remained lower.
SNPs in antigens are expected to be under strong balancing selection, limiting their use to study the underlying population structure. In line with this, many antigens showed strong clustering but in contrast to microsatellites many clusters were shared between continents [34, 35]. Like microsatellites, AMA-1 and MSP-1 alleles from South Pacific populations showed very little admixture with any other parasite populations [34, 60]. DBP-II alleles in contrast were more evenly distributed across continents [35]. A study using putatively neutral SNPs covering a 200-kb genomic region confirmed subdivision between Brazil and SE-Asia [61], and a barcode of 42 SNPs across the genome was recently published and tested on a small number of clinical samples from three continents [62]. However, this barcode was not yet tested for assessing local population structure or for typing asymptomatic, low-density infections.
Continuous malaria control is expected to reduce parasite diversity and effective population size, and to increase differences between populations due to clonal expansion of remaining parasite strains [63]. Indeed, near-clonal expansion of parasites has been observed for P. falciparum in the highlands of PNG [8], in Solomon Islands [64] and in South America [7]. Likewise, Artemisinin-resistant clones have expanded in SE-Asia [65].
In striking contrast to these findings, nearly all studies assessing P. vivax diversity found high parasite diversity, even in countries now aiming to eliminate malaria [66–69]. The clonal expansion in Uzbekistan, a country that had successfully eliminated malaria in the 1960-ies, is the first such population structure reported for P. vivax. Low microsatellite diversity was also found in South Korea, where transmission has been low for decades and the parasite population is relatively isolated [70], as well as from a rural, isolated site in Peru [71]. The high P. vivax diversity in countries with low transmission likely indicates a high underlying effective population size and thus a large number of infected individuals. Two hallmarks of P. vivax biology add to this, namely hypnozoites in the liver, and a large proportion of asymptomatic, low-density infections that escape screenings conducted by light microscopy or rapid diagnostic test and thus a substantially underestimated parasite reservoir [4].
The isolates studied here were collected prior to the renewed call for malaria elimination. Only few studies have typed samples collected after up-scaling control measures, but diversity remained high [72–74]. It seems that control has little short-term effect on population size, and diversity measures changes slowly as long as the effective population size remains high (above 100 genetically distinct parasite clones) [75]. Therefore diversity measures will only be useful to assess the impact of control programs once transmission is very low after several years of intensified control.
In recent years malaria control has been intensified reducing prevalence and incidence in PNG [76], many Asian countries [69, 77] and South America [78]. It will be important to evaluate whether reduced prevalence is paralleled by increased sub-structuring on small scale, i.e. breaking up of the South Pacific and SE-Asian clusters, indicating local hotspots of transmission. A pronounced reduction of genetic diversity and increase in population structure will implicate success of control and interruption of parasite gene flow from neighboring populations.
In previously malaria-free regions, microsatellite typing can help to study outbreaks. Because of their high discrimination power between clones, genotyping outbreak samples can clarify whether a single clone was imported and spread across a local region, or whether steady gene flow from neighboring regions with ongoing transmission occurs, resulting in a diverse parasite population [9].
Microsatellite typing remains an important tool to study P. vivax, as it can be done in any lab equipped for PCR. For epidemiological studies and drug trials, a limited set of 2–6 markers, depending on transmission intensity, provides sufficient resolution to distinguish individual clones. The full panel of 11 microsatellite markers showed clear population structure on a global scale, and differences in diversity reflect transmission intensity and isolation of parasite populations. These population genetic measures could potentially be used as tools to measure the impact of control programs; however, due to the large effective population size even in countries of moderate endemicity, these parameters are likely to change slowly.
|
10.1371/journal.ppat.1002764 | The Interdomain Linker of AAV-2 Rep68 Is an Integral Part of Its Oligomerization Domain: Role of a Conserved SF3 Helicase Residue in Oligomerization | The four Rep proteins of adeno-associated virus (AAV) orchestrate all aspects of its viral life cycle, including transcription regulation, DNA replication, virus assembly, and site-specific integration of the viral genome into the human chromosome 19. All Rep proteins share a central SF3 superfamily helicase domain. In other SF3 members this domain is sufficient to induce oligomerization. However, the helicase domain in AAV Rep proteins (i.e. Rep40/Rep52) as shown by its monomeric characteristic, is not able to mediate stable oligomerization. This observation led us to hypothesize the existence of an as yet undefined structural determinant that regulates Rep oligomerization. In this document, we described a detailed structural comparison between the helicase domains of AAV-2 Rep proteins and those of the other SF3 members. This analysis shows a major structural difference residing in the small oligomerization sub-domain (OD) of Rep helicase domain. In addition, secondary structure prediction of the linker connecting the helicase domain to the origin-binding domain (OBD) indicates the potential to form α-helices. We demonstrate that mutant Rep40 constructs containing different lengths of the linker are able to form dimers, and in the presence of ATP/ADP, larger oligomers. We further identified an aromatic linker residue (Y224) that is critical for oligomerization, establishing it as a conserved signature motif in SF3 helicases. Mutation of this residue critically affects oligomerization as well as completely abolishes the ability to produce infectious virus. Taken together, our data support a model where the linker residues preceding the helicase domain fold into an α-helix that becomes an integral part of the helicase domain and is critical for the oligomerization and function of Rep68/78 proteins through cooperative interaction with the OBD and helicase domains.
| Viruses have to optimize the limited size of their genomes in order to generate the proteins required for infection and replication. Several mechanisms are used to accomplish this including the use of multiple promoters and alternative splicing. These processes generate gene products with diverse functions through the combinatorial assembly of a small number of protein domains. The small genome of the adeno-associated virus has two major open reading frames that generate seven proteins, four non-structural Rep proteins and three capsid proteins. The non-structural Rep proteins share a motor domain that uses hydrolysis of ATP to generate the conformational changes that drive DNA replication, transcriptional regulation, site-specific integration and the packing of viral genome into capsids. These functions depend upon the oligomerization of Rep proteins on specific DNA sites through the cooperation of the N-terminal origin binding domain and the C-terminal helicase domain. We provide evidence that the linker that connects the two domains is an integral feature of the helicase domain and contains a conserved aromatic residue that is critical for oligomerization. This residue emerges to be a signature motif of SF3 helicases and is also present in a subset of bacterial Rep proteins that support rolling circle replication mechanism.
| The four adeno-associated virus (AAV) Rep proteins are generated from a single open reading frame by the transcriptional use of two different promoters (p5 and p19) and subsequent alternative splicing mechanisms [1], [2], [3]. These reactions produce proteins that share three functional domains: an origin binding domain (OBD), a SF3 helicase domain and a putative zinc-finger domain [4], [5]. The combination of these domains imparts these proteins with striking multifunctionality. In particular, the larger proteins Rep78 and Rep68 function as initiators of DNA replication, transcriptional regulators, DNA helicases and as key factors in site-specific integration [6]. The smaller Rep proteins Rep40 and Rep52, play a critical role during packaging of viral DNA into preformed empty capsids, where they are thought to be part of the packaging motor complex [7], [8], [9]. Although in terms of domain architecture the AAV Rep proteins resemble other members of the SF3 protein family, the peculiar OBD with its additional nuclease activity and the complex character of their oligomeric properties, set them apart from other SF3 helicases such as simian virus 40 large T antigen (SV40-LTag) and papilloma virus E1 (PV-E1) proteins [10], [11], [12], [13]. In both of these proteins, the minimal SF3 helicase domain assembles into a hexameric ring in a process that can be induced by the presence of ATP and/or single-stranded DNA [14], [15]. In contrast, Rep40 containing only the helicase domain and Rep52 with an additional Zn-finger domain, appear to be monomeric [16], [17]. This indicates that oligomerization of AAV Rep proteins requires the presence of both the OBD domain and the helicase domain. This combination imparts both Rep68 and Rep78 with a complex and dynamic oligomeric behavior in-vitro that is modulated in large part by the nature of the DNA substrate [18]. The monomeric behavior of both Rep40 and Rep52 is striking in that they appear to contain the required structural features that are present in other SF3 helicase members. The X-ray structures of both SV40-LTag and PV-E1 show that their helicase domains assemble as hexameric rings and that the oligomerization interface is bipartite [15], [19]. One interface is formed by the interaction of neighbouring N-terminal oligomerization domains (OD). The second interface is formed by the interaction of the C-terminal AAA+ domains and is further stabilized by the presence of nucleotides [11], [15]. In order to understand the structural features that promote AAV Rep oligomerization, we pursued in this study a detailed structural comparison of SF3 helicases. We show that the OD domain in Rep40/52 has been hindered in its ability to oligomerize by the transcriptional use of the p19 promoter. This event generates proteins with a smaller OD domain as compared to other SF3 helicases. More importantly, we show that in the context of Rep68/78 the required oligomerization is supported by the interdomain linker which is directly involved in oligomerization interface and we provide evidence that the tyrosine residue preceding the start of Rep40/52 (Y224) is critical in the oligomerization and therefore activity of the large AAV Rep proteins. Taken together, our results support a model where oligomerization of Rep68/78 is mediated by a composite oligomerization interface formed by the OBD, helicase and linker domains, with the latter playing an essential role in the inducing the oligomerization process.
As a first step in our attempt to determine the structural features that promote oligomerization in AAV Rep proteins, we analyzed the oligomeric interface of SF3 family members SV40-LTag and PV-E1. As previously described, the helicase domain contains two subdomains: a N-terminal helical bundle of four α-helices known as the oligomerization domain (OD) and the C-terminal AAA+ subdomain (Figure 1A). In PV-E1 the oligomerization interface spans both subdomains forming two extended surfaces at opposite faces of the proteins. In the AAA+ subdomain, one face comprises all the catalytic residues, including: the P-loop, its subsequent helix, the β-strands with the associated Walker B residues, sensor 1 motif, and one side of the β-hairpin (Figure 1B). The neighboring subunit interacts through areas that are located in the α-helices “behind” the β-sheet and on the opposite side of the β-hairpin (Figure 1B). Overall, about 20% of the solvent accessible area takes part in the interface and includes about 34% of all residues. In PV-E1, the OD domain consists of 68 residues forming a four helical bundle. The oligomeric interface comes from interaction of residues located in helices 1 and 4 in one monomer, with residues in helices 2, 3 and part of helix 4 in the other subunit (Figure 1B). Most of the interface is hydrophobic with many tyrosine and isoleucine residues. Similar types of interactions are seen in the interface formed by the SV40-LTag OD domains. This domain is a lot bulkier, spanning 89 residues that form a five-helix bundle. The extra helix originates from an additional Zn-finger motif. Significantly, the OD of Rep40, on the other hand, has only 52 aminoacids and, thus, is significantly shorter than PV-E1 and SV40-LTag OD domains. The direct result of this difference is a decrease in the total accessible surface area by more than 1000 Å2. In addition, the packing of the helices is less compact, producing a more dynamic structure (Figure 1C). We hypothesize that the smaller OD domain of AAV Rep proteins imparts these proteins unique oligomeric properties where the smaller Rep40/52 are mostly monomeric while Rep68/78 -with the additional OBD domain- form oligomers. However, the measurable ATPase activity in all Rep proteins, suggest that Rep40/52 should oligomerize in the presence of nucleotides [20].
To determine if the presence of nucleotides can induce oligomerization of Rep40 -containing the minimal helicase domain-, we carried out sedimentation velocity experiments in the presence and absence of nucleotides at different concentrations. The sedimentation velocity profiles offer a complete characterization of the number and type of oligomers in solution. The data were analyzed using the program sedfit [21], [22]. Figure 2A shows plots of the c(s) distribution against the sedimentation coefficient (s) for two concentrations of Rep40 in the absence of nucleotides. A single peak whose s20,w increases slightly with increasing concentrations is observed. The slight but significant increase in s and calculated molar mass is consistent with a weak and transient dimerization (for hydrodynamic reasons, s is expected to decrease with increasing concentrations of an ideal solute). The data where also fitted using the program sedphat to a monomer-dimer association were the process is in rapid exchange on the time scale of the centrifuge [22]. Table 1 shows that the dissociation constant in the absence of nucleotides is ∼10−3 M, which is at the upper end of detection by sedimentation velocity. Similar distributions of Rep40 (at 36 µM) in the presence of either 5 mM ATP or ADP are shown in Figure 2B and 2C. Here an increase is observed in the width of these peaks if compared to those for Rep40 alone. This is a well-understood behavior for a associating system whose exchange kinetics are neither slow of fast on the time scale of the centrifuge, thus, broadening the c(s) distribution peak [23]. The presence of a small shoulder suggest that dimer formation is occurring here as well, although perhaps its rate of dissociation is slower than for Rep40 alone. The s-value of the shoulder is consistent with a transient Rep40 dimer that represents ∼0.2% of the total amount of protein. The relatively low ATPase activity of Rep40 reported in the literature supports our model of transient dimerization promoted by the binding and/or hydrolysis of ATP [20].
In order to assess whether the interdomain linker connecting the OBD domain and the helicase domains contains additional regions of distinct structure that may play a role in promoting oligomerization, we first carried out secondary structure prediction analysis to determined if the linker contains additional regions of structure. The results suggest that the region from residue 215 to 224 has the potential to form an α-helix (Figure 3A). We hypothesized that this region could extend the first helix of the OD domain (Figure 3A) and the ensuing increase in surface accessible area may be sufficient to drive oligomerization. To test this hypothesis, we designed a new Rep construct beginning at the start of the linker region and extending to aminoacid 536 (a truncated version of Rep68 without the OBD domain, Rep68Δ200), and performed sedimentation velocity and cross-linking studies in order to characterize its oligomerization properties. The sedimentation profile of Rep68ΔN200 shows the presence of two peaks, one corresponding to the monomeric species (∼2.53S) and the other to a dimer (∼3.71S). The amount of formed dimer increases at higher concentrations as expected from a monomer-dimer equilibrium system (Figure 3B). Formation of dimers was also observed when we performed cross-linking experiments. Figure 3C shows that the amount of dimeric species has significantly increased in Rep68ΔN200 as compared to Rep40wt. We calculated the dimerization constants of Rep40wt and Rep68ΔN200 from a global fitting of the sedimentation velocity data to a monomer-dimer model (Table 1). In summary, we determined that the presence of the linker region increases the strength of dimerization by about 10-fold relative to that of Rep40.
Next, we sought to determine the minimal length of linker that is needed to promote oligomerization. We generated three additional constructs, named Rep68ΔN209, Rep68ΔN214 and Rep68ΔN219 and tested their ability to oligomerize (Figure 4). Our results indicate that Rep68ΔN214 contains the minimal length of linker that is required to promote detectible oligomerization, although with the shorter construct Rep68ΔN219, a small shoulder is seen at higher concentration (data not shown). These results confirm that the linker region from 215 to 224 may fold into a α-helix, resulting in an increase of the surface accessible area of the OD domain that mediates oligomerization. This increase, however, is not sufficient to produce higher order oligomers.
In order to determine the contribution of ATP and ADP to the oligomerization of the extended linker Rep linker constructs, we performed sedimentation velocity studies in the presence of nucleotides. Our hypothesis was that if oligomerization reflects the functional state of these proteins, the addition of nucleotides should support and induce further oligomerization. Figure 5 shows that the presence of ATP and ADP induces the formation of higher order oligomers. Formation of dimeric species at this concentration can be seen with Rep68Δ214 as well as the longer constructs RepΔN209 and RepΔN200. In the later two, ADP produces two main populations sedimenting at ∼3S and ∼7S with additional intermediate oligomers. ATP on the other hand, seems to generate more stable species at ∼7S. Again, these data show that the presence of the linker region induces oligomerization of the Rep constructs and that the addition of nucleotides, in particular ATP, induces formation of larger oligomers, possibly through the stabilization of the interface formed by the AAA+ domains. This finding is in good agreement with the unique characteristics of the AAV Rep nucleotide binding pocket, which, based on its open conformation together with the presence of an arginine finger predicts the nucleotide contribution to oligomerization [24].
To determine if the linker is critical for the oligomerization of Rep68, we replaced it with an unrelated sequence and examined its effect on oligomerization using sedimentation velocity. The only prerequisite for the substitute linker were a lack of structure and no impact on the native structures of the connected domains. We chose a sequence from the transcription factor Oct-1. This transcription factor has two DNA binding domains connected by a linker of 29 residues. The X-ray structure of this protein shows that the linker is unstructured and flexible. In addition, it has been used to connect different protein domains without affecting their properties [25], [26]. We generated a Rep68 mutant protein (Rep68octlink), where residues 206 to 224 were replaced with 18 residues from the Oct-1 linker and tested its ability to oligomerize. The sedimentation profile of Rep68 typically shows two populations with sedimentation coefficients of ∼3S and ∼13S (Figure 6A). We have determined that the 13S peak corresponds to a mixture of oligomeric rings (data not shown). Figure 6B shows that the replacement of the linker completely abolishes the oligomerization of the mutant protein Rep68octlink. We can detect formation of dimeric species only at the highest concentration tested and in the presence of ATP, (Figure 6C). These results show that replacement of the linker produces a Rep68 protein whose ability to oligomerize has been severely affected.
The above findings indicate that the linker region plays a central role in the oligomerization of AAV Rep proteins. To confirm that the linker region has an intrinsic property to induce oligomerization, we generated a construct that spans the OBD domain and the linker region (OBD-linker residues 1–224) and measured its ability to oligomerize. We first analyzed the OBD domain (1–208) to determine any oligomerization up to concentrations of 1 mg/ml (43 µM). Our results show that while OBD is a monomer (Figure 7A), the OBD-linker protein construct displays formation of dimers at increasing protein concentrations (Figure 7B). These results support the hypothesis that the linker region has an intrinsic property to induce oligomerization
We generated a model of the Rep68ΔN214 construct using the X-ray structure of Rep40 (residues 225–490) and 9 residues of the linker (215–224) that were added as a helical extension to the N-terminus. The model of the α-helix was generated using Robetta [27]. Figure 8A and 8B shows the structural alignment of the OD domain of the Rep68ΔN214 model with the OD domains of PV-E1 and SV40-LTag. The alignment shows that residue Y224 superimposes with aromatic residues F313 and W270 located at the beginning of helix 1 in the OD domains of PV-E1 and SV40-LTag respectively. Analysis of the structures of both proteins reveals that these aromatic residues play a critical role in forming and stabilizing the oligomerization interface. They pack against both the N-terminal end of helix 4 of the same subunit and the C-terminus end of helix 4 of the neighboring subunit. In order to test the hypothesis that Y224 plays an equivalent role in AAV Rep proteins, we mutated it to alanine and tested its effect on the oligomerization of Rep68ΔN200. Mutation to the smaller residue alanine should have a direct effect in the oligomerization of this protein because of the significant reduction of surface exposed area. Figure 8C shows the sedimentation profile of this mutant protein showing that it completely abolishes the formation of dimers. To confirm that residue Y224 plays an important role in the oligomerization of AAV Rep proteins, we generated a Rep68Y224A mutant and compared its ability to form oligomers with respect to wild type Rep68. Analysis of the Rep68Y224A mutant reveals that at low concentration the protein is mostly found as a monomer with a sedimentation coefficient of ∼3S. At higher concentrations, we observed the appearance of multiple peaks that correspond to dimers, trimers and larger oligomers; nevertheless, the majority of the protein is present as a monomer. The presence of ATP induces a small degree of stability to the dimeric species at 5 µM and both the 5S and 11S species at 10 µM. However, the 13S complex observed with the wild type Rep68 is not formed and most of the protein is still found as a monomer (Figure 8E). These results indicate that residue Y224 is critical for the oligomerization of AAV Rep proteins.
To assess if the disruption of oligomerization observed with the Rep68Y224A mutant has any consequences on the AAV viral life cycle, we produced recombinant AAV2 particles expressing the GFP gene in presence of a helper virus containing the Y224A mutation in the Rep ORF. The cells were harvested and lysed, and the crude lysate (treated with an endonuclease) was used to infect Hela cells. Strikingly, the crude lysate from cells transfected with the mutant helper plasmid didn't contain any infectious rAAV2-GFP particles, as determined by FACS analysis of GFP positive cells (Figure 9). These results show that the residue Y224 of AAV Rep proteins, and the oligomeric properties it confers to these proteins, have a crucial role during the AAV life cycle.
In this study we report that the interdomain linker present in the larger AAV Rep68/78 proteins is an integral part of their oligomerization interface. We showed that the linker region is in fact an extension of the OD domain of AAV Rep proteins. Our results have shown that Rep40 constructs containing either a complete or half linker have the ability to oligomerize. This effect is enhanced in presence of ATP or ADP. We hypothesized that the linker region from residues 215 to 224 forms a α-helix that is connected to the first α-helix of the SF3 helicase domain. Secondary structure prediction and modeling of the linker region supports this argument (Figure 3A and 8B). Furthermore, we have identified a critical aromatic residue (Y224) located at the end of the linker region that is conserved in Rep proteins from all AAV serotypes. The bulky nature of this aromatic residue appears to be a conserved feature in SF3 helicases (Figure 8A). Structural alignment of the OD domain of a Rep40 model with an extended helical linker and those of SV40-LTag and PV-E1 shows that residue Y224 aligns with equivalent aromatic residues Trp270 and Phe313 respectively (Figure 8A, 8B). A detailed analysis of the oligomeric interface of these proteins shows that these aromatic residues have a dual role: they stabilize the hydrophobic core of the OD domain helical bundle, and are part of the oligomerization interface between neighboring subunits. Our results reveal the critical role of the OD domain in the formation of stable oligomers in SF3 helicases. The larger OD domains of SV40-Tag and PV-E1 proteins in cooperation with the AAA+ motor domain generate a helicase domain that forms stable hexamers. Constructs of SV40-LTag and PV-E1 without the OD domain fail to oligomerize [14], [19]. Another example that shows the fundamental role of the OD domain in oligomerization comes from the study of the evolutionary related proteins involved in rolling circle replication (RCR) of plasmids. The protein RepB from streptococcal RCR plasmid pMV158 is a hexameric protein that initiates replication of plasmid DNA and has a domain structure that resembles SF3 helicases but lacks the AAA+ subdomain [28]. Its N-terminal OBD domain is structurally and functionally related to the OBD from AAV Rep proteins due to the presence of the HUH motif critical for DNA nicking. Its C-terminal domain only consists of a 4 helical bundle that is similar to the OD domains of SF3 helicases and is responsible for hexamerization. Structural alignment shows that RepB has an aromatic residue (Phe143) equivalent to residue Y224 in AAV Rep68/78. We hypothesize that the role of this residue has been conserved throughout evolution to serve as a modulator of oligomerization in SF3 helicases and related RCR proteins. The smaller AAV Rep proteins Rep40/52 with truncated OD domains are missing the Y224 residue and thus are not able to sustain a stable oligomerization interface and are mostly monomeric. Consequently, the stable oligomerization of AAV Rep proteins requires the cooperative interaction of the OBD domain, the linker and the helicase domain. In this context, the OD sub-domain, and in particular the aromatic residue at the C-terminus of linker, appear to be the triggering element required for the oligomerization of AAV Rep proteins.
The critical role of residue Y224 in the overall AAV-2 viral life cycle is illustrated by the complete abolishment of production of infectious particles from AAV-2 vector constructs produced in the context of Rep carrying the Y224A mutation (Figure 9). This result prompts the question of which specific functions are affected by this mutation. We think that most of the biochemical activities of Rep68/78 will be affected due to the impairment in oligomerization. Remarkably, an earlier report by Walker et al. on the identification of residues necessary for site-specific endonuclease activity showed that a Y224 mutant was defective in AAV hairpin/DNA binding, trs endonuclease, DNA helicase and ATPase activity [29], suggesting that correct oligomerization of Rep proteins may be important in all of these functions.
In agreement with our results, a recent report has shown that the presence of the linker in an AAV5 Rep40 construct induces oligomerization in presence of DNA. However, the authors concluded that the linker effect is primarily due to its interaction with DNA [30]. As we demonstrated in this report, the oligomerization effect is an intrinsic property of the linker due to its critical role in the formation of an oligomerization interface as part of the OD domain. The presence of DNA induces further oligomerization as seen with all helicases [13]. However, it appears that the linker also plays an additional role in protein-DNA interaction that may be important during the assembly of Rep68/78 on DNA substrates such as the AAV origin of replication and AAVS1 integration site.
The use of alternative gene promoters is a common mechanism to generate protein diversity and flexibility in gene expression. At the same time it allows to obtain multiple functions from a limited number of genes, thus optimizing the size of the genome. It is clear that in the case of the Rep proteins from the AAV virus, nature has generated two sets of proteins that differ primarily in their ability to oligomerize. Rep proteins obtained from the AAV P19 promoter generate Rep40 and Rep52 with truncated OD domains and are thus unable to oligomerize. Both proteins play a critical role during DNA packaging into capsids; however, the mechanism of action of monomeric Rep40/52 during packaging remains elusive. Rep proteins generated from the P5 promoter, on the other hand, require the cooperative interaction of three different oligomeric interfaces produced by the OBD domain, the linker and the helicase domain. This feature potentially provides an additional dimension for the regulation of the diverse Rep activities when compared to the related proteins from SV40 and PV. We suggest that the cooperative interactions and the modulation of these interfaces – in particular in the presence of various specific DNA substrates – orchestrate the variety of functions performed by Rep68/Rep78 proteins and may thus represent a key to our understanding of the underlying mechanisms.
Finally, our report introduces the possibility of two distinct helicase modes for the biological functions supported by AAV Rep proteins. In the context of the large Rep proteins, a complete OD domain directs the formation of stable oligomers with a DNA unwinding mode likely to resemble that of the related viral proteins SV40-Tag and E1. The small Rep proteins, however, appear to utilize an incomplete OD domain that retains Rep40/52 in a monomeric state with formation of transitional dimeric complexes required for ATP hydrolysis. It is intriguing to speculate that this unique arrangement allows AAV to utilize two distinct motor activities with a single AAA+ domain. As Rep40/52 have been demonstrated to be required for genome packaging it is feasible to address the question whether this process requires a Rep40/52–mediated dimeric DNA helicase activity by a mechanism that is as yet undiscovered or whether further oligomerization is induced by interaction with capsid proteins.
All mutant proteins were generated using the pHisRep68/15b plasmid, which contains the AAV2 Rep68 ORF subcloned in vector PET-15b (Novagen). Site-directed mutagenesis for mutants Y224A was generated using the QuickChange mutagenesis kit (Stratagene). Rep constructs with different linker extensions were generated by PCR with primers designed to encompass the particular protein region. Primers included restriction enzyme sites NdeI and XhoI, and the sequence of the TEV protease site. The Rep68 protein used in these studies contained a Cys to Ser mutation that prevented aggregation but was functionally identical to the wild type protein (data not shown). The Rep68octlink construct was generated by substitution of residues 206 to 224 of AAV2 Rep68 with the mouse Oct-1 linker residues 328–346 (GeneBank CAA49791) using the gene synthesis services from GeneScript. The sequences of all constructs were confirmed by DNA sequencing (GeneWiz).
All proteins were expressed using the pET-15b vector, expressed in E. coli BL21(DE3) cells (Novagen), and purified as described before [18]. The final buffer contains (25 mM Tris-HCl [pH 8.0], 200 mM NaCl, and 2 mM TCEP). His6-PreScission Protease (PP) was expressed in BL21(DE3)-pLysS at 37°C for 3 h, in LB medium containing 1 mM IPTG. Cell pellets were lysed in Ni-Buffer A (20 mM Tris-HCl [pH 7.9 at 4°C], 500 mM NaCl, 5 mM Imidazole, 10% glycerol, 0.2% CHAPS, and 1 mM TCEP). After five 10-s cycles of sonication, the fusion protein was purified using a Ni-column – equilibrated in Ni-buffer A. Protein eluted was desalted using buffer A and a HiPrep™ 26/10 desalting column (GE Healthcare). His-PP tag was removed by PreScission protease treatment using 150 µg PP/mg His-PP-Rep68. After overnight incubation at 4°C, buffer was exchanged using the same desalting column and Ni-Buffer A. Subsequent Ni-column chromatography using the buffer B (same as buffer A but with 1 M imidazole), was performed to remove the uncleaved fusion protein, and untagged Rep68 was eluted with 30 mM imidazole. Rep68 was finally purified by gel filtration chromatography using a HiLoad Superdex 200 16/60 column (GE Healthcare) and Size Exclusion buffer. N-terminus His6-tagged WT and mutant Rep68 proteins were concentrated to 10 mg/ml, flash-frozen in liquid N2, and kept at −80°C until use.
The cross-linking reactions for Rep40 and Rep68ΔN200 were made according to an adapted protocol from Packman and Perham [31]. The reaction mixture was in cross-linking buffer (25 mM HEPES, 200 mM of NaCl, pH 8.0) and protein concentration was 2 mg/ml. A 30 fold molar excess of 100 mM DMP (dimethyl pimelimidate dihydrochloride, MP Biomedicals, LLC) was added to the reaction and incubated 60 min at room temperature. The reaction was quenched by addition of 1 M Tris, pH 7.5 to a final concentration of 50 mM. The samples were analyzed in an 8% SDS-PAGE.
Hek 293T cells were triple transfected using polyethylenimine (PEI) with an AAV2 ITR-containing plasmid including the GFP gene, a helper plasmid expressing AAV2 Rep (wt or Y224A cloned from the pHisRep68Y224A/15b) and Cap, and a third construct containing the adenovirus helper functions (pXX6, University of North Carolina Vector Core Facility). The presence of the Y224A mutation was confirmed by sequencing (Eurofins). After 72 h, the cells were harvested and lysed in 150 mM NaCl, 50 mM Tris at pH 8.5, followed by three freeze - thaw cycles. The lysate was treated for 30 minutes at 37°C with 150 units/ml of benzonase endonuclease (Sigma). HeLa cells were infected with increasing amounts of crude lysate, and the percentage of GFP-positive cells was determined three days post-infection.
Sedimentation velocity experiments were carried out using a Beckman Optima XL-I analytical ultracentrifuge (Beckman Coulter Inc.) equipped with a four and eight-position AN-60Ti rotor. Rep protein samples were loaded in the cells, using in all cases buffer used in the final purification step. Samples in double sector cells were centrifuged at 25,000 rpm for Rep68 proteins (Rep68 and Rep68Y224A). For Rep40 and linker constructs sedimentation was performed at 40,000 rpm. In all experiments, temperature was kept at 20°C. Sedimentation profiles were recorded using UV absorption (280 nm) and interference scanning optics. For the analysis of the results the program Sedfit was used to calculate sedimentation coefficient distribution profiles using the Lamm [21].
Structures of AAV-2 Rep40 (1S9H), Bovine papillomavirus E1 protein (2GXA), Simian virus 40 T large antigen (1SVO) and plasmid pMV158 RepB (3DKY) were analyzed using the programs COOT [32], PYMOL [33] and CHIMERA [34]. Structural alignment was done using the DALI server [35]. Secondary structure prediction was performed using PredictProtein [36]. Modeling of the linker region was done using ROBETTA [27].
|
10.1371/journal.pntd.0007154 | Etiology and severity of diarrheal diseases in infants at the semiarid region of Brazil: A case-control study | Diarrheal diseases are an important cause of morbidity and mortality among children in developing countries. We aimed to study the etiology and severity of diarrhea in children living in the low-income semiarid region of Brazil.
This is a cross-sectional, age-matched case-control study of diarrhea in children aged 2–36 months from six cities in Brazil’s semiarid region. Clinical, epidemiological, and anthropometric data were matched with fecal samples collected for the identification of enteropathogens.
We enrolled 1,200 children, 596 cases and 604 controls. By univariate analysis, eight enteropathogens were associated with diarrhea: Norovirus GII (OR 5.08, 95% CI 2.10, 12.30), Adenovirus (OR 3.79, 95% CI 1.41, 10.23), typical enteropathogenic Escherichia coli (tEPEC), (OR 3.28, 95% CI 1.39, 7.73), enterotoxigenic E. coli (ETEC LT and ST producing toxins), (OR 2.58, 95% CI 0.99, 6.69), rotavirus (OR 1.91, 95% CI 1.20, 3.02), shiga toxin-producing E. coli (STEC; OR 1.77, 95% CI 1.16, 2.69), enteroaggregative E. coli (EAEC), (OR 1.45, 95% CI 1.16, 1.83) and Giardia spp. (OR 1.39, 95% CI 1.05, 1.84). By logistic regression of all enteropathogens, the best predictors of diarrhea were norovirus, adenovirus, rotavirus, STEC, Giardia spp. and EAEC. A high diarrhea severity score was associated with EAEC.
Six enteropathogens: Norovirus, Adenovirus, Rotavirus, STEC, Giardia spp., and EAEC were associated with diarrhea in children from Brazil’s semiarid region. EAEC was associated with increased diarrhea severity.
| Most childhood diarrheal diseases studies focus on children from health centers or emergency hospitals that contribute to a lack of understanding of the etiology of community diarrhea. We aimed to investigate the etiology and severity of diarrheal diseases in children living in Brazil’s low-income, semiarid region communities. We used a case-control study of diarrhea (1,200 children, 596 cases and 604 controls) and showed that six enteropathogens (norovirus, adenovirus, rotavirus, STEC, Giardia and EAEC) are associated with diarrhea in these communities. Furthermore, enteroaggregative E. coli (EAEC) was associated with a high diarrhea clinical severity score. These findings help further the understanding of diarrheal disease etiology of previously unevaluated children from Brazil’s low-income semiarid region. This study contributes new information that expands our knowledge of diarrheal etiology in non-hospitalized children from developing countries. Moreover, we believe this work will support further research on specific enteropathogens, with special attention to the role of EAEC in severe cases of diarrhea and guide public health policies and physicians in the management of diarrheal diseases in Brazil’s low-income semiarid region.
| Diarrheal diseases remain a prominent cause of morbidity and mortality in developing countries as the second most common cause of death in children under five years old [1,2]. Studies of diarrheal illness among children in Brazil and other developing countries has focused on health centers or emergency hospitals that primarily treat patients with moderate-to-severe diarrhea and have the ability to identify enteropathogens [3–5]. This approach captures only a small subset of Brazil’s diarrheal diseases burden and limits accurate understanding about pathogen prevalence in the poorest semiarid region of Brazil.
The high prevalence of diarrheal disease in developing areas is significant because it inhibits normal growth, impairs cognitive function, and disrupts physical and educational development in children [6–14]. Therefore, enteropathogen specific studies of diarrheal diseases etiologies and clinical severity in Brazil’s semiarid region will contribute to public health preventions and interventions against diarrheal diseases.
This study is a cross-sectional, multisite work that evaluated childhood diarrheal etiology among six cities in Brazil’s semiarid region. We aimed to investigate etiology, severity of diarrheal episodes, and environmental factors associated with diarrheal diseases in children 2–36 months old.
A case-control study was conducted in six cities with a population greater than fifty-thousand that were randomly selected from the five states of Brazil’s semiarid region: Cajazeiras (Paraíba), Crato (Ceará), Ouricuri (Pernambuco), Patos (Paraíba), Picos (Piauí) and Sousa (Paraíba). During the active surveillance period from November 2009 to February 2012, fecal samples were collected from children aged 2–36 months who reside in urban communities near primary health care units. Cases were defined as children with diarrhea (three or more liquid stools in the last 24 hours). A standardized questionnaire was completed during the enrollment interview to collect the following detailed health information: demographic, environmental, socio-economic status, breastfeeding practices, other clinical conditions, vaccination history, frequency of diarrhea episodes and anthropometric measurements.
Diarrhea cases were finding via active surveillance by field workers walk door-door in the vicinity of the primary health care units and investigate the households until the sample size was reached. Diarrhea cases were defined as a child aged 2–36 months with a history of three or more liquid stools in the last 24 hours prior to the arrival of fieldworkers who were responsible to collect the stool samples. Inclusion criteria (diarrhea cases) were: 1) had three or more liquid stools in the last 24-hours; 2) had no chronic illness or hospitalization within 12-hours of study enrollment; and 3) written consent provided by parents or legal guardians. Inclusion criteria for no-diarrhea, controls, were: 1) did not present with diarrhea in the past two weeks; and 2) written consent provided by parents or legal guardians
The semiarid region includes the states of Ceará, Piauí, Rio Grande do Norte, Paraíba, Pernambuco, Alagoas, Sergipe, and Bahia of the Northeastern macro-region, but does not include Maranhão and the area north of Minas Gerais (Fig 1). The semiarid region covers 969,589.4 km2 and has a population of 23.5 million. The estimated population of children under five years old is 2.3 million. The average annual rainfall is less than 800 mm and the aridity index can reach 0.5, which represents the water balance between precipitation and potential evapotranspiration. Drought risk in the semiarid region is greater than 60% [15].
Case report forms were designed to collect information during child enrollment and capture demographic, birthdate, sex, anthropometric measurements such as current weight, length and head circumference, child care practices such as breastfeeding, and the characteristics of the mother/caregiver. Environmental and socio-economic status data capture included: household exterior material, number of rooms, number of people sleeping in the household, number of children less than five years old, source of drinking water, toilet facilities, number/type of animals living in their household, and the average monthly income for the entire household.
Clinical data were collected at the time of enrollment and diarrhea episodes were identified by the fieldworkers. They were defined as a maternal report of three or more liquid stools in a 24-hour period. Discrete episodes had at least two intervening days without diarrhea. A diarrhea severity score was adapted for every episode using elements derived from the Malnutrition-Enteric Disease MAL-ED scores [16]. Dehydration was defined as moderate or severe based on the World Health Organization manual of the treatment of diarrhea [17]. Dysentery was defined as the presence of visible blood in the stool as reported by the child´s mother/caregiver. Diarrhea associated with fever was defined as diarrhea and the mother recording a temperature greater than 37.5°C. Vomiting associated diarrhea was defined as vomiting at any point during an episode of diarrhea. Vaccine administration data was captured for the following: rotavirus (Rotarix G1P[8] GlaxoSmithKline, Wavre, Belgium); BCG: Bacillus Calmette-Guérin; MMR: Measles, Mumps and Rubeola; Hepatitis B; Hib: Haemophilus influenzae type b; DPT: Diphtheria, Pertussis, Tetanus; and OPV: Oral Polio Vaccine. In addition, antibiotic and other medications given to the child during diarrhea episodes were recorded.
The study protocol used a standard recumbent length measuring board (Schorr Productions, Olney, MD) to measure the length of children to the nearest 0.1 cm. Digital scales were also used to measure weight to the nearest 100 g. The weight-for-age (WAZ), length-for-age (LAZ), and weight-for-length (WLZ) z-scores were calculated using the World Health Organization Multi-Country Growth Reference Study [18]. This study used a Seca 212 infant head circumference tape (Seca Deutschland, Hamburg, Germany) made of non-stretch Teflon synthetic material and a range of 5–59 cm marked in 0.1 cm increments
Stool specimens were collected and stored at -80°C until used. DNA and RNA extraction were performed using the QIAamp DNA Stool Mini Kit (Qiagen, USA) and QIAamp Viral RNA extraction kit (QIAGEN, USA), respectively. Nucleic acid was amplified with sequence-specific primer-probe sets (S1 Table). Either forward or reverse primers were biotinylated on the 5’-end and probes were amine-modified at the 5′-end comprised of 12-carbon spacers to enable coupling to the carboxylated fluorescent microspheres. Following Multiplex polymerase chain reaction (PCR) reactions and membrane hybridization procedures, samples were analyzed by the Bio-Plex 200 System (Bio-Rad, CA, USA) [19,20]. The results were reported as microsphere specific media fluorescent intensity (MFI) and corrected for background bead fluorescence. Corrected MFI were calculated as follows: cMFI = (MFIanalyte−MFInegative control) / MFInegative control. Positive samples had cMFI values greater than three. Positive (DNA template from reference organisms) and negative controls (nuclease-free water) were included in every run. We used four distinct Multiplex PCR panels (bacteria 1 and 2, protozoa and virus) to identify 17-different enteropathogens. All PCR conditions and references are described in S1 Table. Bead coupling and hybridization procedures followed descriptions published elsewhere [19,20]. Myeloperoxidase (MPO) biomarkers were measured in stool samples to access gut inflammation using a kit from Immunodiagnostic (Bensheim, Germany).
The study protocol and consent form were approved by the local institutional review board (IRB) at all cities sites and by collaborating IRBs, and approved by Brazil’s National Commission on Ethics in Research and the Research Ethics Committee of the Federal University of Ceará (Craft No. 5502006, Protocol No. 23805). Written informed consent was obtained from the parent or guardian of every child.
The estimated sample size of infantile diarrhea etiology in the Brazilian semiarid region was 980 to 1,400 children. The sample size of 278 cases and 278 controls provided a statistical power of 80% and a statistical significance of P <0.05 for pathogen isolation in at least 6% of cases and 1.5% of controls. We estimated a 10% loss of subjects from the study thereby requiring we have 306 cases and 306 controls for a total of 612 subjects for the study.
The collected data were entered into Excel spreadsheets v.4.0 (Microsoft Corp., Seattle, WA) by two independent data entry persons and then compared to ensure accuracy. Statistical analysis was performed using SPSS (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp.) and used for all analyses. All study subject samples and data were analyzed anonymously. The Shapiro-Wilk test was used to evaluate the normality of the quantitative variable data, and the Levene test was used to evaluate the equality of the variances. The Student's t test was used for normally distributed variables; the Mann-Whitney test (two groups) and Kruskal Wallis test (three or more groups) were used for variables whose distribution was not normal. Qualitative variables were analyzed using the chi-square test or Fisher's test. GraphPad Prism software, version 3.0 for Windows (San Diego, CA, USA), was used for complementary statistical analysis, table formatting and figures. Multivariate logistic regression models were used to access risk factors and identify the enteropathogens most associated with diarrhea episodes. Factors of risk or protection included child anthropometrics, child care, mother/caregiver characteristics, environmental and socio-economic parameters. We also used multivariate logistic regression models to access etiologic association with the following outcomes: diarrhea severity, signs and symptoms, and episode duration. In these models, values for β coefficients (SE) were presented to show the positive or negative relationship between the variable and the outcome. Odds ratios (OR) with 95% confidence intervals (95% CI) were utilized to assess the risk between a variable and its outcome. A significance level of <0.05 was used for all statistical analyses.
A total of 1,600 children were screened, 400 were ineligible and 1,200 children were enrolled (596 cases and 604 controls). All 1,200 children provided stool samples and their data are detailed in Table 1.
The overall prevalence of enteropathogens in diarrhea cases and controls are summarized in Table 2. Pathogen prevalence by first, second and third year of life are shown in Fig 2. Norovirus GII (OR = 8.514; 95% CI: 1.088–66.632) and typical enteropathogenic Escherichia coli (tEPEC), (OR = 4.686; 95% CI: 1.034–21.228) were significantly associated with the diarrhea case group compared to control children in the first year of life (Fig 2A). In the second year of life norovirus GII (OR = 12.758; 95% CI: 1.643–99.061), shiga toxin-producing Escherichia coli (STEC), (OR = 2.497; 95% CI: 1.255–4.967) and Giardia spp. (OR = 1.807; 95% CI: 1.133–2.880) were significantly associated with diarrheal episodes (Fig 2B). In the third year we only identified sapovirus (OR = 5.215; 95% CI: 1.092–24.900), which had a significant association with the diarrhea case group when compared to the controls (Fig 2C).
Since a higher proportion of these stool samples presented with two or more pathogens (55%; 661/1200), we adjusted the analysis using the multivariate logistic regression model to include all enteropathogens (Fig 3 and Table 3). The results show the most likely enteropathogens associated with diarrhea by decreasing odds ratio: norovirus GII (OR = 5.385; 95% CI: 2.196–13.203); adenovirus (OR = 3.476; 95% CI: 1.264–9.558); rotavirus (OR = 1.929; 95% CI: 1.192–3.121); shiga toxin-producing E. coli (STEC; OR = 1.623; 95% CI: 1.033–2.548); Giardia spp. (OR = 1.537; 95% CI: 1.140–2.074); and enteroaggregative E. coli (EAEC; OR = 1.403; 95% CI: 1.092–1.804). Fig 2D–2F) shows the pathogens prevalence by grouping children with one, two and three or more pathogens. In the group with one pathogen isolated, EAEC (50.7%; 180/355) had the highest prevalence (Fig 2D). EAEC (66.8%; 227/340), EPEC (38.2%; 130/340), and most atypical EPEC (35.6%; 121/340) had the highest prevalence when children presented with two pathogens, followed by Giardia spp. (25.1%; 85/338) (Fig 2E). Children with three or more pathogens had EAEC (75.7%; 243/321), EPEC (38.2; 13/340) and Salmonella spp. (36.6%; 117/420) as the most prevalent pathogens (Fig 2F).
To evaluate which of the enteropathogens were associated with higher severity, the type of episode and signs and symptoms of diarrheal episodes were analyzed using the multivariate logistic regression analysis model. Among all enteropathogens, enteroaggregative E. coli (EAEC) was the only one that maintained a significant association with severity of the diarrheal episodes (OR = 2.070; 95% CI 1.391–3.079). EAEC was also associated with moderate-to-severe dehydration (OR = 1.572; 95% CI 1.099–2.249). Norovirus GII was associated with fever (OR = 2.332; 95%CI 1.083–5.024). When we considered acute episodes (code = 0) and prolonged episodes (code = 1) in multiple regression analysis, norovirus GII was associated with prolonged diarrhea episodes (OR = 3.941; 95% CI 1.208–12.851) and Salmonella spp. was associated with acute episodes (OR = 0.044; 95% CI 0.012–0943).
This is the first large study on the broad etiology of diarrheal diseases using a highly sensitive and specific molecular diagnostic to identify causality and access both symptomatic and asymptomatic enteric infections in young children across six cities in the low-income semiarid region in Brazil. We were able to identify enteropathogens in 84.7% of the stool samples regardless of whether they were asymptomatic controls or symptomatic diarrhea cases. The study also showed that bacterial enteric infections were the most prevalent cause of diarrheal diseases, followed by protozoa and viruses. Overall, this report identified eight enteropathogens specifically associated with significant enteric infections in young children from this population: enteroaggregative E. coli, Giardia spp., shiga toxin-producing E. coli, rotavirus, norovirus GII, typical enteropathogenic E. coli, adenovirus, heat-stable and heat-labile producing enterotoxigenic E. coli.
Children in the first year of life had 56.4% of their diarrheal stool samples associated with two or more pathogens. We showed that symptomatic diarrheal cases had a significantly higher burden of two or more enteropathogens compared to asymptomatic controls. Kotloff et al., in a multisite matched case-control study (GEMS) in sub-Saharan Africa and south Asia, also found enteropathogen detection to be more common in diarrheal stools than non-diarrheal stools [10]. Similar results were found in the MAL-ED in South America, Africa, and Asia, where the number of enteropathogens detected was higher in diarrhea stools than non-diarrheal stools [21]. Multivariate logistic regression analysis of six enteropathogens: norovirus, adenovirus, rotavirus, STEC, Giardia and EAEC showed significant odds of being associated with a risk for diarrheal diseases. In the GEMS study they found rotavirus, Cryptosporidium, ETEC, tEPEC and Shigella as the major enteropathogens associated with moderate-to-severe diarrhea. Although we found similar enteropathogen association with diarrheal stools, such as rotavirus, ETEC and tEPEC, there are some differences due to different geographical areas, type of study design and selected child population.
The MAL-ED study identified norovirus GII, rotavirus, Campylobacter spp., astrovirus and Cryptosporidium spp. in the first year and Campylobacter spp., norovirus GII, rotavirus, astrovirus and Shigella spp. in the second year of life in these children. This report consistently showed a norovirus association with diarrheal stools in the first and second year of life in these children. The differences among other enteropathogens could be explained by different study designs and geographical areas.
EAEC was the most prevalent enteropathogen overall and by age at 2–11 months, 12–23 and 24–36 months, either alone or combined with other enteropathogens. In the MAL-ED study, EAEC also had a high prevalence reaching the second and third most prevalent enteropathogen in the first and second year of life, respectively (21). Recently, unpublished data, using a quantitative Real Time PCR approach in the surveillance of stools among 1,469 children from the MAL-ED cohort study identified EAEC as the most prevalent enteropathogen that also had an association with decrement in length at 2 years. In the same cohort study, Lima et al. also showed that EAEC subclinical infection and coinfection impaired child growth identified at the patient’s 6-month follow-up [22]. The children in the diarrhea case group of this report had a significantly lower length and head circumference compared to the control group children. Additional reports and recent studies showed that EAEC infections were associated with significant impact on child nutrition even in asymptomatic children, which is likely caused by gut inflammation and malnutrition [6,23,24]. This report also showed a consistently significant elevation of MPO, a marker of gut inflammation, in diarrhea stool samples compared to control group children. In the adjusted multivariate logistic analysis, EAEC also showed significant association with clinical severity of diarrhea cases, plus a specific association with moderate-to-severe dehydration. Lima et al. showed a combination of virulence genes, aaiC (aggR-activated island), and agg3/4C (usher, AAF/III-IV assembly unit), but lacking agg4A (AAF/IV fimbrial subunit), and orf61 (cryptic protein) with diarrhea stools compared to control samples, which could contribute to understanding the pathobiology of EAEC enteric infection [25].
This study also showed that norovirus GII was associated with fever and prolonged episodes of diarrhea, while Salmonella spp. was associated with acute episodes. Rotavirus was also associated with diarrhea cases compared to controls, but this was in part due to lower vaccine coverage seen in this group even though the overall vaccine coverage rate was proportionally high. This brings attention to the value of increasing rotavirus vaccine coverage to prevent rotavirus enteric infection [26]. Norovirus GII and astrovirus enteric infections had a lower proportion of diarrhea cases, but they were associated with diarrheal stools in the univariate analysis. However, norovirus GII and adenovirus had the highest attributable fractions followed by rotavirus, STEC, Giardia spp. and EAEC. There are no vaccines for these enteric infections, except rotavirus, and the development of additional vaccines to prevent infection by norovirus GII, astrovirus and adenovirus would be an important contribution to global health [27,28]. Giardia spp. was more frequently associated with diarrhea cases when compared to controls in the adjusted analysis. Giardia spp. was the third most prevalent enteric infection after EPEC and EAEC. So, this is an important cause to be considered for enteric infections of both asymptomatic and symptomatic infections because of its negative impact on childhood malnutrition [29]. Consistent with the MAL-ED cohort study this report also showed ETEC (ST_LT producing) and STEC (shiga producing toxin E. coli) as important enteric infections significantly associated with diarrheal stools compared to control stools [21].
In the adjusted multivariate logistic analysis, we showed that only increased length of the child and number of rooms in the household were protective factors for diarrheal diseases. Lima et al. also reported an increased risk for EAEC infection and coinfections in non-diarrheal stools associated with lower socio-economic and sanitation facilities in the households of children (18). Baker et al. found sanitation and hygiene-specific risk factors for moderate-to-severe diarrhea in young children in the Global Enteric Multicenter Study [30].
The main limitation of this study was the case-control design, which by itself, limited the evaluation of diarrheal disease etiologies and asymptomatic carrier impact on nutrition and neurocognitive development among children as previously demonstrated decades ago and recently reported in MAL-ED cohort studies [13,14]. There are several advantages of this study, such as the molecular diagnostic approach utilized to study the etiologies of enteric infections, the sample size calculated to provide statistical significance to potentially lower prevalence enteric infections, and finally the identification of risk factors and common etiologies associated with enteric infections to facilitate planning for prevention and intervention via public health programs in low-income semiarid regions in Brazil.
In summary, these results show that bacterial diarrheal etiologies remain the most prevalent pathogens isolated from stool samples, followed by protozoa and viruses in the setting of young children from low-income semiarid regions in Brazil. In the adjusted multivariate logistic regression analysis, we identified six enteropathogens, norovirus, adenovirus, rotavirus, STEC, Giardia spp. and EAEC and odds associated with diarrhea cases compared to control children. These results suggest the importance of these six enteropathogens as causes of acute diarrhea episodes and of EAEC enteric infections having association with a high clinical severity score plus moderate-to-severe dehydration. Therefore, identifying key preventive measures and interventions that reduce exposure to these enteropathogens, developing and providing education for mother/caregiver, ensuring adequate child nutrition, increasing vaccination coverage, and improving access to better environmental and socio-economic factors are key to reduce diarrheal diseases and the potential consequences on growth and neurocognitive development for these children.
|
10.1371/journal.pntd.0001564 | Enhanced Protective Efficacy of a Chimeric Form of the Schistosomiasis Vaccine Antigen Sm-TSP-2 | The large extracellular loop of the Schistosoma mansoni tetraspanin, Sm-TSP-2, when fused to a thioredoxin partner and formulated with Freund's adjuvants, has been shown to be an efficacious vaccine against murine schistosomiasis. Moreover, Sm-TSP-2 is uniquely recognised by IgG1 and IgG3 from putatively resistant individuals resident in S. mansoni endemic areas in Brazil. In the present study, we expressed Sm-TSP-2 at high yield and in soluble form in E. coli without the need for a solubility enhancing fusion partner. We also expressed in E. coli a chimera called Sm-TSP-2/5B, which consisted of Sm-TSP-2 fused to the immunogenic 5B region of the hookworm aspartic protease and vaccine antigen, Na-APR-1. Sm-TSP-2 formulated with alum/CpG showed significant reductions in adult worm and liver egg burdens in two separate murine schistosomiasis challenge studies. Sm-TSP-2/5B afforded significantly greater protection than Sm-TSP-2 alone when both antigens were formulated with alum/CpG. The enhanced protection obtained with the chimeric fusion protein was associated with increased production of anti-Sm-TSP-2 antibodies and IL-4, IL-10 and IFN-γ from spleen cells of vaccinated animals. Sera from 666 individuals from Brazil who were infected with S. mansoni were screened for potentially deleterious IgE responses to Sm-TSP-2. Anti-Sm-TSP-2 IgE to this protein was not detected (also shown previously for Na-APR-1), suggesting that the chimeric antigen Sm-TSP-2/5B could be used to safely and effectively vaccinate people in areas where schistosomes and hookworms are endemic.
| There are currently no vaccines available to combat helminth (worm) infections in humans. The most devastating of the diseases caused by human helminths are schistosomiasis (or bilharzia) and hookworm disease. By fusing one of the lead schistosomiasis vaccine antigens, Sm-TSP-2, with a protective fragment from one of the lead hookworm vaccine antigens, Na-APR-1, we have produced a chimeric vaccine, termed Sm-TSP-2/5B that might provide protection against two debilitating and co-endemic neglected tropical diseases. Sm-TSP-2/5B provided increased protection compared to Sm-TSP-2 alone when formulated with human approved adjuvants and tested in a mouse model of schistosomiasis. Moreover, IgE against Sm-TSP-2 or Na-APR-1 has not been detected in the blood of residents from an area in Brazil that is endemic for schistosomes and hookworms, indicating that vaccines based on these molecules would be unlikely to generate allergic reactions in recipients from developing countries.
| Schistosomiasis ranks among the most important infectious diseases in tropical regions, resulting in a loss of between 4.5 and 92 million Disability-Adjusted Life Years (DALYs) annually and almost 300,000 deaths in sub-Saharan Africa alone [1], [2], [3]. High rates of post-treatment reinfection [1], the inability of periodic chemotherapy to interrupt transmission [4], the exclusive reliance on praziquantel as the only chemotherapeutic option [5], [6] and the unsustainability of mass drug administration [7] has led to the development of new anti-schistosomiasis control measures, inlcuding vaccines, to complement existing initiatives [5], [8], [9].
Molecules lodged in the apical membrane of the schistosome tegument represent vulnerable targets for immunological attack by host antibodies due to their intimate association with the host immune system. One such family of molecules – predicted by proteomic analyses of the schistosome tegument to be accessible to host immunoglobulin [10] – is the tetraspanin integral membrane proteins. Tetraspanins contain four transmembrane domains and two extracellular loops that are predicted to interact with exogenous ligands [11], [12]. Indeed, the second extracellular loop of one of these schistosome tetraspanins, Sm-TSP-2, has proven to be an effective anti-schistosomiasis vaccine, eliciting 57–64% protection in mice vaccinated with the antigen followed by challenge with S. mansoni cercariae [12]. Other schistosome tetraspanins are protective in mouse models of schistosomiasis [10], including Sm23 [13], [14] and Sj-TSP-2, an S. japonicum orthologue of Sm-TSP-2 [15]. Moreover, Sm-TSP-2 was strongly recognised by IgG1 and IgG3 from putatively resistant but not from chronically infected individuals [12], further highlighting the promise of this antigen as a subunit vaccine against human schistosomiasis.
The tegument of adult and schistosomula of S. mansoni is thinner and distinctly more vacuolated compared to controls after in vitro treatment with Sm-tsp-2 double-stranded RNA (dsRNA) [16]. Moreover, injection of mice with schistosomula pre-treated with Sm-tsp-2 dsRNA resulted in the recovery of 83% fewer parasites from the mesenteries compared to controls [16], highlighting the importance of Sm-TSP-2 in proper tegument development and worm survival, and providing a potential mechanism by which the vaccine exerts its protective effect.
In an earlier study, we reported the production of a chimeric form of Sm-TSP-2, consisting of Sm-TSP-2 fused to the immunodominant and neutralizing 5B region of the hookworm aspartic protease Na-APR-1, termed Sm-TSP-2/5B [17]. Hookworm infection and schistosomiasis caused by S. mansoni are co-endemic in much of sub-Saharan Africa and Brazil, and there is potential interest in developing a vaccine that targets both of these high prevalence and high disease burden helminths [18]. Na-APR-1/5B is a 40 amino acid fragment of the protease that contains an immunodominant alpha helix, A291Y, which is the target epitope recognized by polyclonal and monoclonal antibodies that are capable of neutralizing the catalytic activity of Na-APR-1 [17]. Na-APR-1/5B could not be produced in soluble form, but when fused to Sm-TSP-2, it was produced in soluble form by E. coli and induced antibodies upon vaccination that neutralized the enzymatic activity of Na-APR-1; the chimera is currently under investigation as a hookworm vaccine. Using a mouse model of schistosomiasis, we explored the efficacy of the Sm-TSP-2/5B chimera in comparison to Sm-TSP-2 alone when both antigens are formulated with alum/CpG. Given the recent safety concerns of helminth vaccines that elicit an IgE response in individuals residing in an endemic area [19], we also assessed the recognition of Sm-TSP-2/5B by IgE from individuals chronically infected with S. mansoni, a crucial step in determining whether or not this antigen could be used to safely and effectively vaccinate people in areas endemic for both hookworms and schistosomes.
All work involving experimental procedures with laboratory animals was approved by the animal ethics committee of James Cook University according to the regulations of the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, 7th edition (reference EA16). All work involving human subjects research was approved by the Human Research Ethics Committees or Institute Review Boards of Instituto René Rachou-FIOCRUZ, the Brazilian National Committee for Ethics in Research (CONEP), George Washington University Medical Center, and the London School of Hygiene and Tropical Medicine. Informed written consent was obtained from all adults or the parents and guardians of all children involved in the study.
Oligonucleotide primers incorporating NdeI and XhoI restriction sites (forward primer: GCGCATATGGAAAAGCCCAAGGTCAAAAAACAC; reverse primer GCGCTCGAGGTGCGCTTTGCTTAGATCGCTGAC) and pfu turbo DNA polymerase (Stratagene) were used to amplify the extracellular loop 2 region (Glu-107 – His-184) of the S. mansoni tetraspanin Sm-TSP-2 from the pBAD/TOPO/Sm-TSP-2 plasmid [12] in our laboratory. The amplicon was then cloned into the NdeI and XhoI sites of the pET41a expression vector (Novagen), removing the GST fusion tag to allow for native N-terminal expression of the protein, but retaining the vector's C-terminal 6×his tag to facilitate purification by Immobilised Metal Affinity Chromatography (IMAC). The ensuing plasmid was then transformed into chemically competent E. coli BL21-AI cells (Invitrogen). Sm-TSP-2 was expressed using the auto-induction method and media formulations established by Studier [20]. Briefly, 10 ml of minimal media supplemented with 50 µg/ml kanamycin (MDGkan) was inoculated with a single, recombinant BL21-AI colony and grown overnight at 37°C with shaking (225 rpm). The entire overnight culture was then used to seed 1.0 L of defined media supplemented with 50 µg/ml kanamycin (ZYM-5052-Akan), which was incubated for 24 hours at 37°C with shaking (225 rpm). Bacteria were pelleted, lysed and the resultant homogenate purified by (IMAC) as described previously [17]. Purified Sm-TSP-2 was buffer-exchanged in a dialysis bag (Pierce) with a cut-off size of 3 kDa against two changes of 50 mM sodium phosphate, pH 6.5, 10 mM NaCl (CEX buffer) (2.0 L each) at 4°C for at least 2 hours and then further purified by passing through a pre-packed 5.0 ml Hi-Trap SP-FF column (GE Healthcare) (equilibrated with 10 column volumes of CEX buffer) at a flow rate of 1.0 ml/min using an AKTA Prime UPC FPLC unit (GE Healthcare). Bound protein was purified by washing with resuspension buffers containing a rising concentration (10–500 mM) of NaCl and eluting in 5 column volumes of elution buffer (50 mM sodium phosphate, pH 6.5, 1.0 M NaCl). Sm-TSP-2 was desalted in a dialysis bag (Pierce) with a cut-off size of 3 kDa against two changes of PBS (2.0 L each) at 4°C for at least 2 hours and the final protein concentration was adjusted to 1.0 mg/ml using an Amicon Ultra-15 centrifugal concentration device (Millipore).
Sm-TSP-2/5B was produced in E. coli and purified as previously described [17]. The pMal-4E plasmid encoding Maltose Binding Protein (MBP) was kindly provided by Dr F. Cardoso and MBP was expressed in E. coli and purified on amylose resin according to the manufacturer's instructions (New England Biolabs).
An emulsion containing 100 µg of Sm-TSP-2 or Sm-TSP-2/5B (1.0 mg/ml) and an equal volume of Freund's complete adjuvant was subcutaneously injected into a single New Zealand White rabbit. The same amount of antigen emulsified in an equal volume of Freund's incomplete adjuvant was similarly administered 2 and 4 weeks later. The rabbit was bled 2 weeks later and the serum collected by centrifugation.
Freshly perfused adult S. mansoni were fixed in 100% methanol overnight at 4°C, embedded in Tissue-tek Optimal Cutting Temperature compound (ProSciTech) and cryostatically sectioned into 7.0 µm sections. Sections were rehydrated in PBS and blocked with PBS/0.05% Tween 20 (PBST)/1% Foetal Calf Serum (FCS) for 1 hour at RT. After washing twice (5 minutes each) with PBST, sections were incubated with either anti-Sm-TSP2, anti-Sm-TSP2/5B or naive rabbit sera (8.0 µl in 200 µl PBST/1% BSA) and 5.0 µl methanolic Alexa Fluor 488-Phalloidin (Invitrogen) for 1 hour at RT and then washed again (3×5 minutes each). The sections were then probed with goat anti-rabbit IgG-Cy3 (Jackson) (1∶500 in PBST/1% BSA) for 1 hour at RT. After a further 3 washes with PBST, slides were air dried briefly and mounted with cover slips using PBS/50% Vectorshield mounting medium with DAPI (Vector Industries) to stain nuclei. These were examined using a Leica IM1000 DMIRB inverted fluorescence microscope.
The inhibition of hemoglobin digestion by Na-APR-1 using anti-Sm-TSP2/5B IgG was performed as described previously [17]. An equal amount of anti-Sm-TSP-2 IgG was used as a negative control.
The study was conducted in Americaninhas, a rural community in northeast Minas Gerais state, Brazil and has been described in detail [12]. The study design was a total population survey, with all individuals in a 10 km2 area eligible for inclusion. All participants excluded from the study were offered a fecal exam and treated for all helminth infections, but were not considered part of the data set for analysis. Women who were evidently pregnant, or who tested positive on a urine pregnancy test received treatment for all helminth infections after the end of the pregnancy or the termination of breast-feeding.
The parasitological survey and blood draw were performed during April-July 2004, the results of which can be found in Table 1. Subjects were asked to provide two fecal samples on two separate days, which were examined qualitatively by formalin-ether sedimentation. Helminth-positive samples were then examined by Kato–Katz fecal thick smear to quantify the intensity of infection, as eggs per gram of feces (epg). Two slides were counted from each day's sample, i.e. 2–4 slides from each individual, as some individuals only provided one sample. Individuals who were egg-positive by sedimentation but negative by Kato-Katz were assigned a count of 3 epg, half the Kato-Katz detection limit. Hookworm was exclusively N. americanus. Adults or children positive for gastrointestinal nematodes were offered a single 400 mg dose of albendazole and individuals infected with S. mansoni were treated with praziquantel. Egg-negative individuals were not treated. Treated individuals were examined post-treatment to confirm treatment efficacy, and offered repeat treatment(s) until egg-negative.
Approximately 20 ml of blood was collected from 666 volunteers in siliconized tubes for separation of serum. In brief, the level of IgE against Sm-TSP-2 was measured by indirect ELISA using Polysorp 96-well microtiter ELISA plates (NUNC F96, Fisher Scientific) which were incubated overnight at 4°C with antigen (1 µg/ml in 0.15 M PBS, pH 7.2). After washing with PBST, the plates were blocked for 2 hours at RT with 250 µl of PBST/3% BSA. One hundred microliters of sera (1∶25 in PBST/3% BSA) were added to the wells and incubated overnight at 4°C, then the plates were washed with PBST and 100 µl of mouse biotin-conjugated monoclonal anti-human IgE FC (Human Reagent Laboratory, Baltimore, MD) (1∶200 in PBST/3% BSA) was added to the plates. Plates were incubated for 2 hours at RT and then washed with PBST. Plates were developed by adding o-Phenylenediamine dihydrochloride in 0.05 M phosphate-citrate buffer (pH 5.0) plus 30% hydrogen peroxide H2O2 for 30 minutes at RT in the dark. Fifty microliters of 2N H2SO4 was added to stop the colorimetric reaction, which was read at a wavelength of 490 nm on a SpectraMax 340 PC (Molecular Devices) microplate reader. SOFTmax Pro for Windows was used for the analysis and storage of data.
Approval for the work described in this study was obtained from the James Cook University Animal Ethics Committee. Groups of ten female C57BL/6 mice were immunised with Sm-TSP-2, Sm-TSP-2/5B, or the control protein MBP. Each antigen (25 µg per dose) was formulated with an equal volume (25 µl) of a 13 mg/ml colloidal suspension of aluminium hydroxide gel (alum) (Sigma) and 5 µg of CpG oligodinucleotide 1826 (CpG) (Invivogen) and injected intraperitoneally on days 0, 14 and 28. Mice were challenged on day 42 with 120 S. mansoni cercariae by abdominal penetration [21]. Trials were conducted twice on different dates and with different batches of cercariae. Serum samples were collected at day −2 (pre-immunisation), day 40 (pre-challenge) and day 91 (necropsy) to assess antibody responses.
Mouse necropsy and worm and egg burden assessments were performed as described previously [12]. Reductions in parasite loads were calculated as percentages of the parasite burden in the control group. Statistical significance was assigned a threshold of P = 0.05 and values were determined using the student's t test function in Graph Pad Prism.
Individual anti-Sm-TSP-2 titres (total IgG, IgG1 and IgG2a) were determined for all trial 1 animals just prior to cercarial challenge and at necropsy using standard ELISA techniques. Antigen was coated on microtiter plates at 1.0 µg/ml. Sera were serially diluted (1∶1,000 to 1∶16,384,000 for total IgG and IgG1 measurements and 1∶1,000 to 1∶256,000 for IgG2a assessment) and 100 µl was added to each well. After addition of the appropriate horseradish peroxidase-conjugated goat antibody (Jackson), peroxidase activity was detected with tetramethyl benzidine chromogenic substrate and measured at 655 nm.
Spleens were taken from all animals from trial 2, and single cell suspensions prepared by passing through a 70 µm filter (BD Biosciences). Red blood cell lysis buffer (Sigma) was used to remove red blood cells. Splenocyte preparations were counted, and cultured in duplicate at 1×106 cells/well in 96-well plates. Schistosome egg antigen (SEA) and soluble adult worm antigen preparation (SWAP) were prepared as described respectively [18], [19] and added to the cultures at 10 µg/ml and cultured at 37°C, 5% CO2 for 72 h. Levels of IL-4, IL-10, and IFN-γ in cell-free supernatants were assessed by ELISA (OptEIA, BD Biosciences).
The large extracellular loop of Sm-TSP-2 (Sm-TSP-2) (molecular weight including 6×His tag = 10 kDa) was expressed in E. coli using the auto-induction technique of Studier [20] instead of the more conventional method of IPTG induction normally used to drive protein expression in T7 promoter-based, inducible systems. In addition to producing an increased biomass despite using identical seeding conditions and culture volumes, Sm-TSP-2 was produced by auto-induction and purified by IMAC to a final concentration of 100 mg/L (Fig. 1A), more than twice the yield of Sm-TSP-2 obtained by IPTG-induction (data not shown). To obtain reasonable yields of soluble chimeric Sm-TSP-2/5B (molecular weight including 6×His tag = 16.1 kDa), the protein required expression in the less reductive cytoplasmic environment of the slow-growing Rosetta-Gami strain of E. coli, in addition to being cultured at a sub-optimal growth temperature of 23°C; as a result, auto-induction of Sm-TSP-2/5B was not a feasible production method. Nevertheless, when expressed using IPTG-induction and purified by IMAC, we obtained a yield of 20 mg/L of soluble Sm-TSP-2/5B (Fig. 1B).
The localization of Sm-TSP-2 to the outer tegument of S. mansoni has previously been documented using an antibody raised to the thioredoxin fusion protein [12]. The recognition of native Sm-TSP-2 by anti-Sm-TSP-2/5B antibodies (Fig. 2A) indicated that parasite-derived Sm-TSP-2 epitopes were faithfully reproduced in the recombinant protein and were not disrupted by the addition of the 5B region of Na-APR-1 to the C-terminus of Sm-TSP-2. No reaction was observed with naive rabbit serum (Fig. 2B). Similarly, the ability of anti-Sm-TSP-2/5B IgG to bind (and inhibit) Na-APR-1 hemoglobinase activity demonstrates the preservation of 5B epitopes within the chimeric protein. No hemoglobinase inhibition of the enzyme was observed when anti-Sm-TSP-2 IgG was used in the assay (Fig. 2C).
Sera from 666 individuals from Minas Gerais state, Brazil – an area of high S. mansoni transmission – were assessed for the presence of an IgE response against Sm-TSP-2. No detectable levels of anti-Sm-TSP-2 IgE antibodies were observed, despite the presence of a strong IgE response to SEA in some individuals (Fig. 3).
Mice vaccinated with alum/CpG adjuvanted Sm-TSP-2 and Sm-TSP-2/5B mounted strong Sm-TSP-2-specific IgG responses (Table 2). IgG1 responses dominated and IgG2a responses were generally weak (not shown). Pre-challenge IgG endpoint titers (four-fold serial dilutions) ranged from 256,000–1,024,000 for Sm-TSP-2 vaccinated mice and 256,000–4,096,000 for Sm-TSP-2/5B vaccinated mice. At necropsy (post-challenge), titers had waned to 64,000–256,000 for Sm-TSP-2 vaccinated mice and 64,000–1,024,000 for Sm-TSP-2/5B vaccinated mice. Mean and median anti-Sm-TSP-2 antibody titers were higher in the group vaccinated with Sm-TSP-2/5B (means 486,400 vs 1,450,667; medians 256,000 vs 1,024,000), implying that mice vaccinated with the chimera made a stronger antibody response against the Sm-TSP-2 region of the immunogen, and increased titers were not due to anti-5B antibodies. No obvious association between antibody titer and parasite burden was detected. Of the mice vaccinated with Sm-TSP-2/5B, two mice had no worms, one mouse had two worms and one mouse had four worms. All four mice had the lowest liver egg burdens and high antibody titers (≥1,024,000). However, two other mice had equally strong antibody titers but had higher parasite burdens (29 and 34 worms), precluding determination of a robust correlation between worm burdens and antibody titers.
Sm-TSP-2/5B and Sm-TSP-2 formulated with alum/CpG protected against experimental challenge with S. mansoni. Vaccinated groups had respective decreases in worm burden of 54–58% (Sm-TSP-2/5B, P<0.01) and 25–27% (Sm-TSP-2, P<0.05), compared to controls over two independent trials (Fig. 4A and 4B). A comparative reduction in liver egg burdens was also observed in these groups – 48–56% (Sm-TSP-2/5B, P<0.01) and 20–27% (Sm-TSP-2, P<0.05), respectively (Fig. 5A and 5B). When the data from both trials were combined, significant decreases in worm and liver egg burdens were seen between the group vaccinated with Sm-TSP-2/5B and the group vaccinated with Sm-TSP-2 (P<0.01 and P<0.05, respectively). Liver egg burdens were not disproportionately reduced compared with burdens of worms, suggesting no additional effect on parasite fecundity (Table 2).
Splenocytes from vaccinated and challenged animals were restimulated with SEA and SWAP to assess the cytokine responses to vaccination and parasite challenge. Levels of IL-4, IL-10 and IFN-γ from splenocytes were elevated in all infected animals compared to uninfected MBP-vaccinated animals when restimulated ex vivo with SEA and SWAP (Figure 6), indicating that infection-related cytokine responses were produced, although responses to SEA were generally higher. SEA and SWAP-specific IL-4 responses tended to increase in Sm-TSP-2/5B-vaccinated animals compared to control (MBP-vaccinated) infected animals, however this only reached significance with SWAP restimulation. IL-10 production in response to SWAP, but not SEA, was also increased due to Sm-TSP-2/5B vaccination. IFN-γ production in response to both SEA and SWAP were also highly significantly increased (P<0.01) in response to Sm-TSP-2/5B vaccination.
We have previously demonstrated that the large extracellular loop of the S. mansoni tegument tetraspanin, Sm-TSP-2, when linked to a thioredoxin fusion partner and formulated with Freund's adjuvants, is an efficacious vaccine antigen, eliciting high levels of protection in a murine schistosomiasis model of infection [12]. Herein, we show that modified and chimeric forms of the Sm-TSP-2 vaccine antigen are also protective, even when formulated with a human-approved adjuvant combination, and that a schistosomiasis vaccine based on Sm-TSP-2 (or Sm-TSP-2/5B) satisfies additional selection criteria for progression into clinical trials, such as safety concerns around the potentially deleterious effects of pre-existing IgE responses in helminth endemic populations [1], [19].
There is a paucity of funding - driven by the lack of a commercially viable market - available for the production of vaccines against the neglected tropical diseases, and so a vaccine antigen must be amenable to low-cost manufacture [22]. Despite attempts at optimisation of production of these two antigens being preliminary at best, both Sm-TSP-2 and Sm-TSP-2/5B have been expressed at yields that, at this initial stage, may be indicative of cost-effective up-scaling and clinical development. Indeed, Sm-TSP-2 has been recently produced in Pichia pastoris fermentation cultures in our laboratory at a yield or over 500 mg/L (data not shown) and efforts are currently underway to express Sm-TSP-2/5B in a similar fashion.
We recently suggested that the presence of a pre-existing human serum IgE response to a helminth vaccine antigen is a down-selection criterion [1] when considering a molecule for progression towards clinical trials because of the safety risks involved [19]. No detectable levels of Sm-TSP-2-specific IgE were found in individuals chronically infected with S. mansoni, despite very strong IgE responses to proteins found within SEA. This is also the case for the hookworm antigen, Na-APR-1 [23], the origin of the 5B domain in Sm-TSP-2/5B. Despite the absence of a detectable IgE response, previous studies have shown that humans from schistosome- and hookworm-endemic areas mount IgG1 responses to Sm-TSP-2 [12] and Na-APR-1 [23], indicating that both antigens are recognized by the immune system in a natural infection. What determines the isotype response (IgG vs IgE) mounted by an infected individual to a helminth antigen is multifactorial and an unresolved topic of debate [24]. What is clear, however, is the potential danger of developing a vaccine based on an antigen that is the target of a naturally acquired IgE response in the target population.
Of the two test groups, mice vaccinated with Sm-TSP-2/5B had the highest level of protection against experimental schistosomiasis. We initially hypothesized that this increased protection was due to cross-reactive epitopes within the 5B region of hookworm Na-APR-1 and its S. mansoni orthologue, Sm-catD [25]. However, numerous attempts to show binding of anti-Sm-TSP-2/5B to recombinant Sm-catD and schistosome extracts using Western blotting and immunoprecipitation coupled to tandem mass spectrometry (A. Dougall and A. Loukas, unpublished) were unsuccessful. Anti Sm-TSP-2/5B did, however, bind strongly to recombinant Na-APR-1 and inhibited the ability of the enzyme to cleave a synthetic substrate in a previous study [17] and has likewise been shown to neutralise the hemoglobinase capacity of Na-APR-1 in this study; indeed, the 5B region of Sm-TSP-2/5B is highly immunogenic and was the target of a panel of IgG1 mAbs raised to recombinant Na-APR-1 [17]. Production of a chimeric antigen comprising Sm-TSP-2 and the 5B region of Sm-CatD from S. mansoni instead of Na-APR-1 is currently underway in our laboratory and may have the additional benefit of being able to induce an antibody-mediated neutralization of Sm-TSP-2 in the tegument and Sm-CatD in the gut of the intra-mammalian stages of S. mansoni. Given the absence of an obvious cross-reactive schistosome epitope for antibodies to the Na-APR-1/5B fragment, we therefore sought to confirm whether the increased protection obtained with Sm-TSP-2/5B compared to Sm-TSP-2 alone was due to the increased size and therefore increased immunogenicity of the chimera. When microtiter plates were coated with Sm-TSP-2 and probed with antisera from mice immunized with Sm-TSP-2 or Sm-TSP-2/5B, the IgG endpoint titers were higher on average for the group immunized with Sm-TSP-2/5B, implying that vaccination with the larger immunogen resulted in an increased TSP-2-specific antibody titer. We also noted that individual mice with the highest antibody titers had the fewest worms, as highlighted in Table 2. Studies are also in progress to determine whether the chimeric protein generates similar levels of protection against hookworm infection caused by Necator americanus.
Restimulation of splenocytes from vaccinated and infected mice prior to necropsy showed a general increase in both Th2 (IL-4), regulatory (IL-10) and Th1 (IFN-γ) responses to parasite antigens, which was especially marked in increased IFN-γ production by mice vaccinated with Sm-TSP-2/5B compared to those vaccinated with the control non-parasite protein MBP. This implies that every animal was effectively challenged, indicating that the recovery of very few or no parasites in some mice was not due to an unsuccessful infection but successful vaccination. These data also suggest that Th1 cytokines have a role in the protective response against schistosomiasis, a finding that has been documented in infection studies with the parasite [26], [27] and vaccination experiments with recombinant vaccine candidate antigens from the tegument such as Sm29 [28] and Sm14 [29].
A caveat of using human-approved adjuvants to test vaccine antigens in the early stages of process development is that the full potential of a candidate antigen may not be realized due to the increased immunostimulatory properties of adjuvants containing mycobacteria and other toxic components, such as Freund's adjuvants. However, the levels of protection reported herein for Sm-TSP-2/5B were similar to those reported for Freund's formulated thioredoxin-Sm-TSP-2, and well exceed the 40% benchmark set by the WHO for progression of an antigen into clinical trials irrespective of the adjuvant used [30].
Sm-TSP-2 immunolocalizes to the surface of schistosomula [16] and adult worms [12] and has been found in the outer tegument of mature schistosomes [10] in abundance using proteomic techniques [31]. The ultrastructural morphology of adult worms and schistosomula treated in vitro with Sm-tsp-2 double-stranded RNA displayed a distinctly vacuolated and thinner tegument compared to controls, suggesting that Sm-TSP-2 may play a pivotal role in tegument development in the early stages of intra-mammalian development [16]. These insights into Sm-TSP-2 function, along with the apparent importance of humoral immunity in anti-Sm-TSP-2 vaccination, lead us to hypothesize that the surface of the schistosomulum and adult fluke are potential sites of immune attack where these crucially important membranes are being opsonized by anti-Sm-TSP-2 antibodies for further attack by complement, antibody-dependent cellular mechanisms, or both. We are currently exploring the immunologic mechanisms responsible for vaccine-induced efficacy using genetically modified mice.
The Sm-TSP-2-based vaccine antigens reported in this study appear to exhibit all the early-stage characteristics of a vaccine targeting developing countries where schistosomiasis is endemic, based on their ease of production, absence of IgE reactivity, preferential recognition by resistant humans [12], essential nature of the protein for parasite survival [16] and vaccine efficacy in animal models. These features, coupled with the recent finding of a lack of polymorphism between geographical isolates of Sm-TSP-2 throughout Africa [32] provide a compelling argument for the use of Sm-TSP-2-based antigens as safe and effective anti-schistosomiasis vaccines. These additional studies also open the door to exploring more than a single helminth target with a single antigen.
|
10.1371/journal.pntd.0004556 | New Trypanosoma evansi Type B Isolates from Ethiopian Dromedary Camels | Trypanosoma (T.) evansi is a dyskinetoplastic variant of T. brucei that has gained the ability to be transmitted by all sorts of biting flies. T. evansi can be divided into type A, which is the most abundant and found in Africa, Asia and Latin America and type B, which has so far been isolated only from Kenyan dromedary camels. This study aimed at the isolation and the genetic and phenotypic characterisation of type A and B T. evansi stocks from camels in Northern Ethiopia.
T. evansi was isolated in mice by inoculation with the cryopreserved buffy coat of parasitologically confirmed animals. Fourteen stocks were thus isolated and subject to genotyping with PCRs targeting type-specific variant surface glycoprotein genes, mitochondrial minicircles and maxicircles, minisatellite markers and the F1-ATP synthase γ subunit gene. Nine stocks corresponded to type A, two stocks were type B and three stocks represented mixed infections between A and B, but not hybrids. One T. evansi type A stock was completely akinetoplastic. Five stocks were adapted to in vitro culture and subjected to a drug sensitivity assay with melarsomine dihydrochloride, diminazene diaceturate, isometamidium chloride and suramin. In vitro adaptation induced some loss of kinetoplasts within 60 days. No correlation between drug sensitivity and absence of the kinetoplast was observed. Sequencing the full coding sequence of the F1-ATP synthase γ subunit revealed new type-specific single nucleotide polymorphisms and deletions.
This study addresses some limitations of current molecular markers for T. evansi genotyping. Polymorphism within the F1-ATP synthase γ subunit gene may provide new markers to identify the T. evansi type that do not rely on variant surface glycoprotein genes or kinetoplast DNA.
| Trypanosoma (T.) evansi causes surra in various animal species in Africa, Latin America and Asia. Despite inducing important animal suffering, economic losses and being a World Animal Health Organisation (OIE) notifiable disease, surra is severely neglected in terms of awareness, control interventions and research into improved control tools. Most serological tests can only detect T. evansi type A, while molecular tests rely on detection of highly variable genes or on fragile kinetoplast DNA. Even more, the obscure T. evansi type B, first isolated decades ago in Kenya, totally escapes surveillance due to absence of reliable diagnostic tools. In the present study we isolated new type B stocks from Ethiopia, thus suggesting that this type of T. evansi is probably more widely distributed than previously thought. We further report on an alternative molecular marker for both types of T. evansi and present data on the drug sensitivity of the Ethiopian isolates.
| Surra, a wasting disease caused by Trypanosoma (T.) evansi, is one of the non tsetse-transmitted Animal African Trypanosomoses (AAT) occurring in Ethiopia. The disease imposes significant financial losses due to reduced fertility and mortality and is prohibiting the import of highly productive yet trypanosusceptible cattle breeds [1–3]. T. evansi belongs to the subgenus Trypanozoon, that also comprises T. brucei and T. equiperdum [4–6]. In terms of geographical distribution, Trypanosoma equiperdum and T. evansi, causing respectively dourine in horses and surra in livestock in Africa, Asia, and South America, have been far more successful than T. brucei, a parasite confined to sub-Saharan Africa where its vector, the tsetse fly, is present [7]. Recent phylogenetic studies suggest that T. evansi and T. equiperdum evolved from T. brucei on several occasions and from genetically distinct T. brucei strains and therefore could be considered as subspecies of T. brucei [8,9].
Trypanosomes are characterised by the presence of a structure called kinetoplast that corresponds with the DNA (kDNA) of their unique mitochondrion. T. brucei kDNA contains 20–50 copies of maxicircles (about 23 kb) and a highly diverse set of thousands of minicircles (about 1 kb). Maxicircles contain rRNA coding regions and genes coding for subunits of the respiratory chain complexes while minicircles code for guide RNAs required for editing [10].
T. equiperdum and T. evansi are dyskinetoplastic (kDNA-) since they lack part of the kDNA [8–11]. T. equiperdum typically has retained maxicircles, in some cases with substantial deletions, but has lost its minicircle diversity. T. evansi does not have maxicircles and either shows minicircle homogeneity or are akinetoplastic (kDNA) [10,12–14]. Based on their minicircle restriction digestion profile, T. evansi can be divided into type A and type B [15,16].
T. evansi type A is the most abundant and is found in Africa, South America and Asia. It is characterised by the presence of the gene for the variant surface glycoprotein (VSG) RoTat 1.2. This RoTat 1.2 VSG is expressed early during infections resulting in the detectability of anti-RoTat 1.2 antibodies in animals infected with T. evansi type A [17,18]. In contrast, T. evansi type B is far less common and has so far been isolated only from camels in Kenya [16,19]. More recently, serological and molecular evidence for the presence of T. evansi type B in Sudan, Ethiopia and Chad was published [20–24]. T. evansi type B lacks the RoTat 1.2 gene and as a consequence, infections with this type are not detected with serological and molecular tests based on RoTat 1.2 VSG, such as the CATT/T. evansi and RoTat 1.2 PCR [15,18,19,25]. So far, three molecular tests have been developed for the identification of T. evansi type B: the EVAB PCR, targeting a type B-specific minicircle DNA sequence, and a PCR and a LAMP targeting a type B-specific VSG JN 2118Hu [15,19,26]. T. equiperdum is the least known parasite of the Trypanozoon group, with very few isolates available for research, albeit new stocks were isolated from Ethiopian and Venezuelan horses recently [24,27].
Unlike T. brucei, T. evansi and T. equiperdum cannot develop in tsetse flies due to their inability to transform into the procyclic life stage. They can only survive in a mammalian host where they produce ATP exclusively through glycolysis. In contrast to bloodstream forms, ATP production in procyclic trypanosomes relies on oxidative phosphorylation and, therefore, on the capacity to express the full set of corresponding mitochondrial genes, including some which are encoded by the kDNA [10,28]. Bloodstream forms of T. evansi, T. equiperdum and laboratory-generated T. brucei strains that have lost all or critical parts of their kDNA, can survive without kDNA due to specific single amino acid mutations in the gamma (γ) subunit of the mitochondrial F1-ATP synthase [28]. Interestingly, the specific mutations/deletions in the C-terminal region of F1-ATP synthase γ subunit enable differentiation among the Trypanozoon strains [8]. Furthermore, when the F1-ATP synthase γ subunits of T. evansi type A (A281del), T. equiperdum (A273P) and the laboratory-generated T. brucei (L262P) strains are overexpressed in a T. brucei γ subunit knock out strain, the latter can survive after loss of its kinetoplast after treatment with DNA intercalating drugs such as acriflavin or ethidium bromide [28,29]. Once the genetically modified T. brucei are independent from kDNA maintenance and expression, they become multidrug resistant to the diamidine and phenanthridine class of drugs [30].
In T. evansi, drug resistance has been reported in several type A strains originating from Africa, Asia and Latin America [31–34]. Some Chinese strains appear to be innately resistant to the phenanthridine class of drugs [35]. In contrast, nothing is known on the drug susceptibly of the T. evansi type B strains. In a previous study, we reported that T. evansi infections are very common in camels, equines, cattle and small ruminants in Tigray and Afar provinces in Northern Ethiopia [20]. We also provided molecular and serological evidence that both T. evansi type A and type B occur in these provinces. In that study, of those dromedary camels that were parasitologically positive, buffy coat samples were collected and cryopreserved in liquid nitrogen for later isolation of the parasite. We here report on the isolation, adaptation to in vitro culture, genetic and phenotypic characterisation and in vitro drug sensitivity of T. evansi type A and B from Northern Ethiopia.
The Animal Experimentation Ethics Committee (AEEC) of the Institute of Tropical Medicine (ITM) advised on the protocol for collection of blood samples from dromedary camels (EXT2012-1) and for the isolation of trypanosomes via inoculation of mice (EXT2012-2) at the College of Veterinary Medicine, Mekelle University. The study protocol for in vivo expansion of trypanosomes at ITM was approved by the AEEC (BM2013-1). Collecting blood from camels and experiments on mice were conducted according to the national guidelines of the Ethiopian Ministry of Livestock and Fishery Development and the Institutional Review Board of the Ministry of Science and Technology.
Details on the collection and cryopreservation of buffy coat samples from dromedary camels that were parasitologically confirmed in the micro haematocrit centrifugation technique have been fully described elsewhere [20]. Two hundred μl of thawed buffy coat were inoculated intraperitoneally (IP) in two 25–30 g Swiss albino mice that were immunosuppressed with 0.16 μg kg-1 body weight dexamethasone (Shanghai Central Pharmaceutical, China) one day prior to inoculation [36]. Parasitaemia was checked in 5 μl of tail blood using the matching method [37], starting from day 7 post-infection and subsequently on every third day. As soon as trypanosomes were detected in at least one mouse, the animal was anaesthetised (the other kept as a backup), its blood was collected on heparin by heart puncture, diluted in an equal volume of phosphate buffered saline glucose (PSG; 7.5 g/l Na2HPO42H2O, 0.34 g/l NaH2 PO4H2O, 2.12 g/l NaCl, 10 g/l D-glucose, pH 8) and subinoculated into four naïve mice (200 μl each) which were monitored for parasitaemia as described above. Mice used as backup were euthanised when the newly infected mice became positive. When parasitemia reached about ± 107.8 cells ml−1 of blood, two of these parasitaemic mice were euthanised (the other two were kept as back up) and blood was taken for subinoculation into four other naïve mice. This protocol was repeated until the parasitaemia reached about 108.4 cells ml−1. At this stage the stock was considered in vivo adapted. All four mice were anaesthetised and exsanguinated by heart puncture in an equal volume of Triladyl-egg yolk-phosphate buffered saline glucose (TEP) cryomedium [38] for cryopreservation in 1 ml aliquots.
Cryostabilates were thawed in a water bath at 37°C and diluted in PSG to 1 trypanosome per field (± 105.7 cells ml−1). Two-hundred μl volumes were injected IP in two naïve 20–30 g female OF-1 mice (Charles River, Belgium). Starting from three days post infection (DPI), parasitaemia was monitored daily and harvested at first peak parasitemia, typically at day 4 to 5 post-infection, as described above. Volumes of 0.5 ml of the blood were run over a mini Anion Exchange Centrifugation Technique (mAECT) column to separate the trypanosomes from the blood [39]. The trypanosomes eluted from the column were washed twice with 5 ml ice-cold PSG by centrifugation at 1500 g for 15 min. After the last centrifugation, the supernatant PSG was discarded and the trypanosome sediment was re-suspended in 100 μl of PSG. Part of this suspension was used for in vitro culture adaptation. The remainder was centrifuged at 1500 g for 5 min and the sediment was frozen at -80°C until DNA extraction. The isolates used for in vivo isolation and expansion and the corresponding T. evansi type A and B specific PCR result on their corresponding buffy coat DNA are indicated in Table 1.
The highly concentrated trypanosome suspension in PSG was diluted to 2 x 105 cells ml−1 in Hirumi’s modified Iscove’s medium 9 (HMI-9), complemented with 15% (v/v) heat-inactivated foetal bovine serum (Gibco, Belgium) and 5% (v/v) heat-inactivated horse serum (Gibco, Belgium) (abbreviated as HMI-9 (HS)) [40,41]. Parasites were seeded at 2 x 104, 2 x 103 and 2 x 102 cells ml−1, in a total volume of 500 μl in a 48-well plate (Nunc, Denmark) and incubated at 37°C and 5% CO2. After 72 hours, a well, where trypanosome density had increased above 2 x 105 cells ml−1, was used for further subpassage in 500 μl of HMI-9 (HS). The well with the highest density of viable parasites was then further maintained in HMI-9 without horse serum [40]. When possible, log phase growing in vitro cultures were scaled up in flasks (Nunc, Denmark) to obtain larger numbers of parasites for cryostabilisation, DNA extraction and in vitro drug sensitivity testing [42]. The in vitro growth curves of the different stocks were generated by seeding cells at 1 x 104 cells ml−1 in 500 μl of HMI-9 in three replicate wells that were counted every 24 h. The doubling times (Td) were calculated from the exponential part of the curve using non-linear regression fitted with an exponential equation in GraphPad Prism 6 (GraphPad, version 6, USA).
DNA extraction of trypanosome sediments prepared from the in vivo expanded and the in vitro adapted populations was performed with DNA Isolation Kit (Roche Diagnostics, Germany) following the protocol recommended for isolation of DNA from mammalian tissue. From T.b. brucei AnTat 1.1E, T.b. gambiense LiTat 1.3, T.b. gambiense type II ABBA and T. equiperdum Dodola 940, DNA was extracted using the Maxwell 16 Tissue DNA Purification kit on a Maxwell 16 instrument according to the manufacturer's instructions (Promega, Belgium). DNA concentrations were measured using the Nanodrop ND-1000 UV-Vis spectrophotometer (NanoDrop Technologies, USA) and adjusted to 10 ng μl-1. A set of PCRs targeting VSG genes (RoTat 1.2 and JN 2118Hu), maxicircle genes (ND4, ND5, ND7 and A6), class A minicircles (miniA PCR) and class B minicircles (EVAB PCR) minisatellites (MORF-2REP), P2 adenosine transporter (AT1) and the F1-ATP synthase γ subunit were adopted to characterise the studied parasite populations [4,15,19,28,43–45]. Where applicable, the published PCR protocols were adjusted to the requirements of the HotStarTaq Plus DNA polymerase (Qiagen, Germany). Primer sequences, reaction mixture contents, cycling conditions and expected amplicon size are described and referenced in Table 2. All PCR amplifications were carried out in 200 μl thin-wall PCR tubes (ABgene, UK) in a T3 thermocycler 48 (Biometra, Germany). Ten μl of amplified products were electrophoresed in 1 to 2% agarose gel at 135 V for 30 min and afterwards stained with ethidium bromide for visualization under UV light. For direct sequencing, PCR was performed in 50–100 μl volumes and amplicons were cleaned up and concentrated using a PCR cleanup kit (QIAquick PCR Purification Kit, Qiagen, Germany) and sent out for bidirectional direct sequencing at the Genetic Sequencing Facility (VIB, Belgium) using the described PCR primers.
The full length sequence of the F1-ATP synthase γ subunit was cloned into a BamHI and HindIII double digested pHD309 vector using the In-Fusion Cloning kit (Clontech, Japan). Primers contained a F1-ATP synthase γ subunit specific sequence based on the T. evansi sequence of STIB 810 (EU185797) and a 5′ extension of 15 bp specific to the place of integration in pHD309, containing the restriction sites and sequence overlap with the vector, as required for the In-Fusion Cloning reaction. Proofreading-PCR was performed using the Clone-Amp HiFi PCR premix (Clontech, Japan). Amplicons were cleaned up (QIAquick PCR Purification Kit, Qiagen, Germany) before use in the In-Fusion protocol. The reaction products were transformed in Stellar competent cells according to the manufacturer's recommendations (Clontech, Japan). Transformant clones were checked for the presence of insert using colony PCR, cultured in LB medium, plasmid purified (QIAprep Spin Miniprep Kit, Qiagen, Germany) and at least 7 to 12 clones per transformation were bidirectionally sequenced at the Genetic Sequencing Facility (VIB, Belgium) using primers binding to pHD309.
Melarsomine dihydrochloride (Cymelarsan, Sanofi Aventis, France) and isometamidium hydrochloride (Veridium, Ceva Santé Animale, Belgium) were prepared as 10 mg ml−1 stock solutions in distilled water. Dophanil powder (Dophanil, Docpharma, Belgium), containing 445 mg diminazene diaceturate and 555 mg antipyrine per gram, was concentrated to a 10 mg ml−1 diminazene diaceturate solution in DMSO (Sigma, Belgium). Suramin (Germanin, Bayer, Germany) was prepared as a 100 mg ml−1 in DMSO. A method to measure the IC50 values of compounds in 96-well plates was performed as described elsewhere [46]. Briefly, 2 × 104 cells ml−1 from in vitro adapted stocks were exposed to seven threefold drug dilutions, ranging from 5000 to 7 ng ml−1 for suramin, 500 to 0.7 ng ml−1 for diminazene diaceturate and from 250 to 0.35 ng ml−1 for melarsomine dihydrochloride and isometamidium hydrochloride, in a total volume of 200 μl of HMI-9 medium. Next, the plate was incubated for 72 hours at 37°C with 5% CO2 followed by addition of 20 μl of resazurin (Sigma, Belgium; 12.5 mg in 100 ml PBS) for measuring trypanosomes viability. After a further 24 h incubation at 37°C and 5% CO2, fluorescence was measured (excitation λ = 560 nm; emission λ = 590 nm) with a VictorX3 multimodal plate reader using top reading (Perkin Elmer, Belgium) [42]. The results were expressed as the percent reduction in parasite viability compared to the parasite viability in control wells without drugs. The 50% inhibitory concentration (IC50) was calculated using non-linear regression fitted with a (log) inhibitor versus normalised response (variable slope) equation (GraphPad, version 6, USA). The IC50 values obtained from day 30 and day 60 in vitro cultures were compared using t-tests corrected for multiple testing according to the Holm-Sidak method (α = 0.05) (GraphPad, version 6, USA).
Trypanosome populations at different stages of in vivo and in vitro expansion were examined for the presence of the kinetoplast using 4',6-diamidino-2-phenylindole (DAPI) staining. Briefly, live trypanosomes in PSG or in vitro culture medium were washed in PBS by centrifugation, deposited onto microscope slides, air dried and fixed with methanol for 30 min. Subsequently, the slides were rehydrated in PBS and mounted in 87% glycerol containing 1 μg ml-1 DAPI (Sigma, Belgium) [28]. Images were captured with an epifluorescence microscope (Olympus BX41, Olympus, Japan) equipped with a NU fluorescent cube (excitation: 360–370 nm and emission > 420 nm)) and Cell˄D software (Olympus, Japan). DAPI stained trypanosomes were grouped according to the number of kinetoplasts (K) and nuclei (N) present within each cell. The percentage of kinetoplastic cells in a DAPI stained slide was calculated on the basis of on average 300 examined trypanosomes, by dividing the sum of 1K1N + 2K1N + 2K2N cells by the sum of 1K1N + 2K1N + 2K2N + 0K1N + 0K2N cells. A two-tailed Spearman correlation matrix (using a confidence interval of 95%) was used to find the correlation between the percentage of kinetoplastic cells at day 30 and day 60 of in vitro culture and the respective IC50 value for a particular drug (GraphPad, version 6, USA).
To check the in vivo infectivity of trypanosome populations that were cryostabilised after continuous propagation in vitro for 60 days, 5 x 106 cells in 300 μl were inoculated in a single OF-1 mouse where after parasitaemia was checked as described above.
Thirty cryopreserved buffy coat specimens from parasitologically positive dromedary camels were inoculated in immunosuppressed Swiss albino mice. In total, 22 parasite stocks originating from 22 different animals could be isolated and cryopreserved after 2 to 5 subpassages in mice. They were labelled as MCAM/ET/2013/MU/01 to MCAM/ET/2013/MU/22. Based on positivity in RoTat 1.2 PCR and EVAB PCR of the corresponding cryopreserved buffy coats, 20 of these stocks are T. evansi type A and 2 are T. evansi type B (Table 1) [20]. Copy cryovials of these primary isolates were brought to ITM, Antwerp and 14 were selected for further expansion in mice. The selection was based on their geographical origin and subtype: 12 type A stocks originated from different sampling stations in Afar and Tigray (MCAM/ET/2013/MU/01, 02, 04, 05, 06, 07, 08, 09, 11, 13, 15, 17) and two type B stocks (MCAM/ET/2013/MU/10 and 14) were from Awash Fentale in Afar. At peak parasitaemia, between 4 to 7 DPI, parasites were harvested, purified from blood using a mAECT column, washed with PSG and pelleted for DNA extraction and for in vitro culture adaptation.
DNA extracts of in vivo expanded stocks were subjected to RoTat 1.2 PCR and JN 2118Hu PCR to identify the T. evansi type based on type-specific VSG sequences. In addition, the specificity of these PCRs was tested on DNA of other Trypanozoon strains (T.b. brucei AnTat 1.1E, T.b. gambiense LiTat 1.3, T.b. gambiense type II ABBA, T. evansi type A RoTat 1.2, T. evansi type B KETRI 2479 and T. equiperdum Dodola 940). Results are represented in Table 3. All the in vivo expanded stocks that originated from RoTat 1.2 PCR positive buffy coats, were also positive in RoTat 1.2 PCR (MCAM/ET/2013/MU/01, 02, 04, 05, 06, 07, 08, 09, 11, 13, 15 and 17). Direct sequencing of the 488 bp amplicons from these putative T. evansi type A stocks and the T. evansi RoTat 1.2 strain revealed 100% identity (in a 350 bp sequenced fragment) with the published RoTat 1.2 VSG sequence (AF317914), thus identifying them as T. evansi type A. Only one synonymous polymorphism (C699A) was found in MCAM/2013/ET/MU/04. The gel with the RoTat 1.2 PCR products from the purified trypanosomes showed a faint band of about 400 bp amplified in T. evansi KETRI 2479 and in MCAM/ET/2013/MU/10 and 14. Direct sequencing of these 400 bp amplicons failed. The PCR targeting the T. evansi type B specific VSG JN 2118Hu generated the expected amplicon in T. evansi type B KETRI 2479 and in MCAM/ET/2013/MU/10 and 14. Additionally, an amplicon was generated from MCAM/ET/2013/MU/15. Also for T.b. brucei AnTat 1.1E and T.b. gambiense type II ABBA, amplicons of 273 bp were produced in the JN 2118Hu PCR. Direct sequencing of these amplicons revealed that the Ethiopian T. evansi type B MCAM/ET/2013/MU/10 and 14, T. evansi type B KETRI 2479 and T.b. brucei AnTat 1.1E were 100% identical (in a 190 bp sequenced fragment) to the corresponding sequence of JN 2118Hu VSG (AJ870486). In T.b. gambiense type II ABBA, one synonymous mutation (G300A) was found.
Four PCRs that target maxicircle DNAs, of which three NADH-dehydrogenase subunits (ND4, ND5, ND7) and the ATPase subunit 6 (A6), and two PCRs that target class-specific minicircle sequences (miniA PCR and EVAB PCR) were run on DNA extracts of the purified trypanosomes (Table 3). All Ethiopian T. evansi stocks and T. evansi type A RoTat 1.2 and T. evansi type B KETRI 2479 were negative for all four maxicircle genes, while T.b. brucei AnTat 1.1E, T.b. gambiense LiTat 1.3, T.b. gambiense type II ABBA and T. equiperdum Dodola 940 were positive for all four maxicircle genes.
All stocks that contain RoTat 1.2 VSG, except MCAM/ET/2013/MU/09, were positive in miniA PCR. Additionally, weak amplification was seen in T.b. brucei AnTat 1.1E. MCAM/ET/2013/MU/10 and 14 were positive in EVAB PCR, confirming their identification as T. evansi type B as observed on their corresponding buffy coat specimens (Table 1). Additionally, EVAB PCR amplicons were detected in 3 stocks that were also positive for RoTat 1.2 VSG PCR suggesting a mixed infection with type A and B: a strong amplification was present in MCAM/ET/2013/MU/15, while a weak amplification was visible in MCAM/ET/2013/MU/11 and 17. The presence of kinetoplasts in the trypanosome cells was demonstrated using fluorescence microscopy with DAPI staining on ex vivo isolated trypanosomes (Table 3). T. evansi RoTat 1.2, T. evansi KETRI 2479 and all but one Ethiopian T. evansi stocks show a kinetoplast in > 96% of the cells. Stock MCAM/ET/2013/MU/09 was found to be akinetoplastic since only the nucleus of the trypanosomes was visible with DAPI.
In T. evansi RoTat 1.2, the MORF2-REP locus consists of 4 and 6 repeats, while in T. evansi KETRI 2479, 3 and 5 repeats were found (Table 3). In vivo expanded Ethiopian stocks of type A had either 1 allele (7 repeats) or 2 alleles (6 and 7 repeats), thus displaying a different pattern than T. evansi type A RoTat 1.2. The Ethiopian type B stocks MCAM/ET/2013/MU/10 and 14 contain 3 and 4 repeats, and thus have a pattern different from T. evansi type B KETRI 2479. MCAM/ET/2013/MU/15 showed a clear pattern of the Ethiopian type B (3 and 4 repeats), and double allele pattern of the Ethiopian type A (6 and 7 repeats). The other presumed mixed type A and type B stocks MCAM/ET/2013/MU/11 and 17 showed only the Ethiopian type A T. evansi pattern (Fig 1). DNA extracted from the buffy coats revealed similar MORF2-REP patterns as the in vivo expanded trypanosomes except for the buffy coat of MCAM/ET/2013/MU/15 that revealed only the Ethiopian type A MORF2-REP pattern. The other Trypanozoon strains showed the following patterns: T. b. gambiense LiTat 1.3 had 7 and 11 repeats, T.b. gambiense type II ABBA had 3 repeats, T. equiperdum Dodola 940 had 11 repeats, while no amplicons were generated from T.b. brucei AnTat 1.1E under the giving PCR conditions.
Sequence analysis of in total 136 clones of the full length F1-ATP synthase γ subunit, amplified from DNA of the in vivo expanded Ethiopian stocks MCAM/ET/2013/MU/04, 06, 09, 10, 11, 13, 14, 15 and of T.b. brucei AnTat 1.1E, T.b. gambiense LiTat 1.3, T. evansi RoTat 1.2, T. evansi KETRI 2479, T. b. gambiense type II ABBA and T. equiperdum Dodola 940 revealed diverse homozygous and heterozygous nucleotide polymorphisms spread over the entire coding sequence (Table 4).
The F1-ATP synthase γ subunit of T.b. gambiense LiTat 1.3 (KT934830) appeared homozygous and identical to the T.b. gambiense DAL972 sequence (Tbg972.10.90). T.b. gambiense type II ABBA (KT934831) appeared homozygous and differed in only 2 SNPs (G801T and A882G) from the T.b. gambiense sequence. T. evansi RoTat 1.2 and the Ethiopian stocks MCAM/ET/2013/MU/04, 06, 09,11 and 13 were heterozygous and revealed in one allele (KT934833), identical to the published full length T. evansi STIB 810 (EU185798) sequence, the deletion of nucleotides A841-843del. The second allele contained a C142T polymorphism (KT934832), that is not present in the wild-type T. evansi STIB 810 sequence (EU185797), but that could be identified in the genome sequence of the Chinese akinetoplastic T. evansi STIB 805 strain [9]. For T. evansi KETRI 2479 and the Ethiopian stocks MCAM/ET/2013/MU/10 and 14 we obtained heterozygous alleles, different from the partial sequence of T. evansi KETRI 2479 (EU185794). The first allele had the unique A844T polymorphism (KT934835), and differed from the second allele in 3 additional SNPs (T321C, T807C, T867G) that were also found in some T.b. brucei and T. equiperdum. Interestingly, the in vivo expanded stock of MCAM/ET/2013/MU/15 revealed alleles that belonged to T. evansi type A and type B. In contrast, when the original buffy coat of this stock was tested, only alleles of T. evansi type A were found. Finally, T. equiperdum Dodola 940 (KT934836) appeared homozygous and its single allele was identical to one of the two alleles found in T.b. brucei AnTat 1.1E (KT934837), but differed in 5 SNPs with the sequence from T. equiperdum BoTat 1.1 (EU185793) and in 6 SNPs with T. equiperdum STIB 841 (EU185792). However, for the T. equiperdum STIB 841 strain, 5 of the 6 SNPs were ambiguous polymorphisms that do not rule out similarity to T. equiperdum Dodola 940.
Fourteen Ethiopian T. evansi stocks, T. evansi RoTat 1.2 and T. evansi KETRI 2479 were expanded in mice and purified from blood at peak parasitaemia to initiate primary in vitro cultures in HMI-9 (HS) medium. After 96 hours, the initial 2x104 cells ml−1 inoculum reached concentrations above 2x105 cells ml−1 for all the different stocks. These cells were used for further in vitro propagation by subpassage in fresh medium. Over the next 72 hours, only MCAM/ET/2013/MU/09, 14 and 15, and T. evansi RoTat 1.2 and T. evansi KETRI 2479 showed proliferation. In contrast, slightly increased cell densities were observed for MCAM/ET/2013/MU/01, 04, 06 and 10. For all other strains not a single inoculum proliferated and longer incubation led to growth cessation.
Because the HMI-9 (HS) medium did not support sufficient in vitro culture growth for most of the Ethiopian T. evansi stocks, it was abandoned and replaced with HMI-9 without horse serum. In vitro adapted strains of T.b. brucei AnTat 1.1E and T.b. gambiense LiTat 1.3 were cultured in HMI-9 in parallel. In vitro cultures were only considered adapted to HMI-9 medium when it was possible to maintain the parasites in continuous proliferation. To this extent, dense parasite cultures, containing 2–5 x 105 cells ml−1, were subpassaged into new wells using serial fivefold dilutions in fresh medium. When these subpassages reached densities above 2 x 105 cells ml−1 within a 48–96 hours period, the stock was considered adapted. The five stocks that already grew well in the HMI-9 (HS) medium continued proliferating when inoculated from the dense cultures at serial fivefold dilutions in HMI-9. These five stocks were considered to be in vitro adapted after 15 days of in vitro culture. Out of the four remaining stocks, only MCAM/ET/2013/MU/04 and 10 slowly regained the ability to proliferate in HMI-9 at a reduced subpassaging scheme using serial twofold dilutions. MCAM/ET/2013/MU/04 required 25 days to adapt, while MCAM/ET/2013/MU/10 was only fully adapted after day 35 of in vitro culture. Gradually increasing the culture volume allowed to obtain sufficient parasites from the adapted cultures for in vitro drug testing, DNA extraction, and cryostabilisation at day 30 (all, except MCAM/ET/2013/MU/10) and at day 60 of in vitro culture (all stocks).
DNA of the in vitro adapted stocks was subjected to RoTat 1.2 PCR, EVAB PCR and MORF2-REP PCR. All in vitro stocks had similar molecular profiles as their corresponding in vivo expanded parental stocks, except MCAM/ET/2013/MU/15. While the in vivo expanded stock of the latter was identified as a mixed infection of T. evansi type A and type B, the in vitro adapted stock (at day 30 and day 60 in vitro culture) was identified as pure T. evansi type B with the above mentioned PCRs and confirmed by cloning and sequencing of the F1-ATP synthase γ subunit. Thus, beside T. evansi RoTat 1.2 and T. evansi KETRI 2479, we achieved the in vitro adaptation of 2 Ethiopian type A stocks, 2 Ethiopian type B stocks and additionally ended up with a pure T. evansi type B in vitro adapted stock originating from a mixed type A and type B in vivo adapted stock. Growth curves were generated for T.b. brucei AnTat 1.1E and all seven in vitro adapted stocks (Fig 2). T.b. brucei AnTat 1.1E and T. evansi RoTat 1.2 had the shortest Td, 7.5 ± 0.3 h-1 and 7.7 ± 0.2 h-1 respectively, and reached the highest maximum population density (MPD) of ± 3–4 x 106 cells ml-1, while T. evansi KETRI 2479 had a longer Td, 10.8 ± 0.2 h-1, and a lower MPD of ± 1 x 106 cells ml-1. The Ethiopian type A stocks MCAM/ET/2013/MU04 and MU09 had a Td of 11.2 ± 0.4 and 11.3 ± 0.4 respectively, and a MPD of ± 1 x 106 cells ml-1. Similarly, the Ethiopian type B stocks MCAM/ET/2013/MU10, 14 and 15 had a Td of 12.9 ± 0.5, 11.3 0.5 and 12.1 ± 0.6 respectively, and a MPD of ± 0.7–1 x 106 cells ml-1 (Fig 2).
After day 30 and 60 of in vitro culture, IC50 values were determined for melarsomine dihydrochloride (Cymelarsan) (Fig 3A), isometamidium hydrochloride (Veridium) (Fig 3B), diminazene diaceturate (Dophanil) (Fig 3C) and suramin (Germanin) (Fig 3D). In general, non-significant differences (p > 0.05) were found between IC50 values recorded at day 30 and day 60 of in vitro culture, except for the melarsomine dihydrochloride IC50 values of T. evansi RoTat 1.2 and T. evansi MCAM/ET/2013/MU/14 and for the isometamidium hydrochloride IC50 values of T. evansi KETRI 2479 and T. evansi MCAM/ET/2013/MU/09 (p < 0.05). For comparison between the different stocks, the IC50 values of day 30 and day 60 of in vitro cultures were averaged. All Ethiopian T. evansi stocks had IC50 values for melarsomine dihydrochloride (IC50 1.9–3.3 ng ml-1) that were similar to those of T.b. gambiense LiTat 1.3 (IC50 4.3 ng ml-1), T.b. brucei AnTat 1.1E (IC50 6.8 ng ml-1), T. evansi RoTat 1.2 (IC50 3.0 ng ml-1) and T. evansi KETRI 2479 (IC50 4.1 ng ml-1). For isometamidium hydrochloride, the IC50 values of the Ethiopian T. evansi (IC50 0.6–6.2 ng ml-1) fall within the range of T.b. gambiense LiTat 1.3 (IC50 0.1 ng ml-1), T.b. brucei AnTat 1.1E (IC50 7.3 ng ml-1), T. evansi RoTat 1.2 (IC50 7.1 ng ml-1) and T. evansi KETRI 2479 (IC50 5.5 ng ml-1). However, the two Ethiopian T. evansi type A stocks (IC50 4.3–6.2 ng ml-1) appear to be threefold less sensitive that the three type B stocks (IC50 0.6–1.9 ng ml-1). For suramin, large differences in IC50 values were found among the Ethiopian T. evansi (IC50 15.9–261.5 ng ml-1) stocks and among the other strains: T.b. brucei AnTat 1.1E (IC50 39.5 ng ml-1) and T. evansi RoTat 1.2 (IC50 35.8 ng ml-1) appear highly susceptible, while T.b. gambiense LiTat 1.3 (IC50 134.0 ng ml-1) and T. evansi KETRI 2479 (IC50 222.4 ng ml-1) are less susceptible. The two Ethiopian T. evansi type A (IC50 153.5–261.5 ng ml-1) appear to be tenfold less sensitive than the three type B (IC50 15.9–27.6 ng ml-1). For diminazene diaceturate, the IC50 values of all Ethiopian T. evansi (IC50 17.5–48.5 ng ml-1) are higher than those of T.b. gambiense LiTat 1.3 (IC50 5.2 ng ml-1) and T. evansi RoTat 1.2 (IC50 13.8 ng ml-1), but similar to T.b. brucei AnTat 1.1E (IC50 39.6 ng ml-1) and T. evansi KETRI 2479 (IC50 24.0 ng ml-1). The two Ethiopian T. evansi type A (IC50 37.4–48.5 ng ml-1) appear to be twofold less sensitive than the three type B (IC50 17.5–25.9 ng ml-1). Direct sequencing of the full length TeAT1 PCR amplicons of MCAM/ET/2013/MU/04, 09, 10, 14, and 15, T. evansi type A RoTat 1.2 and T. evansi Type B KETRI 2479 revealed no polmorphisms to the wild-type TeAT1 sequence (AB124588).
DAPI staining was performed on in vivo and in vitro propagated stocks (Fig 4). In vitro culture did not change the percentage of kinetoplastic cells in T.b. gambiense LiTat 1.3 (99%), T.b. brucei AnTat 1.1E (99%) and MCAM/ET/2013/MU/09 (0%). On the other hand, already after 30 days in vitro culture a decrease in the percentage of kinetoplastic cells was observed in T. evansi RoTat 1.2 (89%), T. evansi KETRI 2479 (81%), MCAM/ET/2013/MU/04 (97%), 14 (93%) and 15 (94%) compared to non-in vitro adapted trypanosomes. After 60 days of in vitro culture, the percentage of kinetoplastic cells dropped even further for T. evansi KETRI 2479 (64%), MCAM/ET/2013/MU/04 (89%) and 10 (35%). No significant correlation was observed between the percentage of kinetoplastid cells of all in vitro adapted T. evansi stocks (including day 30 and day 60) and their IC50 values for melarsomine dihydrochloride (ρ = -0.13, p = 0. 67), isometamidium hydrochloride (ρ = -0.324, p = 0.278), suramin (ρ = -0.097, p = 0.752) and diminazene diacetureate (ρ = -0.355, p = 0.233). These data suggest that among the in vitro adapted Ethiopian T. evansi stocks there is no relation between the drug sensitivity and the presence of kinetoplast DNA. Furthermore, their loss of kDNA does not seem to influence rodent infectivity since all cryostabilates made from day 60 in vitro cultures remained infective for mice with detectable parasitaemia at 4–5 DPI.
Previous molecular and serological studies revealed that trypanosome infections in camels from Northern Ethiopia are caused by either RoTat 1.2 PCR or EVAB PCR positive parasites. In some instances amplicons of both PCRs were detected within the same buffy coat extract, suggesting the occurrence of mixed infections [20]. The present study was undertaken to isolate the trypanosomes from camels carrying apparent single infections through inoculation of their buffy coats in immunosuppressed mice. The in vivo inoculation led to the successful isolation of 22 stocks, out of which 14 were selected on the basis of their geographical origins for further investigations (5 stocks from Tigray and 9 stocks from Afar). Next, we performed an in-depth comparative molecular analysis on DNA extracts from the isolated parasite stocks using diverse PCRs. Furthermore, we analysed the specificity of each of these PCRs on a collection of Trypanozoon strains.
The RoTat 1.2 VSG sequence can be used to characterise T. evansi type A [25,43]. In our collection, all buffy coats positive in RoTat 1.2 PCR yielded in vivo isolated stocks that were RoTat 1.2 PCR positive but that were negative in the maxicircle gene targeting PCRs. Furthermore, with the exception of the akinetoplastic stock MCAM/ET/2013/MU/09, all these strains had type A minicircles. MCAM/ET/2013/MU/09 may be naturally akinetoplastic since the DNA extracted from the original buffy coat was negative in all PCRs targeting kinetoplast DNA. The occurrence of naturally akinetoplastic strains was previously documented in Latin America and China [12–14,47]. One stock (MCAM/ET/2013/MU/04) contained a SNP in its RoTat 1.2 VSG PCR amplicon. SNPs in RoTat 1.2 amplicons were previously reported in Egypt but do not necessarily lead to a negative result in RoTat 1.2 based antibody detection tests. This was also the case for the camel from which MCAM/ET/2013/MU/04 was isolated [48,49].
Initially defined by minicircle class B, identification of T. evansi type B is possible with EVAB PCR that amplifies a fragment of this minicircle [15]. Additionally, it was proposed that the VSG JN 2118Hu, first described in a Kenyan T. evansi strain, is a specific marker for T. evansi type B [19].
In our collection, 2 buffy coat extracts that were positive in EVAB PCR yielded in vivo isolated stocks that were EVAB PCR positive as well. Interestingly, an EVAB PCR amplicon was also detected in three additional in vivo expanded stocks that were RoTat 1.2 PCR positive but for which the corresponding buffy coats were EVAB PCR negative. These three stocks might be mixed infections. JN 2118Hu VSG PCR appeared to be less sensitive because it detected only 3 out of 5 EVAB PCR positive isolated stocks. Furthermore, the JN 2118Hu VSG PCR appeared to be less specific since T.b. brucei AnTat 1.1E and T.b. gambiense type II ABBA were also positive in this PCR. None of the EVAB PCR positive isolated stocks contained maxicircle DNA and they were all negative in miniA PCR, except for the three mixed infections. Therefore, we conclude that we isolated at least two “pure” T. evansi type B stocks from Ethiopian camels, decades after the initial isolation of T. evansi type B from camels in Kenya [15].
We used the minisatellite locus MORF2-REP to verify whether both putative mixed stocks, that were positive in RoTat 1.2 PCR and EVAB PCR, were real mixed infections or hybrids between T. evansi type A and B. The Ethiopian isolates clustered in two classes of T. evansi type A, of which one with a previously described heterozygous profile (6 and 7 repeats) and one with a homozygous profile (7 repeats). The Ethiopian T. evansi type B stocks had a heterozygous profile (3 and 4 repeats) differing from the only known profile described for Kenyan type B isolates (3 and 5 repeats) [50]. In one of the mixed infections we observed a profile that can be interpreted as a mixture of Ethiopian type A and type B, while the others only revealed the Ethiopian type A pattern. These results prove that we are dealing with mixed infections and not with hybrids between T. evansi type A and type B. To exclude that these apparent mixed infections represent cross-contamination with genetic material, we attempted in vitro cultivation of the in vivo expanded stocks.
Previously we have shown that addition of 1,1% methylcellulose to HMI-9 greatly helps the in vitro adaptation of Trypanozoon strains, including T.b. gambiense and T. evansi RoTat 1.2 [40]. However, to avoid the use of this highly viscous medium we preferred the use of horse serum to adapt T. evansi stocks as is suggested in previous reports [51–53]. While this approach proved to be successful for all type B stocks, only two out of nine Ethiopian T. evansi type A could be adapted. Interestingly, in the case of mixed stock MCAM/ET/2013/MU/15, this medium selected T. evansi type B out of the mixed population. While only the type A infection was detected in the buffy coat DNA extract, both types could be detected in the in vivo expanded stock DNA, but eventually only type B was detected in the in vitro adapted stock.
Gillingwater and colleagues reported on the drug sensitivity profiles of a panel of T. evansi and T. equiperdum strains where they considered T. evansi STIB 806K to be a reference sensitive strain for suramin (IC50 70.4 ng ml-1), diminazene diaceturate (IC50 4.5 ng ml-1) and melarsomine dihydrochloride (IC50 1.4 ng ml-1). They reported drug resistance in two T. evansi stocks with an IC50 for suramin > 10000 ng ml-1 (STIB 780 and STIB 781), and in the T. equiperdum OVI strain, with an IC50 for diminazene diaceturate of 302 ng ml-1 and an IC50 for melarsomine dihydrochloride of 17.6 ng ml-1 [46]. The only strain that is shared between their panel and our collection is T. evansi RoTat 1.2, which despite different approaches in the experimental testing, yielded corresponding IC50 values, especially for diminazene diaceturate and melarsomine dihydrochloride, thus facilitating comparison between both studies. In our Ethiopian T. evansi collection, no resistance against melarsomine dihydrochloride was found. However, some stocks appeared to have raised IC50 values for suramin (> 200 ng ml-1) and diminazene diaceturate (> 50 ng ml-1). The IC50 values that we observe for T.b. gambiense LiTat 1.3 and the Ethiopian T. evansi type B are similar to the in vitro IC50 value of 0.82 ng ml-1 found by Sahin and coworkers for T. congolense IL3000 which is sensitive to isometamidium (Veridium) in vivo [54]. In the same study, an in vitro IC50 of 11.06 ng ml-1 is reported for T.b. brucei AnTat 1.1 strain, which is slightly higher than the value that we obtained in experiments with our T.b. brucei AnTat 1.1 strain and the other T. evansi stocks [54]. Nevertheless, defining our T. evansi stocks as either sensitive or resistant based solely on the in vitro drug sensitivity results may be too audacious, given the fact that IC50 values were determined in only one assay, the resazurin viability assay [55–57]. Therefore, an in vivo drug sensitivity profile of all our Trypanozoon strains against the commonly used trypanocides remains to be elucidated. Interestingly, both Ethiopian T. evansi type A stocks appear to be less susceptible to suramin, diminazene diaceturate and isometamidium hydrochloride than the three type B stocks. In T.b. brucei, resistance against suramin and isometamidum hydrochloride has been linked to several proteins [58,59], while resistance to diamidine and melaminophenyl classes of drugs is attributed to the transporter protein TbAT1 and the aquaporin AQP2 [60–62]. The lower sensitivity to diminazene diaceturate was not caused by mutations in the T. evansi TeAT1 [63].
Interestingly, DAPI staining of the trypanosomes indicated slight to severe loss of the kDNA in all in vitro adapted T. evansi stocks, when compared to in vivo adapted stocks. The loss of kDNA in in vitro cultured T. evansi is a phenomenon that has been known for a long time [10,15,55,64]. Non-vital loss of the kinetoplast is made possible by mutations in the F1-ATP synthase γ subunit of T. evansi allowing to uncouple from the Fo subunit and effectively circumventing the requirement for mitochondrial gene expression [65]. Furthermore, it has been shown that the expression of certain T. evansi F1-ATP synthase γ subunit coding sequences in T. brucei allows this species to survive loss of its kDNA after chemical treatment [28]. Moreover, in such genetically modified T. brucei, independence of kDNA maintenance and expression is associated with multidrug resistance [30]. In our collection of T. evansi stocks we did not observe differences in drug sensitivity between populations that were partially or completely akinetoplast confirming earlier evidence that the presence or absence of kDNA is irrelevant within this context [30,55].
Recently, Carnes et al. showed that SNPs in the F1- ATP γ subunit could be used to genotypically support the multiple origins of at least 4 dyskinetoplastic T. evansi/T. equiperdum lineages: one major group of RoTat 1.2 VSG positive T. evansi/T. equiperdum type A, and three very small groups each represented by only a single strain: T. evansi type B KETRI 2479, T. equiperdum BoTat and T. equiperdum OVI [9]. All Ethiopian T. evansi type A had the corresponding mutation of the type A group. The Ethiopian type B T. evansi shared a similar profile as KETRI 2479. Finally, the Ethiopian T. equiperdum strain Dodola, which had some maxicircle genes but was negative for both type A and type B markers revealed an F1-ATP synthase sequence similar to T.b. brucei AnTat 1.1E strain, thus likely belongs to the same dyskinetoplastic group as T. equiperdum OVI [9,28].
In conclusion, our study shows that the apparent T. evansi type that is detected in a buffy coat of an infected camel does not necessarily represent the full diversity that is present in the infected animal. Moreover, the fact that 5 out of 22 new T. evansi isolates from camel in Ethiopia contain T. evansi type B may be an indication that is more widespread than currently known. The inoculation of the trypanosomes in immunosuppressed mice may allow the propagation of mixed populations. In contrast, in vitro cultivation seems to reduce the diversity by selecting for only one particular type, in our study T. evansi type B. Secondly, our study addresses some drawbacks of current molecular markers for T. evansi genotyping. To rely solely on VSG markers or kDNA markers for the molecular identification of T. evansi may be misleading due to possible recombinations occurring in VSG genes and to the presence of akinetoplastic T. evansi stocks. In this regard, we confirm that the F1-ATP synthase γ subunit gene, that is not related to the VSG repertoire nor to the presence of kDNA, may become an interesting target for genotyping T. evansi stocks in areas where both types overlap and where mixed infections can occur. Nevertheless, it is not possible to separate the Ethiopian T. equiperdum from T. brucei on the basis of this target gene. Thirdly, no evidence of in vitro drug resistance was found in our collection of T. evansi type A and type B stocks. The presence or partial absence of kDNA in the in vitro adapted T. evansi stocks did not correspond with the drug sensitivity phenotype.
|
10.1371/journal.pntd.0007292 | The landscape of enteric pathogen exposure of young children in public domains of low-income, urban Kenya: The influence of exposure pathway and spatial range of play on multi-pathogen exposure risks | Young children are infected by a diverse variety of enteric pathogens in low-income, high-burden countries. Little is known about which conditions pose the greatest risk for enteric pathogen exposure and infection. Young children frequently play in residential public areas around their household, including areas contaminated by human and animal feces, suggesting these exposures are particularly hazardous. The objective of this study was to examine how the dose of six types of common enteric pathogens, and the probability of exposure to one or multiple enteric pathogens for young children playing at public play areas in Kisumu, Kenya is influenced by the type and frequency of child play behaviors that result in ingestion of soil or surface water. Additionally, we examine how pathogen doses and multi-pathogen exposure are modified by spatial variability in the number of public areas children are exposed to in their neighborhood. A Bayesian framework was employed to obtain the posterior distribution of pathogen doses for a certain number of contacts. First, a multivariate mixed effects tobit model was used to obtain the posterior distribution of pathogen concentrations, and their interdependencies, in soil and surface water, based upon empirical data of enteric pathogen contamination in three neighborhoods of Kisumu. Then, exposure doses were estimated using behavioral contact parameters from previous studies and contrasted under different exposure conditions. Pathogen presence and concentration in soil varied widely across local (< 25 meter radius area) and neighborhood-level scales, but pathogens were correlated among distinct surface water samples collected near to each other. Multi-pathogen exposure of children at public play areas was common. Pathogen doses and the probability of multi-pathogen ingestion increased with: higher frequency of environmental contact, especially for surface water; larger volume of soil or water ingested; and with play at multiple sites in the neighborhood versus single site play. Child contact with surface water and soil at public play areas in their neighborhood is an important cause of exposure to enteric pathogens in Kisumu, and behavioral, environmental, and spatial conditions are determinants of exposure.
| The enteric infections that cause diarrheal disease remain a leading cause of morbidity and mortality in young children in low-income settings where sanitation infrastructure is limited. The role of different exposure pathways in transmission of enteric pathogens to young children is unknown, but accurate information on relative risks posed by different pathways could improve how well intervention strategies are prioritized. New evidence reveals that young children often play at public areas in their neighborhood, including areas used for human and animal feces disposal, potentially exposing them to heavily contaminated environmental fomites. We describe a novel exposure assessment approach that examines how the dose of six different enteric pathogens, and the probability of exposure to multiple pathogens is influenced by behavioral, environmental, and spatial conditions in public areas of three low-income, urban neighborhoods of Kisumu, Kenya. We found that increased frequency of contact with the environment and volume of material ingested, especially for surface water, and increased contact with different neighborhood locations elevates dose and probability of multiple pathogen exposure. Our study demonstrates that playing in public areas poses a risk to children and provides new insights on how pathogen exposures vary across transmission pathways.
| Children living in low income countries with poor sanitary conditions experience an average of 4 to 8 diarrheal episodes per year between birth and 2 years of age, [1] demonstrating that they are chronically exposed to enteric pathogens beginning in the first year of life. Furthermore, recent studies have highlighted wide diversity in the microbial etiology of early childhood (<5 years of age) enteric infection in such settings, suggesting that they are exposed to a variety of pathogenic organisms in the first years of life.[2–9] Little is known about the rate with which children are exposed to and acquire enteric infections over time. While diarrheal incidence may suggest an exposure rate of up to 4 to 8 pathogen types per year, many infections are asymptomatic and go undetected without extensive diagnostic profiling, [10] meaning that diarrhea symptoms are likely an underestimate of how often children acquire new infections. Diarrhea rates may even further underestimate how often children are exposed to pathogens, but remain uninfected due to insufficient exposure dose, lack of pathogen viability, host acquired immunity, or other mediating conditions. Adding to this complexity, co-infection of individual children by two or more types of pathogens–regardless of symptomology–is common.[3, 11] Co-infection, even in the absence of diarrhea, is associated with greater risk of environmental enteric dysfunction (EED), undernutrition, and re-infection by a new pathogen, perpetuating the cycle of disease.[4] Understanding which exposure pathways contribute most to multi-pathogen exposure of children could improve the prioritization of interventions that reduce early childhood enteric disease incidence.
Wagner and Lanoix’s “F-diagram” conceptualized the routes of fecal-oral disease transmission according to the properties of environmental materials (drinking water, food, soil, etc.) that can be contaminated by feces and ingested by humans.[12] There is limited research on how exposure varies across exposure pathways, particularly with respect to the rates at which children experience multi-pathogen exposure and infection. Existing comparisons of exposure pathways have relied on fecal indicator bacteria concentrations, [13–18] or pathogen-specific risks, [19–21] both of which have major methodological limitations in measuring the overall probability of enteric pathogen exposure. Different types of pathogens have been frequently detected in households of India and Tanzania and public play areas in Kenya, revealing that exposure to pathogens in private and public settings is likely.[22–24] Our group has further shown that soil and surface water from public areas where children play in Kisumu are often contaminated simultaneously by multiple types of pathogens, [24] revealing that children ingesting soil or water at some public sites could ingest doses of multiple types pathogens. Only one report to our knowledge has examined exposure from the perspective of ingesting multiple types of pathogens, rather than presence/absence of an indicator.[25] But, the modeling approach summed the individual probabilities of exposure to each type of pathogen from South African surface waters, rather than accounting for interrelatedness of pathogen contamination across sampling locations or exposure pathways. Since multi-pathogen contamination varies across location, exposure models must account for possible pathway- or location-specific differences in multi-pathogen contamination and transference.
The overall probability of exposure to enteric pathogens may be fundamentally different across exposure pathways and across location of exposure. For example, eating soil from the ground, especially in public areas, may be more hazardous than ingestion of household drinking water because humans or animals may defecate directly on the ground whereas drinking water is more likely to be protected and treated for safety.[13, 26, 27] Young children typically have high rates of contact with soil and objects, [28–32] and occasionally surface water, [33, 34] and frequently place their hands in their mouth with no handwashing in between.[34, 35] This results in frequent indirect ingestion of trace amounts of soil, and perhaps surface water. Geophagia (direct ingestion of handfuls of soil) among young children [26, 27, 34, 36] and drinking from surface water also occur, albeit less frequently than hand-to-mouth behaviors.[27, 34, 36, 37] The relative contributions of different behaviors, volume of material ingested, and type of material on cumulative pathogen doses (total number of pathogen types ingested per day) is unknown.
Finally, spatial variability of young children’s play in neighborhoods could influence the dose and diversity of pathogen exposure. Many children play outside the household unattended, while others have developmental- or guardian- driven limitations that restrict distance away from the household and acceptable areas for play. In settings where the landscape is often contaminated by feces from many humans and animals, the children who play in a constrained spatial area (i.e. near their household) may have a lower probability of pathogen exposure than children who roam across a larger spatial area and play at a variety of locations throughout the course of a day, especially if those locations are public areas used for feces disposal. More knowledge on how child behavior, type of environmental fomite, and spatial range of child play influences enteric pathogen exposure is needed for prioritizing interventions.
This novel exposure assessment study utilizes information on fine and macro-scale spatial variability in enteric pathogen detection and co-detection across public play areas in a typical low-income, fecal-contaminated setting to explore the relative importance of different environmental, behavioral, and spatial conditions in pathogen exposure of young children. The first objective of this study was to measure how increased frequency of child contact with soil or surface water and the volume ingested (indirect vs. direct ingestion) by children in public play areas influences the ingestion dose of enteric pathogens, and the probability of exposure to one or more enteric pathogens. Second, we compare how site-constrained (child plays at one public residential location) versus neighborhood (free roaming) spatial range of play for children influences pathogen dose and probability of multi-pathogen exposure. This modeling approach could be adapted to include a variety of setting-specific information on child behaviors and environmental conditions to better assess the relative contribution of various exposure pathways to child infection.
This exposure assessment study utilizes observational and environmental microbiology data on public sites in Kisumu, Kenya that has been described previously.[24] In brief, 166 total public sites in three peri-urban neighborhoods of Kisumu were randomly selected for a cross-sectional study on the role of human and animal sanitary conditions in neighborhood-level pathogen contamination. A “site” was defined as all public area (private households and businesses excluded) falling within a 25-meter radius of a randomly-generated set of central coordinates within each neighborhood. During rapid observation of sites (~10–15 minutes per site), our prior study revealed that at least one child <5 yrs was observed at 40% (66 of 166) of public sites, [24] with 94% (62 of 66 child observations) of these occurring in residential versus industrial, farming, or undeveloped areas. Furthermore, children were observed performing behaviors that resulted in hand or mouth contact with environmental fomites (touching soil, surface water, animals, or objects on the ground, swimming, eating food, eating dirt, mouthing hands). Soil and surface water samples analyzed by qRT-PCR were found to contain multiple types of common enteric pathogens.
This study uses microbial data reported in the parent study, but restricted to 116 residential public sites where children are most likely to spend time. In the parent study, three extra samples were collected at seven randomly-selected sites in each neighborhood to account for anticipated variance in pathogen distributions within the <25 m radius area of individual site. Resamples could have been as much as 50 meters apart, if on opposite ends of the 25 m radius site. A total of 125 soil samples and 34 surface water samples were collected from 116 residential public sites included in the study, including 15 sites where 45 soil and 3 sites where 6 surface waters were sampled.[24]
To ensure sufficient knowledge about the distribution of pathogen concentrations and to obtain numerical stability in the statistical estimation algorithms, only pathogens detected in greater than 5% of both soil and surface water samples from eligible sites were included in the model. Of the 19 pathogens tested during environmental sampling in the parent study, concentration data for 6 pathogens (S1 Table) were eligible for inclusion in the model: Cryptosporidium spp., Giardia lamblia, human adenovirus 40/41, Enteropathogenic E. coli (EPEC bfpA and/or eaeA), Enterotoxigenic E. coli (ETEC estA and/or eltB), and Enteroaggregative E. coli (EAEC aaicA and/or aatA).[24] Detection frequencies for less common pathogens at these 116 residential public sites are reported in S2 Table. If there was a positive detect for more than one bacterial gene marker, and concentrations (Cp) varied between gene markers, concentrations of ETEC-estA, EPEC-bfpA, and EAEC-aatA were prioritized over concentrations of ETEC-eltB, EPEC-eaeA, and EAEC-aaiC, respectively, based on etiological importance in pediatric diarrheal disease.[2, 10] Although the dataset was restricted to 6 of 19 pathogens measured, the number of soil and surface water samples with at least one positive detect (138/159 samples) did not change, indicating that these 6 pathogens collectively are sensitive indicators for the presence of pathogen contamination in the environment in this setting. Repeat detection of each pathogen in <25 m radius multi-sampled sites is reported in S3 Table.
The statistical analyses aimed to estimate the dose distributions of each pathogen type by environmental fomite type (soil vs. surface water), by contact type (indirect hand-to-mouth vs. direct geophagy or drinking), frequency of contact, and spatial range of exposure (site-restricted contact limited to <25 meter radius public environment vs. neighborhood-level contact with multiple randomly selected sites across the neighborhood). We also aimed to characterize fine scale within-site variability in pathogen contamination by sample type to understand how our sampling design might influence between-sample pathogen covariance. All analyses were conducted using R version 3.5.0. A Bayesian framework was employed to obtain the posterior distribution of pathogen doses for a certain number of contacts, denoted as D(k) for k contact events, i.e., the distribution of D(k) implied by information provided by both our data and previous studies.[15, 27, 32, 38–41] The posterior distribution yields point estimates and credible intervals for the parameters of the pathogen concentration distribution, denoted as θ and described in more detail below, for soil and water samples.
There are two parts to the modeling framework. The first part uses environmental microbiology data (S1 Table) to estimate the distribution of each pathogen concentration in soil and surface water.[24] The second part combines the concentration distribution of part one with contact fate parameters provided from previous studies (Table 1) to estimate the exposure pathway-specific dose distribution by fomite type, contact type, and behavior frequency. The posterior distribution of interest, namely that of D(k) and θ given the observed data and information from previous studies, can be decomposed to clearly reveal these two components of the statistical model:
Pr(D(k),θ|data,previousstudies)=Pr(D(k)|θ,previousstudies)⋅Pr(θ|data)
In implementation, we estimated this posterior distribution via a Monte Carlo approach on the joint posterior distribution augmented with the pathogen concentration corresponding to the k events. That is, we may first obtain a sample of θ from the marginal posterior distribution given the data, then draw k pathogen concentrations for the current value of θ, and finally, given those concentrations and information obtained from previous studies on exposure pathways, draw D(k).
Several challenges arose in estimating the parameters θ for the pathogen concentration distributions. First, there was left censoring caused by methodologically-constrained lower limits of detection (S1 Table). Second, there were two important sources of dependency in the data—that which occurs due to the correlations between the different pathogens, and that which occurs due to re-sampling within the 25-meter radius area of an individual site.
To handle data challenges, we fit a multivariate mixed effects (MVME) tobit model to the log transformed concentration data. The first source of dependency in the data was accounted for by modeling all pathogens jointly rather than running many univariate analyses. It was also important that we not neglect to account for the latter type of dependency described above, as the spatial patterns of young children playing in neighborhoods could influence the dose and diversity of pathogen exposure in public areas. Thus, the proposed random effects included in our MVME tobit model account for this spatial dependence. The parameters of the MVME tobit model θ can thus be broken into three components: (1) the mean of the log concentrations for each of the 6 pathogens; (2) the 6 × 6 covariance matrix for the residuals (Σ); (3) and the 6 × 6 covariance matrix of the random effects (Ω). Hence θ contains 48 parameters. The correlation for the pth pathogen between two samples at the same site; equivalently, the proportion of total variance explained by the site-level variance can be found via the intra-class correlation (ICC): Ωpp/(Ωpp + Σpp), and the total variance of the pth pathogen log concentration is Ωpp + Σpp. Of course, samples taken at two different sites are independent and is representative of our neighborhood-level play scenario. See Supporting Information S1 Text for details on this statistical model.
For each environmental sample type, samples of θ were obtained from the posterior distribution using a Gibbs sampler. From these samples, posterior draws of pathogen concentrations of k new events were drawn from a multivariate normal distribution parameterized by the draws of θ. See Supporting Information S1 Text for details on the Gibbs sampling algorithm.
A theoretical model was developed to estimate and compare the dose and diversity of enteric pathogens ingested by young children via indirect and direct exposure to soil and surface water at public play areas. The contact frequency was held at a constant rate, ranging from a minimum of 1 to a maximum of 10 contacts, for pattern comparison purposes, so behaviors in this model are not weighted to account for the likelihood of engaging in the behavior and the rate of contact given a child plays in a public area for a specified time span.[34] Therefore, the results are not cumulative estimates of actual child exposure, but represent possible exposures given a range of possible conditions. To obtain a posterior sample of the final dose for each set of conditions, the k concentrations drawn previously were multiplied by fate parameters, each drawn from a random distribution to account for the inherent variability in such occurrences, and then the k doses ranging from 1 to 10 were summed. The spatial assumption determined whether or not the k contacts were correlated. The formulas used to estimate the dose distribution from indirect and direct contact with soil (1) and surface water (2) are:
The pth pathogen is denoted by a subscript of p and their concentrations in soil and surface water are denoted as CSp and CWp, respectively. * means truncated with a lower bound = 0 and upper bound = 1, † means the surface area of hand-to-mouth contact cannot exceed the surface area that was contaminated during hand-to-object contact, thus the fraction is truncated at 1.
Fate parameters obtained from the extant literature to estimate exposure to pathogens through indirect contact include: the transfer efficiency of the environmental fomite to the hand (soil: TES, water: TEW), the total surface area of the child’s hand (SAHi), the fraction of the child’s hand contacting the environmental object (FHO), the fraction of the hand mouthed by the child (FHM), and the transfer efficiency of environmental residual from hand-to-mouth (TEHM) (Table 1). Total hand surface area (cm2) used in this model was based on estimated surface area parameters for children between the ages of six months to less than six years.[32] Standard deviation for total hand surface area (cm2) per age category (6 to 11 months, 12 to 23 months, and 24 to 72 months) was calculated by dividing the difference of the EPA-reported mean and 95th percentile for hand surface area by the 95th quantile of a standard normal distribution (1.645).[32] Each age category was equally represented during simulation by sampling the probability of obtaining a random child within each of the unequal month spans and respective hand surface area mean and standard deviation (SAHi). The distribution for the fraction of the child’s hand involved in hand-to-object contact (FHO) and hand-to-mouth contact (FHM) was calculated by minimizing the squared differences between theoretical and empirical quantiles.[40] Transfer efficiency of environmental residue from hand-to-mouth (TEHM) was estimated with a single point estimate due to the lack of literature to infer a distribution for all pathogens used in this analysis.[41] Our exposure model assumes that the hand region that contacted the object is the same region that contacted the mouth. This assumption is supported by the finding that regardless of the type of interaction, hand contact predominantly involves the fingers.[40] When summing across estimated indirect doses, hand size was held constant, while soil adherence and hand area that contacted the object and then mouth varied between k contact events. To estimate direct ingestion of soil or surface water, the volume of respective substance placed in the mouth during a geophagia (VS) or drinking occurrence (VW) was estimated with a single parameter because of the lack of literature to describe the distribution of direct ingestion occurrences (Table 1).[15, 27, 33]
We ran 25,000 iterations of the Gibbs sampler thinning by keeping every 5th draw. Of these 5,000 draws, 1,500 were used as a burn in period leaving 3,500 samples for our Monte Carlo analyses. The estimated pathogen concentration (C) log means and back-transformed median concentrations (exponentiated log mean) are reported in Table 2. The log standard deviation (SD) size relative to the log mean was large for pathogens in soil and surface water, except for the frequently detected Cryptosporidium spp. Intraclass correlation (ICC), indicating the correlation between pathogens in different samples at the same site, varied by pathogen and type of environmental material. Moderate-to-high ICCs in surface water indicated low within-site variability in concentration for all pathogens. Low-to-moderate ICC in soil indicated pathogen concentrations varied more within site.
Estimated pathogen concentration (C) distributions (Table 2) were all positively correlated in surface water, but many were not positively correlated in soil (Fig 1). The 95% credible intervals (CI) around the correlation means (S1 Fig) revealed that within a single sample of surface water, twelve of fifteen pathogen comparisons were significantly positively correlated. When comparing different water samples taken at the same site, six pathogen comparisons in surface water were significantly positively correlated: adenovirus 40/41 and EPEC, adenovirus 40/41 and EAEC, ETEC and EPEC, Giardia and EPEC, Giardia and EAEC, and EPEC and EAEC. Many other pathogen comparisons trended towards statistical significance. Within a single sample of soil, we observed a significant negative correlation between Cryptosporidium spp. and adenovirus 40/41 and positive correlations for ETEC and adenovirus, EAEC and adenovirus 40/41, and EAEC and ETEC. However, significant pathogen correlations were not observed between different samples from the same site, reflecting high variability in concentrations within a 25m radius area.
When comparing similar pathway, number of contacts, and spatial scale, adenovirus, EPEC, and EAEC pathogen doses from soil contact (Fig 2) were always lower than doses from surface water (Fig 3), and Cryptosporidium doses were consistently higher from soil contact than surface water (see S2–S5 Figs for pathogen-specific graphs). As the rate of hand-to-mouth contact or direct ingestion increased for site-level play, the dose of Giardia and ETEC from soil exceeded that of surface water (S2 and S4 Figs). All contact with surface water resulted in ingestion of DNA representing at least one pathogen organism of any type, with the exception of Giardia, for one water-hand-mouth contact (Fig 3). If frequency of contact with soil or surface water is held constant, geophagia or drinking surface water always resulted in higher pathogen doses compared to soil/water-hand-mouth contact (Fig 2C/2D vs. 2A/2B; Fig 3C/3D vs. 3A/3B). However, if hand-mouth contact occurs more often than geophagy or drinking surface water, then doses resulting from hand-mouth contact could exceed exposures from direct ingestion. For example, if a child exhibited ten cumulative soil-hand-mouth contacts and one geophagia contact during play at one site, the EAEC dose for soil-hand-mouth (~53 bacteria, Fig 2A, solid box) would exceed the EAEC dose for geophagia (~0.4 bacteria, Fig 2C, dashed box). Overall, when frequency of contact with soil or surface water is held constant, pathogen doses were always greater when children played at multiple sites in the neighborhood, versus just one site (Fig 2B/2D vs. 2A/2C; Fig 3B/3D vs. 3A/3C), but the magnitude of change depended upon the pathogen type (visualized side-by-side in S6–S11 Figs).
Pathogen-specific dose distributions from indirect and direct contact with soil and surface water at site- and neighborhood-level are reported in S6–S11 Figs. The log mean and untransformed median dose D(k) of pathogens ingested during contact with soil were largest for Cryptosporidium spp. (Fig 2). However, Cryptosporidium spp. dose did not considerably increase with increased behavior frequency or increased spatial scale of play from site- to neighborhood-level (S6 Fig). The lowest pathogen dose ingested during soil contact was EPEC—yet the dose of EPEC substantially increased for neighborhood vs. site-level play (S10 Fig). For example, it required over two times more soil-hand-mouth contacts to ingest the same dose of EPEC at site-level play (~7 contacts, Fig 2A, solid circle) as neighborhood-level play (~3 contacts, Fig 2B, dashed circle). Noticeably, the dose of human adenovirus 40/41 from surface water contact exponentially increased as spatial scale expanded from site to neighborhood play and surpassed all other pathogen doses at neighborhood-level exposure.
Pathogen dose and multi-pathogen exposure could differ with age due to differences in frequency of child behaviors and child hand size. In this analysis, behavior frequency was treated either as an experimental factor influencing dose or a constant for comparing doses between soil versus water exposure pathways or site versus neighborhood exposure pathways. However, child hand size was a parameter in our model that increases with age and could influence soil-hand-mouth or water-hand-mouth pathogen transmission. A sensitivity analysis for age group (S12–S15 Figs and S4–S7 Tables) revealed only small differences in pathogen doses for children 6 to <12 month, 12 to <24 month, and 24 to <72 months of age, although older children who had slightly greater doses because of larger hand size.
Fig 4 illustrates the probability of ingesting one or more pathogens for 1 to 10 indirect or direct contact(s) with soil or surface water during site-restricted or neighborhood play, where a successful ingestion is defined as DNA of 1 or more pathogen types. Across all behaviors, the probability of ingesting more than one pathogen type intensified as spatial scale expanded from site to neighborhood play. For example, the probability of ingesting six pathogens from two water-hand-mouth contacts during site-restricted play (43%, Fig 4E, solid box) increased by about 15% if the child exhibited the same behavior and frequency during neighborhood-level play (58%, Fig 4F, dashed box). Soil-hand-mouth contact resulted in the lowest probability of ingesting diverse pathogen types compared to all other behaviors practiced at the same frequency. This is especially evident for soil-hand-mouth contact during site-restricted play where the probability of ingesting all 6 pathogens did not exceed 35% for 10 contacts. Any contact with surface water posed a high probability for ingestion of diverse pathogens and is demonstrated by a > 90% probability of ingesting 6 pathogens during ≥5 water-hand-mouth contacts and ≥3 drinking water contacts during neighborhood-level play.
Our prior work has demonstrated that children < 5 yrs living in low-income neighborhoods of Kisumu, Kenya are exposed during play in public residential areas to soil and surface water contaminated by human and animal feces and a diverse range of enteric pathogens.[24] This study addresses the next question as to how children are impacted by playing in these settings. Specifically, we quantified the impact of different child exposure behaviors, environmental transmission pathways, and spatial situations on the dose and diversity of enteric pathogens ingested by young children as a result of contact with public residential areas. When holding frequency of behaviors constant, the dose and probability of multiple enteric pathogen exposure were typically greater when a child ingested: (1) surface water versus soil, (2) greater volumes of soil (geophagy) or surface water (drinking small mouthfuls) versus soil/water-hand-mouth contact, and (3) soil or surface water from multiple neighborhood locations versus just one spatially-restricted site. Evidence that children have an increased probability of simultaneous exposure to multiple enteric pathogens during certain play conditions indicates that exposure pathways may be more important than others in elevating the risk of infection or even co-infection by different pathogens.
Our studies in Kenya and Haiti have confirmed that children engage in the exposure behaviors chosen for this exposure assessment (hand contact with soil and surface water, hand-to-mouth contact, and geophagy) in public areas and is consistent with extant literature in domestic settings. [26–28, 33, 34, 36, 42] In our studies of child play in public areas in Haiti, geophagy was 6 times more common than drinking surface water (0.9/hr vs 0.15/hr), and hand-to-mouth was roughly 9 (child) to 20 (infant) times more common than geophagy.[34] If we assume that child behavior frequencies are generalizable across geographic contexts, then child exposure to low doses (100−102) of at least one type of pathogen from short durations of play outside the home in Kisumu is certain, due to the pervasiveness of pathogen contamination in these neighborhoods and frequent hand-to-soil and hand-to-mouth contacts. Multi-pathogen exposure was common for even the lowest risk behaviors; for example, the probability for exposure to 2 or more pathogens for one soil-hand-mouth contact during site-level play was 71%, although exposure to all pathogens was unlikely at even 10 contacts (< 35%). Contact with surface water was more dangerous, with only a few contacts resulting in exposure to all six pathogens. In reality the rate of different child behaviors likely vary between children and across populations of children, so the ranking of actual exposure risks associated with different behavioral, environmental, and spatial conditions should be evaluated based upon specific cultural settings. In the absence of quantitative data on child behavior in Kisumu public areas, we chose to keep contact frequency of behaviors at a constant rate to examine relative relationships between dose and diversity. Nonetheless, this data shows the threshold of safe contact between children and public domains in Kisumu for exposure to pathogens is low.
Another notable discovery was that exposure of the child to multiple public locations in their neighborhood, such as with free roaming child play, significantly increased pathogen doses and probability of multi-pathogen ingestion. Consideration of space as a determinant of environmental exposure to fecally-transmitted pathogens is rarely explicitly included as a function of exposure modeling, and when it is, spatial units tend to be defined using households or clusters of households as units of study. Our prior studies suggest that young children in urban, low-income settings–even those less than 24 months–are not immobile elements whose environmental exposures can be defined by these architectural boundaries.[34] While this study provides theoretical insight on potential child exposure conditions if children’s range of movement is considered the unit of study, very little is actually known about how much time children in low income settings spend in public areas and how far they roam. Even within a <25 m radius area, such as area just outside a child’s compound, our evidence of extreme covariance in pathogen concentrations and pathogen diversity in soil highlights the need for caution in estimating exposure risks from soil-hand-to-mouth and geophagy pathways with small sample sizes and with single organism microbial indicators. Accounting for microbial covariance in the environment during study design and modeling could improve reporting on model uncertainty and prevent conclusions from being biased by inadequate study design.
There were some pathogen-specific differences in exposure dose concentration curves. Cryptosporidium spp. was the most common pathogen in soil, at relatively stable concentrations across the neighborhood, which led to higher exposure doses. In this study we did not determine how many of these samples were C. parvum or C. hominus types, which are typically considered responsible for human infections. Although the public health importance of these exposures remains unclear, the 18S subunit gene indicator used in this study to detect Cryptosporidium was the same indicator used in a study of diarrhea in children less than two years of age in Bangladesh, which found C. meleagridis species are also common causes of child infection.[43] Pathogenic E. coli dose curves generally behaved similarly to each other, although soil appears to be a less common transmission pathway for EPEC. Human adenovirus and Giardia concentrations were highly varied between sites, compared to other pathogens, which indicates that child exposure to multiple public sites was an important determinant of greater doses.
This study has several important limitations. Concentrations of enteric viruses, bacteria, and protozoan pathogens were estimated by quantitative reverse transcription Polymerase Chain Reaction (qRT-PCR). While use of a multi-pathogen qRT-PCR process reduced methodological sources of variability in concentration estimates, and ergo exposure doses, it does not distinguish between viable (culturable), non-viable (live but non-culturable), and dead organisms. Pathogen exposure doses presented in this study may overestimate the number of infectious pathogens children ingest through contact with soil and surface water. There is a lack of information on how well qRT-PCR correlates with other approaches for quantifying pathogen environmental exposure, and with child disease outcomes in settings like Kisumu. PCR detection of microbial source tracking (MST) markers in Indian households was associated with diarrhea symptoms in one study.[22] However, MST markers may occur in the environment more frequently than infectious pathogens, and it is unclear whether MST presence is a reliable proxy for pathogen dose. Our model assumes pathogen concentrations follow a lognormal distribution; further the joint posterior distribution was largely reliant on the Tobit formulation and LLODs used to estimate the distribution of a high proportion of truncated values (non-detects). These modeling assumptions are difficult to verify due to the high proportion of non-detects. Because it is unlikely that very low doses result in meaningful exposures, we excluded doses below 1 DNA when estimating the probability of exposure to multiple pathogens. An additional limitation is our reliance on contact fate parameters from the extant literature. An important facet for future research would be to test the sensitivity of our conclusions to variations in these fate parameters.
Although validating health outcomes with exposure doses in Kisumu children was outside the scope of this study, our findings of pervasive environmental contamination and exposure to high doses of diverse pathogens are analogous to the county’s high prevalence of diarrhea in children under the age of five (18% of children experiencing diarrhea in a two-week prevalence).[44] Additionally, each of the pathogens common in soils and surface waters of Kisumu have been reported as leading causes of enteric infection in children in Kenya and elsewhere.[3] Further efforts should consider the importance of public exposure pathways in pediatric enteric infections. More information on dose-response in children in low income settings that accounts for potential interdependencies in the presence of pathogens would advance our understanding of exposure and infection outcomes.
The findings of probable pathogen exposure among young children who play in public areas provide one plausible explanation for why recent large scale trials of household WASH interventions have reported little [45] or no impact on child diarrhea [46] and why the global diarrhea disease burden remains high despite marked improvements in basic water and sanitation access in recent decades.[47] Child play outside the household is far more common than appreciated [34] and is likely to continue as long as families live in crowded conditions. Children playing unaccompanied or being cared for by older siblings may contribute to hazardous scenarios where children play in neighborhood areas and contact objects that are unsafe or unsanitary. Household WASH interventions do not typically address the topic of where children play, and do not install barriers between children and soil and water contaminated by the feces of one’s neighbors and domestic animals. Thus, while household WASH improvements can reduce child exposure and infection from some pathways, they may not sufficiently reduce exposure across all pathways to observe differences in diarrhea rates. To our knowledge, no household WASH trials have explicitly assessed whether children in intervention households have contact with public areas outside their compound. Thus, it remains unclear whether public exposures diminish the impact of improved household WASH conditions. Most of the recent large trials were conducted in rural areas where children’s probability of exposure to human feces in the neighborhood may be lower than for children in crowded urban Kisumu. [46] Yet, while human sanitation issues may vary across rural and urban settings, domestic animal feces is a universal environmental contamination issue across rural and urban settings.
In conclusion, our results suggest that exposure to public residential areas poses an important risk for enteric pathogen ingestion for children living in low-income settings with poor sanitary conditions, especially when those behaviors are frequent and of high volume, involve contact with surface water, and occur at multiple locations in the child’s neighborhood. Addressing these exposures will require broader WASH interventions targeting both child play behaviors and the environmental conditions in which play occurs. Future research should address the scarcity of information about spatial patterns of child behavior and the importance of public exposures in child infection outcomes.
|
10.1371/journal.pgen.1007959 | Spatio-temporal regulation of nuclear division by Aurora B kinase Ipl1 in Cryptococcus neoformans | The nuclear division takes place in the daughter cell in the basidiomycetous budding yeast Cryptococcus neoformans. Unclustered kinetochores gradually cluster and the nucleus moves to the daughter bud as cells enter mitosis. Here, we show that the evolutionarily conserved Aurora B kinase Ipl1 localizes to the nucleus upon the breakdown of the nuclear envelope during mitosis in C. neoformans. Ipl1 is shown to be required for timely breakdown of the nuclear envelope as well. Ipl1 is essential for viability and regulates structural integrity of microtubules. The compromised stability of cytoplasmic microtubules upon Ipl1 depletion results in a significant delay in kinetochore clustering and nuclear migration. By generating an in silico model of mitosis, we previously proposed that cytoplasmic microtubules and cortical dyneins promote atypical nuclear division in C. neoformans. Improving the previous in silico model by introducing additional parameters, here we predict that an effective cortical bias generated by cytosolic Bim1 and dynein regulates dynamics of kinetochore clustering and nuclear migration. Indeed, in vivo alterations of Bim1 or dynein cellular levels delay nuclear migration. Results from in silico model and localization dynamics by live cell imaging suggests that Ipl1 spatio-temporally influences Bim1 or/and dynein activity along with microtubule stability to ensure timely onset of nuclear division. Together, we propose that the timely breakdown of the nuclear envelope by Ipl1 allows its own nuclear entry that helps in spatio-temporal regulation of nuclear division during semi-open mitosis in C. neoformans.
| Unlike the model ascomycetous budding yeast Saccharomyces cerevisiae, microtubule organizing centers (MTOCs) coalesce to form the spindle pole body (SPB) in C. neoformans. This process also ensures unclustered kinetochores to gradually cluster in this organism. As C. neoformans cells enter mitosis, the nuclear envelope ruptures and the nucleus eventually moves to the daughter bud before division. Here, we combine cell and systems biology techniques to understand the key determinants of nuclear division in C. neoformans. We show that the evolutionarily conserved Aurora B kinase Ipl1 enters the nucleus during the mitotic phase as cells undergo semi-open mitosis. Ipl1 regulates dynamics of cytoplasmic microtubules, cytosolic proteins such as Bim1 and dynein-mediated cortical forces and integrity of the nuclear envelope to ensure timely kinetochore clustering and nuclear division in this medically relevant human pathogenic budding yeast.
| High-fidelity chromosome segregation ensures faithful transmission of the genetic material to subsequent generations. This process is powered by the dynamic interactions of the centromere-kinetochore complex and the mitotic spindle. The microtubule organizing centers (MTOCs), microtubule fibers, microtubule-associated proteins (MAPs) and molecular motors belonging to kinesin and dynein superfamilies influence microtubule (MT) dynamics and functioning of the mitotic spindle. Among the three types of MTs emanating from MTOCs, the cytoplasmic MTs (cMTs) ensure proper spindle positioning and nuclear migration [1, 2]. Several MAPs and MT-based motor proteins participate in organizing the MT cytoskeleton by regulating polymerization-depolymerization kinetics, dynamics of cross-linking, and motility of MT fibers [3, 4]. The plus-end tracking proteins (+TIPs), a subgroup of MAPs is characterized by their preferential association with the MT plus-ends. Among the plethora of +TIPs, Bim1, the EB1 homolog in yeast, localizes at the distal ends of cMTs, establishes contacts with the cortical polarity determinant Kar9 and guides the plus-ends of cMTs along the cortical actin cables for penetration of MTs into the budding daughter cell [5, 6]. In contrast to Kar9-mediated interactions, dynein molecules are transported from the less dynamic minus-end towards the plus-end of MTs, empowering spindle movement through the mother-daughter bud neck [7, 8]. Perhaps, dynein molecules undergo spatio-temporal regulation during spindle movement by regulated targeting of She1 to cMTs [9]. In addition, cytoplasmic dynein molecules in yeast can also potentially crosslink two MT fibers physically and slide them along each other [10].
An interplay between kinases and phosphatases spatio-temporally regulate kinetochore-MT interactions (reviewed in [11]). The evolutionarily conserved Ipl1/Aurora B kinase, the best-studied regulator of the kinetochore-MT attachments, primarily senses the tension at the kinetochore [12, 13]. Aurora B kinase Ipl1 stabilizes bi-oriented kinetochore-MT attachments and destabilizes incorrect kinetochore-MT attachments [14, 15]. In addition, Ipl1 ensures the integrity and timely disassembly of the mitotic spindle by phosphorylating She1 and Bim1 respectively in Saccharomyces cerevisiae [16–18]. Ipl1 regulates dynein activity along the cMTs by phosphorylating She1 and influences movement of the pre-anaphase spindle into the mother-daughter bud neck [8].
Unlike hemiascomycetous budding yeasts such as S. cerevisiae, the process of chromosome segregation is less known in the phylum Basidiomycota [19, 20] that shared a common ancestor with Ascomycota more than 500 million years ago [21]. The best-studied basidiomycete is the human pathogen Cryptococcus neoformans, that causes life-threatening pneumonia and meningitis, and is the 5th most leading cause of death in immunocompromised patients [22]. Azole-associated acquisition of aneuploidy is a well-elucidated mechanism that contributes to the emergence of transient azole resistance in C. neoformans [23, 24]. Clones that emerged at the highest drug concentration tested were found to be disomic for multiple chromosomes [24]. Thus, aneuploidy provides an increased fitness to C. neoformans under the azole stress [25].
Although C. neoformans divides by budding, a number of striking variations are observed in the dynamics of MTOCs, the site of nuclear division and the timing of kinetochore clustering as compared to the ascomycetes such as S. cerevisiae and C. albicans. In ascomycetous budding yeast species, a single MTOC, also known as the spindle pole body (SPB), is embedded in the nuclear envelope (NE) [26]. Once the SPB is duplicated during the G1/S transition phase, the kinetochore-MT interaction is established although the NE never breaks down, resulting in a closed mitosis [27, 28]. In contrast, C. neoformans cells have several MTOCs present throughout the cytoplasm during interphase and undergo semi-open mitosis characterized by transient rupture of the NE during metaphase to anaphase transition [20, 29]. In ascomycetes, the nucleus migrates close to the mother-daughter cell junction and divides into two equal halves [19, 30], while in C. neoformans, the nucleus first migrates completely to the daughter cell and divides [19, 20]. The kinetochores are clustered and tethered to SPBs by MTs during most of the cell cycle in ascomycetous yeast species [28, 31]. In contrast, the unclustered kinetochores gradually coalesce into a single cluster with the help of MTs as cells progress towards mitosis in C. neoformans [20]. We previously demonstrated that these fundamental variations in the process of nuclear division in these two fungal phyla are determined by the populations of cMTs and cortical dyneins [19]. Here, we combined cell biology studies and computational simulations to understand the molecular basis of unconventional nuclear division in C. neoformans. We used the ploidy sensor Ipl1 as a tool to study this process. Overall, our results uncover a distinct mechanism leading to atypical mitosis and the role of Ipl1 in this process to maintain ploidy in a basidiomycete C. neoformans.
The ORF CNAG_01285 is the homolog of Ipl1 in C. neoformans in the FungiDB (http://fungidb.org/fungidb/) having an evolutionarily conserved kinase domain (S1A and S1B Fig). To study the localization of Ipl1, we functionally expressed it as a fusion protein with mCherry at its N-terminus under the GAL7 promoter in the strain CNNV114 co-expressing GFP-tagged histone H4. Strikingly, overexpressed Ipl1 shows a distinct localization to the cytosol throughout the cell cycle. However, Ipl1 is also nuclear localized only during mitosis (Fig 1A). Ipl1 colocalizes with GFP-tagged histone H4 from the time of migration of nucleus to the daughter bud till the nucleus is divided into two equal halves. We further validated the localization of Ipl1 when expressed at the cellular level by functionally expressing it as a fusion protein with a triple GFP epitope at its C-terminus under the native promoter in the strain CNNV113. A reduced Ipl1 localizes to the nucleus only during specific stages of mitosis. The cytosolic signal is barely visible possibly due to low and dispersed signal intensities spread across the cytoplasm (Fig 1B). In fact, Ipl1’s localization in the nucleus at certain stages of the cell cycle also remained undetected when expressed under the native promoter. While nuclear localization of Ipl1 in mitosis is evolutionary conserved, its cytosolic localization remained speculative in S. cerevisiae [32] and unknown in other yeast species.
The NE transiently ruptures during mitosis in C. neoformans [20]. To test whether rupturing of the NE during mitosis helps in nuclear entry of Ipl1, we tagged nuclear localized proliferating cell nuclear antigen (PCNA), an S-phase specific nuclear protein, with GFP in CNNV112 cells expressing mCherry-Ipl1. Strikingly, Ipl1 shows a distinct cytosolic localization throughout the cell cycle (S2A Fig). However, during disassembly of the NE in mitosis marked by the diffusion of nuclear PCNA to the cytosol, Ipl1 is also found to be nuclear localized, suggesting rupturing of the NE during mitosis helps in the timely nuclear entry of Ipl1. Further, cytosolic and nuclear localization dynamics of Ipl1 was examined by performing time-lapse microscopy with CNNV112 cells grown in permissive conditions (Fig 1C, S1 movie). At the time of nuclear migration to the daughter bud, Ipl1 gains access to the nucleus due to the NE rupture and thus localized both in the cytosol and as well as in the nucleus. This suggests that the NE rupture event facilitates the nuclear Ipl1’s tension sensing function during kinetochore-MT attachments. In addition, cytosolic localization of Ipl1 throughout the cell cycle is possibly important for migration of the nucleus from the mother to the daughter bud. Incidentally, depletion of Ipl1 resulted in a premature rupture of the NE and leakage of PCNA into the cytosol in 18% of budded population indicating that Ipl1 also regulates the timely disassembly of the NE to facilitate its own entry into the nucleus during mitosis (Figs S2B and 1D).
To study the essentiality of IPL1 for cell viability, we constructed the conditional promoter shut-down mutant strain by placing the ORF under the regulatable GAL7 promoter which is repressed in the presence of glucose (non-permissive) but expressed when galactose (permissive) is used as the carbon source [33]. The inability of two independent conditional ipl1 mutants, CNNV101 and CNNV102, to grow under non-permissive conditions confirmed that Ipl1 is essential for viability in C. neoformans (Fig 2A). Mutants of ipl1 were first identified in the increase-in-ploidy screen of mutants in S. cerevisiae, an ascomycete [34]. To test this function of Ipl1 in C. neoformans, a basidiomycete yeast, we performed imaging of CNNV114 cells co-expressing Ipl1-mCherry and histone H4-GFP. Although Ipl1-mCherry signals could not be detected after growth of cells in non-permissive conditions for 4 h (S3A Fig), Ipl1-depleted cells exhibited the most severe phenotype of accumulation of cells having unsegregated nuclei (56% of the population) after 8 h of growth under non-permissive conditions (Fig 2B). Therefore, we performed subsequent experiments by growing the conditional ipl1 mutant in non-permissive conditions for 8 h. To test the occurrence of unsegregated nuclei in the Ipl1-depleted cells, we imaged nuclear migration in the live wild-type CNVY108 and mutant CNNV104 cells (Fig 2C) that suggested while a fraction of Ipl1-depleted cells had the nucleus remained unsegregated in the mother cell or present at the mother-daughter cell junction, a proportion of the mutant cells had nuclei segregated with a delay in the nuclear migration (mean = 47 min) as compared to the wild-type cells (mean = 35 min) (Fig 2D and 2E; S2 Movie). Taken together, we conclude that Ipl1 is required for timely nuclear migration during cell division in C. neoformans.
Nuclear migration from the mother cell to the daughter is governed by cMTs in yeasts [2, 19, 35]. Therefore, we analyzed the length of the cMTs in Ipl1-depleted cells. The length of the cMTs was severely affected in the mutant CNNV105 cells with an average length of the cMTs of 1.24 μm as compared to 2.7 μm found in the wild-type CNVY107 cells (Fig 3A and 3B). More than 80% budded Ipl1-depleted cells possessed shorter MTs as opposed to the majority of the cells having intact MTs in the wild-type (Fig 3C), although the total cellular pool of tubulin remains unaltered even after depletion of Ipl1 (S3B Fig).
Co-localization of MTs with kinetochores as well as aberrant kinetochore clustering in response to nocodazole suggested that MTs are indispensable for kinetochore clustering in C. neoformans [20]. We investigated kinetochore clustering in Ipl1-depleted cells having compromised integrity of MTs by localizing mCherry-CENP-A signals (inner kinetochore) in the wild-type, CNVY107 and Ipl1-depleted strain CNNV105 (Fig 3D). Approximately equal proportion of Ipl1-depleted cells exhibited four distinct phenotypes (a) unclustered kinetochores in the mother cell with no signal in the daughter cell (ratio of the daughter/mother cell size < 0.4) (19%), (b) unclustered kinetochores in the mother cell with no signal in the daughter cell (ratio of the daughter/mother cell size > 0.5) or unclustered kinetochore at the neck of mother-daughter cell (24%), (c) clustered kinetochores that were stuck at the neck of the mother-daughter cell junction (26%), and (d) unsegregated or aberrantly segregated kinetochores (31%) between the mother and daughter cells (Fig 3D). While the kinetochores were found to be unclustered in the wild-type cells with an average ratio of the daughter /mother cell size of 0.4, kinetochores remained unclustered in Ipl1-depleted cells at a much higher average ratio of 0.75 (S3C Fig), indicating a delay in kinetochore clustering in the Ipl1-depleted cells. Thus, we quantified the time required for kinetochore clustering in the wild-type, CNVY107 and the Ipl1-depleted mutant strain CNNV105 (Fig 3E). Strikingly, the Ipl1-depleted cells displayed a 10 min delay in time (35 min) required for kinetochore clustering as compared to the wild-type (25 min). It is noteworthy that not all the cells with disintegrated MTs harbor defective kinetochore clusters in the mutant (S3D Fig). Since an outer kinetochore protein Dam1 was shown to have a compromised localization in the absence of the mitotic spindle in S. cerevisiae [36, 37], we examined the localization of two outer kinetochore proteins, Mtw1 and Dad1, in CNNV108 and CNNV109 respectively. Localization of both the proteins was apparently unaltered in the Ipl1-depleted cells (S3E Fig). Next, we compared the delayed kinetochore clustering phenotype observed in the Ipl1-depleted cells with the wild-type cells treated with the MT-depolymerizing drug such as nocodazole. While the kinetochores were found to be unclustered in the wild-type cells with an average ratio of the daughter /mother cell size of 0.5, kinetochores remained unclustered in nocodazole-treated cells at a higher average daughter /mother cell size of 0.86 (S4A and S4B Fig), similar to the Ipl1-depleted cells. Together, these results confirmed that without affecting the kinetochore localization, depletion of Ipl1 affects the MT stability resulting in an alteration in spatio-temporal regulation of kinetochore clustering in C. neoformans. Thus, we focussed our subsequent studies on how Ipl1 might be regulating kinetochore clustering and nuclear migration in C. neoformans.
MTOCs conglomerate into a compact mass to form the fully functional mother SPB [38, 39], which simultaneously facilitates the clustering of MTOC-bound kinetochores in C. neoformans [40]. To elucidate the Ipl1-mediated process of kinetochore clustering, we studied the dynamics of MTOCs by localizing Spc98, an MTOC component, in the strain CNNV118 (Fig 4). In unbudded interphase cells, MTOCs are unclustered and puncta of Spc98-3xGFP signals are scattered (Fig 4A, left panels). As the cell cycle progresses, Spc98-3xGFP signals gradually cluster to form a mature SPB. Subsequently, signals representing clustered MTOCs segregate into two halves during mitosis, one of which is retained in the daughter bud and the other is delivered to the mother cell (Fig 4A, left panels). Next, we investigated MTOC clustering in Ipl1-depleted CNNV118 cells by localizing Spc98-3xGFP upon their growth in non-permissive conditions for 8 h (Fig 4A, right panels). The Ipl1-depleted cells exhibited four distinct phenotypes (a) unclustered MTOCs in the mother cell with no signal in the daughter cell (ratio of the daughter/mother cell size < 0.4) (12%), (b) unclustered MTOCs in the mother cell with no signal in the daughter cell (ratio of the daughter/mother cell size > 0.5) or unclustered MTOCs scattered at the neck of mother-daughter cell (40%), (c) clustered MTOCs or clustered segregated SPBs present either in the mother or daughter cell (12%), and (d) unsegregated or aberrantly segregated SPBs (34%) between the mother and daughter cells (Fig 4B). While the MTOCs were found to be unclustered in the wild-type cells with an average ratio of the daughter /mother cell size of 0.4–0.5, MTOCs remained unclustered in Ipl1-depleted cells at an average ratio of >0.5, indicating a delay in MTOC clustering in the Ipl1-depleted cells. Further, using an in silico model, we canvassed various plausible MT-driven MTOC clustering schemes (Fig 5A) characterized by parameters (Table 1) [41, 42]. We assumed that a single kinetochore is associated with a single MTOC prior to the clustering. The cortical interaction of the cMTs mediated by dyneins and other motor proteins generates a directed mechanical force on various intracellular components including MTOCs [43–45]. As an additional component to cortical interaction, we introduced a form of bias on the cortically interacting cMT plus ends towards the septin ring that is triggered by Bim1 yielding an enhancement in the force transmission to the MTOCs and leading to a complete fusion (~98%) of MTOCs within 23 min (Fig 5B). Migration of the nucleus from the mother to the daughter cell is partially concomitant with the process of MTOC clustering during mitosis in C. neoformans [20]. Henceforth, it is essential to validate whether the proposed clustering scheme adequately corroborates with the timely nuclear migration [6, 43]. We found that the recruitment of cortical interaction with bias at the onset of the MTOC clustering and sustaining it all through takes ~38 min for proper nuclear migration (Fig 5C), close to the experimental time scale of ~35 min. Hence, we employed the cortical interaction with bias scheme to explore the attributes of the mitotic cell cycle in the wild-type (S3A Movie).
The cMTs are required to be long enough to establish contacts with the cell cortex to render the directional preference of cMTs via cortical interaction with bias [58, 59]. A combined action of Bim1 and dynein has been shown to be essential to impart an ‘effective’ cortical bias on the cMT plus-ends penetrating into the cortex towards the septin ring [19, 60]. It has been reported previously that C. neoformans lacks a Kar9 homolog (Table 2) [61], we probed for the function of Bim1 and dynein in generating an effective cortical bias at the plus-ends of the cMTs in C. neoformans. We assumed a uniform distribution of dynein in the mother and daughter bud cortices, while Bim1 was distributed only in the mother cell. Our simulation connotes that reduced Bim1-mediated bias on the tips of the growing MTs is not favorable for nuclear migration (Fig 5D). Our in silico study also predicts a differential activity of dynein in the mother and daughter cell cortex. While a strong dynein-pull on the cMTs from the mother cortex is not favorable for nuclear migration, an optimized dynein-pull from the daughter cortex transmitted via cMTs is essential for nuclear migration (Fig 5E). Together, our in silico model characterized that an effective cortical bias mediated by Bim1 and dynein in the mother cortex towards the daughter cell is essential for nuclear migration.
Further, to validate the model’s prediction of Bim1’s function in imparting cortical bias on the tips of the cMTs penetrating the daughter buds, we studied the localization of Bim1 at different stages of the cell cycle in the wild-type cells. We functionally expressed Bim1-3xGFP under the native promoter in the strain CNNV119. Localization of Bim1 at the mother bud from the time of its emergence of the daughter bud possibly suggests the Bim1-mediated bias of the cMTs interacting with the mother cortex towards daughter bud (Fig 5F). Subsequently, we tested the model’s prediction of the differential activity of dynein in mother and daughter cells, by studying the localization of dynein during different stages of the cell cycle. We functionally expressed dynein-3xGFP under the native promoter in the strain CNNV116. Dynein puncta are found to be distributed between the mother and daughter cell during the cell cycle in a stage-specific manner (Fig 5G). In unbudded/small-budded cells, multiple dynein puncta are visible. As the cells proceed to the next stage of cell cycle, dynein possibly localizes to the tip of the migrating cMTs and gets localized to the spindle poles. Subsequently, dynein relocalizes to the cortex of both the mother and daughter cells (Fig 5G). Clearly, there are many randomly distributed dynein puncta in the mother bud compared to the daughter bud (large condensed dynein puncta). The spatial distribution of dynein puncta across the whole mother bud and localized dynein puncta in the daughter bud of large-budded cells suggests differential activity/force transduction of dynein in the mother and daughter cell.
Next, to test the model’s prediction of impaired nuclear migration due to a reduced bias and higher dynein activity in the mother cortex, we examined the dynamics of nuclear migration in the deletion mutant of BIM1 CNNV107 (Δbim1) and a strain over-expressing DYN1, CNNV110 (DYN1OE). As predicted by the model, we observed that in the absence of Bim1 or in the presence of heightened dynein activity, the nuclear migration was delayed (Fig 6A and 6B). In silico model predicts that proper dynein activity is required for timely kinetochore clustering. The effect of the heightened and reduced activity of dynein on the dynamics of kinetochore clustering was tested in strain CNNV111 carrying GFP-CENP-A. Altered expression of dynein indeed leads to impaired dynamics of kinetochore clustering (S4C, S4D, S4E and S4F Fig). Taken together, in silico and experimental data, the synergistic presence of an effective bias due to the interplay between cortical determinants within the mother cortex and a more directed dynein-pull from the daughter cortex is indispensable for the proper nuclear division.
Experimental results suggested that the depletion of Ipl1 shortened the average cMT length from 2.63 μm to 1.3 μm, delayed kinetochore clustering time by 40% (10 min) and nuclear migration by 18% (12 min). Employing a characteristic mean cMT length (1.3 μm) in the model over non-identical sets of homogeneous cell population yields inconsistencies in the time required for nuclear migration and MTOC clustering (S5A and S5B Fig). The model predicted the time required for MTOC clustering and nuclear migration by introducing heterogeneous length of cMTs and perturbing the effective cortical bias mediated by Bim1 and dynein in the mother which also disagreed with the experiment. To account for the synergistic disruption of the cMT length and nuclear migration upon Ipl1 depletion in the model, we assumed that the level of effective cortical bias (a key factor for proper nuclear migration) diminishes exponentially with the average cMT length (Fig 6C). A combined effect of disintegrating MTs and diminishing bias in the mother cortex resulted in clustering of MTOCs in ~29 min and proper nuclear migration within ~47 min (Fig 6D and 6E; S3B–S3D Movie). In contrast to a uniform bias across population scenario (S6C Fig), in silico distribution for the neck to nucleus distance in presence of a length-dependent bias also closely emulated the experimental observation (S6D Fig). To test the model’s prediction of diminishing effective cortical bias in Ipl1-depleted cells, we first studied the localization of dynein in Ipl1-depleted cells of CNNV116 after their growth in non-permissive conditions for 8 h. We observed a reduction in the punctate localization of dynein in the Ipl1-depleted cells as compared to the cells grown in permissive conditions (S6 Fig). Upon increasing the exposure time using the definite intensity module, we observed aberrant spatial distribution of dynein puncta in Ipl1-depleted cells like clustered dynein puncta in unbudded or budded cells (20%), multiple dynein puncta in the daughter bud of large-budded cells (39%) or no dynein puncta in the daughter bud of small-budded cells (60%) (Fig 6F and 6G). Aberrant spatial distribution of dynein in the Ipl1-depleted cells possibly results in the net reduction in the directed dynein-mediated pull from the daughter bud required for proper nuclear migration. However, the decreased MT stability could also result in the aberrant localization of dynein puncta in the Ipl1-depleted cells.
To understand the significance of proper spatial distribution of dynein or Bim1 in nuclear migration, we introduced an additional refinement on the earlier in silico template having uniform dynein/Bim1 distribution in mother and daughter cortices (Fig 6H). We find that the higher Bim1 concentration in the mother bud imparts a greater bias to the cMTs interacting with the mother cortex, required for nuclear migration in to the daughter bud. Our model findings allude that there is a preferable range of effective dynein densities along the cortex resulting in timely nuclear migration (~35–40 min) across the overall population of cells (Fig 6I). In the configurations having the higher effective dynein density in the condensed patch in daughter cortex compared to the dynein density in the mother and elsewhere in the daughter cortex (demarcated in Fig 6I), the randomly distributed dynein puncta in the mother cortex seem to lose the tug-of-war with respect to the focused dynein puncta in the daughter, facilitating timely nuclear migration in wild-type cells. Data obtained from the refined model shows that the parameter regime for proper nuclear migration becomes constrained when additional spindle characteristics are mapped with the experiments suggesting, apart from the puncta, uniformly distributed low-density dynein along the cortex might be crucial (Fig 6I and 6J). Further, we simulated the model considering the altered MT parameters and effective spatial distribution of dynein observed in Ipl1-depleted scenario. We find that a net reduction in the directed-pull on the nucleus from the daughter bud results in the delayed nuclear migration in the Ipl1-depleted cells as observed previously, when the uniform spatial profile of dynein and Bim1 was considered. Combining model refined predictions and experimental validations, we conclude that the differential spatial profile of dynein in the mother and daughter bud also orchestrates proper nuclear migration. Ipl1 spatio-temporally regulates cytoplasmic MTs and spatial distribution of Bim1 and/or dynein in the cell. Absence of Ipl1 impairs the effective cortical bias and the MT lengths leading to defects in kinetochore clustering and nuclear migration.
Next, we estimated the fidelity of chromosome segregation during mitosis in Ipl1-depleted cells due to altered nuclear migration and kinetochore clustering. The flow cytometric analysis (FACS) confirmed the existence of aneuploidy in a population of Ipl1-depleted cells within 8 h of protein depletion (Fig 7A). Volume rendering by Imaris 7.6.4 software (Bitplane, Zurich, Switzerland) reveals a biased distribution pattern of the chromatin mass in the Ipl1-depleted large budded cells. In most cases, a greater amount of the chromatin mass was present in the mother cell as compared to the daughter cell (Fig 7B). C. neoformans is known to acquire aneuploidy as an adaptive mechanism to exhibit heteroresistance against azole drugs [24]. Indeed, two independent ipl1 conditional mutants CNNV101 and CNNV102, grown in the non-permissive condition for 8 h followed by growth on plates containing permissive media, yielded a higher number of fluconazole (32μg/ml) resistance colonies as compared to the wild-type (Fig 7C). Together, these results provide an evidence of the generation of aneuploidy upon depletion of Ipl1 in C. neoformans.
In this work, we show that Ipl1 is essential for viability and it orchestrates pre-anaphase nuclear movement from the mother cell to the daughter in the human pathogenic budding yeast C. neoformans. Ipl1 plays a critical role in maintaining MT stability to facilitate timely kinetochore clustering and nuclear migration. Our in silico study predicts that dynein and/or Bim1 activity regulates MTOC/kinetochore clustering, a pre-requisite for nuclear migration. By regulating the Bim1/dynein activity or localization, Ipl1 creates a net force bias necessary for proper nuclear migration to the daughter cell followed by its division during mitosis in C. neoformans. Cytosolic Ipl1 regulates these processes. It also regulates the disassembly of the NE to enter into the nucleus during mitosis and monitors the kinetochore-MT attachments to ensure high-fidelity chromosome segregation. Finally, we show that malfunctioning of Ipl1 can lead to aneuploidy-induced drug resistance in C. neoformans. Taken together, we demonstrate that dynamic localization of Ipl1 during semi-open mitosis spatio-temporally regulates the process of nuclear division and maintains proper ploidy in the human pathogen C. neoformans that belongs to the fungal phylum of Basidiomycota (Fig 7D).
In ascomyceteous yeast S. cerevisiae, Ipl1 displays a dynamic subcellular localization during the cell cycle [18]. In metaphase, Ipl1 associates with the centromeres and facilitates biorientation by preventing syntelic attachments [62]. In anaphase, Ipl1 localizes to the spindle midzone and regulates timely disassembly of the spindle followed by cytokinesis [9, 17, 18]. However, whether Ipl1 plays any role in regulating the dynamics of cMTs or associated cortical proteins remained unknown. In this study, we elucidated the role of Ipl1 in regulating cortical interactions in C. neoformans having atypical mitotic features [20]. A greater number of cMTs and an increased dynein activity at the daughter cortex in C. neoformans provide a stronger pulling force on the migrating nucleus as compared to the S. cerevisiae [19]. Biased polarisation of cMTs towards the daughter cell and asymmetric recruitment of accessory cytoplasmic proteins such as Myo2, Bim1, Kar9, dynein etc. have been shown to influence nuclear migration [8, 19, 59, 60]. In the absence of a functional homolog of Kar9 [61], which imparts a cortical bias on the cMT tips penetrated into the cortex [60] and She1 which inhibits dynein activity along the MTs in S. cerevisiae [8], it is conceivable that an altered mechanism is operational to drive the process of nuclear migration in C. neoformans. Owing to significant differences in the dynamics of nuclear migration and kinetochore clustering as compared to well-studied yeast species, and the absence of Kar9 and She1, C. neoformans provides a unique system to study the evolution of chromosome segregation across budding yeast species.
Majority of Ipl1-depleted cells in C. neoformans exhibit defects in nuclear migration in contrast to most ipl1-2 cells with unequally segregated chromosomes in S. cerevisiae [32]. Kinetochores are closely associated with MTOCs (and SPBs) throughout the cell cycle in C. neoformans. Ipl1-depleted cells exhibit altered MTOC clustering dynamics which in turn possibly results in delayed kinetochore clustering. The compromised structural stability of MTs suggests that cytoplasmic Ipl1 possibly functions either as an MT stabilizer or as a negative regulator of MT-severing enzymes and regulate nuclear migration in C. neoformans. It is also possible that gross MT abnormalities and formation of “tubulin” blobs in Ipl1-depleted cells could be an indirect consequence of Ipl1 depletion through the altered activity of MAPs or kinesin or improper tubulin folding. A series of evidence indicated that motor proteins such as the kinesin-8 homolog, Kip3 is involved in kinetochore clustering in S. cerevisiae [63], but its function remains uncharacterized in C. neoformans. While MTs are stabilized by end-binding proteins such as Bim1 [60], Ipl1 triggers spindle disassembly by phosphorylating Bim1 in S. cerevisiae [9, 17]. Recently, it has been shown that Ipl1 inhibits spindle stabilization by negatively regulating the MT-stabilizing protein She1 in S. cerevisiae [16]. Using spatio-temporal simulations, we find that pathways involving cortical bias triggered by Bim1 and cytoplasmic dynein facilitate nuclear migration and MTOC clustering in C. neoformans.
A constitutively high activity of Bim1 and/or dynein throughout the cell cycle results in defective spindle disassembly and nuclear migration [16, 17, 64]. In C. neoformans, dynein/Bim1 puncta are distributed in the mother cortex while these puncta are condensed and localized in the daughter bud. Since the spatial profile of the dynein in the mother cell is less condensed and distributed, the resultant opposing pull on the nucleus in the mother cell is substantially reduced, hence lacks in competition against the pull from the daughter for proper migration. Furthermore, the net pull from the mother bud is also suppressed by the enriched spatial profile of Bim1 in the mother cortex contributing to a net effective bias of the pulling force on the nucleus towards the daughter bud. Thus, this spatial arrangement facilitates a net directed-pull on the nucleus towards the daughter bud through the cMTs interacting with the daughter cortex. The spatial distribution of dynein puncta in the mother and daughter bud is found to be significantly altered in Ipl1-depleted cells. Our model findings subscribe that for a suitable range of effective dynein densities along the cortex and the higher effective dynein density in the condensed patch in daughter cortex is important for proper nuclear migration. It is possible that Ipl1 regulates the timely nuclear migration by maintaining the effective spatial distribution profile of dynein/Bim1 in the mother and daughter cell. Irrespective of whether the in silico structural template contains uniform dynein distribution at the mother and the daughter cortex or punctated distribution, our model findings consistently refer to the importance of stronger dynein pull from the daughter bud compared to the mother bud in orchestrating proper nuclear migration.
Presence of cytosolic Ipl1 throughout the cell cycle indicates that Ipl1 is involved in regulation of the activity of cytosolic proteins in C. neoformans. Indeed, a varying cytosolic to the nuclear pool of Ipl1 across various stages may explain a cell cycle-dependent differential regulation of Bim1 and/or dynein in C. neoformans. In the absence of the direct experimental evidence on how Ipl1 alters the characteristics of dynein/Bim1/MTs in C. neoformans, we used an in silico approach and reproduced the Ipl1-depleted phenotypes observed in the experiments. The purpose of the model is to predict plausible mechanism(s) by which the nuclear migration and kinetochore clustering is delayed in Ipl1-depleted cell. Assuming Ipl1 depletion alters effective dynein/Bim1-mediated bias and enhances MT catastrophe, we could reproduce the mutant phenotypes observed in the experiment using our simulated model. Applying perturbations in dynein/Bim1 profiles and densities along with variations in MT stability as observed in the experiment, we could successfully capture the semi-quantitative characteristics of spindle architecture in both wild-type and Ipl1-depleted cells. Although, the simplified in silico framework provides a plausible scenario of the underlying molecular interactions, the model has certain limitations. Specifically, the model neglects considering explicit mechanism how dyneins/Bim1 stochastically attaches and detaches with MTs at the cortex and move along the MT filament toward the SPBs. The model further ignores the dynamic clustering of dyneins that might spatio-temporally regulate pulling on the nucleus via MTs along the course of nuclear migration. Such limitations of a coarse-grained approach invite a more refined approach to be addressed in the future.
The depletion of Ipl1 using the well characterized GAL7 promoter system over a period of many cell cycles in an unsynchronized cell culture is likely to create heterogenous phenotypes of defective kinetochore clustering, nuclear migration and microtubule length in C. neoformans. Alternatively, this effect could be an indirect consequence of chromosome mis-segregation in the previous cell cycles. Recently, condensin has been shown to play an indirect role in the regulation of the gene expression in fission yeast [65]. It is important for the maintenance of proper gene expression by regulating accurate chromosome segregation during mitosis [65]. Therefore, development of a single cell cycle depletion assay such as with cell cycle arrest and release or auxin-based depletion system would be more useful to rule out the indirect consequences of depletion of Ipl1 in C. neoformans. In this study, we propose that semi-open mitosis maintains Ipl1’s function in time and space by regulating the timely entry of Ipl1 into the nucleus that is allowed only when the NE ruptures at during mitosis in C. neoformans. Taken together, we posit that cytosolic pool of Ipl1 regulates the overall MT integrity, and Bim1 or/ and dynein activity to ensure high-fidelity chromosome segregation and prevent aneuploidy-associated drug resistance in C. neoformans.
Strains and primers used in this study are listed in the S1 Table and S2 Table respectively.
The conditional mutant strains of IPL1 were constructed by replacing the promoter of the IPL1 ORF (CNAG_01285) with the GAL7 promoter [33]. The 5’ coding region of the IPL1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into KpnI and XhoI sites of the plasmid pSHG7 that contained the GAL7 promoter and the hygromycin (HygB) marker, to construct pG7IPL1. The 5’ untranslated region (UTR) of the IPL1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the NotI site of pG7IPL1 to obtain pUSG7IPL1. After confirmation of the orientation of the insert, the resulting plasmid was used to amplify the entire deletion cassette using the primers (NV123 and NV118) and introduced into each of H99 [66], CNVY107 and CNVY108, using biolistic transformation [67] to obtain CNNV101, CNNV102, CNNV105, and CNNV104 respectively. All the conditional mutants were confirmed by PCR using oligos listed in the S2 Table.
To visualize the signals of outer kinetochore proteins such as Mtw1 and Dad1 under Ipl1-depleted conditions, we constructed conditional ipl1 mutant strains in CNVY103 and CNVY120 strain backgrounds. The cassette containing the upstream homology region from IPL1, the GAL7 promoter, HygB and the 5’ UTR of the ORF was amplified using primers mentioned above and transformed into the Mtw1- and Dad1- epitope-tagged strains to obtain CNNV108 and CNNV109 respectively. These conditional mutants were confirmed by their inability to grow under non-permissive media conditions.
To study the dynamic subcellular localization of Ipl1, we constructed the strain expressing Ipl1 tagged with mCherry under the GAL7 promoter (CNNV114). The mCherry coding DNA sequence was amplified from the plasmid pS2M using oligos listed in the S2 Table and cloned into BamHI and XhoI sites of the plasmid pUSG7IPL1 that contained the 5’ coding region of the IPL1 ORF, 5’ UTR of the IPL1 ORF, GAL7 promoter, and HygB marker, to construct pUSG7mchIPL1. The resulting plasmid was used to amplify the entire tagging cassette using the primers (NV132 and NV118) and introduced into H99 and CNVY108 using biolistic transformation [67] to obtain CNNV103 and CNNV114 respectively.
To the distinguish the nuclear localization of Ipl1 from that of cytosol, we constructed the strain co-expressing proliferating cell nuclear antigen (PCNA) tagged with GFP and Ipl1 tagged with mCherry under the GAL7 promoter (CNNV112). The 5’ coding region of the PCNA ORF (CNAG_06079) was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the BamHI site of the plasmid pCIN19 that contained the GFP and the nourseothricin (NAT) marker. The resulting plasmid was transformed into CNNV103 to obtain CNNV112.
To study the subcellular localization of Ipl1 under the native promoter, we constructed the strain expressing Ipl1 tagged with triple GFP under the native promoter (CNNV113). The 5’ coding region of the IPL1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into ClaI and XhoI sites of the plasmid pB1586 that contained the triple GFP epitope, to construct pIPL1GFP3. The 3’ untranslated region (UTR) of the IPL1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the SacII site of pIPL1GFP to obtain pUSIPL1GFP3. After confirmation of the orientation of the insert, the resulting plasmid was used to clone the neomycin (NEO) resistance gene. A fragment (2 kb) containing NEO was amplified from pLK25 [20] using oligos listed in the S2 Table and cloned into the NotI site of pUSIPL1GFP3, to obtain pUSIPL1GFP3N. The resulting plasmid was digested with XhoI and NdeI and introduced into H99 using biolistic transformation [67] to obtain CNNV113.
To visualize the nucleus by localizing histone H4 in the absence of Bim1, we constructed the bim1Δ strain in a GFP-H4-expressing strain (CNVY107) background. The BIM1 ORF (CNAG_03993) was replaced with the neomycin (NEO) resistance gene cassette to generate bim1Δ mutants. The BIM1 knockout cassette containing NEO was generated by the overlap PCR strategy described earlier [20] using oligos listed in the S2 Table. Approximately 1 kb each of 3’and 5’ UTR of the gene was amplified from the H99 genomic DNA. A fragment (2 kb) containing NEO was amplified from pLK25 [20]. All the three fragments were mixed to perform overlap PCR for generating the BIM1 deletion cassette. The amplified bim1Δ: NEOR cassette was introduced into CNVY108 (expressing GFP-H4) using biolistic transformation [67] to obtain CNNV107.
The conditional mutant strains of DYN1 were constructed by replacing the promoter of the DYN1 ORF (CNAG_05894) with the GAL7 promoter [33]. The 5’ coding region of the DYN1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into BamHI and SalI sites of the plasmid pSHG7 carrying the GAL7 promoter and HygB to construct pG7DYN1. The DYN1 5’UTR was amplified from H99 using oligos listed in the S2 Table and cloned into the NotI site of pG7DYN1 to obtain pUSG7DYN1. After confirmation of the orientation of the insert, the resulting plasmid was used to amplify the cassette using the primers (NV404 and NV356) and introduced into each of H99 [66], CNVY108 and CNVY113 strains, using biolistic transformation [67] to obtain CNNV110 and CNNV111 respectively. All the conditional mutants were confirmed by PCR using oligos listed in the S2 Table.
To study the subcellular localization of Dyn1 under the native promoter, we constructed the strain expressing Dyn1 tagged with triple GFP under the native promoter (CNNV116). First, a fragment (2 kb) containing NEO was amplified from pLK25 [20] using oligos listed in the S2 Table and cloned into the NotI site of pIPLGFP3, to obtain pCI3GFPN. The 5’coding region of the DYN1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into ClaI and KpnI sites of the plasmid pCI3GFPN that contained the triple GFP epitope and NEO resistance gene, to construct pCD3GFPN. The 3’untranslated region (UTR) of the DYN1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the SacI site of pCD3GFPN to obtain pCD3GFPNDS. After confirmation of the orientation of the insert, the resulting plasmid was used as a template to amplify DYN1-3xGFP DS cassette using oligos NV452 and NV455 and introduced into CNNV101 using biolistic transformation [67] CNNV116.
To study the subcellular localization of MTOCs under the native promoter, we constructed the strain expressing Spc98 tagged with triple GFP under the native promoter (CNNV118). The 5’ coding region of the SPC98 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into ClaI and KpnI sites of the plasmid pCI3GFPN that contained the triple GFP epitope and NEO resistance gene, to construct pCS3GFPN. The 3’ untranslated region (UTR) of the SPC98 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the SacI site of pCS3GFPN to obtain pCS3GFPNDS. After confirmation of the orientation of the insert, the resulting plasmid was used as a template to amplify SPC98-3xGFP DS cassette using oligos, NV478 and NV477 and introduced into CNNV101 using biolistic transformation [67] to obtain CNNV118.
To study the subcellular localization of Bim1 under the native promoter, we constructed the strain expressing Bim1 tagged with triple GFP under the native promoter (CNNV119). The 5’ coding region of the BIM1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into ClaI and KpnI sites of the plasmid pCI3GFPN that contained the triple GFP epitope and NEO resistance gene, to construct pCB3GFPN. The 3’ untranslated region (UTR) of the BIM1 ORF was amplified from the H99 genomic DNA using oligos listed in the S2 Table and cloned into the SacI site of pCB3GFPN to obtain pCB3GFPNDS. After confirmation of the orientation of the insert, the resulting plasmid was used as a template to amplify BIM1-3xGFP DS cassette using oligos, NV470 and NV473 and introduced into H99 using biolistic transformation [67] to obtain CNNV119.
The conditional mutant strains CNNV101, CNNV102, CNNV110 and CNNV111 carrying IPL1 and DYN1 respectively, under the control of the GAL7 promoter were grown in YPG (1% yeast extract, 2% peptone, 2% galactose) as a permissive medium and YPD (1% yeast extract, 2% peptone, 2% dextrose) as a non-permissive medium. The deletion mutant strains were grown in YPD. All the C. neoformans strains were grown at 30°C.
The conditional mutant strains, CNNV104 and CNNV105, were grown overnight in the permissive medium and re-inoculated in the non-permissive media for 4 h. The deletion mutant strain, CNNV107 was grown in YPD overnight. These cells were pelleted at 4,000 rpm and washed once with 1x phosphate buffered saline (PBS). The cell suspension was placed on the slide containing a thin 2% agarose patch prepared in dextrose and the patch was covered with a coverslip. Live-cell imaging was performed at 30°C on an inverted confocal microscope (ZEISS, LSM-880) equipped with a temperature-control chamber (Pecon incubator, XL multi SL), a Plan Apochromat 100x NA oil 1.4 objective and GaAsp photodetectors. Two-color images were taken by the sequential switching between RFP and GFP filter lines (GFP/FITC 488, mCherry 561 for excitation and GFP/FITC 500/550 band-pass, mCherry 565/650 long-pass for emission). For time-lapse microscopy of mCherry-CENP-A, histone H4-GFP, and GFP-Tub1, images were collected at a 2-min interval with 1.5% intensity exposure, a 60-s interval with 0.2% intensity exposure and a 60-s interval with 1% intensity exposure respectively, with 0.5 μm Z-steps. All the images were displayed after the maximum intensity projection of images at each time using ImageJ.
The conditional mutant strains, CNNV104, CNNV105, CNNV108, CNNV109, and CNNV116 were grown till OD600 = 1 in the permissive medium and re-inoculated in the non-permissive media for 8 h. The deletion mutant and CNNV113 strains were grown in YPD overnight. These cells were pelleted at 4,000 rpm and washed once with 1x phosphate buffered saline (PBS) before the cell suspension was placed on a thin growth medium containing 2% agarose patch present on the slide. A coverslip was placed on the patch and processed for imaging. The images were acquired at room temperature using laser scanning inverted confocal microscope LSM 880-Airyscan (ZEISS, Plan Apochromat 63x, NA oil 1.4) equipped with highly sensitive photo-detectors or the DeltaVision System (Applied Precision) or Axio Observer Calibri (ZEISS). The filters used were GFP/FITC 488, mCherry 561 for excitation and GFP/FITC 500/550 band pass, mCherry 565/650 long pass for emission. Z- stack images were taken at every 0.3 μm and processed using ZEISS Zen software/ImageJ. The Airyscan images were subjected to the super-resolution mode processing using ZEISS zen software. All the images were digitally altered with minimal adjustments to levels and linear contrast till the signals were highlighted.
The length of an MT fiber was measured by tracking it manually using a freehand line tool in ImageJ. The end-to-end distance was measured for the intact MTs while the longest possible distance along the clustered MT structure was measured for the disintegrated MTs. The nuclear volumes were measured after the 3D rendering of confocal images with IMARIS 7.6.4 software (Bitplane, Zurich, Switzerland). Images were filtered by Gaussian smoothing and a surface was created using a threshold of absolute intensity. Radii of the nuclei spots were determined by taking a half of the longest diameters for each nucleus spot measured in an individual stack in Imaris. The nuclear volumes were obtained in statistics of processed images and these values were used to calculate nuclear volume ratio of the mother versus daughter cell.
Wild-type and mutant cells were grown under permissive and non-permissive conditions for the indicated time points. The cells of OD600 = 3 were harvested and washed with 1x PBS. The cells were resuspended in 12.5% TCA and disrupted using acid-washed glass beads (Sigma, Cat. No. G8772) by vortexing for 30 min at 4°C. Cell lysates were precipitated at 13000 rpm for 10 min and washed with 80% acetone. The pellet was dried and resuspended in the lysis buffer (0.1 N NaOH, 1% SDS). The samples were diluted in 5x SDS loading dye (5% β-mercaptoethanol, 0.02% Bromophenol blue, 30% glycerol, 10% SDS, 250 mM Tris-Cl pH 6.8) and denatured. Denatured samples were subjected to electrophoresis using 10% SDS PAGE and transferred to a nitrocellulose membrane for 45 min at 25 V by semi-dry method (Bio-Rad). The membranes were blocked with 5% skim milk containing 1x PBS (pH 7.4) for 1 h at the room temperature followed by its incubation with primary antibodies in 2.5% skim milk overnight at 4°C. After three 10 min washes in PBST (1x PBS, 0.05% Tween) solution, the membranes were incubated in solutions containing secondary antibodies in 2.5% skim milk for 2 h. The membranes were washed with PBST (1X PBS, 0.05% Tween) solution thrice and the signals were detected using chemiluminescence method (SuperSignal West Pico Chemiluminescent substrate, Thermo Scientific, Cat. No. 34080).
Primary antibodies used for western blot analysis were mouse anti-GFP (dilution 1:2500) (Santa Cruz Biotech, Cat. No. 9996) and mouse anti-PSTAIRE (dilution 1:2000) (Abcam, Cat. No.10345). Secondary antibodies used are goat anti-mouse HRP conjugated antibodies (dilution 1:10,000) (Bangalore Genei, Cat. No. HP06).
The wild-type and Ipl1-depleted cells grown in the non-permissive condition for 8 h were 10-fold serially diluted (105 to 102) and spotted on plates containing the permissive medium having indicated concentrations of fluconazole (Sigma, Cat. No. F8929-100 MG) and incubated further at 30°C for the antifungal drug resistance assay.
The values of the MT length, budding indices, mother/daughter nuclear volume ratios, kinetochore clustering and nuclei migration time durations were plotted as violin plots overlaid with the column scatter graph in a column type mean with the SEM using the Prism 7 software. Statistical differences were determined using a paired Student’s t-test to calculate the statistical significance. In all the in silico results reported, the quantification of various attributes were averaged over 2000 samples and plotted as column type means with SEM.
|
10.1371/journal.pgen.1004569 | The Not5 Subunit of the Ccr4-Not Complex Connects Transcription and Translation | Recent studies have suggested that a sub-complex of RNA polymerase II composed of Rpb4 and Rpb7 couples the nuclear and cytoplasmic stages of gene expression by associating with newly made mRNAs in the nucleus, and contributing to their translation and degradation in the cytoplasm. Here we show by yeast two hybrid and co-immunoprecipitation experiments, followed by ribosome fractionation and fluorescent microscopy, that a subunit of the Ccr4-Not complex, Not5, is essential in the nucleus for the cytoplasmic functions of Rpb4. Not5 interacts with Rpb4; it is required for the presence of Rpb4 in polysomes, for interaction of Rpb4 with the translation initiation factor eIF3 and for association of Rpb4 with mRNAs. We find that Rpb7 presence in the cytoplasm and polysomes is much less significant than that of Rpb4, and that it does not depend upon Not5. Hence Not5-dependence unlinks the cytoplasmic functions of Rpb4 and Rpb7. We additionally determine with RNA immunoprecipitation and native gel analysis that Not5 is needed in the cytoplasm for the co-translational assembly of RNA polymerase II. This stems from the importance of Not5 for the association of the R2TP Hsp90 co-chaperone with polysomes translating RPB1 mRNA to protect newly synthesized Rpb1 from aggregation. Hence taken together our results show that Not5 interconnects translation and transcription.
| In this work we show that, both in the nucleus and in the cytoplasm, Not5 plays a “bridging” role for RNA Polymerase II. In the cytoplasm, Not5 interacts with the mRNA encoding the largest subunit of RNA polymerase II Rpb1 and supports the association of a co-chaperone to newly produced protein, to keep it soluble and assembly competent. In the nucleus, Not5 interacts with the Rpb4 subunit of polymerase that is known to readily dissociate from the rest of the polymerase, and it is essential for Rpb4 to associate with mRNAs at the completion of transcription to contribute to translation and mRNA degradation in the cytoplasm. Hence our data define Not5 as a key player in the cross-talk between different stages of eukaryotic gene expression: Not5 impacts on production of polymerase, hence transcription, during translation, and on Rpb4 mRNA association, hence translation and mRNA degradation during transcription.
| The life of an mRNA molecule in eukaryotic cells is considered to be the sum of distinct events separated in time and space. Precisely, this separation seems to constitute the characteristic difference distinguishing eukaryotes from prokaryotes, where translation is co-transcriptional and occurs in a single cellular compartment. Several studies in recent years, however, have challenged this simple view. First, the heptapeptide repeat-containing C-terminal domain (CTD) of the largest subunit of eukaryotic RNA polymerase II (RNA Pol II) was found to direct post-transcriptional RNA processing events. It serves as a landing platform for components of the machines involved in mRNA capping, splicing, and mRNA export [1], [2], [3]. More recently and provocatively, an RNA Pol II subunit, Rpb4, has been suggested to play roles not only in the nucleus during the transcription process, but also subsequently in the cytoplasm, contributing to both the RNA degradation and translation processes [4], [5].
The conserved eukaryotic Ccr4-Not complex also contributes to both transcription and mRNA decay and is found both in the cytoplasm and nucleus (for reviews see [6], [7]). The complex consists of 9 subunits in the yeast Saccharomyces cerevisiae (Ccr4, Caf1, Caf40, Caf130, and Not1-5). The single CNot3 protein of higher eukaryotes, whether human [8] or fly [9] corresponds to yeast Not3 and Not5, which share 44% identity in their N-termini. In these eukaryotes, the complex also carries additional subunits CNot10 and CNot11, and lacks Caf130 [10], [11], [12], [13]. The Ccr4-Not complex plays roles at several stages of gene expression. Many subunits of Ccr4-Not can be cross-linked to genes being transcribed [14], [15], [16], the complex interacts with RNA Pol II and contributes to transcription elongation [17], and the Not subunits impact on the distribution of general transcription initiation factors across the genome [14], [18]. The Ccr4 and Caf1 subunits comprise the major eukaryotic deadenylase and catalyze the first and rate-limiting step of RNA degradation ([19] and for review see [20]). Recent studies in yeast have established that some subunits of the Ccr4-Not complex are present at translating ribosomes (polysomes) [21], [22] and that the level of polysomes is reduced in certain Ccr4-Not deletion mutants. This coincides with an accumulation of aggregated proteins in the mutants [21], [23], and with the importance of the Ccr4-Not complex for the assembly of the multi-subunit proteasome complex [24].
The functional implication of both the Ccr4-Not complex and the Rpb4 subunit of RNA Pol II to all stages of the mRNA life cycle was supported by a recent study revealing that transcription and mRNA degradation rates have co-evolved oppositely and that this coincides with single nucleotide changes in either RPB4 or CCR4-NOT genes [25]. These studies indicate that Rpb4, which connects transcription to downstream events, may somehow connect the polymerase also to the Ccr4-Not complex. Intriguingly however, it has been reported that the interaction of polymerase with the Ccr4-Not complex, does not require Rpb4 [17].
RNA Pol II consists of 12 subunits, 10 of which compose the catalytic core. These are shared with RNA Pol I and RNA Pol III or related to subunits of these other polymerases [26]. Several recent studies have provided insight into the mechanism by which the 12-subunit RNA Pol II is assembled (reviewed in [27]). The finding that partially assembled polymerase complexes accumulated in the cytoplasm under conditions of imbalanced levels of different RNA Pol II subunits suggested cytoplasmic assembly of this enzyme. For instance, treatment of cells with α-amanitin leads to specific degradation of Rpb1 from elongation-stalled polymerase and to the accumulation of a cytoplasmic Rpb2 sub-complex containing Rpb3, Rpb10, Rbp11 and Rpb12 [28]. In contrast, inhibition of the de-novo synthesis of any Pol II subunit besides Rpb1 by siRNA leads to the accumulation of cytoplasmic Rpb1. This pool of Rpb1 is mostly unphosphorylated and insensitive to α-amanitin suggesting that it is newly synthesized Rbp1, which has not been engaged in transcription. The cytoplasmic assembly complexes have been characterized by mass-spectrometry-based proteomics [28] and found to represent two intermediates. The Rpb2 sub-complex contains the Gpn1/Npa3, Gpn2 and Gpn3 GTP binding proteins and several chaperones. Another sub-complex contains Rpb1 with the Hsp90 chaperone and its R2TP co-chaperone. In yeast, R2TP is composed of the Tah1 tetratricopeptide repeat (TPR) protein, Pih1 (protein interacting with Hsp90) and the two AAA+ ATPases, Rvb1 and Rvb2. Pih1 binds Tah1, the Rvb proteins and the yeast Hsp90 chaperones Hsp82 and/or Hsc82 [29], [30]. The two RNA Pol II assembly intermediates join, and then enter the nucleus mainly via the nuclear transport of fully assembled polymerase [28]. The transport requires association of the assembled polymerase with an NLS-containing protein Iwr1 [31]. GTP binding might also play a role in assembly and/or nuclear import of assembled RNA Pol II, since depletion of the human GTP binding protein Gpn1, or mutation of its yeast ortholog Npa3, leads to cytoplasmic accumulation of polymerase subunits [32], [33]. Another protein recently identified as playing a role in assembly of RNA Pol II and also RNA Pol I and RNA Pol III is the Bud27 prefoldin [34]. It is required for the correct integration of Rpb5 and Rpb6 into all three polymerases and acts prior to nuclear import.
Rpb4 forms a hetero-dimeric sub-complex with Rpb7, which extends like a stalk from the core of RNA Pol II. The Rpb4/7 dimer can dissociate from the rest of RNA Pol II and is in excess to the other RNA polymerase II subunits. It shuttles between the cytoplasm and the nucleus, interacts with the translation scaffold factor eIF3 and is required for wild-type levels of translating polysomes [4], [35], [36] and for review see [37]. Curiously, while Rpb4 plays an essential role in connecting transcription to downstream steps in gene expression, it is not essential for yeast viability.
In this work we investigated how Rpb4 and the Ccr4-Not complex are functionally connected. We determined that Rpb4 shows a very tight two-hybrid interaction with the Not5 subunit of the Ccr4-Not complex. In addition, we found that the presence of Rpb4 in translating ribosomes, and more globally the association of Rpb4 with mRNAs, was dependent upon nuclear Not5. We observed that not only Rpb4, but also several other RNA Pol II subunits were present in polysomes. These findings are consistent with cytoplasmic assembly of RNA Pol II occurring on translating ribosomes, as suggested for protein complexes quite generally [38]. Moreover our data indicates that cytoplasmic Not5 contributes to RNA Pol II assembly at least in part by supporting interaction of de-novo synthesized Rpb1 with Hsp90-R2TP co-chaperone and preserving a soluble pool of Rpb1 that is apt to interact with Rpb2 to form new RNA Pol II complexes. We determined that this role of Not5 for Rpb1 solubility is conserved in Drosophila melanogaster. Hence, Not5 is in a central position for the bidirectional communication between transcription and translation.
The importance of the Ccr4-Not complex during the entire life of mRNAs (for review see [6]) is reminiscent of the extended function described for the Rpb4 subunit of RNA Pol II [37]. We hence investigated by the two-hybrid assay whether the Ccr4-Not complex interacts with Rpb4. We used Rpb4 as bait and tested its interaction against each of the Ccr4-Not subunits as preys. As positive control we used Nip1, an eIF3 subunit with which Rpb4 has been shown to interact [4]. We observed strong two-hybrid interactions of Rpb4 with both Not3 and Not5. In both cases the detected interaction was more remarkable than that between Rpb4 and its known partner Nip1 (Fig. 1A). A weaker interaction between Rpb4 and the other Ccr4-Not subunits was also evident (Fig. S1A) and they could be confirmed by co-immunoprecipitation experiments (Fig. S1B) even in the presence of RNase indicating that the interaction is not bridged by RNA. Not3 and Not5 share extended sequence homology and may have evolved in yeast from a common ancestral gene, since in higher eukaryotes there is a single gene encoding this subunit of the Ccr4-Not complex (reviewed in [39]). The deletion of either Not3 or Not5 is lethal when combined with the deletion of Not4 [40] so we tested whether this genetic interaction was shared by Rpb4 that is not a subunit of the Ccr4-Not complex, but interacts with both Not3 and Not5. Indeed, the deletion of Rpb4 displayed a striking slow growth phenotype when combined with Not4 (Fig. 1B). A synthetic slow growth phenotype was also detected when rpb4Δ, like not5Δ [40], was combined with ccr4Δ, but not when it was combined with the deletion of Caf40, another Ccr4-Not subunit (Fig. S1C).
Rpb4 has been connected to translation and found in polysomes [4], and this has also been established for certain Ccr4-Not subunits (Not4 and Not5) [21], [22], [23]. This led us to test whether the presence of Rpb4 and Ccr4-Not subunits in polysomes was interdependent. Not2 and Not5 were essential for the detection of Rpb4 in polysomes (Fig. 1C). When trying to do the reverse experiment we were unable to create an rpb4Δ strain expressing tagged Not5, probably because of synthetic lethality issues (see below). Hence instead we followed the fractionation of tagged Not3 in cells lacking Rpb4. Not3 was present in polysomes even in cells lacking Rpb4 (Fig. 1D).
Rpb4 was first connected to translation through its interaction with the translation initiation factor eIF3 [4]. We could recapitulate this interaction by co-immunoprecipitating a subunit of eIF3, Prt1, with Rpb4 (Fig. 1E). Since we observed that Rpb4 was not present in polysome fractions in not2Δ and not5Δ, it was of interest to determine whether Rpb4 could still interact with eIF3 in these mutants. We could not detect co-immunoprecipitation of Prt1 with Rpb4 in not5Δ cells, suggesting that the interaction of Rpb4 with eIF3 is dependent upon Not5 (Fig. 1E). Consistent with a role for the Ccr4-Not complex in mediating the interaction of Rbp4 with eIF3, we found that Prt1 co-immunoprecipitates with Not1 (Fig. 1F). In addition two-hybrid experiments revealed interactions between another eIF3 subunit, Nip1 and many Ccr4-Not subunits (Fig. S1D). The Not1-eIF3 interaction was independent of Not5 (Fig. 1F) in good correlation with the observation that Not1 association with polysome fractions does not depend upon Not5 but is dependent upon intact polysomes (Fig. S1E).
Rpb4 has been shown to shuttle between the nucleus and the cytoplasm and accumulates in the cytoplasm under stress conditions that can be mimicked by fixing yeast cells with formaldehyde [41]. Under such conditions, we could confirm a largely cytoplasmic localization of Rpb4 in wild-type cells (Fig. 1G). In contrast, in cells lacking Not5, Rpb4 was present in the cytoplasm in a lesser amount and accumulated in nuclei (Fig. 1G). Moreover, the nuclei tended to display aberrant morphologies when cells lacked Not5 and expressed tagged Rpb4. This synthetic phenotype was consistent with our observation that no viable spores lacking both Not5 and Rpb4 germinated upon dissection of diploids obtained from crossing rpb4Δ with not5Δ and that attempts to create an rpb4Δ strain expressing a C-terminally tagged Not5 was unsuccessful as mentioned above. The expression of a derivative of Not5 maintained in the cytoplasm by a nuclear export signal (NES) in the not5Δ strain expressing tagged Rpb4 (Fig S2) could not rescue the stress-induced cytoplasmic localization of Rpb4, in contrast to its wild-type counterpart (Fig. 1H).
The presence of Rpb4 in the cytoplasm was reported to result from its interaction with mRNAs occurring at the completion of transcription [36]. We thus analyzed the interaction of Rpb4 with a couple of mRNAs, namely RPB1 and NIP1, in wild-type cells or in cells lacking Not5. Rpb4 is expressed at similar levels in both strains, and is immunoprecipitated to similar extents in both strains (Fig. S3). RPB1 mRNA was significantly enriched in the Rpb4 immunoprecipitates from wild-type cells but not from not5Δ cells, and in parallel significant binding of Not5 to RBP1 mRNA was detected (Fig. S4). In contrast, no significant enrichment of NIP1 mRNA could be detected in the Rpb4 immunoprecipitates from either strain, but it is nevertheless noteworthy that less NIP1 mRNA was immunoprecipitated with Rpb4 from not5Δ than from the wild-type (Fig S4). The relative presence of NIP1 mRNA in the immunoprecipitate versus total mRNA was similar for Rpb4 and Rpl17, a ribosomal protein expected to be associated with all translated mRNAs (Fig. S4). This indicates that Rpb4 is probably associated with NIP1 mRNA, but that the representation of NIP1 mRNA in the pool of mRNAs associated with Rpb4 is not higher than its representation in the pool of total mRNAs. This is in contrast to RPB1 that is more enriched in the immunoprecipitates of Rpb4 than Rpl17, and significantly enriched in the immunoprecipitate versus total mRNA of Rpb4, but not Rpl17. Neither RPB1 nor NIP1 mRNAs were immunoprecipitated from an untagged strain confirming that the immunoprecipitations were specific.
Previously not only Rpb4, but also its partner protein Rpb7 has been implicated in translation and cytoplasmic mRNA degradation [4], [42]. The Rpb4 and Rpb7 proteins are thought to shuttle between nucleus and cytoplasm as a heterodimer [36] and are believed to act together in mRNA degradation and translation, connecting different stages of gene expression [4]. Consistently, Rpb7 was also detectable in polysome fractions (Fig. 2A). However, its presence in these fractions was not dependent upon Not5 in contrast to that of Rpb4, and it was less extensively localized in polysomes than Rpb4 (compare Fig. 2A to Fig. 1C). Furthermore, localization of Rpb7 under the stress conditions, which resulted in a mostly cytoplasmic localization of Rpb4 in wild-type cells, was mostly nuclear and was not affected by Not5 (Fig. 2B). These findings suggest that the presence of Rpb7 in the cytoplasm does not follow the same rules as the presence of Rpb4, neither in the extent of its cytoplasmic localization nor in the Not5-dependence of its polysome association. Moreover, in contrast to Rpb4, Rpb7 does not display a significant two-hybrid interaction with Not5 (Fig. S5).
The different behavior of Rpb4 and Rpb7 concerning the extent of their polysome association and its dependence upon Not5 led us to study other RNA Pol II subunits in this respect. The other subunits that we investigated, including the two largest subunits, Rpb1 and Rpb2, as well as the RNA pol II-specific subunit Rpb11, were present in polysome fractions regardless of the presence or absence of Not5 but dependent upon polysome integrity as defined by polysome disruption with EDTA (Fig. 3A). The exception was Rpb3, which was not detectable in polysome fractions (Fig. 3B) as has previously been described [4].
These unexpected findings led us confirm that RNA Pol II subunits interact with ribosomes. We treated cellular extracts with RNase to disrupt polysomes in a manner distinct from EDTA treatment, and confirmed that loss of the heaviest polysomes resulted in the disappearance of Rpb2 from the heaviest fractions (Fig. 3C). We also immunoprecipitated Rpb2 from cells expressing a tagged ribosomal subunit Rpl25 and could co-immunoprecipitate Rpl25 with Rpb2 (Fig. 3D).
The finding that in addition to Rpb4 and Rpb7, core polymerase subunits are present in polysomes is compatible with reports that RNA Pol II assembly takes place in the cytoplasm, and with the claim that many protein complexes are assembled co-translationally [38]. In addition, since in absence of Not5 Rpb4 fails to localize to polysomes, and does not even accumulate in the cytoplasm, Rpb4 may not assemble optimally with RNA Pol II. To determine whether Rpb4 might dissociate more readily from RNA Pol II in the absence of Not5, we analyzed Rpb4 complexes from total extracts of wild-type or mutant cells on a native gel (Fig. 4A). An Rpb4 complex of the size of core RNA Pol II (around 700 kDa) was detected in both extracts. In addition, many faster migrating forms of Rpb4 were detected, compatible with our knowledge that this subunit readily dissociates from RNA Pol II. However, the extent of these additional smaller complexes was greater in the absence of Not5.
These findings led us to question whether Not5 might be globally affecting RNA Pol II assembly. We looked at complexes of several other RNA Pol II subunits (Rpb2, Rpb3, Rpb9 and Rpb11) from wild-type and mutant cell extracts on native gels. In all cases, a complex of the size of core RNA Pol II was observed with the 4 subunits (Fig. 4B) as with Rpb4 (Fig. 4A). The 4 subunits were detected in at least one additional smaller complex of a similar size in extracts from not5Δ (Fig. 4B). These two complexes could readily be purified via any of the 4 subunits as shown for Rpb9 on Fig. 4C. Western blotting revealed that the larger complex contains Rpb1, as expected if it contains core RNA Pol II, but the smaller complex from not5Δ lacks Rpb1 (Fig. 4C). This was also observed for the Rpb2, Rpb3 and Rpb11 complexes (Fig. S6). Interestingly, both complexes were sensitive to RNase but not to DNase treatment of the extracts (Fig. 4D) suggesting that they could be newly assembled RNA Pol II or assembly intermediates stabilized with RNA rather than RNA Pol II extracted from chromatin or released after transcription. It is notable that RNase treatment of extracts had an impact on very large forms of the polymerase subunits that without treatment tended to remain at the top of the native gels, or did not even enter the gel (see for instance Rpb11 complexes on Fig. 4B) in both wild-type and mutant extracts. New large heterogeneous forms of the polymerase subunits (shown for Rpb11 on Fig. 4D) were detected in the native gels. It could be that RNase digestion of polysomes released RNA Pol II in forms that could then enter native gels.
Taken together, these results suggest that RNA Pol II assembly is co-translational and that Not5 is important for co-translational assembly of RNA Pol II. This model could be confirmed by pulse labeling wild-type and not5Δ cells with 35S-methionine and purifying RNA Pol II via Tap-tagged Rpb2 immediately, and after 1 and 2 h chase. We followed the Rpb1 co-purifying with Rpb2. This experiment showed delayed association of labeled Rpb1 with Rpb2 and then subsequently delayed chase of this newly labeled Rpb1 in the Rpb2 purification, in the not5Δ strain compared to the wild-type strain (Fig. S7).
The Rpb2, Rpb3, Rpb9 and Rpb11 complexes lacking Rpb1 detected in extracts from cells lacking Not5 (see above) are reminiscent of an RNA Pol II assembly intermediate that has been described [28]. Its accumulation in not5Δ cells might indicate that the Rpb1 intermediate, with which it should join, is present in limiting amounts in mutant cells. To address this issue, we compared Rpb1 levels in wild-type and mutant soluble extracts. The level of Rpb1 was lower in extracts from not5Δ compared to wild-type, particularly Rpb1 in very large complexes (Fig. 5A and B). Though the difference was not always as dramatic as shown on Fig. 5A and B, it was nevertheless very consistent.
We thus compared the half-life of Rpb1 in wild-type and not5Δ cells. No significant reduction of Rpb1 levels after 2 h of protein synthesis arrest was observed suggesting that Rpb1 is not particularly unstable in either wild-type or not5Δ cells (Fig. 5C). This finding is in accord with results of a previous study, which measured Rpb1 levels in wild-type cells up to 4 h after protein synthesis arrest [43]. We obtained a similar result after 8 h of protein synthesis arrest (Fig. S8). Hence, increased degradation of Rpb1 is unlikely to explain the low levels of Rpb1 in soluble extracts of not5Δ.
RPB1 mRNA levels are not significantly altered in not5Δ cells ([44] and Fig. S4). To determine whether there might nevertheless be a difference in the levels of de-novo synthesized Rpb1 between wild-type and not5Δ that could explain the reduction of Rpb1 in total extracts of the mutant, we performed a pulse-chase experiment and immunoprecipitated Rpb1 from extracts of labeled cells (Fig. 5D). Surprisingly, for a similar pulse, the amount of soluble proteins labeled was generally higher in not5Δ than in wild-type cells. Indeed, the same amount of label was provided by 4 times less total protein from not5Δ than from wild-type (compare Coomassie panels of Fig. 5D). Moreover similar amounts of Rpb1 were immunoprecipitated from 4 times less total protein from not5Δ than from wild-type (Fig. 5D, IP-Rpb1 upper panel), suggesting that immunoprecipitation of Rpb1 was more efficient from not5Δ than from the wild-type. In addition, more of the Rpb1 immunoprecipitated from the mutant was labeled (Fig. 5D, IP-Rpb1 lower panel), indicating that de-novo synthesized Rpb1 was more efficiently immunoprecipitated from the mutant, possibly because there was less unlabeled Rpb1 to compete for antibody binding in the mutant. Finally, the amounts of labeled Rpb1 detected in the immunoprecipitate from wild-type or not5Δ were not much reduced after 60 min of chase under conditions of protein synthesis arrest. Taken together, these results indicate that newly synthesized Rpb1 is mostly stable and not limiting in not5Δ but that it seems to be more accessible to immunoprecipitation in the mutant.
Because of the possible differential competing immunoprecipitation of labeled and unlabeled Rpb1 in the 2 strains in these experiments, we could not definitively define whether translation of RPB1 mRNA was altered or not in not5Δ. We thus checked the relative representation of RPB1 and other mRNAs in polysomes of not5Δ relative to the wild-type. We evaluated the level of RPB1, NIP1 and several other mRNAs in the same amount of RNA prepared from total extracts and polysomes of wild-type and not5Δ. We found no significant difference in levels of RPB1 (Fig. 5E) or NIP1 (Fig. S9) mRNAs in total extracts or polysomes between wild-type and not5Δ cells. In contrast, several mRNAs were significantly reduced in polysomes relative to their representation in total extracts in not5Δ compared to the wild-type. For instance levels of RPS8A and GAR1 mRNAs were higher in total extracts but lower in polysomes of not5Δ whilst RPS22A was expressed at equal levels in both strains, but much less present in polysomes of not5Δ (Fig. 5E). In contrast UBP10 mRNA was under-expressed in not5Δ but present at equal levels in polysomes. Taken together, these results indicate that distribution of mRNAs in the translating pool of mRNAs is modified in not5Δ.
If neither the stability nor the de novo synthesis of Rpb1 is reduced in not5Δ cells, then why are Rpb1 levels in cellular extracts from these cells reduced? The majority of soluble Rpb1 in wild-type extracts is present in heterogeneous complexes, much larger than the major Pol II complex that can be purified via the other Pol II subunits (Fig. 5B). These Rpb1 complexes were severely reduced in not5Δ extracts, while the Rpb1-containing complex that was purified via other RNA Pol II subunits (such as Rpb9, see above Fig. 4C) was present in roughly equal amounts in wild-type and mutant cells.
These heterogeneous Rpb1 complexes reduced in not5Δ are too large to be mature RNA Pol II, but might include newly synthesized assembly-competent and soluble Rpb1 associated with the Hsp90 chaperone (Hsp82 and Hsc82 in yeast) and the R2TP co-chaperone [28]. If such complexes are reduced in not5Δ, one might expect newly produced Rpb1 to fall out of soluble extracts and aggregate. Analysis of total soluble extracts and protein aggregates from wild-type and not5Δ showed that this is indeed the case (Fig. 6A). To confirm these observations, we purified the R2TP complex via its subunits Rvb1 and Rvb2, from wild-type and not5Δ. The R2TP subunits were expressed at equal levels in both strains, whereas, as expected, the levels of Rpb1 were lower in the mutant (Fig. 6B, left panel). Like Rpb1, the R2TP subunits were purified much more efficiently from not5Δ strains (Fig. 6B, right lower panel). Nevertheless Rpb1 co-purified with both Rvb1 and Rvb2 from wild-type cells, but much less from not5Δ (Fig. 6B, right upper panel). A similar observation was made when R2TP was purified via the Pih1 subunit (Fig. S10). Whilst interaction of R2TP with Rpb1 appeared reduced in cells lacking Not5, in contrast Rpb1 similarly co-immunoprecipitated Hsp90 (Fig. 6C), and we found that Hsp90, like Rpb1, accumulated in protein aggregates in not5Δ (Fig. 6D). Expression of a Not5 derivative that carries a nuclear localization signal and complements the slow growth of not5Δ, failed to rescue aggregation of Rpb1 and accumulation of Hsp90 in the aggregates (Fig. 6D), or accumulation of RNA Pol II assembly intermediates (Fig. 6E, left panel), consistent with a role of Not5 in the cytoplasm, at the site of translation. It did however rescue the presence of Rpb4 in polysomes, and it partially rescued polysome levels (Fig. S11A and S11B). In turn, expression of the Not5 derivative that carries a nuclear export signal and that could not rescue accumulation of Rpb4 in the cytoplasm (see Fig. 1H) did rescue levels of soluble Rpb1 and RNA Pol II assembly (Fig. 6E, right panel), and it fully recovered wild-type levels of polysomes (Fig. S11C) but not presence of Rpb4 in polysomes (Fig. S11D).
If R2TP and Hsp90 need to associate with newly produced Rpb1, one might expect that these proteins are also present at the site of translation. We hence tested for the presence of Rvb1, Rvb2 and Hsp82 in polysome fractions of extracts separated on sucrose gradients (Fig. 6F). Indeed, all three proteins were detected in heavy fractions containing polysomes. The presence of these proteins in heavy fractions was clearly dependent in part upon polysome integrity, as determined by disrupting polysomes with EDTA (Fig S1E), supporting the idea that these proteins are associated to some extent with polysomes. Interestingly, in not5Δ, while Rvb1 and Hsp82 had sedimentation patterns similar to the wild-type, Rvb2's presence in polysomes was reduced. We also observed that Hsp82 associated significantly with RPB1 mRNA in wild-type cells, but even more so in the absence of Not5 (Fig. S12). These findings are consistent with a Not5-dependent role in co-translational assembly of R2TP, Hsp90 and Rpb1.
Since Not5 is important for cytoplasmic localization of Rpb4, we wondered whether this could be indirectly the cause for Not5 relevance in appropriate interaction of Rpb1 with R2TP and expression of assembly-competent soluble Rpb1. To address this question, we compared Rpb1 levels in wild-type, not5Δ and rpb4Δ and how it might affect complexes of other RNA Pol II subunits. The level of Rpb1 was increased in total extracts of rpb4Δ rather than decreased as in not5Δ and the deletion of Rpb4 had relatively little impact on Rpb9 complexes compared to not5Δ (Fig. 6G).
To test whether aggregation of Rpb1 and the role of Not5 could be demonstrated in living cells in higher eukaryotes, we studied Drosophila nurse cells. In these cells active transcription in multiple polythenic nuclei serves the production of the maternal mRNA dowry for the early development of the embryo. We analyzed cells that are heterozygous nulls for RPB2 and should therefore accumulate cytoplasmic Rpb1 in assembly intermediates, and determined the consequence of making such cells heterozygous nulls for CNOT3 (the ortholog of yeast NOT5). Wild-type cells were used as a control. All cells were stained with antibodies against Rpb1 (Fig. 7). As expected, the absence of one RPB2 allele led to accumulation of cytoplasmic Rpb1, whilst in wild-type cells Rpb1 was mostly visible in nuclei. In CNOT3+/− cells, no major difference could be seen compared to wild-type cells. In trans-heterozygotes, however, cytoplasmic speckles, which correspond probably to Rpb1 aggregates, were clearly visible (Fig. 7 and Fig. S13).
These data indicate that the role of Not5 for the association of newly synthesized Rpb1 with the R2TP co-chaperone to protect newly synthesized Rpb1 from aggregation is likely to be conserved from yeast to flies.
In this work we show that the Not5 subunit of the Ccr4-Not complex interacts with the Rpb4 subunit of RNA Pol II, and that Not5 is required for the interaction of Rpb4 with the eIF3 translation factor, for Rpb4 presence in polysomes and globally for Rpb4 cytoplasmic localization, because Not5 is required for Rpb4 association with mRNAs. These findings indicate that Not5 contributes importantly to the linkage of transcription to translation by Rpb4.
The mechanism by which Not5 exerts this effect on Rpb4 is unclear. It has been shown that Not5 is recruited to transcribed ORFs, and that the Ccr4-Not complex can interact with RNA Pol II and impact on elongation (for review see [6]). This interaction was first reported not to require Rpb4 [17], but this question is now revisited [45]. We find that localization of Not5 to the nucleus can rescue Rpb4 cytoplasmic accumulation. Therefore we imagine that Not5 contributes either to prevent Rpb4 from dissociating from transcribing RNA Pol II so that it can associate with mRNAs at the completion of transcription or to directly promote Rpb4 association with mRNAs. We demonstrate a very strong interaction of Not5 with Rpb4 in the two-hybrid assay. Similarly, we see strong interaction between Rpb4 and Not3 that is 44% similar to Not5 in its N-terminal domain. Hence, one can assume that Rpb4 interacts with the N-terminal domain of Not5. Consistently, the deletion of this domain of Not5, when combined with the deletion of Not3, leads to a temperature sensitive growth phenotype [46], as does the deletion of Rpb4 [47].
Besides being important for transcription under stress conditions, Rpb4 has also been reported to impact on mRNA export, mRNA degradation and translation [48], [49]. Not5 shares all of these functions, and moreover a recent study reported that single nucleotide changes in RPB4 or NOT5 correlate with opposite co-evolution of transcription and mRNA degradation rates [25]. Our current study revealing that Not5 is important for cytoplasmic functions of Rpb4 provides now a good explanation for this functional similarity of the two proteins.
Rpb4 is believed to fulfill its cytoplasmic function as part of the Rpb4-Rpb7 heterodimer. Our results argue against this idea. First, while the interaction of Not5 with Rpb4 is strong, the interaction between Not5 and Rpb7 is weak. Furthermore, while Not5 is important for the cytoplasmic and polysome localization of Rpb4, it is not required for these localizations of Rpb7, which in any event are less prominent than those of Rpb4. It could be that the roles attributed to Rpb7 in the cytoplasm by the analysis of mutant phenotypes are due to its importance to connect Rpb4 with the core RNA Pol II, since this connection is important for the cytoplasmic functions of Rpb4 [4], [35], [36]. The presence of Rpb7 in polysomes, previously argued to indicate a translation function for Rpb7 [4], might instead be related to cytoplasmic co-translational assembly of RNA Pol II. This issue obviously still needs to be clarified.
Our study reveals that Not5, besides playing a role in the nucleus for the cytoplasmic localization and cytoplasmic functions of Rpb4, contributes in the cytoplasm to co-translational RNA Pol II assembly. Indeed, we observed that in the absence of Not5, a complex consisting of Rpb2, Rpb3, Rpb9 and Rpb11 but lacking Rpb1 accumulates. This correlates with a critical importance of Not5 for the association of Rpb1 with the R2TP Hsp90 co-chaperone and a tendency of Rpb1 to aggregate in not5Δ cells. Interestingly, in the absence of Not5, Hsp90 shows increased association to RPB1 mRNA and Hsp90 is detected in the not5Δ aggregates together with Rpb1. This observation is compatible with a model in which Hsp90 associates with RPB1 mRNA during Rpb1 production, but in the absence of Not5 has a tendency to remain “stuck” to this mRNA.
The inefficient formation of the soluble assembly-competent Rpb1 intermediate consistently correlates with an accumulation of an Rpb2 intermediate, with which it joins to form RNA Pol II. Though we did not specifically define the composition of this Rpb2 sub-complex, Gpn1/Npa3 that is known to participate in RNA Pol II assembly and to associate with Rpb2, and Iwr1 that binds cytoplasmically assembled RNA Pol II for its nuclear import, were present in similar sub-complexes as Rpb2 in not5Δ (Fig. S14).
The importance of Not5 for formation of soluble Rpb1 assembly intermediates does not result from the importance of Not5 for localization of Rpb4 to polysomes. Indeed, the former needs cytoplasmic Not5 that does not rescue presence of Rpb4 in polysomes. Moreover Rpb1 does not aggregate in cells lacking Rpb4, and in fact is present at enhanced levels in extracts from rpb4Δ. It is interesting to note that the presence of Rpb4 in polysomes does not entirely rescue polysome levels if Not5 is expressed in the nucleus. Inversely, expression of Not5 in the cytoplasm fully rescues polysome levels despite the absence of cytoplasmic Rpb4.
Our findings argue that Not5 is important for translation at least in two different ways: one via its importance in the nucleus to support Rpb4 presence in the cytoplasm and a second one in the cytoplasm for co-translational events. The exact connection between these roles of Not5 in two separate cellular compartments remains to be determined. The importance of Not5 for translation is further exemplified by a change in specific mRNA translatability.
The importance of Not5 for co-translational RNA Pol II assembly via production of soluble-assembly-competent de novo synthesized Rpb1 seems distinct from the reported role of the Bud27 prefoldin, since no reduction of soluble Rpb1 was described in cells lacking Bud27 [34] and it is conserved. Indeed, in Drosophila cells, a reduction of the Not5 ortholog, CNot3, results in accumulation of cytoplasmic Rpb1 in speckles.
Our findings raise the question of how Not5 contributes to the interaction of R2TP with newly synthesized Rpb1, as neither the level of R2TP subunits nor that of newly made Rpb1 is reduced in absence of Not5. We know that Not5 is present at polysomes and interacts with RPB1 mRNA, and that a component of the Rpb1 assembly complex, in particular Rvb2, is not detectable in polysome fractions in the absence of Not5. Our mass spectrometry analysis of proteins co-purifying with Not5 identified both Rvb1 and Rvb2 (Table S1). Hence Not5 might interact with R2TP subunits and bring them to productively interact with Hsp90 and newly synthesized Rpb1 in ribosomes translating RPB1 mRNA. This is compatible with the role of the Ccr4-Not complex in co-translational quality control that has been suggested by several recent studies (for review see [50]).
Rvb1 and Rvb2 are not only components of R2TP but also of several important protein complexes (for review see [51]). We observed that both proteins were more accessible for immunoprecipitation in not5Δ extracts suggesting that R2TP may be globally more accessible in this strain. It could be that this co-chaperone is globally less well associated with its client proteins, not only with Rpb1. This is compatible with our observation that protein aggregation in not5Δ is quite prominent [21] and does not only concern Rpb1, and it is compatible with a reduced presence of Rvb2 in polysomes in the absence of Not5. If true, this would indicate that the cytoplasmic function of Not5 will affect many different protein complexes besides RNA Pol II.
At the same time, Not5 in the nucleus, by being critical for the association of Rpb4 with mRNAs and Rpb4 cytoplasmic functions, will globally affect many different cellular components because Rpb4 has wide-spread roles in translation and mRNA decay. Not5 itself is a component of the same complex as the major yeast deadenylase, and it remains to be defined if the cytoplasmic functions determined for Rpb4 are mediated via its interaction with Not5 in the cytoplasm. In any event, taken together our work identifies Not5 as an essential cellular regulator connecting transcription, mRNA degradation and translation in eukaryotic cells.
The Saccharomyces cerevisiae strains used in this work are listed in Table S2. The plasmid encoding NLS-Not5 (pJG4-5-NOT5) has already been described [46]. To generate a plasmid expressing Myc-tagged Not5 we used pGREG516 [52] and inserted the RPS7A promoter and NOT5 coding sequences by the drag and drop technique leading to pMAC763. To fuse a NES (LALKLAGLDI) at the C-terminus of Not5 we digested pMAC763 with BsiWI and XhoI and amplified NOT5 sequences with a forward primer (TTT GCC TCA CCC AAC GTC AAT C) located prior to the BsiWI site and a reverse primer: (GAG GTC GAC TTA TAT GTC CAA ACC AGC TAA TTT AAG TGC TAA CAG TTT TTC GAA ATC TTC TTC AT) including a SalI site, a stop codon, the NES sequence and the end of the NOT5 ORF, digested this fragment with BsiWI and SalI and cloned it to the BsiWI and XhoI site of the digested pMAC763. The plasmid obtained was verified by sequencing.
Ribosomes were fractionated on a 12 ml 7–47% sucrose gradient as in [21]. To analyze comparable amounts of polysomes we applied 2 mg of wild-type and 4 mg of not5Δ extracts. The polysomes were disrupted by adding 25 mM EDTA or by 1 µg/ml RNaseA and incubation for 5 min at room temperature. RNasin Plus (Promega) at 0.2 unit/µl was added to stop the nuclease digestion. RNA was isolated from heavy polysome fractions by the Trizol reagent (Invitrogen) following the recommendations of the manufacturer, pellets were washed two times by 75% ethanol to remove sucrose, RNA concentration was measured by nanodrop.
These assays were performed as described [53], [54]. Relevant ORFs were amplified by PCR and cloned into pJG4-5 and pLex202. The relevant combination of plasmids was transformed into MY290.
Aggregated proteins from total extracts were isolated as previously described [21].
One hundred ml of cells grown exponentially to an OD600 of 0.8 were resuspended in 50 ml of medium lacking methionine and incubated for 15 min at 30°C with agitation. Cells were pelleted and resuspended in 5 ml of the same medium and then incubated in the presence of 5 µCi/OD600 unit of 35S methionine for 5 min at 30°C. To stop the labeling reaction, 25 ml of ice cold H2O with (for Rpb1 immunoprecipitation) or without (for Rpb2 purification) 100 µg ml−1 of CHX was added to the reaction. Both labeling experiments were done in biological duplicates. Cells were pelleted, resuspended in YPD with or without 100 µg ml−1 of CHX and placed at 30°C. 10 ml of cells were pelleted and frozen at 30 and 60 min for Rpb1 immunoprecipitation or at 60 and 120 min for Rpb2 purification, for extract preparation. 5 µl of total extracts at 2 mg/ml were TCA precipitated with 300 µl of ice-cold 25% TCA containing 2% of casamino acids for 30 min on ice. The precipitate was collected by vacuum filtering of 250 µl the TCA reaction mix on Whatman GF/A glass fiber filters. Amino acids were removed by rinsing the filter 3 times with 1 ml of ice-cold 5% TCA. For determination of 35S incorporation into translation products the filter was put into scintillation fluid (NOCS 104; Amersham) and counted in a Wallac 1409 liquid scintillation counter. The same amount of labeled protein from each extract in a volume of 800 µl was incubated with 0.5 µg anti-CTD antibody and magnetic Protein G beads (Invitrogen) for Rpb1 immunoprecipitation or directly with IgG beads for Rpb2 purification, that were pretreated with 200 µl of 5 mg/ml not5Δ total protein extracts in IP buffer (40 mM HEPES-KOH pH 8.0, 100 mM KCl, 150 mM potassium acetate, 1 mM EDTA, 10% glycerol and protease inhibitors) to saturate unspecific binding. For Rpb1 immunoprecipitation, after 3 washes beads were boiled with 50 µl of SB, and for Rpb2 purification the beads were washed additionally 3 times with TEV cleavage buffer and exposed to TEV cleavage as described below. Both preparations were analyzed by western blotting and by Coomassie staining. Coomassie stained gels were dried and revealed by Phosphorimager (Typhoon Phosphorimager 8600).
One hundred milliliter of cells collected at OD600 0.8–1.0 was treated with CHX (100 µg ml−1) for 10 min at 4°C and cells were harvested. Single-step affinity purification was done in the presence of CHX (100 µg ml−1) and 80 units/ml of RNase inhibitor (RNasine, Fermentas). RNase inhibitor and CHX were applied in similar concentrations for all the subsequent steps of the RIP. TEV cleavage was done in IP buffer (described above) for 1 h at 30°C. One fourth of the TEV eluate and 25 µg of total protein were subjected to western blotting to verify the affinity purification. 0.5 mg of total extracts and the rest of the TEV eluates were treated with phenol-chloroform for nucleic acid extraction. The nucleic acids were precipitated at -20°C with ethanol upon addition of sodium-acetate (100 mM) and 3 µl of linear acrylamide (Fermentas). Pellets were resuspended in H2O and were DNaseI treated (RQ1 RNase-free DNase, Promega). For RIPs, 500 ng of the TEV eluates and 500 ng of the RNA from the total extracts were reverse transcribed, for polysomal and total RNA comparisons 1 µg of RNA were reverse transcribed, with M-MLV RT (Promega) using oligo d(T) primers according to manufacturer's instructions. In performing the DNase and reverse transcription experiments we followed the manufacturer's instructions. qPCR primers were constructed to amplify approximately 200 bp long fragments close to the polyA tail of the mRNA. As a positive control we used Rpl17-TT to immunoprecipitate all translated mRNAs. Negative control was a wild-type strain without any protein tagged. We conducted qPCR on the reverse transcribed samples. For each 20 µl reaction, 9 µl first strand cDNA solution, 10 µl ABsolute qPCR SYBR Green Mix (ABgene), 0.5 µl forward primer (10 µM) and 0.5 µl reverse primer (10 µM) were mixed together. PCR parameters were as follows; 95°C, 10 min for heat activation of DNA polymerase mix, followed by 94°C, 15 s (denaturation); 60°C, 1 min (annealing and synthesis) for 38 cycles. Relative enrichment ratios and relative mRNA abundances were determined by the Pfaffl method [55], and normalized to the total RNA input in the case of the RIPs, and to wild-type RNA levels in the case of the polysome-total RNA comparisons in which NIP1 mRNA was used as a loading control. The primers used are: Rpb1 5′: GTC ACC AAG TTA CAG CCC AAC G; Rpb1 3′: AGA TCC TGG GCT GTA GCC TG; Nip1 5′: AGC TGA TGA GCG TGC TAG AC; Nip1 3′: AGG AAC GAC GAA TGG ATT TTG GAG.
Cells were grown to a log phase (OD600 of 0.5), fixed by adding 1 ml of 37% formaldehyde and incubating for 2 h at RT, then pelleted, washed with PBS (10 mM sodium phosphate buffer pH 7.4, 137 mM NaCl, 2.7 mM KCl) and resuspended in 0.5 ml of spheroplasting buffer (20 mM potassium phosphate pH 7.4, 1.2 M sorbitol). 0.2 ml of cell suspension was treated with 3.2 µl of 1.42 M β-mercaptoethanol and 5 µl of 5 mg/ml zymolyase 100T for 60 min at 30°C. Cells were washed once with 1 ml then resuspended in 100 µl of PBS +0.05% Tween 20. 20 µl of spheroplasts were placed on polylysine coated microscope slides and dried. Slides were washed three times with PBS and cells were blocked in 20 µl of PBS containing 1 mg/ml BSA for 30 min. 20 µl of primary antibody diluted 1∶200 in PBS with BSA was placed on the cells and incubated for 1 h at RT. Cells were washed three times with PBS and incubated for 1 h at RT with secondary antibody diluted 1∶1000 in PBS with BSA. After 3 washes with PBS, cells were treated with 20 µl of 1 µg/ml DAPI in PBS and washed again 3 times with PBS. Glass slides were mounted in 90% glycerol containing PBS and analyzed with fluorescent microscopy using an Axio Vert 200 device supplied with cooled CCD camera or with a Nikon Ti-E motorized inverted microscope system equipped with an Orca-Flash 4.0 Digital CMOS camera.
The Drosophila P-element insertion lines used for immunofluorescence were obtained from the Bloomington stock center (stock numbers: 15271 for CNOT3 and 34754 for RPB2). The trans-heterozygous flies were generated by crossing the two different P-element insertion lines. As a wild-type control we used the w1117 strain. For microscopy Drosophila ovaries were fixed in 4% paraformaldehyde. To block nonspecific staining embryos were incubated in 1% BSA (Sigma) in PBST (0.1% (v/v) Tween 20 in PBS) for 120 min at 4°C. Ovaries were incubated with the primary antibodies and 1% BSA in PBST. After several rinses in PBST, ovaries were incubated in secondary antibodies for 3 h at room temperature. To detect DNA, ovaries were stained with DAPI following incubation with the secondary antibody. Following several rinses in PBST ovaries were mounted in Aqua Poly Mount (Polysciences Inc). Optical sections were generated with an Olympus FV1000 confocal microscope.
For small scale tandem affinity purifications of RNA Pol II or R2TP 100 OD600 units of cells were broken with 0.5 ml of glass beads in 0.6 ml of IP buffer during 25 min at 4°C. Beads were washed with 0.5 ml of IP buffer. After clarification, 0.8 ml of the supernatants containing 4 mg of total protein treated when indicated with 1 µg/ml RNaseA for 5 min at room temperature were incubated with 40 µl of IgG sepharose beads (GE Healthcare). The beads were washed three times with 0.2 ml of IP buffer and then 3 times with 0.2 ml TEV buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Tween 20, 0.5 mM EDTA). Beads were incubated for 1 h at 30°C in 40 µl of TEV buffer containing 1 µM DTT and 1 unit of TEV protease (Invitrogen). Beads were sedimented and the supernatant was applied for native 3–12% Bis-Tris gel (Invitrogen) analysis or boiled with SDS sample buffer (SB) and separated by SDS-PAGE (4–12%) followed in both cases by western blotting.
RNA Pol II was single-step purified from 4 mg of total protein obtained from 100 OD600 units of cells. Eluates were concentrated to 25 µl and analyzed by Native PAGE 3–12% Bis-Tris gels (Invitrogen). From total extracts 25 µg proteins were analyzed by Native PAGE. When indicated total extracts were treated either by 1 µg/ml RNaseA for 5 min at RT, or by 20 units/ml of DNase I (New England BioLabs Inc.) for 10 min at 37°C. After Native PAGE samples were analyzed by western blotting.
Primary antibodies used for western blotting were anti CBP (Anti-Calmodulin Binding Protein; DAM1411288; Millipore) used at 1∶5000, anti HA (Anti influenza hemagglutinin; H3663; Sigma) used at 1∶5000, anti CTD which recognizes Rpb1 and will be referred to in the text as anti Rpb1 (ab5408; Abcam) used at 1∶500, or finally anti PAP (Peroxidase-Anti-Peroxidase; P1291; Sigma) used at 1∶10000 and rabbit polyclonal anti Ccr4 which was generated in our laboratory and used at 1∶5000. The secondary antibodies were anti-Mouse-HRP (IgG-Peroxidase conjugate; A9044; Sigma) used at 1∶10000 or anti-Rabbit-HRP (IgG-Peroxidase conjugate; A8275; Sigma) used at 1∶10000. For detection of Rpb1, ovaries were incubated with 7G5 mouse monoclonal antibody used at 1∶1000 (a kind gift of Dr. László Tora, Strasbourg).
|
10.1371/journal.ppat.1005955 | Vaccinia Virus Protein C6 Inhibits Type I IFN Signalling in the Nucleus and Binds to the Transactivation Domain of STAT2 | The type I interferon (IFN) response is a crucial innate immune signalling pathway required for defense against viral infection. Accordingly, the great majority of mammalian viruses possess means to inhibit this important host immune response. Here we show that vaccinia virus (VACV) strain Western Reserve protein C6, is a dual function protein that inhibits the cellular response to type I IFNs in addition to its published function as an inhibitor of IRF-3 activation, thereby restricting type I IFN production from infected cells. Ectopic expression of C6 inhibits the induction of interferon stimulated genes (ISGs) in response to IFNα treatment at both the mRNA and protein level. C6 inhibits the IFNα-induced Janus kinase/signal transducer and activator of transcription (JAK/STAT) signalling pathway at a late stage, downstream of STAT1 and STAT2 phosphorylation, nuclear translocation and binding of the interferon stimulated gene factor 3 (ISGF3) complex to the interferon stimulated response element (ISRE). Mechanistically, C6 associates with the transactivation domain of STAT2 and this might explain how C6 inhibits the type I IFN signalling very late in the pathway. During virus infection C6 reduces ISRE-dependent gene expression despite the presence of the viral protein phosphatase VH1 that dephosphorylates STAT1 and STAT2. The ability of a cytoplasmic replicating virus to dampen the immune response within the nucleus, and the ability of viral immunomodulators such as C6 to inhibit multiple stages of the innate immune response by distinct mechanisms, emphasizes the intricacies of host-pathogen interactions and viral immune evasion.
| In response to a viral infection, infected host cells mount an early, innate immune response to limit viral replication and spread. Type I interferons (IFNs) are produced by a cell when a viral infection is detected and are a crucial aspect of this early immune response. IFNs are released from the infected cell and can act on the infected cell itself or neighbouring cells to initiate a signalling pathway that results in the production of hundreds of anti-viral proteins. In this work we investigated a vaccinia virus protein called C6, a known inhibitor of type I IFN production. Here we show that C6 also inhibits signalling initiated in response to type I IFNs, therefore providing a dual defence against this essential immune response. The results show that, unlike the majority of viral inhibitors of IFN signalling, C6 inhibits the signalling pathway at a late stage once the proteins required for IFN-stimulated gene transcription have reached the nucleus and bound to the DNA. This work illustrates the complex relationship between infecting viruses and the host immune response and further investigation of the mechanism by which C6 inhibits this important immune pathway will likely increase our knowledge of the pathway itself.
| The innate immune response is the first line of defense against invading pathogens. Interferons (IFNs) are one of the key players in this early response to infection and are particularly important to protect against viruses, as can be seen by the increased susceptibility of IFNα/β receptor (IFNAR) knock out mice to viral infections [1]. There are two main branches to the IFN response; their production and the signalling initiated in response to the binding of secreted IFNs to their receptors at the cell surface.
Type I IFNs, which include IFNβ, several IFNα variants and other tissue or species-specific members, are produced directly in response to virus detection by cellular pattern recognition receptors (PRRs). Upon recognition of pathogen associated molecular patterns (PAMPs) such as viral DNA or RNA, PRRs activate several signalling pathways many of which converge on the kinases TANK-binding kinase (TBK1) and IκB kinase-ε (IKKε). These kinases, in complex with adaptor proteins such as TANK, NAK-associated protein 1 (NAP1) or similar to NAP1 TBK1 adaptor (SINTBAD), phosphorylate the transcription factor IFN regulatory factor 3 (IRF-3). Once phosphorylated, IRF-3 dimerises and translocates into the nucleus and, in combination with other transcription factors, drives transcription from promoters containing cognate binding sites, including the IFNβ promoter [2].
Once produced and secreted from cells, type I IFNs can act in a paracrine or autocrine fashion by binding to the IFNAR, which is composed of the two subunits IFNAR1 and IFNAR2. The binding of type I IFN to the receptor complex leads to the cross activation of the two Janus protein kinases, Tyk2 and Jak1 that are bound to the cytoplasmic domains of the IFNAR1 and IFNAR2, respectively. Once activated these kinases phosphorylate the transcription factors signal transducer and activator of transcription 1 (STAT1) and STAT2. These phosphorylated proteins then heterodimerise and bind to IRF-9 to form the IFN stimulated gene factor 3 (ISGF3) transcriptional activator complex. This tripartite complex translocates into the nucleus where it binds to IFN stimulated response elements (ISREs) found in the promoter of IFN stimulated genes (ISGs) and induces their transcription. The type I IFN signalling pathway and its regulation is reviewed in [3].
The importance of the IFN response for protection against viral infections is illustrated by the array of mechanisms and proteins used by viruses to evade and inhibit these cellular pathways, reviewed in [4]. Vaccinia virus (VACV) is a well-studied member of the Poxviridae and was the vaccine used in the eradication of smallpox [5]. It is a large DNA virus, with approximately 200 genes, that replicates exclusively in the cytoplasm of infected cells [6]. Between one third and one half of these 200 genes have been shown to have immunomodulatory or immunoevasive roles [7,8].
Many of these immunomodulatory proteins are able to inhibit type I IFN production, either through inhibition of the NF-κB pathway, for example VACV proteins B14 [9] and A49 [10], or through inhibition of the IRF-3 signalling pathway, as with VACV proteins A46 [11] and K7 [12]. In contrast, very few inhibitors have been identified that act post-IFN production. To date two VACV proteins are known to inhibit IFN signalling after type I IFN has been secreted from cells. B18 is a secreted VACV protein that binds type I IFN in solution and on the surface of cells and prevents its interaction with the IFNAR [13–15] and VH1 is a virally-encoded phosphatase that is packaged within virions [16] and dephosphorylates both STAT1 and STAT2, therefore acting as an intracellular inhibitor of JAK/STAT signalling [17,18].
C6 is a predicted member of the VACV Bcl2-like protein family, a family of 10 proteins whose previously studied members have various innate immune inhibitory functions [19–25]. It is expressed early during infection and its deletion attenuates the virus in both intranasal and intradermal models of infection in the mouse [26]. Despite being attenuated, VACV strains engineered to lack C6 showed enhanced immunogenicity in vivo [27,28]. Previously, C6 was shown to bind to the TBK1/IKKε adaptor proteins, SINTBAD, NAP1 and TANK, and prevents the TBK1/IKKε-dependent activation of IRF-3 and therefore inhibits the induction of type I IFNs [26]. Given many VACV proteins have been shown to have multiple functions, for example N1 that inhibits both NF-κB signalling [29] and apoptosis [23,30], and the observation that the hitherto only known function of C6 occurs in the cytoplasm of infected cells despite C6 being present in the nucleus and cytoplasm, we investigated whether C6 may have additional immunomodulatory functions.
In this study, VACV protein C6 is shown to be a dual function protein that inhibits type I IFN signalling as well as type I IFN production. Data presented show that the inhibition of IFNα-induced JAK/STAT signalling occurs at a late stage in the pathway, downstream of STAT phosphorylation, heterodimerisation, and nuclear translocation and downstream of ISGF3 binding to the ISRE, thus indicating an inhibitory function of C6 at or after formation of the transcriptional complex. Furthermore, C6 is shown to associate with the transcriptional activating domain (TAD) of STAT2, providing a plausible mechanism by which this viral protein could disrupt transcriptional complex formation. Interestingly, C6 was able to inhibit the transcriptional induction of all but one of the IFNα-dependent genes tested. This indicates that the step(s) inhibited by C6 downstream of ISGF3-ISRE interaction is likely to be conserved for a large number of ISGs, rather than a gene-specific transcriptional requirement. To our knowledge, this additional function of C6 makes C6 the first nuclear inhibitor of the IFN response encoded by the cytoplasmic-replicating DNA virus, VACV. The ability of VACV, a virus whose replication cycle takes place entirely within the cytoplasm of infected cells, to extend its influence into the host cell nucleus to inhibit crucial innate immune signalling pathways, highlights the complexity of virus-host interactions.
Previously, C6 was characterised as an inhibitor of IFNβ production through inhibition of the IRF-3/7 signalling pathway [26]. However, many viral proteins are known to have multiple functions. To determine whether C6 was able to also inhibit signalling downstream of type I IFNs, its effect on the expression of an ISRE-dependent luciferase reporter gene (ISRE-luciferase) was assessed. HeLa and HEK293T cells were co-transfected with expression plasmids for ISRE-luciferase and V5- or TAP- tagged C6 or control proteins (S1 Fig). Cells were then stimulated with IFNα for 8 h and the expression of luciferase was measured by luminescence. Treatment of HEK293T and HeLa cells with IFNα led to an induction of luciferase expression and in both cell types this induction was significantly inhibited by the co-expression of either TAP-C6 (p<0.0001 and p<0.01 respectively) or C6-V5 (p<0.001 and p<0.001 respectively, Fig 1A and 1B). As expected the positive controls Nipah Virus V protein (NiV-V) and Parainfluenza virus 5 V protein (PiV5-V) that are known to inhibit IFN signalling [31,32] also inhibited IFNα-induced luciferase expression, whilst expression of GFP or VACV protein B14 had no inhibitory effect.
To confirm the ability of C6 to inhibit type I IFN signalling, the effect of C6 on the induced transcription of endogenous ISGs was assessed next. HeLa cells stably expressing GFP alone (EV) or GFP in combination with V5 tagged- C6, PiV5-V or B14 were stimulated with IFNα for 8 h. RNA was extracted from these cells and used for qPCR analysis of ISG mRNA expression. IFNα treatment of cells resulted in induction of the well-characterised ISGs tested, including interferon-induced protein with tetratricopeptide repeats 1 (IFIT1), IFIT3 and MxA. The presence of C6 significantly inhibited the induction of gene expression when compared to both the EV and B14 controls (Fig 1C). Once again PiV5-V expression also inhibited the induction of ISG expression as expected (Fig 1C).
Finally, these stably transduced cells were used to confirm the ability of C6 to inhibit IFNα-induced gene expression at the protein level using flow cytometry analysis of IFIT1 expression. Cells were stimulated with IFNα for 8 h and then fixed, permeabilised and stained with an anti-IFIT1 antibody. Stained cells were analysed by flow cytometry to assess for IFIT1 protein expression. Once again treatment of EV transduced cells with IFNα resulted in an induction of IFIT1 expression, which was significantly inhibited by expression of C6 or the positive control, PiV5-V (p<0.01, Fig 1D). In contrast, expression of B14 had no effect on IFIT1 expression. Together these data show that C6 inhibits IFNα mediated gene expression at both the mRNA and protein level.
The IRF-3 signalling pathway is initiated in response to detection of viral RNA, DNA or proteins by PRRs in the cell and leads to the activation of the kinases TBK1 and IKKε. These kinases then phosphorylate IRF-3 causing its translocation into the nucleus from where it drives transcriptional activation of a number of target genes including IFNβ. C6 inhibits IRF-3 signalling at the level of TBK1 and IKKε, preventing the nuclear translocation of IRF-3 [26]. Several studies have described potential crosstalk between the IRF-3 and JAK/STAT signalling pathways, which together constitute the type I IFN response. In addition, IRF-3 is known to cause the transcriptional activation of a subset of ISRE containing gene promoters directly [33] and an additional phosphorylation event on STAT1 by IKKε is required for the full induction of approximately 30% of IFNα-responsive ISGs [34].
To rule out that the inhibitory effect of C6 on type I IFN signalling was an indirect consequence of its ability to interfere with TBK1 and IKKε function, and that this inhibitory activity was instead a novel function for C6, the effect of a TBK1/IKKε specific inhibitor, BX795, on the ISRE-dependent reporter gene assay was assessed. Cells transfected with the ISRE-luciferase reporter were treated for 3 h with BX795 before stimulation with IFNα for a further 6 h in the continued presence of BX795. BX795 treatment had no inhibitory effect on the IFNα-dependent induction of firefly luciferase in the cells tested, when compared to a carrier (dimethyl sulphoxide, DMSO)-treated control (S2 Fig). To confirm the effectiveness and specificity of BX795 in this assay, BX795 treatment was also used in conjunction with firefly luciferase under the control of promoters responsive to the IRF-3 and NF-κB signalling pathways. Cells transfected with these luciferase reporters were stimulated with polyI:C and IL-1β respectively. Whilst treatment with BX795 had no effect on IL-1β-induced activation of the NFκB reporter gene, the induction of the IRF-3-responsive promoter by poly I:C was significantly reduced (p<0.001), indicating that the inhibitor was effective at the doses used (S2 Fig). These data indicate that inhibition of TBK1 and IKKε does not cause the inhibition of ISRE-dependent gene expression that is seen in the presence of C6, implying that this is instead a novel function of C6.
To determine how C6 inhibits the cellular response to IFNα, the phosphorylation status of STAT1 and STAT2 in cells expressing C6 following IFNα treatment was examined. HeLa cells stably expressing C6 or control proteins were stimulated with IFNα for 45 mins. Cell lysates were used for immunoblot analysis of levels of phosphorylated STAT1 and STAT2, including quantification of proteins using Odyssey software (LICOR). Neither C6 nor the negative control protein B14 inhibited the IFNα-induced phosphorylation of STAT1 or STAT2, whereas PiV5-V protein inhibited the phosphorylation of both proteins as expected (Fig 2).
Following phosphorylation, STAT1 and STAT2 heterodimerise and bind to IRF9 to form the ISGF3 complex. This complex then translocates into the nucleus where it drives expression of genes with an ISRE in their promoter. To determine whether C6 prevents the dimerisation of STAT1 and STAT2, an immunoprecipitation of STAT1 from IFNα-stimulated cells was performed in cells transiently transfected with C6 or N1, a VACV Bcl-2-like protein that does not inhibit IFNα signalling. Immunoblot analysis showed that following IFNα stimulation an increased amount of STAT2 was found associated with STAT1. However, C6 did not alter the amount of STAT2 bound to STAT1 either before or after stimulation (Fig 3).
Following its formation, the ISGF3 complex translocates into the nucleus. To assess whether this nuclear translocation was inhibited by C6 the localization of STAT1 and STAT2 before and after IFNα stimulation was assessed by confocal microscopy. HeLa cells stably expressing GFP alone (EV) or in combination with V5-tagged C6 or PiV5-V were stimulated with 500 U/ml IFNα for 1 h. Cells were then fixed, permeabilised and stained for endogenous STAT1 (Fig 4A) or STAT2 (Fig 5A). Following IFNα stimulation the percentage of cells showing a nuclear stain for STAT1 and STAT2 increased to approximately 50% (Fig 4B) and 70% (Fig 5B), respectively. These localization patterns were not altered by C6 or the negative control, EV. The localisation of STAT1 could not be assessed in PiV5-V expressing cells due to its degradation by this viral protein, however, PiV5-V expression inhibited the translocation of STAT2 completely (Fig 5A). Together these data indicate that C6 does not inhibit the pathway prior to ISGF3 complex formation and nuclear translocation.
To obtain additional evidence that the inhibitory effect of C6 on IFNα signalling is downstream of ISGF3 complex formation, a plasmid encoding IRF-9 fused to the C-terminal region (amino acids 747–851) of the transcriptional activation domain of STAT2 (referred to as IRF9-S2C) was utilised. Previously, this fusion protein, which overcomes the need for ISGF3 complex formation, has been shown to act as a constitutively active ISGF3-like transcriptional activator in the absence of IFN stimulation [35]. When a plasmid encoding IRF9-S2C was co-transfected into HEK293T cells along with the ISRE-luciferase reporter gene, a large increase in firefly luciferase expression was observed in cells expressing IRF9-S2C relative to those transfected with empty vector (EV) only (Fig 6, columns 1 and 2). Interestingly, when co-transfected into cells, C6 inhibited IRF9-S2C-driven ISRE reporter activity significantly (p<0.0001), whereas, neither PiV5-V protein nor NiV-V protein showed any inhibitory activity (Fig 6). This is in keeping with the known ability of these two viral proteins to inhibit the IFNα signalling pathway upstream of ISGF3 complex formation [31,32]. These data confirm that C6 inhibits IFNα signalling at a late stage following ISGF3 complex formation.
To establish whether C6 inhibited IFNα signalling by preventing the binding of the ISGF3 complex to the ISRE in promoters of ISGs, the ability of the IRF9-S2C fusion protein to bind the ISRE was assessed. HEK293T cells were transfected with plasmids expressing IRF9-S2C and V5-tagged C6 or control proteins and cell lysates were harvested 16 h later. A biotin-labelled ISRE probe optimised previously for ISGF3 binding (ISREcore) [36], or a control biotin-labelled ISRE sequence that was shown to lack ISGF3 binding (ISRErandom) [36] were incubated with cell lysates and streptavidin beads were used to immunoprecipitate the biotinylated DNA probe and associated proteins. C6 or GFP expression did not inhibit the binding of either IRF9-S2C or endogenous STAT2 to the biotin-labelled ISRE (Fig 7). In contrast, NiV-V protein was able to inhibit endogenous STAT2 binding to the ISRE but not binding of the IRF9-S2C construct as expected (Fig 7). Neither IRF9-S2C nor endogenous STAT2 bound to the ISRErandom control sequence as expected. Ku70, a known DNA binding protein, was used here as a control for DNA input and gel loading and was found to bind to both the ISREcore and ISRErandom DNA probes.
To investigate whether C6 interacts with any of the components of the ISGF3 complex, an immunoprecipitation assay using FLAG-tagged STAT1, STAT2 and IRF-9 was performed. Plasmids expressing these proteins were co-transfected into HEK293T cells along with a V5-tagged C6 expression vector. Immunoprecipitation with anti-FLAG beads co-precipitated C6 with STAT2 but not with STAT1 or IRF-9 (Fig 8A).
To determine if the interaction between C6 and STAT2 was affected by IFN stimulation, HeLa cells were transfected with plasmids expressing HA-tagged STAT2 and either TAP-tagged C6 or N1 then mock-stimulated or stimulated with IFNα prior to immunoprecipitation (Fig 8B). This showed that the interaction between C6 and STAT2 did not require prior stimulation with IFN.
To confirm the interaction between C6 and STAT2 and to ascertain whether it occurs at endogenous protein levels and during viral infection, HEK293T cells were infected with VACVs expressing TAP-tagged C6 or N1 under the natural promoters for these genes. Cells were then lysed and immunoprecipitations were performed against the FLAG epitope in the TAP tag of these viral proteins. Once again C6 associated with STAT2 and not STAT1, whilst N1 did not associate with either protein (Fig 8C), confirming the specific interaction between STAT2 and C6 at endogenous protein levels during viral infection.
As C6 inhibits IFNα signalling initiated by IRF9-S2C expression, its ability to interact with this fusion protein, which contains only the C-terminal 104 aa of the STAT2 transactivation domain, was assessed. To this end, HEK293T cells were co-transfected with either IRF9-S2C or HA-IRF-9 and C6-TAP or N1-TAP. Immunoprecipitation with anti-FLAG beads showed association between IRF9-S2C and C6 but not between C6 and HA-IRF-9 (Fig 8D). This indicates that C6 associates with the final 104 aa of the STAT2 transactivation domain, a region known to be important for recruitment of downstream chromatin modifying enzymes and transcriptional machinery.
Finally, the biological importance of C6 for inhibition of the JAK-STAT pathway leading to activation of the ISRE promoter was assessed during VACV infection. The activity of C6 in blocking this pathway was likely to be masked to some degree during infection by the presence of the virus phosphatase VH1, which dephosphorylates STAT1 and STAT2 and is delivered into cells by the invading virion immediately after infection. Therefore, two methods were used to assess if C6 contributed to the inhibition of the JAK-STAT pathway during infection. One method was simply to transfect the ISRE-luciferase reporter plasmid into cells 16 h before the cells were infected with either wt VACV or the vΔC6 mutant and then measure ISRE-luciferase at different times p.i. Preliminary experiments established that infection of cells at 5 PFU/cell for 5 h was optimal before luciferase activity was measured in cell lysates. Under these conditions, infection by vΔC6 induced significantly greater luciferase activity than did wt VACV. Immunoblotting for VACV protein D8 showed that the virus infections were equivalent and immunoblotting for GAPDH showed equal loading of samples. This experiment was conducted using either crude or purified virus preparations (n = 4) and in each case a significant difference between the viruses was observed (Fig 9A).
The second method exploited the ability of the IRF9-S2C protein to activate ISRE promoter within the nucleus and downstream of the position at which the VH1 phosphatase mediates inhibition of the JAK-STAT pathway. Optimisation experiments to determine the amount of the IRF9-S2C plasmid to transfect and the length of time after transfection prior to virus infection showed that this potent inducer of ISRE-dependent gene expression was best only transfected a few hours before infection because overnight transfection induced very high levels of luciferase activity. Transfection of the IRF9-S2C plasmid for 4 h induced a modest 2-fold induction in luciferase activity and this was increased further by virus infection (Fig 9B). However, following virus infection for 5 h, wt VACV induced lower levels of luciferase than vΔC6, as in Fig 9A. These significant differences were seen reproducibly in multiple experiments (n = 4). Immunoblotting confirmed equal infection and protein loading.
Collectively, these data show that during VACV infection protein C6 is able to diminish expression from the ISRE promoter over and above the effect of the VH1 phosphatase.
Previously, C6 was identified as a VACV immunomodulator and virulence factor and was shown to inhibit the induction of type I IFNs through inhibition of the IRF-3/7 signalling pathway at the level of the TBK1/IKKε kinase complex [26]. This study identifies a second function for VACV protein C6 as an inhibitor of the cellular response to type I IFN. Data presented demonstrate that C6 inhibits IFNα-induced expression of ISGs at both the mRNA and protein level (Fig 1). The inability of a pharmacological inhibitor of TBK1/IKKε to reduce IFNα-induced reporter gene expression (S2 Fig) indicates that inhibition of this kinase complex by C6 is unlikely to explain the ability of C6 to also inhibit the cellular response to IFNα. Therefore, to elucidate the mechanism by which C6 has its inhibitory effect on this second pathway, the IFNα-induced phosphorylation and nuclear translocation of STAT1 and STAT2 were examined and C6 was found to have no effect on these early events of this signalling pathway (Figs 2, 4 and 5). Furthermore, both endogenous STAT2 and a constitutively active ISGF3 mimic, IRF-9-S2C, were still able to bind to the ISRE in the presence of C6 (Fig 7), indicating C6 exerts its inhibitory effect after ISGF3 binding to the ISRE. Interestingly, C6 interacts with STAT2 (Fig 8A–8C) and the transactivation domain (aa 747–851) of STAT2 fused to IRF-9 (Fig 8D) but not with STAT1 or IRF-9 (Fig 8A). The STAT2 transactivation domain is known to be required for the recruitment of chromatin modifiers and transcriptional machinery [37]. Therefore, the ability of C6 to interact specifically with this domain gives insight into how this viral protein may inhibit this crucial signalling pathway at such a late stage.
C6 is one of many VACV proteins that inhibit the IFN response, however, it is the first such protein known to inhibit both branches of the IFN response, inhibiting both IFNβ production and the cellular responses to type I IFN. C6 is also the first VACV protein identified to inhibit the response to type I IFN in the nucleus of infected cells; VACV inhibitors to date act early in the JAK/STAT signalling pathway, either extracellularly to prevent binding of secreted type I IFN to their receptor (B18) [13–15], or in the cytoplasm of infected cells to dephosphorylate activated STAT1 and STAT2 (VH1) [17,18]. C6 instead acts at a very late stage in the pathway, after the ISGF3 complex has formed, translocated into the nucleus and bound to the ISRE. Despite this, its inhibitory action was evident on 6 out of 7 of the ISGs examined, suggesting that the protein or step it targets is required for the induction of many ISGs.
The requirement for many protein inhibitors of a single, albeit important, signalling pathway is not well understood. However, VACV shows a similar ‘belt and braces’ approach to other signalling pathways, for example the NF-κB pathway, for which it is currently known to possess 10 inhibitors [9–12,23,29,30,38–42]. These proteins are not completely redundant however, as they cause virus attenuation in vivo when deleted individually [10,11,39,43–47]. Similarly, previous work has shown that deletion of C6 leads to an attenuated phenotype in both intradermal and intranasal models of VACV infection in mice [26], despite the presence of other IFN-signalling inhibitors. The identification of a second function of C6 means that the observed attenuation of the C6 deletion virus cannot be attributed to a single function as yet, but highlights the importance of this immunomodulatory protein. Structure-based mutagenesis of C6 may enable the dissection of these different activities as was done for the related VACV protein N1 [30].
The use of multiple proteins to inhibit a single pathway may be explained in many ways, such as the possibility of incomplete inhibition by any one protein, or the requirement for different immunomodulators in different cell types or infection stages. It could also be explained by possible crosstalk between innate immune signalling pathways meaning that inhibition of a pathway at a certain point could be overcome by activation of another communicating immune signalling pathway. However, the late stage at which C6 inhibits the type I IFN response would suggest that cross talk from other pathways would still be unable to activate ISGF3-driven ISG transcription in the presence of C6. The role of C6 in inhibiting the JAK-STAT pathway during virus infection may be masked to some degree by the effect of VH1 that dephosphorylates STAT1 and STAT2 rapidly after infection. VH1 is an essential gene for VACV replication [16], preventing its elimination by genetic manipulation. Nonetheless analysis of ISRE-driven gene expression early after infection with either wild type virus or a mutant virus lacking C6 showed a functional role for the C6 protein in diminishing ISRE-dependent gene expression (Fig 9). The roles of VH1 and C6 are therefore complementary. VH1 is expressed late in infection and packaged into the virion, whereas C6 is an early VACV protein, expressed from approximately 2 h post infection. These differential expression patterns might explain the requirement of both inhibitors, perhaps with their relative importance differing over the life cycle of the virus.
The principal transcriptional activating complex responsible for type I IFN-induced gene expression, ISGF3, consists of three components, STAT1, STAT2 and IRF-9. In this complex, STAT2 provides a potent and essential transcriptional activation domain [48]. The mechanistic process by which ISGF3, likely through this C-terminal domain of STAT2, signals to and promotes transcription of ISGs by RNA polymerase II remains unclear. However, a number of cellular proteins known to have roles in transcription, such as components of the Mediator complex [49], or chromatin modification, including histone deacetylases (HDACs) [50], histone acetyltransferases (HATs) [51] and chromatin remodeling complexes [52–55], have been identified as being essential for ISGF3-driven transcription. The precise mechanism by which C6 inhibits JAK/STAT signalling remains to be determined, however it is possible that the association of C6 with the STAT2-transactivation domain could prevent or alter the interactions of STAT2 with these or other cellular proteins required for ISG transcriptional induction. Indeed some such proteins have been shown to bind directly to the STAT2 transactivation domain, for example the HATs p300/CBP [51] and GCN5 [56]. In the future, further work is needed to assess the exact consequence of the C6-STAT2 TAD interaction.
The importance of the IFN-signalling pathway in preventing viral replication and spread dictates that the majority of mammalian viruses have one or multiple mechanisms of inhibiting the response of infected cells to type I IFN. There is an array of different mechanisms by which viral proteins inhibit JAK/STAT signalling, some of which are reviewed in [4]. Many such mechanisms focus on inhibiting the early steps in the JAK/STAT signalling pathway, by degradation of either STAT1 or STAT2 as with PiV5 [32] and PiV2 V proteins [57], respiratory syncytial virus NS1 and NS2 proteins [58] and dengue NS5 protein [59], by inhibition of STAT phosphorylation as with Sendai virus C protein [60], or by cytoplasmic sequestration of the ISGF3 complex as with NiV [31] and Hendra virus [61] V proteins.
Fewer viral proteins have been identified that inhibit the later stages of this signalling pathway, once the ISGF3 complex has reached the nucleus. The human cytomegalovirus (HCMV) IE1 protein interacts with STAT2 and inhibits the binding of STAT2 and promyelocytic leukemia protein (PML), a protein that associates with STAT1, STAT2, and HDACs, to ISG promoters [62]. Therefore, HCMV IE1 delivers its inhibitory action within the nucleus and via an interaction with STAT2, but again upstream of the inhibitory action of C6. Conversely, adenovirus E1A protein has its inhibitory effect, like C6, downstream of ISGF3- promoter binding but does so through interacting with and preventing the functioning of HATs and histone ubiquitylating complexes required for full ISG transcriptional activation and not through a direct interaction with STAT2 [63]. Similarly, influenza A virus nonstructural protein 1 (NS1) inhibits the cellular response to type I IFNs in the nucleus through interaction with a complex involved in transcriptional elongation, hPAF1C and once again not by a direct interaction with STAT2 [64]. Interestingly, NS1 inhibits type I IFN production through binding dsRNA produced by the virus and thus preventing its detection by PRRs [65–67]. NS1 and C6 have therefore both evolved to inhibit both type I IFN production and type I IFN-induced signalling but by distinct mechanisms in each pathway.
To our knowledge VACV protein C6 is the first viral protein shown to both associate with the STAT2 transactivation domain and inhibit IFNα-dependent ISG induction after ISGF3 binding to the ISRE. The sequence of events that occur following ISGF3 binding to the ISRE is poorly understood and further elucidating the mechanism by which C6 inhibits this signalling pathway in the nucleus may enhance our knowledge of the late stages of the type I IFN signalling pathway. Lastly, it is notable that although VACV is a cytoplasmic DNA virus, it has evolved mechanisms to inhibit IFN production or activity within both the cytoplasm and the nucleus and indeed outside the infected cell by the expression of soluble type I and type II IFN binding proteins.
HEK293T (ATCC CRL-11268) cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM, Invitrogen) with 10% heat treated (56°C, 1 h) foetal bovine serum (FBS, Seralab) and penicillin/streptomycin (P/S, 50 μg/ml, PAA laboratories). HeLa (ATCC CCL-2) cells were maintained in Minimum Essential Medium (MEM, Invitrogen) supplemented with 10% FBS, P/S and 1:100 non-essential amino acids (Gibco).
HeLa cells stably expressing GFP only or in combination with V5-tagged C6 (V5-C6), V5-Parainfluenza virus 5 V protein (V5-PiV5-V) or V5-tagged B14 (V5-B14), were obtained after transduction of cells with lentiviruses (see below) and sorting to obtain GFP-positive cells in a MoFlo MLS high-speed cell sorter (Beckman Coulter). For each protein two populations were sorted based on GFP expression level; the top 30% of GFP-expressing cells (high expressers) and the next 30% (middle expressers). Immunoblot analysis was used to determine the expression levels of the protein of interest in these two populations. Based on the observed expression of these proteins, the high GFP-expressing populations were chosen for, EV (GFP only), V5-C6, and V5-PiV5-V and the middle GFP-expressing V5-B14 population. For protein expression level of cell lines see S1C Fig. Lentivirus particles for transduction were generated after transient co-transfection of HEK293T cells with entry and packaging vectors and the bicistronic genomic vector encoding GFP and the appropriate V5 tagged protein using PEI (CellnTec).
The recombinant VACV Western Reserve strain C6-TAP and N1-TAP viruses were described [68]. The tandem-affinity purification (TAP) tag used here and in plasmids described below contains 2 copies of the streptavidin-binding sequence and 1 copy of the FLAG epitope [69]. Wild type (wt) VACV strain WR and the deletion mutant lacking the C6L gene were described [26].
Antibodies used were from the following sources; Rabbit (Rb) anti-FLAG (Sigma-Aldrich, F7425, diluted 1:5000), Mouse (Ms) anti-V5 (AbD Serotec Ltd, MCA1360, diluted 1:5000), Rb anti-HA (Sigma Aldrich, H6908, diluted 1:1000), Ms anti-α-tubulin (Millipore, 05–829, diluted 1:5000), Rb anti-actin (Sigma, A2066, diluted 1:1000), Ms anti-Phospho-STAT1 (Invitrogen, 333400, diluted 1:750), Rb anti-phospho-STAT2 (Millipore, 07–224, diluted 1:1000), Rb anti-STAT1 for immunofluorescence (Millipore, 06–501, diluted 1:300), Rb anti-STAT1 for western blotting (Cell signalling, 9172S, diluted 1:1000), Rb anti-STAT1 for immunoprecipitation (Santa Cruz, sc-345, diluted 1:100), Rb anti-STAT2 (Santa Cruz, sc-476, diluted 1:100 for immunofluorescence and 1:500 for western blotting), Ms anti-Ku70 (Abcam, ab3114, diluted 1:1000), Ms anti-IFIT1 for flow cytometry (Abcam, ab70023, 1:500), and Alexa Fluor 546 goat anti-Rb IgG (H+L) (Invitrogen, A-11010, diluted 1:750 for immunoblotting) or Alexa Fluor 647 donkey anti-Ms IgG (H+L) (Invivogen, A- 31571, diluted 1:2000 for flow cytometry).
Reagents used in this study were BX795 (Tocris), poly(I:C) (InvivoGen), Protein G Sepharose 4 Fast Flow (GE Healthcare), High Capacity Streptavidin Agarose Resin (Thermo Scientific), human IFNα and human IL-1β were from Peprotech, poly(dI:dC) and ANTI-FLAG M2 Affinity Gel were from Sigma Aldrich. Biotinylated DNA for immunoprecipitations were synthesised by Integrated DNA technologies. The sequences were ISREcore; TGCCTCGGGAAACCGAAACTGAAGCCA and ISRErandom ACTGATCGGAAACCGAAACGATCTATG. These sequences were taken from [36].
Codon-optimised TAP-C6 and N1-TAP, were described previously [68] and [30]. B14-TAP was kindly provided by Dr. Brian Ferguson (Department of Pathology, University of Cambridge, UK). The sequence of PiV5-V and NiV-V were amplified by PCR from plasmids kindly provided by Prof. Richard Randall (University of St Andrews, UK) and then subcloned into mammalian expression vectors pcDNA3.1 (Invitrogen) with an N-terminal V5 tag. V5-C6 was produced by PCR amplification of C6 from VACV WR DNA and cloned into pcDNA3.1. GFP-V5 was provided by Dr. Christian Ku (Department of Pathology, University of Cambridge). The sequences of IRF-9, STAT1 and STAT2 were amplified by PCR from HeLa cDNA and subcloned into mammalian expression vector pcDNA4/TO with a C-terminal TAP tag and/or vector pcDNA3.1 with a N-terminal HA tag. pcDNA3 IRF9-STAT2C was a gift from Prof. Curt Horvath (Addgene plasmid 37544) [35]. pcDNA4/TO (Invitrogen) was used in luciferase reporter assays as EV. ISRE-luciferase, NF-κB-Luciferase, and TK renilla were obtained from Dr. Andrew Bowie (Trinity College, Dublin, Ireland), and ISG56.1-Luciferase was from Ganeth Sen (Lerner Research Institute, Ohio, USA).
Reporter gene assays were performed in HeLa or HEK293T cells seeded in 96-well plates. Cells were transfected with 100 ng firefly luciferase reporter plasmid, 10 ng GL3-renilla luciferase plasmid and 100 ng of expression plasmid for the protein of interest. For the BX795 reporter gene assay only the firefly report plasmid and GL3-Renilla plasmids were transfected. For the IRF9-S2C reporter gene assays 100 ng firefly reporter plasmid and 10 ng GL3-Renilla plasmid were transfected along with 50 ng IRF9-S2C and 50 ng C6 expression vector or control plasmids, except in the empty vector only control where 100 ng pcDNA4 was transfected only. Transit-LT1 (Mirus, 2 μl per 1 μg DNA) was used for transfection of HeLa cells and PEI (CellnTec, 2 μl per 1 μg DNA) for HEK-293T cells. Sixteen hours post transfection cells were stimulated as indicated in the figure legends. Cells were harvested in passive lysis buffer (Promega, 100 μl/well). The firefly-luciferase readings of each sample were normalised to the renilla-luciferase readings and fold inductions were calculated relative to the non-stimulated controls for each plasmid. Experiments were performed in triplicate and conducted at least 3 times.
HeLa cell lines stably expressing the proteins of interest were grown in 12-well plates and RNA was extracted using the RNeasy kit (QIAGEN). One μg of each RNA sample was used to synthesise cDNA using Superscript III reverse transcriptase according to the manufacturer’s protocol (Invitrogen). ISG mRNA was quantified by real-time PCR using a ViiA 7 Real-Time PCR System (Life Technologies), fast SYBR Green Master Mix (Applied Biosystems) and the following primers, IFIT1 (Fwd: CCTGAAAGGCCAGAATGA GG, Rev: TCCACCTTGTCCAGGTAAGT) IFIT3 (Fwd: ACACAGAGGGCAGTCATGAGTG, Rev: TGAATAAGTTCCAGGTGAAATGGC) MxA (Fwd: ATCCTGGGATTT TGGGGCTT, Rev: CCGCTTGTCGCTGGTGTCG) GAPDH Fwd: ACCCAGAAGACTGTGGATGG, Rev: TTCTAGACGGCAGGTCAGGT). Amplification of ISGs was normalised to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) amplification from the same sample, and the fold induction of genes in response to IFNα was calculated relative to the unstimulated control of the cell line. Experiments were performed in biological triplicate and conducted three times.
HeLa cells stably expressing the proteins of interest were grown in 6-cm dishes and stimulated with 500 U/ml IFNα or mock-stimulated for 8 h. Cells were removed from the dishes by addition of trypsin (GIBCO), transferred to Eppendorf tubes and washed twice with ice-cold PBS. Cells were fixed in 4% paraformaldehyde and permeabilised with 0.1% Triton X. Cells were then incubated with anti-IFIT1 (Abcam, ab70023) in Triton buffer (0.5% BSA, 0.02% sodium azide, 0.1% Triton X-100 in PBS) for 1 h at 4°C. Cells were washed twice with Triton buffer and then incubated with Alexa 647 Donkey anti-Mouse (Invivogen) in Triton buffer for 1 h in the dark at room temperature. Cells were washed twice in Triton buffer, once in 0.5% BSA in PBS and then analysed on a CyAn ADP Analyser (Beckman coulter). Collected data were analysed using Summit (Beckman Coulter).
HeLa cells stably expressing the proteins of interest were grown on glass coverslips in 6-well plates. Cells were stimulated with 1000 U/ml IFNα for 1 h. Cells were fixed with 4% paraformaldehyde. Auto-fluorescence was quenched in 150 mM ammonium chloride in PBS and the cells were then permeabilised in 0.1% Triton X-100 in PBS. Cells were incubated in blocking buffer (0.5% BSA in PBS) for 30 min, stained with primary antibody for 1 h (STAT1: 1:300, STAT2 1:100 in blocking buffer) and for 1 h with secondary antibody (Alexa 546, Invitrogen). Coverslips were mounted onto microscope slides in Mowiol 4–88 containing 4',6-diamidino-2-phenylindole (DAPI). Slides were visualised and imaged using a Zeiss LSM 780 Confocal microscope. Images were viewed using LSM Image Browser (Zeiss).
HEK293T cells were grown in 10-cm dishes and transfected with the constructs outlined in the figure legends using either Transit LT1 (Mirus) or calcium phosphate transfection. Sixteen hours later cells were stimulated with IFNα or mock-treated as described in figure legends, then lysed in lysis buffer (150 mM NaCl, 20 mM Tris-HCl pH 7.4, 10 mM CaCl2, 0.1% (v/v) Triton-X, 10% (v/v) glycerol and protease (cOmplete Mini, Roche) and phosphatase inhibitors (PhosSTOP, Roche)) and cleared by centrifugation. Samples were then incubated with 30 μl Protein G Sepharose 4 Fast Flow (GE Healthcare) and anti-STAT1 (Santa Cruz, sc-345) for 6 h, or ANTI-FLAG M2 Affinity Gel (Sigma Aldrich) or Anti-HA Agarose (Sigma Aldrich) and 2 h. Immunoprecipitations were washed 3 times in lysis buffer and bound proteins were eluted by boiling in buffer containing SDS. Samples were then analysed by SDS-PAGE (polyacrylamide gel electrophoresis) and immunoblotting with the stated antibodies.
HEK293T cells were grown in 10-cm dishes and co-transfected with expression plasmids for IRF9-S2C, and V5-NiV-V, V5-C6 or V5-GFP using calcium phosphate transfection in triplicate for each condition. Sixteen hours later cells were lysed in lysis buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP40, 5% glycerol and protease (cOmplete Mini, Roche) and phosphatase (PhosSTOP, Roche) inhibitors). Lysates were incubated firstly with 10 ng/ml poly(dI:dC) for 30 min, then with 100 pmol biotin-labelled ISREcore or biotin-labelled control DNA for 1.5 h and finally with 30 μl High Capacity Streptavidin Agarose Resin (Thermo Scientific) for 3.5 h. Immunoprecipitations were washed four times with lysis buffer and proteins were eluted by boiling in buffer containing SDS. Samples were then analysed by SDS-PAGE and immunoblotting with the stated antibodies.
Un-paired student’s T-tests were used to analyse data, with Welch’s correction applied when variances differed significantly between samples. Statistical significance is expressed as: *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.
|
10.1371/journal.ppat.1004305 | Diverse Host-Seeking Behaviors of Skin-Penetrating Nematodes | Skin-penetrating parasitic nematodes infect approximately one billion people worldwide and are responsible for some of the most common neglected tropical diseases. The infective larvae of skin-penetrating nematodes are thought to search for hosts using sensory cues, yet their host-seeking behavior is poorly understood. We conducted an in-depth analysis of host seeking in the skin-penetrating human parasite Strongyloides stercoralis, and compared its behavior to that of other parasitic nematodes. We found that Str. stercoralis is highly mobile relative to other parasitic nematodes and uses a cruising strategy for finding hosts. Str. stercoralis shows robust attraction to a diverse array of human skin and sweat odorants, most of which are known mosquito attractants. Olfactory preferences of Str. stercoralis vary across life stages, suggesting a mechanism by which host seeking is limited to infective larvae. A comparison of odor-driven behavior in Str. stercoralis and six other nematode species revealed that parasite olfactory preferences reflect host specificity rather than phylogeny, suggesting an important role for olfaction in host selection. Our results may enable the development of new strategies for combating harmful nematode infections.
| Parasitic worms are a significant public health problem. Skin-penetrating worms such as hookworms and the human threadworm Strongyloides stercoralis dwell in the soil before infecting their host. However, how they locate and identify appropriate hosts is not understood. Here we investigated the host-seeking behavior of Str. stercoralis. We found that Str. stercoralis moves quickly and actively searches for hosts to infect. We also found that Str. stercoralis is attracted to human skin and sweat odorants, including many that also attract mosquitoes. We then compared olfactory behavior across parasitic worm species and found that parasites with similar hosts respond similarly to odorants even when they are not closely related, suggesting parasitic worms use olfactory cues to select hosts. A better understanding of host seeking in skin-penetrating worms may lead to novel control strategies.
| Skin-penetrating nematodes such as the threadworm Str. stercoralis and the hookworms Ancylostoma duodenale and Necator americanus (Figure 1A) are intestinal parasites that infect approximately 1 billion people worldwide. Infection with skin-penetrating worms can cause chronic gastrointestinal distress as well as stunted growth and long-term cognitive impairment in children. Moreover, Str. stercoralis infection can be fatal for immunocompromised individuals and infants [1]. Str. stercoralis is endemic in tropical and sub-tropical regions throughout the world, including the United States, and is estimated to infect 30–100 million people worldwide [2]. Infection rates in rural and semi-rural areas are often high, particularly among children. For example, a recent study found that 25% of school children in semi-rural Cambodia were infected with Str. stercoralis [3]. A better understanding of how skin-penetrating worms target human hosts could lead to new strategies for preventing infections.
Skin-penetrating nematodes are infective only during a particular stage of their life cycle called the infective juvenile (IJ), a developmentally arrested third larval stage analogous to the C. elegans dauer [4]. IJs inhabit the soil and infect by skin penetration, often through the skin between the toes. Inside the host, IJs migrate through the circulatory system to the lungs, are coughed up and swallowed, and develop to adulthood in the intestine [1]. IJs may also reach the intestine using other migratory routes [5]. Adult nematodes reproduce in the intestine, and eggs or young larvae are excreted in feces. In the case of hookworms, young larvae develop into IJs, which find and infect new hosts (Figure 1B). In the case of Strongyloides species, some larvae develop into IJs and others develop into free-living adults. In the human parasite Str. stercoralis and the rat parasite Str. ratti, which are subjects of this study, all progeny of free-living adults develop into IJs (Figure 1C). Some species of Strongyloides, such as the dog and cat parasite Str. planiceps, can undergo a limited number of sequential free-living generations [6]. Thus, Strongyloides can develop through at least one free-living generation outside the host. Str. stercoralis can also cycle through multiple parasitic generations in the same host, resulting in a potentially fatal disseminated infection [1].
Little is known about the process by which skin-penetrating nematodes find hosts [7]. IJs of some skin-penetrating species respond to heat and sodium chloride [8]–[12], suggesting a role for thermosensation and gustation in host seeking. In addition, Str. stercoralis is attracted to human blood serum and sweat [10], [12], while Str. ratti is attracted to mammalian blood serum [13]. It has long been speculated that olfaction may be important for host seeking, since animals emit unique odor blends that could confer species-specificity [7]. However, the only specific odorant that has so far been found to elicit a response from a skin-penetrating nematode is urocanic acid, a component of mammalian skin that attracts Str. stercoralis [14]. Thus, the extent to which skin-penetrating nematodes use olfactory cues to locate hosts is unclear.
Here we examined the host-seeking strategies and sensory behaviors of the human parasite Str. stercoralis as well as two other species of skin-penetrating nematodes, the rat parasites Str. ratti and Nippostrongylus brasiliensis (Figure 1A, D). We compared their behaviors to those of five other nematode species with diverse lifestyles and ecological niches: the passively ingested ruminant-parasitic nematode Haemonchus contortus; the entomopathogenic nematodes (EPNs) Heterorhabditis bacteriophora, Steinernema glaseri, and Steinernema carpocapsae; and the free-living nematode Caenorhabditis elegans (Figures 1A, D). This across-species analysis was used to fit the behaviors of skin-penetrating nematodes into an ecological framework, and to identify species-specific behavioral differences that reflect differences in phylogeny, host range, or infection route. We found that different species of mammalian-parasitic nematodes employ diverse host-seeking strategies, with the human parasite Str. stercoralis being a cruiser that actively seeks out hosts. We found that Str. stercoralis and the other skin-penetrating nematodes are attracted to skin and sweat odorants, while the passively ingested ruminant parasite Ha. contortus is attracted to the smell of grass. By comparing odor response profiles across species, we found that olfactory preferences reflect host specificity rather than phylogeny, suggesting a critical role for olfaction in the process of host finding and appropriate host selection. Our results provide insight into how skin-penetrating nematodes locate hosts to infect.
To gain insight into the host-seeking strategies used by mammalian-parasitic nematodes, we first examined their movement patterns in the absence of chemosensory stimuli. We compared their movement patterns to those of EPNs, which use well-characterized host-seeking strategies: some are “cruisers” that actively search for hosts, some are “ambushers” that wait for passing hosts, and some use an intermediate strategy [9], [15]. We first examined motility using an assay in which IJs were allowed to distribute on an agar plate in the absence of chemosensory stimuli for one hour and the location of IJs on the plate was recorded. We found that the motility of skin-penetrating IJs resembled that of EPN cruisers, with the human parasite Str. stercoralis being the most active (Figure 2A). By contrast, the motility of Ha. contortus resembled that of the ambushing EPN Ste. carpocapsae (Figure 2A). Thus, skin-penetrating IJs appear to be more active than passively ingested IJs.
To investigate the host-seeking strategies of skin-penetrating nematodes in more detail, we examined unstimulated movement of IJs using automated worm tracking [16]. We found that parasitic IJs vary dramatically in their crawling speeds, with the human parasite Str. stercoralis moving more rapidly than the other species tested (Figure S1A). The mean speeds of the skin-penetrating rat parasites were comparable to that of the most active EPN, Ste. glaseri, while the mean speed of Ha. contortus resembled that of the less active EPNs (Figure S1A). Turn probability also varied among species but did not correlate with speed (Figure S1B). Some but not all species crawled significantly faster following mechanical stimulation, and in fact the maximum speeds attained by Str. stercoralis, Str. ratti, and Ste. glaseri following mechanical stimulation were similar (Figure S1C–D, Movies S1 and S2). Thus, at least some of the differences in basal crawling speeds among species reflect differences in movement strategy rather than differences in the inherent speeds at which the IJs are capable of crawling.
The fact that Str. stercoralis has a higher basal speed than Str. ratti and N. brasiliensis is consistent with the possibility that host-seeking strategy evolved independently in these species to accommodate host behavior and ecology. Str. ratti and N. brasiliensis are parasites of nesting rodents, which are highly focal with circumscribed resting places. Since parasite transmission likely occurs within the confines of the nest, rapid mobility may not provide an adaptive advantage for these parasites. By contrast, Str. stercoralis is a parasite of humans, primates, and dogs, all of which are highly mobile. Rapid mobility may be necessary for Str. stercoralis to accommodate the mobility of its hosts.
Heat is emitted by all mammals and is a known sensory cue for some mammalian-parasitic nematodes, including Str. stercoralis [11]. We therefore examined the responses of the mammalian-parasitic IJs to a 37°C heat stimulus. We found that the skin-penetrating nematodes increased their crawling speed in response to thermal stimulation, while the passively ingested nematode Ha. contortus did not (Figure 2B). Skin-penetrating nematodes may increase their speed in response to heat to maximize the likelihood of encountering host skin.
A comparison of IJ movement patterns at room temperature versus 37°C revealed that skin-penetrating IJs show dramatically different movement patterns at the different temperatures. The trajectories of individual IJs were relatively straight at room temperature but highly curved at 37°C (Figure 2C). To quantify these differences, we calculated a distance ratio consisting of the total distance travelled divided by the maximum displacement achieved. We found that all three species of skin-penetrating nematodes showed greater distance ratios at 37°C compared to room temperature (Figure 2D). These results suggest that heat may act as a cue that signifies host proximity and stimulates local searching. However, we note that the temperature at the surface of human skin is 32–35°C [17], and IJ movement within this temperature range remains to be examined.
An important component of host-seeking strategy for many parasitic nematodes is nictation, a behavior in which the worm stands on its tail and waves its head to facilitate attachment to passing hosts [9]. We examined the nictation behavior of mammalian-parasitic nematodes by performing nictation assays on an “artificial dirt” substrate consisting of dense agar with near-microscopic pillars [18], since IJs are not capable of standing on standard agar plates due to the high surface tension on the plates [18]. We found that nictation frequencies varied among species. N. brasiliensis showed a high nictation frequency comparable to that of the ambushing EPN Ste. carpocapsae (Figure 2E and Movie S3), suggesting that it spends most of its foraging time nictating. By contrast, the Strongyloides species showed much lower rates of nictation (Figure 2E and Movie S4), suggesting they spend most of their foraging time crawling. Ha. contortus did not nictate on the artificial dirt substrate or any other substrate tested (see Materials and Methods), suggesting it may not be capable of nictating.
Taken together, our results suggest that mammalian-parasitic nematodes employ diverse host-seeking strategies. The skin-penetrating Strongyloides species appear to be cruisers that are highly mobile and tend to crawl rather than nictate. By contrast, the passively ingested nematode Ha. contortus appears to be an ambusher that displays little unstimulated movement. N. brasiliensis can exhibit rapid, prolonged movement comparable to that of the cruisers but tends to nictate rather than crawl, suggesting it is also an ambusher. However, we note that foraging strategy is in some cases substrate-dependent, and different strains of a species can exhibit different host-seeking behaviors [19], [20]. Thus, we cannot exclude the possibility that the host-seeking strategies of these species may vary under conditions not tested here.
EPNs have been shown to use a diverse array of insect volatiles and herbivore-induced plant volatiles for host finding [21]–[30]. By contrast, only one odorant has so far been identified as an attractant for Str. stercoralis [14]. We therefore tested the extent to which Str. stercoralis displays directed movement in response to human-emitted volatiles. We examined the responses of Str. stercoralis IJs to a large panel of odorants, most of which are known to be emitted by human skin, sweat, and skin microbiota (Table S5). Responses were examined using a chemotaxis assay (Figures S2 and S3) [21], [22]. We found that Str. stercoralis was strongly attracted to a number of these odorants (Figure 3A). Nearly all of the attractants we identified for Str. stercoralis also attract anthropophilic mosquitoes (Figure 3A), suggesting that nematodes and mosquitoes target humans using many of the same olfactory cues. While many of the human-emitted odorants that attracted Str. stercoralis are also emitted by other mammals, 7-octenoic acid is thought to be human-specific [31] and may be used by Str. stercoralis to target humans. Str. stercoralis and disease-causing mosquitoes are co-endemic throughout the world [2], and our results raise the possibility of designing traps that are effective against both parasites.
We also examined responses to carbon dioxide (CO2), which is emitted by aerobic organisms in exhaled breath and is an attractant for many parasites, including EPNs [9], [21], [22]. We found that Str. stercoralis was repelled by CO2 at high concentrations and neutral to CO2 at low concentrations, suggesting that CO2 is not a host attractant (Figure 3A and Figure S4A). These results are consistent with the fact that Str. stercoralis infects by skin penetration, and only low levels of CO2 are emitted from skin [32]. However, some EPNs respond synergistically to mixtures of CO2 and other odorants [33], and we cannot exclude the possibility that Str. stercoralis is attracted to CO2 in mixtures or under conditions not tested here.
The fact that Str. stercoralis responds to human-emitted odorants suggests that olfaction plays an important role in host finding. However, the extent to which Str. stercoralis or any other mammalian-parasitic nematode uses olfactory cues for host selection is not known. To gain insight into whether olfaction contributes to host choice, we compared the olfactory responses of Str. stercoralis to those of six other species: Str. ratti, N. brasiliensis, Ha. contortus, He. bacteriophora, Ste. carpocapsae, and C. elegans. We found that all species responded to a wide array of odorants, indicating that as is the case for EPNs [21], [22], even ambushers are capable of robust chemotaxis (Figure 3B and Figure S4). Moreover, each species exhibited a unique odor response profile, indicating that olfactory responses are species-specific even among closely related species such as Str. stercoralis and Str. ratti (Figure 3B). CO2 response varied greatly among species. Like Str. stercoralis, Str. ratti and N. brasiliensis were repelled by CO2 at high concentrations and neutral to CO2 at low concentrations (Figure 3B and Figure S4B–C). By contrast, Ha. contortus IJs, like EPN IJs and C. elegans dauers [21], [22], were attracted to CO2 (Figure 3B and Figure S4D). To confirm that the observed responses to odorants were olfactory rather than gustatory, we examined the responses of Str. stercoralis and Str. ratti to a subset of odorants in a modified chemotaxis assay in which odorants were placed on the plate lid rather than the plate surface. We found that attractive responses were still observed when the odorants were placed on the plate lid, although the response of Str. stercoralis to one odorant was slightly reduced (Figure S5). Thus, the observed behavioral responses are primarily olfactory, but in some cases may include a gustatory component.
The olfactory preferences of the passively ingested mammalian parasite, Ha. contortus, are consistent with its known ecology. Ha. contortus IJs migrate from the feces of their ruminant hosts to grass blades, where they are ingested by grazing ruminants [34]. The fact that 5% CO2, which approximates the concentration found in exhaled breath [35], was strongly attractive to Ha. contortus (Figure S4D) suggests that Ha. contortus may use exhaled CO2 to migrate toward the mouths of potential hosts. By contrast, Ha. contortus was repelled by many of the skin and sweat odorants tested (Figure 3B), consistent with a lack of attraction to mammalian skin. Of the few attractive odorants we identified for Ha. contortus, two – methyl myristate and myristic acid – are known constituents of cow and goat milk [36]–[38] and may be used by Ha. contortus to migrate toward cows and goats. To test whether Ha. contortus also responds to plant-emitted odorants, we examined responses to freshly cut grass. We found that Ha. contortus is attracted to the smell of grass, while Str. stercoralis and Ste. carpocapsae are not (Figure 3C). These results suggest that Ha. contortus uses CO2 in combination with other ruminant-emitted odorants and grass odorants to position itself for passive ingestion.
We then quantitatively compared odor response profiles across species, and found that species with similar hosts responded more similarly to odorants despite their phylogenetic distance (Figure 3D). For example, the distantly related rat parasites Str. ratti and N. brasiliensis responded similarly to odorants, as did the distantly related insect parasites He. bacteriophora and Ste. carpocapsae. The three skin-penetrating species responded more similarly to each other than to the other species tested, while the passively ingested mammalian parasite Ha. contortus responded very differently from all of the other species tested (Figure 3D). These results indicate that olfactory preferences reflect host specificity and infection mode rather than phylogeny, consistent with a key role for olfaction in host selection.
Skin-penetrating nematodes exit from hosts in feces as eggs or young larvae and subsequently develop into infective larvae outside the host. Thus, both infective and non-infective life stages are present in the environment (Figure 1B–C). This raises the question of whether host attraction is specific to the infective stage. We compared olfactory responses of free-living larvae, free-living adults, and IJs for both Str. stercoralis and Str. ratti in response to a subset of host odorants. We found that all three life stages were robustly attracted to host odorants, suggesting that host attraction is not downregulated in non-infective life stages (Figure 4). The free-living life stages of skin-penetrating worms are thought to reside primarily on host fecal matter, where they feed on bacteria present in the feces [39]. We therefore compared the responses of free-living larvae, free-living adults, and IJs to host feces. We found that responses differed dramatically across life stages: free-living larvae and adults were strongly attracted to feces, while IJs were neutral to host feces (Figure 4). Moreover, while Str. ratti IJs were neutral to both host and non-host feces, Str. stercoralis IJs were neutral to host feces but repelled by non-host feces (Figure 4).
Our results suggest a model in which all life stages are attracted to host skin odor, but strong attraction to host fecal odor by the free-living life stages causes them to remain on feces. Attraction to fecal odor is downregulated at the infective stage, enabling the IJs to migrate away from the feces in search of hosts. Repulsion of Str. stercoralis IJs from non-host feces may serve as an additional mechanism to prevent foraging in close proximity to non-hosts. To gain insight into the individual odorants that confer changes in sensitivity to feces, we examined responses to two components of fecal odor, skatole and indole [40]. We found that the free-living stages of Str. ratti were highly attracted to both skatole and indole, while the IJs were neutral to both odorants (Figure S6A). Thus, altered sensitivity to these odorants may contribute to the developmental change in the response to fecal odor. By contrast, Str. stercoralis IJs were more attracted to skatole than the free-living life stages and all three life stages were relatively unresponsive to indole (Figure S6B), suggesting that other as yet unidentified odorants mediate the sensitivity of Str. stercoralis to fecal odor.
Str. stercoralis infection is a worldwide cause of chronic morbidity and mortality. Current drugs used to treat nematode infections are inadequate for nematode control: some are toxic, drug resistance is a growing concern, and reinfection rates are high [41]. Our data suggest that Str. stercoralis IJs are fast-moving cruisers that actively search for hosts using a chemically diverse array of human-emitted odorants. The identification of odorants that attract or repel Str. stercoralis and other parasitic nematodes lays a foundation for the design of targeted traps or repellents, which could have broad implications for nematode control.
Gerbils were used for host passage of Str. stercoralis, and rats were used for host passage of Str. ratti and N. brasiliensis. All protocols and procedures were approved by the UCLA Office of Animal Research Oversight (Protocol No. 2011-060-03B), which adheres to the AAALAC standards for laboratory animal use, and were in strict accordance with the Guide for the Care and Use of Laboratory Animals.
Strongyloides stercoralis UPD strain and Strongyloides ratti ED321 strain were provided by Dr. James Lok (University of Pennsylvania). Nippostrongylus brasiliensis was provided by Dr. Edward Platzer (University of California, Riverside). Haemonchus contortus was provided by Dr. Adrian Wolstenholme and Mr. Bob Storey (University of Georgia). Heterorhabditis bacteriophora Oswego strain and Steinernema glaseri VS strain were provided by David Shapiro-Ilan (USDA). Steinernema carpocapsae were from the ALL strain [21], [22], [42]. C. elegans dauers were from the wild isolate CB4856 (“Hawaii”). Male Mongolian gerbils for culturing Str. stercoralis were obtained from Charles River Laboratories. Male or female Long-Evans or Sprague Dawley rats for culturing Str. ratti and N. brasiliensis were obtained either from Harlan Laboratories or second-hand from other investigators at UCLA through the UCLA Internal Animal Transfer supply system for surplus animals. Galleria mellonella larvae for culturing EPNs were obtained from American Cricket Ranch (Lakeside, CA).
Str. stercoralis was serially passaged in gerbils and maintained on fecal-charcoal plates. Inoculation of gerbils with Str. stercoralis was performed essentially as previously described [43]. Briefly, Str. stercoralis IJs were isolated from fecal-charcoal plates using a Baermann apparatus [43]. Each gerbil was subcutaneously injected with 2000 IJs in 200 µl sterile PBS. Gerbils became patent (as defined by the presence of nematodes in gerbil feces) on day 12 post-inoculation and remained patent for approximately 70 days. At 28 and 35 days post-inoculation, each gerbil received 2 mg methylprednisolone (Depo-Medrol, Pfizer) subcutaneously to induce an auto-infective cycle. To harvest infested feces, gerbils were housed overnight in cages containing a wire rack on the bottom of the cage. Fecal pellets fell below the rack onto damp cardboard and were collected the following morning. Feces were mixed with dH2O and autoclaved charcoal (bone char from Ebonex Corp., Cat # EBO.58BC.04) in an approximately 1∶1 ratio of charcoal to feces. The fecal-charcoal mixtures were poured into Petri dishes (10 cm diameter, 20 mm height) lined with wet filter paper, and were stored at 23°C until use. Nematodes used for behavioral analysis were isolated from fecal-charcoal plates using a Baermann apparatus [43] or from plate lids. To obtain free-living larvae (primarily post-parasitic L2s) for chemotaxis assays, nematodes were collected from fecal-charcoal plates after approximately 18 hrs. To obtain free-living adults for chemotaxis assays, nematodes were collected from fecal-charcoal plates after 48 hrs. To obtain IJs, nematodes were collected from fecal-charcoal plates starting at day 5 post-collection. IJs were used for behavioral assays within 2 weeks of fecal collection.
Str. ratti was serially passaged in rats and maintained on fecal-charcoal plates. Inoculation of rats with Str. ratti was performed essentially as previously described [44]. Briefly, Str. ratti IJs were isolated from fecal-charcoal plates using a Baermann apparatus. Each rat was subcutaneously injected with 700 IJs in 300 µl sterile PBS. Rats became patent on day 6 post-inoculation and remained patent for up to 28 days post-inoculation. To harvest infested feces, rats were housed overnight in cages containing a wire rack on the bottom of the cage. Fecal pellets fell below the rack onto damp cardboard and were collected the following morning. Fecal-charcoal plates were prepared as described above for Str. stercoralis and stored at 23°C until use. Nematodes used for behavioral analysis were isolated from fecal-charcoal plates using a Baermann apparatus [43] or from plate lids. Free-living larvae, adults, and IJs were obtained from fecal-charcoal plates as described above for Str. stercoralis.
N. brasiliensis was serially passaged in rats and maintained on fecal-charcoal plates. To inoculate rats, N. brasiliensis IJs were isolated from fecal-charcoal plates using a Baermann apparatus. Each rat was subcutaneously injected with 4000 IJs in 300 µl sterile PBS. Rats became patent on day 6 post-inoculation and remained patent for up to 14 days. Infested feces were collected as described above for Str. ratti. Fecal-charcoal plates were prepared as described above for Str. stercoralis, except that vermiculite (Fisher catalog # S17729) was added to the feces and charcoal in an approximately 1∶1∶1 ratio of vermiculite to charcoal to feces. Plates were stored at 23°C until use. In some cases, either Nystatin (Sigma catalog # N6261) at a concentration of 200 U/ml or Fungizone (Gibco catalog #15290-018) at a concentration of 1 µg/ml was added to the filter paper on the bottom of the plate to inhibit mold growth. Nematodes used for behavioral analysis were isolated from fecal-charcoal plates using a Baermann apparatus [43] or from plate lids. To obtain IJs, nematodes were collected from fecal-charcoal plates starting at day 7 post-collection. IJs were used for behavioral assays within 2 weeks of fecal collection.
Ha. contortus was stored in dH2O at 8°C prior to use. IJs were tested within 6 months of collection. No differences in IJ movement or behavior were observed in freshly collected versus 6 month old IJs. IJ behavior declined after 6 months, so IJs older than 6 months were not tested.
EPNs were cultured as previously described [21]. Briefly, 5 last instar Galleria mellonella larvae were placed in a 5 cm Petri dish with a 55 mm Whatman 1 filter paper acting as a pseudo-soil substrate in the bottom of the dish. Approximately 250 µl containing 500–1000 IJs suspended in water was evenly distributed on the filter paper. After 7–10 days the insect cadavers were placed on White traps [45]. Emerging IJs were collected from the White trap, rinsed 3 times with dH2O, and stored in dH2O until use. Ste. carpocapsae and He. bacteriophora were maintained at 25°C, while Ste. glaseri was maintained at room temperature. IJs were used for behavioral assays within 7 days of collection from the White trap.
C. elegans was cultured on NGM plates seeded with E. coli OP50 according to standard methods [46]. Dauer larvae were collected from the lids of plates from which the nematodes had consumed all of the OP50 and stored in dH2O at room temperature prior to use. Dauer larvae were used for behavioral assays within 2 weeks of collection from plate lids.
30–100 IJs were placed in the center of a chemotaxis plate [47]. IJs were allowed to distribute over the agar surface for 1 hr, after which the percentage of IJs in the outer zone (Zone 2) was determined. Zone 1 was a 4 cm diameter circle centered in the middle of the plate. Zone 2 consisted of the rest of the plate and included the edges of the plate, which acted as a trap since IJs that crawled onto the plate edge desiccated and could not return to the agar surface.
Recordings of worm movement were obtained with an Olympus E-PM1 digital camera attached to a Leica S6 D microscope. To quantify unstimulated movement, 4–5 IJs were placed in the center of a chemotaxis plate [47] and allowed to acclimate for 10 min. 20 s recordings were then obtained. Worms that either did not move, that stopped moving during the recording, or that crawled off the assay plate during the recording were excluded from the analysis. To quantify movement before and after mechanical stimulation, IJs were placed on chemotaxis plates and allowed to acclimate for 10 min. prior to tracking. Baseline movement was recorded for approximately 15 s. The plate lid was then removed, the IJ was gently agitated using a worm pick, and post-agitation movement was recorded for approximately 30 s. 5 s recording clips directly following agitation were used to calculate the maximum speeds shown in Figure S1D, and 5 s recording clips directly preceding and following agitation were used to generate the sample tracks shown in Figure S1C. Maximum speeds were calculated in WormAnalyzer (see below) based on changes in worm position over a seven frame (or 0.23 second) window. To quantify movement following thermal stimulation, assays were performed in a 37°C warm room. Chemotaxis assay plates were kept in the warm room prior to use. Individual IJs were transported into the warm room, transferred to assay plates, and immediately recorded for 20 s. For the room temperature control, IJs were similarly transferred to assay plates and immediately recorded for 20 s. Locomotion was quantified using WormTracker and WormAnalyzer multi-worm tracker software (Miriam Goodman lab, Stanford University) [16]. The following WormTracker settings were adjusted from the default settings (designed for C. elegans adults) for analysis of IJ movement: min. single worm area = 20 pixels; max. size change by worm between successive frames = 250 pixels; shortest valid track = 30 frames; auto-thresholding correction factor = 0.001. To calculate turn frequencies, the following WormAnalyzer settings were adjusted from the default settings for analysis of IJ speed: sliding window for smoothing track data = 30 frames; minimum run duration for pirouette identification = 2.9 s for Str. stercoralis, 5.3 s for Ste. glaseri, and 6 s for all other species (to compensate for differences in speed among species). All turns were confirmed by visual observation of worm tracks; turns not confirmed by visual observation were not counted. For calculations of maximum displacement in Figure 2D, the distance between the worm's start point and the farthest point the worm reached during the 20 s recording was calculated in ImageJ.
Nictation was quantified on “micro-dirt” agar chips cast from polydimethylsiloxane (PDMS) molds as previously described [18], except that chips were made from 5% agar dissolved in dH2O and were incubated at 37°C for 2 hr and then room temperature for 1 hr before use. The micro-dirt chip consisted of agar with near-microscopic pillars covering its surface (pillar height of 25 µm with a radius of 25 µm and an interval between pillars of 25 µm), which allowed IJs to nictate on top of the pillars. For each assay, 3–10 IJs were transferred to the micro-dirt chip and allowed to acclimate on the chip for 10 min. Each IJ was then monitored for 2 min. An IJ was scored as “nictating” if it raised its head off the surface of the chip for a period of at least 5 s during the 2 min assay period. Nictation behavior was also tested on sand. Sand nictation assays were performed essentially as previously described [21], [48]. Sand (silicon dioxide, >230 mesh, CAS 60676-86-0) was distributed onto the surface of a chemotaxis plate using a sieve. IJs were transferred to the plate surface and allowed to acclimate for 10 min. Nictation behavior was then observed for two minutes. In all cases, nictation behavior on sand was consistent with nictation behavior on micro-dirt chips. In the case of Ha. contortus, we also tested for nictation on grass and vermiculite; no nictation was observed on any substrate tested. To test for nictation on grass, grass samples were collected from a lawn seeded with UC Verde Buffalo grass and perennial rye grass (the same lawn as for sample 1 below). The grass was cut into small chunks (∼2.5 mm×2.5 mm) and distributed onto the surface of a chemotaxis plate. IJs were transferred onto the plate surface or directly onto blades of grass, and nictation was scored after a 10 min. acclimation period. Nictation was also scored after 20, 30, or 60 min., or the next day. No nictation was observed with Ha. contortus at any time point.
Odor chemotaxis assays were performed essentially as described [21], [22] (Figure S2). Assays were performed on chemotaxis assay plates [47]. Scoring regions consisted of 2 cm diameter circles on each side of the plate along the diameter with the center of the circle 1 cm from the edge of the plate, as well as the rectangular region extending from the edges of the circle to the edge of the plate. Either 2 µl (for mammalian-parasitic IJs) or 1 µl (for insect-parasitic IJs and C. elegans dauers) of 5% sodium azide was placed in the scoring region as anesthetic. 5 µl of odorant was then placed on the surface of the assay plate in the center of one scoring region, and 5 µl of control (paraffin oil, dH2O, or ethanol) was placed on the surface of the assay plate in the center of the other scoring region. Approximately 200 worms were placed in the center of the assay plate and left undisturbed on a vibration-reducing platform for 3 hours at room temperature. A chemotaxis index (CI) was then calculated as: CI = (# worms at odorant−# worms at control)/(# worms at odorant+control) (Figure S2). A positive CI indicates attraction; a negative CI indicates repulsion. A 3 hour assay duration was used because 3 hour assays were found to be most effective for EPNs [21], [49]. However, 1 hour assays were also performed with Str. ratti, and no significant differences were observed in 1 hour vs. 3 hour assays (Table S6). Two identical assays were always performed simultaneously with the odor gradient in opposite directions on the two plates to control for directional bias due to room vibration; assays were discarded if the difference in the CIs for the two plates was ≥0.9 or if fewer than 7 worms moved into the scoring regions on one or both of the plates. Liquid odorants were tested undiluted unless otherwise indicated. Solid odorants were prepared as follows: 1-dodecanol, methyl palmitate, and methyl myristate were diluted 0.05 g in 2.5 ml paraffin oil; palmitic acid was diluted 10 g in 200 ml ethanol; myristic acid, skatole, and indole were diluted 0.05 g in 2.5 ml ethanol; and L-lactic acid was diluted 0.05 g in 2.5 ml dH2O. Ammonia was purchased as a 2 M solution in ethanol. Solid odorants were tested at these concentrations unless otherwise indicated. For assays in which odorants were placed on the plate lid rather than the plate surface (Figure S5), filter paper squares of approximately 0.5 cm in width were attached to the plate lid using double-stick tape. Odorant or control was then pipetted onto the filter paper, and chemotaxis was examined as described above.
CO2 chemotaxis assays were performed essentially as described [21], [22]. Assays were performed on chemotaxis assay plates [47], and scoring regions were as described above for odor chemotaxis assays (Figure S2). Gases were delivered at a rate of 0.5 ml/min through holes in the plate lids from gastight syringes filled with either a CO2 mixture containing the test concentration of CO2, 10% O2, and the balance N2, or a control air mixture containing 10% O2 and 90% N2. Certified gas mixtures were obtained from Air Liquide or Airgas. Assays were performed and scored as described above for odor chemotaxis assays, except that the assay duration was 1 hour.
Fresh grass samples were collected from the campus of the University of California, Los Angeles. Sample 1 was collected from a lawn seeded with UC Verde Buffalo grass and perennial rye grass, and sample 2 was collected from a lawn seeded with a custom blend of annual ryegrass, Festuca, Bonsai dwarf fescue, Bermuda grass, and bluegrass. 200 µl of dH2O was added to 0.1 g grass. Grass was then ground in a small weigh boat, and 5 µl of the grass suspension was used in a chemotaxis assay with 5 µl dH2O as a control. Grass was either used immediately for chemotaxis assays or stored at 4°C for no more than 3 days.
Uninfected rat or dog feces was collected from animals in the UCLA vivarium. Responses to feces were tested using a modified chemotaxis assay in which feces was placed on the plate lid rather than the plate surface. Filter paper squares of approximately 0.5 cm in width were attached to the plate lid using double-stick tape. Fecal matter was moistened with dH2O, smeared onto filter paper, and tested in a chemotaxis assay as described above for odor chemotaxis assays. We note that similar attraction to feces was observed when filter paper with feces was tested against filter paper with dH2O, and no attraction was observed to wet filter paper when wet filter paper was tested against dry filter paper (data not shown).
Statistical analysis was performed using either GraphPad Instat, GraphPad Prism, or PAST [50]. The heatmap was generated using Heatmap Builder [51].
|
10.1371/journal.pcbi.1005008 | Collective Signal Processing in Cluster Chemotaxis: Roles of Adaptation, Amplification, and Co-attraction in Collective Guidance | Single eukaryotic cells commonly sense and follow chemical gradients, performing chemotaxis. Recent experiments and theories, however, show that even when single cells do not chemotax, clusters of cells may, if their interactions are regulated by the chemoattractant. We study this general mechanism of “collective guidance” computationally with models that integrate stochastic dynamics for individual cells with biochemical reactions within the cells, and diffusion of chemical signals between the cells. We show that if clusters of cells use the well-known local excitation, global inhibition (LEGI) mechanism to sense chemoattractant gradients, the speed of the cell cluster becomes non-monotonic in the cluster’s size—clusters either larger or smaller than an optimal size will have lower speed. We argue that the cell cluster speed is a crucial readout of how the cluster processes chemotactic signals; both amplification and adaptation will alter the behavior of cluster speed as a function of size. We also show that, contrary to the assumptions of earlier theories, collective guidance does not require persistent cell-cell contacts and strong short range adhesion. If cell-cell adhesion is absent, and the cluster cohesion is instead provided by a co-attraction mechanism, e.g. chemotaxis toward a secreted molecule, collective guidance may still function. However, new behaviors, such as cluster rotation, may also appear in this case. Co-attraction and adaptation allow for collective guidance that is robust to varying chemoattractant concentrations while not requiring strong cell-cell adhesion.
| To get from one part of the body to another, single cells often follow chemical signals. Sometimes, though, isolated cells ignore these signals, but a group of cells still manages to travel in a directed way. How can this happen? We argue that if the signal changes how the cells interact with their neighbors, a cluster of cells can detect signals single cells ignore. We use computational models to study how this can happen, and show that the speed of the cluster will depend on how the cells process the signal, as well as whether or not the cells are tightly connected to one another. We also show if the cells are only loosely connected, and are attracted to a secreted molecule, cell clusters may develop rotation and other effects that will change how effectively they can sense signals.
| Many individual cells, including white blood cells and bacteria, chemotax—sensing and following gradients of signals. Some cells, though, are not loners—they migrate collectively—and cells traveling in clusters and sheets during development must chemotax together. Many experiments [1–5] have shown that clusters can have capabilities that single cells lack: in particular, clusters of cells can follow a gradient even when single cells do not. How can cells work together to follow a gradient that each individual cell is incapable of sensing? How can cells integrate data from across the cluster to improve their gradient sensing abilities? Is cluster chemotaxis essentially different than single-cell chemotaxis? The simplest possibility, that cells just spatially average the gradient signal acting independently on each of them and thereby achieve a more accurate sensing capability, is ruled out, at least for lymphocytes, by experiments that show clusters can travel in the direction opposite to that of single cells [1]. A different possible explanation relies on the qualitative idea of collective guidance [6], in which a cluster of cells can gain a direction even though each of its individual cells senses only the level of signal, and not its gradient. To make this notion more quantitative, we have recently introduced such a model of collective guidance in the context of neural crest cells where the cluster’s directionality comes from a regulation of contact inhibition of locomotion (CIL) [7]; a related model was also proposed for clusters of lymphocytes [1] and extended for studying border cell migration [8].
However, our current understanding of collective guidance and how collective chemotaxis occurs without single-cell gradient sensing does not account for the possibility of response coordinated by chemical signaling between cells. Our minimal model of collective guidance posits that each cell reacts only to the local chemoattractant and the physical presence of its neighbors [7]. More complicated signal processing could take place on the cluster scale if cells use signaling molecules to communicate with each other to collectively process the information contained in the chemoattractant gradient, as was recently suggested to be the case in branching morphogenesis [4, 9]. It is therefore important to ask: What experimental signatures would tell us if this were happening, and how would this signal processing change the efficiency of the cluster’s movement? Can collective signal processing overcome shallow gradients seen in vivo (e.g. [10]), amplifying differences in cluster behavior between the front and the back? In minimal models of collective guidance [1, 7], the cluster moves by a tug of war, and is likely under tension. Nevertheless, collective chemotaxis can also occur in the absence of strong adhesion [2]. How does this happen? We will address all of these questions in this paper. Our initial focus will be on understanding in vitro experiments in relatively controlled environments [1, 2], especially experiments on explants of neural crest cells, and using these results to develop a useful quantitative framework for the study of collective guidance more generally, including collective chemotaxis in vivo [11, 12].
To understand why clusters of explanted neural crest cells chemotax where single cells do not [2], we will analyze both short-range interactions between cells and long-range interactions mediated by chemical secretions. The primary short-range interaction between neural crest cells are cadherin-mediated adhesion and contact inhibition of locomotion (CIL). CIL results in cells repolarizing away from each other after contact. CIL in tissues may be regulated by the type of cadherin expressed, as well as being linked to mechanical force between cells [13–16]. Many possible molecular mediators of CIL have been established, including the non-canonical Wnt-planar cell polarity pathway and ephrin signaling [17, 18]. Within this paper, we will take a phenomenological approach to modeling CIL, describing its consequences rather than its molecular origin.
We first study models of biochemical processing of the chemoattractant signal within the cell cluster, assuming strong cell-cell adhesions as in our earlier model [7]. We treat the possibility of gradient sensing via cell-cell communication, using a mechanism that allows adaptation, i.e. the cluster’s response becomes insensitive to the overall level of the signal S(r). We do this using a local excitation, global inhibition (LEGI) scheme [19]. This model is supported by recent experiments on collective gradient sensing in branching morphogenesis, which identify gap-junction-mediated communication between cells as a critical aspect of collective gradient sensing [4, 9]. We also consider the possibility of cluster-level amplification of a sensed gradient, where relatively small changes in the chemoattractant signal S(r) across the cluster are amplified into much larger changes in the response level. With both adaptation and a switch-like amplification, we find that clusters of an optimal size are more efficient at chemotaxing than either smaller or larger clusters. Amplification of the external signal allows clusters to develop a large velocity even in a shallow gradient. We argue, based on simple scaling principles, that sufficiently large clusters with only short-range adhesion undergoing collective guidance would be expected to either fragment or become increasingly slow.
We then show that if the cohesion of a cluster is not controlled by local cell-cell adhesion, but rather by chemotaxis toward a secreted signal (or “co-attraction” [20]), a cluster of cells can undergo collective guidance by regulation of CIL even if cells are not in continuous contact. We show how co-attraction and regulated CIL interact in order to create robust chemotaxis. In the presence of co-attraction, new behaviors, including persistent cluster rotation, may emerge. We provide an extensive characterization of the transition to rotation, and how rotation can alter the efficiency of gradient-sensing clusters.
We want to model the collective guidance of a cluster of cells exposed to a chemical gradient S(r). We use the experiments of [2] on neural crest explants responding to Sdf1 gradients as a guide to determine the features we include as well as the model parameters, though we expect our results to be more generally applicable as well. There are four major elements of a model of this process: 1) single-cell dynamics, 2) physical interactions between cells and contact-range effects like contact inhibition of locomotion, 3) the response of the cells to the chemical S(r), and 4) chemical communication and signaling between cells.
We use a two-dimensional stochastic particle model to describe cells exposed to a chemical gradient S(r). We describe each cell i with a position ri and a polarity pi. The cell polarity indicates the cell’s direction and propulsion strength, i.e. the velocity with which it would travel in the absence of additional forces; we thus define pi so that an isolated cell with polarity pi has velocity pi. The cell’s motion is overdamped, so physical forces like cell-cell adhesion and exclusion change the cell’s velocity—the velocity of the cell is pi plus the net force the other cells exert on it, ∑ j ≠ i F i j. We model chemically-induced effects like CIL as altering a cell’s biochemical polarity pi. Our model is then:
∂ t r i = p i + ∑ j ≠ i F i j (1) ∂ t p i = - 1 τ p i + σ ξ i ( t ) + β i ∑ j ∼ i r ^ i j + χ ∇ c ( r i ) | ∇ c | Θ ( | ∇ c | - g 0 ) (2)
where Fij are intercellular forces, e.g. cell-cell adhesion and volume exclusion, and ξi(t) are fluctuating, temporally uncorrelated noise terms that are Gaussian with 〈 ξ μ i ( t ) ξ ν j ( t ′ ) 〉 = 2 δ μ ν δ i j δ ( t - t ′ ), where the Greek indices μ, ν run over the dimensions x, y. The first two terms on the right of Eq 2 are a standard Ornstein-Uhlenbeck model [28, 29]: pi returns to zero with a timescale τ, but is pushed away from zero by the fluctuating noise ξ(t). This models a cell that has a motion that is only persistent over a time of τ.
Throughout this paper, we choose our units to be defined by the typical parameters of neural crest cells. With this in mind, we take our length scale to be the typical equilibrium cell-cell separation and our time scale to be the relaxation time—this corresponds to setting the cell diameter to be unity and the relaxation time τ = 1. To convert between these simulation units and real units, we use values estimated from the experiments of [2]: typical equilibrium cell-cell separation is 20 μm and the typical time over which a cell reorients is roughly 20 minutes, i.e. τ = 20 minutes in real units. Within the simulation units we have chosen, measured neural crest cell velocities are on the order of 1, so we choose σ = 1. This choice means that the root mean square speed of an isolated cell is 〈 | V | 2 〉 1 / 2 = 2 1 / 2 σ τ 1 / 2 ≈ 1 . 4 microns/minute, in good agreement with, e.g. [2].
When we include adaptation, we assume that the kinetics of Eqs 7 and 9 are fast compared with the dynamics of interest, and set them to their steady states, assuming k−R ≫ kR and thus Ri = Ai, ss/Ii(t). We set the diffusion rate kD = 4 in our units, corresponding to a time for equilibration of a few minutes, consistent with experiments using FRAP to see equilibration of fluorescent dyes across gap junctions [39]. However, we note that this rate can depend on the identity of the inhibitor, and may also be regulated [40, 41]. We set the rates of generation and decay of the inhibitor to be kI = k−I = 1; this is discussed more in the adaptation section. A complete list of parameters and their justifications is included in the Supplementary Information, Table S1.
We integrate Eqs 1, 2 and 7–9 explicitly with an Euler-Maruyama integrator [42]. The time step varies: for rigid clusters with high adhesion, we choose Δt = 1 × 10−4, and for co-attraction simulations we choose Δt = 1 × 10−3. Further details about time step selection as well as source code are available in the Supplementary Information.
In our earlier paper [7], we studied a minimal version of the model described above, with no co-attraction (χ = 0) and no adaptation or amplification, i.e. β i = β ¯ S ( r i ). We briefly note a few results from that paper here, as in some limits, our more complex model will reduce to this model. Under assumptions of cluster rigidity and slow reorientation, the mean drift of a cluster of cells obeying Eqs 1 and 2 is given by
〈 V 〉 c ≈ β ¯ τ M · ∇ S (10)
with the approximation true for S(r) ≈ S0 + r · ∇S. 〈⋯〉c is an average over the fluctuating pi but with fixed configuration and orientation of cells ri. The matrix M depends only on the configuration of cells; formulas for many cluster shapes and sizes are given in [7]. Mean cluster velocity 〈Vx〉 saturates at large number of cells N. This arises because we have the difference in signal between the front and the back growing as the cluster radius (∼ N), while the perimeter of the cluster also grows as N. The force on the cluster then grows as N at large N, while the effective friction of the cluster grows independently with the number of cells, as N—hence the net velocity should behave as ∼N1/2 × N1/2/N ∼ 1 at large N. (Similar scaling arguments are found for the circular cluster limit in [1].) As we move beyond the minimal model, these scaling assumptions may break down, and therefore larger clusters will not necessarily have saturating velocities.
Ref. [7] also provides analytic results for the chemotactic index CI– a measure of the directionality of the cluster motion. This is commonly defined as the ratio of the distance traveled in the direction of the gradient (the x direction) to the total distance traveled. To clarify how we average over many realizations of a path, we define CI = 〈Vx〉 / 〈|V|〉.
Until this point, we have only looked at highly adherent, effectively rigid clusters. However, collective cell migration can also occur with a high degree of fluidity and cell-cell rearrangement [12, 53–60]. In addition, we have until now assumed that the only attraction between cells is short-range, representing cell-cell adhesion. However, neural crest cells also attract one another through chemical secretions, which can control the extent of cluster directionality and cohesion [20, 27]—and many other cell types also chemotax toward secretions [61, 62]. We extend our model to allow for this possibility, and show that clusters of cells that cohere via co-attraction can also be directed by collective guidance. These clusters need not be rigid, and can have significant re-arrangement or even only transient contacts.
In this section, we will treat clusters with co-attraction (χ ≠ 0), but assume only the minimal model of signal processing, with the CIL susceptibility β i = β ¯ S ( r i ).
In our earlier work [7], we provided a minimal quantitative model that embodied the collective guidance hypothesis [2, 6] and provides a plausible initial model for collective chemotaxis when single cells do not chemotax. However, this model made two fairly strict assumptions: first, that the cluster does not perform any relevant internal processing of the chemoattractant gradient, and second, that the cluster is highly adherent with cohesion from short-range interactions. We and others have used similar models to analyze collective chemotaxis in many biological contexts [1, 7, 8]. In this paper, we have relaxed both of these assumptions, and showed that new qualitative behaviors may arise, emphasizing potentially necessary extensions to these models.
We find, consistent with our earlier model and the experimental results of [2], that small clusters of cells can chemotax, even if single cells cannot. However, while our minimal model predicts that both velocity and chemotactic index increase as cluster size increases, we find that adaptation to and amplification of the chemoattractant signal can lead to cluster velocity to decrease and chemotactic index to saturate at a value less than one at large cluster sizes. This effect can arise purely from the finite rate of contact mediated diffusive transfer of chemicals between contacting cells. However, even if transfer is fast, if the cell’s response to the signal is switchlike, cluster velocities can be non-monotonic and chemotactic indices can saturate. This non-monotonic behavior is a sign of the way in which the cluster processes the chemoattractant signal. Non-monotonicity of cluster velocity was recently observed in border cell chemotaxis, and was interpreted in terms of additional hydrodynamic resistance at large cluster sizes [8]. Our results here show that alternate mechanisms, like amplification, could potentially also create this qualitative signature, and should be considered.
How do our results compare to experimental data? Theveneau et al. [2] find that chemotactic indices of small (2–3 cell) and large clusters of neural crest cells are similar, but do not observe a large variation in cluster speed. Within our models with adaptation and amplification in adherent clusters, we see that chemotactic indices of 7 cell clusters and 61 cell clusters are often similar—but we generally observe that 2–3 cell clusters have smaller chemotactic indices. This is in part because of orientational averaging—under our assumption that single cells do not chemotax, pairs of cells have a chemotactic behavior that varies strongly with orientation [7]. However, we highlight some difficulties with a direct comparison between model and experiment in this case. First, if cells have a distribution of CIL strengths β ¯, we might expect that cells with higher CIL strength β ¯ form smaller clusters; thus the strength of the collective guidance mechanism could be different between small and large clusters. Secondly, the concentration of the chemoattractant Sdf1 is not well-characterized in the bead assay used by [2]; more extensive studies using microfluidic chambers would aid in pinpointing differences between our model and experiments in neural crest. Non-monotonicity can also make experimental results difficult to interpret. Within our model with adaptation and amplification, a given cluster could have either a larger or smaller velocity than a smaller cluster, depending on the sizes of the clusters and the details of the adaptation mechanism (e.g. the rate of diffusion of the inhibitor). This suggests that solely comparing large and small clusters could potentially be misleading, and in general more detailed experiments as a function of cluster size are needed.
How generic is the result that sufficiently large tightly bound clusters decrease in speed? We have found that this occurs both with adaptation and switch-like amplification, but not with the minimal model of [7], where clusters will break apart if they are not sufficiently tightly bound. We expect this behavior to be relatively broadly present in sufficiently large, and sufficiently tightly bound clusters. In collective guidance in tightly adherent clusters, there is a tug-of-war, and the cluster is under tension, and would tear itself apart in the absence of the strong adhesion. In order for the cluster to maintain its speed as the number of cells increases, which increases friction between the cluster and the surface, the difference in protrusion strength between the front and back must also increase. In adaptation and switchlike amplification, this increase in protrusion strength is slow or absent, and the cluster slows at large N. However, if the protrusion strength βi increases with cluster size, it will eventually overcome cell-cell adhesion—and we expect the cluster to eventually scatter. This argument is not specific to the models we have studied, but should generalize to any model where the cluster is under tension and driven solely by the edge: large tightly adherent clusters will either scatter as the edge force increases, or slow, if the edge can no longer robustly pull the entire cluster.
Our model for strongly adherent clusters has assumed that the primary driver of the dynamics of strongly adherent clusters are cells at the edge. This is consistent with the observations of [2] on neural crest, who observe that only the edge cells develop strong protrusions: there are no cryptic protrusions. In small (2–30 cells) adherent clusters of epithelial cells, traction forces are also mainly found to be large at the edge [70–72]. However, in other systems, most notably the classic example of migration of large epithelial sheets in a wound healing geometry, traction forces are also exerted significantly away from the edge [73]. If these interior traction forces are relevant, we would expect the scaling of cluster velocity with cluster size to be significantly altered. Ultimately, experimental traction force measurements may be crucial in determining whether our assumption of edge-driven dynamics is appropriate; however, this assumption is consistent with the currently available data. We have also recently shown that a similar model of CIL can reproduce some of this traction force data, lending support to our assumptions here [74].
Adaptation in single-cell chemoresponse is a ubiquitous and well-tested principle, but its existence is not established for clusters of neural crest or lymphocytes; applying a step response would be a straightforward test of adaptation, and we would expect protrusions and traction forces to peak and then adapt (Fig 2A).
We argue that the gap junction-mediated LEGI model we have used is a reasonable expectation for collective signal processing. Recent papers have independently suggested that gap junctions play a role in gradient sensing and proposed a similar LEGI model [4, 9], though solely in one dimension, and without any effects of CIL, cell motility, or amplification. Our results suggest that gap-junction mediated gradient sensing across the cluster may be effective—though with characteristic effects on the cluster velocity, as discussed above. As noted earlier, our hypothesis of gap junction mediated communication is consistent with the experimental observation that gap junctions modulate neural crest cell motility in vivo [35, 36].
In this paper, we also modeled the possibility that the coherence of the cluster is not provided by strong physical adhesion, but rather by chemoattraction to a secreted signal, i.e. co-attraction. This co-attraction mechanism is known to be relevant in neural crest [20] (and see also the model [27]). Our model with co-attraction also shows that, consistent with experiments on neural crest [2], that the collective guidance mechanism proposed here can guide cells even with only transient contacts. However, we also see that if co-attraction is too large, new emergent behaviors can appear, including cluster rotation. Persistent rotation of cell clusters is not observed in neural crest, and only transient rotations appear to occur in lymphocytes undergoing collective chemotaxis [1]. Rotating droplets are, however, observed in bacteria and in the social amoeba Dictyostelium discoideum; vortex formation in bacteria has been speculated to also occur in part via chemotaxis to a secreted molecule [63] though chemotaxis is not necessary and may not be relevant in Dictyostelium [75].
Similar to our model of co-attraction, chemotaxis to a secreted molecule and relayed response to a secreted molecule have been modeled extensively in the literature, with varying degrees of biological specificity and detail [27, 63, 76–82]. We note particularly [27, 81, 83] who also focus on neural crest. Our approach here has been to tend toward minimalism where possible, while still respecting the experimental facts on the neural crest explant experiments of [2]. Many of our assumptions could be generalized, including, e.g. explicitly modeling chemical details of the co-attraction sensing [77], or changing our assumption that cells respond equally strongly to weak and strong gradients of co-attractant. Changing the details of the co-attraction model will change the cluster structure, which will quantitatively change the cluster velocity. Our initial studies in this direction have not yet revealed important qualitative changes.
Other variants of stochastic particle models have been used to model collective cell migration, ranging from models that use single particles to represent cells [30, 53, 63, 84–86] to those that use more detailed representations of cells with either multiple particles or additional details of cell shape [87–91]. Other techniques, such as the Cellular Potts Model [58, 75, 92] and phase field models [23, 93, 94] have also been developed to study collective cell migration with significantly greater levels of detail on the cell’s shape and its internal biochemistry. Because emergent collective guidance has had only limited quantitative models in the past [1, 7], we have chosen our cell models to be as minimal as possible, in an attempt to focus on the essential aspects of collective guidance. Earlier models have been created to study neural crest chemotaxis in vivo [81, 83, 95]; however, these have explicitly described chemotaxis arising from a “follow-the-leader” mechanism where single leader cells can sense a gradient [96], rather than through the collective mechanism we study here, where individual cells need not sense the gradient level.
We also mention that unlike many of the models discussed above, our model does not include an interaction designed to align a cell’s polarity with its neighbors’ motion [53, 63, 75] or its own velocity or displacement [30, 58, 87, 88], and these mechanisms are not necessary for the effects we describe here. Competition between the collective guidance mechanism and alignment mechanisms may be an interesting area for future study.
Our stochastic interacting particle model is relatively simple, which allows us to in some cases derive analytic results [7]. Many extensions of this approach are possible. Our model could be developed further for more quantitative comparisons by careful measurement of single-cell statistics in or out of a chemoattractant gradient [28, 97]; this could lead to nonlinear or anisotropic terms in Eq 2. Our description of contact inhibition of locomotion has also assumed, for simplicity, that contact with both the front and back of the cell is inhibitory; other possibilities may alter the collective dynamics of the cell cluster [23].
Our main findings are: 1) Cluster velocity and chemotactic index may reflect internal signal processing, and provide an experimental window into these processes. 2) We expect sufficiently large clusters undergoing collective guidance to either become increasingly slow or break up. 3) Strong adhesion between cells is not necessary for collective guidance to function if cells chemotax to a secreted molecule. 4) A balance of this co-attraction and graded contact inhibition of locomotion are necessary for efficient chemotaxis. 5) Co-attraction may also induce cluster rotation, and we have explicitly characterized the transition to rotation. 6) The combination of cluster rotation and cluster chemotaxis may induce systematic drifts that depend on cluster rotation.
|
10.1371/journal.pntd.0004926 | Using a Human Challenge Model of Infection to Measure Vaccine Efficacy: A Randomised, Controlled Trial Comparing the Typhoid Vaccines M01ZH09 with Placebo and Ty21a | Typhoid persists as a major cause of global morbidity. While several licensed vaccines to prevent typhoid are available, they are of only moderate efficacy and unsuitable for use in children less than two years of age. Development of new efficacious vaccines is complicated by the human host-restriction of Salmonella enterica serovar Typhi (S. Typhi) and lack of clear correlates of protection. In this study, we aimed to evaluate the protective efficacy of a single dose of the oral vaccine candidate, M01ZH09, in susceptible volunteers by direct typhoid challenge.
We performed a randomised, double-blind, placebo-controlled trial in healthy adult participants at a single centre in Oxford (UK). Participants were allocated to receive one dose of double-blinded M01ZH09 or placebo or 3-doses of open-label Ty21a. Twenty-eight days after vaccination, participants were challenged with 104CFU S. Typhi Quailes strain. The efficacy of M01ZH09 compared with placebo (primary outcome) was assessed as the percentage of participants reaching pre-defined endpoints constituting typhoid diagnosis (fever and/or bacteraemia) during the 14 days after challenge. Ninety-nine participants were randomised to receive M01ZH09 (n = 33), placebo (n = 33) or 3-doses of Ty21a (n = 33). After challenge, typhoid was diagnosed in 18/31 (58.1% [95% CI 39.1 to 75.5]) M01ZH09, 20/30 (66.7% [47.2 to 87.2]) placebo, and 13/30 (43.3% [25.5 to 62.6]) Ty21a vaccine recipients. Vaccine efficacy (VE) for one dose of M01ZH09 was 13% [95% CI -29 to 41] and 35% [-5 to 60] for 3-doses of Ty21a. Retrospective multivariable analyses demonstrated that pre-existing anti-Vi antibody significantly reduced susceptibility to infection after challenge; a 1 log increase in anti-Vi IgG resulting in a 71% decrease in the hazard ratio of typhoid diagnosis ([95% CI 30 to 88%], p = 0.006) during the 14 day challenge period. Limitations to the study included the requirement to limit the challenge period prior to treatment to 2 weeks, the intensity of the study procedures and the high challenge dose used resulting in a stringent model.
Despite successfully demonstrating the use of a human challenge study to directly evaluate vaccine efficacy, a single-dose M01ZH09 failed to demonstrate significant protection after challenge with virulent Salmonella Typhi in this model. Anti-Vi antibody detected prior to vaccination played a major role in outcome after challenge.
ClinicalTrials.gov (NCT01405521) and EudraCT (number 2011-000381-35).
| Typhoid fever is a common cause of febrile illness in tropical countries. Although currently available typhoid vaccines are moderately effective, they are not suitable for use in young children. Development of new vaccines is complicated as Salmonella Typhi, the causative bacteria, only infect humans. In this study, we used a recently developed human typhoid challenge model to directly assess the efficacy of a new oral vaccine candidate, M01ZH09, compared to placebo. A parallel group of participants were given 3-doses of licensed oral Ty21a vaccine as a positive comparator. We found that a single dose of M01ZH09 was not effective in preventing typhoid infection in our model, although significant effects were seen in delaying onset of infection and reducing bacterial numbers. Ty21a also failed to significantly protect against infection suggesting our model was particularly stringent. We discovered that anti-Vi antibodies, present in some individuals prior to vaccination, contributed significantly to preventing infection in some individuals, and when this effect was taken into account, M01ZH09 halved the risk of developing typhoid after being challenged. These results demonstrate the utility of human challenge models in assessing the efficacy of new typhoid vaccine candidates, and suggest that further development of M01ZH09 dosing or delivery strategies may produce better results. These results also support further development of Vi-based vaccines as a potentially preventive intervention.
| Typhoid fever, caused by Salmonella enterica serovar Typhi (S. Typhi), continues to be a major cause of global morbidity and poverty, particularly in areas without basic sanitation and limited access to clean water, and among non-immune travellers to those settings [1–3]. Despite causing an estimated 22 million new cases each year, vaccination to prevent typhoid infection has been little implemented [4, 5]. Licensed vaccines have demonstrated moderate efficacy in preventing infection in older children and adults, but are not suitable for use in young children and infants less than 2 years of age [6, 7]. Development of new efficacious vaccines is complicated by the human host-restriction of S. Typhi, the lack of clear correlates of protection, the scale required to run field trials of efficacy and uncertainty about estimation of vaccine impact due to suboptimal diagnostics. A human challenge model can be used to overcome some of these difficulties and can provide some direct estimation of efficacy in vaccine recipients who are deliberately exposed to the pathogen in a controlled setting [8].
M01ZH09 is a live attenuated oral vaccine constructed from the parent Ty2 strain by defined, independently attenuated deletion of the ssaV and aroC genes [9]. A single dose of M01ZH09 vaccine has proven to be well-tolerated and highly immunogenic in six previous phase I and IIa studies [10–13]. In particular, high levels of anti-lipopolysaccharide (LPS) antibodies were generated in response to vaccination in studies conducted in both low- and high-endemicity areas and in diverse age groups [10–13]. Evidence to support anti-LPS response as a useful protective parameter is limited, and mostly derived from observations made in endemic settings [14, 15]. Evaluation of typhoid vaccines in previous human challenge studies has been instrumental in their development, notably for Ty21a, which is also a live attenuated vaccine derived from Ty2 but does not constitutively express the Vi (Virulence) capsular polysaccharide and contains multiple additional genetic attenuations [16–18].
The aim of this study was to assess whether a single dose of oral M01ZH09 could protect healthy adult volunteers against developing typhoid infection in a challenge model, 28 days after vaccination. In our recently developed challenge model, ingestion of 104 CFU virulent S. Typhi Quailes strain bacteria resulted in a 65% attack rate in unvaccinated adult volunteers [19]. Suitability of the model for vaccine efficacy (VE) evaluation was assessed in parallel by using the 3-dose schedule of oral Ty21a vaccine as an open-label comparator group.
A randomised, double-blind, placebo-controlled trial was performed at the Oxford Vaccine Group in the Centre for Clinical Vaccinology and Tropical Medicine (Churchill Hospital, Oxford, UK) to assess the protective efficacy of a single dose of M01ZH09 compared to placebo against S. Typhi challenge 28 days after vaccination.
This phase 2b trial was sponsored and monitored by the Oxford University Clinical Trials and Research Governance Department, approved by NRES South Central–Oxford A (11/SC/0302) and conducted in accordance with the principles of the International Conference of Harmonisation, Good Clinical Practice guidelines. After study initiation (November 2011), an independent Data and Safety Monitoring Committee reviewed clinical and laboratory data relating to patient safety (months 1, 5 and 8) and interim unblinded analyses of VE (months 5 and 8). No changes to the study protocol or participant eligibility were recommended.
Potential participants from the community were approached using a variety of media including postal leaflets, e-mails, advertising posters and local newspaper and football programme advertising. Interested individuals were then invited to contact the study centre for further discussion and to receive written study information prior to invitation for eligibility assessment and enrolment. Eligible participants were healthy men and non-pregnant women, aged 18 to 60 years with no previous history of typhoid vaccination, infection or likely exposure to S. Typhi. All eligible volunteers were provided with detailed pre-study counselling and provided written informed consent. Following consent, participants were thoroughly evaluated for health problems by history, physical examination, blood screening and ultrasound examination of the gall bladder. A full description of the inclusion/exclusion criteria can be found in the study protocol (S1 Protocol).
We randomised participants to a double blind or open-label (Ty21a) arm in a 2:1 ratio, with further randomisation of the double blind arm to M01ZH09 or placebo (1:1 ratio). Randomisation lists were computer generated by permuted block randomisation with variable block sizes. Blinding was effected through identical packaging and allocation concealment by sequentially numbered sealed opaque envelopes. An independent study statistician (LMY) generated the randomisation sequence and provided sealed envelopes containing allocation codes. Participants and study staff remained unaware of group allocation until four weeks after the completion of the challenge phase, at which point specific, unblinded study team members revealed vaccine assignment to the participant only. These study team members took no part in performing the laboratory assays or data analysis.
A single dose of the study vaccine, M01ZH09 (Emergent BioSolutions, Wokingham, UK), containing 1x1010 CFU of live attenuated S. Typhi (Ty2 ΔaroC ΔssaV) ZH09 strain was re-suspended in 20mL NaHCO3[aq] solution prior to ingestion. Placebo, containing excipients only (M9S basal medium plus 10%(w/w) sucrose), was re-suspended in an identical fashion. Participants were fasted for one hour before and after vaccine ingestion. Participants in the open-label arm ingested three capsules of enteric-coated Ty21a vaccine (Crucell UK Ltd, High Wycombe, UK), containing not less than 2x109 CFU on alternate days, in accordance with manufacturer’s instructions [20].
Antibody secreting cell (ASC) responses to LPS, flagellin and Vi were measured at baseline and 7 days after vaccination by ELISpot assay. Briefly, peripheral blood mononuclear cells (PBMC) separated from venous blood, were dispensed in concentrations of 2.5 or 5.0x106 cells/mL to nitrocellulose plates pre-coated with lipopolysaccharide (S. Typhosa LPS, L2387; Sigma-Aldrich, Dorset, UK), Vi (Sanofi Pasteur, Maidenhead, UK), flagellin (prepared by isolation from S. Typhi Quailes strain and purification at the University of Maryland School of Medicine) or buffer only (negative control). After overnight incubation and wash steps, alkaline-phosphatase goat anti-human IgG, IgM and IgA antibodies were added. After further incubation spots were developed with alkaline phosphatase substrates. Spots were manually counted by two independent observers and expressed as spots/1x106 PBMC.
Antibody responses were measured 28 days after vaccination (immediately prior to S. Typhi challenge) and compared with those collected at baseline (pre-vaccination). Specific immunoglobulin G (IgG), IgA and IgM isotype responses to LPS and flagellin were measured in serum as previously described [19]. In addition, immunoglobulin G (IgG) responses to Vi were measured using a commercial ELISA kit (VaccZyme, The Binding Site Ltd, Birmingham, UK) according to the manufacturer’s instructions.
Four weeks after vaccination participants were challenged with S. Typhi, Quailes strain, as described previously [19]; day 0 was defined as the day of challenge. A target challenge dose of 1-5x104 CFU was used to achieve a 65% attack rate in participants allocated to placebo (S1 Fig). After fasting for 90 minutes participants ingested 1·2g/120mL NaHCO3 [aq] followed two minutes later by the challenge inoculum suspended in 0·53g/30mL NaHCO3 [aq]. Following challenge, participants were observed for a further 90 minutes prior to leaving the clinic.
Participants were reviewed at least daily at the study site for 14 days, except for days 2 and 4 after challenge when they were telephoned twice instead, or if additional visits were required. Assessments performed included clinical evaluation and microbiological assay of blood and stool samples as described below and in the study protocol (S1 Protocol).
The primary study objective was to assess the efficacy of M01ZH09 or Ty21a vaccines compared with placebo in preventing infection during the 2-weeks after S. Typhi challenge. Typhoid diagnosis (TD) was defined as either, a) oral temperature ≥38°C sustained for ≥12 hours or more after day 5 of challenge, b) a blood culture positive for S. Typhi taken after day 7 of challenge, or c) a blood culture positive for S. Typhi collected after day 5 plus objective symptoms or signs (including fever) of typhoid infection.
Severe typhoid fever was defined as a case fulfilling the criteria for TD with the addition of one or more of the following features: oral temperature recorded ≥40°C, systolic blood pressure ≤85mmHg, significant lethargy or confusion, a GI bleeding event or suspected/confirmed perforation, or any Grade 4 (‘life threatening’) laboratory abnormality.
Participants fulfilling the criteria for typhoid diagnosis were assessed by a physician and initiated on antibiotic treatment and other medication required for symptom control. Antibiotic treatment given either at TD or at day 14 (in those not developing features of infection) was ciprofloxacin 500mg twice daily for 14 days (first-line), or azithromycin 500mg once daily for 7 days (second-line). Following diagnosis, participants were reassessed at 6, 12, 24, 48, 72 and 96 hours, to ensure resolution of clinical symptoms and bacteriological cure. In the event that a first positive blood culture result was received beyond Day 14 after challenge (and thus after commencement of antibiotic treatment), a TD assessment was made and further visits were arranged as determined by the study investigator. Compliance with antibiotic treatment was determined by direct observation at each study visit and by daily telephone/text message reminders. Following completion of an antibiotic course, clearance of the challenge strain was confirmed by microbiological culture of at least two stool samples obtained at least one week apart, collected at least 3 weeks after completion of antibiotics.
Routine haematological and biochemical monitoring was performed using blood samples collected at challenge and at each visit thereafter. Blood (10mL) and stool samples were collected for bacterial culture at each visit, while quantitative blood culture was performed using 10mL of blood at the TD visit only. Cultures were performed by the local hospital accredited pathology laboratories according to national standard operating procedures [21–24], and as previously described [19].
In addition to regular clinical review, participants collected symptom data (solicited for headache, feeling generally unwell, loss of appetite, abdominal pain, nausea/vomiting, myalgia, arthralgia, cough, diarrhoea and constipation and any unsolicited symptoms) daily and twice-daily self-recorded oral temperature readings using supplied written diary cards for 7 days after vaccine and 28 days after challenge ingestion. Data from diary cards was confirmed and clarified with the participant at each review.
Safety measures instituted included 24-hour contact with a study doctor, involvement of the participant’s general practitioner, notification of the participant’s close-contacts of involvement in the study, and provision of a 24-hour emergency contact who could be approached if the participant could not be contacted. All screened individuals were consented for inclusion onto The Over-volunteering Prevention System Database (TOPS)[25].
Based on previous findings a challenge inoculum of 1-5x104 CFU was used to achieve an attack rate of 60–75%. Assuming a similar attack rate in the placebo group using the same TD definitions, then to demonstrate a protective effect of 83%, resulting in a reduction in attack rate to 10%, 21 participants would be needed per group. If the attack rate in the placebo group were to fall to 50%, then 30 participants would be needed per group to demonstrate a protective effect of vaccination of 80% with 90% power (1-β) at the 5% significance level (α). To include an additional 10% dropout rate, target enrollment was 33 individuals per group.
Statistical comparisons were made between vaccine groups and placebo using the per protocol population, which included all participants completing the 14-day challenge period. The study was not powered to compare the two active vaccine groups with each other and thus it was pre-specified that these comparisons would not be performed.
The frequencies of TD in M01ZH09 (primary endpoint) or Ty21a recipients were compared with placebo to calculate vaccine protective efficacy (VE, i.e. the percentage reduction in attack rate in those vaccinated compared to the unvaccinated group) and presented along with 95% confidence intervals. Sensitivity analyses compared VE under different definitions of typhoid fever. Time to diagnosis (hours), time to development of fever (≥38°C, hours), and time to bacteraemia (point at which blood culture was collected, hours) were summarised using the Kaplan-Meier method. A post hoc proportional hazards model assessed factors associated with time to typhoid infection or time to bacteraemia, including vaccine receipt, baseline antibody status and age, sex and travel to endemic regions.
Vaccine immunogenicity was assessed by log10-transformation of antibody levels measured by ELISA and ASC counts (with 0 counts given a nominal value of 0.25). Comparisons between groups were analysed using analysis of covariance (ANCOVA) models to adjust for pre-vaccination titres. Comparisons of bacterial load and other non-normally distributed variables were conducted using Mann-Whitney U tests. All reported p-values are two-tailed with statistical significance at p<0.05. Analyses were performed using Stata version 13.0 (StataCorp, Texas, USA) and SPSS Statistics version 22 (IBM, Portsmouth, UK).
The UK National Research Ethics Service provided ethical approval for the trial (Oxfordshire Research Ethics Committee A, 11/SC/0302), which was performed in accordance with the principles of the ICH-Good Clinical Practice guidelines and amendments. All trial participants provided written informed consent in accordance with the Declaration of Helsinki (S1 Checklist).
The study was registered at Clinicaltrials.gov (NCT01405521) and with the European Clinical Trials database (EudraCT 2011-000381-35).
Ninety-nine participants were enrolled and randomised to one of three vaccine groups between November 28, 2011, and June 27, 2012 (Fig 1, S1 Dataset). The demographic characteristics of each group were similar (Table 1), with an overall median (range) participant age of 30.2 (19–60) years, 64.6% male sex and 88.9% self-declared white British ethnicity. Of note, 29.6% of all participants reported previous short periods of travel (<6months duration) to areas known to be endemic for typhoid.
Of those randomised, 94/99 participants attended for vaccination. Compliance with fasting requirements and vaccine schedules was high (98.5%). Four weeks after vaccination 92/94 participants remained eligible and consented to challenge with S. Typhi. Challenge was performed a median (range) of 27 (21–33) days after vaccine course completion and the median (range) dose of S. Typhi bacteria ingested was 1.82 (1.46 to 2.66) x104 CFU (S1 Fig). Ninety-one participants successfully completed the 2-week challenge period, and were included in the efficacy and outcome assessments presented below.
Overall, 51/91 (56.0% [95% CI 45.2 to 66.4]) participants met the pre-specified criteria for typhoid infection during the 2-week period post-challenge. Of these, 47/51 (92.2% [95%CI 81.1 to 97.8]) diagnoses were confirmed by positive blood culture. In the placebo group, 20/30 (66.7% [95%CI 39.1 to 75.5]) were diagnosed with typhoid producing an attack rate similar to previous challenge studies (Table 2)[19]. After challenge, typhoid was diagnosed in 18/31 (58.1% [95%CI 39.1 to 75.5]) participants in the M01ZH09 and 13/30 (43.3% [95%CI 25.5 to 62.6]) in the Ty21a group, resulting in a calculated VE [95%CI] of 13% [-29% to 41%]) and 35% [-5% to 60%], respectively (Table 2).
Overall, a similar number of participants in each group were diagnosed by either clinical (i.e. temperature ≥38°C for ≥12 hours) or blood culture criteria. This split in diagnosis type was largely due to the time required for blood culture incubation and clinician notification; the majority of participants (40/47, 85%) were bacteraemic at sampling points prior to the onset of fever (≥38°C).
Estimates of VE were sensitive to changes in the definition of TD, ranging from 3% to 23% for M01ZH09 and from 22% to 67% for Ty21a (Table 2). Of particular clinical interest, the finding of fever (≥38°C) with a subsequent positive, confirmatory blood culture (an approximation for passive surveillance in field-testing conditions) resulted in VEs of 52% [-25% to 81%] and 80% [16% to 95%] for either single dose M01ZH09 or 3-doses of Ty21a, respectively.
By day 8 post-challenge, 8/31 (26%) M01ZH09 recipients and 12/30 (40%) Ty21a participants had reached the infection endpoint compared with 15/30 (50%) placebo recipients. The Kaplan-Meier median [95%CI] time elapsed between challenge ingestion to development of fever (≥38°C) or positive blood culture sampling was 265 [127 to 403] hours in M01ZH09 recipients or 172 [109 to 236] hours in placebo recipients, respectively (p = 0.249, log-rank; Fig 2).
Time to first positive blood culture in participants receiving M01ZH09 or Ty21a was delayed compared with placebo (p = 0.042 and 0.047 respectively, log-rank; Fig 3A). Time to onset of fever in participants was similar across groups (p = 0.1; Fig 3B).
Severe typhoid infection was diagnosed in 11/91 (12%) study participants, with rates of 3/31 (10%) in M01ZH09, 4/30 (13%) in placebo and 1/30 (3%) in Ty21a recipients, respectively. Most severe diagnoses (6/8) were defined by abnormalities in measured blood parameters only (see Table 2).
Frequent and marked physiological changes and symptoms were found in the majority of participants after challenge, particularly in those developing infection (Fig 4). Diary card symptoms in blinded participants showed that participants receiving M01ZH09 had fewer, milder symptoms compared to placebo (Fig 5, S1 Table). There were no clear differences in the physiological parameters of participants diagnosed with infection among the different vaccine groups (Figs 6 and 7).
Over 1000 blood culture samples were collected from participants after challenge; due to the later clinical presentation, M01ZH09 participants tended to have a greater number of samples obtained than those in either of the other two groups (Table 3). The average number of positive blood cultures collected from each typhoid-diagnosed participant was similar across groups (2.25, 2.70 and 2.63 for those in M01ZH09, placebo and Ty21a groups, respectively). Of note, the median [IQR] duration of blood culture positivity (measured from time of first positive to last positive sampling) was similar between the M01ZH09 and placebo groups (28.7 [0 to 41.5] and 27.7 [21.4 to 31.3] hours, respectively) but longer in Ty21a vaccine recipients (46.3 [28.4 to 53.3] hours).
The quantitative S. Typhi load (available for 41/51 diagnosed participants) at diagnosis (prior to antibiotic initiation) was significantly lower in the blood of M01ZH09 and Ty21a recipients compared with placebo (median [IQR] bacterial load CFU/mL, M01ZH09: 0.13 [0.05–0.80], placebo: 1.30 [0.30–5.40]; p = 0.012; Ty21a: 0.05 [0.05–0.88]; p = 0.011, Mann-Whitney U test; Fig 8).
Early shedding of S. Typhi in participant stool samples (within the first 72 hours after challenge) was frequent (44/91, 49% participants) and similar between vaccine groups (M01ZH09 55%, placebo 49%, Ty21a 47% of participants; Table 4). Identification of early shedding was significantly associated with subsequent diagnosis of typhoid infection (relative risk [95% CI] 1.71 [1.15 to 2.53], p = 0.005, Chi-square test). From 72 hours after challenge onwards, S. Typhi was cultured from 92/790 (12%) stool samples collected from 41/91 (45%) participants. Overall, no differences in numbers of participants shedding S. Typhi were found in those diagnosed compared with participants who did not develop evidence of infection, either after day 4 or at any time point overall (p = 0.089 and p = 0.370 respectively, Chi-square test).
Antibiotic initiation rapidly terminated stool shedding, with no positive stool cultures being obtained after the first dose of treatment had been taken. No evidence of convalescent or long-term carriage of S. Typhi was found in any of the two follow-up stool cultures obtained.
Both active vaccines and placebo were well tolerated and no Serious Adverse Events were identified related to vaccine receipt. Participants from each group reported a similar number and severity of symptoms during the 7 days after vaccine receipt (Fig 9 and S1 Table).
Pre-vaccination (day -28) anti-LPS, -H and -Vi ASC levels measured by ELISpot and antibody titres measured by ELISA were similar between groups (S2 Table and S3 Table). One week after a single dose of M01ZH09 or three doses of Ty21a, most participants showed a significant increase in anti-LPS and anti-H ASC isotype assays compared with placebo (Fig 10A and 10B, Table 5). Corresponding significant increases were seen in all anti-LPS and anti-H antibody isotypes between day -28 and day 0 (prior to challenge ingestion) in response to M01ZH09 vaccine when compared with placebo (Fig 10C and 10D, Table 6). In contrast, vaccination with Ty21a resulted in a borderline significant increases in anti-LPS IgG only (p = 0.047, ANCOVA).
No significant increases in anti-Vi IgG antibody levels were found in response to vaccination. Of note, a range of anti-Vi IgG antibody titres was found at baseline. These included 6/32 (19%), 12/30 (40%) and 8/29 (28%) participants in the M01ZH09, placebo and Ty21a vaccine groups respectively, with levels detectable above the lower detection limit (LLD, 7.4EU/mL) and the 75% percentile measurement found in a UK adult blood donor population sample (n = 81)[19].
Despite finding robust overall humoral anti-LPS and anti-H responses to vaccination, these responses failed to confer protection against challenge (Fig 11A and 11B). While anti-Vi IgG antibody titres were unaffected by the vaccines used in this study, baseline titres were significantly higher in those subsequently found to be protected after challenge (Fig 11C). In an exploratory proportional hazards analysis, baseline parameters were assessed for their impact on developing typhoid during the 2-week challenge follow-up period. Anti-Vi antibody titre prior to vaccination was the only variable found to be predictive of TD. Challenge dose, fold increase in anti-LPS or anti-H antibody titres due to vaccination, sex and prior travel to endemic regions were not significant. When baseline anti-Vi antibody levels were accounted for in the model, the hazard rate of TD in the M01ZH09 group was approximately half that of the placebo group (0.513, p = 0.048) and, similarly, time to bacteraemia rates were 60% lower in both active vaccine groups compared with placebo (Table 7).
For the first time in over 40 years we have demonstrated the utility of a human challenge study in the assessment of a new typhoid vaccine candidate, M01ZH09. In phase I and IIa studies performed to-date, in low and high transmission settings and in both adults and children, a single dose of M01ZH09 has proven to be well tolerated and highly immunogenic [10–13]. In this study, we found that neither a single dose M01ZH09 nor three doses of Ty21a given 28 days prior to challenge resulted in significant overall protection. Both vaccines caused a significant reduction in the microbiological burden of infection and alteration of the clinical disease profiles, however, when compared with those participants receiving placebo only. With adjustment for baseline anti-Vi titre, M01ZH09 vaccine receipt resulted in significant protection against developing typhoid fever during the two-week period after challenge.
We successfully demonstrated the reproducibility of this model for vaccine assessment by recreating the attack rate of 67% in placebo recipients, identical to that described in our preliminary dose-escalation study [19]. The selected challenge dose (and anticipated attack rate) was chosen to avoid exposing excessive numbers of participants to a potentially infectious pathogen while also attempting to reduce the risk of overwhelming potentially protective vaccine-induced responses, as has been observed in some of the historical studies performed [26]. While we were able to document many of the features of clinical typhoid fever in our volunteers at this dose, this high attack rate in placebo recipients means that our model is likely to be a more stringent test of VE than the historical Maryland model. In these studies, challenge with 105 CFU S. Typhi Quailes strain 5–9 weeks after vaccination with 5–8 doses of Ty21a resulted in high rates of anti-LPS antibody seroconversion and 87% protective efficacy, albeit without the constraints of a two-week follow-up period [16].
To provide some measure of the ability of our challenge model to evaluate VE, we incorporated an open label Ty21a arm using the standard European 3-dose schedule. Despite resulting in expected levels of immunogenicity [27], we found that Ty21a vaccination resulted in a protective efficacy of only 35% after challenge, a point estimate that did not reach significance. Reasons for lower efficacy compared with that found in the historic Maryland studies might include different dosing schedules and vaccine formulations used, the background immunity of study participants, challenge doses and the methods used and the availability of automated blood culture technology. It is interesting to note however, that a 35% VE corresponds to that shown for the 3-dose schedule of Ty21a at Year 1 in a recent meta-analysis including 20,543 participants [6]. Likewise, with a similar definition for TD (fever with subsequent microbiological confirmation) as was used in the original vaccine/challenge efficacy study by Gilman et al, Ty21a VE reached 80% [95%CI, 16 to 95] in this study compared to 87% [95% CI, 47 to 96][16].
Lower Ty21a efficacy is also likely to reflect the higher challenge dose (and thus attack rates) used in our study. Of note, the corresponding attack rate in the Maryland Ty21a studies was 53% after challenge with 105 CFU in non-vaccinated volunteers [16]. During the field trials of Ty21a, VE was lowest in areas with higher infection rates and therefore probably higher exposure doses. In Indonesia, where rates of infection were 1,206/100,000 for example, VE ranged from 52.7% [95%CI. 23.9 to 58.6] in 3–19 year olds given three doses of liquid Ty21a to 23.6% [95%CI, -78.8 to 67.3] in 20–44 year olds given 3 doses of enteric-coated Ty21a during 30 months of follow-up [28]. Reasons given for this lower efficacy in the older age group included a lower number of cases in the placebo arm, possible variation in circulating S. Typhi strain types or vaccine production. It is interesting to note both that the same enteric-coated formulation was used in our study, and the relatively close phylogenetic relatedness of the Quailes strain to more recently found Indonesian strains [19].
In keeping with the previous studies, we found M01ZH09 to be highly immunogenic and well tolerated when given as a single oral dose. In addition to anti-LPS responses, significant increases in anti-H ASC and antibody titres were also seen in M01ZH09 recipients, in contrast to those vaccinated with Ty21a. Of note, a single dose of M01ZH09 contained approximately twice as many CFU than 3-doses of Ty21a (1x1010 versus 6x109 CFU) although both vaccine formulations would also contain non-viable bacteria. Despite the immune responses seen, and against the expectation that high LPS antibody levels might correlate with protection against Salmonella infection [14, 15, 26], vaccination failed to confer significant protection against typhoid infection after challenge. Other beneficial effects of M01ZH09 vaccination were seen however, suggesting that active vaccine-mediated mechanisms were in effect. These included a noticeable delay in infection onset (effectively extending the clinical incubation period) characterised as a delay in the appearance of fever and symptoms. It is noteworthy that delay in onset of infection is used in many vaccine evaluation challenge models as evidence of VE [8]. There was also a measurable reduction in the microbiological burden of infection, both in time to bacteraemia, the level of bacteraemia at TD and some reduction in stool shedding in the few days preceding TD. While the precise immune mechanisms responsible for protection against typhoid infection are still under investigation, these data confirm that, at least in typhoid-naïve individuals, anti-LPS and anti-H responses may moderate the onset or severity of infection symptoms, but are not sufficient alone to prevent invasion and systemic dissemination.
Generation of anti-LPS antibodies is frequently used as an indicator by which to select potential oral vaccine candidates in addition to being the focus of several Salmonella-based vaccine programmes. Support for the protective role of anti-LPS responses comes from both field trials, in which a correlation has been found between rates of anti-LPS seroconversion and subsequent risk of typhoid infection [14], and mouse/S. Typhimurium models in which monoclonal antibodies confer protection against homologous challenge [29]. A central argument to using live attenuated vaccines in endemic settings is that protective immune responses may be boosted by background exposure to S. Typhi in food or water supplies or due to persistence of antigen in reticuloendothelial niches. This is likely responsible for the finding in field trials that VE often increases over time. The two consequences for our study relate to the very low levels of background LPS exposure, possibly resulting in suboptimal vaccine responses and thus less protection to challenge, and the short window between vaccination and challenge. It is worth noting that infants and young children are less likely to have experienced this background exposure and therefore multiple vaccine doses may be more efficacious.
Post-hoc multivariate analyses to explore factors contributing to the development of typhoid after challenge revealed the apparent protective contribution of vaccination when adjusted for baseline anti-Vi IgG antibody titre. Participants were carefully selected to be both typhoid and typhoid-vaccine naïve, by taking self-reported histories and confirming individuals’ vaccine and medical histories with their general practitioners. The finding that 29% of participants had some measurable Vi antibody detectable (albeit at low levels) at baseline was therefore unexpected. Reasons for this may include colonisation/exposure to cross-reactive or homologous antigens (Citrobacter freundii 5396/38) or that those with very short travel histories may have been exposed, albeit briefly and sub-clinically to S. Typhi whilst abroad. Notably there was no correlation found between overseas travel and detectable anti-Vi IgG. If confirmed in future planned studies, evidence of the protection afforded by even low levels of anti-Vi IgG supports current Vi polysaccharide vaccine recommendations by the World Health Organisation and prevention strategies in travellers and high-risk populations in endemic regions [7], and for the development of newer more immunogenic Vi-conjugate vaccines.
Alternatively, low-level anti-Vi responses may be a marker for other cross protective immune responses. Additional host mucosal and adaptive immune responses are likely to play a major role in the protection afforded against infection by live-attenuated vaccine strains. These include secretory IgA, classical and non-classical (HLA E-restricted) CD8+ cytotoxic T cells [30], mucosal associated invariant T (MAIT) cell responses [31], regulatory T cell function activation [32], and functional properties of monocyte and dendritic cells [33], and, as yet incompletely characterised alterations in the gut microbiome [34]. Maturation of several of these non-humoral responses may take longer than the 28-day period allowed between vaccination and challenge in this study [30]. Additionally, thus far we have only explored the major surface antigenic determinants (LPS, Vi and flagellin). Additional serum and secreted antibody responses (to, for example, outer membrane proteins or GroEL) are also likely to play a role in protection against infection. Likewise, functional antibody activity and antibody avidity may be important in prevention of disease [15, 35].
To provide a uniform repeatable model of typhoid challenge, non-immune, typhoid-naïve adults were specifically selected. While these findings may therefore not be directly extrapolated to endemic field settings, they are likely to be relevant to travellers venturing to those regions. The VE findings are therefore also likely to be an underestimate of efficacy in an endemic setting, due to reasons of background exposure and antigen durability.
A major advantage of performing vaccine evaluation in a closed human challenge study, is recognition of subclinical phenotypes of disease. While this allows a more accurate estimation of vaccine effect, it also makes calculation of VE more susceptible to the disease endpoint definitions used. Our study incorporated demonstrable bacteraemia in the primary endpoint definition; previous challenge studies and field efficacy studies predated the development of more sensitive automated culture techniques and used a passive surveillance system to detect illness in the community participants. We therefore likely ‘overcalled’ those with infection, resulting in an apparently higher attack rate in our highly-controlled, intensively sampled participants than may have been found in these historical studies and had the vaccine been assessed in under field conditions.
In this unique study, we demonstrate the safety and utility of an ambulant outpatient human challenge study in evaluating the efficacy of a new typhoid vaccine candidate. While M01ZH09 failed to demonstrate significant protection in the per protocol analysis, post hoc analyses provide the intriguing possibility that this single dose vaccine may provide up to 50% protection in this stringent challenge model. That low level anti-Vi antibody appears to be protective supports current efforts to develop a conjugate-Vi vaccine suitable for use in younger children, however the emergence of Vi-negative S. Typhi strains and S. Paratyphi as underscores the importance of pursuing alternative strategies, including the development of live-attenuated vaccines. If a single dose of M01ZH09 reduces the risk of infection after challenge by half and has an impact on shedding, and therefore potentially transmission, oral vaccination could be a readily delivered public health control strategy.
|
10.1371/journal.pbio.2003769 | Extreme genome diversity in the hyper-prevalent parasitic eukaryote Blastocystis | Blastocystis is the most prevalent eukaryotic microbe colonizing the human gut, infecting approximately 1 billion individuals worldwide. Although Blastocystis has been linked to intestinal disorders, its pathogenicity remains controversial because most carriers are asymptomatic. Here, the genome sequence of Blastocystis subtype (ST) 1 is presented and compared to previously published sequences for ST4 and ST7. Despite a conserved core of genes, there is unexpected diversity between these STs in terms of their genome sizes, guanine-cytosine (GC) content, intron numbers, and gene content. ST1 has 6,544 protein-coding genes, which is several hundred more than reported for ST4 and ST7. The percentage of proteins unique to each ST ranges from 6.2% to 20.5%, greatly exceeding the differences observed within parasite genera. Orthologous proteins also display extreme divergence in amino acid sequence identity between STs (i.e., 59%–61% median identity), on par with observations of the most distantly related species pairs of parasite genera. The STs also display substantial variation in gene family distributions and sizes, especially for protein kinase and protease gene families, which could reflect differences in virulence. It remains to be seen to what extent these inter-ST differences persist at the intra-ST level. A full 26% of genes in ST1 have stop codons that are created on the mRNA level by a novel polyadenylation mechanism found only in Blastocystis. Reconstructions of pathways and organellar systems revealed that ST1 has a relatively complete membrane-trafficking system and a near-complete meiotic toolkit, possibly indicating a sexual cycle. Unlike some intestinal protistan parasites, Blastocystis ST1 has near-complete de novo pyrimidine, purine, and thiamine biosynthesis pathways and is unique amongst studied stramenopiles in being able to metabolize α-glucans rather than β-glucans. It lacks all genes encoding heme-containing cytochrome P450 proteins. Predictions of the mitochondrion-related organelle (MRO) proteome reveal an expanded repertoire of functions, including lipid, cofactor, and vitamin biosynthesis, as well as proteins that may be involved in regulating mitochondrial morphology and MRO/endoplasmic reticulum (ER) interactions. In sharp contrast, genes for peroxisome-associated functions are absent, suggesting Blastocystis STs lack this organelle. Overall, this study provides an important window into the biology of Blastocystis, showcasing significant differences between STs that can guide future experimental investigations into differences in their virulence and clarifying the roles of these organisms in gut health and disease.
| Blastocystis are unicellular eukaryotic organisms related to algae and some plant pathogens. They are common constituents of the human gut microbial community, colonizing approximately 1 billion humans worldwide. Whether their presence is harmful or not continues to be hotly debated. Part of the uncertainty stems from the fact that at least 17 subtypes have been identified from various mammalian hosts, including 9 from humans. To better characterize and understand Blastocystis, we have sequenced and annotated the genome for subtype 1 and compared it with previous genomic results for subtypes 7 and 4. The comparisons revealed considerable differences between the 3 sequenced subtypes for a number of genomic features like DNA base composition, size of genome, number of genes, and number of introns. We also examined various biochemical pathways and cellular systems in the context of the full gene complement to better understand the biology of Blastocystis, including some of its more unusual features like a mitochondrion related organelle. We also identified subtype-specific gene family expansions that may be related to virulence. Finally, we showed that Blastocystis appears to have most of the genes necessary for sexual reproduction. This study provides resources and hypotheses for future investigations into the biology and potential pathogenicity of these common gut microbes.
| Blastocystis is a genus of atypical, nonflagellated, anaerobic stramenopiles commonly inhabiting the intestinal tract of humans and other animals. The majority of stramenopiles, which, along with the Alveolata and Rhizaria, are members of the eukaryotic supergroup known as SAR, are marine biflagellated cells with tubular hairs on their surface—Blastocystis has none of these characteristics. The number of humans infected with Blastocystis globally has been estimated at over 1 billion [1], with prevalence being higher in developing than in developed countries [2]. The infective stage of Blastocystis is an environmentally resistant cyst, with the most common mode of transmission being the fecal-oral route. The pathogenicity of Blastocystis is controversial, but colonization with this organism has been linked to gastrointestinal symptoms, including diarrhea, abdominal pain, nausea, and irritable bowel syndrome [3]. Blastocystis is capable of becoming established in the gut and is difficult to eradicate via pharmacological interventions [3]. However, a causal link between the presence of the organism and disease symptoms has not been established [1,3], and some authors argue that Blastocystis is a part of a healthy gut microbiota [4].
At the genetic level, Blastocystis is remarkably heterogeneous. Many morphologically similar but genetically distinct lineages of Blastocystis have been identified, based primarily on sequences of their small subunit (SSU) ribosomal RNA genes [5]. Seventeen lineages have been isolated from mammals and birds to date and are referred to as “subtypes” (STs); the inter-ST divergence of the SSU rRNA gene is at least 3%. Blastocystis STs can have a remarkably broad host range and are almost never found exclusively in 1 host [5,6]. Only STs 1–9 have been found in humans to date, with STs 1–4 being responsible for around 90% of all human cases examined [6].
The high degree of genetic diversity is a confounding factor in establishing whether Blastocystis is a pathogen. Recent in vitro and in vivo molecular investigations have identified hydrolases and proteases as candidate virulence factors. Blastocystis proteases cause cleavage and degradation of immunoglobulin A (IgA) secreted by the host, may disrupt the intestinal epithelial barrier, and increase production of pro-inflammatory cytokines [2,7,8]. However, there is substantial inter-ST variation in adhesion to enterocytes, disruption of intestinal epithelial tight junctions, activation of pro-inflammatory cytokines, and the ability to scavenge nitric oxide. Virulence factor variability presumably extends to the intra-ST level [9] because the same ST is commonly found in both symptomatic and asymptomatic hosts. The highly variable clinical presentations attributed to Blastocystis could potentially be due to colonization with different STs or strains of the organism and untangling the relationships remains an area of active research and concern [10–12].
Recently, researchers have begun to assess the relationship between Blastocystis colonization and the composition of the prokaryotic gut microbiota. While several studies indicate Blastocystis may be associated with a more diverse and “healthy” microbiota [13,14], others have reported an association between Blastocystis colonization and a decrease in protective bacteria in the gut [15] or no differences in microbiota between Blastocystis-positive and -negative patients [16].
Despite the unanswered questions regarding its potential clinical relevance, studies of Blastocystis genomes are still in their infancy. Currently, several mitochondrial genomes (STs 1–4 and 6–9 [17–20]) and 2 high-quality draft nuclear genomes (STs 4 and 7 [21,22]) are available, as are genome survey data for additional STs (STs 2, 3, 6, 8, and 9 [13]).
The ST7 Blastocystis nuclear genome, obtained using a culture established from fecal matter from a symptomatic human, was the first to be sequenced [21]. The study reported a genome size of 18.8 Mb with 6,020 protein-coding genes. The authors described, among other things, a complex "mitochondria-like" organelle, effector proteins possibly involved in adaptation to a parasitic lifestyle, and a suite of secretory proteins that could have the potential to alter host physiology. They also detailed the genome architecture of ST7, finding it to be highly compact and having numerous duplicated blocks of genes.
More recently, an ST4 genome obtained from a laboratory rodent in Singapore was published [22]; its size was 12.91 Mb and a set of 5,713 protein-coding genes was predicted. There were no detailed analyses of the genome structure or of the genes themselves. In 2015, draft assemblies for the nuclear genomes of STs 2, 3, 4, 6, 8, and 9 were released to public databases. None of these assemblies have been annotated with predicted genes and no analyses have been reported to date beyond their use in a microbiome analysis [13].
Here, we report a draft genome sequence and transcriptome analysis of Blastocystis ST1, NandII strain from a symptomatic human, and compare it to the published genomes of Blastocystis ST7 [21] and ST4 [22]. We first provide a general overview and statistics for the NandII genome and perform high-level comparisons between the 3 genomes. We then present findings derived from extensive manual annotation, with a focus on genes that are potential host effectors, and highlight significant genomic differences between the STs. Our study provides a strong framework for subsequent molecular and cellular investigations of the role of Blastocystis in gastrointestinal health and disease and of its impact on the microbial communities of the gut.
The assembled size of the Blastocystis ST1 genome is 16.5 Mb spread across 580 scaffolds (Table 1). In comparison, the assembly of ST7 is composed of 54 scaffolds and 18.8 Mb, while the ST4 assembly is 12.9 Mb and has 1,301 scaffolds. The substantially lower number of scaffolds in the ST7 assembly reflects the longer reads of the Sanger sequencing technology used versus the shorter reads generated in next-generation sequencing (454 and Illumina) used in the cases of ST1 and ST4. Although the assembly for ST7 is substantially larger with longer scaffolds, the fold coverage, gene complement, and transcriptome coverage suggest that the difference in size between the strains is real.
The substantial genomic differences between ST1, ST4, and ST7 are not limited to size. The GC content varies by 15% between the 3 STs (39.6% in ST4, 45.2% in ST7, and 54.6% in ST1), which is significantly different when compared to GC-content variation in some parasites: the GC content difference among 3 strains of Giardia is only 2% [23], as is the difference among 4 species of Leishmania [24]. More biologically relevant is the variation among the Blastocystis STs in terms of overall gene numbers and gene structure. ST1 has 524 more protein-coding genes than were identified in ST7 and 831 more than in ST4 (Table 1). ST1 has the largest average gene size, the highest percentage of genes with introns, the highest average number of exons per gene, and the largest percentage of its genome devoted to encoding proteins, whereas ST7 has the lowest figures for these same attributes (Table 1). Although ST7 was characterized as being a compact genome [21], ST1 is demonstrably more compact, with an average intergenic spacer size of 615 bps versus 1,687 bps in ST7. ST4 tends to fall in the middle for these various measures, although it has more genes/kb, 0.44, than ST1 (0.39) or ST7 (0.32) because it has the smallest genome size.
While these differences in the number and organization of genes in different STs likely reflect real differences between their genomes, they may to some extent be impacted by the procedures used to predict gene models in the 3 genome assemblies. Particularly noteworthy is that ST7 and ST4 were annotated before the recognition of the phenomenon of polyadenylation-mediated creation of stop codons that occurs in Blastocystis (see below). Hence, the statistics concerning the different genomic elements in ST7 and ST4 (e.g., genes, exons, introns, intergenic spaces) reported in this and subsequent sections must be considered current estimates that may change after revised annotation.
A striking structural feature of Blastocystis STs that is unique amongst eukaryotic nuclear genomes is the polyadenylation-mediated generation of termination codons [25]. This process involves the creation of a functional stop codon by 3′-polyadenylation of mRNAs in a subset of genes, whereby the first 1 or 2 adenines of the poly-A tail appended to terminal uracil-adenine/uracil-guanine or uracil respectively, complete a termination codon missing in the actual gene sequence. Importantly, the addition of the poly-A tail occurs upstream of any possible canonical stop site in the underlying genomic sequence in these genes. Because most gene-finding algorithms are predicated on the presence of standard stop codons, this unique feature in Blastocystis can lead to erroneous gene models with problems such as overly long protein sequences, overlapping genes, chimeric genes, and the introduction of false introns to reach the next downstream stop codon.
The phenomenon was first investigated in ST7 using expressed sequence tag (EST) data [25]. The authors concluded that potentially 15% of ST7 genes have stop codons generated by polyadenylation. They also found some examples of the same process in a preliminary draft genome of ST1. With the full genome of ST1 available, as well as RNA-sequencing (RNA-Seq) data, we investigated the extent of the phenomenon. Appropriate stop codons generated by polyadenylation were found in 1,693 protein-coding genes, representing 26% of the protein-coding genes in ST1. The 15% suggested for ST7 may be an underestimate because mapped RNA-Seq data are far superior to ESTs for confirming whether the stop codon is generated via polyadenylation.
Because this phenomenon was not identified prior to the gene-finding process for the Blastocystis ST7 genome, a number of the initial gene models were incorrect [25]. At present, it is unclear whether the gene models for ST4 suffer from similar problems or even whether the same polyadenylation-mediated generation of termination codons occurs. The sites for polyadenylation appear to be linked to a highly conserved motif (TGTTTGTT) usually found 5 bases downstream of the nucleotide preceding the poly-A tail. A search for this motif indicates that it is abundant (Table 2) in Blastocystis. All 3 genomes have roughly the same number of motifs when genome size is taken into account. ST1 has 1 site per 3,097 bps, ST4 has 1 per 2,751 bps, and ST7 has 1 per 2,782 bps. A random selection of other stramenopile genomes showed that Thalassiosira pseudonana has 1 motif per 17,754 bps, while Phytophthora infestans has 1 site per 37,441 bps, suggesting that in these 2 genomes, the presence of the motif is random.
Because ST4 has a very similar motif complement to ST1 and ST7, it is likely that it too uses the polyadenylation process to generate some of its termination codons. Also suggestive are the 95 gene model pairs that overlap in ST4 (Table 1). An examination of these overlapping coding regions found motifs at locations appropriate to allow separation of the genes in most of the 95 cases. For example, the 3′-ends of KNB46045 and KNB46046 overlap by 37 bps (S1 Fig). Forty-four bps upstream of the annotated stop codon for KNB46045 is the conserved motif TGTTTGTT, which, if used to direct the generation of a new termination codon, would eliminate the overlap between the 2 genes.
Future studies of additional Blastocystis STs need to take polyadenylation-mediated stop codons into account. The mechanism has been demonstrated to be active in ST1 and ST7, suggesting that the mechanism evolved prior to the divergence of these 2 STs.
The ST1 genome is substantially more intron rich than that of ST7 or ST4. It has 35,412 predicted introns compared with 18,200 in ST7 and 24,093 in ST4 (Table 1). Consequently, ST1 has the highest percentage of genes with introns and the highest number of introns per gene. The size of introns in ST1 is also less variable (Fig 1), with over half being 30 bps in length. In ST4 and ST7, the percentage of introns with a length of 30 bps is considerably lower. ST7 shows greater variation in intron size, with a much lower peak at 30 bps and higher percentages of introns in the size range 31–38 bps. ST1's relative intron richness may in part reflect methodological differences in gene calling. Exon/intron boundaries in ST1 were corrected using RNA-Seq transcriptome data, which were not available for ST4 and ST7. Depending on the parameters used to identify a "typical" gene and its exon/intron structure during the automated gene-calling process, having RNA-Seq data would tend to result in finding more real introns and annotating more introns with the correct boundaries.
Of the approximately 35,000 annotated introns in the Blastocystis ST1 nuclear genome, the vast majority (98.3%) represent the standard (or U2 type) spliceosomal introns characterized by GT-AG boundaries and spliced by the so-called major (U2) spliceosome containing U1, U2, U4, U5, and U6 small nuclear RNAs (snRNAs) containing small nuclear ribonucleo proteins. Two additional, much less abundant intron categories are also apparently spliced by the major spliceosome. One has GC instead of GT as the 5′ intron boundary and constitutes about 0.5% of all introns in both ST1 and ST7 (Fig 1), which is on par with the proportion of this intron found in metazoan or plant genomes [26]. The second category, characterized by the GA dinucleotide at the 5′ border, is even more sparse, with 59 such introns (i.e., only around 0.16% of all introns) identified in ST1 and only 9 supported by EST data in ST7. The existence of such introns is not without precedent, as such GA-AG introns were previously identified in the genome of the dinoflagellate Symbiodinium minutum [27].
The final intron category in Blastocystis corresponds to the minor, or U12-type, introns, characterized by boundaries typically exhibiting the dinucleotides AT-AC (Fig 1) and spliced by the minor (U12) spliceosome containing U11, U12, U4atac, and U6atac snRNA molecules. A systematic analysis of the Blastocystis ST1 genome revealed 346 U12-type introns in 319 genes. In addition, genes for all 4 snRNAs and for all 7 protein subunits (20K, 25K, 31K, 35K, 48K, 59K, and 65K) specific to the U12 spliceosome were identified [28] (S1 Table). U12-type introns and U12 spliceosome-specific components were not noted in the report on the ST7 genome [21], but both can be identified in the genome sequence of that ST (Fig 1, S1 Table). Hence, Blastocystis represents a previously missed eukaryotic lineage that has retained U12-type introns and the associated splicing machinery. In stramenopiles, only oomycetes have so far been known to exhibit both types of spliceosomal introns, whereas only the major “standard” type has been retained in the other lineages [29].
Blastocystis ST1 was found to have 6,544 protein-coding genes, in contrast to 6,020 in ST7 and 5,713 in ST4 (Table 1). An examination of the amino acid identities between putative orthologs reveals an extraordinary degree of dissimilarity among the Blastocystis STs (Fig 2). The median sequence identity of aligned regions of orthologs among STs 1, 4, and 7 ranges from 59% to 61% (S2 Table). The differences among the 3 STs exceed those observed for orthologs from pairs of species from parasitic protistan genera such as Cryptosporidium (C. parvum–C. hominis, Alveolata), Leishmania (L. major–L. infantum, Excavata), and Theileria (T. parva–T. annulata, Alveolata) (Fig 2) and indeed among Giardia strains (WB, GS, P15; Excavata) (Fig 2, S2 Table). The dissimilarity is comparable to that between species of Plasmodium (P. falciparum–P. knowlesi, Alveolata) and Trypanosoma (T. cruzi–T. brucei, Excavata). The extent of the dissimilarity among the Blastocystis STs supports the contention that they should be considered at least equivalent to separate species, particularly when placed in the context of other protistan pathogens. The lack of morphologically distinguishing traits and low correlation between ST and host [30] will continue, however, to make the taxonomy of Blastocystis challenging.
ST1 and ST4 genes are the least alike at the protein level while ST7 protein-coding sequences are usually more similar to ST1 than to ST4 genes (Fig 2). This is in line with recent phylogenetic analyses [5,20,31] that place ST4 in a clade sister to one that includes ST1 and ST7. There are also marked differences in gene content between STs. The percentage of genes present in ST1 but not in ST4 is significantly lower than the percentage of genes in ST1 not present in ST7 (Fig 3). Similarly, ST4 has far fewer unique genes when compared with ST1 than with ST7. The percentage of unique genes in ST7 is virtually identical in comparisons with ST1 and ST4. The differences in gene complement between the Blastocystis STs greatly exceed the differences between selected parasitic protistan species pairs (Fig 3). Beyond a common core of genes presumably devoted to common housekeeping tasks, the Blastocystis STs possess a substantial number of genes unique to each ST. The consequences of these major genetic differences between the STs are discussed below.
Kinase enzymes modify target proteins via phosphorylation and participate in the regulation of cellular pathways, particularly those involved in signal transduction [32]. Our analysis of Blastocystis ST1 identified 221 kinases, which were classified according to kinase.com [33]. Representatives of most kinase groups are similarly distributed among STs 1, 4, and 7 (Fig 4) except for 2 clades showing lineage-specific gene expansion. The first is a small clade of calcium/calmodulin-dependent-like (CAMKL) kinases almost exclusive to STs 4 and 7 (Fig 4). The second and more striking example of ST differences is a clade of STE20/7 kinases. While ST4 and ST7 both encode representatives of the STE family, they appear to completely lack members of a cluster of 58 closely related STE20/7 kinases found only in the ST1 genome. This seems to be a spectacular case of gene family expansion specific to ST1.
The exact roles of the various kinases in Blastocystis STs and, in particular, the large group of STE20/7 kinases exclusive to ST1 are currently unclear. The STE20/7 family includes members of the mitogen-activated protein kinase (MAPK) signaling pathway, which regulates responses to extracellular stimuli [34]. The ST1 cluster does not exhibit a particularly close relationship with any specific gene of the STE20/7 family. In Giardia, 2 members of the MAPK family have been implicated in initiation of encystation [35], while in T. brucei, a MAPK is involved in mediating its interferon-γ-induced proliferation in the host [36]. MAPK pathways have been characterized, at least in pathogenic fungi, as a "functional nervous system" that controls virulence and modulates the outcome of the disease [37]. General “housekeeping” responses to stimuli would presumably be similar across the STs, so the highly developed exclusive MAPK pathway in ST1 must have some unique consequences and could be worthy of future functional studies.
A stringent, 4-step approach identified a total of 89 genes confidently predicted to encode secreted proteins in Blastocystis ST1 (S1 Data). The corresponding predicted number for ST7 was 307 [21], although those results were based only on SignalP predictions, a less stringent approach than the one used here. Most of the 89 secreted proteins predicted in ST1 are also present in ST7, with the exception of a metallophosphoesterase and GH2 and GH33 glycosyl hydrolases. Several of the 89 genes were not predicted as secreted in ST7 because their signal peptides were not part of the predicted gene models. More than half of these putative secreted proteins have a predicted role in posttranslational modification and protein turnover (Fig 5). We compared the secretory signal peptides of the different STs and were unable to identify a unifying characteristic apart from a core of 10 hydrophobic amino acids (S2 Fig). The majority of the secreted protein-coding genes in the 3 STs were present in more than 1 copy, with a notable expansion of cysteine proteases (see below).
Intriguingly, ST1 has 4 genes that are predicted to code for secreted collagen-like proteins. These 4 genes are very different at the amino acid level, with pairwise percentage identities only in the mid to upper 30s, but all have signal peptides that suggest they are secreted and all have the distinctive collagen-like motif GXY (glycine, second and third residues can be anything but frequently proline and hydroxyproline) repeated 63 times. ST4 also has 4 copies of genes encoding these collagen-like proteins, with only 1 predicted to be secreted, whereas ST7 has a single gene encoding a related protein that lacks a signal peptide. Based on sequence similarity and repeat characteristics, the Blastocystis collagen-like proteins are of the bacterial type [38]. A number of bacterial pathogens have collagen-like proteins that are involved in pathogenicity, immune response elicitation, and host–parasite interactions. For example, collagen-like proteins are able to bind to the human extracellular matrix, thereby aiding adhesion and colonization [39–41]. It is possible that collagen is one of the factors mediating adhesion of Blastocystis to enterocytes. Another possibility is that collagen may be part of a mechanism used by Blastocystis to trap bacteria and other microbial eukaryotes for nutritional purposes. This process has been observed by electron microscopy, although an exact mechanism has not been described [42–44]. At present, the roles played by collagen-like proteins in Blastocystis are purely speculative, based on their similarities to bacterial proteins that are clearly involved in pathogenicity. However, it suggests fertile ground for further investigation, specifically into whether differences in both number of genes and potential for secretion are implicated in variable virulence.
Another class of proteins demonstrating clear ST differences is proteases or peptidases. Proteases are crucial for many biological processes and constitute potential virulence factors in parasitic protists [45]. Cathepsin B, a cysteine protease, has been linked to increased intestinal cell permeability [8], while other cysteine proteases have been reported to cleave human secretory IgA [46] and induce up-regulation of interleukin 8 cytokine transcription and secretion [47] in intestinal epithelial cells.
Blastocystis has a large number of proteases, with the 3 STs having mostly similar profiles, but with some notable differences. The total number of proteases encoded in ST1 is 243, with 198 in ST4 and 210 in ST7 (S2 Data) (Fig 6). All 3 STs encode aspartic, cysteine, metallo, serine, and threonine proteases. The most prevalent protease genes are those of the cysteine type, constituting between 39% (ST7) and 47% (ST4) of the degradome. Undoubtedly, many cysteine proteases play roles that are conserved in eukaryotes, such as lysosomal function, autophagy, and ubiquitination, but others have been implicated in host–parasite interactions [48–50]. All 3 STs show extensive gene expansions of the cysteine protease families C1, C13, and C19. A large number of C1 genes is fairly common. Less common is the number of C13 genes found in ST1 (16), ST4 (11), and ST7 (11). Most protist genomes contain fewer than 5 C13 genes and many just have a single version (see MEROPS database, merops.sanger.ac.uk, [51]). The only protist known to approach the number of C13 genes seen in Blastocystis is the sexually transmitted human excavate Trichomonas vaginalis (10), in which at least 1 type of C13 protein has been implicated in trichomonal cytoadherence [52]. While Blastocystis appears to have an elevated number of C19 genes compared with other protists (MEROPS database), currently there is no indication that these genes are involved in anything other than standard intracellular removal of ubiquitin molecules.
Other protease families that exhibit large differences between STs are C56, C95, C97, and M16 (S2 Data). The biological significance, if any, of these differences is currently unknown. The most striking disparity between the STs is seen in metallo-type proteases (Fig 6, S2 Data). The divergence in the number of genes belonging to this type is entirely attributable to the complete absence of subfamily M23B genes in the ST4 genome, while both ST1 and ST7 have 29 members. The M23 family is composed of metallopeptidases involved in the lysing of peptidoglycans in bacterial cell walls for either defense or feeding (MEROPS database). What functions they possess in Blastocystis and why they are absent from ST4 are open questions. Differences in the number of the S54 rhomboid serine proteases were also identified, with ST1 having 4 genes, ST4 having 3, and ST7 having only 1. These transmembrane peptidases play a role in the invasion of host cells in the alveolates Toxoplasma, Cryptosporidium, and Plasmodium, while in the amoebozoan Entamoeba histolytica, the single rhomboid protease is vital to immune evasion via the cleavage of lectins during receptor capping [53–55].
The membrane-trafficking system enables transport of proteins and lipids between intracellular locations and is an interface with the extracellular environment. Crucial to the healthy working of eukaryotic cells, it underpins the pathogenic mechanisms of diverse eukaryotic parasites through the release of virulence factors as well as the uptake of metabolites from the host. Trafficking is enabled by a suite of proteins involved in vesicle formation and fusion [72]. Comparative genomic and molecular phylogenetic analyses have established that a relatively complex complement of membrane-trafficking machinery was present in the last eukaryotic common ancestor (LECA) [73].
Our analyses showed that homologs of nearly all components of the vesicle-formation and -fusion machinery present in the LECA are also encoded in Blastocystis STs 1, 4, and 7 (S3 Fig). Indeed, there is evidence for somewhat expanded paralog numbers in several of the complexes, most notably in the Sec24 subunit of the COPII coat; the Arf GAPs; the HOPS and GARP tethering complexes; and the adaptor protein complexes 1, 2, and 4 (S3 Fig). This is also where much of the variability between the STs is observed (along with the numbers of some endosomal sorting complexes required for transport [ESCRT] paralogs).
A number of Ras superfamily proteins (or small GTPases) known to be involved in membrane and protein trafficking have orthologs in Blastocystis (S4 Data). Notable among them are 2 (ST7) or 3 (ST4, ST1) paralogs of a GTPase that serves as the membrane-anchored β subunit (SRβ) of the signal-recognition particle receptor on the ER that is involved in the cotranslational import of proteins into the ER [74]. Most eukaryotes, including other stramenopiles, employ only 1 SRβ protein, so the functional significance of the varying number of SRβ paralogs in Blastocystis is unclear.
Blastocystis STs also harbor some Ras superfamily members that appear to have emerged in this lineage through extensive divergence of existing or duplicated genes. These can be considered “novel” genes and may potentially underpin significant evolutionary innovations. For instance, ST1 and ST4 both encode a divergent Rab7-like paralog (Rab7L in S4 Data) that is apparently missing from the genome of ST7. Although the cellular function is difficult to predict for such divergent paralogs, the similarity of Rab7L to standard Rab7 proteins suggests that it may be involved in some trafficking processes associated with lysosomes or vacuoles. The Blastocystis genomes encode 5 additional divergent Rab-related proteins whose evolutionary origin cannot be deduced with confidence and hence are labeled RabX1 to RabX5 (S4 Data). These proteins are presumably involved in specialized membrane-trafficking pathways in Blastocystis, but their functions cannot be predicted from sequence analyses alone. The 2 Rab1 paralogs in Blastocystis, annotated as Rab1A and Rab1B, differ from the previous examples in that their origin is more ancient, as they apparently stem from a Rab1 duplication previously suggested to be an evolutionary novelty of the SAR supergroup [75]. The functional significance of the 2 differentiated Rab1 paralogs in organisms of the SAR clade remains unknown despite an attempt to characterize both paralogs in Toxoplasma gondii cells [76]. Most recently, Rab1A was found in association with rhoptries in schizonts of P. falciparum and suggested to be involved in regulating vesicular trafficking from the ER to the former secretory organelles of the parasite [77].
Control of the cell cycle and cell proliferation is mediated extensively by the anaphase promoting complex/cyclosome (APC/C) [78]. It functions as an E3 ubiquitin ligase that coordinates the degradation of specific substrates via the 26S proteasome at specific points in the cell cycle [79]. Typically, the complex is composed of about a dozen subunits with a combined mass of about 1.5 MDa. It can be divided into 3 functional parts or subcomplexes [80]: (i) a structural complex serving as a scaffold, (ii) a catalytic arm, and (iii) a tetratricopeptide repeat (TPR) arm designed to position the substrate for successful transfer of ubiquitin. Surprisingly, Blastocystis seems to lack genes for all the subunits that make up the scaffold (i.e., APC1, APC4, and APC5), and to encode a reduced TPR arm composed of only 2 subunits versus typically 4 in other stramenopiles and up to 7 in other eukaryotes [78]. Particularly notable is the loss of the APC3 subunit, which is found in virtually all eukaryotic organisms. However, Blastocystis appears to be unusual in that it has several paralogs encoding the TPR subunit APC8, which potentially may compensate for the absence of the other subunits in the TPR arm (S5 Data).
The APC/C interacts with a number of adaptors and coactivators that modulate its activity and specificity. The most important of these adaptors are cell-division cycle protein 20 (Cdc20) and cadherin-1 (Cdh1) because, at various stages of the cell cycle, they are essential for activation of the complex and selecting which substrates to interact with [81]. In particular, APC/C-Cdh1 plays a role in DNA synthesis during the G1/S phase because it allows the 26S proteasome to degrade several DNA replication inhibitors [80]. Surprisingly, while the vast majority of eukaryotes possess both Cdc20 and Cdh1, Blastocystis only has a homolog of Cdc20. How the end of anaphase is regulated in Blastocystis remains an open question.
In addition, genes encoding 2 of the main APC/C targets that are involved in the integrity and regulation of the cohesion complex, Scc2/Scc4 and Eco1, were absent. This protein complex keeps sister chromatids together and regulates their separation during cell division. Scc2/Scc4 aids in loading the complex onto the chromosome, while Eco1 is responsible for the establishment of cohesion between cohesin and chromatin [80]. The absence of these components in Blastocystis and other stramenopiles suggests that an alternative route may exist to achieve a properly functioning cohesion complex and the separation of sister chromatids.
Potential homologs of proteins involved in DNA damage response and repair, chromatin structure relevant to repair, and meiosis, a process not previously attributed to the life cycle of Blastocystis, were identified (S4 Fig, S6 Data). Homologs of genes encoding 9 out of 11 meiosis-specific proteins required for the progression of meiosis in other organisms [82,83] are found in Blastocystis ST1 and ST4; these include Hop1, Spo11-2, Top6BL, Dmc1, Hop2, Mnd1, Msh4, Msh5, and Mer3. Msh5 is absent from ST7, and Rec8 and Spo11-1 were not identified in any of the STs. Similar to T. vaginalis [84], the Blastocystis genomes apparently do not encode components of the nonhomologous end-joining machinery, suggesting that homologous recombination is the principal mechanism for the repair of double-stranded DNA breaks.
A total of 203 carbohydrate active enzymes (CAZymes) were identified in the Blastocystis ST1 genome (S7 Data). The most interesting cases are discussed below and additional details are provided in S1 Text.
Unlike the intestinal parasites Giardia and Entamoeba, which rely on the host as a source of purines and pyrimidines [95,96], Blastocystis ST1 possesses complete pathways for the de novo synthesis of these compounds (S8 Data). Its capacity for de novo amino acid biosynthesis is limited to alanine, aspartate, and glutamate, while serine and glutamine can be produced via conversion from other amino acids. Blastocystis ST1 has a mostly complete folate biosynthesis pathway, lacking only the alkaline phosphatase responsible for the dephosphorylation of 7,8-dihydroneopterin 3′-triphosphate. However, this enzyme is also lacking in many probiotic bacteria such as bifidobacterial, which appear to use another NUDIX enzyme [97]. While genes encoding homologs of this protein could not be identified, it seems likely that Blastocystis ST1 uses an uncharacterized alkaline phosphatase to complete the pathway.
Unlike most parasites, Blastocystis ST1 appears to have a nearly complete de novo thiamine (vitamin B1) biosynthesis pathway. In this pathway, 4-amino-2-methyl-5-hydroxymethylpyrimidine (HMP) is pyrophosphorylated by HMP kinase sequentially to form HMP-PP, which is in turn condensed with thiazole to form thiamine phosphate by thiamine-phosphate pyrophosphorylase. Only the thiazole salvage enzyme hydroxyethylthiazole kinase (ThiM) could be identified, suggesting thiamine biosynthesis in Blastocystis ST1 is either dependent on exogenous thiazole or synthesizes thiazole by an unknown mechanism. The latter is not unprecedented, as P. falciparum has been shown to synthesize vitamin B1 de novo despite having the same repertoire of thiamine synthesis-related genes as Blastocystis [98].
Cytochrome P450 (CYP) is a large and versatile heme-containing protein superfamily containing at least 317 families and numerous subfamilies (https://cyped.biocatnet.de/sequence-browser). These proteins are found in all domains of life and in all major eukaryotic lineages. However, no homologs were found of any of the CYP families in the 3 Blastocystis genomes. Genomes available for other stramenopiles do encode various CYPs. The diatom P. tricornutum has 3 CYP genes, consisting of 2 CYP97 and 1 CYP51 enzymes, whereas T. pseudonana encodes 9 family members [99]. P. falciparum was the first documented eukaryotic species without a CYP-encoding gene [100]. However, it should be noted that Kinetoplastida such as T. brucei have a limited set of CYPs devoted to sterol synthesis (e.g., CYP51A). Another apicomplexan parasite, T. gondii, has a single CYP gene, encoding a steroid 11-β hydroxylase. Since sterols are essential for eukaryotic membranes, the lack of CYP51 in Blastocystis suggests that it obtains sterols from its host, as does Giardia [101].
The intestinal eukaryote Blastocystis continues to be of significant clinical interest because of it widespread prevalence. Despite years of study, however, its pathogenicity and role in the gut remain controversial. This ambiguity is compounded by the presence of STs that, while morphologically indistinguishable, are nevertheless genetically heterogeneous. The extent of the differences and the degree to which they matter clinically are still unclear, but the increase in genomic data reported here opens up the possibility of answering some of the outstanding questions through comparative analyses.
Structurally, there is considerable variation among the 3 Blastocystis ST genomes now available in terms of genome size and GC-content. ST1 also displays sizeable differences compared to ST7 and ST4 with regard to the number of genes identified and general gene characteristics, such as intron numbers and gene size. However, the use of a novel mechanism by which stop codons are generated was examined thoroughly in ST1 and led us to conclude that some of the differences in gene numbers and characteristics could be due to previous annotation efforts in ST4 and ST7 not taking the stop codon issue into account.
While the need for reannotation of ST4 and ST7 affects the quality of the protein data sets from these STs to some extent, nevertheless, divergence at the amino acid level between homologous proteins has been demonstrated that is almost an order of magnitude greater than that seen between species within other genera of parasitic protistans. Analyses of protein classes of interest revealed numerous differences. Of interest will be the variation in the number and type of protease genes identified among the STs, since this class of protein has been linked to pathogenicity. Another intriguing area ripe for exploration and experimentation is the kinome. ST-specific expansions of protein kinase genes were identified. In particular, ST1 encodes a large number of novel proteins from the STE20/7 family, which is typically involved in responding to extracellular stimuli.
General aspects of the Blastocystis genome were examined that help in understanding its biology and how it relates to other stramenopiles and other parasitic protistans as well. Some of the highlights (see S1 Text for additional analyses) include an expanded proteome of the MRO, the absence of peroxisome-related genes, the presence of α-glucan metabolism, and intriguing expansions and additions to the repertoire of membrane-trafficking machinery proteins. The extensive structural and gene complement differences between the genomes of the 3 STs suggests the need for a revised taxonomy of Blastocystis. Since Blastocystis has been linked to disease in humans, the primary focus of any genomics analyses will likely be medically oriented. The availability of Blastocystis genomes will greatly assist future functional studies. In particular, the development and successful implementation of a transformation system, to allow finely tuned control over constructs as well as localization experiments, will depend on mining the genome for appropriate primers, promoters, and genes of interest. Resequencing to evaluate the success of transformations and any off-target responses will be facilitated by preexisting genomes. Pathogenomic studies of Blastocystis will benefit greatly from access to a genome of ST1. Comparisons with existing and future genomes from other STs and other isolates of ST1 will help in finding genes that are correlated with virulence as well as the development and testing of drug targets. Equally as important to elucidating the roles of Blastocystis STs in the human gut is understanding the biology of these anaerobic protists and their repertoires of metabolic pathways, which allow them to thrive in the gut environment.
Blastocystis sp. NandII (ST1) was obtained from the American Type Culture Collection (ATCC 50177) and maintained in Locke's medium and horse serum egg slants at 35.6°C in an anaerobic chamber [102]. Blastocystis ST1 cells were harvested by centrifugation at 850xg for 15 min at 4°C. Total DNA was extracted with the standard CTAB protocol [103]. Subsequently, Hoechst dye cesium chloride (CsCl) centrifugation was used to obtain purified nuclear DNA. Extracted DNA was diluted with TE buffer to a final volume of 10 mL and resuspended in 11.5 g of CsCl. Ten mg of Hoechst dye was added and the mix was homogenized by shaking for 3 hours. The solution was then transferred into Quick-Seal centrifuge tubes and centrifuged in a fixed-angle Ti-75 rotor at 40,000xg for 44 h. The resulting DNA bands were visualized under long-wave UV light. The AT-rich organellar band was removed and the remaining nuclear DNA band retrieved with a 30-gauge needle. Hoechst dye was removed with 3 successive washes of water-saturated butanol. DNA was precipitated with 100% ethanol and suspended in water.
Total RNA was isolated using a modified Trizol protocol [104], in which the supernatant, after the first round, underwent another Trizol extraction. The resulting RNA was used to create a cDNA paired-end library (Vertis Biotechnologies AG [Germany]) that was then sequenced on a 4SLX Titanium Platform (Genome Quebec). The reads were filtered for primer, vector, and adaptor sequences. Following filtering, 69,155,000 read pairs remained. A second cDNA library was constructed (Beijing Genomics Institute) and RNA-Seq data was generated using the Illumina HiSeq 2000 system. As above, the reads were filtered for primer, vector, and adaptor sequences. The reads were also filtered by quality scores using FASTX-Toolkit (version 0.0.13) http://hannonlab.cshl.edu/fastx_toolkit/index.html. To be retained, a minimum of 70% of the bases had to have quality scores of 20 or better [99]. Ultimately, 370,469,956 reads remained.
The miraEST [105] and Trinity [106] assembly software programs were used to assemble the paired-end reads into contigs. 20,426 contigs were assembled. Only contigs above 200 bps were considered for further analysis. The longest contig was 18,604 bps and N50 was 793 bps.
The Blastocystis ST1 genome was sequenced using 454 pyrosequencing and Illumina technologies. Illumina sequencing entailed generation of a mate paired-end library with an insert size of 3 kb. Following the filtering steps described above, 1,865,740 454 and 55,617,444 Illumina reads remained. The 454 and Illumina reads were assembled using the Ray genomic assembler software [107]. Only contigs above 200 bps were retained for further analysis.
Using different combinations of 454 and Illumina reads, 3 different assemblies were created. 2F was generated from just Illumina reads and contained 13,338 contigs. 2E was generated from all reads and contained 29,673 contigs. 2D was generated just from 454 reads. The software Minimus2 from the AMOS package [108] was used to merge the 3 into a single assembly with 14,304 contigs.
The genome assembly for Blastocystis ST1 was interrogated for the presence of contigs composed of bacterial reads. For contigs of less than 500 bps, BLASTx and BLASTn [49] searches were performed. All contigs with a percent identity of 95% and higher against entries in the nr and nt NCBI databases were considered contaminants. In total, 305 contigs were removed. Large contigs were split into 5,000-bp pieces while those less than 5,000 bps in length were kept intact. All the pieces were compared, using BLASTx, against a specialized database consisting of protein sets from all publicly available bacterial genomes and protein sets from 28 eukaryotic genomes drawn mainly from Chromalveolata and Archaeplastida. Pieces that only had matches for bacterial proteins were putatively designated as derived from bacterial contamination. Because of the possibility of LGT from bacteria, all of the 5,000-bp pieces from a large contig were required to be “bacterial” to warrant removal of the contig. In total, 459 contigs were set aside as being bacterial in nature and most probably resulting from bacterial contamination. 150 were also removed as clearly derived from contaminating Saccharomyces cerevisiae and E. histolytica DNA.
After the removal of bacterial/yeast/Entamoeba contamination, attempts were made to assemble the ST1 contigs into larger scaffolds. SSPACE [109], which is specially designed to work with preassembled contigs, was used on those contigs greater than 500 bps. The 2,494 contigs were assembled into 693 scaffolds with the N50 increasing from 8,801 bps to 50,205. The Blastocystis ST1 genome sequences are available from the NCBI with the accession LXWW00000000.
In the absence of an experimentally determined genome size for ST1, the extent to which the final assembly represents the full genome was determined roughly using 2 methods. The first method entails taking a set of 679 conserved protein sequences derived from RNA-Seq data and matching them against the genome assembly. Of the conserved transcripts, 5.4% were not found in the genome assembly, suggesting that the current assembly encompasses 94.6% of the genome or at least the gene space of the genome [110].
The second strategy for estimating genome coverage was based on a method first used in the Conus bullatus genome paper [111] to estimate genome size. A database of 31.8 million 90-bp Illumina reads was used in BLASTn searches of the assembled contigs. Read coverage for contigs was estimated as (number_of_hits*read_length)/length_of_contig. The mode was determined from a histogram of the read coverages of the contigs and used to estimate genome size: (number_of_reads*read_size)/read_coverage_mode ((31,826,528*100)/153 = 18.7 Mb).
The estimated size was then compared with the assembled genome size (17.4 Mb/18.7 Mb = 93%).
Additionally, the protein data sets for the 3 genomes were used for BUSCO [112] analyses to assess their completeness based on a set of 429 highly conserved single copy eukaryotic genes. An initial BUSCO analysis was done using the ST protein data set. Any conserved genes that were considered missing were then compared against the appropriate ST genomic scaffolds to determine if the gene had been missed during annotation. Given the evolutionary distances between Blastocystis and the organisms used to generate the BUSCO test set, the likelihood of achieving 100% coverage was low. More useful was a comparison of the results for the 3 STs.
The average level of heterozygosity for ST1 was determined by aligning trimmed genomic reads to the genomic assembly with Bowtie 2.3.1 [113]. The program mpileup from SAMtools [114] was used to determine the read depth at each position of the assembly as well as the breakdown of the 4 possible bases. An in-house perl script was used to parse the mpileup data. Positions with less than 10X coverage were ignored. A site was deemed to be heterozygous if at least 2 different bases were present and there were at least 2 (or 3) reads with the different bases. The number of heterozygous sites was then divided by the total number of sites (with ≥10X coverage).
A set of 679 protein sequences was generated from the ST1 RNA-Seq clustered transcripts and used to train AUGUSTUS [115] on a repeat masked assembly [116]. The training set contained protein sequences with the following criteria: at least 1 intron, less than 80% identical at the amino acid level to any other sequence in the set, and a start site based on highly similar eukaryotic protein sequences. The script optimize_augustus.pl was also run to create species-specific meta parameters. A set of 32,269 Trinity [106] assembled RNA-Seq transcripts were used to generate a hints file for donor/acceptor intron sites. Subsequent gene modeling generated 5,637 gene models, which were then corrected with PASA [117] based on RNA-Seq transcripts. Because of the unusual nature of termination codons in Blastocystis [25], it was necessary to set the parameter stopCodonExcludedFromCDS to true. Of the final gene model set, 58% was examined manually at some point to check various aspects such as intron boundaries, stop codons, and chimeric models. The degree to which intron boundaries were confirmed by RNA-Seq data was determined by mapping RNA-Seq data to the genome assembly using Bowtie2 [113]. Using the intron boundary positions and the alignment positions of individual reads from the SAM file, an in-house perl script calculated that 70.3% of the intron boundaries were confirmed by 5 or more RNA-Seq reads (60.1% with 10 or more RNA-Seq reads).
To assist in the manual annotation of the gene models, various information was generated for each model. The models were compared against the conserved domain database [118] and domain hits as well as Pfam [119] hits were extracted from the results. The models were also compared against the NCBI protein database as well as the protein data set for ST7 using BLASTp. SignalP results were generated using a locally installed version of SignalP. The genome browser Genomeview [120] was chosen for annotation because of its relative ease in setting up and populating with information via gff files and the ability to alter the information using custom scripts. The genome browser also included tracks for genome reads mapped to the scaffolds (Bowtie2 [113] and SAMtools [114]), genome read map coverage, RNA-Seq reads mapped to the scaffolds, and RNA-Seq map coverage.
Because of the possibility of missed genes, the intergenic spacers were identified and the sequences compared against the NCBI protein database (nr) as well as the ST7 protein data set using BLASTx. Potentially missed genes were mapped to the scaffolds and displayed in the genome browser for confirmation and correction through manual annotation. The ST7 protein data set was also compared to the gene models for ST1 using BLASTp and the results mapped to the scaffolds as a genome browser track.
Automated annotation for eukaryote genomes, particularly those from nonmodel organisms and poorly sampled lineages, are prone to missed genes, i.e., the gene-finding algorithms fail to find all of the protein-coding regions. While having a transcriptome is invaluable for reducing the level of missed genes, invariably, some are not found. Therefore, as a general rule, the various analyses detailed below that involve presence/absence of genes, particularly when comparing STs, used techniques such as tBLASTn to interrogate the full genome sequence before declaring that a gene was missing from an ST. Additional steps to confirm the presence/absence of a gene are indicated in the individual sections.
The genomic scaffolds for STs 1, 4, and 7 were ordered according to their lengths. They were further stripped of their Ns and header lines. The program GC-Profile [121] as implemented at http://tubic.tju.edu.cn/GC-Profile/ was used to create GC profile and GC content graphs for the 3 genomes. As recommended by the authors, a halting parameter of 300 and a minimum segmentation length of 3,000 bps were used.
GC-Profile, a windowless method, was chosen from many different algorithms to avoid some of the disadvantages present in other methods to calculate GC content such as GC patterns dependent on window size, and lack of resolution [121,122].
The SAMtools option depth [114] was used to determine the read-depth coverage at each position of the genome scaffolds. These values were used to calculate median read-depth coverage and normalized median read-depth coverage using in-house scripts. SNPs were detected using the SAMtools options mpileup and bcftools.
The level of amino acid divergence between Blastocystis protein sequence data sets was analyzed using reciprocal best BLAST hit protein sequence pairs. Because of concerns about retained introns in some of the STs as well as issues with incorrect stop codons, the calculation of amino acid identity scores for gene pairs was restricted to high-scoring segment pairs without the inclusion of gaps. The values were generated using an in-house perl script.
For Blastocystis ST1, the analysis of different intron categories was based on the final annotation of the genome, incorporating data from RNA-Seq and extensive manual curation. To analyze introns in the Blastocystis ST7 genome, the available annotation of the genome could not be relied on because it proved to be inaccurate in many cases. Therefore, analysis was restricted to introns that have direct support from EST sequences. Available ESTs were aligned to the genome sequence using STAR [123] and introns indicated by the alignments were identified and manually corrected if needed. Homologs of protein subunits specific for the minor spliceosome were identified by BLAST (BLASTp, tBLASTn) using sequences of previously defined subunits from Homo sapiens and plants. The identity of the candidate orthologs was confirmed by reciprocal BLASTp searches against the nr database at NCBI. Incorrect or missing gene models in Blastocystis ST1 were corrected or added to the annotation of the genome. Most of the identified homologs in Blastocystis ST7 are either represented by incorrect gene models or completely missing from the annotated genes but could be found in the assembly with tBLASTn. Models for snRNA in both STs were identified using the program cmscan from the Infernal package [124].
A multipronged approach was used to predict mitochondrial proteins. (1) Each gene model and translated transcript was queried for N-terminal mitochondrial targeting sequences using Mitoprot and TargetP [125,126]. Sequences that returned a score greater than 0.5 were further investigated as MRO candidates. (2) The top 10 BLAST hits retrieved from the nr database at NCBI were surveyed for any mitochondrial annotation and manually investigated. (3) All gene models and translated transcripts were queried against a local curated MitoMiner data set [127] and significant hits (e-value < 1e-10) were further investigated. (4) Finally, a local MRO protein data set from various anaerobic protists including Pygsuia biforma, T. vaginalis, and Mastigamoeba balamuthi was queried against the gene models and translated transcript data sets. Mitochondrial targeting sequence scores and protein annotations are provided in S3 Data. Those proteins previously unrecognized as mitochondrial in Blastocystis species are shown in color in Fig 7.
Homologs of proteins suspected to be involved in cytokinesis were identified by BLASTp against the Blastocystis ST1 predicted proteome (e-value cutoff: 10e-10), using as queries previously identified mammalian orthologs [128]. Absence of homologs was verified by tBLASTn against the genomic scaffolds and transcriptome. Blastocystis hits were then used as queries to run reversed BLASTp searches against nr and results were manually inspected to confirm assignation to a given protein family. Identification of homologs of the APC/C complex and targets was carried out in a similar fashion, using as queries previously identified homologs in Blastocystis ST1 or, if these did not exist, various other eukaryotic homologs [78]. For calcium-binding proteins, H. sapiens homologs of protein families identified in [129] were used as queries to carry out BLASTp searches against the Blastocystis ST1 predicted proteome but also against all eukaryotic protein data present in GenBank as of February 2016 ("Eukaryota[Organism]" option of the BLAST+ package).
A query database including 135 DNA repair genes, meiosis-specific genes, and meiosis-related genes from a wide range of eukaryotes was established using literature and key word searches of the NCBI database. Genes with functional experimental evidence were used as queries to retrieve similar proteins using PSI-blast. After retrieval, protein sequence alignments were created and used to build HMM profiles with HMMER3.1b2 [130] to enable gene searches in Blastocystis STs 1, 4, and 7. Retrieval of genes was restricted to e-values below 1e-04. Phylogenetic trees were constructed with FastTree [131] using closely related paralogs and/or reconstructing entire gene families for each gene of interest. Protein domains were mapped to the trees to facilitate gene recognition. In a few cases, secondary structure predictions were made to confirm the gene identification.
221 kinases were identified from the protein-coding gene set of Blastocystis ST1 using HMMER [130]. The STYKc profile (SM00221) from the SMART database [132] was used, as was the Pfam Pkinase profile (PF00069) [119]. Blastocystis kinases were classified by their positions in phylogenetic trees inferred from the alignments of the HMMER output (both neighbor joining and maximum likelihood trees were taken into consideration). Blastocystis gene models with 2 kinase domains were classified by the phylogenetic position of the domain with a better HMMER score. The kinase classification follows the webite www.kinase.com [33].
The 221 kinases identified from Blastocystis ST1 were used to query the protein-coding gene sets from STs 4 and 7. The annotations for STs 4 and 7 were also searched for the key word “kinase” and putative protein kinases were checked against the kinase database (www.kinase.com) using the BLAST search tool.
The resulting sequences from STs 1, 4, and 7 were aligned using MUSCLE [133], followed by manual inspection of the alignment. A maximum likelihood tree was inferred by the RAxML program [134] with the PROTGAMMALG model and visualized using Figtree (http://tree.bio.ed.ac.uk/software/figtree/).
The Blastocystis STs 1 and 7 protein models were searched by BLASTp with CYP protein sequences representing the CYP2 (DmelCYP307A1, DmelCYP303A1, and DmelCYP18A1), CYP3 (HsCYP3A4, HzCYP321A1, DmelCYP6A2, DmelCYP6A8, and DmelCYP6G1), and CYP4 (DmelCYP4G15 and DmelCYP4G1) animal clades [135]. CYPs from the mitochondrial clan were searched using DmelCYP314A1, DmelCYP302A1, DmelCYP12A1, DmelCYP301A1, DmelCYP49A1, and DmelCYP315A1. Additional animal and fungi CYP clans [136], clan 51, clan 7, clan 26, clan 20, clan 46, clan 19, and clan 74, were also used to search for CYPs in the Blastocystis protein models as well as the CYPome from sequenced genomes of Guillardia theta, T. pseudonana, and Phytophthora sojae. CYP51 clan involved in sterol biosynthesis was further studied using evolutionarily close CYP51s including Galdieria sulphuraria CYP51, Porphyridium purpureum CYP51R1, Batrachochytrium dendrobatidis CYP51F1, Cyanidioschyzon merolae CYP51G1, Dictyostelium discoideum CYP516A1, L. major CYP51E1, Monosiga brevicollis CYP51A1, T. pseudonana CYP51C1, and Chlamydomonas reinhardtii CYP51G.
CYP searches for other organisms were done as above but using NCBI, CYPED (https://cyped.biocatnet.de/), and the eukaryotic pathogen database EuPathDB (http://eupathdb.org/eupathdb/) [137–139].
The components of the membrane-trafficking system in Blastocystis STs 1, 4, and 7 were identified using both comparative genomic and phylogenetic methods. To identify potential homologs, functionally characterized membrane-trafficking components from H. sapiens and S. cerevisiae were used as queries to search the Blastocystis predicted proteins using BLASTp with default search parameters. Hits with an e-value ≤5e-02 were considered candidate homologs and were used as BLAST queries to reciprocally search predicted proteins of the H. sapiens and S. cerevisiae genomes. The Blastocystis protein was considered homologous to the initial H. sapiens query if it retrieved the query or a clear ortholog as (i) the top hit and (ii) with an e-value ≤5e-02.
If a component could not be positively identified using BLAST searching, a hidden Markov model (HMM) was created using HMMER version 3 [140], including clear homologs from combinations of the following taxa: H. sapiens (NCBI, http://www.ncbi.nlm.nih.gov/), S. cerevisiae (NCBI), D. discoideum (dictyBase, http://dictybase.org/), Arabidopsis thaliana (NCBI), C. reinhardtii (Phytozome, http://www.phytozome.net/), P. sojae (JGI, http://www.jgi.doe.gov/), T. pseudonana (JGI), Naegleria gruberi (JGI), Bigelowiella natans (JGI), and Emiliania huxleyi (JGI). The HMMs were used to search the Blastocystis STs 1, 4, and 7 genomes (hmmsearch). Hits with an e-value ≤5e-02 were used as BLASTp queries to search the H. sapiens genome, and homologs were identified using the same criteria as were used in the initial BLASTp searches.
If BLASTp or HMMer failed to identify a membrane-trafficking component, tBLASTn searches were performed, using the H. sapiens and S. cerevisiae queries to search the Blastocystis scaffolds. The region of the scaffold with a hit that had an e-value ≤5e-02 was excised and used as a BLASTx query to search the H. sapiens genome, and homologs were identified using the same criteria as were used in the BLASTp and HMMer searches.
The more permissive e-value cutoff of 5e-02 was intended to reduce the number of false negatives from divergent Blastocystis genes when searching with the experimentally characterized opisthokont queries. Nonetheless, we found in post hoc assessment that over 93% of the orthologs identified had e-values of lower than 1e-05 in either the forward or reverse BLAST searches.
In searching for the components of a protein family in Blastocystis, Bayesian and maximum likelihood phylogenetics methods were used to more rigorously determine orthology. For each phylogeny, a combination of homologs from H. sapiens, S. cerevisiae, T. pseudonana and P. sojae, P. infestans, A. thaliana, D. discoideum, and N. gruberi were collected and aligned with the Blastocystis homologs using MUSCLE version 3.8.31 [133] and manually adjusted. Positions of ambiguous homology were removed (alignments available upon request) and ProtTest 3.2 [141] was used to determine the best-fit model of sequence evolution, which, in all cases, was the LG model [142], incorporating rate among site and invariant site corrections where relevant.
Phylobayes 3.3 [143] was used to produce the optimal topology and posterior probability values. Analyses were run until the average standard deviation of the split frequencies fell below 0.1 and the effective sample size was at least 100. Once convergence occurred, the first 20% of sampled trees were removed. Additionally, RAxML and, in some cases, PhyML version 3 [144] were used to obtain maximum likelihood bootstrap values (100 pseudoreplicates). Resultant phylogenetic trees were viewed using FigTree v1.4.0.
To identify putative PEX genes in Blastocystis STs 1, 4, and 7, protein sequences from H. sapiens, S. cerevisiae, and/or Neurospora crassa were used as queries for BLASTp and tBLASTn searches against locally hosted Blastocystis protein sequences and scaffolds, respectively. Candidate homologs with e-values less than or equal to 0.05 were subjected to reciprocal BLASTp searches against the locally hosted query genome as well as the locally hosted nr database. The retrieval of the original query as the top reciprocal BLASTp hit with an e-value less than or equal to 0.05 was the criteria for identification of a putative homolog.
To identify putative PEX genes in A. thaliana and the stramenopile genomes of P. ramorum, T. pseudonana, and P. tricornutum, H. sapiens, S. cerevisiae, and/or N. crassa protein sequences were used as queries for pHMMer [145] searches against locally hosted query genomes. Candidate homologs with e-values less than or equal to 0.05 were subjected to reciprocal pHMMer searches against the locally hosted genome and NR database. Thus, the retrieval of the original query as the top reciprocal pHMMer hit with an e-value less than or equal to 0.05 was the criteria for identification of a putative homolog. Results were compared to previously listed PEX genes in A. thaliana [146] and P. tricornutum [68]. Newly identified A. thaliana, P. ramorum, T. pseudonana, and P. tricornutum PEX genes were subsequently used as queries for BLASTp and tBLASTn searches against locally hosted Blastocystis protein sequences and scaffolds, respectively. Candidate homologs were subjected to reciprocal BLASTp searches according to the criteria described above.
The CAZy annotation pipeline was used to analyze 5,966 predicted proteins from the Blastocystis ST1 genome using a 2-step procedure of identification and annotation [147]. Sequences were subjected to BLASTp analysis against the CAZy database, composed of full-length proteins. Hits with an e-value <0.1 were then subjected to a modular annotation procedure using BLASTp against libraries of catalytic and carbohydrate binding modules and profile HMMs [148]. The results were complemented with signal peptide, transmembrane, and glycosylphosphatidylinisotol (GPI) anchor predictions [149,150]. Fragmentary gene models and all models suspected of containing errors were identified and flagged. A final functional annotation step involved performing BLASTp comparisons against a library of protein modules derived from biochemically characterized enzymes [147]. Three other stramenopile species, the diatom T. pseudonana, and the oomycetes, Albugo laibachii Nc14 and P. infestans T30-4, are present in the CAZy database and were used for comparison. The predicted CAZymes encoded by the genome of the brown alga Ectocarpus siliculosus (http://bioinformatics.psb.ugent.be/orcae/overview/Ectsi) were also annotated using the same procedures for comparison with other stramenopiles.
To predict various pathways related to intermediary metabolism (e.g., nucleotide, amino acid, cofactor metabolism, etc.), the predicted gene models and RNA-Seq transcripts were annotated using the KEGG automated annotation server (KAAS) using the single-directional best hit approach (see http://www.genome.jp/tools/kaas/). Each pathway was individually searched for completeness and results are summarized in S8 Data. In some cases, in which KAAS did not predict an ortholog capable of a reaction, query sequences were retrieved from KEGG and manually searched against the scaffold and transcriptome data.
All predicted gene models were run through SignalP with a threshold of 0.70. The selected models were then run through TargetP in order to identify a secretory pathway signal peptide with a cutoff of 0.70. Subsequently, the remaining models were run through WoLF PSORT [151] and only those predicted as extracellular were retained. As a final step, TMHMM [152] was used as a final checkpoint to find and remove any model with predicted transmembrane domains. Therefore, the secretome was identified as having a secretory pathway signal peptide, extracellular localization, and absence of transmembrane regions.
All gene models were run through the MEROPS database [51]. Only the models with predicted active sites were considered. Annotation and naming of the models follows the MEROPS database terminology (S2 Data).
A reciprocal best BLAST analysis was performed for the protein data sets from Blastocystis ST1 and C. reinhardtii using the program orthoparahomlist.pl with default settings (Stanke M. Orthoparahomlist.pl script. 2011. https://github.com/goshng/RNASeqAnalysis/blob/master/pl/orthoparahomlist.pl). The resulting list of orthologs was compared against the 213 C. reinhardtii conserved ciliary proteins [153] to determine which of the orthologs found in Blastocystis ST1 had matches.
Members of the Ras superfamily were searched for in the 3 Blastocystis genomes using well-characterized representative sequences from other eukaryotes (H. sapiens, P. sojae, and others) as queries. Both the databases of predicted protein sequences (using BLASTp) and the genome sequence assemblies (using tBLASTn) were searched to ensure that no gene was missed because of the lack of a corresponding protein prediction. Indeed, a number of genes that were missing in the protein sequence databases were identified in all 3 STs. The missing or inaccurate gene models in ST1 were manually predicted or corrected using various lines of evidence (transcriptomic data and comparisons to homologs) and incorporated into the main genome annotation. Models missing for the ST4 and ST7 genomes were constructed only when the phylogenetic relationship to ST1 genes was unclear from the tBLASTn searches. Orthologous and paralogous relationships among Blastocystis Ras superfamily genes were established primarily on the basis of reciprocal BLAST searches, but phylogenetic analyses were required in a few cases. Orthology to conserved eukaryotic groups of GTPases was, in most cases, obvious from BLAST comparisons, except for several highly divergent paralogs related or similar to Rab GTPases, which most likely represent lineage-specific, rapidly evolving genes derived from duplications of some common Rab genes. A more detailed analysis was carried out for the Blastocystis family of Miro proteins. These sequences and Miro homologs from 2 newly analyzed stramenopile lineages (Labyrinthulea, represented by Aurantichytrium limacisporum, and Placididea, represented by Cafeteria sp. Caron lab isolate) were added to a selection of Miro sequences previously analyzed by Vlahou et al. (2011) [63], excluding the very divergent sequences of N. gruberi, ciliates, and trypanosomatids. The sequences were aligned using MAFFT (version 7, http://mafft.cbrc.jp/alignment/server/, [154]) and the alignment was trimmed using the Gblocks server (http://molevol.cmima.csic.es/castresana/Gblocks_server.html, [155]) with the least stringent parameters, keeping 245 aligned amino acid positions. The tree was calculated using RAxML-HPC (8.2.8) run at the Cyberinfrastructure for Phylogenetic Research (CIPRES) Portal (http://www.phylo.org/sub_sections/portal/), employing the LG+Г+F substitution model and rapid bootstrapping followed by a thorough search for the optimal tree.
|
10.1371/journal.pbio.0060072 | Riding the Wave: Reconciling the Roles of Disease and Climate Change in Amphibian Declines | We review the evidence for the role of climate change in triggering disease outbreaks of chytridiomycosis, an emerging infectious disease of amphibians. Both climatic anomalies and disease-related extirpations are recent phenomena, and effects of both are especially noticeable at high elevations in tropical areas, making it difficult to determine whether they are operating separately or synergistically. We compiled reports of amphibian declines from Lower Central America and Andean South America to create maps and statistical models to test our hypothesis of spatiotemporal spread of the pathogen Batrachochytrium dendrobatidis (Bd), and to update the elevational patterns of decline in frogs belonging to the genus Atelopus. We evaluated claims of climate change influencing the spread of Bd by including error into estimates of the relationship between air temperature and last year observed. Available data support the hypothesis of multiple introductions of this invasive pathogen into South America and subsequent spread along the primary Andean cordilleras. Additional analyses found no evidence to support the hypothesis that climate change has been driving outbreaks of amphibian chytridiomycosis, as has been posited in the climate-linked epidemic hypothesis. Future studies should increase retrospective surveys of museum specimens from throughout the Andes and should study the landscape genetics of Bd to map fine-scale patterns of geographic spread to identify transmission routes and processes.
| Once introduced, diseases may spread quickly through new areas, infecting naive host populations, such as has been documented in Ebola virus in African primates or rabies in North American mammals. What drives the spread of the pathogenic fungus Batrachochytrium dendrobatidis (Bd), which causes chytridiomycosis, is of particular concern because it has contributed to the global decline of amphibians. We modeled the spatiotemporal pattern of the loss of upland amphibian populations in Central and South America as a proxy for the arrival of Bd and found that amphibian declines in Central and South America are best explained by Bd spreading through upland populations; we identified four separate introductions of Bd into South America. Climate change seriously threatens biodiversity and influences endemic host–pathogen systems, but we found no evidence that climate change has been driving outbreaks of chytridiomycosis, as has been posited in the climate-linked epidemic hypothesis. Our findings further strengthen the spreading-pathogen hypothesis proposed for Central America, and identify new evidence for similar patterns of decline in South American amphibians. Our results will inform management and research efforts related to Bd and other invasive species, as effective conservation actions depend on correctly identifying essential threats to biodiversity, and possible synergistic interactions.
| Amphibian populations are declining across the globe at an alarming rate, with over 43% of species in a state of decline [1,2]. In addition to long-recognized threats such as habitat loss, overexploitation, and exotic species introductions, amphibians in all biogeographic regions face several new threats, including climate change, emerging infectious diseases, and chemical contaminants [3]. Since amphibian declines were first noticed in the late 1980s [4], many studies have attributed certain declines to a particular cause, often through a process of elimination and using limited data [5], rather than through a rigorous testing of hypotheses [6,7]. This type of approach is particularly a problem for remote tropical areas for which even baseline biotic inventories are scarce or nonexistent, and for countries with limited funding or scientific infrastructure [8]. Yet, these areas harbor the greatest number of amphibian species and are suffering the greatest numbers of declines and extinctions [2]. From a scientific standpoint, it is important to evaluate and test likely hypotheses of declines, inform conservationists with data-based predictive scenarios, and design research to provide additional data from new and understudied sites and taxa. From a practical standpoint, we need solid data and strong hypotheses to better plan conservation activities [9,10]. From an ethical standpoint, we need to understand, as quickly as possible, the global patterns and causes of amphibian declines to prevent further losses of biodiversity. In this spirit, we examine alternative hypotheses regarding the relationship between climate change and amphibian declines, including extinctions.
One cause of amphibian declines is chytridiomycosis, hypothesized to be an invasive disease [11] recently introduced into the Americas and Australia, that is caused by the fungal pathogen Batrachochytrium dendrobatidis (“Bd”). Bd is widespread throughout South America [12–16], and its role in population declines has been linked to interactions with climate change, although no studies have explicitly considered evidence of spatiotemporal spread of disease as an alternative to the recently proposed climate-linked epidemic hypothesis (CLEH) [17,18]. Regardless, both field studies on amphibians [6,19] and on fungal population genetics [20–21] strongly suggest that Bd is a newly introduced invasive pathogen. These case studies, from Central America and western North America [6,22], describe healthy amphibian populations when Bd is absent, but show acute die-offs and subsequent population declines immediately following detection of Bd at study locations. Results in both regions parallel other well-studied disease systems (e.g., mammalian rabies, Lyme disease, and Ebola virus [23–27]), where invasive diseases spread across the landscape, invading new areas and affecting naive populations. Such results (although not framed as rigorous tests of hypotheses) suggest an outcome for future declines with some degree of certainty [23,28].
Predicting the impacts of an invasive disease will require an understanding of the biotic and abiotic factors that influence the interactions between the host and pathogen. Temperature and moisture influence most aspects of the biology of amphibians [29,30]. Thus, climate change likely will impact amphibians as a result of both direct effects on physiology and as a result of indirect effects following changes in interactions with other species. Indeed, responses to climate change by amphibians are numerous (reviewed by [31]) and may occur on the scale of individuals [32], populations [33–35], and species [13]. Documented effects of climate change on amphibians typically have been in the form of population fluctuations or long-term declines, rather than sudden die-offs and subsequent rapid declines in local populations.
Just as temperature, rainfall, and humidity influence the biology of amphibians, so too might these factors affect the growth, persistence, and ecology of a potential pathogen. Changes in regional or local climate may directly or indirectly alter pathogen development and survival rates, disease transmission, and host susceptibility [36], thereby influencing the degree of host population response. Like amphibians, much of the basic biology of Bd is affected by temperature and moisture. In the lab, optimal growth occurs between 17–25 °C, and death occurs at temperatures above 29 °C or below 0 °C, or after prolonged desiccation [37]. Such lab results are reflected in field studies from Australia, Panama, South Africa, and the western US (e.g., California and Colorado) in which prevalence of infection varies by season, elevation, or region [38], with increased prevalence associated with cooler temperatures and moister conditions [39–43]. All of these studies occurred at locations where Bd had been introduced prior to the study, and was endemic during the course of research. Although environmental factors likely influence the survival and growth of Bd, there is no evidence that climatic factors cause outbreaks of chytridiomycosis from resistant spores or from saprophyte forms. The lack of evidence for climate-induced outbreaks is especially important to note, given recent suggestions that promote climate change as a potential mechanism for amphibian declines [17].
Recent global climate change is well documented [44] and is confounded in time with the recent declines of amphibians [2], which requires careful analyses to distinguish correlation from causation. In Central America, temperature is predicted to increase, and rainfall is predicted to decrease [44], making many of these areas less favorable for Bd. Paradoxically, the CLEH predicts that amphibians should decline in years that follow an unusually warm year because “shifts in temperature influence disease dynamics” ([17] p. 161). Under this scenario, Bd is proposed to emerge from a dormant state, or switch from a facultative saprobe into a pathogen as the environment becomes warmer. Specifically, the CLEH proposes that outbreaks of chytridiomycosis are triggered by a shrinking thermal envelope, in which maximum temperatures become cooler, and minimum temperatures become warmer [17]; this effect is hypothesized to be most pronounced at mid-elevations. The CLEH was based on relating timing in changes of regional temperature to the timing of the disappearance of species of harlequin frogs (genus Atelopus) in Lower Central and northern South America [14]. Pounds et al. [17] presented analyses to assess associations of pan-Neotropical weather patterns and the putative timing of the disappearance of Atelopus species [14]. Oddly, the authors did not examine the database for evidence of simple spatiotemporal patterns in the disappearance of Atelopus species, which might be suggestive of epidemic movement of disease.
In this paper, we evaluate the data regarding the declines of Atelopus species [14], as well as additional data from other amphibian species in Lower Central America and Andean South America derived from published literature and from our own field work. We search for spatial and temporal patterns of amphibian declines to determine whether evidence supports an alternative hypothesis that Bd is an invasive pathogen. If Bd represents a disease epidemic, then spatiotemporal patterns would indicate that after an initial introduction, the pathogen spreads systematically through environments with favorable climates, geography, and host populations. Further, if Bd is an exotic invasive [6] in South America, then spatiotemporal patterns should be similar to those documented in Central America and Australia [6,45].
Because few die-offs and population declines of amphibians have ever been observed directly, especially prior to the 1990s, attribution of the causes and the timing of these events is often only a rough estimate, and in many cases, no additional data will ever be available. Furthermore, in the case of Atelopus, taxonomic studies are still discovering “new species” decades after the organisms have gone extinct [46]. The conservation status of this group was summarized [14] against a backdrop of taxonomic confusion, widespread habitat destruction, remote localities, sociopolitical challenges, and in the absence of population data for most Atelopus species. These uncertainties led to the conclusion that “in many cases the available information did not permit quantitative analysis.” ([14] p.192).
Findings [14] were presented in the form of “Last Record” to indicate the last known sighting of a live individual in the field, and were often based on the last known museum specimens collected. The conservation status of each species was classified as “Stable,” “Decline,” or “Data Deficient,” and well-qualified causes (e.g., habitat loss, introduced predatory fish, Bd, and other, unknown factors) of the apparent declines were presented [14]. Many of the detailed notes and comments regarding the suspected causes of declines did not appear in the final publication, but a final draft of the full database (dated 15 March 2004) was circulated among the coauthors and to other members of the RANA network (http://rana.biologia.ucr.ac.cr/). These data were used [17] to test the proposed CLEH. They renamed “Last Record” as “Last Year Observed” (LYO), and used those dates as a proxy for the actual date of disappearance of individual species. Realizing that the data contained uncertainties, the following justification was presented:
“Undoubtedly, the LYO does not accurately represent the timing of a disappearance in some cases, especially in tier two. Thus uncertainty is high for any particular species, and the strength of our conclusions lies in the broad patterns. Errors in the data could generate these patterns [of correlation of timing of disappearance with unusually warm years] only if sampling were biased so that the LYO tends to follow a relatively warm year irrespective of the timing of disappearances. A decline in a cool year might be misclassified as having occurred in a warm year, but the reverse is no less probable.” (see Supporting Online Material p. 18 in [17]) (emphasis ours).
Ultimately, testing hypotheses such as the CLEH or Bd as an invasive pathogen requires data about the spatiotemporal patterns of Bd occurrence. Unfortunately, little sampling of Bd from nondeclining populations or from the environment (independent of frogs) has been undertaken, with the result that frog decline has been used as a proxy for Bd appearance. This leads to an important distinction between LYO data [17] and the actual time when Bd arrived at a location. When using amphibian population declines to study spatiotemporal patterns of Bd, the datum of interest is the date of the actual decline (DOD, hereafter) because it more accurately reflects the timing of the arrival of Bd to a site than does LYO (Figure 1).
LYO, as the year of last record [14], may not reflect Bd dynamics for a number of reasons. First, because the entire range of Atelopus species could not possibly be surveyed in a systematic fashion, LYO might indicate the last time scientists visited a particular site and noted a particular species [14]. Second, species occurring in remote or politically unstable areas have not been surveyed in decades [14]. Third, rediscovery of a few individuals changes LYO, despite a known DOD. For example, A. mucubajiensis declined between 1988 and 1990 [47–49], but a rediscovery of a few individuals in 2004 [17] produces an LYO 14–16 y after the DOD. A similar example is the case of A. bomolochos in Ecuador. The species declined in 1980 [50], but has an LYO of 2002, based on a singular sighting decades after its DOD, as noted in the original 15 March 2004 database: “One individual seen in 2002 in the PN Sangay (D. Almeida), but otherwise this formerly abundant species has disappeared from its range.” This species had an initial status of Decline [14], and infected individuals were found in 1980 and 1991. It was later reclassified as Stable [17], although data indicate this species declined precipitously in the 1980s (likely due to Bd), with a singular sighting in 2002.
Pounds et al. [17] acknowledged the Atelopus dataset contained uncertainties and errors (see above). However, their analytical approach did not explicitly test how error around LYO—at any quantified level—affected the regression analyses that were performed to assess relationships between temperature and timing of disappearance. The bootstrapping techniques used by [17] calculate confidence intervals for estimated parameter values from a given set of data [51]. Though used on a wide array of subsets from the original data (e.g., separate tiers of frogs, rediscovered species, and species in protected areas), this approach did not explicitly address the potential effects of the acknowledged errors in LYO on the estimated correlation coefficients.
Elementary statistical theory indicates that increases in sampling error, independent of bias in the parameter estimator, will reduce model fit simply by increasing the variance in the variables. In the dataset, DOD differed from LYO for 24–27 of 54 species, ten of these were Stable despite a reported LYO, and for those species with different dates for LYO and DOD, the mean difference was 11.2 y ± 8.2 (standard deviation [SD]) (Table 1). Thus, previous analyses [17] using LYO data had large amounts of sampling error relative to the 1-y time lag used in the regression models [14,17]. Given the low accuracy of the data to estimate Bd in time, we wanted to determine whether the data could support such fine resolution of statistical relationships between species declines and temperature. As such, we performed an explicit examination of how the observed sampling error in LYO would affect correlations between population declines and air temperature (AT) [17]. We perform similar simulations on our new analyses correlating time since probable infection and distance of spread.
We tested the robustness of the relationship between AT and LYO data reported in [17] (see Figure 3 in [17]), using a Monte Carlo analysis to examine the effect of systematic error on LYO date that might be expected from sampling error for Tier 1 species. First, we added bidirectional error around LYO data sampled from a normal distribution, with a mean of zero and an increasing standard deviation of up to 6 y (a conservative amount given the observed differences). This type of error would imply that either a DOD (not necessarily an extinction event) occurred sometime around the LYO date, or that because searches for individuals used in LYO dates were not systematic, the true LYO date could have been missed. Next, we added error forward in time only, indicating that the real date of extinction occurred sometime after the last observed individual was noticed by observers in the field. This error could especially be likely given that individuals from populations at low density might be difficult to detect and because, as above, searches for individuals were not systematic. This directional error was modeled in three ways. First, we added error to LYO sampled from a uniform distribution, with error ranging from 0 to 6 y (i.e., LYO was as likely to occur some time after the reported date as it was likely to occur at some time close to the reported date). Second, we added error sampled from a Poisson distribution, with a mean error from 0 to 4 y (i.e., the actual extinction date likely occurred closer to the reported LYO date than later). Third, we similarly added error sampled from an exponential distribution, with a mean error from 0 to 6 y. For the Monte Carlo simulations, we ran 10,000 trials for each year of error that was added to LYO, for all types of error additions. For each trial, we calculated the correlation coefficient of the relationship between the resultant LYO and AT following the time interval 1970–1998 [17]. For each randomized trial, we report the mean correlation coefficient and a 95% confidence interval. Further, we report the number of trials out of 10,000 that produced a result equal to or greater than the magnitude of the result reported in [17]. A low number of such trials would suggest that the result from the original analysis is not a reliable estimate for the true relationship between climate and LYO.
We gathered records of approximate DOD and Bd infection (Table 1) from published studies based in South and Central America, from the Global Amphibian Assessment (GAA) [1], from the Atelopus database [14], and from our own field work. In the case of the Atelopus database [14], we excluded species that persisted at typical levels of abundance (Stable [14]), species that were Data Deficient, and lowland species for which evidence of disease- or climate-induced declines was lacking. In these analyses, we used species (or groups of species at a site) for which there was a reasonable estimate of the year of actual decline. For purposes of comparison, we present (Table 1) both our estimate of DOD and the estimate of LYO ([14,17] and other sources cited in Table1). Geographic coordinates for each species (Table 1) represent the geographic centroid of the known range of the species derived from the GAA database (V. Katariya, personal communication). Elevational ranges were taken from the GAA. In the case of site-specific declines, elevational data and geographic coordinates represent the data for that site.
Rates (kilometers/year) of spread were calculated by dividing the distance between pairs of locations by the interval between the DODs. Rates of spread in Central America (our Wave 1) have been described previously [5,6]. These data were updated and recalculated by supplementing the original database with DOD data with additional published records (Table 1; Wave 1).
We propose Bd as an exotic introduction with subsequent spread as an alternative hypothesis to the CLEH. We created a map of known locations of Bd by date using DOD and geographical data (Table 1). Starting with the earliest record, we built temporal demarcations of Bd movement that depicted the fastest observed spread by finding locations that were both the farthest away and with the shortest time interval from the previous earlier DOD. We created maps along each of five primary cordilleras assuming Bd could not move through lowland areas, given its thermal requirements. By creating demarcations using consecutive years of data, we could determine whether the area covered by Bd increased in size through time, as expected if Bd were to spread via a wave-like expansion. Our null hypothesis was that little or no wave-like pattern would indicate that Bd presence was randomly scattered across space and time (e.g., [27]).
Further, for each wave, we ran additional regression analyses [27] to quantitatively test for a linear relationship between time and space in Bd spread. Distances between the origin of disease introduction and DOD location were regressed against the differences in time between the earliest DOD and the date of introduction within a particular wave. We did not account for autocorrelation among geographic locations in our analysis because spatial relationships were modeled directly as the causal factor underlying the temporal patterns of disease. Such a correction would have been necessary if we had sought to explain temporal patterns of disease with some factor other than spatial distance (e.g., altitude, climate, and habitat type). Significance for the regression was determined through randomization techniques [52] to ensure robustness against any violations of normality in the data [52]. Ten thousand random datasets were constructed by pairing values of distance with values of time that were drawn randomly (with replacement) from the pooled distribution of times in the dataset within a wave. Regressions were then calculated for each of these random datasets. The significance of the original regression was determined by comparing the β from the original regression with those from regressions of the randomly constructed datasets. The significance level of the test was calculated as the proportion of values in the distribution of β's from the randomized datasets that were as extreme as or more extreme than the β calculated in the original regression. A significant (p < 0.05) regression coefficient supports disease spread through space and time whereas a nonsignificant regression indicates random occurrences of disease across space and time. This approach allows for testing the competing hypotheses that Bd is either a spreading, invasive pathogen or an emerging, endemic pathogen, and is similar to issues and statistical methods regarding Ebola outbreaks in African primates and humans [27].
Similar to our approach for LYO, we developed a Monte Carlo procedure that simulated error in DOD to determine how robust the regression analyses of each wave were to sampling error in DOD. For each wave, we added error sampled from a uniform distribution to each datum backward in time for up to 20 y. For the error simulations on each wave, we ran 10,000 trials for each year of error that was added to the time since earliest DOD. From these trials, we calculated the mean and 95% confidence interval on the slope of the relationship (i.e., β) between the number of years since the earliest DOD against distance from the earliest DOD using linear regressions. We considered the amount of error necessary to falsify the hypothesis of an epidemic wave as that when the 95% confidence interval of the slope overlapped zero.
The earliest records of Bd in South America are from A. bomolochos in Cañar Province, Ecuador, in 1980 [12]; another relatively old record from 1986 (associated with a DOD ca. 1977; Table 1) exists from A. cruciger, near Caracas, Venezuela [53]. Consequently, we used these as two starting points, with the first assuming a Bd introduction in Cañar, Ecuador with subsequent spread, and the second assuming an introduction near Caracas, Venezuela (Wave 2). We calculated the rates of spread from Cañar (1) northward into Colombia along the Cordilleras Occidental, Central, and Oriental (Waves 3a); (2) northeastward into Venezuela along the Cordillera Oriental of Ecuador and Colombia and into Los Andes de Merida (Wave 3b); and (3) southward into Peru along the Cordillera Occidental and Oriental (Wave 4).
We calculated rates of spread assuming the Venezuela record represents a second, independent introduction of Bd into Andean South America (Wave 2). If this is the case, the Venezuelan introduction likely represents the third introduction onto the continent, as the second oldest record of Bd in South America is from the Atlantic coastal forest of Brazil in 1981 [16]. This study focuses on the Lower Central America and Andean Region of South America, so we do not include these Atlantic Brazilian records in the calculations here; evidence of an epidemic wave in Brazil was discussed previously [5].
According to the CLEH, extinction risk for Atelopus species is greatest at mid-elevations, between 1,001 and 2,399 m [17] (see Figure 1 in [17]). Using the same dataset, LaMarca et al. [14] originally concluded that no species of Atelopus above 1,500 m was Stable. The differences in conclusions between these two analyses were striking and were likely caused by two factors. First, the use of LYO [17] excluded information about status [14]. Second, while using LYO data as a proxy for the percentage of species missing by elevation, Pounds et al. [17] used a cut-off of 1998 to determine status from the LYO data. Species observed between 1999 and 2004 (final year in the database) were categorized as having neither declined in abundance nor gone missing. The justification for 1998 as a cut-off date was apparently that no LYO of 1999 was recorded, thus representing “a natural break in the data” (see Supporting Online Material, p. 24, of [17]).
The use of LYO instead of DOD, in combination with the arbitrary 1998 cut-off date [17], could have obscured possible elevational patterns of decline in Atelopus, so we analyzed Status data [14] with no temporal cutoffs to determine elevational patterns in declines. We used the same elevation categories as [17] and included only species categorized as Stable or Decline, excluding species considered Data Deficient. We considered A. tricolor as Decline (contra [17]) based on expert commentaries in the original Atelopus database [14] (see comments above). We calculated an exact Pearson chi-square test statistic for the four elevation classes in a two-way (Stable or Decline) contingency table because some cells had zero species.
Although it is widely assumed that the decline of amphibians in 1987 at Monteverde Cloud Forest Reserve, Costa Rica, was the result of an outbreak of Bd, direct evidence of such does not exist. A positive record of Bd in 2003 (Supporting Online Material, p. 3, of [17]) indicates that Bd is now endemic to the area. Because the CLEH was based on patterns of climate change and the loss of A. varius at Monteverde [54,55], we examined museum specimens for evidence of presence of Bd prior to 1986. We collected skin samples from the pelvic patch of 64 frogs collected between 1979 and 1984 from mid-elevations (985–1,645 m). The specimens represented 18 species of anurans, 12 of which are riparian and 11 of which have been shown to be infected by Bd elsewhere in their range (Table S1). Skin samples were stored in 70% alcohol, transported to the histology lab at Southern Illinois University, Carbondale, where we followed histological procedures as described in [56]. We calculated Bd infection prevalence and 95% Clopper-Pearson binomial confidence intervals for the 64 samples.
Merino-Viteri [57] conducted a similar histological survey of 174 museum specimens representing more than 16 species, collected between 1953 and 2000, from throughout Ecuador. We calculated Bd infection prevalence and 95% Clopper-Pearson binomial confidence intervals to determine the likelihood of detection if Bd were present prior to evidence of die-offs. For Ecuador, we used two subsamples of this database [57]: (1) 32 specimens (29 A. ignescens and three other Atelopus spp.) collected prior to 1980, and (2) 89 specimens of Atelopus ignescens sensu strictu [46] collected between March 1953 and November 1987.
We found evidence of directional spread of Bd along most of the principal cordilleras of Lower Central America and the Andean region (Figures 2 and 3), supporting the hypothesis that Bd is an exotic pathogen that was introduced into South America in the late 1970s–early 1980s, and has caused multiple amphibian declines in the past 30 y [6,12–14,58]. The maps indicate localized areas of initial introductions and subsequent expansion of moving wave fronts.
Although the oldest record of Bd is in the environs of Cañar, Ecuador, we find questionable the hypothesis that this record represents a singular introduction of the pathogen to all of South America. If this were the case, then the subsequent northward and eastward spread toward Caracas, Venezuela, in a mere 5 y left no trace of itself in the intervening regions (e.g., Colombia). This rate of movement (∼300 km/y; Table 1) would be much higher than any of the observed rates of spread for the region using more accurately defined intervals (Table 1). Instead, we prefer the more parsimonious hypothesis of dual introductions near Cañar, Ecuador [12], and near Caracas, Venezuela (Figure 3). The hypothesized introduction into Venezuela is based on the earliest record there of Bd (A. cruciger, 1986), at a site located 115 linear kilometers from Caracas, and reports of population declines in that species in the mid-1970s [53]. Given the proximity of the geographical centroid of A. cruciger, and the potential error in DOD, the data support the hypothesis of a second Andean introduction near Caracas in the mid-1970s, and subsequent westward spread along the Merida Andes–Cordillera Oriental axis towards Colombia and Ecuador. Our analyses indicate that Bd invaded the Cordilleras Occidental and Central of Colombia, moving northward towards Panama, and moved northeast towards Venezuela along the Cordillera Oriental–Los Andes de Merida axis. Precisely where the eastward expansion from Ecuador and the westward expansion from Caracas meet—likely in Colombia—cannot be determined accurately given the available data.
Overall, there is support for directional spread of Bd (Figure 4). The relationship between time since the earliest DOD within a wave and distance of spread was significant for Wave 1 (p = 0.0067; β = 21.17; R2 = 0.97), Wave 3a (p = 0.0013; β = 43.32; R2 = 0.47), Wave 3b (p = 0.0050; β = 56.57; R2 = 0.40), and Wave 4 (p = 0.0279; β = 61.95; R2 = 0.49). The relationship for Wave 2 was not supported (p = 0.32). Note that Wave 1, sampled with nearly no error in DOD, had much higher R2 values than the other waves, suggesting that sampling error in DOD likely reduced our ability to measure spatiotemporal patterns in Bd spread in the Venezuelan region (Wave 2, see error analyses below).
Bd can spread rapidly, moving across entire countries in less than 5 y and across northern South America in approximately 20 y. The estimated rates of spread of Bd in the Andes are generally consistent with rates reported from Central America [6], viz., 25–282 km/y (Figure 3 and Table 1). The upper range of these rate estimations (202 and 282 km/y) results from the record of Bd in Peru (A. tricolor, 1998) and a 1-y difference between A. eusbianus and A. quimbaya approximately 280 km apart.
The previously published [17] relationship between AT and LYO was not robust to the incorporation of sampling uncertainty in LYO data (Figure 5). The original analysis [17] suggested a correlation coefficient of 0.65, relating LYO and AT from the previous year. However, using a normal error distribution of mean zero, the correlation coefficient quickly fell to 0.14 after the standard deviation of LYO was increased to 6 y. Even when the standard deviation was only 2 y, the 95% confidence interval overlapped zero, and only 1.4% of the simulated trials resulted in a correlation coefficient equal to or greater in magnitude to that of the original analysis. The decrease in the magnitude of correlation coefficients was not as great using Poisson, exponential, or uniform error forward in time, but suggested an overall poor relationship. Using a Poisson error distribution, the correlation coefficient dropped to an average of 0.30 for error ranging from 1.5 to 5 y. Further, no more than 2.6% of simulated correlation coefficients showed a value equal to or greater than the original result after 1.5 y, and the confidence interval of the result overlapped zero. After 2 y of error sampled from a uniform distribution, the correlation coefficient averaged 0.34 in simulated results, and only 2.6% of the simulated correlation coefficients were greater than or equal to the original result when including any amount of error. Similarly, after 2 y of error sampled from an exponential distribution, the correlation coefficient averaged 0.42 in simulated results, and only 5.7% of the simulated correlation coefficients were greater than or equal to the original result when including any amount of error. Interestingly, all of the correlation coefficients remained positive when error was added. This positive relationship is likely the result of both Bd spread and AT being positively correlated with time.
Unlike the relationship between AT and LYO, the relationship between time since the earliest DOD and distance of spread of Bd was robust to additional sampling error (Figure 6). When adding error backward in time from a uniform distribution, the relationship in Wave 1 remained statistically significant when random error up to 16 y was applied, in Wave 3a and Wave 4 up to 20 (+) y, and in Wave 3b up to 18 y.
An increasing percentage of Atelopus species declined as elevation increased (chi-square = 13.16, df = 2, p = 0.0014), with 100% of species occurring above 1,000 m having declined prior to 2004, while only 30% of those species near sea-level declined (Figure 7).
None of the 64 individual frogs (Table S1) examined histologically that were collected from Monteverde prior to the documented declines in 1987 were infected with Bd (p = 0, 95% confidence interval [CI] = 0%–5.6%). Although not conclusive, this suggests that Bd was not endemic in frogs at the site prior to that time.
Our statistical analyses of the histological results of Ecuadorian frogs [57] support our hypotheses that it was unlikely (1) that Bd was present in Ecuadorian amphibians prior to 1980 (N = 32, all negative, 95% confidence limits [CL] = 0–0.108881), or (2) that Bd was present and infecting Atelopus ignescens in Ecuador prior to this species' decline in 1988 (N = 89, all negative, 95% CL = 0–0.040601).
Our results make two main points. First, we present analyses supporting a classical pattern of disease spread across naive populations, at odds with the CLEH proposed by [17]. Second, our analyses and re-analyses cast doubt on CLEH. We discuss these findings in more detail below.
Our analyses of data for declines of amphibians in the Andean region of South America identified spatiotemporal patterns that were robust to sampling error in DOD and consistent with the spread of an introduced, invasive pathogen [23–26], as has been documented for Bd in Central America ([6] and additional results presented here). The oldest records of Bd in South America are 1980 in Ecuador [12], 1981 in Brazil [16], and 1986 in Venezuela [53], although the earliest reported declines of amphibian populations are in the “mid-1970s” in Venezuela [53], mid-1980s in Ecuador [12], and 1979 in Brazil [59]. Targeted searches for Bd in museum specimens collected from these areas at these times may produce older records.
Results of histology support our hypothesis that Bd was absent from Ecuador prior to 1980 and was introduced as an exotic pathogen. We use that date as the original introduction from which we modeled the spread along Waves 2, 3a, 3b, and 4. Within 7 y, Bd was found throughout Ecuador, supporting the hypothesis that it is an exotic pathogen invading naive populations.
Of particular importance when evaluating our results in light of the known errors in DOD are the results of the Monte Carlo simulations. The statistically significant spatiotemporal regressions of Bd dynamics were robust to even large amounts of error in DOD. These results strongly indicate that even if some of our DOD data were incorrect (which is likely given the quality of the Atelopus database), the wave-like pattern would remain robust.
Interestingly, the pattern of wave-like spread was more robust to sampling error than the CLEH. This result is likely due to resolution of LYO and DOD. Both LYO and DOD have inherent error because neither was collected with foreknowledge of pending declines—unlike data in Central America [6]. As such, the resolution of data is on the scale of roughly a decade (on average 11.2 y ± 8.2 SD). Because the CLEH predicts that declines occur one year following a relatively warm year [17], one should not expect the trend to hold up to the sampling error inherent in LYO or DOD. Indeed, LYO and DOD would need to have a resolution of 1 y or less to adequately evaluate the CLEH. The hypothesis of an epidemic wave, on the other hand, does not require such fine-scale resolution in time. There is no minimum distance that the disease needs to spread in a single year, just that the distance increases with an increasing number of years. In this case, data quality and the number of points in the relationship become especially important. This point can be demonstrated well with Wave 2. If A. sorianoi is excluded from Wave 2, the relationship for spread goes from not statistically significant (p = 0.32) to statistically significant (p = 0.0153; β = 17.2; R2 = 0.92). In this case, Wave 2 would be nearly as well supported as Wave 1, but note that the slope of the relationship is rather low, so outliers will greatly affect this wave. Such a finding underscores the need for quality data.
In addition to identifying four regions with wave-like spread, results indicated that declines, extinctions, and epidemics occurred years after the leading edge of the wave had passed. This has likely prevented others from recognizing the invasion pattern, such as the logistic regressions using the entire dataset based on geographic coordinates performed by Pounds et al. [17]. Some of these events are likely real, and are typical of spreading pathogens, described as “great leaps forward” [60,61]. Other jumps likely are not real, but result from a lack of accurate field data, human facilitation of movement [61], effects of small-scale geography, effects of population dynamics, or from limitations of our large-scale approach. For example, each location represents the estimated geographic centroid of a species' distribution, not necessarily the precise location of Bd infection. We note that the Central American wave, measured with almost no error, had the strongest relationship between distance and time, and showed the least amount of variation in rates of spread despite being measured over the smallest spatial scale. Thus, the sampling error in the South America data may be large, and given the relatively rapid rate of Bd spread, makes smaller-scale detections of patterns, such as the spread of Bd through Ecuador, impossible to discern.
Geography may influence gene flow in plants and animals [62], including amphibians [63] and their pathogens [20]. In the case of chytridiomycosis, it is likely that certain habitats will promote the survival and spread of Bd (e.g., mountain chains and river valleys), other habitat features are likely to slow its spread (e.g., deserts and lowlands), and some are sufficiently remote or isolated such that invasions may be delayed (e.g., isolated highland ponds and upper reaches of minor tributaries). Finally, anthropogenic facilitation of Bd spread is also likely, perhaps along highways, seasonal routes of livestock herding, and other travel routes that might produce a pattern of nonlinear declines.
Our estimates of Bd movement across all regions are quite crude, and we note that studies at small scales have found the slowest rates (e.g., 1.1 km/y in California; V. Vredenburg, personal communication), regional scales provide intermediate estimates (e.g., 26 km/y in Costa Rica and Panama [6]), and continental comparisons produce the highest rates of spread (282 km/y South America; this study). The scale of our study and, especially, the nature of the data available cannot provide robust estimates of fine-scale disease spread—either historical or predictive—for the region. However, these results can guide searches for historic records, and direct future research into finding evidence of multiple invasions or investigating potential mechanisms of transmission and spread. Ultimately, this suggests that different processes drive Bd spread at different scales. For example, rapid movement between and across continents may be caused by human-facilitated movements [64], whereas regional spread within continents is by natural dispersal through riparian corridors, and local spread via amphibian dispersal and transfer of Bd among individuals.
Our analyses and re-analyses of data related to the CLEH all fail to support that hypothesis. First, our simulations indicate that the correlations reported in [17] are not robust and fail with the inclusion of even small levels of error. This is especially important when considering the 1-y time lag between climate events and frog declines that is central to the CLEH model; that approach is very sensitive to slight errors in dates, and the effects of such error were not explicitly examined in that study. While acknowledging the uncertainties in the frog data [17], the potential effects of error in LYO were not explicitly tested at any level of analyses. We note that our simulations not only explored the effects of bias in estimates of LYO (i.e., forward sampling), but also unbiased random errors (errors from a uniform distribution). We consider our tests of error on LYO conservative in that our maximum levels of error (6 y), was smaller than the observed discrepancies between DOD and LYO.
Second, we show the elevational pattern in Bd prevalence [17], a cornerstone of CLEH, was not a real biological pattern but instead was caused by the combined use of LYO data and an arbitrary time cutoff of 1998. Reanalysis of these data using the more accurate Status information [14] shows an elevational pattern at odds with CLEH. Our results are essentially the same as those originally reported [14], in which there is an increasing percentage of Atelopus species declining as elevation increases, so that 100% of species occurring above 1,000 m declined prior to 2004. In contrast, the use of 1998 as a cutoff date [17] completely changed the results of elevational patterns of extinction in the Atelopus database, and incorrectly produced an artifactual “mid-elevation peak” in extinction that was then used to support the CLEH.
Third, it was suggested [17] that Bd might be an endemic saprophyte that emerges as a pathogen as a result of climatic anomalies. This was a key element of the CLEH hypothesis and the conceptual basis for much of the analysis. If true, Bd should be present, but nonpathogenic, in populations prior to declines. Although we cannot rule this out, currently no evidence supports this hypothesis, including our surveys for Bd in museum specimens collected at Monteverde, Costa Rica, and those collected from throughout Ecuador [57]. In the case of A. ignescens in Ecuador, a large number of individuals were examined with histology; another sample was available for additional species throughout Ecuador prior to the oldest record of 1980 [57]. These data indicated only a 4%–11% probability that Bd was present in these populations, but undetected, prior to known declines. For Monteverde, that probability was estimated at 5.6%, although we had to combine results from many species. Given differences in species susceptibility to Bd, this approach does not estimate true prevalence, although this approach produced robust estimates for a decline in Panama [6]. It is never possible to prove that Bd was absent from a population, but demonstrating that the upper 95% confidence limit for its prevalence is low [6] may be adequate.
Bd is hypothesized to be an emerging infectious disease [7,11,22] that has been introduced at various times and sites around the globe; vectors for such geographic spread are assumed to be nonnative frogs such as Xenopus laevis or Lithobates catesbeianus [64,65], but conclusive evidence of historical introductions are unlikely to be identified. A more compelling indicator of multiple introduction events lies in the genetic signature of the pathogen itself, as described in fine-scaled studies in the Sierra Nevada of western North America [20]. The debate as to whether Bd is an exotic pathogen spreading through naive populations, or whether it is an endemic pathogen whose emergence is stimulated by climate change is strikingly similar to the discussion of the origin of Ebola Zaire (ZEBOV; [27]). In the case of ZEBOV, the addition of genetic data to spatial and temporal patterns of outbreaks strongly supported the hypothesis that ZEBOV had recently spread across the region. This approach [27] is necessary to understand the historical patterns of introduction and spread in South America, and at Bd-endemic sites in Lower Central America, where a single epidemic wave has been hypothesized [6].
Although we propose at least three independent introductions of Bd into South America (Ecuador, Venezuela, and Brazil), we note that our estimated rate of spread from A. peruensis to A. tricolor, in Peru, is high (∼202 km/y, suggesting perhaps that another introduction has occurred somewhere to the South. Bd has been present in neighboring Argentina since at least 2002 [66], but has yet to be reported from Chile [1]. However, one of the earliest enigmatic [2] declines in the Neotropics was that of Rhinoderma rufum, which was regularly seen in temperate forests and bogs in Chile until 1978 when it completely disappeared. Like the US and Europe, Chile received shipments of Xenopus from South Africa, and introduced populations were established there in the wild by 1944 [64]. We encourage surveys for Bd in preserved and living amphibians from Chile to test this hypothesis.
Our analyses support a hypothesis that Bd is an introduced pathogen that spreads from its point of origin in a pattern typical of many emerging infectious diseases. Furthermore, although we acknowledge that climate change represents a serious threat to biodiversity, and likely influences endemic host–pathogen systems, the available data simply do not support the hypothesis that climate change has driven the spread of Bd in our study area.
That Bd is spreading into new populations and new regions is not surprising. What has been less clear is whether climate change has played a role in the emergence or spread of the disease. It has been mentioned [17] that patterns of decline could indicate simple wave-like spread of epidemic disease—apparently to reconcile the evidence of such in Central America [6]. The simple spatiotemporal pattern of Bd epidemics in Central America [6] has never been refuted, and has been supported by predictions by conservationists working in the region [28], and our analyses here corroborate a generalized pattern of spread in South America.
Disease dynamics are the result of a complex process involving multiple factors related to the hosts, the pathogen, and the environment that may affect disease patterns on many spatial scales. Disease dynamics are affected by micro- and macro-climatic variables [36], and such synergistic effects likely act on Bd and amphibians, because life histories of both the host and pathogen are directly influenced by both humidity and temperature. For example, environmental changes may cause some enzootic pathogens to become epizootic by causing them to expand their distribution into new areas, alter host behavior, or affect transmission rates [36,67]. Global climate change will directly influence amphibians but will also affect them indirectly through synergisms with habitat alteration, environmental contamination, diseases, and other challenges. A singular failure of recent studies involving climate and the amphibian crisis has been to seek evidence of interaction between temperature and disease across large areas that include sites where Bd is nonendemic, where it is endemic, and where it is epidemic, and to do so against a confounding backdrop of steadily increasing temperature regimes [17,18].
A straightforward example of how climate change might drive amphibian extinctions from invasion of Bd into new environments was described for the Peruvian Andes [15]. Glacial retreat produced new meltwater ponds that were colonized by three species of frogs [15] and later by Bd. This event was followed by die-offs of adults, and declines in metamorphic juveniles and tadpoles. Regional climate analyses [38] indicated this area of frozen ice and snow was not within the appropriate thermal envelope for Bd, but measurements of water temperature indicated that solar heating was capable of warming ponds to levels tolerable by Bd.
Future studies need to first determine whether or not Bd is present at the site, and then determine how long it has been there. Scientists must acknowledge that amphibians are proxies for the presence of Bd and carefully examine data related to declines to accurately portray the timing of infection. LYO data likely are inaccurate, and researchers should attempt to discover the time of Bd arrival or use DOD as more accurate estimates. Where Bd is endemic, complex synergistic interactions exist among the hosts, the pathogen, and climatic variables [68]. If Bd is already endemic (either naturally or introduced), then populations should be studied and monitored to evaluate relative susceptibility to fatality and catastrophic declines [69]. Where Bd is not yet present (e.g., currently Panama east of the Panama Canal), key challenges will revolve around preventing introductions and developing plans to conserve species- and genetic-level diversity in the event of epidemics (e.g, the Madagascar Amphibian Action Plan).
|
10.1371/journal.pgen.1007104 | A case-control collapsing analysis identifies epilepsy genes implicated in trio sequencing studies focused on de novo mutations | Trio exome sequencing has been successful in identifying genes with de novo mutations (DNMs) causing epileptic encephalopathy (EE) and other neurodevelopmental disorders. Here, we evaluate how well a case-control collapsing analysis recovers genes causing dominant forms of EE originally implicated by DNM analysis. We performed a genome-wide search for an enrichment of "qualifying variants" in protein-coding genes in 488 unrelated cases compared to 12,151 unrelated controls. These "qualifying variants" were selected to be extremely rare variants predicted to functionally impact the protein to enrich for likely pathogenic variants. Despite modest sample size, three known EE genes (KCNT1, SCN2A, and STXBP1) achieved genome-wide significance (p<2.68×10−6). In addition, six of the 10 most significantly associated genes are known EE genes, and the majority of the known EE genes (17 out of 25) originally implicated in trio sequencing are nominally significant (p<0.05), a proportion significantly higher than the expected (Fisher’s exact p = 2.33×10−17). Our results indicate that a case-control collapsing analysis can identify several of the EE genes originally implicated in trio sequencing studies, and clearly show that additional genes would be implicated with larger sample sizes. The case-control analysis not only makes discovery easier and more economical in early onset disorders, particularly when large cohorts are available, but also supports the use of this approach to identify genes in diseases that present later in life when parents are not readily available.
| Trio exome sequencing and de novo mutation (DNM) analysis has been the main approach to discovering genes responsible for severe sporadic disorders, including a range of neurodevelopmental disorders. This approach requires sequencing parents, identifying DNMs from trio sequence data, and comparing the observed rate of DNMs to the expected. In this study, we adopted a case-control design, performed a gene-based collapsing analysis, and rediscovered several of the epileptic encephalopathy (EE) genes originally implicated by DNM analysis of EE trios. Our collapsing analysis focused on ultra-rare, highly impactful variants (“qualifying variants”) by filtering against large-scale population datasets, and this approach revealed that most of the standing variation can be filtered out and DNMs are enriched in “qualifying variants”. Our study suggests that a case-control analysis approach can be used to identify disease genes with causal mutations that are predominantly de novo in place of trio-based analysis methods. This offers an efficient and cost effective alternative approach when large-scale trio sequencing is not possible.
| One of the most important recent developments in human genomics is the use of a trio sequencing paradigm to implicate new disease genes in sporadic disease by evaluating patterns of de novo mutations (DNMs). This framework compares the observed pattern of DNMs in probands to the expected based on the size of the protein-coding sequence and the estimated tri-nucleotide mutation rate[1], and has implicated scores of genes conferring risk of epilepsy[2, 3], intellectual disability[4–6], autism[7–10], and other neurodevelopmental conditions[4]. This approach is costly because of the need to sequence complete trios and often is not practical or possible for conditions that present after childhood where parents may not be available for sequencing. Moreover, a precise estimate of mutation rate is not available for small insertion/deletions (indels)[1], limiting the ability to assess the significance of genes harboring de novo indels.
In parallel to these developments, collapsing analyses, which typically compare the burden of rare, presumably deleterious variants gene by gene in cases versus controls, have proven increasingly successful in implicating diseases genes, for example in amyotrophic lateral sclerosis[11, 12], idiopathic pulmonary fibrosis[13, 14], and monogenic disorders[15]. Surprisingly, however, it has not yet been assessed whether the collapsing framework can identify the genes implicated by analysis of trio sequencing data. We addressed this question by implementing a genome-wide gene-based collapsing analysis using whole exome sequencing (WES) data generated from 488 epileptic encephalopathy (EE) patients, including those previously analyzed using the trio-based DNM analysis framework, and a large cohort of unrelated control individuals to assess the efficacy of case-control analysis to identify disease genes implicated by DNM analysis for EE. Strikingly, despite a modest sample size, we identified three known EE genes achieving genome-wide significance (p<2.68×10−6), and found that the majority of the known EE genes (17 out of 25) originally implicated in trio sequencing are nominally significant (p<0.05). While not all known EE genes reached genome-wide significance, the significant enrichment of known genes among nominally significant p-values genome-wide suggests that with larger samples sizes many of these genes will reach p-values that will exceed that threshold. Collectively, our results show that collapsing analysis can effectively implicate genes carrying causal DNMs, and trio sequencing is not the only effective strategy for gene discovery even in genes that confer risk largely due to DNMs. We argue that the fundamental reason for this is that existing filtering strategies are increasingly accurate in identifying very young mutations including those that are de novo in the proband.
The collapsing analysis compared a total of 488 cases with 12,151 controls (S1 Fig). Three genes (Fig 1, Table 1, S1 Table, and S2 Fig), KCNT1, SCN2A and STXBP1, showed enrichment of qualifying variants in EE patients and achieved genome-wide significance (p<2.68×10−6). No other genes were found to be genome-wide significant by both Fisher’s exact test and logistic regression p-values, but 17 of the 25 genes (68%, including the three above) known to be associated with dominant EE (https://www.omim.org/phenotypicSeries/PS308350) were nominally significant (logistic regression p<0.05) in this dataset, all showing enrichment of qualifying variants in EE patients (Table 1). This is in contrast to the total of 885 nominally significant (logistic regression p<0.05) genes out of all the 18,503 genes tested (Fisher’s exact p = 2.33×10−17). We used a hypergeometric test to assess whether these 25 known dominant EE genes tend to have lower p-values in our case-control gene-based collapsing analysis compared with the rest of the genome. Specifically, at each observed ranking of the 25 epilepsy genes (based on logistic regression p-values), we performed a hypergeometric test to assess whether there were more epilepsy genes at this ranking, or lower, than one would expect if the ranks were randomly assigned to all 18,503 genes tested (Table 1). There was a consistent pattern that known dominant EE genes tended to have smaller p-values in our dataset (Table 1).
In the 25 genes known to cause dominant forms of EE, 74 of the 488 cases (15.16%) had at least one qualifying variant, compared to 302 of the 12,151 controls (2.49%, Fisher’s exact p = 1.95×10−32). Among the 64 of the 74 cases with trio WES data, a total of 73 qualifying variants were found in these 25 EE genes, and 47 of these qualifying variants (64.4%) were confirmed to be de novo in our previous DNM analyses (Table 1 and S2 Table), including all the qualifying variants in STXBP1 (n = 6), DNM1 (n = 5), KCNQ2 (n = 3), GNAO1 (n = 2), CDKL5 (n = 3), ALG13 (n = 1) and SLC35A2 (n = 1) identified in the 488 cases (no inherited qualifying variant was observed in these genes in all cases; Table 1).
Comparing 488 EE cases and 12,151 controls using a gene-based collapsing analysis of “qualifying variants”, we successfully identified three known EE genes at genome-wide significance level. In addition, known EE genes were found to have smaller than expected association p-values compared with the rest of the genome. We showed that DNMs contributed to the majority of qualifying variants in the 25 known dominant EE genes identified in cases, and in several genes they accounted for all of them. As most of these 25 EE genes are originally implicated by sequencing trios and analyzing DNMs, our results clearly demonstrate the efficacy of case-control gene-based collapsing analysis to identify genes without spending effort specifically ascertaining DNMs by sequencing trios.
Several factors affect the power of case-control gene-based collapsing analysis, including locus heterogeneity, penetrance, and how “qualifying variants” are defined as a class to represent the properties of bona fide pathogenic mutations. Because most if not all known EE-causing mutations are not observed in ExAC, we required the qualifying variants to be absent in ExAC. Remarkably, because of the large sample size of ExAC, most standing variation is essentially filtered out (except mutations arising in recent generations, including DNMs), and indeed 64.4% of the qualifying variants in the 25 known EE genes are confirmed to be de novo in 64 cases, thus recovering many of the EE genes originally implicated by DNM analysis. Notably, all the six STXBP1 and five DNM1 qualifying variants in cases are de novo, highlighting the power of using ExAC to filter out standing variation. However, even at the sample size of ExAC, where widespread mutational recurrence is observed[16], background variation in controls may still prevent a gene that is securely implicated in DNM analysis from reaching genome-wide significance in case-control analysis. For example, in DNM1, even with five qualifying variants (all DNMs) in unrelated cases, there are 18 qualifying variants in controls unfiltered by ExAC. These 18 qualifying variants may be private but not DNMs, and may be further filtered out by a larger and more genetically diverse control datasets. Indeed, although most genes known to cause EE (and other neurodevelopmental disorders) are intolerant to standing functional variation[17], implying a lower rate of background variation than the genomic average, our empirical data shows considerable variability in the frequency of qualifying controls across the 25 EE genes (Table 1). Versions of collapsing that focus on subregions of genes will likely allow finer discriminations amongst pathogenic variants and background variation.
As a class, disease-causing DNMs clearly represent the extreme of rare variation by typically not being able to pass even one generation due to extremely strong negative selection. However, this does not mean every DNM identified in an individual is pathogenic, and there are DNMs presenting as standing variation in human population datasets like ExAC and these DNMs are unlikely to be pathogenic[18]. By focusing on qualifying variants absent in ExAC, such presumably benign DNMs can be excluded from collapsing analysis. Conversely, if a pathogenic variant is inherited and the parent is not known to be affected (e.g., due to incomplete penetrance or variable phenotype), it would not be identified in trio-based analyses focused on DNMs but may be captured in case-control analyses.
The DNM analysis framework typically compares observed rate of DNMs in cases with expectation relying on estimates of the mutability of genes since very large populations of control trios are not available for direct comparisons. Precisely estimating mutation rate across the human genome is difficult and the current DNM analysis framework cannot effectively accommodate indels well due to lack of accurate estimations of mutation rate for this class of variants. However, case-control analysis directly compares the pattern of qualifying variants empirically observed in both cases and controls and is not affected by mutation rate estimates.
When a disease gene is securely implicated using a case-control framework, caution is needed to interpret the causality of qualifying variants identified in that gene. Importantly, an excess of qualifying variants in cases versus controls does not imply all qualifying variants in cases are pathogenic or all qualifying variants in controls are benign. Instead, interpretation should be performed per variant per individual after the case-control association testing is performed. Certainly, for an individual case, knowledge of whether a variant is de novo or not remains an important consideration in diagnostic interpretation[19]. However, our work clearly shows that a collapsing analysis using only probands can also discover genes that cause disease due to DNMs. This not only makes discovery easier and more economical in early onset disorders, but opens up the possibility of identifying genes that carry causal DNMs in diseases that present later in life when parents are not readily available. These results have clear implications for discovery strategies in a range of different genetic diseases.
We started with WES or whole genome sequencing (WGS) data generated from 496 cases selected from several genetic studies of EE and 12,916 controls selected from other studies and not known to have neurodevelopmental, neuropsychiatric, or severe pediatric diseases. The cases were originally recruited and studied by groups including the Epi4K Consortium, the Epilepsy Phenome Genome Project (EPGP), the Epilepsy Genetics Initiative (EGI)—a signature program of Citizens United for Research in Epilepsy (CURE), and EuroEPINOMICS-RES Consortium.
Written informed consent was collected at the time of recruitment at each of the clinical sites. Patient collection and sharing of anonymized specimens for research was approved by site-specific Institutional Review Boards and ethic committees. Details of the IRB and approval numbers are available from S3 Table.
To maximize sample size, both cases and controls included individuals with diverse ancestries including African, Caucasian, East Asian, Hispanic, Middle Eastern, and South Asian. After relatedness check and principal component analysis, a total of 488 cases and 12,151 controls remained for association analysis, and 75.6% of cases (n = 369, S4 Table) had been analyzed previously in trio or single-patient interpretation analyses.
Sequencing was performed at multiple sites (S2 Table). All data starting from either FASTQ or BAM files were processed through the alignment and annotation pipeline at the Institute for Genomic Medicine at Columbia University Medical Center (formerly Center for Human Genome Variation at Duke University). Case (S2 Table) and control samples were sequenced after exome capture using a variety of technologies (Agilent Clinical Research Exome, IDT xGen Exome Research Panel V1.0, Illumina Nextera Rapid Capture—Expanded Exome [62MB], SeqCap EZ Exome v2, SeqCap EZ Exome v3, SeqCap EZ MedExome, SureSelect Human All Exon - 50MB, SureSelect Human All Exon - 65MB, SureSelect Human All Exon V4, SureSelect Human All Exon V4 - 50MB, SureSelect Human All Exon V4 + UTR, SureSelect Human All Exon V5, SureSelect Human All Exon V5 + UTR, and VCRome2_1) or whole genome sequenced according to standard protocols.
After quality filtering the raw sequence data using CASAVA (Illumina, Inc., San Diego, CA), the Illumina lane-level FASTQ files were aligned to the Human Reference Genome (NCBI Build37/hg19) using the Burrows-Wheeler Alignment Tool (BWA).[20] Picard (http://picard.sourceforge.net) was used to remove duplicate reads and process these lane-level SAM files, resulting in a sample-level BAM file that was used for variant calling. Variant and genotype calling was performed using the GATK software with local re-alignment around insertion/deletion variants and base quality recalibration for variants[21].
Variants for analysis were restricted to the consensus coding sequence public transcripts (CCDS release 14) plus 2 base pair intronic extensions[22]. Variants were further required to have: i) at least 10-fold coverage, ii) quality score (QUAL) of at least 30, iii) genotype quality (GQ) score of at least 20, iv) quality by depth (QD) score of at least 2, v) mapping quality (MQ) score of at least 40, vi) read position rank sum (RPRS) score greater than -3, vii) mapping quality rank sum (MQRS) score greater than -6, viii) indels were required to have a maximum Fisher’s strand bias (FS) of 200, ix) variants were screened according to VQSR tranche calculated using the known SNV sites from HapMap v3.3, dbSNP, and the Omni chip array from the 1000 Genomes Project to “PASS” SNVs were required to achieve a tranche of 99.9% for SNVs in genomes and exomes and 99% for indels in genomes, x) for heterozygous genotypes, the alternate allele ratio was required to be ≥25%. Finally, variants were excluded if they were among a predefined list of known sequencing artifacts or if they were marked by EVS (http://evs.gs.washington.edu/EVS/)[23] or ExAC (http://exac.broadinstitute.org/about)[16] as being problematic variants. Variants were annotated to Ensembl 73[24] using SnpEff[25].
Any exomes with gender discordance between clinically-reported and X:Y coverage ratios were removed, as were contaminated samples according to VerifyBamID[26].
Before running gene-based collapsing analysis, we implemented both sample- and site-level pruning procedures to minimize the systemic bias in data that might lead to spurious association or reduced power to detect real association. The site-pruning procedure (coverage harmonization) is described in the section below. Here, we described the sample-level pruning procedure including removing related individuals and population outliers identified in principal component analysis (PCA).
To identify related individuals, we generated genotype data in PLINK format[27] and then used KING[28] to calculate pairwise kinship coefficients for all case and control subjects. We used the kinship coefficient 0.1 as a cutoff and removed samples introducing relatedness while preferentially retaining cases; we retained samples with a higher overall coverage in the CCDS regions to break ties if applicable. After this step, 492 of the 496 cases and 12,248 of the 12,916 controls were kept for further analysis.
Next we ran PCA using EIGENSTRAT[29] on the 492 cases and 12,248 controls with a LD-pruned (r2 threshold 0.1) list of single-nucleotide polymorphisms (SNPs) extracted from exomic sequencing data. After removing outliers given a sigma threshold (6.0 along the top10 principal components) for 5 iterations, a total of 488 cases and 12,151 controls entered gene-based collapsing analysis (S1 Fig).
For the 488 cases and 12,151 controls entering association analysis, at least 10-fold coverage was achieved for an average of 93.20% in cases and 95.19% in controls of the 33.27 MB of the consensus coding sequence (CCDS release 14) plus 2 base pair (bp) intronic extensions (to accommodate canonical splice site variants). To address the confounding effect introduced by imbalance of coverage between cases and controls, we pruned out sites with uneven coverage in cases and controls using our previously described site-pruning procedure[30]. Specifically, for each site in CCDS plus 2 bp extensions, we determined the percentages of cases and controls that had at least 10-fold coverage, and that site was excluded from further analysis if the percentages differed by >11.97% between cases and controls. This site-pruning procedure removed 8.58% of the CCDS (+2bp intronic extensions) bases from the analysis. After site pruning, at least 10-fold coverage was achieved for an average of 88.27% in cases and 88.12% in controls of the 33.27 MB CCDS (+2bp intronic extensions) bases. These sites entered the association analysis where case and control populations had a comparable coverage to accurately compare patterns of variation gene by gene.
To identify genes associated with EE under the case-control association analysis framework, we performed a genome-wide search for an enrichment of “qualifying variants” in protein-coding genes in cases compared to controls looking for risk alleles. A “qualifying variant” was determined by a set of criteria, based on allele frequency and functional predictions, designed to capture the characteristics of pathogenic variants associated with EE. Specifically, in this study, we focused on “ultra-rare”, highly impactful variants, and a variant was determined to be qualifying if it: 1) was absent in the Exome Variant Server (EVS) and Exome Aggregate Consortium (ExAC release 0.3); 2) had ≤4 copies of variant allele in the 488 cases plus 12,151 controls; and 3) was predicted to be loss-of-function (stop_gained, frame_shift, splice_site_acceptor, splice_site_donor, start_lost, or exon_deleted) or missense “probably damaging” by PolyPhen-2 (HumDiv). We focused on this subset in an effort to try to capture the de novo variant signal that has been previously reported to play a role in a range of epilepsies and in particular EE subtypes [2, 31, 32]. For each gene, an indicator variable (1/0 states) was assigned to each individual based on the presence of at least one qualifying variant in the gene (state 1) or no qualifying variant in that gene (state 0); this was equivalent to a dominant genetic model. Accordingly, for a given gene, a qualifying case (or control) was defined to be a case (or control) subject carrying at least one qualifying variant in that gene. We used two-tailed Fisher’s exact test to evaluate statistical significance of genic association. To address the potential confounding effect of background rate of “qualifying variants,” we further constructed a logistic regression model including the total number of “ultra-rare” (absent in EVS and ExAC and having ≤4 copies of variant allele in the 488 cases plus 12,151 controls) synonymous variants per individual as covariate. To account for bias due to small counts of qualifying variant, we employed a Firth correction with profile likelihood based tests [33, 34]. With 18,668 CCDS genes we aimed to test, we adopted the genome-wide significance level of p = 2.68×10−6 using Bonferroni correction (0.05/18,668).
Quantile-quantile plots were generated using a permutation-based expected probabilities distribution. To achieve this, for each model (matrix) we randomly permuted the case and control labels of the original configuration: 488 cases and 12,151 controls and then recomputed the Fisher’s Exact test for all genes. This was repeated 1,000 times. For each of the 1,000 permutations we ordered the p-values and then took the mean of each rank-ordered estimate across the 1,000 permutations, i.e., the average 1st order statistic, the average 2nd order statistic, etc. These then represent the empirical estimates of the expected ordered p-values (expected -log10(p-values)). This empirical-based expected p-value distribution no longer depends on an assumption that the p-values are uniformly distributed under the null. For comparison we have provide QQ plots for the actual p-values (S2 Fig) and empirically-based expected p-value distribution (S3 Fig).
To compute the permutation-based expected p-value distribution for Firth logistic regression, due to the presence of the covariate (the total number of “ultra-rare” synonymous variants per individual), we implemented permutation using the R package “BiasedUrn” (https://cran.r-project.org/web/packages/BiasedUrn/) to maintain the confounding role of covariate in each permuted data set while the association between genotype and disease was broken[35]. Permutation was performed 1,000 times and the empirical-based expected p-value distribution was calculated in the same way as described above. For comparison to the BiasedUrn permuted p-values, we have provided QQ plots for the actual p-values generated from the Firth logistic regression (S4 Fig).
|
10.1371/journal.pntd.0006006 | Impaired anti-fibrotic effect of bone marrow-derived mesenchymal stem cell in a mouse model of pulmonary paracoccidioidomycosis | Bone marrow-derived mesenchymal stem cells (BMMSCs) have been consider as a promising therapy in fibrotic diseases. Experimental models suggest that BMMSCs may be used as an alternative therapy to treat chemical- or physical-induced pulmonary fibrosis. We investigated the anti-fibrotic potential of BMMSCs in an experimental model of lung fibrosis by infection with Paracoccidioides brasiliensis. BMMSCs were isolated and purified from BALB/c mice using standardized methods. BALB/c male mice were inoculated by intranasal infection of 1.5x106 P. brasiliensis yeasts. Then, 1x106 BMMSCs were administered intra venous at 8th week post-infection (p.i.). An additional group of mice was treated with itraconazole (ITC) two weeks before BMMSCs administration. Animals were sacrificed at 12th week p.i. Histopathological examination, fibrocytes counts, soluble collagen and fibrosis-related genes expression in lungs were evaluated. Additionally, human fibroblasts were treated with homogenized lung supernatants (HLS) to determine induction of collagen expression. Histological analysis showed an increase of granulomatous inflammatory areas in BMMSCs-treated mice. A significant increase of fibrocytes count, soluble collagen and collagen-3α1, TGF-β3, MMP-8 and MMP-15 genes expression were also observed in those mice. Interestingly, when combined therapy BMMSCs/ITC was used there is a decrease of TIMP-1 and MMP-13 gene expression in infected mice. Finally, human fibroblasts stimulated with HLS from infected and BMMSCs-transplanted mice showed a higher expression of collagen I. In conclusion, our findings indicate that late infusion of BMMSCs into mice infected with P. brasiliensis does not have any anti-fibrotic effect; possibly because their interaction with the fungus promotes collagen expression and tissue remodeling.
| This is the first study that evaluates the effect of BMMSCs therapy for lung fibrosis induced by the fungal pathogen Paracoccidioides brasiliensis, the causative agent of paracoccidioidomycosis, one of the most important systemic endemic mycosis diagnosed in South America and Central America. Our findings showed an impaired anti-fibrotic effect of BMMSCs transplantation. This effect could be triggered by either the chronic inflammatory microenvironment induced by P. brasiliensis or by a direct interaction between BMMSCs and the fungus, resulting in an exacerbation of the pulmonary fibrosis. In fact, the pro-fibrotic effect exerted by BMMSCs was toned-down by the usage of the antifungal ITC.
| Bone marrow-derived mesenchymal stem cells (BMMSCs) are adult stem cells capable of both renew themselves and differentiate in vitro into multiple cell lineages [1]. These cells can also modulate the inflammatory response and induce tissue regeneration through release of cytokines, chemokines, growth factors and genetic material (e.g. miRNAs) [2–6]. Likewise, their microbicidal properties have been already described [7–10]. Cell-based therapies in regenerative medicine using syngeneic or autologous BMMSCs are considered a promising approach, because they do not induce tissue rejection and exert a localized effect than the systemic classical pharmacological strategies [11].
BMMSCs have also been evaluated in animal models of acute lung injury induced by chemicals, such as bleomycin [12, 13] and hydrochloric acid (HCl) [14]. These studies have shown that BMMSCs secrete cytokines, chemokines, growth factors and extracellular matrix proteins, and that can influence the magnitude and quality of the immune response (e.g. modulating the inflammatory response), and promote tissue repair. Likewise, BMMSCs can differentiate into pulmonary stromal cells (e.g. lung fibroblasts and myofibroblasts) [12, 14, 15].
Pulmonary fibrosis (PF) is a process characterized by excessive deposition of collagen and extracellular matrix components that results in a pathological remodeling of the pulmonary architecture. Thus, patients with PF exhibit radiographic, but also functional and clinical alterations in the lung [16]. From the pathological perspective, PF is a dynamic process involving immune system cells and soluble factors including leukotrienes, cytokines (IFNγ, TNFα, IL1, IL4, IL6, IL17), chemokines (CCL2, CCL3, CXCL12), reactive oxygen species (ROS), growth factors [platelet-derived growth factor (PDGF), vascular endothelial growth factor (VEGF), insulin-like growth factor (IGF)], and membrane-bounded and soluble molecules such as prostaglandins, metalloproteinases (and their tissue inhibitors), among others [17]. An imbalance between pro-fibrotic responses and anti-inflammatory and pro-tissue repair agents, results in the differentiation and activation of myofibroblasts which once activated produce abundant amounts of collagen, thus inducing fibrosis of the pulmonary parenchyma [18]. The participation of all the previous components in the PF has been extensively studied in animal models [16, 18, 19].
Pulmonary fibrosis can be induced by microbial agents including the dimorphic fungal pathogen Paracoccidioides brasiliensis, the causative agent of paracoccidioidomycosis (PCM), disease that is considered one of the most important endemic systemic mycosis in South America and Central America [20–23]. Brazil, Colombia and Venezuela are the countries with the highest number of cases reported so far, with an estimated of 10 million people infected [20, 21]. The chronic form of PCM is the most frequent clinical presentation (90% of the cases), and it is characterized by a granulomatous inflammatory response with fibrosis development and loss of respiratory function, which is observed in 60% of the patients [22]. Itraconazole (ITC) is the treatment of choice in PCM [23]. Nonetheless, it exerts a fungistatic effect against P. brasiliensis in vivo, and it does not attenuate the pulmonary alterations induced by the fungal infection, including fibrosis [24, 25]. Animal models of PCM have allowed characterization of the mechanisms involved in the development of pulmonary fibrosis, and evaluate diverse strategies to treat it. Thus, combined therapies including pentoxifylline plus ITC [26], or an anti- neutrophil monoclonal antibody alone or in combination with ITC [27, 28], have showed that such treatment strategies reduced substantially the PF. However, the potential risk for using immunosuppressive drugs or biological agents, mainly in host with other unknown latent infections, should be considered [29].
The objective of this study was to investigate the regenerative effect of BMMSC on PF induced by the fungal pathogen P. brasiliensis in an in vivo experimental model of PCM; nonetheless, our findings indicated that these cells impaired the anti-fibrotic effect and by the contrary, an exacerbated fibrotic process was observed.
BMMSCs were obtained from four weeks-old BALB/c mice from the breeding colony maintained at the Corporación para Investigaciones Biológicas (CIB, Medellín-Colombia). The BMMSCs isolation and purification protocols were adapted from a protocol previously described by Rojas et al [12]. Briefly, mice were anesthetized with a solution of ketamine (80 mg/kg) and Xylazine (8 mg/kg) via intramuscular. Femurs and tibias were removed and bone marrow cells were isolated by flushing with Dulbecco's Modified Eagle Medium (DMEM)-low glucose (GIBCO, Invitrogen Corporation, Carlsbad, CA, USA) containing penicillin/streptomycin1% (vol/vol) (GIBCO). Cells were transferred to cell culture flasks (Eppendorf, Hamburg, Germany) with DMEM-low glucose supplemented with 10% (vol/vol) fetal bovine serum (FBS) (GIBCO) and nonessential amino acids 1% (vol/vol) (GIBCO), followed by incubation at 37°C in 5% CO2. Non-adherent cells were removed after 48 hours, and maintained in standard culture media for 7 days.
In order to exclude hematopoietic stem cells and leucocytes, a magnetic bead-based mouse cell depletion kit (Miltenyi Biotec, Bergisch Gladbach, Germany) containing anti-CD45, anti-CD11b, anti-CD5, anti-Gr1 (Ly-6/C), and anti-Ter 119 monoclonal antibodies was used. The BMMSC surface markers expression profile was determined by flow cytometry. The following antibodies were used: isothiocyanate (FITC) anti-CD45 (BD Pharmingen, San Diego, CA, USA), phycoerythrin (PE)-Cy5-anti-CD44, allophycocyanin (APC) anti-CD105, PE-Cy7-anti-CD106, APC-anti-TER-119, Pacific blue-anti-SCA-1, and PE-anti-CD73 (Biolegend, San Diego, CA, USA). Cells were analyzed using a FACS Canto II system (BD Biosciences, San Jose, CA, USA) and FlowJo V10 software (FlowJo, LLC, Data Analysis software, Ashland, OR, USA). In addition, a differentiation assay to demonstrated the BMMSCs plasticity (differentiation to chondrogenic, adipogenic and osteogenic lineages) was performed using a differentiation commercial kit, and following the manufacturer’s instructions [StemPro (Waltham, MA, USA)]. Finally, the purified cells were kept in standard culture media until the day of transplant.
This study was carried out following the Colombian (Law 84/1989, Resolution No. 8430/1993), European Union, and Canadian Council on Animal Care regulations. The protocol was approved by the Institutional Ethics Committee of the CIB (Acta No.95).
A highly virulent strain of P. brasiliensis (Pb18) was used in order to develop the experimental pulmonary fibrosis model as described previously [28]. Briefly, BALB/c male mice (8 weeks old) were intranasally infected with 1.5 x 106 P. brasiliensis yeast cells contained in 60 μl of phosphate-buffered saline (PBS). The total inoculum was split into two equal doses, which were instilled within a 5–10 minutes (min) period. Non-infected (control) mice were inoculated with 60 μl of PBS.
Infected and non-infected mice were intravenously injected with 1x106 BMMSCs at 8th week post-challenged given in a single dose. Six week post-inoculation, an additional group of infected animals was treated with 100 μl of Itraconazole (ITC) oral solution (Sporanox, Janssen-Cilag S.A., Mexico) administered at a dose of 1mg/day in order to achieve serum levels equal to 1 μg/mL. The above treatment was administrated daily and uninterruptedly for 6 weeks by gavage. All animals included in the various experimental groups were sacrificed at week 12th p.i. and their lungs harvest for further studies.
Lungs of mice were removed, homogenized and sequentially filtered through 70 and 40μm sterile cell strainers (Thermo Fisher Scientific Inc, Waltham, MA, USA) in RPMI cell culture medium plus 1% (vol/vol) FBS (Sigma-Aldrich, Saint Louis, MO, USA). Cells suspension were centrifuged at 500 G, 10°C for 10 min, and red blood cells were lysed using the ACK Lysing Buffer (GIBCO). Viability of the cells was determined by trypan blue exclusion test with samples being used if they were 95% of viable. Cells were resuspended in RPMI plus 10% FBS and counted using a hemocytometer. Fc receptors were blocked using a purified rat anti-mouse CD16/CD32 (BD Pharmigen, San Diego, CA, USA). Then, cells were treated with Cytofix/Cytoperm and Perm/Wash solution (BD Pharmigen, San Diego, CA, USA) [28]. Fibrocytes were determined using FITC anti-collagen I (Rocklad inc Limerick USA), PE anti-CD45 (Biolegend San Diego USA), and APC anti-CD34 (BD Pharmingen, San Diego USA). Anti-mouse IgG-FITC (Rocklad), anti-mouse IgG2aκ-PE (Biolegend) and anti-mouse IgG1κ-APC (BD) were used as isotype controls. The stained cell suspensions were fixed with FACS buffer/1% (vol/vol) PFA (Carlo Erba, Barcelona, Spain). Assays were performed using a FACS Canto II system (BD Biosciences, San Jose, CA, USA), while information analysis were done using FlowJo V10 (FlowJo, LLC, Data Analysis software, Ashland, OR, USA). Fibrocyte population was analyzed as follows: (a) cell events in region 1 (R1) were gated by forward scatter versus side scatter areas; (b) CD45+ events were gated from R1 by side scatter area versus CD45 staining to establish the R2 region, from which (c) cell events were gated to determine fibrocytes by collagen 1+ (intracellular) and CD34+ (surface). The number of fibrocytes was determined by multiplying the percentage of the gated population by the total number of leukocytes (CD45+ population).
Lungs were processed and analyzed as described by Puerta-Arias et al [28]. Briefly, lungs were perfused with 1X PBS to wash out red blood cells. Tissue fixation was completed in a 4% buffered formalin solution. Then, fixed tissues were embedded in paraffin and sections stained with Masson trichrome, and examined using a Nikon Eclipse Ci-L microscope—Nikon DS-Fi2 digital camera. A morphometric analysis was performed using NIS Elements 4.30.02 Laboratory Image Software (Nikon Instruments Inc., Melville, USA). The percentage of occupied area by the inflammatory response was calculated by dividing the total inflamed area, which includes cellular infiltrates and granulomatous lesions by the total area of the lung.
Homogenized lung suspensions were treated with acid neutralizing reagent (0.5M acetic acid, 0.1 mg/ml pepsin) (Sigma-Aldrich, Saint Louis, MO, USA). Then, colorimetric detection of soluble collagen content was performed according to the manufacturer's protocol of a sircol collagen assay kit (Biocolor, Northern Ireland, U.K.). A calibration curve was constructed using bovine collagen-I in the range of 1–10 μg.
Human lung fibroblasts were obtained from Rojas’ Lab repository, collected under an established protocol from the University of Pittsburgh Center for Organ Research Involving Decedents (CORID). Cultures of human fibroblasts (2x104 cells/200uL, pass 4) were treated with soluble lungs supernatants (protein concentration 10ug/mL) from all experimental groups, for 24h at 37°C. Then, fibroblast activity was determined by measuring the expression of collagen type-I gen using reverse transcriptase real-time-PCR (RT-qPCR) assays, as previously described [29]. As controls, we used PBS and TGF-β [(5ng/ml final concentration) Peprotech Rocky Hill, United States].
All real time PCR assays were performed as previously described [28]. Briefly, RNA was obtained from lungs of mice using Trizol (Invitrogen, Carlsbad, CA, USA). Samples were treated with DNase I (Thermo Fisher Scientific Inc, Waltham, MA, USA), and cDNA was synthesized using 500ng of total RNA using cDNA synthesis kit for RT-qPCR according to the manufacturer’s instructions (Thermo Fisher Scientific Inc, Waltham, MA, USA). Real-time PCR was done using Maxima EVAGreen/Fluorescein qPCR Master Mix according to the manufacturer’s instructions (Applied Biological Materials ABM Inc, Richmond, Canada). The CFX96 Real-Time PCR Detection System (Bio-Rad, Headquarters Hercules, California, USA) was employed to measure gene expression levels. Melting curve analysis was performed after the amplification phase of real time PCR assays to eliminate the possibility of non-specific amplification or primer-dimer formation. Validation of housekeeping genes for normalization mRNA expression was performed before gene expression analysis. Expression of fibrosis-related genes encoding for collagen, transforming growth factor beta (TGF-β), matrix metalloproteinases (MMP) and tissue inhibitor of metalloproteinases (TIMP) were evaluated. Fold changes in the target gene mRNA expression were quantified relative to glycer-aldehyde-3-phosphate dehydrogenase (GAPDH the housekeeping gene previously defined) [28]. Each experiment was repeated twice using 5 mice per each one of the groups with gene expression analysis being conducted by triplicate.
Data analysis was performed using Graph Pad Prism software version 7 (GraphPad Software, Inc., La Jolla, CA, USA). Normality for all values was calculated by the Shapiro-Wilk test and when comparisons between three or more groups were required, the ANOVA test was employed. On the other hand, comparisons between two specific groups were determined by Student-t test. Mean and standard error of the mean (SEM) were calculated for all analyses. We considerate P<0.05 values as significant.
We determined the granulomatous inflammatory areas through histopathological analysis in lung of experimental mice. We observed that lungs of P. brasiliensis infected-mice developed a granulomatous inflammatory response with collagen fibers surrounding granulomas (Fig 1A). Interestingly, administration of BMMSCs in infected mice showed an exacerbation of the inflammatory process, with a higher granulomatous inflammation and fibrosis areas with loss of parenchyma (Fig 1B). In contrast, the infected animals treated with ITC alone showed inflammatory and fibrotic responses similar to those infected and non-transplanted mice (Fig 1C). Moreover, the combined administration of BMMSCs/ITC in P. brasiliensis-infected mice considerably decreased the inflammatory response and fibrosis (Fig 1D), in comparison with those infected animals that only received cell-based therapy. A morphometric analysis revealed that occupied area by granulomatous inflammation in infected and transplanted mice was twice higher when compared with infected non-treated mice (p<0.001), or three time that infected and BMMSCs/ITC-treated animals (p<0.001) (Fig 1E). There was statistically significant difference in the average of occupied area by granulomas between infected and ITC-treated mice and those that received combined therapy (p<0.005).
Fibrocytes are bone marrow-derived fibroblast progenitor cells that have been implicated in tissue remodeling or repairing process, including the development of fibrosis. Following flow cytometry analysis we found a significantly increased number of fibrocytes (CD45+/CD34+/Collagen1+) in lungs from infected mice (p<0.005) (Fig 2) relative to PBS instilled controls. Moreover, infected and BMMSCs-treated animals showed almost twice the number of fibrocytes when compared with infected non-treated mice (p<0.005) (Fig 2). Interestingly, ITC treatment, in combination with BMMSCs, reduced the fibrocytes counts, versus P. brasiliensis infected mice (p<0.001) or infected BMMSCs-treated animals (p<0.001) (Fig 2).
Collagen is considered the most important extracellular matrix protein involved in fibrosis. Accordingly, we determined the effect of BMMSCs therapy on lung soluble collagen content. Significantly increased levels of collagen in lungs from mice infected with P. brasiliensis were observed when compared with PBS instilled animals (p<0.005) (Fig 3). Infected and BMMSCs-treated mice exhibited an increased significantly in collagen content relative to infected non-treated animals (p<0.001) (Fig 3). Remarkably, ITC treatment reduced the amount of soluble collagen in the lungs from both, P. brasiliensis infected-mice (p<0.005), or those with BMMSCs transplantation (p<0.005) (Fig 3).
To assessment the capability of lung homogenized to activate human lung fibroblasts, we performed in vitro assays stimulating fibroblasts with lung supernatants from experimental animals. We observed that lung supernatants from infected and BMMSCs-treated mice induced a higher expression of the gene encoding for collagen type-I in human fibroblast, in comparison with the respective homogenized lung supernatants from the infected non-treated mice (p<0.005) (Fig 4). The homogenized lung supernatants from mice infected with P. brasiliensis and treated with ITC, alone or in combination with BMMSCs transplantation, in human fibroblast also induced a collagen gene expression similar to that found in human fibroblast stimulated with lung homogenates from infected non-treated-mice (Fig 4).
Our next step was to determine if the BMMSCs administration affects the related-fibrotic response genes expression in lungs from mice infected with P. brasiliensis. We observed that P. brasiliensis infected-mice showed a higher expression of almost all genes evaluated (Col1α3, Col3α1, TGF-β1, TGF-β3, MMP-8, MMP-12, MMP-13, MMP14, TIMP-1 and TIMP-2) when compared with uninfected mice. Moreover, after BMMSCs treatment, a significantly higher expression of Col3α1, TGF-β3, MMP-8, and MMP-15 was observed in comparison with those P. brasiliensis infected-mice, while a reduction on MMP-13 gene expression was also observed (Fig 5). Infected mice that received the ITC treatment showed a slight but higher expression of the MMP-15 and TIMP-2 genes relative to infected non-treated animals. Remarkably, the combined therapy BMMSCs/ITC induced a synergistic reduction of Col3α1, TGF-β-3, MMP-8, MMP-12, and TIMP-1, as well as an increase of TIMP-2 gene expression, when compared to infected mice that received cell transplantation (Fig 5).
Pulmonary fibrosis (PF) is a serious disease triggered by chemical, physical, or infectious agents, but also it could be an idiopathic or cryptogenic process [16, 18]. The fungus P. brasiliensis is the etiological agent of paracoccidioidomycosis (PCM), a disease endemic in Latin America. In the chronic form of the disease, more than 60% of patients develop fibrotic sequelae compromising lung parenchyma, even after the completion of treatment [22]. In fact, the current therapeutic strategy for PCM is based on azole compounds as itraconazole (ITC), an antimycotic that reduces fungal load but not PF development [20]. Therefore, in recent years, we had been focused to evaluate alternative therapies in an attempt to reduce the fibrotic response in this disease [22, 28]. In the present study, we investigated for the first time the effect of a BMMSCs-based cellular therapy on PF in a murine model of PCM. Nonetheless, contrary to other reports showing a beneficial effect of BMMSCs transplants in PF due to chemical or physical agents [30–32], we observed that this type of cellular therapy exacerbated the fibrotic response.
BMMSCs are considered as promising in the development of cellular therapies due to their capacity to regenerate tissue, as well as for their immunomodulatory properties, which include the release of paracrine or endocrine signaling molecules [5, 32, 33]. Moreover, different studies have shown that BMMSCs are able to sense the microenvironment and respond to both physical (i.e. mechanical) and chemical stimulus [34, 35]. Namely, it has been recognized that extracellular matrix (ECM) influences stem cell lineage commitment [35]. As an example, Li et al. [34, 35] described that fibronectin, a glycoprotein that connects integrins in cell surface with collagen fibers in the ECM, playing a role as mechanotransducer signal that regulate human mesenchymal stem cells (hMSC) differentiation [34]. Our group have previously shown that lungs of mice infected with P. brasiliensis exhibit an increased expression and re-arrangement of ECM components (e.g. collagen, fibronectin, laminin, proteoglycans), even after two days post-infection, but fully established after 4th week post-infection [36, 37]. Accordingly, that early deposition could favors the BMMSCs differentiation to fibrocytes, who in turn can proliferate and undergo phenotypic conversion to fibroblast or myofibroblast, the cellular culprits of fibrosis [38]. In fact, after BMMSCs administration, we observed a rise of fibrocytes counts and soluble collagen content in lungs from infected-mice.
In addition to their ability to interact with ECM proteins, BMMSCs may also recognize microbial compounds through pattern recognition receptors (PRRs), interactions that may induce a pro-inflammatory response [35, 38]. Bernardo et al [39], just as Waterman et al [40], have demonstrated that the activation of Toll-like receptors (TLRs) in mesenchymal stem cells promotes their polarization into a pro-inflammatory phenotype, named MSC1, which can fuel inflammation and subsequent fibrosis [39, 40]. In this sense, it has been reported that the interaction between P. brasiliensis and TLR4 lead to a severe fungal infection, associated with an enhanced exacerbated proinflammatory response [41]. All these reports support both the fungal proliferation and tissue damage observed after BMMSCs administration, and suggest an immunoregulatory role of these stem cells; thus, the deleterious effect observed maybe triggered by the interaction of BMMSCs with either P. brasiliensis compounds, extracellular matrix and the inflammatory microenvironment developed during the chronic course of PCM. However, more studies are needed to evaluate the interaction between BMMCSs and this fungal pathogen, as well as its implications not only for the immune response, but also for tissue repair.
Besides that, it is worth to highlight that myofibroblasts might be derived from other cellular sources beyond bone marrow fibrocytes. In fact, these collagen-producing cells, and main effectors of fibrosis, may arise from resident fibroblast, or as a result of epithelial/endothelia mesenchymal transitions [42]. In this study, the administration of BMMSCs was associated with an increase in the number of fibrocytes. Although, the origin of these cells could not be confirmed, we may suppose that they could come from bone marrow or pericytes, which subsequently differentiate into fibroblasts and then into myofibroblasts collagen-producer cells, thus contributing to the increase of PF in those P. brasiliensis infected- and BMMSCs treated-mice. Accordingly, we evaluated the collagen I gene expression in human fibroblast stimulated with homogenized lung supernatants from P. brasiliensis infected mice. We found that supernatants from infected and BMMSCs-transplanted mice induced a higher production of type I collagen transcript in human fibroblast, in comparison with those cells stimulated with infected non-treated animals. These results clearly indicate the presence of molecules from the lung microenvironment able to stimulate the collagen production in human fibroblast. Among the major stimuli to activate fibroblasts, IL6, TGFβ, IL13 and FGF are found, and once activated, these cells differentiate into myofibroblasts or could produce pro-fibrotic molecules such as IL1, VEGF, Insulin-like growth factor 2 (IGFII), Insulin-like growth factor-binding protein (IGFBP), IL6 and IL33 [42]. In fact, we observed an increased expression of Col 3α1 and TGFβ-3 genes in lungs of P. brasiliensis–infected mice treated with BMMSCs. In concurrence with these reports, more recently we have observed a significant increase of IL-1α, IL-1β, IL-6, TNF-α and IL17 levels in homogenized lung supernatants from mice infected with P. brasiliensis [28]. All these cytokines have also been implicated in the pathogenesis of PF. Nonetheless, a direct activation of fibroblast by P. brasiliensis compounds could not be ruled out.
Matrix metalloproteinases (MMPs) are a family of zinc- and calcium-dependent endopeptidases (around 25 members are know so far) that are either secreted or membrane-bound enzymes [43]. MMPs have long been considered to be essentials for ECM remodeling, which is critical in embryonic development and tissue homeostasis, including inflammatory response and tissue repair [43]. In this context, as stated by Pardo et al, MMPs not only degrade ECM components, but also release, cleave and active a wide range of growth factors, cytokines, chemokines and cell surface receptors affecting numerous cell functions (e.g. adhesion, proliferation, differentiation, migration, cell death) [44]. Thus, MMPs and their tissue inhibitors (TIMPs) play a central role in the extracellular pathways of ECM degradation and, therefore, in fibrosis development or resolution [44]. Namely, MMP8, a collagenase, can also cleave the chemokines CXCL8 and LIX, resulting in enhanced chemoattractant activities, which could be associated with fibrosis development [43, 44]. These results show that after BMMSC administration there was not only an increase in the expression MMP8 gene but also elevated neutrophils counts, findings noticed before in association with development of fibrosis as observed in our PCM model previously reported [28]. BMMSCs transplantation also induced a decrease in gene coding for MMP-13, another collagenase, but not changes in the expression of TIMP-1 and TIMP2 genes were observed. Meanwhile, the ITC/BMMSC transplantation therapy decreased synergistically the expression of TIMP-1 and MMP-13. Over expression of TIMP-1 is associated whit liver fibrosis [45], while MMP-13 cleaves and inactivates CCL2, CCL7 and CXCL12 leading to reduction in chemotaxis, as well as to a decrease in the fibrosis process [43, 45]. However, our data interpretation relative to MMPs or TIMPs gene expression is limited, as the current knowledge concerning pathological tissue repair in PMC is scarce. In addition, although most of the fibrosis-related genes analyzed showed small fold-changes with statistical significant, a possible meaningful biological effect cannot be ruled-out.
ITC is the antifungal treatment of choice in PCM. Of note, additionally to its antifungal effect, it has been recently documented that this antifungal medication also exhibits immunomodulatory properties [46]. Moreover, in a previous work, we found that ITC reduces the expression of certain genes encoding for pro-inflammatory cytokines (IFN-γ, IL-6, IL-17, TGF-β1, TNF-α), transcriptional factors (T-bet, GATA-3, Spi-1, RoRc, Ahr, FoxP3), and fibrosis development (MMP-1, MMP-8, MMP-13, and Col3α1). Additionally, it also diminished the number of inflammatory cells—including neutrophils—in the lungs of mice infected with P. brasiliensis [27]. In the present study, it was observed that the ITC regulates the expression of the MMP-15 and TIMP-2 genes that have been recognized as inducers of pulmonary fibrosis [27].
Overall, our results demonstrated an exacerbating effect of the BMMSCs therapy on pulmonary fibrosis induced by P. brasiliensis infection. We hypothesized that this outcome could be triggered by either the interaction with P. brasiliensis compounds or by the inflammatory microenvironment induced by this process. Nonetheless, the combined therapy ITC/BMMSCs showed promising results since synergistically it reduced TIMP-1 and MMP-13. Thus, the use of BMMSC under different conditions or combined with other treatments (e.g. ITC) opens the possibility to new therapeutic approaches for this type of fibrosis resulting from an infectious disease.
PCM is considered a neglected tropical disease mostly affecting low income individuals who live in underdeveloped Latin American rural regions where the technology and the resources needed to administer the immunotherapeutic measures here suggested would probably not be available. Nonetheless, the implementation of cellular therapies is progressing and the prospects are to arrive in a few years to the administration of autologous bone marrow or stem cells obtained from adipose tissues even in these regions.
|
10.1371/journal.pcbi.1000854 | Intrinsically Disordered Regions May Lower the Hydration Free Energy in Proteins: A Case Study of Nudix Hydrolase in the Bacterium Deinococcus radiodurans | The proteome of the radiation- and desiccation-resistant bacterium D. radiodurans features a group of proteins that contain significant intrinsically disordered regions that are not present in non-extremophile homologues. Interestingly, this group includes a number of housekeeping and repair proteins such as DNA polymerase III, nudix hydrolase and rotamase. Here, we focus on a member of the nudix hydrolase family from D. radiodurans possessing low-complexity N- and C-terminal tails, which exhibit sequence signatures of intrinsic disorder and have unknown function. The enzyme catalyzes the hydrolysis of oxidatively damaged and mutagenic nucleotides, and it is thought to play an important role in D. radiodurans during the recovery phase after exposure to ionizing radiation or desiccation. We use molecular dynamics simulations to study the dynamics of the protein, and study its hydration free energy using the GB/SA formalism. We show that the presence of disordered tails significantly decreases the hydration free energy of the whole protein. We hypothesize that the tails increase the chances of the protein to be located in the remaining water patches in the desiccated cell, where it is protected from the desiccation effects and can function normally. We extrapolate this to other intrinsically disordered regions in proteins, and propose a novel function for them: intrinsically disordered regions increase the “surface-properties” of the folded domains they are attached to, making them on the whole more hydrophilic and potentially influencing, in this way, their localization and cellular activity.
| Intrinsically disordered proteins and protein segments carry out a wide range of important biological functions despite their lack of permanent tertiary structure. Using advanced computational methods we study the biophysical properties of the intrinsically disordered regions in the enzyme nudix hydrolase from the desiccation- and radiation-resistant bacterium D. radiodurans. Interestingly, these regions are absent in homologue proteins in non-extremophile bacteria, suggesting that they might be involved in helping the key rescue-and-repair proteins in D. radiodurans, such as nudix hydrolase, adapt to the extreme absence of water. We show that the disordered regions in nudix hydrolase enlarge the overall surface of the enzyme, and most importantly, increase its overall affinity for water (i.e. its hydrophilicity). We suggest a novel hypothesis that this, indeed, may be the principal function of disordered regions in some cases: they increase the chances of the protein to be located in the remaining water patches in the desiccated cell, where it is protected from the desiccation effects and can function normally.
| The dominant paradigm for describing the functioning of proteins is that of well-defined, structured molecular machines undergoing concerted, conformational changes while carrying out their function [1]–[3]. However, over the past few years it has become clear that the reality is much more complex, and that there are proteins that simply do not have a defined tertiary structure, and yet still carry out multitudes of different important functions. These intrinsically disordered proteins and protein segments (IDPs), also called “natively unfolded” or “natively disordered”, are by definition difficult to study using classical methods of structural biology, but they have in recent years received significant attention, largely due to two facts [4]–[14]. First, it has become clear that these proteins are extremely abundant. Using mostly bioinformatics approaches, it has been shown that in eukaryotes about 30% of all proteins are largely intrinsically disordered (ID) and that about 50% have long ID stretches [15]. Among signaling proteins, in particular, about 70% have long disordered segments [16]. These numbers are large: even if some of the predicted disordered regions actually prove to be structured, it is likely that IDPs in eukaryotes by far exceed, for example, the entire population of membrane proteins.
The second motivating factor is the fact that IDPs are involved in a host of extremely important cellular functions such as molecular recognition, assembly, protein modification and entropic chain functions [6]–[8], [11], [14], [17]. For example, in complex cellular activities such as cell signaling or regulation, it is often required that actions of many key molecular players be tightly controlled and coordinated through interaction and recognition based on unique identifying features [8], [18], which, in the case of signaling proteins, are often located in ID regions. When recognition of multiple binding partners and high-specificity/low-affinity binding is required, the choice on the molecular level often involves IDPs or ID regions [6], [7], [19]. Finally, large numbers of IDPs are known to be involved in human diseases such as cancer, neurodegenerative diseases, diabetes, cardiovascular diseases and amyloidoses [14].
Thus far, some of the biggest advances in the study of IDPs have been accomplished through bioinformatics and data mining approaches [11], [12], [20]–[24]. It has been shown that amino-acid sequences of IDPs tend to exhibit high hydrophilicity, low sequence entropy, and lack of the so-called order-promoting amino acids and bulky hydrophobic residues. Based on such features, several bioinformatics tools have been developed that can reliably predict the level of intrinsic disorder in a given sequence [11], [20]. Moreover, using mostly NMR, small-angle X-ray scattering, and different spectroscopic methods [5], [7], [25], basic structural features of IDPs have been elucidated, including their molecular size, level of structural heterogeneity, role of transient structure in coupled binding-and-folding events, aggregation tendencies and presence of persistent structure [6], [7], [25]. In particular, a combination of computer simulation and experimentally-determined restraints has provided some of the first ensemble-level pictures of IDPs [13], [25], [26]. However, compared with our knowledge of the structure and mechanism of ordered proteins, our understanding of IDPs is still extremely rudimentary and incomplete, primarily because of their structural and dynamic complexity. In the present study, we use molecular dynamics simulations and free energy calculations to focus on the role of IDPs in an extremophile bacterium: Deinococcus radiodurans.
D. radiodurans is a non-motile, non-spore-forming bacterium that belongs to the Deinococcaceae family [27], [28]. It is characterized by an extreme ability to withstand high doses of desiccation and ionizing radiation. For example, this bacterium can survive a dose of 5000 Gy of ionizing radiation, inducing more than 200 DNA double-strand breaks, with no effect on its viability [27], [28]. However, the molecular mechanisms underlying the high radiation resistance of D. radiodurans are thought to have evolved primarily as a side effect of mechanisms to counter extreme desiccation, as the bacterium thrives in dry, arid environments [27]. Over the years, evidence has accumulated suggesting that there exists no single, dominant mechanism responsible for the extremophilic nature of D. radiodurans, but that rather a combination of different mechanisms is at play [27], [28]. These range from passive structural contributions, such as the increased genome copy number [29], compact nucleotide organization [30], high intracellular concentration of the ROS-scavenger manganese [31], to active enzymatic repair mechanisms, including nucleotide and base excision repair and DNA double-strand break repair [27], [28], [32].
However important these mechanisms may be, it is also possible that the bacterium's proteome has undergone major structural adaptations in order to cope with environmental stresses. One potential strategy to address this possibility is to compare the proteome of D. radiodurans with those of its non-extremophile relatives and look for conspicuous differences. Recently, one of us has taken exactly this approach and focused on the presence and the putative biological role of ID regions in the proteome of D. radiodurans (Krisko et al., 2010, manuscript submitted to Proteins: Structure, Function and Bioinformatics). A subset of proteins in D. radiodurans was identified that contain highly hydrophilic stretches with low sequence complexity, indicative of intrinsic disorder, that are absent in non-extremophile homologues. Interestingly, this list includes a preponderance of housekeeping and rescue-and-repair proteins, including DNA polymerase III, a nudix hydrolase, rotamase, ABC transporters, adenine deaminase and LEA proteins. To further probe the significance of this finding, we here focus on a variant of nudix hydrolase, occuring naturally as a dimer, and analyze the properties of its ID regions. Nudix hydrolases [33] are a large class of enzymes present in all organisms that hydrolyze a wide range of pyrophosphates, including nucleotide di- and triphosphates, dinucleotides and nucleotide sugars. Importantly, some members of the family degrade oxidized nucleotides and in this way prevent potentially mutagenic effects these would have if incorporated into nucleic acids. Other family members regulate the concentration of metabolic intermediates and signaling molecules [33]. Interestingly, D. radiodurans exhibits 26 different types of nudix hyrolase, which is about three times more than what would be expected given the size of its genome, and is the highest number per Mbp of any bacterial genome known. In particular, it has been reported that the majority of D. radiodurans nudix genes are strongly induced during the stationary phase, which has been implicated in metabolic reprogramming.
IDPs, such as the tail regions of nudix hydrolase, are extremely dynamic and are very difficult to study by X-ray crystallography or nuclear magnetic resonance (NMR). However, atomistic computer simulations and molecular dynamics (MD) techniques are ideally suited to provide a complete, atomistic picture of the diverse, dynamic ensembles characterizing the IDPs. Here, using sequence analysis methods and molecular dynamics simulations in conjunction with hydration free energy calculations, we show that the ID regions in this protein significantly alter its solvation properties, making the whole protein significantly more hydrophilic. We hypothesize that this, in fact, is the principal functional role of these ID regions, and that they have evolved to keep the key housekeeping and repair enzymes solvated, and consequently functional, under extreme desiccation conditions.
We have analyzed the propensity for intrinsic disorder of the UniProtKB Q9RWW5_DEIRA sequence of nudix hydrolase from D. radiodurans (denoted hereafter as DRNH), using several different algorithms for predicting intrinsic disorder in proteins (Figure 1). Despite their different heuristics for determining disorder, all predictors classify the N-terminal region of DRNH, comprising approximately the first 80 residues, as intrinsically disordered. In particular, both the IUPRED and DisProt predictors associate disorder probabilities of over 90% with all residues below 75, with DisEMBL following suit at a somewhat lower level of significance. This convergence of different intrinsic disorder prediction algorithms is not surprising, given the sequence properties of the N-terminal tail of DRNH. Namely, it features a preponderance of disorder-inducing amino acids, such as glycine and proline, and extremely hydrophilic, polar amino acids, such as arginine and lysine, all of which are strong determinants of intrinsic disorder.
Compared to the sequences of homologue nudix hydrolases in non-extremophile bacteria, the sequence of DRNH studied here possesses a 40-residue C-terminal tail in addition to the N-terminal tail. Unlike the N-terminus, the C-terminus exhibits weaker signatures of intrinsic disorder (Figure 1). While DisEMBL and DisProt assign appreciable levels of disorder to this region (>50%), the IUPRED predictor considers it to be significantly less disordered, except for the very last 10 amino acids or so. Secondary structure prediction of both tails suggests that they are either in disordered or in coil conformation. For example, prediction of α-helical content using AGADIR [34] suggests that neither region is appreciably helical (helical propensity less than 1% for both tails). Similarly, JPRED [35] secondary structure predictor assigns no major secondary structure content to the N-terminus, and only a short putative 8-residue helical stretch to the C-terminal tail (residues 221–228). Based on these predictions, we consider both tails to be largely disordered, with the caveat that the N-terminus is likely significantly more disordered.
There is no high-resolution structure of nudix hydrolase from D. radiodurans. However, there exists an X-ray structure of the core region of a highly homologous (E-value of 8e-05) nudix hydrolase from T. thermophilus, including all residues except the N-terminal 80 and C-terminal 40 residues of the D. radiodurans nudix hydrolase. Using homology modeling (see Methods), we have generated a model of the nudix hydrolase from D. radiodurans in its dimeric form, with the missing regions modeled in a physically realistic, yet extended conformation. This model was then used as a starting structure for simulating a set of 200 independent molecular dynamics (MD) trajectories, each initiated using a different random number seed, and each exploring the phase space independently. Due to their innate flexibility and heterogeneity, the properties of ID regions can only be described on the level of ensembles. Consequently, we have run 200 independent MD trajectories to sample the possible conformations of the nudix tails in low-viscosity implicit solvent to further enhance sampling.
As expected, the extended N- and C-terminal regions of the molecule collapse to a more compact form during the simulations, while the core domain remains largely unchanged. In Figure 2 we show the radius of gyration, Rgyr, and the solvent-accessible surface area of DRNH as a function of time and averaged over all 200 independent trajectories. Rgyr decreases monotonically, leveling out at about 550 ps, from which point on it decreases only slightly by another 2 Å. The standard deviation increases up to 10 Å during the simulation, indicating large variability in the generated trajectories. At 2 ns, the curves still exhibit a marginal downward trend, suggesting that some minor additional collapse may still be possible. Importantly, the collapsed structures of both tails are extremely varied, and share little structural similarity with each other. For example, the ensemble averaged pair-wise backbone root-mean-square deviation for the 200 final structures at 2 ns is 17.1±5.8 Å for the N-terminal 82 residues, and 10.4±2.8 Å for the C-terminal 40 residues, indicating significant structural diversity despite the compactness (Figure 2b). In contrast, the ensemble-average pair-wise backbone root-mean-square deviation for the dimer cores of the 200 final structures is 6.0±1.2 Å. However, if one focuses on individual cores and excludes the 5 relatively dynamic residues at either end of each core and a particularly mobile short loop (residues 147–151 in core 1 and residues 396–400 in core 2), these values drop to 3.7±1.0 Å (core1) and 4.4±0.8 Å (core2). This should be compared to the average backbone RMSD of 5.0±1.2 Å relative to the modeled starting structures of the cores, after the exclusion of the floppy termini and the mobile loop (see above). The latter value is relatively large compared to similar values of typical MD simulations, and can likely be explained by the expected structural relaxation away from the initially modeled structure. However, the secondary structure and the compactness of the core remain largely unchanged, suggesting that the resulting hydration free energies (see below) are not adversely affected by these changes. Together with Rgyr, the solvent-accessible surface area of nudix hydrolase also decreases over time as the tails collapse and some of the initially exposed amino acids become buried (Figure 2).
As there is no experimental high-resolution information on the structure of DRNH, it is at present difficult to fully assess the accuracy of our simulated model of its unstructured tails, which is a general problem with most IDPs. To partially address this challenge, we have calculated the partial molar volume (PMV) of the simulated DRNH and its segments using the 3D-RISM formalism [36], [37], and compared it with calculated and experimental values for several different proteins (Figure S1). PMV is extremely sensitive to protein structure, and can be taken as a low-resolution test for the accuracy of our model [38]. Reassuringly, the PMV values for DRNH (full dimer, 41042±1825 cm3/mol) and its segments (N-tail single, 6579±43 cm3/mol; C-tail single, 3221±24 cm3/mol) fall closely on the line in the PMV versus molecular weight graph, defined by experimental as well as computed values for several different proteins (Figure S1). That many unstructured proteins have a similar partial specific volume as structured proteins (reflected in the linear dependence of the PMV on molecular weight) is a known experimental observation [39], and the fact that our results agree with it can be taken as partial evidence about the accuracy of our simulated model of DRNH.
What is the influence of the nudix hydrolase tails on the solvation properties of the entire molecule? The set of structures resulting from our simulations was used to compare the absolute hydration free energy of nudix hydrolase with the hydration free energies of a set of representative reference structures from the PDB (Figure 3a). The hydration free energies (HFE) of proteins in the reference set are all below 0 kJ/mol and centered around moderate values (<HFEref.set> = −17508±8219 kJ/mol) with a heavy tail in the direction of low hydration free energies. Since we excluded membrane proteins from the reference set, there are no structures with positive hydration free energies. Interestingly, the average hydration free energy of nudix hydrolase is more than 20000 kJ/mol below the average energy in the reference PDB set (−41009±3580 versus −17508±8219 kJ/mol). This is a significant deviation, with only very few proteins in the reference set being as hydrophilic as DRNH (Figure 3a). These results are reflected if one calculates the total free energy of the molecules, including their hydration free energy, but excluding their conformational entropy (Figure S2).
To test whether this large deviation in hydration free energy of nudix hydrolase is due to the ID tails, we created a set of structures with the N-terminal tail shortened by 44 residues from the N-terminus, and a set with neither N nor C-terminal tails at all. The hydration free energies of these structures show that a major contribution to the total hydration free energy of nudix hydrolase indeed comes from the two tails. For example, the average hydration free energy of the N-terminal Δ44 truncation mutants was −30438±3355 kJ/mol. Moreover, if one completely removes both ID tails of DRNH, its average hydration free energy is in the middle of the distribution of hydration free energies of the reference proteins (Figure 3a, “cores”).
The hydration free energy of proteins depends on their size. For this reason, we also compare the size-normalized hydration free energies of DRNH and the reference data set (Figure 3b). Albeit less dramatically, the size-normalized HFE of DRNH still remains significantly more negative than that of an average protein (with 90.5% of the reference proteins exhibiting a more positive hydration free energies than the average size-normalized HFE of DRNH).
Being extremely charged and hydrophilic, the composition of an ID stretch of amino acids is expected to share significant similarity with the composition of the water-exposed surface of a typical globular protein. In a way, one can think of ID regions as protein segments consisting of only surface and no hydrophobic core. How does this view agree with our simulated ensembles of nudix hydrolase from D. radiodurans? In Figure 4, we analyze the sequence complexity of different segments of DRNH as a function of relative solvent accessibility. In other words, we create mock sequences containing only those residues in the N-terminal tail, C-terminal tail, the core, or the complete DRNH sequence that are solvent-exposed to a given degree, and evaluate their sequence entropy (see Methods). Sequence entropy is a measure of the diversity (variability) of amino acids in a given sequence. In our case, it captures the heterogeneity of solvent-exposed residues in the tail and core regions of nudix hydrolase (the lower the value, the more homogenous are the types of residues on the surface). Importantly, the fact that all surface residues are strung together in one sequence plays no role – sequence entropy does not depend on the ordering of amino acids in a given stretch. We compare the complexity of nudix hydrolase' surface residue with the equivalent values obtained for the reference PDB set. First, looking at an accessibility cutoff of 0.0 (i.e. including all residues, both the buried ones and the ones at the surface), the sequence entropies of the complete N and C-terminal tails are significantly lower than that of either the complete nudix hydrolase core or the average protein from the reference PDB set. However, as one focuses more and more on just the exposed residues (i.e. with the accessibility cutoff increasing), the distribution of sequence complexities starts converging to a common value of about 3.0 at a solvent accessibility cutoff around 0.4 (Figure 4a). Importantly, this cutoff is the typical value accepted in the literature as a definition for surface residues. This means that if one looks at just the surface residues in the nudix core, or in a typical protein, their compositional complexity matches that of, for example, the complete nudix N-terminal tail. In other words, the complete N-terminal tail, by its sequence composition, is very similar to the surface of a typical globular protein. This trend is further exemplified in Figure 4b, where we show the dependence of the sequence complexities on the solvent accessibility cutoff.
The results presented in this paper show on several levels a correlation between the N and C-terminal tails' presence and an increased hydrophilicity of nudix hydrolase in D. radiodurans. We have shown a significant deviation of the tails' sequence complexity and hydrophilicity from those of the DRNH core. Furthermore, the significance of these results is enhanced by the fact that the tails' solvent-accessible area constitutes 64.91±1.41% of the total solvent-accessible area of the protein in the collapsed state. Finally, the results of the hydration free energy calculations show an even clearer picture: the hydration free energy of DRNH is twice as large as an average protein's hydration free energy, with the tails contributing the major part to this difference.
Although explicit-solvent molecular dynamics simulations and free energy calculations are typically more accurate, explicitly accounting for effects such as solute-solvent hydrogen bonds, the GB/SA implicit solvent [40]–[42] was the method of choice in our calculations for the following reason. Namely, implicit solvent methods are computationally less expensive and allow faster sampling of the conformational space, especially if performed at reduced values of solvent viscosity, as here. While the kinetic information coming from such simulations is likely compromised, the thermodynamic values, such as hydration free energies, should not be affected. Consequently, using implicit solvation allowed us to simulate a large system, which would hardly be accessible by any other means. In particular, ID regions are typically more flexible and less compact than average globular proteins, and therefore, at this time, implicit solvent simulations, along with Monte Carlo strategies, are likely the only feasible methods to study their structural and dynamic properties in atomistic detail. Finally, the 200-member set at 2 ns for a dimer of this size represents a state-of-the-art level of sampling on fast processors. Moreover, due to the solvent viscosity that is approximately 2 orders of magnitude lower than that of water, one can argue that the 2 ns trajectories actually capture events that would actually occur on the 200 ns time-scale. Namely, following the Kramers' relation, the rate of activated processes is inversely proportional to solvent viscosity in the high-friction limit [43]. While this relation does not hold for all values of solvent viscosity, it has been shown that for protein simulations it is valid for values down to approximately 1/1000 that of water [44], including the one used herein. That said, it should still be noted that implicit solvent models such as GB/SA may not be as accurate as explicit solvent models in capturing thermodynamic quantities such as hydration free energies [45]–[48]. In particular, it has been shown that the surface area dependent component in the GB/SA model, as well as in some other implicit solvent models, may not capture the non-polar part of the hydration free energy accurately enough for all applications in biomolecular simulations [49]–[51]. Most notably, the contributions of dispersion and solvent-accessible volume terms are missing in such models [52]. What is more, the accuracy of the GB/SA model critically depends on the accurate estimation of the effective Born radii [53]. An exciting novel approach for calculating solvation properties uses the three-dimensional reference interaction site model theory (3D-RISM) [36], [37], an integral equation theory for the behavior of molecular liquids, and has shown good agreement with other theoretical methods and experiments. Recent advances with this method even allow simulating molecular dynamics trajectories using the 3D-RISM formalism. It will be instructive to compare the results obtained here using the GB/SA formalism with those obtained using other methodologies. However, it is our belief that the quantitative effects observed in this study are so pronounced that they are independent of the potential inaccuracies in the specific methodology used for calculating hydration free energies.
While low viscosity has arguably allowed us to simulate events on the hundred nanosecond time-scale, it is still possible that with longer simulation time, one would observe major structural rearrangements of the tails. In particular, it is possible that some parts of the tails would actually adopt structured secondary and tertiary folds. This is particularly relevant for the C-terminal tail as its sequence exhibits lower levels of intrinsic disorder than the N-terminal tail. Would our conclusions be different if the tails adopted unique, ordered, folded structure? One way to address this question is to compare the hydration free energy of the complete nudix hydrolase dimer with those of similarly-sized folded globular proteins from our representative PDB set. In Figure S3, we summarize the results of such an analysis for nudix hydrolase and 23 globular proteins from our representative PDB set with sizes between 490 and 510 amino acids. Indeed, the hydration free energy of the nudix hydrolase dimer including the disordered tails is significantly more negative than those of the equivalent-size globular proteins (a difference of more than 5000 kJ/mol on average). In other words, if the tails were to adopt regular secondary and tertiary folds, it is likely that their hydration free energy would not be as negative as in the case for the unstructured tails. That said, any addition of extra hydrophilic amino acids, be it in the structured or unstructured conformation, would lower the hydration free energy of any protein. It is just that unstructured regions would likely have a stronger influence.
Note that the hydration free energy, as calculated and discussed here, is the free energy needed to transfer a protein from vacuum to water, and is only a first-order approximation to the actual free energy required to move a given protein from a desiccated region of a cell to a hydrated region. However, recalling the generalized Born equation, this approximation is likely only inaccurate up to a multiplicative constant. Namely, in the generalized Born formalism, the polar part of the hydration free energy of transfer between two media is proportional to the difference in the inverse dielectric constants of the two media. The part that depends on the environment is just a pre-factor dealing with dielectric permittivities. One can assume that these are the same for all proteins, provided they find themselves in similar environments both in water and in desiccated aggregates. In this sense, the relative ordering of HFEs that we calculated (water - vacuum) is the same as it would be in the case of the difference in HFE between water and desiccated aggregates. The only thing we assume here is that has the same sign as , which is likely the case, since aggregates are expected to have a lower permittivity than water. Another argument one can make is that in the GB/SA formulation, the environment enters the equations only through the dielectric permittivity, which one could claim to be similar in vacuum and in desiccated proteins. For this reason, we believe that the relative ordering of hydration free energies calculated here is the same as the ordering of free energies for transferring proteins from the aqueous environment into desiccated parts of the cell.
Our results show that the hydration free energy of DRNH is significantly more negative than that of a typical protein. This effect may be functionally important when water is scarce. Namely, under extreme desiccation conditions, D. radiodurans loses large amounts of water and experiences oxidative stress, causing a number of changes. The removal of the hydration shell from phospholipids in the membrane causes an increase in their phase transition temperature and their transition to gel phase at environmental temperatures [54]. In the cytoplasm, sugars such as trehalose can replace the shell of water molecules around proteins [55]. At environmental temperatures, this leads to the formation of a glassy matrix within cell (with mechanical properties of a plastic solid) with increased viscosity limiting in such way all processes that require diffusion and ensuring stability [56]. Proteins known to be expressed in stages of extreme desiccation are those belonging to the late embryiogenesis abundant (LEA) family of proteins [57]. Their function and mode of action are far from understood and call for further investigation. They are characterized with long stretches of highly hydrophilic and charged amino acid residues that are disordered in structure and are similar to the tails studied herein [57].
Under the conditions of extreme desiccation, it is important that housekeeping proteins and proteins required for efficient recovery after desiccation be additionally protected to help the organism recover in the first phase after rehydration. We hypothesize that precisely this is the function of the hydrophilic tails of D. radiodurans' nudix hydrolase – they increase the likelihood for the protein to stay in patches of residual water in the cell (i.e. in the more hydrated areas of the heterogeneous cellular matrix), while other proteins denature due to the removal of water (Figure 5). Note that this model, while supported by the data presented here, is preliminary and hypothetical – further experimental verification should shed light on its correctness. The ID regions at the termini of DRNH are unique to D. radiodurans, giving the bacterium a clear evolutionary advantage in harsh environments. One way to increase the overall hydrophilicity of a given protein is to change its surface residues in the course of an evolutionary process. However, if selective pressure on this trait is extremely strong, changing surface residues will at some point likely compromise the structural integrity of the whole protein. Therefore, we suggest that adding disordered tails at either the N or C-terminus may be a successful strategy to address this challenge without endangering the structural (and therefore enzymatic) integrity of the protein core.
IDPs have traditionally been associated with a number of particular functional classes: The function of entropic chains emerges completely from their lack of structure, effectors modify or inhibit the activity of their binding partners, scavengers store and neutralize small ligands, assemblers help assemble and stabilize protein complexes, and display sites mediate regulatory posttranslational modifications, such as phosphorylation or limited proteolysis [4]. Following the findings of the present work, we propose a new functional class for IDP regions, namely that of hydrators. The ID regions increase the probability for a protein to remain solvated in case of dehydration. Finally, as this function can be understood as a general mechanism for protecting proteins from denaturation, it can also be classified as a generalized chaperone activity. Future experimental and theoretical research should demonstrate the functional biological relevance of this hypothetical proposal.
There is no high-resolution structure of the nudix hydrolase studied herein (UniProtKB accession number Q9RWW5_DEIRA). We looked for possible homologues by performing a BLAST [58] search using the non-redundant data set (nr). A known structure of nudix hydrolase in T. thermophilus (PDB code 1VC8) was thus identified as the closest homologue of nudix hydrolase in D. radiodurans and was subsequently used to model the structure of the latter using Modeller [59]. As input, Modeller was also provided with secondary structure assignments for 3 residues that do not exist in the 1VC8 sequence, but are located in the middle of segments with well-defined secondary structure, as follows: GLY 165: β-strand, ASP 171: β-strand, ARG 205: α-helix. The N-terminal and C-terminal tails, unique to DRNH, were initially assigned physically realistic extended conformations. The 3D structure generated in this way was further refined using SQWRL3 [60], to optimize the position of amino-acid side-chains, and used as a starting structure for molecular dynamics simulations.
ID regions in the DRNH sequence were determined using three different algorithms: IUPRED [61] in the long-disorder mode, DisEMBL [21] 1.5 using the loops/coils definition, and the neural-network-based DisProtVL3H [62].
Nudix hydrolase was simulated in its naturally occurring form as a dimer using the Amber 9 molecular simulation package. After steepest descent energy minimization, we generated 200 independent trajectories, each 2 ns long, from the starting Modeller-generated model structure using implicit, low-viscosity solvent and the Onufriev et al. [41] generalized Born/surface area (GB/SA) model. The simulations were carried out using Langevin dynamics with a collision frequency of 1 ps−1, with each independent trajectory initiated from a different random number seed. A cutoff of 16.0 Å was chosen for all non-bonded interactions, and SHAKE bond-restraints were applied to all bonds. The hydration free energies in this study were calculated using the generalized Born/surface area method (GB/SA) in the Amber package with the same parameters as the simulations. All structural and thermodynamic analysis of nudix hydrolase was carried out on a composite set containing 200 final structures from 200 independent trajectories (at 2 ns).
Partial molar volumes (PMV) of the set of 200 final structures of DRNH, as well as a set of 7 proteins with sizes ranging from 50 up to 602 amino acids, were calculated using the rism3d_pmv routine available in AmberTools 1.4. The routine was parameterized to use the Kovalenko-Hirate closure [63]. The 7 proteins were chosen from a larger reference PDB set (see below).
The absolute and relative solvent accessibilities (SA) of each amino acid in the DRNH sequence, as well as in the reference PDB set, were calculated using DSSP [64]. The relative SA was determined by dividing the absolute SA of a given residue by the SA of the completely hydrated individual amino acid. The computed values allowed us to determine surface-exposed and buried residues as a function of variable cutoff.
The local sequence complexity of surface residues was quantified by the Shannon entropy, defined as:(1)where pi is the relative frequency of occurrence of each amino acid in a composite sequence containing all surface residues.
In order to be able to compare the properties of nudix hydrolase in D. radiodurans to those of typical proteins, we selected and examined a set of representative proteins with known tertiary structure. Using the PDB-REPR [65] tool, we filtered the protein databank (PDB) (November 2007) for proteins whose structures were solved in X-ray experiments with a resolution ≤2.5 Å, an R-factor ≤0.3 Å, and having all side chains resolved. Furthermore, we did not allow structures with chain breaks, non-standard amino acids, and structures shorter than 40 amino acids. We also excluded all non-natural mutants of a given structure, complexed proteins, protein fragments, and membrane proteins. Finally, we manually verified the set and removed all misclassified structures and structures containing residues with more than one rotamer, thus obtaining a final set of 800 proteins, including multimeric ones. As a control, we compared the hydration free energies of this representative PDB set with the corresponding energies of the reference set of structures used by Feig et al. [66]. Tinker's [67] pdb2xyz and xyz2pdb utilities were used in preparing the Feig et al. [66] data set for Amber calculations.
Before evaluating any energies of the PDB reference set, we used Amber's molecular dynamics tool to minimize the potential energy of all structures, removing possible steric clashes due to different parameterizations used in the original PDB data.
|
10.1371/journal.ppat.1002305 | Mechanisms of Candida albicans Trafficking to the Brain | During hematogenously disseminated disease, Candida albicans infects most organs, including the brain. We discovered that a C. albicans vps51Δ/Δ mutant had significantly increased tropism for the brain in the mouse model of disseminated disease. To investigate the mechanisms of this enhanced trafficking to the brain, we studied the interactions of wild-type C. albicans and the vps51Δ/Δ mutant with brain microvascular endothelial cells in vitro. These studies revealed that C. albicans invasion of brain endothelial cells is mediated by the fungal invasins, Als3 and Ssa1. Als3 binds to the gp96 heat shock protein, which is expressed on the surface of brain endothelial cells, but not human umbilical vein endothelial cells, whereas Ssa1 binds to a brain endothelial cell receptor other than gp96. The vps51Δ/Δ mutant has increased surface expression of Als3, which is a major cause of the increased capacity of this mutant to both invade brain endothelial cells in vitro and traffic to the brain in mice. Therefore, during disseminated disease, C. albicans traffics to and infects the brain by binding to gp96, a unique receptor that is expressed specifically on the surface of brain endothelial cells.
| During hematogenously disseminated infection, the fungus Candida albicans is carried by the bloodstream to virtually all organs in the body, including the brain. C. albicans infection of the brain is a significant problem in premature infants with disseminated candidiasis. To infect the brain, C. albicans must adhere to and invade the endothelial cells that line cerebral blood vessels. These endothelial cells express unique proteins on their surface that are not expressed by endothelial cells of other vascular beds. Here, we show that C. albicans infects the brain by binding to gp96, a heat shock protein that is uniquely expressed on the surface of brain endothelial cells. Gp96 is bound by the C. albicans Als3 invasin, which induces the uptake of this organism by brain endothelial cells. The C. albicans Ssa1 invasin also mediates fungal uptake by brain endothelial cells, but does so by binding to a receptor other than gp96. Thus, during hematogenously disseminated infection, C. albicans traffics to and infects the brain by binding to gp96, a receptor that is expressed specifically on the surface of brain endothelial cells.
| Hematogenously disseminated candidiasis is a serious disease that remains associated with approximately 35% mortality, even with currently available treatment, and Candida albicans is the infecting organism in approximately 50% of patients [1], [2]. During this disease, C. albicans is carried by the bloodstream to virtually all organs of the body, including the brain. Although candidal infection of the brain may not be clinically evident in adults with disseminated candidiasis, it is frequently found at autopsy in patients who die of this disease [3]. Even more importantly, candidal brain infection, especially meningitis, is a significant problem in premature infants who have risk factors for disseminated candidiasis, even in the absence of detectable candidemia [4], [5].
To invade the brain parenchyma, blood-borne C. albicans cells must adhere to and traverse the endothelial cell lining of the blood vessels within the central nervous system. Brain endothelial cells are significantly different from those lining systemic blood vessels. For example, they have tight junctions that are not present in the endothelial cells in other vascular beds. Forming the blood brain barrier, brain endothelial cells restrict the diffusion of large or hydrophilic molecules into the central nervous system, while allowing the diffusion of small hydrophobic molecules [6]. More importantly, some microbial pathogens, such as Neisseria meningitidis, Streptococcus pneumonia, Escherichia coli K1, and Cryptococcus neoformans have an enhanced capacity to adhere to and invade human brain microvascular endothelial cells (HBMECs), which enables them to preferentially infect the central nervous system via the hematogenous route [7]–[11]. Thus, these pathogens can exploit the unique characteristics of HBMECs to specifically infect the brain.
Studies using human umbilical vein endothelial cells (HUVECs) as representative systemic endothelial cells have demonstrated that C. albicans adheres to, invades, and damages these cells in vitro [12], [13]. One mechanism by which C. albicans invades these cells is by stimulating its own endocytosis, which is induced when the C. albicans invasins, Als3 and Ssa1, bind to receptors such as N-cadherin and HER2 on the endothelial cell surface [14]–[17]. C. albicans yeast and hyphae can also invade HBMECs by inducing their own endocytosis [18], [19]. However, the mechanism by which this pathogen invades these endothelial cells and infects the brain is poorly understood.
Recently we discovered that C. albicans VPS51 is up-regulated by contact with HUVECs in vitro, and that a vps51/vps51 insertion mutant is defective in damaging these endothelial cells [20]. In Saccharomyces cerevisiae, Vps51 is known to bind to the Vps52/53/54 complex and is required for the retrograde transport of proteins from endosomes to the late Golgi [21], [22]. Although the function of Vps51 in C. albicans has not been studied in detail, the vps51/vps51 insertion mutant has a fragmented vacuole, similar to the corresponding S. cerevisiae mutant [20]–[22]. Thus, Vps51 likely plays a role in protein trafficking in C. albicans.
In the current study, we investigated how deletion of VPS51 affects the virulence of C. albicans during hematogenous infection. We found that the vps51Δ/Δ null mutant exhibits a preferential tropism for the brain. This tropism is mediated in part by the enhanced exposure of Als3 on the surface of the vps51Δ/Δ mutant, which binds to gp96 on the surface of HBMECs and mediates invasion of these endothelial cells. We further discovered that gp96 functions as a receptor for wild-type C. albicans on HBMECs, but not HUVECs, indicating that this organism invades the central nervous system by binding to a receptor that is expressed specifically on HBMECs.
To investigate the role of Vps51 in the virulence of C. albicans, we inoculated mice via the lateral tail vein with a wild-type strain, a vps51Δ/Δ mutant, and a vps51Δ/Δ+pVPS51 complemented strain and then monitored their survival over time. We found that all mice infected with the vps51Δ/Δ mutant survived for the entire 21-day observation period, whereas all mice infected with the wild-type strain died within 7 days after inoculation (Figure 1A). Complementing the vps51Δ/Δ mutant with an intact copy of VPS51 restored its virulence to wild-type levels, thus confirming that Vps51 is required for the maximal virulence of C. albicans. The greatly reduced virulence of the vps51Δ/Δ mutant was further verified by infecting mice with a 6-fold higher inoculum. As expected, mice infected with the wild-type strain at this higher inoculum died rapidly, with a median survival of only 3 days (Figure 1B). However, all mice infected with the vps51Δ/Δ mutant still survived. Therefore, Vps51 is necessary for the full virulence of C. albicans.
The mouse model of hematogenous disseminated candidiasis mimics many aspects of this disease in humans, particularly the formation of microabscesses in most organs [23], [24]. We therefore investigated the effects of deleting VPS51 on organ fungal burden. During the first 4 days of infection, the kidneys and livers of mice infected with the vps51Δ/Δ mutant contained significantly fewer organisms than those of mice infected with either the wild-type or vps51Δ/Δ+pVPS51 complemented strain (Figure 1C and D). Furthermore, the kidney fungal burden of mice infected with the vps51Δ/Δ mutant progressively declined after the first day of infection. In contrast, the kidney fungal burden of mice infected with the wild-type and vps51Δ/Δ+pVPS51 complemented strains progressively increased for the first 4 days post-infection, after which these mice began to die. These results further demonstrate that the overall virulence of the vps51Δ/Δ mutant is decreased.
A surprising result was that during the first 4 days of infection, the brain fungal burden of mice infected with the vps51Δ/Δ mutant was significantly greater than that of mice infected with either the wild-type or vps51Δ/Δ+pVPS51 complemented strain (Figure 1E). Indeed, after 3 days of infection, the brains of mice infected with the vps51Δ/Δ mutant contained a median of 50-fold more organisms than those of mice infected with the wild-type strain. Despite having a high brain fungal burden, the mice infected with the vps51Δ/Δ mutant did not appear to be sick and had no obvious signs of neurological disease. Moreover, beginning on the fourth day of infection, these mice progressively cleared the organisms from their central nervous system. These results suggest that while the overall virulence of the vps51Δ/Δ mutant is decreased, it has a distinct tropism for the brain.
We verified these quantitative culture results by performing histopathologic analysis of the brains of the infected mice. Foci containing multiple organisms were visible in the brains of mice infected with the vps51Δ/Δ mutant, especially in the hippocampus (Figure 1F). In sharp contrast, only rare organisms were visible in the brains of the mice infected with either the wild-type or vps51Δ/Δ+pVPS51 complemented strains, and these organisms were typically either solitary or in pairs.
To determine if another member of the Vps51/52/53/54 complex is required for maximal virulence and enhanced brain tropism, we constructed and analyzed a vps53Δ/Δ mutant. This strain also caused no mortality in mice following tail vein inoculation. (Figure 2A) Furthermore, it accumulated at significantly higher levels in the brain than did the wild-type and vps53Δ/Δ+pVPS53 complemented strains (Figure 2B). Collectively, these results demonstrate that the Vps51/52/53/54 complex plays a key role in virulence, and that C. albicans strains that lack components of this complex preferentially infect the brain.
To cross the endothelial cell lining of the vasculature, C. albicans must first adhere to these endothelial cells and then invade through them. We hypothesized that the vps51Δ/Δ mutant had increased capacity to infect the brain because it preferentially adhered to and invaded the unique endothelial cells that line the blood vessels of central nervous system. To test this hypothesis, we compared the interactions of this mutant with HUVECs and HBMECs in vitro. We found that the adherence of the vps51Δ/Δ mutant to HUVECs was increased by only 22% compared to the wild-type strain (Figure 3A). However, the adherence of this mutant to HBMECs was increased by 95% (Figure 3B). There was an even greater difference in the capacity the vps51Δ/Δ mutant to induce its own endocytosis by HUVECs compared to HBMECs. The endocytosis of the vps51Δ/Δ mutant by HUVECs was 58% lower than that of the wild-type strain (Figure 3C). In contrast, the endocytosis of this mutant by HBMECs was 39% higher than the wild-type strain (Figure 3D). Complementing the vps51Δ/Δ mutant with an intact copy of VPS51 restored its interactions with both types of endothelial cells to wild-type levels. The increased capacity of the vps51Δ/Δ mutant to adhere to and invade HBMECs compared to HUVECs provides a likely explanation for the enhanced tropism of this mutant for the brain.
Next, we sought to identify the HBMEC receptor for both wild-type C. albicans and the vps51Δ/Δ mutant. HBMECs are known to express high amounts of the heat shock protein, gp96 on their cell surface, whereas HUVECs do not [9]. Furthermore, gp96 functions as an HBMEC-specific receptor for E. coli K1 strains that cause neonatal meningitis [9]. We used multiple complementary approaches to evaluate whether gp96 expression is required for C. albicans to invade HBMECs. First, we tested the capacity of an anti-gp96 antibody to inhibit HBMEC endocytosis of C. albicans. This antibody reduced the endocytosis of wild-type C. albicans by 24% and the vps51Δ/Δ mutant by 48% (Figure 4A). Second, we determined the effects of siRNA-mediated knockdown of gp96 on endocytosis. HBMECs transfected with gp96 siRNA endocytosed 52% fewer wild-type C. albicans cells and 82% fewer vps51Δ/Δ cells than did HBMECs transfected with control siRNA (Figure 4B). Importantly, the effect of gp96 knockdown on endocytosis was specific for HBMECs because knockdown of gp96 in HUVECs had no effect on their capacity to endocytose C. albicans (Figure 4C). Also, knockdown of gp96 did not inhibit HBMECs endocytosis of transferrin (Figure 4D), demonstrating that reducing gp96 protein levels did not cause a global decrease in receptor-mediated endocytosis. Collectively, these results indicate that gp96 is required for maximal HBMEC endocytosis of both wild-type C. albicans and the vps51Δ/Δ mutant.
To further explore these findings, we investigated the effects of overexpressing gp96 on the endocytosis of C. albicans. We found that overexpression of gp96 in HBMECs enhanced the endocytosis of the wild-type strain and the vps51Δ/Δ mutant by 100% and 115%, respectively (Figure 4E). Similarly, heterologous expression of human gp96 in Chinese hamster ovary (CHO) cells resulted in a 42% increase in the endocytosis of wild-type C. albicans and a 102% increase in the endocytosis of the vps51Δ/Δ mutant compared to control CHO cells transfected with the empty vector (Figure 4F). Therefore, these combined results demonstrate that gp96 functions as an HBMEC receptor that mediates the endocytosis of both wild-type C. albicans and the vps51Δ/Δ mutant.
Our previous studies revealed that the C. albicans proteins Ssa1 and Als3 function as invasins that induce the endocytosis of this organism by HUVECs [14], [15]. To investigate the roles of these fungal proteins in HBMEC invasion, we analyzed ssa1Δ/Δ and als3Δ/Δ single mutants, as well as vps51Δ/Δ ssa1Δ/Δ and vps51Δ/Δ als3Δ/Δ double mutants. Approximately 30% fewer hyphae of the ssa1Δ/Δ single mutant were endocytosed by HBMECs as compared to wild-type parent strain and the ssa1Δ/Δ+pSSA1 complemented strain (Figure 5A). Similarly, the endocytosis of the vps51Δ/Δ ssa1Δ/Δ double mutant was significantly lower than the vps51Δ/Δ single mutant. Thus, Ssa1 is required for the maximal endocytosis of both wild-type and vps51Δ/Δ mutant strains of C. albicans by HBMECs in vitro.
Als3 played a greater role than Ssa1 in stimulating the endocytosis of C. albicans by HBMECs in vitro. Both the als3Δ/Δ single mutant and the vps51Δ/Δ als3Δ/Δ double mutant were endocytosed extremely poorly by these endothelial cells (Figure 5B), indicating that Als3 is essential for the endocytosis of C. albicans by HBMECs in vitro.
To determine whether Ssa1 and Als3 mediate the endocytosis of C. albicans by directly interacting with endothelial cells, we used a heterologous expression strategy in which we expressed C. albicans SSA1 or ALS3 in the normally non-invasive yeast, Saccharomyces cerevisiae [25]. Expression of C. albicans SSA1 in S. cerevisiae resulted in a 300% increase in the endocytosis of this organism by HUVECs and a 43% increase in its endocytosis by HBMECs, as compared to the control strain of S. cerevisiae (Figure 6A and B). Moreover, expression of C. albicans ALS3 in S. cerevisiae resulted in a 2050% and 1880% increase in endocytosis by HUVECs and HBMECs, respectively (Figures 6C and D). Collectively, these data demonstrate that Ssa1 is a more potent inducer of fungal endocytosis by HUVECs than by HBMECs, whereas Als3 can induce endocytosis by HUVECs and HBMECs with similar efficacy. As HUVECs do not express gp96 on their surface [9], HUVEC endocytosis of S. cerevisiae expressing C. albicans SSA1 or ALS3 is mediated by receptors other than gp96, such as N-cadherin and HER2 [14]–[17].
The above results suggested a model in which Als3 on the surface of C. albicans hyphae binds to gp96 on the surface of HBMECs and induces endocytosis. To test this model, we analyzed the effects of siRNA knockdown on the endocytosis of the S. cerevisiae strain that expressed C. albicans Als3. As predicted, knockdown of gp96 in HBMECs reduced the endocytosis of the Als3 expressing strain of S. cerevisiae by 79% compared to control HBMECs (Figure 6E).
We also tested the capacity of different C. albicans mutants and strains of S. cerevisiae to bind gp96 in HBMEC membrane protein extracts. As predicted by our endocytosis results, the vps51Δ/Δ mutant bound more gp96 than did the wild-type strain (Figure 7A). Also, the ssa1ΔΔ mutant bound slightly less gp96 than did the wild-type strain, and the als3Δ/Δ mutant bound very poorly to this protein. Finally, the strain of S. cerevisiae that expressed C. albicans Als3 bound to gp96, whereas the control strain of S. cerevisiae did not (Figure 7B), thus indicating that Als3 directly interacts with gp96.
Next, we used flow cytometric analysis of C. albicans hyphae stained with either anti-HSP70 or anti-Als3 antibodies to quantify the levels of Ssa1 and Als3 that were exposed on the surface of the various strains. Although the vps51Δ/Δ mutant had normal Ssa1 surface exposure (data not shown), it had greater surface exposure of Als3 than did the wild-type and vps51Δ/Δ+pVPS51 complemented strains (Figure 7C). The greater surface exposure of Als3 by the vps51Δ/Δ mutant likely contributes its enhanced capacity to induced HBMEC endocytosis.
Lastly, we investigated the roles of Ssa1 and Als3 in mediating brain invasion in vivo by both wild-type and vps51Δ/Δ mutant strains of C. albicans. Mice were inoculated with the various C. albicans strains via the tail vein and their brain fungal burden was determined 3 days later. Similar to our previous results [14], the brain fungal burden of mice infected with the ssa1Δ/Δ single mutant was significantly less than that of mice infected with either the wild-type strain or the ssa1Δ/Δ+pSSA1 complemented strain (Figure 8A). However, the brain fungal burden of mice infected with the vps51Δ/Δ ssa1Δ/Δ double mutant was only 1.7-fold lower than that of mice infected with the vps51Δ/Δ single mutant, a difference that did not achieve statistical significance (p = 0.053). Taken together, these results indicate that Ssa1 is necessary for wild-type C. albicans to cause maximal brain infection, but that it plays a relatively minor role in enhanced brain tropism of the vps51Δ/Δ mutant.
Different results were obtained with strains that lacked Als3. The brain fungal burden of mice infected with the als3Δ/Δ single mutant was similar to that of mice infected with the wild-type strain (Figure 8B). In contrast, mice infected with the vps51Δ/Δ als3Δ/Δ double mutant had 5.5-fold fewer organisms in their brain compared to mice infected with the vps51Δ/Δ single mutant. Therefore, although Als3 is dispensable for wild-type C. albicans to infect the brain, it is important for the vps51Δ/Δ mutant to achieve maximal brain fungal burden.
In the mouse model of disseminated candidiasis, kidney fungal burden is directly correlated with mortality [23], [26]. Thus, many studies of this disease have used kidney fungal burden as the primary endpoint when analyzing either the virulence of mutant strains of C. albicans in mice or the susceptibility of mutant strains of mice to disseminated candidiasis [27]–[31]. However, during disseminated candidiasis in both mice and humans, C. albicans infects virtually all organs in the body. To do so, the blood-borne organisms must adhere to and invade the vascular beds of these organs. Importantly, there are significant differences among the endothelial cells that line the vasculature of the different organs, as well as the immunologic milieu of these organs [32], [33]. These differences provide a compelling rationale to investigate the capacity of C. albicans to traffic to and persist in organs other than the kidney. The brain is a particularly important target organ in neonates with hematogenously disseminated candidiasis [4], [5], and its blood vessels are lined with the unique endothelial cells that form the blood-brain barrier. Our studies with a vps51Δ/Δ mutant strain of C. albicans led us to discover that C. albicans traffics to the brain and invades cerebral blood vessels in part by binding to gp96 that is expressed on the surface of brain endothelial cells.
We had previously identified C. albicans VPS51 through a microarray study that was designed to discover genes that were up-regulated when the organism adhered to HUVECs [20]. In that study, we determined that a vps51/vps51 insertion mutant had reduced capacity to damage HUVECs and increased susceptibility to antimicrobial peptides [20]. These in vitro findings led us to predict that VPS51 would be required for the maximal virulence of C. albicans during disseminated disease. In the current study, we verified this prediction by determining that mice infected with a vps51Δ/Δ deletion mutant had no mortality and progressively cleared this strain from their kidneys and liver.
A unique and unexpected phenotype of the vps51Δ/Δ mutant was its marked propensity to infect the brain. In the few previous studies in which the brain fungal burden of mice infected with mutant strains of C. albicans was determined, the fungal burden in the brain generally paralleled the fungal burden in the kidney. For example, mice infected with ecm33Δ/Δ and hog1Δ/Δ mutants had improved survival and reduced fungal burden in both the kidney and the brain, as compared to mice infected with the wild-type strain [34], [35]. Thus, it was unusual to find that mice infected with the vps51Δ/Δ mutant had reduced kidney fungal burden, yet significantly increased brain fungal burden.
Our finding that the enhanced capacity of the vps51Δ/Δ mutant to adhere to and invade HBMECs, as compared to HUVECs, provides a likely explanation for its brain tropism. One difference between HBMECs and HUVECs is that the former cells express gp96 on their surface, whereas the latter cells do not [9]. Multiple lines of evidence indicate that gp96 functions as an HBMEC receptor for both wild-type C. albicans and the vps51Δ/Δ mutant. For example, an anti-gp96 antibody and siRNA knockdown of gp96 inhibited HBMEC endocytosis of C. albicans. Furthermore, overexpression of gp96 in HBMEC and the heterologous expression of human gp96 in CHO cells increased the endocytosis of C. albicans. Finally, wild-type C. albicans cells bound to gp96 in extracts of HBMEC membrane proteins, and the highly endocytosed vps51Δ/Δ mutant bound even more of this protein. Collectively, these data indicate that gp96 is an HBMEC receptor for C. albicans.
It was notable that in both the anti-gp96 antibody studies and the gp96 siRNA experiments, inhibition of gp96 function or expression had greater effect on the endocytosis of the vps51Δ/Δ mutant than the wild-type strain (78% reduction for the vps51Δ/Δ mutant vs. 38% reduction for the wild-type strain; p<0.0001). These results indicate that the vps51Δ/Δ mutant preferentially utilizes gp96 as a receptor to invade HBMECs. They further suggest that the enhanced brain tropism of the vps51Δ/Δ mutant is likely due to its increased binding to gp96 on the surface of brain endothelial cells.
Although these results strongly indicate that gp96 is important for HBMEC endocytosis of C. albicans, the findings that neither the anti-gp96 antibody nor siRNA knockdown of gp96 completely blocked the endocytosis of this organism suggest that it can invade HBMECs by additional mechanisms. Such mechanisms include the induction of endocytosis by binding to one or more receptors, such as N-cadherin that are independent of gp96 and active penetration, in which hyphae physically push their way into host cells by progressively elongating [16], [36].
Because gp96 also functions as a molecular chaperone [37], it is possible that it could be involved in the endocytosis of C. albicans by altering the expression or function of other proteins on the surface of HBMECs. Our data indicate that this possibility is remote because HBMEC endocytosis of C. albicans was inhibited by the anti-gp96 antibody, which is unlikely to affect the chaperone function of gp96. In addition, siRNA knockdown of gp96 inhibited the endocytosis of C. albicans by HBMECs, but not HUVECs, in which gp96 is located intracellularly. Moreover, gp96 knockdown did not affect transferrin uptake in HBMECs, a process that is mediated by the transferrin receptor. Thus, the role of gp96 in inducing the endocytosis of C. albicans is likely due to its function as a cell surface receptor rather than a chaperone.
Gp96 has been reported to be expressed on the surface of some epithelial cells where it functions as a receptor for Listeria monocytogenes, Neisseria gonorrhoeae and bovine adeno-associated virus [38]–[40]. In addition, gp96 on the surface of HBMEC is known to be bound by E. coli K1 OmpA [9]. This binding induces the endocytosis of E. coli by activating signal transducer and activator of transcription 3 (STAT3), which functions upstream of phosphatidylinositol-3 kinase and protein kinase C-α [41]–[43]. Whether the binding of C. albicans to gp96 activates a similar signaling pathway remains to be determined.
C. albicans possesses at least two invasin-like proteins, Ssa1 and Als3. Both of these proteins induce the endocytosis of C. albicans by HUVECs by binding to N-cadherin and other endothelial cell receptors [14], [15]. These two invasins may function cooperatively because the endocytosis defect of an ssa1Δ/Δ als3Δ/Δ double mutant is not greater than that of an als3Δ/Δ single mutant [14]. Our current studies with the C. albicans ssa1Δ/Δ and als3Δ/Δ mutants and strains of S. cerevisiae that overexpress C. albicans Ssa1 and Als3 demonstrate that both of these proteins can induce HBMEC endocytosis. The results of these in vitro experiments also indicate that Als3 is more important than Ssa1 in inducing HBMEC endocytosis, probably because it plays a greater role in binding to gp96.
Our mouse studies suggest that Ssa1 is required for the maximal trafficking of wild-type C. albicans to the brain because the brain fungal burden of mice infected with the ssa1Δ/Δ mutant was significantly less than that of mice infected with the wild-type strain. These results are similar to our previous data [14]. However, deletion of SSA1 in the vps51 mutant had only a minor effect on brain trafficking. It is probable that in the vps51Δ/Δ mutant, the effects of deleting SSA1 were masked by the increased surface expression of Als3.
A paradoxical finding was that although the endocytosis of the als3Δ/Δ mutant by HBMECs was severely impaired in vitro, this mutant had normal trafficking to the brain in mice. The normal virulence of an als3Δ/Δ mutant in the mouse model of disseminated candidiasis has recently been reported by others [44]. It is unclear why there is such a large discrepancy between the host cell interactions of the als3Δ/Δ mutant in vitro and its virulence in mice, especially because ALS3 is highly expressed in vivo [45], [46]. The most probable explanation for these paradoxical results is that other invasins, such as Ssa1 and perhaps other proteins, compensate for the absence of Als3. Because the in vitro experiments were performed using human endothelial cells and the virulence experiments were performed in mice, it is theoretically possible that differences between human and mouse gp96 may account for the differences between the in vitro and in vivo results. However, human and mouse gp96 are 97.5% identical at the amino acid level, making this possibility unlikely.
Importantly, our results indicate that Als3 does play a role in the enhanced brain tropism of the vps51Δ/Δ mutant because the brain fungal burden of mice infected with vps51Δ/Δ als3Δ/Δ double mutant was significantly lower than that of mice infected with the vps51Δ/Δ single mutant. Because protein trafficking is likely abnormal in the vps51Δ/Δ mutant, we speculate that this strain has reduced expression of compensatory proteins in response to deletion of ALS3. On the other hand, the vps51Δ/Δ als3Δ/Δ double mutant still had greater tropism for the brain compared to the wild-type strain. This result suggests that the overexpression of additional proteins, other than Als3, contributes to the brain tropism of the vps51Δ/Δ single mutant.
The combined results of these experiments support a model in which C. albicans invades the brain during hematogenously disseminated infection by binding to proteins that are specifically expressed on the surface of brain endothelial cells. One of these proteins is gp96, which is bound predominantly by C. albicans Als3 (Figure 9). At least one other brain endothelial cell protein functions as receptors for C. albicans Ssa1. As the endothelial cells of other vascular beds also express unique surface proteins, it is highly probable that blood-borne C. albicans utilizes different endothelial cell surface proteins to infect different organs. Identification of these organ-specific receptors for C. albicans may lead to novel approaches to block these receptors and thereby prevent hematogenous dissemination.
The fungal strains used in this study are listed in Supplemental Table S1. All C. albicans mutant strains constructed for this study were derived from strain BWP17 [47]. Deletion of the entire protein coding regions of both alleles of VPS51 was accomplished by successive transformation with ARG4 and HIS1 deletion cassettes that were generated by PCR using the oligonucleotides vps51-f and vps51-r (The oligonucleotide sequences are listed in Supplemental Table S2) [47]. The resulting strain was subsequently transformed with pGEM-URA3 [47] to re-integrate URA3 at its native locus. The vps53Δ/Δ mutant was constructed similarly, using the oligonucleotides vps53-f and vps53-r. To construct the VPS51 complemented strain (vps51Δ/Δ+pVPS51), a 2.6 Kb fragment containing VPS51 was generated by high fidelity PCR with the primers vps51-rev-f and vps51-rev-r using genomic DNA from C. albicans SC5314 as the template. This PCR product was digested with NcoI, and then subcloned into pBSK-Ura, which had been linearized with NcoI. The resulting construct was linearized with NotI and PstI to direct integration at the URA3 locus of a Ura– vps51Δ/Δ mutant strain. The vps53Δ/Δ-complemented strain (vps53Δ/Δ+pVPS53) was generated similarly, except that primers vps53-rev-f and vps53-rev-r were used to PCR amplify a 3.3 Kb DNA fragment containing VPS53.
To delete the entire protein coding region of ALS3 in the vps51Δ/Δ mutant, deletion cassettes containing ALS3 flanking regions and the URA3 or NAT1 selection markers were amplified by PCR with primers als3-pgem-KO-f and als3-pgem-KO-r, using pGEM-URA3 [47] and pJK795 [48] as templates, respectively. These PCR products were then used to successively transform a Ura- ssa1Δ/Δ strain. The resulting als3Δ/Δ vps51Δ/Δ double mutant was plated on 5-fluoroorotic acid to select for a Ura- strain, which was then transformed with pGEM-URA3 as above. The als3Δ/Δ vps51Δ/Δ+pVPS51 complemented strain was generated the same way as was the vps51Δ/Δ+pVPS51 complemented strain. The ssa1Δ/Δ vps51Δ/Δ double mutant and its VPS51-complemented strain (ssa1Δ/Δ vps51Δ/Δ+pVPS51) were generated similarly to the als3Δ/Δ vps51Δ/Δ double mutant and its complemented strain, except that primers ssa1-pgem-f and ssa1-pgem-r were used to amplify the SSA1 deletion cassettes.
The construction of the S. cerevisiae strain that expressed C. albicans ALS3 under the control of the ADH1 promoter and its control strain containing the backbone vector was described previously [25]. To express C. albicans SSA1 in S. cerevisiae, a 2.0 kb fragment containing the SSA1 protein coding region was generated by PCR with primers ssa1-exp-bglii-f and ssa1-exp-xhoi-r using pRP10-SSA1ORF as template [49]. The resulting SSA1 fragment was cloned downstream of the GAL1 promoter of pYES2.1/V5-His-TOPO using the pYES2.1 TOPO TA Expression Kit (Invitrogen) following the manufacturer's instructions. The control strain of S. cerevisiae was transformed with the backbone vector alone. Expression of C. albicans SSA1 was induced by growth in SC minimal medium containing 2% galactose following the manufacturer's protocol.
Male BALB/c mice weighing 18–20 g (Taconic Farms) were used for all animal experiments. For survival studies, 10 mice per strain were injected via the tail vein with either 5×105 or 3×106 yeast of the various C. albicans strains [50] and then monitored for survival three times daily. All inocula were confirmed by colony counting. In the organ fungal burden studies, the mice were inoculated with 5×105 yeast as above. At various time points, 7 mice per strain were sacrificed and the kidney, liver, and brain were harvested. These organs were weighed, homogenized and quantitatively cultured. For histopathological analysis, a portion of the excised tissue was fixed in zinc-buffered formalin followed by 70% ethanol. The tissue was then embedded in paraffin, after which thin sections were prepared and stained with Gomori methenamine silver. They were examined by light microscopy. All mouse experiments were approved by the Animal Care and Use Committee at the Los Angeles Biomedical Research Institute and carried out according to the National Institutes of Health (NIH) guidelines for the ethical treatment of animals.
HUVECs were harvested from umbilical cords with collagenase and grown in M-199 medium supplemented with 10% fetal bovine serum and 10% defined bovine calf serum (Gemini Bio-Products), and containing 2 mM L-glutamine with penicillin and streptomycin (Irvine Scientific) as previously described [51]. HBMECs were isolated from the capillaries in small fragments of the cerebral cortex, which were obtained by surgical resection from 4- to 7-year-old children with seizure disorders at Children's Hospital Los Angeles. HBMECs were harvested from these capillaries and maintained in a mixture of M-199 and Ham's F-12 media (1∶1 v/v) supplemented with 10% fetal bovine serum, 1 mM sodium pyruvate, and 2 mM glutamine as described previously [52]. More than 98% of these cells were positive for Factor VIII-rag and carbonic anhydrase, and negative for GFAP by flow cytometry. In addition, 99% of the cells took up Dil-Ac-LDL by immunocytochemistry. CHO K-1 cells expressing human gp96 were generated and grown as outlined before [9]. All cell types were grown at 37°C in 5% CO2.
HBMECs or HUVECs were grown to 60% confluence in six-well plates, then transfected with either gp96 siRNA (Catalog number HSS110955; Invitrogen) or a random control siRNA using lipofactamine 2000 (Invitrogen), according to the manufacturer's instructions. Gp96 knockdown was verified by Western blotting of total endothelial cell lysates with an anti-gp96 monoclonal antibody (Santa Cruz Biotechnology).
HBMEC membrane proteins from host cells were isolated using octyl-glucopyranoside exactly as described previously [16]. Next, 2×108 hyphae of the various C. albicans strains or 8×108 yeast of the different S. cerevisiae strains were incubated on ice for 1 h with 250 µg of HBMEC membrane proteins in PBS with calcium and magnesium and containing 1.5% octyl-glucopyranoside and protease inhibitors. The unbound proteins were removed by extensive rinsing in the same buffer. Next, the proteins that had bound to the hyphae were eluted with 6M urea. The eluted proteins were separated by SDS-PAGE and detected by immunoblotting with the anti-gp96 antibody using enhanced chemiluminescence (Pierce).
The adherence of C. albicans to HUVECs and HBMECs grown in 6-well tissue culture plates was measured by a modification of our previously described method [25]. Briefly, germ tubes of the various strains were generated by a 1-h incubation in RPMI 1640 medium (Irvine Scientific) at 37°C. The germ tubes were enumerated with a hemacytometer and suspended in HBSS at 200 cells/ml. After rinsing the endothelial cell monolayers twice with HBSS, 1 ml of the germ tube suspension was added to each well. The cells were incubated for 30 min, after which the nonadherent organisms were aspirated and the endothelial cell monolayers were rinsed twice with HBSS in a standardized manner. Next, the wells were overlaid with YPD agar and the number of adherent organisms was determined by colony counting. The adherence results were expressed as a percentage of the initial inoculum, which was verified by quantitative culture. Each strain was tested in triplicate on three different days.
The number of organisms internalized by the endothelial cells was determined using our standard differential fluorescence assay [15], [16]. Briefly, endothelial cells on glass coverslips were infected with 105 yeast phase cells of each strain of C. albicans in RPMI 1640 medium. After incubation for 3 h, the cells were fixed with 3% paraformaldehyde. The noninternalized cells were stained with anti-C. albicans rabbit serum (Biodesign International) that had been conjugated with Alexa 568 (Invitrogen). Next, the endothelial cells were permeablized in 0.1% (vol/vol) Triton X-100 in PBS, after which both the internalized and the noninternalized organisms were stained with anti-C. albicans rabbit serum conjugated with Alexa 488 (Invitrogen). The coverslips were mounted inverted on a microscope slide and viewed under epifluorescence. The number of organisms endocytosed by the endothelial cells was determined by subtracting the number of noninternalized organisms (which fluoresced red) from the total number of organisms (which fluoresced green). At least 100 organisms were counted on each coverslip, and all experiments were performed in triplicate on at least three separate occasions.
HBMECs were grown to 70% confluency in 6-well tissue culture plates and then incubated for 3 in serum-free medium to deplete endogenous transferrin. Next they were incubated for 45 min in serum-free medium containing AlexaFluor 555-labeled transferrin (Invitrogen; 10 µg/ml). The unincorporated transferrin was removed by rinsing, after which the cells were incubated for an additional 30 min. Any remaining surface bound transferrin was removed by rinsing the cells twice with ice-cold PBS containing Ca++ and Mg++ (PBS++) followed by two, 5 min incubations with ice-cold acid wash buffer (0.2 M acetic acid (pH 2.8) 0.5 M NaCl). Finally, the cells were washed three times with ice-cold PBS++, detached with Cell Dissociation Buffer (Invitrogen), and suspended in PBS++. Their transferrin content was determined by flow cytometry, analyzing at least 10,000 cells.
Flow cytometry was used to analyze the surface expression Als3p on hyphae of the various strains using a minor modification of our previously described method [14]. Briefly, hyphae of the different strains of C. albicans were fixed in 3% paraformaldehyde and blocked with 1% goat serum. The hyphae were then incubated with either a rabbit polyclonal antiserum raised against rAls3-N or purified rabbit IgG. After extensive rinsing, the cells were incubated with a goat anti-rabbit secondary antibody conjugated with Alexa 488. The fluorescent intensity of the hyphae was measured by flow cytometry. Fluorescence data for 10,000 cells of each strain were collected.
The capacity of the various strains of C. albicans and S. cerevisiae to adhere to, and be endocytosed to endothelial cells was compared using analyses of variance. Differences in the fungal burden of mice infected with these strains were analyzed using the Wilcoxon Rank Sum test. Differences in survival were analyzed using the Log-Rank test.
The protocol for collecting umbilical cords for the harvesting of HUVECs used in these studies was approved by the Institutional Review Board of the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center. This protocol was granted a waiver of consent because the donors remained anonymous. The protocol for using fragments of the cerebral cortex, obtained by surgical resection from 4- to 7-year-old children with seizure disorders, for isolation of HBMECs was approved by the Institutional Review Board of Childrens Hospital Los Angeles. These fragments were obtained from anonymous donors in 1992-1993 and the HBMECs used in the current studies were isolated at that time and stored in liquid nitrogen. The use of HBMECs in our studies is exempted because the donors are unknown and there is no information linking the HBMECs with the donors. The mouse studies were carried out in accordance with the National Institutes of Health guidelines for the ethical treatment of animals. This protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center.
|
10.1371/journal.pgen.1004060 | Subtle Changes in Motif Positioning Cause Tissue-Specific Effects on Robustness of an Enhancer's Activity | Deciphering the specific contribution of individual motifs within cis-regulatory modules (CRMs) is crucial to understanding how gene expression is regulated and how this process is affected by sequence variation. But despite vast improvements in the ability to identify where transcription factors (TFs) bind throughout the genome, we are limited in our ability to relate information on motif occupancy to function from sequence alone. Here, we engineered 63 synthetic CRMs to systematically assess the relationship between variation in the content and spacing of motifs within CRMs to CRM activity during development using Drosophila transgenic embryos. In over half the cases, very simple elements containing only one or two types of TF binding motifs were capable of driving specific spatio-temporal patterns during development. Different motif organizations provide different degrees of robustness to enhancer activity, ranging from binary on-off responses to more subtle effects including embryo-to-embryo and within-embryo variation. By quantifying the effects of subtle changes in motif organization, we were able to model biophysical rules that explain CRM behavior and may contribute to the spatial positioning of CRM activity in vivo. For the same enhancer, the effects of small differences in motif positions varied in developmentally related tissues, suggesting that gene expression may be more susceptible to sequence variation in one tissue compared to another. This result has important implications for human eQTL studies in which many associated mutations are found in cis-regulatory regions, though the mechanism for how they affect tissue-specific gene expression is often not understood.
| Transcription is initiated through the binding of transcription factors (TFs) to specific motifs that are dispersed throughout the genome. Genomics methods have helped to discern which motifs for a TF are occupied and which are not, yet it is poorly understood why certain combinations of bound motifs, and not others, drive specific patterns of expression. Here, we take a bottom-up approach to address this question: We constructed simple, synthetic elements containing motifs for only one or two TFs in different orientations and integrated them into the Drosophila genome. By assessing when and where these elements drive expression, we could model specific rules governing tissue-specific enhancer activity. Despite the general importance of TF combinatorial interactions during development, elements with a single TF's motif were often sufficient to drive complex expression. By combining motifs for two factors, we observed non-additive expression in the heart. While the enhancer's activity could tolerate changes in motif spacing and orientation in many tissues, the robustness of heart expression was very sensitive to subtle sequence changes. These results highlight an important property of enhancers—as their readout is context-specific, so too are the effects of mutations within them, including small insertions that may alter a gene's expression in one tissue, but not in another.
| Gene expression is initiated by the binding of transcription factors (TFs) to cis-regulatory modules (CRMs) such as enhancer elements, which give rise to specific patterns of temporal and spatial activity [1]. Recent years have seen a dramatic increase in the ability to identify the location of regulatory elements using genome-wide information on TF occupancy [2], [3], [4], [5], [6], [7], cofactor recruitment [8], [9] and chromatin modifications [10], [11], [12], [13], [14], [15]. These studies have identified thousands to hundreds-of-thousands of regulatory elements that could be potentially active at a given time and in a given cell-type during development. Given this extensive regulatory landscape, it has become an enormous challenge to decode how CRMs function in terms of their spatio-temporal activity. Detailed dissection of individual elements has revealed developmental enhancers whose function is dependent on the presence of individual TF motifs [16], [17], combinations of motifs [18], [19], [20], and the specific arrangement [21], [22], [23], [24], [25], [26] or not [2], [27], [28], [29] of those motifs. These examples have inspired much debate over the relative importance of each of these variables to enhancer function, but systematic rules to understand their contribution have not emerged.
While dissection of endogenous enhancers has proved to be an extremely powerful approach to understand enhancer function [19], [20], [29], [30], [31], [32], individual enhancers instantiate only one of many possible solutions that can give rise to a specific gene expression pattern [7], [28], [33], thereby limiting the range of functional rules that are generally explored. Synthetic elements offer the possibility to examine a wide range of motif compositions and motif positioning rules while ensuring, as much as possible, the neutrality of non-motif (i.e. spacer) sequences. Synthetic promoter-YFP libraries in yeast, for example, were used to quantify and model the effect of different promoter motif configurations on the levels of YFP expression [34], [35]. Similarly, synthetic constructs combined with massively parallel sequencing were used to dissect the relationship between DNA sequence and activity of constructs transiently transfected into cell lines [36] or injected into mouse tail veins [37]. These approaches offer the clear advantage of scale, as thousands of elements with different DNA sequence combinations can be examined simultaneously. However, they are limited by the simplicity of the read-out, which is a relative measure of the levels of the CRM's activity at a single time point or condition. For most developmental enhancers, the impact of sequence changes on the timing or tissue-specificity of gene expression is equally important. It is also not clear to what extent episomal DNA from transient transfection or tail-vein injections recapitulates the impact of chromatin context and nucleosome positioning on gene expression. In these respects, reporter assays in transgenic animals provide invaluable information. Although generally difficult to scale, multi-stage assays using stable transgenic embryos yield precise information about when and where an enhancer is active in an in vivo chromatinized context.
In this study, we systematically engineered 63 synthetic elements that differ with respect to the number and kinds of TF motifs they contain, as well as the relative spacing and orientation of these motifs. These elements were specifically designed to assess three properties of enhancer motif organisation relevant to metazoan development: (1) the ability of homotypic clusters of individual motifs to function as developmental enhancers, (2) the ability of combinations of different kinds of motifs (heterotypic motif clustering) to generate new emergent activity, and (3) the effect of changes in motif organisation, such as number, spacing and orientation, on the robustness of enhancer activity in both space and time. While many studies have focused on the effect of point mutations on enhancer activity, there is an enormous amount of structural variation within natural populations, including the sequence of Drosophila [38], [39] and humans [40]. To assess the influence of small insertions or deletions, we have systematically changed the spacing between motifs for four heterotypic pairs of TF motifs.
To examine these properties, we focused on ten motifs recognized by TFs that form part of a highly studied regulatory network that governs Drosophila mesoderm development [41], [42], including the downstream effector TFs of Wingless (known as Wnt in vertebrates) and Dpp (BMP in vertebrates) signaling pathways. We generated very simple elements consisting of six motifs in either a homotypic or heterotypic combination, which were stably integrated into the Drosophila genome. The resulting patterns of expression driven by these elements suggest a number of interesting features governing CRM function. First, despite the known importance of combinatorial activity to refine spatial patterns of expression, elements with multiple copies of an individual motif can drive complex patterns of expression. For example, although pMad is known to act cooperatively with lineage TFs in both Drosophila and vertebrates, it is sufficient to transduce Dpp activity when its motifs are in the appropriate configuration. Second, combining motifs for as few as two TFs can lead to novel emergent expression. Although the importance of cooperative DNA binding in the regulation of development has long been supported by other studies, the sufficiency of so few sites speaks to the extent to which even minor changes in cis-regulatory sequence can lead to the evolution of novel expression profiles. Third, the spacing and orientation of motifs is not only essential for enhancer activity in terms of binary on-off effects, but also has more quantitative effects on the robustness of gene expression, including inter- and intra- embryo variability. Fourth, these effects of motif organization, often referred to as motif grammar, are tissue-specific. The same ‘two-TF’ enhancer can function using very flexible motif spacing in one tissue, yet have rigid constraints in another, demonstrating an additional way in which the organization and function of CRMs acts to reduce the constraints of pleiotropy for regulatory mutations.
For each factor, synthetic CRMs were generated by combining six motifs, separated by a spacer sequence of defined length. Therefore, there were two aspects to the design of the synthetic elements; the choice of motif instance used for each TF and the sequence of the spacer. First, for the TF motifs, we selected high affinity motifs for each factor, as much as possible. TFs often recognize the same or highly similar sequences. This includes not only members of the same family of TFs, e.g. GATA factors (such as Pnr), but also TFs with apparently unrelated DNA binding domains, e.g. the bHLH factor Twist and the zinc finger TF Snail. While we cannot change this inherent property of TFs, we did try to increase the potential specificity of the motif (or word instance) used here for the particular TFs we are interested in by using motifs derived from in vivo occupancy data for nine of the ten factors [2], [7]. For clarity, we refer to each construct by the name of the TF from which ChIP data was used to learn the motif. However, the sequence of all motif instances used, as well as their similarity to other TF motifs is provided in Table S1.
Second, we tried to select a neutral spacer sequence that does not include known TFBS, based on our current knowledge. This is not trivial, as in addition to the spacer sequence itself, the border sequence bridging the spacer and the known motif (red bar, Figure 1A) can in itself create an additional binding site. To minimize this possibility, rather than using a common spacer, we computed an optimal spacer sequence for each combination of motifs to minimize the chance of inadvertently generating additional binding sites, based on current information (Supplemental Methods (Text S1)). During the course of this study, our results show that the spatio-temporal activity of the designed elements is primarily driven by the intended TF motifs and not from the spacer sequence, as indicated by two lines of evidence: (1) The concordance between the activities of homotypic and heterotypic elements for the same TFs, which were designed with different spacer sequences (described below) and 2) the very similar spatio-temporal activity of two heterotypic CRMs, which were specifically designed with identical TF motifs but with two different spacer sequences (pMad-Tin, see below).
Each synthetic CRM, which were on average 105 bp in length, was cloned into a common minimal lacZ reporter vector and stably integrated into the same location in the Drosophila genome using the phiC31 system [43], to allow for a direct comparison of enhancer activities. The ability of each CRM to drive spatio-temporal lacZ expression was assessed during all stages of embryogenesis by fluorescent in situ hybridization to examine the full regulatory potential of the DNA sequence.
The combinatorial binding of TFs provides complexity to regulate refined spatial patterns of expression, and forms the basis for logical operations within Gene Regulatory Networks driving development [44]. However, in addition to combinatorial activity, homotypic clusters of an individual TF's motif are also present in regulatory regions in the vicinity of developmental genes in both Drosophila [45] and vertebrates [46]. Although prevalent in vivo, the role of these clusters in regulating gene expression, and the properties governing how they function, is currently not clear. Studies examining the ability of clusters of motifs to regulate gene expression in transgenic reporter assays have had varied success: this includes (although not exhaustive) homotypic clusters of motifs that were sufficient to drive activity [47], [48], [49], [50], [51], [52] and those that were not [48], [53], [54], [55], [56]. However, as the elements in each of these individual studies were designed and tested in different ways, it is impossible to deduce any functional inference across studies.
We initiated this study by systematically examining the activity of elements composed of a cluster of six identical motifs. In total we tested homotypic CRMs with motifs for ten essential factors, encompassing multiple types of DNA binding domains: bHLH (Twist (Twi)), homeobox (Tinman (Tin), Bagpipe (Bap)), T-box (Dorsocross (Doc)), GATA zinc finger (Pannier (Pnr)), MADS box (Myocyte enhancer factor 2 (Mef2)), FoxF (Biniou (Bin)), HMG-domain (T-cell factor (TCF)), Ets-domain (Pointed (Pnt)) and the MH1 domain (pMad). Seven out of ten of these very simple elements were able to drive sequence-specific spatio-temporal expression (Figures 1B–F, 2, S1). In six cases, the expression profiles driven by clusters of a single motif were sufficient to partially (and in the case of two, almost completely) recapitulate the expression of the TF whose in vivo occupied motif was used to construct the CRM. For example, the synthetic CRMs containing Doc enriched and GATA motifs drove expression in the presumptive amnioserosa, cells where the endogenous Doc and pnr genes are expressed (white boxes, Figure 1B″,C″). Similarly the synthetic CRM containing Tin motifs drove expression in a subset of the dorsal mesoderm, colocalizing with the expression of the endogenous tin gene (Figure 1D″, arrow). Although multimerizing the preferred Twist E-box motif was not sufficient to drive mesoderm expression at early stages of development, this synthetic CRM could activate expression in the hindgut visceral mesoderm (VM) at later stages, overlapping the endogenous twist gene's expression (Figure 1E″, arrow).
The activity of two homotypic CRMs, containing motifs for either pMad (discussed below) or Bin, stood out as they were sufficient to recapitulate almost the entire domain of the TF's activity. For the Bin CRM, this included the foregut, midgut and hindgut VM (Figures 1F″, S2) through multiple stages of development. This result suggests that motifs for this FoxF factor are sufficient to regulate expression throughout the VM at multiple stages of development, consistent with Bin's essential requirement for VM development [57] and extensive enhancer occupancy [7], [58]. In contrast to the six CRMs that gave ‘expected’ activity, the activity of the homotypic dTCF CRM only partially overlaps wingless expression, (the ligand that activates the cascade leading to dTCF activation). The CRM's activity is restricted to segmental groups of cells that are in close proximity to, but not always adjacent to, the wg stripes (Figure S1A″), which may reflect more ‘long’-range signaling from Wg, or alternatively the activity of an additional TF that can occupy this dTCF motif.
Taken together, these results indicate that homotypic clusters of an individual motif are often sufficient to regulate specific patterns of spatio-temporal activity, analogous to the more commonly studied combinatorial elements. Multimerizing single TF motifs can provide remarkable specificity, as demonstrated by the non-random overlap of CRM activity with part or all of the TF's expression in six out of ten cases tested. Conversely, it is interesting that not all activator clusters are sufficient to drive expression, in contrast to what might be assumed from the activity of yeast GAL4 sites. For example, three synthetic CRMs (multimers of Pnt, Mef2 and Bap motifs) yielded no activity (Figure S1B–D). In the case of Mef2 and Twist, the lack of general mesoderm activity is surprising given that both TFs have very well characterised DNA binding specificities in both Drosophila and vertebrates. This diversity in CRM output may reflect inherent differences in the regulatory potential among different TFs or in their ability to act cooperatively in a homotypic manner to potentiate their activity.
The dpp gene, coding for the ligand of the Dpp signaling cascade, is expressed in the foregut, hindgut and midgut VM in parasegment 3 (PS3) and 7 (PS7) at stage 13/14 of embryogenesis [59]. dpp expression is restricted to these VM domains through the integration of activating inputs from Ubx and Bin [57], [59], and repression via Wg signaling [60]. The downstream effector TF of Dpp signaling is the phosphorylated form of Mad (pMad). Our synthetic CRM containing six pMad motifs could recapitulate almost the entire expression of dpp in the VM during these stages of embryogenesis (Figure 2A). Therefore, despite the complexity in the network of upstream factors regulating dpp expression, once dpp is expressed, sites for its downstream effector (pMad) alone are sufficient to activate enhancer activity in these sub-tissue domains. The spatial boundaries of the enhancer's activity are most likely refined through the action of Brinker (Brk). Brk is a transcriptional repressor whose expression is negatively regulated by Dpp signaling [61]. This results in cells with high levels of pMad and low Brk (Dpp responding cells, where our synthetic CRM is active) and those with high Brk and low pMad (neighbouring cells outside the Dpp signaling domain, where our CRM is inactive). As Brk and pMad can recognize the same motif [62], there is often direct competition between the two TFs to regulate enhancer activity, where the relative levels of both TFs serve to limit the spatial domain of Dpp target gene expression, as nicely demonstrated for the endogenous Ubx enhancer [62].
The sufficiency of pMad motifs alone to activate enhancer activity was unexpected given previous studies in both Drosophila [56] and vertebrates [63], [64], indicating that Mad proteins bind to enhancers cooperatively with other tissue-specific ‘lineage’ factors, and have little activity alone. We therefore further investigated the regulatory potential of pMad sites in isolation by directly comparing the activity of elements containing six, four or two motifs, in the same orientation and with the same spacer sequence (Figures 2, S3). The expression of lacZ in PS7 of the midgut VM (Figure 2) is particularly informative of pMad activity: the Dpp signaling cascade activates pMad in an autocrine [65] and paracrine [66] fashion, which likely results in the highest levels of pMad activation in VM cells in PS7 and lower levels in adjacent parasegments. The homotypic CRM containing six pMad motifs reflects this paracrine signaling, having a larger domain of expression covering neighboring cells compared to the dpp expressing cells (Figure 2A). In contrast, the synthetic CRM containing four motifs had more restricted activity to a narrow domain anterior and posterior to the dpp expressing cells (Figure 2B), while the CRM containing only two motifs was active only in the Dpp producing cells (autocrine signaling) (Figure 2C). Therefore, pMad sites alone, when present in close proximity, are sufficient to activate enhancer activity in the absence of specific lineage TFs. However, the extent of the activity is dependent on the number of available pMad motifs and on the balance between pMad and Brk concentration.
In addition to the number of cells in which the enhancer was active, we also observed a strong correlation between the number of sites and the strength of the CRM. In the amnioserosa, for example, the CRM with six pMad sites drove robust expression at stage 11 (Figure S3A″), while four sites resulted in reduced activity, and the CRM with two sites drove only very weak expression in the amnioserosa (Figure S3B,C). This trend was even more dramatic at stage 14, when the pMad concentration in the amnioserosa decreases [67]. While the CRM with six pMad sites drove strong activity at stage 14 (Figure S3D), the CRMs with either four or two sites were inactive in the amnioserosa at this stage. The number of pMad sites, thus, appears to be able to compensate for a decrease in the amount of accessible pMad protein at stage 14, presumably by providing a larger platform for cooperative binding.
Cooperative interactions between TFs can sometimes occur through direct protein-protein interactions (PPI), which may introduce constraints in the organization of the TFs' motifs within CRMs [1]. Taking advantage of the design flexibility of synthetic elements, we generated heterotypic CRMs composed of motifs for TFs known to exhibit protein-protein or genetic interactions: namely motifs for Tin with either Pnr, Doc, dTCF, or pMad [68], [69], [70], [71], to systematically explore two properties: 1) The influence of changes in the relative spacing and orientation of motifs on the robustness of CRM activity in different tissues, and 2) The ability of different combinations of TF motifs to generate emergent expression patterns in a particular tissue, not observed for CRMs containing only one kind of motif.
Each heterotypic CRM contained three pairs of TF motifs for either Doc-Tin, GATA-Tin, dTCF-Tin, or pMad-Tin. For each TF pair, on average nine constructs were tested in transgenic animals, in which the spacing and/or orientation of the Tin motif was systematically altered (Table S2). Heterotypic CRMs containing GATA-Tin or Doc-Tin combinations resulted in a pattern of expression that was largely the sum of the activity of each of the corresponding homotypic CRMs alone (data not shown). Changing the spacing and orientation of the sites (Table S2) had little or no effect on CRM activity, indicating a minimal role for motif positioning, or grammar, in these instances. The dTCF-Tin heterotypic CRMs did not have any activity, despite the fact that the homotypic CRMs containing six dTCF or Tin sites drove specific activity in segmental groups of cells in the ectoderm and mesoderm, respectively (Figures 1D, S1A). This lack of activity in the heterotypic context most likely reflects the reduction in the number of motifs used, from six in the homotypic constructs to three motifs for each factor in the heterotypic CRM. Correspondingly, homotypic construct containing four dTCF sites [53], [54] had no embryonic activity.
In contrast to the GATA-Tin and Doc-Tin heterotypic pairs, whose activity was robust to changes in motif organisation, the activity of the pMad-Tin heterotypic CRMs changed depending on the motif spacing. As discussed above, pMad motifs alone are sufficient to regulate expression in PS3 and PS7 of the midgut VM. In these cases, the strength of activity was dependent on the number of motifs present, with a dramatic reduction in expression observed going from four to two sites (Figure 2). A homotypic CRM containing three pMad sites separated by 13 bp also showed a dramatic reduction in VM activity (Figure 3A), a result that may stem from either or both the reduction in sites or the increase in spacing (from 6 bp to 13 bp). Interestingly, when a Tin motif is inserted within each 13 bp ‘spacer’ sequence, this pMad-Tin heterotypic design can restore activity in the midgut VM (Figure 3B, white square), giving a pattern of expression similar to the pMad homotypic CRM with 6 bp spacing. The ameliorative effects of the Tin sites were limited by distance. Increasing the spacing between adjacent pMad and Tin motifs from 2 to 8 bp, which increases the spacing between pMad motifs from 13 to 25 bp, caused a progressive reduction in lacZ expression in the midgut VM (Figure 3). This VM activity is primarily driven through the pMad and Tin motifs and not the spacer sequence, as changing the spacer sequence had a minimal effect on VM enhancer activity, while a similar distance effect was observed by altering the length of spacer between the two sites (compare Figures 3 and S5). These results suggest that the occupancy of Tin sites acts either to bridge the distance between pMad sites (up to ∼20 bp) facilitating cooperative pMad regulation or alternatively to facilitate pMad-Tin combinatorial regulation of VM expression.
To distinguish between these two possibilities, we examined the effect of altering the relative orientation of the sites, reasoning that if Tin occupancy acts indirectly to help pMad recruitment altering the orientation of the Tin site should have little effect on activity. However, we observed that changes in motif orientation did influence CRM activity: one orientation of the Tin motif (arbitrarily referred to as antisense (A)) typically drove stronger VM expression compared to CRMs with the motif in the opposite orientation (referred to as sense (S)) (Figure 3, compare C to G). This result indicates that the Tin motifs are not just neutral sequences bridging the spacing between homotypic pMad motifs, but rather suggest a specific mechanism of cooperative Tin-pMad DNA binding, and therefore transcriptional regulation.
Combining pMad and Tin sites resulted in CRM activity in the dorsal mesoderm, VM and amnioserosa –representing essentially the summation of the expression profiles of the respective homotypic CRMs (Figures 1D, 2A, S4). However, in contrast to GATA-Tin and Doc-Tin CRMs, whose joint expression profiles were largely additive, the heterotypic pMad-Tin CRMs were also sufficient to drive expression in a new domain, the developing heart (Figure 4). Cardioblast specification requires the action of a large number of TFs and signaling cascades (for review, see [72]). Tin, Pnr (GATA factor), and Doc together with Dpp and Wg signaling are essential for heart development [73], [74], [75], [76], [77], [78], a tissue in which these factors regulate each others' expression [70], [75], [76], [77], [78], [79], [80], and act as a collective unit to regulate enhancer activity [2], [27]. Given this complex regulation, we did not anticipate that a simple element containing only pMad and Tin sites would be able to drive expression in the heart. Importantly, the emergence of this pattern was highly dependent on the CRM architecture: Among our constructs, heart activity was only observed in CRMs containing three pMad and three Tin motifs placed in close proximity (within 2–4 bp) to one another (Figure 4B–E). In contrast, no heart activity was observed in CRMs where the spacing between motifs was increased to 6 bp or 8 bp.
Even in the ‘optimal’ motif configuration, we observed embryo-to-embryo variability, indicating that expression in the heart is significantly less robust (i.e. more influenced by stochastic events) than is expression driven by pMad-Tin constructs in the VM and other tissues. To better assess the robustness of these constructs activity, we made use of P-element transgenesis to integrate our CRMs into random locations in the genome. We reasoned that non-robust CRM activity would be more highly influenced by variation in chromatin context at different genomic positions compared to more robust expression profiles. Across these random-insertion sites, the pMad-Tin CRMs drove highly consistent expression in the VM, amnioserosa, and dorsal mesoderm, while heart activity in those same embryos varied dramatically as a function of genomic location (Table S3). In fact, the pMad-Tin A2 CRM was only active in the heart in the context of a transgenic fly line obtained using a P-element (Figure 4B), and not with the phiC31-mediated integration (Figure 3B). These results indicate that the heart activity of pMad-Tin CRMs is teetering on the edge of activation, being highly sensitive to both the motif context within the enhancer and the enhancer context within its chromatin environment. Thus, while the combined activity of these two TFs can give rise to emergent activity in the developing heart, this is not a robust mechanism to generate heart expression.
Examining CRM activity in the heart revealed that some CRMs exhibit varied activity among embryos, and even within an embryo, as demonstrated by the pMad-Tin S4 construct, which drove expression throughout the entire heart in some embryos (Figure 3G), but only in a posterior portion of the heart in others (Figure 4E). To assess this variability in a systematic manner, and to provide quantitative data to which we could fit a model explaining CRM activity (see below), we applied two measures of the robustness of CRM activity: (1) Penetrance, defined as the fraction of embryos within a population that show CRM activity in the relevant tissue (in this case the VM or heart), and determined by any spatial overlap of lacZ expression with dpp (VM) or tin (heart) at a defined developmental stage (Figures 5A, S6A). (2) CRM expressivity, defined as the fraction of tissue-specific regions within an embryo (i.e. the proportion of the VM or heart) that display CRM expression (Figures 5A, S6A). In the midgut VM, for example, there are four domains of CRM activity (Figure 5A). If the CRM is active in all four domains, it has an expressivity of 1, while activity in two out of four domains has an expressivity of 0.5. To minimize systematic error, we implemented automated image analysis (see Materials and Methods) of embryos at a consistent stage of development (Figures 5A, S6A). In using penetrance and expressivity as quantitative metrics of CRM activity, we avoided issues arising from directly comparing non-linearly amplified signals from standard in situ hybridization of lacZ levels between CRMs. Penetrance provides a reliable measure of enhancer embryo-to-embryo variability, while expressivity provides a readout of intra-embryo enhancer variability – both of which we exploit to assess the effect of motif organization on the robustness of activity in two tissues.
The CRM penetrance of around one hundred embryos was measured for each of the 8 pMad-Tin synthetic CRMs (Table S4). With this quantification, differences in activity became more striking in several regards. First, within a tissue, subtle differences in activity between CRMs with different motif configurations become clear. For example, in the antisense orientation, changing the motif spacing from 2 bp to either 4 bp or 6 bp spacing resulted in only a slight decrease in penetrance, from 1 to 0.91 in the VM. However, the expressivity of these CRMs was quite distinct. The CRM with a 4 bp spacing had an expressivity of 1, while this number was nearly halved (0.59) for the CRM with a 6 bp spacing (Figure 5B; Table S5). The extreme effect of a small (2 bp) change in motif spacing suggests that direct, and potentially cooperative, interactions among bound factors have been disrupted, leaving the enhancer's activity more prone to variation.
A second striking observation is the extent to which changes in CRM architecture impacts activity in a tissue-specific manner. This is made clear by relative differences in CRM penetrance between the VM and heart. In the antisense orientation, for example, CRM penetrance remains almost unchanged in the VM when the motif spacing between the pMad and Tin sites is changed from 2 to 6 bp (Figure 5B). In contrast, the penetrance in the heart drops from 0.89 at a 4 bp spacing to zero when the motifs are spaced by 6 bp. A similar dramatic effect was observed by flipping the orientation of the Tin motif from antisense to sense, at a 4 bp spacing, which caused the penetrance of heart activity to drop from 0.89 to 0.48 (Figure 5C). These switch-like transitions in penetrance indicate that heart activity can only occur with a very restricted motif organization, which relies on close proximity of all TF motifs.
Taken together these results highlight an important property of cis-regulatory activity in multicellular organisms: An enhancer element (e.g. composed of only two types of motifs as described here) can require a very restricted motif configuration to regulate expression in one tissue (heart), but yet be much more flexible in its motif organisation to drive robust activity in another (VM).
The extent to which cooperative interactions, including higher-order interactions across multiple proteins, contribute to enhancer activity is difficult to assess by simply visualizing expression patterns. Here, computational models can be extremely helpful to explore the effect of potential interactions on CRM function [33], [81]. To better understand the contribution of higher-order interactions among TFs on our synthetic CRMs, we used fractional site occupancy modeling (Figures S7; Methods (Text S1)), which describe DNA-protein and protein-protein interactions as thermodynamic processes, an approach that has been successfully used to understand other regulatory elements [82], [83], [84], [85], [86], including Drosophila enhancers [26], [87].
As fractional site occupancy models analyze the probability of every possible configuration of binding events, the complexity of the models increases exponentially with the number of TFBSs. To avoid these complications, we aimed to identify the simplest model that recapitulates the observed CRM activity. In line with our observations (as seen in Figures 3 and 5), and with research in another tissue [71], the model assumes the presence of direct cooperative interactions between neighbouring pMad and Tin proteins. Protein-protein interactions were modeled as the extent of overlap between spheres of “interaction space” around each bound protein, with sense and antisense orientations having different effective spherical radii (Figure S7, and includes parameters for the strength of possible cooperativity between bound TFs (Supplemental Methods (Text S1)). The model specifically explores whether additive interactions between Tin-pMad pairs are sufficient to recapitulate the observed experimental data or if potential ‘higher order’ cooperativity among pMad-Tin pairs with nearby pMad or Tin bound proteins are required to drive robust CRM activity.
For VM activity driven by the pMad-Tin heterotypic CRMs, a model that includes an additional degree of cooperativity beyond pMad-Tin pairs fit the data better compared to a model in which pMad-Tin pairs act in an independent additive manner (Figures 6A, S8A). A key difference between the models lies in their predictions of the robustness of shorter CRMs: if higher order interactions are central for robust CRM activity then shorter CRMs will have a sharper decrease in robustness compared to pMad-Tin interactions alone. To experimentally test this, we halved the size of two heterotypic CRMs, generating CRMs with pMad-Tin-pMad motifs in S2 and A4 configuration. Such small CRMs could drive expression in the midgut VM (Figure 6B,C), albeit at a reduced level. In contrast, CRMs with only one pMad and Tin site, representing the smallest possible cooperative binding configuration (a pMad-Tin pair), drastically reduced all VM activity (Figure 6D,E). Incorporating higher level cooperativity into the model, without any further fitting, significantly improved the quality of the prediction of the shortened CRMs (Figures 6F,G, S8B). This suggests that pMad-Tin-pMad is the minimal configuration essential for robust VM expression. Importantly, this model was robust to a ‘leave-two-out’ cross validation iterated over all possible orderings of the 12 CRMs, arguing against over-fitting (Figure S8C). Finally, to test the model's prediction that additive interactions between pMad-Tin pairs with other bound pMad or Tin proteins are insufficient to drive robust VM expression, we tested the activity of a CRM with the motif configuration Tin-pMad-Tin. This CRM had very weak VM activity (Figure S8D), consistent with a requirement for higher order interactions between multiple pMad-Tin pairs for robust CRM activity (Figure 6G). In summary, increasing the complexity of TF cooperativity resulted in significantly improved consistency with experiment compared to considering only independent pMad-Tin cooperative pairs (Figure 6G).
Next, we addressed how heterotypic pMad-Tin CRMs lead to activity in the heart. Simple models with cooperativity between bound Tin and pMad can recapitulate the observed penetrance and expressivity in the heart for CRMs (Figures 6H, S8E,F). However, only the model including higher-order cooperative interactions between three neighboring units of Tin-pMad-Tin was consistent with both the shorter constructs and the six TF motif CRMs (Figures 6H, S8E,F). We note that the true minimal motif arrangement to generate robust heart activity is likely to be much more complex. In line with this, the Tin-pMad-Tin CRM has no heart activity. Interestingly, the effective range of cooperative TF interactions learned by the model for the heart was considerably lower (<5 bp) than for the VM (<9 bp) (Table S6).
In summary, while our experimental data is suggestive of cooperative TF interactions being likely necessary for CRM activity, the modeling has formalized this and systematically identified the range of interactions and the likely minimum level of higher-order TF cooperativity required for activity in both the VM and heart. Taken together, the modeling provides regulatory rules that explain how the same two TF motifs can give rise to activity in two different tissues (VM and heart) depending on the motif organization within the CRM.
In this study, we stably integrated simple synthetic CRMs into transgenic Drosophila embryos, and then combined a quantitative analysis of enhancer activity with fractional site occupancy modeling to determine the contribution of motif organization to activity in two tissues. The results of these analyses highlight a number of organizational features that contribute to an enhancer's activity during development.
While quantifying the activity of a simple ‘two-TF motif’ CRM (pMad-Tin), our results show that enhancer activity can exhibit very different sensitivity to motif organization in one tissue compared to another (Figure 7). Several mechanisms could account for this interesting effect, including different concentrations of the TF (i.e. pMad or Tin) in the different tissues, the availability of tissue-specific co-factors, or tissue-specific priming of the enhancer, which may increase the ease by which the enhancer is activated.
An elegant dissection of the endogenous spa enhancer demonstrated that completely rearranging the relative order and spacing of TF binding sites could switch its cell type-specific activity from cone cells to photoreceptors in the eye [20]. In comparison, the changes in motif organisation introduced in our study were much more subtle such that the relative order of motifs was completely preserved. Yet only changing the spacing or orientation of motifs altered the robustness of enhancer activity in a tissue-specific manner. This result indicates that small insertions or deletions in CRMs, that do not affect the TF motifs themselves, could still have significant effects on gene expression in one tissue while having no effect in another. A study examining the activity of neuroectoderm enhancers between Drosophila species supports this model, where reduced spacing between Dorsal and Twist sites results in broader neuroectodermal stripes of CRM activity, while increased motif spacing resulted in progressively narrower stripes [88]. Studies of both endogenous enhancers and the synthetic CRMs described here provide compelling evidence that the exact positioning of motifs within CRMs is crucial for the robustness of their activity in one tissue, while it may be largely dispensable in another. Different cell types can therefore interpret the same motif content of a given enhancer in different manners.
The Drosophila heart is composed of two cell types, cardioblasts and pericardial cells, each of which requires the integration of many regulatory proteins for proper specification and diversification [72]. A characterized pericardial enhancer, eve MHE, for example, contains pMad and Tin binding sites in addition to sites for dTCF, Twi, Ets proteins, and Zfh1 [17], [89]. Given this complexity, it was surprising that a simple element built from pMad and Tin sites alone was sufficient to drive expression in the heart, albeit at a later developmental stage. Our analyses indicate that this activity is due to cooperativity binding between Tin and pMad, facilitated by a very specific motif arrangement. Using crystal structure data from close homologues of pMad [90] and Tin [91], we modelled the two TFs interaction on DNA, using a similar range of motif spacing (Figure S9). This 3D structural model indicates that it is possible for the DNA binding domains of these two proteins to both bind to DNA at a 2 bp spacing and to physically interact at a 2 bp and 4 bp spacing, but not at 6 bp spacing. Although done by homologue mapping, this structural data is consistent with our functional analyses of CRM activity, and further supports direct DNA binding cooperativity between these two TFs.
It is interesting to note, that although pMad and Tin sites are sufficient to drive expression in the heart from stage 13 to 14 (when placed in a limited motif arrangement), nature appears to use other enhancer configurations to regulate this critical function. There are two important aspects to this finding. First, heart activity arising from CRMs containing pMad and Tin sites alone is not robust. The enhancers are on ‘the edge’ of activation, where subtle changes in motif positioning or enhancer location switch activity between embryos and within embryos. Second, endogenous enhancers that are bound only by pMad and Tin – with no known input from other factors – direct expression in the dorsal mesoderm and not in the heart, at stage 10 [2], [92]. In our synthetic situation, pMad and Tin sites also drive robust expression in the dorsal mesoderm, in addition to variable weak expression in the heart. Therefore, although pMad and Tin sites alone are sufficient to drive heart activity in limited motif contexts, this mechanism is most likely not robust enough to be generally used to drive heart expression in vivo. This is consistent with recent studies showing that heart enhancer activity is elicited by the collective action of many TFs, which can occupy enhancers with considerable flexibility in terms of their motif content and configuration [2], [27]. Our pMad-Tin synthetic elements uncovered a very simple, although not very robust, alternative mechanism to regulate heart activity, and represent a nice example of how combinatorial regulation can lead to emergent expression profiles more than the simple sum of its parts.
The expression of key developmental genes is generally buffered against variation in genetic backgrounds and environmental conditions. This may occur at many levels including RNA polymerase II pausing [93], [94] and the presence of partially redundant enhancers [95], [96], [97], [98]. However, robust expression may also be buffered by the motif content within an enhancer to ensure a stable regulatory function. CRMs, for example, often include additional binding sites to those that are minimal and necessary [99]. In the context of the pMad-Tin synthetic CRMs, the motif organization can also act to ensure robust activity. Our results demonstrate that even in situations where the composition of motifs and their relative arrangement are maintained, subtle changes in the spacing between the motifs could have dramatic effects on enhancer output. Interestingly, this effect seems to be very tissue-specific, with some tissues maintaining robust activity whilst others lost all enhancer activity.
Taken together, the data presented in this study demonstrate that subtle alterations in motif organization can affect the ability of different tissues to ‘read’ an enhancer, which in turn may allow each tissue to fine-tune enhancer activity based on fluctuations in its molecular components.
Binding affinity models (PWMs) for Twi, Tin, Mef2, Bap, Bin, dTCF, pMad, and Doc2 and GATA were derived from ChIP-chip data analyses [2], [7]. The model for Pnt was generated using published footprints (Supplemental Methods (Text S1)). PWMs were first trimmed on each side to remove positions with an information content (IC) of less than 0.4 (trimming stopped at the first IC position > = 0.4). The sequence that best fits the PWM model was then determined for each trimmed PWM and is referred to hereafter as “TFBS”. All TFBS sequences used to design the synthetic CRMs are available in Table S1. For each CRM, a ‘neutral’ spacer sequence (a linker sequence placed between motifs) was heuristically determined by minimizing the sequence affinity for known TF PWM models (Supplemental Methods (Text S1)).
Synthetic CRMs were generated from long oligonucleotides synthesized by Eurofins MWG Operon with compatible cohesive ends upon annealing for cloning. The forward and reverse strand oligonucleotides were phosphorylated, annealed and subsequently ligated into the pDUO2n [7], to generate stable, transgenic Drosophila lines using the phiC31 site-specific integrase [43]. The pH-Pelican [100] vector was use to test the robustness of enhancer activity at different genomic locations by random P-element transgenesis. The sequence of each CRM was verified to ensure that there were no synthesis errors and is provided in Table S2.
CRM activity was assessed in embryos from transgenic flies using fluorescent in situ hybridization as described previously [101]. The following ESTs or full length cDNAs from the Drosophila Gene Collection (DGC) were used to generate probes: RE13967 (bap), RE40937 (doc2), RE20611 (dpp), GM04312 (dTCF), SD02611 (pnr) and AT15089 (twi). cDNAs used for bin and tin, lacZ, Mef2 and pnt were generous gifts from M. Frasch, U. Elling and M. Taylor respectively. Double or triple in situ hybridizations were performed using anti-fluorescein-POD, anti-DIG-POD and anti-biotin-POD antibodies (Roche, 1∶2000 dilution) and were developed sequentially with Cy3, fluorescein, and Cy5 tyramide signal amplification reagents (Perkin Elmer TSA kit). The lacZ expression patterns were imaged using Zeiss LSM 510 FCS or LSM 510 META confocal microscopes with A-Plan 10×/0.25 PH1 objective.
Background subtraction of both the CRM and tissue-specific channels was performed using a morphological opening with disk size greater than the largest relevant VM region (typical disk size of 25 pixel radius). The tissue-specific subset of images (e.g. dpp or tin in situs) were segmented using the Ilastik software package (www.ilastik.org). The segmented images were analysed using Matlab. The segmented regions in each image were smoothened by performing dilation (disk size of 5 pixel radius) followed by equivalent erosion. An area threshold (>200 pixels) was used to remove small, segmented regions. Finally, the perimeter of the segmented regions was calculated (using Matlab function bwperim) and overlaid onto the CRM expression data.
The penetrance was calculated for approximately 100 embryos for each of the twelve lines (Table S4). Bootstrapping was used to estimate the error in the penetrance measurements. The expressivity was calculated from around 16 carefully staged and positioned embryos (based on morphology and markers for VM (dpp) and heart (tin) tissues) for each line (Table S5). Embryos aligned dorsally were imaged and four regions of midgut VM were assigned, as shown in Figure 5A. The observed heart expression (also viewed dorsally) occurs in two rows of cells along either side of the embryo. We separated each row into an anterior and posterior segment, resulting in four heart regions (Figure S6A). The posterior segments, to the right of the PS7 VM region, correspond roughly to the heart proper, while the defined anterior heart segments correspond roughly to the region often referred to as the ‘aorta’. The penetrance in both the VM and heart was therefore measured as signal in one to four different regions of the tissue.
A fractional occupancy model was used to analyze the experimental data [102]. Our methodology was similar to other thermodynamic models used to understand CRM activity in Drosophila (e.g. [26], [87]). The model had at most four parameters: two parameters described the relevant protein-protein interactions; and two parameters were used to distinguish sense and antisense binding effects. Mathematical details are provided in the Supplemental Methods (Text S1).
This work is carried out in Drosophila, and was conducted in compliance with EMBL's guidelines.
|
10.1371/journal.pcbi.1005101 | Processing Oscillatory Signals by Incoherent Feedforward Loops | From the timing of amoeba development to the maintenance of stem cell pluripotency, many biological signaling pathways exhibit the ability to differentiate between pulsatile and sustained signals in the regulation of downstream gene expression. While the networks underlying this signal decoding are diverse, many are built around a common motif, the incoherent feedforward loop (IFFL), where an input simultaneously activates an output and an inhibitor of the output. With appropriate parameters, this motif can exhibit temporal adaptation, where the system is desensitized to a sustained input. This property serves as the foundation for distinguishing input signals with varying temporal profiles. Here, we use quantitative modeling to examine another property of IFFLs—the ability to process oscillatory signals. Our results indicate that the system’s ability to translate pulsatile dynamics is limited by two constraints. The kinetics of the IFFL components dictate the input range for which the network is able to decode pulsatile dynamics. In addition, a match between the network parameters and input signal characteristics is required for optimal “counting”. We elucidate one potential mechanism by which information processing occurs in natural networks, and our work has implications in the design of synthetic gene circuits for this purpose.
| From circadian clocks to ultradian rhythms, oscillatory signals are found ubiquitously in nature. These oscillations are crucial in the regulation of cellular processes. While the fundamental design principles underlying the generation of these oscillations are extensively studied, the mechanisms for decoding these signals are underappreciated. With implications in both the basic understanding of how cells process temporal signals and the design of synthetic systems, we use quantitative modeling to probe one mechanism, the counting of pulses. We demonstrate the capability of an Incoherent Feedforward Loop motif for the differentiation between sustained and oscillatory input signals.
| From Ca+2 signaling to coordination of cell fates, oscillatory signals are essential to regulation of cellular processes [1–4]. The dynamic properties of such signals are crucial for controlling behaviors of single cells and cell populations [5]. As such, the mechanisms underlying the generation of these signals are well-established [2, 6, 7]. For instance, the network constraints governing the circadian clock elucidate design principles dictating the generation of both natural and synthetic pulses [8–10]. Some general requirements for the generation of oscillations include ‘nonlinear’ reaction rates and negative feedback [9]. A systems-level approach to oscillation characterization examines the topologies in natural systems that give rise to pulse generation [9]. This demonstrates the necessity of ‘nonlinear’ kinetic rate laws for the destabilization of the steady state in the generation of oscillations [9]. While this constraint allows the generation of pulses with a diverse set of network motifs, negative feedback (especially negative feedback with a time delay) is found in all these cases. This component is used to reset the network to its initial state [2, 9]. Engineered systems based on such design constraints demonstrate the capability to generate synthetic oscillators mimicking those found in nature [6]. Even in the absence of any apparent regulation, transient oscillations in gene expression can emerge from cell-size control [11].
Despite the ubiquity of oscillations in biology, much less is known about how cells process these signals. In particular, how do cells distinguish between oscillatory and sustained inputs? For a given oscillatory input, how do cells retrieve encoded information from the frequency and amplitude? For signal processing in the frequency domain, computational methods illustrate one potential mechanism, where a critical frequency defines the bandwidth for high fidelity signal propagation for each network [3]. This capacity can be changed with an increased oscillation amplitude or with increased kinetic rates. Regardless of the strategies that give rise to signal encoding, it is important to further understand how cells process oscillatory signals.
Many natural biological networks exhibit the ability to distinguish oscillatory and sustained signals. While several studies describe the contrasting downstream phenotypes, the architectures that give rise to such outcomes remain unclear. One common motif shared by such networks is the Incoherent Feed-Forward Loop (IFFL), in which an input both activates and represses a single output (Fig 1A) [4, 12, 13]. For example, oscillations in the transcription factor Ascl1 play a critical role in driving the proliferation of multipotent neural progenitor cells (NPCs) [14, 15]. In contrast, the sustained expression of Ascl1 promotes neuronal fate differentiation in NPCs [15, 16]. In social amoeba Dictyostelium discoideum, 3’ 5’-cyclic adenosine monophosphate (cAMP) oscillations result in the optimal gene expression for development while continuous stimulation inhibits transcription (Fig 1B) [17–19]. cAMP directly induces the expression of contact site A gene (csaA) while repressing the transcriptional activity of GtaC on csaA [17]. Additionally, the number of cAMP pulses regulates the coordination of development.
An IFFL also underlies the regulation of neuronal regeneration by cAMP response element binding protein (CREB). CREB activates regeneration-associated genes (RAGs) and nuclear factor interleukin-3-regulated protein (NFIL3), which negatively regulates RAGs (Fig 1C) [20]. In Drosophila, the constitutive activation of CREB upregulates a single RAG, Arg1, while CREB oscillations induced by neuronal injury increase the transcription of all members [20, 21].
The IFFL motif is also found downstream of p53 in the regulation of cell fate decisions. The tumor suppressor p53 oscillates in response to γ-radiation, and the number of oscillations increase the level of damage. In contrast, UV radiation induction results in a sustained pulse of p53 [22]. While p53 oscillations lead to cell cycle arrest, sustained p53 activation induces apoptosis [23, 24]. Here, p53 activates P53 upregulated modulator of apoptosis (PUMA), which plays a role in the induction of apoptosis, and Slug, a transcription factor which represses the activity of PUMA (Fig 1D) [25].
In mammalian cells, pulsatile dynamics of extracellular signal-regulated kinases (ERK) in response to epidermal growth factor (EGF) result in proliferation [26–28]. However, nerve growth factor (NGF) induces sustained ERK, driving differentiation [16]. Here, the transcription factor Myc and the cyclin-dependent kinase inhibitor p21 are both activated by ERK [29, 30]. Myc, an inducer of proliferation, represses the activity of p21, an inhibitor of cell cycle progression (Fig 1E) [31–35].
Tumor necrosis factor-α (TNFα) can stimulate oscillations in the transcription factor NF-κB, which in turn triggers nonspecific inflammatory response genes in immune cells [36]. In addition, persistence of the NF-κB oscillations dictates the resulting transcriptional profile [37, 38]. In contrast, bacterial liposaccharides (LPS) promote sustained NF-κB activation, which induces a LPS-specific expression program [39]. We identify a potential IFFL where NF-κB activates the expression of both IL8, a cytokine, and BCL3, a competitor for NF-κB target promoters including IL8 (Fig 1F) [40].
The common occurrence of IFFLs in these networks suggests a role in processing transient or oscillatory signals. This motif is ideal for decoding oscillatory signals due to its ability to react to fold-changes rather than the absolute concentration of an input [12, 13, 41, 42]. In addition, IFFLs are one of the major core topologies that can drive temporal adaptation, the generation of a pulse in response to a sustained input [43]. In essence, this network is able to desensitize a system to a sustained stimulus while maintaining the ability to respond to periodic stimulation. This property is the fundamental principle behind the ability to differentiate pulsatile and sustained signals.
Here, we examine a mechanism by which IFFLs process oscillations and the conditions that facilitate counting of pulses. This network exists in two states—one in which the system resets when the counting mechanism fails due to a conflict between the signal requirements and the actual input, and one where pulses result in a stepwise increase of the output. To differentiate between sustained and oscillatory signals, the network must display both states. The system must exhibit a stepwise increase of the output in response to oscillatory signals while having no response to sustained signals. Analogous to the use of radios to transform information from the frequency and amplitude domains of electromagnetic waves, the IFFL acts as a decoder, recognizing pulsatile signals.
We model an IFFL consisting of three components (Fig 2A): an input node (S), an intermediate node (X), and an output node (R). S activates the production of both X and R, while X induces the degradation of R through Hill kinetics. The dynamics of the motif can be described with two dimensionless ordinary differential equations (ODE):
drdτ=βΦ(τ)−(γRxnxn+1+γo)r,
(1)
dxdτ=βΦ(τ)−x,
(2)
where Φ(τ) represents the input signal, which can be sustained, oscillatory, or transient. Unless noted otherwise, we define Φ(τ) as a periodic square wave function with k pulses: Φ(τ) = 1 when iT + τo ≤ τ < iT + D + τo and Φ(τ) = 0 otherwise. Here, i is the index of a pulse (0 ≤ i < k), τo is the start time of the first pulse, T is the cycle period, and D (< T) is the pulse duration. For all simulations we assume that τo = 1. x denotes the concentration of X, r denotes that of R, and τ denotes time. γo represents the basal degradation rate constant of R, γR is the maximal induced degradation rate of R by X, and β is the maximal synthesis rate of X and R. We require that the induced degradation of R have an ultrasensitive dependence on X, as indicated by a high Hill coefficient (n = 110). Such ultrasensitive dependence can be achieved with oligomerization, tandem binding sites, increased cascade lengths, covalent modification cycles, or titration by inhibitors [44–48]. For titration, competitive binding of the inhibitor sets an activation threshold to be overcome, thereby increasing the Hill coefficient.
We consider Φ(τ) with varying durations and cycle periods: a longer duration corresponds to a more sustained signal. We first consider the case where the degradation of R is solely induced by X (γo = 0). Fig 2B illustrates typical time courses of X and R in response to two different oscillatory signals. In either case, X peaks in response to each individual pulse and returns to a basal level before the subsequent pulse. When the duration is sufficiently small (Fig 2B: left column), r exhibits a linear stepwise increase with an increasing number of pulses (for a constant β and pulse duration). For a long duration (Fig 2B: right column), however, R exhibits transient pulses similar to that of X. That is, only an input with a sufficiently short duration generates a more sustained output. The critical determinant of these divergent outcomes is the dynamics of X, which triggers the effective degradation of R when it reaches its half-activation threshold (x = 1). For a sufficiently short duration, x never reaches 1, and thus is unable to trigger the degradation of R. As a result, r increases with each additional input pulse. In contrast, for a long duration, x exceeds 1 for each pulse, leading to the periodic resetting of R.
For an input signal with short pulses, R can count the number of pulses: the level of R at a fixed time point (r(τE)) is approximately proportional to the number of pulses for a fixed pulse duration and amplitude (Fig 2C). A least-squares linear regression on r(τE) versus the number of pulses provides a quantitative measure for the quality of counting. The slope of this line measures the strength of the response. For instance, the slope is near zero for the signal with long pulses (S2 Fig), indicating no response (Fig 2B). The quality of the fit, as measured by R2, quantifies the performance of counting. In all our simulations, we consider R2 = 0.99 as the cut-off for high-quality counting. With 0 < R2 < 0.99, low-quality counting emerges as r saturates (reaches a steady state) over time. This characteristic impairs the ability to predict r(τE) with an increasing number of pulses. In addition, the absolute failure of counting occurs when R2 = 0.
In the base model (Fig 2), we assume that R is degraded solely by X. In general, however, a basal level degradation of R can deteriorate the counting quality (Fig 3A). A highly unstable reporter (γo = 0.05; Fig 3B) decreases the counting quality such that R2 < 0.99. The decreased slope also indicates the weakened response to the oscillating input. This loss of counting is attributed to the decreased timespan for which r can be maintained in the absence of the input. Therefore, to reduce the impact of endogenous degradation of the reporter on the counting quality, γo can be minimized through the use of a highly stable reporter.
Relaxing this assumption, we reach the same conclusions for reliable counting when γo is sufficiently small: our simulations demonstrate that the counting capability can be maintained in the general condition (γo > 0), like in the base case (Fig 4). IFFL kinetics constrains the ability to process oscillatory signals; therefore, we predict that both the production rate (β) and the induced degradation rate of the output (γR) are core parameters. In Fig 4A, we evaluate the range of the counting capability with respect to two parameters, β and D (sample time courses shown in Fig 4B). Even in the presence of basal degradation (γo = 0.01), counting can be maintained for appropriate combinations of β and D. We observe that as β increases, the range of durations for which counting occurs decreases while increasing the slope for different numbers of pulses (for R2 ≥ 0.99). On the other hand, increasing D increases the slope in cases when R2 ≥ 0.99 due to the increased time for R production.
To differentiate between pulsing and sustained inputs, X must overcome the activation threshold in the case of a sustained input. Therefore, the network cannot discern signals when β ≤ 1 because the threshold can never be reached. In this case, the network is unable to display the two necessary states to discern signals, illustrated by R2 ≥ 0.99 at all pulse durations (Fig 4A). However, when β > 1, the system has the potential to discriminate inputs (illustrated by a range of durations where R2 = 0 and a range where R2 ≥ 0.99) and the pulse duration restricts the counting capability. To count, the duration of input’s OFF state in each cycle (T − D) needs to be long enough to reset the basal level of the intermediate such that the subsequent pulse will not push x to 1. Generally, shorter pulse durations are more likely to exhibit this capability due to the longer duration of the OFF state (Fig 4B). The boundary between counting and no counting is determined by whether each pulse can drive X to the threshold, leading to complete degradation of R. The level of X at the end of the first pulse is (1 − e−D). Therefore, the boundary between the two regions is set by β=11−e−D, which approximates the border observed in numerical simulations (Fig 4A).
For a small γo, another critical parameter for counting is the rate constant for induced degradation, γR. Fig 4C and 4D depicts the case when β > 1 with a varying γR and pulse duration for a constant cycle period. With a constant β, we expect the counting range to remain constant. Instead, numerical simulations demonstrate a monotonically decreasing trend of the counting range as γR increases. Above a point, increasing γR has no impact on counting because a higher γR provides no additional benefit on the ability to degrade R in the case of a sustained input. Below this point, however, the range of pulse durations that can count increases, but the ability to differentiate pulsing and more sustained inputs diminishes due to a higher range of low-quality counting (0 < R2 < 0.99). For a sufficiently low γR, the system cannot discriminate between the two types of inputs (Fig 4C). We attribute this characteristic to the decreased impact of the intermediate on the degradation of the reporter, which is required for maintaining temporal adaptation.
The kinetics of the system prevent the complete degradation of x to 0 after each pulse, leading to the progressive accumulation of X with each additional pulse. Starting from a basal level, X reaches the maximum concentration (peak) at the end of the pulse, and returns to a new basal level before the next pulse:
xbasal,i=xpeak,i−1e−(T−D),
(3)
xpeak,i=β−(β−xbasal,i)e−D,
(4)
where xpeak, i is the peak concentration of X after the ith pulse, xbasal, i is the basal concentration of X before the ith pulse, and T = cycle period. To start, xpeak, 0 = 0 while both xbasal, i and xpeak, i increase alongside i. Counting fails at pulse nc, when xpeak, nc > 1. We term nc the counting capacity of an IFFL for a given input signal, which sets the maximum number of pulses the network can count. As nc approaches ∞, the network can maintain counting when the negligible accumulation of X allows for indefinite counting. This exists when the pulse duration satisfies 1−e−D1−e−T<1β. In addition, as T approaches ∞, the analytical border converges to 1 = β(1 − e−D) such that counting is maintained when 1 > β(1 − e−D).
Fig 5 shows that the duration at which counting exists decreases with increasing β. Consistent with the numerical analysis where slopes are generally higher at shorter pulse durations with an increasing β, higher synthesis rates of X shorten the duration range for which counting can occur (Fig 4). Thus, a higher β means that X will reach the threshold at shorter pulse durations, resulting in the loss of the counting capability beyond this duration. The curve of the boundary signifies that increasing the cycle period has little impact on the duration at which counting is lost (Fig 5). More specifically, at the analytical border, the duration at which 1−e−D1−e−T=1β is maintained relies on β. However, at low periods, the duration at which counting exists relies more heavily on the cycle period. As the pulse duration approaches the cycle period, there is less time for X to return to the basal value before subsequent pulses. This effect has a greater impact in the case of low cycle periods. Although the basal level of X will never return to the initial value, if x remains below the activation threshold, negligible increases can maintain the counting capability.
The analytical solution 1 > β(1 − e−D) defines the conditions under which indefinite counting occurs based on the peak concentration of the intermediate. When γo > 0, however, the number of pulses any particular system can count becomes finite due to the saturation of the reporter. This is described in S1 Text, which gives the final level of R at time τE (> (k − 1)T + D + τ0) after any arbitrary number of pulses (k > 0) when 1 > β(1 − e−D) as r(τE)=βγo(1-e−γoD)∑i=0k−1e−γo(τE−τ0−(iT+D)). τ0, the time at which the pulses start, can impact the parameter range with the counting capability. Depending on γo, a series of pulses that start at the beginning of the simulation interval can have a different r(τE) than those that start near the end of the interval. To analyze the effect of the other parameters, we constrain our analysis by assuming a constant τ0 for all simulations. Here, r(τE) is a monotonically increasing function of the number of pulses, regardless of the other parameters (as long as X does not accumulate beyond its threshold to trigger the degradation of R). When γo = 0, r(τE) is proportional to k: r(τE) = kβ. However, the dependence of r(τE) on k deviates from the proportionality as γo increases. This is consistent with the numerical simulation results illustrated by Fig 3.
The key constraint for IFFLs to distinguish oscillatory signals resides in its ability for temporal adaptation where the network responds with a pulse to a sustained input. As long as an IFFL maintains this characteristic, the motif can respond to periodic stimulation and act as a pulse counter. Each time the input is removed, the intermediate is reset and the system can respond to subsequent pulses. We illustrate this principle by considering alternative implementations of the IFFL satisfying this constraint.
We examine the reliance of counting on the current architecture by shifting the ultrasensitive response to the production of X through a positive feedback loop (S1 Text). Like the initial model, this alternative model maintains the ability to count in the base case with no endogenous R degradation (S3 Fig). Here, when β is large, the linear dependence of the degradation of R on both X and R causes any X to induce the degradation of R, preventing counting. However, when β is small, R degrades slowly because of low concentrations of X, preventing the differentiation between sustained and pulsing inputs. Like in the initial model, γo dictates the length of time during which R remains constant and should be minimized to maximize the duration range in which counting can occur. For a large γR, the presence of a small concentration of R or X will lead to the efficient degradation of R, preventing the ability to count. Meanwhile, a small γR will be unable to induce the effective degradation of R even at high concentrations of X. To demonstrate the robustness of this counting mechanism, we develop additional models to take into account common properties of biological systems including time delay, modifications to the current architecture, and noise (S1 Text; S1, S3, S5, S8 and S9 Figs).
Here, we analyze an underappreciated phenomenon in information processing by signaling networks: processing of oscillatory signals. In this context, we illustrate the unique ability of incoherent feedforward loops (IFFL) with proper network parameters to distinguish transient (or oscillatory) and sustained signals. For oscillatory signals, we analyze the extent to which the network is able to count the number of pulses. Additionally, we establish the dynamic constraints of the input signals and network parameters such that counting can occur. With the appropriate parameters, the ability of an IFFL motif to respond to transient signals but not to sustained signals relies upon the property of temporal adaptation in response to a sustained signal. Insights from our model can be used to identify natural information decoding systems and understand the effect of perturbations to such networks.
Our results can be applied to GtaC shuttling in social amoeba, which differentiates between pulsing and sustained cAMP and matches optimal signals to the ideal transcript accumulation. Cai et. al. show that when the network regulating CsaA is subjected to high frequency inputs, GtaC is unable to respond to each pulse, resulting in a decreased transcript accumulation. Similarly, a lower than optimal (natural) cAMP frequency results in a decrease in the overall transcript concentration. This phenomenon suggests the presence of a band-pass filter. As in Fig 5, this system illustrates the idea that a decreased frequency (higher period) corresponds to a reduced fraction of the period during which counting can occur. Here, the parameters of the endogenous circuit dictate the range of frequencies that can be decoded resulting in dynamics similar to those explained by our IFFL model.
In a similar manner, the frequency of ERK activation has been implicated in cell-fate determination. While previous studies have demonstrated the dependence of differential cell fates (differentiation or proliferation) on specific growth factors, Ryu et. al. illustrate that this decision relies more heavily on the frequency of ERK activity than on the growth-factor identity [49]. Unlike with the sustained application of growth factors, synthetic pulses provide control over the frequency of the ERK activity. This results in differentiation at intermediate frequencies and proliferation at high and low frequencies with a growth factor traditionally known to solely induce proliferation. While this shift in cell fates has been shown experimentally, the mechanisms underlying this frequency decoding are still unknown.
Evolution may have selected for an optimal pulse frequency for social amoeba and in mammalian cell differentiation for different purposes. However, despite differences in signaling dynamics, the IFFL is a general motif found downstream of divergent signals in both cases. Generally, IFFLs are known for acting as sign-sensitive accelerators by increasing the response time of gene expression, dictated by the specific combination of kinetic parameters [13]. While some parameter sets allow the IFFL to act as pulse generators or band-pass filters, others speed up an output response [13, 50]. In addition, the IFFL has been shown to generate both time-dependent and dose-dependent biphasic responses [51]. Our analysis defines the quantitative constraints of another property of IFFLs, the capability for counting.
In natural systems, motifs do not exist in isolation and the dynamics of a motif in isolation may not persist when placed in the context of a larger network. Despite this caveat, our analysis is valuable for two major reasons. First, we identify a new property of IFFLs—the potential for information processing. The importance of IFFLs in signal processing is implied with its presence in many natural networks that are able to respond differentially to oscillatory and sustained signals (Fig 1). Considering this caveat, we explore the extent by which the counting mechanism, can persist in the presence of different perturbations in S1 Text (e.g. introduction of time delay, alterations to the current architecture, and cellular noise). Second, our analysis is useful for guiding the design and implementation of synthetic gene circuits, which are often intended to operate in relative isolation from other regulatory networks in the host cell.
Understanding the constraints of this network for decoding mechanisms is important in the design of synthetic circuits with more complex functions [52, 53]. The conclusions from our study can be used in the design of a synthetic pulse counter [54, 55]. Many synthetic gene circuits respond to the characteristics of static signals. In contrast, circuits adapted from our model would respond to the temporal dynamics of input signals. This would provide a wider range of input attributes, giving rise to a larger spectrum of potential circuit responses. In addition, by matching IFFL characteristics to natural cellular oscillations, synthetic signal decoding circuits would have the capability to respond to natural as well as synthetic stimuli.
Natural systems including those involved in cell proliferation, cell death, and neural regeneration convert oscillatory and sustained signals into distinct biological outcomes. In the case of disease, these circuits become deregulated, altering the function of networks required for information encoding or decoding. Insights derived from our model for the pulse counting mechanism will be important for future work in understanding the ways in which these perturbations impact the way cells decode information.
We use the following system of ODEs to define the IFFL network:
d[R]dt=k1Φ(t)−dR[X]n[X]n+Kxn[R]−dRo[R]
(5)
d[X]dt=k2Φ(t)−dx[X]
(6)
where k1 and k2 are rate constants for the production of R and X respectively, and dx and dRo are rate constants for the endogenous degradation of X and R respectively. The term [X]n[X]n+Kxn represents the threshold response through Hill kinetics for the induction of the degradation of R by X, with a maximal degradation rate of dR. Unless noted otherwise, we define Φ(t) as a periodic square wave function with k pulses: Φ(t) = 1 when iT′ + to ≤ t < iT′ + D′ + to and Φ(t) = 0 otherwise. Here i (0 ≤ i < k) is the pulse index, to is the start time, T′ is the cycle period, and D′ is the pulse duration.
We assume a square waveform input to simplify analysis. However, this assumption can be relaxed; numerical simulations indicate that other wave forms, such as sinusoidal input signals (S1 Fig), can also be counted. In our model, we assume first-order kinetics for the endogenous degradation of R and X. We assume that the induced degradation by X occurs through an ultrasensitive threshold response. In general, this arm can be found in other positions while maintaining the network structure. Our analysis can be generalized to other forms of the motif, as illustrated by the alternative models (S1 Text).
We non-dimensionalize the model by defining:
x=[X]Kx,r=[R]k2k1Kx,τ=tdx,τ0=t0dx,D=D′dx,T=T′dx,γR=dRdx,γo=dRodx,andβ=k2Kxdx
This leads to Eqs 1 and 2.
|
10.1371/journal.pcbi.1004784 | Elucidation of Genetic Interactions in the Yeast GATA-Factor Network Using Bayesian Model Selection | Understanding the structure and function of complex gene regulatory networks using classical genetic assays is an error-prone procedure that frequently generates ambiguous outcomes. Even some of the best-characterized gene networks contain interactions whose validity is not conclusively proven. Founded on dynamic experimental data, mechanistic mathematical models are able to offer detailed insights that would otherwise require prohibitively large numbers of genetic experiments. Here we attempt mechanistic modeling of the transcriptional network formed by the four GATA-factor proteins, a well-studied system of central importance for nitrogen-source regulation of transcription in the yeast Saccharomyces cerevisiae. To resolve ambiguities in the network organization, we encoded a set of five interactions hypothesized in the literature into a set of 32 mathematical models, and employed Bayesian model selection to identify the most plausible set of interactions based on dynamic gene expression data. The top-ranking model was validated on newly generated GFP reporter dynamic data and was subsequently used to gain a better understanding of how yeast cells organize their transcriptional response to dynamic changes of nitrogen sources. Our work constitutes a necessary and important step towards obtaining a holistic view of the yeast nitrogen regulation mechanisms; on the computational side, it provides a demonstration of how powerful Monte Carlo techniques can be creatively combined and used to address the great challenges of large-scale dynamical system inference.
| Gene regulatory networks underlie all key processes that enable a cell to maintain long-term homeostasis in a changing environment. Understanding the structure and function of complex gene networks is an experimentally difficult and error-prone procedure. Mechanistic mathematical modeling promises to alleviate these problems, as we demonstrate here for the yeast GATA-factor network, the central controller of the cellular response to nitrogen source quality. Despite years of targeted studies, the interaction pattern of this network is still not known precisely. To resolve several still-remaining ambiguities, we generated a set of alternative mathematical models, and compared them against each other using Bayesian model selection based on dynamic gene expression data. The top-ranking model was then validated on a separate, newly generated dataset. Our work thus provides new insights to the mechanism of nitrogen regulation in yeast, while at the same time overcoming some key computational inference problems for large models in systems biology.
| Decades of research on gene regulatory networks have provided us with a wealth of knowledge on their topologies. However, even the best characterized networks contain many ambiguous interactions, discovered using a variety of experimental techniques that often cannot validate their presence conclusively. Moreover, knowledge of a “static” gene regulatory interaction pattern consisting of multiple feedback and/or feedforward loops cannot provide insight into which regulatory interactions are functionally relevant at a given time and cellular context. Dynamic mechanistic modeling informed by quantitative, time-resolved experimental data can provide discriminatory resolution and is thus an indispensable tool for understanding the structure and function of complex gene networks.
The GATA gene regulatory network in the yeast Saccharomyces cerevisiae is an example of a well-characterized transcriptional network that contains multiple feedback loops. This feedback has confounded the inference of regulatory interactions from experiments and led to several speculative, unverified regulatory hypotheses. The network is composed of four transcription factors (TFs) that respond to the quality of the available nitrogen source and regulate the transcriptional response of around 90 genes related to nitrogen catabolism. Specifically, the network comprises the transcriptional activators Gat1 and Gln3 and the transcriptional repressors Dal80 and Gzf3, all four of which recognize the same core motif in the promoter regions of their gene targets, including the promoters of GAT1, DAL80 and GZF3. This cross-regulation provides tight control over the transcription of genes encoding for permeases and catabolic enzymes required for the utilization of poor nitrogen sources when more preferred sources are available. This phenomenon is generally referred to as Nitrogen Catabolite Repression (NCR) [1]. Depletion of rich nitrogen sources (e.g. glutamine) results in the relief of NCR, providing cells with the metabolic repertoire to scavenge for and utilize non-preferred nitrogen sources (e.g. proline). Yeast cells monitor the nitrogen availability by a yet unknown mechanism involving the rapamycin-sensitive TORC1 pathway, among possibly other signaling pathways, and accordingly control the NCR activity by modulating the subcellular localization of the two GATA activators [1–4]. In particular, TORC1 is known to mediate the localization of Gln3 and Gat1 through phosphorylation: during growth on poor nitrogen sources, Gln3 and Gat1 are not phosphorylated and localize in the nucleus to activate transcription, while in the presence of a good nitrogen source they are phosphorylated and remain predominantly cytoplasmic [5–8], although their phosphorylation pattern does not always correlate with their localization [9]. TORC1 inhibition with the antifungal agent rapamycin results in a nitrogen starvation phenotype that induces NCR-sensitive gene expression even in the presence of a good nitrogen source [10–12], a property frequently explored to mimic a downshift from a good to a poor nitrogen source, with concomitant relief of NCR [13].
Despite many years of targeted studies, parts of the GATA network topology remain obscure, since the complex interaction pattern complicates the interpretation of available experimental data. Various transcriptional interactions have been suggested over the years, but have remained unverified by subsequent observations. For example, results in [14–16] suggest Dal80 self-repression, yet its binding to the DAL80 promoter remains unverified. Moreover, the available experimental data (Northern blots [14] and LacZ assays [15]) cannot preclude the possibility that the observed increase in Dal80 expression in a Δdal80 background is due to indirect regulatory interactions. Similarly, the negative regulation of DAL80 by Gzf3 has been inferred from assays (LacZ [15] and Northern blots [16] in a Δdal80 background) that cannot differentiate between direct and indirect effects. Overall, a careful examination of the experimental evidence reported in the literature revealed in total five interactions whose validity cannot be unambiguously concluded. These hypothesized interactions are indicated with dashed lines on Fig 1. A detailed literature-based justification for the consideration of these interactions as hypotheses is presented in Section 1.2 of S1 Text.
To resolve such ambiguities we used Bayesian model selection combined with dynamic gene expression data. Based on an extensive literature search, we first compiled a set of five interactions that have been hypothesised in the literature, but remain unvalidated. We next encoded these biological hypotheses into alternative mathematical model structures and formulated a Bayesian model selection problem [17–21]. Exploiting special structures present in the resulting dynamical models and by creatively using Monte Carlo-based inference, a workflow to carry out inference for dynamical systems with very high-dimensional parameter spaces was developed. This allowed the systematic comparison of alternative models against each other and the selection of the best candidates based on the measured dynamic mRNA responses of target genes under a nitrogen upshift perturbation and rapamycin treatment. The top-ranking model was subsequently validated using experimental data generated in GATA factor deletion strains carrying a GFP reporter. Our results provide strong insights into the long standing open issues surrounding the transcriptional regulation of NCR. They provide strong evidence for Gat1 positive autoregulation, for Dal80 repression of GZF3 and that the two activators do not interact on the GATA-factor promoters. On the other hand, repression of DAL80 by Gzf3 appears not to be essential, and there is no strong support in favor of Dal80 self-repression. The top-ranking model structure was subsequently used to provide quantitative insights into network function that would be hard to obtain experimentally. With our system being among the largest and most complex considered for Bayesian model selection to date, we were also able to demonstrate how powerful Monte Carlo estimation methods can be efficiently used to address large-scale inference problems in computational biology.
To gain a better understanding of the transcriptional control of NCR by the yeast GATA gene regulatory network, we compiled a literature-based list of its components and their interactions. The established knowledge of how the GATA-factors regulate the expression of each other is depicted with solid lines on Fig 1 (a list of relevant references is provided in Section 1.1 of S1 Text), while hypothesized interactions are indicated with dashed lines and presented in detail in Section 1.2 of S1 Text.
To encode mathematically the established biological knowledge on the GATA network, as well as the hypothesized interactions, we generated a set of ordinary differential equation models that capture the evolution of all chemical species involved (mRNAs, proteins and protein complexes). The models account for mechanistic details that describe the rates of mRNA transcription, protein production, protein degradation, nuclear-cytosolic translocation and dimerization, formalized in a total of 13 dynamical states and embedding three input variables. Moreover, they take as input an external signal that reflects the quality of the nitrogen source and determines the translocation rates for the two activators. An additional, secondary input of the system is the Gln3 mRNA concentration. The state variables contained in the model describe the mRNA concentrations, the nuclear / cytosolic concentration of the activators, and the monomeric / dimeric concentration of the repressors. Further details can be found in Materials and Methods, and Section 2 of S1 Text.
The basic model structure based solely on the well-established GATA network interactions comprises 41 parameters. To determine if any, or a combination, of the hypothesized interactions are more plausible given the experimental observations, we next encoded the five biological hypotheses into alternative mathematical model structures. Since the five hypothesized interactions are not mutually exclusive, a total of 25 − 1 = 31 additional alternative model structures, M k, were generated, each encoding a particular combination of interactions. Each model structure accounted for 41 to 50 parameters, depending on the combination of hypotheses. The structures were named as follows: starting from the full model (M 0) that contains all hypothesized interactions, we denoted each subsequent model by the interactions it is missing. For example, model M 124 misses the interactions suggested by hypotheses 1, 2 and 4, according to the enumeration of interactions presented on Fig 1. In order to verify the plausibility of the hypothesized interactions based on the improved predictions of an augmented model structure relative to others, we proceeded with two rounds of model selection and an experimental model validation step as summarized in Fig 2.
Model selection was based on an existing dataset of mRNA abundances previously quantified for wild type yeast subject to an upshift from proline to glutamine (Pro→Gln) and to a downshift induced by rapamycin addition to glutamine-grown cells (Gln+Rap) [22].
To assess which network topology among the alternatives is supported by the GATA-factor gene expression data, denoted DTF, we performed a first round of Bayesian model selection. According to the Bayesian approach, detailed in Section 3.1 of S1 Text, all 32 alternative model structures were initially assigned an equal level of plausibility (prior probability), P ( M k ). Subsequently, the prior model probabilities were updated using the experimental data to estimate P ( D T F | M k ) (called the evidence for model M k) to obtain the posterior model probabilities, P ( M k | D T F ), using Bayes’ formula: P ( M k | D T F ) ∝ P ( D T F | M k ) P ( M k ). These quantities, shown on Fig 3(a), encode the plausibility of each model structure after incorporating the experimental observations.
To enable the calculation of posterior probabilities on the high-dimensional model parameter spaces, we used a Sequential Monte Carlo (SMC) sampler [23] (S1 Text, Section 3.2), which was developed based on a comparison of different advanced sampling methods [24]. Sequential Monte Carlo is a family of powerful algorithms that tackle the problem of sampling from an intractable (i.e. hard-to-sample) distribution by starting from a tractable one and moving through a sequence of artificial intermediate distributions. The algorithms include several user-defined settings that can greatly affect their performance, and successful application of these methods had never been reported for dynamical systems of size comparable to the one treated here. Our SMC sampler was able to explore efficiently the parameter spaces thanks to an adaptive sampling mechanism based on density estimation via Gaussian mixtures, which is able to overcome the common problems faced by traditional sampling approaches in high-dimensional settings. The algorithm was thus able to provide low-variance estimates (Section 5.2 and Fig. I and Fig. N in S1 Text) that enabled us to reliably rank the alternative model structures according to their posterior probabilities (Fig 3(a)). Following the interpretation of model evidence ratios given in [25] and given that all model priors are equal, a ratio of posterior probabilities greater than 100 can be interpreted as decisive support of the data in favor of one model against another. Based on these posterior probabilities no model stands out clearly from the rest: the ratio of posterior probabilities between the top-ranking and the rest of the models is not great enough to provide decisive support in its favor (Fig 3(a)).
Although the available gene expression data alone could not provide unambiguous evidence in favor of a single model structure, we observed a set of candidate models whose posterior probabilities are clearly higher from the rest. Interestingly, all these structures contain the repression of GZF3 by Dal80 (hypothesis 4). We therefore eliminated all 16 model structures missing this interaction, and proceeded to discriminate among the remaining 16 models that account for the repression of GZF3 by Dal80.
An indirect way to observe the changes in the GATA-factor transcription activities, is to consider their regulatory effect on known target genes. With the aim of obtaining additional model resolution to sharpen the model selection results, we extended the core model to account for additional target genes regulated by the GATA factors and for which gene expression data is also available. Yeast GATA factors are the main regulators of around 90 genes involved in nitrogen catabolic gene expression and core nitrogen metabolism [26, 27]. Of these, we selected six targets that are known to be mainly controlled by the GATA factors during NCR—DAL1 (allantoinase), DAL5 (allantoin permease), GLN1 (glutamine synthetase), GLT1 (glutamate synthetase), MEP2 (ammonium permease) and PUT4 (proline permease) (Fig 1) -, and used them in the subsequent model selection process. The exact regulatory influence of each GATA factor on each target is still elusive and seems to differ depending on the structure of their promoter, such as the number and spacing of binding sites. More information about these genes and a justification for their choice is given in Section 1.3 of S1 Text.
To account for the gene expression data from these GATA targets (denoted Dtargets and previously obtained in [22]), we expanded the initial GATA-factor model by six additional states, representing the target mRNAs. Since the precise regulation pattern (number of GATA regulators and interaction strengths) of each target is uncertain, each target equation contributes seven unknown parameters to the extended model (cf. S1 Text, Section 2.4). This leads to a significant increase in computational cost of the model selection process, as the total number of parameters rises to 92 in the case of the extended model M 0 *. To the best of our knowledge, no currently available Monte Carlo algorithm is able to reliably sample parameter spaces for dynamical systems of this size, a computational challenge even when compared to existing studies with thousands of variables for static Bayesian hierarchical models [28]. We have been able to circumvent this limitation by employing a novel modular sampling approach, in which we exploit the unidirectional flow of state information in the extended system. This property allowed us to decompose the total model evidence calculation into a product of several factors, each of which can be obtained with much smaller computational effort. The theoretical justification and the practical implementation of our approach are provided in Materials and Methods, and Section 4 of S1 Text respectively.
To further discriminate among the remaining 16 hypotheses, we applied a second round of Bayesian model selection to the extended model formulation. Following the modular sampling approach described in Section 4 of S1 Text, the posterior probability of the k-th augmented model, M k * can be obtained from the formula
P ( M k * | D T F , D t a r g e t s ) ∝ P ( M k | D T F ) F ( D T F , D t a r g e t s , M k ) ,
where DTF and Dtargets denote the TF and target gene expression datasets respectively, M k is the k-th TF model structure corresponding to a combinatorial topology of four possible interactions (hypotheses 1, 2, 3 and 5), and F is a multiplicative factor that can be estimated by Monte Carlo integration, as described in Section 4 of of S1 Text. Note that the posterior probability of the original model, P ( M k | D T F ), was already available from the first model selection round. Table I in S1 Text summarizes the estimates of the multiplicative factors F for the 16 model structures considered in this second round.
Putting together the estimates for P ( M k | D T F ) with the estimates of F, we obtained the model posteriors shown on Fig 3(b). We clearly observe that all structures lacking hypothesis 2 are strongly penalized, as their posterior probabilities are the lowest among all structures considered. This result suggests that Gat1 self-activation drastically changes a model’s capacity to accommodate the target gene expression data, while the TF dataset is less discriminatory by itself. Overall standing out as the most plausible models were M 135 and M 35, both missing interactions corresponding to hypotheses 3 and 5. The presence or absence of hypothesis 1 does not make a significant difference between the two models, since the posterior probabilities differ only by a small factor (3.3). This may arise from the fact that the Bayesian methodology implicitly penalizes Model M 35 relative to M 135 because of its extra free parameter. Thus, after two rounds of Bayesian model selection, the initial list of 32 candidate models was reduced to two top-ranking topologies. These two top models strongly support the role of Gat1 self-activation and of GZF3 repression by Dal80 (hypotheses 2 and 4), and discard the relevance of DAL80 repression by Gzf3 and Gln3-Gat1 interaction (hypotheses 3 and 5) in regulating the yeast NCR response. In the subsequent sections, the top-ranked model (M 135) will be used, due to its reduced complexity relative to M 35.
To validate the results of the final model selection round, we challenged model M 135 to predict the outcome of additional experiments. To this end, we designed an experiment to dynamically monitor GFP expression from GATA promoters in the absence of each of the four GATA factors during the same two shifts used for model selection (Pro→Gln and Gln+Rap). Specifically, we constructed a collection of GFP-reporter plasmids expressing the yeast Enhanced Green Fluorescent Protein (yEGFP) gene immediately downstream of the native promoter of each GATA factor (S2 Text). Each of the four plasmids and the control vector were transformed into the wild type and all four GATA single deletion mutants, yielding a total of 25 yeast strains (S2 Text). The strains were cultivated in liquid culture in microtiter plates and monitored online for biomass and GFP evolution (S1 Dataset). Glutamine and rapamycin were added to cells growing exponentially in proline or glutamine, respectively. The fluorescence and biomass measurements were background-corrected and processed following the approach described in [29] to obtain the relative concentration of GFP, as well as the time-dependent growth rate.
In parallel, we simulated the GFP response of each GATA promoter under the experimentally defined conditions, using the topology of the top-ranked model M 135 and an adjusted set of differential equations that account for the extra species involved (GFP mRNA, immature and mature GFP). Further details can be found in Section 2.5 of S1 Text.
The outcome of the GATA-factor model M 135, augmented with the GFP reporter dynamics, was used for a qualitative comparison between the predicted GFP evolution and the experimental data. Experimental and predicted results for strains harboring the DAL80 and the GZF3 reporter GFP are shown in Figs 4 and 5, respectively (very similar predictions were obtained with model M 35). Strains harboring the GLN3 reporter showed no significant changes in GFP production rate (S1 Dataset), in line with the previously described observations that GLN3 is regulated in a NCR-independent manner. The plasmid harboring the GAT1 reporter did not show any GFP signal for unclear technical reasons that could not be addressed, while the GZF3 promoter signal was very close to background in most deletion strains. Overall, despite some caveats that preclude their quantitative comparison (S1 Text, Subsection 2.5.1), predictions with model M 135 match experimental outcomes well in terms of the ordering and general trend of the responses, reinforcing our model-based conclusions on the presence of the hypothesized interactions 2 and 4 depicted on Fig 1.
To gain further insights into open questions regarding the functioning of the GATA network, we next explored in detail the dynamic behavior of model M 135 to extract key quantitative variables that describe the dynamics of the GATA regulatory interactions during the nutritional upshift and the rapamycin-induced downshift (Figs 6 and 7).
The most obvious output of the model is its ability to describe the mRNA levels of the GATA factor targets, for which the model has been fitted during model selection. Figs 6(c) and 7(c) depict the experimental and described mRNA trajectories during the upshift and rapamycin treatment, respectively. Key open questions that remain elusive are (i) what are the dynamics of nuclear translocation/degradation of the GATA factors and how does that dictate their nuclear abundance, (ii) what is the nuclear abundance of each GATA factor and how does that dictate their TF activity, and (iii) which GATA factor is mainly responsible for the regulation of each target promoter. To address these questions, we extracted the model variables on the concentrations of nuclear and cytoplasmic GATA factor species and used them to calculate (i) the abundance of the active forms of Gat1, Gln3, Dal80 and Gzf3 (that is, nuclear Gln3 and Gat1, as well as Dal80 and Gzf3 homodimers, shown on Figs 6(a) and 7(a)), and (ii) the relative contribution of each of the four TF active forms to the regulation of the target gene expression (Figs 6(b) and 7(b)).
The inferred abundances for the active forms of Gat1, Gln3 and Dal80 show a drastic reduction within the first minutes upon glutamine addition to proline-grown yeast (Fig 6(a)). The nuclear depletion of Gat1 and Gln3, caused by their translocation to the cytosol as defined by the model, is completed within 5 minutes, while the nuclear depletion of Dal80, due to protein degradation after the shut-down of its expression, has a longer half-life of ∼15 minutes. The drastic depletion of Dal80 dimers is consistent with the fact that under nitrogen-rich conditions it is practically undetectable [30]. By contrast, nuclear abundance dynamics in the rapamycin-induced downshift reveal a clear difference between Gln3 and Gat1 (Fig 7(a)). While Gat1 increases its nuclear abundance monotonically to a saturation level after rapamycin treatment, Gln3 shows a transient overshoot to a lower steady state level. A similar trend has been observed experimentally, albeit with very coarse quantification and sparse sampling over time [9]. The abundance of Gzf3 remains practically constant in both shifts, as Gzf3 responds weakly and returns to steady-state levels after a transient change during the first 30 minutes of the shifts (Figs 6(a) and 7(a)).
To determine the relative contribution of each of the four TF active forms to the regulation of the target gene expression, we estimated the contribution of each TF to the fractional occupancy of each target gene promoter (Figs 6(b) and 7(b), Section 2.2 of S1 Text). The relative contributions of the GATA TFs to their target promoters during the upshift suggest that all target genes reduce their expression mainly because of the nuclear exit of the activators Gln3 and Gat1 (Fig 6, in particular panel b). The particular behavior of the Gzf3 mRNA (Fig 6(c)) seems however to arise from the interplay between Gln3 and the repressor Dal80: as Gln3 exits the nucleus and Dal80 remains around 15 minutes longer, GZF3 expression transiently drops repressed by Dal80. Upon disappearance of Dal80, the repression effect disappears, and the basal expression of GZF3 together with the small amount of nuclear Gln3 take over and restore the Gzf3 mRNA level. In contrast, the different nuclear behavior of the two activators in the rapamycin-induced downshift is reflected in the more diverse gene expression patterns of targets (Fig 7): those that are predicted by the model to be jointly regulated by Gln3 and Gat1 (e.g. DAL80, DAL5, GLN1, GLT1, MEP2, PUT4) according to the results of Fig 7(b), maintain a high expression level after the shift, while those affected mostly by Gln3 (e.g. DAL1, GAT1, GZF3) show a burst of expression followed by a lower steady state level. Interestingly, the latter group of genes also shows a high contribution of Dal80 in the later downregulation phase, which confirms the role of Dal80 as an important modulator of nitrogen catabolite repression relief [1]. Regardless of the condition, Figs 6(a) and 7(a) show that the role of Gzf3 in target expression seems to be that of a constant repressor, acting almost independently of the nitrogen source, possibly to assure full repression even in the presence of traces of nuclear Gln3 and Gat1 [1, 31].
Determination of functional gene regulatory interactions using currently available experimental techniques is still a time-consuming and non-trivial process, particularly difficult to resolve in networks containing feedback and/or feedforward loops. The yeast GATA gene regulatory network, the central transcriptional controller of nitrogen catabolite repression (NCR) in S. cerevisiae, is an example of a relatively well-characterized network with only four TFs but comprising several feedback/feedforward loops, which have so far hindered conclusive validation of several hypothesized interactions. In this work, we tackled the problem of identifying the most plausible interactions from existing hypotheses by applying mathematical modeling and Bayesian model selection to determine the support that experimental data lends to five yet unverified interactions within the GATA network. Overall, our model selection results provided strong evidence in favor of two of the hypothesized interactions, Gat1 self-activation and GZF3 repression by Dal80 (hypotheses 2 and 4 on Fig 1), while further biological evidence is necessary to conclude on the requirement of Dal80 self-repression (hypothesis 1). The remaining hypotheses—DAL80 repression by Gzf3 and Gln3-Gat1 interaction—appear dispensable according to our model, either because they are too weak to have significant impact on the measured system variables, or because they arose due to indirect regulatory effects.
Our approach relied on two rounds of Bayesian model selection applied to a system of ordinary differential equations describing the mechanistic details of transcription, translation and translocation of the members of the yeast GATA network. A basic model structure was first developed based on the current established regulatory interactions, and subsequently augmented to 32 structures corresponding to all possible topologies determined by combinations of the five hypothesized interactions. At this point we should note that our model selection approach (that is, considering the model structure corresponding to each combination of hypotheses in isolation) is equivalent to including a mass at zero in the priors of the full model that correspond to parameters that are “switched off” when certain interactions are missing and inferring the posterior parameter distribution over this complex multimodal prior. Further details are provided in Subsection 3.1.1 of S1 Text. Evaluation of the model structure that best described the dynamic mRNA data experimentally obtained in two distinct perturbations was enabled by a careful design of a computational pipeline that allowed us to efficiently handle models of great size and complexity, and which can prove to be generally useful for model-based inference problems with similar features. To overcome the great difficulties of sampling from complex, high-dimensional parameter distributions, particularly important here was the efficient design of our SMC sampler and our modular sampling approach that enabled the reduction of a high-dimensional sampling problem into two easier sub-problems. The applied Bayesian model selection procedure allowed us to identify a top-ranking model structure, M 135, that was able to reproduce the experimental data with the minimal necessary complexity, as well as to predict responses from an independent validation experiment.
The top-ranking model structure strongly supported the regulatory relevance of Gat1 self-activation and GZF3 repression by Dal80, while the remaining three hypotheses did not substantially improve predictions relative to the basic model (Fig 3). When challenged to predict the outcome of a validation experiment comprising the GFP screening of each GATA-factor promoter activity in the absence of each of the regulators during the same two shifts, the top-ranking model performed well and qualitatively predicted the responses and sequence of events (Figs 4 and 5). Adding to its interest for model validation, the performed experiment offers a valuable dataset to systematically evaluate how each GATA-factor impacts each other’s gene expression during either an upshift or a downshift in NCR activity.
Many aspects of the functioning of the GATA network under NCR-repressive (glutamine-grown yeast) or NCR-relieved (proline-grown yeast) conditions can be confirmed simply based on the initial steady-state points of the validation experiments (initial points in Figs 4(a), 4(c) and 5(a), 5(c)), and can be better understood in light of the model structure. During exponential growth in glutamine, DAL80 is derepressed in Δgzf3, while GZF3 is derepressed in Δgat1 and repressed in Δgln3. During growth in proline DAL80 is derepressed in Δdal80 and repressed in Δgln3, while GZF3 is derepressed in Δgat1 and Δdal80, and repressed in Δgzf3. These observations generally agree with the established and here suggested regulatory interactions controlling DAL80 and GZF3 gene expression, as depicted on Fig 1. We noticed however that two of the observed results corresponded to hypotheses that were not validated by the top-ranking model: repression of DAL80 by Gzf3 and self-repression of Dal80. While the latter needs further biological validation (it was part of the second-ranked model), our results suggest that the apparent repression of DAL80 by Gzf3 is mediated through GAT1. Consequently, the increase of DAL80 transcript levels in a Δgzf3 strain is attributed to the relief of repression on GAT1, which in turn activates DAL80. Another apparent contradiction was the derepression of GZF3 in Δgat1, an unexpected behavior considering that Gat1 is an activator, and which contrasts with the result for Gln3, the other activator. This counterintuitive behavior is however predicted by the model (Fig 5(b) and 5(d)): Gat1 deletion leads to DAL80 downregulation, which in turn causes an increase of Gzf3, since Dal80 is a direct inhibitor of GZF3 expression (hypothesis 4). This contradiction further suggests that Gln3 is the main activator of GZF3, since only deletion of Gln3 (but not Gat1) lowers GZF3 transcription. Also unexpected was the experimental observation that GZF3 levels are repressed in Δgzf3, an observation also explained by the model: when Gzf3 is deleted, GAT1 expression increases and, due to the relatively weak effect of Gat1 on GZF3, the concomitant increase of Dal80 ultimately reduces the transcription of GZF3. As a final observation, we noticed from our experiments that Gzf3 mainly exerts its repressor activity specifically under NCR-repressive conditions, while it gets overshadowed by Dal80 once NCR is relieved, in agreement with previous reports from the literature [15, 32]. In fact, deletion of Dal80 did not result in a behavior different from the wildtype in glutamine-grown cells, supporting the view that DAL80 is tightly switched off under NCR. Overall, the experimental data reflected well the current knowledge of the GATA-network in regulating NCR, and could offer several model-guided insights.
In addition to explaining experimental observations and helping to resolve the plausibility of the five hypothesized interactions, the top-ranking model structure was also explored to bring insights into the dynamics and operation of the yeast GATA network. To this end, we extracted from the model the variables that described the concentrations of nuclear/cytoplasmic TFs, and the relative contribution of each active TF to regulation of the different target gene expression (Figs 6 and 7). Our results regarding the differing nuclear localization responses of the two activators in the downshift agree with recent experimental observations suggesting that the nuclear localization of the GATA activators is likely to be regulated by two distinct pathways, of which one is more responsive to rapamycin, and the other to nitrogen source quality [33–36]. One particularly difficult question to resolve experimentally is the determination of the relative contribution of each GATA-factor to the regulation of their targets, since all GATA-factors share the same (or very similar) binding motifs on the promoter of the targets. By extracting from the model the fractional occupancy of each TF on each target gene (Figs 6(b) and 7(b)), we produced plausible predictions for the main responsible GATA-factor regulating each of the GATA targets considered in this study.
Altogether, our modeling exercise brought several insights into the function of the GATA network. First, the presence of Gat1 self-activation appears to confer greater independence from the other activator, Gln3, as suggested by the high levels of nuclear Gat1 following the rapamycin-induced downshift, when Gln3 is predominantly cytoplasmic (Fig 7(a)). Such independence seems to offer more fine tuning possibilities for yeast cells to regulate the balance between activators and repressors in the nucleus. Second, we provide strong evidence that Dal80 is indispensable to negatively regulate GZF3, and that this is not constitutively expressed as previously suggested by some groups [1, 31, 32], though contradicted by others [15, 16]. In fact, the experimentally measured Gzf3 mRNA clearly showed that GZF3 is transiently regulated following the perturbations, before returning to a steady-state similar to initial levels. This transcriptional regulation however does not lead to great changes in abundance of Gfz3, rather suggesting that Gzf3 behaves like a constant repressor.
In conclusion, our work constitutes a necessary and important step in the direction of mathematical modeling of the yeast GATA gene regulatory network, a small system with a complex interaction pattern that has hampered clear interpretation of experimental observations related to NCR. Further accumulation of experimental data will enable our model to be expanded and connected with existing signaling models of the TOR pathway [37], nitrogen transport [38] and core metabolism [39], to gain a more holistic view and a better understanding of NCR.
The GATA system equations (Section 2, S1 Text) are based on several assumptions supported by the literature and listed below for completeness:
All GATA factors recognize the same core motif (5’-GATAA-3’ or 5’-GATTA-3’), found in several copies upstream of NCR-controlled targets, as well as at the GAT1, DAL80 and GZF3 promoters. Gln3 is the only GATA factor whose expression is not nitrogen-regulated to any significant extent [1], while the rest of the GATA factors display a complex interaction pattern ([1, 2, 10, 16, 45] and references therein). From the interactions summarized in Figs 1 and 8, the following chemical reactions were derived, based on the list of assumptions given above (proteins are denoted by capital first letter, mRNA by small):
Transcription factor activation and translocation
mRNA production/degradation
Protein production/degradation
Protein-protein interactions
The above reactions are described by a set of ordinary differential equations given in Section 2 of S1 Text. They are all assumed to follow mass-action kinetics, except mRNA transcription and TF activation. The role of each regulator on the production rate of a given mRNA is clarified in Fig 1. The transcription rate of a specific mRNA is assumed to be proportional to the fractional occupancy of its promoter, i.e. the fraction of time that the promoter is active. The fractional occupancy at any given time is a function of the regulator amounts present at that time (following the common quasi-steady-state assumption for promoter occupancy). The form of the fractional occupancy function is determined using the thermodynamic approach of [43, 46]. An example of a fractional occupancy function for two activators is given on Fig 8.
Depending on the type of shift modeled (i.e. upshift or downshift) a separate activation/inactivation signal from the upstream signaling components is considered for each activator, and serves as an external input to the system (functions k1w(t) and k1x(t) in the reactions above). Each signal belongs to a class of sigmoid functions, which is biologically plausible and can capture step-like activity changes. The parameters of our sigmoids have to be estimated from the available transcription data, along with the rest of system parameters. More concretely, the parameterized functional forms we assume, also displayed on Fig 8, are the following:
The role of each parameter in the above functions is intuitively obvious. Each GATA activator is assigned its own set of parameter values, which also vary between the different shifts and have to be estimated from the available transcription data, along with the rest of system parameters. The assumed time dependence of the activation rate is reasonable, given recent experimental readouts of TOR pathway activity, which show a) a fast, step-like decrease in TOR activity upon rapamycin treatment [47] b) a very fast, step-like increase in TOR activity during a nutrient upshift (proline to glutamine) [47] c) a very fast, step-like increase in Gln3 phosphorylation (which controls its cytoplasmic localization) upon a nutrient upshift (proline to glutamine) [8].
Finally, to obtain the Gln3 mRNA input signal the available mRNA timecourse measurements for each experiment (Section 2.8 and Fig. A in S1 Text) were linearly interpolated and fed into the model simulator.
The generated ordinary differential equation models encode mathematically the existing biological knowledge about the GATA network and enable us to use statistical methods for selecting the model with the optimal complexity that can reproduce the available experimental data. In this work we chose to carry out model selection in a Bayesian framework [48]. Contrary to the commonly used Akaike and Bayesian Information Criteria (AIC and BIC), which are valid only asymptotically [49] (i.e. as the amount of data tends to infinity), Bayesian model selection is applicable with a limited amount of data. Moreover, it naturally penalizes model complexity without explicitly referring to the number of model parameters, as AIC and BIC do. This is especially important for large nonlinear models considered in Systems Biology, as practical unidentifiability of parameters [50] is very common and implies that the “effective” number of parameters (“degrees of freedom”) in a given model does not correspond to the actual number of parameters. Finally, Bayesian model selection incorporates our prior beliefs about parameter values and model plausibility in a consistent way, whereas this is impossible with AIC and BIC.
Given a set of competing biological hypotheses { H k } k = 1 K, each encoded in a mathematical model M k, Bayesian model selection works by computing the posterior probability P ( M k | D ) of each model given the available experimental data D. This involves the computation of the marginal likelihood (also called evidence) P ( D | M k ), which, being an integral over the high-dimensional parameter space of M k, forms the main computational bottleneck of the process. Further details on Bayesian model selection are provided in Section 3.1 of S1 Text.
Since the evidence P ( D | M k ) cannot be evaluated analytically in all but the simplest cases, Monte Carlo-based numerical integration methods are typically employed for its computation. Due to the high dimensionality of the parameter spaces considered, simple estimators based on the Laplace approximation of the posterior and importance sampling estimators have been shown to result in highly variable and/or biased results [51]. After a detailed comparison of different sophisticated sampling methods [24], we chose to implement a Sequential Monte Carlo (SMC) sampler, described in more detail in Section 3.2 of S1 Text.
Briefly, the SMC sampler can provide samples from the posterior distribution of parameter values, P ( θ k | D , M k ) (where θk denotes the parameter vector of the k-th model), as well as an estimate of the evidence integral. P ( θ k | D , M k ) expresses the conditional distribution of the model parameters after taking the observed dataset D into account [48] and, according to Bayes’ theorem, it is proportional to P ( D | M k , θ k ) P ( θ k | M k ), where is P ( D | M k , θ k ) the likelihood function and P ( θ k | M k ) the prior parameter distribution (definitions and details are provided in Section 3.1 of S1 Text).
SMC generates samples from the posterior parameter distribution and estimates the evidence using a sequence of bridging distributions, fβ, defined according to a “cooling schedule”:
f β i ( θ ) ∝ P ( D | M , θ ) β i P ( θ | M ) , (1)
for 0 = β0 < β1 < … < βN = 1. The algorithm works by propagating a population of particles sampled from the diffuse prior through this sequence of intermediate distributions that gradually “morph” into the (typically much more concentrated and complex) target posterior.
As it is practically impossible to verify SMC convergence in a rigorous way for the problem at hand, we repeatedly ran the algorithm for a few different models to monitor the variability of the estimated quantities and detect any anomalous behavior. The algorithm was thus iteratively tuned so that the variance of the estimates was small enough to permit safe conclusions about model ranking (further details can be found in Section 5.2 of S1 Text).
When the dynamical system of interest displays a modular structure without feedbacks, a simple rewriting of the evidence integral can prove very helpful for carrying out the computation in a sequential manner. We have used this evidence decomposition to speed up the computation in the second model selection step by defining the transcription factor network as the “upstream” module, and the six GATA targets as the “downstream” modules, as described in Section 4.2 of S1 Text.
Here, we briefly describe the concept of evidence decomposition for modular systems: as an example, consider a dynamical system of the form
x ˙ = F ( x , θ ) ,
where x ∈ R n and θ ∈ R m is the parameter vector. We make the following assumptions:
If we denote by π(θ1) and π(θ2) the priors on the two parameter sets and by P(D1, D2|θ1, θ2) the likelihood function of the parameters, we can immediately write
P ( D 1 , D 2 | θ 1 , θ 2 ) = P ( D 1 | θ 1 ) P ( D 2 | θ 1 , θ 2 ) . (2)
The form of the likelihood thus encodes the flow of state information between the two subsystems, and can be easily generalized to the case of a cascade of n subsystems, each affecting the next.
In the simple case of two modules, the evidence integral becomes
P ( D 1 , D 2 ) = ∫ ∫ P ( D 1 | θ 1 ) P ( D 2 | θ 1 , θ 2 ) π ( θ 1 ) π ( θ 2 ) d θ 1 d θ 2 (3) = ∫ P ( D 1 | θ 1 ) π ( θ 1 ) d θ 1 ︸ P ( D 1 ) ∫ P ( D 1 | θ 1 ) π ( θ 1 ) ∫ P ( D 1 | θ 1 ) π ( θ 1 ) d θ 1 P ( D 2 | θ 1 , θ 2 ) π ( θ 2 ) d θ 2 (4) = P ( D 1 ) ∫ P ( D 2 | θ 1 , θ 2 ) P ( θ 1 | D 1 ) π ( θ 2 ) d θ 2 . (5)
In the above equations, P(D1) denotes the evidence of the module corresponding to F1, based only on the D1 dataset by ignoring the downstream subsystem. Apart from P(D1), we also need P(θ1|D1), which is the parameter posterior for the upstream module, based again on D1. According to this rewriting of the total evidence, its calculation can then proceed in two steps: first, the upstream module is treated in isolation, and the results of this computation (evidence and parameter posterior) are then fed into the calculation of the evidence for the downstream module. In effect, numerical estimation of this second integral amounts to integrating the likelihood for D2 with respect to the posterior of θ1 in place of the prior, and multiplying by the evidence P(D1).
The same procedure can be generalized when multiple subsystems are jointly affected by the first one, but do not interact with each other. Further details on how this decomposition can be exploited in the SMC sampling algorithm are provided in Section 4 of S1 Text.
All models were implemented using SBTOOLBOX2 [52] (http://www.sbtoolbox2.org/main.php), a freely available Matlab toolbox that is best suited for simulation and analysis of ODE-based models. The SBPD extension of the toolbox is particularly useful, as it enables high-speed simulation (∼100x faster than the built-in Matlab integrators) of high-dimensional ODEs by converting models to C code and using the powerful CVODEs integrator [53] from the SUNDIALS package [54].
At each temperature step, the SMC sampler requires the likelihood evaluation of b ⋅ M parameter points, where M is the size of the particle population and b the number of Metropolis-Hastings iterations used in our proposal kernel (Section 3.4, S1 Text). Since the likelihood evaluation requires the integration of the model ODEs, this is a very computationally demanding task, even if a single model run takes a small fraction of a second. For this reason, all SMC runs in this work were performed on 64 cores of the ETH Brutus cluster (https://www1.ethz.ch/id/services/list/comp_zentral/cluster/index_EN), using custom-written and speed-optimized parallel Matlab code. With this setup, an SMC run of the first model selection round with M = 15000, b = 15 and 70 temperature steps, takes around 2 hours to complete for each model structure. Additional speedup can be achieved by converting into C code the second most time-consuming step of the SMC, the fit of the Gaussian mixture model (Section 3.4, S1 Text).
The full GATA-factor model in SBML and SBTOOLBOX2 formats is provided in S1 File.
We used time-course mRNA microarray data previously obtained by us in two different perturbation experiments: a nitrogen quality upshift from proline to glutamine (Pro→Gln) and a rapamycin-induced downshift during growth in glutamine (Gln+Rap) [22] (NCBI GEO accession numbers GSE54844 and GSE54851). Briefly, wildtype Saccharomyces cerevisiae was grown in well-controlled bioreactor operated in batch mode using a defined minimal media with glucose as sole carbon source and a defined nitrogen source composition. In the Pro→Gln upshift, yeast was grown exponentially in proline as sole nitrogen-source and a dynamic upshift was induced by addition of glutamine. In the rapamycin-induced downshift (Gln+Rap), the downshift was induced by the addition of rapamycin to yeast growing exponentially in glutamine. Gene expression was quantified using Affymetrix DNA microarrays at eight timepoints (-10, 3, 7, 10, 14, 24, 56 and 120 minutes after the perturbation), with triplicate measurements taken at -10, 7, and 24 minutes from three independent biological replicates. Further replicates are cost-prohibitive for such dynamic experiments [22]. The triplicates were used to assess both the biological and microarray variability and define a measurement noise model (S1 Text, Section 3.3). Since Affymetrix DNA microarrays do not allow comparison of intensities across different transcripts species, we worked with fold-changes normalized relative to the steady-state sample taken before the time of the shift. Experimental and data processing details can be found in [22].
Wildtype S. cerevisiae FY4 and four isogenic single gene-deletion yeast strains lacking each of the four GATA-factors were transformed with the low-copy plasmid pRS41H harboring the promoter region of each GATA-factor (-600 to -1 bp upstream of the beginning of the ORF) immediately upstream of a GFP reporter gene (see S2 Text for details). Plasmid inserts containing the GATA promoter, the yGFP3 sequence and the yeast CDC28 terminator were synthesized by GeneArt AG (Regensburg, Germany) as described in S2 Text. This resulted in a total of 25 strains (five backgrounds—wildtype, Δdal80, Δgat1, Δgln3 and Δgzf3—each transformed with one of the possible five plasmids harboring the promoter GATA-GFP—empty vector, pDAL80-GFP; pGAT1-GFP, pGLN3-GFP and pGZF3-GFP). All strains were cultivated in microtiter plates in Biolector, grown under the same conditions and subjected to the same shifts used to generate the mRNA data (details in S2 Text). Cell fluorescence (GFP filter) and biomass accumulation was monitored in real time (S1 Dataset). The fluorescence (I(t)) and biomass (A(t)) measurements were background-corrected and processed following the approach described in [29] to obtain the relative concentration of GFP, r(t) ∝ I(t)/A(t), as well as the time-dependent growth rate μ(t) = dln(A(t))/dt.
|
10.1371/journal.pgen.1007911 | Plasma membrane architecture protects Candida albicans from killing by copper | The ability to resist copper toxicity is important for microbial pathogens to survive attack by innate immune cells. A sur7Δ mutant of the fungal pathogen Candida albicans exhibits decreased virulence that correlates with increased sensitivity to copper, as well as defects in other stress responses and morphogenesis. Previous studies indicated that copper kills sur7Δ cells by a mechanism distinct from the known resistance pathways involving the Crp1 copper exporter or the Cup1 metallothionein. Since Sur7 resides in punctate plasma membrane domains known as MCC/eisosomes, we examined overexpression of SUR7 and found that it rescued the copper sensitivity of a mutant that fails to form MCC/eisosomes (pil1Δ lsp1Δ), indicating that these domains act to facilitate Sur7 function. Genetic screening identified new copper-sensitive mutants, the strongest of which were similar to sur7Δ in having altered plasma membranes due to defects in membrane trafficking, cortical actin, and morphogenesis (rvs161Δ, rvs167Δ, and arp2Δ arp3Δ). Consistent with the mutants having altered plasma membrane organization, they were all more readily permeabilized by copper, which is known to bind phosphatidylserine and phosphatidylethanolamine and cause membrane damage. Although these phospholipids are normally localized to the intracellular leaflet of the plasma membrane, their exposure on the surface of the copper-sensitive mutants was indicated by increased susceptibility to membrane damaging agents that bind to these phospholipids. Increased copper sensitivity was also detected for a drs2Δ mutant, which lacks a phospholipid flippase that is involved in maintaining phospholipid asymmetry. Copper binds phosphatidylserine with very high affinity, and deleting CHO1 to prevent phosphatidylserine synthesis rescued the copper sensitivity of sur7Δ cells, confirming a major role for phosphatidylserine in copper sensitivity. These results highlight how proper plasma membrane architecture protects fungal pathogens from copper and attack by the immune system, thereby opening up new avenues for therapeutic intervention.
| The transition metal copper is used by the innate immune system to attack microbial pathogens. To better understand how the human fungal pathogen Candida albicans resists this type of stress, we screened for mutants that were more susceptible to killing by copper. Interestingly, we identified a new class of copper-sensitive mutants whose plasma membranes are more readily permeabilized by copper. The common characteristic of these new copper-sensitive mutants is that they have an altered cell surface, which weakened their resistance to copper. These results help to explain the toxic effects of copper and suggest novel therapeutic strategies for fungal infections.
| The human fungal pathogen C. albicans typically grows as a commensal organism on human mucosa. However, C. albicans can cause severe mucosal infections or lethal systemic infections when the immune system is impaired [1, 2]. Serious infections also occur when conditions promote an overgrowth of C. albicans that overwhelms the immune system. This can happen as a consequence of the use of antibacterial antibiotics that disrupt the microbiota or as the result of biofilm formation on medical devices and catheters. In order to survive in a human host, C. albicans must be able to resist a wide range of stressful conditions promoted by the immune system. This includes elevated temperature, antimicrobial peptides, oxidation, and nitrosylation [3–5]. Copper has recently been recognized as a form of stress encountered by microbes in vivo [6–8], and C. albicans cells have been reported to experience copper stress at sites of infection [9, 10]. For example, stimulation of macrophages with interferon-γ leads to increased expression of the copper importer CTR1 and translocation of the ATP7A copper transporter to the phagosomal membrane where it pumps copper into the phagosome [11]. Copper can react with H2O2 produced by the oxidative burst in phagosomes to form a broader array of damaging reactive oxygen species [6, 12, 13].
Cells have a complex relationship with copper in that excess copper is toxic, but too little copper is also bad, as cells require this metal as an essential cofactor for many enzymes. The ability of copper to transition between cuprous (Cu1+) and cupric (Cu2+) oxidation states facilitates the catalysis of many important electron transfer reactions in the cell. Fungal cells therefore use several mechanisms to tightly regulate the levels of copper such that there appears to be less than one free copper molecule per cell [14, 15]. One mechanism is that excess copper inactivates the Mac1 transcription factor to turn off expression of the copper importer CTR1, thereby decreasing copper uptake [16]. Once inside the cell, intracellular copper is typically sequestered by a chaperone protein for delivery to an appropriate metalloprotein, such as the role of Ccs1 in delivering copper to the Sod1 superoxide dismutase [17]. Excess cytoplasmic copper is bound by scavenger proteins, such as the Cup1 and Crd2 metallothionein proteins in C. albicans, or stored in the vacuole [18–21]. In Cryptococcus neoformans, metallothioneins play a key role in copper resistance and are important for virulence [7, 22]. In C. albicans, the major role in resisting copper is carried out by Crp1, a plasma membrane exporter which pumps excess copper out of the cell, while metallothioneins play a less important role [18, 19]. Consistent with this, CRP1 is important for virulence of C. albicans [9]. The C. albicans copper resistance genes CUP1 and CRP1 are induced in response to excess copper by the Cup2 transcription factor [18, 19, 23, 24], and copper resistance also appears to be regulated by the PKA pathway [25].
Recent studies revealed that the Sur7 plasma membrane protein promotes resistance of C. albicans to copper by a mechanism that is distinct from the known pathways mediated by Crp1 and Cup1 [26]. Sur7 is a tetraspan integral membrane protein that localizes to punctate patches in the plasma membrane known as MCC domains, each of which is associated with a complex of cytoplasmic proteins known as an eisosome [27–29]. These domains correspond to ~250 nm long furrows in the plasma membrane created by the Pil1 and Lsp1 proteins [30–32]. The sur7Δ mutant showed strong defects in virulence and a decreased ability to grow in macrophages that correlated with increased sensitivity to copper [26]. The sur7Δ mutant has additional defects in morphogenesis, plasma membrane, and cell wall composition that likely contribute to the virulence defects [31, 33–35].
Many of the toxic effects of copper are attributed to its redox properties, which can lead to oxidation of protein, lipids and DNA [6–8, 36]. Excess copper can also inhibit growth of microbes by having direct toxic effects on a wide range of macromolecules, especially those containing thiols, such as the iron sulfur cluster proteins [6, 7, 37]. Previous studies showed that copper can permeabilize the plasma membranes of diverse cell types, including bacterial, fungal, plant, and animal [12, 36, 38–41]; however, the mechanism is not understood. To better define the novel process by which Sur7 promotes resistance to copper, we used genetic approaches to screen for C. albicans mutants that are more susceptible to killing by copper. The strongest mutants identified are similar to sur7Δ cells in having defects in membrane trafficking and cortical actin organization that cause broad changes in plasma membrane architecture (pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ). Interestingly, these mutants were all permeabilized more readily by copper. Analysis of cell surface lipids implicated exposure of phosphatidylserine (PS) in copper sensitivity, which correlates with studies on model membranes demonstrating that copper binds PS with high affinity and promotes membrane damage [42–44]. These results provide important new insight into how plasma membrane architecture is organized to protect the cell surface from attack by copper. These results also have significance for the development of novel therapeutic approaches and for the effective use of metallic copper surfaces as an antimicrobial strategy [45, 46].
To determine whether other proteins present in the MCC/eisosome domains of the C. albicans plasma membrane contribute to copper resistance, we tested 8 different mutant strains for their ability to grow on agar medium containing 500 μM CuSO4 (Fig 1A). All of the mutants exhibited wild-type growth, except the pil1Δ lsp1Δ mutant, which was similar to the sur7Δ strain in failing to show detectable growth. Pil1 and Lsp1 are membrane binding BAR domain proteins that function to form the eisosome domains [27, 30, 32]. MCC/eisosome domains facilitate Sur7 function by promoting its stability and localization in the plasma membrane [31]. We therefore tested the effects of overexpressing SUR7 in the pil1Δ lsp1Δ mutant and found that it strongly rescued the growth of pil1Δ lsp1Δ cells on copper-containing medium (Fig 1B), similar to the way that overexpression of SUR7 was found previously to rescue many of the cell wall and morphogenesis mutant phenotypes of pil1Δ lsp1Δ cells [31]. These results indicate that Sur7 plays the key role in copper resistance, and that Pil1 and Lsp1 act to facilitate this role.
To quantify the sensitivity to copper, mutant strains were grown in liquid cultures containing different concentrations of copper. Both the sur7Δ and the pil1Δ lsp1Δ mutants showed greatly increased sensitivity to copper, as they were inhibited by about a 100-fold lower dose of copper under these growth conditions (Fig 1C). Significantly, they were nearly as susceptible to copper as the highly sensitive cup1Δ crp1Δ mutant, which lacks both the metallothionein and the copper exporter genes [18]. The concentrations of copper needed to inhibit cell growth were affected by the growth medium, as the inclusion of components such as amino acids or yeast extract elevated the concentration of copper needed to kill cells, presumably because these media additives can chelate the copper. Therefore, these tests were done with minimal medium.
Time course studies were carried out by treating cells with 0.5 μM copper in water and then measuring the viable colony forming units (CFUs). Interestingly, whereas wild type cells slowly lost viability over a 2 hr incubation, the sur7Δ and pil1Δ lsp1Δ mutant cells rapidly lost viability as most of the cells were dead after 30 min (Fig 1D). The cup1Δ crp1Δ strain appeared to give intermediate sensitivity in this short-term assay, although variability in the CFU assays limited the statistical significance of these results. This was more readily observed in an independent set of time course assays performed under more acidic conditions that prolong the time course (S1 Fig).
To better define how cells resist copper toxicity, we screened libraries of C. albicans mutant strains [23, 47, 48], along with deletion strains in our own collection, by replica-plating cells onto copper-containing medium. The most sensitive mutants were also tested for growth in the presence of 10 other metals to examine the specificity of the copper phenotype. A heat map summary (Fig 2A) of the 18 most copper-sensitive mutants revealed that they mainly fell into three categories. One expected category included mutants lacking known copper resistance genes (crp1Δ and cup2Δ) that were characterized by high sensitivity to copper, but not the other metals. A second group contained HOG MAP kinase pathway mutants (pbs2Δ, ssk2Δ, hog1Δ), and the third was comprised of sur7Δ and other mutants with defects in morphogenesis and membrane trafficking (Fig 2A). These latter two mutant categories could be readily distinguished by their sensitivity to other metals. For example, the HOG pathway mutants were inhibited by calcium and manganese, whereas many of the morphogenesis mutants were sensitive to cadmium and chromium (Fig 2A). Although sur7Δ cells appeared more sensitive to cadmium, chromium, and cesium when assayed for growth by replica-plating onto agar plates, growth in liquid cultures containing different concentrations of metals indicated that the sur7Δ mutant was only weakly more susceptible to these other metals (Fig 2B). Thus, the sur7Δ mutant is preferentially affected by copper.
We next examined whether the copper sensitivity of sur7Δ and other C. albicans mutants identified in the screen was due to oxidative stress generated by copper. Copper is known to react with compounds such as H2O2 to generate diverse reactive oxygen species (ROS) that damage cells [6]. This is due in part to the ability of copper to transition between cuprous (Cu1+) and cupric (Cu2+) oxidation states, enabling it to undergo Fenton-like chemical reactions. Interestingly, the sur7Δ mutant showed at most a weak increase in sensitivity to oxidative stress when grown on medium containing H2O2 (Fig 3A). In contrast, control strains showed that a catalase mutant (cat1Δ) and the HOG pathway mutants (pbs2Δ, ssk2Δ,and hog1Δ) displayed increased sensitivity to H2O2, as expected based on previously published data [49, 50]. The cat1Δ catalase mutant grew well on medium containing copper, indicating that the presence of this metal in the medium did not cause high levels of oxidative stress (Fig 3B). In addition, the cup1Δ crp1Δ strain, which is very sensitive to copper, grew similar to the wild type control strain on medium containing H2O2. Interestingly, the pil1Δ lsp1Δ mutant was more sensitive to H2O2, which is likely to be due to effects on a family of four related antioxidant proteins that localize to eisosomes [51]. Altogether, these results indicate that the copper sensitivity of the sur7Δ mutant is not due to increased susceptibility to oxidative stress.
Since the members of the largest group of copper-sensitive mutants we identified were similar to sur7Δ in having defects in membrane trafficking and morphogenesis, we examined three other mutants that are known to have strong defects in these processes and to also show altered plasma membrane organization. Two mutants, rvs161Δ and rvs167Δ, lack BAR domain proteins that are needed for proper cortical actin organization and play an important role in the scission phase of endocytosis [52]. In addition, we tested a double arp2Δ arp3Δ mutant of the actin-related proteins Arp2 and Arp3 that form a complex that is needed for cortical actin localization and efficient endocytosis [53]. All three strains showed increased inhibition of growth when replica-plated onto agar medium containing copper (Fig 4A). Incubation in a range of copper concentrations in liquid media showed that all three mutants were similar to sur7Δ in that they could only resist low concentrations of copper. The arp2Δ arp3Δ strain was the most sensitive, as it could only grow in copper concentrations up to 3.2 μM, while the wild type grew at ≥400 μM (Fig 4B).
A common feature of the copper-sensitive mutants sur7Δ, pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ is that they have abnormal actin organization [33, 52, 53]. We therefore examined the possibility that resistance to copper is dependent upon an intact actin cytoskeleton. This was tested by treating the wild type control strain DIC185 with the actin polymerization inhibitor cytochalasin A for 2 hr to disrupt the actin cytoskeleton, and then incubating the cells in the presence or absence of different concentrations of copper. Interestingly, over the range of copper concentrations used, cells incubated in water showed about a 5-fold decrease in viability, whereas cytochalasin A-treated cells showed about a 100-fold decrease in viability (Fig 4C). This synergistic effect on the loss of viability when cytochalasin A-treated cells were exposed to copper supports the conclusion that normal actin cytoskeleton and plasma membrane organization promote resistance of C. albicans to copper.
The observation that sur7Δ cells have altered plasma membrane architecture, including mislocalization of cortical actin, septins, and the lipid PI4,5P2 [26, 31, 33], suggested the possibility that copper reduces the viability of these cells by permeabilizing the plasma membrane. Copper has been shown to permeabilize the plasma membranes of a wide range of cell types [38–41], but the mechanism is not known and no mutants have been described previously that are hypersensitive to this process. To examine how copper affects the plasma membrane barrier function, we incubated C. albicans cells with different concentrations of CuSO4 for 2 hr, and then treated the cells with the membrane impermeable dye SYTOX Green that can only stain intracellular nucleic acids if the plasma membrane has been compromised. Interestingly, the sur7Δ cells showed significantly increased staining in a dose-dependent manner compared to the wild type and sur7Δ + SUR7 complemented strains (Fig 5A and S2 Fig). A time course analysis demonstrated that sur7Δ cells experienced more membrane damage than other strains even after a short 30 min incubation in copper (Fig 5B). Similar results were obtained with other membrane-impermeable stains, such as propidium iodide and FM4-64, and copper reduced the fluorescence of sur7Δ cells carrying the plasma membrane localized Pma1-GFP fusion protein, further supporting the conclusion that copper was permeabilizing the plasma membrane and having toxic effects (S3 Fig).
The other copper-sensitive morphogenesis mutants rvs161Δ, rvs167Δ, and arp2Δ arp3Δ showed strong SYTOX Green staining after incubation in copper, with arp2Δ arp3Δ exhibiting the highest level of staining (Fig 5C). This correlated with arp2Δ arp3Δ being the most copper-sensitive strain (Fig 4B). In contrast, the mutant strains that lack the Crp1 copper exporter or the Cup1 metallothionein that sequesters intracellular copper (crp1Δ, cup1Δ crp1Δ, and cup2Δ) showed only a small increase in SYTOX Green staining after copper treatment, which was significantly lower than that seen for sur7Δ and the other morphogenesis mutants (Fig 5D). This indicates that plasma membrane permeabilization by copper is separate from the toxic effects of intracellular copper.
The relationship between cell morphology and SYTOX Green staining was examined since sur7Δ cells are known to range from typical-looking budding cells to abnormally large cells [33]. Interestingly, about 65% of large cells (≥ 7 μm) stained with SYTOX Green after copper treatment compared to only about 25% of normal sized cells (Fig 6A and 6C). Although the larger cells were permeabilized more readily, it is significant that the typical-sized cells were also susceptible to copper. The cell cycle stage of the sur7Δ cells also affected permeabilization by copper, as about 60% of budded cells stained compared to only 10% of unbudded cells (Fig 6B and 6C). This correlates with increased function of the actin cytoskeleton and vesicle trafficking during polarized bud morphogenesis.
The increased ability of copper to permeabilize the mutants suggested that their cell surfaces may be altered, so plasma membrane organization was probed by assaying sensitivity to compounds that target specific plasma membrane lipids. The sur7Δ, pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ mutants all showed essentially the same sensitivity as wild type cells to the polyene antibiotic amphotericin B (Fig 7A), which binds ergosterol in the plasma membrane [54]. In contrast, the mutant cells were more sensitive to the lanthionine antibiotics cinnamycin and duramycin (Fig 7B and 7C), which bind phosphatidylethanolamine (PE) if it is in the outer leaflet of the plasma membrane [55]. Similarly, all of the mutants were more susceptible to the depsipeptide antibiotic papuamide A (Fig 7D), which binds phosphatidylserine (PS) in the outer leaflet [56, 57]. PE and PS are normally enriched in the inner leaflet of the plasma membrane by proper membrane trafficking and the action of flippase proteins that promote inward translocation of these lipids [58]. The increased sensitivity to these drugs therefore indicates an altered surface of the plasma membrane. Interestingly, the relative increased susceptibility of the sur7Δ, pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ mutants to cinnamycin, duramycin, and papuamide A correlated with their relative sensitivity to permeabilization by copper, consistent with an altered presentation of lipids on the cell surface contributing to the ability of copper to permeabilize a membrane.
To confirm these results, we mutated members of the phospholipid flippase gene family in C. albicans, which encode P4-ATPases that promote inward translocation of lipids. Flippases are responsible for maintaining phospholipid asymmetry in the plasma membrane by translocating PS and PE from the outer leaflet to the inner leaflet of the plasma membrane [58]. Little or no analysis of flippases has been carried out in C. albicans, but in S. cerevisiae NEO1 is essential and the other four have redundant functions (DNF1, DNF2, DNF3 and DRS2). Surprisingly, although the apparent C. albicans ortholog of NEO1 (C1_04630C) is not essential, a neo1Δ mutant showed a detectable increase in sensitivity to both duramycin and copper (Fig 7E). Similar results were obtained for the dnf1Δ (lacking C5_00570W). Interestingly, the drs2Δ mutant (lacking C3_07230W) showed the strongest increase in sensitivity to both duramycin, and copper. These results provide additional evidence that altered plasma membrane organization contributes to the increased sensitivity to copper.
To test whether the sur7Δ strain was altered in response to other agents that compromise plasma membrane integrity, we treated cells with the polycations DEAE dextran hydrochloride (500 kDa), and poly-L-lysine hydrobromide (30 kD). Poly-L-lysine with molecular weight > 25 kDa was reported to permeabilize yeast cells, as were DEAE dextrans with masses from 5 to 380 kDa [59]. The sur7Δ mutant showed slightly increased viability after exposure to either 0.1 μg/ml poly-L-lysine or 0.5 μg/ml DEAE dextran compared to the wild type control, indicating it was more resistant to this type of membrane disruption (S4 Fig) These results further confirm that copper has a specific effect on the sur7Δ plasma membrane.
Previous studies indicated that introducing polyunsaturated fatty acids (PUFAs) into the S. cerevisiae plasma membrane, which normally lacks PUFAs, increased the susceptibility to permeabilization by copper [60]. However, this did not seem likely to be a factor for C. albicans, as fatty acid analysis indicated that there were no significant differences in PUFA content in C. albicans strains between the wild type control, sur7Δ, and the pil1Δ lsp1Δ mutant (S1 Table). Furthermore, in S. cerevisiae the susceptibility to copper permeabilization plateaued at 20% PUFAs, and C. albicans naturally contains about 30% PUFAs [61].
The increased sensitivity to papuamide indicated that PS was inappropriately exposed on the plasma membrane surface. Previous studies demonstrated that PS binds copper with very high affinity, and that this interaction can promote membrane damage that could lead to permeability [43, 44, 62]. To test the role of PS, we deleted the CHO1 gene that encodes PS synthase [63]. However, the sur7Δ cho1Δ cells that lack PS were more resistant to copper than were the parental sur7Δ cells (Fig 8A). The sur7Δ cho1Δ double mutant consistently showed decreased permeabilization by copper relative to the sur7Δ single mutant (Fig 8B). Although there was some day to day variability in the total number of cells stained with Sytox Green, the percent of stained sur7Δ cho1Δ cells subtracted from the sur7Δ single mutant value averaged 9.9 + 2.3%. Altogether, these results indicate that abnormal plasma membrane organization leads to exposure of PS on the plasma membrane where it can be bound and attacked by copper.
Copper resistance is important for a wide range of bacterial and fungal pathogens [8, 11, 22, 64, 65]. Previous studies indicated that the increased sensitivity of sur7Δ mutant cells to copper-mediated killing was not due to a defect in the Crp1 copper exporter, which carries out the major known copper resistance mechanism in C. albicans [18, 19, 26]. Therefore, this study was initiated to better understand the role of the MCC/eisosome protein Sur7 in promoting resistance to copper. Analysis of a set of mutants lacking different MCC/eisosome proteins revealed that a pil1Δ lsp1Δ mutant, which is defective in forming the furrow-like membrane invaginations corresponding to MCC/eisosomes, was also more sensitive to copper (Fig 1). Sur7 is not stably maintained in the plasma membrane of pil1Δ lsp1Δ mutants [31], so we overexpressed SUR7 and found that this strongly rescued the copper defect of the pil1Δ lsp1Δ cells, indicating that Sur7 plays the key role in copper resistance, rather than eisosome furrows that can form in the absence of Sur7. Further analysis of the sur7Δ cells showed that they were not highly sensitive to other metals (Fig 2), which indicates that the defect is specific to copper and not due to alteration in vacuolar function that is associated with increased sensitivity to multiple metals [20, 66, 67]. The effect of copper on the sur7Δ cells was not due to a general ability of copper to catalyze creation of reactive oxygen species, since the sur7Δ mutant did not show a correspondingly large increase in susceptibility to oxidative stress (Fig 3). These results indicate that Sur7 carries out a novel role in copper resistance.
Genetic screening of C. albicans mutants was used to gain further insight into the mechanisms underlying the increased copper sensitivity of sur7Δ cells. Interestingly, most of the copper-sensitive mutants that were identified have defects in membrane organization or trafficking (Fig 2). Targeted analysis of three mutants known to have strong defects in plasma membrane function and cortical actin (rvs161Δ, rvs167Δ, and arp2Δ arp3Δ mutants) revealed that they all showed increased susceptibility to killing by copper that was similar to or greater than the sur7Δ mutant (Fig 4). In support of a role for the actin cytoskeleton, wild type cells treated with the actin inhibitor cytochalasin A also displayed increased sensitivity to copper (Fig 4C). Furthermore, overexpression of SUR7, which rescued the copper defect of pil1Δ lsp1Δ cells (Fig 1B), was shown in a previous study to improve plasma membrane organization of cortical actin [31]. Altogether, these results indicate that altered plasma membrane organization caused by defects in the MCC/eisosome protein Sur7 or disruption of the actin cytoskeleton increases the susceptibility to killing by copper.
Copper binds strongly to membranes, which suggested it could have a direct effect on lipids. In particular, copper binds to PS with picomolar affinity and to PE with micromolar affinity [43, 44]. Copper is thought to bind PS with such high affinity relative to other divalent metals because it can interact in a special configuration with two molecules of PS [42–44]. Usually PS and PE are sequestered on the inner leaflet of the plasma membrane by the action of phospholipid flippases [68]. To determine whether these phospholipids were exposed on the cell surface of the copper-sensitive mutants, the mutant cells were tested for increased sensitivity to cinnamycin and duramycin, which bind to PE, and to papuamide A, which binds to PS (Fig 7). There was a good correlation between the relative increases in sensitivity to copper and to these drugs that target PS and PE, consistent with copper sensitivity reflecting the degree to which the surface exposure of PS and PE is altered in the mutants. The mutants that were most susceptible to being permeabilized by copper (sur7Δ, pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ) all had broad defects in plasma membrane structure that could result in the cell surface exposure of PS and PE [33, 52, 53, 69]. Furthermore, three different phospholipid flippase deletion mutants displayed increased sensitivity to copper, especially the drs2Δ mutant (Fig 7). We therefore examined the effects of deleting the gene encoding PS synthase (CHO1) to deplete cells of PS. Remarkably, the sur7Δ cho1Δ double mutant showed greatly improved resistance to copper (Fig 8). These results strongly support the conclusion that copper has a direct effect on the plasma membrane, and implicate interaction with PS as part of the underlying mechanism for copper-mediated killing of the mutant cells.
Although plasma membrane permeabilization can be a consequence of cell death, several lines of evidence indicate that copper kills sur7Δ cells and the other new copper-sensitive mutants we identified by promoting membrane permeabilization. Previous studies have shown that copper is distinct from other divalent metal ions in its ability to bind and damage model membranes containing PS or PE [42–44]. Atomic force microscopy studies showed copper promotes lipid reorganization and affects membrane fluidity that could weaken the membrane barrier function [70, 71]. The ability of excess copper to permeabilize red blood cell membranes was correlated with a loss of membrane fluidity and decreased ability of the cells to tolerate deformation [72]. Copper has been known to permeabilize the plasma membranes of fungi and other cell types from plants to animals for 30 years, and has been used as a tool to form pores in the plasma membrane to selectively release cytoplasmic pools of amino acids [36, 38–41], but the mechanism by which copper permeabilizes the plasma membrane is not known. It is therefore significant that the sur7Δ, pil1Δ lsp1Δ, rvs161Δ, rvs167Δ, and arp2Δ arp3Δ mutants represent the first mutants reported for being more susceptible to copper permeabilization (Fig 5), as they provide new insights into the mechanisms underlying this process. Once cells become permeabilized, copper would be expected to enter cells and cause further lethal damage. In contrast, control studies showed that copper did not cause a significant increase in plasma membrane permeability for the crp1Δ, cup1Δ crp1Δ, and cup2Δ mutants in spite of their greatly increased sensitivity to this metal (Fig 5D). This supports the conclusion that there are independent pathways for killing by copper.
The special redox properties of copper suggest that it damages membranes by promoting oxidation. Studies with copper-treated model membranes demonstrate that this metal can promote oxidation of both PS and PE [42, 44]. In our studies, the oxidation-sensitive C. albicans mutant pst1Δ pst2Δ pst3Δ ycp4Δ mutant (Fig 1A) and a cat1Δ catalase mutant (Fig 3) did not show a significant increase in susceptibility to copper. However, it may be that these intracellular proteins do not have a direct effect on oxidation events in the outer leaflet of the plasma membrane. Furthermore, other data showed that the oxidative effects of copper on PS are differentially affected by different subsets of anti-oxidant molecules, indicating that not all antioxidant pathways are likely to be equally effective [62]. Thus, copper may be acting in a specialized microenvironment at the cell surface to promote oxidative damage to membranes leading to permeabilization.
The plasma membrane is an effective target for therapy, as the commonly used antifungal drugs alter the plasma membrane directly or indirectly (i.e. fluconazole, amphotericin, caspofungin) and many antimicrobial peptides act by permeabilizing the plasma membrane [54, 73]. One reason the antifungal drugs are effective is that even at sublethal doses they can broadly impact plasma membrane organization [74]. Proper architecture of proteins and lipids is required to form a strong protective barrier around the cell, and it plays key roles in virulence by coordinating a wide range of dynamic functions including secretion of virulence factors, cell wall synthesis, invasive hyphal morphogenesis, endocytosis, and nutrient uptake [74]. The identification in this study of new copper-sensitive mutants and the role of phosphatidylserine provides insights for new modes of antifungal therapy. Copper is amenable to incorporation into therapeutic strategies [75, 76]. Drugs that alter the plasma membrane surface may be used to increase the susceptibility of cells to copper in the host [6–10], and conversely copper could be used to enhance the membrane perturbing effects of currently used antifungal drugs [77]. In addition, our results have important implications for understanding microbial killing by metallic copper surfaces, which have been explored for preventing the spread of nosocomial infections [45, 46]. Thus, copper could be used in a synergistic fashion with other therapeutic strategies to more effectively kill fungal pathogens.
The C. albicans strains used in this study appear in Table 1. Cultures were grown either in rich YPD medium (2% dextrose, 1% peptone, 2% yeast extract, 80 mg/L uridine), or synthetic medium (yeast nitrogen base, 2% dextrose) [78]. Synthetic medium was supplemented with amino acids and uridine, if required. The pbs2Δ, ssk2Δ, hog1Δ, cup2Δ, and neo1Δ strains [47] were made prototrophic by transformation with a PCR-amplified ARG4 DNA fragment that was integrated into the native locus. Primers used to amplify the cassette included 18 base pairs of homology approximately 500 base pairs upstream and downstream of the ARG4 open reading frame. The sla2Δ, vps28Δ, vps16Δ, cka2Δ, ckb2Δ, and ssn8Δ strains [48] were transformed with a similarly constructed HIS1 cassette to correct the remaining auxotrophy. PMA1-GFP strains were constructed with the photostable GFPγ variant, as described previously [79]. Homozygous dnf1Δ (orf19.932/C5_00570W) and drs2Δ (19.6778/ C3_07230W) mutants were constructed from C. albicans parental strain BWP17 by the sequential deletion of both copies of the targeted gene [80]. Deletion cassettes were generated by PCR amplification of the ARG4 or HIS1 selectable marker genes, using primers that included 70 to 80 base pairs of DNA sequence homologous to the upstream or downstream regions of the targeted open reading frame. Cells which had undergone homologous recombination to delete the targeted gene were identified by PCR analysis using a combination of primers flanking sites of cassette integration and internal primers. Homozygous cho1Δ mutants were constructed in C. albicans parental strain DIC185 and the previously described sur7Δ deletion (33), using the transient CRISPR-Cas9 system [81]. The gene-deletion construct was synthesized using the SAT1 flipper as a template and primers that included 80 base pairs upstream and downstream of the targeted gene [82]. Homozygous deletion clones were identified by rescue of the ethanolamine auxotrophy phenotype on synthetic medium with 1mM ethanolamine [63], and by PCR analysis using primers flanking the CHO1 gene, as well as SAT1 internal primers.
Mutant strain libraries of S. Noble [47], A. Mitchell [48], and O. Homann [23], along with strains in our own collection were replica-plated onto synthetic agar medium containing 500 μM, 250 μM, 100 μM, 20 μM, or 0 μM CuSO4, plus added arginine, histidine, or uridine, if necessary. Other metals were tested similarly by replica-plating onto solid agar medium containing 1.5 M NaCl, 1.5 M KCl, 150 mM LiCl, 800 mM RbCl, 300 mM CsCl, 3 mM CoCl2, 300 μM CdCl2, 100 mM MnCl2, 20 mM ZnCl2, 4 mM NiCl2, 2 mM FeSO4, 200 μM CrO3, and 500 mM CaCl2. The concentrations of these metals were determined by identifying a dose at which the growth of wild type cells was reduced, but not lethal. Plates were incubated at 30°C for 48-72 hr, and then the extent of growth inhibition was recorded daily on a scale of 0 to 5, with 0 indicating the level of growth of the wild type strain and 5 indicating no growth. The results represent the average of at least two assays for screening each of the different metals, and then the susceptible mutants were confirmed in additional assays. The mutant strains were clustered using the program Cluster, and then a heat map was generated. The mutants that were not detectably sensitive to copper are listed in the Supplementary Information (S2 Table).
Strains were grown overnight at 30°C in synthetic medium. Cells were then diluted to 5 x 103 cells/ml in synthetic medium and 300 μl was added to the first row of wells of a 96-well plate (Costar, Corning, Inc., Kennebunk, ME). 240 μl of cells was applied to subsequent rows of the plate. CuSO4 was then added to wells of the first row to a final concentration of 10 mM and serial dilutions were made by transferring 60 μl from one row to the next. Plates were then covered with AeraSeal oxygen permeable sealing film and incubated at 37°C for 48 hr. The highest concentration of copper that supported growth was recorded. Similar methods were used to quantify the sensitivity of C. albicans strains to other metals. Viability following exposure to copper was assayed by growing strains overnight in synthetic medium to log phase at 30°C. The cells were then washed with sterile H2O and diluted to 3.0 x 103 cells/ml in sterile 1 mM MES buffer pH 6 in the presence or absence of 0.5 μM CuSO4. Cells were then incubated at 37°C for the indicated time, 100 μl of each sample was then spread onto YPD agar plates, incubated at 30°C for 48 hr, and then CFUs were quantified. The effects of the actin polymerization inhibitor cytochalasin A (Sigma-Aldrich, St. Louis, MO) were examined with cells that were grown overnight in synthetic medium at 30°C to log phase, diluted to 1 x 106 cells/ml in synthetic medium, and then cytochalasin A was added to a final concentration of 10 μg/ml every 30 min for 2 hr, with incubation at 30°C. Cytochalasin was added every 30 min to maintain a high level of this drug as it is not stable under these conditions [83]. Cells were then washed with sterile H2O, diluted to 3 x 103 cells/ml in sterile H2O, a final aliquot of cytochalasin A was added, and cells were incubated with the indicated concentrations of CuSO4 for 2 hr at 37°C. 100 μl of each sample was then spread onto YPD agar plates and incubated at 30°C for 48 hr to determine the viable CFUs. Statistical analysis of the data was carried out using Prism software (GraphPad Inc.).
SYTOX Green (Invitrogen, Molecular Probes, Eugene, OR) is a membrane impermeable nucleic acid stain that can be used to assay plasma membrane integrity [84]. For the analysis, strains were grown in synthetic medium overnight at 30°C to log phase, washed in sterile H2O, and diluted to 1 x 107 cells/ml. Following incubation in CuSO4 with 1 mM MES buffer pH 6 at 37°C, cells were washed in sterile H2O, SYTOX Green was added to a final concentration of 2.5 nM, and cells were incubated at room temperature for 5 min. Cells were washed again in sterile H2O and then analyzed by fluorescence microscopy. Propidium iodide (Invitrogen, Molecular Probes, Eugene, OR) staining was carried out in a similar manner, except the stain was added to a final concentration of 4.2 μg/ml, and incubation was carried out for 15 min at room temperature. Cells were stained with the lipophilic dye FM4-64 (Invitrogen, Molecular Probes, Eugene, OR) by adding it to a final concentration of 20 μM [85]. Cells were incubated on ice for 20 min, washed in cold sterile H2O, and viewed immediately by fluorescence microscopy. Images were obtained using an Olympus BH2 microscope (Melville, NY) equipped with a Zeiss AxioCam digital camera (Thornwood, NY).
Cells were grown overnight in synthetic medium at 30°C, diluted to 1.0 x 106 cells/ml in synthetic medium, and 250 μl was spread onto the surface of a synthetic medium agar plate. Paper filter disks (Becton Dickinson and Company, Sparks, MD) containing 10 μl of the concentration of drug to be assayed were applied onto the surface of plates containing the indicated strains. After incubation for 48 hr at 30°C, the diameters of the zones of growth inhibition were measured. Drugs tested included duramycin (Sigma Aldrich, St. Louis, MO), cinnamycin (Santa Cruz Biotechnology, Santa Cruz, CA), papuamide A (Flintbox, British Columbia, Canada), and amphotericin B (Sigma-Aldrich, St. Louis, MO). Determination of sensitivity to the oxidizing agent H2O2 (Fisher Scientific, Fair Lawn, NJ), was performed by spot assay. Cells grown overnight in synthetic medium at 30°C were diluted to 1.0 x 107 cells/ml in synthetic medium and a 10-fold dilution series was made. 2 μl of each dilution was spotted onto YPD containing the indicated concentrations of H2O2 and plates were incubated at 30°C for 48 hr and then photographed. Strains tested for H2O2 sensitivity were also spotted onto SD medium containing CuSO4 for confirmation of copper sensitivity. To assay cells for reaction to polycations, strains were grown overnight in synthetic medium, washed in H2O, diluted to 3 x 103 cells/ml in H2O, and then incubated with DEAE dextran hydrochloride (500 KDa, Sigma-Aldrich, St. Louis, MO) or poly-L-lysine hydrobromide (30 kDa, Sigma-Aldrich, St. Louis, MO) for 2 hr at 37°C. 100 μl of each sample was then spread onto YPD medium, incubated at 30°C for 48 hr, and then CFUs were quantified to assess viability.
Cells were grown overnight in synthetic medium at 30°C, diluted to 1.0 x 106 cells/ml in 50 ml synthetic medium, and then incubated with shaking at 37°C until a concentration of approximately 1.0 x 107 cells/ml. Cells were then harvested by centrifugation, washed in sterile H2O, and then frozen cell pellets were shipped to Microbial ID, Inc. (Newark, DE) for fatty acid analysis. Sample preparation involved saponification, methylation, and extraction of fatty acid methyl esters. Gas chromatography was used to create a cellular fatty acid profile identifying the percent of each fatty acid species and localization of unsaturated bonds. For a second analysis, cells were streaked onto YPD medium, incubated at 30°C, and then submitted to Microbial ID, Inc. for testing.
|
10.1371/journal.pbio.1001789 | Metabolic Erosion Primarily Through Mutation Accumulation, and Not Tradeoffs, Drives Limited Evolution of Substrate Specificity in Escherichia coli | Evolutionary adaptation to a constant environment is often accompanied by specialization and a reduction of fitness in other environments. We assayed the ability of the Lenski Escherichia coli populations to grow on a range of carbon sources after 50,000 generations of adaptation on glucose. Using direct measurements of growth rates, we demonstrated that declines in performance were much less widespread than suggested by previous results from Biolog assays of cellular respiration. Surprisingly, there were many performance increases on a variety of substrates. In addition to the now famous example of citrate, we observed several other novel gains of function for organic acids that the ancestral strain only marginally utilized. Quantitative growth data also showed that strains with a higher mutation rate exhibited significantly more declines, suggesting that most metabolic erosion was driven by mutation accumulation and not by physiological tradeoffs. These reductions in growth by mutator strains were ameliorated by growth at lower temperature, consistent with the hypothesis that this metabolic erosion is largely caused by destabilizing mutations to the associated enzymes. We further hypothesized that reductions in growth rate would be greatest for substrates used most differently from glucose, and we used flux balance analysis to formulate this question quantitatively. To our surprise, we found no significant relationship between decreases in growth and dissimilarity to glucose metabolism. Taken as a whole, these data suggest that in a single resource environment, specialization does not mainly result as an inevitable consequence of adaptive tradeoffs, but rather due to the gradual accumulation of disabling mutations in unused portions of the genome.
| Adaptation to a single constant environment is commonly expected to result in decreased performance in alternative conditions, or specialization. It has been proposed that, rather than occurring through the neutral accumulation of mutations in unused alternative pathways, this happens because loss of these pathways enhances fitness in the constant environment via “tradeoffs.” We examined growth rates across a variety of nutrients for 12 independent lineages of Escherichia coli that had evolved in the laboratory for decades in a glucose-containing medium. Surprisingly, after 20,000 generations there were actually widespread improvements in the use of alternative nutrients, rather than the expected declines. After 50,000 generations, however, we find that this trend reversed for those populations that evolved a much higher mutation rate. This indicates that high mutation rate, and not adaptive tradeoffs per se (as had been previously proposed), is the primary driver of specialization. These results caution against general assumptions about the importance of adaptive tradeoffs during evolution, and emphasize the key role that newly evolved changes in mutation rate can play in promoting niche specialization.
| Evolving populations face the fundamental dilemma that there is no single phenotype that is optimal in all environments. When an evolving population occupies the same selective environment for an extended period of time, no advantage is realized by maintaining fitness on resources it no longer encounters. Adaptation to a selective environment can result in correlated responses in alternative environments. Although these can be synergistic improvements, a response that decreases fitness in other environments is known as specialization. This prevents the rise of “Darwinian demons”: single supergenotypes that are optimized across all conditions [1]. It is critical to understand the origin of specialization because it underlies the origin and maintenance of diversity—it is why “the jack of all trades is a master of none” [2].
Specialization can result from either selective or neutral processes. Antagonistic pleiotropy describes when natural selection favors changes that are beneficial in the current environment but reduce function in other environments. Alternatively, specialization may result from mutations that decrease fitness in alternative environments that are neutral in the selective environment. These mutations have the potential to either drift or hitchhike to fixation via “mutation accumulation.” As neutral mutations accrue in proportion to the mutation rate, the clearest evidence of mutation accumulation can come from excess specialization in mutator lineages, which contain defects in mutational repair that can elevate mutation rates ∼100-fold [3]. In contrast, where specialization is rapid and occurs in parallel across lineages, a pattern commonly seen for adaptation itself, this has been cited as support of selection-driven antagonistic pleiotropy [4].
The experimental evolution of 12 populations of Escherichia coli grown for thousands of generations on a single substrate has been used to distinguish whether selective or neutral processes drive metabolic specialization [4]. The populations were part of the Lenski Long-Term Evolution Experiment (LTEE) [5], in which wild-type E. coli B have been diluted 1∶100 daily and regrown in well-mixed medium containing glucose as the sole usable carbon source. After 20,000 generations (20k), competitive fitness on glucose had increased by ∼70%. However, respiration assays in static 96-well Biolog plates (Hayward, CA) [6] suggested dramatic decreases in metabolic performance on alternative substrates. These declines occurred rapidly and in parallel across populations, coincident with the largest gains in fitness. This was suggested to indicate selection-driven antagonistic pleiotropy as the main mechanism of specialization. Furthermore, because there was only a weak, nonsignificant excess of declines by the populations which had become mutators earlier in the experiment [3], this suggested neutral mutation accumulation played little role, if any, in glucose specialization by 20k.
In this study we have readdressed the basis of specialization in the LTEE populations, motivated by our discovery that the growth rates of isolates in well-mixed media are poorly captured by assays of cellular respiration in static, proprietary media. Given this surprising finding, we analyzed the selected (i.e., glucose) and correlated responses of isolates from both 20k and 50,000 generations (50k) from four perspectives:
As a first step to readdressing specialization in the LTEE populations, we sought to replicate the Biolog respiration results at 20k presented by Cooper and Lenski [4], as well as extend this analysis to the populations at 5ok. Despite changes in the Biolog assay itself, since the previous study, we recovered a similar pattern for the panel of substrates (Figure S1).
The validity of Biolog assays as a proxy for strain improvement came into question after finding decreases for the very substrates on which selection occurred (Figure 1A). Despite abundant evidence of improvement from competitive fitness assays [4] and growth rates [8], respiration on glucose consistently decreased over the course of the experiment. Furthermore, although one lineage in the A-3 population evolved to utilize as a carbon source the citrate included in Davis Minimal (DM) medium [7], it produced a statistically indistinguishable respiration value from the other Cit− isolates at 50k (Figure S2).
Given that the selective substrates with known fitness improvement had decreased cellular respiration, we turned to direct measurements of growth rates across a wide panel of substrates using a robotic growth analysis system [9],[10]. By measuring growth rate we capture the demographic metric best correlated with competitive fitness in the evolutionary environment [11]. As the LTEE was performed in shaken, fully aerated flasks, these well-mixed 48-well plates were a closer match to the evolutionary environment than unshaken 96-well plates, as unshaken plates commonly exhibit subexponential growth due to oxygen limitation [10]. Although Biolog uses a proprietary minimal media, for the growth rate measurements we used the same DM media as the LTEE experiment. We omitted citrate, however, as this choice allowed us to include the Cit+ A-3 population in our analysis. We chose carbon sources based upon the substrates for which significant, parallel decreases were previously observed via respiration assays [4], as well as citrate and several sugars on which growth tradeoffs were previously measured after 2,000 generations [12].
Comparing growth rates to respiration data, it becomes evident that the latter is not an accurate assay for growth (Figure 1B). There were many cases where respiration occurred without growth, as 156 out of the 702 strain/substrate combinations measured did not permit growth but did have measurable respiration—a known feature of Biolog assays [6]. Even after removing these categorical disagreements, and a smaller number of instances of growth without respiration (11 strain/substrate combinations), Biolog respiration values were a poor predictor of growth rate [R2 = 0.18, linear regression F test(1,499) = 108.8].
Growth rate data across substrates revealed a surprising degree of correlated gains in performance (Figure 2). Indeed, at 20k, there were actually more correlated increases in rate than decreases (165 versus 99, respectively, p<0.0001 for binomial two-sided test with null of random gains and losses). By 50k, the picture had reversed, now with more decreases than increases (167 versus 119, p = 0.005, binomial two-sided test).
In addition to many quantitative improvements of growth rates, there were several examples where isolates acquired the ability to grow on substrates that the ancestral strain could not utilize over the 48 h time-course of the growth experiments. Only one such example was previously known: the aforementioned gain of citrate utilization by the A-3 population [7]. We found that this strain also gained the ability to grow within 48 h on three C4-dicarboxylate tricarboxylic acid cycle intermediates (succinate, aspartate, and malate). Three other 50k isolates from different replicate populations gained the ability to use this same set of three C4 dicarboxylate intermediates, as well as fumarate.
We compared mutators to nonmutators to ask whether mutation accumulation contributed to the observed decreases in growth (Figure 2). At 20k generations, despite increasing in growth rate more than decreasing (47 versus 41 cases), the mutators were marginally worse, on average, than nonmutators (p = 0.03, Pearson's chi-squared test comparing proportion of growth rate reductions). By 50k there was a stark pattern of mutators declining in catabolic ability compared to nonmutators (p<0.0001, Pearson's chi-squared test). This can be seen in the large block of blue (decreases in rate) for five of the mutators. Nonmutators at 50k still increased in growth rate more often than they decreased (80 versus 52 cases, p = 0.018 binomial two-sided test). Because some strains are known to be affected by the absence of citrate even though they cannot use it as a carbon source [13], we also tested the growth rate of the 50k strains on alternative substrates supplemented with 1.7 mM sodium citrate, as in LTEE growth media. Although the reductions in growth rate relative to the ancestor were ameliorated in some cases by the addition of citrate, the mutator strains still suffered significantly more growth rate decreases than the nonmutators (p = 0.003, Pearson's chi-squared test).
Given the trend of both increased and decreased growth rate on alternative carbon sources, we assessed the degree of parallelism with which metabolic erosion occurred. We segmented the data by substrate and asked how many evolved strains decreased in growth rate or cellular respiration on each substrate. We took as a null expectation that decreases in metabolic function are equally as likely as increases, and plotted the observed pattern against this null distribution (Figure S3).
In no case do our observations of metabolic decreases closely match the null distribution. As previously, cellular respiration was reduced for nearly all strains on all substrates. The observed average number of strains with reduced respiration on a substrate was 11.3 at 20k and 12.8 at 50k, 5.3 out of 12 and 6.3 out of 13 more strains than would be expected given the null distribution (p<0.0001, binomial two-tailed exact test). For growth rate, at 20k on average 5.3 strains reduced in growth rate on each substrate, in fact 0.7 fewer strains than expected given the null distribution (p = 0.01). However, this average somewhat masks the bimodal pattern seen in the distribution, with some substrates showing nearly no strains reducing growth rate and others nearly all. At 50k, an average of 8.3 strains lost function on each substrate, 1.8 more than expected (p<0.0001). Clearly there is some parallelism in decreases in growth rate, but it is worth emphasizing that the substrates used in this study were those for which widespread, parallel losses in cellular respiration were previously observed.
We asked whether the correlated changes in performance on alternative substrates could be predicted based on the similarity of the catabolic network for growth on that compound compared to that for glucose. There are two rationales that would support this hypothesis. First, there are more loss-of-function mutations available for a nonglucose substrate if it uses many unique enzymes, and we might expect to see metabolic specialization scale with mutational target size under mutation accumulation. These mutations may either simply be unguarded by purifying selection, or perhaps even selectively advantageous to lose. Second, the balance and direction of flux through various pathways will lead to a different optimal allocation of enzymes to balance the needs of catalysis versus expression costs. As such, antagonistic and synergistic pleiotropy suggest that highly overlapping metabolic flux patterns might be expected to suffer fewer declines, or possibly even synergistic gains, relative to a very differently used substrate. A rough approximation of the similarity between different substrates is to simply group them as sugars or “nonsugars” that require gluconeogenesis for anabolism. To frame this more quantitatively, however, we also used genome-scale metabolic models to make predictions about specialization.
In order to approximate internal metabolic states, we used flux balance analysis (FBA) to generate predicted flux patterns for each compound [14]. This approach generates a vector of internal flux values that describes the relative flow through every reaction in a cell were it to optimize biomass production per substrate molecule. Although selection in batch culture largely acts upon rate, biomass production per unit substrate has been shown to effectively capture the growth of the LTEE ancestor on glucose, and 50k evolved strains deviated only slightly from this pattern [15]. We therefore compared FBA-derived flux vectors using a number of metrics to determine their degree of dissimilarity to the flux vector for glucose (see Materials and Methods).
We first tested whether evolved decreases in growth rate scaled with mutational target size. There is no expected behavior under this hypothesis for increases in growth rate, so we limited our analysis to combinations of strains and substrates for which growth rate had decreased. By identifying the reactions necessary for optimal growth on alternative substrates that are not necessary for growth on glucose, and determining the number of coding nucleotides necessary for those reactions, we were able to approximate the number of available mutations that would decrease growth rate on a substrate. Contrary to our hypothesis, we found no significant relationship between mutational target size and reduction in growth rate [p = 0.15, linear regression F test(1, 146) = 2.1] (Figure 3B).
Our hypothesis also suggests that substrates used more similarly to glucose would permit more rapid growth. Starting with the simple categorization of substrates as sugars and nonsugars, we found no correlation between these groupings and changes in growth rate (Figure 3A). Indeed, there were many reductions in growth rate for sugars other than glucose. To frame this hypothesis in a more quantitative way, we compared the Hamming distance between the vector of predicted fluxes for an alternative compound and that of glucose. Contrary to our hypothesis, we found no significant relationship between metabolic similarity to glucose and correlated responses [p = 0.26, linear regression F test(1, 323) = 1.27], and any relationship measured was in fact in the opposite direction as predicted (Figure 3D). As a confirmation that Hamming distance between flux vectors for alternative substrates is biologically relevant, we found that it was a significant predictor of some of the variance in the relative growth rate of the ancestor [p = 0.0001, R2 = 0.26, linear regression F test(1, 48) = 17.2] (Figure 3C). Alternative metrics to Hamming distance performed similarly poorly in predicting patterns of tradeoffs (Table S4). These data suggest that the similarity of overall flux patterns is a surprisingly poor predictor about which substrates would experience correlated increases or decreases in performance.
Linking the mutator-driven metabolic specialization to their vastly elevated mutation rate itself, we hypothesized that their abundance of amino acid substitutions may generate trends indicative of the types of effects they had upon their gene products. The mutator lineages acquired mutations with a rate up to 0.06 per generation [16]. For the A-1 lineage, by 40,000 generations there were 627 SNPs, 599 of which were in coding regions, and 513 of those were nonsynonymous [17]. This is a tremendous load of amino acid substitutions, which are viewed as likely to be deleterious due to destabilizing proteins [18]–[20]. We therefore hypothesized that, if protein destabilization was a dominant factor affecting growth at 37°C, we could predictably ameliorate these defects by lowering the growth temperature. Nonmutators will have a small number of such mutations, but the ∼100-fold greater rate of such mutations in the mutator genomes should make growth more temperature sensitive than for nonmutators.
Growth rate data support the hypothesis that mutators have general temperature-sensitivity. For 50k isolates, growth rate relative to the ancestor at 30°C was higher than at 37°C in 98 cases, compared to only 37 cases where it was reduced relative to the ancestor (p<0.0001, binomial two-tailed exact test) (Figure 4B). That is, despite the fact that these strains have adapted for 50,000 generations at 37°C, the ratio of their growth rate to that of the ancestor is higher at the foreign 30°C than their native temperature. This general improvement at the lower temperature was not present for nonmutators (56 improved relative to the ancestor by moving to 30°C, 57 worse—p = 0.99, binomial two-tailed exact test), and the difference between mutators and nonmutators was significant (p = 0.0002, Pearson's chi-squared test). Furthermore, in the cases where evolved 50k isolates completely lost the ability to grow on substrates, when grown at 30°C these losses were ameliorated more than half the time for mutators (35 of 61), significantly more than for nonmutators (6 out of 23, p = 0.01, Pearson's chi-squared test) (Figure 4C).
An alternative hypothesis for the elevated temperature sensitivity of mutators is that the phenotype is directly caused by the mutation in mismatch repair rather than the accumulation of destabilizing mutations that it caused. For the 20k isolates, in most cases growth was better relative to the ancestor at the native 37°C than at 30°C (119 versus 75, p = 0.002, binomial two-tailed exact test) (Figure 4A). There was no significant difference at 20k between this pattern for mutators and nonmutators (p = 0.20, Pearson's chi-squared test), and no significant difference in the number of rescues from complete loss of growth (p = 0.31, Pearson's chi-squared test), ruling out that mutator status itself generates temperature sensitivity.
Replicate experimentally evolved populations such as the LTEE are ideal for studying the processes leading to specialization. Here we directly measured growth rates in well-mixed conditions rather than cellular respiration in static media. This seemingly minor choice generated qualitatively different results, leading to the opposite conclusions from those previously made about the mechanisms and patterns of specialization during tens of thousands of generations of growth on a single substrate.
The earlier report [4] of widespread, parallel tradeoffs after extended growth on a single compound fit comfortably with the general notion that unused capacities will tend to degrade after an extended period of disuse. However, the suitability of cellular respiration assays for growth performance was seriously challenged by our data. There were many cases of “false positives,” with respiration on substrates that do not support growth, and a weak predictive ability for growth even after removing these. The parallel decreases in function observed across compounds in the respiration assay seem to be due to cultures becoming worse at reducing the dye in the Biolog assay environment. This generic effect was strong enough to mask both adaptation on glucose and the novel gain of citrate use in the A-3 50k isolate.
Growth rates indicated little evidence in support of widespread antagonistic pleiotropy, with more increases than decreases in growth on alternative compounds through 20k, and very few parallel declines. There were individual counterexamples observed, such as the previously characterized universal loss of ribose utilization early in adaptation [21], and the tendency for reduced or loss of growth on maltose [22]–[24]. As such, it is clear that examples of antagonistic pleiotropy do exist in the data. However, relatively few other substrates showed this pattern at 20k or 50k, despite the fact that these substrates were those where parallel reductions in respiration were observed. Because selection drives antagonistic pleiotropy, it is commonly expected that the early period of rapid adaptation would coincide with the most tradeoffs in alternative environments, and that the frequent parallelism in the targets among early beneficial mutations would drive parallel losses [4]. Given these criteria, the growth data do not support antagonistic pleiotropy as the primary driver of specialization.
There are three implicit assumptions about antagonistic pleiotropy, however, that if not met alter the expectations for specialization driven by selection. First, if different beneficial mutations occur across lineages, they will not necessarily lead to the same pleiotropic tradeoffs. As of 20k, out of the 14 genes screened in all of the populations, there were three genes with mutations in all populations and two more in a majority. The other screened genes had mutations in a minority or none of the other populations, suggesting that a variety of different beneficial mutations occurred across lineages [17]. Second, beneficial mutations in the same target may have differing pleiotropic effects in different lineages due to other mutations present. This “epistatic pleiotropy” [25] has been found to be common in multiple model systems [26]–[29]. Third, the early large-effect beneficial mutations may or may not be responsible for greater pleiotropic effects than later, smaller effect mutations. Yeast morphological pleiotropy scaled with fitness, for example, but the correlation explained only 17% of the variation [30]. The first and second scenarios above—distinct mutations or epistatic pleiotropy—would lead to a scenario whereby parallel metabolic declines are no longer necessarily expected from antagonistic pleiotropy. The third scenario—pleiotropy not scaling with selective effect—would mean the temporal dynamics of fitness gain in the selective conditions and the rate of performance losses in alternative environments need not be tied.
These caveats underscore our limited ability to make conclusions about the role of antagonistic pleiotropy in the observed metabolic declines. The only sure determinant of whether a correlated change is the result of pleiotropy or neutral mutation is to genetically manipulate the strains to isolate the effect of individual mutations. This suggests that future experiments, for example, test early and parallel mutations previously screened for epistatic effects [31] for pleiotropy. Ultimately, as we discuss below, the key determinant of the role of mutation accumulation is whether metabolic specialization was substantially affected by mutation rate.
Rather than a general pattern of metabolic specialization, these data revealed an unexpected extent of correlated improvements in growth on alternative compounds. Why would E. coli maintain or improve performance on substrates that had not been supplied for decades? There are three general classes of explanations, two of which mirror the processes considered for pleiotropic tradeoffs.
The first explanation for the correlated improvements, and undoubtedly the least likely, would be neutral performance gains through mutations that had no selective consequence in glucose: the beneficial analog of mutation accumulation. If this were the case, mutators might have more increases in rate than nonmutators, and more improvements would have occurred by 50k than 20k, which is the opposite of what was observed.
The second explanation of the correlated improvements is that the same mutations that were beneficial during growth on glucose may have led to gains in alternative environments—that is, “synergistic pleiotropy.” There are known examples of this occurring for early mutations in E. coli evolving in these conditions [12],[32], and it is a common pattern seen across organisms (for example [33]). These synergistic mutations may be generally beneficial in the laboratory environment of the LTEE and thus unrelated to carbon source metabolism. Indeed, there are examples of both adaptation to generic aspects of a selective environment, such as the trace metal formulation [34], and removal or down-regulation of costly genes or genome regions [21],[35],[36]. Synergistic pleiotropy could also result from mutations that directly improved glucose metabolism, such as mutations in the phosphotransferase-mediated uptake system that also increased growth on the other sugars imported by this system [12].
A third hypothesis for correlated gains of function on alternative compounds is that there were additional compounds besides glucose (and citrate) available from cell excretions or lysis. The serial transfer regime of the LTEE creates a scenario whereby populations use all of their glucose resources within the first few hours, and remain in stationary phase the remainder of the day. The ancestral E. coli excrete a small amount of acetate in this environment, and this increased on average 2-fold by 50k generations [15]. It is thus unsurprising that the strongest, most universal gain in alternative compounds by 20k was on acetate (Figure 2A). In terms of cell lysis, this has allowed one population (A-2) to maintain a long-term polymorphism for over 40,000 generations. A “large” colony lineage that grows fast on glucose but lyses substantially in stationary phase cross-feeds a “small” colony lineage that is not as fast on glucose as the larges but has specialized as a “cannibal” [37]–[39]. This results in a stable, negative frequency-dependent fitness effect between these strategies. Although an earlier study of the other populations at 20k failed to reject that fitness interactions were transitive through time [40], these competitions were performed at a 50∶50 ratio and thus may have missed interactions that occur when one partner is rare.
The most remarkable correlated increases were the several examples of “novel” gains of function by evolved isolates on substrates where the ancestor failed to grow. The citrate example has been reported previously [7], but we did not expect to find other such substrates. These novel gains of function are distinct from what was seen for citrate, as over a longer duration (>100 h) the ancestral E. coli seem to grow to measurable density on these substrates. We are currently exploring whether these long lags represent slow physiological acclimation or the emergence of evolved genomic changes. In the case of the Cit+ A-3 lineage, succinate is likely excreted during citrate import [41]; thus, selection for its use is perhaps unsurprising. The fact that several other strains experienced similar gains across the same range of C4 dicarboxylic acids, and that this included the cross-feeding “small” phenotype clone from the A-2 population, appears to suggest that these compounds may be excreted, or present during stationary phase from lysed cells.
With the availability of whole-genome scale metabolic models for E. coli, we asked whether we could predict the trend of correlated responses by comparing their pattern of use to that of glucose. We proposed a generic, seemingly obvious hypothesis that the more different the metabolism of an alternative substrate was from the metabolism of glucose, the more likely it would be that populations would have decreased (or lost) their ability to use it. As described, this logic holds regardless of whether or not selection drove metabolic specialization. In order to quantify the similarity of substrates, we applied FBA to compare the predicted optimal metabolic flux states for each compound. The data, however, did not support our hypothesis: there was almost no relationship between similarity to glucose and correlated response. Recent in silico attempts to predict growth capability on substrates based on metabolic similarities have had some success [42], suggesting that evolution may be acting here on functions not included in the model. For example, mutations may have occurred in functions related to differential regulation that distinguish these sugars, rather than central metabolic enzymes for which their use is nearly identical. These mutations are known to have occurred in the LTEE, for example in spoT and nadR [35],[43]. These predictions are also based on the assumption that glucose is the only available carbon source. If growth on other carbon sources is under selection due to their excretion or presence after cell lysis, it may explain some of the lack of predictive power here.
Although the identity of the substrates that experienced tradeoffs (or improvements) were not those we expected, we reasoned that the biophysical effects of the deluge of mutations in mutators might lead to a predictable pattern of temperature sensitivity in these strains. The genomic sequences and data available to date [16],[17] suggest mutators will have on the order of 500–2,000 nonsynonymous mutations, perhaps more. Random amino acid substitutions have been shown to be mildly deleterious in general due to destabilizing proteins [20]. To ask whether tradeoffs observed in the mutators were at least partly due to destabilizing mutations in proteins needed for alternative substrates, we tested whether mutators would be more sensitive to changes in incubation temperature than nonmutators. Consistent with this hypothesis, we found that the 50k mutators performed better relative to the ancestor at 30°C than at the 37°C temperature where they have evolved. One alternative hypothesis that was ruled out is that the mutator allele itself leads to temperature sensitivity, as the 20k mutators performed better at 37°C than at 30°C, and the changes with temperature were not distinguishable from nonmutators. Although other alternative hypotheses may explain some of the temperature sensitivity, these data are consistent with the hypothesis that neutral degradation of protein-coding sequences in these strains proceeded via partial destabilization on the way to eventual loss of function.
The comparison of mutators and nonmutators at 50k strongly suggests that neutral mutation accumulation was the primary driver of metabolic specialization. The difference in metabolic erosion allows us to distinguish the overall trend from forms of antagonistic pleiotropy that could have lacked parallelism (different mutations or epistatic pleiotropy) or that may have arisen late relative to fitness gains (if pleiotropy did not scale with selective effects). Despite a significant difference from nonmutators in the proportion of growth rate reductions, after 20,000 generations and a decade of adaptation the mutators still increased growth rate in more cases than they decreased, and only by 50k did mutators as a group have more decreases than increases. By the later time point, five of the seven mutators had decreased growth (or complete loss) for essentially every single alternative compound (except citrate and C4 dicarboxylic acids for A-3, which were under selection for this strain). Interestingly, the other two (A-1, A-2S) do not show this pattern. These counterexamples may be due to the fact that A-1 acquired its mutator status late [17], and A-2S is the cross-feeding generalist described above that adapted to grow upon lysed cell material [38].
The late appearance of metabolic erosion argues for the unparalleled utility of truly long-term experiments. A neutral process such as mutation accumulation needs time to become apparent, although hitchhiking with beneficial mutations can speed their fixation (i.e., “draft” [44],[45]). With a reduced effective population size, the window of selective effects that behave neutrally grows. As such, the effects of elevated mutation rates and mutation accumulation become apparent much more quickly with evolution regimes with small bottlenecks, such as single colonies [46],[47]. The late appearance of specialization also contrasts sharply with abundant evidence that lineages can diversify and specialize in mere tens or hundreds of generations. In addition to population size, this difference in timescale appears to correlate with the type of selective environment, and thus the evolutionary process that was responsible. Whereas the Lenski LTEE is notable as an environment with a single nutrient resource at high concentration, the cases of rapid diversification have involved spatial heterogeneity [48], rate-limiting resources in a chemostat [49], or the presence of multiple substrates simultaneously [50]. In those scenarios, selection is actively pulling on different performance features of an organism and antagonistic pleiotropy appears to dominate. The relatively slow degradation of catabolic capacity in the LTEE suggests that E. coli faces comparatively little tension between improving upon glucose and maintaining performance on other substrates, even those which are predicted to be utilized in a very distinct manner. Specialization in this case appears not to have been a requisite tradeoff of adaptation, but rather a result of the degradation of unneeded proteins.
Given the severity of metabolic erosion for mutators even in large laboratory populations, it is remarkable just how common mutator lineages are in nature. Mutators have been isolated at frequencies over 1% and seem to be particularly common in organisms such as pathogens [51]. The frequency of mutators in nature, despite the associated costs, may be partially explained by increased evolvability, shown in laboratory medium [52] and in mice [53].
Our results add substantially to the idea that an elevated mutation rate is an ill-fated long-term strategy even for large populations, as declines in performance in alternative environments will eliminate previously occupied parts of the niche space. Recent findings have suggested that mutators may tend to attenuate their increased mutation rate over time [16],[54],[55], perhaps to avoid the harmful effects of Muller's Ratchet. Thus both tradeoffs in alternative environments and mutation load in selective environments may contribute to the paradox that over the short term lineages often benefit from elevated mutation rates, but the long-term trend across phylogenies has been for stability in mutation rates of free-living microbes [56].
E. coli B isolates were obtained from the LTEE [5] after 20,000 and 50,000 generations. Briefly, in the evolution experiment 12 populations of E. coli B were founded with either the arabinose-negative strain REL606 (populations A-1 to A-6) or the otherwise isogenic arabinose-positive derivative, REL607 (A+1 to A+6). These have been evolved since 1988 in 50 mL flasks containing 10 mL of DM media with 139 µM glucose (25 mg/L) as a carbon source. The cultures were grown at 37°C while shaking at 120 rpm, and were transferred daily via 1∶100 dilutions (∼6.64 net doublings per day).
The isolates analyzed in this experiment consisted of the ancestral lines REL606 and REL607, as well as the “A” clone frozen at 50k and 20k generations for the 12 populations. The A-2A clones at 20k and 50k were from the “large” lineage that has coexisted with a cross-feeding “small” lineage for tens of thousands of generations [57], and thus here we refer to them as A-2L. At 50k we also examined A-2C (REL11335), a “small” clone that we refer to here as A-2S. All evolved strains are listed in Table S1.
For growth rate measurements, we acclimated out of the freezer by inoculating 10 µL frozen cultures into 630 µL modified DM250 media in 48-well micotiter plates (Costar) and growing overnight on a plate shaking tower (Caliper). All growth rates were measured at the LTEE selective temperature (37°C) unless otherwise described. The modified DM media is the same as previously used throughout the evolution of these strains [5], except it contained 250 mg/L glucose and no sodium citrate. Following acclimation, saturated cultures were transferred into new plates with a 1∶64 dilution in DM media supplemented with 5 mM of a single carbon source. Under these conditions, growth in plates correlates well with growth in flasks, both with and without citrate [p<0.0001, R2 = 0.74, linear regression F test(1, 28) = 79.3]. The substrates analyzed were those where consistent reduction in cellular respiration were previously observed, as well as several sugars for which fitness changes had been previously measured after 2,000 generations [12] and citrate. Between 3 and 11 biological replicates were run for each strain/carbon source combination.
Optical densities were obtained every 30 min to 1 h on a Wallac Victor 2 plate reader (Perkin-Elmer), until 48 h had passed or cultures reached saturation, using a previously described automated measurement system [9],[10]. Growth rates were determined by fitting an exponential growth model using custom analysis software, Curve Fitter (N. F. Delaney, CJM, unpublished; http://www.evolvedmicrobe.com/Software.html). Representative growth curves and fitted growth rates are shown in Figure S4. The growth rate for all strains relative to the ancestor was calculated for each plate (averaging over the two ancestors, REL606 and REL607), and averaged across plates. Mean growth rates relative to the average of the ancestors were used throughout. When quantifying the number of increases and decreases in rate for evolved strains, we used all of the data for the substrates for which the ancestor exhibited growth—a necessary criterion for the evolved strains to demonstrate reductions.
Biolog assays for respiration capacity were run as in Cooper and Lenski [4]. Briefly, cultures were grown from freezer stocks for two cycles of 1∶100 dilution and 24 h of growth in 10 mL LB in flasks shaking at 37°C. LB was used to avoid catabolite repression due to growth in minimal media, which could result in fewer positive results on nonglucose carbon sources. These cultures were inoculated 1∶100 into fresh LB and grown for 6 h before being spun down at 12,000 g for 10 min and rinsed in saline to remove residual medium. Rinsed cells were resuspended in IF-0 buffer with dye added (Biolog) to a constant density of 85% transmittance, and all wells of Biolog PM1 plates were inoculated with 100 µL of this suspension. The plates were incubated, unshaken, at 37°C. OD600 was measured at 0, 4, 12, 24, and 48 h, and all well readings were adjusted by subtracting the reading of the well at 0 h. A trapezoidal area approximation [4] combined the five measurements for each well into one value, which reflects the area under the curve (AUC) of optical density versus time. One replicate plate experiment was performed for each evolved strain at 20k and 50k generations, and four replicates were run for each ancestor (REL606 and REL607). Tests for tradeoffs for the evolved strains as a group on a substrate had were one-sample t tests against the ancestral distribution with significance cutoff of p = 0.002 to adjust for multiple comparisons. Tests for individual strains were the same but with a cutoff of p = 0.0005.
Flux analysis was carried out with a genome-scale model of E. coli metabolism (iAF_1260 [58]). The model incorporates 2,382 reactions and 1,668 metabolites. The default minimal media environment and reaction bounds were used. Fluxes were predicted for each individual carbon source provided, normalized by number of carbon atoms to 10 units of glucose. Maximal biomass per substrate was used as the objective criterion as previously described [15]. To determine whether a reaction was necessary for optimal growth on a substrate, each reaction flux predicted was individually constrained to zero. Only the necessary reactions, those for which constraining the flux resulted in a reduction in final biomass, were considered in the analysis of differences between flux vectors (Table S2). Reaction differences between substrates, considered for the Hamming distance, are listed in Table S3. Table S4 summarizes alternative distance metrics that were used to assess the difference between flux vectors.
|
10.1371/journal.pgen.1003633 | Sociogenomics of Cooperation and Conflict during Colony Founding in the Fire Ant Solenopsis invicta | One of the fundamental questions in biology is how cooperative and altruistic behaviors evolved. The majority of studies seeking to identify the genes regulating these behaviors have been performed in systems where behavioral and physiological differences are relatively fixed, such as in the honey bee. During colony founding in the monogyne (one queen per colony) social form of the fire ant Solenopsis invicta, newly-mated queens may start new colonies either individually (haplometrosis) or in groups (pleometrosis). However, only one queen (the “winner”) in pleometrotic associations survives and takes the lead of the young colony while the others (the “losers”) are executed. Thus, colony founding in fire ants provides an excellent system in which to examine the genes underpinning cooperative behavior and how the social environment shapes the expression of these genes. We developed a new whole genome microarray platform for S. invicta to characterize the gene expression patterns associated with colony founding behavior. First, we compared haplometrotic queens, pleometrotic winners and pleometrotic losers. Second, we manipulated pleometrotic couples in order to switch or maintain the social ranks of the two cofoundresses. Haplometrotic and pleometrotic queens differed in the expression of genes involved in stress response, aging, immunity, reproduction and lipid biosynthesis. Smaller sets of genes were differentially expressed between winners and losers. In the second experiment, switching social rank had a much greater impact on gene expression patterns than the initial/final rank. Expression differences for several candidate genes involved in key biological processes were confirmed using qRT-PCR. Our findings indicate that, in S. invicta, social environment plays a major role in the determination of the patterns of gene expression, while the queen's physiological state is secondary. These results highlight the powerful influence of social environment on regulation of the genomic state, physiology and ultimately, social behavior of animals.
| The characterization of the genomic basis for complex behaviors is one of the major goals of biological research. The genomic state of an individual results from the interplay between its internal condition (the “nature”) and the external environment (the “nurture”), which may include the social environment. Colony founding in the fire ant Solenopsis invicta is a complex process that serves as a useful model for investigating how the interplay between genes and social environment shapes social behavior. Unrelated, newly mated S. invicta queens may start a new colony as a group, but ultimately only one queen will survive and gain full reproductive dominance. By uncovering the genetic basis for founding behavior in fire ants we therefore provide useful insights into how cooperative behavior evolved in a context that might be considered primitively eusocial, because newly mated queens in a founding association are morphologically, physiologically and genetically very similar and display no evident division of labor. Our results suggest that social environment (founding singly or in pairs, switching dominance rank vs. maintaining rank) is a much greater driver of gene expression changes than social rank itself, suggesting that social environment, and not reproductive state, is a key regulator of gene expression, physiology and ultimately, behavior.
| Behavior is a complex phenotypic trait, which results from the interactions of multiple intrinsic and extrinsic factors that associate in a nonlinear, often unpredictable fashion [1]. Intrinsic factors include the genetics, the physiology or the phenotype of an organism, while the most typical extrinsic factor is the external environment. In social systems like insect societies, environmental cues primarily are the result of the social environment, i.e. the nature and patterns of interactions among individuals within the colony [2]. The “nature-versus-nurture” debate has long been the major driver of the discussion as to whether internal state of an animal or the external environment (e.g., the social environment) regulates gene expression more [3]. Regardless, extrinsic and intrinsic factors clearly are reciprocally interconnected: the social environment influences the neurogenomic state of the animal, which is responsible for the social behavior performed [4], [5]. A hallmark of advanced social behavior is altruistic behavior, which is achieved through a reproductive division of labor in which few individuals develop into the reproductive caste while most of the colony members become non-reproductive workers and perform all tasks related to colony maintenance and growth. Both fixed (developmental pathways) and plastic (behavioral strategies) factors contribute to this division of labor (reviewed in [6]). Consequently, there has been great interest in studying genes and biological processes that regulate the reproductive and worker divisions of labor [7]. In the advanced eusocial systems examined thus far, differences between queens and workers are largely the result of developmental factors, while differences among workers are often triggered by social signals [8]. However, primitively social systems display reproductive division of labor between females that are anatomically, physiologically and genetically very similar and this reproductive division of labor seems to be primarily established and maintained by social environment. The genes underlying this process have not yet been examined, and potentially may function as core genes associated with sociality.
Variation in colony founding among ant queens is an ideal model to examine the interplay between genes and social environment that has shaped the evolution of cooperative behavior in primitively social systems. Colony founding can occur in two modalities: haplometrosis, where a single queen independently starts a new colony, and pleometrosis, where multiple queens associate and cooperate to start a new colony [9]. Pleometrosis is a fascinating example of cooperative behavior that is not fostered by kin selection, because these groups often comprise unrelated individuals (reviewed [10]). Among social insects, pleometrosis exists in halictine bees [11], termites [12], paper wasps [13], [14], [15] and in several species of ants [8]. In ants, pleometrosis is known to be associated with division of labor in the leaf-cutter ant Acromyrmex versicolor [16] and in the harvester ant Pogonomyrmex californicus [17]. Pleometrotic associations produce a complex social environment, where individuals simultaneously are in cooperation and conflict, and social and reproductive dominance hierarchies are established. These associations represent relatively primitive social systems in which individuals with equivalent anatomical and physiological features develop a division of labor through their behavioral interactions. Thus, identification of the genes underlying establishment of these hierarchies will not only provide insight into the effects of social environment on an individual, but also into the evolution of social behavior.
The red imported fire ant Solenopsis invicta is an excellent system for studying the genes associated with haplometrotic and pleometrotic behaviors, because queens from the monogyne social form (characterized by a single egg-laying queen per nest once established) can adopt either approach for colony founding depending on multiple factors, e.g. the density of newly mated queens in nesting sites [18]. However, monogyne fire ants ultimately only tolerate a single reproductive queen such that the initial cooperation among unrelated pleometrotic cofoundresses slowly transitions to competition and rivalry, which will inevitably produce only one winner and one or multiple losers [19], [20], [21]. Once the first workers emerge, pleometrotic queens engage in open fights where they injure or kill rival cofoundresses and workers actively participate in this process until all the queens are executed but one (see Movie S1). In both haplometrosis and pleometrosis, founding queens initially face a critical period (claustral period) where they are sealed in their nest and must defend it from enemies and competitors, e.g. other fire ant colonies that populate the same area [22], [23]. During the claustral period, fire ant queens rely exclusively on their body mass reserves to produce the first generation of workers. There are physiological and behavioral differences between haplometrotic queens, pleometrotic “winners” and pleometrotic “losers”. Haplometrotic queens lose more weight during the claustral period, and produce more brood per individual than queens in pleometrotic associations [24], [25]. In pleometrotic associations, winners tend to have larger head size, lose less weight [20], [21], [26] and occupy the top of the brood pile while losers are usually found outside the nest chamber [19], attempting to avoid any interaction with the winner or with workers. However, nothing is known about the genes and molecular pathways that underlie these processes.
We performed two separate experiments to characterize the genomic basis for haplometrotic and pleometrotic founding behavior in fire ants. We developed a microarray platform using the official gene set of the fire ant genome [27] plus a set of ESTs obtained from assemblies of the fire ant transcriptome to examine genome-wide expression patterns across founding queens. In the first experiment, we compared whole body gene expression patterns among haplometrotic queens and paired pleometrotic winners and losers that were collected shortly after emergence of the first workers (but prior to execution of the loser). We predicted that haplometrotic queens would be more similar to pleometrotic winners than to pleometrotic losers, because they both will serve as the single queen for the mature colony. For the comparison between winners and losers alone, our expectations were less well-defined: on one hand, we expected to find substantial differences given that their physiology, behavior and fate differ significantly, but on the other hand, winners and losers are not anatomically distinct and there is only weak correlation between morphological measures and outcome of the conflict [20]. In the second experiment, we manipulated queen rank in pleometrotic pairs to determine how changing social rank and social environment would affect an individual's gene expression patterns. This was accomplished by pairing pleometrotic queens with a new partner at the end of the claustral period in order to switch putative winners to losers and vice-versa. Controls in which partners were altered (and social environment was changed) but social rank remained the same were also included. We hypothesized that final social rank would be the primary regulator of gene expression patterns. However, for both experiments our results indicated that social environment (pleiometrosis vs. haplometrosis, switched rank vs. maintained rank) was a much greater driver of gene expression changes than social rank itself, suggesting that social environment, and not reproductive state, is a key regulator of gene expression, physiology and ultimately, behavior.
Haplometrotic queens (haplo) and paired pleometrotic winners (win) and losers (los) were collected shortly after emergence of the first workers (see methods, N = 8 haplo, 8 win and 8 los). Microarray analysis of gene expression patterns (see Methods for design and validation of microarrays) in whole bodies of these queens revealed that 4080 of the 9388 transcripts included in the analysis were differentially regulated at FDR<0.001 (Table S1). A principal components analysis (PCA) of the differentially regulated transcripts revealed that the social environment is more important than social rank in driving the patterns of gene expression in founding queens (Figure 1A). Differences between haplometrotic and pleometrotic queens accounted for 91.8% of the variation in gene expression while differences between win and los accounted for only 8.2%. Pairwise comparisons of transcripts differentially regulated (FDR<0.001) among the three groups of fire ant queens demonstrated that expression patterns in haplo are more similar to win than to los, since there are fewer genes differentially regulated uniquely between haplo and win (404) than haplo and los (477; Nominal Logistic Fit: df = 1, ChiSquare = 6.78, P = 0.0092; Figure 1B).
We performed Gene Ontology analysis on the 3003 differentially regulated transcripts (out of the initial pool of 4080) that have Drosophila orthologs with FlyBase annotations using DAVID [28], [29]. 517 GO terms were significantly enriched at a p-value<0.05 (Functional Annotation Chart, see Table S2 for the complete list of GO terms). Additionally, 6 KEGG molecular pathways (Kyoto Encyclopedia of Genes and Genomes, [30]) were significantly enriched (P<0.05): aminoacyl-tRNA biosynthesis, basal transcription factors, dorso-ventral axis formation, endocytosis, RNA degradation and ubiquitin mediated proteolysis (Table S2). To cluster the GO categories into larger functional groups, the 517 significantly enriched GO terms were mapped to the GO_slim2 file in CateGOrizer [31]: 440 GO terms were assigned to one of the ancestor terms by single count (Figure 2). The functional groups containing the greatest number of GO terms were metabolism (19% of the significantly enriched GO terms), cell organization and biogenesis (11%) and development (10%; for a complete list of all ancestor terms represented in this analysis see Table S3).
To further characterize the genes that were differentially regulated between haplo and pleometrotic queens (pleo), we examined the overlapping set of 3192 transcripts (of which 2541 had Drosophila orthologs with FlyBase annotations) that were differentially regulated between both haplo vs. win and haplo vs. los (Figure 1B). For clearer graphical presentation, we used k-means Clustering in GENESIS [32] to separate these transcripts into two large clusters according to expression patterns: 2280 transcripts that were upregulated in haplo and downregulated in pleo (cluster 1, Figure S1) and 912 transcripts downregulated in haplo and upregulated in pleo (cluster 2, Figure S2). We performed GO analysis on both groups, using Functional Annotation Clustering with medium stringency. For cluster 1 (1925 FlyBase matches), we obtained 88 significantly enriched GO terms (see Table S4; P<0.05) and 1 KEGG pathway (basal transcription factors, P = 0.01). Several of the GO terms were related to aging (determination of adult life span, death), immunity (immune system development, JNK cascade, hemopoiesis), reproduction (reproductive developmental process, oocyte development, eggshell formation, morphogenesis of follicular epithelium, regulation of oocyte development), response to stimuli (response to stress, regulation of response to stimulus, negative regulation of response to stimulus, response to ecdysone), lipid biosynthetic process, locomotion and neurological system processes (neurotransmitter secretion, neurogenesis, central nervous system development, regulation of nervous system development). In cluster 2 (616 FlyBase matches), 34 GO terms (Table S5, P<0.05) and 1 KEGG pathway (glycerophospholipid metabolism, P = 0.01) were significantly enriched. Many GO terms were similar to those in cluster 1, and included determination of adult life span, olfactory behavior, lipid metabolic process, detection of light stimulus, as well as several related to morphogenesis or development of organs and apparatuses like sensory organ, muscle, limb, wing disc, gut and respiratory system.
Interestingly, only 43 transcripts were differentially regulated between win and los queens. A GO analysis performed on this small set of transcripts revealed that fatty acid and hormone metabolic processes were significantly enriched GO terms (Functional Annotation Clustering, P<0.001 and P<0.01, respectively, Table S6). Several transcripts in this group have interesting functions (Table 1). Transcripts upregulated in win included: G protein-coupled receptor kinase 2 (Gprk2), which is involved in the Toll signaling pathway during the response against Gram positive bacteria [33]; endosulfine (endos), which functions in the insulin-signaling pathway during oogenesis [34]; Pheromone-binding protein-related protein 3 (Pbprp3), a member of the odorant binding proteins responsible for chemoreception [35]; and bubblegum (bgm), which is involved in the metabolism of very long-chain fatty acids and prevents neurodegeneration [36]. Transcripts upregulated in los relative to win included: I'm not dead yet (Indy), associated with aging [37]; pale (ple), which plays a role in the response to wounding [38] and in the metabolism of dopamine [39]; desat1, a major regulator of cuticular hydrocarbon biosynthesis involved in pheromone emission and detection [40]; and juvenile hormone acid methyltransferase (jhamt), a key enzyme in the biosynthesis of JH, the major endocrine regulator in insects [41].
GO categories related to aging and longevity were significantly enriched in sets of transcripts that were differentially regulated in haplo vs. pleo (clusters 1 and 2; see Tables S4 and S5). Out of the 129 genes included in the Drosophila aging GO term, oligos representing 93 putative orthologs were present on the fire ant microarray. Of these, 90 were expressed at high enough levels to be included in the microarray analysis and 46 were significantly differentially regulated across the three groups of queens (Figure 3A). The majority of these genes (34) were upregulated in haplo: this number was significantly higher than expected by chance (Nominal Logistic Fit: df = 2, ChiSquare = 29.58, P<0.0001). In addition to their role in regulating aging pathways in Drosophila, several genes in this group have been linked to queen vs. worker caste differentiation and behavioral maturation in honey bees [42]. These include: forkhead box (foxo) and target of rapamycin (TOR), two major players in the insulin-signaling pathway which is associated to caste determination in honey bees [43] and to the workers' transition from nursing to foraging behavior [44], [45], and Peroxiredoxin 5037 (Prx3), which is associated with enhanced learning ability when expressed at higher levels in honey bee workers [46].
To further investigate the patterns of expression of aging genes in haplo and pleo queens, we compared our study to a study on aging in Drosophila [47]. In this study, which was aimed at investigating the temporal and spatial (tissue-specific) transcriptional profiles in Drosophila, the authors listed all the age-related GO terms that were significantly enriched and classified them based on the tissue where they were expressed and on their directional expression. We found that 106 GO terms were upregulated in fire ant haplo queens and old Drosophila, while only 36 were shared between old flies and pleo: however, the difference was not statistically significant (Fisher's Exact Test: P = 0.67). When we compared downregulated GO terms, we found that 11 were shared between haplo and old flies while 67 were shared between old flies and pleo: the difference was statistically significant (Fisher's Exact Test: P = 0.0029). Most of these 67 overlapping GO terms (see Table S7 for details) encompassed genes that were regulated in the gut (32) and fat bodies (23), followed by brain (13), muscles (11), malpighian tubules (7) and accessory glands (6).
Immune-related GO terms were significantly enriched in cluster 1 (genes upregulated in haplo vs. pleo, see Table S4). To better examine the overall expression profiles of genes involved in immune pathways, we obtained a list of significantly enriched GO terms for both cluster 1 and cluster 2 (Functional Annotation Chart in DAVID, P<0.05, see Tables S8 and S9) and we mapped these lists to the list of “Immune system gene classes” available on the CateGOrizer website (http://www.animalgenome.org/bioinfo/tools/catego/slims.html). Thereafter, we compared the relative proportions of the parent/ancestor immune categories between the two groups (Figure 4). This analysis confirmed a significant overrepresentation of immune-related classes in cluster 1 relative to cluster 2 (Nominal Logistic Fit: df = 1, ChiSquare = 61.16, P<0.0001), clearly visible in terms of total number of immune categories present and number of GO terms within common categories. These results suggest that haplometrotic queens overall have higher expression levels of immune-related genes and therefore may be better equipped in terms of immune response.
Next, we examined the expression of the fire ant orthologs of the 177 canonical immune-related genes annotated in honey bees [48]. Orthologs for 83 of these genes were included in our array; 82 were expressed at high enough levels to be included in the analysis, and 34 were within our list of 3003 significantly differentially regulated transcripts (Figure 3B). Expression levels of these genes are not strongly coordinated, with similar numbers of up- vs. downregulated genes in haplo vs. pleo queens. Several genes in the Immune-deficiency (IMD) pathway were differentially regulated, including Inhibitor of apoptosis 2 (Iap2), TGF-beta activated kinase 1 (Tak1), immune deficiency (imd), bendless (ben) and Death related ced-3/Nedd2-like protein (Dredd). Furthermore, several members of the Immunoglobulin (IG) Superfamily were differentially regulated, including bent (bt), turtle (tutl), sidekick (sdk) and Down syndrome cell adhesion molecule (Dscam).
We examined gene expression levels of candidate genes that were included in one or more GO terms that were significantly enriched in our GO analyses (Figure S3 and Table S10). Expression patterns of all 13 candidate genes were consistent with what observed for haplo and los in the arrays, and these expression differences were significant for 11 genes. We validated Indy and Sod2 for determination of adult life span. In the arrays, Indy was downregulated in haplo and Sod2 was upregulated in haplo: qRT-PCR analysis confirmed these trends and the difference between groups was statistically significant for both genes (P<0.001). For immune response, we validated Dredd and kay, which were both upregulated in haplo in the arrays: qRT-PCR analysis confirmed this trend and the difference between the two groups of queens was statistically significant for kay (P<0.05) but not for Dredd (P = 0.29). Desat1, ifc and putative fatty acyl-CoA reductase CG5065 were analyzed for their involvement in the synthesis and metabolism of fatty acids. In the arrays, desat1 and putative fatty acyl-CoA reductase CG5065 were downregulated in haplo while ifc was upregulated in haplo: these trends were confirmed by qRT-PCR analysis and the difference in the expression levels was statistically significant for putative fatty acyl-CoA reductase CG5065 (P<0.001) and ifc (P<0.05) but not for desat1 (P = 0.11). Reproductive genes included br and Btk29A, both upregulated in haplo in the arrays: this was confirmed by qRT-PCR (P<0.001). We validated Sema-5c and Mer because they play a role in olfactory behavior: these genes were downregulated in haplo in the arrays and in the qRT-PCR analysis (P<0.001 and P<0.05, respectively). Finally, we analyzed fru for aggressive behavior and woc for neurogenesis: these genes were upregulated in haplo in the arrays and in the qRT-PCR analysis (P<0.05 and P<0.001, respectively).
We further examined the role of social rank on gene expression patterns by manipulating social rank of individuals in pleometrotic pairs. We swapped winners and losers between groups to generate four groups of queens: winners switched to losers (win/los), losers switched to winners (los/win), continuing winners (win/win) and continuing losers (los/los). Very few transcripts were differentially regulated among these groups, with a total of 616 transcripts at a relatively low significant threshold (FDR<0.1, see Table S11 for a listing of these transcripts). Principal components analysis demonstrated that 48% of the variation in gene expression was associated with switching social rank (win/los and los/win were clustered relative to win/win and los/los), 37% of the variation was associated with the final rank (i.e., win/los and los/los were clustered), while 15% was associated with the initial rank (i.e. win/win and win/los were clustered; Figure 5).
We performed GO analysis with Functional Annotation Clustering on the 527 differentially regulated transcripts that have Drosophila orthologs with FlyBase annotations. 21 GO terms were significantly enriched at p-value<0.05 and three survived Benjamini correction: cellular metabolic process, cellular ketone metabolic process and maintenance of protein location (see Table S12). Among the other GO terms, of particular interest was lipid metabolic process, which includes several genes involved in the metabolism of fatty acids such as Helix loop helix protein 106 (HLH106) [49], scully (scu) [50] and two putative fatty acyl-CoA reductases. Additional genes with significant differences in expression included Coenzyme Q biosynthesis protein 2 (Coq2), which plays a role in the response to pathogens, aging and in the insulin-signaling pathway [51], [52], juvenile hormone acid methyltransferase (jhamt), which was also significantly differentially regulated between win and los in experiment 1 (see above), and radish (rad), which is involved in learning and memory [53]. Finally, the GO term “response to stress” was significantly overrepresented, which includes key immune response genes such as immune response deficient 5 (ird5) [54], Ras-related protein Rac1 (Rac1) [55], Hemolectin (Hml) [56], Argonaute 2 (AGO2) [57] and caspar (casp) [58].
Ninety-three transcripts were differentially regulated both between win and los in experiments 1 (548 transcripts, FDR<0.1) and in experiment 2 (616 transcripts, FDR<0.1): this is significantly more than expected by chance (Hypergeometric Test: Representation factor: 2.2, p<1.009−13). These transcripts corresponded to 80 Drosophila orthologs, which were used to perform a GO analysis using Functional Annotation Clustering: 6 GO terms appeared to be significantly enriched, including lipid metabolic process (P<0.001) and regulation of hormone levels (P<0.05; Table S13). The expression patterns of the 9 differentially regulated transcripts involved in lipid metabolic process across the two experiments are shown in Table 2.
This study demonstrates that social environment greatly influences gene expression in founding queens of the fire ant Solenopsis invicta, and that the effects of social rank are secondary. Social environment in the first experiment (haplometrosis vs. pleometrosis) strongly influenced expression of thousands of genes, and the difference between pleometrotic winners and losers was much smaller. However, in terms of gene expression differences, pleometrotic winners were more similar to haplometrotic queens, suggesting that reproductive and social rank still does impact, albeit more subtly, gene expression patterns. In the second experiment, we manipulated both the social environment and social rank of queens in pleometrotic pairs. Switching social rank significantly affected gene expression patterns more than the initial or final social rank of the individual. Several categories of genes were differentially regulated among these groups of queens, including genes involved in core processes such as metabolism, aging and immunity.
In a recent study Ferreira et al. [59] explored the genetic basis of the early phases of social evolution in a primitively eusocial Polistes wasp. These authors found that 75% of the 2,442 genes differentially expressed between queen and worker phenotypes were novel, having no significant similarity with described sequences. This result supports the hypothesis that novel genes are likely important for eusocial evolution, as previously suggested by other studies [60], [61]. Interestingly, within our pool of 9,388 genes initially analyzed for experiment 1 (haplometrotic vs. pleometrotic queens), 41% were novel but this percentage decreased to 26% (which is a significantly smaller percentage, Fisher's Exact Test, P<0.0001) when we consider only the genes that were differentially regulated between the two phenotypes of founding queens. Thus, while caste differences in Polistes may be associated with novel genes, plasticity in founding behavior in fire ants seems to rely predominantly on more conserved genes.
Differences in gene expression between haplometrotic and pleometrotic queens were likely due to differences in the physiological demands placed on singly- vs. multiply-founding queens and differences in the costs associated with social environment, where pleometrotic queens are more likely to incur in higher levels of stress due to the establishment of social ranks. We found that genes involved in core physiological processes, including metabolism, cellular processes, development, morphogenesis and biosynthesis were significantly differentially regulated between these groups of queens. Haplometrotic queens produce more eggs and lose more weight than pleometrotic queens during the claustral period of colony founding [20]: this seems to be due to queen-queen reciprocal reproductive inhibition and oophagy/cannibalism of larvae in pleometrotic associations [62]. Genes associated with reproductive functions (including development of reproductive tissues and production of oocytes and eggs) were upregulated in haplometrotic queens. Furthermore, in order to produce eggs, newly mated queens degrade wing muscle tissues and metabolize fat body storage proteins to produce free amino acids [63]. We found 58 protein-related GO terms and 10 amino acid-related that were upregulated in haplometrotic queens versus 5 and 0, respectively that were upregulated in pleometrotic queens (Functional Annotation Chart, see Tables S8 and S9).
Genes associated with stress response were differentially regulated between haplo and pleo queens. Stress tolerance may be achieved by reducing the production of reactive oxidant species (via improved regulation of mitochondrial processes) and/or by increasing the antioxidant activity [64], [65]. In our study, we found that two mitochondria-related GO terms, namely mitochondrial electron transport, NADH to ubiquinone (15 genes) and mitochondrion organization (18 genes) were upregulated in haplo and none in pleo. Moreover, 9 antioxidant genes were upregulated in haplo, including two superoxide dismutases (Sod2 and CCS), two Peroxiredoxins (6005 and 5037), Glutathione S transferase S1 and PTEN-induced putative kinase 1 (Pink1), which plays an essential role in maintaining neuronal survival by preventing neurons from undergoing oxidative stress [66]. These results suggest that haplo queens may experience lower levels of oxidative stress either by producing less ROS or by keeping the levels of antioxidants high. Higher stress levels in pleo queens could be correlated to their social environment, dominated by queen-queen aggressive interactions and competition. Stress tolerance is positively correlated with lifespan [67] and this trait has been used as a proxy for long-lived phenotypes in studies that examine the genetic basis of lifespan [68]. Only SOD was upregulated in pleo. Interestingly, overexpression of SOD has been correlated to increased organismal longevity in Drosophila [69], but this was not confirmed in Lasius niger, where long-lived queens expressed lower levels of this gene than short-lived males and workers [70]. It is evident that the effect of SOD on longevity is highly dependent upon the sex and genetic background [71] and also the social environment [72].
The overrepresentation of GO terms associated to biosynthesis and metabolism (in particular those related to lipids) prompted us to look closer at the nutritional state of founding queens. Nutrition is closely linked to fertility and longevity [73]. In insects, the insulin-signaling pathway regulates nutrient-sensing [74] while juvenile hormone and ecdysone mediate reproductive processes [75]. In honey bees, long-lived queens have low levels of insulin and juvenile hormone, while they have high levels of FOXO, vitellogenin and ecdysone; opposite patterns are found in sterile short-lived workers [76]. Our results show that haplo had higher levels of FOXO and of the ecdysone receptor. Haplo queens also presumably had lower levels of JH, since levels of juvenile hormone acid methyltransferase (jhamt), an enzyme that converts inactive precursors of JHs to active JHs [41], were downregulated, and levels of juvenile hormone epoxide hydrolase 2 (Jheh2), involved in juvenile hormone catabolic process, were upregulated.
There is no clear prediction about which group of queens should have a longer life-span. Our analyses show that a large set of aging-related GO terms was upregulated in haplometrotic queens, while a smaller set was upregulated in pleometrotic queens. This result is not sufficient to establish which group of queens is expected to have longer life-span, since ageing is a quantitative trait determined by both environmental and genetic components. Previous studies of the genetics of longevity in Drosophila melanogaster, identified sets of genes in which upregulated expression either accelerates or decelerates the aging process [65]. However, in our study, genes from both categories were equally up- and downregulated across haplo and pleo queens (see Figure 3A). Therefore, the knowledge of the genetics of longevity in the insect model D. melanogaster cannot be transferred directly to our study system.
Immune-related genes were overexpressed in haplometrotic vs. pleometrotic queens. Most of the overrepresented immune-related GO terms were associated to cellular immunity: endocytosis, phagocytosis, cell adhesion, apoptosis, cytokinesis, the cascade regulating mitogen-activated protein kinase (MAPKKK) and the c-Jun amino-terminal protein kinase (JNK) cascade. In particular, the JNK pathway controls the rapid up-regulation of cytoskeletal genes in response to infection and plays a major role in wound healing [77]. Key genes in the JNK pathways [78] were upregulated in haplo, namely kayak, hemipterous, misshapen, anterior open and Cdc42. Hemopoiesis is the process that is responsible for production and differentiation of immune cells [79]: two key genes involved in this process, serrate and serpent, were upregulated in haplo. Haplo queens may have better constitutive immune responses perhaps because they experience less social stress than pleo queens do: in fact, once initial cooperation transitions into open competition, pleo queens frequently engage in reciprocal aggressions which can lead to serious injuries or death. It is hypothesized that there is a trade-off between reproduction, nutrition and immunity [80], suggesting that highly reproductive haplo queens should have overall reduced immune responses during colony foundation period when food sources are limited. However, previous studies in honey bees demonstrated that reproductive queens have higher expression of immune genes than non-reproductive workers [81], [82], and thus this trade-off may not exist in social insect queens, perhaps because queens have more energy resources than workers.
Only 43 transcripts were significantly differentially regulated between winners and losers in couples of pleometrotic queens from experiment 1. Although surprising, this result might be explained by the small phenotypic differences between the two types of queens. Previous studies showed that some phenotypic traits such as head width are weakly correlated with the reproductive investment and survival (hence the rank) of pleometrotic cofoundresses [20]. It has been suggested that the relatively weak association between these parameters stems from selection to maintain cooperation [20]. If phenotypic differences strongly correlate with the chances of surviving, smaller queens with lower fighting abilities would be selected not to cooperate and feed the brood in the colony. Thus, the small differences at the genomic level between winners and losers may reflect selection for a system where differences between cofoundresses is sufficiently small so that all of them have a chance of surviving, and thus an interest to cooperate with unrelated individuals.
The two GO terms that were differentially regulated between winners and losers were related to metabolism of lipids and metabolism of hormones. Four transcripts, bubblegum, desat1, Dmel_CG17374 and Dmel_CG31522, which function in fatty acid metabolism, were differentially regulated. Long-chain fatty acids are the precursors of cuticular hydrocarbons in insects, which can function as nestmate recognition cues and social pheromones in many insect species (reviewed [83]). Interestingly, bgm, which encodes a very long-chain fatty acid CoA ligase [36], was downregulated in losers relative to winners and haplo: thus, this gene may be involved in regulating chemical cues related to dominance. Similarly, desat1, which is expressed at higher levels in losers than in winners or haplo functions in pheromonal communication [40]. Altered bgm expression has also been associated with infection (and correlated with changes in cuticular hydrocarbon profiles) in honey bees [84], while desat1 appears to play a role in autophagic responses [85]. Thus, these genes may also be involved in signaling infection, nutrient deprivation or other stress responses.
Behavioral manipulation of the social rank in pairs of pleometrotic queens demonstrated that manipulation of social environment (i.e., conditions in which the social rank of the individual changes) had a much larger effect on gene expression than the initial or final social rank of the individual. Note, however, that all individuals in the study switched social partners, which may have elicited additional (undetected) changes in gene expression. Studies in vertebrates have demonstrated that social interactions and changes in the social environment can be one of the most potent stressors [86]. Indeed, genes associate with ‘response to stress’ were significantly enriched, with a set of 30 transcripts differentially regulated among the four groups of manipulated queens (see Results and Table S12). The effects of restructuring social ranks have not been considered broadly in other species [87], but decreased social rank in dark-eyed junco birds is associated with increased metabolic rates, while increased social rank results in a much lower physiological change [88]. Similarly, for dominant, but not for subordinate, birds there is a measurable metabolic cost to joining a new social group [88].
In both experiments, genes involved in lipid biosynthesis and metabolism were differentially regulated, suggesting that these processes play a key role in mediating fire ant founding behavior and foundress associations. Lipids such as cuticular hydrocarbons play a role in advertising the fertility state in many ant species: these compounds are usually more abundant in reproductive queens and egg-laying workers (reviewed in [89]). Indeed, ‘lipid biosynthetic/metabolic process’ was differentially regulated in haplo vs. pleo and in win vs. los in experiment 1 (Tables S4, S5 and S6) and in experiment 2 (Table S12). These results support the hypothesis that lipids (and in particular fatty acids) are of great importance in regulating social interactions between queens and among nestmates in general. In fire ant pleometrotic associations, the pheromones and nestmate recognition chemicals derived from these fatty acids are most likely an important component of the individual's chemical profile, which is used by nestmate queens to decipher the physiological condition and thus the social rank of the partner.
We used newly developed genomic tools to examine the gene expression patterns associated with complex social behaviors involved in colony founding by fire ant queens. Our results suggest that social environment (haplometrotic vs. pleometrotic, switched vs. maintained social rank) is more important than the social rank or internal condition of the individual in regulating gene expression patterns, and presumably downstream physiological and behavioral traits. Furthermore, because the process of pleometrotic colony founding in fire ants has all the features of a primitively social system in which morphologically, physiologically, and genetically similar individuals perform cooperative behavior to form social groups of unrelated individuals, this is an excellent model to examine the genes that underlie these social behaviors. We found that several core processes were significantly differentially regulated, including metabolism, stress response, aging, reproductive processes, and immunity. Interestingly, lipid metabolic processes were regulated across experiments; these may play a role in both nutrient storage/mobilization and chemical communication. In the future, it will be interesting to investigate whether the molecular pathways characterized in this study also are operating at earlier stages of the co-founding process (e.g., before the emergence of workers). Such studies will help elucidate the mechanisms responsible for the transition from cooperation to conflict in pleometrotic founding queens. Finally, fire ants also display genetically distinct monogyne (colonies headed by a single queen) and polygyne (colonies headed by multiple queens) social forms. It will be of great interest to determine if the same genes that regulate haplometrosis and pleometrosis also are involved in regulating queen number in mature colonies.
A total of 787 fire ant queens were collected immediately after a nuptial mating flight on May 5th, 2010 in a large parking lot (Target, 3970 SW Archer Rd, Gainesville, FL) and weighed. Since the area of collection has a high prevalence of monogyne colonies, we expected these queens to belong to the monogyne social form; this was subsequently confirmed by screening 108 queens for social form using Gp-9 as a marker following the protocol as described in [90]. The remaining 679 queens were split into two groups: 308 queens were set up in pairs (pleometrosis) based on having similar weights (range ±0.2 mg) and paint-marked with different colors, while 371 queens were set up individually (haplometrosis) and paint-marked as well. All the queens were provided with a nesting chamber consisting of a glass tube half-filled with water, which was covered by a cotton ball and a layer of dental plaster: this keeps the chamber moist but avoids an excess of water which is deleterious for the young brood. Tubes were sealed with a loose cap to provide air flow. Specimens were reared in the dark at 28°C, 70% relative humidity under claustral conditions (no food and no water) for 1 month.
After eclosion of the first batch of workers (minims), incipient colonies were provided with water, sugar water and frozen crickets. Glass tubes were set open in pencil boxes coated with Fluon to prevent escape. Queens were subsequently monitored daily until it was possible to identify the social rank of the two cofoundresses in pleometrotic couples. Previous studies have found that the initial cooperation between the two cofoundresses turns into conflict after the emergence of minims, resulting in the execution of one queen [19]. Queens that survive the competition (winners) are usually located at the top of the brood pile within the nest chamber and they are generally tended by workers; conversely, queens that will be executed (losers) are normally seen outside the nest chamber, hiding from workers in order to avoid being attacked (Figure S4). We used these observations to establish the social rank of the two pleometrotic queens, i.e. winner and loser. We collected 25 pleometrotic couples and 25 haplometrotic queens in dry ice and stored them at −80°C to be later processed.
This assay was performed with 34 couples of pleometrotic queens from the same pool of newly mated queens as experiment 1. The queens were paired and placed in nesting chambers as before. After emergence of minims, queens' behavior was monitored as before. Once the behavioral observation revealed the social rank of the two cofoundresses, queens were weighed again and re-paired with a different partner. We created the following three groups of queens: a) winner+winner (similar weight), b) loser+loser (similar weight), and c) winner+loser (different weights). Again, we monitored the behavior until the social rank of the newly coupled specimens was evident and we collected 4 new behavioral phenotypes in the same way as above: a) winners switched into losers (win/los, N = 7), b) losers switched into winners (los/win, N = 11), c) continuing winners (win/win, N = 12) and d) continuing losers (los/los, N = 5).
Individual fire ant queens were thawed and dissected under cold RNAlater (Qiagen, Valencia, CA) to confirm the mating status: unmated queens were not included in the analysis. Total RNA was extracted using the RNeasy Plus kit (Qiagen) combined with a RNase-Free DNase step (Qiagen) to remove any possible contamination by genomic DNA. Subsequent steps in the microarray analysis were performed at the Penn State Genomic Core Facility. RNA concentration and purity were assessed using NanoDrop and Qubit and RNA quality was assessed using RNA Nano Chips on the Agilent Bioanalyzer. 1 µg of each sample was amplified using the Ambion (Life Technologies) Amino Allyl MessageAmp II aRNA Amplification Kit (AM1753). 15 µg of aRNA were dyed with Cy3 or Cy5 (GE Health Care #RPN5661) and subsequently purified according to the Ambion Kit instructions. 1.5 µg of a Cy3 labeled sample were combined with 1.5 µg of a Cy5 labeled sample and fragmented using RNA Fragmentation Reagents (Ambion AM8740) according to the manufacturer's instructions. Samples were hybridized with mixing in a MAUI hybridization instrument overnight at 42°C. Arrays were scanned using Axon GenePix 4000B.
For the first microarray developed to validate the efficiency of probe sequences, we pooled RNA samples (2 µg total) from different castes, developmental instars and social forms as follows: 3 female alates, 15 workers, 5 larvae and 5 pupae from both monogyne and polygyne social forms and 5 males from monogyne colonies only.
The fire ant genome includes an official gene set of 16,569 protein-coding genes that were generated by a combination of ab initio, EST-based, and sequence similarity-based methods [27]. For our microarray studies, we combined the official gene set with a set of ESTs obtained from assemblies of the fire ant transcriptome for a total set of 63,436 sequences (“transcripts”). We successfully designed 60-mer probes for 51,531 of these transcripts (Roche NimbleGen, Inc., Madison WI). These sequences/probes were grouped into four categories: ESTs with gene models (EWGM, 7433 transcripts), ESTs without gene models (EWOGM, 40,613), gene models (GM, 3246) and gene models redundant with other models (GMRWOM, 239).
We developed and used a first microarray (1-plex 385,000 probe capacity, Roche NimbleGen, Inc., Madison WI) to validate the probe design and test multiple probes per transcript. On average, we designed 7 probes per transcript for a total of 355,930 probes. Each probe was tested for both the red (Cy5) and the green (Cy3) dyes. For transcripts with only one probe (N = 296), we verified that the probe had acceptable intensities for both dyes. For the other transcripts we examined the performance of the probes with the green dye only, because these showed consistently higher intensity compared to the red dye. Probes were ranked in the follow manner: a) if there were only 2 probes per transcript (N = 230), we selected the one with higher intensity; b) if there were 3 to 6 probes (N = 744), we calculated the ratio “probe intensity/median intensity of all probes for that transcript” and selected the probe with highest ratio if the value was <3, otherwise we selected the probe with the second highest ratio; c) for transcripts with 7 probes (N = 50,261), we followed the procedure as in “b” but, in case the probe with the highest ratio was >3, we removed that probe, calculated new ratios and selected a new probe with highest ratio. This procedure allowed us to select the probes with highest intensity that were not outliers.
Selected probes were printed in pairs on two 12-plex microarrays (each array had a 135,000 probe capacity, Roche NimbleGen, Inc., Madison WI). We used a loop design with dye swaps incorporated, allowing us to hybridize 24 RNA samples to each array. For experiment 1 we hybridized 8 haplometrotic queens, 8 pleometrotic winners and 8 pleometrotic losers (Figure S5) and for experiment 2 we compared 6 win/los, 6 los/win, 6 win/win and 5 los/los (Figure S6).
Any spots with an intensity of less than 300 (the background level on the arrays) were removed from the analyses, as were spots present on less than 20 out of 24 arrays. Expression data were log-transformed and normalized using mixed-model normalization (proc MIXED, SAS, Cary, NC) with the following model:where Y is expression, dye and block are fixed effects, and array, array*dye and array*block are random effects. Transcripts with significant expression differences between groups were detected by using a mixed-model ANOVA with the model:where Y represents the residual from the previous model. Treatment, spot and dye are fixed effects and array is a random effect. P-values were corrected for multiple testing using a false discovery rate of <0.001 for experiment 1 and <0.1 for experiment 2 (proc MULTTEST, SAS). Because the number of differentially regulated transcripts for experiment 1 was very high (∼13,000 out of ∼50,000), and to avoid an excess of redundancy among the different groups of transcripts, we included only probes corresponding to GM and EWGM (see above).
Hierarchical clustering, using the Ward method, and principal component analysis (PCA) for global patterns of gene expression were performed in JMP 9.0.2 (SAS, Cary, NC). We used Genesis 1.7.6 (Graz, Austria) to cluster differentially regulated genes based on average linkage and to perform k-means clustering in experiment 1. Gene Ontology analysis was performed using functional annotation chart/clustering in DAVID version 6 [28], [29] using DAVID default population background and a cutoff of p<0.05. For all Gene Ontology (GO) analyses, fire ant genes were matched to their Drosophila orthologs in FlyBase (http://flybase.org/). CateGOrizer [31] was used to count the occurrences of significantly enriched GO terms within each of the pre-defined set of parent/ancestor GO terms. The array data were deposited on the ArrayExpress website according to MIAME standards (ArrayExpress accession: E-MEXP-3886 for experiment 1, E-MEXP-3898 for experiment 2).
We compared the results from experiment 1 to the following studies:
We performed overlaps between list of transcripts and GO terms with Venny [91]. In the first comparative study we overlapped fire ant transcripts directly, while in the second study we used Drosophila orthologues (FlyBase numbers) to compare fire ant transcripts to the genes of the fruit fly. Statistical significance of the overlap was calculated using a hypergeometric test (http://nemates.org/MA/progs/overlap_stats.html). Selected GO analyses based on study overlap were performed in DAVID as above. In the second study, to test for the significant agreement in the patterns of expression between two studies we performed Fisher's Exact Tests in JMP.
We examined gene expression levels of the following candidate genes (Table S10): Indy and Sod2 (determination of adult life span); Dredd and kay (immune response); desat1, ifc and Putative fatty acyl-CoA reductase CG5065 (synthesis and metabolism of fatty acids); br and Btk29A (reproductive functions); Sema-5c and Mer (olfactory behavior); fru (aggressive behavior) and woc (neurogenesis). We used the total RNA extracted from fire ant queens for the microarray analysis and compared gene expression between haplo and los on an ABI Prism 7900 sequence detector (Applied Biosystems, Foster City, CA, USA). cDNA was made using SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen-Life Technologies, Carlsbad, CA, USA) and Random Hexamers according to the manufacturer's protocol. The cDNA was then diluted 2(x) with ultra-pure water. Amplification was performed in a 10 µl reaction mixture containing 5 µl of 2× SYBR Green Master Mix (Applied Biosystems-Life Technologies, Carlsbad, CA, USA), 1 µl of each primer (10 µM) and 2 µl of cDNA at the following conditions: 50°C for 2 min, 95°C for 10 min, 40 cycles of 95°C for 15 sec and 60°C for 1 min, a dissociation step of 95°C for 15 sec and 60°C for 15 sec. We used 8 queens per group: triplicate reactions were performed for each of the samples and averaged for use in statistical analysis. Expression levels of candidate genes were normalized to the geometric mean of two housekeeping genes, Rp-9 and Rp-37 [27]. Negative control (cDNA reaction without RT enzyme) was also used. Primer sequences were developed in Primer3Plus (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) and primer efficiency was first validated using standard curves. Statistical analysis was performed with nonparametric Kruskall-Wallis rank sums in JMP 10 (SAS, Cary, NC). The data were shown normalized to the haplo group.
|
10.1371/journal.pbio.1001679 | eIF4EBP3L Acts as a Gatekeeper of TORC1 In Activity-Dependent Muscle Growth by Specifically Regulating Mef2ca Translational Initiation | Muscle fiber size is activity-dependent and clinically important in ageing, bed-rest, and cachexia, where muscle weakening leads to disability, prolonged recovery times, and increased costs. Inactivity causes muscle wasting by triggering protein degradation and may simultaneously prevent protein synthesis. During development, muscle tissue grows by several mechanisms, including hypertrophy of existing fibers. As in other tissues, the TOR pathway plays a key role in promoting muscle protein synthesis by inhibition of eIF4EBPs (eukaryotic Initiation Factor 4E Binding Proteins), regulators of the translational initiation. Here, we tested the role of TOR-eIF4EBP in a novel zebrafish muscle inactivity model. Inactivity triggered up-regulation of eIF4EBP3L (a zebrafish homolog of eIF4EBP3) and diminished myosin and actin content, myofibrilogenesis, and fiber growth. The changes were accompanied by preferential reduction of the muscle transcription factor Mef2c, relative to Myod and Vinculin. Polysomal fractionation showed that Mef2c decrease was due to reduced translation of mef2ca mRNA. Loss of Mef2ca function reduced normal muscle growth and diminished the reduction in growth caused by inactivity. We identify eIF4EBP3L as a key regulator of Mef2c translation and protein level following inactivity; blocking eIF4EBP3L function increased Mef2ca translation. Such blockade also prevented the decline in mef2ca translation and level of Mef2c and slow myosin heavy chain proteins caused by inactivity. Conversely, overexpression of active eIF4EBP3L mimicked inactivity by decreasing the proportion of mef2ca mRNA in polysomes, the levels of Mef2c and slow myosin heavy chain, and myofibril content. Inhibiting the TOR pathway without the increase in eIF4EBP3L had a lesser effect on myofibrilogenesis and muscle size. These findings identify eIF4EBP3L as a key TOR-dependent regulator of muscle fiber size in response to activity. We suggest that by selectively inhibiting translational initiation of mef2ca and other mRNAs, eIF4EBP3L reprograms the translational profile of muscle, enabling it to adjust to new environmental conditions.
| Most genes are transcribed into mRNA and then translated into proteins that function in various cellular processes. Initiation of mRNA translation is thus a fundamental control point in gene expression. Working in a zebrafish model, we have found that muscle activity (or inactivity) can differentially regulate the translation of specific mRNAs and thereby control the growth of skeletal muscle. Emerging evidence suggests that control of translational initiation of particular mRNAs by an intracellular signaling pathway acting through TORC1 is a major regulator of cell growth and function. We show here that muscle activity both activates the TORC1 pathway and suppresses the expression of a downstream TORC1 target—the translational inhibitor eIF4EBP3L. This removes a brake on translation of certain mRNAs. Conversely, we show that muscle inactivity can up-regulate this translational inhibitor, thereby causing reduced translation of these mRNAs. One of the mRNAs targeted in this manner by eIF4EBP3L is Mef2ca, which encodes a transcription factor that promotes assembly of muscle contractile apparatus. Our work thus reveals a mechanism by which muscle growth can be differentially influenced depending on the context of muscle activity (or lack thereof). If this pathway operates in people, it may help explain how exercise regulates muscle growth and performance.
| Control of gene function at the level of mRNA translation is emerging as a major regulator of cell and developmental biology, with medical relevance in cancer and elsewhere [1]–[5]. Several broadly acting signaling pathways appear to control translational modules in cultured cells, whereby thousands of mRNAs are coordinately regulated through specific recognition of motifs in the mRNA that remain to be fully defined [6],[7]. One major regulator of such translational modules is the TOR (Target of Rapamycin) pathway [8],[9]. In organisms from yeast to man, TOR regulates protein synthesis, cell size, and general metabolism through several distinct mutliprotein complexes [10]–[12]. A major function of these complexes is to target TOR's protein kinase activity to particular substrates, among them the proteins p70 S6 kinase (p70S6K) and eukaryotic Initiation Factor 4E binding proteins (eIF4EBPs), which regulate protein synthesis [13],[14]. But how TOR achieves specific regulation of diverse cell behaviors in so many different cell types is unclear. Here we define a mechanism that sensitizes specific mRNAs to TORC1 activity in skeletal muscle.
Muscle has advantages for the study of TOR signaling and its role in cell size control. Muscle fibers are among the few cell types that undergo dramatic and reversible changes in size during the normal life of an organism. Moreover, as postmitotic cells, muscle fibers are unaffected by changes in cell cycle that complicate analysis of cell growth and size in proliferative cells [15]. A major regulator of muscle fiber size is exercise, known experimentally as “activity” [16]–[18]. For example, 14 days of mouse hind limb suspension, which decreases electrical activity and the force it elicits, led to reduction of 25%–55% in muscle mass [19],[20]. In humans, reduction of activity leads rapidly to muscle wasting with huge societal and health implications in hospitalized patients and the ageing population. Electrical inactivity has been shown to promote protein breakdown through activation of atrogenes, some of which encode components of both the proteasomal and autophagic degradation pathways [21]–[23]. Inactivity also reduces TORC1 signaling in muscle, which has been suggested to reduce protein synthesis [24],[25] and lead to atrophy.
Zebrafish muscle provides a particularly good system in which to study the effects of activity on muscle growth and in neuromuscular disease [26]–[28]. Two types of muscle fibers are formed in each segmented myotome during the first day postfertilization, a superficial layer of mononucleate slow fibers and a larger number of underlying multinucleate fast fibers [29]–[33]. Over ensuing days, both fiber types undergo significant growth, but the slow fibers remain mononucleate. Zebrafish first move at 17 hours postfertilization (hfp) and undergo repeated contractions during the embryonic and larval stages [34]. Such contractions can readily be blocked without preventing development [35]–[39], providing an opportunity to examine the effect of activity on muscle growth in the developmental context, which is hard to achieve in other vertebrate models.
Here we show that inactivity prevents translation of a specific set of muscle mRNAs, including that encoding Mef2ca, a transcription factor essential for normal muscle growth. Inactivity acts in two ways, first by promoting accumulation of the inhibitor of protein synthesis initiation eIF4EBP3L and second by reducing TORC1 signaling and thus permitting eIF4EBP activity. Active eIF4EBP3L blocks translational initiation of Mef2ca, preventing normal myofibrilogenesis and muscle growth.
To test the role of activity in zebrafish muscle growth, we examined the reduction in myofiber size in two inactive conditions: the immotile mutant chrndsb13/sb13, which lacks the acetylcholine receptor δ subunit (Figure S1), and after exposure to tricaine mesylate (MS222), a zebrafish anesthetic drug. Both treatments block electrical activity and therefore contraction in muscle. Three parameters were assayed, the width of myofibrillar bundles within slow myofibers, myosin heavy chain (MyHC) immunofluorescence detected with F59 antibody, and the content of various myofibrillar proteins by Western analysis. In the absence of activity, myofibrillar bundle width diminished by 30% in chrndsb13/sb13 mutants and 22% after MS222 treatment, compared to their respective controls (Figure 1A,B). Similarly, there were decreases in slow MyHC immunofluorescence of 61% and 36% and reductions in slow MyHC by Western of 12% and 52%, respectively (Figure 1A,C,D,F). To investigate whether the decrease in MyHC immunoreactivity was unique to slow fibers or whether it also occurred in fast fibers, we assessed total somite MyHC. MyHC immunoreactivity was decreased by 44% in chrndsb13/sb13 relative to its siblings (Figure 1E). The fast muscle protein myosin light chain, recognized by F310, was reduced by 42%, as were general muscle proteins such as Actin (40%) and MyBPC (35%) (Figures 1D,F and 2A). Thus, in the zebrafish, as in other models, inactivity reduces myofibril content, permitting the use of zebrafish to investigate in vivo the mechanisms that regulate muscle size following disuse.
We have previously shown that lack of Mef2 activity leads to poor myofibril assembly and that lack of Myod reduces early muscle growth [40],[41]. We therefore tested whether inactivity reduced production of these proteins in parallel with the reduction in muscle growth. MS222 treatment for just 17 h led to a dramatic loss of Mef2 protein without any appreciable change in Myod relative to other control muscle proteins such as Actin and Vinculin (Figure 2A,B). Several Mef2 genes are expressed in differentiated zebrafish muscle, particularly mef2ca, mef2cb, mef2d and mef2aa [42]–[44]. Of these, mef2ca and mef2d are the most abundant at 48 hpf (Figure S2). Mef2C has pleiotropic anabolic effects on murine myogenesis [45],[46]. Consistent with this observation, we found that Mef2c protein was reduced by MS222 treatment and in chrndsb13/sb13 mutants (Figure 2C). Moreover, whole-mount immunofluorescence confirmed a decrease of Mef2c in myotome nuclei following MS222 treatment (Figure 2D).
To test whether a decrease in Mef2c contributes to the reduction in myofiber width, mutant and sibling embryos from a mef2catn213/+ incross were compared. A 20% decrease in myofibril bundle width and in somite volume was observed in 48 hpf (Figure 2E,G) and 5 dpf mutants (Figure 2F,H). These findings show that Mef2ca activity is required for normal muscle growth.
We next asked whether Mef2c mediates the effect of muscle activity on myofiber width. MS222 causes an even greater reduction of myofibril bundle width than a complete loss of Mef2ca, indicating that muscle activity does more than promote Mef2ca activity (Figure 2E,G). Nevertheless, in the absence of activity, loss of Mef2ca has no effect on myofibril bundle width (Figure 2E,G), suggesting that the Mef2ca pathway is inactive when muscle itself is inactive. These results argue that Mef2ca contributes to muscle activity-induced growth, but is not the sole mechanism.
The preferential reduction in Mef2c protein in inactive muscle could stem from either increased proteolysis or reduced synthesis. In adult muscle, inactivity potently activates both the proteasomal and autophagy-lysosomal proteolysis pathways [47],[48]. To test the role of proteasomal degradation in Mef2 regulation, muscle activity was abolished in the presence of MG132, an inhibitor of the proteasome. MG132 treatment of fish embryos up-regulated known targets of the proteasome, p53 and Sqstm1 [49],[50], providing positive controls for MG132 efficacy (Figures 3A and S3A). Inactivity led to decrease in Mef2 levels even in the presence of MG132 (Figure 3A,B), indicating that Mef2c decline is not due to proteasomal degradation. Congruently, rather than increasing Mef2c, MG132 alone decreased Mef2c, suggesting that proteasomal degradation is not a significant Mef2c turnover pathway in embryonic muscle (Figure 3A,B).
Autophagy was then assessed by the level of Sqstm1, which is a target to autophagic degradation and marker of autophagic flux [51]. The TORC1 inhibitor rapamycin, a known and potent inducer of autophagy, reduced zebrafish full-length 50 kd Sqstm1, serving as a positive control by showing that autophagy reduces Sqstm1 in zebrafish (Figure 3A,B). The level of zebrafish Sqstm1 was not decreased by muscle inactivity (Figure 3A,B). As a further control, we found no increase in sqstm1 mRNA that could have compensated for a loss of Sqstm1 protein ([52]; Figure S3B). Therefore, muscle inactivity did not induce autophagy in embryonic zebrafish. Moreover, inhibition of autophagy with the lysosomal inhibitor chloroquine did not block the reduction in Mef2c protein caused by inactivity (Figure S3C). Furthermore, Mef2c level appeared unchanged under rapamycin-induced autophagy, indicating that Mef2c is not a target for autophagic degradation (Figure 3A,B). Thus, the reduction of Mef2c in inactive muscle was apparently not due to enhanced degradation by the ubiquitin-proteasome or autophagy pathways, which suggested that its synthesis is regulated.
To determine whether the reduction in Mef2 was due to a reduction in mRNA synthesis, we used qPCR. Upon MS222 exposure, no decrease in mRNA level for any of the four Mef2 genes was detected (Figure 3C). Thus, reduced mRNA levels does not account for loss of Mef2 in inactive muscle.
The lack of demonstrable change in Mef2 mRNAs or Mef2c proteolysis in inactive embryonic muscle prompted us to examine changes in protein synthesis, particularly because muscle activity is known to promote protein synthesis [53],[54]. We therefore developed a method to quantify RNA associated with polysomes in fish embryos. Polysomes are complexes that contain actively translating mRNA associated with two or more ribosomes and represent the densest RNA-containing fraction of cytoplasm. Embryo cytoplasmic lysate was subjected to sucrose density gradient separation and the location of the zebrafish polysome-containing fraction determined (Figures 3D,E and S4). As controls, polysomes were diminished by addition of EDTA to the extract, which dissociates polysomes, and enhanced by addition of cycloheximide, which stabilizes polysomes by inhibiting translational elongation, thereby stalling ribosomes on the mRNA (Figure S4). The effect of activity on polysomes was then determined. Following 17 h MS222 exposure, the amount of nucleic acid in the polysome fraction was reduced compared to that in the 80S monosome and 60S and 40S ribosomal subunit peaks (Figure 3E). Polysome-associated RNA decreased by ∼10% in MS222 compared to active controls. This finding shows that muscle inactivity reduced protein synthesis in zebrafish, providing a potential explanation for the reduction in Mef2c in inactive muscle.
To determine whether specific mRNAs such as those encoding Mef2 are subject to translational regulation, the proportion of mef2ca mRNA in polysomes was assessed. mRNA from the polysome (heavy fraction; P) was purified and the change in mRNAs following MS222 treatment determined by qPCR (Figure 3F). Translation of mef2ca reduced by 55% in the absence of activity, whereas translation of various controls, such as dmd (dystrophin) and actn3b (α-actinin 3b), differed significantly, showing no reduction (Figure 3F). We did not detect significant reduction in translation of other mRNAs tested by qPCR. As expected, relative reduction of mRNAs in the polysomal fraction was complemented by a relative increase in the subpolysomal fraction (Table S2). These findings show that there is preferential loss of specific mRNAs from polysomes in inactive muscle, and that Mef2ca protein translation is specifically promoted by muscle activity.
TORC1 was previously shown to differentially regulate translation [1],[2],[55],[56], and is a major player in activity-related muscle growth [18],[19]. We therefore asked whether muscle activity affects TORC1 activity in the embryonic zebrafish. TORC1 pathway activity was assessed by examining the phosphorylation of eIF4EBP and ribosomal protein S6, known downstream targets of the TORC1 complex involved in translational regulation [57]. When zebrafish embryos were exposed to MS222, Western analysis revealed that phosphorylation of S6S240/244 and eIF4EBPT37/46 was decreased by 40% and 30%, respectively, compared to untreated controls (Figure 4A). As these proteins are widely expressed, immunostaining of phopho-S6 (pS6) and total S6 was used to assess changes occurring in muscle tissue. MS222 caused a decrease in muscle pS6 compared to controls (Figure 4B). The levels of S6 itself also appeared somewhat down-regulated, although no loss of ribosomal 18S or 28S rRNA was detected in inactive embryos (Figure S5). Thus, muscle activity activates the TORC1 pathway, making it a good candidate as a regulator of the translational response to activity.
TORC1 is inhibited by rapamycin, which is known to reduce muscle growth [18],[19]. Rapamycin treatment of zebrafish embryos reduced muscle growth, but to a lesser extent than inactivity (Figure S6A–C). Strikingly, although rapamycin reduced muscle growth, it was much less effective than inactivity at reducing Mef2c (Figure 3A,B; see also [58]) or Myosin and Actin protein accumulation (Figure S6D). Congruently, rapamycin was more effective at reducing S6 phosphorylation than eIF4EBP phosphorylation (Figure S7), whereas inactivity appeared equally effective on each TORC1 target (Figure 4A). Thus, the greater effectiveness of inactivity compared to rapamycin on reducing muscle growth correlates with the ability of inactivity to down-regulate Mef2c translation.
TORC1 can affect both global translation by promoting rRNA synthesis via S6K [59],[60], and differential translation by blocking eIF4EBPs [1],[2]. eIF4EBPs prevent translational initiation by binding eIF4E, inhibiting its interaction with eIF4G, and thereby preventing recruitment of the 40S ribosomal subunit to the mRNA [61],[62]. To determine which TOR-dependent translational regulation operates in inactive muscle, we first looked at rRNA. We tested the levels of rS28 and rS18; after 16 h MS222 they were not reduced (Figure S5). On the other hand, reduction in protein level following inactivity occurred rapidly and varied from protein to protein (Figures 1, 2, and 3), which implied that TORC1-dependent differential translational regulation occurs, potentially via eIF4EBP.
Active TORC1 phosphorylates and inhibits eIF4EBP, thereby promoting translational initiation of specific mRNAs [1],[2],[63],[64]. The vertebrate eIF4EBP protein family consists of three members—eIF4EBP1, eIF4EBP2, and eIF4EBP3—that share >55% amino acid identity (Figure S8A). Each harbors a conserved eIF4E binding site, YDRKFLL, two canonical TOR phosphorylation sites, TPGGT, and several transregulatory phosphorylation sites (Figure S8C). Zebrafish have four eIF4EBPs, with an eIF4EBP3 duplicate named eIF4EBP3-like (eIF4EBP3L; Figure S8B). eIF4EBP3L has greater homology (78% identity) to human eIF4EBP3 than between any other zebrafish∶human eIF4EBP pair (Figure S8B). Two eIF4EBP3s are found in many teleosts (www.ensembl.org), suggesting they arose from the early teleost genome duplication. The zebrafish eIF4EBP genes have distinct tissue expression: eif4ebp1 and eif4ebp2 widely and highly in head and neural tissue, eif4ebp3 only abundant in pancreas, and eif4ebp3l in the somite, eye, and branchial arch region [65]. Thus, eif4ebp3l appeared to have a particular role in muscle tissue.
The effect of muscle activity on eIF4EBP gene expression was assessed. Upon exposure to MS222, the level of eif4ebp3l mRNA increased 2-fold, particularly in muscle, but did not otherwise change tissue mRNA distribution. MS222 also induced eif4ebp1 mRNA 1.9-fold (Figures 4C and S8D). To rule out effects of MS222 on neural activity and to examine the effect of activity specifically in muscle tissue, we assessed mRNA levels in the chrndsb13 mutant, in which muscle alone is inactive, and found a 2.5-fold increase in eif4ebp3l mRNA but without change in eif4ebp1 mRNA (Figure 4D). These data indicate a specific role for eif4ebp3l, and not eif4ebp1, in the response to muscle inactivity.
Separately, we analyzed the effect of inactivity on so-called atrogenes, genes known to be induced in muscle atrophy caused by malnutrition, denervation, sepsis, or ageing in adult mammalian muscle [66],[67]. We found, as in the case of eif4ebp3l, a 2-fold increase in atrogin-1, murf2, and trim32 mRNAs induced by MS222 (Figure 4E) and similar induction in chrndsb13 mutants (Figure S9). These results show that inactivity up-regulates a suite of mRNAs regulating protein turnover in muscle, including eif4ebp3l.
When translation of eif4ebp3l mRNA was analyzed by polysomal fractionation, a significantly increased fraction was in polysomes (Figure 3E). These increases in both eif4ebp3l mRNA level and its translation rate suggest that eIF4EBP3L protein will increase in inactive muscle. Lacking an eIF4EBP3L-specific antibody, we could not show such up-regulation directly, but total eIF4EBP protein did appear higher in inactive embryos (Figure 4A). We hypothesized that increased eIF4EBP3L might cause certain mRNAs to become more sensitive to TORC1-regulation in inactive muscle. Consistent with this idea, more significant reduction in myofibrilogenesis was caused by MS222 than by inhibiting TORC1 with rapamycin alone (Figure S6). Thus, muscle inactivity both increased eIF4EBP level and reduced eIF4EBP phosphorylation, raising the possibility that eIF4EBP3L drives the differential translational repression in inactive muscle.
As inactivity increased active (unphosphorylated) eIF4EBP3L and reduced translation of mef2ca mRNA and myogenesis, we tested whether Mef2ca protein is down-regulated when eIF4EBP3L is active. Embryos were injected with mRNA encoding eIF4EBP3L to achieve a 2-fold increase, comparable to that induced by inactivity (Figure S10A). First, the intensity of Mef2c immunofluorescence was measured in confocal stacks of skeletal muscle nuclei (Figure 5A). Overexpression of eIF4EBP3L in active control muscle did not significantly affect Mef2c levels, in contrast to the reduction observed with MS222 (Figure 5A,B). However, overexpression of eIF4EBP3L in inactive MS222-treated embryos, a condition in which eIF4EBP is hypophosphorylated (Figure 4A), significantly decreased Mef2ca protein levels by a further 20% below that caused by MS222 alone (Figure 5B). We conclude that eIF4EBP3L can inhibit Mef2c accumulation in muscle, but that muscle activity normally suppresses eIF4EBP3L activity.
To determine whether inhibition of TORC1 activity can mimic the effect of MS222 on eIF4EBP3L, rapamycin was applied to embryos injected with eif4ebp3l mRNA and controls. Rapamycin suppressed TORC1 activity in zebrafish embryos, reducing both phosphoS6 and phospho4EBP (Figure S7). As in Figure 3A, rapamycin had no effect on Mef2c level in mock-injected control active muscle (Figure 5A,C), presumably because the low endogenous eIF4EBP3L level was insufficient to cause a significant reduction in initiation of Mef2ca translation. However, when eIF4BP3L was overexpressed, rapamycin caused a 40% reduction in Mef2c level (Figure 5C), comparable to that triggered by MS222 (Figure 5B). Thus, TORC1 activity suppresses eIF4EBP3L function in active skeletal muscle.
To test the role of endogenous eIF4EBP3L, we knocked down eIF4EBP3L by injecting 1–2 cell stage embryos with morpholino antisense oligonucleotides (MO) targeting the exon1-intron1 splicing site (BP3LMO1) and against the start codon (BP3LMO2) of eif4ebp3l. In the absence of specific antibody against eIF4EBP3L, we verified the efficacy of BP3LMO1 by qPCR and found a decrease of 80% in eif4ebp3l mRNA level, whereas mef2ca mRNA was not affected, relative to actin (Figure 5D). In embryos injected with control MO, inactivity led to a 40% decrease in Mef2c protein. This decrease was prevented when eIF4EBP3L was knocked down (Figure 5E). Thus, eIF4EBP3L is necessary for the reduction in Mef2c in inactive muscle.
The involvement of eIF4EBP3L in Mef2c reduction caused by inactivity is most simply explained by altered translational initiation. To test this hypothesis, embryos injected with eif4ebp3l RNA were subjected to polysomal fractionation. Overexpression of eIF4EBP3L led to a 40% decrease in mef2ca mRNA in the polysome fraction, whereas dystrophin mRNA, a negative control that was unaffected by activity (Figure 3F), showed no change (Figure 5F). Moreover, addition of rapamycin to inhibit TORC1 enhanced the effect of eIF4EBP3L overexpression (Figure 5F), and decreased Mef2c protein (Figure 5C). Whereas either eif4ebp3l RNA injection or 6 h rapamycin treatment alone reduced mef2ca mRNA in polysomes, neither was sufficient to lower Mef2c protein (Figure 5A,C,F). This finding suggested that, in order for the translation inhibition to lead to reduction of Mef2c protein level, such inhibition must exceed a threshold, as occurs when both eIF4EBP3L is up-regulated and TORC1 is inhibited.
To confirm that endogenous eIF4EBP3L specifically regulates mef2ca translation, we reduced eIF4EBP3L level with morpholino and performed polysome fractionation on active and inactive muscle. In the presence of control MO, muscle inactivity reduced mef2ca translation by 37% (p = 0.0035, while having no effect on dmd mRNA; Figure S11). In contrast, in the presence of eIF4EBP3L MO, muscle inactivity had no significant effect on mef2ca translation (Figure S11). In addition, knockdown of eIF4EBP3L in both active and inactive muscle led to an increase in mef2ca mRNA in polysomes, but did not affect dmd mRNA (Figure S11). Nevertheless, Mef2ca protein did not accumulate upon eIF4EBP3L knockdown (Figure 5E). These data suggest that eIF4EBP3L is required for the reduction in mef2ca translation in inactive muscle and retains the ability to target specific mRNAs in active muscle.
As Mef2ca regulates myofibrilogenesis and fiber growth (Figure 2E,F), we asked whether eIF4EBP3L also does so. In active control muscle, eif4ebp3l morphant myofibrilogenesis appeared similar to that in controls (Figure 6A). Inactivity caused a 75% decrease in slow MyHC signal in embryos injected with control MO. However, this decline was prevented when eIF4EBP3L was knocked down with either BP3L MO (Figure 6A). We conclude that eIF4EBP3L prevents normal myogenesis under inactive conditions.
Inactive muscle has both increased eif4ebp3l mRNA and decreased phosphorylation of TORC1 targets, both of which cooperate to specifically suppress translation (Figures 4 and 5). To test whether eIF4EBP3L can influence myofibrilogenesis independent of TORC1 activity, we generated constitutively active eIF4EBP3L by mutating the five potential threonine/serine phosphorylation sites to alanine, thereby generating 5A3L (Figure S8C). However, injecting 5A3L RNA led to early embryonic death, possibly due to widespread translational repression.
To overcome the early lethality, we cloned 5A3L into a zebrafish heat-shock–inducible expression vector also containing a GFP marker to achieve controlled mosaic expression in muscle. Slow myofibrilogenesis was assayed at 54 hpf by comparing GFP-expressing fibers to unmarked neighboring fibers within the same somite. Overexpression of 5A3L led to a 36% decrease in myofibril bundle width (Figure 6B), whereas controls expressing GFP alone showed no significant change. Moreover, a variety of other cell types in epidermis, notochord, and neural tube showed no detectable morphological change upon 5A3L overexpression (Figure S10B). Thus, active eIF4EBP3L reduces slow myofibrilogenesis.
The current work provides three new insights into translational control in the context of skeletal muscle tissue in vivo: first, that muscle activity differentially regulates translation of specific mRNAs; second, that the conserved protein eIF4EBP3L, a downstream target of TORC1 signaling, regulates translation and myofibrilogenesis in response to activity; and third, that Mef2ca is among the translationally controlled targets that mediate the effect of electrical activity on myofibrilogenesis and thereby muscle growth.
By applying polysomal mRNA fractionation to whole zebrafish embryos, we have developed a method to analyze changes in their translational control. We find that muscle activity promotes translation of a specific mRNA that is normally expressed in muscle and encodes an important muscle regulator, namely Mef2ca (Figure 7). The translational changes we highlight occur in relation both to total RNA content of the embryo and to other muscle-specific mRNAs, such as Dystrophin and α-Actinin-3b. The reduced translation triggered by inactivity leads to rapid loss of Mef2 protein.
Muscle electrical activity controls many aspects of adult muscle character, including contractile and metabolic properties and fiber size. In developing amniote primary muscle, however, the major effect of reduced activity is reduced growth rate [68],[69] (reviewed in [70]). We show here that growth is also reduced in electrically inactive zebrafish muscle, consistent with previous comments [39]. Without activity, initial differentiation of slow and fast muscle fibers appears normal, but accumulation of myofibrillar protein is diminished. In the mononucleate slow fibers, which form in normal numbers, myosin fails to assemble into myofibrils at the normal rate, leading to a reduced width of myofibril bundles.
The pathways that mediate the effects of activity on muscle growth are many and fall into two broad categories, catabolic pathways triggered by inactivity and anabolic pathways promoted by activity. The relative contribution of proteolysis or protein synthesis changes to muscle mass decrease in the inactive condition is a controversial topic. Although catabolism is a major driver, altered protein synthesis may also contribute [54],[71],[72]. Our data reveal eIF4EBP3L as a new regulator of muscle growth that is up-regulated in inactive muscle and inhibits translational initiation through targeting a specific subset of mRNAs. A high-throughput approach will be needed to define the full set of mRNAs showing altered translation in inactive muscle. We also found that muscle inactivity causes a 2-fold increase in E3 ligase atrogenes, known targets of FoxO, that trigger proteasomal degradation of muscle constituents [73],[74]. Thus, our data suggest that both decreased protein synthesis and increased catabolism are triggered in response to inactivity in developing muscle.
How does inactivity up-regulate eif4ebp3l expression? In other systems, eIF4EBP mRNAs are up-regulated by the FoxO pathway [75]–[79]. Considering the increase in transcription of both eif4ebp3l and known atrogenes in inactive muscle, we hypothesize that FoxOs are activity-dependent regulators of transcription of both eif4ebp3l and the E3 ligases in growing zebrafish muscle.
We show that loss of eIF4EBP3L function prevents atrophic changes caused by inactivity during growth. If a human eIF4EBP functions similarly in adults, it could constitute a novel therapeutic target for prevention of acute muscle wasting. Changes in other eIF4EBPs, such as eIF4EBP1, have been observed in muscle atrophy caused by fasting, and have been suggested to regulate general translation rate [80]. Interestingly, we observed significant increase in eif4ebp1 mRNA in our zebrafish inactivity model only when whole body electrical activity, not just muscle activity, was blocked. As eif4ebp1 is highly expressed in the central nervous system, these data suggest that eIF4EBP1 may respond to activity by mediating growth-related changes, such as synaptic elaboration, during neural development [81],[82].
eIF4EBPs block translation by binding to eIF4E and preventing eIF4G attachment to the 5′ cap-dependent mRNA initiation complex. TORC1 pathway activity phosphorylates eIF4EBP, inhibiting its interaction with eIF4E and thereby promoting translational initiation. The TORC1-4EBP pathway was previously thought of as a general regulator of 5′ cap-dependent translation [61],[63],[83]. However, recent large-scale screens revealed differential mRNA translation by TORC1-4EBP in cultured cells. Specific sequences within an mRNA, known as TOP-like or Pyrimidine-Rich Translational Elements (PRTEs), correlate with sensitivity to translational regulation [1],[2]. We find similar differential control by TORC1-4EBP3L in zebrafish muscle in vivo. The mef2ca mRNA that is translationally repressed by eIF4EBP3L has a PRTE in the 5′ UTR, as does mef2d (Table S3). However, as eif4ebp3l itself also has a PRTE, defining an mRNA sequence consensus for eIF4EBP3L translational repression in muscle will require further analyses.
Muscle inactivity reduced phosphorylation of eIF4EBP and ribosomal protein S6, suggesting that loss of TORC1 pathway activity triggered the translational and growth reduction. However, in zebrafish, inactivity reduces Mef2c levels and muscle growth more effectively than the TORC1 inhibitor rapamycin, as previously observed in mammals [19]. We found that up-regulation of eIF4EBP3L in inactive muscle helps explain why inactivity is more effective than rapamcyin alone. In active muscle, eIF4EBP3L levels are low and the activity of eIF4EBP3L is further suppressed by phosphorylation by TORC1. In contrast, when eIF4EBP3L is induced in inactive muscle, translation becomes more sensitive to TORC1 activity. In this situation, the inactivity also lowers eIF4EBP3L phosphorylation, leading to a high level of active eIF4EBP3L, which specifically inhibits translation of certain mRNA targets, such as mef2ca. We confirm this hypothesis by showing that when eIF4EBP3L is overexpressed, Mef2c level becomes more sensitive to rapamycin. Thus, the level of eIF4EBP3L acts as a gatekeeper, controlling the sensitivity of muscle to TORC1 activity.
In contrast to eIF4EBP1 and eIF4EBP2, the function of eIF4EBP3 has been unclear [84],[85]. We show that eIF4EBP3L blocks translational initiation of specific mRNAs and simultaneously sensitizes the inactive muscle to TORC1. One advantage of such control is that, when activity returns, protein synthesis can rapidly resume as soon as TORC1 activity increases and phosphorylates eIF4EBP3L. Thus, if the pathway we have revealed in developing muscle were to also operate in inactive adult muscle, eIF4EBP3L might constitute a fast switch for recovery of atrophic muscle.
Other eIF4EBPs also appear to function in the reaction of muscle to distinct stresses. Stress induces Drosophila eif4ebp as a metabolic brake [86],[87]. Likewise, either eIF4EBP1−/− or eIF4EBP1−/−;eIF4EBP2−/− mice show defects in fat and muscle metabolism under stress conditions, but not under normal conditions [86],[88],[89]. Our data suggest that eIF4EBP3L acts as a metabolic brake preventing anabolism in inactive muscle, and also regulates cell growth by specific translational control in response to TORC1. The presence of high levels eIF4EBP3L sensitize muscle to TORC1, in agreement with the recent finding that eIF4E/eIF4EBP ratio is a key determinant of TORC1 action [90]. Activity, nutritional status, and other factors influencing TORC1 activity would be expected to have stronger effects when eIF4EBPs are up-regulated. Strikingly, many manipulations that trigger either muscle growth or atrophy alter expression of an eIF4EBP [78],[80],[91],[92]. In the brain, eIF4EBP2 appears to act downstream of TORC1 signaling to control the translation of specific mRNAs involved in synaptogenesis and linked to autism [5]. As there are at least three eIF4EBP genes in all vertebrates examined, it is possible that each becomes a gatekeeper of TORC1 in response to distinct stresses, and could then select which particular mRNAs become TORC1 targets.
The specific regulation of mef2ca mRNA translation begs the question of the role of Mef2 in growing muscle. Myofibril assembly is critically dependent upon Mef2 activity, which seems to be particularly important for thick filament biogenesis [41],[46]. We show that Mef2ca by itself is essential for normal fiber growth. Mef2ca mRNA first accumulates as muscle undergoes terminal differentiation [41], and we show that myosin is reduced and myofibrilogenesis is inefficient in mef2ca mutants. The additional presence of Mef2d, Mef2cb, and Mef2aa in skeletal muscle may contribute to overall Mef2 activity, permitting lower but significant rates of myofibrilogenesis in mef2ca mutants [44]. Up-regulation of Mef2 level by electrical activity will contribute positively to myofibrilogenesis. Loss of Mef2ca alone has less effect upon myofibrilogenesis than loss of activity. It is likely that other translational and/or transcriptional targets of activity, possibly including Mef2d, contribute to this difference.
The Mef2 protein is itself a transcription factor activated by muscle electrical activity via calcineurin and inhibited in cancer-induced muscle wasting [93],[94]. Our discovery of a TORC1-4EBP-Mef2 pathway that regulates muscle mass reveals an additional level of activity-dependent regulation of Mef2. The TORC1-4EBP-Mef2 pathway might be involved in other muscle-wasting conditions that affect TOR activity, such as fasting, ageing, and cachexia. By showing that Mef2c regulates muscle fiber growth regardless of any earlier role in myoblast terminal differentiation, our data support the view of Mef2 as a muscle homeostatic regulator [95],[96]. It will be interesting to determine whether translational regulation of human MEF2C is also important in the nervous system, where MEF2C is associated with autism and synaptic regulation [97].
In the heart, Mef2c is required for cardiomyocyte differentiation and expression persists in differentiated cardiomyocytes [44],[98]. Our finding of activity-dependent accumulation of Mef2 in skeletal muscle suggests a mechanism that could regulate heart muscle cell size and/or function. At later developmental stages, Mef2a becomes a major Mef2 required for normal heart development and prevention of heart attack in mouse [99] and zebrafish [42],[44],[100]. The precise role of Mef2a and other Mef2s in the adult heart and skeletal muscle is unclear, but our data suggest a role in controlling the balance between anabolism and catabolism.
Strikingly, both murine Mef2c and zebrafish mef2ca mRNAs accumulate at muscle fiber ends [41],[101]. We speculate that this mRNA localization may facilitate their translational regulation by activity-dependent signals, including, but not necessarily restricted to, TORC1-4EBP. Such regulation would provide a novel paradigm in regulation of muscle gene expression, whereby changing translation sends a signal back to the nucleus to regulate transcription. In this scenario, the gatekeeper function of eIF4EBP3L could control the sensitivity of localized mRNAs to local signals generated at the muscle fiber end. For example, mechanical force, transmitted from muscle to its attachments at the fiber end, could trigger signals mediating the effects of activity on translation. Such regulation via TORC1 may contribute to the potent hypertrophic effect of high-force stimuli on muscle [102],[103]. Regulated translation of transcription factors could mediate mechanical or other signals elsewhere in biology.
There is great societal importance to determining the mechanisms by which physical activity enhances well-being both in the elderly and throughout life. Our findings show that one effect of activity is to control translation of specific muscle proteins that themselves influence muscle growth. As a major metabolic tissue constituting almost half our body mass, skeletal muscle and its energy balance are increasingly understood to be a significant endocrine regulator of whole body physiology. It will be important to determine whether and at what life stages the pathway we have revealed regulates the response of human muscle to activity.
All work described was performed under licenses conforming to UK Animals (Scientific Procedures) Act 1986.
D. rerio wild-type and mef2catn213 [104], chrndsb13 [39] mutants were maintained on Tübingen background and reared according to [105]. 0.016% MS222 (3-amino benzoic acid ethyl ester, 0.64 mM) was added to embryo medium for 17–24 h. Embryos were injected at 1–2 cell stage with 1–2 ng of MOs against eif4ebp3l; exon1-intron1 junction (BP3L MO1) and start codon (BP3L MO2) or with standard control MO (Gene Tools) or with 100 pg eif4ebp3l RNA made using Ambion Megascript kit from CloneJET plasmid containing full-length zebrafish eif4ebp3l. The plasmids pCS2-ZF-HSP70-5A3L-IRES-GFP or empty vector were injected at 1 cell stage. Heat shock of 39°C for 2 h was applied at 30 hpf and 48 hpf, and myofiber width was measured at 54 hpf. Cell Trace BODIPY (#C34556 Invitrogen) was added during the second heat shock, washed, and imaged live 4 h later. Dechorionated 2 dpf embryos were treated for 6–8 h with 1–5 µM rapamycin or 100 nM MG132 (Cayman #13697).
eIf4ebp3l cDNA was PCR cloned into CloneJET pJET1.2 (Thermo Scientific) using primers listed in Table S1. The 5A3L mutant was generated by substituting T33,T42,T46,S61,T66 to alanine using QuikChange II Site-Directed Mutagenesis Kit #200523 (Agilent Technologies) with primers listed in Table S1. The 5A3L mutant eIf4ebp3l was cloned into XbaI/SalI digested hsp70-4-MCS-IRES-mGFP6, kindly donated by S. Gerety and D. Wilkinson [41].
Immunodetection in whole embryos was as previously described [41]. Confocal stacks were collected on a Zeiss LSMExciter confocal microscope and signal intensity was calculated using Volocity software. All embryos are shown in lateral view, anterior to left, dorsal to top. Myofibrillar bundle width was measured on slow MyHC in 10 fibers in somite 17 by assessing transverse dorsoventral width of three well-bundled regions in anterior, middle, and posterior of each fiber from maximal intensity projections. To ensure fair fiber size comparison in mosaic analyses, the myofibrils of GFP+ fibers were compared with the four immediately surrounding GFP− slow fibers. For Westerns, protein extracts were made from 20–50 embryos using SDS loading buffer. Samples were incubated for 5 min at 100°C, sonicated, and stored at −20°C. Proteins from ∼5 embryo-equivalents were separated by PAGE, blotted to PVDF membrane which was blocked with 5% low fat milk for 1 h RT, incubated overnight with primary antibody and for 1 h RT with secondary antibody, developed with ECL Super Signal® #34080 (Thermo Scientific), and assessed on ImageJ software. Primary antibodies used were: pS6 #5364, S6 #2317, p4EBPT37 #2855, 4EBPT46 #4923 (Cell Signalling), Vinculin #V9131, Actin #A2066 (Sigma), Mef2c #55913 (Anaspec), Mef2 C-21 #SC313, MyoD C-20 #SC302 (Santa Cruz), SQSTM1 #ab56416, Tropomyosin #ab7786 (Abcam), general MyHC (A4.1025) [106], slow MyHC (F59) [31], MyLC #F310 (DSHB), and Myosin binding protein C (MyBPC, a kind gift of E. Ehler).
In situ RNA hybridization was performed as previously described [41]. Probes for eif4ebp1, eif4ebp2, eif4ebp3, and ei4ebp3l were prepared by T7/T3 RNA polymerase with primers listed in Table S1. Other probes were as described: mef2ca, mef2d [42], mef2aa, mef2cb [44], mef2ab (NCBI AL918279), and mef2b (Exelixis 3573227, NCBI JX292158). qPCR analysis used SYBR green MESA Blue (RT-SY2X-03 WOUB, Eurogentec) with specific primers listed in Table S1 and were performed in triplicate to ensure reproducibility. cDNA was generated from 20 embryos using Invitrogen SuperScript III.
We adjusted the protocol of [107] to zebrafish. Briefly, dechorionated embryos were incubated with 400 µg/ml cycloheximide for 10 min, snap frozen in liquid nitrogen, pulverized by pestle and mortar, and stored at −80°C. Powder was resuspended in 500 µl lysis buffer/70 embryos and the sample pipetted 10 times. Nuclei were removed by centrifugation (12,000× g, 10 s, 4°C). Supernatant was supplemented with 250 µl of extraction buffer and centrifuged (12,000× g, 5 min, 4°C) to remove mitochondria and membranous debris. 75 µl (1/10th) of the supernatant was stored at −80°C for total RNA. Remaining supernatant was layered onto a 10 ml linear 10%–50% sucrose gradient and ultracentrifuged for 2 h in Beckman SW40Ti at 38,000 rpm at 4°C, brake off. Samples were collected passing a UV 254 nm detector (Gilson UV/VIS-155) to detect the polysome profile and fractionated, generally into three ∼4 ml fractions. Samples were incubated with 200 µl/ml proteinase K, 1% SDS, 10 mM EDTA pH 8, for 30 min at 37°C. RNA was recovered by 1∶1 phenol∶chloroform extraction followed by ethanol precipitation with glycogen. RNA pellet was washed overnight with 2 M LiCl, and resuspended in 500 µl RNAse free water, treated with DNAse I for 30 min, ethanol precipitated, and 200 ng RNA was used per reverse transcription reaction.
|
10.1371/journal.pgen.1000629 | Establishment of Cohesion at the Pericentromere by the Ctf19 Kinetochore Subcomplex and the Replication Fork-Associated Factor, Csm3 | The cohesin complex holds sister chromatids together from the time of their duplication in S phase until their separation during mitosis. Although cohesin is found along the length of chromosomes, it is most abundant at the centromere and surrounding region, the pericentromere. We show here that the budding yeast Ctf19 kinetochore subcomplex and the replication fork-associated factor, Csm3, are both important mediators of pericentromeric cohesion, but they act through distinct mechanisms. We show that components of the Ctf19 complex direct the increased association of cohesin with the pericentromere. In contrast, Csm3 is dispensable for cohesin enrichment in the pericentromere but is essential in ensuring its functionality in holding sister centromeres together. Consistently, cells lacking Csm3 show additive cohesion defects in combination with mutants in the Ctf19 complex. Furthermore, delaying DNA replication rescues the cohesion defect observed in cells lacking Ctf19 complex components, but not Csm3. We propose that the Ctf19 complex ensures additional loading of cohesin at centromeres prior to passage of the replication fork, thereby ensuring its incorporation into functional linkages through a process requiring Csm3.
| During cell division, chromosomes must be distributed accurately to daughter cells to protect against aneuploidy, a state in which cells have too few or too many chromosomes, and which is associated with diseases such as cancer and birth defects. This process begins with the generation of an exact copy of each chromosome and the establishment of tight linkages that hold the newly duplicated sister chromosomes together. These linkages, generated by the cohesin complex, are essential to resist the pulling forces of the spindle, which will pull the sister chromosomes apart into the two new daughter cells. Here we examine the establishment of cohesin at the pericentromere, the region surrounding the site of spindle attachment and where its forces are strongest. We find that a dedicated pathway promotes cohesin establishment in this region through a two-step mechanism. In the first step, a group of proteins, known as the Ctf19 complex, promote the association of cohesin with this region. In the second step, the Csm3 protein, which is coupled to the DNA replication machinery, ensures its conversion into functional linkages. We demonstrate the importance of this process for accurate chromosome segregation during cell division.
| The accurate transmission of the eukaryotic genome requires that the two copies of each chromosome are held together following their synthesis in S phase until the time of their segregation in mitosis. This chromatid cohesion, which facilitates the biorientation of sister chromatids on the mitotic spindle, is achieved by a multi-subunit complex known as cohesin (reviewed in [1]). Once proper bipolar attachment is achieved, a protease, separase, cleaves the Scc1/Mcd1 subunit of cohesin and destroys the linkages, thereby triggering the movement of sister chromatids to opposite poles [1].
The establishment of cohesion between sister chromatids is coupled to their replication in S phase. In budding yeast, cohesin is loaded onto chromosomes before DNA replication in a manner dependent on, and at the binding sites of, the cohesin-loading complex Scc2/Scc4 [2],[3]. Subsequently, cohesin is thought to translocate from these sites as a result of passage of the transcriptional apparatus [3]. Transformation of this loaded cohesin into functional linkages between sister chromatids requires a second step that takes place during S phase. Scc1 produced after S phase associates with chromosomes but fails to generate cohesion [4]. Several proteins that travel with the replication fork function in this second step. Among the replication fork-associated factors that have been implicated in cohesion function is the Tof1-Csm3 complex which is required for replication fork pausing at replication barriers [5]–[11]. These observations suggest a tight coupling between cohesion establishment and passage of the replication fork.
Analysis of cohesin distribution along both mitotic and meiotic chromosomes of budding yeast has revealed that the highest levels of cohesin are found in a ∼50 kb domain surrounding the ∼120 bp centromere sequence, called the pericentromere [3], [12]–[14]. In fission yeast, pericentromeric heterochromatin is important for cohesin association with the pericentromere during mitosis and meiosis [15]–[18]. Budding yeast lacks pericentromeric heterochromatin but a functional kinetochore is required for pericentromeric cohesin enrichment [13],[19]. The high levels of cohesin in the pericentromere raised a paradox because sister centromeres are known to separate under tension over an approximately 20 kb domain without cohesin cleavage, a phenomenon known as “centromere breathing” [20]–[22]. A possible solution to the paradox was provided by the observation that exertion of tension across sister kinetochores causes a dramatic reduction in the amount of cohesin associated with this 20 kb domain where breathing is observed [19],[23]. However, this region of the pericentromere is thought to form an intramolecular loop, so cohesin may not link sisters in this region [24]. Therefore, although pericentromeric cohesin plays a role in chromosome segregation that cannot be fulfilled by arm cohesins [19], it exhibits unique behavior, the functional relevance of which is unclear.
A clear role for pericentromeric cohesin has, however, emerged in meiosis. In contrast to mitosis, where cohesins are removed along the length of chromosomes simultaneously, meiosis requires that loss of arm cohesins and pericentromeric cohesins are temporally separated (reviewed in [25]). The Shugoshin protein (Sgo1) localizes to the ∼50 kb pericentromeric domain of enriched cohesin and protects it from separase-dependent cleavage [14], [26]–[29]. This pericentromeric cohesin is essential for the accurate segregation of sister chromatids during meiosis II. Two kinetochore proteins, Iml3 and Chl4, are important for preventing non-disjunction of sister chromatids during meiosis II [27],[30] and for proper localization of Sgo1, a role shared with cohesin [14]. Iml3 and Chl4 are components of a conserved kinetochore subcomplex, known as the Ctf19 complex (see [31] for review) (Figure 1A). Ctf19 is required for the enhancement of cohesin in the pericentromere [19], but whether this function is shared with other members of the complex was not known. Similarly, Csm3, and its fission yeast counterpart, Swi3, have been implicated in pericentromeric cohesion through screens in both meiosis and mitosis [9]–[11],[32].
Here we further investigate the role of Csm3 and the Ctf19 complex in the establishment of cohesion at the pericentromere. We show that two of the more peripheral Ctf19 complex proteins, Iml3 and Chl4, direct increased cohesin association with the pericentromeric region during both mitosis and meiosis. In the absence of Iml3 and Chl4, cohesin binding at the pericentromere is reduced and this has important implications for sister chromatid cohesion and chromosome segregation. Conversely, we find that the replication-fork associated protein, Csm3, is not required for cohesin association with the pericentromere, but plays a role in cohesion establishment at the pericentromere that is non-overlapping with Iml3 and Chl4. Our results indicate that Iml3 and Chl4 ensure the association of cohesin with the pericentromeric region, to facilitate its subsequent incorporation into functional linkages through a replication-coupled step, which requires Csm3.
Iml3 and Chl4 are peripheral components of the Ctf19 kinetochore subcomplex [33],[34] (Figure 1A). We asked if Iml3 and Chl4 share the reported role of Ctf19 [19] in promoting the enhancement of cohesin in the pericentromere. We examined the localization of the cohesin subunit, Scc1, at 11 sites on chromosome IV (Figure 1B) by chromatin immunoprecipitation (ChIP) followed by real-time quantitative PCR (qPCR). For the purposes of this paper “centromere” refers to the ∼120 bp region where the kinetochore assembles, “pericentromere” we define as the ∼50 kb enriched cohesin domain and “inner pericentromere” describes the domain in which “breathing” and cohesin removal is observed under tension. We first treated cells with the microtubule-depolymerizing drugs, nocodazole and benomyl, to generate a metaphase arrest where sister kinetochores are not under tension. This has been previously shown to result in high levels of cohesin association with centromeric and pericentromeric regions in wild type cells [19],[23]. In wild type cells (Figure 1C, SCC1-6HA, black bars), Scc1 levels were high at cohesin-enriched chromosomal arm sites (A1, A3), but low at the cohesin-poor A2 site. Scc1 association with all pericentromeric sites was increased in wild type cells over the no tag control, although we observed particularly high enrichment at the centromere (C1, C2) and pericentromeric sites P2–4 (Figure 1C). In iml3Δ and chl4Δ mutants, Scc1 levels at chromosomal arm sites (A1–3) were comparable to that seen in wild type cells (Figure 1C), indicating that Iml3 and Chl4 do not affect cohesin association with chromosome arms. In contrast, Scc1 was reduced at centromeric (C1, C2) and pericentromeric (P2–4) sites and the enrichment over chromosome arms was lost. Similar results were obtained for chromosome V (Figure S1). Therefore, the role of Ctf19 in promoting assembly of a cohesin-rich pericentromeric domain is shared with other components of the complex.
To investigate the effect of tension across sister kinetochores on cohesin association with the pericentromere in cells lacking IML3 or CHL4, we arrested cells by depletion of CDC20 (under control of the methionine-repressible promoter, MET-CDC20; Figure 1D). This generates a metaphase arrest where sister kinetochores attach to opposite poles and are under tension, resulting in a reduction in cohesin levels within the ∼20 kb inner pericentromere [19],[23]. Again Scc1 association on chromosomal arm sites was unchanged in iml3Δ or chl4Δ mutants (Figure 1D, A1–3). In contrast to the situation in nocodazole (Figure 1C), however, we observed little enrichment of Scc1 at the centromere (C1, C2) or at the majority of pericentromeric sites (P3–6), even in wild type cells (note the scales in Figure 1C and 1D are different). Outside this region, cohesin is retained upon the exertion of tension in wild type cells (site P2). However, we observed a clear reduction in Scc1 association with this site in iml3Δ and chl4Δ mutants in the presence of tension. This indicates that Iml3 and Chl4 may also restrict the region in the pericentromere where cohesin shows tension-dependent association.
We investigated the importance of pericentromeric Scc1 for sister chromatid cohesion during mitosis by inducing a metaphase arrest in strains with tetO arrays integrated 2.4 kb from the centromere of chromosome IV and which express TetR-GFP (+2.4CEN4-GFP; [21]). Previous reports have indicated that microtubule forces cause sister centromeres to separate specifically at inner pericentromeric regions during metaphase, so that 2 GFP dots are observed in a fraction of wild type cells [20]–[22]. Wild type, iml3Δ and chl4Δ cells were released from a G1 block into a metaphase arrest by depletion of CDC20, either in the presence of nocodazole (in DMSO) to depolymerize microtubules or DMSO alone. Depletion of CDC20 caused a metaphase arrest in all three strains, as judged by spindle morphology in the strains treated with DMSO (Figure S2A) or nuclear morphology in the nocodazole-treated strains (not shown). We scored separation of sister chromatids at CEN4 by monitoring the numbers of cells in which 2 GFP dots were visible (Figure 2A). Consistent with previous reports [20]–[22], we observed separation of sister +2.4CEN4-GFP in ∼30% of wild type cells in a manner dependent on microtubules (Figure 2A). This fraction was increased to ∼60% in iml3Δ and chl4Δ mutants and was also microtubule-dependent (Figure 2A). This indicates that Iml3 and Chl4 are required to generate cohesion at the pericentromere that is effective in resisting microtubule pulling forces.
Ctf19 complex components are required for accurate chromosome transmission during mitosis [33], [35]–[40]. We used a colony-sectoring assay [41] to measure chromosome loss in mutants lacking Ctf19 complex components or Csm3. All mutants tested, with the exception of nkp1Δ and nkp2Δ, exhibited chromosome loss rates that were elevated compared to the wild type, however the frequencies were not equivalent (Table S1). The highest chromosome loss rates were observed in mcm21Δ and ctf19Δ mutants, with other mutants exhibiting lower, but non-equivalent rates of loss.
We compared cohesion defects in Ctf19 complex mutants by examining the separation of +2.4CEN4-GFP foci in cells arrested in metaphase by CDC20 depletion. Figure 2C shows that 120 mins after release from G1, GFP foci separation was similarly increased over wild type in all single and double mutant combinations tested (see also representative time courses Figure S3), despite their different chromosome loss rates. Previous analyses have revealed that the association of Iml3, Chl4, Ctf3, Mcm16 and Mcm22 with kinetochores depends on Ctf19, but not vice-versa [33],[39]. We found that Chl4 also fails to associate with CEN4 in a mcm21Δ mutant and its levels are reduced in both mcm16Δ and iml3Δ mutants (Figure S4). These findings suggest that Ctf19 and Mcm21 could mediate cohesion establishment through recruitment of the more peripheral components of the sub-complex.
We next asked if the distance between the separated centromeres is increased in iml3Δ, chl4Δ and ctf3Δ mutants arrested in metaphase by depletion of CDC20. An SPC42-dtTomato construct was used to visualize spindle pole bodies (SPBs) together with +2.4CEN4-GFP (Figure 2B) and directly monitor the accumulation of cells in metaphase (Figure S2B) alongside sister centromere separation (Figure S2C). Measurement of distances between separated +2.4CEN4-GFP foci from the 60–120 min time points revealed that the majority of separated sister centromeres were greater than 0 but less than 1 µm apart in wild type cells, consistent with previously published observations (Figure 2D) [21]. However, in iml3Δ, chl4Δ and ctf3Δ mutants, most of the separated sister centromeres were greater than 1 but less than 4 µm apart. Perhaps as a consequence of this increased centromere stretching, we found that the distance between SPBs was also increased in iml3Δ, chl4Δ and ctf3Δ mutants compared to wild type cells (Figure 2E). We conclude that both the frequency and distance of sister centromere separation is increased at metaphase in the absence of Ctf19 complex components.
To ask if the cohesion defects observed in Ctf19 complex mutants at CEN4 also apply to other centromeres we used GFP labels in which tetO arrays are integrated 4.5 kb from CEN6 (Figure 2F) or 1.4 kb from CEN5 (Figure 2G; [22]). Figure 2F shows that CEN6-GFP behaves in a manner reminiscent of +2.4CEN4-GFP in both wild type and iml3Δ cells. That is, sister centromeres separated at an appreciable frequency in wild type cells (∼50%), but with a greater occurrence (∼75%) in the iml3Δ mutant cells. CEN5-GFP, however, exhibited a different behavior, because sister centromere separation was nearly complete (∼80%) in wild type metaphase-arrested cells, perhaps as a result of the close proximity of this label to the centromere [2] and no increase in CEN5-GFP separation was observed in iml3Δ, chl4Δ or ctf3Δ mutant cells (Figure 2G). Because the Ctf19 complex is required for normal levels of the cohesin Scc1 at the pericentromere of chromosome V during mitosis (Figure S1) we reasoned that cohesion defects might be apparent at loci more distant from CEN5. Indeed, two intermediate sites in the pericentromere, at ∼12.6 kb and ∼17.8 kb to the left of CEN5 (-12.6CEN5-GFP and -17.8CEN5-GFP, respectively), which split rarely in wild type cells [21], showed an increased separation in the absence of IML3 or CHL4 (Figure 2H and 2I). However, cohesion at a site just outside the pericentromere (URA3-GFP; 38.4 kb from CEN5 [42]) was virtually unaffected in iml3Δ and chl4Δ mutants (Figure 2J) and splitting of GFP signals at a telomeric locus on chromosome V (TEL5-GFP) was not observed in any of the strains (Figure 2K). This is consistent with our ChIP results and indicates that arm cohesion is intact in iml3Δ and chl4Δ mutants. Our results suggest that the Ctf19 complex directs cohesion establishment at the pericentromere of budding yeast chromosomes and that this is essential to restrict the region where sister chromatids separate when their kinetochores are under tension.
Pericentromeric cohesion is particularly important during meiosis. Following the segregation of homologous chromosomes during meiosis I, sister chromatids lack all arm cohesion and are solely reliant on pericentromeric cohesion. This property of meiosis exposes weak general cohesion defects [43] and also led us to identify IML3 and CHL4 as important mediators of pericentromeric cohesion [27]. To assess the importance of other Ctf19 complex components for meiosis II, we examined the fate of a GFP label at URA3 (38.4 kb from CEN5) on either one or both copies of chromosome V (heterozygous or homozygous GFP dots, Figure 3A and 3B, respectively) after meiosis. Examination of the fate of heterozygous dots after meiosis II revealed that sister GFP labels failed to segregate into different nuclei in approximately 20% of tetranucleate cells of all Ctf19 complex mutants tested, with the exception of nkp1Δ and nkp2Δ (Figure 3C). Meiosis II non-disjunction in Ctf19 complex mutants with homozygous GFP dots was also evident from the high frequency of tetranucleate cells with a GFP dot in just 3 of the 4 spores, which was comparable also in iml3Δ chl4Δ or chl4Δ ctf3Δ double mutant combinations (Figure 3D). Although most of the mutants with homozygous GFP dots showed only the expected modest number of tetranucleate cells with GFP label in just two spores, ctf19Δ and mcm21Δ mutants were a notable exception (Figure 3D). Examination of GFP dot segregation in binucleate cells (Figure S5) revealed that Ctf19 and Mcm21, but not other Ctf19 complex components, are also required for accurate segregation during meiosis I. These findings are consistent with our conclusion from the mitotic chromosome loss data (Table S1) that Ctf19 and Mcm21 promote accurate chromosome segregation also in ways other than cohesion establishment.
To address whether meiosis II non-disjunction in Ctf19 complex mutants could be due to a failure to enrich cohesin in the pericentromere, as in mitosis, we examined localization of the meiosis-specific counterpart of Scc1, Rec8, during meiosis. Wild type, iml3Δ and chl4Δ cells were arrested in metaphase I (by depletion of CDC20 [44]), and the localization of the meiotic cohesin, Rec8, was examined at 10 sites on chromosome IV (Figure 3E) by ChIP followed by qPCR. Because sister kinetochores are uniquely mono-oriented during meiosis I (Figure 3F), this leads to a situation where they are not under tension, enabling the retention of high levels of cohesin in the centromere and pericentromere. Consistent with previous reports, Rec8 localization closely resembled that of the mitotic cohesin, Scc1, in wild type cells [12],[14]. As in mitosis, deletion of IML3 or CHL4 caused a reduction in Rec8 levels at the centromere (C1) and pericentromeric sites (P1–6), but not at chromosomal arm sites (A1–3). We also observed a similar requirement for Ctf19 in localizing Iml3 and Chl4 during meiosis, and a partial dependence on Mcm16 and Mcm22 (Figure S6). Furthermore, we found that Iml3 and Chl4 are localized specifically at the core centromere, despite their ability to influence cohesin in the surrounding pericentromere (Figure S6). We conclude that the Ctf19 complex plays similar roles in pericentromeric cohesion establishment during mitosis and meiosis.
To understand more about cohesion establishment in the pericentromere, we extended our analysis to Csm3, which is required for proper chromosome segregation in meiosis [10] and has been implicated in cohesion establishment during mitosis [9],[11]. Furthermore, Swi3, the fission yeast homolog of Csm3, was shown to have a role in cohesion establishment [32]. Examination of +2.4CEN4-GFP separation in csm3Δ cells arrested in metaphase by CDC20 depletion confirmed that csm3Δ mutants exhibit a cohesion defect that is more severe than that of Ctf19 complex mutants (Figures 2C and 4A). Consistently, csm3Δ cells showed a more pronounced cohesion defect than chl4Δ cells at loci 12.6 and 17.8 kb away from CEN5 (-12.6CEN5-GFP and -17.8CEN5-GFP; Figure 4B and 4C, respectively). Furthermore, in contrast to chl4Δ mutants and in agreement with previous reports [9],[11], cells lacking CSM3 exhibited a small defect in cohesion at the URA3-GFP locus (Figure 4E). Examination of URA3-GFP separation in csm3Δ iml3Δ double mutants, revealed an additive effect (Figure 4E and 4F). Similarly, we observed additive effects on meiotic chromosome segregation in csm3Δ iml3Δ mutants (Figure S7). These results indicate that Csm3 plays a role in cohesion establishment that is non-overlapping with the Ctf19 complex and which may not be restricted to the pericentromere.
To ask if the cohesion defects of csm3Δ mutants could, like that of Ctf19 complex mutants, be due to a failure to recruit proper levels of cohesin to chromosomes, we examined the localization of Scc1 in nocodazole-arrested csm3Δ mutants (Figure 4G and 4H). However, we found that Scc1 levels in the csm3Δ mutant were comparable to wild type at all sites tested, unlike a Ctf19 complex mutant control, ctf3Δ (Figure 4H). This indicates that Csm3 does not promote cohesion establishment by directing cohesin association with chromosomes.
Csm3 is part of a complex with Tof1 that travels with the replication fork and is required for fork stalling at DNA-protein barriers, including centromeres [5]–[8]. Csm3 is thought to achieve fork stalling by counteracting the Rrm3 helicase and deletion of RRM3 restores fork stalling in csm3Δ mutants at replication termination sites [45]. If Csm3 facilitates cohesion establishment by promoting fork stalling, we reasoned that the cohesion defect might also be rescued in csm3Δ rrm3Δ mutants. Indeed, deletion of RRM3 in csm3Δ mutants reduced the frequency of +2.4CEN4-GFP separation at metaphase (Figure 4I and 4J), although not to wild type levels. Interestingly, the rrm3Δ single mutant exhibited a lower frequency of GFP dot separation that wild type (Figure 4I and 4J). Since rrm3Δ mutants exhibit increased fork stalling, particularly at hard to replicate sites such as centromeres [46], these results are consistent with the idea that fork stalling facilitates cohesion establishment.
Next we investigated the role of the Ctf19 complex in pericentromeric cohesin recruitment in more detail. Our results show that while cohesin is reduced at the pericentromere of iml3 Δ and chl4Δ cells, a low level remains during a metaphase arrest in the absence of tension (Figure 1C). Because cohesin can associate with chromosomes after S phase, but is not normally cohesive [4], we analyzed the kinetics of cohesin loading in chl4Δ cells in a synchronized mitotic cell cycle. Wild type and chl4Δ cells carrying MET-CDC20 and SCC1-6HA were released from a G1 arrest into methionine and nocodazole at 18°C and samples were taken at 15 min intervals for analysis of Scc1 localization at 4 sites on chromosome IV (Figure 5A). FACS analysis indicated that both strains entered into S phase after 60 min under these conditions (Figure S8). In wild type cells, cohesin associated with the two centromeric (C1 and C2) sites after only 30 min (Figure 5B and 5C). Cohesin began to be recruited to the chromosome arm (A1) and pericentromeric (P2) sites a little later, at 45 min, in wild type cells (Figure 5D and 5E). The early loading of cohesin at centromeres is consistent with a specialized mechanism for cohesin recruitment operating in this region. In chl4Δ cells, cohesin loaded onto the chromosome arm (A1) site with the same kinetics as in wild type cells (Figure 5E), but was barely detectable at the two centromeric (C1 and C2) sites, even after 75 min (Figure 5B and 5C). Cohesin recruitment was also both delayed and inefficient at the pericentromeric (P2) site in chl4Δ mutants (Figure 5D). These findings suggest that Chl4 is required for the loading of cohesin at the centromere. Furthermore, the almost complete absence of cohesin at centromeres as chl4Δ cells enter S phase explains why cohesion establishment fails, given that centromeres replicate early in S phase [47]. It follows that the low, but appreciable, levels of pericentromeric cohesin observed in nocodazole-arrested chl4Δ mutants (Figure 1C) must arrive after passage of the replication fork and therefore may not be functional in holding sister chromatids together.
Previous observations have indicated that a pathway that promotes cohesin loading at the centromere might be particularly active in mitotic cells, although the significance of this is unknown [19],[23]. Furthermore, the removal of microtubule forces in metaphase-arrested cells allows cohesin to re-associate with the pericentromere apparently independently of the Scc2/4 complex that is normally responsible for the loading of cohesin at sites throughout the genome [23]. To ask if the Ctf19 complex directs the accumulation of cohesin at the pericentromere during metaphase, we arrested wild type and chl4Δ cells in metaphase with sister kinetochores under tension by depleting CDC20 and examined the association of Scc1 with sites on chromosome IV (Figure 5F). As before, the presence of tension caused cohesin levels to be very low at the centromere in both strains, similarly high at a chromosomal arm site (A1) and enriched only in the presence of Chl4 at a pericentromeric site (P2) just outside the tension-sensitive region (Figure 5G). We then treated these cells with nocodazole to depolymerize microtubules, thereby eliminating tension, and re-examined cohesin association with these sites (Figure 5H). As expected, cohesin levels did not increase greatly over background upon the eradication of tension at the chromosomal arm site (A1) or the pericentromeric site (P2) in either wild type or chl4Δ mutant cells (Figure 5H). However, cohesin levels increased substantially at both centromeric sites (C1 and C2) in wild type cells after microtubule depolymerization (Figure 5H). In contrast, cohesin accumulation at the centromere during metaphase was more modest in chl4Δ mutants and did not show enrichment over the chromosomal arm site (Figure 5H).
To distinguish between de novo loading of cohesin at the pericentromere and redistribution of cohesin from other chromosomal sites, we restricted our analysis to cohesin that was produced only in metaphase. We used MET-CDC20-containing strains carrying SCC1-3HA under control of the galactose promoter (pGAL-SCC1-3HA) but with untagged endogenous cohesin. Cells were released from G1 in the presence of raffinose (to prevent pGAL induction) and methionine (to deplete CDC20) to allow cells to accumulate in metaphase with endogenous (untagged) cohesin and sister kinetochores under tension. After 90 min, anti-HA ChIP was performed to determine the background signal in the absence of SCC1-3HA expression (Figure 5I). Subsequently, nocodazole and galactose were added to depolymerize microtubules and induce expression of SCC1-3HA. After a further 60 min, association of Scc1-3HA at 3 chromosomal sites was analyzed. Figure 5J shows that, in wild type cells, only a low level of metaphase-produced cohesin associated with a chromosomal arm site, whereas levels at pericentromeric and centromeric sites were elevated well above background. This is consistent with a previous report [23] showing that de novo produced cohesin preferentially associates with the pericentromere during metaphase. In the chl4Δ mutant, however, metaphase-produced cohesin at the C2 and P2 sites was hardly elevated above its levels at the A1 arm site (Figure 5J). These results indicate that Chl4 contributes to the mechanism that drives cohesin loading at the pericentromere during metaphase, although the purpose and functionality of this cohesin remains unknown.
Next we investigated the relationship between Ctf19 complex-dependent recruitment of cohesin to the pericentromere and DNA replication. Our findings indicate that cohesin is barely above background at the centromere prior to the initiation of DNA replication in chl4Δ mutants (Figure 5B and 5C), however, during a metaphase arrest without microtubules a low, but appreciable, level of cohesin accumulates in this region (Figure 1C). If cohesion establishment at the pericentromere occurs through the normal replication-coupled mechanism in S phase, then the late arrival of cohesin to this region in chl4Δ cells would exclude its conversion into functional inter-sister linkages. We reasoned that delaying replication fork passage in Ctf19 mutants might allow this late-arriving cohesin to be converted into functional cohesion. The experimental set-up is shown in Figure 6A. We released wild type, chl4Δ, iml3Δ, ctf3Δ and csm3Δ cells carrying MET-CDC20 from a G1 arrest into medium containing hydroxyurea (HU) to inhibit DNA replication and containing 8 mM methionine to deplete CDC20. After 90 min in the presence of HU, the drug was washed out and cells were allowed to accumulate in metaphase (as a result of CDC20 depletion). FACS analysis confirmed an approximately 90 min delay in bulk DNA replication in all strains (Figure S9). As expected, strains activated the DNA damage checkpoint kinase, Rad53 in HU [48], as evidenced by its hyperphosphorylation (Figure S10A), but the slowest migrating forms disappeared rapidly upon release in all strains, although in the case of the csm3Δ there was a slight delay (Figure S10A). Nevertheless, none of the mutants showed sensitivity to HU (Figure S10B), indicating that they are able to respond to and recover from the damage caused. In control cells that were not treated with HU, we observed a cohesion defect at CEN4, in chl4Δ, iml3Δ, ctf3Δ and csm3Δ mutants, as before (Figure 6B). However, remarkably, delaying cells in S phase by HU treatment reduced the separation of sister +2.4CEN4-GFP foci to near wild type levels in iml3Δ, chl4Δ and ctf3Δ mutants, but not csm3Δ mutants (Figure 6C). In the HU-treated cells, all mutants separated SPBs with approximately the same kinetics (note that SPB duplication is not prevented by HU treatment) (Figure 6D). Therefore, a HU-induced S phase delay rescues the pericentromeric cohesion defect of iml3Δ and chl4Δ and ctf3Δ mutants, but not csm3Δ mutants.
We examined the association of Scc1 with chromosomes in wild type, chl4Δ and csm3Δ cells arrested in metaphase in the absence of microtubules, following a HU-induced S phase delay. Consistent with our finding that Csm3 does not influence cohesin levels in the pericentromere at metaphase in the absence of an S phase delay (Figure 4H), we observed similar levels of cohesin in csm3Δ and wild type cells (Figure 6G) after a HU delay. In the absence of CHL4, after the HU delay, cohesin levels were, however, reduced in the centromere and pericentromere to a similar extent as without a HU delay (compare Figure 6G with Figure 1C). We also did not observe increased levels of pericentromeric cohesin in iml3Δ or chl4Δ mutants arrested in metaphase with sister kinetochores under tension after HU treatment (Figure 6H). These results rule out the possibility that a HU delay rescues cohesion at the pericentromeres of Ctf19 complex mutants by reversing the cohesin recruitment defect in these mutants. Rather, delaying S phase in Ctf19 complex mutants appears to overcome the failure of the late-arriving cohesin to become cohesive.
As a further test of the ability of a replication delay to allow pericentromeric cohesion establishment in Ctf19 complex mutants, we examined the separation of +2.4CEN4-GFP dots in metaphase-arrested cells lacking the S phase cyclins, CLB5 and CLB6, which are known to delay origin firing [49],[50] (Figure 7A–7D). FACS analysis revealed a 45 min delay in completion of DNA replication in cells lacking the S phase cyclins compared to wild type, iml3Δ or chl4Δ mutants (Figure S11). This delay in DNA replication was increased to 60 min and 75 min in clb5Δ clb6Δ mutants lacking IML3 or CHL4, respectively (Figure S11), the reasons for which are unclear. We also observed a corresponding delay in spindle pole body separation in the clb5Δ clb6Δ background (Figure 7B). As before, chl4Δ and iml3Δ mutants separated +2.4CEN4-GFP foci rapidly and with a greater frequency than wild type (Figure 7A). As expected, due to the SPB separation delay, +2.4CEN4-GFP separation was delayed in cells lacking the S phase cyclins, however, deletion of CHL4 or IML3 did not increase the rate or frequency of separation (Figure 7A). This finding has to be interpreted with caution because clb5Δ clb6Δ iml3Δ and clb5Δ clb6Δ chl4Δ cells show an increased delay in spindle pole body separation (Figure 7B) as compared to clb5Δ clb6Δ cells, which could account for slower +2.4CEN4-GFP separation in these cells. However, taking the difference in the timing of SPB duplication into consideration, +2.4CEN4-GFP separation in chl4Δ or iml3Δ cells was reproducibly lower in the absence of the S phase cyclins, even though clb5Δ clb6Δ cells showed more separation than wild type, using these criteria (Figure 7C and 7D, compare time points with asterisks). Therefore, removal of the S phase cyclins appears to at least partially rescue the cohesion defect of iml3Δ and chl4Δ cells.
How does a replication delay allow the inefficiently loaded cohesin in Ctf19 complex mutants to become cohesive? As hypothesized above, it could restore the normal order of cohesin loading and replication fork passage, to allow cohesion generation through the usual replication-fork coupled mechanism in S phase. However, an alternative mechanism of global cohesion establishment has been described in G2 cells subjected to DNA damage [51]–[53]. Because HU is a DNA damaging agent and clb5Δ clb6Δ mutants have also been found to activate the DNA damage checkpoint [54], cohesion could be restored in iml3Δ and chl4Δ mutants through DNA damage-dependent cohesion establishment. To distinguish between the replication delay and DNA damage hypotheses we took advantage of the fact that some DNA replication origins are known to replicate early in S phase, whereas others replicate late. We used centromeric (CEN5) plasmids that carry a single origin that fires early (p306.10; ARS306) or late (p12; ARS1412) [55],[56]. Both plasmids also carry the URA3 selectable marker, which allows growth on medium lacking uracil, enabling us to examine their ability to propagate in wild type, chl4Δ, iml3Δ and csm3Δ cells. Strains containing each of the plasmids were allowed to grow for 2.5 h in non-selective medium before plating onto selective (lacking uracil) and non-selective (rich) medium in parallel (Figure 7E). The ratio of colonies that are able to grow on medium lacking uracil compared to rich medium are given in Figure 7F. Although both plasmids were well maintained in wild type cells, they showed reduced stability in all three mutants. In the case of csm3Δ cells, we observed an equivalent decrease in stability of both the early and late-replicating plasmid, suggesting that replicating timing does not affect their ability to propagate in this mutant. Replication timing did however influence plasmid stability in chl4Δ and iml3Δ cells because the early-replicating plasmid displayed a greatly reduced stability compared to the late-replicating plasmid in these cells. Although we have not directly tested whether the decreased stability of the p306.10 plasmid in iml3Δ and chl4Δ mutants is the result of defective cohesion, these findings are consistent with the idea that pericentromeric cohesion establishment fails in cells lacking Ctf19 complex components because cohesin is not recruited prior to the passage of the replication fork.
In budding yeast, an approximately 50 kb region around the centromere, the pericentromere, is enriched for cohesin binding. Enhanced binding of cohesin in the pericentromere requires the ∼120 bp centromere sequence and is dependent on a functional kinetochore [13],[57]. Previous findings identified Ctf19 as being an important component of the kinetochore in this process [19]. We and others [58] have extended these findings and shown that the Ctf19 kinetochore subcomplex is an important mediator of cohesion establishment at the pericentromere. Since all single and double Ctf19 complex mutants show similar cohesion defects, the critical role of the Ctf19 and Mcm21 components in pericentromeric cohesion establishment may be to recruit the more peripheral components. However, our findings also indicate that Ctf19 complex components may have additional functions in chromosome segregation. In particular, Ctf19 and Mcm21 are unique in their requirement for chromosome segregation also during meiosis I. Whether this reflects a greater general requirement for these proteins in chromosome segregation (Table S1) or a specific role in, for example, generating linkages between homologs is unknown. However, Ctf19 components do not appear to play a major role in mediating kinetochore-microtubule attachments, since sister centromeres are pulled further apart in these cells. In addition, visualization of all kinetochores in Ctf19 complex mutants using Mtw1-GFP revealed no obvious attachment defect (JF and AM, unpublished observations). We note that human Mcm21 and Ctf19 proteins promote bipolar spindle assembly and chromosome congression, respectively [59] and that yeast Chl4 has been found to play a role in converting naked centromere DNA into an established centromere that is heritable [38]. Whether these functions of the Ctf19 complex are mediated solely through a role in cohesion establishment will be important to investigate in the future.
How does the Ctf19 complex promote cohesin enrichment within the pericentromere? The simplest explanation is that the Ctf19 complex promotes the loading of cohesin at the centromere, which then spreads bidirectionally into the surrounding pericentromere. This could explain how the Ctf19 complex, which is localized within the ∼125 bp core centromere, is able to influence cohesin association over a much greater region. Support for this model comes from our finding that cohesin is detected at centromeric sites earlier in the cell cycle than at a pericentromeric site (Figure 5B–5D) and the observation that the Scc2 cohesin loader component is reduced at the centromere, but not other pericentromeric sites in the absence of MCM21 [58]. However, Chl4, at least, may be able to influence cohesin association with the pericentromere independently of Scc2/4. We found that Chl4 is required for high levels of newly-synthesized cohesin to associate with the pericentromere during metaphase after the eradication of tension, a process that appears to be independent of Scc2 [23]. Although the functional relevance of metaphase-loaded cohesin remains unknown, these data provide evidence that the Ctf19 complex promotes cohesin association with the pericentromere throughout the cell cycle.
We compared the contribution of Iml3-Chl4 to pericentromeric cohesion establishment with that of Csm3. The finding that cells lacking IML3 and CSM3 show synthetic cohesion defects suggests that these genes promote cohesion establishment through different pathways. Indeed, we observed no defect in cohesin association with the pericentromere in the absence of Csm3. This suggests that Csm3 facilitates cohesion establishment in a step after cohesin loading. Interestingly, Csm3 and its binding partner, Tof1, travel with the replication fork, and the Tof1-Csm3 complex is required for stable fork pausing at protein-DNA barriers including those at centromeres [5], [6], [45], [60]–[63]. Like Csm3, Tof1 has been implicated in cohesion [9],[11]. Tof1-Csm3 is conserved, being homologous to the Timeless-Tipin complex in humans and the Swi1-Swi3 complex in fission yeast [8],[9],[64]. Furthermore, the fission yeast Swi1-Swi3 complex is also required for the stabilization of stalled forks and efficient cohesion generation [8],[32]. Taken together, these observations indicate that the stalling and stabilization of replication forks may be the critical function of Csm3 in cohesion establishment. Indeed, we found that deletion of the helicase RRM3, which restores fork stalling to csm3Δ mutants [45], partially rescued the cohesion defect. Perhaps Tof1-Csm3 is required to maintain the association of essential cohesion establishment factors with the replisome upon encountering protein-DNA barriers. Such a hypothesis could explain why cells lacking CSM3 show chromosome segregation defects in meiosis II, but not meiosis I. Perhaps there is a greater requirement for Csm3 in generating cohesion at protein-DNA barriers, such as centromeres, where kinetochore binding could impede replication fork progression. Alternatively, this could simply reflect the susceptibility of meiosis II to general cohesion defects due to the absence of arm cohesion [43].
Our analysis of cohesin association in a synchronized cell cycle showed that cohesin loads at centromeres early and in a Chl4-dependent manner. Given that centromeres replicate early in S phase [47], one important function of kinetochore-driven cohesin loading could be to ensure that cohesin is in place at centromeric regions prior to passage of the replication fork, thereby ensuring its incorporation into functional cohesion. In support of this model, allowing more time for cohesin loading by delaying DNA replication, either by HU treatment or deletion of the S phase cyclins, in iml3Δ and chl4Δ mutants improved pericentromeric cohesion. We cannot at present rule out that DNA damage-induced cohesion [51]–[53] is responsible for the restoration of pericentromeric cohesion in these cells. However, evidence that replication timing in the absence of DNA damage does play a role in centromere function in iml3Δ and chl4Δ cells, came from our observation that a CEN5-containing late-replicating plasmid is more stable than a similar early-replicating plasmid in iml3Δ and chl4Δ cells. Preliminary observations indicate that early and late replicating, but otherwise identical, CEN4 plasmids behave in a similar manner (JF and AM, unpublished observations). It is interesting to note that while all centromeres replicate early in S phase, among centromeres, CEN4 and CEN6 replicate relatively late in their normal chromosomal context, whereas CEN5 replicates early [47]. In wild type cells we saw the highest frequency of GFP dot separation with a probe close to CEN5. Whether this is due to the increased distance of the CEN4 and CEN6 probes from the centromere (2.4 kb and 4.5 kb, respectively, compared to 1.4 kb for CEN5), or an effect of their replication timing needs further investigation.
A crucial role for pericentromeric cohesion has emerged during meiosis, where its protection during meiosis I ensures the fidelity of chromosome segregation during meiosis II [25]. However, it is becoming increasingly clear that pericentromeric cohesion plays specialized roles in mitosis too [19]. Being proximal to the site of microtubule attachment, it is required to resist spindle forces (Figure 2). Recent findings in fission and budding yeast have also demonstrated its importance in specifying the geometry of kinetochore-microtubule attachment [58],[65]. However, generation of cohesion at the centromere represents a considerable challenge, since kinetochore assembly generates a protein-DNA barrier to the replication fork whose passage is intimately linked to cohesion establishment. Furthermore, centromeres are known to replicate early during S phase [47], so that a mechanism must exist to ensure that cohesin is in place prior to passage of the replication fork. We propose that a two-step dedicated pathway of cohesion establishment involving the Ctf19 complex and Csm3 overcome these challenges to ensure the reinforcement of cohesion at the pericentromere. In the first step, the Ctf19 complex, through its Iml3 subunit, directs cohesin loading at the centromere, which subsequently spreads throughout the pericentromeric region. This dedicated pathway of cohesin loading is essential to ensure that cohesin is in place prior to passage of the replication fork early in S phase. Csm3 is crucial in this second step to ensure the integrity of the replisome and its association with cohesion establishment factors as it traverses the protein-DNA barrier that surrounds the centromere. The resultant enhanced cohesion in the vicinity of the centromere safeguards the fidelity of chromosome segregation. The Ctf19 complex and Csm3 are highly conserved. Whether such a mechanism of kinetochore-driven cohesion establishment operates in organisms where pericentromeric heterochromatin is known to attract cohesin [15]–[18],[66] is an important question for the future.
All strains used for meiotic experiments were derivatives of SK1 and all strains used for mitotic experiments were derivatives of W303. The iml3Δ::KanMX6 and chl4Δ::kanMX6 deletions were described in [27]. The pCLB2-3HA-CDC20 fusion was described in [44]. The REC8-3HA and SCC1-6HA tags were described in [67] and [57], respectively. pGAL-SCC1-3HA was described in [4]. The CHL4-6HA and IML3-6HA tags were created using a PCR-directed method [68]. Other deletions were generated using genomic DNA from the yeast deletion collection [69] as template for PCR, followed by transformation. The pMET3-CDC20 construct was constructed using a PCR based method as described by [70]. To create the CEN6-GFP strain, a 1 kb region to the right of CEN6 was cloned into pRS306tetO112 [42] and integrated into a strain carrying tetR-GFP, resulting in the generation of a GFP label 4.5 kb to the right of CEN6. The Spc42-tdTomato construct was described in [71]. Other GFP chromosome labels were described previously [21],[22],[42],[72]. Strains are listed in Table S2. Plasmids p12 and p306.10 were described in [56] and [55], respectively.
Growth conditions for individual experiments are given in the figure legends. All cultures were grown at room temperature unless otherwise stated. Meiosis was performed at 30°C as described in [73]. Drugs and alpha factor were removed by filtration, washing with at least 5 volumes of medium lacking sugar. Methionine was used at 8 mM and readded to 4 mM every hour. To depolymerise microtubules a mixture of benomyl (30 µg/ml) and nocodazole (15 µg/ml) were used and nocodazole (7.5 µg/ml) was readded every hour. Hydroxyurea was used at 10 mg/ml in liquid medium and 100 mM in plates.
ChIP was carried out as described by [14]. qPCR was performed in a 20 µl SYBR Green reaction using a BioRad iCycler machine. To calculate ChIP enrichment/input, ΔCT was calculated according to: ΔCT = (CT(ChIP)−(CT(Input)−logE (Input dilution factor))) where E represents the specific primer efficiency value. Enrichment/input value was obtained from the following formula: E∧−ΔCT. The average of 3 independent experiments for which qPCR was performed in triplicate are shown with error bars indicating standard deviation. We designed primers corresponding to cohesin-enriched (A1, A3) or cohesin-poor (A2) sites on chromosome arms as well as a centromeric site (C1) and several pericentromeric sites (P1–6), based on published ChIP data after microarray hybridization for the mitotic cohesin, Scc1 [23]. We were not able to analyze site C2 during meiosis since this primer set did not give a product in the SK1 strain background, presumably due to divergence from the published S. cerevisiae sequence. Sequences of primers are available on request. For the ChIP shown in Figure S1, primer sets were as described in [14] and Image J software was used to quantify ethidium bromide-stained gels.
To measure chromosome loss or plasmid loss, strains carrying the SUP11 artificial chromosome, or carrying the early or late replicating plasmids (p306.10 and p12), were grown overnight in minimal media lacking uracil (SD/-ura) before transferring to rich media (YPD) for 2.5–3 h. For the chromosome loss assay, approximately 2000–5000 cells were then plated out on YPD media. Loss of the artificial chromosome causes colonies to appear red. The percentage of sectored colonies that were at least half red, indicating loss in the first division after plating, were scored. To avoid amplification of earlier loss events, entirely red colonies were excluded from the analysis. For the plasmid loss assay, equal culture volumes totaling 1000–2000 cells were plated onto each of YPDA and SD/-URA and the percentage of colonies able to grow on SD/-URA, indicating retention of the plasmid, was calculated.
Fixing cells for visualization of GFP-labeled chromosomes was performed as described by [67]. Indirect immunofluorescence methods were as previously described [74]. Microscopy was performed on a Zeiss Axioplan 2 microscope and for measurements of inter-CEN and –SPB distance, images were grabbed using a Hamamatsu ORCA-ER camera and analyzed using Zeiss Axiovision software. In the cohesion assay for mitosis, GFP dots were scored as separated if two dots were clearly visible in the same cell. A total of 200 cells were scored and all cells in the field were scored. To confirm metaphase arrest in each experiment, spindle pole body separation was analyzed either using Spc42-dtTomato or after anti-tubulin immunofluorescence or parallel samples. The leakiness of the MET-CDC20 construct allowed spindle elongation in a few cells (usually<5% and never more than 10%) only at later timepoints (120 min) and was comparable in all strains in a given experiment. To score GFP dots in meiosis, cells were co-stained with DAPI to visualize nuclear morphology and 200 binucleate or tetranucleate cells were counted from the 5 and 8 h timepoints, respectively.
For flow cytometry, cells were fixed in 70% ethanol at 4°C over night. The cells were then treated with RNase over night, then digested with pepsin (Sigma). The cells were finally treated with propidium iodide (Sigma) to stain the DNA. Samples were briefly sonicated before analysis. FACS analysis was performed according to the manufacturer's instructions (BD FACS Calibur). FACS data were analyzed using CellQuest software.
Samples for immunoblot analysis were prepared after TCA fixation as described by [75]. A goat Rad53 antibody (yC-19; Santa Cruz Biotechnology, inc.) was used at a dilution of 1∶1000.
|
10.1371/journal.ppat.1000320 | Modulation of Neutrophil Function by a Secreted Mucinase of Escherichia coli O157∶H7 | Escherichia coli O157∶H7 is a human enteric pathogen that causes hemorrhagic colitis which can progress to hemolytic uremic syndrome, a severe kidney disease with immune involvement. During infection, E. coli O157∶H7 secretes StcE, a metalloprotease that promotes the formation of attaching and effacing lesions and inhibits the complement cascade via cleavage of mucin-type glycoproteins. We found that StcE cleaved the mucin-like, immune cell-restricted glycoproteins CD43 and CD45 on the neutrophil surface and altered neutrophil function. Treatment of human neutrophils with StcE led to increased respiratory burst production and increased cell adhesion. StcE-treated neutrophils exhibited an elongated morphology with defective rear detachment and impaired migration, suggesting that removal of the anti-adhesive capability of CD43 by StcE impairs rear release. Use of zebrafish embryos to model neutrophil migration revealed that StcE induced neutrophil retention in the fin after tissue wounding, suggesting that StcE modulates neutrophil-mediated inflammation in vivo. Neutrophils are crucial innate effectors of the antibacterial immune response and can contribute to severe complications caused by infection with E. coli O157∶H7. Our data suggest that the StcE mucinase can play an immunomodulatory role by directly altering neutrophil function during infection. StcE may contribute to inflammation and tissue destruction by mediating inappropriate neutrophil adhesion and activation.
| Enterohemorrhagic Escherichia coli (EHEC) poses a significant threat to the U.S. food supply, causing foodborne gastrointestinal disease in humans that can progress to hemolytic uremic syndrome (HUS), a potentially fatal kidney disease. Research suggests that EHEC strains are growing more virulent, resulting in a higher incidence of hospitalization and development of HUS from recent produce-associated outbreaks. Although immune dysregulation is a feature of HUS disease, the specific mechanisms contributing to altered immune function require investigation. Furthermore, the contribution of the immune response to early intestinal disease is not known. StcE is a secreted protease of EHEC that is expressed during infection and may contribute to virulence via cleavage of mucin-like glycoproteins. In this study, we define mucinase activity toward glycoproteins on the surface of human neutrophils and find that StcE alters neutrophil activity by interacting with these proteins. StcE affected crucial neutrophil functions including oxidative burst production and migration. The effects of StcE were both cleavage-dependent and cleavage-independent, providing insight into a novel mechanism for mediating neutrophil function via mucin interactions. Our study reports an immune-modulating role for a potential EHEC virulence factor and provides a possible explanation for altered neutrophil phenotypes observed during E. coli O157∶H7-induced disease.
| Enterohemorrhagic Escherichia coli (EHEC) of serogroup O157∶H7 is an emerging human diarrheal pathogen associated with numerous food-borne outbreaks in the U.S. Infection of the colon by EHEC causes mild diarrhea that proceeds to bloody colitis and can be acquired following ingestion of fewer than 100 organisms [1]. In 15% of childhood cases, EHEC gastroenteritis progresses to the more serious hemolytic uremic syndrome (HUS), characterized by red blood cell fragmentation, low platelet count, and acute renal failure. HUS can cause severe kidney damage and have outcomes ranging from full recovery to death [2]. EHEC virulence factors include the locus of enterocyte effacement, which confers the ability to form attaching and effacing lesions, and the phage-encoded Shiga toxin, which causes termination of protein synthesis in the microvascular endothelium leading to cell death and tissue destruction [3]. Efforts to study the pathogenesis of EHEC are complicated by the lack of a suitable animal model that fully recapitulates human disease. Although models of EHEC-induced diarrhea in rabbits or injection of purified components leading to HUS-like symptoms in mice and baboons have been described, no model exists that follows the natural progression from EHEC infection to development of HUS [4]–[6].
The immune response is involved in the development of HUS, but it is less clear how the interaction between bacteria and the host immune system influences early EHEC disease progression. Colonic damage can include hemorrhage and edema within the lamina propria with focal necrosis and neutrophil influx [7], but leukocyte infiltration into the intestinal lumen occurs in only ∼50% of EHEC cases and is rarely severe [2]. Patients who progress to HUS demonstrate clear indicators of an inflammatory response with neutrophil involvement, and increased circulating blood leukocytes are correlated with development of disease [8],[9]. Increased levels of interleukin-8 (IL-8) and complexed elastase are found in the blood [10]–[12]. Children with HUS demonstrate infiltration of monocytes and neutrophils into the kidney glomeruli [13],[14]. The onset of HUS occurs 5–7 days after initiation of diarrhea, and it has been suggested that inappropriate immune cell activation in the gut could lead to renal pathology and explain the lag time in the development of disease [15].
The majority of EHEC isolates in the United States carry the 92 kb pO157 virulence plasmid. Carriage of the plasmid is associated with increased incidence of hemorrhagic colitis, HUS, and colonization of the bovine recto-anal junction mucosa [2],[16],[17]. pO157 encodes StcE (Secreted protease of C1-esterase inhibitor), a type II-secreted, 95 kDa zinc-dependent glycoprotease that is produced during EHEC infection [18],[19]. StcE recognizes O-glycan-induced protein conformations in order to cleave the protein backbone of mucin-type glycoproteins [L. Walters, unpublished, [18],[20]].
Mucins are large glycoproteins that coat numerous surfaces in the body and play important roles in cell-cell interactions within the immune system. CD43 and CD45 are large mucin-type glycoproteins expressed exclusively and abundantly on the surface of nearly all hematopoeitic cells including neutrophils [21],[22]. CD45 is a protein tyrosine phosphatase that can exist as several isoforms which vary primarily in the length of the terminal O-glycosylated portion of the extracellular domain. Dephosphorylation of Src family kinases by CD45 regulates the signaling threshold in T cells [21]. Little is known about CD45 function on neutrophils, but it may modulate chemotaxis and the oxidative burst [23],[24]. CD43 possesses a large extracellular domain that contains 60–80% of its total molecular mass in sialylated O-glycans. Extensive glycosylation causes the protein to assume a rod-like conformation that protrudes ∼45 nm from the cell surface [25]. The combination of steric hindrance, negative charge, and relative abundance on the cell surface provides an anti-adhesive force [26], and CD43-deficient leukocytes demonstrate increased adhesion in vitro and in vivo [22], [27]–[29]. The intracellular domain of CD43 interacts with cytoskeletal linker proteins, allowing neutrophils and T cells to cluster CD43 at the rear of the cell, or “uropod”, during adhesion and migration. This removes anti-adhesive force from the leading edge to promote adhesion and/or migration, while providing a useful anti-adhesive force at the uropod [30]–[34].
In this study we report the interaction of the StcE protease with CD43 and CD45 on the neutrophil surface. StcE altered neutrophil function via both cleavage-dependent and cleavage-independent effects. Proteolytic activity of StcE led to increased neutrophil oxidative burst production, while binding of StcE was sufficient to increase neutrophil adhesion, leading to impaired migratory capacity. We propose that the interaction between StcE and CD43 prevents the sialoglycoprotein from providing crucial anti-adhesive force, preventing uropod-mediated detachment leading to impaired migration. Oxidative burst production and migration defects leading to increased neutrophil retention could contribute to tissue destruction and inflammation, as well as bacterial evasion of the immune response. Interaction with neutrophil surface mucins by StcE might therefore represent a novel way of dysregulating the immune response during EHEC disease.
StcE binds to and aggregates cells of the Jurkat leukemic T cell line and binds to the undifferentiated HL-60 promyelocytic cell line [18]. To determine if StcE interacted with neutrophils, primary human neutrophils isolated from whole blood and neutrophil-like differentiated HL-60 cells (dHL-60s) were treated with StcE or a proteolytically inactive StcE point mutant, E435D [18],[35]. Binding was detected by flow cytometry using polyclonal anti-StcE antibody. Both wild-type StcE and E435D bound to neutrophils (Figure 1A) and dHL-60s (data not shown). We next sought to identify the surface determinant responsible for this interaction. The cell-bound mucin-like glycoproteins CD43 (leukosialin) and CD45 (leukocyte common antigen) were previously identified as StcE substrates on Jurkat T cells (unpublished data), and we investigated whether these proteins served as ligands for StcE on the neutrophil surface. As we were unable to immunoprecipitate StcE using available antibodies, we performed direct precipitations with StcE fused to a chitin binding domain (StcE-CBD) and bound to chitin beads (CB). StcE precipitated both CD43 and CD45 from dHL-60 lysates (Figure 1B). Staining of precipitation reactions for total glycoprotein or total protein did not reveal other significant binding partners, suggesting that CD43 and CD45 are the major relevant binding partners on the neutrophil surface (Figure S1).
The pulldown assays described above were performed in the presence of EDTA, which inhibits the metalloprotease activity of StcE. We next investigated whether CD43 and CD45 could be proteolytically cleaved by StcE. Intact dHL-60s were treated with purified, endotoxin-free StcE or proteolytically inactive E435D and analyzed by immunoblottting. Recognition by the anti-CD43 L10 monoclonal antibody (mAb), which recognizes an extracellular epitope near the N-terminus, completely disappeared upon treatment with StcE but not E435D (Figure 2A). An antibody to the intracellular domain of CD43, sc-7052, recognized a StcE-cleaved product of ∼28 kDa (Figure 2B), consistent with a fragment containing the intracellular and transmembrane domains. No L10-reactive cleavage product of any size appeared in the supernatant (Figure 2A), suggesting that the extracellular domain of CD43 was degraded while leaving the intracellular and transmembrane domains intact. Selective cleavage of mucin-like domains is consistent with StcE activity toward other known substrates [18]. Flow cytometric analysis confirmed that presence of the CD43 extracellular epitope recognized by L10 was reduced on the surface of primary human neutrophils following treatment with StcE but not E435D (Figure 2C).
We examined potential cleavage of CD45 using two antibodies. The HI30 mAb recognizes an epitope in the CD45 extracellular domain that is found in all isoforms. Treatment of dHL-60s with StcE resulted in a small size shift from 180 kDa to ∼165 kDa in the band recognized by HI30 (Figure 2D). This suggested that StcE cleaved within the very N-terminal O-glycosylated extracellular portion of CD45. CD45RO is the major isoform present on the neutrophil surface. We next examined cleavage of CD45 using the UCHL1 mAb, which is specific for CD45RO, suggesting that it binds somewhere in the O-glycosylated extracellular region that defines this isoform. Flow cytometric analysis of neutrophils demonstrated no change in staining of total CD45 with HI30, (Figure 2E), while staining of CD45RO with UCHL1 was reduced following StcE treatment (Figure 2F). This suggested that StcE cleaves within the membrane-distal, O-glycosylated extracellular portion of CD45 but does not degrade the protein further.
CD43 and CD45 are found uniquely on the surface of immune cells and may be important for neutrophil function [21],[22]. To determine if StcE modulates neutrophil function, we examined the effect of StcE treatment on the oxidative burst using a flow-based assay for hydrogen peroxide and superoxide production. Treatment with StcE increased the respiratory burst in the absence of other stimuli (Figure 3A and 3B), while neutrophils treated with proteolytically inactive E435D did not differ significantly from control samples. These findings suggest that StcE modulates neutrophil oxidative function through its protease activity. Both CD43 and CD45 have been suggested to contribute to the neutrophil oxidative burst, and we did not identify the specific mediator of this effect [24],[33].
Because CD43-deficient leukocytes are more adherent, we examined whether removal of the CD43 extracellular domain by StcE altered the ability of human neutrophils to adhere to the extracellular matrix. Treatment with StcE significantly increased neutrophil adhesion on a fibrinogen (Fbg)-coated surface in a dose-dependent manner (Figure 3C). Surprisingly, proteolytic activity of StcE was not required, as binding by E435D induced a similar phenotype. Treatment with either protein resulted in a 2–3 fold increase in adhesion, similar to stimulation with fMLP, which served as a positive control.
As proteolytic activity of StcE was not essential to increase adhesion, we sought to confirm that binding activity of E435D was responsible for this effect. Purified protein was heat-inactivated, which eliminates both binding and proteolysis by StcE and controls for the presence of heat-stable contaminants in the protein preparation. Heat-inactived StcE (hiStcE) had no effect on neutrophil adhesion. E435D is proteolytically inactive but retains substrate binding activity, unlike hiStcE, indicating that binding is the minimal function necessary to induce neutrophil adhesion. To account for the possibility that StcE and E435D increased adhesion by serving as a bridge between CD43 or CD45 and the extracellular matrix, binding of StcE to Fbg was tested by ELISA. StcE alone did not bind appreciably to Fbg (Figure S2), suggesting that effects were specific to interaction of StcE and E435D with the neutrophil surface.
CD43 is localized to the uropod during neutrophil adhesion and migration, removing its anti-adhesive force from the front of the cell [30]. The observation that both StcE and E435D induced neutrophil adhesion led us to hypothesize that protein binding to CD43 could interfere with its anti-adhesive function even in the absence of cleavage. This hypothesis is consistent with reports that antibodies to CD43 can induce clustering of the protein at the uropod and increase cell adhesion [30],[32]. We therefore examined how treatment with StcE and E435D affected CD43 localization in adherent neutrophils via confocal immunofluorescence microscopy. Although the anti-StcE antibody exhibited some background staining of vehicle-treated cells, we observed specific and diffuse surface staining of bound StcE (Figure 4A), consistent with flow cytometry data. StcE-treated neutrophils demonstrated reduced membrane staining for the CD43 extracellular domain with L10, consistent with results of immunoblotting and flow cytometry. No change in localization of the CD43 intracellular domain as detected by sc-7052 was observed, confirming that it remained intact and membrane-associated (Figure 4B). In contrast, treatment of neutrophils with E435D caused readily observable relocalization of CD43. Both extracellular and intracellular staining revealed clustering of CD43, and E435D staining was co-localized with the CD43 extracellular domain (Figure 4). Immunofluorescence staining for total CD45 revealed no change in localization induced by StcE or E435D (data not shown).
These data suggest that E435D promotes neutrophil adhesion by clustering CD43 to the uropod and removing its anti-adhesive force from the rest of the cell membrane. To further confirm that E435D bound to native CD43 on the neutrophil surface, we used flow cytometry to investigate the ability of E435D to compete with the L10 mAb for binding to the extracellular domain. Increasing concentrations of E435D led to decreased L10 staining (Figure S3), indicating that E435D bound to CD43 and blocked antibody accessibility. Reduction of L10 binding by StcE was evident at a much lower protein concentration, demonstrating that cleavage by StcE was more efficient than blocking with E435D in preventing L10 antibody binding. Together these findings suggest that while StcE causes loss of CD43 from the neutrophil surface, E435D clusters CD43 in the uropod, leading to removal of anti-adhesive force from the cell membrane and providing an alternative mechanism by which binding alone can induce neutrophil adhesion.
Neutrophil adhesion regulates the development of cell polarity and is required for cells to become migration-competent, but too much adhesion can interfere with migration [36]. We examined how the StcE-induced increase in adhesion affected migratory capabilities of human neutrophils using transwell assays. In the absence of chemoattractant, StcE treatment had no effect on migration across transwell inserts (Figure 5A). In the presence of fMLP as a chemoattractant in the lower chamber, treatment with StcE or E435D caused a significant, 1.7-fold reduction in migration across the filters. Consistent with the results of adhesion experiments, binding but not proteolytic activity of StcE was required to inhibit neutrophil migration, and heat-inactivated protein had no effect.
Circulating leukocytes that detect chemotactic signals first adhere to the vasculature and then transmigrate across the endothelium in order to reach effector sites. To verify that results obtained with purified Fbg extended to interactions with the endothelium, migration experiments were conducted using monolayers of primary human lung microvascular endothelial cells (HMVEC-L). Both StcE- and E435D-treated neutrophils demonstrated decreased migration across HMVEC-L toward fMLP (Figure 5B). These findings confirmed that the defect in neutrophil migration results from the action of StcE on the neutrophil surface and is not specific to the migratory barrier.
We further examined the effect of StcE on neutrophil migration using time lapse microscopy. The adhesion and migration experiments described above were conducted with PBS as a vehicle control. Although the experimental media contained 5% serum, the formal possibility remained that increased total protein concentration was responsible for the effect of StcE or E435D addition. We therefore evaluated migration in the presence of an equivalent concentration of human serum albumin (HSA), and cell behavior was identical to vehicle treatment (data not shown and Video S1). Neutrophils were treated with HSA, StcE or E435D on a Fbg-coated surface and non-directional migration was imaged in the absence or presence of interleukin-8 (IL-8). Consistent with adhesion and transmigration assays, StcE-treated and E435D-treated neutrophils demonstrated increased adhesion and were visibly impaired in their migratory capabilities (Figure 6A and Videos S2 and S3). IL-8 treatment caused an increase in random migration of neutrophils (Video S4), and both StcE and E435D reduced migration even in the presence of IL-8 (Figure 6A and Videos S5 and S6) and fMLP (data not shown). Quantitation of neutrophil migration was performed on IL-8 treated samples, as these had comparable numbers of adherent cells for control and experimental conditions. Treatment with StcE or E435D significantly reduced neutrophil migration velocity compared to HSA control (Figure 6B). Although they traveled shorter distances over time, StcE- and E435D-treated neutrophils did not appear deficient in production of forward protrusions. However, cells seemed unable to retract their rearward edge and move forward, suggesting that StcE interfered with neutrophil migration by preventing de-adhesion of the uropod. During migration, StcE-treated neutrophils displayed striking morphological differences, with formation of elongated tails at the uropod (Figure 6B). E435D treatment also resulted in elongated morphology, but the phenotype was not as severe. Analysis of cell length confirmed that unstimulated, StcE-treated neutrophils were significantly longer than control cells, while E435D treatment did not cause a significant difference in cell length (Figure 6C). Both StcE-treated and E435D-treated neutrophils exhibited increased cell length in the presence of IL-8, although the increase was not significant compared to IL-8 stimulated controls (Figure 6C). Together our findings suggest that both StcE and E435D interfere with CD43-based anti-adhesion to alter adhesion and migration, but do so via different mechanisms dependent on binding and cleavage of cell surface mucins (StcE) or binding alone (E435D).
Zebrafish share many features of the mammalian immune system and have recently been utilized as a model to study neutrophil migration and chronic inflammation [37]–[39]. We took advantage of the zebrafish model to examine the effects of StcE on neutrophil-mediated inflammation in vivo. Zebrafish embryos at 3 days post-fertilization (dpf) were wounded in the ventral tail fin in the presence of a bath of StcE protein. At this stage in development, neutrophils are normally located in the caudal hematopoeitic tissue and circulating in the bloodstream. Wounding of the tail fin induces neutrophil recruitment to the wound, and resolution of this response is generally observed after 24 hours [37]. Wounded embryos were fixed after six or 24 hours and localization of myeloperoxidase (mpo), a neutrophil-specific marker, was examined by immunofluorescence microscopy. Differences in neutrophil recruitment were not observed six hours after wounding (data not shown). At 24 hours, treatment with StcE caused visible neutrophil mislocalization (Figure 7A), resembling chronic inflammation recently observed in zebrafish mutants [39]. Neutrophil mislocalization was quantified by counting neutrophils present in the fin, confirming that treatment with StcE caused a significant increase in number of mislocalized neutrophils (Figure 7B). HiStcE had no effect on neutrophil localization. The inflammation-like phenotype observed in zebrafish embryos suggests that StcE may affect neutrophil motility and trafficking in vivo to regulate inflammatory responses.
In this study, we report cleavage of CD43 and CD45 on the human neutrophil surface by StcE, a secreted glycoprotease of E. coli O157∶H7. We found that StcE exerted both cleavage-dependent and cleavage-independent effects on neutrophil migration and activation, suggesting that it may modulate the immune response during infection. We have previously reported specific cleavage of mucin-type O-glycoproteins by StcE [18],[20]. CD43 and CD45 share characteristics of these substrates and were the major StcE ligands on the neutrophil surface. We found that StcE cleaved specifically within the O-glycosylated domains of these proteins. The majority of the CD43 extracellular domain is heavily O-glycosylated, and the resultant negative charge and steric hindrance serve to inhibit non-specific cell-cell interactions. StcE degraded this domain, leading to loss of anti-adhesive function. In contrast, the majority of the CD45 extracellular domain is N-glycosylated, and the N-terminal portion, which varies in length by differential splicing, is O-glycosylated. StcE cleaved within this terminal portion, leaving the majority of the protein intact. It is not known what effect the terminal O-glycosylated region has on function of the intracellular phosphatase domain, but cleavage of this region by StcE could promote inhibitory dimerization.
In order to fight infection, neutrophils must leave the bloodstream to reach effector sites. Cells first adhere to the vascular endothelium and then migrate across the endothelial cell layer and through the tissues by sensing and responding to chemoattractant gradients [40]. Adhesion is required to initiate this process, but excessive adhesion can inhibit migration [36]. Neutrophils counter the anti-adhesive function of CD43 by shedding it from their surface when they become activated [34],[41]. The remaining surface-associated CD43 is redistributed to the uropod at the cell rear [30],[31],[42], where it may provide useful anti-adhesive force. Treatment with StcE led to increased neutrophil adhesion that interfered with random migration as well as chemotaxis across filters and endothelial monolayers. Surprisingly, this effect was cleavage-independent. The proteolytically inactive mutant, E435D, caused similar effects to StcE, suggesting that binding was the minimal function required. Our data suggest that removal of CD43 anti-adhesion is the mechanism by which StcE and E435D interfere with migration, although we cannot rule out a role for CD45. StcE degraded the extracellular domain of CD43, while E435D clustered CD43 at the uropod and blocked antibody binding to the extracellular domain. Cleavage of CD43 by StcE reduces anti-adhesive force at the promigratory leading edge, but also relieves the anti-adhesive force in the uropod that could promote rear detachment. E435D, like crosslinking antibodies, induces CD43 relocalization to the uropod, reducing anti-adhesive force at the leading edge. Binding of E435D may also mask the negative charge of CD43 and interfere with anti-adhesion at the rear, causing a similar outcome to StcE via a slightly different mechanism. Increased adhesion can be induced by crosslinking antibodies to CD43 [30], and it is unclear if this results from simple masking of the protein or because antibody binding transduces a pro-adhesive signal. Our results support the hypothesis that the anti-adhesive function of CD43 at the uropod is an important component of cellular migration.
Although it has been proposed that CD45 may regulate chemotactic signaling in neutrophils [23], only recently has it been suggested that CD45 may be directly important for cell adhesion. Shivtiel and colleagues found that bone marrow mononuclear cells deficient in CD45 were more adherent to fibronectin as a result of increased activation of β1 integrins [43]. It is possible that CD45 signaling may be important for neutrophil adhesion and that interaction with CD45 contributes to StcE-mediated effects on adhesion and migration. It is unknown whether StcE cleavage of the terminal O-glycosylated portion of CD45 will affect its signaling capacity. Investigation of the effect of StcE on CD45 signaling is a potential topic for future study.
The observation of a cleavage-independent function for StcE in neutrophils parallels findings with C1-esterase inhibitor (C1-INH), another StcE substrate. StcE-cleaved C1-INH retains its ability to inhibit the complement cascade, and E435D is equally capable of potentiating C1-INH function. StcE binds to C1-INH and the bacterial surface simultaneously, increasing the local concentration of C1-INH and protecting the cell from complement-mediated lysis [35]. If binding of StcE to its substrates is sufficient, what is the purpose of proteolytic activity? Proteolysis may be more important for some substrates than others. For example, cleavage of intestinal mucins might be required to promote colonization, whereas only binding is necessary to affect activity of C1-INH and CD43. Alternatively, proteolytic activity of StcE might be dispensable for its interaction with all substrates but provide enhanced turnover. StcE displays high affinity and low turnover of C1-INH and MUC7 [44], and it is possible that proteolysis provides a mechanism for it to detach from one substrate molecule in order to bind another. This would facilitate interaction of the same StcE protein molecule with different substrates, allowing it to interact with multiple glycoproteins during infection.
Proteolytic activity was not completely dispensable to modulate neutrophil function, as the activities of E435D and StcE were not identical. Active StcE more potently induced an elongated cell morphology during neutrophil migration than did E435D, suggesting that cleavage was more efficient than binding in preventing rear detachment. This could be explained if binding of E435D to CD43 blocked charge repulsion, but removal of the CD43 extracellular domain by StcE eliminated both charge repulsion and steric hindrance. Furthermore, StcE exerted a cleavage-dependent effect on neutrophil oxidative burst production. The specific substrate mediating these effects was not identified, but CD45 is an attractive candidate because it has previously been reported to modulate production of the oxidative burst [24]. CD43 may also play a role in oxidative burst production, although evidence for this is less clear [33]. Regardless of the exact mechanism, the finding that StcE increased the oxidative burst in a cleavage-dependent manner provides further evidence that it may play an immunomodulatory role during infection.
Neutrophils are the first responders to bacterial infection, and pathogens have strategies to combat this response that include inhibition of chemoattractant receptors and inactivation of C5a and IL-8 [45]. To our knowledge, proteolysis of cell surface mucins by StcE represents a novel mechanism for altering neutrophil function. Respiratory burst production and the ability to migrate are crucial capabilities of neutrophils in fighting infection, and alteration of these functions by StcE may lead to a dysregulated immune response during EHEC infection. Neutrophils isolated from children with HUS are more adherent to the vascular endothelium [46], and StcE could contribute to this phenotype. The effect of StcE on neutrophil migration was both rapid and persistent. Increased adhesion and impaired migration were evident after 30 minutes of StcE treatment. After 210 minutes, fewer StcE-treated neutrophils had migrated across transwell filters, suggesting this was not a transient defect that could be overcome with time. The longevity of the effect further suggests that the interaction between StcE and neutrophils may be physiologically relevant. Whether alteration of neutrophil function leads to pro- or anti-inflammatory outcomes may depend on the site of activity and presence of other stimuli. Neutrophils that remain stuck to the endothelium may be unable to migrate into the intestine in response to infection, and StcE may thus protect the bacteria from clearance by the host immune response. The observed decrease in neutrophil migration across endothelial cell monolayers supports the conclusion that StcE could impair migration out of the vasculature and into the intestine. Inhibition of complement activation at sites of infection by StcE-localized C1-INH would reduce production of the chemotactic C5a fragment, further contributing to a migration defect. Alternatively, neutrophils that are stuck to the endothelium may contribute to inflammation and tissue destruction, as seen in the kidneys during HUS. Zebrafish treated with StcE exhibited neutrophil mislocalization to the tissues that resembled recently described inflammatory mutants, lending credence to this hypothesis. The fact that mislocalization occurred at a late but not early time point after wounding may be explained by accumulation of inflammatory signals from neutrophils that are retained in the tissues, leading to progressive neutrophil infiltration and retention over time. Enhancement of the neutrophil oxidative burst by StcE could further contribute to inflammatory tissue damage in the intestines and during HUS as a result of inappropriate neutrophil retention.
The pO157 plasmid is associated with EHEC disease incidence and severity, suggesting that plasmid-encoded genes might contribute to pathogenesis [2],[3]. The plasmid-encoded StcE protein has a dedicated type II secretion system, is co-regulated with known virulence factors, and is produced in detectable amounts during infection, making it a likely virulence candidate. In a recently described disease model of EHEC in rabbits, mutation of the type II secretion system led to decreased colonization, and the authors conclude that lack of StcE secretion might contribute to this defective colonization [47]. Potential contributions of StcE to virulence including inhibition of complement-mediated lysis and promotion of pedestal formation have been described [18],[20]. Cleavage of CD43 and CD45 by StcE on the neutrophil surface, leading to increased adhesion, defective migration, increased respiratory burst production, and overall dysregulation of the immune response, may provide another mechanism by which StcE enhances the progression of disease caused by enterohemorrhagic E. coli. Moreover, as CD43 and CD45 are expressed on many cells of the immune system, it is unlikely that the immunomodulatory activity of StcE is limited to neutrophils.
For studies with neutrophils from human subjects, informed consent was obtained from healthy donors at the time of blood draw with approval of the University of Wisconsin-Madison Center for Health Sciences Human Subjects committee.
Recombinant StcE and E435D protein were expressed and purified as described [20]. Endotoxin was removed using Endotrap blue columns (Lonza) according to manufacturer's instructions. Samples were evaluated by Lonza Endotoxin Testing Services, and endotoxin levels were routinely <1 EU/mL. Dulbecco's Phosphate-Buffered Saline with Ca2+/Mg2+ (DPBS+/+, Mediatech) of an equivalent volume to soluble protein was used as a vehicle control. Heat-inactivated StcE (hiStcE) was produced by incubation at 65°C for 10 minutes (Grys 2006). Interleukin-8 (IL-8), f-Met-Leu-Phe (fMLP), fibrinogen (Fbg), bovine serum albumin (BSA), and phorbol myristate acetate (PMA) were purchased from Sigma. Human serum albumin (HSA) was from ZLB Bioplasma AG (Berne, Switzerland). Phycoerythrin (PE)-conjugated anti-CD43 clone L10 and FITC-conjugated anti-CD45 clone HI30 were obtained from Caltag (Invitrogen). The anti-CD43 C-terminal domain antibody (sc-7052) was obtained from Santa Cruz Biotechnology. PE-conjugated anti-CD45RO clone UCHL1 was obtained from Ebioscience (San Diego, CA). EM-grade 16% paraformaldehyde (PFA) was from Electron Microscopy Sciences (Hartfield, PA) and 16% formaldehyde was from Polysciences (Warrington, PA).
Peripheral blood neutrophils were purified from human blood using Polymorphprep according to manufacturer's recommendations (Nycomed, Sheldon, UK). HL-60 cells (ATCC) were maintained in Iscove's Modified Dulbecco's Medium (IMDM) according to ATCC guidelines. HL-60 cells were differentiated (dHL-60s) as previously described [48]. Primary human lung microvascular endothelial cells (HMVEC-L) were maintained in EGM-2MV media according to manufacturer's instructions (Lonza). For transmigration assays, cells were seeded on collagen-coated 3 µm pore, 0.33 cm2 polycarbonate transwell inserts (Costar) at a density of 1×105 cells.
For cleavage reactions, 1×106 dHL-60s in IMDM were treated with 1 µg/mL StcE, E435D or vehicle control for 30 min at 37°C, 5% CO2, followed by separation of supernatants and cell pellets. Supernatants were precipitated with 10% trichloracetic acid (TCA) on ice. Samples were separated by SDS-PAGE, transferred to nitrocellulose or PVDF and immunoblotted using standard methods [49]. L10 and HI-30 antibodies were used at 1∶500 and sc-7052 was used at 1∶200. Samples were detected by enhanced chemoluminescence (ECL) using Immobilon HRP substrate (Millipore). For direct precipitation of binding partners, StcE was expressed with an uncleavable intein-chitin binding domain fusion tag (CBD) in vector pTYB11 and purified using the IMPACT system as previously described [20]. Purified StcE-CBD bound to chitin beads (CB) was incubated with 1×107 lysed dHL-60s. CB alone served as a negative control, and samples were processed using standard methods for co-immunoprecipitations [49].
Neutrophils (1×106/mL) were treated with 1 µg/mL StcE or E435D (unless otherwise indicated) or vehicle control in EGM-2MV for 30 min at 37°C, 5% CO2. Cells were blocked in DPBS+/− with 0.05% BSA and 0.01% HSA and incubated with primary antibody per manufacturer instructions. Samples were read using an LSRII flow cytometer (Becton Dickinson). For the flow oxidative burst (OB) assay, neutrophils (2.5×106) were labeled with dihydrorhodamine 123 (DHR) (Molecular Probes) as described [50],[51] and treated with 5 µg/mL StcE or E435D, vehicle control, or 30 ng/mL PMA for 30 min at 37°C, 7.5% CO2. Oxidation by H2O2 and O2− of DHR to rhodamine was measured as fluorescence of live cells in the FITC channel. Data were analyzed using FlowJo (Treestar) and individual values from five independent experiments were combined and analyzed by one-way ANOVA using GraphPad Prism.
Neutrophils (1×105/mL) in EGM-2MV were plated on Fbg-coated (10 µg/mL) glass coverslips for 30 min at 37°C, 5% CO2 in the presence of 20 µg/mL StcE, E435D or vehicle control, and further incubated for 10 min in the presence of 100 nM fMLP. Samples were fixed in 1% PFA in DPBS, post-fixed in 1% formic acid, and permeabilized in 0.1% Triton-X 100 (Sigma). Primary antibodies were used at 1∶200 (anti-StcE), 1∶100 (L10 and HI30), or 1∶25 (sc-7052). Goat anti-rabbit Alexa Fluor 633, goat anti-mouse Alexa Fluor 488, and rabbit anti-goat Alexa Fluor 488 secondary antibodies were used at 1∶200, and samples were mounted in ProLong Gold antifade reagent with DAPI (Molecular Probes). Coverslips were imaged using a 63× oil immersion lens on a Zeiss LSM510 confocal microscope. Data were obtained and analyzed using LSM 5 Image Software (Zeiss).
Non-tissue culture-treated dishes were coated with 10 µg/mL Fbg, and 5×105 neutrophils were plated in the presence of 8.33 µg/mL StcE or E435D or vehicle control in EGM-2MV for 30 min at 37°C, 7.5% CO2. 1.25 nM IL-8 or 100 nM fMLP was included for the last 5 or 10 min as indicated. Dishes were placed in The Box closed system (Life Imaging Services, Reinach Switzerland) at 37°C and imaged on an Olympus IX-70 inverted microscope (Olympus America) using a 20× phase objective. Images were collected using a Coolsnap fx cooled charged-coupled device (CCD) video camera (Photometrics, Huntington Beach, CA) and captured into Metaview v6.2 (Universal Imaging Corp., Downingtown, PA) every 15 s for 10 min. To obtain measurements of cell velocity, cell centroids were tracked for the first 21 frames using MetaMorph v7.0r2. For quantitation of cell length, all cells in two random frames were measured using MetaMorph. Means of at least three independent experiments were combined and analyzed by one-way ANOVA using Prism with Dunnett posttest.
Neutrophils were fluorescently labeled as described [52] and brought to 2×106/mL in EGM-2MV. 50 µL of cell suspension were added to StcE, E435D, hiStcE, or vehicle control serially diluted in 50 µL EGM-2MV in a Fbg-coated 96-well black plate (Greiner, Kremsmuenster, Upper Austria). Cells were allowed to adhere for 40 min at 37°C, 7.5% CO2. Positive control cells were treated with 100 nM fMLP for the final 10 min. Samples were washed 3 times and the fluorescence of remaining adherent cells measured using a Gemini EM microplate spectrofluorometer (Molecular Devices) with excitation/emission at 485/530 nm. A standard curve was included on each plate, and linear regression was performed with Prism to determine number of neutrophils adhered in each well. Relative adhesion was calculated by normalizing the number of adherent cells to the average values for vehicle control in the absence of fMLP. Means of at least three independent experiments performed in duplicate were combined and analyzed by two-way analysis of variance (ANOVA) using Prism with Bonferroni post test.
Neutrophil migration was determined using transwell assays essentially as described (Lokuta 2005). 3 µm transwell filters were coated with 2.5 µg/mL Fbg or a monolayer of HMVEC-L, and 4×105 calcein-AM-labeled neutrophils were placed in the top chamber with 25 µg/mL (filters) or 50 µg/mL (monolayers) StcE, E435D, hiStcE, or vehicle control. EGM-2MV alone or containing 100 nM fMLP was placed in the bottom chamber and samples were incubated at 37°C, 5% CO2 for 210 min. 50 mM EDTA was added to the lower chamber, transwells were removed and the fluorescence of migrated cells quantitated as described for adhesion assays. Numbers of transmigrated cells were normalized to vehicle control with chemoattractant, and means of at least three independent experiments were combined and analyzed by one-way ANOVA using Prism with Bonferroni post test.
Zebrafish were bred and maintained as previously described (Mathias 2006). At 3 days post fertilization (dpf), zebrafish embryos were anesthetized in 3 mL embryo water (E3) containing 0.1 mg/mL tricaine and 25 µg/mL StcE, hiStcE, or vehicle control. Zebrafish were wounded in the dorsal tail fin with the tip of a 25 gauge needle, and the bath replaced with protein treatment in E3 without tricaine. After 24 hours at 28°C, zebrafish were fixed and stained for whole-mount immunofluorescence as previously described (Mathias 2006). Rabbit polyclonal anti-myeloperoxidase antibody and goat anti-rabbit Alexa Fluor 488 (Molecular Probes) were used at 1∶500. Images were acquired with a Nikon SMZ-1500 zoom microscope with epifluorescent illumination using MetaMorph software. For quantitation of inflammation, images of individual fish were compiled and blinded, and the number of neutrophils in the dorsal and ventral fin caudal to the yolk sac were counted. Means of four independent experiments were combined and analyzed by one-way ANOVA using Prism with Dunnett's post test.
Swissprot ID numbers for proteins described in the text are as follows: StcE (O82882); CD43 (P16150), CD45 (P08575).
|
10.1371/journal.pcbi.1005085 | Efficient Reconstruction of Predictive Consensus Metabolic Network Models | Understanding cellular function requires accurate, comprehensive representations of metabolism. Genome-scale, constraint-based metabolic models (GSMs) provide such representations, but their usability is often hampered by inconsistencies at various levels, in particular for concurrent models. COMMGEN, our tool for COnsensus Metabolic Model GENeration, automatically identifies inconsistencies between concurrent models and semi-automatically resolves them, thereby contributing to consolidate knowledge of metabolic function. Tests of COMMGEN for four organisms showed that automatically generated consensus models were predictive and that they substantially increased coherence of knowledge representation. COMMGEN ought to be particularly useful for complex scenarios in which manual curation does not scale, such as for eukaryotic organisms, microbial communities, and host-pathogen interactions.
| Many large-scale mathematical models describe metabolism to understand how microbes and other organisms (including humans) function and interact with each other and with their environment. Making these models is extremely time- and effort-intensive; it requires gathering and combining information from many sources, including the organism’s genome sequence, biological databases, scientific literature, and expert advice. The exact procedure and resources used depend on the model creators’ expertise and research interests, such that independently created models for the same organism are often very different and can hardly be compared. However, each model typically contains unique information that is ‘lost’ when working with a different model. To integrate the available knowledge, we developed a computational tool to build consensus metabolic models. Our tool—COMMGEN- combines independently generated models by matching identical parts and resolving differences between inconsistent parts. We apply our tool to four sets of models of different organisms. In all these sets, COMMGEN identified and resolved hundreds of inconsistencies. COMMGEN can be generally applied to standardize and improve models of metabolism, in particular for complex scenarios, such as those involving microbial communities and host-pathogen interactions.
| Genome-scale constraint-based metabolic models (GSMs) are curated organism-specific knowledge repositories [1]. They integrate many distinct (bio)chemical entities and typically account for thousands of metabolites, reactions and genes. When assuming that metabolism is in a steady state, GSMs also enable metabolic simulations with applications in genome annotation [2,3], analysis of omics data [4–6], phenotype predictions [7–9], organism comparison [9–12], drug discovery [7,13,14], and metabolic engineering [8,15]. GSMs thereby quantitatively reconstruct the internal metabolic and transport wiring of the modeled organism and thus increase our systems level understanding.
Genome-scale metabolic reconstructions consist of metabolites, metabolic reactions (including boundary reactions and a biomass reaction), cellular compartments, and genes [1,16]. The reactions are organized according to the cellular compartments in which they are active. Enzyme-driven (as opposed to spontaneous) reactions are associated with Gene-protein-reaction rules (GPR), which include one or more genes. For multiple genes, the GPR indicates whether alternative isozymes or enzyme complexes catalyze the reaction [17]. A reaction’s equation consists of substrates and products with their corresponding stoichiometries. A reaction’s reversibility describes whether the reaction operates forward, backward, or bi-directionally. The reaction flux bounds specify the reaction’s capacity, that is, the absolute upper and lower bounds of the reaction flux. Transport reactions transfer metabolites between cellular compartments, whereas boundary reactions define nutrient uptake and secretion. The biomass reaction, finally, reflects the molecular composition of a cell or organism and represents cell or organism growth. Together, these entities and their encoding in a GSM aim to represent the current knowledge of the organism’s metabolism.
However, even for well-studied organisms such as Saccharomyces cerevisiae or Bacillus subtilis, many uncertainties remain during GSM construction. These uncertainties are typically manually addressed based on expert knowledge and scientific literature, which involves a laborious iterative process that can take several years, for example, for eukaryotes [1]. The main sources of uncertainties are: (i) incomplete and erroneous information from heterogeneous and potentially contradictory data sources such as insufficiently curated and inconsistent gene annotations [18], alternative naming and spelling variants of metabolites (different namespaces) [18–21], and conflicting reaction reversibilities [2,22]; (ii) subjectivity in interpreting literature sources; (iii) integration of qualitative and quantitative data (e.g., inconsistent growth data); and (iv) incompatible levels of detail between and among (reference) databases; for example, databases may represent metabolic pathways by detailed individual reactions or by a single lumped reaction [18], and they may use varying structural definitions for metabolite classes such as lipids and polymers [21,23].
As a consequence, when several GSMs for the same organism are developed independently, they are complementary and only partially overlapping [24,25]. The extent of variation between models for the same organism can be dramatic. For example, the well-established human and yeast GSMs agree only on 3% [18] and 35% [20] of their reactions, respectively, when ignoring electron, proton, and water imbalances. Differences between GSMs resulting from different modeling frameworks and model authors can even be more substantial than biological differences between organisms [26]. Any GSM-driven analysis, which needs to (somewhat arbitrarily) select one GSM when several are available, thus, only operates on a subset of the available information.
To represent metabolism more comprehensively, and thereby improve our understanding of a target organism, alternative GSMs of a target organism can be integrated into a so-called consensus model of the respective organism, one per organism. Consensus models have an increased scope (by combining unique parts of initial GSMs) and they are more consolidated (by identifying shared parts of initial GSMs that are likely to be reliable). When discrepancies exist between GSMs, these must be carefully examined to select the most appropriate modeling alternative. However, while consensus models have been generated successfully for several (model) organisms such as budding yeast and human, this required extensive manual curation by communities of domain experts [10,20,24,25,27]. To alleviate this bottleneck and render GSMs truly useful for the understanding of cellular function and evolution, community function, and host-pathogen interactions, semi-automatic consensus model generation approaches have been proposed. It has been shown that the combination of complementary GSMs of the same organism reduces existing gaps in individually reconstructed GSMs [28,29]. These approaches focused mainly on reconciling namespaces (a particularly important challenge for matching metabolites) or on curating the underlying databases [18,21]. Thereby, existing methods address only a small subset of the problems in consensus model generation described above. For example, they do not identify and curate cases when two initial GSMs represent the same metabolic process at different levels of granularity [30].
Here, we present COMMGEN, a tool for COnsensus Metabolic Model GENeration that reconciles two or more distinct GSMs of the same organism beyond a common namespace. COMMGEN automatically identifies similarities, dissimilarities, and complements of the metabolic networks based on an extensive classification of problems that typically arise during GSM integration and on novel algorithms to resolve these problem classes. For several model organisms, we show that semi-automatically created consensus GSMs in a standardized namespace [31] are substantially more consolidated than achievable by a common namespace alone, and that they retain or even improve on the initial GSMs’ predictive capabilities. Because the consensus GSMs contain the information from each initial GSM, they comprehensively represent our best understanding of the organisms’ metabolic networks.
Our analyses addressed model building, testing and refinement in a stepwise fashion. We started by identifying the classes of inconsistencies that exist between models for four widely different albeit representative microbes. We subsequently set up the framework for COnsensus Metabolic Model GENeration, and tested it on the four case studies for functionality and predictability.
To systematically resolve inconsistencies between two or more Initial GSMs (IGSMs) to be integrated, we defined three main (coupled) inconsistency categories: metabolites, reactions, and compartments. We explain these categories and the inconsistency classes they contain using examples from four sets of IGSMs that cover gram-positive and gram-negative bacteria as well as yeast (Fig 1a).
COMMGEN is a software tool that is designed to address the above problems in consensus model generation, leading to a semi-automatic reconciliation of two or more GSMs for a given organism. In terms of software architecture, COMMGEN operates on GSMs in SBML format [32], the standard modeling language for systems biology (Fig 2a). The IGSMs are first converted into a common chemical naming system using the MnXRef namespace [31]. Next, COMMGEN combines all reactions of the IGSMs into a Basic Consensus Model (BCM). The BCM is used to identify and reconcile inconsistencies between and within the IGSMs, ultimately yielding a Refined Consensus Model (RCM) in SBML format. Because many inconsistencies are interconnected, it is difficult to identify a consensus between IGSMs, to distinguish between conflicting and complementary model parts, and to resolve all inconsistencies automatically. COMMGEN therefore resolves all unambiguous cases automatically, and it guides the user to decide on the remaining cases. COMMGEN records all changes such that the user can automatically repeat the procedure with minimal effort, including manual alterations of previously made choices.
To identify and address all the different inconsistency classes described above, COMMGEN iteratively applies a set of independent methods (Fig 2b). All methods automatically identify instances of their respective inconsistency classes. Metabolite matching is a core element of model merging. We developed a novel algorithm to identify sets of metabolites that represent the same chemical compound based on their network context, that is, their neighboring metabolites and reactions, thereby addressing the issue of different granularity in IGSMs for metabolites (see Methods for details). Performance tests for P. putida networks revealed very high sensitivity and specificity of the algorithm, even when only a minority of the network is used to infer matching metabolite sets (Fig 2c). Metabolite matching allows COMMGEN subsequently to reconcile the associated reactions: metabolites are merged, through which novel pathways and branching points can be formed, and alternative representations of biochemical reactions become apparent. Specifically, COMMGEN matches sets of reactions in the following categories (see Methods for the respective algorithms): (i) reactions with identical metabolites but different stoichiometries; (ii) nested reactions; (iii) reactions that differ only in redox pairs; (iv) partially overlapping reactions; and (v) lumped reactions. Furthermore, it deals with differences in subcellular compartmentalization by (i) facilitating the removal of transporters; (ii) enabling the removal of entire compartments; (iii) resolving differences in the modeling of boundary reactions; (iv) identifying different transport reactions for the same metabolite across the same membrane; and (v) identifying identical biochemical conversions in different compartments.
COMMGEN’s methods differ in the extent to which identified inconsistencies can be resolved automatically (Fig 2b). For some categories, the user can choose to automatically handle inconsistencies, for example, to deal with differences in reaction directionality. Conditionally automatic refers to inconsistency classes where some instances can be addressed automatically, but others cannot: if two matched reactions differ only in stoichiometric coefficients, COMMGEN can automatically select the elementally balanced reaction, but only when exactly one reaction is balanced. Manual intervention is always possible, and it is required when inconsistencies are too complex and diverse for a well-performing heuristic for automation. Manual curation is also advisable when an erroneous choice may substantially impact model performance. For example, a single incorrect match between two metabolites with different chemical sum formulas can have severe consequences for the correctness of model predictions. Hence, although the COMMGEN method for network-based metabolite matching performs extremely well (Fig 2c), we recommend manual confirmation of predicted matches.
To describe COMMGEN operation in detail and to evaluate the framework’s performance, we focus on consensus model generation for Pseudomonas putida, for which the two GSMs iJP962 [8,10] and iJN746 [33] have been developed independently (Fig 1a). The initial overlap between these two models is surprisingly low: they only have 58% of their genes, 33% of their metabolites and 2% of their reactions in common. Conversion into the MnXRef namespace [31] only increases the common part to 44% for metabolites and 11% for reactions.
To quantitatively determine the occurrences of inconsistencies and their resolution, we classify reactions as consensus reactions (shared between the GSMs) and unique reactions. We further categorize unique reactions according to whether they are unrelated to any inconsistency, related to a single inconsistency, or related to multiple inconsistencies (a reaction may appear in the last category because COMMGEN methods are not mutually exclusive in the inconsistencies they identify). Because the identified inconsistencies ultimately depend on namespace consistency, user-defined settings, and user choices, we quantified the resolution of inconsistencies by automatic processing to remove user bias as much as possible. After creating the BCM from the IGSMs and merging the identical reactions, the fraction of consensus reactions was low (11%) and approximately half of the unique reactions were associated with at least one inconsistency (Fig 3a; S1 Protocol). The inconsistencies exemplified in Fig 1 are, thus, not isolated cases; they merely illustrate the main problems in consensus model generation.
Next, we employed a four-step automatic process to reconcile inconsistencies between the IGSMs and to converge to an automatically generated RCM (Fig 3a). First, COMMGEN increased the namespace consistency through our network context-based metabolite matching method (note that we manually confirmed the proposed matches such that subsequently identified inconsistencies were not overestimated). This increased the overlap to 53% for metabolites and 16% for reactions. In the second step, COMMGEN addressed the difference in cellular compartments in the P. putida GSMs (Fig 1a). In particular, transport reactions from iJP962 that immediately take up metabolites from the extracellular space into the cytoplasm were split such that they match the transport processes from iJN746, and periplasmic instances of the involved metabolites were added. Next, COMMGEN identified and merged sets of reactions with practically (ignoring protons and water) identical net formula. These sets include reactions that have different GPR rules or different reaction directionalities, or that did not have identical net formulas prior to the splitting of transport reactions or the COMMGEN-based metabolite matching. In this step, we processed inconsistent reaction reversibilities using our previously published method to predict reaction directionalities based on metabolite patterns [2], and we processed inconsistent gene associations by combining the GPR rules with a ‘strict’ heuristic (see S2 Protocol). Finally, COMMGEN identified and merged reactions that involve the same metabolites, but differ in stoichiometric coefficients; directionality and GPR inconsistencies were handled as above.
The detailed data shown in Fig 3a emphasize the interdependencies of inconsistencies that may arise in model merging, in particular, that resolving inconsistencies may facilitate subsequent identification of more inconsistencies, resulting in an increased number of identified inconsistent reactions. The four automated steps increased the share of reactions that are consensus reactions originating from both IGSMs from 11% (in the BCM) to 39% (in the RCM), while also substantially reducing the number of reactions associated with inconsistencies (Fig 3a). We evaluated the significance of the metabolite matching step by re-running the process without it, which lead to only 23% consensus reactions (Fig 3b). In addition, we used the automatically generated RCM as the starting point for manual curation guided by COMMGEN methods. This allowed us to reconcile most of the remaining inconsistencies and to obtain a consensus for 50% of the reactions (Fig 3c). In summary, our detailed case study for P. putida therefore provides evidence for the efficiency of the COMMGEN framework, and in particular of its novel methods such as network context-based metabolite matching.
We next asked, to what extent automated consensus model generation preserved or even extended functionality of the IGSMs, initially focusing on the P. putida models. Our automated method involved the probabilistic prediction of reaction directionalities [2] to resolve reaction inconsistencies, instead of simply setting all reactions with conflicting directionalities to reversible, which would tend to overestimate the organism’s metabolic capabilities. It maintained reaction directions in case of consensus between the IGSMs, although the prediction method is agnostic to matches between models; it constrained directions in many cases when such constraints existed in only one IGSM (Fig 3d). The benefits of this approach are best exemplified with a concrete example (Fig 4a). The P. putida BCM contains a small set of reactions that together allow for non-physiological CO2 fixation. This incorrect CO2 fixation cycle was automatically removed when inconsistent directionalities of a reaction present in both IGSMs were processed, thereby preventing a major error in the RCM. Note that direction prediction also identified a reaction assigned with a direction that is not consistent with the remainder of the network (see also Fig 1i), namely a directed lumped reaction common to both IGSMs, and a bidirectional non-lumped reaction set present in only one model. Another important aspect of model consolidation is the extent to which active reactions in the IGSMs (that is, reactions that can carry metabolic flux in principle) are preserved. As shown in Fig 3e, essentially all active reactions in one of the networks remained active in the RCM, and only reactions that were non-functional in both IGSMs remained inactive. In growth phenotype predictions, the RCM occasionally disagreed with all IGSMs, suggesting ‘new’ metabolic functions. For example, while neither of the IGSMs captured that P. putida can grow on L-quinate as sole carbon source, complementation of reactions in the RCM enabled a biologically consistent model behavior (Fig 4b). These aspects together indicate overall functionality of the automatically generated consensus model.
The performance of GSMs as mathematical models for cellular metabolism is typically evaluated by assessing their ability to correctly predict wild type and mutant growth phenotypes across different growth conditions [34]. We performed corresponding simulations for automatically refined consensus models as well as for their ancestors (IGSMs and BCM) for each of the four evaluated organisms (Fig 1a). Specifically, we computed sensitivity, specificity, accuracy, and Matthew’s correlation coefficient (MCC; unlike accuracy it takes the total numbers of true and false test cases into account) [35] for growth phenotype predictions (see S3 Protocol for details). Fig 5a shows the performance indicators for the IGSMs, the BCMs, and the automatically refined consensus models for each organism. In nearly all metrics, the IGSMs outperformed the BCM (except for P. putida), and they were outperformed by the RCM (except for B. subtilis). For B. subtilis, resolving inconsistencies in the BCM decreased all scores except sensitivity. This can be explained by one IGSM (iBSu1103) being largely based on a predecessor (iYO844); in addition, iBSu1103 was optimized for correct growth predictions using GrowMatch [34,36]. Information from iYO844 can thus include errors that were deliberately removed from iBSu1103 and it can reverse changes made by the performance optimization. Thus, although the prediction profiles of the RCMs largely resemble the IGSM profiles, RCMs on average outperform both the IGSMs and the BCMs, indicating efficiency of the automated consensus model generation methods in COMMGEN even in terms of prediction capabilities. Notably, user choices of the biomass reaction do not influence the performance substantially (Fig 5a), pointing to robustness of the methods as well.
Finally, we wanted to evaluate how automatic consensus model generation compares to its (largely) manual counterpart. We focused on the community approach to establish a yeast consensus model [20] based on the IGSMs iMM904 [37] and iLL672 [38] because this first model reconciliation effort is especially well documented. Fig 5b shows that transfer of the IGSMs into a standardized namespace alone identifies only small subsets of common metabolites and reactions. COMMGEN’s automated reconciliation method, in contrast, achieves nearly the same extent of matching between the IGSMs as reported for the manual curation. The automatically generated RCM showed good performance in mutant phenotype predictions (sensitivity = 0.98, specificity = 0.28, accuracy = 0.87 and MCC = 0.42; note that a comparison to the manual consensus model is impossible because the community effort did not aim at establishing a model suitable for FBA). In addition, COMMGEN directly identifies many inconsistencies between model reactions that result, for example, from different numbers of compartments in the IGSMs (Fig 5c). These would be clear starting points for domain experts for subsequent COMMGEN-assisted manual curation. We believe that the combination of automated procedures with close-to-manual quality and of support for targeted manual curations would substantially enhance future community efforts.
Genome-scale constraint-based metabolic models are both integrated knowledge repositories and predictive mathematical models. In terms of knowledge representation, a consensus model should be more consolidated than individual GSMs due to shared parts, more comprehensive due to unique parts, and more accurate due to reconciliation of inconsistencies in similar parts. A consensus model, however, can propagate errors in the initial models’ unique parts, and it may be less consistent than the initial models, especially when inconsistencies in similar model parts were not identified or reconciled.
Inconsistencies in GSMs are typically nested, not mutually exclusive, and therefore difficult to address, which so far prevented the development of methods for the automated generation of consensus models [30]. Manual network reconciliation, the predominant approach applied today, is difficult and cumbersome because the number of inconsistencies between just two or three IGSMs already runs in the thousands. Based on a systematic classification of inconsistencies, COMMGEN automatically identifies and semi-automatically reconciles inconsistencies between and within two or more IGSMs. The inconsistencies could theoretically be reconciled fully automatically, but automated resolution depends on the used reference databases, which vary to a large extent [18]. Therefore, COMMGEN does not entirely remove the need for manual inspection and curation. For example, our framework relies on network similarity between alternative realizations of metabolites and reactions in order to match them. Because the reactions surrounding biomass formation are often implemented very differently in different GSMs, they are not matched. While our implementation lets the user choose one of the IGSM biomass reactions, a manual update seems necessary as long as COMMGEN does not automatically fetch external information that would enable an automatic reconciliation of the biomass reaction. In addition, there exists a trade-off between sensitivity and specificity for the identification of inconsistent reactions, which limits the detection of lumped and non-lumped pathway representations with a different net reaction. Also, the identification of similar or identical reactions in different cellular compartments is difficult to achieve automatically (but an extension of the current framework could progress in this direction by combining the information from metabolite instances in different compartments prior to metabolite matching). COMMGEN thus forms a necessary bridge between full automation and high-quality manual curation for consensus metabolic model generation.
Regarding a GSM’s predictive mathematical model character, it is important to note that remaining inconsistencies in a consensus model can have severe effects, for example, when inconsistencies resulting from model merging are not adequately addressed. As a consequence, individual GSMs may outperform a consensus model in terms of predictive ability even though the latter is more representative of the available information. COMMGEN’s aim (and design) is to compare and reconcile IGSMs in order to obtain a high-quality representation of the IGSMs’ combined information. In contrast to model optimization methods such as GrowMatch [34], COMMGEN does not create a model optimized for predictive ability, and it does not use corresponding experimental information. However, our example applications also demonstrated that automatically generated consensus models almost always have higher predictive power than the manually curated IGSMs and that these models can be comparable to manually constructed consensus models as shown for yeast. COMMGEN increases coherence with the actual biological system while maintaining predictive power. This balance is of utmost importance for the usability and reliability of GSMs to elucidate cell function interactions.
As demonstrated by our case study for P. putida, we argue that (semi-) automatically generated consensus models provide the basis for additional improvements due to their comprehensiveness and standardized naming system. Gap-filling methods [2,39] may be able to close gaps that are not apparent in the IGSMs. One can use existing methods [2,40] to re-evaluate reaction directionalities, especially for reactions that differed in the IGSMs. Compartment assignment methods [41] can resolve remaining compartmentalization issues and optimization methods [34,42] may alter the model to increase its predictive ability. Finally, a good consensus model is a solid foundation for new models by providing a basis for GSMs of similar organisms, and via its integration into multi-scale whole-cell or tissue models [36].
More generally, the systematic integration of heterogeneous information is an essentially unsolved challenge in (post-)genomic biology. For metabolism, consensus GSMs are formalized means for complementing incomplete information, and for identifying and addressing errors through the comparison of independently generated GSMs for the same organism. COMMGEN automatically identifies and semi-automatically resolves widespread and highly interlinked inconsistencies between initial GSMs, thereby moving beyond existing approaches for manual and computer-aided consensus model generation. It can therefore facilitate the construction of new models by comparing and combining information from automatic model construction tools such as the modelSEED [43] and manual model construction efforts, and facilitate GSM updates using a reference—both tasks are analogous to consensus GSM generation.
While we focus here on the reconciliation of multiple GSMs for the same species, we argue that COMMGEN’s methods and standardization are more widely applicable. The identification of similar, yet distinct, biochemical entities can help to compare metabolic capabilities of organisms in detail via their GSMs, or even to compare entire pathway databases. However, dealing with different species will require new, systematic preprocessing steps to map gene sets in different organisms functionally to each other (e.g., via orthology or enzyme classification numbers), which is a topic of future research. In addition, COMMGEN’s methods for identifying redundancies and hierarchical relationships in networks can be used to further advance standardization of terms and ontologies. We therefore expect COMMGEN to be of substantial aid in future integration of knowledge for metabolic networks, to greatly accelerate model-building processes and to thereby improve subsequent high-throughput model-based network analyses. Although COMMGEN will not directly address the domain-specific problems, these capabilities will lay a solid foundation for the systematic, genome-scale comparison of metabolic spaces within and across genera and will have substantial impact for large-scale evolutionary analyses, design of microbial communities, and understanding of host-microbe (pathogen, microbiome) interactions.
iJN746 and iJP962 were requested from and received by email from the first authors of the corresponding papers. GSMN-TB was downloaded from http://sysbio3.fhms.surrey.ac.uk/. iNJ661 was obtained from the supplementary files of the corresponding paper. The remaining models were taken from the model repository at www.metanetx.org. See S1 Dataset for details.
For comparison to experimental data, the models were loaded into the COBRA toolbox [44]. The bounds of the boundary reactions were adjusted based on the medium composition and, where necessary, additional flexibility was provided to individual models. Gene knockout strains were simulated by removing the reactions requiring the encoded protein. To discriminate growth from no growth for wild type strains a default cut-off value (10−6) was used whereas a minimal relative growth rate (30%) to the wild type was used for mutant strains. See S3 Protocol for details.
In a metabolic network, reaction nodes are only connected to the metabolite and gene nodes that are involved in the corresponding reaction. Similarly, metabolite and gene nodes are only connected to reaction nodes. However, reaction nodes are not informative for the identity of metabolites as two metabolites representing the same chemical compound are non-overlapping in their connected reaction nodes. Therefore, we characterize metabolites by the other metabolite and gene nodes that are connected to the same reactions. We use this information to quantify how similar metabolites from different models are based on their network context. These similarity scores are then compared to the scores of metabolites that are known to match because they are present in both models: pairs of metabolites that score comparable to these shared metabolites may consist of functionally equivalent chemical compounds. We use a user-defined percentile of shared metabolite scores as a threshold to identify similar metabolites. The method is described in the following:
A lumped reaction is an artificial reaction that represents the net effect of multiple individual reactions. Therefore, if the lumped and non-lumped representations carry flux in opposite directions, steady state is maintained as they cancel each other out. We use this property to identify lumped reactions by linear programming. The method is described in the following:
Alternative transport reactions result in the transport of a metabolite between two compartments with a different net reaction. We identify metabolites with alternative transport reactions one metabolite at a time. If a metabolite is present in two or more compartments, we identify all transport reactions for this metabolite by selecting reactions where the metabolite is on both sides of the equation. If two of these reactions transport the metabolite between the same two compartments, these reactions are alternative transport reactions.
Invalid transport reactions are reactions that transport metabolites between two unconnected compartments. We identify these by forming a list of all compartments that are directly connected through transport reactions in the IGSMs and asking the user to indicate if any of these are invalid. For any of the invalid compartment connections, we identify reactions that contain metabolites from both compartments; these reactions are invalid transport reactions.
We create a separate stoichiometric matrix Scmp (m x r) for each compartment. These matrices only contain reactions of which all metabolites are in the same compartment. Columns (reactions) that are identical between these matrices represent identical reactions with an alternative compartmentalization.
In the MnXRef namespace, metabolites with an unclear compartmentalization are placed in the compartment UNK_COMP. For each reaction that contains a metabolite in UNK_COMP, we identify reactions from the other IGSM(s) that involve all metabolites with known compartmentalization similarly to the identification of alternative stoichiometries. These reactions are then filtered for reactions that also involve the metabolite with the unknown compartmentalization.
Boundary (exchange) reactions are artificial reactions that represent the exchange of metabolites with the medium. They only involve a single metabolite, and have no metabolites on the other side of the equation. In some models these reactions are lumped together with transport reactions that import metabolites from the extracellular compartment. After the MnXRef namespace conversion these reactions are still annotated as boundary reactions, and are thus easily identified in COMMGEN by searching for boundary reactions with non-extracellular metabolites.
To combine GSMs with an alternative compartmentalization, it is sometimes most straightforward to remove a compartment ‘RC’ from a GSM and move its reactions to a different target compartment ‘TC’. We defined four categories of reactions in RC, which are treated differently when RC is removed: (i) Reactions that only involve metabolites from RC are moved to TC; (ii) Multi-compartment reactions that transport a metabolite between RC and TC are removed; (iii) Multi-compartment reactions involving RC and TC that involve a chemical conversion are kept, but all metabolites from RC are placed in TC; (iv) Multi-compartment reactions involving RC and a metabolite other than TC are kept, and all metabolites from RC are placed in TC.
Identical net reactions are reactions that involve the same set of metabolites in the same stoichiometries, but they may be defined in opposing directions. Therefore, we create a double stoichiometric matrix Sdbl (m x 2r) that contains the normal stoichiometric matrix S (m x r), as well as its negative -S (m x r). We then identify columns (reactions) in Sdbl that are identical.
We convert the S (m x r) matrix to a Boolean (0/1) representation Slog (m x r). We then identify columns in Slog that are identical; these correspond to reactions involving the same metabolites, but in different stoichiometries.
GSMs often differ in their involvement of redox pairs in any particular reaction. The first step in identifying these inconsistencies is the creation of a list of redox pairs. COMMGEN comes with a list of commonly used redox pairs in the MnXRef namespace, and this list can be expanded by the user. COMMGEN can suggest expansions for this list by selecting metabolite pairs that co-occur frequently (≥ 80% of reactions). We identify reactions that are identical except for their redox pairs by expanding the stoichiometric matrix S (m x r) to Srdx (m+1 x r) by adding an artificial metabolite ‘redox pair’. Then, for each reaction that involves a redox pair, we put the stoichiometric coefficients of the redox metabolites in Srdx to ‘0’, and add a ‘1’ in the ‘redox pair’ row instead. We then use the same approach as for the identification of alternative stoichiometries to identify reactions that only differ in stoichiometries and redox pairs.
We convert the S (m x r) matrix to a Boolean (0/1) representation Slog (m x r). For each column (reaction) we then identify other columns that contain nonzero elements on each row where the respective column has a nonzero element. These sets of columns (reactions) are potentially nested reactions. We then confirm these sets by detecting sets where two or more metabolites that are on the same side of the equation for one reaction, are on the same side of the equation for the other reaction.
Similar reactions are reactions from different IGSMs that share a predefined number of genes, substrates and products. We identify similar reactions by constructing three sets of pairs of reactions: (i) reactions that originate from different IGSMs, (ii) reactions that share the required number of substrates and products, and (iii) reactions that share the required number of genes. All combinations of two reactions in each of these three sets are considered similar reactions.
All computational simulations and analyses were performed using MATLAB [45]. Gurobi [46] was used as linear programming solver for flux balance analysis.
COMMGEN uploads SBML files to MetaNetX.org [47], where the namespace conversion into MnXRef [31] is performed, and downloads the resulting model. Because errors may be introduced at this stage (incorrect namespace conversion of individual metabolites) the mapping is presented to the user who can reject incorrect matches. See S4 Protocol for details.
The COMMGEN version used for this paper is freely available as MATLAB code as S6 Protocol. A current version of COMMGEN can be found at https://gitlab.com/Rubenvanheck/COMMGEN.
|
10.1371/journal.pgen.1001265 | A Novel RNA-Recognition-Motif Protein Is Required for Premeiotic G1/S-Phase Transition in Rice (Oryza sativa L.) | The molecular mechanism for meiotic entry remains largely elusive in flowering plants. Only Arabidopsis SWI1/DYAD and maize AM1, both of which are the coiled-coil protein, are known to be required for the initiation of plant meiosis. The mechanism underlying the synchrony of male meiosis, characteristic to flowering plants, has also been unclear in the plant kingdom. In other eukaryotes, RNA-recognition-motif (RRM) proteins are known to play essential roles in germ-cell development and meiosis progression. Rice MEL2 protein discovered in this study shows partial similarity with human proline-rich RRM protein, deleted in Azoospermia-Associated Protein1 (DAZAP1), though MEL2 also possesses ankyrin repeats and a RING finger motif. Expression analyses of several cell-cycle markers revealed that, in mel2 mutant anthers, most germ cells failed to enter premeiotic S-phase and meiosis, and a part escaped from the defect and underwent meiosis with a significant delay or continued mitotic cycles. Immunofluorescent detection revealed that T7 peptide-tagged MEL2 localized at cytoplasmic perinuclear region of germ cells during premeiotic interphase in transgenic rice plants. This study is the first report of the plant RRM protein, which is required for regulating the premeiotic G1/S-phase transition of male and female germ cells and also establishing synchrony of male meiosis. This study will contribute to elucidation of similarities and diversities in reproduction system between plants and other species.
| Meiosis is a pivotal event to produce haploid spores and gametes in all sexually reproducing species and is a fundamentally different type of cell cycle from mitosis. Thus, the molecular mechanisms to switch the cell cycle from mitosis to meiosis have been studied by many researchers. In yeast and metazoans, RNA-binding proteins are known to play important roles in the post-transcriptional regulation of genes implicated in the meiotic entry and meiosis. In contrast, in the plant kingdom, the mechanisms to control the meiotic entry have largely remained elusive. In this study, we discover a novel RNA-recognition-motif (RRM) protein in rice (Oryza sativa L.), designated MEL2, and demonstrate that MEL2 is required for the faithful transition of germ cells from mitosis to meiotic cell cycle. Rice MEL2 shows partial similarity with human DAZAP1, which is an RRM protein and relates to Azoospermia syndrome in human, while there are critical structural differences between germline-specific RRM proteins of mammals and plants. Our findings will lead the molecular-biological studies of plant meiotic entry to the next steps and will enable a comparison of the systems of meiotic entry between animals and plants.
| The transition from mitotic to meiotic cell cycle is a central issue of reproductive development in all sexually reproducing species. Meiosis is a fundamentally different type of cell cycle from mitosis, and a pivotal event for eukaryotes to halve the chromosome number and form haploid gametes. The basic meiotic processes are evolutionarily conserved among eukaryotic species. In contrast, the signalling cascade that leads to meiosis initiation shows great diversity among species [1].
The mechanism initiating meiotic entry remains largely elusive in plants. Based on experiments using explanted pollen mother cells (PMCs) of Trillium, the commitment of mitotic cells to meiotic division is thought to be established during the premeiotic DNA replication (premeiotic S) or G2 phase in plants [2], [3]. Only Arabidopsis SWITCH1 (SWI1)/DYAD and its maize homolog AMEIOTIC1 (AM1) are known to be required for the initiation of plant meiosis. Both are plant-specific coiled-coil proteins with unknown functions [4]–[6]. The maize am1 mutant displays the replacement of male and female meioses by somatic mitoses, and eventually the degeneration of ameiotic meiocytes [7]–[9], indicating that AM1 is implicated in the decision of germ cells being directed to meiosis or mitosis. Thus, the primary function of AM1 is supposed in the premeiotic S or G2. However, immunocytological analyses revealed that AM1 diffuses inside the nucleus during premeiosis, and its localization shifts to meiotic chromosomes and pericentromeric regions during early meiosis [6], suggesting that AM1 plays a second role in progression of early meiosis. SWI1/DYAD does not seem to act directly to initiate meiosis, because it also acts in the regulation of meiotic chromosome structure and cohesion [4]. Thus, molecular mechanisms specifically underlying meiotic entry have been poorly understood in the plant kingdom.
Flowering plants have evolved an intricate network of regulatory mechanisms to ensure proper timing of the transition to flowering [10]. In addition, to achieve simultaneous fertilization within a limited season, the timing of meiotic entry is also strictly regulated. Male meiosis usually occurs in a large population of synchronously dividing cells to ensure sufficient fertility of the organisms. In plants, it is also synchronous among PMCs within an anther and among anthers within a single flower. The synchrony of male meiosis is thought to be established during premeiotic interphase. This is because the thymidine base analog, bromodeoxyuridine (BrdU), becomes incorporated synchronously into PMCs at premeiotic S [11], [12], while their preceding mitoses seem to occur asynchronously. The Arabidopsis mutant, tardy asynchronous meiosis (tam), exhibits a phenotype of delayed and asynchronous cell divisions during male meiosis [13]. The TAM gene encodes an A-type cyclin that abundantly accumulates in nuclei of male meiocytes during prophase I [14], strongly suggesting that cyclins and cyclin-dependent kinases govern the synchronous progression of plant meiosis.
RNA-recognition-motif (RRM) proteins play crucial roles in the regulation of germ cell development, especially meiosis, in yeast and metazoan species. They participate in the processing, transport, localization, and translation of mRNAs [15]. In fission yeast, the RRM protein, Mei2, is necessary for the initiation of meiosis by turning off the DSR-Mmi1 system for selective elimination of over a dozen meiosis-specific transcripts during the mitotic cell cycle [16]. Deletions encompassing the human Y-linked Deleted in azoospermia (DAZ) gene cluster, encoding RRM proteins, result in a complete loss or severe reduction of germ cells in the testis. In all species examined, the expression of DAZ, DAZ-like (DAZL) and their homologs has been reported only in germline cells [17]–[19]. These RRM proteins target the adenylate-uridylate-rich element (AU-rich element; ARE) found within the 3′ untranslated region (3′UTR) of mRNAs, and control mRNA turnover rate and translation in cooperation with poly(A)-binding proteins [20], [21]. Boule, the Drosophila ortholog of vertebrate DAZL, binds to the 3′UTR of Twine mRNA, which encodes a meiotic-type Cdc25 kinase, and promotes the translation of Twine and the premeiotic G2/M transition [18]. Mouse DAZL also binds to the 3′UTR and promotes the translation of Sycp3 mRNA, which is a component of the synaptonemal complex (SC) [22]. In plants, redundant roles of five members of Arabidopsis mei2-like RNA binding proteins (AMLs) are suggested in meiotic chromosome organization [23]. The AMLs are composed of three RRMs, like fission yeast Mei2, whereas their function is presumably different from that of yeast Mei2 in meiotic entry. Therefore, in plant reproduction, no RRM protein functionally analogous to that of yeast and metazoans has been reported.
In this study, we identified a novel rice RRM protein, MEIOSIS ARRESTED AT LEPTOTENE2 (MEL2). In mel2 anthers at early meiosis, most germ cells failed to enter the premeiotic S and meiosis, and a part escaped from the defect in the premeiotic S and underwent meiosis with a significant delay or continued mitotic cycles aberrantly. Rice MEL2 had partial similarity with human DAZ-Associated Protein1 (DAZAP1) [24]. However, MEL2 carried ankyrin repeats and the RING finger motif in addition to the RRM. This motif combination was conserved among the monocot Poaceae species, but not in dicot plants or in other organisms, despite the control of premeiotic germ-cell cycle essential for the reproduction of all eukaryotes. We will discuss structural differences and functional similarities of rice MEL2 to RRM proteins involved in the mammalian DAZ system mainly by analysis of the mel2 mutant.
Figure 1A illustrates the initiation and differentiation of rice germline cells described by Nonomura et al. [25], [26]. Primordial germ cells, or archesporial cells, are initiated at the hypodermis of the stamen and the ovule primordium. In the stamen, archesporial cells generate sporogenous and parietal cells. Male sporogenous cells undergo several premeiotic mitoses, and many meiocytes are produced in each of the four locules within the anther. Parietal cells continue periclinal divisions and generate three-layered inner-anther walls, the most interior of which become tapetal cells to provide nutrients and pollen-wall materials to male meiocytes and microspores. In the ovule primordium, plural archesporial cells are initiated. Subsequently, only a single archespore which adheres to the nucellar epidermis differentiates into a single sporogenous cell, and matures into a single female meiocyte. During premeiotic maturation, both male and female meiocytes enlarge far more in cytoplasmic and nuclear volumes than somatic cells.
To elucidate the genetic network that supports early germ-cell development, we selected a seed-sterile mutant line, ND00287, in rice. The sterile segregants of this line developed normally throughout their life cycle except for seed production (Figure S1). The sterile phenotype segregated as a single recessive mutation (fertile:sterile = 141∶47, chi-square (3∶1) = 0.00). Microscopic observation revealed that whereas the wild type had equally sized PMCs, sterile mutants produced divergent sizes of PMCs (Figure S2), probably a result of insufficient maturation and enlargement of premeiotic cells. Gametogenesis was disrupted in both male and female organs of the mutant (Figure S1). This phenotype resembled that of the mel1 mutant that we previously identified [26], and thus this gene was designated MEL2.
In anthers, the appearance of PMCs at the premeiotic interphase was unlikely to have been affected by the mel2 mutation, except callose accumulation around the cells was lacking in the mutant (Figure 1B, 1G). Callose is one of the cell wall component, and plays a vital role in the process of pollen development [27]. Interestingly, in mel2-1 mutant anthers, 0.69% of premeiotic germ cells (n = 291) underwent the mitotic metaphase, whereas no PMCs did in the wild type (n = 311) (Figure 1B, 1G). While the wild-type PMCs entered and underwent normal meiosis, the mutant PMCs were drastically hypervacuolated (Figure 1C, 1H). In the wild type, haploid microspores were released from tetrads after the completion of meiosis (Figure 1D). In contrast, highly vacuolated mutant PMCs failed to produce tetrads and microspores (Figure 1I). In addition to a failure in meiosis, tapetal cells also became aberrantly vacuolated and hypertrophic (Figure 1I). Highly vacuolated PMCs underwent apoptotic DNA fragmentation, revealed by the TdT-mediated dUTP-biotin nick end labeling (TUNEL) method (Figure 1E, 1J). Aberrantly hypertrophic tapetal cells also caused apoptosis at a step earlier than the programmed cell death (PCD) in normal process of tapetal development (Figure 1E, 1J). A serious defect in meiosis progression was also observed in the megaspore mother cell (MMC), the female meiocyte. When three of the tetrad spores had been degraded in the wild-type ovule (Figure 1F), the mutant MMC was still before meiotic cell division (Figure 1K) or the tetrad before degradation of three spores (Figure 1L). Surprisingly, in contrast to PMCs, no conspicuous vacuolation was observed in the MMC.
Though the ultrastructure of PMCs was also observed, no remarkable difference was observed between the wild type and mel2 mutant at the premeiotic interphase (data not shown). Howerver, at the meiotic prophase I, mel2 PMCs were hypervacuolated, but not in the wild-type PMCs, and in addition, mitochondria were enlarged in mel2 PMCs extremely more than those in wild types (). The formation of megamitochondria is known to precede apoptosis in the cells treated with various free radical-generating chemicals [28]. Thus, the ultrastructural analysis also suggested that the mel2 PMCs were directed to apoptosis.
Southern blot analysis of the ND00287 population revealed that the Tos17 insertion showed complete genetic linkage with the seed-sterile phenotype (Figure S4). This insertion tagged the gene locus, Os12g0572800, in Rice Annotation Project Database build4 (RAP-DB, http://rapdb.dna.affrc.go.jp/). When the 10-kbp wild-type genomic fragment including this locus was introduced into mel2 homozygous plants, the transformants recovered fertility (Figure S4). Furthermore, NE04525 carrying another allelic Tos17 insertion in this locus (mel2-2) exhibited the same mel2 phenotype (data not shown). Thus, we concluded that the Tos17 insertion into Os12g0572800 caused the mel2 mutation.
Full-length MEL2 cDNA was obtained from young panicles, including germ cells at developmental stages earlier than meiosis, by 5′-rapid amplification of cDNA ends (RACE) technology. To determine the transcriptional start site of MEL2 mRNA, three rounds of 5′RACE were performed. Four independent RACE libraries were produced by a gene-specific antisense primer nearest to the 5′ end (Figure S4B), and 16 of 17 RACE sequences terminated at the same 5′-endpoint. The putative start site predicted in this study mapped to 254 bp upstream from the location annotated in the RAP-DB. The MEL2 gene was composed of 14 exons and 13 introns (Figure S4). The MEL2 cDNA encoded a novel protein of 1,160 amino-acid residues (aa) of previously unknown function (DDBJ, AB522964). An online motif search revealed three conserved domains in the deduced MEL2 sequence: ankyrin repeats (ANKs, PF00023), an RNA recognition motif (RRM, PF00076), and a C3HC4-type RING finger motif (RING, PF00097) (Figure 2, Figure S5). An N-myristoylation consensus sequence, which allows protein binding to the plasma membrane or other intracellular membranes in eukaryotic cells [29], was found at the N-terminal end.
ANKs are implicated in protein-protein interactions [30]. Rice MEL2 contained 10 imperfect and tandemly aligned copies of ANKs (Figure S5). The RRM consisted of 80–90 aa with two highly conserved short motifs, an RNP1 octamer and an RNP2 hexamer, which are found in numerous proteins involved in post-transcriptional processes [31]–[33]. MEL2 contained a single RRM that conserved both RNP1 and RNP2 sequences (Figure 2). The MEL2 peptide sequence excluding the ANKs (451 aa to the end) showed similarity to human DAZAP1 in a BLASTp search (Score = 62.4 bits; E-value = 1e−07) [34]. DAZAP1 contains two RRMs at the N-terminus and a proline-rich domain at the C-terminus [24]. The C-terminal half of rice MEL2 was also rich in proline residues (615 to 1,042 aa, Figure 2).
The rice MEL2 sequence was evolutionarily conserved among Poaceae species; Sorghum bicolor, Brachypodium distachyon, and Zea mays (Figure 2, Figure S6). The Sorghum locus Sb08g018890 and the Brachypodium locus Bd04g03890 encoded putative proteins of 1,083 and 1,076 aa, 77.1% (813/1,055 aa) and 75.5% (791/1,048 aa) identical to rice MEL2, respectively. The Zea locus AC208308.3_FGP002 showed high conservation of the RRM and RING, but not the ANK, probably because the sequence information was incomplete in maize genome. The RRM followed by the proline-rich sequence was also conserved in these three species. No proteins carrying a combination of the three domains, ANK, RRM and RING, were found within the genome of the dicot model plant Arabidopsis thaliana, nor did we detect any proteins with the three-domain combination in genome information from 787 species included in the Archaea, Bacteria, Eukaryota and Viruses by in silico searches of the GTOP web database [35], [36]. Thus, we concluded that the motif combination found in rice MEL2 is unique to Poaceae or monocot plants.
The rice genome contained another predicted gene locus, Os12g0587100, similar to MEL2 (Figure S7). Os12g0587100 was located about 0.9-Mbp from the MEL2 locus toward the telomere side on the long arm of chromosome 12. Putative coding sequences of this locus were highly homologous to those of the MEL2 gene, while the homology was lost in the exon 12 and the following exons.
The spatiotemporal expression of MEL2 mRNA was examined by reverse-transcription PCR (RT-PCR) and in situ hybridization techniques. Based on the results of RT-PCR, the MEL2 mRNA was expressed mainly in young panicles and flowers (Figure S8). In situ hybridization revealed that the MEL2 expression was initiated in male and female archesporial cells at the hypodermis of stamen and ovule primordia (Figure 3A–3C). In the ovule, the MEL2 mRNA was expressed in multiple archesporial cells during early stages (Figure 3B, 3C), and subsequently in a single sporogenous cell (Figure 3D, 3E). In the stamen, strong MEL2 signals were detected in sporogenous cells (Figure 3D, 3F), and in addition, faint signals were also observed in parietal cells, which generate tapetal cells (Figure 3F). The MEL2 signal disappeared before meiosis in both male and female organs (Figure 3G). Thus, we concluded that the MEL2 gene was expressed in male and female germline cells from their initiation to meiosis, and weakly in male nursery cells including tapetal cells. The mel2-1 mutant flowers transcribed only aberrant types of MEL2 mRNA, which contained a 4.0-kbp Tos17 insertion within the seventh exon (Figure S9). This insertion was predicted to cause an in-frame stop codon at the 5′-end of the insert, and to result in a truncated form of MEL2 protein without the RRM and RING, if any was translated. Thus we concluded that the mel2-1 was a null allele.
The MEL2-like Os12g0587100 locus was also transcribed, and its transcripts were detected in young panicles and flowers (Figure S8). Sequencing the RT-PCR product revealed that the MEL2-like cDNA included many nucleotide polymorphisms against MEL2 cDNA, which could induce a shift in reading frame that would result in lack of RRM and RING motifs (data not shown). Therefore, the MEL2-like locus was considered to be a pseudo-gene.
If MEL2 had functioned during premeiotic mitosis prior to premeiotic interphase, the number of PMCs would be decreased in mel2-1 mutant anthers. However, it was not different significantly between PMC numbers in the wild type (103.7±15.3 per anther locule; average of three anthers) and the mutant (97.0±6.6). In addition, no remarkable aberration in the morphology of reproductive tissues and germ cells was observed in the mel2 anthers during the premeiotic-mitosis stage (Figure 1G). These observations indicate that MEL2 function was excluded from the premeiotic-mitosis stage of male germ cells.
To investigate whether the mel2 germ cells could pass through the premeiotic interphase normally, we examined the expression profile of several cell cycle-related genes in mel2 mutant anthers. The rice histone H4 mRNA is abundantly expressed during S-phase, and CDKB2;1 (cdc2Os3) is enriched during G2/M transition, but is less abundant or absent during S-phase in rice mitosis [37]. During premeiotic mitosis, CDKB2;1 was expressed in patches in both wild-type and mel2-1 anthers (Figure 4A, 4F), indicating that premeiotic germline mitoses occur asynchronously in the wild type and are unaffected by the mel2 mutation. At the boundary of premeiotic interphase and meiosis, no expression of CDKB2;1 mRNA was detected in the wild type (Figure 4B), indicating that the cells entered synchronously into premeiotic S, whereas CDKB2;1 was still expressed in patches in the mel2 anther (Figure 4G). At the same stage, H4 mRNA was expressed in all PMCs of wild-type anthers (Figure 4C). In contrast, in mel2 anthers, only a few PMCs exhibited H4 signal (Figure 4H). The asynchronous expression of CDKB2;1 and H4 was observed in anthers of three independent mel2 plants at the stages that the wild type underwent early meiosis. The MEL1 gene, which encodes an Argonaute family protein, is expressed exclusively in germline cells before meiosis [26]. In the wild type, MEL1 mRNA was strongly detected in sporogenous cells undergoing premeiotic mitosis (Figure 4D), and was rapidly downregulated during premeiotic interphase (Figure 4E). Premeiotic mel2 sporogenous cells also expressed MEL1, same as wild-type cells (Figure 4I). However, the expression continued aberrantly in early meiotic stages (Figure 4J).
Next, we performed the BrdU incorporation experiment for the premeiotic athers. In the wild type, ten of 31 flowers at the premeiotic interphase had anthers in which PMCs incorporated BrdU into their nuclei synchronously, whereas the most mel2 flowers had anthers that showed asynchronous incorporation of BrdU into the PMC nuclei (Figure 5A). In most of mel2 anthers, only 20% of PMCs incorporated BrdU simultaneously (Figure 5B, 5C).
These results clearly indicate that MEL2 plays essential roles in the decision for germ cells to enter the premeiotic S-phase.
Next, we examined whether the mel2 mutant PMCs were able to enter meiosis. Telomere clustering or a bouquet structure, a typical feature of zygotene meiocytes in maize [38], was observed in four of 10 wild-type meiocytes, whereas no bouquet was observed in any of 31 mel2-1 meiocytes (Figure S10). In wild-type soma and premeiotic meiocytes, centromeres and telomeres were arranged in peripheral and interior regions of the nucleus, respectively, and their positions were inverted during meiotic entry. However, mel2 PMCs at early zygotene lacked this inversion, retaining a soma-like centromere arrangement (Figure S10). Further, we performed immunofluorescent detection of rice meiotic proteins PAIR2 and ZEP1. PAIR2 transiently associates with meiotic chromosome axes and is required for SC establishment [39]. ZEP1 is a component of the transverse filament of SC [40]. Both genes were transcribed normally even in mel2 flowers (Figure S8). In wild-type meiocytes at early zygotene, PAIR2 associates along meiotic chromosome axes, and filamentous ZEP1 signals begin to elongate between homologous axes (n = 101) (Figure 6A). In contrast, in all mel2 meiocytes (n = 118) at early zygotene, neither PAIR2 nor ZEP1 was detected on chromosomes (Figure 6B). These observations indicate that mel2 meiocytes fail to enter meiosis when the wild type undergoes early zygotene.
At late zygotene in wild-type meiocytes, ZEP1 stretches extended overall meiotic chromosomes and most PAIR2 proteins had been removed from the axes, indicating the completion of homologous synapsis (Figure 6C). In mel2 meiocytes, 79.7% of meiocytes (n = 64) exhibited faint or abnormally dotted signals of PAIR2 in nuclei (Figure 6D). All these meiocytes showed a soma-like centromere arrangement. However, the remaining 20.3% showed an early zygotene-like, filamentous appearance of PAIR2 and a meiotic centromere arrangement (Figure 6E), whereas ZEP1 proteins, which failed to be loaded on homologous axes, accumulated aberrantly in the cytoplasm. It was impossible to observe whether the mutant PMCs emanating filamentous PAIR2 signals underwent subsequent meiotic steps, because of significant cell disruption due to hypervacuolation (Figure 1I) and apoptosis (Figure 1J).
These results strongly suggest that in mel2 mutant, most of male germ cells which show the defect in the premeiotic G1/S transition result in the lack of meiosis, even though meiotic genes have been normally expressed. The 20% cells could escape from the defect in the transition and enter early meiotic stages, but extremely later than usual, and yet they failed to establish the SC between homologous chromosome pairs until the apoptosis.
The seed-sterile phenotype of mel2 homozygous plants was rescued by introducing transgenes expressing the MEL2 protein with a T7-peptide tag at the C-terminus (C-tagged) (Figure S11), indicating that this recombinant protein is functional in vivo. Unfortunately, we failed to obtain any clear signals of the T7-tagged protein in western blotting (data not shown), probably because of the low level and spatiotemporal limitations of its expression. However, indirect immunofluorescence enabled to visualize the subcellular protein localization. In wild-type PMCs undergoing premeiotic mitosis, a faint signal was observed in the cytoplasm (Figure 7A). In premeiotic interphase, signals were found in the cytoplasm, especially concentrated at the perinuclear region (Figure 7B). By early meiosis I, MEL2 had been released to the cytoplasm of PMCs, and in turn, a faint signal appeared at the perinuclear cytoplasmic region of the inner anther-wall cells, including tapetal cells (Figure 7C). MEL2 signals finally disappeared at post-meiotic stages of PMCs and anther-wall cells (data not shown). This localization was observed in seven of eight anthers from two independent plants (C6#2, C9#2 in Figure S11). The C-tagged MEL2 signal was excluded from the nucleoplasm in any of these stages. In transgenic plants expressing N-tagged MEL2 protein, the immunofluorescent signal diffused over all the cytoplasm in premeiotic PMCs (Figure 7D). N-myristoylation is a post-transcriptional protein modification, in which myristic acid is covalently attached to an N-terminal glycine residue, exposed during cotranslational N-terminal methionine removal by N-myristoyltransferase [41]. Thus, the immunofluorescence of N-tagged MEL2 may represent the first methionine residue with the T7 tag that had been removed and diffused throughout the cytoplasm.
This study provides the first evidence that the novel RRM-containing protein plays essential roles in meiotic entry in rice. In the mel2 mutant, the progression of male and female meioses was significantly affected, and the male meiocyte and tapetal cells were hypervacuolated and directed to apoptosis (Figure 1, Figure S3). The mel2 mutation disturbed the most germ cells to transit into the premeiotic S-phase in anthers (Figure 4 and Figure 5). However, twenty percents of the cells could enter the S-phase (Figure 5B) and undergo meiotic processes in which chromosomes showed a typical appearance of early zygotene (Figure 6), while the wild-type cells underwent late zygotene. In these cells, neither precocious separation of sister chromatids nor sister centromeres was observed. This result indicates that the MEL2 function might be excluded from the regulation of meiotic chromosome structure and cohesion, different from the function of Arabidopsis SWI1/DYAD. Thus, the role of rice MEL2 could be specified in the premeiotic cell-cycle control.
The MEL2 function in premeiotic interphase will be tangible in comparison with that of maize AM1, the coiled-coil protein also controlling meiotic entry. AM1 is implicated in the decision of germ cells being directed to meiosis or mitosis [6]. Interestingly, in the am1 mutant, ameiotic mitoses replacing meioses occur synchronously [7], indicating that the synchrony of male meiosis is genetically separable from the meiosis commitment, and also that the AM1 function can be allocated into the meiosis commitment following the establishment of synchrony. In contrast, in the rice mel2 mutant, most of male germ cells could not enter the premeiotic S and lost the synchrony (Figure 4 and Figure 5). Thus, it is strongly suggested that the MEL2 function precedes the establishment of synchrony, the meiosis commitment and the function of maize AM1. The ameiotic mitoses also occurred in the mel2 anther, but only in a small amount of male germ cells (Figure 1G), probably representing that some of the PMCs would return to the mitotic cell cycle before the meiosis commitment. Taken together, we conclude that MEL2 plays an essential role in the premeiotic G1/S-phase transition in rice.
Pawlowski et al. [6] proposed the existence of a novel checkpoint system monitoring faithful transition of leptotene to zygotene based on the degeneration phenotype of am1 mutant PMCs. Our results may support this proposal in rice. The mel2 mutant PMCs initiated hypervacuolation and apoptosis simultaneously when wild-type PMCs underwent early meiosis (Figure 1). This degeneration of meiocytes would be an indirect effect of the mel2 mutation, because it was never observed in the mel2 MMCs (Figure 1K, 1L). In yeast and metazoans, a system referred to as pachytene checkpoint monitors for defects in homologous recombination and synapsis, and meiocytes arrested in pachytene will eventually be eliminated [42]. In contrast, plants are thought to lack the typical meiotic checkpoint [43]. This consideration has been based on most plant meiotic mutants being able to complete meiosis, while fragmentation or nondisjunction of chromosomes takes place. However, most plant materials examined so far are thought to have been mutated in meiotic machinery, but not in premeiotic events. In turn, the function of maize AM1 and rice MEL2 is supposed in premeiotic events, in contrast to the meiotic mutants previously reported in plants.
In this study, we mainly focus on male meiosis, because in rice, it is easier to be observed than female meiosis. However, the mel2 mutation also affected the progression of female meiosis (Figure 1K, 1L). The fundamental role of MEL2 might be in the initiation of premeiotic G1/S transition in the appropriate timing in both male and female cells.
The central region of MEL2 protein resembles human DAZAP1, whereas MEL2 possesses a single RRM, in contrast to the doublet in DAZAP1 (Figure 2). DAZAP1 is a member of the proline-rich RNA-binding proteins (PRRPs) [24], and also of the heterogeneous nuclear ribonuclear proteins (hnRNPs), known to bind to newly synthesized RNA transcripts and participate in their processing and export [44]. A role of human DAZAP1 in transcription is suggested by its specific exclusion from the transcriptionally inert XY body in the nuclei of pachytene spermatocytes, and a requirement for active transcription for its nuclear localization [45], [46]. Mouse DAZAP1 is also detected in the nucleus of pachytene spermatocyte, and its localization dramatically shifts from the nucleus to the cytoplasm during the maturation of spermatids [45], [47]. In male dazap1 mutant mice, spermatogenesis is arrested before the first meiotic division, and the cells are directed to apoptosis, whereas the female has largely normal oogenesis [48].
Human DAZAP1 is an interacting counterpart of DAZ and DAZL proteins [24]. DAZ, DAZL and BOULE are able to form an RNA-protein complex with another RNA-binding protein, PUM2, while they may function in distinct molecular complexes during germ cell development [49], [50]. In addition, yeast two-hybrid screening of testis proteins revealed that human DAZ interacts with DZIP3 (DAZ-interacting protein3)/hRUL138 [49], which has the potential for RNA binding and RING E3-ubiquitin ligase. DZIP3/hRUL138 is expressed ubiquitously in various tissues, and is localized to certain cytoplasmic structures, especially perinuclear regions, but excluded from the nucleoplasm [51]. Thus, in the mammalian system, RNA-binding proteins, such as DAZ, DAZL, DAZAP1 and PUM2, first associate with the target mRNA precursors in the nucleus of germline cells. They export the mature targets to the cytoplasm, form a complex with ubiquitously expressed cytoplasmic proteins, such as DZIP3/hRUL138, on the cytoplasmic nuclear membrane or endoplasmic reticulum, and regulate the translation of target mRNAs.
Rice MEL2 localized the perinuclear region, but it was excluded from the nucleoplasm of germ cells, distinct from mammalian DAZ families (Figure 7). We hypothesize that MEL2 may be a hybrid form of a DAZAP1-like protein and a DAZ-interacting E3 ligase, such as DZIP3/hRUL138, and may have evolved to acquire a germline-specific function in ancestral monocots. This idea raises the possibility that RING E3 ligase-dependent ubiquitination is required for germline development commonly in eukaryotic species. It is also suggested that unknown DAZ-family proteins that transport the target mRNAs from the nucleoplasm to the cytoplasm exist in plant germline cells.
The Arabidopsis locus At5g57740 or XBAT32 encodes a MEL2-like protein composed of ANK and RING motifs at the N- and C-termini, respectively, but not of RRM, and it promotes lateral root formation by inhibiting ethylene biosynthesis [52], [53]. XBAT32 is expressed ubiquitously in various Arabidopsis tissues, but most abundantly in anthers. We hypothesized that domain-shuffling events occurred between an RRM protein and a XBAT32-like protein required for meiotic entry in ancestral monocots after the monocot-dicot divergence around 200 million years ago [54].
In the mel2 mutant, synchronous progression of premeiotic S-phase was completely disrupted (Figure 4 and Figure 5). This mel2 phenotype indicated that the genetic system controlling the premeiotic G1/S-phase transition would closely relate to the system terminating the premeiotic mitosis and establishing the synchronous progression of premeiotic- and meiotic-cell cycles in the rice anther. Figure 7F summarizes a transition of subcellular localization of MEL2 protein. It is plausible by analogy with the mammalian DAZ system that the perinuclear localization of MEL2 functions in the translational inhibition of some cell-cycle related gene(s), cooperating with the perinuclear translational machineries. MEL2 may temporarily arrest the progression of asynchronous germ-cell cycles at premeiotic G1 end or the onset of S-phase at the perinuclear region, and the synchronous release of MEL2 to the cytoplasm allows the cells to enter premeiotic S-phase synchronously within an anther. The identification of binding substrates of MEL2 will contribute to evidence this hypothesis.
According to this idea, unknown signalling factor(s) should be hypothesized to mediate cell-cell communication and promote the synchronous release of MEL2 from the perinucleus. During premeiotic interphase and early meiosis, it is known that male meiocytes form a single coenocyte in an anther locule, in which the cytomictic channels connect each other of the cells [55]. This channel network may help the signalling for synchrony of male meiosis.
In addition to failure of meiotic entry, the mel2 mutation caused the hypervacuolation and hypertrophy of tapetal cells (Figure 1I). This is different from the case of the maize am1 mutant, in which no tapetal-cell degeneration has been reported [7], [9]. Tapetal cells provide nutrients and pollen-wall materials to microspores, and degenerate, probably by PCD [56]. It is demonstrated that several gibberellin (GA)-related rice mutants display a hypertrophy of tapetal cells and result in male sterility [57]. This hypertrophic phenotype is attributed to the absence of PCD in the tapetum, because externally supplied GA can restore the tapetal phenotype of the oscps1 mutation, which causes defects in GA biosynthesis. The hypertrophic tapetum in the GA-related mutations seems to resemble that in the mel2 mutation. However, as opposed to GA mutants, apoptosis identified by the strong TUNEL signal arose in the mel2 tapetum (Figure 1J). In addition, GA-related mutants can undergo meiosis and produce tetrad spores [57], distinct from the mel2 mutant. These observations may suggest that mel2-dependent hypertrophy of the tapetum is independent of the GA-signalling pathway.
Both MEL2 mRNA and protein were expressed weakly in tapetal cells (Figure 2 and Figure 6). Thus, tapetal degeneration in mel2 anthers would be a primary effect of the absence of MEL2 protein, while it is difficult to neglect the possibility that degeneration of PMCs directly causes tapetal-cell hypertrophy. MEL2 expression in tapetal cells appeared during early meiosis I (Figure 7). Tapetal cells are known to become multinucleate or polyploidized by mitoses without cytokinesis, in many cases during meiotic I prophase [58]. In rice, tapetal cells become binucleated, and in Arabidopsis, binucleation occurs synchronously at early leptotene. Thus, MEL2 function may be required not only for meiotic entry, but also for synchronous tapetal-cell binucleation, the disruption of which may induce hypertrophy and precocious tapetal-cell death. However, the synchronous expression of H4 among tapetal cells was frequently observed even in mel2 anthers (Figure 4H). It remains unclear whether this result excludes MEL2 function from the synchronization of tapetal-cell division.
In conclusion, we have proved that the RRM protein plays an essential role in plant germ-cell development in addition to yeast and metazoans, although the protein's structure, function, timing of expression, and subcellular localization differ between rice and non-plant species. This study also suggests that genome shuffling and the generation of a novel motif combination in ancestral monocots may have brought rice MEL2 a unique function in germline cell-cycle control. Further analysis of MEL2 function will contribute to better understanding of post-transcriptional or post-translational regulation of plant germ-cell development, and also to elucidating similarities and differences in reproduction systems between plants and other species.
Seed-sterile mutant lines were selected as described [25]. For cytological and expression analyses, the F2 plants four-times backcrossed with cv. Nipponbare (BC4F2) were used. Non-transgenic plants were grown in a field in the city of Mishima, Shizuoka, Japan. Transgenic plants were grown in the growth chamber, LPH-2HCT (NK system), at 30°C for 14 hrs with the light and at 25°C for 10 hrs in dark.
The linkage relationship between the sterile phenotype and transposed Tos17 fragments was analyzed by DNA gel blot hybridization and polymerase chain reaction (PCR) using the R3 population of 188 plants segregating the mel2 seed-sterility. DNA extraction, DNA gel blotting, cloning and isolation of the Tos17-tagged genome sequence were performed as described [25]. PCR genotyping for the mel2 mutant populations was performed using the mixture of three primers: 868, 869 and T17LTR4MF for mel2-1 allele, or 870, 871 and T17LTR4MF for mel2-2 allele (Table S1). 50- to 100-ng genomic DNAs and above three primers in 5-µL water were added to the same volume of GoTaq Green Master Mix (Promega).
The longitudinal length of flower buds and anthers was measured under the dissection microscopy SMZ645 (Nikon). The anther length is generally used as a criterion to determine developmental stages of germline cells in rice [59]. This criterion was also used in this study, because the anther length was increased proportional to longitudinal flower (or lemma) length, whose elongation was unaffected by mel2 mutation, until the end of meiosis (Figure S12). A precise stage in each flower or anther was determined by the mRNA expression or immunofluorescence of stage-specific gene or protein markers.
The full-length MEL2 cDNA was obtained from 3.0-cm young rice panicles, frequently including flowers in premeiosis. RNA extraction and RACE reaction were according to the methods as described [26]. In addition to the oligo(dT)20 primer, two MEL2 gene-specific, antisense primers, 871 and T2028R were used for three rounds of RTs, followed by the RACE-PCR with adaptor primers (AP1 and AP2) supplied by the manufacturer (Table S1). All products were cloned into pCR-BluntII-TOPO vector (Invitrogen), and sequenced by Dye Terminator Cycle Sequencing kit and ABI PRISM 3130xl Sequencer (Applied Biosystems). Three independent RACE fragments were combined into a single, full-length cDNA sequence by PCRs.
The entire coding region of the MEL2 gene and its 2.0-kbp upstream cis-sequence from the putative transcriptional start site were included within the 10-kbp of single SalI genomic fragment (Figure S4). The 10-kbp fragment was isolated from the rice BAC clone OSJNBa0036A19, and subcloned into the pPZP2H-lac binary vector [60]. This plasmid or the empty vector as a negative control was introduced into mel2-1 homozygous calli in accordance with the method as described [61]. The genotype of calli was determined by PCR, in which the template DNA was extracted from young shoots germinating on the callus-induction medium.
A sequence of the full length MEL2 cDNA was supplied for the BLAST search on RAP-DB (http://rapdb.dna.affrc.go.jp/), and we found Os12g0587100 locus (MEL2like) homologous to MEL2 gene within rice genome. Genomic sequences of MEL2 coding region and MEL2-like locus were compared by HarrPlot program [62]. Specific primer sets for MEL2-like, TMEL2L1402F/TMEL2L1974R and 919/TMEL2L1974R were designed as referencing HarrPlot information and used for RT-PCR against the RNA extract from young flowers. Then we succeeded to amplify the MEL2-like transcript, in which the putative intron sequences were spliced out when compared with the genomic sequence.
Histological analysis of rice reproductive organs was done by using the plastic- embedded sections, the preparation method of which was described [25]. Sections were stained with toluidine blue (Chroma Gesellshaft Shaud) or provided for the TUNEL assay and other immunofluorescent analyses.
TUNEL was performed as described previously [56]. Plastic-embedded sections of rice panicles and flowers were treated with TUNEL apoptosis detection kit (DeadEnd Fluorometric TUNEL system, Promega) according to the manufacturer's instruction. The fluorescent TUNEL signal was detected by FV300 CLSM system and Photoshop.
Electron microscopic observation was done in accordance with the method described previously [39].
In situ hybridization against rice tissues was performed in accordance with the method as described [25]. To avoid a cross hybridization among highly homologous gene families, we adopted the high-stringency condition with 0.3 M NaCl and 50% formamide at 50°C for hybridization and 0.5xSSC at 50°C for wash. For the synthesis of RNA probes, two short ∼500-bp DNA fragments were amplified by PCRs of the MEL2 cDNA with the primer sets, 919/1034 and 1035/1036, respectively (Table S1). Both fragments were cloned into the pCRII-TOPO vector (dual promoter system) (Invitrogen), and transcribed to make antisense or sense RNA probes by SP6 or T7 promoters with DIG RNA labeling kit (Roche). Three PCR fragments against OsCDKB2;1 cDNA were amplified by primer sets of M486F/M718R, M415F/M537R, and M609F/M739R, respectively, and cloned into pCRII-TOPO. The full-length 583-bp cDNA of rice histone H4 (RAP-DB: Os09g0553100) was cloned into pBluescript SK- (Stratagene). Both plasmids were provided for the synthesis of RNA probes as in MEL2.
Fresh young panicles of 3–5 cm in length were curt from stems and placed in 100 µm BrdU solution in the dark for 4 hours. Plastic sectioning and detection of incorporated BrdU were done in accordance with the method described previously [39]. Before the immunization, ten minutes treatment of sections with Proteinase K (0.1 mg/mL, Sigma) often improved the accessibility of antibodies and the intensity of anti-T7-signals.
To investigate the MEL2 expression profile, total RNAs were extracted from various tissues of wild-type rice plants; embryo and endosperm from mature seeds, seedlings, shoot apices, leaf blades, leaf sheathes, roots, flag leaves, 1 cm young panicles, young flowers in 1–2 mm, 2–4 mm and 2–7 mm lengths, and mature flowers. All tissues were cut off and handled with forceps, and immediately transferred into microtubes filled with liquid nitrogen and stored at −80°C. RNeasy Plant Mini kit (QIAGEN) was used for RNA extraction. Total RNAs were reverse-transcribed with the oligo(dT)20 primer and SuperscriptIII reverse-transcriptase (Invitrogen), and provided for semi-quantitative RT-PCR. For MEL2 mRNA, the primers, 918/919, were used. To investigate the structure of MEL2 transcript in the mel2 mutant, the primer sets, 918/919 and 868/869, were used. An expression of rice meiotic genes was examined in the mel2 mutant flowers by using the following primer sets; 496/647 for PAIR1, 555/ 518 for PAIR2, and K180/K183 for ZEP1. The primer set ActinF/ActinR was used to amplify rice Actin cDNAs as a positive control.
Indirect immunofluorescent staining of rice meiocytes was performed in accordance with the method as described [39] with minor modifications. Rat anti-ZEP1 and rabbit anti-POT1 antibodies (Komeda, Kurata, and Nonomura, unpublished) were diluted in 1/1000 and 1/3000, respectively, and detected with AlexaFluor647 goat anti-rat IgG (Molecular Probes) and Cy3 goat anti-rabbit IgG (Amersham). Maximum four channels of fluorescent signals were simultaneously observed by Fluoview FV300 CLSM system, upgraded with LD405/440 laser unit (Olympus). Captured images were enhanced and pseudo-colored by Photoshop CS2 software (Adobe).
Plasmid constructions to produce T7 (MASMTGGQQMG)-tagged MEL2-expressing plants were based on the 10-kbp genomic SalI-fragment same in the complementation test. The 10-kbp SalI fragment (Figure S1) was subcloned into pT7Blue vector (Novagen). To add the T7 tag to the N-terminus of MEL2, the 476-bp fragment including the translational start site was amplified with the primers MEL2gApaI2F/MEL2gNotI2R, directly cloned into pCR-BluntII-TOPO vector (Invitrogen), and provided for site-directed insertion of T7-tag sequence by the inverted tail-to-tail direction PCR with primers MEL2T7NF/MEL2T7NR and for ligation as described [63]. This plasmid was again provided for PCR with MEL2gApaI2F/MEL2gNotI2R (476bp+T7). The ApaI-NotI fragment of MEL2g/pT7Blue plasmid was replaced to the 476bp+T7 fragment by In-Fusion Advantage PCR Cloning Kit (Clontech). Finally, the insert carrying the T7 tag was cut out with SalI and inserted into SalI site of the binary vector pPZP2H-lac [59]. To add the T7 tag to the C-terminus, the middle 4-kbp and the 3′-terminal 400-bp fragments of MEL2 genome were amplified with primer sets, MEL2gSmaIF/MEL2InFu1R, and MEL2InFu1F/MEL2ctransEndR, respectively. The latter 400-bp fragment was cloned into pCR-BluntII-TOPO vector, and provided for site-directed insertion of T7-tag sequence just in front of MEL2 stop codon (400bp+T7). The plasmid used for above complementation test was digested with SmaI to remove the latter half 5.5-kbp genomic fragment. The rest sequence, including pPZP2H-lac and the first half of MEL2 gene, was fused with the middle 4-kbp and the 400bp+T7 fragments by In-Fusion Cloning Kit. Two resultant binary plasmids, the N-tagged MEL2 plasmid (MEL2gT7N/pPZP2H-lac) and the C-tagged one (MEL2gT7C/pPZP2H-lac), were introduced into mel2/mel2 calli, and transgenic plants were regenerated according to the method as described [60].
Immunocytology was done by using plastic sections of transgenic anthers in accordance with indirect immunofluorescense above mentioned, with goat anti-T7 antibody (Bethyl Laboratory) as a primary antibody and AlexaFluor488 donkey anti-goat IgG (Molecular Probes) for detection.
|
10.1371/journal.pbio.1001257 | Histone Deacetylase Complexes Promote Trinucleotide Repeat Expansions | Expansions of DNA trinucleotide repeats cause at least 17 inherited neurodegenerative diseases, such as Huntington's disease. Expansions can occur at frequencies approaching 100% in affected families and in transgenic mice, suggesting that specific cellular proteins actively promote (favor) expansions. The inference is that expansions arise due to the presence of these promoting proteins, not their absence, and that interfering with these proteins can suppress expansions. The goal of this study was to identify novel factors that promote expansions. We discovered that specific histone deacetylase complexes (HDACs) promote CTG•CAG repeat expansions in budding yeast and human cells. Mutation or inhibition of yeast Rpd3L or Hda1 suppressed up to 90% of expansions. In cultured human astrocytes, expansions were suppressed by 75% upon inhibition or knockdown of HDAC3, whereas siRNA against the histone acetyltransferases CBP/p300 stimulated expansions. Genetic and molecular analysis both indicated that HDACs act at a distance from the triplet repeat to promote expansions. Expansion assays with nuclease mutants indicated that Sae2 is one of the relevant factors regulated by Rpd3L and Hda1. The causal relationship between HDACs and expansions indicates that HDACs can promote mutagenesis at some DNA sequences. This relationship further implies that HDAC3 inhibitors being tested for relief of expansion-associated gene silencing may also suppress somatic expansions that contribute to disease progression.
| The human genome contains numerous DNA trinucleotide repeats, which mutate infrequently in most situations. However, in families affected by certain inherited neurological diseases such as Huntington's, a trinucleotide repeat has undergone an expansion mutation that lengthens the repeat tract. This expansion is generally sufficient to cause disease. Further germline and somatic expansions in affected families occur at very high frequencies—approaching 100% in some cases—suggesting that mutation of the trinucleotide repeat becomes the norm rather than the exception, while the rest of the genome remains genetically stable. These observations indicate that trinucleotide repeat expansions are localized in the genome and occur by novel mutational mechanisms. We searched for proteins that favor expansions and identified specific histone deacetylase complexes (HDACs)—comprising enzymes that remove acetyl groups from histones—in budding yeast and in human astrocytes. Interfering with these HDACs by mutation, RNA interference, or small molecule inhibitors blocked 50%–90% of expansion events. We also found that yeast HDACs promote expansions via a downstream deacetylation target, the nuclease Sae2. These results indicate that HDACs promote trinucleotide repeat expansions by modulating key proteins, which in turn catalyze the expansion. We postulate that HDAC inhibitors, currently being tested for relief of the transcription-related consequences of expansions, may have the beneficial side effect of reducing the risk of further somatic expansion.
| The relentless expansion of trinucleotide repeats (TNRs) causes Huntington's disease (HD), myotonic dystrophy type 1 (DM1), and at least 15 other inherited neurological disorders [1]. It is thought that expansions are actively promoted by the presence of key proteins, not their absence, probably due to the “corruption” of their normal biochemical activities by TNR DNA [2]–[4]. Evidence for promoting factors includes the fact that disease alleles expand at high frequencies, sometimes approaching 100% [5], in otherwise normal individuals and in a number of transgenic and knockin mouse models of HD and DM1 [6]–[12]. Using candidate gene approaches, the DNA repair factors Msh2, Msh3, Pms2, Ogg1, and Xpa were identified as promoting proteins in mice, based on the fact that somatic expansions are suppressed ∼50%–90% by homozygous knockout of Msh2, Msh3, Pms2, Ogg1, or Xpa [6]–[13]. Knockout of Msh2 or Msh3 also largely eliminates intergenerational expansions [7],[9],[10],[14]. Thus, key DNA repair components promote expansions in certain mouse models.
The transgenic mice studies described above monitor long, disease-causing TNRs becoming even longer. For example, commonly used HD mouse models carry CAG tracts of 110–120 repeats [10],[12]. A human inheriting an HD allele in this length range would develop the disease as a young child [15]. As an alternative approach, we focus on expansions near the crucial threshold, a narrow range of allele lengths (∼30–40 uninterrupted repeats in humans [2],[4],[16]) that demarcates stable shorter repeats from unstable longer tracts. Expansion risk in humans and in yeast increases sharply once the threshold is crossed [17],[18]. Expansions crossing the threshold are critical initiating mutations leading to enhanced instability and disease [2]–[4]. It is not known whether the mechanism of expansion is the same for threshold-length alleles and long, disease-causing tracts. In this study, we find that yeast mutants lacking the nucleases Sae2 or Mre11 reduce expansion rates for (CTG)20 alleles, whereas sae2 or mre11 mutants show increased expansion frequencies for long (CAG)70 repeats [19]. This new evidence suggests that triplet repeat length helps determine expansion mechanism.
The goal of this study was to identify novel factors in yeast and human cells that promote expansions of TNR alleles near the threshold. We found specific histone deacetylase complexes (HDACs) that promote expansions, plus one human histone acetyltransferase (HAT) that inhibits expansions, and we suggest a mechanistic link between HDACs and DNA repair. These results indicate a causal relationship between HDACs and expansions, and they show that protein acetylation and deacetylation are key modulators of TNR instability.
If specific proteins promote TNR expansions, then mutants deficient in these proteins will have fewer expansions. A large-scale yeast mutant screen was performed to identify mutants with reduced expansion rates. Cells with a (CTG)20-CAN1 reporter (Figure 1A) were randomly mutagenized with a disruption library. A (CTG)20 repeat tract was utilized, as this allele length is near the apparent threshold in yeast [18]. Reduced expansion rates are manifested as fewer canavanine resistant cells (Figure S1). Nine thousand disruptants, covering approximately 50% of non-essential genes, were subjected to several rounds of screening with increasing stringency. Eleven mutant genes were identified that consistently suppressed TNR expansions (Figure S1). Three of the 11 genes were SIN3, PHO23, and HDA3. SIN3 encodes a subunit of histone deacetylases Rpd3L and Rpd3S, whereas the subunit encoded by PHO23 is unique to Rpd3L. HDA3 encodes a subunit of another HDAC, Hda1. The hda3 mutant was found twice, along with single isolates of sin3 and pho23. Thus, a blind screen pulled out three genes encoding components of Rpd3L and Hda1, an enrichment of ∼100-fold compared to random chance. This clustering of mutations in related enzymes suggested a causal relationship between specific HDACs and TNR expansion.
Targeted knockouts of sin3, pho23, and hda3 confirmed the gene assignments and allowed further analysis of expansions. Expansion rates were quantified using two reporters, CAN1 (Figure 1A) and URA3 [18], and all expansions were confirmed by PCR (Figure 1B). If an HDAC mutant primarily affects the instability at the triplet repeat, independently of the readout gene, then similar phenotypes would be expected for assays with CAN1 and URA3. This outcome was observed (Figure 1C and Table S1). Single mutants of sin3, pho23, and hda3 showed 9- to 18-fold reductions in expansion rates for the CAN1 reporter integrated into chromosome II (Figure 1C, left panel). Expansion rates were reduced >1,000-fold in the double mutants pho23 hda3 and sin3 hda3, which are simultaneously deficient in both Rpd3L and Hda1. When the reporter gene was URA3, a similar pattern of suppressed expansion rates occurred (Figure 1C, middle panel). The magnitude of the phenotype was somewhat smaller: 2- to 4-fold suppression in expansion rates for single HDAC mutants, and 10- to 18-fold for the double mutants. Thus, both CAN1 and URA3 reporters integrated at the same locus yielded similar outcomes, suggesting that Rpd3L and Hda1 affect instability of the TNR. To exclude a position effect, the CAN1 reporter was relocated to an integration site on chromosome V. Suppression of expansions was again seen for the HDAC mutants (Figure 1C, right panel). Single mutants reduced expansion rates by 2- to 3-fold, while the pho23 hda3 and sin3 hda3 double mutants yielded 12- to 340-fold effects. In total (Figure 1C), the single mutants sin3, pho23, or hda3 showed significant reduction in CTG expansion rates in seven of nine assays. All six assays using the double mutants, pho23 hda3 or sin3 hda3, consistently gave lower expansion rates, and the double mutant effect was always stronger than for the single mutants. HDAC mutants in a common commercial strain, BY4741, also displayed reduced expansion rates for CAN1 integrated at LYS2. Relative to wild type, expansion rates in the sin3 mutant were strongly suppressed (>100-fold), with a milder phenotype for pho23 (3-fold reduced), and a small but not statistically significant reduction of 1.7-fold for hda3. Overall, targeted knockout of Rpd3L and/or Hda1 suppressed expansion rates in most assays, and expansions were almost completely eliminated in some cases.
Expansion suppression could be phenocopied by treating wild type cells with trichostatin A (TSA), which inhibits many but not all HDACs [20]. TSA reduced expansion frequencies by 2.6-fold (Figure 1D) at a concentration that inhibits most HDAC activity of Rpd3 and Hda1 in vitro [21]. This finding is consistent with a published report showing that TSA-treated Drosophila had ∼3-fold fewer expansions of a (CAG)78 transgene, with preferential modulation of +1 repeat changes relative to other sizes [22]. In yeast, expansion sizes were similar with or without TSA, ranging from +6 to +19 repeats (Figure S2). Cells with impaired HDAC function showed the anticipated accumulation of acetylated histone H3, by nearly 5-fold in the sin3 hda3 mutant and about 2.4-fold in wild type cells treated with TSA (Figure 1E). Compared to the HDAC mutants, TSA gave smaller effects on both expansion levels and the accumulation of acetylated histone H3, presumably due to incomplete inhibition by the drug.
Several control experiments eliminated trivial explanations of the HDAC effect on expansions. The range of expansion sizes was similar in wild type cells, HDAC mutants, and TSA-treated cells (Figures 1F and S2), indicating that HDAC status did not affect the genetic selection for expansions. Rather, the expansion size data suggest that HDACs likely govern initiation of expansions; there are fewer initiation events when HDACs are mutated or inhibited, but once the process is started the final size of the expansion is similar. There was no growth disadvantage of the HDAC mutants, with or without an expanded TNR, under conditions that select for expansions (Figures S3 and S4). CAN1 transcript levels varied by 2-fold or less in the HDAC mutants (Table S2), showing no correlation with changes in expansion rates. Finally, suppression of expansions was primarily attributable to Rpd3L and Hda1, because only modest expansion phenotypes occurred in mutants defective in the alternative HDACs Rpd3S, Hos1, Hos2, Hos3, or Sir2 (Figure S5). In summary, mutation or chemical inhibition of yeast Rpd3L and Hda1 suppresses CTG repeat expansions by 50%–90%, with even greater effects in some mutant strains. These data support a mechanistic link between triplet repeat expansions and the yeast HDACs Rpd3L and Hda1.
To address whether HDACs promote expansions in human cells, we focused on class I human HDACs, the homologs of yeast Rpd3 [23]. The small molecule inhibitor 4b is selective for the class I enzyme HDAC3 but with some activity against HDAC1 [24]. 4b treatment reverses FXN gene silencing in primary cells from Friedreich's ataxia patients [24] and relieves disease phenotype and transcriptional abnormalities in HD transgenic mice [25]. In light of the yeast experiments presented above, we posited that HDAC inhibition by 4b might have the added benefit of suppressing expansions in human cells. To test this idea, CTG repeat expansions were measured in a cultured human astrocyte cell line, SVG-A. Glial cells such as astrocytes show somatic expansions in HD patients [26], and SVG-A cells support expansions in culture, as measured by the assay shown in Figure 2A [27].
4b efficiently suppresses TNR expansions in SVG-A cells at doses that are well tolerated. Treatment with 4b reduced expansion frequencies in a dose-dependent manner (Figure 2B and Table S3). Compared to the DMSO-only control, expansion frequencies were suppressed 70% and 77% by 4b at 10 µM and 20 µM, respectively. In contrast, treatment of SVG-A cells with an HDAC1- and HDAC2-selective inhibitor called compound 3 [28] did not suppress expansion frequencies (Figure 2B; small increases were not significant). Together, the inhibitor results suggest HDAC3 is the relevant target. Confirmation came from RNAi knockdowns. Knockdown of HDAC3 resulted in 76% reduction in expansion frequencies (Figure 2C), the same extent seen at the highest doses of 4b, whereas knockdown of HDAC1 elevated the expansion frequency slightly but not to a statistically significant level. Inhibiting HDAC3 with 4b or knocking it down changed the frequency of expansions, not their sizes (Figure 2D). Expansions added as many as 18 repeats to a starting tract of 22 repeats; thus, some expansions regulated by HDAC3 in SVG-A cells cross the threshold of 30–40 repeats observed in humans [2],[4],[16]. The reduced number of expansions upon 4b treatment could not be attributed to increased cell death, because the SVG-A cells retained ≥83% viability, relative to DMSO-only control, even at the highest dose of inhibitor (Figure 2E). Molecular analysis of global histone H4 acetylation showed the anticipated increase in acetylated H4, up to about 10-fold, when cells were treated with 4b (Figures 2F and S6). The opposite phenotype—increased expansions—was seen with RNAi knockdown of the histone acetyltransferases CREB-binding protein (CBP) and p300 (Figure 2G), consistent with observations in Drosophila [22]. We conclude that HDAC3 and CBP/p300 have opposing effects on expansions in SVG-A cells, with HDAC3 promoting TNR expansions.
We first tested the idea that expansion rates are suppressed in cis by hyperacetylation of histones near the repeat tract, as might occur in HDAC mutants. The approach took advantage of previous studies showing that transcription and histone acetylation at some yeast genes are particularly sensitive to the absence of SIN3. One such locus is the INO1 gene, which we refer to as a “hot” zone. In sin3 mutants compared to wild type, transcript levels increase about 30-fold [20],[29] and histone acetylation increases 3.6- to 5-fold [30],[31] at INO1. If expansions are sensitive to local histone acetylation, then integration of the TNR reporter at INO1 should give an enhanced sin3 phenotype, i.e. show greater suppression of expansions. Similarly, there should be less sin3 phenotype on expansions at a “cold” zone like SPS2 whose expression and histone acetylation is nearly unaffected in a sin3 mutant [20],[29],[30]. The results indicate otherwise (Figure 3A). For both integration sites, hot and cold, the effect of sin3 on expansions was similar (6.4-fold suppression at INO1, 5.7-fold at SPS2). Nearly identical suppression effects were seen when the reporter was integrated at another relatively cold locus, LYS2 (8.8-fold; Figure 1C, left panel), or at another hot zone locus, IME2 (8.8-fold; unpublished data).
Confirmation studies of chromatin acetylation at the TNR locus led to an unanticipated result. Chromatin immunoprecipitation (ChIP) was used to evaluate pan-acetylation of histone H4 compared to total H4 at INO1, SPS2, and the TNR reporter (Figure 3B and C). H4 acetylation at INO1 was increased 3- to 5-fold in the sin3 mutant as expected for a hot zone, while H4 acetylation at SPS2 was low in both the wild type and sin3 strains, typical of a cold zone. These findings are independent of the integration site of the TNR reporter (compare Figure 3B and 3C), indicating that insertion of the reporter does not alter acetylation levels at either integration locus. Unexpectedly, we found that histones near the TNR are hyperacetylated, regardless of SIN3 status, to about the same level as INO1 in the sin3 mutant (Figure 3B and C). Hyperacetylation seems to be conferred in part by the trinucleotide repeat, because a control reporter with a randomized sequence in lieu of the TNR yielded a greater dependence of histone acetylation on SIN3 status (“Rand,” Figure 3B). Although the TNR is not uniquely responsible for hyperacetylation of nearby histones (Figure S7), it does contribute.
We concluded from the results in Figure 3A–C that HDACs most likely promote expansions in trans, perhaps by controlling the expression or stability of factors that expand the TNR. The nuclease Sae2 was investigated because a recent study showed Sae2 is stabilized by deacetylation in an Rpd3- and Hda1-dependent manner [32]. Furthermore, Sae2, along with the Mre11/Rad50/Xrs2 complex, is known to process hairpin DNA in vivo and in vitro [33],[34]. Since TNR expansions are thought to involve structured intermediates such as a hairpin [2]–[4], we tested the idea that an sae2 mutant would suppress expansions. The sae2 mutant partially suppressed expansions when compared side-by-side with a sin3 mutant (Figure 3D), consistent with the idea that Sae2 is one (but not the sole) relevant target of Rpd3. Mutation of the nuclease encoded by MRE11 suppressed expansions as much as the sin3 mutant (Figure 3D). Although Rpd3 is not known to directly regulate Mre11, the expansion phenotype of the mre11 mutant is consistent with the possibility that HDACs stabilize Sae2, which then works together with Mre11 to promote expansions. In support of this idea, the expansion phenotype of the sin3 mre11 double mutant was indistinguishable from those of the sin3 and mre11 single mutants (Figure 3D). In contrast, loss of the Exo1 exonuclease showed no effect on expansions, and the sae2 exo1 double mutant was no more defective than the sae2 single mutant (Figure 3D). Together, the results of Figure 3 suggest that yeast Rpd3L and Hda1 promote expansions in trans through the nucleases Sae2 and Mre11.
This study reveals that yeast Rpd3L and Hda1 and human HDAC3 promote expansions of threshold-length triplet repeats in budding yeast and human astrocytes. Interfering with HDAC function through mutation, RNAi knockdown, or small molecule inhibitors eliminates most expansions. It is striking that yeast Rpd3 and Hda1 elicit opposite effects on genetic stability depending on the genomic context; these HDACs accelerate mutagenesis at triplet repeats, whereas they favor chromosome stability via the DNA damage response and processing of double strand breaks [32]. We also found that the human HATs encoded by CBP and p300 have the contravening effect of stabilizing triplet repeats. The latter finding complements an earlier report that CBP modulates instability of long repeats in Drosophila [22]. The relevant yeast HAT remains to be identified. The identification of HDACs as promoting factors and the protective action of HATs emphasizes the importance of protein acetylation/deacetylation to expansions. The mechanistic and therapeutic implications of these findings are considered below.
As in double strand break processing [32], one downstream target of Rpd3L and Hda1 is likely to be the nuclease Sae2. We propose a model where Rpd3L and Hda1 positively regulate Sae2 by stabilizing it. Sae2 and Mre11 then function together as nucleases to promote expansions (Figure 3E). This model is based in part on the study of Robert et al., who found that acetylated Sae2 is degraded by autophagy, but that Sae2 is stabilized by deacetylation in an Rpd3- and Hda1-dependent manner [32]. Also consistent with the Robert et al. work, we infer that Sae2 is not the only relevant target of these HDACs because the expansion phenotype of a sae2 mutant is not as strong as for sin3 (Figure 3D). Other factors, currently unknown, are also proposed to be regulated by Rpd3 and Hda1 and to contribute to expansions by mechanisms that remain to be elucidated (Figure 3E). Sae2 and Mre11 (acting in the Mre11/Rad50/Xrs2 complex) are known to process hairpin DNA in vivo and in vitro [33],[34]. It remains to be determined whether these enzymes actually process a TNR hairpin intermediate to accelerate expansions. The effects of Sae2 and Mre11 have also been examined for expansions of long (CAG)70 repeats [19]. In this study, expansion frequencies increased in sae2 or mre11 mutants. One likely explanation is that long alleles in yeast break more frequently than do the shorter alleles we utilize; thus, long repeats in yeast rely on double strand break repair to prevent expansions [19]. In support of this possibility, expansions of (CAG)70 are also enhanced by loss of the recombination proteins Rad51 and Rad52 [19], whereas rad51 or rad52 mutants do not affect expansion rates of CTG alleles between 13 and 25 repeats [35],[36]. The outcomes of Sae2 and Mre11 activity could be different in break repair than in putative hairpin processing described above.
We found that yeast HDAC mutants suppress expansions in nearly all assays (Figure 1C), but quantitative differences in phenotype illustrate that some aspects of HDAC regulation of expansions remain unknown. What other factors regulated by yeast Rpd3L and Hda1 or human HDAC3 might contribute to expansions? One possibility is chromatin structure near but not immediately adjacent to the repeat. The triplet repeat literature contains several connections between expansions and proteins that modulate chromatin structure, including Drosophila CBP [22] mentioned above, the insulator protein CTCF [37],[38], and the DNA methyltransferase Dnmt1 [39]. A second possibility is that HDACs promote expansions by controlling the firing of DNA replication origins [40]–[43]. The major finding against the origin firing model is that similar SIN3-dependent promotion of expansions was seen when our yeast reporter was integrated at four different loci (LYS2, INO1, SPS2, and IME2; Figures 1 and [3]), which are 21–130 kb away from the nearest origins that become deregulated in rpd3Δ cells [42]. We feel it is unlikely that Rpd3-dependent origin firing explains suppression of expansions, although HDAC effects on fork progression or fork stalling cannot be ruled out at this time.
HDAC inhibitors are currently being evaluated as therapies to treat the transcriptional defects in several TNR expansion diseases [44],[45]. For example, 4b treatment reverses FXN gene silencing in primary cells from Friedreich's ataxia patients [24] and relieves disease phenotype and transcriptional abnormalities in HD transgenic mice [25]. Our work implies these inhibitors may have a second, beneficial effect of suppressing somatic expansions that contribute to disease progression.
Triplet repeat expansion assays using the URA3 reporter have been described previously [18],[27]. Assays using the CAN1 reporter (Figure 1A) utilized canavanine at 60 µg/ml to select for resistance. All expansions were verified by single-colony PCR across the repeat tract followed by analysis on high-resolution polyacrylamide gels [18]. Details of statistical analysis are provided in Tables S1 and S4.
Whole cell lysates (yeast and SVG-A astrocytes) or histone acid extracts (SVG-A astrocytes) were separated electrophoretically and transferred to PVDF membranes. Primary rabbit antibodies were against histone H3 (A300-823A, Bethyl Laboratories), acetyl-histone H3 (#17-615, Millipore), acetyl-histone H4 (#06-866, Millipore), β-actin (A2066, Sigma-Aldrich), HDAC3 (sc-11417, Santa Cruz Biotechnology), and HDAC1 (CH00218, Coriell Institute for Medical Research). Assessment of HDAC3 expression via Western blot analysis resulted in two bands around 50 kDa, the predicted size of the protein, presumably representing the two reported isoforms of HDAC3 [46]. Throughout all experiments, consistent knockdown of the top band was observed following HDAC3 siRNA treatment, however levels of the bottom band varied between experiments. Quantitation of HDAC3 knockdown was performed by densitometric analysis of the top band only. A mouse antibody was used against histone H4 (ab31830, Abcam). Secondary antibodies conjugated to horseradish peroxidase were 711-035-152 and 115-035-003 from Jackson ImmunoResearch Laboratories. Visualization was by chemilluminescence (Western Lightning Plus-ECL, PerkinElmer).
250 ml yeast cell cultures were grown to A600∼0.8 at 30° in yeast extract/peptone/dextrose. Following cross-linking with 1% formaldehyde (15 min, 22°), cross-linked chromatin was isolated in lysis buffer containing 50 mM HEPES/KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100%, 0.1% sodium deoxycholate and the protease inhibitors 1 mM PMSF, 1 mM benzamidine, 1 µg/ml leupeptin, and 1 µg/ml pepstatin. After sonication (40% duty cycle for seven cycles of 5 s each with 50 s cooling in between; Digital Sonifier EDP 100-214-239, Branson), chromatin fragments were immunoprecipitated with antibodies specific for total histone H4 (5 µg, A300-646A, Bethyl Laboratories) or pan-acetylated H4 (7 µl, # 06-866, Millipore) at 4°C overnight. Immune complexes were captured by incubating with Protein G magnetic beads (S1430S, New England BioLabs) for 4 h at 4°C. After a series of washes, DNA was eluted in 250 µl elution buffer (50 mM Tris-HCl pH 8, 10 mM EDTA, and 1% SDS) and crosslinks were reversed by incubating overnight at 65°C. DNA was purified by phenol-chloroform extraction followed by an ethanol precipitation and analyzed by quantitative PCR (Applied Biosystems, 7500 FAST). Primer sequences used for quantitative PCR are provided in the Supporting Information section. Signals for total H4 and acetylated H4 were quantified by the method of 2−ΔΔCt and normalized using the following calculation: (Ct immunoprecipitate−Ct input)−(Ct background−Ct input). Amplification of the chromosome VI telomere region was chosen as a measurement for background [31],[47]. The normalized IP values obtained for acetylated H4 were divided by the normalized IP values for total H4.
Cells were grown to mid-log phase and then extracted with hot acidic phenol. Following clean-up of the RNA, reverse transcription was performed in triplicate. cDNA levels were analyzed in triplicate by quantitative real-time PCR and normalized to ALG9 levels. Details and primer sequences are provided in Table S2.
SVG-A astrocytes were seeded in 60 mm tissue culture dishes and transfected with 5 µg shuttle vector DNA using Lipofectamine 2000 (Invitrogen Corporation). After 6 h, the DMEM transfection media was replaced by DMEM supplement with 10% fetal bovine serum, plus one of the HDAC inhibitors 4b or compound 3 (kindly provided by Joel Gottesfeld, The Scripps Research Institute) or DMSO only. Cells were incubated for an additional 48 hours, then samples were taken for either expansion assay or histone analysis. To measure expansions, plasmid DNA was extracted and concentrated by using Hirt's alkaline lysis [48] and Amicon Ultra 50 K centrifugal filter units (Millipore). Purified plasmid DNA was digested by DpnI (New England Biolabs) and then transformed into S. cerevisiae for measurement of canavanine resistance or into E. coli for analysis of total plasmid numbers as measured by ampicillin-resistant colonies. Histone extracts were prepared by acid extraction (protocol provided by Abcam).
RNA interference experiments were performed with minor variations. SVG-A cells were seeded and transfected with ON-TARGET plus or siGenome SMARTpool siRNAs (100 nM) against HDAC3 (L-003496, M-003496), HDAC1 (M-003493), or scrambled non-targeting siRNA (D-001810) from Dharmacon using DharmaFECT 1. After 48 h, cells were transfected with 7 µg of shuttle vector and also re-transfected with siRNAs using Lipofectamine 2000. After another 2 d, expansion frequencies were prepared as above, in parallel with immunoblot analysis of whole cell lysates.
All p values were determined by two-tailed Student's t test. p and n values for each data set are specified in Tables S1, S2, S3, S4 unless stated in the figure legend.
|
10.1371/journal.pntd.0005087 | Minimally Symptomatic Infection in an Ebola ‘Hotspot’: A Cross-Sectional Serosurvey | Evidence for minimally symptomatic Ebola virus (EBOV) infection is limited. During the 2013–16 outbreak in West Africa, it was not considered epidemiologically relevant to published models or projections of intervention effects. In order to improve our understanding of the transmission dynamics of EBOV in humans, we investigated the occurrence of minimally symptomatic EBOV infection in quarantined contacts of reported Ebola virus disease cases in a recognized ‘hotspot.’
We conducted a cross-sectional serosurvey in Sukudu, Kono District, Sierra Leone, from October 2015 to January 2016. A blood sample was collected from 187 study participants, 132 negative controls (individuals with a low likelihood of previous exposure to Ebola virus), and 30 positive controls (Ebola virus disease survivors). IgG responses to Ebola glycoprotein and nucleoprotein were measured using Alpha Diagnostic International ELISA kits with plasma diluted at 1:200. Optical density was read at 450 nm (subtracting OD at 630nm to normalize well background) on a ChroMate 4300 microplate reader. A cutoff of 4.7 U/mL for the anti-GP ELISA yielded 96.7% sensitivity and 97.7% specificity in distinguishing positive and negative controls. We identified 14 seropositive individuals not known to have had Ebola virus disease. Two of the 14 seropositive individuals reported only fever during quarantine while the remaining 12 denied any signs or symptoms during quarantine.
By using ELISA to measure Zaire Ebola virus antibody concentrations, we identified a significant number of individuals with previously undetected EBOV infection in a ‘hotspot’ village in Sierra Leone, approximately one year after the village outbreak. The findings provide further evidence that Ebola, like many other viral infections, presents with a spectrum of clinical manifestations, including minimally symptomatic infection. These data also suggest that a significant portion of Ebola transmission events may have gone undetected during the outbreak. Further studies are needed to understand the potential risk of transmission and clinical sequelae in individuals with previously undetected EBOV infection.
| With over 28,000 reported cases, the 2013–16 West African Ebola virus disease epidemic is the largest and longest on record. This study provides further evidence that Ebola, like other viruses, causes a spectrum of clinical manifestations that may include minimally symptomatic infection. The findings also suggest that many episodes of human-to-human transmission of Ebola virus in West Africa may have gone undetected in the recent outbreak. This has implications for the definition of Ebola virus disease survivorship, delineation of transmission chains, and future vaccine studies.
| Despite over 28,000 reported cases of Ebola virus disease (EVD) in the 2013–16 pandemic as of March 27, 2016 [1], we are only beginning to trace the complex biosocial processes that have promoted spread of the virus [2,3]. Important questions remain, including how to best use tools such as new vaccines [4] and rapid diagnostic tests [5] to contain future outbreaks, the extent to which symptomatic individuals do not present for care, how to identify and manage clinical sequelae of EVD [6], and the incidence and transmission dynamics of minimally symptomatic Ebola virus (EBOV) infection.
Evidence for minimally symptomatic EBOV infection is limited. During the 2013–16 outbreak in West Africa, it was not considered epidemiologically relevant to published models or projections of intervention effects [7–10]. Moreover, it is not known if clinical sequelae seen in survivors of EVD (e.g., uveitis) exist in individuals who had minimally symptomatic EBOV infection.
In order to improve our understanding of the transmission dynamics of EBOV in humans, we investigated the occurrence of minimally symptomatic EBOV infection in a recognized Ebola ‘hotspot’ [11], which we defined as an area with a reverse transcription polymerase chain reaction (RT-PCR)-confirmed EVD attack rate above 2% in a two-month period.
To measure the occurrence of EBOV infection in a group of people exposed to confirmed cases, we validated a commercial ELISA kit in a cohort of patients with known Ebola status and then used this tool to determine infection status in previously quarantined individuals.
The study protocol was approved by the Sierra Leone Ethics and Scientific Review Committee and the Stanford University Institutional Review Board (Protocol ID: 33882). We held meetings with Paramount, Sectional, and Town Chiefs to discuss the proposal and supply them with written information. We then held town meetings to describe the project and answer community questions. Individuals provided written informed consent or placed a thumbprint after hearing a consent script read in the Krio or Kono languages (both ethics committees approved the use of an oral script and thumbprint for those participants who could not read or write); parents signed for children under the age of 18, and children and adolescents provided verbal assent. Subjects received 50,000 Leones (~$10 US) for their participation in the study.
Kono is a diamond-rich district in eastern Sierra Leone which reported 301 cases of EVD (that is, RT-PCR confirmed individuals who presented for isolation or were identified by surveillance teams) between August 2014 and February 2015 (Kono District Ebola Response Centre (DERC), personal communication). Several of the authors worked closely with the Kono DERC, British military, and International Federation of the Red Cross on containment, isolation, and surveillance during the time of active transmission in the district. Parts I and II of this study were conducted from October 2015 to January 2016 by a team consisting of two physicians, a physician-anthropologist, a laboratory technician, and two community health workers.
One year after the peak of the Ebola epidemic in Sierra Leone, we identified positive IgG responses to Zaire-EBOV in 14 individuals not known to have had EBOV infection from a village classified as a hotspot for Ebola transmission. Of these, the majority reported having had no symptoms consistent with EVD during the epidemic while two reported only fever. Thus, we provide further evidence that Ebola, like many other viral infections, presents with a spectrum of clinical manifestations, including minimally symptomatic infection. In addition, our data suggest that a significant portion of Ebola transmission events may have gone undetected during the epidemic.
Although it was difficult to verify symptoms through our retrospective interviews (given the considerable denial of EVD during the outbreak due to stigma and the fear of being admitted to an ETU where those admitted were seldom discharged), our data indicate that 25% of EBOV infections may have been minimally symptomatic, which is similar to the data from empiric studies and to recent modeled estimates [14].
The phenomenon of previously undetected, minimally symptomatic EBOV infection was evident around the discovery of the virus in 1976. Using an immunofluorescence assay, the World Health Organization/International Study Team found that 19% of contacts of EVD cases—very few of whom gave any history of illness—had antibodies to the virus [15]. In 2000, Leroy and colleagues published a study (based on ELISA/Western blot) and found that of 24 asymptomatic close contacts of Gabonese patients with EVD, 11 developed both IgM and IgG responses to Ebola Zaire antigens, indicating viral infection [16]. Other investigators have found evidence of seropositive individuals in areas without large outbreaks using ELISA and postulate that there may have been active circulation of filovirus without apparent clinical manifestations [17,18]. Heffernan and colleagues also used ELISA in Gabon and found that 1% of individuals in an epidemic zone had IgG antibodies to Ebola Zaire virus, yet no history of exposure [19]. In another study in Gabon, Becquart and colleagues found a 15.3% Ebola Zaire IgG seroprevalence in 220 randomly selected villages and concluded that most of the seropositive persons identified “probably had mild or asymptomatic infection” [20]; however, they used uninfected individuals in France as negative controls. We found that unexposed expatriates (not included as negative controls) had a significantly lower mean log anti-GP (M = 5.25 U/mL, SD = 0.54, N = 12) than unexposed Sierra Leoneans (M = 6.40 U/mL, SD = 1.06, N = 132) using the two-sample t-test for unequal variances, t(20.05) = 6.40, P< = 0.001 (see Fig 3), potentially due to cross-reactivity of our assay with closely related pathogens circulating in the region.
Our study has several possible limitations. First, if asymptomatic EBOV infection is a common occurrence, it is possible that some of the “true negatives” used for the validation of our ELISA assays could have been infected with EBOV. We did not use a microneutralization assay for comparison purposes, nor did our pool of EVD survivors represent all of the possible antibody titers in the regional population. Although the serologic assay conferred sensitivity and specificity greater than 95%, the assay protocol produced antibody titers that were considered qualitative in nature. Second, we relied on study participants and their household contacts’ memories of events that had taken place up to a year prior when we classified them as symptomatic or asymptomatic (potential recall bias); however, during the outbreak, most quarantined households were monitored daily by surveillance teams that conducted symptom screens and measured temperatures, and individuals were brought to treatment facilities for EBOV testing if they screened positive. Third, our IgG assays indicate previous infection but provide no information on when that infection took place. The 3 negative controls with positive tests either represent false positives due to cross-reactive antibodies or previously infected individuals. (We did not perform IgM ELISAs, as other investigators have demonstrated that Ebola IgM titers largely diminish within 60 to 90 days of symptom onset [21,22].) Fourth, we did not ascertain whether there were non-quarantined individuals who had minimally symptomatic infection.
Our study focused on the quarantined population of one village. Extrapolation of our findings to other villages and generalizability to the epidemic should be approached with caution. Although many of the study participants were followed by surveillance teams during the time of active EBOV transmission in their village, the timing of the possible infection cannot be known from IgG data, and it is improbable that surveillance teams followed each individual throughout the duration of the outbreak in Sierra Leone. Thus, the concern that IgG positive persons indeed had but denied symptoms cannot be excluded.
Despite these limitations, our serosurvey provides a deeper perspective on EBOV transmission, and more village-level serosurveys could enhance our understanding of undetected EBOV transmission at the epidemic level. Furthermore, our findings suggest there would be value in exploring the interaction of seropositive persons and EVD cases to improve our understanding of exposure risk. As a result, we may learn more about how efforts at containment can be improved. The data also have important implications for future vaccine studies that rely on detecting antibody to EBOV. Lastly, the findings support the World Health Organization’s interim guidance on clinical care for survivors of EVD, which defines a survivor as a person:
There is ongoing discussion in West Africa over the definition of survivorship, usually specified by having a positive EBOV RT-PCR result and discharge from an ETU. Plans are underway to develop national registries and provide ID cards to such survivors, so as to delineate individuals eligible for free social and medical support services. Should the notion of survivorship be extended to all those who are IgG positive, including those who had minimally symptomatic infection or who were sick but were never tested at the time of illness? How we define a community of suffering is always problematic and should be revisited given that this definition has implications for identity, stigma, and access to social and medical services.
In conclusion, by using ELISA to measure Zaire-EBOV antibody concentrations, we identified a significant number of individuals with previously undetected minimally symptomatic EBOV infection in a ‘hotspot’ village in Sierra Leone, approximately one year after the village outbreak. Further studies are needed to understand the potential risk of transmission and clinical sequelae in individuals with minimally symptomatic EBOV infection.
|
10.1371/journal.ppat.1002782 | Highly Efficient Prion Transmission by Blood Transfusion | It is now clearly established that the transfusion of blood from variant CJD (v-CJD) infected individuals can transmit the disease. Since the number of asymptomatic infected donors remains unresolved, inter-individual v-CJD transmission through blood and blood derived products is a major public health concern. Current risk assessments for transmission of v-CJD by blood and blood derived products by transfusion rely on infectious titers measured in rodent models of Transmissible Spongiform Encephalopathies (TSE) using intra-cerebral (IC) inoculation of blood components. To address the biological relevance of this approach, we compared the efficiency of TSE transmission by blood and blood components when administrated either through transfusion in sheep or by intra-cerebral inoculation (IC) in transgenic mice (tg338) over-expressing ovine PrP. Transfusion of 200 µL of blood from asymptomatic infected donor sheep transmitted prion disease with 100% efficiency thereby displaying greater virulence than the transfusion of 200 mL of normal blood spiked with brain homogenate material containing 103ID50 as measured by intracerebral inoculation of tg338 mice (ID50 IC in tg338). This was consistent with a whole blood titer greater than 103.6 ID50 IC in tg338 per mL. However, when the same blood samples were assayed by IC inoculation into tg338 the infectious titers were less than 32 ID per mL. Whereas the transfusion of crude plasma to sheep transmitted the disease with limited efficacy, White Blood Cells (WBC) displayed a similar ability to whole blood to infect recipients. Strikingly, fixation of WBC with paraformaldehyde did not affect the infectivity titer as measured in tg338 but dramatically impaired disease transmission by transfusion in sheep. These results demonstrate that TSE transmission by blood transfusion can be highly efficient and that this efficiency is more dependent on the viability of transfused cells than the level of infectivity measured by IC inoculation.
| In the UK, several v-CJD cases have been identified in patients that received blood or blood-derived products prepared from incubating asymptomatic donors. Since there is no screening test to identify infected donors, procedural risk reduction measures remain the only protection against v-CJD transfusion risk. These measures rely, in part, on the assumptions that (i) the level of infectivity in blood is low and (ii) the risk of blood borne transmission is directly correlated with the infectious titer of blood and blood products. Using a transmissible spongiform encephalopathy (TSE) animal model, we have provided evidence that despite a very low infectious titer in blood as measured by inoculation into brain, the transfusion of 0.2 mL of blood from asymptomatic infected donors is sufficient to transmit the disease with a 100% efficacy. We further demonstrated that this high efficiency of disease transmission is crucially dependant on the viability of the transfused white blood cells rather than on their infectious titer. These findings provide new insights into the pathogenesis of TSE diseases and require revision of some of the key assumptions of the v-CJD blood borne risk assessments.
| In 1996, a new form of CJD, named variant CJD (v-CJD) was identified in humans [1], [2]. The presence of the v-CJD agent in lymphoid tissues of patients incubating the disease raised the possibility of blood borne v-CJD transmission. Since then, four v-CJD cases were reported in patients transfused with labile blood products from asymptomatic donors who later developed v-CJD [3].
Current measures for prevention of v-CJD transmission by blood products rely on risk assessments based on infectious titers reported for the blood of rodent models of TSE [4]. These infectious titers were established by inoculating blood components by the intracerebral (IC) route in homologous species. All these experiments concur in indicating that infectivity in the blood of symptomatic rodents varies between 1 and 10 ID50 per mL and that both plasma and leukocytes were infectious [5], [6], [7], [8], [9], [10].
However, the IC inoculation route is unlikely to reflect the specifics of the transfusion exposure route; i.e. the administration of large numbers of living cells and a large volume intravenously to the recipient.
In early 2000, transmission of both experimental BSE and natural scrapie were reported to occur following transfusion of whole blood collected in asymptomatic incubating sheep [11], [12]. The greater similarity in size between humans and sheep allows the transfusion of blood volumes in sheep that are closer to those used in human medicine. Moreover, the pathogenesis of variant CJD in humans is very similar to that in classical scrapie or BSE in sheep. As a consequence sheep TSE models are accepted as relevant for assessing v-CJD transmission risk through the transfusion route [13], [14].
In this study, using a TSE sheep infection model, we investigated the ability of blood and blood components to transmit the disease when administered by the transfusion route and measured their relative infectious titer by IC inoculation in transgenic mice (tg338) over-expressing ovine PrP. Our findings show that the disease is transmitted with high efficiency by transfusion of infected blood and blood components and requires revision of some of the key assumptions used for assessing the risk of blood borne transmission of v-CJD.
All animal experiments have been performed in compliance with our institutional and French national guidelines, in accordance with the European Community Council Directive 86/609/EEC. The experimental protocol was approved by the INRA Toulouse/ENVT ethics committee.
VRQ/VRQ sheep were produced in a biosecure unit from a flock that was sourced originally from New Zealand by Veterinary Laboratory Agency in the United Kingdom (UK). This TSE free flock is a unique source of VRQ/VRQ animals that are considered free of classical scrapie [15]. The scrapie inoculum was prepared from symptomatic VRQ/VRQ sheep experimentally infected with PG127 scrapie isolate. This stock was previously titrated by end-point dilution in tg338 mice inoculated by the intracerebral (IC) route [16]. TSE-free Cheviot sheep (6–10 months old) were orally challenged with a 2 g equivalent of brain (administered as 1% brain homogenate in glucose). Animals were then observed until the occurrence of clinical signs (around 200 days post inoculation) and euthanized when exhibiting locomotor signs of the disease that impaired feeding. In this model clinical phase (first clinical signs occurrence to recumbency) is short (one to two weeks). After culling, each sheep was necropsied and lymphoid tissues (spleen, third eye-lid, mesenteric lymph node, prescapular lymph node, tonsil) were sampled in addition to the CNS for PrPSc detection. The PrP genotype was obtained for every animal by sequencing Exon 3 of the Prnp gene as previously described [17].
Blood was collected in vigil animals from jugular vein into Macopharma MSE3500 collection bags. To separate plasma, buffy coat and red cells, blood units were centrifuged at 3640 rpm for 12 minutes (including acceleration and deceleration) at 20°C in a Jouan LR5-22 centrifuge. For each sample, packed cell volume and cell numeration were measured using IdeXX Vet-station automat and Sysmex XT2000i. White blood cells (WBC) were obtained from buffy coat fraction, by osmotic lysis of erythrocytes with ACK solution (NH4CL 0,15 M, KHCO3 1 mM, Na2EDTA 0,1 mM, pH 7.4). Briefly, buffy coat (1 volume) was gently mixed (by rolling tubes) with ACK (1 volume) and incubated for 5 min RT. WBC were then washed 3 times with 50 mL PBS to remove erythrocytes debris and plasma.
WBC fixation was performed using paraformaldehyde (PFA)/PBS solution. WBC cells were re-suspended in a 50 mL conic tube using 10 mL of PBS and 10 mL of a PBS/PFA 4% solution was added (final concentration 2%). Tubes were then rolled on bench during 20 seconds and incubated for 10 min on ice. Cells were then washed four times in PBS (50 mL) and their morphology (lack of clumps and cellular integrity) was checked by direct microscopic examination and flow cytometry.
Recipient sheep were anesthetized by IV administration of Ketamine/Diazepam (2 mg and 0.1 mg per kg respectively) and a catheter (16G) was installed in the jugular vein. This catheter was used for all transfusions and intravenous administrations of blood and blood products. Whole blood (200 mL), plasma (100–150 mL) or WBC (re-suspended in 100 mL 5% glucose) were transfused/intravenously administered by gravity in individual recipients over a 30–40 min period.
In the whole blood spiking experiment, the titrated PG127 brain homogenate was diluted in isotonic (5%) glucose solution to obtain the desired number of Infectious Dose 50 as measured by IC inoculation in tg338 mice (ID50 IC in tg338) in a 20 mL volume. Each inoculum was then mixed with 200 mL of freshly collected whole blood by 30 minutes of continuous rocking before transfusion back into the same donor sheep. In the whole blood titration by transfusion route, the required whole blood volumes were sampled from the collection bag using a sterile syringe and either transferred into a new pouch or gently re-suspended in a sterile 5% glucose solution (QSP 200 mL), before administration (gravity) to recipients. In the WBC experiment, WBC (fresh or fixed) were gently re-suspended in a 200 mL sterile pouch of isotonic (5%) glucose before infusion (gravity) over a 20–30 min period. All transfusions were performed within the 6 hours following blood collection.
Sheep were housed in a category 2 animal facility. Donor and recipient sheep were kept separated. Contact controls (n = 5) were housed in the same pen as transfusion recipients throughout the experiment. None displayed any TSE symptoms. Orally inoculated sheep (donors) and transfused sheep (recipients) were monitored daily. When displaying symptoms that could impair their ability to feed, the animals were euthanized by an intravenous injection of T61 (Intervet). Transfused animals that had not displayed clinical signs after 400–450 days were also euthanized. Upon completion of the experiment, the contact controls were also euthanized. All animals were systematically necropsied and a panel of tissues (brain, spleen, mesenteric lymph node, tonsil, third eyelid, prescapular lymph node, ileum was collected. Each sample was divided in two parts, the first one was formalin fixed for abnormal PrP (PrPSc) immunohistochemistry investigation while the other was snap frozen at −80°C for WB PrPSc detection.
Bioassay experiments were carried out in transgenic mice (tg338) over-expressing the VRQ allele of the ovine prion protein (PrP) [18]. At least six mice were inoculated intracerebrally with each sample (20 µL). Mice were then monitored for the appearance of TSE clinical signs. Mice that did not succumb to TSE were euthanized at 200 days post inoculation (dpi). The last positive transmission was observed 150 dpi during an end-point dilution titration of a PG127 brain stock [16]. CNS and spleen samples collected from each individual animal were tested by Western blot (WB).
Whole blood, plasma and red blood cell concentrate were high speed homogenized for 90 seconds (Precess 48- Bertin, France) and inoculated neat (20 µl IC) in tg338 (n = 18). No acute toxicity was observed at inoculation during this experiment. The infectious titer per mL of blood (number of ID/mL) was estimated by the limiting dilution titration (application of Poisson model) described by Brown et al to estimate the infectious titer in TSE affected rodents' blood and classically used in this type of study [5], [7]. Percentage volume of packed blood cells and plasma were used to calculate the infectious titer per mL of blood.
WBC corresponding to known quantity of whole blood was high speed homogenized for 90 seconds (Precess 48- Bertin, France) in 5% glucose solution. The resulting suspension was then titrated by end point dilution by intracerebral (IC) inoculation of 1/10 successive dilutions into groups of 6 tg338 mice (20 µL/mouse). The infectious titer (Infectious Dose 50, most likely value and 95% confidence intervals) were determined by the Spearman-Kärber method [19].
According to Fischer [20] and as previously used for Prion infectivity ID/ID50 comparison [21] one ID50 per mL was considered in average equivalent to 0.693 ID.
A Western blot kit (TeSeE Western Blot, Bio-Rad) was used following the manufacturer's recommendations. For each sample, 250 µL of 10% tissue homogenate was subjected to PrPSc extraction. Immunodetection was performed using Sha31 monoclonal antibody conjugated to horseradish peroxidase (0,06 µg per mL) which recognizes amino acids 145–152 (YEDRYYRE) of the sheep PrP sequence. Peroxidase activity was detected using ECL substrate (Pierce). Immunohistochemistry was performed as previously described [22], using 8G8 antibody which specifically recognises amino acids 95–108 (SQWNKP) of the sheep PrP protein.
The first experiment was to determine the minimum infectious dose (ID50 as measured by end point dilution titration in tg338 mice by the IC route) able to transmit TSE infection in sheep by transfusion. This was done by transfusing groups of two TSE free VRQ/VRQ recipient sheep with whole blood that had been spiked with successive 1/10 dilutions of a brain homogenate prepared from scrapie affected sheep (PG127 isolate). This brain homogenate had been previously titrated by end point dilution in tg338 mice, and had a titer of 106.6 ID50 IC/g [16]. The lowest dose resulting in transmission was observed in sheep (1 infection out of 2 challenged individuals) that were transfused with blood containing 103 ID50 IC in tg338 mice (Table 1). No clinical signs or abnormal PrP deposition were observed in animals that were challenged with lower doses and observed for at least 750 days post challenge.
In parallel, seven VRQ/VRQ sheep were orally challenged with 106.7 ID50 IC in tg338 of the same PG127 brain homogenate stock. Whole blood samples were collected in four of these animals (D1–D4) at a late stage of the incubation period (two to five weeks before clinical onset). 200 mL, 20 mL, 2 mL and 0.2 mL of each donor's fresh whole blood were transfused into TSE free VRQ/VRQ recipients (Table 2). All animals that received 200 mL, 20 mL or 2 mL of blood developed clinical TSE. Three out of the four recipients that received the 0.2 mL dose also developed clinical scrapie. The fourth recipient was apparently healthy when euthanized after 450 days of incubation but had abnormal PrP depositions in both lymphoid tissues and the central nervous system.
These results, when considered together, indicate that the transfusion of 0.2 mL of whole blood transmitted scrapie more efficiently than the IV administration of 103 ID50 IC in tg338 (brain homogenate material). It therefore suggested that in scrapie infected sheep (close to symptomatic phase) the infectious load per mL of blood should be higher than 103.6 ID50 IC in tg338.
To confirm this estimate, whole blood from three donors (D1 to D3) and one TSE free control (C1) were each inoculated intracerebrally into eighteen tg338 mice (20 µL per mouse). An incomplete attack rate was observed in the inoculated mice (Table 3) and infectious titers were estimated by a statistical method previously used to calculate the concentration of infectivity in blood of TSE infected rodents models [5], [23]. Whole blood titers ranged between 3 and 16 ID IC in tg338 mice per mL (Table 3).
Plasma, red blood cells and WBC were prepared from the same blood samples. Plasma and RBC were inoculated undiluted intracerebrally into eighteen tg338 mice. Only a single infection was observed (D1 plasma). The infectivity titers in plasma and red blood cells expressed as the equivalent volume of whole blood were computed to have a 0.95 probability of contributing less than 8.6 and 2.7 ID IC in tg338 per mL of blood respectively.
The infectivity in WBC was measured by end point dilution titration by IC inoculation of ten folds serial dilutions in tg338 mice (Table 4). Titers calculated by the Spearman-Karber method and expressed as the equivalent volume of whole blood ranged between 7 and 24 ID50 IC in tg338 per mL of blood, i.e between 4.5 and 13.4 ID per mL (one ID50/mL being in average equivalent to 0.693 ID/mL).
The infectious titers of whole blood in sheep incubating scrapie were comparable with those previously reported in hamster and mouse models [6], [7], [8], [9], [10]. However, these titers were at least 100 fold lower than those initially suggested by the transfusion experiments.
In an attempt to resolve this discrepancy we investigated the relative abilities of whole blood, plasma and white blood cells to transmit the disease by the transfusion route in sheep. Over 700 mL of blood was collected, few days (D4) to three weeks (D7) before clinical onset, from each of four orally challenged VRQ/VRQ sheep (D4 to D7) (Tables 5, 6). 500 mL of each blood was used to prepare plasma and WBC. The WBCs were divided into two equal parts; the first half was untreated while the other half was fixed with PBS-paraformaldehyde (PFA). Fresh whole blood (200 mL), fresh WBC and PFA fixed WBC (each equivalent to 200 mL of starting blood) and plasma prepared from 200 mL and 20 mL of blood were administrated intravenously to TSE free VRQ/VRQ recipient sheep (Table 5).
Whole blood and fresh WBC produced clinical scrapie in all recipients. The incubation periods of recipient sheep receiving whole blood or WBC were similar suggesting comparable virulence of both products (Table 4). In contrast, the intravenous administration of crude plasma resulted in infection of only three of the four recipient sheep. The incubation periods of the plasma infections were substantially longer than in sheep who received whole blood. None of the sheep that received plasma that had been prepared from 20 mL of whole blood were infected. Donor (D4), whose plasma did not transmit the disease (even when prepared from 200 ml of whole blood), was also part of the IV whole blood titration (Table 2) where 0.2 mL of its whole blood was sufficient to infect the recipient. From this it can be concluded that, in this individual, plasma was at least 10,000 fold less efficient than whole blood for transmitting the disease by the intravenous route.
Whereas all recipients that received fresh WBC succumbed to clinical disease, the IV administration of PFA fixed WBCs resulted in TSE infection of only two of four recipient sheep (Table 5) and in incubation periods that exceeded those observed for plasma recipients. This outcome might suggest that PFA fixation inactivated WBC associated infectivity. However, when inoculated IC into tg338 mice, the same fresh and PFA fixed WBC homogenates produced 100% attack rates with similar incubation periods for both groups (Table 6). The end point titration by IC inoculation in tg338 of PFA fixed and fresh WBC homogenates from two of these donors (D4 and D7) further confirmed that the fixation procedure we applied did not substantially affected the WBC's infectious titer (Table 6).
Using TSE infection in sheep as a model, we have shown that the intravenous administration of a few hundred microliters of blood is sufficient to infect a transfusion recipient. This high efficiency of transmission by blood transfusion contrasts with the very low infectious titer measured by IC inoculation of blood into transgenic mice expressing ovine PrP.
This discrepancy shows that the infectious titer as determined by IC inoculation of rodents does not accurately predict the ability of blood and blood-derived products to transmit the disease when administered by the intravenous route. Importantly, this finding indicates that assessments of transfusion transmission risks from v-CJD that are based on IC measurements in rodents may seriously underestimate the risk.
Our results indicate that freshly prepared WBCs have a comparable capacity to whole blood to infect transfusion recipients. However WBC fixation with PFA strongly inhibited their ability to transmit the disease intravenously. Tissue fixation with aldehydes is classically described to reduce by several log10 the TSE infectivity titer (as measured by IC inoculation in rodents) [24], [25]. However this activity is dependent on the fixation procedure and in particular on its duration. For instance, while the fixation of a 263 K scrapie 10% brain homogenate with 3.7% formaldehyde during 4 hours (at room temperature) resulted in a 1.58 log10 drop in the infectivity, a 30 min fixation had no significant impact on the titer [26]. Our bioassay experiments in tg338 mice, confirmed that the fixation procedure we applied (2% PFA during 10 minutes on ice) preserved TSE infectivity. Beside this, PFA fixation stabilizes cellular morphology through cross linking of the cell surface proteins which also makes it impossible for cells to change their cell membrane molecules and impairs their capacity to develop dynamic cellular interactions [27], [28]. We therefore think that the ability of blood (or blood-derived products) to transmit the disease by the transfusion route crucially depends on the ability of blood cells to interact with the host.
This view is consistent with results obtained in in vitro prion infection models in which cell-mediated infection was reported to be more efficient than infection with cell homogenates or PFA fixed cells [29], [30]. It also concurs with the previously reported ability of living splenic cells prepared from scrapie infected C57Bl6 mice to transmit scrapie in RAG-10/0 mice following IV administration whereas no transmission could be observed following the IV administration of splenic cell homogenates to the same mouse model [31]. Whether the high capability of viable WBC to infect blood recipients relies on the capacity of certain cellular populations to deliver TSE agent to specific target site(s) will require further investigation.
In the PG127 scrapie sheep model, plasma was less able than whole blood and fresh WBCs to transmit the disease by the transfusion route. This observation is consistent with data obtained from similar transfusion experiments carried out in BSE infected sheep and in Chronic Wasting Disease infected cervids [32], [33]. The low transmissibility of the disease by the plasma in these animal models contrasts with the results obtained in rodents where plasma contains 40 to 60% of the infectivity in whole blood [7], [8], [34]. Together, with the infectivity measurements that we performed, these data support the contention that the partitioning of TSE agents in blood components differs between sheep and rodents models. However, the differences of infectivity load between plasma and WBC that we observed are unlikely to explain on their own the limited efficacy of plasma to transmit the disease (by comparison to whole blood and WBC). This statement is strongly supported by the WBC fixation experiment that we performed in which the PFA treatment did not alter WBCs infectious titer but reduced their capacity to transmit the disease to recipients at level which is comparable to plasma.
In this study, labile blood products containing viable WBC presented the greatest risk of transmitting Prion disease. This finding strongly supports the continuation of universal leuco-reduction as currently applied in the EU countries and Canada to reduce, amongst other potential adverse effects, the risk of v-CJD transmission [34]. Additional experiments will be necessary to determine the minimal number of WBC (leukocytes and/or platelets) that is sufficient to transmit the disease and to identify the WBC cell population(s) responsible for virulence.
Finally, our results also raise some concerns about the use of the ‘spiking’ models for investigation of blood-borne TSE transmission risk [35], [36]. Whereas this approach is very convenient to measure the TSE infectivity reduction by certain process, it is probably of limited relevance for assessing the efficacy of devices intended to mitigate the risk of Prion disease transmission by blood and blood derived products.
|
10.1371/journal.ppat.1000431 | B7-H1 Blockade Increases Survival of Dysfunctional CD8+ T Cells and Confers Protection against Leishmania donovani Infections | Experimental visceral leishmaniasis (VL) represents an exquisite model to study CD8+ T cell responses in a context of chronic inflammation and antigen persistence, since it is characterized by chronic infection in the spleen and CD8+ T cells are required for the development of protective immunity. However, antigen-specific CD8+ T cell responses in VL have so far not been studied, due to the absence of any defined Leishmania-specific CD8+ T cell epitopes. In this study, transgenic Leishmania donovani parasites expressing ovalbumin were used to characterize the development, function, and fate of Leishmania-specific CD8+ T cell responses. Here we show that L. donovani parasites evade CD8+ T cell responses by limiting their expansion and inducing functional exhaustion and cell death. Dysfunctional CD8+ T cells could be partially rescued by in vivo B7-H1 blockade, which increased CD8+ T cell survival but failed to restore cytokine production. Nevertheless, B7-H1 blockade significantly reduced the splenic parasite burden. These findings could be exploited for the design of new strategies for immunotherapeutic interventions against VL.
| The protozoan parasite Leishmania donovani is the cause of visceral leishmaniasis, a chronic disease that currently affects 12 million people worldwide. We are interested in understanding the immune mechanisms that can control infection. Preliminary studies suggested that CD8+ T cells can kill parasites and limit disease; however, studying these important killer cells has been hindered, because we do not know what parasite molecules they recognize. To overcome this, we engineered parasites to express ovalbumin. Since many tools exist to track and measure immune cells targeted at ovalbumin, we can now track the specific CD8+ T cell responses that develop upon infection with Leishmania. We found that Leishmania initially induced CD8+ T cells to divide and produce molecules such as IFN-gamma that may help them to kill parasites. However, the CD8+ T cells rapidly lost their effector function and died off as infection progressed. More encouragingly, though, we were able to recover some CD8+ T cell function by blocking immune inhibitory molecules that are induced by parasite infection. The recovered T cells killed parasites and controlled infection. These results are important as they could be exploited for the design of new therapeutic vaccine strategies aimed at inducing protective CD8+ T cells.
| Antigen-specific CD8+ T cell responses are essential for protection and clearance of many microbial pathogens. CD8+ T cells recognize peptides which are presented in the context of major histocompatibility complex (MHC) class I via T cell receptor (TCR). Rare naïve CD8+ T cells are activated in secondary lymphoid tissues following encounter with dendritic cells expressing peptide/MHCI complexes [1]. Once activated, antigen-specific T cells typically undergo massive expansion, differentiate into effector cells, and acquire the capacity to kill and produce cytokines [2]–[5]. The magnitude of expansion largely depends on the amount of antigen and/or the number of the naïve precursors [6],[7]. This robust proliferation is then followed by a programmed contraction, which occurs independently of duration of infection, magnitude of expansion or antigen dose [7]. Only 5–10% of the cells present during the peak phase survive the contraction, becoming long-lived memory cells [8]. Memory cells show increased responsiveness and undergo dramatic clonal expansion after reencounter with the same antigen, and thereby confer protection [4],[9].
This paradigm of T cell differentiation and memory formation has been mainly derived from models of acute viral and bacterial infections, such as Lymphocytic Choriomeningitis Virus (LCMV; Armstrong strain), Vaccinia Virus and Listeria monocytogenes [2], [7], [10]–[12]. Yet it may not apply to CD8+ T cell responses generated in the presence of persistent antigen stimulation. Indeed, several degrees of dysfunction, such as delays in expansion and contraction, anergy, and suppression and exhaustion of effector responses, have been observed during chronic diseases [13]–[18]. The inhibitory receptor PD-1 and its ligand B7-H1 have been shown to play an important role in the regulation of CD8+ T cell function in anti-tumour and anti-microbial immunity, and also in the early CD8+ T cell fate decisions [19]–[22]. This pathway appears to induce T cell apoptosis and inhibits proliferation and cytokine production upon TCR engagement in vitro [23],[24]. In vivo, B7-H1/PD-1 interaction was shown to control the initiation and reversion of anergy, to inhibit T cell functions, and to be the key pathway in the induction of exhaustion [21],[25],[26]. This functionally inactivated phenotype has also been described in humans, and shown to be reverted by treatment with blocking antibodies to B7-H1, thereby restoring the capacity of CD8+ T cells to control disease and decrease viral load [21].
Experimental visceral leishmaniasis (VL) represents an exquisite model to study CD8+ T cell responses in a context of chronic inflammation and antigen persistence. In mice, the two main target organs of this disease are the liver and the spleen [27]. While in the liver the infection is self-resolving due to the development of a TH1-dominated granulomatous response, spleens infected with Leishmania donovani, the causative agent of visceral leishmaniasis (VL), stay chronically infected. Together with CD4+ T cells, CD8+ T cells have been shown to be essential for the control of primary infections in various experimental models of Leishmaniasis [28]–[31]. They also appear to be the main mediators of resistance to rechallenge and the major correlates of protection in vaccine-induced immunity against several Leishmania species [30], [32]–[35]. However, the onset of these responses seems to be delayed: polyclonal CD8+ T cell responses are only detectable 3–4 weeks into the infection in both L. major and L. donovani infected mice [29],[30]. Due to a lack of knowledge of Leishmania-specific CD8+ T cell epitopes, antigen-specific CD8+ T cell responses in VL have thus far not been studied.
In this study, transgenic L. donovani parasites expressing ovalbumin [36] were used to characterize the development, function and fate of Leishmania-specific CD8+ T cell responses during the course of infection. We show that L. donovani parasites evade CD8+ T cell responses by limiting their expansion and inducing functional exhaustion and cell death.
To determine the extent and significance of bystander activation and distinguish it from antigen-specific responses, we first compared the expansion of adoptively transferred OT-I CD8+ T cells in mice infected with wild type (LV9) and Ovalbumin-transgenic (PINK) Leishmania donovani parasites. In order to visualize and analyze OT-I CD8+ T cell responses in LV9 infected mice, it was necessary to transfer 105 OT-I CD8+ T cells per mouse. Although at day 4 after infection there were twice as many OT-I CD8+ T cells in the spleen of LV9 infected compared to naive mice, these cells were 20 times fewer than those detected in PINK infected mice (Figure 1A), despite similar splenic parasite burdens (Figure S1). By day 14 the number of OT-I CD8+ T cells in PINK infected mice was still 10 times higher then in LV9 infected mice, which had returned to baseline levels. At day 3 p.i., about 30% of the OT-I CD8+ T cells in LV9 infected mice had undergone 3–4 rounds of division, downregulated CD62L and expressed high levels of CD44. The percentage of dividing and activated cells remained unchanged throughout the course of infection (data not shown). These data indicate that the proliferative response of OT-I CD8+ T cells observed following PINK infection results mainly from antigen-specific stimulation and expansion rather then bystander activation.
We next compared the onset, expansion and dynamic of adoptively transferred OT-I CD8+ T cells in PINK infected mice to OT-I CD8+ T cell responses induced after infection with recombinant Vaccinia Virus expressing SIINFEKL (rVV-SIINFEKL). Whereas L. donovani mounts chronic infections in the spleen (Figure 1B), by contrast, VV is an excellent model for acute viral infections and is mainly cleared by prototypic CD8+ T cell response [10]. Adoptive transfer of 104 OT-I CD8+ T cells in rVV-SIINFEKL infected mice, resulted in peak expansion at day 6, with 9×106 OT-I CD8+ T cells found on average in the spleen. The cells then underwent clonal contraction and by day 21 only about 7% of the cell numbers present during the peak phase were detected (Figure 1C). Following transfer of 104 OT-I CD8+ T cells to PINK infected mice, maximum expansion was reached at day 9, and the expansion was 100–200 times lower compared to rVV-SIINFEKL infections. By day 14, the cell number was reduced by 70%. Similar numbers of OT-I CD8+ T cells were found in the spleen until day 28 (on average 1.4×104–1.6 104 cells per spleen), after which time the number of OT-I CD8+ T cells further decreased, so that by day 41 only 50% of the cells had survived (Figure 1D).
We next characterized the phenotype of PINK induced OT-I CD8+ T cell responses based on the expression of CD62L, CD127, CD44, CD122, and CD69, and compared it to OT-I CD8+ T cell responses induced by rVV-SIINFEKL. As shown in Figure 2A, about 40–60% of the OT-I CD8+ T cells in mice infected with PINK acquired an effector phenotype by downregulating CD62L during the peak expansion, compared to 90% in rVV-SIINFEKL infected mice (Figure 2B). In the latter group, the majority of the cells remained CD62Llo/int and started to slowly upregulate CD62L only after day 14, suggesting that central memory cells were gradually generated. In contrast, in PINK infected mice 85% of the cells expressed high levels of CD62L at day 14, and at day 21, 92% of the OT-I CD8+ T- cells was CD62Lhi. Between day 21 and 28, cells started down regulating CD62L again and by day 37, 64% of the cells were CD62Llo/int, thereby re-acquiring characteristics of an effector phenotype. A similar biphasic pattern of expression was observed with the IL-7R (Figure 2A, central panel). CD127 was down regulated early during the response (day 3–6), but by day 9 about 75% of the cells were CD127+. After day 21, cells started down regulating CD127 again and by day 37, more then the half of the cells were CD127 negative. The timing of this shift in the phenotype of the OT-I CD8+ T cells corresponds to the expansion in splenic parasite load (Figure 1B).
CD44 was upregulated from day 3 on (Figure S2A). The vast majority of the cells were CD44hi during the first 41 days of infection. Similarly, cells started upregulating CD122 at day 3 and remained CD122+ until day 21 (Figure 2A). However, after day 21 about 50% of the cells had downregulated CD122. CD69 has been reported to be transiently expressed during T cell activation and differentiation following antigen-presentation by dendritic cells; however, this molecule has also been shown to be persistently expressed by human and murine T cells in a context of chronic inflammation [37]. CD69 was transiently expressed by all cells present in the spleen of rVV-SIINFEKL infected mice at day 2, while only 3–9% of the cells were CD69+ between day 6 and 21 (Figure 2C, right panel). In contrast, in PINK infected mice about 40% of the OT-I CD8+ T cells expressed CD69 at day 3. This percentage slightly increased during the course of infection with the exception of day 14, when only 20% of the cells were CD69+ (Figure 2C, left panel).
We also monitored the proliferation of OT-1 CD8+ T cells by assessing the CFSE dilution over the course of the infection (Figure S3). Between day 2 and 6 after infection, the cells had undergone several rounds of division, resulting in a complete dilution of the CFSE staining (Figure S3). All OT-I CD8+ T cells present in the spleen were CFSE− until day 21. In mice infected with PINK, OT-I CD8+ T cells had already undergone 4–5 rounds of division at day 3 (Figure S3A), with maximal CFSE dilution observed at day 6. Interestingly, CFSE dilution at day 6 was higher then at days 9, 14, 21, and 28, indicating that the cells present at these later time points had undergone fewer rounds of division then those present at day 6. One possible explanation is that effector cells present in the spleen at day 6 have migrated to the liver, the other site of infection. To test this hypothesis, we enumerated the OT-I CD8+ T cells present in the liver during the course of infection (Figure S3B). A maximum of 1500 cells was detected during peak expansion, suggesting that migration of effector cells to the liver was not responsible for the disappearance of these cells from the spleen. This suggests that effector cells that had expanded at day 6 had possibly died and were replaced by newly recruited and activated cells. Notably, these cells did not undergo more then 5–6 rounds of division, even during peak expansion.
Taken together, these results show that OT-I CD8+ T cells following PINK infection display a biphasic activation pattern. During the first 9 days of infection, before OT-I CD8+ T cells undergo clonal contraction, they exhibit an effector phenotype; this activation results in limited expansion. After the first wave of activation, the majority of the cells that survived clonal contraction were CD62Lhi CD44hi CD127+ CD122+ and KLRG1− (data not shown). This phenotype is similar to that displayed by central memory cells. By week 3 of infection, cells are reactivated, they downregulate CD62L and CD127, but this time they start loosing CD122 expression and their numbers begin to wane.
Given the unique alternation of surface phenotype of OT-I CD8+ T cell responses during L. donovani infections, we were intrigued to investigate the effector function of those cells. After a brief in vitro restimulation with the SIINFEKL peptide, cells were stained for INFγ, IL-2, TNFα, Granzyme B, and CD107a. Surprisingly, L. donovani induced a very strong CD8+ T cell effector response, characterized by a high percentage of cells producing cytokines, of which more then the half were coproducing multiple cytokines (Figure S4A and Figure 3, left panels). 70–90% of OT-I CD8+ T cells expressed IFNγ between day 3 and day 28pi in PINK infected mice (Figure 3A, left panel); 45–55% of those cells were concomitantly producing TNFα; and 18–22% were co-producing IL-2 (Figure S4A). These percentages were much greater then those observed following rVV-SIINFEKL infection (Figure 3, right panels). IL-10 was not detected at any time point during infection (data not shown).
In mice infected with rVV-SIINFEKL, we could observe a gradual increase over time in the percentage of polyfunctional cells and in the amount of IFNγ produced per cell (Figure 3, right panels and Figure S4B). This was not the case in PINK infected mice, where CD8+ T cell responses became less functional after the first 3–4 weeks of infection. After day 28pi, most of the cells stopped producing IFNγ, and those that did showed a decrease in the mean fluorescence intensity of the staining (Figure S4A). A similar loss of production was observed for TNFα (Figure 3B, left panel) and IL-2 (Figure 3C, left panel): cells secreting these cytokines became progressively less functional from day 14 on.
In rVV-SIINFEKL infected mice, 40–60% of the OT-I CD8+ T cells were positive for Granzyme B during the first 2 weeks of infection (Figure 3D). This percentage gradually decreased with the generation of memory cells (Figure 3D, right panel). In contrast, in PINK infected mice the cells displayed a biphasic production pattern: 28–48% of the cells stained positive for Granzyme B during the first 9 days of infection; between day 14 and day 21pi, only 1–2% of the cells were positive (notably, during this period cells display a central memory phenotype); and by day 28pi, cells had reacquired the capacity to produce Granzyme B. We next measured degranulation by cell surface modulation of CD107a (LAMP-1). A pattern of expression similar to that seen for Granzyme B was observed for CD107a (Figure S5).
Thus, following PINK infections OT-I CD8+ T cells appear to become dysfunctional over time and express CD69+CD44hiCD62lo/intCD122 lo/neg, however, they maintained degranulation and cytotoxic capacity. These characteristics are very similar to those described for exhausted cells [12],[38]. Interestingly, functional exhaustion of OT-I CD8+ T cell was not observed in the liver. In this organ, OT-I CD8+ T cells were still producing high levels of IFNγ even at day 41pi (data not shown), suggesting that there might be an organ-specific regulation of CD8+ T cell responses and that exhaustion is most likely the consequence of the suppressive splenic environment.
In order to understand this biphasic activation pattern, we decided to asses whether the antigen presenting capacity of splenic DC changes during the course of infection. An in vitro proliferation assay using conventional splenic CD11chi DC purified from infected mice was carried out at different time points of infection. DCs were coincubated with labelled naïve OT-I CD8+ T cells for 72 h at 37°C. T cell proliferation was used as a read out for antigen presentation. Maximal proliferation was observed at day 6 after infection: at this time point on average 11.8% of OT-I CD8+ T cells had undergone cell division (Figure 4B). In contrast, DC purified at day 9, 14 and 21 induced a poor proliferation of OT-I CD8+ T cells (2–3.7%). Their capacity to present antigen increased then again at day 28 (Figure 4B). However, it is important to note that chronic L. donovani infections result in splenomegaly. Spleens start to visibly enlarge between day 14 and day 21. As the percentage of DC in the spleen remains the same that means that the capacity of DC to present antigen on a population level at day 21 and 28 is greater then appears from our proliferation assay. Thus it appears that, despite the constant presence of the parasite, antigen presentation by DC during the course of infection follows a biphasic pattern, with peaks at day 6 and day 28pi.
As adoptively transferred OT-I CD8+ T cells appeared to acquire characteristics of an exhausted phenotype during chronic infection, we proceeded to examine the B7-H1 expression on dendritic cells from the spleen of infected mice. Conventional CD11chi dendritic cells significantly and increasingly upregulated B7-H1 during the course of infection (Figure 5A). This upregulation was concomitant with increased PD-1 expression by OT-I CD8+ T cells (Figure 5B). These data suggest the potential involvement of the PD-1/B7-H1 pathway in the induction of CD8+ T cell exhaustion observed in the spleen of chronically infected mice.
We also investigated the expression of the costimulatory molecules CD40, CD86 and CD80. CD40 and CD86 were increasingly expressed during the course of infection (data not shown); in contrast, CD80 failed to be upregulated (Figure 5A).
To asses whether the PD-1/B7-H1 pathway was involved in the inhibition of cytokine production and/or cell death of adoptively transferred OT-I CD8+ T cells, we treated chronically infected mice with anti- B7-H1 blocking antibodies. Treatment started at day 15, when the expression of B7-H1 began to increase; T cell responses and parasite load were monitored. The B7-H1 blockade prevented the dramatic reduction in cell numbers that was observed in untreated mice (Figure 6A). The analysis of cell surface markers of OT-I CD8+ T cells in treated versus isotype control group revealed that cells expressing low/intermediate levels of CD62L may be surviving better in treated mice compared to the isotype control group (Figure S6A). In order to investigate whether the PD-1/B7-H1 pathway could be involved in inhibiting the activation of newly migrated naïve OT-I CD8+ T cells in the spleen during chronic infection, we tested the capacity of ex-vivo purified dendritic cells to induce proliferation of naïve OT-I CD8+ T cells in vitro in the presence or absence of the anti-B7-H1 antibody. B7-H1 blockade did not increase the proliferation of OT-I CD8+ T cells (Figure 6B). Taken together, these results imply that the PD-1/B7-H1 pathway could be either involved in the inhibition of the proliferative capacity or in the induction of cell death of effector CD8+ T cells. Further investigations are needed in order to clarify this mechanism of action.
Surprisingly, the in vivo blockade only partially restored cytokine production. In fact, IFNγ production was only transiently restored, but OT-I CD8+ T cells were still increasingly loosing their capacity to produce IL-2 and TNFα (Figure S6B). No differences were observed in the percentage of cells positive for Granzyme B or in the mean fluorescence intensity of the granzyme B staining.
Despite this functional impairment, mice treated with anti-B7-H1 antibodies were able to control parasite growth, with the splenic parasite burden reduced by 70% at day 21, 85% at day 28, and 87% at day 35 (Figure 6C). These results demonstrate that the PD-1/B7-H1 pathway plays a very important role in the suppression of CD8+ T cell responses during chronic L. donovani infections.
We next assessed whether B7-H1 blockade would confer protection against infection with the L. donovani wild type strain LV9 in a non-transgenic model. We therefore treated C57BL/6 mice chronically infected with LV9 with the anti B7-H1 antibody as previously described and monitored the parasite burden at day 21, 28 and 35 after infection (Figure 6D). B7-H1 blockade significantly reduced the splenic parasite burden at day 21 (65% reduction), 28 (57.5%), and 35 (71.4%), suggesting that endogenous CD8+ T cell responses were most likely rescued by the blockade. Interestingly, B7-H1 blockade also conferred protection in the liver, but only at day 21 (53.2% reduction in the parasite burden) and 28 (48.8% reduction), and was ineffective at day 35, when parasite growth in this organ was under control and infection had already significantly decreased (Figure S6C).
To determine whether CD8+ T cells were the main mediators of protection following B7-H1 blockade, we induced OT-I CD8+ T cell responses by superinfecting chronically infected mice with rVV-SIINFEKL at a distant site. Mice were challenged subcutaneously at day 32 pi with wild type vaccinia virus (VV) or rVV-SIINFEKL, and euthanized 2, 6, and 9 days later. As expected, rVV-SIINFEKL induced a strong proliferation of OT-I CD8+ T cells (Figure 7A and Figure S7), whose numbers increased about 40-fold compared to the unchallenged, PINK infected group. Challenge with rVV-SIINFEKL not only induced a proliferative response of OT-I CD8+ T cells, but also restored cytokine production (Figure 7C and 7D). Infection with VV did not have any effect on either proliferation or cytokine production, suggesting that the OT-I CD8+ T cell response was strictly antigen-specific.
We next compared the splenic parasite burden of mice infected with PINK to that of mice superinfected with VV or rVV-SIINFEKL (Figure 7B). Challenging mice with wild type VV did not alter the course of infection in the spleen. In contrast, infection with rVV-SIINFEKL reduced the parasite burden by 80%. This suggests that reviving CD8+ T cell responses during chronic L. donovani infections could be a successful strategy for immunotherapeutic interventions.
Our main findings demonstrate that L. donovani is able to evade the attack from CD8+ T cells by suppressing their expansion and effector function. The data show that despite the constant presence of parasites in the spleen, CD8+ T cells responses exhibited a biphasic activation pattern. The first wave of activation led to limited expansion. The second wave resulted in cell death and exhaustion of CD8+ T cells. B7-H1 blockade rescued CD8+ T cell responses from cell death, but failed to completely restore cytokine production. In spite of this, the parasite burden was considerably reduced after treatment, suggesting that maintenance of effector CD8+ T cell responses is crucial for the control of L. donovani infections in the spleen.
The adoptive transfer experiments demonstrate for the first time that L. donovani induces CD8+ T cell responses early during infection. We were able to visualize this early response only because we transferred 104 OT-I CD8+ T cells. Physiologically, the number of naturally occurring naïve precursors for a determined epitope is estimated to range between 50 and 1000 cells per mouse [6], [39]–[42]. However, due to the limited expansion capacity of CD8+ T cells in this model, transfer of such a low number of cells did not allow us to perform an accurate analysis of endogenous CD8+ T cell responses. This might explain why in previous studies the onset of polyclonal responses has been reported to be substantially delayed and could only be detected, mainly in the liver, after 3–4 weeks of infection [30].
Although we transferred the same number of cells with the same epitope specifcity, OT-I CD8+ T cells increased in numbers by only 5 fold in Leishmania infected mice, compared to 900 fold in mice infected with rVV-SIINFEKL. In other infection models, expansions up to 50,000-fold were observed [2],[5],[39],[40],[43], suggesting that Leishmania induces a very poor CD8+ T cell expansion. When we assessed the antigen-presenting capacity of DC purified from infected animals, we found that even during the early stages of infection, DC were capable of inducing only a weak proliferative response of naïve OT-I CD8+ T cells. This is not surprising, since processing of Leishmania antigens for MHCI presentation has been shown to be TAP-and proteosome-independent [44], a pathway that is much less efficient then the conventional ER-based, TAP-dependent pathway for Class I presentation. We also noted that OT-I CD8+ T cells present in the spleen at day 9 p.i. had undergone fewer rounds of division then those detected at day 6 p.i., implying that between day 6 and 9 effector cells had died and were replaced with newly activated CD8+ T cells. This suggests that expansion could also be limited by cell death of effector cells. We also cannot rule out the possibility of defective recruitment of CD8+ T cells into the spleen. Leishmania infections are known to interfere with chemokine expression [45]–[47], including CCL3 [48], a chemokine that was recently shown to be involved in guiding CD8+ T cells to sites of CD4+ T cell-dendritic cell interaction [49]. Hence, the limited expansion of OT-I CD8+ T cells in L. donovani infected mice might be due to a combination of several factors, including low antigen load, poor recruitment of CD8+ T cells and/or increased cell death of effector cells. Mechanisms responsible for this poor expansion are currently under investigation.
In agreement with the previous literature, OT-I CD8+ T cell expansion was followed by contraction at day 14 despite antigen persistence [7]. This contraction was much steeper then in rVV-SIINFEKL infected mice, suggesting that cells were dying more rapidly. One of the most striking findings was that about 80% of the cells that had survived contraction showed a central memory-like phenotype, by expressing CD62Lhi, CD44hi, CD127+, CD122+, CD69−. These cells produced high amounts of IFNγ upon restimulation and the majority were polyfunctional. Additionally, they did not produce Granzyme B. The remaining 20% displayed an effector phenotype (CD62Llo, CD127−), suggesting that effector memory cells were not generated. A similar population of central memory-like cells has been recently observed in mice infected with Trypanosoma cruzi [50]. Despite being capable of antigen-independent survival, this population was shown to be maintained for over a year in the presence of antigen persistence. A recent report suggested a crucial role of T-bet as a molecular switch between central- and effector memory cells [51],[52]. T-bet deficiency was shown to enhance generation of central memory cells. As T-bet is also involved in the induction of enhanced CD122 expression [53], and CD122 expression by OT-I CD8+ T cells is gradually decreased during the course of VL, it is possible that this molecule might not be properly induced in Leishmania-specific CD8+ T cells.
Another interesting observation was the biphasic activation pattern of OT-I CD8+ T cells, which reflects the variation in the capacity of DC to present antigen during the course of infection. This biphasic pattern can be in part explained by the biology of the Leishmania infections. These protozoan parasites are obligate intracellular pathogens that preferentially reside in macrophages, but they can also be found in other cell types [54],[55], including DC [56]. During the first wave of expansion, the majority of the cells capable of cross-presenting Leishmania antigen via MHCI is most likely killed by CTLs so that at the end of contraction very few DC presenting antigen survive and most of the parasites reside in cells that are unable to cross-present antigen. For the second wave of expansion, parasites will have first to be released from those cells in order to be phagocytosed by DC and then killed and processed for antigen presentation to CD8+ T cells. This explains why between d14 and d21 p.i. DC showed a very poor antigen presenting capacity. However, the amount of antigen presented during this period, although little, could still be enough to restimulate a memory response. Thus, OT-I CD8+ T cell responses might be already impaired at this early stage of infection. Indeed, the second wave of activation did not result in expansion, but in functional exhaustion and cell death of the OT-I CD8+ T cells.
This dysfunctional response could be a consequence of an intrinsic problem following defective priming and/or could result from a suppressive splenic environment. Although we can not rule out that OT-I CD8+ T cell responses in L. donovani infected mice might also have some intrinsic defects, our data support the second scenario. Indeed conventional CD11chi splenic DC seemed to increasingly express the inhibitory molecule B7-H1 and failed to upregulate the costimulatory molecule CD80. B7-H1 is constitutively expressed on subsets of macrophages, B-cells and thymocytes, and can be induced on dendritic cells, endothelial and epithelial cells [19],[57]. Upregulation of B7-H1 on DC has been observed during several chronic infections and in a wide range of tumors [20],[26],[58],[59]. Our results show that in vivo blockade of B7-H1 during chronic L. donovani infection increased the survival of OT-I CD8+ T cells. B7-H1 is thought to inhibit T cell proliferation and cytokine production by ligation with the PD-1 receptor [23]. Through ligation with a yet unknown receptor, B7-H1 can also induce programmed cell death of effector T cells [60]. Increased survival of OT-I CD8+ T cells after B7-H1 blockade could therefore result from restoration of the proliferative capacity or inhibition of induced cell death of effector CD8+ T cells.
In contrast to what has been recently reported in the literature [21],[26], in vivo blockade of B7-H1 during chronic VL did not completely restore the functional capacity of exhausted OT-I CD8+ T cells. This suggests that suppression of cytokine production by CD8+ T cells during L. donovani infection might be induced by mechanisms other than through the B7-H1/PD-1 pathway. A recent report has demonstrated a synergistic effect between TGFβ and the B7-H1/PD-1 axis in suppressing CD8+ T cell responses [61]. As TGFβ does not seem to play an important role during chronic L. donovani infections [62], the possibility that IL-10, which is elevated in both mouse and human VL [63]–[69], could synergistically act with the B7-H1/PD-1 axis, needs to be investigated. To our surprise, B7-H1 blockade resulted in significant decrease in the parasite burden even if it failed to fully restore IFNγ production. While CD4+ T cells are clearly an important source of IFNγ in VL, recently we have shown that therapeutic intervention with antigen-specific CD8+ T cells in chronically infected mice dramatically reduced the parasite burden [36], indicating that CD8+ T cells might play a much more important role than previously thought. The current data reinforce these findings by showing that OT-I CD8+ T cells rescued from cell death by blocking B7-H1 or by superinfecting mice with rVV-SIINFEKL resulted in host protection. The mechanism of protection is not clear and might not merely rely on IFNγ production, as only 20% of the OT-I CD8+ T cells were producing low amounts of IFNγ at d35 pi. Nonetheless, most of the cells were granzyme B positive and were degranulating upon restimulation, suggesting that they have retained their cytotoxic capacity. To date there is no evidence that CD8+ T cells can mediate protection against L. donovani through their cytotoxic activity.
In summary, this study shows that restoration of dysfunctional CD8+ T cell responses induced by chronic L. donovani infections results in disease control and host protection. This implies that targeting CD8+ T cell responses by therapeutic vaccination could be beneficial against chronic L. donovani infections. Moreover, these findings might provide insights into the development of novel strategies for therapeutic vaccination or other interventions aimed at inducing CD8+ T cell responses, which might circumvent and/or neutralize the immunosuppressive environment of the spleen.
C57BL/6-Tg(OT-I)-RAG1tm1Mom mice were purchased from Taconic; B6-Ly5.2 congenic mice were obtained from The National Cancer Institute (Frederick, MD, USA), and B6.129S7-Rag1tm1Mom/J from The Jackson Laboratory. All mice were housed in the Johns Hopkins University animal facilities (Baltimore, MD) under specific pathogen-free conditions and used at 6–8 weeks of age. All experiments were approved by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine.
Ovalbumin-transgenic parasites were a gift from P.Kaye and D.F. Smith (University of York, UK) and were generated as previously described[36]. Wild type and ovalbumin transgenic Leishmania donovani (strain LV9) parasites were maintained by serial passage in B6.129S7-Rag1tm1Mom/J mice, and amastigotes were isolated from the spleens of infected animals. Mice were infected by injecting 2×107 amastigotes intravenously via the lateral tail vein. Hepatic and splenic parasite burdens were determined either by limiting dilutions [31]or by examining methanol-fixed, Giemsa stained tissue impression smears[70]. Data are presented as number of parasites per spleen or as Leishman Donovan Units (LDU).
The recombinant vaccinia virus (rVV) encoding SIINFEKL (chicken ovalbumin 257–264) was a gift from F. Zavala (School of Public Health, JHU, Baltimore) [71]. Mice were infected intravenously or subcutaneously with 2×106 pfu.
OT-I/RAG1 mice, transgenic for a T cell receptor specific for chicken ovalbumin 257–264 presented by the MHC class I molecule H-2 Kb, were used as T cell donors. CD8+ T cells were enriched from splenocytes of naïve OT-I/RAG1 animals using magnetic cell sorting (MACS), following manufacturers instructions (Miltenyi Biotech). Naïve CD8+ T cells were then sorted to >98% purity using FACSVantage (Becton Dickinson) based on their expression of CD44 and CD62L. After sorting, cells were labelled with CFSE. Briefly, cells were resuspended at 5×107/ml in PBS and incubated with 2.5 µg/ml CFSE (Molecular Probes, USA) for 10 min. at 37°C. The reaction was stopped by addition of ice cold RPMI. Samples were then analyzed using a FACSDiva (Becton Dickinson) for CFSE uptake prior to adoptive transfer. Depending on the experiment, 1×104 or 5×104 cells were injected into the lateral tail vein of B6-Ly5.2 congenic mice. Animals were infected the day after with rVV-SIINFEKL and/or with wild type or ovalbumin-transgenic Leishmania donovani.
1×104 sorted naïve OT-I CD8+ T cells were adoptively transferred into B6-Ly5.2 congenic mice prior to infection with 2×107 ovalbumin expressing L. donovani amastigotes. At day 32pi mice were superinfected subcutaneously at the base of the tail with 2×106 PFU of Vaccinia Virus (VV) or with recombinant VV expressing the SIINFEKL peptide (rVV-SIINFEKL). Animals were sacrificed at d2, d6 and d9 after infection with rVV-SIINFEKL.
OT-I CD8+ T cells were identified by staining splenocytes, lymphnode cells and hepatic mononuclear cells with biotinylated anti-CD45.2 antibody followed by PerCP-streptavidin (BD Biosciences). The following antibodies were used to further characterize the OT-I response: APC-conjugated anti-CD44 and anti-CD8, PE-conjugated anti-CD62L, anti-CD69, anti-CD122, anti-CD127 (all obtained from BD Biosciences), and anti-PD-1 (eBioscience). Splenocytes were also stained with APC-conjugated anti-CD11c, FITC-conjugated anti-MHCII, PE-conjugated anti-CD86, PE-Cy5.5 conjugated anti-CD80, and biotinylated anti-B7H1 and anti-CD40, followed by PerCP-conjugated streptavidin (all purchased by BD Biosciences). For all surface markers, cells were directly stained following standard protocols. For intracellular staining, splenocytes were stimulated with the SIINFEKL peptide for 4 hours in the presence of Brefeldin A and then stained with biotinylated anti-CD45.2, followed by PerCP-conjugated strepavidin. After fixation, cells were permeabilized and stained with anti-Granzyme B (Invitrogen) or APC-conjugated anti-INFγ (BD Biosciences), PE-conjugated anti IL-2 (BD Biosciences), and PE-Cy7-conjugated anti-TNFα (eBioscience). Cells were also stained with PE-conjugated anti-CD107 (eBioscience) following the protocol described by Betts et al. [72].
Flowcytometric analysis was performed with a LSRII (Becton Dickinson). One to two millions cells per sample were acquired and analysed with the FACSDiva or with CellQuest software.
Spleen of naïve and ovalbumin-transgenic L. donovani infected mice were digested with 0.4 mg/ml collagenase D for 30 minutes at room temperature. Conventional CD11chi dendritic cells were then enriched by MACS using CD11c microbeads (purity 80–85%).
Dendritic cells were seeded at a concentration of 2×104 cells/well in a 96 wells plate.
After negative selection with anti-CD11c microbeads, CD8+ T cells were purified from the spleen of naïve OT-I/RAG1 mice using magnetic cell sorting (Miltenyi Biotech) (85–90% purity). CD8+ OT-I T cells were then labelled using a red fluorescent cell linker PKH26 (Sigma) in order to track proliferation. They were then added to the ex-vivo purified dendritic cells at a concentration of 105/well. 1 ng/ml of recombinant human IL-2 was also added to the wells. The proliferation of OT-I T cells was assessed 72 h later by flowcytometry using FACSDiva (BD Biosciences) and analysed with the FACSDiva software. Results are expressed as percentage of OT-I CD8+ T cells that have undergone one or more rounds of division. The percentage of cells that entered division when incubated with DCs from a naive animal was subtracted from this value.
Antagonistic mouse B7-H1 monoclonal antibody (clone 10B5) was purified on a protein G column from the supernatant of the hybridoma cell line. The hybridoma cell line was a gift from L.Chen (Johns Hopkins University, School of Medicine, Baltimore). Hamster IgG (Sigma) was used as isotype control. Mice were treated every 4 days with 100 µg of antibody i.p. The first treatment started at day 15 p.i. Before treatment, antibodies were tested for functionally relevant LPS contamination, by assaying their ability to synergize with IFNγ for the induction of inducible NO synthase [73]. No activity was detectable in such assays (sensitivity,1 ng/ml LPS; data not shown).
Results were analyzed using an unpaired Student t-test. P<0.05 was considered significant. All experiments were repeated at least twice.
|
10.1371/journal.pbio.1002098 | Relatedness, Conflict, and the Evolution of Eusociality | The evolution of sterile worker castes in eusocial insects was a major problem in evolutionary theory until Hamilton developed a method called inclusive fitness. He used it to show that sterile castes could evolve via kin selection, in which a gene for altruistic sterility is favored when the altruism sufficiently benefits relatives carrying the gene. Inclusive fitness theory is well supported empirically and has been applied to many other areas, but a recent paper argued that the general method of inclusive fitness was wrong and advocated an alternative population genetic method. The claim of these authors was bolstered by a new model of the evolution of eusociality with novel conclusions that appeared to overturn some major results from inclusive fitness. Here we report an expanded examination of this kind of model for the evolution of eusociality and show that all three of its apparently novel conclusions are essentially false. Contrary to their claims, genetic relatedness is important and causal, workers are agents that can evolve to be in conflict with the queen, and eusociality is not so difficult to evolve. The misleading conclusions all resulted not from incorrect math but from overgeneralizing from narrow assumptions or parameter values. For example, all of their models implicitly assumed high relatedness, but modifying the model to allow lower relatedness shows that relatedness is essential and causal in the evolution of eusociality. Their modeling strategy, properly applied, actually confirms major insights of inclusive fitness studies of kin selection. This broad agreement of different models shows that social evolution theory, rather than being in turmoil, is supported by multiple theoretical approaches. It also suggests that extensive prior work using inclusive fitness, from microbial interactions to human evolution, should be considered robust unless shown otherwise.
| The evolution of sterile worker castes in social insects has fascinated biologists ever since Darwin; how can selection favor a trait that decreases reproductive fitness? W. D. Hamilton solved this dilemma in the 1960s with a theory showing that reproductive altruism could evolve if it increased the worker’s inclusive fitness, which included effects that it had on increasing the fitness of its relatives. This solution to a crucial evolutionary problem, sometimes called kin selection, was challenged in a recent paper. The paper generated much controversy, but no one has contested its new theoretical model of the evolution of eusociality, which appeared to overturn much of what was previously thought to be true from kin selection theory. Here we examine this model in greater depth, showing that its apparently novel conclusions are overgeneralized from narrow and often inappropriate assumptions. Instead, this modeling strategy yields results that confirm important insights from kin selection and inclusive fitness, such as the importance of relatedness and the existence of conflicts in social insect colonies.
| The eusocial insects have occupied an important place in biology because of their extraordinary levels of cooperation [1–4]. In ants, termites, some bees, some wasps, and a few other taxa, certain individuals, called workers, give up their own reproduction in order to help others reproduce. Darwin was vexed over the question of how such reproductive altruism evolves or indeed how any traits of sterile workers evolve, but he believed that such sterility was due to some form of selection at the family level or at the group level [5]. Hamilton provided the first rigorous treatment of this idea, with a key insight being the importance of genetic relatedness [1]. A conditional gene causing a worker to give up reproduction could be favored if it provided sufficient help to a relative who would share that gene at above-random levels. He showed that this process, which became known as kin selection, could be analyzed by summing up an actor’s fitness effects, each multiplied by the actor’s relatedness to the individual receiving the fitness effect. When this sum, called the inclusive fitness effect, is positive, the trait should be favored by selection. For giving up one’s reproduction (fitness cost c) to benefit other individuals (total fitness gain b) related by r, the inclusive fitness condition is −c + rb > 0.
Kin selection and inclusive fitness became the dominant modes of thinking about the evolution of eusocial insects [4,6,7], and their success in this area has led to them being applied to many other problems in social evolution [8–12]. Recently, this paradigm was criticized by Nowak et al. [13], who argued that inclusive fitness was an inaccurate and unnecessary method and that kin selection was not a very useful way to think about social evolution. Both of these conclusions have in turn been extensively criticized as depending on multiple misconceptions [14–22]. We concur with many of these criticisms but do not revisit them here. Instead, we offer a different kind of critique of the Nowak et al. paper. To provide an example that bolstered their general arguments, Nowak et al. [13] also developed their own mathematical model of the evolution of eusociality, presenting it as an example of a modeling approach that is superior to inclusive fitness modeling. However, as has been recently pointed out [23], this eusociality model has scarcely been addressed.
We do not contest this modeling approach. Instead, we accept it as valid and use it to show that its implementation in Nowak et al. [13] led to errors of interpretation that greatly overstated any differences with standard inclusive fitness results. We do not address the exact quantitative match of the two approaches but instead focus on large apparent discrepancies of interest to empiricists. Because their model is claimed to be superior to inclusive fitness, we focus on three of their conclusions that seem at greatest variance with the conventional inclusive fitness and kin selection view of the evolution of eusociality. In each case, we will show that the kin selection view is essentially confirmed. Nowak et al. [13] also make other assertions about eusociality that are consistent with inclusive fitness theory, such as the importance of grouping and preadaptations. We ignore these in order to focus on the seemingly novel conclusions of the Nowak et al. model. The first two of these are fundamental qualitative differences from inclusive fitness, while the last is more a difference in degree.
First, Nowak et al. [13], following earlier work by Wilson [24,25], claimed that relatedness was not an essential element in the evolution of eusociality. They wrote that “relatedness is better explained as a consequence rather than as the cause of sociality,” that “grouping by family hastens the spread of eusocial alleles but it is not a causative agent,” and that “relatedness does not drive the evolution of eusociality” [13]. In the same vein, they also contest empirical evidence that relatedness is important [13]. We take causality to mean that variation in relatedness leads to variation in the likelihood of evolving eusociality. As has previously been pointed out, the Nowak et al. model could not test this because it was based on groups of relatives, with no comparable model of unrelated individuals being presented [15,20]. Nowak et al. appear to have partially accepted this point: “One, we do not argue that relatedness is unimportant. Relatedness is an aspect of population structure, which affects evolution” [26]. However, this response leaves unanswered exactly how it affects evolution. At least one of the authors [27] continues to assert that relatedness only hastens the spread of alleles and that it is not causal. To test these claims, we extend their model to cases in which relatedness can vary.
Second, whereas inclusive fitness theory has emphasized that cooperation occurs in the face of potential and actual conflicts among colony members with different interests [4,7,28,29], Nowak et al. [13] assert that the colony as a whole is all that matters. They argue that “the workers are not independent agents,” that “their properties are determined by the alleles that are present in the queen (both in her own genome and in that of the sperm she has stored),” that “the workers can be seen as ‘robots’ that are built by the queen,” and that they “are part of the queen’s strategy for reproduction” [13]. Nor, contrary to earlier work by Wilson [24,25], do they brook any conflicts between levels of selection: “there is only one level of selection, the hymenopteran colony, which is treated as an extension of the queen, whose genes are the units of selection” [13]. To test whether workers and queens are independent agents that are selected differently, we construct parallel models in which the genes determining whether their offspring stay and help are expressed in mothers or expressed in offspring.
Finally, Nowak et al. [13] claim that eusociality is harder to evolve than has been appreciated. They write that “a key observation of our model is that it is difficult to evolve eusociality, because we need very favorable parameters” and that “despite the obvious and intuitive advantages of eusociality, it is very hard for a solitary species to achieve it” [13]. If there is any novelty in this conclusion, it must be that eusociality is harder to evolve than has been thought previously; that is, it is harder to evolve than predicted from inclusive fitness effects (–c + rb > 0). We explore how this conclusion changes with reasonable alterations in the fitness functions and the worker decision rules.
If the three apparently novel conclusions of Nowak et al. are correct [13], then inclusive fitness theory could be said to have made some serious errors, and we might have to throw out or rethink important elements of the last 50 years of social evolution theory. If instead our models reject those apparently novel conclusions in favor of results consistent with those obtained through inclusive fitness, it would show that different theoretical approaches yield broadly consistent results, as they ought to in a healthy science.
We modify the Nowak et al. [13] haploid model, which is simpler than their haplodiploid one but sufficient to demonstrate the important points. Our goal is not to exactly model eusociality in any particular organism but to examine the logic and truth of three general claims in Nowak et al. [13], claims that pertain to both the haploid and haplodiploid models. The basic model includes solitary and eusocial genotypes expressed in offspring, where solitaries always leave to reproduce, while eusocials stay and help their mother with probability q and leave to reproduce with probability 1 – q. Mothers and offspring are genetically identical. Differential equations describe changes in the numbers of solitary individuals and eusocial colonies based on colony-size–specific queen birthrates (bi) and death rates (di), as well as worker death rates (α) and density dependence (η) (see Methods, Equation 1). If larger colony size (more workers) sufficiently increases the queen’s birthrate and/or decreases her death rate, the eusocial type can be favored over solitary reproduction under some probabilities of staying q. Using these equations, we recovered results indistinguishable from those of Nowak et al. [13] (e.g., their Figure 4). We then explored the effects of various assumptions by changing them one by one.
First, the models of Nowak et al. [13] assumed eusocial offspring stay with their mother so that there was always genetic relatedness among participants. In the haploid model, this meant that helpers were genetically identical (r = 1) to their mother and to the siblings they raised. To vary genetic relatedness in the haploid model, we allowed some offspring mixing between mothers before implementing their genetic helping rules. Each offspring has a probability r of being with her own mother before deciding whether to help her or leave to reproduce and a probability 1 – r of being with a random mother. This could result from offspring movement between nests, from mothers laying a fraction of their eggs in other nests, or from nest usurpation [30,31]. r is equivalent to relatedness to the new mother (after movement) because it represents identity to that mother above chance levels; a fraction r is identical to the head of their colony and her offspring (r = 1), while the remainder are randomly associated with colonies (r = 0). After this temporary mixing, offspring execute the original Nowak et al. strategies: offspring with the solitary genotype always leave to reproduce alone, and offspring with the eusocial genotype stay and help their colony with probability q. Differential equations implementing this model are given in the Methods (Equation 2).
The filled circles in Fig. 1 show when selection on offspring favors eusociality under varying relatedness r, worker-assisted queen birthrate b, and probability of staying q (other parameters continue to match the standard Nowak et al. Figure 4 parameter values). Lowering relatedness clearly makes it more difficult for eusociality to evolve; with lower r, a higher b is required to favor eusociality. In the extreme, when offspring are randomly associated with colonies so that relatedness is zero, even b = 500 (a 1,000-fold increase in the queen’s birthrate due to helpers) is insufficient to favor eusociality. As expected from inclusive fitness theory, relatedness is causal in the sense that some relatedness is necessary for eusociality and increasing relatedness increases the range of conditions allowing eusociality to evolve.
Second, to address the issue of whether worker offspring are independent agents or simply robots carrying out the queen’s interests, we need to compare models of control by different agents. This means comparing models in which the decision to stay and help is made by genes in offspring bodies to models in which it is made by genes in the resident queens’ bodies. Though Nowak et al. [13] seem to argue for queen control, their models are for offspring control because they generally assume that genes expressed in worker bodies determine the decision to stay or leave.
However, inclusive fitness theory predicts that when queen control is possible, it will generally be more favorable for evolving eusociality [7] unless relatedness is one, in which case no conflict is expected. To model queen control under varying relatedness in the haploid model, we allowed offspring to mix exactly as in the offspring control model above but then allowed the resident queen’s genotype to determine if her mixed offspring pool helps or not. If the mother has the solitary genotype, all of her mixed pool disperses to become reproductives; if the mother has the eusocial genotype, she causes a fraction q of her offspring pool to stay and help her, independent of offspring genotype. Differential equations governing this system are given in the Methods (Equation 3). As predicted by inclusive fitness theory, eusociality evolves much more easily under queen control (Fig. 1, all circles). The only exception, as expected under inclusive fitness theory, is when there is no mixing between nests so r = 1 and the two decision rules are selected identically. In fact, assuming that queens can control the trait, we see the expected opposite relationship with relatedness; the less related the queen is to the offspring in her colony, the more the queen is selected to cause them to be workers.
The final claim that we examine is that eusociality is hard to evolve [13]. This depends on what is meant by “hard,” but we can usefully ask whether eusociality is as difficult to evolve as is implied in the Nowak et al. [13] paper. Their claim seems based on particular and odd choices for fitness functions and worker decision rules. The fitness function that they generally explored was a threshold function in which workers add no fitness gains to the queen below a colony of size m and add a fixed gain (increasing queen b or decreasing d) in colonies at or above size m, regardless of how many workers are added. This means that workers in colonies below that threshold contribute nothing until enough further workers join and that workers above the threshold also add nothing extra unless other workers die, returning the colony to the threshold. If most workers are contributing nothing, then it is not surprising that eusociality would be hard to evolve. In the example most explored, the threshold colony size m was set at 3 (their Figure 4), such that two workers were needed to raise the queen’s birthrate from b0 = 0.5 to b = 4 and to lower her death rate from d0 = 0.1 to d = 0.01 (they also let α = 0.1 and η = 0.01) [13]. This 8-fold increase in the queen’s birthrate allowed eusociality to evolve for some values of q, but lower values of b did not allow eusociality to evolve. Not surprisingly, requiring more workers before the queen increased fitness (higher m thresholds) made eusociality even more difficult to evolve.
As noted above, the assumption that workers must stay with probability q, regardless of the state of the colony, means they may be maladaptively staying in colonies that are too large to gain further benefits. It should be easy for workers to avoid this problem. For example, they might instead implement the rule to stay when the colony is below some threshold size w and leave when it is at or above that size. We implemented differential equations to model this change of assumption (see Methods, Equation 4) in the original Nowak et al. model with worker control and r = 1 (i.e., independently of the other changes explored above). Eusociality does evolve more readily. For example, for the same parameter values as in Figure 4 of Nowak et al., eusociality can now be favored under a somewhat lower benefits threshold (b = 3), that is, when helped queens get a 6-fold advantage.
In addition, the threshold fitness function assumed by Nowak et al. [13] prevents the earliest workers from contributing anything. However, it is easy to envision advantages that would come from having only a single worker [25,32]. To view this effect in isolation, we return to the Nowak et al. [13] decision rule (stay with probability q) and to their parameter values given above but allow a single worker to add half the contribution to the queen that two workers add (for both birthrate and death rate) (m = 3, b0 = 0.5, d0 = 0.1, d = 0.01, α = 0.1, η = 0.01). This simple change (implemented in Equation 1) makes it much easier to evolve eusociality, with b = 1.5 or only a 3-fold increase required (Fig. 2) versus 8-fold with the threshold model. This analysis does not resolve what actual fitness functions and decision rules apply in nature, but we note that evolution tends to take the easiest paths available and eschew the difficult ones.
This result appears very close to what is expected under inclusive fitness when r = 1: if two workers increase queen birthrate from 0.5 to 1.5, each raises it by 0.5, exactly the amount that the worker gives up by helping. However, the comparison is not accurate for two reasons. First, this comparison of birthrates neglects the workers’ effect on queen death rate in the model. Second, having gone back to the stay-with probability q decision rule, some workers waste their efforts by joining large colonies. In order to compare more closely with inclusive fitness, we altered both of these: the queen death rate is now unchanged by workers, and the stepwise birthrate function is implemented together with the stay-below-colony-size-w decision rule. For w = 3, eusociality is not favored at b = 1.5 (where inclusive fitness predicts it to be neutral [workers giving up 0.5 and adding 0.5 to the queen]) but is favored to evolve at b = 1.6. It is still possible to argue that eusociality is hard to evolve, depending upon one’s standard for what is hard, but it is considerably easier to evolve than implied by the initial Nowak et al. model and, not surprisingly hard relative to inclusive fitness predictions.
The controversy over the Nowak et al. paper has mostly been conducted at rather abstract levels; different researchers prefer different modeling strategies and may also interpret the evidence differently [13–20,26]. We take a different and more concrete approach by investigating their model for the evolution of eusociality more deeply. If their methods are superior and raise novel insights, we should welcome them and perhaps question our older theories. If instead their methods lead to no novel insights, it undermines the larger claims that the model is used to buttress, specifically that inclusive fitness has not been useful.
We have therefore followed the recommendation of Nowak et al. [13] for modeling social evolution, and in particular eusociality, using deterministic evolutionary dynamics described by ordinary differential equations. However, stimulated by inclusive fitness thinking, we have sought to understand apparent differences between their results compared to previous models. In every case, we find that their rejection of accepted results is incorrect and that in fact the insights known from inclusive fitness theory also emerge using their method.
The claims that relatedness only hastens the spread of eusocial alleles and that relatedness is not causal [13,27] are shown by our models to be false. The proposition could not be tested in the Nowak et al. [13] models because they did not examine any low-relatedness case [15,20]. We have modeled variable relatedness and shown that, under offspring control, high relatedness broadens the range of conditions allowing eusociality to evolve. Relatedness affects not just speed of selection but whether it is favored at all; when relatedness is zero, eusociality does not evolve even with very high benefits (increasing queen birthrate 1,000-fold). This shows that relatedness plays an essential and causal role. Of course, these are not surprising findings because the importance of relatedness was previously well understood from many kinds of models using inclusive fitness [1,7], population genetics [33–35], quantitative genetics [36–38], and game theory [39,40], as well as being supported by much empirical evidence [7,9,41,42].
An alternative interpretation of Nowak et al.’s [13] views is that relatedness is not causal because high relatedness does not always drive the evolution of eusociality. However, this is a rather empty view since no one has ever asserted the contrary and Hamilton’s rule explicitly includes other factors that interact with relatedness. In addition, this view would negate most biological causality of any kind, as no single factor ever completely determines outcomes. Finally, if Nowak et al. agreed that variation in relatedness is an important determinant of eusociality, which is widely regarded as the most important contribution to the topic in 50 years, why did they not say so, instead consistently arguing against its significance? This pattern extends beyond the Nowak et al. [13] paper to Wilson’s earlier and later papers [24,25,27] and to work from Nowak’s group purporting to show new pathways to cooperation [43,44] that in fact depended critically on relatedness and could be interpreted via inclusive fitness [45–47]. Whatever view of causality is taken, it is important to be clear that the Nowak et al. [13] modeling strategy is just like others in showing that higher relatedness is an important factor promoting higher cooperation.
A second claim of Nowak et al. [13], that workers are robots and simply part of the queen’s reproductive success, cannot be made without testing and contrasting queen and worker decision rules. Nowak et al. [13] tested only offspring control models because the decisions are controlled by genes expressed in workers. It is a longstanding result of inclusive fitness theory that parents and offspring are agents with different interests that can be in conflict [28,48]. In particular, in the eusociality context, inclusive fitness predicts that offspring will be selected to help their mothers under a narrower range of conditions than the mothers would favor (eusociality evolves more readily if mothers control the helping of their offspring) (pp. 58–63 of [7]). This follows from differences in relatedness. Workers should gain less from helping less-related kin, but queen inclusive fitness improves if she is less related to the workers who pay the fitness cost.
To examine this question, one must compare selection of offspring agency (genes expressed in the offspring determine whether she becomes a worker) versus maternal agency (genes expressed in the mother determine whether her offspring become workers). We therefore constructed haploid models for maternal control to compare with the results under offspring control. As predicted by inclusive fitness theory, the two cases evolve differently and can be in conflict: mothers favor helping by their offspring under a much broader range of conditions than the offspring themselves favor, except when mothers and offspring are genetically identical (Fig. 1, all circles). And as predicted, when relatedness is low and eusociality is very difficult to evolve under worker control, it is very easy to evolve if the queen has control, because the queen is unrelated to most of the workers who pay the fitness cost. If queens really were in control from the origin of eusociality, and if they could exert that control on unrelated offspring, that would be the easiest path to eusociality. However, this is contradicted by phylogenetic studies showing that relatedness was always high at the various origins of eusociality [41]. In contrast, the standard kin selection model of worker control predicts this observation.
Finally, the claim that eusociality is difficult to evolve [13] is less fundamental than the other two claims and also less wrong because its truth necessarily depends on how one defines “hard to evolve.” Eusociality has evolved a modest number of times and therefore could be viewed as hard to evolve, but their model does make it appear that eusociality is harder to evolve than has been believed. We show that this result hinges on assumptions that are heavily biased towards that conclusion. Little justification was given for why we should accept these particular assumptions. In particular, assumptions are made that imply that many workers waste their efforts. First, their model assumed that offspring stay with probability q, independent of any information that might be available about the need for workers. One advantage of inclusive fitness thinking is that it induces researchers to think of workers as agents being selected to get better outcomes (higher inclusive fitness) using whatever information is available to them. One such piece of information is the number of workers already present on the nest. In the threshold fitness model, there is no inclusive fitness gain to be had from staying above that threshold, unless some workers die, so we asked if there was some obvious better decision rule than stay with probability q. We therefore tested decision rules that have workers staying when the colony is below a threshold size (not necessarily the same as the fitness threshold) and leaving when the colony is above that size. Not surprisingly, we find that this class of decision rules makes it easier to evolve eusociality, because fewer workers are making wasteful decisions to stay in large colonies. Such a rule seems well within the capabilities of workers. They need not count adults. They simply need to be able to assess some reasonable correlate of the count, something that even microbes do when using quorum sensing to change their behavior. For social insects, the mechanism might involve the degree of comfort with contacting other adults or the hunger demands of offspring.
Similarly, the threshold fitness model assumed by Nowak et al. devalues worker behavior at the other, low, end of colony sizes. In most of their model examples (though not their general model), it was assumed that it was necessary to have two workers to provide any benefit at all to the queen (m = 3). That means that the first worker to join a colony provides nothing. However, it is easy to envision situations in which the first worker to join would provide real benefits [32]. The simplest is that at this point one individual can guard the nest while the other forages [25]. Empirical evidence suggests that first helpers do provide benefits [49–54]. If we modify the Nowak et al. threshold model to a step model in which each worker below the threshold adds an additional fixed benefit up to the maximum at colony size m, so that the efforts of unjoined first workers are not wasted, eusociality evolves much more easily. Thus, two modifications—the stepped fitness function and the altered worker decision rule—independently make it easier for eusociality to evolve. When we implemented these two rules together so that no workers waste their efforts and assumed workers affect only queen birthrates, eusociality evolved when predicted by inclusive fitness effects on birthrates. We do not know if this is general; the exact correspondence of the two methods may deserve additional study, but our goal here is to address the apparent major discrepancies.
The method advocated by Nowak et al. [13] offers the advantage of specifying parameters like birth and death rates explicitly and following their effects over time while allowing some features, like colony size, to change. We expect that these methods can be used to generate interesting results. However, they are more complex and less intuitive than inclusive fitness thinking, so considerable care is needed to fully understand them. The common thread in the three errors pointed out in this paper is overgeneralization from narrow assumptions or particular parameter values. Relatedness was said to be unimportant even though the models did not vary relatedness. The assertion that workers are not independent agents was made in the absence of models that compared decision rules of different agents. Eusociality was said to be difficult to evolve based on specific and questionable assumptions about the fitness function and offspring decision rules. The more complex the model, the easier it is to be misled by particular results that are not general. In this case, the initial Nowak et al. model [13] missed not just minor details but perhaps the most important generalizations known from the last five decades of theory and empirical study: the importance of relatedness and conflict. Apparent lack of agreement with prior results should have triggered more than a quick rejection of inclusive fitness and kin selection; it should have led to a questioning of why the results were, or seemed to be, different. When examined more closely, models of the type advocated by Nowak et al. [13] do not overturn but instead reaffirm principles of social evolution discovered through inclusive fitness. To have multiple theoretical approaches converging on similar results attests to the robustness of social evolution theory.
Our models are all based on the haploid model of Nowak et al. [13]. They modeled the evolution of eusociality with systems of differential equations tracking the number of solitary queens (x0) and eusocial colonies of size i (xi). We use a modified notation because our low-relatedness models require us to also keep track of colonies headed by solitary-genotype queens. We therefore let ei be the number of colonies of size i headed by a eusocial queen (that is with i – 1 workers) and si be the number of colonies of size i headed by a solitary queen. With this modified notation, equation set 58 of Nowak et al. [13] can be written as:
s˙1=(b1ϕ−d1)s1e˙1=∑i=1∞biϕ(1−q)ei−b1ϕqe1−d1e1+αe2e˙i=bi−1ϕqei−1−biϕqei−diei−α(i−1)ei+αiei+1fori>1,
(1)
where bi and di are the birth and death rates of colonies of size i, q is the probability that an offspring of a eusocial colony stays as a worker (offspring of solitary colonies never stay), α is the worker mortality rate, and ϕ is a density-dependent correction factor equal to 1/(1 + ηX), with X being the total population size including workers and η scaling the size of the system.
For specific examples, Nowak et al. [13] usually assumed birthrates and death rates were governed by a simple threshold function: below some threshold colony size m, bi = b0 and di = d0 and at or above colony size m, bi = b and di = d. Using two numerical methods (see below), we used Equation 1 to reproduce the results of Figure 4 in Nowak et al. [13] (see S1 Code).
The Nowak et al. models all assumed high and fixed relatedness. We modify their haploid model to incorporate a parameterized mixing step, which allows us to vary the degree of relatedness between queens and workers. The mixing occurs before offspring decide to be workers or reproductive queens. We allowed offspring to move to other mothers, eusocial or solitary, with probability 1 – r. Each moving offspring is replaced by a eusocial or a solitary offspring with probabilities fe and fs, which are simply the proportions of such offspring produced in the population:
fe=∑ibiei/∑ibi(si+ei)fs=∑ibisi/∑ibi(si+ei).
After mixing, offspring execute their staying rule (leave for solitaries and stay with probability q for eusocials). r is relatedness to the mother they help because r of the time she is identical, and 1 – r of the time she is genetically random or unrelated. For this offspring decision model, the equations describing changes in colony types are as follows:
s˙1=∑i(biϕ(1−r)fsei)+∑i(biϕ(r+(1−r)fs)si)−d1s1+αs2 s˙i=bi−1ϕ(1−r)feqsi−1−biϕ(1−r)feqsi−disi−α(i−1)si+αisi+1e˙1=∑i(biϕ(r+(1−r)fe)(1−q)ei)+∑i(biϕ(1−r)fe(1−q)si)−b1ϕ(r+(1−r)fe)qe1−d1e1+αe2e˙i=bi−1ϕ(r+(1−r)fe)qei−1−biϕ(r+(1−r)fe)qei−diei−α(i−1)ei+αiei+1.
(2)
Here ei and si still represent numbers after decision rules are executed and do not reflect numbers in the transient mixing stage. The equations were numerically solved using S3 Code.
For maternal control, we implemented the same offspring mixing model but allowed the mother’s genotype to determine whether the offspring in her colony (some of them resulting from mixing from other colonies) stay and help. Thus, if the queen is eusocial, her (mixed) offspring will become new workers with probability q or new queens with probability 1 – q. If the queen is solitary, then all offspring will become new queens. The equations now become the following:
s˙1=b1ϕ(r+(1−r)fs)s1+∑i(biϕ(1−r)fs)(1−q)ei)−d1s1 e˙1=∑i(biϕ(r+(1−r)fe)(1−q)ei)+b1ϕ(1−r)fes1−b1ϕqe1−d1e1+αe2e˙i=bi−1ϕqei−1−biϕqei−diei−α(i−1)ei+αiei+1.
(3)
Note that, unlike the worker model, there are no solitary colonies larger than one (after the transient mixing stage) because a solitary queen always causes her offspring pool to disperse and become reproductive. The equations were numerically solved using S4 Code.
To examine if eusociality is easier to evolve than suggested in Nowak et al. [13], we tested alternative worker decision rule and fitness functions. First, instead of staying with probability q, eusocial offspring always stay when colony size i < w and always leave when i ≥ w. The equations are as follows:
s˙1=(ϕb1−d1)s1e˙1=∑i=w∞ϕbiei−ϕb1e1−d1e1+αe2e˙i=ϕbi−1ei−1−ϕbiei−diei−α(i−1)ei+αiei+1for 1< i < we˙i=ϕbi−1ei−1−diei−α(i−1)ei+αiei+1fori = we˙i=−diei−α(i−1)ei+αiei+1for i > w.
(4)
These equations were numerically solved using S5 Code.
We also altered the fitness functions from single thresholds to step functions. Now each added worker adds the same amount, up to the maximum b attained at colony size m. The maximum gain in both models is the same, but now each worker up to size m adds something. We can model this with Equation 1: if b0 is the birthrate of a solitary queen and b is the birthrate of a eusocial queen in colony size m, then we let the birthrate of queens in smaller colony sizes 1 < i < m be b0 + (i −1)(b − b0)/(m −1). Similarly, we let the queen death rate for colony sizes 1 < i < m be d0 + (i −1)(d − d0)/(m −1). This implementation of the Nowak et al. model was numerically solved using S2 Code.
To solve the ordinary differential equations, we used two numerical methods. For Equations 1–4, Euler's method was used in R to numerically determine the equilibrium population of the system, using a time step of h = 0.1 and a maximum colony size of n = 50 and terminating when either E or S population/number of individuals was less than ε = 0.1 or after a maximum of 50,000 time steps. Equation 1 was also solved with a first-order numerical procedure with the step size 0.1 implemented in MATLAB. The procedure was started with equal numbers of solitary females and eusocial queens (n = 100) and was terminated when either the solitary or eusocial populations were extinct (defined as less than 0.05) or both the solitary and eusocial populations stabilized at a maximum of 200,000 time steps. Both numerical methods successfully reproduced Figure 4 of Nowak et al. [13].
|
10.1371/journal.pntd.0006116 | Combinations of registered drugs reduce treatment times required to deplete Wolbachia in the Litomosoides sigmodontis mouse model | Filarial parasites can be targeted by antibiotic treatment due to their unique endosymbiotic relationship with Wolbachia bacteria. This finding has led to successful treatment strategies in both, human onchocerciasis and lymphatic filariasis. A 4–6 week treatment course using doxycycline results in long-term sterility and safe macrofilaricidal activity in humans. However, current treatment times and doxycycline contraindications in children and pregnant women preclude widespread administration of doxycycline in public health control programs; therefore, the search for shorter anti-wolbachial regimens is a focus of ongoing research. We have established an in vivo model for compound screening, using mice infected with Litomosoides sigmodontis. We could show that gold standard doxycycline treatment did not only deplete Wolbachia, it also resulted in a larval arrest. In this model, combinations of registered antibiotics were tested for their anti-wolbachial activity. Administration of rifamycins in combination with doxycycline for 7 days successfully depleted Wolbachia by > 2 log (>99% reduction) and thus resulted in a significant reduction of the treatment duration. Using a triple combination of a tetracycline (doxycycline or minocycline), a rifamycin and a fluoroquinolone (moxifloxacin) led to an even greater shortening of the treatment time. Testing all double combinations that could be derived from the triple combinations revealed that the combination of rifapentine (15mg/kg) and moxifloxacin (2 x 200mg/kg) showed the strongest reduction of treatment time in intraperitoneal and also oral administration routes. The rifapentine plus moxifloxacin combination was equivalent to the triple combination with additional doxycycline (>99% Wolbachia reduction). These investigations suggest that it is possible to shorten anti-wolbachial treatment times with combination treatments in order to achieve the target product profile (TPP) requirements for macrofilaricidal drugs of no more than 7–10 days of treatment.
| Over the past years, more attention has been brought to neglected tropical diseases including lymphatic filariasis and onchocerciasis. The latter are caused by helminthic parasites and lead to chronic and debilitating symptoms and present a major health burden that also affects the economy of endemic countries. It has been suggested that disease elimination may be possible but an accelerated implementation of proven and cost-effective interventions are needed if the targets for elimination are to be achieved. Recently, an indirect mode of action has been identified, targeting bacterial Wolbachia endosymbionts within the filariae, which also kills the adult parasites, an advantage over the drug currently used for mass drug administration, i.e. ivermectin. Doxycycline has been successfully used in clinical trials, however due to its long regimen as well as restrictions of use in children and pregnant women new drugs or drug combinations are required that overcome these obstacles. Here, we present the filarial parasite Litomosoides sigmodontis as suitable model for the preclinical testing of anti-wolbachial drugs against filariae and show that combinations of already registered drugs with anti-wolbachial efficacy are able to reduce the treatment time dramatically.
| More than 200 million humans are parasitized by filarial nematodes, causing the neglected tropical diseases: lymphatic filariasis, loiasis and onchocerciasis. The lymphatic, ocular and dermatological damages have severe economic and social consequences including poor school performance, low productivity, low income, higher health related costs among infected adults, and a reduced life span [1]. The achievements of onchocerciasis and lymphatic filariasis (LF) mass drug administration (MDA) programmes have considerably reduced transmission and led to the formulation of the goal to eliminate these diseases [2]. However, research is still needed to develop safe and easy to administer macrofilaricides (drugs that kill the adult filarial parasites) to reduce the treatment time needed for MDA programmes and to be used in “problem areas”, e.g. areas in which suboptimal responses and potential resistance to ivermectin (IVM) have emerged or Loa loa is co-endemic. In the latter, MDA programmes have not been implemented due to the risk of severe adverse events in co-infected patients due to rapid killing of L. loa microfilariae (MF) by IVM [3]. The development and implementation of macrofilaricidal drugs will increase cost effectiveness by avoiding unnecessary IVM treatments, particularly in “end-game” scenarios or when shifting from MDA to “test & treat” strategies [4]. Furthermore, implementation of macrofilaricidal drugs will reduce the program times required to eliminate onchocerciasis and LF, which is the aim of the Millennium Development Goals.
Due to their mutualistic association with filarial worms, Wolbachia have been discovered as new target for chemotherapy in filariasis [5, 6]. Treatment of filarial infected animals with tetracycline resulted in the elimination of Wolbachia from filarial tissues in Litomosoides sigmodontis, Brugia malayi, B. pahangi, Onchocerca ochengi, O. lienalis, O. gutturosa and Dirofilaria immitis, prevented parasite establishment and filarial growth and rendered adult worms sterile [7–12]. Wolbachia-negative filarial species such as Acanthocheilonema viteae and Loa loa are not influenced by antibiotic treatment, suggesting that filarial viability and fertility is not affected directly [9, 13]. We have established the first safe macrofilaricide in humans [14–16], doxycycline (DOX), which targets essential Wolbachia endosymbionts, and has proven efficacy in both, lymphatic filariasis and onchocerciasis. An advantage of this mode of action is the slow onset of anti-parasitic activity, thereby avoiding adverse reactions caused by rapid micro- and/or macrofilaricidal activity [17]. However, treatment duration lasts 4–6 weeks and is not an option for children and pregnant or lactating women. The anti-Wolbachia (A-WOL) consortium aims to identify novel anti-Wolbachia drugs, compounds or combinations that are suitable for use in humans [4]. In vitro experiments using a Wolbachia-containing insect cell culture [18] have provided a valuable tool for high throughput screening for anti-wolbachial compounds [19]. As a second step, candidates successful in in vitro Wolbachia-depletion assays have to be tested for their in vivo efficacy, their bioavailability within the mammalian host as well as within the parasite. Choosing the right parasite model and the optimal protocol / time-line is a critical issue to balance high-throughput screening potential and ability to translate results for efficacy or shortening of treatment time against the human parasite. Despite general similarities, each of the existing filarial models has its strengths and limitations [20] and the discovery of macrofilaricidal drugs, i.e. killing directly the adult parasite, has been a critical issue due to varying parasite sensitivity to the drug or bioavailability of the drug within the different hosts used.
Since in anti-wolbachial therapy, the mode of action has been elucidated and it is known that the depletion of Wolbachia eventually leads to parasite death [15], the first step of proving a drug´s in vivo activity is to show that indeed Wolbachia are depleted with a given regimen. Here we present natural infection of BALB/c mice with the filarial nematode L. sigmodontis ([21], formerly known as Litomosoides carinii and later re-named by Bain et al. [22]) as a rodent model for an in vivo screening of anti-wolbachial compounds. The choice of this parasite / host combination was driven by the following advantages: i) immunocompetent mice are susceptible to L. sigmodontis, ii) parasites develop much faster compared to O. ochengi in cattle, iii) the model allows a quick optical screen on day 35 for worms with a stunted growth as a correlate to anti-wolbachial activity, iv) if a macrofilaricidal activity of hits in an already patent infection shall be analysed, the same parasite species can be used in larger rodents, such as Mongolian gerbils, which allow longer follow up times suitable for the slow macrofilaricidal action of anti-wolbachials. In the present study, we use the larval L. sigmodontis model for anti-wolbachial screening and describe the effects of doxycycline at doses that provide sufficient and suboptimal drug exposure in vivo. This enabled us to compare the treatment time of compounds to deplete Wolbachia directly against DOX and revealed that minocycline has a superior efficacy. In addition, rifamycins were tested and rifapentine (RPT) was proven to be more efficacious compared to rifampicin. Finally, combination therapies were investigated and the combination of RPT and moxifloxacin (MOX) showed the strongest reduction of treatment time using both intraperitoneal (ip) and oral administration routes and allowed a strongly reduced treatment time from 14 days to only 2 and 4 days, respectively. These investigations suggest that it may be possible to shorten anti-wolbachial treatment times to match the current target product profile (TPP) for macrofilaricidal drugs [4].
Animal housing conditions and the procedures used in this work complied with to the European Union animal welfare Directive 2010/63 EU. All protocols were approved by the Landesamt für Natur, Umwelt und Verbraucherschutz, Cologne, Germany (AZ 8.87–50.10.35.08.024, AZ 84–02.04.2012.A140).
Experiments were designed to compare Wolbachia reduction in L. sigmodontis worms between different treatment groups, a “doxycycline gold standard" (50 mg/kg/day administered ip for 14 days) and the vehicle control, in a randomized design with multiple arms and a shared vehicle control. The primary endpoint was the Wolbachia single gene FtsZ copy number compared to the vehicle control. As secondary objective FtsZ levels as well as the FtsZ/worm actin ratio were further compared to the gold standard treatment [23]. Sample size calculation was based on the mouse as experimental unit using the mean and SD of the FtsZ copy numbers of 10 worms per experimental unit. At an alpha-level of 0.05 and a power of 0.95 with an effect size of d = 8.18, the estimated groups size was 2 due to the strong Wolbachia reduction efficacy of the gold standard. This calculation was based on a preliminary experiment with five worms per animal and three animals per group. To compensate for possible variation in infections, at least three animals were used in the experimental and control groups. All possible measures were taken to minimize nuisance or the effects of subjective bias when allocating animals to treatment (computer-generated random sequence numbers). Using 4–6 week old female BALB/c mice from one supplier, the bias of the experimental unit was minimized. Furthermore, all animals in one experiment were infected at the same time with the same batch of parasites in order to reduce the variation of infection to the minimum. After the infection, animals were randomly assigned to the treatment groups. Results were assessed in a blinded manner, as the person involved in animal treatment and the person analysing the FtsZ levels were not aware of the treatment groups.
Female BALB/c mice were purchased from Janvier Labs, Saint-Berthevin, France, at the age of 4–6 weeks and were allowed to adjust for 7 days prior to infection. Animals were housed in the animal facilities (conventional, OHB (optimized hygiene barrier)) of the University Hospital of Bonn in individually ventilated cages with access to food and water ad libitum and enrichment. The L. sigmodontis life cycle was maintained at the Institute for Medical Microbiology, Immunology and Parasitology as described earlier [24]. L. sigmodontis lives naturally in the cotton rat (Sigmodon hispidus). During the blood meal on an infected cotton rat the arthropod intermediate host (tropical rat mite Ornithonyssus bacoti) ingests the microfilariae (MF) which moult twice and develop to the infective stage 3 larvae (L3) within 10 days. In a following blood meal the mites transmit the L3 to cotton rats. Similarly, in the case of murine infection the L3 containing mites are allowed to take blood from mice and thereby transmit the L3. In the rodent host the L3 migrate to the thoracic cavity and reach sexual maturity within 25–33 days.
Six to eight week old female BALB/c mice were infected with L. sigmodontis by natural infection as described above. Beginning one day after infection, the mice were given ip injections of 10% DMSO (vehicle control), drugs as indicated in the text and figures, or were left untreated. If not indicated otherwise, all substances were obtained from Sigma Aldrich with a purity of >90% and diluted in PBS, 10% DMSO. Gold standard (DOX ip, 50 mg/kg/day,) and vehicle control were given for 14 days. So adverse events have been observed during treatment. Thirty-five days post infection, the animals were euthanized using isoflorane and worms were recovered from the thoracic cavity by PBS lavage. The female worms were sorted, their lengths measured and individually frozen for DNA extraction. At least 3 mice and 5–10 L. sigmodontis females per mouse were used for each treatment group.
Genomic DNA was extracted from individual worms using the QIAamp DNA mini kit (Qiagen, Hilden, Germany). The Qiagen protocol was used with the following changes: the worms were incubated with proteinase K overnight at 56°C; and Wizard SV96 DNA binding plates (Promega) and vacuum manifold instead of DNA columns were used to bind, wash, and elute the DNA in 50 μL of 10-mM Tris, 0.5-mM EDTA, pH 9. Elution plates were sealed and stored at −20°C.
Reduction/depletion of Wolbachia was monitored by qPCR using Rotorgene (Qiagen) and primers for Wolbachia FtsZ (GenBank Accession No.: AJ010271), a single copy number gene. Filarial actin (GenBank Accession No.: GU971367) was determined to normalize worms of a different size. The genes were quantified from the purified DNA by real-time duplex PCR (qPCR) using the Qiagen’s QuantiTect Multiplex NoROX Kit with the following conditions: 10x HotStar Taq Polymerase buffer (Qiagen), 200 μM dNTP, Primers: LsFtsZ forward primer (5´-cgatgagattatggaacatataa-3´), LsFtsZ reverse primer (5´-ttgcaattactggtgctgc-3´), LsActin forward primer (5´-atccaagctgtcctgtctct-3`), LsActin reverse primer (5´-tgagaattgatttgagctaatg-3´), LsFtsZ taqman probe (5’6-FAM cagggatgggtggtggtactggaa 3’ TAMRA), LsActin taqman probe (5’HEX actaccggtattgtgctcgatt 3’TAMRA), 2.5 units HotStar Taq, and 2 μl DNA in a 20 μl reaction. Final primer concentrations were 500 nM for LsFtsZ and 400 nM for LsActin, final taqman probe concentrations were 25 nM for LsFtsZ and 50 nM for LsActin and 6 mM for the MgCl2 concentration. Genes were amplified in a Rotorgene 3000 (Qiagen) using the following conditions: 1X 15 min at 95°C, 45 cycles of 95°C for 15 sec, 58°C for 30 sec. Fluorescence was acquired on the FAM and JOE channel. Copy numbers for each gene were calculated using a modification of the comparative quantification formula as described in [23]. FtsZ and Actin copy numbers are given per 50 μl single worm DNA extract, of which 2 μl were used for each PCR run.
Data were distributed in a non-parametric fashion, median and interquartile ranges are presented. For comparing the level of Wolbachia depletion in worms, the Mann-Whitney-U test was used to calculate statistical differences either against the vehicle treated or gold standard groups. All p-values are given in S1 Table. Analyses between different experimental groups are given in the figures, when additionally performed. P-values <0.05 were considered to represent significant differences. All statistics were calculated using Graph- Pad Prism version 5.02 for Macintosh.
Experiments were conducted in BALB/c mice, infected naturally with the rodent filarial parasite L. sigmodontis via the bite of tropical rat mites (Ornithonyssus bacoti). Starting the following day, animals were subjected to a daily ip injection of 50 mg/kg (mg/kg) of DOX for 14 days (Fig 1a) and filariae were recovered and analysed after 7 (only 7 days of dosing), 14, 21, 28 and 35 days post infection (dpi) for their Wolbachia content. Fig 1b shows that in DOX treated female worms Wolbachia FtsZ copy numbers were decreased from 7 dpi throughout day 35 pi (Fig 1b). In untreated animals Wolbachia copy numbers increased between day 7 to 14 and 28 to 35 pi, indicating a rapid multiplication of the bacterial endosymbionts during larval development. The increase of bacteria in untreated controls coincides with their moult from L3 into L4 and L4 into young adults. We further investigated the size of the female worms starting at day 14 pi and found a growth retardation after DOX treatment with parasites that did not develop beyond the size of untreated filariae at day 21 pi (Fig 1c). Accordingly, filarial actin copy numbers increased in vehicle treated controls from 7 to 35 dpi (Fig 1d). As a result, the ratio of Wolbachia FtsZ over filarial actin showed an increase between days 7 to 14 and 28 to 35 pi in vehicle treated animals, which was reduced by 96.4 to 99.9% by DOX treatment (Fig 1e).
With this model as a screening tool for anti-wolbachial compounds, it is possible to detect effective compounds by day 14 pi. In the following experiments we chose to use day 35 pi for Wolbachia measurements as this time point has the advantage of a second parameter, i.e. growth inhibition of the worms for a rapid evaluation of compounds.
Next, we investigated optimal as well as suboptimal DOX treatments for Wolbachia depletion in order to set up standards for identification of compounds that are more potent than DOX. To this, we subjected infected animals to a 7, 10 and 14 day treatment of daily ip DOX doses ranging from 6.3 to 50 mg/kg and performed the analysis of female worms at 35dpi. Wolbachia levels were reduced in a dose and treatment time dependent manner (Fig 2a), whereas parasite actin DNA levels were not significantly affected by the treatment (Fig 2b). The ratio of Wolbachia FtsZ and filarial actin DNA was also calculated and is presented in Fig 2c. Examination of the size and developmental stage of the worms revealed a dose and treatment time dependent inhibition of filarial growth (Fig 2d) in accordance to the loss of Wolbachia. In addition, parasites that were recovered from animals with higher DOX dose regimens and longer treatment times remained in the L4 larval stage as determined based on the length of the buccal capsule (white bars, Fig 2e and 2f), suggesting that growth and development is impaired by depletion of Wolbachia. Interestingly, light microscopy analysis indicated no morphological damages, e.g. with regard to the integrity of male spiculae (personal communication Dr. Coralie Martin). Growth inhibition occurred already at lower drug doses that did not achieve the threshold of >99% Wolbachia reduction, indicating that the treatment affected the functional activity of the bacteria and growth inhibition could be achieved with suboptimal Wolbachia reduction. In addition, we analysed the FtsZ and actin signal in the vehicle and gold standard groups (14 days ip 50 mg/kg/d DOX) from all our experiments (n = 23). After DOX, FtsZ levels were reduced on average by 5 log drops, whereas actin levels varied by a maximum of one log. We found that for FtsZ the median was 36 (±8.15x104) in the worms recovered after DOX and 6.6x106 (±1.35x106) in the vehicle group and for the nematode actin 3,4x104 (±7.1 x104) in the DOX and 5.5 x105 (±1.1 x106) in the control group.
Taken together these data show that 50 mg/kg DOX ip for 14 days is suitable to deplete Wolbachia by >99%. For the detection of compounds that are more potent than DOX, a suboptimal regimen, e.g. 25 mg/kg DOX for 10d, is recommended. This dose results in a small but significant reduction of FtsZ copy numbers and more prominently a clear inhibition of growth, suggesting that functional impairment of Wolbachia (e.g.production of molecules that enable parasite development) precedes their disappearance from the parasite tissue. Having set the time point of analysis at day 35 pi, it is possible to identify active compounds quickly by reduction in worm length, even if FtsZ is not or only marginally reduced. Thereby potential drug candidates can be identified with higher potency and treatment regimens can be optimized.
According to the current target product profiles, new macrofilaricidal drugs should have good oral bioavailability. Therefore we performed a dose titration with oral DOX. A dose response in FtsZ reduction was observed when DOX was administered orally (Fig 3a). Also the FtsZ/actin ratio (Fig 3b) and the length reduction (Fig 3c) showed a dose response. Despite the fact that treatment duration may not have been long enough for complete absence of the Wolbachia signal, growth inhibition occurred at 200 mg/kg for 7/10/14 days and 100 mg/kg for 14d and indicates effective inactivation of Wolbachia.
We tested the ability of the L. sigmodontis model to identify compounds with improved activity compared to DOX, using a series of registered antibiotics with known or unknown anti-wolbachial activity. Among these were tetracycline derivatives (tigecycline, minocycline, methacycline) and gyrase inhibitors (sparfloxacin, ciprofloxacin). Animals were subjected to a 10 day (all compounds at 25 mg/kg per day) or 4 day ip treatment course (tetracyclines at 50 mg/kg per day). All tetracyclines administered for 10 days inhibited filarial growth, whereas both gyrase inhibitiors were less effective (Fig 4). Sparfloxacine was reported to be effective at higher doses (160 mg/kg, 14 days) [19].
FtsZ copy numbers further show that tigecycline and minocycline were more potent than DOX (Fig 4a and 4b). Shortened treatment for 4 days with the tetracyclines methacycline and minocycline resulted in a less pronounced Wolbachia decline (83.8 and 88.0% reduction of FtsZ; Fig 4a) and partial growth inhibition (Fig 4c).
We next investigated whether rifamycins are equivalent or faster acting than DOX and analysed rifapentine (RPT) and rifampicin (RIF). Fig 5a–5c show that at a dose of 50 mg/kg ip for 7 days, RPT showed a slight but significant higher activity against Wolbachia compared to RIF (Fig 5a and 5b). Administration of both rifamycins reduced female worm length (Fig 5c).
A combination of drugs can reduce the time of treatment regimens, such as the Denver regimen against TB (MOX in combination with RPT, [25]) and the efficacy of a tetracycline / rifampin combination has already been shown against Onchocerca ochengi, albeit with longer treatment times [26]. We investigated the treatment time of such combinations in this model. We also included minocycline, as it was more potent than DOX (Fig 4). As we expected a high efficacy against Wolbachia, we administered the combination therapy for four days ip. Fig 5d–5f show that similar to the single administration, RPT is more potent than RIF in double combinations. Furthermore, minocycline (MIN) has similar activity as DOX in combination with either RPT or RIF. We also tested whether the MOX/rifamycin combination, as used against tuberculosis, is effective in this model. Whereas MOX alone was only partially effective at 7 dpi, MOX/RPT treatment for only four days showed Wolbachia depletion equivalent to the DOX gold standard administered for 14d (Fig 5d and 5e).
Since some of the double combinations showed efficacy when administered for 4 days, we investigated whether the addition of a triple combination of tetracycline, rifamycin and MOX further reduced the treatment time. We compared DOX (50 mg/kd/d), MIN (50 mg/kd/d), RPT (15 mg/kd/d), RIF (15 mg/kd/d) and MOX (200 mg/kd/bid) in a triple ip combination for 2 and 4 days. As seen before, all combinations were more efficacious if they contained RPT instead of RIF and all treatments tested inhibited filarial growth (Fig 6a–6f). Combining a tetracycline with RPT and MOX resulted in the greatest reduction of Wolbachia. Even a treatment time of 2 days depleted Wolbachia FtsZ by 99.9% when used in combination with RPT, but not with RIF. This finding indicates that anti-wolbachial drugs combinations reduce the treatment time from 14 days to 2 days in the L. sigmodontis larval model. We further tested the triple combination using the oral route and found that 4, 6 or 8 days of treatment reduced the number of bacteria significantly (>99.8%, Fig 7a and 7b) and impaired filarial growth (Fig 7c).
In Fig 5d–5f we observed that also the double combination of ip RPT and MOX was effective in reducing Wolbachia. We compared this combination against a triple combination that included an additional tetracycline. Interestingly, when administered ip for 3 days at a dose of 15 mg/kg (RPT) and 200 mg/kg/bid (MOX), the double combination was equally effective to the triple combination including either 50 mg/kg DOX or MIN (Fig 8a–8c). Similarly, oral treatments for 4 or 7 days with the double combination of MOX and RPT was as effective in reducing Wolbachia as the triple combinations (Fig 9a–9c).
For the previous combinations a high MOX dose regimen was used and we investigated whether a lower MOX dose leads to the same results. Therefore we applied 15 mg/kg (RPT) and 100 mg/kg (MOX) and compared 7 and 14 days of treatment with the single components to 4 and 7 days of treatment with the combination. Fig 10a–10c shows that even at a lower MOX dose a significant Wolbachia reduction was achieved by the combination after 4 days of treatment (99.8%) and a reduction of 99.99% at 7 days of treatment. The single components MOX and RPT were both partially active with 7 days of treatment (96.8% and 83.6%) and highly active with 14 days of treatment (99.99% and 99.6%).
Taken together, a double combination of MOX and RPT, as used in the standard Denver regimen against TB [25], is highly potent in depleting Wolbachia in the L. sigmodontis larval model.
Innovative treatments are urgently needed to improve individual clinical care for patients, reduce human suffering, and ensure advancement towards the WHO and ESPEN (Extended Special Project For Elimination Of Neglected Tropical Diseases) goals for LF and onchocerciasis elimination (http://apps.who.int/gb/ebwha/pdf_files/WHA70/A70_35-en.pdf). Depletion of Wolbachia with DOX has resulted in the development of an alternative anti-filarial chemotherapy, with a 4–6 week treatment course resulting in long-term sterility and most importantly a strong macrofilaricidal effect (even though temporally delayed) [15, 27]. The indirect mode of action results in a slow death of the adult worms and this “soft kill” is a highly desired outcome in order to avoid the rapid release of parasite and Wolbachia products that can induce a strong pro-inflammatory response and thereby adverse side effects [17]. The A-WOL consortium has focused on the discovery and development of new or repurposed antibiotics without the caveats of DOX (long-term treatment, not approved for use in children and pregnant women) [4].
Here we describe the use of the rodent L. sigmodontis model as an in vivo screening system for the discovery and development of anti-wolbachial compounds. This screening procedure that examines anti-wolbachial drug efficacy during L3 to adult worm development is a very sensitive tool to assess anti-wolbachial drug candidates. It has the advantage to identify also non-optimal candidates that need further treatment or formulation adjustments rather than missing potential hits. At this stage parasite replication occurs at a relatively high level and we could show that bacterial function is acutely needed for the further larval development and growth. Using this sensitive stage therefore allows for the identification also of non-optimal candidates that may only need further dose or formulation adjustments, which would be missed otherwise. However one must be cautious as to whether treatment of the larvae is a good predictor for the treatment of adult parasites in humans. It may well be that dose and duration of treatment may be higher and/or longer for an anti-wolbachial drug to deplete Wolbachia also in adult worms. Therefore within the development of a hit/lead into a preclinical candidate, studies must include those that treat animals after the onset of patency. This requires the use of jirds, which without treatment stay patent for at least 6 months.
Also, modelling has to be applied involving PK/PD data from different animals such as mice and jirds (and for PK, larger animals such as dogs) in order to predict appropriate bioavailability and efficacy in humans. Such models have been developed [28]. When starting the treatment (DOX, 50 mg/kg/14d) one day after infection, Wolbachia numbers remain suppressed in the treatment group until the last follow-up in the larval screen at day 35, the time point when larvae have otherwise passed their final moult into adult worms. In contrast, a larval stage (L3-L4, L4-L5) and worm size dependent increase of Wolbachia occurred in the control group (Fig 1b, [29]). Absence of Wolbachia was not only associated with the inhibition of embryogenesis [8, 10] and obvious growth retardation (Fig 1c, [10, 30], but also a developmental arrest at the L4-stage (Fig 2e), an observation that was noticed earlier in Brugia pahangi infected gerbils [31]. This finding suggests a fundamental role for Wolbachia not only for reproduction and survival of the parasites but also for growth and development. In addition to the biological importance itself, this finding adds another benefit to this model. Terminating the experiment at a time point where control worms have moulted into adults allows a quick evaluation of efficacy of a given compound by assessing growth retardation. This is supported by our data using suboptimal concentrations of antibiotics suggesting that growth inhibition precedes absence of Wolbachia, indicating functional inhibition of bacteria-derived molecules required to enable parasite development. This is particularly important in order not to prematurely reject drugs administered at suboptimal dose during screening, or hits that need further hit to lead optimization, or drugs that show bacteriostatic rather than bactericidal effects. In the latter, biological effects of partial Wolbachia depletion like growth inhibition are present, although Wolbachia are still detected by PCR (Fig 4a–4c). The concept of functional rather than numerical impairment of Wolbachia is of biological importance, but difficult to be directly translated into the therapy of human infections, as a different stage is targeted, i.e. the adult worm. The model focussed upon in this study targets the larval stage, a phase in which bacterial replication occurs at very high levels and a potential repopulation of Wolbachia may occur. In human clinical trials however it could be shown that a greater than 90% reduction of Wolbachia does not lead to reoccurrence of bacteria, a threshold that is needed for successful treatment efficacy.
To establish this model for different routes of administration, we compared the ip and oral administration routes and performed dose titration experiments. After both, oral and ip application of DOX, a dose dependency occurred with regard to bacteria depletion and inhibition of filarial growth (Figs 2 and 3). Despite the fact that the oral application is less effective than the ip administration due to DOX’s known poor oral bioavailability [32], oral or ip application of a compound can be tested in this model, depending on the requirements of the research question and the pharmacokinetic profile of the compound. We further suggest that for the identification of compounds with better potency than DOX, a DOX regimen reduced in both dose and time should be used (25 mg/kg/10d) and compared to the gold standard (50mg/kg/14d). An example for such a drug is MIN, another broad-spectrum tetracycline antibiotic that has already been found to be more active than DOX in vitro using Onchocerca gutturosa adult males [33]. MIN was among the top hits of the A-WOL screening activities [4] and its efficacy was recently confirmed using the Wolbachia containing insect cell line C6/36 [19] and adult Brugia malayi screens [34] and in a human clinical trial [35].
Using the suboptimal regimen (25mg/kg/d/14d), DOX showed intermediate levels of Wolbachia reduction, whereas the biological activity of the bacteria was already blocked, as seen by filarial growth inhibition. In contrast, tigecycline and MIN were even more potent in Wolbachia reduction compared to DOX. Whereas the application of tigecycline is intravenous and therefore does not conform to the current TPPs for macrofilaricidal drugs, MIN is a potential candidate, although it has, as a tetracycline, the same contraindications as DOX [34]. Rifamycins may overcome those contraindications and our study showed that ip treatments with RIF and RPT alone for one week reduced the Wolbachia burden by 97 and 99%. Recent PK/PD analysis and in vivo B. malayi and O. ochengi studies suggest that orally administered RIF at an elevated dose of 35 mg/kg leads to a Wolbachia reduction greater than 90% predicting cure in B. malayi and O. ochengi after 7 and 14 days of treatment, respectively [28]. Due to this increased in vivo potency elevated RIF dose treatment is thought to achieve equivalent efficacy as the long-course DOX treatment and is proposed for phase II trials.
For some important infectious diseases, combination therapies with antibiotics have been established. An important example is the standard Denver regimen for treatment of tuberculosis, consisting of a 6-months course of a combination of isoniazid and RIF administered daily, supplemented by ethambutol and pyrazinamide in the first two months. Approaches to shorten the treatment time or reduce the number of daily observed treatments is an important goal that will increase adherence to the treatment and thereby treatment success. Investigations have compared daily RIF versus RPT and have shown that once weekly RPT and daily RIF have the same efficacy with similar cure rates [36], however, with an increase of the risk of bacteriological relapse [37]. Further studies focused on the increase of the dose of RPT from 10 to 15 mg/kg to compensate for its high protein binding that may be partially responsible for the suboptimal activity observed in once-weekly regimens [38]. The 15 mg/kg dose of RPT has been well tolerated in humans and increasing the dose of RPT has demonstrated enhanced sterilizing activity in a mouse model of tuberculosis [39]. In our studies, we showed that RPT was superior to RIF depletion with regard to Wolbachia depletion in the L. sigmodontis model. Additional improvement was achieved by the inclusion of MOX, a fluorquinolone of the fourth generation with potent bactericidal activity [25]. This regimen has been investigated recently in a clinical trial in humans for TB and it has been shown that the 6-month regimen with a weekly administration of high-dose RPT and MOX is as active as the control regimen and the 4 month regimen is non-inferior to the control regimen [40]. However this treatment is not recommended for general use other than progressing active TB in children and pregnant women.
We tested in our model whether a combination of a tetracycline, rifamycin and MOX were able to reduce the treatment duration and found high efficacy in the reduction of bacterial loads with RPT and MIN being slightly more potent than RIF and DOX.
Testing all double combinations that could be built from the triple combinations revealed that the combination of RPT (15mg/kg) and MOX (2 x 200mg/kg) had the best efficacy in both administration routes (oral or ip). The RPT plus MOX combination was equivalent to the triple combination (>99% Wolbachia reduction).
These investigations suggest that it may be possible to shorten anti-wolbachial treatment times to 7 days or less in humans, meeting the criteria of the current TPP for macrofilaricidal drugs. In addition, those shortened treatment times have the advantage that they are unlikely to cause resistance in Mycobacterium tuberculosis. Further PK/PD studies similar to Sharma et al. [34] and screening in other models of adult filarial infection [41] should be performed to confirm pan-filarial activity and to better identify the human bioequivalent doses to translate these results successfully to clinical trials.
|
10.1371/journal.ppat.1004672 | The Intracellular Bacterium Wolbachia Uses Parasitoid Wasps as Phoretic Vectors for Efficient Horizontal Transmission | Facultative bacterial endosymbionts are associated with many arthropods and are primarily transmitted vertically from mother to offspring. However, phylogenetic affiliations suggest that horizontal transmission must also occur. Such horizontal transfer can have important biological and agricultural consequences when endosymbionts increase host fitness. So far horizontal transmission is considered rare and has been difficult to document. Here, we use fluorescence in situ hybridization (FISH) and multi locus sequence typing (MLST) to reveal a potentially common pathway of horizontal transmission of endosymbionts via parasitoids of insects. We illustrate that the mouthparts and ovipositors of an aphelinid parasitoid become contaminated with Wolbachia when this wasp feeds on or probes Wolbachia-infected Bemisia tabaci AsiaII7, and non-lethal probing of uninfected B. tabaci AsiaII7 nymphs by parasitoids carrying Wolbachia resulted in newly and stably infected B. tabaci matrilines. After they were exposed to infected whitefly, the parasitoids were able to transmit Wolbachia efficiently for the following 48 h. Whitefly infected with Wolbachia by parasitoids had increased survival and reduced development times. Overall, our study provides evidence for the horizontal transmission of Wolbachia between insect hosts by parasitic wasps, and the enhanced survival and reproductive abilities of insect hosts may adversely affect biological control programs.
| Vertically-transmitted facultative bacterial endosymbionts are common in invertebrates, and affect traits as diverse as the mode of sexual reproduction, speciation, and susceptibility to pathogens. Horizontal transmission of endosymbionts is thought to be infrequent in most species, and not to contribute to their spread through populations. Here we demonstrate that parasitoid wasps can act as vectors, transmitting the endosymbiont Wolbachia between whitefly hosts at a high rate. The ovipositors and mandibles of parasitoids can be contaminated with Wolbachia when probing infected whitefly. If these parasitoids then probe Wolbachia-free hosts and the whitefly survive, it will result in a stably infected line with increased fitness. Such vector-borne transmission may explain why endosymbionts are so widely distributed, and why genetically similar symbionts are often found in phylogenetically distant organisms.
| Vertically transmitted intracellular bacteria often live in symbioses with their arthropod hosts [1]. These endosymbionts may be obligate (essential for host survival) or facultative, in which case they can increase or decrease host fitness [1–3]. When endosymbionts are obligate they typically share a long evolutionary history with their hosts and are found within specialized cells [4]. The facultative endosymbionts tend to have a more recent association with arthropods, but are nonetheless very common among arthropods. For instance, 40% of insect species are estimated to be infected with Wolbachia [5]. Similar to obligate endosymbionts, facultative endosymbionts are also transmitted vertically from mother to offspring with high fidelity and this is considered the primary transmission pathway [4]. However, through their evolution these endosymbionts have also been transmitted horizontally between different species [1,6,7], with closely related endosymbionts occurring in phylogenetically distant insects [1,7–12]. In the last two decades there have been multiple studies reporting evidence of Wolbachia transmission within and between both phylogenetically close and more distant species either through phylogenetic or transinfection studies [8,13–18].
The spread of endosymbionts in field populations by horizontal transmission, however, has received comparatively little attention, and the mechanisms driving horizontal transmission are only recently becoming apparent [1,3]. Horizontal transmission of symbionts has been documented when infected and uninfected parasitoid wasps develop within the same host insect [19–21], when infected males mate with uninfected females [22], when infected and uninfected whiteflies feed on the same host plant and the symbiont moves through the phloem sap [23], and when symbionts are acquired from the environment [24]. It is also possible for endosymbionts to be transmitted by vectors. Hamiltonella defensa and Regiella insecticola symbionts can be efficiently transmitted when a parasitoid wasp sequentially stabs an infected then an uninfected aphid [25]. Similarly, Spiroplasma can be transmitted between Drosophila species by ectoparasitic mites [26]. Here we report the efficient phoretic transfer of Wolbachia by the parastoid Eretmocerus sp. nr. furuhashii from infected whitefly Bemisia tabaci AsiaII7 to uninfected individuals..
The whitefly Bemisia tabaci is a small hemipterous insect that feeds on phloem sap of numerous host plants. It is currently considered as a complex of at least 24 distinct cryptic species that are morphologically indistinguishable but markedly differ in host range, ability to transmit viruses, insecticide resistance and the endosymbionts they are infected with [27–30]. The B. tabaci species complex harbours various bacterial symbionts, including Wolbachia. Wolbachia has been found in the ovarian cells of the host and at the circumference of and inside the bacteriocytes [31]. The AsiaII7 B. tabaci is indigenous to China and is one of the common cryptic species in South China (formerly “Cv” biotype) [32]. It was first discovered on variegated laurel Codiaeum variegatum in Guangzhou in 2006, and further studies indicated that this whitefly can damage various economically important ornamental plants [33]. Eretmocerus sp. nr. furuhashii is one of the dominant parasitoids of whitefly B. tabaci in South China [34,35]. It is a primary, solitary parasitoid which oviposits externally between the nymphal host and the leaf surface, but nonetheless penetrates whitefly nymphs with its mouth parts and ovipositor to examine them before laying and to feed.
The current study was motivated by a six-year long observation from 2007–2012 of two subcolonies, one for rearing whitefly B. tabaci AsiaII7 (hereafter “WR”, 5 cages) and the other for rearing the parasitoid E. sp. nr. furuhashii using AsiaII7 as hosts (hereafter “PR”, 6 cages). We found that the Wolbachia infection rate of whitefly housed with parasitoids was higher than that without parasitoids, gradually increasing during the surveillance period (S1 Fig.). Meanwhile, the prevalence of Wolbachia in E. sp. nr. furuhashii wasp was lower than that of the whitefly, but also gradually increased (S1 Fig.). Although this is not a replicated experiment, it led us to hypothesise that the coexistence of parasitoid and whitefly may promote the transmission of Wolbachia between different individuals of whitefly. Here we show that non-lethal host inspection of the whitefly nymphs by the wasp can transfer Wolbachia. The acquisition of Wolbachia benefited its host’s fitness in terms of faster immature development and increased adult survival.
To test whether parasitoids can vector Wolbachia, we allowed them to oviposit and feed on Wolbachia-infected whitefly and then transferred them to a cage of uninfected whitefly. By observing the behavior of E. sp. nr. furuhashii in the laboratory, we found that among the uninfected AsiaII7 nymphs which were visited by parasitoids for feeding or oviposition checking, 35.8% (38 of 106) died, adult parasitoids emerged from 34.0% (36 of 106), and 30.2% (32 of 106) resulted in adult whitefly. Wolbachia screening by PCR revealed that 93.8% (30 of 32) of the newly emerged whitefly individuals became infected after surviving the parasitoid penetration.
In order to confirm infection by Wolbachia, samples of both AsiaII7 nymphs (including the donor whitefly from PR cage, the newly infected nymphs and their offspring) and E. sp. nr. furuhashii were selected randomly for fluorescence in situ hybridization. We found different distributions of Wolbachia in the two-level trophic system. For the parasitoid E. sp. nr. furuhashii, Wolbachia was found both in their mouthparts and ovipositor after they fed on or penetrated whitefly hosts, but not in the ovaries (Fig. 1). Among the donor AsiaII7 nymphs there were two distinct patterns of infection. In some individuals Wolbachia was “scattered” in both the ovaries and somatic tissues (Fig. 2 A, B; 6 of 30 individuals), while in others it was “confined” to the bacteriocytes and ovaries (Fig. 2 C, D; 24 of 30 individuals).
In uninfected whitefly that had been visited by a contaminated parasitoid, Wolbachia was not visible during the first 72 h. This is presumably due to its low density, as qRT-PCR showed a continuous increase of Wolbachia during this time, indicating that the insect had been infected and the bacterium could replicate (S2 Fig.). In the offspring of whitefly visited by contaminated wasps, Wolbachia is clearly visible in 3rd instar nymphs. In all cases the symbiont had the scattered pattern (15 of the 16 AsiaII7 offspring harboured scattered Wolbachia, while Wolbachia was not detected in the other one).
To identify the Wolbachia strain in AsiaII7 whiteflies and Eretmocerus parasitoids, we sequenced five MLST genes and the wsp gene. This revealed that the Wolbachia strain from E. sp. nr. furuhashii was the same as in AsiaII7 (sequence type 388). By comparing the sequences those in the Wolbachia MLST database (http://pubmlst.org/wolbachia/) and by constructing phylogeny, this was identified as a new sequence type “ST388” (S1 Table, S3 Fig.).
To be stably transmitted between generations Wolbachia must infect the female germ line. We established populations of whitefly from individuals that had been probed by contaminated parasitoids and measured the prevalence of Wolbachia for five generations. We found that 85.0 to 90.0% of the F1 to F5 progeny of the newly-infected whiteflies were Wolbachia-positive (Fig. 3), indicating the vertical transmission and long term persistence of Wolbachia in the newly-infected AsiaII7 populations. While we did not directly measure vertical transmission rates (uninfected whitefly may have been among the parents of each generation), these results indicate efficient transmission from parent to offspring. The negative control was uninfected in all 5 generations.
Wolbachia DNA can be detected by PCR for at least 5 days in E. sp. nr. furuhashii after they fed or oviposited on Wolbachia-infected AsiaII7 nymphs (Fig. 4A). However, parasitoids contaminated with Wolbachia were only able to transmit it for the first 48 h, indicating that the bacteria are gradually losing their infectivity with time (Fig. 4B).
We established populations of whitefly with and without Wolbachia, and found that they had significant biological differences. Compared to uninfected AsiaII7, infected whiteflies developed significantly faster (Fig. 5, General Linear Mixed Model: t = 3.38, df = 6, P = 0.01) and had significantly increased longevity (Fig. 5, Cox proportional hazards mixed model: z = 3.86, P = 0.0001). There was no significant effect of Wolbachia on either juvenile survival (Fig. 5, Generalised linear mixed model: z = 1.6, P = 0.11) or fecundity (Fig. 5, General Linear Mixed Model: t = 0.64, df = 6, P = 0.55). Wolbachia was associated with a slightly higher proportion of female offspring, but this was not significant (Fig. 5, Generalised linear mixed model: z = 1.7, P = 0.09).
To understand how horizontal transmission will affect the prevalence of a symbiont, we simulated the spread of Wolbachia through a population. To do this we modified existing models of Wolbachia dynamics to include an additional parameter w, which is the product of the number of new cases generated by horizontal transmission from a single infected host in an otherwise uninfected population and any reduction in vertical transmission efficiency of Wolbachia in the newly infected hosts relative to those infected from their mother. The value of w will be determined by factors including the frequency with which a vector visit hosts, the time that contaminated vectors remain infectious, and the probability that a contaminated vector infects an uninfected whitefly. Field data shows that the majority of wasps are contaminated by Wolbachia DNA (S4 Fig.), and our lab data suggests many of these will be infectious (DNA is detectable for 5 days, the infectious period is 2 days). We also find that about a third of nymphs survive parasitism. Therefore, w will be below the parasitism rate, but not dramatically so.
We first investigated the effect of horizontal transmission in the absence of any reproductive manipulation or effect on host fitness (Fig. 6A). The outcome depends on the rate of horizontal transmission relative to the rate at which the infection is lost due to imperfect vertical transmission. When horizontal transmission is low, the infection will tend to be lost, but moderate rates of parasitism (w = 0.06) can result in Wolbachia invading and reaching a stable equilibrium frequency. Fixation is prevented by imperfect vertical transmission.
Next, we examined how horizontal transmission will affect a strain of Wolbachia that causes the reproductive manipulation cytoplasmic incompatibility (CI; Fig. 6B). Combining CI and horizontal transmission can dramatically alter the outcome, with horizontal transmission allowing the near fixation of a strain that would otherwise be lost. The reason for this is that there is a threshold prevalence that CI strains must exceed in order to invade populations. Horizontal transmission can lower or eliminate this threshold. For the parameters used in Fig. 6B, if the starting prevalence is high, then Wolbachia can invade and reach a high prevalence without horizontal transmission.
Finally, we examined how horizontal transmission affects a strain that provides a fitness benefit to the host (Fig. 6C). Fitness benefits alone can allow the bacterium to invade the population as a mutualist and reach an intermediate equilibrium frequency maintained due to imperfect vertical transmission. However, horizontal transmission can accelerate the invasion and increase the equilibrium prevalence.
Wolbachia has well-described effects on host physiology and reproduction that increase its prevalence in populations [1,5,36–38]. For instance, cytoplasmic incompatibility, female-biased offspring sex ratios and providing protection against viral infection are potent methods that may increase the frequency of infected relative to uninfected matrilines. These mechanisms all rely on bolstering the vertical transmission of the bacterium. On the other hand our study shows that within populations Wolbachia may spread horizontally as well. If horizontal transmission proves to be important in other insects, it is tempting to speculate that it may sometimes explain the observation of Wolbachia infections with no known phenotypic effect [39]. Furthermore, parasitoids could also potentially vector Wolbachia to novel species, such as the other cryptic species of B. tabaci which share parasitoids with AsiaII7. This could partly explain the plethora of phylogenetic evidence of discordance between host and symbiont phylogenies [8,11,12,40,41].
Parasitoid-vectored horizontal transmission of Wolbachia may be important in natural populations. AsiaII7 is an indigenous cryptic species of B. tabaci in south China, and we have surveyed field populations of this whitefly species and its parasitoids from 2006–2012. Wolbachia was detectable in about 30–40% of wild E. sp. nr. furuhashii throughout this time (S4 Fig.). Furthermore, most AsiaII7 individuals were Wolbachia-infected, and the prevalence increased during this time (S4 Fig.). Numerous factors may have contributed to this increase, including fitness benefits and horizontal transmission (see results).
It has been suggested that horizontal transmission from parasitoids to their hosts would be unlikely as parasitized hosts die [8]. The mode of transmission we have described here relies on the fact that parasitoids do not always kill hosts with which they interact. Given mixed-instar nymphs, Eretmocerus parasitoids of whitefly B. tabaci usually exhibit a clear preference for feeding on 1st and 4th instar nymphs, leaving 2nd and 3rd instars for oviposition, while Encarsia parasitoids prefer to feed on 1st and 2nd instars, leaving 3rd and 4th instars for oviposition [42,43]. Our FISH and PCR screening revealed that the ovipositors and mouthparts of parasitoids get contaminated with Wolbachia when they probe or feed on infected whitefly nymphs. When these parasitoids probe uninfected whitefly nymphs, about one third survived, and of those 93.8% became infected with Wolbachia. Similar to Eretmocerus emiratus parasitoids developing in Rickettsia-infected B. tabaci hosts [7], Wolbachia failed to penetrate the oocytes of the parasitoid, but this is not required for parasitoids to transmit the bacterium effectively.
Subsequent to horizontal transmission, the bacteria have to be transmitted efficiently from mother to daughter in order to persist in the population. In many such cases infections fail to persist [1,7] or are transmitted with poor fidelity [16,17,23,44,45]. For example, Wolbachia was transmitted at a low rate (3.2%) from an infected Drosophila simulans host to the parasitoid Leptopilina boulardi, and subsequently was lost within four generations [15]. In our experiment, Wolbachia persisted at a stable prevalence (85.0–90.0%) for at least five generations in AsiaII7 whitefly, suggesting that the vertical transmission rates were high. Our result support the recent study of Gehrer and Vorburger [25] in which they demonstrated that parasitoids can transfer the bacterial endosymbionts Hamiltonella defensa and Regiella insecticola by sequentially stabbing infected and uninfected individuals of the black bean aphid Aphis fabae. Similar to our results, this established new, heritable infections.
Vectors that transmit infection can either be biological vectors, where the infective agent replicates or develops in the vector, or phoretic (mechanical) vectors where it does not. Three lines of evidence indicated that the parasitoids are acting as phoretic vectors. First, FISH indicates that the Wolbachia is only found on the surface of the wasp and does not infect its tissues or ovaries. Second, the infectivity of the wasps rapidly declines after they have become contaminated with Wolbachia. Third, the Wolbachia infection was routinely monitored in population cages, and wasps were always uninfected unless they were recently exposed to infected AsiaII7 whiteflies. Therefore, it seems most likely that the wasp is acting as little more than a dirty needle. Our FISH experiments revealed that in all the newly infected AsiaII7 individuals, Wolbachia had a scattered distribution across tissues, while vertically infected insects tended to have Wolbachia largely confined to the ovaries. Therefore we cannot be sure that the confined distribution of Wolbachia in whitefly ovaries can also be horizontal transmitted from one individual to another by parasitoids.
Facultative endosymbionts, including Wolbachia, can change the fitness or biology of their hosts. For example, facultative symbionts in the pea aphid (Acyrthosiphon pisum) can protect their hosts against entomopathogenic fungi and parasitoid wasps, ameliorate the detrimental effects of heat, and influence host plant suitability [1,46–48]; B. tabaci Mediterranean cryptic species infected with Wolbachia showed decreased juvenile development time, increased juvenile survival, increased adult life span and an increased percentage of female progeny [49]. Similarly, our study also showed that Wolbachia-infected AsiaII7 survived for longer and developed faster. There was also an increase in the proportion of daughters produced, but our sample size was small and the effect was not statistically significant. Wolbachia always had imperfect vertical transmission, with infected females producing uninfected daughters, which may explain why natural populations are a mixture of infected and uninfected individuals.
Our study revealed changes in fitness caused by Wolbachia that can potentially increase the rate of population growth of whitefly, leading to more severe damage to crops. Fitness measurements can often prove context-dependent and hard to replicate, so it is important to reproduce our results in the field. Furthermore, our laboratory results should be replicated in case there are unaccounted for differences between our Wolbachia-infected and Wolbachia-free populations. However, assuming similar effects are found in nature, our results suggest that parasitoids used for biological control could result in unintended negative consequences when the control agent also allows the horizontal transmission of Wolbachia between matrilines or even species of pests. With increased chemical resistance of whitefly in many countries, parasitoids have become an important biocontrol agent to manage whitefly infestations, especially in greenhouses. The B. tabaci species complex is known to be host to at least 56 species of parasitoids, mostly from the genera Eretmocerus and Encarsia [35], and some of these have been commercially produced and applied. Usually one B. tabaci cryptic species can be parasitized by dozens of parasitoid species, and one parasitoid species can attack several B. tabaci cryptic species, or even more distantly related species of whitefly such as the greenhouse whitefly Trialeurodes vaporariorum. This may provide many opportunities for endosymbionts to horizontally transmit between different whitefly species, potentially causing the detrimental fitness changes. We propose that this unintended negative impact of parasitoids in pest biological control cannot be ignored.
Adult AsiaII7 whiteflies were first collected from variegated laurel plants, Codiaeum variegatum (L.), in Guangzhou in 2006 [50]. They were then reared on hibiscus in two separate greenhouses (15×7m) in the South China Agricultural University (SCAU). These were covered with PVC film on the top and nylon net (70 mesh) on the sides. Two subcolonies were set up using 5–6 cages in the laboratory, one of which was used for AsiaII7 whitefly rearing only and the other was used for parasitoid E. sp. nr. furuhashii rearing. In the parasitoid rearing cages, new and clean hibiscus plants were added to replace the old plants when needed. This allows both the B. tabaci hosts and Eretmocerus parasitoids to stably coexist because removing old plants takes the parasitoid pupae away. Both the AsiaII7 whitefly and Eretmocerus parasitoids were reared in an insect growth chamber at 26.0±0.5°C, 70–80% relative humidity, 14:10 h (L:D) photoperiod.
AsiaII7 samples were collected three times per year (Apr-May, Aug-Sep, Dec-Jan) from both the two subcolonies between 2007–2012. In each survey, 30–50 samples per plant and 3 plants were collected in each subcolony using an aspirator. All samples were preserved immediately in 95% ethanol. Whitefly species identity was confirmed using mtCOI sequences [51,52]. From each survey, 20 whitefly adults from each subcolony were selected randomly for PCR-based detection of Wolbachia following the methods of Ahmed et al. [53]. Briefly, Wolbachia was detected using primers for the wsp, ftsZ and 16S rRNA genes. Both negative controls (ddH2O) and a template DNA quality control (primary endosymbiont Portiera 16S rRNA gene to indicate the DNA quality of extraction) were included. Results were further confirmed using FISH. Alongside the whitefly collection, 20 individuals of E. sp. nr. furuhashii adults were also sampled for Wolbachia detection with PCR and FISH. To identify the Wolbachia strain in parasitoids and whiteflies, we used multi locus strain typing methods described in Baldo et al. [54].
To establish Wolbachia-positive lines of AsiaII7, male/female pairs from the AsiaII7 rearing cages were allowed to reproduce on hibiscus, one pair per plant. Once the F1 progeny emerged, ten pairs of newly emerged adults from one parent were selected at random. Five pairs were tested for the presence of Wolbachia using PCR; the remaining five pairs were caged individually on hibiscus leaves. Any line that contained uninfected individuals was discarded until a line in which all five of the tested pairs were positive for Wolbachia was identified. This screening process, based on five pairs being selected from each generation, was continued for 10 generations prior to use in experiments. After the initial selection of the line, all individuals screened were positive for Wolbachia.
To establish an outbred population, which we used for measuring fitness and in other experiments, we mixed together a large number of these lines. In total 50 to 80 pairs of parents were randomly selected, and all the validated Wolbachia positive couples were released into a rearing cage to reproduce. The progeny that emerged were used for the fitness measurements below.
To establish a pure culture of Wolbachia-free whitefly, the process followed that used to establish the infected line except that here the initial line was the one where all five pairs tested negative. The infected and uninfected populations were sampled from the same original greenhouse population. Both the infected and uninfected lines of AsiaII7 were then kept in rearing cages (60×60×60 cm) in separated air-conditioned insect growth chamber at 26.0±0.5°C, 70–80% relative humidity, 14:10 (L:D) photoperiod and light intensity of approximately 3000 Lux. Wolbachia-infection status in the cages was monitored on a monthly basis by PCR.
In the laboratory, we examined horizontal transmission of Wolbachia from the infected AsiaII7 nymphs to the uninfected nymphs via parasitoids. About 10 pairs of uninfected AsiaII7 adults were released into a leaf cage to reproduce on hibiscus for 48 h. Progeny from these adults developed to 2nd-3rd instar, at which point all but 60 nymphs were removed. Before the experiment, three mated parasitoid females of E. sp. nr. furuhashii (2-day age) were first introduced into a leaf cage in which there are 2nd-3rd instar Wolbachia-positive AsiaII7 nymphs for 2h of feeding and oviposition. After that, the parasitoids were transferred into the leaf cages with the 60 Wolbachia-free 2nd-3rd instar nymphs. The parasitoids were allowed to feed and oviposit for another 2 h, oviposition behaviors were observed using a binocular dissecting microscope, and nymphs visited by the wasp were marked with indelible ink. We used 240 nymphs of AsiaII7 in total (60 in each repeat), of which 106 were visited by parasitoids. Seven days later, the number of nymphs that were parasitized (larval parasitoids are visible in whitefly hosts), the number of nymphs that were subjected to the insertion of an ovipositor but survived, and nymphs that were parasitized but died were recorded. Meanwhile, the emerged whiteflies and parasitoids were both collected for Wolbachia detection by PCR following the methods of Ahmed et al. [53]. All the transmission experiments were done at 26.0±0.5°C, 70–80% relative humidity, 14:10 h of L:D photoperiod, and the four repeats were replicated contemporaneously.
The FISH procedure to detect Wolbachia in whitefly AsiaII7 and parasitoid E. sp. nr. furuhashii followed the method of Xue et al. [49] and Sakurai et al. [55]. About twenty Eretmocerus female adults (2-day old) were recaptured 24–48 h after they had been released to attack abundant Wolbachia positive AsiaII7 nymphs in a leaf cage. For the whitefly samples, we investigated forty 3rd instar AsiaII7 nymphs from the Wolbachia-infected donor PR cages (thirty were finally FISH photographed successfully), a similar number of the Wolbachia newly infected AsiaII7 nymphs via parasitoid and their F1 generation offspring were randomly selected for FISH detection respectively.
Two 5’ rhodamine labeled Wolbachia probes (targeted 16S rRNA of Wolbachia) were used to increase the signal: W2: 5’-CTTCTGTGAGTACCGTCATTATC-3’ and W3: 5’-TCCTCTATCCTCTTTCAATC-3’. Stained samples were washed thoroughly twice in the same buffer at 48°C for 20 min. All the samples were observed using an inverted fluorescence microscope (Nikon Eclipse Ti-U). Specificity of the detection was confirmed using the uninfected whiteflies as controls.
In order to quantify the increase of Wolbachia after it was transmitted into Wolbachia negative AsiaII7 whiteflies, 60 early 3rd instar AsiaII7 nymphs that have been fed or probed by those Wolbachia-carrying Eretmocerus parasitoids were screened out and divided into three groups. Then the three groups were used for Wolbachia qRT-PCR detection 24–72 h after they were fed or probed by parasitoids using the primers and protocols described in Xue et al. [49].
In order to know the stability of Wolbachia infection in the newly infected AsiaII7 B. tabaci, a pair of newly infected adults of whitefly were randomly selected and introduced into a leaf cage for 48 ho. Eggs were allowed to develop to F1 adults, at which point 10 of them were randomly selected and tested for Wolbachia. Then another pair of F1 generation adults were randomly selected again and released into a new leaf cage to reproduce F2 generation, among which 10 individuals of F2 adults were selected randomly for PCR detection. This procedure was repeated through to the F5 and only female adults were selected for PCR detection in each generation. A control experiment was conducted in which uninfected AsiaII7 whitefly was reared and checked for Wolbachia infection from F1 to F5 generations in separated cages. Both the Wolbachia infected and uninfected whitefly lines were reared in the insect growth chamber and four replicates were conducted in each generation contemporaneously.
After E. sp. nr. furuhashii fed or penetrated whitefly nymphs, FISH showed that Wolbachia was found in their mouthparts and ovipositors, but not in their ovaries. We therefore hypothesized that these wasps were contaminated temporarily by Wolbachia and can transfer this endosymbiont to other hosts by feeding or oviposition. To confirm this hypothesis, 70 to 80 mated females of E. sp. nr. furuhashii were divided into seven groups and released into seven leaf cages with abundant 2nd-3rd instar Wolbachia-positive AsiaII7 nymphs. Parasitoids were allowed to oviposit or feed on the whitefly hosts for 48 h, then all of them were recaptured, introduced into seven Petri dishes (9 cm diameter), and fed with 10% honey water. Hereafter one group of the female parasitoids per day was used to detect the presence of Wolbachia DNA by PCR according to the method of Ahmed et al. [43]. To extract DNA, 10 parasitoids were homogenized in one tube due to the tiny titer of Wolbachia.
In order to confirm the time that Wolbachia on the mouthparts and ovipositors of Eretmocerus wasps is alive and transmittable, five two-day-old female wasps were collected after they parasitized several Wolbachia-positive AsiaII7 nymphs. They were then released into a 5 cm diameter Petri dish, and fed with 10% honey solution but not given hosts to parasitize. After 24, 48, 72 and 96 h, these parasitoids were released into leaf cages to parasitize 2nd-3rd instar nymphs of Wolbachia-negative AisaII7. After 4–6 days, DNA was extracted from individual whitefly adults that emerged, and these samples were tested for the presence of Wolbachia by PCR. The experiment at each time point (24–96 h) was repeated four times contemporaneously.
To compare the biology of Wolbachia uninfected and infected AsiaII7 populations, their development time, immature survivorship, sex ratio, fecundity and adult longevity were investigated according to Qiu et al. [34]. Briefly, 10 pairs of newly emerged AsiaII7 adults were selected at random from the uninfected or infected populations (see above for how these were established). Each pair was introduced into a leaf cage on a hibiscus plant to lay eggs for 48 h (10 leaf cages in each replicate of the experiment), after which all but 20 eggs per leaf cage were removed. The twenty F1 eggs were allowed to develop to adults in the leaf cage, and their emergence were checked daily until all nymphs completed their development. The developmental time of one randomly-selected individual in each leaf cage was recorded (in total 10 individuals were selected in each replicate).
Meanwhile, a pair of Wolbachia positive and negative F1 emerged adults (0–12 h age, which is before whitefly begin laying eggs) were randomly selected from each of the 10 leaf cages, and each pair was introduced into a new leaf cage on a hibiscus plant for reproduction. Their eggs were counted daily until the female died, and newly emerged males were supplemented if the original male died before the experiment ended. The fecundity and longevity of F1 generation adults was recorded. In addition, about 100 F1 adults were randomly selected and their sex ratio recorded. All the experiments were performed in the insect growth chamber (26.0±0.5°C, 70–80% relative humidity, 14:10 h of L:D photoperiod). Care was taken to select uniform hibiscus plants. We replicated the experiments measuring fitness trait four times contemporaneously (i.e. there were four replicates of the Wolbachia-infected and four replicates of the Wolbachia-free treatments).
The data on the fitness effects of Wolbachia was analysed using a series of statistical models. In all the experiments measuring fitness components, both the Wolbachia-infected and Wolbachia-free treatments were replicated four times, with 10 pairs per replicate (80 pairs in total). The 8 replicates (4 Wolbachia-infected and 4 Wolbachia-free) were performed contemporaneously. In all cases the model included the full replicate structure of the experiment and no model simplification was performed.
The data on adult longevity consisted of observations on the lifespan of a single adult offspring from each pair (10 observations per replicate). Each adult was observed until it died, so there is no right censoring of the data. We analyzed this data using a Cox proportional hazard model, which included a fixed effect of Wolbachia infection status and a random effect of replicate (also see S2 Table). The model was fitted using the function coxme in R.
The data on fecundity consisted of a count of the number of eggs laid by each pair (10 egg counts per replicate). The data on immature development consisted of the development time of a single nymph from each pair (10 nymphs per replicate). We treated both of these datasets as Gaussian and analyzed them using a general linear model. The model included a fixed effect of Wolbachia infection status, a random effect of replicate and a residual that accounts for variation between pairs within each replicate.
The data on immature survival consisted of counts of nymphs from each pair that died or survived (10 ratios of dead: alive nymphs per replicate). The data on sex ratio consisted of counts of male and female offspring that were produced by each replicate. Both immature survival and sex ratio were treated as ratios and analysed using generalized linear mixed models with a binomial error structure. The model included a fixed effect of Wolbachia infection status, a random effect of replicate for the survival data, and a residual variance to allow for over-dispersion (i.e. variation between pairs within each replicate that was greater than that expected under binomial sampling).
The figures are plotted using parameters estimated from these models, with the exception of adult longevity where mean survival times were estimated for the figure using a model similar to that described above for the fecundity data. For the binomial data, the means and standard errors were back-transformed from the logit scale into proportions for plotting. The model parameters were estimated by maximum likelihood using the R functions lme and glmer in the package lme4 [56].
Multi locus sequence typing (MLST) was used to identify Wolbachia in both AsiaII7 whitefly and E. sp. nr. furuhashii parasitoids. Five MLST genes, gatB (wMel locus WD_0146), coxA (WD_0301), hcpA (WD_0484), ftsZ and fbpA (WD_1238), together with the Wolbachia surface protein gene (wsp) were used for Wolbachia strain identification. The process is described in Baldo et al. [54].
Concatenated sequences of the seven most closely related Wolbachia sequence types (STs) associated with different host species were selected from the MLST database and used for comparisons with the STs isolated from other whitefly cryptic species, AsiaII7 and the parasitoid, E. sp. nr. furuhashii. Three STs from supergroup A, D and F Wolbachia were used as outgroups. The best model was chosen using the Bayesian Information Criterion (BIC) in MEGA5 [57]. Phylogenetic analysis based on the concatenated data of the five Wolbachia MLST loci (2079 bp) was undertaken using maximum parsimony (MP) and maximum likelihood in MEGA5 [57] and PAUP version 4.0b [58]. T92+G model was used during ML tree constructions.
To understand how horizontal transmission can affect the dynamics of Wolbachia within insect populations we adapted the model of Hoffman et al. [59], who provide a full explanation of the model, the parameters and its assumptions. This model assumes discrete generations. Infected females produce a fraction μ of uninfected ova. The fecundity of Wolbachia-infected females relative to Wolbachia-free females is denoted by F. When the strain induced cytoplasmic incompatibility, the relative hatch rate of incompatible crosses is H. Let sh = 1-H and sf = 1-F. The prevalence (proportion of infected adults) at generation t is pt. In the absence of horizontal transmission, the prevalence at generation t+1 is denoted k. Following Hoffmann et al. [59]:
k=pt(1−μ)(1−sf)1−sfpt−shpt(1−pt)−μshpt2(1−sf)
We then added an extra step in which vectors can transmit Wolbachia horizontally. We assume that Wolbachia in newly infected whitefly has the same effect on female fecundity and cytoplasmic incompatibility as in stably infected lines. We also assume that the rate of hosts being visited by vectors is constant across generations, that wasps visit whitefly randomly with respect to their Wolbachia infection status, and that a contaminated wasp can only infect the next whitefly that it visits (i.e. Wolbachia is lost by probing an uninfected host). Male and female whiteflies are assumed to be equally likely to be visited by wasps and to transmit or be infected by Wolbachia. For simplicity, we also modelled the effects of a vector that never kills its host, although unpublished simulations suggest this produces similar results to a parasitoid that can kill its hosts. We define a new parameter w as the number of new cases generated by horizontal transmission from a single infected host in an otherwise uninfected population. It is possible that females that are infected by horizontal transmission may transmit Wolbachia at a reduced rate. If this is the case, then w can be considered as the number of product of the number of new cases generated by horizontal transmission and the proportionate reduction in the transmission rate in these females. Therefore, the prevalence at generation t+1 is given by:
Pt+1=k+wk(1−k).
The value of w will be determined by factors including the frequency with which vectors visit hosts relative to the time that the contaminated vectors remain infectious and the probability that a contaminated wasp infects an uninfected whitefly.
|
10.1371/journal.pgen.1000600 | RNA Gain-of-Function in Spinocerebellar Ataxia Type 8 | Microsatellite expansions cause a number of dominantly-inherited neurological diseases. Expansions in coding-regions cause protein gain-of-function effects, while non-coding expansions produce toxic RNAs that alter RNA splicing activities of MBNL and CELF proteins. Bi-directional expression of the spinocerebellar ataxia type 8 (SCA8) CTG CAG expansion produces CUG expansion RNAs (CUGexp) from the ATXN8OS gene and a nearly pure polyglutamine expansion protein encoded by ATXN8 CAGexp transcripts expressed in the opposite direction. Here, we present three lines of evidence that RNA gain-of-function plays a significant role in SCA8: 1) CUGexp transcripts accumulate as ribonuclear inclusions that co-localize with MBNL1 in selected neurons in the brain; 2) loss of Mbnl1 enhances motor deficits in SCA8 mice; 3) SCA8 CUGexp transcripts trigger splicing changes and increased expression of the CUGBP1-MBNL1 regulated CNS target, GABA-A transporter 4 (GAT4/Gabt4). In vivo optical imaging studies in SCA8 mice confirm that Gabt4 upregulation is associated with the predicted loss of GABAergic inhibition within the granular cell layer. These data demonstrate that CUGexp transcripts dysregulate MBNL/CELF regulated pathways in the brain and provide mechanistic insight into the CNS effects of other CUGexp disorders. Moreover, our demonstration that relatively short CUGexp transcripts cause RNA gain-of-function effects and the growing number of antisense transcripts recently reported in mammalian genomes suggest unrecognized toxic RNAs contribute to the pathophysiology of polyglutamine CAG CTG disorders.
| We describe several lines of evidence that RNA gain-of-function effects play a significant role in spinocerebellar ataxia type 8 (SCA8) and has broader implications for understanding the CNS effects of other trinucleotide expansion disorders including myotonic dystrophy type 1, Huntington disease like-2, and spinocerebellar ataxia type 7. The SCA8 mutation is bidirectionally transcribed resulting in the expression of CUGexp transcripts from ATXN8OS and CAGexp transcripts and polyglutamine protein from the overlapping ATXN8 gene. These data suggest that SCA8 pathogenesis involves toxic gain-of-function effects at the RNA (CUGexp) and/or protein (PolyQ) levels. We present three lines of evidence that CUGexp transcripts play a significant role in SCA8: 1) CUGexp transcripts accumulate as ribonuclear inclusions that co-localize with MBNL1 in selected neurons; 2) loss of Mbnl1 enhances motor deficits in SCA8 mice; 3) SCA8 CUGexp transcripts trigger alternative splicing changes and increased expression of the CUGBP1-MBNL1 regulated CNS target, GABA-A transporter 4 (GAT4/Gabt4) which is associated with the predicted loss of GABAergic inhibition within the granular cell layer in SCA8 mice. Additionally, alternative splicing changes and GAT4 upregulation are induced by CUGexp but not CAGexp transcripts. From a therapeutic viewpoint, it is promising that this change is reversed in cells overexpressing MBNL1.
| Spinocerebellar ataxia type 8 (SCA8), is a slowly progressive neurodegenerative disease caused by a CTG•CAG expansion that primarily affects the cerebellum [1],[2]. The disease is transmitted in an autosomal dominant pattern with reduced penetrance and the expansion mutation was originally shown to be expressed as a CUG expansion (CUGexp) transcript in the 3′region of an untranslated gene, ATXN8OS. BAC transgenic mice expressing the SCA8 expansion (SCA8 BAC-EXP, [CTG]116) but not a control repeat (SCA8 BAC-CRTL, [CTG]11) from the endogenous human promoter develop progressive motor deficits and a loss of cerebellar GABAergic inhibition [3]. Unexpectedly, SCA8 patients and BAC-EXP mice were found to have 1C2-positive intranuclear inclusions in Purkinje cells and brainstem neurons that result from the expression of a nearly pure polyglutamine protein (ataxin 8) from a novel gene spanning the repeat in the opposite CAG direction (ATXN8). The expression of CUGexp transcripts from ATXN8OS in addition to CAGexp transcripts and a polyglutamine protein from ATXN8 suggests that SCA8 involves toxic gain-of-function effects at both the RNA (CUGexp) and protein (PolyQ) levels. An alternative hypothesis is that the SCA8 expansion affects the expression of an overlapping gene, KLHL1 and although Klhl1 knockout mice have a subtle phenotype, the relevance of this model to the human disease is unclear [4].
Substantial evidence that CUGexp RNAs are toxic comes from studies of the neuromuscular disease myotonic dystrophy (DM) where CUGexp (DM1) and CCUGexp (DM2) RNAs alter the activities of two families of alternative splicing factors, the MBNL and CELF proteins [5],[6]. Multiple lines of evidence support the model that these expansion transcripts cause disease specific clinical features. First, transgenic mouse models of DM1 in which CUGexp transcripts are expressed in the 3′ UTR of the DM1 (DMPK), or the unrelated human skeletal muscle actin (HSA), gene cause skeletal muscle myotonia and myopathy similar to human DM skeletal muscle pathology [7],[8]. In addition, CUGexp or CCUGexp transcripts accumulate as ribonuclear inclusions in DM1 and DM2 patient skeletal muscle [9]–[11] and alter the localization or regulation of RNA binding proteins CUGBP1 [12],[13] and MBNL1 [14],[15]. Additional studies show expression of CUGexp transcripts induce alternative splicing changes in skeletal muscle genes associated with disease symptoms including the chloride channel (CLCN1) and insulin receptor genes (IR) [16]–[19]. Similar alternative splicing events in numerous genes have now been shown to be misregulated in myotonic dystrophy, often causing aberrant expression of fetal isoforms in adult tissue [5].
Evidence that MBNL1 loss-of-function plays a role in DM was suggested by studies showing MBNL1 co-localizes with CUG or CCUG ribonuclear inclusions [10] with additional strong support coming from Mbnl1 isoform knockout mice (Mbnl1ΔE3/ΔE3) that recapitulate several aspects of the multisystemic disease pathology and misregulated splicing events characteristic of DM [20]. Furthermore, MBNL1 has been shown to bind directly to intronic elements of genes misregulated in DM cardiac and skeletal muscle (cTNT and IR) and to promote alternative splicing patterns normally found in adult tissue [21]. Taken together, these data suggest that expression of CUGexp or CCUGexp transcripts induces alternative splicing changes in DM skeletal and cardiac muscle by sequestration and functional loss of MBNL1 protein.
In addition to skeletal and cardiac muscle disease, behavioral and cognitive changes in DM suggest that CUGexp and CCUGexp transcripts also cause CNS effects. DM1 CUGexp transcripts form ribonuclear inclusions that co-localize with MBNL1 in temporal lobe neurons of DM1 patients [22]. Additionally, alternative splicing changes in a number of CNS transcripts (NMDA R1, APP, MAPT/Mapt, Mbnl1) are found in humans and mice [22],[23]. Although the mechanism for these CNS splicing changes is unknown, these data suggest that DM, and possibly the much shorter SCA8 CUG expansion transcripts, cause RNA gain-of-function effects in brain.
In support of the model that SCA8 CUGexp transcripts affect MBNL1 regulated pathways, expression of human SCA8 cDNA transcripts (ATXN8OS) in Drosophila photoreceptor neurons induce a late-onset, neurodegenerative phenotype genetically enhanced by the loss of the fly MBNL1 orthologue muscleblind [24]. This result is similar to fly models of DM1 [25],[26] and consistent with additional evidence for RNA gain-of-function effects in DM.
In this study, we present the first evidence that SCA8 CUGexp transcripts cause “RNA gain-of-function” effects in the brain. First, we demonstrate that SCA8 CUGexp transcripts form hallmark ribonuclear inclusions that co-localize with MBNL1 in humans and mice and that genetic loss of Mbnl1 enhances motor coordination deficits in SCA8 BAC-EXP mice. Additionally, we show that expression of ATXN8OS CUGexp transcripts dysregulate MBNL1-CUGBP1 pathways in the CNS and trigger downstream molecular changes in GABA-A transporter 4 (Gabt4) regulation through an RNA gain-of-function mechanism.
To determine if SCA8 ATXN8OS transcripts form ribonuclear inclusions similar to those found in DM, fluorescence in-situ hybridization (FISH) was preformed on SCA8 human and SCA8 BAC-EXP mouse cerebella using a Cy5-labeled (CAG)10 oligonucleotide probe. In human SCA8 autopsy tissue, CUG positive inclusions were found in the cerebellar cortical layers (red dots) by confocal microscopy. These CUG ribonuclear inclusions are distinguishable from background lipofuscin auto-fluorescence which appears as a yellowish-brown perinuclear staining (Figure 1A). Although ribonuclear inclusions were seen in all three SCA8 autopsy brains examined, the foci varied in distribution, size, and number between SCA8 cases and compared to ribonuclear inclusions found in DM1 cerebellum. In cerebellar tissue from SCA8 patients with 1000 and 400 CTG repeats, single CUG foci were frequently found in the nuclei of molecular layer (ML) interneurons and the Bergmann glia surrounding the Purkinje cells in the granule cell (GC) layer, while multiple smaller foci were typically found in Purkinje cells (PC). Although qRT-PCR shows CUGexp transcripts are expressed at comparable levels in all three SCA8 autopsy cases, ribonuclear inclusions were not reproducibly found in the brain with 109 CTG repeats, and were only detected in a single molecular layer interneuron (not shown). While these data suggest that ribonuclear inclusions are less likely to form in patients with shorter expansions, CUG RNA foci were readily detectable in SCA8 BAC transgenic mice which express similarly sized SCA8 ATXN8OS (CUG)116 transcripts under the control of the endogenous human promoter (Figure 1A, bottom row). Similar to the RNA foci found in SCA8 autopsy tissue with 400 and 1000 repeats, the foci in these mice have a similar distribution in the cerebellar cortex (Purkinje cells, Bergmann glia, and molecular layer interneurons). Additionally, ribonuclear inclusions were found in the deep cerebellar nuclei in our mice, an area of the brain we were not able to examine in the human autopsy cases. No foci were seen in mice expressing normal repeat (CTG)11 or non-transgenic littermates and additionally, no CAG foci were detected using a CTG oligonucleotide probe (not shown).
FISH combined with immunofluorescence (IF) was performed to determine if the SCA8 ribonuclear inclusions co-localize with MBNL1. In human cerebellar sections, MBNL1 co-localizes with the CUG foci in molecular layer interneurons (Figure 1B). In contrast, SCA8 and DM1 Purkinje cells have clearly detectable nuclear CUG foci but these foci do not co-localize with MBNL1, which is predominantly expressed in the cytoplasm (Figure 1B, middle and bottom row arrows). Similar to the human results, SCA8 mice expressing the expansion, but not the control repeat, have CUG-positive ribonuclear inclusions that co-localize with Mbnl1 in molecular layer interneurons and deep cerebellar nuclei (Figure 1C) but not in Purkinje cells (not shown).
Because co-localization of MBNL1 with CUGexp ribonuclear inclusions is thought to lead to functional impairment of nuclear MBNL1 activity in DM1, we tested the hypothesis that RNA gain-of-function effects in SCA8 contribute to the motor deficits in SCA8 BAC-EXP mice via Mbnl1 depletion. Mice from a low copy SCA8 BAC-EXP5 line (SCA8 BAC-EXP5+/−) were crossed to heterozygous Mbnl1 isoform knockout mice (Mbnl1+/ΔE3). Mice from the SCA8 BAC-EXP5+/− line were selected for these studies because these animals, which have normal rotarod performance at 26 weeks of age, do not develop a movement disorder phenotype until >1 year of age [3]. Additionally, although homozygous Mbnl1ΔE3/ΔE3 knockout mice model the multisystemic features of DM pathology heterozygous Mbnl1+/ΔE3 mice are similar to wild type and do not develop myotonia or other skeletal muscle changes [20]. To test if genetic Mbnl1 loss enhances the SCA8 CNS phenotype we crossed heterozygous SCA8 BAC-EXP5+/− mice to heterozygous Mbnl1+/ΔE3 knockout mice and tested the F1 offspring [(SCA8+/− (n = 11); Mbnl1+/ΔE3 (n = 13); SCA8+/−; Mbnl1+/ΔE3 (n = 17) and non-transgenic littermates (n = 9)] for motor deficits at 26 weeks of age by rotarod analysis. The latency to fall in seconds (sec) was recorded for 4 trials per day over 4 consecutive days. Mean differences between groups for each testing day were determined by taking the average of the last 3 trials. Consistent with previous results [3], no significant difference in mean latency to fall was found between the SCA8 BAC-EXP5+/− mice (409.52±22.51 sec) and non-transgenic littermates (414.83±22.18 Sec.; P = 0.86). Although heterozygous Mbnl1+/ΔE3 mice did have a significantly different mean latency to fall (342.15±18.75 sec.) compared to non-transgenic littermates [F(1, 86) = 6.22; P = 0.012], double mutant offspring (SCA8+/−; Mbnl1+/ΔE3) performed significantly worse (273.69±14.13 Sec) than either the singly mutant Mbnl1+/ΔE3 (P = 0.003) or SCA8 BAC-EXP5+/− littermates [F(1,102) = 31.27; P<0.00001] (Figure 2). These data provide genetic evidence that loss of Mbnl1 plays a role in SCA8 pathogenesis and support the hypothesis that SCA8 CUGexp transcripts affect Mbnl1 regulated pathways in the brain.
The discovery that CUGexp RNA and MBNL1 protein co-accumulate in ribonuclear foci in neurons suggests that, similar to the dysregulation of MBNL/CELF pathways in DM muscle, the accumulation of SCA8 CUG expansion transcripts might lead to the dysregulation of developmentally regulated splicing patterns in SCA8 brain. Further support for this hypothesis comes from alternative splicing changes found in MBNL1 and NMDAR1 in SCA8 autopsy brains (Figure 3A) which mimic previously reported changes in DM1 [22],[27]. To identify novel splicing targets that might be affected by Mbnl1 loss or increased Cugbp1 activity we used cross-linking and immunoprecipitation (CLIP) analysis [28] on mouse postnatal day 8 (P8) hindbrains. CLIP was successful in identifing RNA targets of Cugbp1 but not Mbnl1. Because CUG-BP1 and MBNL1 have been shown to be antagonistic regulators of alternative splicing of a number of different targets from work done in the myotonic dystrophy field, we reasoned that CUG-BP1 CLIP tags would be also be good candidate targets for MBNL1 regulation. At P8, hindbrain can be readily dissociated into a cell suspension which was then exposed to UV-light to fix RNA-protein complexes formed in vivo. This procedure avoids artifacts associated with immunopurification of unfixed proteins which may redistribute during in vitro manipulation. Cross-linked RNA-protein complexes were treated with RNase T1, to generate relatively short (60–200 nt) CLIP RNA tags, and immunopurified with a monoclonal antibody against Cugbp1. Complexes were then subjected to electrophoresis/electroblotting and filter-retained RNA tags were identified following cDNA conversion, amplification and DNA sequencing. A total of 315 Cugbp1-associated RNA tags were identified representing 206 genes with 53 multi-hit tags (Figure 3B and Table S1). While the majority of these tags were intronic (64%), in agreement with a previous study on the splicing regulator Nova1 [28], a significant number (25%) were positioned within 3′ untranslated regions (UTRs). Sequence analysis revealed that Cugbp1 CLIP tags were enriched in UG repeats consistent with previous three-hybrid and SELEX studies which have indicated that the highest affinity sites for CUGBP1, CUGBP2 and CELF4 are (UG)n, and (UGUU)n repeats [29],[30]. A recent study identified a GU-rich 11-mer element (GRE) UG(U)3G(U)3GU, which is enriched in the 3′-UTRs of short-lived transcripts, as a CUGBP1 binding motif [31]. Interestingly, only one Cugbp1 3′ UTR CLIP tag (Slc22a5), and two intronic tags (Lrrc9, Samd12), contained this GRE motif (see Table S1).
From this list of potential Cugbp1 targets we focused on the gamma-aminobutyric acid (GABA-A) transporter 4 gene, Gabt4, because previous in vivo functional imaging studies showed a loss of cerebellar GABAergic inhibition in our SCA8-BAC-EXP mice [3]. Further analysis of the Gabt4 CLIP tag showed that the putative Cugbp1 binding site maps to exon 7 (and overlaps the exon 7 5′ splice site) which is highly conserved in human and mouse. The gene name in mouse is Gabt4/Slc6a11 and in human GAT3 but for clarity referred to here as Gabt4 and GABT4, respectively.
Since Gabt4 was identified as a putative target of CUGBP1 by CLIP analysis and because increased Gabt4 expression could explain the increased cortical loss of GABAergic inhibition previously reported in our mice by reducing GABA at the synapse, we hypothesized that SCA8 CUGexp RNA dysregulates Cugbp1 and Mbnl1 pathways resulting in an increase in Gabt4 expression. Consistent with this idea, we found significant increases in cerebellar Gabt4 protein and RNA levels by protein blot (5.36±1.11 fold; p = 0.003) and qRT-PCR (2.72±0.68 fold p = 0.0015) in SCA8 BAC-EXP1 mice compared to non-transgenic littermates (Figure 4A and 4B) while no increase was seen in SCA8 BAC-CTRL animals (Figure 4B). Further supporting the hypothesis that Gabt4 increases in SCA8 are caused by sequestration of Mbnl1 by SCA8 CUGexp transcripts, Gabt4 protein and transcript levels are also higher in Mbnl1ΔE3/ΔE3 knockout mice compared to strain matched (129/B6) non-transgenic littermate controls with a 2.49±0.083 (P<0.001) and 2.29±0.32 (P = 0.013) mean fold increase by protein blot and qRT-PCR, respectively (Figure 4A and 4B). The increases in Gabt4 protein were reproducible in both high copy number SCA8 BAC-EXP lines studied (BAC-EXP1 and BAC-EXP2) but more variable in mice from the lower copy number BAC-EXP5 line (data not shown) consistent with the decreased penetrance reported previously in this line. In addition, no similar changes were seen in Gabt1 levels, another member of GABA-A transporter family expressed in cerebellum indicating that changes seen in Gabt4 are not caused by a general upregulation of GABA-A transporters (Figure 4C).
Immunofluorescence studies show expression of Gabt4 protein is primarily localized to the granular cell layer of the cerebellar cortex and the deep-cerebellar nuclei (DCN) and that no overt change in this distribution is seen (Figure 4D) between wildtype and SCA8 BAC-EXP1 mice. To further characterize the loss of inhibition phenotype and the possible role of Gabt4 we examined the SCA8 BAC mice for functional changes in the granular layer of the cerebellum, the site of highest Gabt4 expression using flavoprotein optical imaging in response to whisker pad stimulation in SCA8 mice [32],[33]. Activation of cerebellar granule cells by mossy fibers is in part controlled by Golgi cell mediated feedback that produces GABAergic inhibition of granule cells [34]. Because the clearance of synaptically released GABA in the granular layer is predominately mediated by Gabt4, up-regulation of Gabt4 would be expected to reduce this Golgi mediated inhibition and enhance the responses to mossy fiber input. The imaging data collected confirm this prediction. Whisker pad stimulation evokes a patch-like response (Figure 4E) consistent with previous electrophysiological and imaging studies [32],[35]. There is a significant increase in both the intensity and area of the response in Crus II in the SCA8 BAC-Exp (n = 5) compared with FVB mice (n = 7) (Figure 4, p<0.05). Therefore, up-regulation of Gabt4 in the granular layer is associated with the expected increase in the response to mossy fiber inputs activated by peripheral stimulation.
In summary, these data demonstrate that Gabt4, a gene identified by CLIP analysis as a Cugbp1 target, is upregulated at the RNA and protein levels in both our SCA8 BAC-EXP1 and Mbnl1ΔE3/ΔE3 knockout animals but not in the SCA8 BAC-CTRL mice expressing a normal length CUG11 repeat. Additionally, Gabt4 upregulation in these mice is associated with the predicted loss of GABAergic inhibition within the granular cell layer.
To determine if GABT4 upregulation also occurs in humans, we examined steady-state RNA and protein levels in human SCA8 autopsy brains. While expression of the SCA8 CUGexp and CAGexp transcripts and GABT4 overlap in both the cerebellum and the frontal lobe, only frontal lobe tissue was suitable for analysis because of significant cell loss in the cerebellum caused by neurodegeneration in SCA8 patients. Total RNA extracted from adult SCA8, DM1 and control and from 26-week fetal frontal cortex was examined by qRT-PCR using primers to exons 1 and 2. SCA8 autopsy brains (n = 3) showed increased GABT4 transcripts levels compared to adult control (p<0.01) (Figure 5A) and similar levels to those found in fetal tissue. A similar trend of increased GABT4 protein was also seen in SCA8 and fetal brain compared to adult control or DM1 tissue (Figure 5B).
Because exon 7 of Gabt4 was identified as a potential Cugbp1 binding site, we investigated if increases in GABT4 expression could be related to misregulation of exon 7 alternative splicing. Consistent with this hypothesis, all three human SCA8 autopsy brains showed a shift in alternative splicing favoring inclusion of GABT4 exon 7 containing transcripts compared to adult control (Figure 5C). Preferential exon 7 inclusion was also seen in control human fetal autopsy tissue but not in adult DM1 tissue and is correlated with the increases in GABT4 RNA and protein seen in SCA8 adult and control fetal brain (Figure 5). Interestingly, no exon 7 splicing shifts or increases in GABT4 expression were found in DM1 autopsy tissue, possibly reflecting differences in spatial or temporal expression of the SCA8 and DM1 CUGexp transcripts.
Similar to the dysregulation of other genes in DM, SCA8 CUGexp transcripts trigger a shift in the ratios of alternatively spliced human GABT4 (+/− exon 7) transcripts that resemble those found during fetal development. Sequence analysis shows that transcripts skipping exon 7 would lead to the introduction of a premature termination codon (PTC) which would be predicted to target (−) exon 7 transcripts for nonsense mediated decay (NMD). These data suggest a model in which developmental alternative splicing changes in human normally lead to lower levels of GABT4 in adult tissue and higher levels in SCA8 and during fetal development.
Similar to humans, Gabt4 is also up-regulated in the SCA8 BAC-Exp and Mbnl1ΔE3/ΔE3 mice. Although alternative splicing of exon 7 has not yet been detected in the mouse by RT-PCR, exclusion of mouse exon 7 would also create a PTC (in exon 10) that would be predicted to lead to NMD. Further studies are needed to determine if NMD in the mouse leads to more efficient degradation of (−) exon 7 transcripts which prevent their detection or if Gabt4 upregulation in the mouse occurs via another mechanism.
To test directly if increases in GABT4 are induced by SCA8 CUGexp or CAGexp transcripts, we examined their effects in human neuroblastoma SK-N-SH cells. Transient transfections were performed using minigenes (Figure 6A) designed to express SCA8 CUG (SCA8-CTGexp) or CAG expansion transcripts (SCA8-REV CAGexp). Cells expressing SCA8 CUGexp transcripts show significant increases in GABT4 RNA levels by qRT-PCR relative to untransfected cells (p = 0.007) or to cells transfected with vector alone (p = 0.006) while expression of the SCA8 CAGexp construct compared to vector alone had no effect (p = 0.72) (Figure 6B). Additionally, transient transfections of minigenes without ATXN8 and ATXN8OS flanking sequence show poly-CUG105 but not poly-CAG105 transcripts up-regulate GABT4 RNA (p = 0.001) (Figure 6C). Further analysis using primers flanking exon 7, show cells expressing higher levels GABT4 transcripts also preferentially express higher ratios of exon 7 included transcripts. (Figure 6B and 6C).
To test if GABT4 expression is antagonistically regulated by CUGBP1 and MBNL1, we transfected SK-N-SH cells using GFP-tagged human MBNL1/41 and CUGBP1 minigenes capable of inducing alternative splicing changes in cell culture [21],[36]. As above, alternatively-spliced exon 7 transcripts were assayed by RT-PCR with primers located in GABT4 exons 6 and 8 and primers spanning exons 1 and 2 were used to assess endogenous levels of GABT4 by qRT-PCR. Cells overexpressing CUGBP1 (p<0.0001) or SCA8 exon A CTGexp minigenes show increases in GABT4 RNA (p = 0.003) and a concomitant increases in protein and a splicing shift favoring exon 7 inclusion compared to vector alone (Figure 7A–7C). While no change in GABT4 RNA or protein was seen in cells overexpressing MBNL1/41 alone, overexpression of MBNL1/41 and SCA8 CUGexp transcripts reverses the increase in GABT4 RNA (p<0.0001) and protein (Figure 7A and 7C) triggered by SCA8 CUGexp transcripts alone and restores exon 7 alternative splicing ratios to the normal adult pattern (Figure 7B). Sequence analysis of the GABT4 RT-PCR products confirm the upper and lower bands include and exclude exon 7 respectively and that the (-) exon 7 transcripts have a premature stop codon in exon 8. Taken together, these results are consistent with a model in which SCA8 CUGexp transcripts alter the regulation of GABT4 by sequestration of MBNL1 and/or an increase in the expression or activity of CUGBP1.
The SCA8 CTG CAG mutation is bidirectionally expressed and produces both CUG and CAG-expansion transcripts and a nearly pure polyglutamine expansion protein [3]. We investigated RNA gain-of-function effects in SCA8 and present three lines of evidence that CUGexp transcripts play a role in SCA8. First, we demonstrate CUGexp transcripts accumulate as ribonuclear inclusions in selected cells and that these RNA foci co-localize with Mbnl1 in a subset of neurons in SCA8 patients and mice. Second, we show genetic loss of Mbnl1 enhances motor coordination deficits in low-copy SCA8 BACexp mice. Third, we demonstrate SCA8 CUGexp transcripts trigger increased expression of a CUGBP1-MBNL1 regulated CNS target, GABT4, in both mice and humans as well as in a human cell culture model and demonstrate the predicted loss of GABAergic inhibition within the granular cell layer occurs in these animals.
Although CNS effects are a clinically important feature of myotonic dystrophy, mechanistic studies have focused on skeletal and cardiac muscle and little is known about the effects of CUGexp transcripts in brain. In addition, the antagonistic relationship between CUGBP1 and MBNL1, previously documented in heart and skeletal muscle, has not been demonstrated in the CNS. In this study we used CLIP analysis to identify putative Cugbp1 CNS targets. To explore the molecular basis of the loss of GABAergic inhibition we investigated changes in expression levels and alternative splicing of the Cugbp1 CLIP target, Gabt4, in our mice. We show that overexpression of CUGBP1 in human SK-N-SH cells, or depletion of MBNL1, result in an upregulation of GABT4 that mimics the in vivo changes in steady state levels caused by CUGexp transcripts. These data provide the first evidence that CUGexp RNA gain-of-function effects in the brain involve the dysregulation of CUGBP1-MBNL1 pathways. These data also suggest that MBNL1 overexpression, which has been demonstrated to be therapeutic in skeletal muscle, might also be an effective treatment to reverse pathological changes associated with expression of CUGexp transcripts in the CNS. Additionally, the long list of other putative Cugbp1 targets identified in mouse brain provides an important future resource for the identification of additional CNS genes dysregulated in SCA8 and DM.
Previous studies in myotonic dystrophy suggest that the expression of CUGexp transcripts, together with MBNL1 and CELF proteins and their downstream target genes need to be coordinated temporally and spatially for disease pathogenesis. Therefore, defining the temporal and cell specific ATXN8OS expression pattern will be crucial for understanding disease pathogenesis and interpreting future results. We identified for the first time in SCA8, CUG-RNA foci in Purkinje cells, molecular layer interneurons and the deep cerebellar nuclei. In addition, we show MBNL1/Mbnl1 colocalizes with SCA8 CUG foci in molecular layer interneurons and the DCN. Interestingly, although Purkinje cells had nuclear CUG foci in both DM1 and SCA8, co-localization with nuclear MBNL1 was not observed in these cells. Further studies will need to be done to determine if MBNL1 is expressed in the nucleus of these cells or if changes in MBNL1 splice forms affect its ability to bind to the CUG expansion transcripts. Taken together, these results suggest that temporal and spatial expression patterns of expansion transcripts and the overlap in expression of specific RNA binding proteins and their downstream target genes are likely to underlie the susceptibility of specific cells to RNA gain-of-function effects and the clinical difference between DM and SCA8.
To investigate the broader significance of MBNL1 in SCA8, we tested the effects of Mbnl1 depletion on a behavioral phenotype in our SCA8 mice by crossing heterozygous Mbnl1+/ΔE3 animals with a low-copy SCA8-BAC-EXP5 line. The phenotypic enhancement of the rotarod deficits found in the doubly transgenic animals with reduced Mbnl1 suggests that expression of CUGexp transcripts in SCA8 and the subsequent downstream effects on Mbnl1 are sufficiently significant to contribute to the movement disorder phenotype found in SCA8.
GABT4 upregulation in human SCA8 autopsy tissue and cell culture studies supports a model in which GABT4 levels are regulated by alternative splicing changes and the NMD pathway. Similar to the dysregulation of other genes in DM1 (CLCN-1, IR) [5], the expression of CUGexp but not CAGexp transcripts [37] triggers a shift in the ratios of alternatively spliced GABT4 transcripts that resembles those found during fetal development.
Further studies are needed to determine if GABT4 alternative splicing changes occur throughout the brain and to directly show if alternative splicing of exon 7 and NMD regulate GABT4/Gabt4 expression levels in both humans and mice. Additional studies are also required to directly test the hypothesis that GABT4 upregulation causes the decrease in GABAergic inhibition seen in our SCA8 BAC-EXP mice.
While these data provide the first evidence for RNA gain-of-function effects in SCA8, it is also possible that the polyQ expansion protein contributes to the disease. In contrast to other disorders in which polyQ expansions are expressed as part of a mature protein, the SCA8 CAGexp is expressed as a nearly pure polyQ tract. While previous studies have shown that pure polyQ repeats are toxic in Drosophila and mice, the relative contribution of RNA and protein gain-of-function effects in SCA8 still needs to be assessed.
In this study we provide several lines of evidence that RNA gain-of-function effects play a significant role in SCA8 and show that SCA8 CUGexp transcripts affect alternative splicing patterns controlled by MBNL1 and CUGBP1 in the mammalian brain. While SCA8 is the first reported disease in which a single expansion mutation expresses both a polyglutamine protein and CUGexp transcripts, bidirectional expression has also been recently described in other triplet expansion disorders. For example, sense (CUGexp) and antisense (CAGexp) transcripts have been reported at the DM1 locus [38]. Similarly, Huntington disease Like 2 (HDL2), another disease in which CUGexp transcripts form ribonuclear inclusions [39], also has 1C2-positive inclusions (1C2 is an antibody that recognizes polyglutamine expansions), suggesting bidirectional expression may also occur in that disease [40],[41]. Additionally, bidirectional expression across the FMR1 CCG CGG repeat in Fragile X tremor ataxia patients has also been reported [42]. Our data showing that relatively short (∼110 repeats) CUGexp transcripts can cause dysregulation of MBNL1/CUGBP1 regulated pathways and the growing number of antisense transcripts recently reported in mammalian genomes [43], highlight the need to look for CUG transcripts expressed at other loci traditionally associated with polyglutamine expansion disorders.
Fluorescent in-situ hybridization (FISH) and immunofluorescence (IF) was performed on frozen parasagittal cerebellar sections (6 µm) as described [10]. Sections were fixed in 4% paraformaldehyde for 30, permeabilized in 2% acetone for 5′, incubated in 30% formamide/2XSSC pre-hybridization for 1 hr at RT and hybridized with a Cy5-(CAG)10 for 2 hr at 42°C and post-hybridized at 45°C for 30′. Sections were coverslipped and stained with DAPI to identify CUG RNA foci or incubated overnight at 4°C with Mbnl1 antibody (polyclonal A2764 gift of C.A. Thornton) at 1∶1000 and visualized by Alexa 488 secondary antibody (1∶2000) at RT for 30′. A2764 is a polyclonal antibody, directed against a C-terminal peptide of Mbnl1, specificity was demonstrated by lack of reactivity on immunoblots from Mbnl1ΔE3/ΔE3 mice and the antibody recognizes all known alternative splice isoforms of Mbnl1 [27]. Co-localization of Cy5-(CAG)10 RNA foci with MBNL1 was done by confocal microscopy (Olympus, Fluoview 1000) in 3 fluorescent channels with ≥5 layers (0.5 µm) compressed along the Z-axis and then merged. Gabt4 immunofluorescence staining was done using 10 µm frozen sections fixed using 4% paraformaldehyde for 30′, permeabilized with 2% acetone for 5′, briefly washed with PBS and then blocked in 5% goat serum, 0.3% Triton X-100 in PBS for 2 hr at 4°C. Sections were incubated with anti-Gabt4 polyclonal antibody (human GAT3, Sigma, St. Louis, MO) diluted 1∶1000 in the same blocking solution for 24 hr at 4°C, washed in PBS four times for 20′ and incubated with secondary antibody (Alexa 488 goat anti mouse IgG, Molecular Probes, Eugene, OR) using a 1∶2000 dilution for 2 hr at RT. Sections were washed 5 times for 5′ each in PBS and DAPI counterstained and coverslipped using Vectastain with DAPI (Vectashield, Ca)
Rotarod training was performed at 26 weeks of age using an accelerating rotarod (Ugo Basille, Comerio, Italy) as described [3]. All mice tested were F1 littermate progeny of single copy integrant SCA8 BAC-EXP5+/− and Mbnl1+/ΔE3 mice. Four trials were run per day for four days: averages of the four trials on day four are presented. rmANOVA followed by post-hoc analysis (Tukey's HSD) was performed to assess differences in rotarod performance between groups (SCA8 BAC-EXP5+/−; Mbnl1+/ΔE3; SCA8 BAC-EXP5+/−/Mbnl1+/ΔE3 and non-transgenic littermates).
For RNA analysis, 15 µg of total RNA isolated from non-transgenic FVB or Bl6/129 wild type, SCA8 BAC-EXP1, SCA8 BAC-CTRL and Mbnl1ΔE3/ΔE3 cerebellum was separated on a Northern Max-Gly glyoxal gel (Ambion), transferred to a nitrocellulose membrane, cross-linked by ultraviolet radiation and hybridized at 65°C in Rapid-Hyb buffer (Amersham) using a [32P]dUTP in vitro transcribed RNA probe to nucleotides 576–996 (NM_172890.3) of the mouse Gabt4 gene. Expression analysis was performed relative GAPDH by densitometry and analyzed by one-way ANOVA with Tukey's HSD post-hoc comparisons when necessary.
Two step qRT-PCR was performed on an ABI Prism 7500 Real Time PCR System (Applied Biosystems, Foster City, CA). Total RNA was isolated from mouse cerebellum, human frontal lobe autopsy tissue or transiently transfected SK-N-SH cells. cDNA was generated from 5 µg of total RNA using 1st Strand Synthesis Supermix primed with random hexamers (Invitrogen). Relative qRT-PCR was performed on 1 µl of cDNA with qRT-PCR SYBER Green Master Mix UDG with ROX (Invitrogen) using mouse specific primers (RSD1013 5′- CCT CTG AAG GCA TCA AGT TCT ATC TGT ACC-3′) (RSD1014 5′-GTT GTT GTA ACT CCC CAG AGC GGT TAG-3′) or human specific primers for autopsy and SK-N-SH cells (RSD1009 5′- AAC AAG GTG GAG TTC GTG CT-3′) (RSD1010 5′- ACT TGT GAA CTG CCC CAG AG-3′). Two stage PCR was performed for 40 cycles (95°C – 15″, 60°C– 1′) in an optical 96 well plate with each sample cDNA/primer pair done in triplicate. Relative quantification compared to strain specific control was estimated using the threshold cycle (Ct) of GABT4/Gabt4 normalized to the Ct of the housekeeping gene Hprt or GAPDH for mouse and human, respectively. Dissociation curve analysis and ethidium bromide gel analysis was used to assess PCR product purity at the end of each qRT-PCR run. Statistical analysis was done using rmANOVA on the mean normalized Ct value of the 3 trials per sample and compared by Tukey's HSD post-hoc analysis for differences between groups when necessary.
Animals were sacrificed and half of the cerebellum was rinsed with PBS and lysed in 450 µl of RIPA buffer (150 mM NaC1, 1% sodium deoxycholate, 1% Triton X-100, 50 mM Tris-HC1 pH 7.5, 100 ug/ml PMSF) for 45′ on ice. Cell lysates were centrifuged at 16,000×g for 15′ at 4 °C and the supernatant was collected. 20 µg of protein were separated on a 10% NuPAGE Bis-Tris gel (Invitrogen), transferred to nitrocellulose membrane (Amersham), blocked in 5% dry milk in PBS containing 0.05% Tween 20 and probed with an anti-GAT4 antibody (human GAT3; Sigma; 1∶1000) or anti-GAT1 antibody (ABcam; 1∶1,000) in blocking solution and then incubated with anti-rabbit or anti-mouse HRP conjugated secondary antibody (Amersham). Mean fold increase in protein levels were determined by densitometry normalized to GAPDH and compared to non-transgenic littermates for each blot. Statistical significance was determined by one-way ANOVA with between group comparisons evaluated by Tukey's HSD when necessary.
ExonA of containing the SCA8 expansion was amplified by PCR from the BAC transgene construct, BAC-exp, using the 5′ primer (5′CGAACCAAGCTTATCCCAATTCCTTGGCTAGACCC-3′) containing an added HindIII restriction site and the 3′ primer(5′ACCTGCTCTAGATAAATTCTTAAGTAAGAGATAAGC-3′) containing an added XbaI restriction site. This HindIII/XbaI fragment was cloned into the pcDNA3.1/myc-His A vector (Invitrogen, CA). The SCA8 ExonA cDNA was placed under the control of the CMV promoter of plasmid pcDNA3.1/myc-His. To construct ExonA-Rev, the HindIII/XbaI fragment of SCA8 ExonA was subcloned in the reverse orientation into pcDNA3.1/myc-His vector. The pCTGexp and pCAGexp clones (108 repeats) were generated by PCR amplification of SCA8 Exon A with added EcoRI restriction sites. This EcoRI/EcoRI fragment was cloned into the pcDNA3.1/myc-His A vector. The integrity of all constructs was confirmed by sequencing. Both pEGFP-N1-CUGBP1 and pEGFP-C1-MBNL1/41 have been described [27],[44],[45].
SK-N-SH cells were cultured in DMEM medium with 10% fetal bovine serum at 37°C with 5% CO2. Transient transfections were performed using 1 µg of SCA8 repeat expressing plasmids, pEGFP-N1-CUGBP1, or pEGFP-C1-MBNL1/41 minigenes and Lipofectamine 2000 Reagent (Invitrogen, Carlsbad, CA). Cells were collected 48 hrs post-transfection for expression analysis.
Analysis of NMDAR1 exon 5 and MBNL1 exon 7 alternative splicing were conducted using total RNA collected from SCA8, DM1 and control human brain autopsy tissue using Trizol (Invitrogen, CA) reagent according to the manufactures procedures. Human NMDAR1 exon 5 splicing was determined by amplification of exon 4-5-6 using PCR primers hsGRIN1 ex4 For 5′- GCGTGTGGTTTGAGATGATG -3′; hsGRIN1 ex6 Rev5′-GGTCAAACTGCAGCACCTTC -3′. Similarly, exon 7 alternative splicing of human MBNL1 was determined by amplification of exon 6-7-8 using PCR primers hsMBNL1 ex6 For 5′-GCTGCCCAATACCAGGTCAAC -3′; hsMBNL1 ex8 Rev 5′-TGGTGGGAGAAATGCTGTATGC -3′. Determination of GABT4 exon 7 alternative splicing was done by reverse transcribing 2 µg total RNA from frontal lobe or SKN-S-H cells as described above. 10% of this reaction was subjected to PCR using primers (RSD1004 5′-GTTGTATACGTGACTGCGACATT-3′; RSD1011 5′-GTTCAGGCAACAGAGCATGA-3′) to amplify nucleotides 791 – 1057 of human GAT3 (NM_014229.1) for 25 cycles at 94°C-45″, 54°C-30″, 72°C-1′ followed by 72°C for 6′. PCR products were run out on a 1.5% agarose gel with ethidium bromide and bands containing exons 6-7-8 (267 bp) or 6–8 (163 bp) were cut and verified by sequencing.
The crosslinking and immunoprecipitation protocol (CLIP) was performed as described [28],[46] with minor modifications. Hindbrains were dissected from mouse postnatal day 8 (P8) C57BL/6J pups followed by dissociation in 1X Hank's balanced salt solution containing 10 mM HEPES, pH 7.3 and UV-irradiated to crosslink RNA-protein complexes. Cells were lysed and RNA was partially digested with RNase T1 to produce 30–200 nt fragments. Lysates were cleared by ultracentrifugation and Cugbp1-bound fragments were immunoprecipitated using mAb 3B1 and Protein G Dynabeads (Invitrogen, Carlsbad, CA). Following 3′-end addition of RNA linkers, 5′ ends were labeled with g32P-ATP, protein-RNA complexes were eluted, separated by electrophoresis and protein-RNA complexes transferred to nitrocellulose. Bands corresponding to 60–70 kDa (10–20 kDa larger than the 50 kDa Cugbp1 protein) were excised from the nitrocellulose and the RNAs released by proteinase K digestion. RNA fragments were size fractionated by denaturing PAGE followed by 5′-end RNA linker ligation, RT-PCR and DNA sequencing.
All animal experimentation was approved by and conducted in conformity with the Institutional Animal Care and Use Committee of the University of Minnesota. Experimental details on the animal preparation, optical imaging and stimulation techniques are only briefly described as these have been provided in previous publications [32],[33]. Mice (3–8 months old) were anesthetized with urethane (2.0 mg/g body weight), mechanically ventilated, and body temperature feedback-regulated. The electrocardiogram was monitored to assess the depth of anesthesia. Crus I and II of the cerebellar cortex were exposed and the dura removed. An acrylic chamber was constructed around the exposed folia and superfused with normal Ringer's solution.
The animal was placed in the stereotaxic frame on an X-Y stage mounted on a modified Nikon epifluorescence microscope fitted with a 4×objective and a 100 W mercury-xenon lamp. Images of Crus I and II were acquired with a Quantix cooled charge coupled device camera with 12 bit digitization (Roper Scientific). The images were binned (2×2) to 256×256 pixels with a resolution of ∼10 µm×10 µm per pixel. Flavoprotein autofluorescence was monitored using a band pass excitation filter (455±35 nm), an extended reflectance dichroic mirror (500 nm), and a>515 nm long pass emission filter [33].
To evoke peripheral responses, the ipsilateral 3C vibrissal pad was stimulated with a bipolar electrode (tips ∼1 mm apart) using 20 V, 300 µs pulses at 10 Hz for 10 s [32]. Parallel fiber stimulation was performed throughout the experiment to test the general physiological condition of the cerebellar cortex. To activate parallel fibers and their postsynaptic targets (Purkinje cells and interneurons), an epoxylite-coated tungsten microelectrode (∼5 M) was placed just into the molecular layer [33].
The basic imaging paradigm consists of collecting a series of 10 control frames (background) followed by a series of 500 experimental frames with an exposure time of 200 ms for each frame (Metamorph Imaging System, Universal Imaging Corp.). Whisker pad stimulation was initiated at the onset of the experimental frames. This was repeated 4 times and the 4 series were averaged. Each pixel in this series of average images was converted into the change in fluorescence above background (ΔF/F) [32]. The maximal response to whisker stimulation occurred in frames 31–75 and these frames were averaged to generate the response image. A pixel was defined as responding to the peripheral stimulation by the following threshold procedure. First, the response image was low-pass filtered (3×3) and the mean and standard deviation (SD) of the pixels in a control region (usually a corner of the image) were determined. The pixels above the mean + 1 SD of the control region were considered to respond to the stimulus. The response area was defined as total area of all the pixels responding and the response intensity as the sum of the DF/F of all responding pixels. Differences in the area and intensity of the responses between the SCA8 and FVB mice were evaluated using a Student's t-test (α = 0.05). For display the pixels above or below this mean±1 SD were pseudo-colored and superimposed on an image of the folia studied.
|
10.1371/journal.ppat.1005775 | Sterol Biosynthesis and Azole Tolerance Is Governed by the Opposing Actions of SrbA and the CCAAT Binding Complex | Azole drugs selectively target fungal sterol biosynthesis and are critical to our antifungal therapeutic arsenal. However, resistance to this class of drugs, particularly in the major human mould pathogen Aspergillus fumigatus, is emerging and reaching levels that have prompted some to suggest that there is a realistic probability that they will be lost for clinical use. The dominating class of pan-azole resistant isolates is characterized by the presence of a tandem repeat of at least 34 bases (TR34) within the promoter of cyp51A, the gene encoding the azole drug target sterol C14-demethylase. Here we demonstrate that the repeat sequence in TR34 is bound by both the sterol regulatory element binding protein (SREBP) SrbA, and the CCAAT binding complex (CBC). We show that the CBC acts complementary to SrbA as a negative regulator of ergosterol biosynthesis and show that lack of CBC activity results in increased sterol levels via transcriptional derepression of multiple ergosterol biosynthetic genes including those coding for HMG-CoA-synthase, HMG-CoA-reductase and sterol C14-demethylase. In agreement with these findings, inactivation of the CBC increased tolerance to different classes of drugs targeting ergosterol biosynthesis including the azoles, allylamines (terbinafine) and statins (simvastatin). We reveal that a clinically relevant mutation in HapE (P88L) significantly impairs the binding affinity of the CBC to its target site. We identify that the mechanism underpinning TR34 driven overexpression of cyp51A results from duplication of SrbA but not CBC binding sites and show that deletion of the 34 mer results in lack of cyp51A expression and increased azole susceptibility similar to a cyp51A null mutant. Finally we show that strains lacking a functional CBC are severely attenuated for pathogenicity in a pulmonary and systemic model of aspergillosis.
| Aspergillus fumigatus is the most important airborne mould pathogen and allergen worldwide. Estimates suggest that >3 million people have invasive or chronic infections that lead to >600,000 deaths every year. Very few drugs are available to treat the various forms of aspergillosis and we rely predominantly on the azole class of agents which inhibit sterol biosynthesis. Resistance to the azoles is growing alarmingly, primarily driven by strains with two principal genetic signatures (TR34/L98H and TR46/Y121F/T289A). In this study we identify that the transcriptional mechanism governing resistance in this group of isolates is linked to the opposing actions of 2 transcriptional regulators, SrbA and the CBC, and uncover a role for the CBC in sterol regulation and virulence in A. fumigatus. We propose targeting SrbA would provide an effective avenue for therapeutic intervention for resistant strains.
| Sterols are components of most eukaryotic cell membranes playing key roles in sustaining membrane integrity and fluidity. The de novo synthesis of sterols by fungi is essential for their viability and numerous antifungal drugs have been developed that exploit the differences between enzymes in the sterol biosynthetic pathway of fungal pathogens and their hosts. The most notable sterol biosynthetic inhibitors are the azoles which are extensively used in crop protection and have been the cornerstone of systemic antifungal therapy in man for the last 30 years [1].
Triazoles such as voriconazole (VORI), itraconazole (ITRA) or posaconazole (POSA) represent the main antifungal drug class employed to treat disease caused by Aspergillus spp. Although in general Aspergillus fumigatus is susceptible to these drugs, resistance is emerging and reaching levels that have prompted some health centres to move away from the use of azoles as a sole first line therapeutic, opting instead for high cost combination therapies and/or less effective agents [2,3,4,5]. Azole drugs act by inhibiting the function of the sterol C14-demethylase Cyp51, leading to ergosterol depletion and simultaneous accumulation of toxic sterol compounds [6]. A principal cause of azole resistance in clinical strains of A. fumigatus is modification of the cyp51A gene, one of two genes that encode isoforms of sterol C14-demethylase in this pathogen. A particular family of pan-azole resistant isolates dominates. Typified by the TR34/L98H modification, but also including TR46/Y121F/T289A and TR53, they harbor a tandem repeat in the promoter of cyp51A along with a non-synonymous mutation resulting in one or more amino acid changes in the Cyp51A protein. In the case of TR34/L98H this is manifest as a duplication of a 34 mer within the 5’ non-translated (5’ NTR) region of cyp51A, combined with a lysine to histidine substitution at position 98 in the protein. Understanding the mechanisms by which the TR34/L98H family and non-cyp51A type mutations lead to resistance is critical to formulating strategies to both detect and treat resistant infections.
What drives resistance in the TR34/L98H family is only partially understood. Introduction of the L98H modification into a hitherto wild-type (wt) isolate of A. fumigatus results in a modest but significant increase in triazole tolerance. Introduction of the TR34 variant into a conventional cyp51A promoter results in an approximate doubling of cyp51A gene expression and an associated increase in tolerance to azoles [7]. Only when these modifications are combined can isolates reach tolerance levels that exceed clinically relevant breakpoints as defined by EUCAST (European Committee on Antimicrobial Susceptibility Testing) [8]. Although the mechanistic basis of L98H driven increase in azole tolerance has been linked to modification of the tertiary structure of Cyp51A [7], the cause of increased expression resulting from the TR34 promoter variant has not been elucidated.
In mammalian systems, sterol production is transcriptionally regulated by sterol regulatory element binding proteins (SREBPs). SREBPs belong to the basic helix-loop-helix (bHLH) transcription factors. In their inactive form they are membrane-bound, once activated the regulator is released and accumulates in the nucleus where it binds to sterol regulatory elements (SREs) in the promoters of target genes, including sterol C14-demethylase and activates expression [9,10]. At sterol excess SREBPs remain inactive causing decreased transcript levels of their targets [10]. The activating role of the SREBPs is facilitated by the action of the CCAAT-binding complex (CBC) NF-Y which is made up of three subunits NF-YA, NF-YB and NF-YC. The SREBPs and NF-Y synergistically activate expression of essentially all genes involved in sterol metabolism [11].
Orthologues to mammalian SREBPs have been found in several fungi including A. fumigatus. The A. fumigatus SREBP homologue, termed SrbA, is a non-redundant transcription factor which, like its mammalian counterparts, plays a key role in the regulation of sterol biosynthetic genes, including cyp51A. In keeping with this role, A. fumigatus strains lacking SrbA have reduced tolerance to the azoles [12,13,14].
Similarly, A. fumigatus has an orthologue of the CBC. As with the mammalian regulator, the A. fumigatus CBC is a multimeric transcription factor complex comprising three subunits (HapB/HapC/HapE) and is highly conserved from yeast (Hap2p/Hap3p/Hap5p) to man (NF-YA/NF-YB/NF-YC) [11,15,16,17,18,19]. A mutation in the HapE subunit (P88L) was recently identified as the causative modification that led to azole resistance in an A. fumigatus strain isolated from a patient in the Netherlands [20].
As both SrbA and the CBC have been directly implicated in modified azole tolerance in A. fumigatus, and their orthologues have positively acting sterol regulatory functions, we sought to understand the role of the CBC in sterol biosynthesis and both regulators in clinical azole resistance. Particularly we were interested in uncovering any role of these regulators in TR34 mediated resistance.
The A. fumigatus CBC functions as a heterotrimer comprising HapB, HapC and HapE and is known to interact with the highly abundant consensus motifs CCAAT and CGAAT [21,22]. A number of potential binding sites are evident within the cyp51A promoter, therefore we sought to identify if the CBC interacted directly with the promoter of cyp51A using chromatin-immunoprecipitation followed by next-generation sequencing (ChIP-seq). To this end, we replaced the native hapC gene in isolate A1160P+ with a chimeric gfp-construct. The respective strain expresses C-terminally GFP-tagged HapC protein (hapCGFP, Fig 1A). We sequenced anti-GFP bound DNA fractions from cell extracts and identified a single region within the promoter (within 1kb upstream of the translational start site) of cyp51A to which the GFP-CBC was bound (S1 Table). The sequencing peak identified by ChIP-seq was at position -297 relative to the translation start placing it at the 3’ end of the 34 mer. This result was validated by ChIP-qPCR by comparing enrichment of the region identified in the ChIP-seq analysis with region not bound by the CBC in the promoter of actA (Fig 1B). By examining the promoter region around the ChIP-seq peak, we found the CBC binding motif CGAAT at position -293 to -289 (Fig 2A). To further analyse the interaction of the CBC to this motif, binding kinetics and affinity were monitored by surface plasmon resonance spectroscopy (SPR) protein-DNA interaction analysis. Kinetic CBC binding responses to a 37-bp cyp51A promoter duplex revealed a dissociation constant (KD) of 74.4 nM, demonstrating that the CBC effectively recognised the CGAAT-293 motif within the original 34 mer in vitro (Fig 2B, upper panel 2). Importantly, although the CGAAT sequence is duplicated in TR34 (CGAAT-327), the binding affinity of the CBC at the duplicated sequence is about 8-fold decreased and consequently not effectively bound (Fig 2C panel 2).
Cyp51A is known to use iron as a catalytic co-factor and the CBC is a known interaction partner of HapX, an iron responsive basic region leucine zipper (bZIP) transcription factor. We therefore screened the nucleotide sequence downstream the CGAAT-293 motif for potential HapX binding sites [19,21,24]. We identified a pseudo-palindromic TTACTAA sequence at position -275 to -269 in the cyp51A promoter that corresponds exactly to the yeast AP-1 (YAP1) bZIP consensus binding site [25]. Employing SPR co-injection analysis we confirmed that this region of the cyp51A promoter is synergistically bound by the CBC/HapX complex (Fig 2B panel 4 and 5). Consistent with the fact that the HapX-binding site is not duplicated in the 34 base tandem repeat, combinatorial CBC/HapX recognition was not detectable on a DNA duplex that contained the duplicated CGAAT motif (Fig 2C, panel 4 and 5).
Together these data demonstrate that only the wt 34 mer of the cyp51A promoter is a direct target of the CBC and its interaction partner HapX.
To determine if the CBC is functioning as a repressor, or an activator of cyp51A gene expression and to assess the role of the HapEP88L mutation we generated CBC gene deletion mutants of each subunit of the CBC namely, HapB (ΔhapB), HapC (ΔhapC) and HapE (ΔhapE), a strain expressing HapEP88L (hapEP88L), and a deletion mutant of HapX (ΔhapX) (S2 Table). Disruption of any of the CBC subunits, HapX or mutation of HapE resulted in increased tolerance to ITRA, VORI and POSA (Table 1). Additionally, similar to HapEP88L [20] cyp51A expression was increased by about 2-fold in the ΔhapC strain (Fig 3A). This result indicates that the CBC and HapX act as repressors of cyp51A and that the HapEP88L mutation abolishes this repressing function. The loss of repressing activity of HapB/HapC/HapEP88L seems to be linked to a reduction in binding affinity of the mutated CBC to the CGAAT-293 motif as indicated by SPR analysis (Fig 2B; compare panel 1 to panel 2).
Combined, these results suggest that azole resistance caused by HapEP88L is linked to impaired CBC DNA-binding function.
The discrepancy between the increase in expression of cyp51A (c. 2-fold) and azole resistance (>32-fold) suggested that cyp51A-independent factors may be influencing azole resistance in CBC deficient strains. This is supported by the fact that a ΔhapCΔcyp51A is more resistant to the azoles than a single cyp51A null mutant (Table 1). We therefore hypothesised a possible function of the CBC in the repression of other genes in the ergosterol biosynthetic pathway.
Interrogation of our ChIP-seq data allowed us to identify CBC binding sites in the promoter regions (< 1.0 kb 5’ NTR) of around half of the ergosterol biosynthetic genes (14 out of 27, S1 Table). We validated the results of the ChIP-seq analysis by ChIP-qPCR for four of the genes erg13B, erg7B, cyp51A and cyp51B (Fig 1B). To assess if binding of the CBC was affected by azole treatment, which would mimic cellular sterol depletion, we performed ChIP-qPCR using A. fumigatus cultures grown in the presence or absence of ITRA. We found enrichment of the binding regions in both sterol replete (-ITRA) and sterol deplete (+ITRA) conditions indicating that the CBC is constitutively bound to all four promoters under these conditions (Fig 1B).
To investigate if binding of the CBC was linked to regulation of gene expression, we monitored expression levels of several genes associated with the ergosterol biosynthetic pathway in our CBC mutant (ΔhapC). Specifically we monitored genes coding for enzymes of the first and committed steps in ergosterol biosynthesis, HMG-CoA synthase (paralogs erg13A and erg13B) and HMG-CoA reductase (paralogs hmg1 and hmg2), the genes encoding the azole drug target sterol C14-demethylase (paralogs cyp51A and cyp51B) as well as erg7B, that encodes lanosterol synthase. For four of these genes, erg13A, erg13B, hmg1 and erg7B, expression was significantly increased by at least 3-fold in ΔhapC when compared to the isogenic host strain (Fig 3A). For cyp51B we detected a modest increase in transcript levels (>1.5 fold). We were unable to detect expression of hmg2 in either strain due to the very low abundance of the transcript. It is noteworthy that our ChIP-seq results did not indicate binding of the CBC to the 1.5 kb region upstream of the erg13A translation start site, suggesting an indirect regulation of this gene by the CBC.
Based on the idea that increased expression of several ergosterol biosynthetic genes in CBC mutants might lead to increased activity of the enzymes, we determined MIC levels of ΔhapC for simvastatin and terbinafine (Fig 3B). These drugs, belonging to the families of statins and allylamines, are ergosterol biosynthesis inhibitors that inhibit the enzymes HMG-CoA reductase (Hmg) and squalene epoxidase (Erg1) respectively. Both of these enzymes perform functions upstream of the C14-demethylation carried out by the Cyp51 enzymes (Fig 3C). In keeping with our hypothesis, ΔhapC displayed a significant increase in tolerance to both of these drugs.
Taken together, these data illustrate that loss of the CBC leads to upregulation of genes involved in several steps of ergosterol biosynthesis, and increases the activity of several enzymatic steps in this pathway. This results in resistance to ergosterol biosynthesis inhibitors belonging to the azoles, statins and allylamine drug classes.
Resistance to simvastatin, terbinafine and azole drugs suggests that CBC mutation leads to increased activity of the whole ergosterol biosynthetic pathway. To analyse the effect of loss of CBC function on ergosterol biosynthesis, sterol levels were quantified using GC-MS. To confirm the validity of our results we analysed the ergosterol levels in a srbA null mutant (ΔsrbA). Loss of SrbA function is documented to result in decreased ergosterol levels and increased C4-methylated sterols (Table 2, columns 6 and 7) [14]. In keeping with this we demonstrated that ergosterol levels were decreased (Table 2, column 2: 3.13-fold) in ΔsrbA and an accumulation of eburicol (Table 2, column 6:, 2.37-fold) and 4,4-dimethylergosta-8,24(28)-dien-3β-ol (Table 2, column 7: 11.03-fold) was observed. In line with our expectations, ΔhapC showed a 2.4-fold increase in ergosterol levels (Table 2, column 2). In addition to ergosterol, which constitutes around 90% of the total sterol content in ΔhapC, we found four further sterols to be elevated in this mutant, including the Cyp51 substrate molecules lanosterol (Table 2, column 5: 1.67-fold) and eburicol (Table 2, column 6: 2.09-fold). The Δcyp51AΔhapC isolate displayed increased levels of ergosterol (Table 2, column 2: 1.59-fold) but also ergosta-5,7,24(28)-trien-3β-ol (Table 2, column 3: 1.65-fold) and episterol (Table 2, column 4: 2.51-fold). The hapEP88L strain showed ergosterol levels similar to that of the wt however, ergosta-5,7,24(28)-trien-3β-ol (Table 2, column 3: 3.57-fold) and episterol (Table 2, column 3: 3.57-fold) were increased in this mutant indicating that the regulatory defect in this mutant differs somewhat from the null mutant.
These findings suggest that transcriptional upregulation of ergosterol biosynthetic genes in ΔhapC leads to increased sterol levels, particularly ergosterol, which is likely to contribute to resistance to ergosterol biosynthesis inhibitors in CBC mutants.
The high abundance of CCAAT and CGAAT motifs in eukaryotic promoters suggests that the expression of a large proportion of protein coding genes might be affected by CBC mutation. Hence, it seemed likely that CBC defective mutants would have significant phenotypic defects. We assessed growth of CBC mutants (ΔhapB, ΔhapC, ΔhapE, hapEP88L) on Aspergillus minimal medium (AMM), RPMI as well as Sabouraud dextrose agar (SAB, rich-nutrient medium). Lack of either subunit resulted in significant growth defects on all media with individual deletion mutants showing identical phenotypes (Fig 4A). The hapEP88L isolate revealed a less dramatic phenotype compared to hap-deletion mutants. Growth of this isolate, carrying a single amino acid mutation, was impaired particularly on RPMI.
To assess the requirement of the CBC in infection, we compared the virulence of ΔhapC and hapCREC with that of the isogenic control strain in systemic and pulmonary murine models of aspergillosis. The pathogenicity of the isogenic control and hapCREC strains were indistinguishable, causing 100% mortality (n = 10) between 9 and 10 days after systemic infection and between 12 and 14 days after pulmonary infection. Systemic infection by ΔhapC caused no mortality while in the pulmonary model only one animal succumbed to the infection (Fig 4B). Tissue burden studies showed a reduced fungal load in animals infected i.v. (n = 10) and i.n. (n = 8) with ΔhapC in comparison to those challenged with wt and hapCREC (S1 Fig). Pulmonary infection resulted in high fungal burden in all animals challenged with wt and hapCREC while ΔhapC was cleared in all mice except two. Kidneys from animals inoculated with wt or hapCREC showed low viable fungal elements, however no colonies were detected in ΔhapC-infected mice. With exception of the brain, the systemic infection resulted in high fungal loads in all tissues after wt and hapCREC inoculation. Animals challenged with ΔhapC showed significantly lower CFU/g in all organs except for the liver and brain when compared to those challenged with wt or hapCREC.
Taken together, these results demonstrate the crucial role of the CBC for A. fumigatus growth and show that transcriptional circuits mediated by this regulator are critical during infection of an immunocompromised host.
Our analysis of the CBC suggests that it is associated with the 34 mer but not directly responsible for the transcriptional enhancement of cyp51A observed in strains with the TR34 mutation. This led us to assess if the other known transcriptional regulator of sterol biosynthetic genes, SrbA, was involved.
In A. fumigatus the sterol regulatory element binding protein SrbA has been described as direct positive regulator of sterol biosynthesis in general and cyp51A in particular [13,27]. A previous study identified a motif similar to the reported sterol regulatory element (Sre1) binding motif of S. pombe upstream of the TR34 site [13]. In an attempt to corroborate the location of this site, we performed a retrospective analysis of a recent genome-wide ChIP-Seq evaluation of SrbA binding [27] and identified a sequencing peak summit at position -302 from the cyp51A coding sequence start site, which in contrast to the previous predicted location of the binding site, is within the duplicated 34 mer. By comparing the 34 bases of TR34 with the SrbA consensus binding motif [27] we identified two putative SREs (SRE1 and SRE2, Fig 5A). These SREs are also evident in the overlapping TR46 and TR53 found in other resistant clinical isolates (Fig 5A) [28,29,30].
To test if SrbA directly interacts with the predicted SREs, we assessed binding of recombinant SrbA to DNA-duplexes containing either SRE1, SRE2 or both sites by SPR (Fig 5B). SrbA showed interaction with both of the predicted binding sites within the 34 mer, however affinity for SRE1 (KD = 22.8 nM) was 12-fold higher than that for SRE2 (KD = 272.6 nM) (Fig 5B; panels 1 and 2 respectively). Employing a DNA duplex containing both SREs we measured a 2-fold increase in the saturating SrbA response (Rmax value of 197.6) combined with an apparent KD of 22.0 nM, which does not represent the simple average of the SrbA affinities measured for the single SREs. In conclusion, we propose cooperative binding of two SrbA homodimers to the 34 mer (Fig 5B; panel 3). Notably, the affinities of SrbA and the CBC to their partially overlapping binding motifs are at a similar level (74.4 vs. 22.0 nM), indicating that both regulators bind competitively to the 34 mer within the wt cyp51A promoter. By contrast, the 8-fold lower affinity for CBC at the CGAAT motif duplicated in TR34 will favor cooperative SrbA binding to both of the SREs in the 5´-TR34 region.
This finding suggests that the increased expression of cyp51A in strains harboring TR34, TR46 or TR53 could be the result of increased SrbA activity as a result of the duplication of its consensus binding sites and a lack of CBC and accordingly CBC/HapX repression.
Based on the hypothesis that the 34 mer is required for SrbA mediated cyp51A activation and consequent azole resistance, we generated a cyp51A promoter variant lacking the 34 mer (cyp51AΔ34, S1 Fig). In order to confirm the previous results described by Snelders in 2011 [7] and to draw a direct comparison to the cyp51AΔ34 isolate in an isogenic background, we also generated a strain carrying the tandem repeat TR34 (cyp51ATR34). Deletion of the 34 mer (cyp51AΔ34) led to a >90% reduction in cyp51A expression similar to the srbA deletion mutant (ΔsrbA) (Fig 6), which is consistent with our hypothesis. A similar result has been observed in a previous study using a cyp51A promoter based luciferase reporter approach [31]. In addition, the cyp51AΔ34 isolate displayed a 4-fold reduction in MIC (from 0.50 mg/L to 0.13 mg/L, Fig 6), which mimics the azole susceptibility phenotype of the isogenic cyp51A deletion mutant (Δcyp51A). MIC levels for the azoles were even lower for ΔsrbA, which is likely a result of lack of activation of several ergosterol biosynthetic genes in this mutant [12,27].
Taken together, our data demonstrates that cyp51A activation relies on the presence of the 34 mer. This region is required for SrbA mediated activation of cyp51A gene expression, hence, for azole resistance in A. fumigatus.
In this study, we report a novel mechanism for the regulation of ergosterol metabolism in A. fumigatus and highlight the interplay between the transcriptional regulators SrbA and CBC. We also describe how perturbation of this regulatory mechanism through numerous routes can induce resistance to the primary class of agent used to treat fungal disease, the triazoles, which act by inhibiting ergosterol biosynthesis through depleting sterol C14-demethylase activity.
In A. fumigatus, the sterol regulatory element binding protein SrbA has been shown to be a positive acting transcriptional regulator of sterol biosynthesis and, aligned with this, has a crucial role in azole tolerance [14]. A recent study revealed that several ergosterol biosynthetic genes, including those encoding the two paralogs of Cyp51 (Cyp51A and Cyp51B) and Erg25 (Erg25A and Erg25B), are under direct control of SrbA [27], and strains lacking the regulator have significant reduction in the expression of the aforementioned genes. The role of SREBPs in activating expression of sterol biosynthesis is highly conserved.
In mammals, two genes encode three distinct isoforms SREBP-1a, SREBP-1c and SREBP-2, which are involved in positively regulating various aspects of lipid and sterol metabolism [10]. The activating role of the SREBPs in mammalian cells is facilitated by the action of the CCAAT-binding complex NF-Y.
The discovery that an amino acid alteration P88L within the HapE subunit of the CBC in A. fumigatus leads to resistance to all clinically used azole antifungals provided a potential link between CBC function and ergosterol biosynthesis and suggested the role of the CBC may also be conserved. Indeed in keeping with a role of the CBC in regulation of ergosterol biosynthesis we have demonstrated that the CBC binds the promoters of several genes in the pathway including cyp51A and cyp51B, erg13B and erg7B. Intriguingly, we identified that the CBC binds within the 34 mer in the promoter of cyp51A commonly duplicated in strains that are pan-azole resistant (TR34 family strains).
If the CBC were a positive regulator of ergosterol biosynthesis this may explain the mechanistic link between regulation of cyp51A and TR34 associated resistance. Our findings however are contrary to this hypothesis and suggest that the CBC of A. fumigatus is playing the opposite role to that of the mammalian orthologue and acting as a negative regulator of sterol biosynthesis. Lack of CBC activity (ΔhapC) leads to derepression of genes in the ergosterol biosynthetic pathway, including genes coding for enzymes of the initial and committed steps of sterol biosynthesis, erg13A, erg13B and hmg1, as well as the azole drug target cyp51A and a concomitant increase in sterol production and tolerance to drugs inhibiting the ergosterol biosynthetic pathway (azoles, allylamines and statins). Additionally, we have shown that the CBC binds efficiently only to the original CGAAT motif (Fig 2, CGAAT-293) but not to the duplicated version within TR34 (Fig 2, CGAAT-327). The reason for the differential binding affinities at these two sites were not investigated further however several lines of evidence point towards the importance of the bases outside this core motif as being critical. First, the structure of the Aspergillus nidulans CBC in complex with DNA demonstrated that hydrogen bonds between the phosphate groups of the DNA backbone outside the core motif, and the protein main or side chain atoms of HapC and HapE, significantly stabilise the protein:DNA complex and induce DNA bending. Second, alignment of the 23 bp cycA promoter sequence from which this structure was derived with the native and repeat cyp51A promoter sequences revealed different nucleotides at positions at which HapE amino acid residues mediate DNA sequence-independent interactions with phosphate oxygen atoms (S2 Fig). Finally our SPR data clearly indicate that an 8-fold lower CBC affinity to the duplicated TR34 CBC site is due to the presence of eight different nucleotides at the 3´end of the motif. Therefore, our working hypothesis is that altered DNA-bending might be the major cause for the decreased CBC affinity.
We have determined that the azole resistance exhibited by isolates with the HapEP88L modification is linked to an inability of the modified CBC to bind effectively to its recognition site in the cyp51A promoter, leading to increased expression. Interestingly, the growth phenotype exhibited by a strain with the HapEP88L mutation is less severe than that of the HapE null, suggesting only a partial loss of function. One could hypothesise two potential mechanisms for this partial phenotype, 1) the modified CBC retains binding capability for all targets but with reduced efficacy or 2) the modified CBC loses the ability to bind a subset of its targets. These scenarios are not mutually exclusive and more likely than not, a combination of both of these factors are responsible. Interestingly, structural analysis of the closely related A. nidulans CBC indicates that HapE stabilises the protein:DNA complex by binding through residues immediately adjacent to P88, namely L89 to K94 [23]. These residues are thought to interact in a DNA sequence-independent manner. Therefore it is likely that the modified CBC exhibits a general reduction in binding capability across all of its targets, with other factors such as the strength of binding of the CBC to the remaining sections of its binding site, or the presence of stabilising protein interaction partners dictating the strength of binding of the modified CBC.
Our genome-wide binding data revealed that the CBC interacts with the cyp51A promoter during sterol depletion as well as repletion. We found this to be true for almost all the ergosterol biosynthetic genes tested, which indicates that the complex might constitutively bind to these targets. This could suggest a simplistic binary regulatory mechanism at the promoters of sterol regulatory genes with the CBC carrying out a repressing role in the absence of an activator, e.g. SrbA. The regulation of sterol metabolism is however far more complex. We have demonstrated that the regulatory element HapX also has a role in the regulation of cyp51A and azole tolerance. As HapX is known to govern the regulation of genes involved in iron acquisition as well as iron consumption its association with cyp51A, a heme-iron containing enzyme is logical. We propose that HapX enhances or stabilises binding of the CBC at the cyp51A promoter in iron limiting conditions and by doing so provides a more nuanced way of regulating ergosterol biosynthesis in environments that are unfavorable for sterol production. It is possible therefore that further as yet undefined factors will have a role in the regulation of ergosterol biosynthesis.
Our data suggests a role for the CBC in TR34 family mediated azole resistance as its repressing function is not duplicated effectively in the repeat. This alone however, could not explain an increase in cyp51A expression observed in TR strains. Our hypothesis was that a positive regulator must be binding here. We therefore re-evaluated the sequence of the tandem repeat and identified two sequences that matched the consensus of the positive regulator SrbA and which were close to a putative binding site highlighted in a recent ChIP-seq study [27]. We confirmed SrbA interaction to both of these sequences in the 34 mer and cooperative binding at a region on the TR34. This leads to the conclusion that TR34 family driven upregulation of cyp51A is a result of effective duplication of SrbA binding sites in combination with ineffective duplication of the CBC site. It is interesting to note that tandem repeats of transcriptional enhancer elements are known mechanisms of azole resistance in several phytopathogenic moulds [32]. For example Hamamoto et al. identified a 126-bp tandem repeat in the cyp51 5’ NTR of the plant pathogen Penicillium digitatum, which caused resistance to Cyp51 inhibitors [33]. Whether resistance in this case is linked to duplication of SREs and/or ineffective duplication of repressor elements is unclear, however we have identified sequences within this repeat that resemble the SrbA binding consensus defined in A. fumigatus [27].
In contrast to A. fumigatus and most other eukaryotes, yeasts such as Saccharomyces cerevisiae and Candida albicans SREBPs have been replaced by Upc2p, a zinc-finger transcription factor which positively regulates sterol synthesis in these species [34]. In C. albicans antifungal azole resistance caused by overexpression of target genes has been extensively studied identifying Upc2p gain-of-function mutations. These types of mutations result in upregulation of the Cyp51 encoding gene and appear to be frequent in azole resistant C. albicans clinical isolates [35]. In A. fumigatus similar mutations in SrbA have not been identified so far. However, our study demonstrates an alternative route to enhance the activity of the positive regulator, SrbA, through increasing its DNA-binding ratio by duplicating its binding site.
The cause of the growth deficiency in the CBC null mutants is difficult to pinpoint exactly. The CBC is a global regulator, its consensus site is present in >30% eukaryotic promoters [22]. In A. nidulans and A. fumigatus several iron metabolic genes are known to be under control of the CBC in complex with HapX [19,24,36]. Additionally the CBC has also been shown to play a prominent role in the regulation of secondary metabolism, development and oxidative stress response [37,38]. These roles and probably the involvement of the CBC in many other biological processes are likely to be the cause for reduced fitness in the mutants. Although the same could be said for the role of the CBC in virulence, its interconnection with the transcription factor HapX is probably more important in this context. HapX mediated transcriptional adaptation to low iron environments has been demonstrated to be crucial for virulence in A. fumigatus and other pathogenic fungi such as C. albicans, Cryptococcus neoformans and Fusarium oxysporum [36,39,40,41]. Hence, we consider one major reason of the reduced pathogenicity to be a direct result of inefficient adaptation to low iron in the CBC mutant. It is interesting to note that despite a clear attenuation in virulence, an azole resistant CBC loss of function mutant, namely HapEP88L, was the likely cause of death in a chronic granulomatous disease patient [42].
This work provides novel insight into the molecular basis of sterol regulation and TR34 mediated azole drug resistance in A. fumigatus and suggests that if the binding of SrbA could be countered one could reverse drug resistance. The direct targeting of transcriptional binding is fraught with difficulties. Unlike many enzymes transcription factors have large surface areas for protein-protein and protein-DNA interactions which are difficult to disrupt with a small molecule. However modulation of these interactions may be possible using oligonucleotide or peptide therapeutics if these molecules can be effectively delivered into fungal cells [43]. Some, albeit limited, data suggests that this may be possible, at least in the case of oligonucleotides [44] and as significant differences exist between human SREBP and SrbA (less than 40% sequence identity at the DNA binding domain) selective targeting could be achieved. Alternatively, it should be possible to disrupt upstream activators of SrbA, for example the protease RbdA which was recently linked to SrbA cleavage and activation [45]. It is encouraging that a number of small molecules that inhibit Upc2-dependent transcriptional signaling in vivo have been identified in yeast, providing evidence that a similar, unbiased approach may be successful in A. fumigatus [46].
All animal care procedures were supervised and approved by the Universitat Rovira i Virgili Animal Welfare and Ethics Committee.
Strains and oligonucleotides used in this study are listed in S2 and S3 Tables. Conidia were grown on SAB agar plates. For phenotypic analysis, A. fumigatus liquid cultures were grown at 37°C in AMM according to Pontecorvo et al. [47]. AMM included 1% glucose as carbon source (w/v) and 20 mM ammonium tartrate as nitrogen source. In vitro susceptibility testing of mutant strains was carried out using the EUCAST broth microdilution reference method [26]. Strains are defined in this manuscript as resistant where their MIC exceeds the relevant breakpoint as defined by EUCAST [8].
Coding sequence of hapB, hapC, hapE, srbA and cyp51A was disrupted in A1160P+. The deletion fragments for homologous recombination were generated using the FusionPCR approach previously described [48]. Around 1 kb of 5’ NTR and 1kb 3’ NTR were PCR amplified and subsequently linked to an antibiotic resistance cassette via PCR Fusion using primers listed in S3 Table. The P88L causing mutation of HapE was carried out using a similar FusionPCR strategy. hapE coding sequence and 1kb 3’ NTR were amplified from wt genomic DNA. The respective amplicons were linked to a hygromycin resistance conferring cassette via FusionPCR using a mutation generating forward primer.
To complement Δcyp51A and generate strains harboring modified cyp51A promoters, the basic plasmid pcyp51AREC, comprising a pyrithiamine resistance cassette, was generated. The plasmid for the reconstitution of ΔhapC, phapCREC, was generated following the same principle described for pcyp51AREC. The backbone of the hygromycin resistance cassette carrying plasmid pAN7-1 was amplified using primers pAN7-1-hapC-f/pAN7-1-hapC-r. hapC coding sequence plus 1.3 5’ NTR and 0.7 3’ NTR were amplified with primers hapCrec-f/hapCrec-r. Both fragments were gel-purified and linked via Gibson assembly as described above.
To tag hapC coding sequence at its 3’ end with gfp, phapCREC was PCR amplified using primers phapC-GFP-FW/phapC-GFP-RV. The gfp encoding gene was amplified employing primers GFPhapC-FW/GFPhapC-RV using the recently described plasmid pgfpcccA as template [49]. Gel-purified fragments were combined using Gibson assembly yielding phapCGFP.
Mutation of specific DNA sections was carried out according to the PCR based Q5 site-directed mutagenesis protocol (NEB). pcyp51AREC was used as template DNA. TR34 was introduced into the promoter region employing primers TR34-FW/TR34-RV. Δ34-FW/Δ34-RV were used to delete the 34 mer. After amplification of each construct yielding linearized PCR amplicons, PCR fragments were gel-purified and circularised using the Quick Ligation Kit (NEB). Plasmids carrying cyp51A promoter mutations were designated pcyp51ATR34 and pcyp51AΔ34.
Generally 2 μg of the deletion constructs or 2 μg of the respective plasmids were used for transformation. Prior to transformation pcyp51AREC and promoter mutated plasmids were linearised using BsrGI, phapCREC and phapCGFP were digested with PmlI (S3 Fig). For transformation 1M sucrose was supplemented to AMM or SAB. Depending on the resistance cassette transformed, 0.1 mg/L pyrithiamine (ptrA), 0.2 g/L hygromycin B (hph) or 0.15 g/L zeocin (ble) were used for selection. AMM was used for pyrithiamine based transformation, SAB was used for hygromycin B and zeocin selection at pH 6 and pH 8, respectively.
Liquid cultures for sterol measurements were grown in AMM for 24 h at 37°C, 200 rpm. Mycelia were harvested through filtration, shock frozen using liquid nitrogen and freeze-dried. The lyophilisate was ground and dissolved in 2M NaOH to obtain a suspension of 3.0 mg/mL. The work-up procedure can be taken from Müller et al. [50]. The residue was dissolved in 800 μL MtBE, 100 μL cholesterol solution (calibration standard, 10 μg/L), and 100 μL of silylation reagent MSTFA/TSIM (9:1) was added. The sample was gently shaken and stored for complete silylation at room temperature for at least 30 min, before being subjected to GC-IT-MS analysis [51,52].
Each sample was prepared in triplicate and measured in duplicate.
Sterols were analysed as trimethylsilyl (TMS) ethers. The sterol TMS ethers were identified by mass spectra and relative retention times (RRT) according to Alcazar-Fuoli et al. and Müller et al. [50,53]. The sterol TMS ether peaks were referred to the TMS ether peak area of the base peak of cholesterol TMS ether. The base peaks of each sterol TMS ether were taken as a quantifier ions for calculating the peak areas for cholestane m/z 217, cholesterol m/z 368, ergosta-5,7,9(11),22-tetraen-3β-ol (1) m/z 251, ergosterol (2) m/z 363, 5,6-dihydroergosterol (3) m/z 343, ergosta-5,7,24(28)-trien-3β-ol (4) m/z 363, episterol (5) m/z 343, lanosterol (6) m/z 393, eburicol (7) m/z 407, and 4,4-dimethylergosta-8,24(28)-dien-3β-ol (8) m/z 408. The content for each sterol [μg/mg] was calculated according to Müller and Bracher [52].
RNA was isolated using TRI Reagent (Sigma). 10 μg extracted total RNA were digested using RQ1 RNase-Free DNase (Promega) and further purified using the RNeasy Mini Kit (Qiagen). qPCR was performed in a 7500 Fast Real-Time PCR System (Applied Biosystems) using the iScript One-Step RT-PCR kit with SYBR Green (Cat# 170–8893)(Bio-Rad). Primers used for RT-qPCR analysis are listed in S3 Table. Amplification reactions were performed in a final volume of 25 μL using (1.0μL) 0.4μM forward primer, (1.0μL) 0.4 μM reverse primer, and 5 ng (5 μL) of total RNA.
The A. fumigatus CBC consisting of HapB(230–299), HapC(40–137) and HapE(47–164) as well as HapX(24–158) were produced and purified as described by Gsaller et al. [19]. A HapE(47–164) subunit carrying the clinically relevant P88L mutation was generated employing a synthetic gene. The heterotrimeric HapB/HapC/HapEP88L complex was purified to homogeneity by subsequent cobalt chelate affinity and size exclusion chromatography (SEC) just as the wt CBC (S4A Fig). The heterotrimeric status of the HapB/HapC/HapEP88L complex, or in other words, its stability was proven by analytical SEC coupled light scattering measurements (S4B Fig). Therefore, an Äkta Explorer system (GE Healthcare) was connected to a miniDawn TREOS static light scattering (SLS) detector equipped with an internal quasi-elastic light scattering (QELS) system in series with an OPTILab T-rEX differential refractometer (Wyatt). Absolute molar mass and hydrodynamic radius (Rh) were determined using the ASTRA 6 software (Wyatt). The CBC and CBCP88L were chromatographed in 20 mM Tris/HCl, 400 mM NaCl, 1 mM DTT pH 7.5 using a Superdex 200 Increase 10/300 GL column (GE Healthcare).
The basic region/helix-loop-helix/leucine zipper (bHLHZ) region of A. fumigatus SrbA (amino acids 161–267) was purified following the procedure described in Linde et al. [54]. Real-time SPR protein-DNA interaction measurements were performed by using protocols published previously [19,54].
Microscopy images were taken on the Leica TCS SP8 inverted confocal microscope (Leica Microsystems CMS, Mannheim, Germany) using a 63×/1.2 NA objective. Images were captured with LAS AF V3.3. Images were processed using ImageJ and Adobe Photoshop CS6.
For ChIP-qPCR analysis 1x106 spores/ml of hapCGFP were grown in 50 mL of AMM for 18 h at 37°C and 200 rpm.
ChIP was performed as previously described by Chung et al. [27] in this work using an Anti-GFP antibody (A-11122, Life technologies) or an anti-IgG control (ab46540, abcam) on Dynabeads Protein A magnetic beads (Thermo Fischer Scientific). Immunoprecipitated DNA was reverse cross-linked, treated with RNase A (Sigma), and then purified using a PCR purification kit (QIAGEN). ChIP’d DNA was eluted with 50 μL of elution buffer and 1 μL of the elution was used for ChIP-qPCR. All ChIP experiments were performed in biological duplicates.
ChIP-qPCR was carried out in an Applied Biosystems 7500 Fast Real-Time PCR System (Thermo Fischer Scientific). Amplification reactions were performed in a final volume of 20 μL using the iTaq Universal SYBR Green Supermix (Bio-Rad) with 0.4 μM forward primer, 0.4 μM reverse primer, and 1 μL of immunoprecipitated DNA. Oligonucleotides used in ChIP-qPCR are listed in S3 Table. Percent input method was applied to analyse the enrichment of the promoter region for each gene and the values were calculated according to the Thermo Fischer Scientific web site (https://www.thermofisher.com/uk/en/home/life-science/epigenetics-noncoding-rna-research/chromatin-remodeling/chromatin-immunoprecipitation-chip/chip-analysis.html). ChIP-qPCR experiments were run in triplicates and the results are presented together with the background signal and standard error.
Virulence of wt, ΔhapC and hapCREC was compared in a model of systemic and pulmonary infection, both developed in four-week old OF-1 male mice (Charles River; Criffa SA, Barcelona). For the pulmonary model, groups of 20 animals (10 for survival and 10 for fungal burden studies) were immunosuppressed with 125 mg/kg of cortisone acetate, given intraperitoneally (i.p.), administered four days prior infection and then 3 days per week. Animals were anaesthetized by inhalatory sevofluorane and challenged by nasal instillation (i.n) with conidial suspensions of each strain containing 1x105 CFU/animal in a volume of 25 μl. Systemic infection was performed in groups of 16 animals (8 for survival and 8 for tissue burden studies) by intravenous (i.v.) inoculation into the lateral tail vein of 3x104 CFU/animal of each strain. Five days after i.v. or i.n. infection, animals included in the tissue burden study were euthanatised by CO2 anoxia. Liver, lungs, kidneys, spleen and brain from animals challenged i.v. were aseptically removed for CFU determination, while lungs and kidneys were used in the pulmonary model. Approximately, one half of each organ was weighted and mechanically homogenized in 1 to 1.5 mL of PBS. Homogenates were serially 10-fold diluted and placed onto potato dextrose agar plates for CFU/g determination.
|
10.1371/journal.pgen.1000137 | Positional Cloning of “Lisch-like”, a Candidate Modifier of Susceptibility to Type 2 Diabetes in Mice | In 404 Lepob/ob F2 progeny of a C57BL/6J (B6) x DBA/2J (DBA) intercross, we mapped a DBA-related quantitative trait locus (QTL) to distal Chr1 at 169.6 Mb, centered about D1Mit110, for diabetes-related phenotypes that included blood glucose, HbA1c, and pancreatic islet histology. The interval was refined to 1.8 Mb in a series of B6.DBA congenic/subcongenic lines also segregating for Lepob. The phenotypes of B6.DBA congenic mice include reduced β-cell replication rates accompanied by reduced β-cell mass, reduced insulin/glucose ratio in blood, reduced glucose tolerance, and persistent mild hypoinsulinemic hyperglycemia. Nucleotide sequence and expression analysis of 14 genes in this interval identified a predicted gene that we have designated “Lisch-like” (Ll) as the most likely candidate. The gene spans 62.7 kb on Chr1qH2.3, encoding a 10-exon, 646–amino acid polypeptide, homologous to Lsr on Chr7qB1 and to Ildr1 on Chr16qB3. The largest isoform of Ll is predicted to be a transmembrane molecule with an immunoglobulin-like extracellular domain and a serine/threonine-rich intracellular domain that contains a 14-3-3 binding domain. Morpholino knockdown of the zebrafish paralog of Ll resulted in a generalized delay in endodermal development in the gut region and dispersion of insulin-positive cells. Mice segregating for an ENU-induced null allele of Ll have phenotypes comparable to the B.D congenic lines. The human ortholog, C1orf32, is in the middle of a 30-Mb region of Chr1q23-25 that has been repeatedly associated with type 2 diabetes.
| Type 2 diabetes (T2D) accounts for over 90% of instances of diabetes and is a leading cause of medical morbidity and mortality. Twin studies indicate a strong polygenic contribution to susceptibility within the context of obesity. Although approximately ten genes making important contributions to individual risk have been identified, it is clear that others remain to be identified. In this study, we intercrossed obese, diabetes-resistant and diabetes-prone mouse strains to implicate a genetic interval on mouse Chr1 associated with reduced β-cell numbers and elevated blood glucose. We narrowed the region using molecular genetics and computational approaches to identify a novel gene we designated “Lisch-like” (Ll). The orthologous human genetic interval has been repeatedly implicated in T2D. Mice with an induced mutation that reduces Ll expression are impaired in both β-cell development and glucose metabolism, and reduced expression of the homologous gene in zebrafish disrupts islet development. Ll is expressed in organs implicated in the pathophysiology of T2D (hypothalamus, islets, liver, and skeletal muscle) and is predicted to encode a transmembrane protein that could mediate cholesterol transport and/or convey signals related to cell division. Either mechanism could mediate effects on β-cell mass that would predispose to T2D.
| Type 2 diabetes (T2D) afflicts ∼246 million people worldwide, including ∼21 million in the United States (7% of the population); another 54 million Americans are pre-diabetic. If the incidence of T2D continues to increase at the present rate, one in three Americans, and one in two minorities born in 2000 will develop diabetes in their lifetimes [1]. Direct medical costs associated with diabetes in the United States exceed $132 billion a year [2], and consume ∼10% of health care costs in industrialized nations.
Peripheral hyporesponsiveness to insulin increases metabolic demands on the insulin-producing β-cells of the pancreatic islets. Many obese individuals are insulin-resistant, but do not become overtly diabetic provided that the increased demand for insulin is effectively met [3],[4]. However, if β-cell mass and/or function are insufficient to meet this requirement, overt hyperglycemia and T2D ensue [5]. In autopsy series of subjects with T2D, total β-cell mass is decreased [6],[7]. Primary reductions of β-cell mass predispose to diabetes in rodent models [8],[9],[10] and in autosomal dominant forms of diabetes (e.g., MODY; maturity onset diabetes of youth) [11]. Such primary reductions might predispose to some instances of T2D.
Susceptibility to T2D is strongly inherited as evidenced by the >80% concordance rates in monozygotic twins [12],[13],[14],[15], familial aggregation, and ethnic predispositions [16]. Heritability of sub-phenotypes related to T2D, e.g. insulin resistance and β-cell hypofunction is even higher [17]. Environmental factors are also important [17],[18]. Although several genes for relatively rare monogenic forms of diabetes, including MODY, syndromic (Wolfram syndrome), lipoatrophic, and mitochondrial-inherited diabetes have been identified [2],[19], the underlying genetic bases for the genetically complex T2D, accounting for >95% of diabetes patients, have remained elusive. The identification of susceptibility genes is made difficult by the polygenic nature of the phenotype [20], its reflection of convergent, distinct metabolic processes producing identical phenotypes (phenocopies), and the potent gene-gene and gene-environment (e.g. obesity) interactions that characterize the disease. Clear genetic influences on the endophenotypes (intermediate phenotypes) of β-cell mass/function and insulin resistance have been shown, and vary among racial groups. [21],[22],[23],[24]. Some notable earlier successes (e.g. PPARG, CAPN10), and a recent series of genome-wide association studies of large numbers of well-phenotyped subjects [25],[26],[27],[28],[29],[30],[31] have identified T2D susceptibility loci/genes (e.g. TCF7L2) whose functions with regard to the implicated phenotypes are uncertain. As no single implicated gene or allele accounts for more than a small fraction of risk to develop T2D, there are still many genes/molecular mechanisms awaiting identification.
In mice, there is striking strain-dependent susceptibility to T2D in the context of obesity [32]. We exploited the differential diabetes susceptibilities of the B6 and DBA strains segregating for the obesity mutation Lepob [32] to identify a diabetes susceptibility QTL in B6xDBA progeny and then used congenic lines derived from the implicated interval to clone a candidate gene accounting for the QTL. Similar strategies have been used to identify QTLs (and responsible genes) for other complex phenotypes in mice [33] such as type 1 diabetes [34], diet-induced obesity [35], tuberculosis susceptibility [36], atherosclerosis [37], epilepsy [38], schizophrenia [39] and, also, T2D [40],[41],[42],[43].
We identified, “Lisch-like” (Ll), a novel gene, encoding multiple, tissue-specific transcripts in brain, liver and islets. The functional consequences of the hypomorphic DBA allele (diabetes-prone) in Lepob/ob mice appear to be late embryonic to early postnatal reductions in β-cell mass due to diminished rates of β-cell replication, some “catch-up” of β-cell mass by 2–3 months, followed by mild glucose intolerance at >6 months of age. These phenotypes are recapitulated in mice with an ENU-induced null allele of Ll.
We identified a QTL for diabetes-related phenotypes in obese F2 and F3 progeny of an intercross between diabetes-resistant C57BL/6J (B6) and diabetes-susceptible DBA/2J (DBA) mice segregating for Lepob. Phenotypes including fasting blood glucose, HbA1c and islet histology mapped with LOD >8 around D1Mit110 on distal Chr 1 at 169.6 Mb (details in Methods: Mapping T2D-related Phenotypes). By producing congenic and sub-congenic B6.DBA lines also segregating for Lepob, we refined the interval to 5.0 Mb between rs31968429 at 168.1 Mb and rs31547961 at 173.1 Mb where all four congenic lines overlap for DBA (Figure 1; details in Methods: B6.DBA Congenic Lines: Creation and Fine Mapping).
We further restricted the search (Figure 1) by identifying a haplotype block [44] conserved between B6 and DBA that extends 3.2 Mb from rs30708865 at 169.9 Mb to rs31547961 at 173.1 Mb. Only eleven unvalidated B6 vs. DBA single nucleotide polymorphisms (SNPs) in this interval are listed in the Mouse SNP database (www.ncbi.nlm.nih.gov/SNP/MouseSNP.cgi); however, among fragments we could amplify containing nine of these putative SNPs, we detected no sequence variants. Moreover, we found no coding sequence/expression difference between B6 and DBA among all genes and transcripts in the “conserved” interval by computation, direct sequencing, and quantitative mRNA expression analysis. Thus, it is unlikely that the variant(s) in the genetically-defined interval with peak at 169.6 Mb mediating differential diabetes susceptibility between these two strains is within the “conserved region.” We sequenced the 3 kb interval between rs31968429 and rs33860076 at the centromeric end of subcongenic line 1jcdt and detected no variants between the two strains. Therefore, we focused our efforts on the 1.8 Mb B6 vs. DBA “variable” interval, between rs33860076 at 168.1 Mb and rs30708865 at 169.9 Mb.
The congenic/sub-congenic lines shown in Figure 1 displayed phenotypes of hypoinsulinemic hyperglycemia in association with relative reductions in β-cell mass due to reduced β-cell proliferation (see Islet Morphology and β-cell Replication and Apoptosis). Phenotypes were generally more salient in male animals. Genotype in the congenic interval (B6 or DBA) per se did not affect their body weight or composition. Supporting experiments are described below.
By 4 weeks of age, fasting plasma glucose was elevated in Lepob/ob males who were D/D (DBA/DBA) for the congenic interval 1jcd and fed standard (9% fat) chow; glucose concentrations were higher up to 120 days. After 120 days, there were no significant differences in fasting glucose between D/D (DBA/DBA) and B/B (B6/B6) mice (Figure 2A). The decline in pre-prandial blood glucose levels in Lepob/ob males between 90 and 200 days is probably attributable to a slight expansion of β-cell mass in response to transient insulin resistance occurring as a normal consequence of sexual maturation (∼60 days of age) [9],[45]. To examine diabetes susceptibility in D/D animals that were obese independent of leptin deficiency, we fed lean (Lep+/+) 1jcd males a high-fat diet (60% kcal from fat) for 13 weeks, starting at 7 weeks of age. These mice became more hyperglycemic than B/B mice (Figure 2B), showing a persistence of this difference – similar to the animals in 2A – up to age ∼140 days when the study ended.
To delineate differences in acute glucose handling in D/D vs. B/B animals, we used intraperitoneal glucose tolerance testing (ipGTT). At 60 days (Figure 2C), and even up to 200 days, when the study ended (Figure 2D), Lepob/ob 1jcdc males were less glucose tolerant than B/B. The relative reduction in glucose tolerance in D/D vs. B/B animals that are not overtly diabetic is likely related to reduction in the number of islets. The occurrence of the diabetes-related phenotype is independent of Lepob, since 100-day old Lep+/+ 1jc D/D males fed the Surwit (high fat, high sucrose) diet for 10 weeks were also less glucose tolerant than littermate B/B males (Figure 2E).
Hyperglycemia due to relative hypoinsulinemia, was evident in 1jc Lepob/ob D/D animals fed a chow diet as early as 4 weeks (Figure 3A). At mean ages of 30- and 62-days, age-adjusted plasma insulin concentrations per mg blood glucose were lower in D/D than in B/B animals. This difference was due to lower plasma insulin in D/D (p = 0.0004) and not higher blood glucose in D/D (p = 0.916). Consistent with these ratios, D/D Lep+/+ males showed a 40% decrease in insulin secretion when clamped at a blood glucose level of 250 mg/dl for an hour (Figure 3B). No difference in insulin sensitivity was detected by euglycemic – hyperinsulinemic clamping (data not shown).
Consistent with their hypoinsulinemic hyperglycemia, 21-day old 1jcd D/D males had smaller islets than their B/B counterparts (Figure 4A). A qualitative cell-autonomous β-cell defect in insulin secretion, however, is unlikely to be the primary functional defect in D/D animals, since islets isolated from 28-day old 1jcd D/D males responded to graded glucose concentrations (2.8 mM–16.8 mM) or 10 mM arginine by secreting amounts of insulin comparable to age- and sex-matched B/B littermates (Figure 4B). Also consistent with insulin/glucose ratios and hyperglycemic clamp results, isolated islets from 60-day old 1jc Lepob/ob males fed normal chow and 100-day old 1jc Lep +/+ on the Surwit diet showed reduced insulin secretion at 2.8 mM and 5.6 mM [glucose] in D/D vs. B/B littermates. For reasons indicated below, the early glucose intolerance of D/D mice is probably due, in part, to a deficiency of β-cell mass.
The fractional area of the pancreas accounted for by β-cells [46] in Lepob/ob 1jcd males was examined in 20-, 60- and 150-day old mice. By 60 days a trend to reduced β-cell area was apparent in D/D, and by 150 days of age, β-cell mass of the 1jcd D/D sub-congenics was about half that of B/B littermate controls. B/D animals had β-cell masses that were about two-thirds of B/B littermate controls (Figure 5A). These findings are consistent with in vivo data showing onset of elevated blood glucose (see Figure 2A) and lower circulating insulin concentrations (relative to glucose) in D/D sub-congenics at ∼60 days of age (see Figure 3A), and persistence of decreased glucose tolerance at 200 days of age. The lower relative β-cell mass in D/D animals reflects fewer numbers of β-cells, rather than smaller sized β-cells. There were no differences in pancreatic weight between D/D and B/B male animals.
To assess the basis for the difference in β-cell mass by 60 days, we measured rates of β-cell replication and apoptosis. We co-stained pancreatic sections in 1jcd congenic 1- and 21-day old Lepob/ob male mice with antibodies to insulin and Ki67 antigen, a nuclear marker of proliferation expressed during all stages of the cell cycle except G0 [47]. To estimate the proportion of dividing β-cells, we normalized the number of Ki67 positive β-cells to the total number of insulin positive cells. Groups consisted of 4 B/B and 4 D/D 1-day old mice or 4 B/B, and 8 D/D 21-day old mice. In 1-day old D/D males, the rate of β-cell replication was ∼1/3 that of B/B littermates, whereas there was no difference in 21-day old animals due to normally reduced β-cell replication by the time of weaning (Figure 5B) [48],[49],[50].
The proportion of small islets (250–2000 µm2) in 21-day old Lepob/ob males was greater in D/D (1jc and 1jcd) mice (73%) than in B/B (60%); whereas the proportion of large islets (10,000–50,000 µm2) was lower (9% in D/D and 14% in B/B). This finding is consistent with the β-cell replication studies in P1 mice (Figure 5B), and recently reported evidence that new β-cells are derived from replication of pre-existing β-cells [51].
In 13-day old 1jc mice, when β-cell apoptosis is active [52], we did not detect significant differences between B/B and D/D islets in β-cell apoptosis using a TUNEL assay [53] and caspase-3 staining [54] (data not shown). Thus, the lower number of β-cells in D/D mice is primarily a result of lower rates of proliferation of β-cells in the perinatal period.
To identify all genes in the minimal DBA variable interval, (see above for definition) we screened 277 genes and transcripts, computationally predicted by GenScan, TwinScan, FGeneSH, Otto, or SGP2 that map to the interval. We excluded 50 single-exon transcripts (probably pseudogenes [55]) that did not belong to a transcript cluster and were not homologous to transcripts in the syntenic human interval, and 16 ribosomal gene transcripts, unique to this interval, that could not be specifically amplified due to their genomic redundancy, and manually curated the remaining 211 predicted transcripts. We rejected 63 that did not amplify in RNA/cDNA pools from multiple organs/ages of B6 and DBA mice (see Methods: Testing for Predicted Transcripts in cDNA Pools) and, using BLASTn, clustered the remaining 148 transcripts into 14 groups. These, correspond to 11 known genes and 3 predicted genes that we validated by amplification in cDNA pools.
A map of the “variable” interval shows 14 genes, flanked by Mael and Pbx1 (Figure 6). We analyzed all transcripts in the entire “variable” region.
The genetic variation accounting for differential diabetes-susceptibility in mice segregating B/B vs. D/D in the congenic intervals could be due to: 1) coding sequence variant(s) that alter the amino acid sequence of a protein(s); 2) regulatory variants, including anti-sense transcripts that affect expression and stability, and 3′ untranslated region (UTR) variants; or 3) splicing variants. We investigated all hypotheses.
To identify all non-synonymous B6 vs. DBA sequence variants in the “variable” interval, we collected genomic sequence for B6 and DBA strains from databases at NCBI and Celera [56], filled gaps using bi-directional sequencing to achieve 100% coverage of all coding sequences in both strains, and validated coding sequence variants by bi-directionally re-sequencing gene fragments encompassing each variant in both B6 and DBA strains. Consequently, we identified five non-synonymous single nucleotide variants: one in each of three FMO-like (flavin mono-oxygenase) genes, and two in chr1.1224.1 (Figure 6). The latter gene, we designated “Lisch-like” (Ll) because of its sequence similarity to a gene in mouse and rat, formerly known as Lisch7 (http://rgd.mcw.edu/), but now known as Lsr (lipolysis stimulated receptor).
Computational analysis of LL and the three FMO-like proteins using SNAP [57], PolyPhen [58], SIFT [59], PAM250 matrix substitution weights [60] and PROFacc [61] predicted that all of the amino acid substitutions were benign with respect to function. The SNAP scores obtained for our variant alleles, -1 (FMO13, K282E), -2 (FMO12, V239I), -3 (LL, A647V), and -6 (LL, T587A; FMO9, Q5R), indicate that there is a ∼60%, ∼69%, ∼79%, and ∼90% respective chance of the non-synonymous variants being neutral. Similarly, PolyPhen classified all variations as “benign” and SIFT scores were well above 0.05 (neutral). PAM weights of 0 and above suggest interchangeability of the respective amino acids throughout evolution. The % differences were low, suggesting that the DBA and B6 variants are equally likely to occur in related sequences (see Methods: Computational Methods for Evaluating Effects of nsSNPs).
We used Affymetrix microarrays to quantify those transcripts in the minimum congenic interval that we had validated by PCR-amplification (see Methods: Testing for Predicted Transcripts in cDNA Pools). We examined hypothalamus, islets, liver, soleus and EDL (extensor digitorum longus) skeletal muscle from DD and BB Lepob/ob congenic animals (see Methods: Microarray Gene Expression Analysis). These arrays did not contain elements for all of the 14 genes we confirmed in the interval: missing from the array were the 3 FMO genes. Therefore, we also used real-time qPCR, to quantify expression of each gene and confirmed transcript in tissues and organs central to diabetes (pancreatic islets, liver, skeletal muscle, adipose tissue and hypothalamus) in 90-day old male Lepob/ob 1jc D/D and B/B animals (see Methods: real time qPCR). Results of the microarray and qPCR experiments are shown in Table 1 and summarized in Figure 7A.
Among genes in the region, including Lmx1a [62], and Rxrg [63], that constitute candidates for susceptibility to T2D, we identified no non-synonymous SNPs (nsSNPs) and no multi-organ differences in expression levels between B/B and D/D animals. The most prominent and consistent differences in expression we did observe were for chr1.1224.1 (Ll), which was two to four-fold lower in 21-day old Lepob/ob D/D mice than in B/B mice in the diabetes-relevant tissues/organs by microarray analysis and up to twenty-fold lower by qPCR (Figure 7A). (We later show that Ll protein in hypothalamus is strikingly reduced in 1jc D/D vs. B/B; see Figure 11A). The difference in Ll gene expression in liver persists with age (Figure 7B) as does the difference in glucose tolerance in response to overt glucose challenge (see Figure 2D). Whether the differences in hepatic Ll expression are mechanistically related to differences in glucose homeostasis are unknown at this point; LL may influence hepatic gluconeogenesis, or the hepatic differences could simply mirror parallel and more physiological relevant changes in β-cells.
We also detected (by PCR) Ll transcripts in e7, e11, e15, and e17 whole mouse embryos, and in testis, kidney, heart, lung, uterus, eye, thymus and spleen. For the anti-sense interval between intron 9 and intron 7 (see below and Figures 1 and 8), we found higher expression levels in liver and hypothalamus of D/D v. B/B animals. This difference is consistent with a possible suppressive role for the D/D anti-sense transcript (see below). The Aldh9a1 gene, known to be highly expressed in human embryonic brain and involved in glycolysis and fatty acid metabolism, showed qualitative changes comparable to those seen in Ll. The mapping experiment that identified the interval of mouse Chr1 containing statistical signals related to T2D phenotypes would be expected to enrich for regions in which several genes might contribute to the phenotypes. Although Aldh9a1 may be such a gene, we chose to focus initially on Ll, since it showed the most striking quantitative differences in expression between D/D and B/B animals.
From the Ensembl database, we identified zebra fish orthologs of Ll and Lsr. The clustalW pair-wise similarity scores for the predicted protein coded for by the zebra fish gene zgc:114089 (Lsr ortholog) is 42 vs, the mouse LSR protein, and 29 vs. the mouse LL protein. The similarity scores for the predicted protein coded for by the zebra fish gene zgc:110016 (Lisch-like ortholog) are 36 vs. LL and 28 vs. LSR. We performed clustalW analysis (Figure 9) between the mouse LL-iso1 protein and three related proteins: 1) the human C1orf32 protein at 1q24.1 (chr.1 165,154,620–165,211,185; NCBI Build 36.1), which is the product of a gene highly expressed in the developing human retina and brain [64]; 2) the predicted protein sequence for the zebra fish Lisch-like ortholog, zgc:110016 located on zebra fish chromosome 9 at 31.6 Mb; and 3) the mouse LSR protein, transcribed from a gene on chromosome 7 at 30.7 Mb. Pair-wise similarity scores for the intact proteins and major domains are shown in the legend. The human homolog is similar throughout, but diverges slightly in the putative ICD. The zebra fish Lisch-like ortholog and mouse LSR proteins are most alike in the TMD, less so in the Ig-like domain, and most dissimilar in the ICD. The Lsr protein has a short extension to exon 6, and no exon 8 equivalent. Ll and Lsr also have splicing patterns similar to the mouse Ildr1 (Ig-like domain receptor 1) gene [65], and the proteins they encode all belong to the Lisch7 family (IPR008664; www.ebi.ac.uk/interpro).
To assess the function of Ll in islet/β-cell ontogenesis, we examined expression patterns and the effects of morpholino-mediated knockdown in zebra fish embryos. Morpholinos are modified anti-sense oligonucleotides that produce a strong hypomorphic “knockdown” phenotype [66] either by inhibiting proper splicing of the pre-RNA transcript [66] or by ATG-blocking of translation [67]. Morpholino knockdown has been used to demonstrate a role for the endocrine hormones GnRH, GHRH and PACAP during development [68],[69],[70],[71]. Many of the molecular mechanisms regulating pancreas development appear to be conserved among zebra fish and other vertebrates [72], and the single zebra fish islet provides an excellent model of vertebrate development.
Using whole mount in situ hybridization (Figure 10A), we observed that the Lisch-like ortholog zgc:110016 was expressed in the brain and otocyst by 48 hours post fertilization (hpf), and by 72 hpf expression was evident in the intestine. The Lsr ortholog zgc:114089, located on Chr 15 at 39.0 Mb, was expressed in pancreas at 48 and 72 hpf, (similar to our postnatal observations in mouse with Ll), intestine, liver, pharynx, pronehphros and otocyst for 48 hpf (72 hpf not shown), and, at 34 hpf, in both pancreatic buds. Since the anterior bud gives rise to exocrine tissue, pancreatic duct, and a small number of endocrine cells, while the posterior bud gives rise only to endocrine tissue [69], expression of the Lsr-like paralog throughout this stage is consistent with a role in the ontogeny of pancreatic endocrine tissue.
The close structural similarities among Lisch-related genes (see Figure 9) suggested that functional data on both zebra fish genes could be physiologically relevant and, therefore, we studied the involvement in islet development of both orthologs. We injected (in separate experiments) morpholinos for both genes into embryos homozygous for the gut-GFP (green fluorescent protein) transgene to visualize developing endodermal organs (Figure 10B) [73]. We assessed β-cell development with an anti-insulin antibody at 48 hpf or by insulin in situ hybridization at 24 hpf (not shown). To assess morpholino specificity, we analyzed the effects of two separate, non-overlapping morpholinos for each gene. Both morpholinos for each ortholog independently produced similar phenotypes, providing evidence that the effects (described below) were the result of specific gene knockdown and not due to nonspecific morpholino-related effects.
Figure 10B shows that both Lsr-like and Ll morpholinos injected at 15 ng/embryo produced general developmental delay in the endodermal organs, evidenced by a smaller liver, a smaller, straighter intestine, and a smaller pancreas that does not extend as much as in wild-type. The Lsr-like morpholinos disrupt β-cells more severely (note ectopic insulin-positive cells in the cephalad region of the pancreas) than do the Ll morpholinos (note the milder local dispersion of insulin-positive cells); 48/72 and 25/144 embryos injected with morpholinos targeting Lsr-like and Ll, respectively, displayed a scattered β-cell phenotype. These effects were rarely observed in uninjected sibling embryos (0/25) or embryos injected with a control morpholino (1/35). Lower doses of Lsr-like and Ll morpholinos (∼7–10 ng) resulted in a lower frequency of β-cell scattering and higher doses (∼20–25 ng) resulted in embryonic toxicity, which is common with high doses of morpholinos. The efficacy of the splice-blocking Lsr-like and Ll morpholinos was assessed via RT-PCR and all were found to strongly and specifically inhibit proper splicing of their respective target transcripts at the 15 ng dose (not shown). In combination, the expression analyses and morpholino knockdown studies provide support for a role of Lisch gene family members in endodermal development, and suggest specific effects on the embryonic β-cell. The relevance of such zebra fish studies to mammalian pancreas development has been shown earlier for Ptf1a [74],[75] and for Pdx1 [76].
To examine phenotypes of mice segregating for a null allele for Ll, we screened a repository of ENU-generated (N-ethyl-N-nitrosourea) mutant sperm DNAs from 18,000 C3HeB/FeJ G1 males (Ingenium; http://www.ingenium-pharmaceuticals.com/) for mutations in Lisch-like [77]. We detected a G/A substitution that encodes an amber stop mutation at threonine-87 [W87*] and also creates an EcoN1 cleavage site, which we used to genotype for the mutation. By in vitro fertilization, we generated W87* heterozygotes on the C3HeB/FeJ background, and bred these animals to generate progeny that were homozygous wild-type (+/+), homozygous mutant (−/−) or heterozygous (+/−) for the W87* mutation. Progeny were born at the anticipated Mendelian ratios, and the −/− animals did not appear grossly compromised.
To verify that the W87* homozygous mutant was hypomorphic for LL protein, we compared a Western blot of hypothalamic extracts prepared from C3HeBFeJ wild-type (+/+) and mutant (−/−) mice, with a second blot of hypothalamic extracts prepared from B/B and 1jc-D/D congenic mice. We probed both sets of filters with a polyclonal rabbit antibody generated to a conjugated polypeptide, corresponding to exons 7 and 8 of isoform 1, in the predicted ICD of LL. As anticipated, LL protein was greatly reduced in the brains of D/D vs B/B congenics and in the ENU-treated W87* homozygotes vs. the wild-type animals (Figure 11A).
In mice at 14 days of age we can detect reductions in β-cell replication rates that are similar to those seen in the DD congenic lines (Figure 5B) There is a >2-fold difference in the proportion of Ki67-positive β-cells in 14-day old wild-type (3.75%) vs. homozygous W87* mice (1.75%), with heterozygotes intermediate (2.5%) (Figure 11B). Plasma insulin concentrations in Ll W87* homozygotes are reduced by the time of sexual maturation (Figure 11C) and, consistent with this difference, at 50 days of age, homozygous W87* males show an increased glucose AUC during iPGTT (Figure 11D). A significant decrease in β-cell mass is also detected in W87* homozygotes (1.05%±.117, n = 3, p = .0113) v. +/+ littermates (2.74±.364; n = 3) at 150 days of age.
It is important to note that these phenotypes were detected despite the segregation of the mutation on a different background strain (C3HeB/FeJ) than our congenics (C57BL/6J), and in the absence of co-segregation of the Lepob. These preliminary data strongly support the candidacy of Ll as the gene accounting for the diabetes-related phenotypes of the DD congenic lines.
Based upon a QTL analysis of modifiers of T2D in B6xDBA F2 Lepob/ob mice, we identified a novel gene, Lisch-like (Ll), whose apparent effect on β-cell development, and possibly other aspects of β-cell/islet biology, qualify it as a strong candidate mediator of susceptibility to T2D. On the C57BL/6J strain background, the presence of the DBA/2J congenic interval(s) produced mild hypoinsulinemic hyperglycemia (in association with reduced β-cell replication and mass). Our preliminary data in ENU-mutagenized mice with a null Ll allele are consistent with a role for LL in β-cell development.
Three of the Ll subcongenic lines (1jcd, 1jcdt and 1jcdc) contain only DBA DNA 3′ of exon 7, while line 1jc is DBA for the entire gene and extends DBA for another 3 Mb 5′ of Ll. We infer, therefore, that coding and/or non-coding DBA vs. B6 variant(s) in the region of DBA overlap accounts for the phenotypic differences between the DBA congenic lines and animals segregating for B6 alleles in this region. In the region of overlap that includes the DBA vs. B6 “variable region” (Figure 6), Ll is the only gene showing anticipated differences in coding sequence and gene expression. These findings strongly support, but do not prove, the putative role of Ll alleles in conveying the phenotypic differences seen between the various DD and BB congenic lines. The phenotypes of the Ll W87* C3H mice also support our inferences regarding the candidacy of Ll based upon the B.D congenics.
There are two non-synonymous SNPs in Ll within the region of overlap among the congenic lines, in exon 9. However, their effects on protein function are predicted to be minor and it is unlikely that they determine the differences in either transcript abundance or protein level seen in the congenics. Variants in other regions of the gene are likely more relevant.
In the 5′ UTR, all but one of the eight variants are in simple repeats, where they are likely less significant. The interval underlying the anti-sense transcript contains 45 D/B variants, including a long, unique insertion. A regulatory role for the Ll anti-sense transcript is suggested by the similar location of anti-sense transcripts at the 3′ ends of the human C1orf32 (human ortholog of Ll) gene (e.g., DA322725 from hippocampus), the human LSR gene (DA320945, also from hippocampus), the human ILDR1 gene (AW851103), and the mouse Lsr gene (BY747866). Moreover, comparative inter-species transcriptomic analysis has identified the 3′ regions of transcripts as important in anti-sense regulation, and conserved overlap between species may be evidence of function [78]. For a recent review of anti-sense regulatory mechanisms, see [79].
We identified 52 B/D variants in the 3′ UTR, and it is estimated that the stability of 35% of yeast transcripts are regulated by motifs in the 3′ UTR [80]. Regulatory motifs, at a similar density, have been identified in the 3′ UTRs of several mammals, including mice [81]. A 3′ UTR polymorphism between two putative mRNA destabilizing motifs in PPPIR3 (muscle-specific glycogen-targeting regulatory PP1 subunit) has been genetically [82] and functionally [83] related to T2D. Variants in the 3′ UTR may also affect regulation by microRNAs (miRNAs). The 3′ UTR is the target of mammalian microRNAs (miRNAs) [84] and their relevance to diabetes is underscored by the finding that mouse islet-specific miR-375 affects insulin secretion [85].
The physiological role of Ll is unknown. Based upon the effects of D alleles of Ll on β-cell proliferation rates, β-cell mass, in vivo insulin release and glucose tolerance, (Figure 5) it is likely that Ll influences early β-cell differentiation/turnover in a manner that predisposes obese animals to later failure of β-cells by effects on mass and possibly function [86],[87]. The fact that these phenotypes are substantially recapitulated in W87* Ll C3H mice supports this inference.
In the neonatal rodent, extensive remodeling of β-cells occurs as a result of simultaneous activation of both apoptosis and β-cell replication [49]. Between 4 and 24 weeks, postnatally, β-cell mass is estimated to increase 10 fold, related in part to increased body mass [49]. Compensation for β-cell stress/loss in adult rodents is primarily by β-cell hypertrophy and β-cell proliferation [51]. In rats, β-cell proliferation rates decline from ∼20% per day in pups, to ∼10% per day at 6–8 weeks, and to ∼2% shortly thereafter [88]. However, even this low rate of turnover apparently does not persist in adulthood. Using continuous long term BrdU labeling in B6x129Sv and BALB/C one year-old mice, replacement rates as low as ∼1/1400 mature β-cells/day have been reported [89]. Consistent with this finding, pancreas mass in the mouse was recently shown to be irreversibly constrained by the size of a progenitor pool in the embryonic pancreatic bud [87]. These data suggest that β-cell mass established in the first 6–8 weeks of life may be critical to the ability to meet subsequent stresses on β-cell function imposed by e.g. obesity, hyperglycemia, and dyslipidemia. The molecular regulation of these processes is incompletely understood, but even transient interruptions may, based upon this formulation, result in permanent effects on cell mass, or function, or both [90]. Hypoactivity of the candidate T2D modifier gene (Ll) reported here could mediate such effects on establishment of initial β-cell mass, and/or later responses of cell hypertrophy/replication by β-cell-autonomous effects or in response to an exogenous ligand for this putative receptor.
Observations that expression levels of Ll are most strikingly affected in liver, the effects of the zebra fish knockdowns on general endodermal development, and structure/function considerations raised by the homologous LSR molecule [91], are consistent with the possibility that the mechanism(s) by which Ll conveys effects on cell mass/function might relate, in part, to consequences of putative effects on hepatic development/ function. IGF1 [92] and hepatic growth factor [93] are examples of such β-cell “hepatokines” affecting β-cell function.
Insight into the function(s) of the mouse Lisch-like protein may be gained from similarities in structure, expression, and cellular location with the human paralog, C1orf32, and with genes encoding related trans-membrane receptors, Ildr1 [65] and Lsr [91]. Splicing patterns of these genes generate isoforms, similar to those of Ll. Each gene's largest isoform includes an extra-cellular Ig-like domain, a single TMD, and a similar set of ICDs in related order. In one isoform of each protein, the TMD and cysteine-rich domains are absent. An evolutionary, regulatory relationship is suggested by the observation that the Ll-paralog and lldr1 are adjacent in the zebra fish genome (Zv6 assembly, UCSC Genome Browser). All three genes are abundantly expressed in the brain, liver and pancreas (and islets, where studied), and all are predicted to have 14-3-3 interacting domains (thus far experimentally verified for the human LSR) [94]. Although 14-3-3 interacting domains may be present on as many as 0.6% of human proteins, their occurrence on all of these Lisch-related proteins is notable, since among known 14-3-3-interacting proteins is phoshodiesterase-3B, which is implicated in diabetes and pancreatic β-cell physiology [95],[96],[97], and others, such as the Cdc25 family members, important in regulating cell proliferation and survival [98],[99].
The human ortholog of Ll, C1orf32, which is 90% identical to Ll at the amino acid level, maps to a region of Chr1q23 that has been implicated in T2D in seven ethnically diverse populations including Caucasians (Northern Europeans in Utah) [100], Amish Family Study [101],[102], United Kingdom Warren 2 study [103], French families [104], and Framingham Offspring study [105], Pima Indians [106], and Chinese [96] with LOD scores as high as 4.3. The mouse congenic interval examined here is in the middle of, and physically ∼10× smaller than, the 30 Mb human interval. Recent analysis of the broad interval ascertained in Utah identified two peaks, one of which, at D1S2762 (at 163.6 Mb), is just 12 kb telomeric to the 5′ end of C1orf32 [107]. The genes, and gene order, are generally conserved between mouse and human in the region syntenic to the congenic interval. The metabolic phenotypes documented in human subjects with T2D linked to 1q23 resemble diabetic phenotypes observed in congenic mice segregating for the DBA interval in B6.DBA congenics examined here [108], suggesting that the diabetes-susceptibility gene in congenic mice and human subjects may be the same gene, or among the genes, acting in the same genetic pathway(s). The syntenic interval in the Goto-Kakizaki (GK) rat also correlates with diabetes-susceptibility [109].
We report the molecular cloning and preliminary characterization of a candidate gene for a mouse QTL modifying T2D phenotypes in mice. The gene, Lisch-like, is novel in structure among diabetes susceptibility genes, and appears to modify β-cell development. Amino acid sequence analysis is consistent with the possibility that hypomorphism for this gene could affect β-cell development by a number of possible molecular mechanisms. Proof of the role of this gene in the imputed phenotypes and molecular processes awaits its further analysis in transgenic animals and cell-based systems.
Mice were housed in a barrier facility in ventilated Plexiglas cages under pathogen-free conditions at room temperature (22±1°C) with a 12 h light/dark cycle. Mice were weaned at 21 d and given ad libitum access to water and 9% Kcal fat Picolab Rodent Chow 20 (Purina Mills; www.purinamills.com/).The high fat diet protocol used in some animals is described below. Columbia University's Institutional Animal Care and Use Committee (IACUC) approved all protocols. After a 4 h morning fast, mice were sacrificed by carbon dioxide asphyxiation and phenotyped for weight, naso-anal length, and glycosuria. Blood was collected by cardiac puncture and aliquoted into microfuge tubes containing an anticoagulant cocktail of 10 µl of 1 mM EDTA and 1.5 mg/ml aprotinin (Sigma A-6279). Plasma and red blood cell pellets were used to measure glucose, insulin, and glycosylated hemoglobin as previously described [110]. Tissues (skeletal muscle, pancreas/pancreatic islets, liver, brain, hypothalamus, kidney, spleen, heart, visceral fat, retroperitoneal fat) were collected and immediately frozen in liquid N2, and stored at −80°C for further studies. Pancreata were dissected under stereoscope, weighed, and fixed in Z-fix zinc-formalin fixative (Anatech; www.anatechltdusa.com/).
Liver tissue or tail tips were used for genomic DNA isolation according to standard procedures [111]. A mutation-specific assay was used to confirm that all phenotypically obese animals were Lepob/Lepob and all lean animals either +/+ or heterozygous at the Lep locus [112] Animals were genotyped using MapPairs Microstaellite Markers (Invitrogen; www.invitrogen.com/) as previously described [113].
Maps were created using MapMarkerQTL (www.broad.mit.edu/genome_software/other/qtl.html) on a dataset representing 404 obese F2 progeny of a B6xDBA cross segregating for Lepob at 120–150 days of age. The QTL for T2D was most significantly associated with fasting blood glucose, glycosylated hemoglobin, and islet histology in male mice to a region of Chr1, with peak statistical significance at D1Mit110 at 169.6 Mb from the centromere (p<10−8) (Figure 12). Other QTLs were identified on other chromosomes (for example Chr5 at 78cM), but none had as great an effect on the phenotype or demonstrated consistent effects on all aspects of the phenotype. We tested for interactions for QTLs and identified a modest interaction between the locus on Chr1 and a second locus at D4Mit286 (p = 0.008).
B6.DBA congenic mice were generated by intercrossing Lepob/Lep+ B6 X DBA mice from Jackson Laboratory (www.jax.org/) to generate F1 progeny, followed by backcrossing to the recurrent B6 strain using a “speed congenic” approach in subsequent generations [114]. At the eighth backcross, a genome scan was performed in all breeders using polymorphic markers at 20 cM intervals. In the mouse line that was continued, all non-contiguous markers outside the DBA interval were homozygous B6. Over the next two generations, there were two recombination events, one that eliminated a telomeric portion of the DBA interval (line 1jc) and one that preserved approximately half of the originally defined DBA interval (line 1jcd). The 1jcd mouse was bred repeatedly to B6 mice, giving rise, by meiotic recombination, to two additional subcongenic lines (1jcdt and 1jcdc) (see Figure 1). Preservation of the phenotypes present in the original B6xDBA and DBAxB6 F2/F3 progeny was assessed by longitudinal and end-point measurements of fasting glucose, insulin, glycosylated hemoglobin and islet morphology. At N12, Lepob/+ mice B6/DBA (B/D) for the respective congenic intervals were intercrossed to produce N12F1 progeny. Obese progeny were used for fine mapping and phenotyping experiments. Lepob/+ animals D/D for the congenic interval were recurrently intercrossed or crossed to B6 Lepob/+ animals to generate ob/ob Lepob/Lepob animals with D/D and B/D genotypes for the Chr1 interval, respectively.
For longitudinal phenotyping studies, mice were fasted for 4 h and restrained for blood collection by a trained individual. Blood was collected from unanesthetized animals by capillary tail bleed into heparinized tubes and stored at −80°C. Glucose was measured with a FreeStyle Flash Blood Glucose Monitor (Abbott; www.abbottdiabetescare.com/). Insulin was measured by ultra-sensitive rat insulin ELISA (ALPCO; www.alpco.com/). HbA1c was measured by affinity chromatography (Mega Diagnostics; www.mega-dx.com/). Urine ketones were measured using Chemistrip Test Strips (Roche Diagnostics; http://us.labsystems.roche.com/index.shtml). For ipGTT, mice were fasted overnight and 0.5 g/kg body weight of 50% dextrose was administered intra-peritoneally at time 0. Plasma glucose was measured at 15–30 min intervals for 3 h, as above. Terminal phenotypic characterization consisted of measurements of fasting glucose, insulin, glycosuria, and glycosylated hemoglobin as previously described [110]. To control for stress-induced hyperglycemia at the time of sacrifice, tail blood glucose was also measured by glucometer one day prior to sacrifice.
High fat chow pellets (#D12492i: 60% kcal from fat, 20% kcal from protein, 20% kcal from carbohydrate) and “Surwit” [115] (#D12331i; 58% kcal from fat, 16.4% kcal from protein, 25.5% kcal from carbohydrate) (Research Diets; www.researchdiets.com/) were used as described in the text.
Pancreatic tissues were dissected under stereoscope to avoid contamination with adipose tissue, and weighed.
Non-overlapping images of longitudinal pancreatic sections were acquired and analyzed using ImageProPlus software version 5.0 (Media Cybernetics; www.mediacy.com/) to calculate insulin-positive area, insulin-positive area as % total area, and number of islets (defined by an area containing a minimum of 8 contiguous insulin-positive cells). For β-cell replication studies, we recorded the number of Ki67-positive or negative, insulin-positive cells. Replication of β-cells was expressed as % of cells (Ki67-positive and insulin-positive)/ total insulin-positive. For replication studies, ∼100 islets were examined per animal from several different non-overlapping sections through the pancreas. ImageProPlus or Image J (1.37 V; NIH) were used to determine the relative area of each section occupied by β-cells or the actual of number of β-cells for each representative longitudinal pancreatic section (50 µm apart) that had been immunochemically stained for insulin as previously described [116]. We analyzed 5–7 sections from different regions of the pancreas. Apoptosis rates were assessed using the DeadEnd Fluormetric TUNEL System G3250 (Promega; www.promega.com/) TUNEL assay and cleaved Caspase-3 (Asp175) Antibody 9661S (Cell Signaling Technology; www.cellsignal.com/).
Pancreatic perfusion and islet collection were performed as previously described [117]. Each pancreas was perfused via the bile duct with 1.5 mg/ml collagenase P (Roche Applied Science; www.roche-applied-science.com/) and incubated at 37°C for 17 min. Following disaggregation of pancreatic tissue, pancreata were rinsed with M199 medium containing 10% NCS. Islets were collected by density- gradient centrifugation in Histopaque (Sigma-Aldrich; www.sigmaaldrich.com/) [117], and washed several times with M199 medium. For glucose-stimulated insulin release studies [118],[119], islets were incubated overnight in RPMI medium 1640 (Invitrogen).
The GSIS procedure has been described previously [120]. Islets were hand-picked into tissue culture dishes containing cold Kreb's buffer (118.5 mM NaCl, 2.54 mM CaCl2, 1.19 mM KH2PO4, 1.19 mM MgSO4, 10 mM HEPES, pH 7.4), and 2% BSA (Sigma-Aldrich), 5.5 mM glucose, and incubated overnight at 37°C. Islets were hand-picked and incubated another 15 min. in Kreb's buffer+BSA, containing 11.2 mM glucose. Hand-picked islets are then resuspended in Kreb's buffer plus BSA, supplemented with 2.8 mM glucose, and shaken at 37°C for 15 min. The pellet was spun down gently and resuspended in triplicate (5–10 islets each) in 500 µl Kreb's buffer, supplemented with glucose at final concentrations of 2.8 mM, 5.6 mM, 11.2 mM or 16.8 mM, or supplemented with 10 mM arginine and incubated for 1 h in a water bath at 37°C with constant shaking (300 rpm). After 1 h incubation, islets were gently pelleted and the supernatant collected and assayed for insulin by ELISA. Islet pellets were dissolved in high salt buffer (2.15 M NaCl, 0.01 M NaH2PO4, 0.04 M Na2HPO4, EDTA 0.672 g/L, pH 7.4) and sonicated at 4–5 W for 30 s and DNA concentration was measured using a TKO100 fluorometer (Hoefer; www.hoeferinc.com/) with Hoechst #33258 dye (Polysciences; www.polysciences.com). Results were expressed as concentration of secreted insulin/[DNA]/h.
Putative transcripts, identified from public annotation and local sequencing, were validated by PCR-amplification from tissue-specific cDNA pools prepared from male and female B6 mice. Two cDNA pools were used: 1. An inclusive cDNA pool was prepared from E7 and E20 fetuses and P1 pups, and included the following tissues of 60-day old mice: eyes, large intestine, skin, tongue, spinal cord, kidney, testes/ovaries, pancreatic islets, whole brain, hypothalamus, skeletal muscle, and liver. This pool was used for transcript validation. 2. A diabetes-relevant cDNA pool, from 90-day old mice, was comprised of only the following tissues and organs: pancreatic islets, whole brain, hypothalamus, skeletal muscle, liver, and adipose tissue. This pool was used to quantify transcripts identified by computational approaches and the microarrays. Nominal intron-spanning primers were generated using the Primer3 program (www.genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi). Amplification was first performed on the diabetes-relevant pool at an annealing temperature of 60°C. If we detected no PCR-product, we performed gradient temperature PCR on the same pool using eight different annealing temperatures from 58–68°C. Gradient temperature PCR was then used to amplify the inclusive cDNA pool. If no product was detected in this pool, a 2nd set of intron-spanning primers was used before we interpreted negative amplification as failure to substantiate a predicted transcript. Positive amplification products of predicted sizes, and those that did not match the expected sizes, were gel-purified and sequenced for confirmation. The final set of primer-pairs is listed in Real-time qPCR.
RNA extraction, purification, labeling, hybridization and analysis were performed as described [121]. 10 BB and 10 DD 21-day old Lepob/ob 1jc males were dissected and RNA was extracted from hypothalamus, liver, isolated islets, EDL muscle, and soleus muscle. Individually labeled RNA (by mouse and organ) was interrogated with Affymetrix MOE-430A expression arrays. For further details, see legends to Table 1 and Figure 7. For all transcripts in the region of interest, where possible, only probes that spanned multiple exons and clearly represented each of the 14 genes in the interval were used. If >1 probe met these conditions, we used only, the probe that gave the strongest signal. Organs were grouped into two groups by genotype and were compared using a two tailed T-test. The Affymetrix probe IDs selected for this analysis are shown in Table S3.
Effects of the DBA congenic interval on the levels of confirmed transcripts expressed in diabetes-relevant organs were assessed on an organ-specific basis. We made separate pools from 90-day old Lepob/ob 1jc D/D and B/B mice for each of the diabetes-relevant organs (see above). Each individual organ pool was generated on 2 occasions from 5 mice. RNA was extracted from organs with TRIzol acid-phenol reagent (Invitrogen). 2 µg of RNA were reverse-transcribed using SuperScript III reverse transcriptase (cDNA First Synthesis Kit, Invitrogen) with random hexamer priming. The cDNA was diluted 4-fold using nuclease-free water (QIAGEN; www.qiagen.com). 2 µl of diluted cDNA were amplified by PCR in Roche LightCycler. A standard curve for each transcript was generated using cDNA diluted 1∶1, 1∶10, and 1∶100. We assessed the number of mRNA molecules in each sample using the slope and intercepts of PCR product appearance during the exponential phase of the PCR reactions optimized for transcript-specific product using specific primers. Each sample was run in triplicate in the same LightCycler run. Using LightCycler Software, we calculated the crossing point (CP) for each sample. The CP is the first maximum of the second derivative of the fluorescence curve, and is equivalent to the number of cycles at which the fluorescence first exceeds background. In the exponential phase, the relationship between CP and initial transcript concentration is linear. We calculated relative concentration ratios, normalized to actin, as follows:
In this expression, ΔCPgene is the CP of the gene in the sample minus the CP of the gene in the relevant reference; ΔCPhg is the CP of the housekeeping gene in the sample minus the CP of the housekeeping gene in the reference (“ref”) sample; and η is the efficiency (where 2 is perfectly efficient) as determined by the negative slope of the plot generated when CP is plotted as a function of the log of initial concentration determined in the standard curve. Each CP listed is the mean of CP values of the triplicates for each sample. Results are summarized in Table 1. Primers used are listed in Table S4 (A).
We amplified full-length Ll cDNAs from either B6 islets (isolated by us) or from Clontech MTC Panels 1 #636745 and 3 #636757, containing pooled multiple tissue cDNAs from 8–12 week old BALB/c mice and from Swiss Webster embryos. In a final volume of 50 µl, we added 0.5 µl LA Taq (TaKaRa; www.takara-bio.com/) to a cocktail containing TaKaRa GC Buffer II, 400 µm each dNTP, 1 µl cDNA and 1 µl each primer (300 ng/µl). Primers are listed in Table S4 (B). Samples were cycled in an MJ Tetrad Thermalcycler (BioRad; www.bio-rad.com) using a Touchdown protocol of a 2 min. extension and decreasing annealing temperature from 60°C to 55°C for 10 cycles, followed by 25 cycles with an annealing temperature of 55°C. Each sample was TOPO TA cloned (Invitrogen) and plated. From all three libraries, a total of 140 colonies were picked and grown overnight in LB buffer. Inserts were amplified by colony PCR and sized by gel-fractionation. Inserts representing each unique size were then sequenced. The isoforms and the exons deleted (Δ): iso1 (intact 10 exons); iso2, Δ6; iso3, Δ4,5,6; iso4, Δ4; iso5, Δ5,6; iso6, Δ9; iso7, Δ5,6,7,8,9.
We used five methods to compute the likelihood of a functional change due to single amino acid substitutions (see Figure 9). SNAP, PolyPhen, and SIFT predict changes in protein function due to a single amino acid substitution. SNAP [57] is a neural-network based method that considers protein features predicted from sequence (e.g., residue solvent accessibility and chain flexibility). Scores from −9 to +9 are estimates of accuracy of prediction, computed using a testing set of ∼80,000 mutants. A low negative score indicates confidence in prediction of neutrality (functional change absent), whereas a high positive score indicates confidence in prediction of non-neutrality (functional change present). Accuracy was computed for neutrals using the equation below:
PolyPhen considers structural and functional information and alignments. Predictions are sorted into 4 classes: benign, possibly damaging, probably damaging, and unknown.
SIFT predictions. SIFT [59] is a statistical method that only considers alignments. Scores range from 0 to 1. Scores >0.05 indicate neutrality of a substitution.
PAM250 matrix substitutions. PAM matrix [124] (Percent Accepted Mutations) reflects frequency of amino acid interchange throughout evolution (by evaluating alignments of proteins in a family). Scores range from a low of −8 for rare substitutions (e.g. W to C) to a high of 17 (same residue found in almost all proteins in alignment).
Percentage in alignment (PROFacc). The score is reported as the difference in observed percentages of wild-type and mutated residues in alignments against a non-redundant UniProt [125] and PDB [126] database (at 80% sequence identity).Scores range from −100 (if the mutant is observed in all instances) to +100 (if the wild type is observed in all instances); 0 if the mutant is observed as often as the wild type. Scores near 0 favor the likelihood of a mutation being neutral.
BAC 95f9 DNA (5 µg) was fragmented to 1–5 kb using a nebulizer supplied with the TOPO Shotgun Subcloning kit (Invitrogen) and checked for size and quantity on an agarose gel. The shotgun library was constructed with 2 µg of sheared DNA. Blunt-end repair, dephosphorylation, ligation into PCR 4Blunt-TOPO vector, and transformation into TOP10 Electrocompetent E. coli were performed with the TOPO Shotgun Subcloning kit, following the manufacturer's protocol. Phenol∶chloroform extraction of the dephosphorylated DNA was replaced with Qiagen QIAquick PCR Purification spin columns (QIAGEN). Recombinant colonies were selected by blue/white screening and incubated in LB medium supplemented with 50 µg/ml ampicillin for 20 h at 37°C in 96-well deepwell plates. Plasmid miniprep was conducted in 96-well plates using QIAGEN Turbo Miniprep kits on a QIAGEN BioRobot 9600. DNA sequencing was performed on a 3730xl Genetic Analyzer (Applied Biosystems; www.appliedbiosystems.com/) using BigDye® Terminator v3.1 Cycle Sequencing Kits with M13 forward and reverse sequencing primers.
ANOVA and ANCOVA were used to assess effects of genotype in congenic interval. Comparisons at individual time points, or pairs of means were performed using Student's t-test. P values are 2-tailed. The Statistica package (StatSoft; www.statsoft.com/) was used for ANOVAE; Excel (Microsoft, http://office.microsoft.com/en-us/default.aspx) for t-testing.
Hypothalamic extracts were prepared using M-PER Mammalian Protein Extraction Reagent (Pierce Biotechnology, www.piercenet.com/). Hypothalamic extracts (85 mg for B/B and D/D congenics and 175 mg for wild-type and mutant ENU mice) were resolved by 8% SDS-PAGE, transferred to nitrocellulose membrane (Invitrogen). We generated a set of polyclonal rabbit antibodies (Covance Research Products; www.covance.com) against the predicted ICD, spanning residues 298–401 (exons 7,8) and verified that the α-ICD rabbit antibodies detected the appropriate fusion proteins, with only minor cross-reactivity in cultured cells. We hybridized the blot with anti-LL anti-sera at a dilution of 1∶5,000 in TBS/0.05%Tween/5% milk (TBSTM) or with blocked anti-LL anti-sera diluted 1∶10,000 in TBSTM. To prepare blocked anti-sera, liver sections from C3HeB/FeJ knock-out mice were fixed overnight in phosphate-buffered paraformaldehyde at 4°C and rinsed in PBS. Sections equivalent to one-third of a liver were fragmented and mixed with 1 ml anti-sera diluted 1/1000 in PBS/0.1% Triton. Liver fragments were spun out and the supernatant was used to probe filters from ENU mice. We detected bound antibody with horseradish peroxidase-coupled antibody against rabbit IgG (Amersham Biosciences; www.amershambiosciences.com) at a dilution of 1∶5,000 using the SuperSignal West Pico Chemiluminescent Substrate kit (Pierce Biotechnology).
Genbank (www.ncbi.nlm.nih.gov/(Genbank) accession numbers for the M. musculus genes: Lisch-like, lipolysis-stimulated remnant receptor-related (XM_001473525); Lsr (NM_017405); Ildr1 (NM_134109); Tada1l SPT3-associated factor 42 (NM_030245); Pogk pogo transposable element with KRAB domain (NM_175170); FMO13, flavin-containing monooxygenase family; FMO-like (XM_136366); FMO9, flavin-containing monooxygenase family; FMO-like (NM_172844) FMO12, flavin-containing monooxygenase family; FMO-like (XM_136368); C030014K22Rik, unknown (NM_175461); Uck2, uridine monophosphate kinase (NM_030724); Tmco1, membrane protein of unknown function (NM_001039483); Aldh9a1, aldehyde dehydrogenase 9, subfamily A1 (NM_019993); Mgst3, microsomal glutathione-S-transferase 3 (NM_025569); Lrrc52, leucine-rich repeat (LRR) protein of unknown function (NM_00103382); Rxrg, retinoid X receptor, gamma (NM_009107); Lmx1a, LIM homeobox transcription factor 1, α (NM_033652); Pbx1 (NM_008783); H.sapiens C1orf32 (NM_199351); LSR (NM_015925); ILDR1 (NM_175924); D.rerio Ll ortholog, zgc:110016 (NM_001030192.1); D. rerio Lsr ortholog, zgc:114089 (NM_001025472.1); R. rattus Lsr (NM_032616).
The Genbank accession numbers for protein sequences: M. musculus Lisch-like (amino acid residues 150–795, XP_001473575); (Lsr) (NP_059101); Ildr1 (NP_598870); H. sapiens C1orf32 (NP_955383); LSR (NP_057009); ILDR1 (NP_787120); D. rerio Lisch-like (NP_001025363); D. rerio Lsr (NP_001020643); R. rattus Lsr (NP_116005)
|
10.1371/journal.pgen.1003047 | A Combination of H2A.Z and H4 Acetylation Recruits Brd2 to Chromatin during Transcriptional Activation | H2A.Z is an essential histone variant that has been implicated to have multiple chromosomal functions. To understand how H2A.Z participates in such diverse activities, we sought to identify downstream effector proteins that are recruited to chromatin via H2A.Z. For this purpose, we developed a nucleosome purification method to isolate H2A.Z-containing nucleosomes from human cells and used mass spectrometry to identify the co-purified nuclear proteins. Through stringent filtering, we identified the top 21 candidates, many of which have conserved structural motifs that bind post-translationally modified histones. We further validated the biological significance of one such candidate, Brd2, which is a double-bromodomain-containing protein known to function in transcriptional activation. We found that Brd2's preference for H2A.Z nucleosomes is mediated through a combination of hyperacetylated H4 on these nucleosomes, as well as additional features on H2A.Z itself. In addition, comparison of nucleosomes containing either H2A.Z-1 or H2A.Z-2 isoforms showed that significantly more Brd2 co-purifies with the former, suggesting these two isoforms engage different downstream effector proteins. Consistent with these biochemical analyses, we found that Brd2 is recruited to AR–regulated genes in an H2A.Z-dependent manner and that chemical inhibition of Brd2 recruitment greatly inhibits AR–regulated gene expression. Taken together, we propose that Brd2 is a key downstream mediator that links H2A.Z and transcriptional activation of AR–regulated genes. Moreover, this study validates the approach of using proteomics to identify nucleosome-interacting proteins in order to elucidate downstream mechanistic functions associated with the histone variant H2A.Z.
| Within the cell's nucleus, DNA closely associates with histone proteins, forming a structure known as chromatin. Packaging DNA into chromatin allows for efficient storage of the genome, and it also provides an additional means of regulating processes, such as gene expression, that require access to DNA. Two copies each of the four core histones (H2A, H2B, H3, H4) associate with approximately 150 base pairs of DNA to make up the basic unit of chromatin, the nucleosome. In addition to the core histones, variants exist that have specialized functions within chromatin. One such variant is H2A.Z, which is essential for cell viability. Here, we describe an approach by which to characterize proteins that interact with H2A.Z-containing nucleosomes. Our findings reveal that many of the identified proteins may interact with H2A.Z nucleosomes by recognizing specific chemical modifications uniquely present on H2A.Z nucleosomes. One such protein, Brd2, interacted in a manner dependent on recognition of acetylated histone residues that are enriched on H2A.Z nucleosomes. Furthermore, this interaction is required for expression of hormone-responsive genes in prostate cancer cells. By this approach, we uncovered a key mediator linking H2A.Z to transcriptional regulation and found a potentially targetable step to regulate prostate cell proliferation.
| H2A.Z is a variant of the canonical histone H2A. Amongst the different variants of core histones that have been identified to date, H2A.Z is unique in being the only variant that is essential for viability and development in a number of organisms [1], [2], [3], [4]. H2A.Z has been implicated to function in multiple cellular pathways, including maintenance of chromosome stability and segregation [5], [6], [7], prevention of the spread of heterochromatin [8], as well as regulation of transcription (for review see [9], [10]). Currently, the essential function of H2A.Z is unknown. Although H2A.Z participates in diverse cellular pathways, its functional mechanisms have yet to be fully elucidated. Like all other histones, H2A.Z is subjected to post-translational modifications (PTMs), which may further modulate its function in the different pathways. For example, acetylation of lysine residues at the N-terminus has been linked to transcriptional activation [11], [12], [13], [14], whereas ubiquitylation of the C-terminus is associated with transcriptionally inactive chromatin [15], [16]. These studies have led us, and others, to propose that H2A.Z physically poises chromatin at transcription start sites and that PTMs on H2A.Z further specify its function in transcriptional activation or repression [9], [17].
At the amino acid level, mammalian H2A and H2A.Z share about 60% identity. Knockout of the H2A.Z gene is lethal in mice, which suggests that the unique regions of H2A.Z are required for the essential function in complex eukaryotes. Moreover, these unique regions likely engage effector proteins that mediate H2A.Z-specific functions. A number of studies have examined and identified H2A.Z-interacting proteins; however, these studies have focused on purifying soluble, tagged H2A.Z, often from whole-cell extracts, to identify proteins interacting directly with this variant [18], [19], [20], [21]. Not surprisingly, these studies mostly identified H2A.Z chaperone proteins and chromatin remodeling complexes that deposit H2A.Z into the chromatin fibre. While these studies have yielded invaluable information regarding the biology and regulation of the targeting and incorporation of H2A.Z into chromatin, they were not as informative in elucidating the functions of H2A.Z downstream of its deposition. To that end, we are interested in identifying effector proteins that engage H2A.Z on chromatin since these proteins likely contribute to the physiological functions associated with H2A.Z. For this purpose, we used a mass spectrometry-based approach to identify proteins that preferentially associate with H2A.Z nucleosomes. We specifically chose mono-nucleosomes, instead of soluble H2A.Z, as the bait for two reasons: First, histones exist in the context of nucleosomes in vivo and there are now many examples of proteins that contact multiple sites on nucleosomes to stabilize their interactions with chromatin (for recent examples see [22], [23], [24], [25]). Second, we have previously found that the general H3 methylation status on H2A.Z-nucleosomes is distinctly different from H2A-nucleosomes [16]. This suggested that H2A.Z-nucleosomes have unique histone PTM signatures that could collectively influence the engagement of downstream effector proteins. This idea is consistent with the concept of multivalency whereby multiple histone PTMs contribute to the overall binding and stability of chromatin-binding proteins or complexes that contain multiple histone PTM-binding motifs [26].
Using our nucleosome purification-mass spectrometry analyses, we indeed identified a number of proteins that preferentially associate with H2A.Z-nucleosomes over H2A-nucleosomes. Gene ontology (GO) analyses showed that the majority of the interacting proteins are chromosome- or chromatin-associated proteins. Consistent with the transcription-related functions of H2A.Z, many of the identified proteins have putative transcription-associated functions. Of the top 21 identified proteins, we focused our follow-up studies on Brd2 because of its transcriptional co-activator function and because it contains bromodomains that bind acetyl-lysines.
Brd2 belongs to the BET family of proteins, all of which contain tandem bromodomains in their N-termini and an extraterminal domain of unknown function in their C-termini [27]. Brd2 has essential cellular functions as evidenced by the early embryonic lethality of homozygous null mice [28], [29]. Moreover, Brd2 hypomorhpic mice become extremely obese when placed on a regular diet, and yet they avoid the development of insulin resistance and diabetic disease [30]. In contrast, over-expression of Brd2 specifically in the B-cell compartment of mice leads to development of leukemia [31]. Although the exact role of Brd2 in these cellular processes is not clear, the importance of its function may have to do with its ability to act as a transcriptional co-activator. It has been reported that Brd2 is involved in the transcriptional activation of cell cycle regulatory genes cyclin A, D1 and E in combination with Ras or MEKK [32]. Furthermore, Brd2's role in transcription is also evidenced by its reported associations with components of transcriptional machinery such as E2F, TBP, and the largest subunit of Pol II [33], [34], [35], [36], and it has also been shown to facilitate transcription through acetylated chromatin [37].
In this study, we found that Brd2's association with H2A.Z nucleosomes is enhanced upon treatment of cells with trichostatin A (TSA) and is dependent on its bromodomains. Peptide competition assays suggest that acetylated H4 is the primary site of interaction between Brd2 and H2A.Z nucleosome and, indeed, we found that the overall H4 acetylation levels are higher on these nucleosomes than on H2A nucleosomes. However, experiments using re-assembled H2A- and H2A.Z-nucleosomes that contain equivalent amounts of H4 acetylation suggest that additional features specific to H2A.Z further augment Brd2's interaction with nucleosomes. This conclusion is further supported by the fact that more Brd2 co-purifies with H2A.Z-1, compared to H2A.Z-2, -containing nucleosomes, even though they both have similar levels of H4 acetylation. In vivo experiments showed that, following hormone stimulation, Brd2 is recruited to androgen receptor (AR)-regulated genes in an H2A.Z-dependent manner. Finally, the small molecule inhibitor JQ1 blocked recruitment of Brd2 to AR–regulated genes, prevented transcriptional activation of these genes, and inhibited prostate cancer cell proliferation. All together, our analyses identified new biologically relevant interactions between nuclear factors and H2A.Z nucleosomes, and yielded insights into how H2A.Z containing-chromatin engages and recruits Brd2 to mediate downstream functions in transcriptional regulation.
To understand the physiological functions of chromosomal H2A.Z, we took an unbiased proteomics approach to identify proteins preferentially interacting with H2A.Z-containing nucleosomes. To do this, we first transfected 293T cells with Flag-tagged H2A.Z (or Flag-tagged H2A as a control), digested the chromatin to mononucleosomes with micrococcal nuclease (MNase) and then immunoprecipitated intact nucleosomes with anti-Flag antibody. By this method, we isolated and analyzed the co-purifying proteins by LC-MS/MS (Figure 1A; see Materials and Methods for details). Following three independent purifications and MS analyses, the cumulative data were filtered (see Materials and Methods for details) to generate a list of proteins that preferentially interact with H2A.Z nucleosomes (Figure 1B). Gene ontology analysis of the top 21 identified proteins (Figure 1D) revealed that most of the proteins are chromatin-associated proteins involved in chromatin organization and transcription. For example, our screen identified components of the H2A.Z remodeling complex SRCAP, including DMAP1, RUVB1, RUVB2 and VPS72. In addition, the transcriptional regulator WDR5 was one of the top hits. WDR5 is involved in transcription via its association with mammalian H3K4 methyltransferase complexes [38], [39], [40], [41]. Insofar as WDR5 preferentially binds di- and tri-methylated H3K4 [42], this is consistent with our previous finding that H2A.Z nucleosomes are enriched for di- and tri-methylated H3K4 [16].
Interestingly, many of the other identified proteins have conserved domains that can recognize various histone PTMs. For example, PHF6 and PHF14 have PHD fingers, which are motifs that can recognize methylated lysine residues [43]. Similarly, PWWP2A contains a PWWP domain, which is part of the Royal Family of domains that also recognize methylated lysine residues [44]. Finally, PHIP and Brd2 both have bromodomains, which are well-characterized acetyl-lysine binding motifs [45], [46]. Given that many of the interacting proteins contain motifs known to bind post-translationally modified histones, the overall PTM signature on H2A.Z nucleosomes likely mediates and contributes to these interactions.
Brd2 is a double-bromodomain protein known to be involved in transcriptional activation. Given that both H2A.Z and Brd2 are known to have transcription-related functions, we chose to perform follow-up experiments to further validate and characterize the interaction between Brd2 and H2A.Z nucleosomes. First, we confirmed this interaction by repeating our mononucleosome IP experiments and compared the amounts of endogenous Brd2 that co-immunoprecipitated with the H2A/H2A.Z nucleosomes by Western blots (Figure 2A; two additional proteins from our list, USP39 and PWWP2A, were also validated as shown in Figure S1). Since Brd2 contains 2 bromodomains [47], [48], [49], [50], [51], we next tested whether hyperacetylating the histones, prior to harvest and immunoprecipitation, would enhance the interaction between Brd2 and chromatin. In control cells (treated with DMSO), Brd2 was immunoprecipitated with H2A.Z nucleosomes, whereas no detectable levels of Brd2 were observed on H2A nucleosomes (Figure 2A—compare lanes 1 & 3 from IP fraction). Treatment of cells with the histone deacetylase inhibitor trichostatin A (TSA) resulted in an overall increase in histone acetylation as confirmed by acetyl-H4 Western blots (lanes 2 & 4 in both the INPUT and IP fractions). More importantly, TSA-treatment greatly enhanced the interaction of Brd2 with H2A.Z nucleosomes. The hyperacetylated H2A nucleosomes also pulled down some Brd2 but the amount is minor compared to that found on H2A.Z nucleosomes (compare lanes 2 & 4 in the IP fraction). All together, these results validated our MS data and showed that Brd2 is a novel H2A.Z-nucleosome-interacting protein. Moreover, this interaction is enhanced when the histone acetylation levels are increased.
To determine if the enhancement of Brd2 binding to H2A.Z nucleosomes following hyperacetylation is specific to Brd2, we compared the binding properties of Brd2 to that of a known H2A.Z-binding protein, VPS72. VPS72 is the human homologue of Swc2, which is a component of the H2A.Z deposition complex Swr1, and directly interacts with H2A.Z, within the Swr1 complex [52]. We first co-transfected Flag-H2A.Z together with YFP-Brd2 or HA-VPS72, and then performed our mononucleosome IPs. Like the endogenous protein, YFP-Brd2 was immunoprecipitated with H2A.Z nucleosomes, and the amount of protein immunoprecipitated was greatly increased in cells treated with TSA (Figure 2B). In contrast, the same amount of HA-VPS72 was immunoprecipitated with H2A.Z nucleosomes regardless of the acetylation status of the chromatin. Therefore, increased binding under hyperacetylated conditions is not a general property of H2A.Z nucleosome-interacting proteins. Interestingly, we also note that over-expression of the YFP-Brd2 construct increases levels of H4Ac, which has been reported previously [53].
Since the interaction between Brd2 and H2A.Z nucleosomes greatly increased when the chromatin is hyperacetylated, this suggests that the interaction is mediated through the bromodomains of Brd2 and the acetylated lysines on histones. To test this, we compared the ability of H2A.Z nucleosomes to immunoprecipitate wild type (WT) Brd2, to a mutant version (BD) of Brd2 that contains point mutations in each of its bromodomains, which renders the domains incapable of binding acetylated lysine residues [53]. Even though the BD mutant expressed at a much higher level than the WT-Brd2 in the transfected cells (Input panel in Figure 2C), the BD mutant was not immunoprecipitated with H2A.Z nucleosomes under basal or hyperacetylated conditions (compare IP fractions lanes 3 & 4 with lanes 5 & 6 in Figure 2C). It is noteworthy that endogenous Brd2 was also immunoprecipitated in both sample sets at comparable levels, as detected by anti-Brd2 antibody. Therefore, the interaction of Brd2 with H2A.Z nucleosomes is directly dependent on the bromodomains of Brd2.
Previously, it has been reported that the bromodomains of Brd2 bind acetylated lysine residues on H4 [47], [48], [49], [50], [51]. Inasmuch as the H2A.Z N-terminal tail also harbours multiple acetylated lysines at similar intervals as those found on the H4 tail [11], [12], [54], [55], it is therefore possible that the bromodomains of Brd2 may also bind acetylated H2A.Z. To test this, we initially performed peptide pull-down assays using recombinant BD2 of Brd2 and peptides corresponding to acetylated H4, H2A.Z, and H2A. However, under our assay conditions, we could not detect binding of BD2 to any of the tested peptides, including the positive control H4 peptides (data not shown). It is possible that our recombinant BD2 did not include enough flanking sequences for proper folding. Alternatively, the reported binding co-efficient for Brd2 BD2 binding to AcH4 peptides is in the mM range [47], [50], which suggests that this interaction is very weak and, therefore, our peptide pull-down conditions may not have been optimized for efficient detection. As an alternative approach, we used the same series of peptides in peptide competition assays to ask whether any of them can compete Brd2 binding from the purified nucleosomes (see competition scheme depicted in Figure 3A). Consistent with the reported binding of Brd2 to AcK12 on H4 [47], [48], [49], [50], [51], and the reported preference of Brd2 for tri- or tetra-acetylated H4 peptides [56], the addition of acetylated H4 peptides (K12, or tetra acetylated at K5, K8, K12, & K16) efficiently competed away binding of Brd2 to H2A.Z nucleosomes (Figure 3B). However, neither Ac-H2A.Z nor Ac-H2A peptides were able to disrupt the interaction between Brd2 and H2A.Z nucleosomes. Therefore, these data suggest that the acetylated H4 on H2A.Z nucleosomes is a critical contact site mediating the interaction between Brd2 and H2A.Z nucleosomes.
In light of this finding, we examined the acetylation levels on H2A.Z and H2A nucleosomes. We immunoprecipited either H2A.Z- or H2A-nucleosomes and we compared the levels of acetylated H4 and H3 residues by Western blot. As shown in Figure 3C, the acetylation level of all residues tested on H4 and H3 was indeed higher on H2A.Z nucleosomes compared to H2A nucleosomes. This difference was apparent under both basal (mock-treated with DMSO), and hyperacetylated (TSA-treated) conditions, suggesting that H2A.Z nucleosomes inherently have higher levels of overall acetylation. As histone acetylation is generally associated with euchromatin (or “open” chromatin) and transcription, this observation is consistent with our previous work that showed that the H3 on H2A.Z nucleosomes, compared to those associated with H2A, have distinct methylation profiles that correspond to an enrichment of euchromatin and depletion of heterochromatin [16]. In summary, our data suggest that Brd2 preferentially binds to H2A.Z nucleosomes, and that this interaction is primarily mediated through the recognition of acetylated H4 residues in the H2A.Z nucleosome. Furthermore, the preference of Brd2 for H2A.Z nucleosomes over H2A nucleosomes is likely due, in part, to the higher levels of acetylated H4 in the H2A.Z nucleosomes, providing increased binding sites for Brd2. On the other hand, since H2A nucleosomes also contain significant levels of H4 acetylation, the much higher preference of Brd2 for H2A.Z-nucleosomes over H2A-nucleosomes suggests that additional regions of contact or stabilization may be present in H2A.Z nucleosomes.
To directly test this possibility, we generated H2A- and H2A.Z-nucleosomes that have equivalent amounts of H4 acetylation and asked whether Brd2 binds differentially to these modified nucleosomes (Figure 4). In brief, mononucleosomes harvested from 293T cells were dialyzed into a high-salt buffer (1.2 M NaCl), causing the H2A-H2B/H2A.Z-H2B dimers to dissociate from the H3–H4 tetramers. Step-wise dialysis of the buffer back down to 140 mM NaCl results in random re-assembly of H2A–H2B and H2A.Z-H2B dimers with H3–H4 tetramers, and thus normalizing the distribution of H3 and H4 PTMs between H2A- and H2A.Z-nucleosomes (see Figure 4A for flow chart). As shown in Figure 4B, in contrast to mononucleosomes not subjected to dialysis (non-scrambled), which show the expected enrichment of H4K12ac and H3K4me3 on H2A.Z nucleosomes, dialyzed (scrambled) nucleosomes have comparable amounts of these two PTMs on the immunoprecipitated H2A- and H2A.Z-nucleosomes (compare lanes 2 & 3 with lanes 5 &6, right panel, Figure 4B). We then incubated the scrambled and non-scrambled nucleosomes with nuclear lysates containing endogenous levels of Brd2, and measured the amount of Brd2 that binds to these nucleosomes. Consistent with our previous experiments, Brd2 shows preferential binding to H2A.Z nucleosomes in the non-scrambled mononucleosome preparations (Figure 4B, lane 3 versus lane 4 in the IP panel). More importantly, in the “scrambled” nucleosome preparation, where H4 acetylation levels are normalized between H2A- and H2A.Z-nucleosomes, Brd2 showed a small yet still distinct preference for H2A.Z nucleosomes. This suggests that Brd2 recognizes additional properties on H2A.Z that augment its interaction with H2A.Z nucleosomes.
Recent studies found that there are in fact two isoforms of H2A.Z (H2A.Z-1 and H2A.Z-2) that differ by 3 amino acids [57], [58], [59], [60]. One of these studies also suggested that, within the nucleosome context, these two isoforms are associated with histones that have slightly different PTM levels [58]. To test whether either or both isoforms of H2A.Z associate with Brd2, we expressed Flag-H2A.Z-1 and Flag-H2A.Z-2 in separate pools of 293T cells and performed the same mononucleosome IP analyses. Interestingly, in spite of the high degree of identity between the two isoforms, significantly more Brd2 co-immunoprecipitated with H2A.Z-1 than H2A.Z-2 nucleosomes under both basal (DMSO-treated) and hyperacetylated (TSA-treated) conditions (Figure 4C). This finding is particularly intriguing given that the H2A.Z-1 and H2A.Z-2 nucleosomes have comparable levels of H4 acetylation (both H4K12Ac and tetra-acetylated H4, Figure 4C). Therefore, it corroborates our earlier conclusion that Brd2's preference for H2A.Z nucleosomes is mediated not only through the hyperacetylated H4, but also through additional factors unique to H2A.Z (more specifically H2A.Z-1). At present, the exact molecular determinant that mediates Brd2's preference for H2A.Z-1 over H2A.Z-2 nucleosomes is not known, but will be the subject of future investigation. Nevertheless, this result raises the possibility that these two isoforms can engage nuclear factors differently, leading to distinct downstream functions.
To test the biological significance of the interaction between H2A.Z nucleosomes and Brd2, we examined their relationship in the context of androgen receptor (AR)- regulated genes since we and others have previously reported that H2A.Z is required for the full activation of AR-regulated genes in the prostate cancer cell line LNCaP [15], [61]. First, we tested and confirmed that the interaction between Brd2 and H2A.Z nucleosomes is conserved in LNCaP cells (Figure S2). Next, we tested whether Brd2 is also recruited to AR-regulated genes in a hormone-dependent manner. We focused on the enhancer and promoter regions of the PSA gene since we previously found H2A.Z enriched at these elements, and we also included a control region that is located between the promoter and enhancer (approximately 2 kilobases upstream of the transcriptional start site). By ChIP analyses, we examined the recruitment/enrichment of AR, H2A.Z, H4Ac and Brd2 to these regulatory regions over a time period of 120 minutes following hormone stimulation (Figure S3 and Figure 5B). In agreement with our previous study, a net loss of H2A.Z occurs at the PSA promoter, but not at the enhancer, following hormone stimulation of the cells. Analysis of AR recruitment during this process reveals a previously well-characterized pattern: Namely, AR is rapidly recruited to the enhancer and promoter, initially peaking at 60 minutes, followed by a second peak at 120 minutes ([62] and Figure S3). Analysis of Brd2 revealed that it is recruited to the hormone-activated PSA gene, with specific enrichment at the enhancer and promoter elements, and its recruitment follows a similar pattern as AR, with two peaks of enrichment at 60 minutes and 120 minutes post-stimulation (Figure 5B). Since Brd2 binds to acetylated H4 (see Figure 2B and [56]), we also examined the dynamics of this histone modification throughout our ChIP time-course experiment. Although H4Ac levels generally increased during gene activation, the most significant increase occurred after the 30-minute time point, suggesting that the gene activation process involves the recruitment of one or more histone acetyltransferases (HATs).
As we discovered that Brd2 is recruited to AR-regulated genes upon hormone activation, we next asked whether this recruitment is dependent on H2A.Z. To that end, we examined Brd2 recruitment by ChIP analysis in the H2A.Z knockdown cells that we have previously generated [15]. Also, we focused our analysis on the enhancer and promoter regions at the 60 minutes post-treatment time point since that corresponds to the first peak of Brd2 recruitment and H4 acetylation. Finally, we examined two AR-regulated genes, PSA and KLK2, to test for consistency. As shown in Figure 6B, compared to the cells expressing a control shRNA, knockdown of H2A.Z reduced recruitment of Brd2 to both the promoter and enhancer regions. Similarly, enrichment of the H4Ac mark was also reduced at both regions in the H2A.Z knockdown cells. The reduction of Brd2 recruitment and H4Ac levels we observe by ChIP analyses is not due to reduced overall levels in H2A.Z knockdown cells, since total levels of Brd2 and H4Ac by Western blot were comparable in control and H2A.Z KD cells (see Figure 6A). Overall, these findings provide biological validation of the biochemical interactions we identified earlier in this study.
The recruitment of Brd2 to AR-regulated genes has not been reported before. Moreover, we discovered that the recruitment is dependent on the presence of H2A.Z. To test the importance of Brd2 in the transcriptional activation of AR-regulated genes, we initially attempted to generate stable Brd2 knockdown LNCaP cells using Brd2-targeting shRNA constructs. However, cells that showed good knockdown of Brd2 also displayed growth defects, which confounded the AR-activation studies. Brd2 is an essential gene and published literature has alluded to the fact that studying Brd2 function by shRNA knockdown is problematic [63]. To address this question using a different approach, we took advantage of the recently developed small molecule inhibitor JQ1, which specifically binds the bromodomains of BET family members and interferes with their binding to acetylated lysines [64]. By pre-treating the hormone-stimulated cells with JQ1, we could disrupt Brd2 binding to chromatin and ask whether this would also affect AR-regulated gene expression. To first confirm that JQ1 perturbs interaction between Brd2 and H2A.Z nucleosomes, we performed H2A.Z mononucleosome IPs from cells treated with JQ1 or vehicle control (DMSO) and compared the levels of Brd2 immunoprecipitated between the two sample sets. As seen in Figure 7A, pre-treatment of cells with JQ1 prior to immunoprecipitation reduces the amount of Brd2 associated with H2A.Z nucleosomes.
Given that JQ1 disrupts the interaction between Brd2 and H2A.Z nucleosomes, we next tested whether JQ1 also blocks recruitment of Brd2 to the AR-regulated genes. Analysis of whole-cell lysates from LNCaP cells showed that JQ1 treatment does not affect total levels of Brd2, AR, or H2A.Z, and only causes a slight reduction in total levels of acetyl H4 (Figure 7B). ChIP assays showed that pre-treatment of DHT-stimulated LNCaP cells with JQ1 significantly reduced binding of Brd2 to both PSA and KLK2 enhancers and promoters at the 60 min peak of Brd2 recruitment (Figure 7C). JQ1 treatment also resulted in a small reduction of H4 acetylation levels at these genes (Figure 7C), but it is unclear whether this is a secondary effect upon reduction of Brd2 recruitment. To test the effects of JQ1 treatment on androgen-stimulated gene expression, we examined the accumulation of PSA and KLK2 mRNA by quantitative RT-PCR. As shown in Figure 7D, pre-treatment of LNCaP cells with JQ1 prior to hormone stimulation greatly reduced expression of both PSA and KLK2 in a dose-dependent manner. Finally, given that AR regulates growth and viability of prostate cells, we also examined the effects of JQ1 on the proliferation of LNCaP cells by MTS assay. As shown in Figure 7E, JQ1 inhibited LNCaP cell proliferation in a dose-dependent manner as well. Taken in sum, these results show that JQ1 has profound effects on the expression of AR-regulated genes and proliferation of LNCaP cells. We note that JQ1 inhibits the binding of all BET family members to acetylated histones and, therefore, the observed effects of JQ1 may not be completely attributable to Brd2 function alone. Indeed, we found that Brd4 is also normally recruited to AR-regulated genes and also shows a preferential interaction with H2A.Z nucleosomes (Figure S4). Nevertheless, these results provide the first evidence that BET family members have critical functions in AR-regulated gene expression and raise the possibility that chemical inhibition of their recruitment to chromatin may have therapeutic potential in blocking prostate cancer cell growth.
In order to gain a deeper understanding of the physiological functions of H2A.Z, we reasoned that proteins acting downstream of H2A.Z must first engage this histone variant in the context of chromatin. Therefore, we specifically chose to perform a proteomics screen to identify proteins that preferentially associate with H2A.Z nucleosomes, as compared to nucleosomes containing the core histone H2A. This is a departure from previously published approaches that focused on identifying and characterizing soluble H2A.Z-interacting proteins, which led to the identification of histone chaperones and remodeling complexes, such as the Swr1 complex, which deposits H2A.Z into chromatin [18], [19], [20], [21]. Our approach is also different from the recent study by Zlatanova and colleagues whereby they reconstituted H2A.Z nucleosomes using recombinant histones that lacked any PTMs, and incubated them with cell lysates to identify proteins that associate with the in vitro reconstituted nucleosomes [65]. The top proteins identified by our screen represent mostly chromosomal proteins and a number of them are either known regulators of transcription, or are predicted to be involved in the regulation of gene expression. Moreover, many of the proteins contain conserved structural motifs that are predicted to recognize and bind histone PTMs, which is consistent with the idea that these proteins are recruited to H2A.Z-containing chromatin to mediate downstream functions. The concept of histone modifications acting to recruit downstream effector proteins/complexes to mediate specific biological outcomes was the basis for the proposed histone code hypothesis [66]. While the use of the word “code” in this context has been a matter of debate, it is nevertheless clear that specific combinations of PTMs do cluster and function together. Moreover, many examples have shown that effector proteins can engage chromatin in a multivalent manner, recognizing multiple PTMs and multiple histone proteins at once [22], [23], [24], [25]. Previous work from our lab has shown that H2A.Z nucleosomes contain a unique set of methylation marks on H3, compared to H2A nucleosomes [16]. In the current study, we have expanded this observation and showed that H2A.Z nucleosomes are also enriched for various acetylation marks on H3 and H4. Furthermore, earlier structural studies of the H2A.Z nucleosome revealed that the presence of the H2A.Z-H2B dimer altered the docking domain with the H3–H4 tetramer, and the H2A.Z-H2B dimer contains an extended acidic patch, which is displayed on the surface of the octamer and could serve as a unique site of interaction with other proteins [67]. Taking all these findings together, we propose that H2A.Z nucleosomes display a combination of unique surfaces, sequences, and PTMs that together recruit and stabilize the binding of downstream effector proteins.
In support of this multivalency model, we found that the binding of Brd2 to H2A.Z-nucleosomes is primarily mediated through the interactions between the bromodomains on Brd2 and the high levels of H4 acetylation on H2A.Z nucleosomes. In addition, there are likely other non-H4-acetylation-dependent contact points. This conclusion is based on two separate lines of evidence: First, using high salt/low salt dialysis to re-assemble nucleosomes that have randomly mixed histone compositions, we found that even though these H2A/H2A.Z nucleosomes now have almost equivalent amounts of H4 acetylation, Brd2 still shows a small but distinct preference for H2A.Z-containing nucleosomes. Second, significantly more Brd2 co-purified with H2A.Z-1 nucleosomes, compared to H2A.Z-2 nucleosomes, even though both types of nucleosomes have similar levels of H4 acetylation. Therefore, these findings strongly suggest that additional features or surfaces on H2A.Z (H2A.Z-1 in particular) provide further interaction sites to stabilize the binding of Brd2 to H2A.Z nucleosomes.
At present, we have not characterized, nor identified, the exact determinants that result in the differential binding of Brd2 to the H2A.Z-1 and H2A.Z-2 nucleosomes. This finding is intriguing given that the two isoforms differ only by 3 amino acids. Nevertheless, it is not without precedence since H3.3 and H3.1 only differ by 4 amino acids and yet they are physically associated with distinct chaperone complexes [68], [69], [70], [71]. Moreover, the few publications specifically examining these isoforms of H2A.Z suggested that there may be subtle differences in the PTMs either on the two isoforms or on the other histones within the nucleosome context [58], [60]. Therefore, these variations could also influence the binding of Brd2. The possible functional differences between these H2A.Z isoforms are unknown and currently under investigation. Characterization studies showed that there is extensive overlap between the isoforms in terms of their nuclear localization, association with the SRCAP chaperone complex, and sites of acetylation. However, since the original lethal phenotypes of the H2A.Z gene knockouts in Drosophila and mice were specific to H2A.Z-1, this suggests that there are non-redundant functional differences between the isoforms. Also, a recent study examining the effects of knocking out H2A.Z-1 or H2A.Z-2 in chicken DT40 cells showed that they have differential effects on the regulation of at least one candidate gene [60]. Therefore, our finding that the two isoforms may recruit different amounts of Brd2 suggests that differential recruitment of downstream effector proteins could contribute to their functional differences. Finally, it has recently been reported that there is also a splice variant of the H2A.Z-2 mRNA in mammalian cells [72]. The H2A.Z protein encoded by this splice variant has a shorter C-terminal tail and forms highly unstable nucleosomes. The biology associated with the H2A.Z variant is clearly complicated and the functional distinctions of these isoforms are currently not well defined. Our approach of purifying nucleosomes of the transfected H2A.Z variants to study the associated histone PTMs within the nucleosome context and also to identify nucleosomal binding partners could easily be applied to these different isoforms. Such future studies could be highly informative of their respective biological functions.
In addition to characterizing Brd2 as an H2A.Z nucleosome-binding protein, we also demonstrated that Brd2 is a novel regulator of androgen responsive genes in LNCaP cells. Our ChIP analyses of the PSA gene in LNCaP cells showed that Brd2 is recruited to the promoter and enhancer regions following stimulation of cells with hormone. More importantly, we found that recruitment of Brd2 is dependent on H2A.Z since knockdown of H2A.Z resulted in reduced levels of Brd2 recruitment, as well as H4 acetylation levels, at the enhancers and promoters of PSA and KLK2 genes. We note that the knockdown construct we used is designed to target H2A.Z-1, and it is unlikely that this shRNA also targets H2A.Z-2 since the DNA sequences of the two genes are quite divergent. Nevertheless, given that Brd2 preferentially associates with H2A.Z-1 nucleosomes, the effects seen with just knocking down H2A.Z-1 is not surprising. Currently, there are no antibodies, nor shRNAs, developed that are specific for the H2A.Z-2 isoform. Once such reagents are available, it would be interesting to test whether H2A.Z-2 has a functional role for the expression of AR-regulated genes.
Prior to this study, the role of Brd2 in AR-regulated gene expression had not been reported. The importance of Brd2, and possibly other BET proteins, in this process is supported by our studies using the small molecule inhibitor JQ1. JQ1 treatment not only disrupted binding of Brd2 to H2A.Z nucleosomes, but also strongly affected the expression of AR-regulated genes and proliferation of LNCaP cells. We note that JQ1 inhibits other BET family members as well [64] and, therefore the JQ1 effects we observed could be due to its inhibition of multiple BET proteins. Indeed, we tested the recruitment of Brd4 to PSA and KLK2, and found a similar trend observed for Brd2 (Figure S4). Furthermore, we also found that Brd4 shows a clear enrichment on H2A.Z nucleosomes compared to H2A nucleosomes. However, since we have not identified Brd4 in our proteomics screens, it is possible that Brd4 is enriched on H2A.Z nucleosomes through indirect interactions. Future studies further clarifying the roles and recruitment of individual BET proteins in AR-regulated gene expression will be highly informative. As JQ1 has such a dramatic effect on AR-regulated transcription, and LNCaP cell proliferation, these findings further raise the potential of therapeutic use of this compound in the treatment of prostate cancer. The use of JQ1 has already shown great promise in the treatment of leukemia and other cancers [73], [74]. Therefore, the usefulness of JQ1 underscores the importance of further understanding how various docking domains of effector proteins engage chromatin and nucleosomes as a mechanism for translating cell signalling pathways to nuclear functions.
Our data all together led us to propose a model whereby, at AR-regulated genes, H2A.Z establishes a unique platform for the recruitment of transcriptional co-activators, such as Brd2 (see Figure 8 for model). The recruitment of Brd2 to H2A.Z nucleosomes depends on the recognition of acetylated lysines; however, a complex multivalent interaction is likely involved in vivo. Indeed, recent studies have reported that bromodomains often exhibit low affinities for their acetylated targets [75], [76]. As suggested by Voigt and Reinberg [77], a chromatin template within the cell nucleus could provide local high concentrations of PTMs. In turn, this could enhance the binding of low affinity interactions, yet would maintain their dynamic nature and therefore allow for rapid response to cellular signals. Therefore, we hypothesize that the unique binding surfaces of H2A.Z nucleosomes act as multivalent platforms for early critical nucleation events, such as the recruitment of transcriptional co-activators. Based on our data, H2A.Z nucleosomes may provide localized sites of H4Ac, for example, for the early recruitment of factors such as Brd2, and this interaction is stabilized by additional elements on the H2A.Z nucleosome. It has previously been reported that Brd2 is associated with acetyltransferase activity towards H4 and H2A [36]. Therefore, additional recruitment of histone acetyltransferases (HATs) would then allow for subsequent increases and spreading of histone acetylation, which would provide a positive feedback loop to promote further recruitment of Brd2 and other factors. Indeed, Brd2 (and Brd3) has been reported to bind hyperacetylated chromatin, facilitating transcription by RNA pol II [37]. This model would be compatible with our observation that there is a loss of H2A.Z at the promoters of AR-regulated genes following hormone stimulation.
In conclusion, our study has provided novel insights into the physiological functions of H2A.Z and its ability to engage chromatin-binding proteins through its influence on PTMs within the nucleosome. Our characterization of the interaction between Brd2 and H2A.Z nucleosomes also furthered our understanding of the role H2A.Z plays in promoting AR-regulated transcription in prostate cancer cells, yielding potential new molecular targets for therapy. The data generated from our MS analysis of proteins binding to H2A.Z nucleosomes will serve as a useful tool in future studies of H2A.Z's role in various chromatin-templated processes.
293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. LNCaP cells were obtained from ATCC and were grown in RPMI 1640 media supplemented with 10% fetal bovine serum. For culturing in the absence of hormone, cells were grown in phenol-red free RPMI 1640 supplemented with 5% charcoal-stripped fetal bovine serum (Invitrogen) for 72 hrs prior to treatment with hormone. Dihydrotestosterone (DHT) was obtained from Sigma and re-suspended in absolute ethanol; DHT was added to cells at a final concentration of 10 nM, or for control samples, an equivalent volume of ethanol was added. For treatment of cells with trichostatin A (TSA), cells were treated with TSA (200 nM) or an equivalent volume of DMSO for 2 hours prior to harvest. The JQ1 reagent was kindly provided by Dr. Jay Bradner, Dana-Farber Cancer Institute. All transfections were carried out using Lipofectamine 2000 (Invitrogen). All expression constructs used were based on the pcDNA 3.1 (+) (Invitrogen) backbone with the Flag tag cloned in-frame. H2A.Z antibody directed against the L1 loop was described previously [16], and the H4 tetra acetyl antibody was a kind gift from the lab of C. David Allis. Other antibodies were obtained as follows: H3 (ab1791) and Brd2 (ab3718) were from Abcam; GFP was from Santa Cruz (sc-8334); H4K12ac, H3K9ac, and H3K14ac were from Upstate; pan-acetyl was from Cell Signaling; AR antibody (PG-21) was from Millipore; and Flag M2 monoclonal antibody was from Sigma.
Generation of mononucleosomes was performed as described previously [16] with slight modifications. 293T cells were grown in 15 cm-diameter plates and transfected with a construct that expresses either Flag-H2A or Flag-H2A.Z. 48 hrs following transfection, cells were trypsinized, counted, and washed in 1× PBS. Cellular pellets were resuspended in Buffer A (20 mM HEPES, pH 7.5, 10 mM KCl, 1.5 mM MgCl2, 0.34 M sucrose, 10% glycerol, 1 mM dithiothreitol, 5 mM sodium butyrate, 10 mM NEM, and protease inhibitors), pelleted and then resuspended in buffer A containing 0.2% Triton X-100 and incubated on ice for 5 min. The nuclear suspension was centrifuged at 1300×g; nuclei were then washed once in Cutting Buffer (10 mM Tris-HCl, pH 7.5, 15 mM NaCl) then resuspended in Cutting Buffer with 2 mM CaCl2. Microccocal nuclease (MNase; Worthington) was added at a concentration of 10 units/1.0×107 cells the incubated at 37°C for exactly 30 min. The reaction was stopped by the addition of EGTA to a final concentration of 10 mM. The MNase-digested nuclei were centrifuged at 1300×g and then subjected to hypotonic lysis by resuspension in TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA). Samples were incubated on ice for 30 min, with occasional mixing by pipette. The suspension was then centrifuged at 16 000×g and the supernatant was transferred to a new tube. Salt was adjusted to 150 mM NaCl by adding 3× Buffer D (60 mM HEPES, pH 7.5, 450 mM NaCl, 4.5 mM MgCl2, 0.6 mM EGTA, 0.6% Triton X-100, 30% glycerol) dropwise, with constant mixing on a vortex set to low speed. Insoluble material was pelleted via centrifugation. The clarified supernatant was then used for immunoprecipitation by adding M2-agarose beads and incubated overnight at 4°C on an end-over-end rotator. Beads were washed 4 times in 1× Buffer D, followed by 3 washes in 1× Buffer D containing 0.5% Triton X-100. Proteins were eluted from the beads by resuspension in 2× SDS sample buffer and boiled for 10 min. For Western blot analysis, samples were run on SDS-polyacrylamide electrophoresis gels according to standard practices. Due to a difference in expression between Flag-H2A and Flag-H2A.Z, immunoprecipitated samples were normalized by total nucleosome content using H3 Western blotting for mass spec analysis. Consequently, an equivalent of approximately 3.0×106 cells were loaded on NuPAGE Novex 4–12% Bis-Tris (1.5-mm thick, 10-well) pre-cast polyacrylamide gels, and separated by molecular mass. Gel lanes were cut into 10 gel blocks of equal size and in-gel digested as previously described [78]. Extracted peptides were C-18 purified using Varian OMIX cartridges (Mississauga, ON, Canada) and analyzed by 1D-LC-MS/MS on a LTQ-Orbitrap XL as previously described [79].
Raw data was converted to m/z XML using ReAdW and searched by X!Tandem against a locally installed version of the human UniProt complete human proteome protein sequence database (release date 2009, 20,323 sequences). The search was performed with a fragment ion mass tolerance of 0.4 Da and a parent ion mass tolerance of ±10 ppm. Complete tryptic digest was assumed. Carbamidomethylation of cysteine was specified as fixed and oxidation of methionine as a variable modification. A target/decoy search was performed to experimentally estimate the false positive rate and only proteins identified with two unique high quality peptide identifications were considered as previously reported [79] (FDR∼0.5%). An in-house protein-grouping algorithm was applied to satisfy the principles of parsimony [80], [81]. A data ranking strategy was applied to select to most promising candidates for biochemical/functional validation. First, semi-quantitative spectral counts were used to remove proteins found in the control (GFP) sample. Triplicate analyses of GFP and H2A.Z were compared, and only proteins with a 10-fold increase in spectral counts in the H2A.Z sample were selected. Next, only proteins detected in at least two out of three MS analyses were considered—72 proteins passed these criteria. This strategy was then repeated on the 72 proteins in comparing H2A.Z versus H2A samples. In comparing H2A.Z and H2A samples, only proteins with a two-fold increase in H2A.Z, compared to H2A, samples, and detected in at least two out of three runs, were considered—21 proteins passed these criteria.
IPs were performed essentially as described above, with the following modification: Peptides were added to the input material giving a final peptide concentration of 30 µg/ml, then incubated at 4°C with rotation for 30 min. M2-agarose beads were then added as described above.
293T cells were transfected with a construct expressing Flag-H2A, Flag-H2A.Z, or GFP-NLS, as described previously. Mononucleosomes were prepared as described above, and dialyzed against 1.2 M NaCl, 10 mM Tris pH 8.0, 0.2 mM EDTA, 1 mM DTT, 5 mM sodium butyrate overnight at 4°C. Nucleosomes were then reconstituted in the same buffer by salt-gradient dialysis at 4°C as follows: 0.9 M NaCl, 2 h; 0.6 M NaCl, 2 h; 0.3 M NaCl, 2 h. The final dialysis was for 3 h against 140 mM NaCl, 20 mM Tris pH 7.6, 5 mM sodium butyrate. Flag immunoprecipitation was carried out as described. Following washes, nucleosomes bound to Flag beads were subsequently incubated overnight at 4°C with salt-extracted nuclear lysates prepared as described in [16]. Following a second round of washes, eluted material was analyzed by Western blot.
Cells stably expressing the H2A.Z (cagctgtccagtgttggtg) shRNA target sequence were generated as described previously [15]. Cells were maintained in media as described above, with the addition of puromycin (0.6 µg/ml).
RT-qPCR or ChIP analysis of LNCaP cells was performed as previously described [15].
LNCaP cells were seeded into 96-well tissue culture plates and grown for 24 hrs as described. Cells were treated with various concentrations of JQ1, or an equivalent volume of DMSO, for 24 hrs, followed by the addition of CellTitre 96 AQueous One Solution reagent (Promega) according to manufacturer's instructions. Absorbance was measured at 490 nM using a Synergy H4 Hybrid Microplate Reader (BioTek).
|
10.1371/journal.ppat.1005787 | MyD88 Shapes Vaccine Immunity by Extrinsically Regulating Survival of CD4+ T Cells during the Contraction Phase | Soaring rates of systemic fungal infections worldwide underscore the need for vaccine prevention. An understanding of the elements that promote vaccine immunity is essential. We previously reported that Th17 cells are required for vaccine immunity to the systemic dimorphic fungi of North America, and that Card9 and MyD88 signaling are required for the development of protective Th17 cells. Herein, we investigated where, when and how MyD88 regulates T cell development. We uncovered a novel mechanism in which MyD88 extrinsically regulates the survival of activated T cells during the contraction phase and in the absence of inflammation, but is dispensable for the expansion and differentiation of the cells. The poor survival of activated T cells in Myd88-/- mice is linked to increased caspase3-mediated apoptosis, but not to Fas- or Bim-dependent apoptotic pathways, nor to reduced expression of the anti-apoptotic molecules Bcl-2 or Bcl-xL. Moreover, TLR3, 7, and/or 9, but not TLR2 or 4, also were required extrinsically for MyD88-dependent Th17 cell responses and vaccine immunity. Similar MyD88 requirements governed the survival of virus primed T cells. Our data identify unappreciated new requirements for eliciting adaptive immunity and have implications for designing vaccines.
| Despite several million new systemic fungal infections annually worldwide, there are no commercial vaccines available. The development of effective vaccines requires a fundamental understanding of how protective immune responses are induced. Using experimental vaccine strains, we previously demonstrated that populations of T helper cells producing interleukin 17 (Th17 cells) and interferon gamma (Th1 cells) mediate vaccine resistance to systemic dimorphic fungi of North America. Here, we report how the immune system recognizes the fungal vaccines and induces the development of protective T cells. We delineate the role of pathogen recognition receptors (PRRs) and their common signaling pathway in host immune cells that recognize the fungal vaccine. While the signaling pathway studied is essential for the development of vaccine-induced T cells, the mechanism of action is novel and included T cell death after activation. The findings could be extended to virus-specific T cells suggesting that the mechanism is conserved among the microbial kingdom. Our work sheds new light on how protective T cells are induced and can be harnessed by vaccine strategies tailored against fungal and other microbial infections.
| The soaring rates of systemic fungal infections worldwide have spurred interest in developing vaccines [1,2,3,4,5,6,7]. We and others have engineered vaccines that protect against experimental infection with the primary pathogenic fungi Coccidioides posadasii [8], Histoplasma capsulatum [9] and Blastomyces dermatitidis [10], which cause the major systemic mycoses of North America and account for an estimated one million new infections annually [11]. CD4+ T cells are the primary effector cells that control fungal infections in healthy hosts [12,13] and Th17 cells are requisite for vaccination against the endemic mycoses of North America [14]. Vaccine induced Th17 cells confer resistance independent of Th1 cells by recruiting and activating neutrophils and macrophages to the alveolar space to reduce the burden of infection.
The development of effective vaccines requires a fundamental understanding of how protective immune responses are induced. We previously reported that the differentiation of Th17 cells and acquisition of vaccine immunity requires innate recognition and signaling through Card9 and MyD88 [14,15]. The innate immune system senses invading microbes through germline-encoded pattern-recognition receptors (PRRs) that bind conserved and invariant structures, termed pathogen-associated molecular patterns (PAMPs) [16]. Fungal PAMPs such as the cell-wall components chitin, α- and β-glucans, and mannans are unique to fungi and distinguish them from the host [17]. The PRRs that are best described for the recognition of fungi include the C-type lectins and Toll-like receptors (TLRs). Vaccination with B. dermatitidis requires Dectin-2 recognition and signaling for the development of Th17 cells [15], whereas the related dimorphic fungi H. capsulatum and C. posadasii require Dectin-1 and Dectin-2 for the induction of protective Th17 cell responses. Most TLRs (except for TLR3) and IL-1R family members trigger pathways via the adaptor protein myeloid differentiation primary-response gene 88 (MyD88) to activate NF-κB and mitogen-activated protein kinases (MAPK) [18,19].
While TLRs and MyD88 have been implicated in the development of Th1 and Th2 cells [20,21,22,23], their role in inducing Th17 cells is unexpected and poorly understood. The regulation of Th1 and Th2 cells by MyD88 is linked to TLR-dependent cytokine production by antigen presenting cells (APCs) that influence T cell differentiation [20,21,22,23].
Both T cell-extrinsic and -intrinsic MyD88 signaling promotes adaptive immune responses. T cell-extrinsic signaling activates dendritic cells (DCs) and macrophages to produce pro-inflammatory cytokines and promote antigen presentation to initiate adaptive immunity during viral, bacterial and parasitic infections [24]. Impaired MyD88 signaling increases susceptibility to fungal infections such as candidiasis, cryptococcosis, aspergillosis, paracoccidioidosis, pneumocystis and coccidioidomycosis [25,26,27]. However, the mechanisms by which MyD88 mediates adaptive immunity are not well understood. In addition to the extrinsic role of MyD88 signaling in immunity to fungal infections, T cell- intrinsic expression of MyD88 is required for resistance to infections with Toxoplasma gondii, LCMV and B. dermatitidis. In experimental T. gondii infection, T cell-intrinsic MyD88 is required for Th1 mediated resistance independent of IL-1R and IL-18R signaling, implying a role for TLRs [28]. During LCMV infection, IFN-γ-producing CD8+ T cells require intrinsic MyD88 signals for differentiation and survival [29]. Finally, CD8 T cell-intrinsic MyD88 signals are required for Tc17 cell responses and immunity to B. dermatitidis infection [30].
In the current study, we uncovered a novel mechanism by which MyD88 enables the development of vaccine-induced anti-fungal Th17 cells and resistance to infection. Instead of regulating the production of priming cytokines by APCs that shape T cell differentiation [20,21,22,23], MyD88 extrinsically regulates the survival of CD4+ T cells during the contraction phase under non-inflammatory conditions. T cell-intrinsic MyD88 signals were largely dispensable for the development of anti-fungal CD4+ T cells. Moreover, TLR3, 7, and 9 served as the extrinsic upstream sensors and signaling receptors that initiate T cell survival signals under non-inflammatory conditions. Similar MyD88 requirements extrinsically governed the survival of virus-primed T cells, implying a general mechanism across microbial kingdoms.
We previously reported that vaccine-induced Th17 and Th1 cells were necessary and sufficient to protect mice against the three major systemic mycoses in North America [12,14]. Myd88-/- mice are highly susceptible to primary infections by B. dermatitidis, P. brasiliensis, A. fumigatus, C. neoformans, and C. albicans [14,31,32,33]. Here, we investigated whether Myd88-/- mice can acquire vaccine immunity against infection with the systemic dimorphic fungi B. dermatitidis and H. capsulatum. Myd88-/- mice were unable to control the live, attenuated #55 vaccine strain of B. dermatitidis and succumbed to dissemination and infiltration of the lungs by the yeast (Fig 1A). To circumvent susceptibility of Myd88-/- mice to the vaccine strain, we immunized them with heat-killed yeast and tested their ability to resist a lethal pulmonary infection with wild-type yeast. Vaccinated Myd88-/- mice failed to acquire resistance and had similar numbers of lung CFU when compared to unvaccinated littermates, which was 4–5 logs higher than in vaccinated wild-type control mice (Fig 1B). Vaccinated Myd88-/- mice had significantly fewer numbers and frequencies of endogenous Th1 and Th17 cells in their lungs on recall than did vaccinated wild-type controls (Fig 1C), which correlated with reduced resistance. To investigate whether these findings apply to other systemic dimorphic fungi, we vaccinated mice with H. capsulatum. Although Myd88-/- mice were able to control vaccination with live H. capsulatum yeast, they were significantly less resistant than vaccinated wild-type controls to lethal pulmonary challenge (Fig 1D). Likewise, vaccinated Myd88-/- mice recruited lower frequencies of endogenous Th17 and Th1 cells to the lung on recall (Fig 1E), which again correlated with reduced resistance. Thus, vaccinated Myd88-/- mice fail to recruit endogenous Th17 and Th1 cells to the lung, which are required to confer vaccine-induced resistance to systemic dimorphic fungi.
T cell-intrinsic MyD88 is necessary for resistance to LCMV and T. gondii infections [28,29,34] and survival of CD8+ T cells [28,29,35]. To investigate whether MyD88 affects the development of vaccine induced CD4+ T cells in a T cell-intrinsic manner, we used two approaches. First, we generated B. dermatitidis-specific TCR Tg (1807) mice that lack MyD88, as described in the methods. We adoptively transferred naïve Tg T-cells from these mice into wild-type recipients before subcutaneous vaccination. Thus, recipient host APCs are wild-type and transferred fungus-specific CD4+ T cells lack MyD88. As a positive control, we also transferred naïve wild-type 1807 into wild-type recipients before vaccination. At serial time intervals post-vaccination, we analyzed the behavior of the two transferred 1807 T cell populations. The number of activated (CD44+), Myd88-/- vs. Myd88+/+ 1807 cells was not reduced at day 7 post-vaccination (burst of T cell expansion) (Fig 2A), at day 35 (contraction of T cells) (Fig 2B) and at day 4 post-infection (recall to the lung) (Fig 2C). Likewise, cytokine production by Myd88-/- vs. Myd88+/+ 1807 cells in the lung after recall was not reduced (Fig 2C). Thus, MyD88 does not intrinsically regulate the development of CD4+ T cells in our vaccine model.
We sought to validate the results observed with TCR Tg cells by also investigating polyclonal, endogenous CD4+ T cells. Here, we vaccinated Myd88∆T mice in which MyD88 expression is absent only in αβT cells [35], and enumerated the number of cytokine producing lung CD4+ T cells upon recall. The numbers of activated (CD44+) and cytokine-producing T cells were similarly increased in vaccinated Myd88∆T and wild-type control mice, as compared to unvaccinated mice (Fig 2D). Vaccinated Myd88∆T mice acquired resistance similar to wild-type controls at day 4 post-infection, although resistance was modestly reduced in the Myd88∆T mice at two weeks post-infection (Fig 2D). Thus, T cell-intrinsic MyD88 may have a small (if any) impact on vaccine resistance mediated by CD4+ T cells in our model, but it does not explain the profound impairment in resistance mediated by the cells in vaccinated MyD88-/- mice.
To investigate whether extrinsic MyD88 is required for the development of Th17 and Th1 cells, we adoptively transferred purified, naïve CD4+ T cells from wild-type 1807 mice into Myd88-/- and Myd88+/+ recipients before subcutaneous vaccination. At serial intervals post-vaccination, we compared development of wild-type 1807 cells in the two recipient hosts. T cell expansion, activation and proliferation of 1807 cells were similar in the recipient hosts, as indicated by the number of CD44+ 1807 T cells and loss of CFSE (Fig 3A and 3B).
To assess T cell differentiation, we used several approaches. Since T cell differentiation is orchestrated by the cytokine milieu produced by APCs, we first determined the n-fold change in cytokine transcripts after co-culture with vaccine yeast. The n-fold increase in the Th17 (IL-6, TGF-β, IL-1β and IL-23p19) and Th1 (IL-12p35) priming cytokines was comparable with bone marrow DCs from Myd88-deficient and -sufficient mice (Fig 3C). Second, we determined the frequencies of Th17 and Th1 1807 T cells at the site of vaccine injection (subcutaneous tissue) at day 11 post-vaccination and upon recall of cells to the lung at day 4 post-infection. The frequencies of cytokine-producing 1807 T cells (and numbers at vaccine injection site) were comparable in both strains of vaccinated recipients (Fig 3D and 3E). Third, we stimulated primed endogenous CD4+ T cells ex vivo with CW/M antigen and measured cytokines in the cell-culture supernatant. The amount of IL-17 and IFN-γ produced by T cells from vaccinated Myd88-/- mice was significantly reduced compared to the cells from wild type mice (Fig 3F).
Since reduced cytokine production by endogenous T cells could be due to either a lack of intrinsic MyD88 or reduced numbers of antigen-experienced T cells after contraction, we enumerated activated (CD44+) T cells in Myd88-/- and wild type mice serially after vaccination. For this, we transferred 1807 cells into both recipients. We found that by 30 days post-vaccination the number of activated 1807 T cells was significantly reduced in the skin draining lymph nodes (sdLN) and in the lung upon recall (Fig 3G and 3H), indicating enhanced contraction when MyD88 is lacking extrinsically (in APCs). These data suggest that MyD88 extrinsically regulates the survival of activated CD4+ T cells, but does not affect their activation, expansion, proliferation or differentiation.
Since MyD88 extrinsically regulates T cell survival and not activation or differentiation (Fig 3G and 3H) we investigated whether it programmed activated T cells during the expansion phase (to make them fit to survive contraction) or exclusively during the contraction phase. To test these two possibilities, we adoptively transferred naïve wild type 1807 cells into Myd88-deficient and–sufficient recipient mice prior to vaccination (Fig 4A). At day 7 post-vaccination, we harvested and sorted CD44+ CD4+ T cells from the sdLN and spleen of these initial recipients, adoptively transferred them into new sets of naïve Myd88-/- and wild type mice and let the activated T cells rest for 4 weeks. To accumulate the transferred T cells for analysis, we challenged the mice and enumerated 1807 T cells recalled to the lung at day 4 post-infection. The number of CD44+ 1807 T cells was high when MyD88 was present during the contraction phase (Fig 4B). These results also indicate that MyD88 promotes T cell survival under homeostatic, resting conditions in the absence of vaccine-induced inflammation. When MyD88 was absent during the contraction phase, the number of activated 1807 cells was as low as when MyD88 was missing throughout the experiment. The absence of MyD88 during the expansion/programming phase did not affect T cell survival. Thus, MyD88 is dispensable during the first seven days post-vaccination and required during the contraction.
We next studied endogenous cells to validate our findings with TCR Tg 1807 T cells. We harvested lymph nodes and spleens from vaccinated wild type Thy1.1+ mice at the peak of expansion and transferred sorted CD44+ CD4+ T cells into naïve wild type and Myd88-/- mice. After four weeks of rest, the transferred T cells were recalled into the lung (with challenge) and enumerated in the lymph nodes and spleen. The number of transferred CD44+ CD4+ T cells was reduced in the lung, lymph nodes and spleen of Myd88-/- recipient vs. wild type recipient mice (Fig 4C). Thus, MyD88 extrinsically regulates the survival of activated CD4+ T cells during the contraction phase.
The net number of T cells during the contraction phase is largely governed by apoptosis of effector cells. Thus, we asked whether adoptively transferred 1807 T cells undergo enhanced apoptosis in Myd88-/- mice. At day 12 and 19 post-vaccination, which is a few days before we noted increased T cell contraction in Myd88-/- mice (Fig 5A), we found increased expression of active caspase 3, but not caspase 8, in activated 1807 T cells (Fig 5B). Because Bcl2 and Bcl-xL promote survival of effector T cells [36], we assessed whether increased contraction of activated CD4+ T cells in Myd88-/- mice is linked with reduced expression of the anti-apoptotic molecules Bcl-2 and Bcl-xL. The expression of both anti-apoptotic markers in transferred 1807 T cells was comparable in vaccinated wild type and Myd88-/- recipients at time points before the T cells underwent enhanced contraction in the latter group (Fig 5C). Thus, reduced 1807 T cell numbers in the absence of extrinsic MyD88 signaling is linked to enhanced apoptosis, but not reduced expression of Bcl-2 and Bcl-xL.
To determine whether MyD88 regulates T cell contraction through the extrinsic vs. intrinsic death pathway, we generated 1807 mice and T cells that lack Fas or Bim, respectively, and adoptively transferred wild type 1807 cells or crossed (pathway-deficient) 1807 cells into Myd88-/- and Myd88+/+ recipient mice (S1A Fig). If MyD88 regulates T cell contraction through Bim, we would expect that the absence of Bim would repair the MyD88 phenotype and the number of Bim-deficient 1807 cells would be similar in the Myd88-/- and Myd88+/+ recipients after contraction. If the number of Bim-deficient 1807 cells were lower in Myd88-deficient mice, this would indicate that MyD88 regulates contraction through a different pathway. Conversely, the same reasoning would hold true for Fas-deficient 1807 cells.
At day 7 post-vaccination, the expansion of all three lines of adoptively transferred 1807 cells was comparable in the sdLN of wild type and Myd88-/- mice (Fig 5D). At day 35 post-vaccination, the contraction of all three lines of 1807 cells was still enhanced in Myd88-deficient vs. MyD88-sufficient recipients, indicating either that MyD88 does not mediate enhanced contraction through the Bim or Fas apoptotic pathways, or alternatively that MyD88 regulated contraction occurs through both pathways, so that when one pathway is eliminated, the other pathway compensates.
The IL-2 cytokine family members IL-7 and IL-15 promote the survival of naïve, activated and memory T cells [37,38] and type I interferons IFN-α/β prevent the death of activated T cells by directly acting on them [39]. To investigate whether MyD88 mediates T cell contraction by regulating the expression of IL-2 family members and type I IFNs, we determined cytokine transcripts in naïve mice. We found no significant reduction in transcripts of IL-7, IL-15, IFN-α, IFN-β (not detected), IFN-α receptor 1 (IFNAR1), IFNAR2, and interferon regulatory factory 8 (IRF8) (S1B and S1C Fig), which impacts type I IFN expression and function, in skin draining lymph node cells and splenocytes of Myd88-/- vs wild type mice. The master regulator of type I IFN expression IRF7 [40,41] was consistently and significantly reduced in naïve Myd88-/- vs. wild type mice (S1C Fig). However, the IRF7 phenotype was not sufficient to explain the T cell survival defect (see below).
Since antigen presentation and T cell stimulation is mostly conducted by professional, myeloid APCs that express CD11c on their surface, we hypothesized that CD11c+ cells are required to mediate MyD88-dependent T cell contraction. To test our hypothesis, we primed naïve 1807 T cells in vaccinated wild type recipient mice and adoptively transferred them into naïve CD11cCre+-Myd88fl/fl mice that lack MyD88 expression in CD11c+ cells and CD11cCre—Myd88fl/fl controls. After four weeks of rest, we found no difference in 1807 T cell numbers from the two groups of recipient mice (S1F Fig) indicating that myeloid CD11c+ cells are dispensable for MyD88-dependent T cell contraction.
MyD88 is an adaptor molecule that serves signaling of most TLRs and members of the IL-1R and IL-18R family. To investigate receptors upstream of MyD88 that extrinsically regulate contraction of T cells, we primed naïve 1807 cells in vaccinated wild type recipient mice with three systemic dimorphic fungi in parallel (B. dermatitidis, H. capsulatum and C. posadasii), purified CD4+ T cells from primary recipients and adoptively transferred them into mice that lack IL-1R; TLR2,3,4,7,9; TLR3,7,9; or TLR2,4 (as shown for MyD88 in Fig 4A). After four weeks of rest in all three fungal priming models, activated (CD44+) 1807 T cells were sharply diminished in TLR2,3,4,7,9-/- and TLR3,7,9-/- mice, similar to vaccinated Myd88-/- mice (Fig 6A and 6B). In IL-1R-/- and TLR2,4-/- mice the number of primed 1807 cells was reduced modestly in some cases, but it was not found to be statistically significant.
To investigate whether increased T cell contraction of effector T cells will yield reduced T cell numbers upon recall, we adoptively transferred naïve 1807 T cells into Myd88-/-, IL-1R-/-, TLR2,3,4,7,9-/-, TLR3,7,9-/-, TLR2,4-/- and wild type mice prior to vaccination. After recall, the number of primed 1807 T cells in the lung was reduced in vaccinated TLR2,3,4,7,9-/- and TLR3,7,9-/- vs. wild type mice (S2A and S2B Fig). In vaccinated IL-1R-/-, the recruitment to the lung and differentiation of activated 1807 T cells was also reduced compared to vaccinated wild type mice as indicated by the number of CD44+ 1807 cells (S2A and S2B Fig) and lower frequency of IL-17 producing 1807 cells (S2C Fig). As a consequence, vaccinated IL-1R-/- mice failed to recruit Th17 cells to the lung upon recall (S2A and S2B Fig). Similar numbers of activated and cytokine producing 1807 T cells migrated to the lungs of vaccinated TLR2,4-/- vs. wild type controls upon recall.
To validate the above results with Tg 1807 T cells, we analyzed endogenous, Ag-specific T cells and measured contraction using a recently generated tetramer that recognizes Calnexin (Cnx)-specific T cells in mice exposed to B. dermatitidis or related ascomycetes [42]. We vaccinated mice with heat-inactivated vaccine yeast and assessed the expansion and contraction of endogenous Cnx-specific CD4+ T cells. Tetramer+ CD44+ CD4+ T cells from vaccinated wild type and all knockout groups of mice expanded similarly, but contracted in an enhanced fashion in Myd88-/-, TLR3,7,9-/- and IL-1R-/- mice (Fig 6C). Thus, the contraction of endogenous Cnx-specific T cells in these strains of mice mirrors the defect observed with activated 1807 T cells.
To determine whether the enhanced contraction of vaccine-induced T cells in these strains of mice impacts resistance, we assessed the burden of lung infection two weeks after challenge with B. dermatitidis and C. posadasii. Vaccinated TLR2, 3, 4, 7, 9-/-, TLR3, 7, 9-/- and IL-1R-/- mice had increased lung CFU compared to vaccinated wild type controls and TLR2,4-/- mice (Fig 6D and S2D Fig). Thus, the number of vaccine-induced Ag-specific T cells that survive contraction forecasts resistance to fungal challenge on recall.
Since IRF7 expression is consistently reduced in Myd88-/- vs. wild type mice (S1C Fig) and TLR3, 7, 9-/- and TLR2, 3, 4, 7, 9-/- but not TLR2, 4-/- mice showed enhanced T cell contraction (Fig 6A and 6B), we sought to investigate whether IRF7 transcripts are consistent with T cell survival in these strains of mice. If IRF7 regulates type I interferon expression and T cell survival downstream, then we would expect that IRF7 expression is similarly reduced in naïve TLR3, 7, 9-/- and TLR2, 3, 4, 7, 9-/- but not in TLR2, 4-/- mice. In whole lymph node and spleen homogenates, IRF7 expression was reduced in naïve Myd88-/-, TLR3, 7, 9-/-, TLR2, 3, 4, 7, 9-/-, TLR2, 4-/- vs. wild type mice (S1D Fig). The fact, that IRF7 expression is reduced in TLR2, 4-/- vs. wild type mice is not compatible with the lack of enhanced T cell contraction in TLR2, 4-/- mice. Since plasmacytoid DC (pDC) are the professional type I interferon producing DC subset, we sought to investigate whether IRF7 expression in pDC is compatible with the T cell survival phenotype in the knockout vs. wild type mice. Thus, we negatively enriched pDC from the spleen and analyzed IRF7 expression by real time RT-PCR. IRF7 expression in pDC was reduced in Myd88-/- vs. wild type mice but not in TLR3, 7, 9-/- and TLR2, 4-/- mice (S1E Fig). The lack of reduced IRF7 expression in TLR3, 7, 9-/- vs. wild type mice is not compatible with enhanced T cell contraction in these mice. In sum, we were able to exclude IRF7 and type I IFNs as the pro-survival signal for activated T cells.
To investigate whether our findings are specific to vaccination with fungi or apply to other classes of microbes, we studied the contraction of LCMV primed CD4+ T cells after infection. We elicited LCMV primed T cells in congenic wild type mice, purified CD4+ T cells on day 8 post-infection at the peak of T cell expansion, and adoptively transferred the cells into naïve Myd88-/-, IL-1R-/-, TLR2, 3, 4, 7, 9-/-, TLR3, 7, 9-/- and wild type mice. After four weeks of rest, we enumerated tetramer-positive, gp66-specific T cells in the spleen. The number of transferred tetramer-positive CD4+ T cells was significantly reduced in Myd88-/-, IL-1R-/-, TLR2, 3, 4, 7, 9-/- and TLR3, 7, 9-/- mice vs. wild type controls (Fig 7A and 7B). These results are similar to those in the fungal vaccine model and suggest a global role of extrinsic MyD88 and upstream TLR379 in promoting the survival of CD4+ T cells upon activation.
T cell memory is the cornerstone of vaccination and vaccine-induced immunity to anti-microbial infection including fungi. Thus, understanding the mechanisms that govern long-term survival of activated T cells is germane for the rational design of vaccines. We previously reported that MyD88 is required for acquisition of anti-fungal vaccine immunity and development of protective Th17 cells [14]. Herein, we dissected where, when and how MyD88 regulates the generation of vaccine-induced anti-fungal Th17 cells.
Here, we describe a novel mechanism by which MyD88 extrinsically regulates the generation of vaccine induced T cell immunity and resistance. Our current understanding of how MyD88 affects adaptive T cell responses to fungi and other microbes is as follows. Extrinsic MyD88 signaling within APCs is thought to regulate the production of priming cytokines for Th1 and Th2 cells [20,21,22,23], whereas MyD88 within T cells can intrinsically regulate their survival and differentiation [29]. Our results demonstrate that MyD88 and TLR 3, 7, and 9 signaling extrinsically regulate the survival of activated T cells during the contraction phase, but not the expansion or priming phase. Moreover, extrinsic MyD88 was unexpectedly dispensable during the expansion and differentiation of vaccine-induced anti-fungal CD4+ T cells in our model and myeloid CD11c+ cells were dispensable for the survival of activated T cells during the contraction phase implying that the stroma might be the source of MyD88 dependent survival.
T cell intrinsic expression of MyD88 was largely dispensable for the generation of vaccine-induced Th17 and Th1 cells as assessed by the development of adoptively transferred Myd88-/- and Myd88+/+ 1807 T cells in wild type recipients, and of endogenous CD4+ T cells in Myd88∆T mice. These results contrast with the intrinsic requirement of MyD88 during the priming of B. dermatitidis vaccine-induced Tc17 (CD8+ T cells producing IL-17) cells [30]. In the absence of CD4+ T cells, intrinsic MyD88 signals were indispensable for Tc17 cell responses, whereas Tc1 cells were less affected in the absence of these signals. MyD88 also has a T cell-intrinsic role for responses during LCMV, Vaccinia and Toxoplasma infections [29,43,44]. In contrast to the above findings, we found here that a loss of T cell-extrinsic MyD88 signals, rather than intrinsic signals, was largely responsible for the lack of protective CD4+ Th17 and Th1 cells during pulmonary recall and resistance to fungal challenge.
By using adoptive transfer of wild type 1807 T cells into Myd88-/- and Myd88+/+ mice, we pinpointed the exact stage at which MyD88 regulates the development of Ag-specific T cells following vaccination. Extrinsic MyD88 regulated the contraction and was required for the survival of activated T cells during that phase, whereas it was dispensable for expansion, proliferation and differentiation of vaccine induced Th17 and Th1 cells. This conclusion is supported by the fact that absence of MyD88 in adoptive transfer recipients during the initial expansion and programming phase enabled the priming and differentiation of 1807 T cells, whereas absence during the maintenance phase led to accelerated contraction of 1807 cells. These results imply that extrinsic MyD88 provided a survival signal during non-inflammatory, resting conditions after adoptive transfer of effector cells into naïve recipient mice. The fact, that MyD88 regulates the survival of activated T cells under homeostatic conditions, in the absence of inflammation sets our findings apart from previously published studies in which LPS induced inflammation mediated by TLR4 signaling also promoted T cell survival [45]. LPS-TLR4-MyD88 dependent pro-memory signals were also reported for CD8+ T cells [46] that have different requirements than CD4+ T cells for T cell activation, proliferation, and differentiation and do not require MyD88 for survival in our vaccine system [30].
The mechanisms that guide transition from effector cells to memory cells are not well characterized. During the priming phase, T cell memory frequencies can be determined by the clonal burst size [47] and inflammatory signals [48,49,50] or duration of antigen availability [51,52]. During the contraction and maintenance phases, members of the IL-2 family and type I interferons IFN-α/β prevent the death of activated T cells [37,38,39,53,54]. IL-2 and its relatives induce Bcl-2 synthesis and proliferation in responsive T cells [37,38,53,54]. IFN-α/β do not increase Bcl-2 levels in T cells to stimulate T cell division [39]. IFN-α/β act directly on the T cells and not indirectly through the induction of other molecules like IL-15 and Bcl-X. We did not find evidence of differential expression of the IL-2 family members IL-7 and IL-15 and type I IFN-α/β IFNAR1, IFNAR2, IRF8 and IRF7 by naïve splenocytes and lymph node cells from Myd88-/-, TLR3, 7, 9-/- and wild type mice and the anti-apoptotic factors Bcl-2 and Bcl-xL by wild type 1807 T cells transferred into Myd88-/- and Myd88+/+ recipients over the course of the priming and contraction phase. Since enhanced T cell contraction does not occur when MyD88 is absent in CD11c+ cells, our data are compatible with the idea that non-myeloid stromal cells could play a role in MyD88 mediated survival of activated T cells.
Since activated T cells underwent enhanced contraction in the absence of extrinsic MyD88, we investigated whether T cells showed enhanced apoptosis. We found that 1807 T cells from vaccinated Myd88-/- vs. wild type recipients showed increased expression of active caspase 3, but not caspase 8. Since caspase 3 is an effector caspase that funnels signals from the extrinsic and intrinsic apoptosis pathways [55], we deleted the signature molecules Fas (CD95; extrinsic) or Bim (intrinsic) from these pathways in adoptively transferred 1807 cells by crossing Bim-/- and Fas-/- mice with 1807 mice. Knocking out Fas or Bim in 1807 cells did not repair the exaggerated contraction in vaccinated Myd88-/- mice, indicating that MyD88 either does not mediate enhanced contraction through these two apoptotic pathways, or that the pathways are redundant.
By using adoptive transfer of 1807 cells and enumerating endogenous calnexin-specific T cells in knockout recipients, we identified the receptors upstream of extrinsic MyD88 mediated T cell survival. In the absence of TLRs 3, 7 and 9, activated T cells underwent enhanced contraction similar to that in Myd88-/- mice. The engagement of TLR3, 7 and 9 distinguishes our findings from a previous report in which LPS induced TLR4 signaling promoted T cell survival under inflammatory conditions [45]. Impaired T cell survival also correlated with the loss of vaccine-induced resistance in the corresponding knockout mice, indicating that our mechanistic analysis is functionally relevant. While others have reported that TLR signaling on T cells or myeloid cells can promote the survival of the respective cells, we are unaware of stromal cell-derived, TLR-mediated survival factors that rescue activated T cells from death. For example, T cell intrinsic MyD88 signaling has been shown to promote T cell survival in LCMV and T. gondii infection models [28,29,56]. TLRs on T cells can promote survival of activated T cells and thereby directly modulate adaptive immune responses without APC. Treatment of purified, activated CD4+ T cells with the dsRNA synthetic analog poly(I:C) and CpG DNA, respective ligands for TLR3 and TLR9, directly enhanced T cell survival without augmenting proliferation. Enhanced survival was associated with Bcl-xL up-regulation [57]. Porin of Shigella dysenteriae type 1 upregulates TLR2 to activate mitogen-activated protein kinase (MAPK) and NF-κB to induce T cell expansion by promoting both proliferation and survival of CD4+ T cells. The proliferation, survival and effector function of CD4+ T cells through TLR2 co-stimulation show the ability of porin to directly call T cells into action [58].
Collectively, we show that MyD88 signals extrinsically regulate anti-fungal vaccine immune responses by preventing apoptosis of activated CD4+ T cells, and that these signals likely originate with upstream TLR 3, 7 and 9, which likewise preserve the memory pool of T cells by stemming exaggerated contraction. Our findings may guide the rational design of vaccines against fungal and other microbial infections.
Inbred wild type C57BL/6, IL-1R1-/- B6.129S7-Il1r1tm1Imx/J mice (stock # 003245) [59] and congenic B6. PL-Thy1a/Cy (stock #00406) mice carrying the Thy 1.1 allele were obtained from Jackson Laboratories, Bar Harbor, ME. Blastomyces-specific TCR Tg 1807 mice were generated in our lab and were backcrossed to congenic Thy1.1+ mice as described elsewhere [60]. Myd88-/- x 1807 mice that carry the congenic marker Thy1.1 were generated by crossing Thy 1.1+ 1807 mice with Myd88-/- mice twice. TLR24-/-, TLR379-/- and TLR23479-/- [61] mice were a generous gift from Dr. Carsten Kirschning from the University of Duisburg-Essen in Germany and were bred at our facility. Myd88-/- [62] and Myd88∆T mice in which MyD88 expression is selectively deleted in all αβT cells [35] were a generous gift from Drs. Doug Golenbock and Laurence Turka. All mice were 7–8 weeks at the time of experiments. All mice above were housed and cared for according to guidelines of the University of Wisconsin Animal Care Committee, who approved all aspects of this work. CD11cCre+Myd88fl/fl [63] and CD11cCre-Myd88fl/fl were housed in specific-pathogen free conditions at the University of California at San Francisco (UCSF).
To enrich epitope-specific T cells in mice we generated an anti-Cnx-specific MHC class II tetramer [42] and used a magnetic bead-based procedure that results in about a 100-fold increase in the frequency of the target population [64,65,66]. Enriched cells were stained with a cocktail of fluorochrome-labeled antibodies specific for B220, CD11b, CD11c, F4/80, CD3, CD8, CD4 and CD44. The entire stained sample was collected on an LSRII flow cytometer and live cells analyzed by FlowJo software (Treestar) following the gating strategy described [66]. The total number of tetramer positive cells from a mouse was calculated from the percent of tetramer-positive events multiplied by the total number of cells in the enriched fraction as described [66] and in the enriched plus unbound fraction when larger numbers of tetramer positive cells are present.
Mice were vaccinated as described [10] twice, two weeks apart, subcutaneously (s.c.) with 106 to 107 live or heat killed BAD1 null B. dermatitidis yeast strain #55 [67], 107 live H. capsulatum strain G217B or 5 x 104 live attenuated vaccine strain (∆T) that lacks the chitinases 2,3 and D-arabinotol-2-dehydrogenase as described previously [8]. Mice were infected intratracheally (i.t.) with 2 x 103 or 2 x 104 wild-type yeast of B. dermatitidis strain 26199, 2 x 105 H. capsulatum G217B, 2 x 105 FKS or 100 spores of the virulent C. posadasii isolate C735 [10,14]. To assess the infiltration of primed CD4 T cells into the lungs, challenged mice were analyzed at day 4 post-infection. To analyze the extent of lung infection, homogenized lungs were plated and yeast colony forming units (CFU) enumerated on BHI agar (Difco, Detroit, MI), sheep-blood containing Mycosel plates, or GYE plates containing 50 μg/ml of chloramphenicol [68].
To assess the T helper cytokine phenotype of Calnexin-specific CD4+ T cells after vaccination we purified and transferred 106 CD4+ T cells from naïve 1807 Tg cells into C57BL/6 wild-type mice before vaccination. On the same day, recipients were vaccinated, and challenged four weeks post-vaccination.
Congenic Thy1.1+ wild type mice were infected with 105 pfu of LCMV Armstrong. 8 days later, CD4+ T cells were purified from the spleen and adoptively transferred into naïve Myd88-/-, IL-1R-/-, TLR23479-/-, TLR379-/- and wild type Thy1.2+ mice. After 4 weeks of rest the number of GP66-specific CD4+ T cells were assessed by tetramer that were generously provided by the NIH tetramer core facility.
Mice were vaccinated once s.c. with 107 heat-killed strain #55 yeast at a single dorsal site. On day 10 post-vaccination inflamed s.c. tissue was excised from the site of vaccination, placed in ice-cold collagenase buffer and minced into fine pieces. To digest the tissue, 5 ml of dissociation buffer (0.025 mg/ml Liberase (Roche Diagnostics) and 50 μg/ml DNAse I (Sigma-Aldrich) in collagenase buffer) was added and samples were incubated at 37°C and 5% CO2 for 30 minutes. To further release single cells, the tissue was mashed with the back of a 10-ml syringe plunger through a 70-μm cell strainer with an additional 5 ml of dissociation buffer and incubated for another 30 minutes. The dissolved tissue was washed with ice-cold PBS containing 5 mM EDTA and 1% BSA and strained again. Filtered cells were spun at 1500 rpm, the supernatant was carefully aspirated and cells were resuspended in complete media for stimulation.
Lung cells were harvested at day 4 post-infection. Cells (0.5 x 106 cells/ml) were stimulated for 4 to 6 hours with anti-CD3 (clone 145-2C11; 0.1μg/ml) and anti-CD28 (clone 37.51; 1μg/ml) in the presence of Golgi-Stop (BD Biosciences). Stimulation with fungal ligands yielded comparable cytokine production by transgenic T-cells compared to CD3/CD28 stimulation as shown previously [42,60]. After cells were washed and stained for surface CD4 and CD8 using anti-CD4 PerCP, anti-CD8 PeCy7, and anti-CD44-FITC mAbs (Pharmingen), they were fixed and permeabilized in Cytofix/Cytoperm at 4°C overnight. Permeabilized cells were stained with anti-IL-17A PE and anti-IFN-γ Alexa 700 (clone XMG1.2) conjugated mAbs (Pharmingen) in FACS buffer for 30 min at 4°C, washed, and analyzed by FACS. Cells were gated on CD4 and cytokine expression in each gate analyzed. The number of cytokine positive CD4+ T cells per organ was calculated by multiplying the percent of cytokine-producing cells by the number of CD4+ cells. Intracellular Caspase 3 Alexa647 (9602S, Cell Signaling), Caspase 8 PE (12602S Cell Signaling), Bcl2 Alexa 647 (633510 Biolegend) and Bcl-xL PE (13835 Cell Signaling) were stained from ex vivo derived lymph node cells and splenocytes.
Cell-culture supernatants were generated in 24-well plates in 1 ml containing 5 x 106 splenocytes and lymph node cells and 5 μg/ml of Blastomyces CW/M antigen [10]. Supernatant was collected after 72 hours of co-culture. IFN-γ and IL-17A were measured by ELISA as above.
Differences in the number and percentage of activated, proliferating or cytokine-producing T cells were analyzed using the Wilcoxon rank test for nonparametric data or the t-test (using GraphPad Prism) when data were normally distributed [69]. A P value < 0.05 is considered statistically significant.
The studies performed were governed by protocols M00969 and AN101733 as approved by the IACUC committees of the University of Wisconsin-Madison Medical School and University of California at San Francisco (UCSF), respectively. Animal studies were compliant with all applicable provisions established by the Animal Welfare Act and the Public Health Services (PHS) Policy on the Humane Care and Use of Laboratory Animals.
|
10.1371/journal.ppat.0030171 | Release of Hepatic Plasmodium yoelii Merozoites into the Pulmonary Microvasculature | Plasmodium undergoes one round of multiplication in the liver prior to invading erythrocytes and initiating the symptomatic blood phase of the malaria infection. Productive hepatocyte infection by sporozoites leads to the generation of thousands of merozoites capable of erythrocyte invasion. Merozoites are released from infected hepatocytes as merosomes, packets of hundreds of parasites surrounded by host cell membrane. Intravital microscopy of green fluorescent protein–expressing P. yoelii parasites showed that the majority of merosomes exit the liver intact, adapt a relatively uniform size of 12–18 μm, and contain 100–200 merozoites. Merosomes survived the subsequent passage through the right heart undamaged and accumulated in the lungs. Merosomes were absent from blood harvested from the left ventricle and from tail vein blood, indicating that the lungs effectively cleared the blood from all large parasite aggregates. Accordingly, merosomes were not detectable in major organs such as brain, kidney, and spleen. The failure of annexin V to label merosomes collected from hepatic effluent indicates that phosphatidylserine is not exposed on the surface of the merosome membrane suggesting the infected hepatocyte did not undergo apoptosis prior to merosome release. Merosomal merozoites continued to express green fluorescent protein and did not incorporate propidium iodide or YO-PRO-1 indicating parasite viability and an intact merosome membrane. Evidence of merosomal merozoite infectivity was provided by hepatic effluent containing merosomes being significantly more infective than blood with an identical low-level parasitemia. Ex vivo analysis showed that merosomes eventually disintegrate inside pulmonary capillaries, thus liberating merozoites into the bloodstream. We conclude that merosome packaging protects hepatic merozoites from phagocytic attack by sinusoidal Kupffer cells, and that release into the lung microvasculature enhances the chance of successful erythrocyte invasion. We believe this previously unknown part of the plasmodial life cycle ensures an effective transition from the liver to the blood phase of the malaria infection.
| The malaria parasite Plasmodium undergoes one large round of multiplication in the liver before beginning the blood phase of the life cycle, the phase that causes the typical episodes of fever and chills. Using intravital microscopy and fluorescent parasites, we studied the mode and dynamics of parasite release from the liver, a critical stage in the malaria life cycle. Earlier work had indicated that infected liver cells could release packets of dozens to hundreds of parasites enveloped by host cell membrane, structures now known as merosomes. We report here that this is the predominant mechanism of parasite release from the liver. The host-derived merosome membrane lacks a marker for phagocytic engulfment, thus allowing safe passage through the gauntlet of Kupffer cells, highly active liver macrophages. Merosomes remain intact during passage through the heart and become sequestered within lung capillaries where the membrane eventually disintegrates liberating the parasites into the lung circulation. We propose that this previously unknown part of the life cycle of Plasmodium facilitates red blood cell invasion, thus jump-starting the blood phase of the life cycle and the onset of clinical malaria.
| Two billion people, more than one third of the world's population, live at risk for malaria and about 1 billion are infected. Each year there are 300 million to 500 million new cases with 2–3 million deaths, the vast majority young children in Africa. We are now forty years past the discovery that radiation-attenuated sporozoites protect against malaria [1], but we still lack an efficient malaria vaccine to combat this deadly parasitic disease, and drug resistance is wide-spread [2].
The malaria infection begins with the introduction of sporozoites from the bite of an infected Anopheles mosquito [3,4]. The sporozoites travel to the liver and develop in hepatocytes to large exoerythrocytic forms (EEFs) [5,6]. Schizogonic division of the EEF then results in the formation of thousands of first-generation merozoites, which are responsible for the initiation of clinical malaria. Merozoites have a short life span and must infect erythrocytes immediately after release into the bloodstream [7]. Merozoites are also highly susceptible to phagocytosis and must therefore avoid contact with macrophages [8]. Acute danger of phagocytic elimination is presented in the form of Kupffer cells [8], the resident phagocytes of the liver that comprise by far the largest population of tissue macrophages of the body [9]. Kupffer cells are predominantly located at sinusoidal bifurcations, largely within and often spanning the sinusoidal lumen [9–11], thereby presenting significant obstacles for non-self particulate material. This strategic position of Kupffer cells makes it difficult for free merozoites to exit the liver without being trapped by these surveillance cells of the innate immune system.
The first evidence suggesting that merozoites can be released from hepatocytes as clusters, held together by host cell cytoplasm, was presented several decades ago in Garnham's ultrastructural examination of Plasmodium yoelii–infected murine livers and described in more detail in Meis' extensive electron microscopic studies on P. berghei infection of the mouse [5,12,13]. More recently, we and others reported that merozoites are released as “extrusomes” or “merosomes” that contain hundreds to thousands of parasites [14,15] (reviewed in [16,17]). Our initial intravital observations using green fluorescent P. yoelii and BALB/c mice revealed extensive movement within EEFs nearing completion of merozoite maturation culminating in budding and release of merosomes into the hepatic bloodstream [14]. An elegant series of in vitro studies described the differentiation of P. berghei merozoites in the human hepatoma cell line HepG2 [15]. While developing into hepatic schizonts, the intracellular parasites prevent the initiation of a death program in their host cells, but leave them to die once merozoite formation is complete. Underlying molecular details remain to be determined, but the data suggest that host cell death in this in vitro model shares more features with autophagy than apoptosis or necrosis [18]. However, information on the viability of hepatocytes releasing merozoites into the sinusoidal blood is lacking to date.
Because P. yoelii infection of the mouse represents an accepted model closely reflecting human malaria [19], we used a variety of microscopic techniques to study the dynamics of merosome budding from infected hepatocytes and the fate of hepatic merozoites in the body. Confocal images provided measurements of merosome volume, merosomal merozoite content, and EEF volume, and appropriate mathematic processing of these data allowed us to calculate the number of hepatic merozoites produced by P. yoelii sporozoites in the murine host. Using intravital and ex vivo microscopy, we found that the vast majority of hepatic P. yoelii merozoites leave the liver camouflaged as merosomes, disseminate within the cardiovascular system, and arrest in the lungs. Molecular markers revealed that merosomal merozoites remain viable and infectious until being released into the pulmonary microcirculation. In contrast, various in vivo and ex vivo assays suggest that unreleased merozoites and the exhausted host cell eventually succumb to necrosis. The resulting inflammatory stimulus attracts neutrophils, and mononuclear phagocytes thus give rise to the formation of microgranulomata. Overall, this systematic temporal and quantitative analysis indicates that merosome formation and release by host hepatocytes, merosome transport to and sequestration in the lungs, and release of merozoites into the pulmonary microvasculature are parts of a previously unrecognized phase of the Plasmodium life cycle.
P. yoelii–infected mice have been suggested to represent a suitable model for human malaria [20]. We also consider P. yoelii an appropriate rodent model for liver stage analysis because it induces less inflammation in murine livers than P. berghei and produces more EEFs [21], which in addition are generally larger and contain more merozoites [12,22,23] (Table S1). While available for other species such as P. berghei, information is scarce regarding ultrastructural changes during P. yoelii EEF maturation in the liver and the subsequent release of first generation merozoites [12,24]. To help fill this gap and to expand our previous investigation of Plasmodium merosomes in live mice [14], we used several light and electron microscopy techniques to examine this process. Mature Plasmodium EEFs contained thousands of merozoites enclosed in a parasitophorous vacuole (PV). Up to the final developmental stage and onset of merozoite release, infected hepatocytes remained in close contact with neighboring uninfected parenchymal (Figure 1A and 1B) and sinusoidal cells (Figure 1C). Shortly before merosome formation, the PV membrane (PVM) disintegrated so that host cytoplasm contained a mixture of mature merozoites, morphologically intact hepatocyte organelles (Figure 1D), parasite remnant bodies (or pseudocytomeres [5]), and parasite stroma left over from schizogonic merozoite formation (Figure 1E). Some of the sinusoids adjacent to infected hepatocytes remained filled with erythrocytes indicating preservation of function, but others were compressed by the expanding parasite and lacked erythrocytes suggesting local obstruction of blood flow (Figure 1C). To calculate the merozoite content of mature EEFs (see below), we needed accurate measurements of the EEF size. Compared to tissue sections, intravital microscopy of green fluorescent protein (GFP) Plasmodium yoelii parasite (PyGFP)–infected mouse livers (Figure 1F) offered the advantage of examining live tissue within an intact animal, thus avoiding artifacts associated with both fresh and fixed sections. Mature EEFs within the liver typically have a slightly ellipsoid shape with the minimum and maximum diameters ranging from 40 to 75 μm (with averages of 49.2 ± 10.3 μm to 55.6 ± 9.0 μm), respectively (n = 16).
Detailed intravital examination of 30 mice at times ranging from 30 to 74 h after intravenous infection with PyGFP sporozoites allowed us to follow the complex series of events involved in merozoite liberation from hepatocytes. We monitored more than 60 EEFs over this period and observed the earliest merosome budding at 46 h (Figure 2A, Videos S1–S3), a time in general agreement with earlier work reporting the first appearance of P. yoelii in the blood at 45.5 h [12]. Of these 60 EEFs, 20 reached maturity during the observation period and released merozoites, while the rest remained immature. The majority (13) of these 20 EEFs released merozoites by merosome formation. Merosome formation continued until 56 h after infection, thus confirming the asynchronous nature of P. yoelii EEF maturation, a common observation in Plasmodium-infected livers [5,25]. Because we infected by intravenous sporozoite injection, the well-known slow release of sporozoites from the mosquito bite site [3,26] alone cannot account for the asynchronicity observed here. For individual EEFs, the process of merosome budding and release lasted several hours during which time the host cell gradually decreased in size and separated from neighboring cells (Figure 2B). In addition to fully formed green fluorescent merozoites, released merosomes contained non-fluorescent remnant bodies and host cell organelles, thus providing further evidence that merosome budding occurs after rupture of the PVM. Eventually, the host cell membrane appeared to lose its integrity and allowed some leftover merozoites to enter the bloodstream singly and without protection by a merosomal membrane (Figure 2C and 2D, Video S4). GFP radiated out from the disintegrating EEF into the surrounding tissue, implying that parasite antigens and host cell cytoplasm were set free as well. Indeed, electron microscopic examination showed free mitochondria in the sinusoidal lumen (Figure 2E). Size and shape of these organelles revealed hepatocyte origin. Eventually, inflammatory cells were attracted to the site of the disintegrating EEF. During phagocytic removal of debris from dead merozoites and host cells, neutrophil granulocytes and mononuclear phagocytes transformed the site of the former EEF into a small granuloma (Figure S1), a structure commonly reported at late stages of Plasmodium liver infection [5,8,27–31]. Thus, merosome formation in the liver occurs over a period of about 10 h and is followed by disintegration of the host cell and some leftover parasites, clearance of the remains by infiltrating phagocytes, and production of a small granuloma.
Merozoites were also liberated by a less frequent mechanism. Starting earlier than merosome formation (42 h post inoculation), some infected hepatocytes rapidly discharged their content of merozoites and cell organelles by a mechanism appearing to involve rupture of the cell membrane (Figure 3A–3E, Video S5). In some cases, the process was complete in as little as 5 min; in others it lasted as long as 60 min. Of the EEFs rupturing in this manner, 80% harbored mature merozoites, but 20% had a homogeneous cytoplasm; thus, schizogony had not even begun (Figure S2). Occasionally, electron micrographs showed immature merozoites incompletely separated from remnant bodies yet released into the sinusoidal bloodstream (Figure 3F). This apparent rupture-release left large faintly fluorescent EEF ghosts at the site of the former host cell. Because our intravital observations were based on confocal microscopy, we considered the possibility of phototoxicity playing a role in this rupture-release mechanism. However, since EEF ghosts identical to those resulting from observed rupture were detectable at the very beginning of intravital examination, we could reject that possibility. Because the EEFs did not decrease in volume prior to transformation into a ghost, and we did not find erythrocytes associated with these ghosts, we suspect that the remains of the host cell cytoskeleton, the surrounding extracellular matrix, and/or the sinusoidal cell layer resealed the ghost after merozoite release; thus, preventing the formation of hemorrhages. Similar to the end of the merosome release mechanism (see above), EEF ghosts were infiltrated by inflammatory cells that gave rise to small granulomata. When we combine results from intravital microscopy, showing that both mature and immature EEFs undergo this rapid decay, with our electron microscopy data, showing that some of the rupturing EEFs were immature, we conclude that this rapid release process is a result of abortive EEF development that, in the absence of host cell membrane protection, exposes the parasites to Kupffer cell phagocytosis.
To demonstrate that merozoites within merosomes are alive and to help exclude the possibility that merosome release represents an abnormal development, we injected infected mice with markers that reveal cell viability in vivo. At points ranging from 51 to 74 h post inoculation, mice were injected with a mix of the membrane-permeable DNA stain Hoechst 33342 and the dead cell marker propidium iodide (PI). Subsequent intravital confocal microscopy revealed that PI does not enter merosomes or intact EEFs (Figure 4A and 4B), but does stain some of the merozoites left behind in EEF ghosts and also in EEFs that had disintegrated after merosome budding (Figure 4C and 4D). These findings support the interpretations above in that they suggest that merozoites that fail to exit the host cell eventually succumb to necrosis.
Efforts to determine the mode by which merosomes breach the sinusoidal cell layer failed so far due to insufficient numbers of suitable events for analysis. We suspect that budding occurs through the endothelial fenestration rather by a paracellular route, because of the extreme natural variability of the diameter of the fenestrae in response to changes in blood pressure and other physiologic stimuli. Interestingly, mature EEFs were frequently surrounded by a layer of flattened cells that had incorporated the dead cell stain PI (Figure 4A and 4B). Perhaps the death of these cells is due to extreme compression by the extensive expansion of the EEF during the final stage of development. Occasionally, merosomes were found budding into such dead cells, but the immobility of the parasites indicated that they were trapped (Figure 4A).
When hepatic merosomes initially bud from infected hepatocytes (Figure 5), they are highly variable in size and contain hundreds to thousands of mature merozoites, while merosomes in blood draining from the liver were smaller and more uniform in size. Intravital microscopy showed very large merosomes moving far more slowly than small ones, which leave the liver lobules at a velocity close to that of blood cells (unpublished data). We frequently observed merosomes hindering the free flow of the blood as they moved along a sinusoid (Figure 5C–5G) as well as being hindered by the vascular architecture. The speed of merosome transport at any instant depended on the diameter and local structure of the sinusoid as well as the size of the merosome. We recorded large merosomes being arrested at sinusoidal bifurcations where they occasionally even reversed direction of movement (Figure 5A and 5B and Videos S6 and S7). Because morphological measurements taken in vitro are subject to artifact and do not reveal in vivo dynamics, we sought a better understanding of sinusoidal architecture using intravital analysis of uninfected transgenic Tie2-GFP mice that have fluorescent vascular endothelia [32]. We found sinusoidal diameters to range from 3.4 μm to 14.1 μm (6.7 ± 1.9 μm; n = 94) under normal blood pressure conditions. Although large merosomes greatly exceed this size range, their considerable deformability allowed them to gradually wind their way towards the central vein and exit the liver without rupture and release of merozoites, a process aided by resizing (Figure 5A and 5B). We occasionally observed large merosomes subdividing into smaller ones while traveling through sinusoids (Videos S8 and S9), but we suspect that shear forces associated with the faster blood velocity in larger vessels caused merosomes in the hepatic effluent and inside lung capillaries to be generally smaller and uniform in size compared to those in the liver. The importance of mechanical forces for resizing is demonstrated by another set of experiments in which PyGFP-infected mouse livers were removed from the animals and analyzed ex vivo by confocal microscopy, i.e., in the absence of blood flow. The sinusoids of such livers contained merosomes of an unusually large size (Figure S3A). When livers were perfused with medium prior to ex vivo confocal microscopy, the sinusoids contained even larger merosomes (Figure S3B). We contend that lack of blood flow prevents subdivision of large merosomes into smaller ones and that liver perfusion hastened merosome budding and liberation from the host cell.
Merosome formation results in packaging a mixture of parasites, remnant bodies, and host cell cytoplasm within host cell membrane for release into the sinusoidal lumen (Figures 5H and 6A). Ultrastructurally, the merosomal matrix contained well-preserved merozoites and morphologically intact host cell mitochondria (Figure 6B) suggesting that these organelles are viable at the time of merosome budding. Merosomes also typically contained remnant bodies (Figure 6A) suggesting that these leftovers from EEF schizogony represent a natural component of the merosomal cytoplasm. In the absence of better viability markers, we interpret the presence of MSP-1 on the surface of merozoites in both mature EEFs and merosomes (Figure 6D and 6E) to indicate intactness and complete differentiation of the parasites, and propose that merosomes are linked to productive infection of erythrocytes. Disintegration of the PVM prior to merosome formation indicates the merosome membrane is derived from hepatocyte cell membrane. Asialoglycoprotein receptor 1 (ASGR1), a protein expressed only on parenchymal liver cells [33–37], was detectable by immunofluorescence lining the basal hepatocyte surface within the space of Disse (Figure 6C). ASGR1 clearly surrounded mature EEFs (Figure 6D), but it was absent from the merosomal membrane (Figure 6E). The lack of this hepatocyte surface protein could be due to dedifferentiation of the infected host cell or modification by the intracellular parasite at late stages of EEF maturation. However, because the ASGR1 label was located predominantly in the space of Disse rather than on the hepatocyte membrane, more work is needed to define the composition of the merosome membrane.
Because intravital observations showed that merosomes remain intact during transport towards the central vein, we examined the hepatic venous effluent for membrane-enveloped parasites. To do this, we opened the inferior Vena cava at its point of entry into the diaphragm and collected blood from the peritoneal cavity. Thick smears were prepared from 5 μl blood, and the concentrations of venous merosomes from three separate experiments were measured. In hepatic venous blood collected 52 h post infection with 2.5 × 106 sporozoites, we found 28.7 ± 4.3 merosomes per μl, and 69% of these merosomes contained between 100 and 200 merozoites (unpublished data).
While much information is available on P. berghei–infected HepG2 cells [15], in vivo data on the molecular composition of the merosome membrane, for example phosphatidylserine (PS) exposure, are lacking to date. To obtain more detail on merosome structure, another set of experiments was performed in which the parasite material available for examination was enhanced by liver perfusion. Beginning 52 h after infection with PyGFP or wild-type (wt) P. yoelii, livers were perfused with culture medium and the perfusate collected. Cells were immobilized by attachment to Alcian blue–treated glass-bottom dishes and immediately examined by confocal microscopy using conditions that maintain viability. Perfusate merosomes typically adapted a spherical shape in vitro (Figure 7A) and 3-D images from confocal stacks demonstrated a relatively uniform size containing several hundred merozoites. Labeling by the phospholipid marker FM 4–64 FX verified that the parasites were held together by a membrane (Figure 7B).
Immediately after harvesting, the majority of the parasites appeared viable and merosome membranes were negative for annexin V labeling (Figure 7C); thus they do not display PS that targets cells for phagocytosis. However, with increased time in vitro, the presence of PS gradually became apparent (Figure 7D). Merozoites in freshly isolated merosomes did not stain with the dead cell marker PI, but those that became positive for PS also lost the ability to exclude PI (Figure 7D). A further viability assessment utilized YO-PRO-1, a DNA stain that selectively passes through the (intact) plasma membrane of apoptotic cells. Again, merozoites in freshly isolated merosomes did not label with YO-PRO-1, but as time in vitro progressed, they began to incorporate YO-PRO-1 along with PI (Figure 7E and 7F). Within roughly 60 min of in vitro examination, all merosomes were positive for annexin V, YO-PRO-1, and PI. Attempts to quantify the time course of these processes more precisely were prevented by the sensitivity of the merosomes to the various steps of isolation from the liver and concentration by centrifugation. Taken together, these results suggest that P. yoelii merosomes leaving the liver contain viable merozoites, and, similar to P. berghei–infected HepG2 cells [15], lack PS as a membrane marker that signals “eat-me” to phagocytes. Considering that Kupffer cells are located largely within and often spanning the sinusoidal lumen, thus presenting a significant obstacle for non-self particulate material and damaged host cells [9–11], the lack of PS on the merosome membrane is likely critical for merozoite escape from this defense mechanism of the host.
Because the entire hepatic effluent must pass through the right ventricle and the pulmonary microcirculation before reaching any other capillary bed, we suspected merosomes might sequester in the lungs. To address this, we used ex vivo confocal microscopy to examine the alveolar microvasculature immediately after lung removal while the tissues were intact and the cells alive. At time points from 46 to 58 h after inoculation with PyGFP sporozoites, we found numerous intact merosomes as well as individual parasites (Figure 8A–8C). We did not find pulmonary merosomes earlier than 46 h post inoculation nor later than 65 h, timing consistent with our observation that merosome release begins and ends at roughly 46 h and 56 h. MSP-1 labeling confirmed the maturity of merozoites, both those within merosomes and those already released into pulmonary capillaries (Figure 8D–8G and Video S10). The small liver stage protein UIS-4, which localizes to the PVM [38], was not detected (unpublished data); thus providing more evidence that the PVM is not involved in merosome formation [5]. As for merosomes in the liver, pulmonary merosomes were negative by immunofluorescence for the hepatocyte receptor ASGR1 (Figure 8F and 8G).
In confocal images we often observed an asymmetric arrangement of individual merozoites in relation to lung merosomes and the pattern suggested some of the merosomes were in the process of disintegrating and releasing merozoites into the pulmonary microvasculature just as blood circulation was stopped by lung removal (Figure 8B). Electron microscopy supports the notion of merozoite release by merosomal membrane degradation. Pulmonary merosomes typically contained morphologically well-preserved merozoites, but the cytoplasmic matrix was swollen, and host cell organelles were clearly degenerating. The membrane of lung merosomes was frequently disrupted or barely detectable (Figure 8H) suggesting that free merozoites found in nearby pulmonary microvasculature had just been released before fixation (Figure 8I). The presence of erythrocytes containing newly invaded merozoites (Figure 8J) supports the notion that blood infection occurred in the lungs.
To determine whether merosomes leaving the liver can pass through the lungs and disseminate throughout the body, we analyzed thick smears of blood collected from the aorta and tail vein for merosomes. We also used intravital microscopy, ex vivo imaging, and immunofluorescence microscopy to examine capillary beds of spleen, kidney, and brain of the same mice. While individual small parasites were occasionally detectable, merosomes were completely absent from aorta and tail vein blood and the microcirculation of these organs (unpublished data). These results demonstrate effective retention of hepatic merosomes in the lungs.
At 52 h after infection with PyGFP, mice were injected with Hoechst 33342 and PI, and the lungs were removed and analyzed ex vivo. Confocal microscopy revealed that pulmonary merosomes and free merozoites excluded PI (Figure 9) and were also TUNEL-negative (unpublished data); thus providing evidence of viability. Because infectivity is the ultimate criterion for viability, we tested merosomal merozoites for their ability to induce a parasitemia in naïve mice. However, interpretation of results from inoculation with blood containing merosomes is complicated by the presence of infected erythrocytes. To circumvent this, we initially attempted to eliminate parasitized erythrocytes using selective hypotonic lysis, but this also affected the integrity of the merosomes. Our solution was to control for infected erythrocytes by comparing the infectivity of two types of blood taken from the same mouse: hepatic effluent (with merosomes and some infected erythrocytes) and tail vein blood (without merosomes, but with the same number of infected erythrocytes). At 52 h after intravenous infection with 2.5 × 106 wt P. yoelii sporozoites, hepatic effluent and tail vein blood samples were collected for inoculation into other mice. Parasitemia and merosome concentration were determined by analysis of thin and thick blood smears, respectively. Preliminary studies showed that the parasitemia in recipient mice injected with hepatic effluent blood rose significantly faster compared to control mice injected with tail vein blood (unpublished data) suggesting that merosomes exiting the liver are infectious. This conclusion can be confirmed once a method for merosome purification is available.
Because of the large number of parasites and their high packing density, counting merozoites in EEFs is not feasible, although estimates have been published [6,22,23]. We therefore used measurements obtained from merosomes to calculate the number of merozoites contained in mature P. yoelii EEFs. To do this, we first quantified the merozoite content of a subset of smaller merosomes. We found that merosomes with a diameter of 13.4 ± 2.0 μm contained 134.7 ± 51.6 merozoites. Then, we determined the average effective volume merozoites take up inside merosomes. P. berghei merozoites measure 1.0–1.2 × 1.5–1.7 μm [39–41], but because the parasites are embedded in cytoplasm that also contains parasite remnant bodies and host cell organelles, the effective volume the parasites occupy is larger than their actual volume of 0.78–1.23 μm3 (based on the ellipsoid volume v = 4/3 π r1 r2 r3). Using a mathematical algorithm for optimal packing of small spheres in a large sphere [42], we found that P. yoelii merozoites have an effective diameter of roughly 2.2 μm and occupy an effective volume of 5.56 μm3 in merosomes. Because intravital and electron microscopy showed that the merozoite packing density and the composition of the cytoplasm was basically identical in merosomes compared to mature EEFs (after rupture of the PVM and mixing of parasites and host cell organelles), we then used the sphere packing algorithm to determine the merozoite content of mature P. yoelii EEFs. Based on the measured EEF diameter of 40–75 μm (see above), the effective merozoite diameter of 2.2 μm, and assuming a round EEF shape, we calculated that P. yoelii sporozoites produce 4,200–29,000 hepatic merozoites (Table S1).
We present here a new model for the transition from the liver to the blood phase of the malaria life cycle (Figure 10): large merosomes of various sizes bud from infected hepatocytes, enter the hepatic circulation, exit the liver intact, subdivide into smaller more uniform sizes, but otherwise withstand bloodstream shear forces during passage through the right ventricle, and accumulate in the lungs where the merosomes disintegrate and release merozoites to initiate the erythrocytic phase of the malaria cycle. While EEF of avian and reptilian malaria parasites develop in the reticulo-endothelial or hematopoietic systems [43–45], a major evolutionary change occurred with the mammalian malaria parasites, whose EEF mature in hepatocytes. Perhaps the nutritionally rich and immunologically privileged hepatic environment offers advantages, but it also presents a problem for merozoites released from EEFs into hepatic sinusoids: unless they invade an erythrocyte very quickly they face a gauntlet of highly phagocytic Kupffer cells. The location of most EEFs in the periportal area of the liver lobule [46] means they must travel almost the full length of the sinusoid and pass by a large complement of Kupffer cells before escaping into relative safety outside the liver. As proposed previously by us and others [14,15], our premise is that evolution produced a countermeasure to this threat: release of merozoites within large packets that are initially hidden from the host's innate immune system by envelopment with a hepatocyte-derived membrane. Here we show that merosomes are delivered to the pulmonary microcirculation where they are released. We propose that release of merozoites into the lung microvasculature rather than into larger blood vessels is advantageous, because the low macrophage density and the reduced blood velocity with reduced shear forces will enhance the ability of merozoites to invade erythrocytes.
Merosome disintegration in the lungs appears to be the predominant mechanism of merozoite liberation into the bloodstream for the following reasons: (1) In confirmation of previous reports on the asynchronous nature of EEF maturation [5,25], we observed P. yoelii merosome formation in the liver from 46 h to 56 h after sporozoite infection. Assuming a 10-h window of merosome release, roughly 3 ml total blood volume in a 40 g mouse, and a 100% rate of sporozoite infection and EEF development, 2.5 million sporozoites would generate 4,167 maturing EEFs per minute, corresponding to 1.4 merosome-releasing EEFs per μl blood. (2) Assuming that extrahepatic merosomes contain on average 150 merozoites, the roughly 29 merosomes we found per μl venous liver blood should have contained 4,350 merozoites. Since P. yoelii EEFs contain 4,200–29,000 merozoites (Table S1), up to 74% of the total number of merozoites released by 1.4 EEFs per min and μl would have been enclosed in merosomes. (3) A large number of merosomes was arrested in alveolar capillaries suggesting that many merosomes withstand the shear forces inside the central cardiovascular system. Together, these data indicate that a major proportion of the merosome population arrives intact in the lungs and then gradually disintegrates, thus liberating merozoites into the microvasculature. Pulmonary merosomes were detectable in the lungs at least up to 58 h after infection, i.e., beyond the period of release from the liver (46–56 h), suggesting that they remained intact for at least many minutes. Similar to hepatic merosomes, which appeared to be infectious and did not stain with annexin V, YO-PRO-1, or PI, pulmonary merozoites were ultrastructurally well preserved, TUNEL-negative, and did not incorporate PI. Together, these data suggest that merosomal merozoites remain viable until their release into the pulmonary microvasculature. Based on the above assumptions, we propose that merozoite liberation in the lungs represents an integral part of the Plasmodium life cycle.
Further support for our premise was found in the following observations and suggestions derived from them. The notion that merozoites shuttled out of the liver within merosomes that are protected from phagocytosis by Kupffer cells [8] was confirmed by demonstrating that murine Kupffer cells do not phagocytoze PyGFP merosomes in vitro (unpublished data), in agreement with the finding that P. berghei merosomes are not ingested by a murine macrophage cell line in vitro [15]. Trager and Jensen's finding that P. falciparum merozoite invasion is enhanced by lack of flow and dense erythrocyte packing [47,48] supports our hypothesis that merozoites released within capillary beds have a better chance to invade erythrocytes than those released into larger vessels. We can imagine that capillary occlusion by arrested merosomes could be helpful by causing local stagnation of the pulmonary blood flow. We can also speculate that merosome arrest in lung septal capillaries allows Plasmodium to exploit the unique microenvironment of the blood-air barrier. Virtually nothing is known about the biology of the first-generation (hepatic) merozoites, but perhaps transient residence in the lungs provides these parasites with time and a suitable microenvironment to gain infectivity for erythrocytes. The well-oxygenated milieu of the terminal airways and the anastomozed nature of the pulmonary microvasculature [49] likely allow local occlusion of septal capillaries by merosomes without causing the necrotic tissue damage associated with infarction of microvessels in other organs.
Many aspects of the process of merosome formation and release we describe are in agreement with earlier work, but others are not. For example, we found that similar to P. berghei–infected HepG2 cells, which detach in toto from the culture vessel after merozoite differentiation is complete [15], merosomes exiting P. yoelii–infected mouse livers contain viable merozoites and initially do not expose PS on their surface. This confirms earlier predictions [14,15] that merozoites are safely shuttled out of the liver disguised as merosomes. The presence of intact mitochondria in mature EEFs indicates that Plasmodium liver stages are able to manipulate hepatocytes in a way that useful organelles (such as mitochondria as a source of energy) are preserved, even after merosome budding. Our interpretation, namely that Plasmodium controls certain host cell functions to the last minute, differs from the P. berghei HepG2 cell model, in which the parasites induce death and detachment of their host cells followed by merosome budding [15]. Further, the cell membrane of P. yoelii–infected hepatocytes remains in close apposition to that of neighboring parenchymal and endothelial cells until the very end of EEF differentiation, i.e., up to the onset of merosome budding, as reported [5,12,13,50,51]. As merosomes are produced, the host cell gradually decreases in size and loses contact with neighboring cells as reported [15]. We observed that after releasing merosomes over several hours, the exhausted host cell eventually disintegrates. Some free merozoites still escaped and entered the sinusoidal lumen, thus being exposed to attack by Kupffer cells. In contrast, others proposed that the remaining host cell remnant is rapidly expelled in toto from the tissue with the resulting void immediately filled by neighboring cells [15,18]. We find that the necrotic remnant attracts neutrophils and mononuclear phagocytes, which eventually produce a small granuloma. Such granulomata are a frequent observation in P. yoelii– and particularly in P. berghei–infected mouse livers [5,8,27–31]. Rather than the void created by expulsion of an EEF being filled quickly, our in vivo observations suggest that hours, if not days, are required for phagocytic removal of parasite and host cell debris with subsequent repair of the structural damage before normal tissue architecture is restored.
Although we found merosome formation to be the predominant mode of merozoite release from the liver, we observed a less frequent but still common alternative: EEFs undergoing what we interpret as decay. This alternative process of EEF ghost formation was rapid and typically complete within minutes to an hour. In contrast to merozoite release by merosome formation, ghost-forming EEFs did not detach from the surrounding tissue. EEF decay was accompanied by leakage of GFP into the surrounding tissue suggesting damage to the host cell membrane. It occurred in immature EEFs (recognizable by a homogeneous green fluorescent cytoplasm) and also in mature EEFs (containing fully formed merozoites) without merosome formation regardless of maturity. Sometimes it was found as early as 42 h after sporozoite infection, hours before merozoite differentiation begins. The end result of this alternative process was the formation of large faintly fluorescent EEF ghosts containing some cellular debris and a few dead merozoites. We interpret this rapid conversion of EEFs to ghosts as abortive liver stage development.
Merozoite content of EEFs has historically been difficult to estimate due to the large number of parasites and their high packing density. Based on measurements of the size and merozoite content of small merosomes combined with size measurements of EEFs and an appropriate mathematical algorithm [42], we were able to calculate the number of merozoites in an EEF (Table S2). Under intravital imaging conditions, mature P. yoelii EEFs measured 40–75 μm and the calculated space effectively occupied by a merozoite is a sphere of 2.2 μm diameter. Using this effective size, we calculated that individual P. yoelii sporozoites produce roughly 4,200–29,000 merozoites per EEF. This number is in general agreement with older estimates of EEF merozoite content [5,12,22,23,52–57] (Table S1). An exception is P. falciparum, which produces considerably larger numbers of hepatic merozoites, most likely because of the small size of the parasites. As far as we know, our analysis of the number of merozoites produced in hepatocytes is the first such analysis based on actual merozoite counts and host cell measurements. Precision is limited by variations in measurements, but basing calculations on direct in vivo measurements enhances accuracy.
Earlier studies conducted by us and others had suggested that merosome budding may precede completion of merozoite differentiation [14,15]. One factor that helped lead to this interpretation is that GFP expressed in the parasite stroma can obscure the parasites in mature EEFs. We now show that prior to merosome formation, the signal of the stromal GFP fluorescence equaled that of the merozoite cytoplasm, thus preventing clear definition of parasites enmeshed in the stroma. At the onset of merosome budding, the stromal GFP emission signal decreased abruptly thus revealing the presence of the already formed fluorescent parasites (Figure 3A–3E and Video 5). Two factors contribute to this reduction in fluorescence of material surrounding the parasites: dilution and loss of cytosolic GFP. Dilution of GFP results from PV disassembly and mixing of fluorescent parasite stroma with non-fluorescent host cytoplasm. Loss of GFP is caused by leakage of the fluorochrome into the environment. In agreement with reports that the hepatocyte membrane becomes permeable at late stages of infection with P. berghei [5], we found that merosome-forming EEFs are typically surrounded by a halo of green fluorescence. Optimization of the imaging conditions allowed us to visualize the parasites inside mature EEFs and revealed that merosomes always contain mature merozoites. Thus, merozoites maturation precedes merosome formation.
Depending on the approach used for measurement, the reported diameters of hepatic and pulmonary capillaries vary greatly. For example, when measured in perfusion-fixed liver tissue, the sinusoidal diameter ranged from 4–6 μm to 9–12 μm [58–60]. A crucially important factor is the pressure applied during perfusion fixation, because the sinusoidal diameter is known to vary with changes in blood pressure [61,62]. To determine the sinusoidal diameter under normal blood pressure conditions, we used live Tie2-GFP mice [32], whose fluorescent endothelia clearly delineate the boundaries of the sinusoidal lumen [31]. In agreement with earlier in vivo microscopic studies, which reported a diameter of 6 μm for portal sinusoids and 7 μm for central sinusoids [58], we found by intravital imaging that liver sinusoids measure 6.7 ± 1.9 μm in diameter. Similar differences between fixed and live specimens were reported for the size of alveolar capillaries. While vascular casts of the lung suggested that alveolar capillaries measure 6.69 ± 1.39 μm in diameter [63], intravital measurements determined a functional diameter of only 1–4 μm [64,65]. Regardless which liver sinusoid and lung capillary measurements are relied upon and regardless of the drastic reduction in merosome size after leaving the liver, merosomes still exceed the size of the lumen of the microvasculature of both liver and lung. Since even the largest merosomes were eventually transported out of the liver, the much smaller extrahepatic merosomes would be expected to be malleable enough to be able to pass though the pulmonary capillary bed. Therefore it is somewhat surprising that the lungs effectively clear the blood of all merosomes, so virtually none were detectable in arterial blood harvested from the left ventricle, in the capillary beds of spleen, brain and kidney, or in tail vein blood. The fact that the velocity in pulmonary capillaries is somewhat higher than hepatic sinusoids [66–69] makes this more unexpected. Consequently, the possibility of a receptor-mediated mechanism for pulmonary merosome arrest cannot be excluded.
Anopheles stephensi mosquitoes were used to propagate wild-type P. yoelii (strain 17 XNL) or PyGFP [14,70]. Sporozoites were purified from the salivary glands of female A. stephensi mosquitoes [71].
Mice were (1) Balb/c (Taconic Farms, Incorporated), (2) Swiss Webster (Taconic Farms, Incorporated), or (3) Tie2-GFP mice, a transgenic strain that expresses GFP in vascular endothelial cells under control of the Tie2 promoter (STOCK Tg(TIE2GFP)287Sato/J; Jackson Laboratory) [31,32]. Animals were maintained and used in accordance with recommendations in the guide for the Care and Use of Laboratory Animals.
Mice were inoculated into the tail vein with 0.3–1.5 × 106 PyGFP sporozoites. At 30–66 h p.i., the animals were surgically prepared for intravital imaging of liver and spleen as described [31] and anesthetized by intraperitoneal injection of a cocktail of 50 mg/kg ketamine (Ketaset, Fort Dodge Animal Health), 10 mg/kg xylazine (Rompun, Bayer), and 1.7 mg/kg acepromazine (Boehringer Ingelheim Vetmedica). Reinjection of the anesthetics at 30-min intervals allowed intravital microscopic examination of the animals for at least 3 h [31].
After surgical preparation for intravital imaging, mice were placed onto the stage of an inverted Zeiss DMIRE2 microscope, equipped with a temperature-controlled Ludin chamber, and analyzed with a Leica TCS SP2 AOBS confocal microscope. Appropriate laser lines were used to excite GFP, various other fluorochromes, and the natural autofluorescence of the mouse tissues. Laser power was reduced to a minimum to avoid phototoxicity and bleaching. These optimized conditions allowed continuous scanning of live PyGFP for a period of up to 6 h without any apparent effect on viability. To assess parasite and host cell viability, some mice were i.v. injected with 1–2 μg/ml of the membrane-permeable nuclear dye Hoechst 33342 prior to confocal microscopy. Other mice received 1 μg/ml PI in addition to detect dead host cells and/or parasites.
Mice were intravenously inoculated with 3 ×106 purified wt P. yoelii or 1 × 106 PyGFP salivary gland sporozoites and various organs were removed at 52 h after infection. Tissue slices were snap-frozen in liquid nitrogen or fixed with PBS containing 4% paraformaldehyde for immunofluorescence labeling of cryosections and with PBS containing 4% paraformaldehyde and 1% glutaraldehyde for electron microscopic examination [72,73].
At 30–66 h after infection with PyGFP, major organs such as spleen, brain, kidney, or lung were removed, placed into glass-bottom dishes, and kept moist with medium for confocal microscopy analysis.
Blood was harvested from (1) the terminal hepatic vein, (2) the aorta, or (3) a tail vein. To increase the probability of detection, ten aliquots of 5 μl blood from each of these sites were spread over an area of 1 cm2, allowed to dry, and stained with Giemsa without prior fixation. Merosomes were counted and expressed as average number ± STD. In parallel, the number of merozoites per merosome was determined accordingly.
Two days after infection with 1.5 × 106 wt P. yoelii sporozoites, hepatic effluent and tail vein blood was harvested from the same animal and parasitemia and merosome content were determined using thin and thick blood smears, respectively. 20-μl hepatic effluent, containing 1 × 105 infected erythrocytes plus 167 merosomes, or tail vein blood containing the same number of infected erythrocytes but no merosomes, was intravenously inoculated into Swiss Webster mice (three mice per group) and the parasitemia was monitored daily by Giemsa staining.
To improve the recovery of parasite material from the liver, merosomes were dislodged from hepatic sinusoids by perfusing mouse livers via the portal vein with oxygenated medium at 5 ml/min for 10–30 min. The effluent was collected in two fractions: fraction 1 was collected from the Vena cava inferior and contained mainly red blood cells; fraction 2 was collected from the Vena cava superior after ligation of the Vena cava inferior. The cells were washed and allowed to settle onto cover slips or glass-bottom dishes (WillCo Wells) treated with Alcian blue [74] for live cell imaging. Nuclei of merosomes were visualized with the membrane permeable nucleic acid stains Hoechst 33342 (1–2 μg/ml) or SYTO-64. Nuclei of dead parasites were determined with membrane impermeable PI (1 μg/ml). Merosome membranes were stained with 5 μg/ml FM 4–64 FX (Molecular Probes). Annexin V Alexa Fluor 488 conjugate or YO-PRO-1 (0.1 μM) were used to detect evidence of programmed cell death in live merosomes. Tissue sections were stained with a BrdU TUNEL assay kit (Molecular Probes) according to manufacturer's guidelines.
Alcian blue–immobilized PyGFP merosomes were fixed and labeled with the red nuclear dye SYTO-64. 3-D stacks were scanned by confocal microscopy and the number of merozoite nuclei was counted using a 3-D object count plug-in of ImageJ (NIH freeware). Merozoite number and merosome diameter were then entered into a formula for efficient packing of equal small spheres in a large sphere ( n = 0.7405 [1–2D] / D3 + 1 / [2D2]; D = dmerozoite / dliver stage ) [42] to determine the effective diameter/volume merozoites occupy inside merosomes. Based on these calculations and the diameter of PyGFP liver stages measured by intravital microscopy, the merozoite content of P. yoelii liver stages was estimated in relation to size.
Frozen sections of 10-μm thickness were prepared with a Reichert-Jung Frigocut cryostat. Parasites were labeled with a mAb directed against the P. yoelii merozoite surface protein MSP-1, a kind gift from W. Bergman [75]. A rabbit antiserum, which was originally generated against the PVM-associated protein from P. berghei, but exhibits cross-reactivity with P. yoelii UIS4 [38,76], was used to label the PV in P. yoelii–infected hepatocytes. Affinity-purified goat IgG against the murine asialoglycoprotein receptor ASGR1 was from R&D Systems. Incubation with the primary antibodies was followed with protein A conjugated to fluorescein isothiocyanate (PA-FITC; Molecular Probes), anti-goat IgG conjugated to Texas Red (GAR-TR; Molecular Probes), or goat anti-rabbit IgG conjugated to Texas Red (GAM-TX; Molecular Probes) in color-matching fluorochrome combinations. In case of a single FITC label, the specimens were counterstained with 0.1% Evans blue in PBS. Immunofluorescence-labeled frozen tissue sections were examined by confocal microscopy.
Mouse liver or lung tissue was fixed with 1% glutaraldehyde and 4% paraformaldehyde in PBS, post-fixed with 1% osmium tetroxide and 1.5% potassium hexacyanoferrate, stained en bloc with 1% uranyl acetate, dehydrated in ethanol, and embedded in Epon as described [72,73]. Semithin sections were cut with an RMC MT-7 ultramicrotome and photographs were taken with Kodak Ektachrome 160T slide film using a Nikon FX-35DX/UFX-DX camera/exposure system. Thin sections were post-stained with uranyl acetate and lead citrate and viewed with a Zeiss EM 910 electron microscope [73].
Electron microscopy negatives and Ektachrome slides were scanned with a Hewlett Packard Scanjet 5370C. All digital, electron, or confocal microscopy images were processed using Image-Pro Plus (Media Cybernetics), Adobe Photoshop (Adobe), and AutoDeBlur (AutoQuant Imaging, Incorporated) software. |
10.1371/journal.ppat.1002970 | How the Fly Balances Its Ability to Combat Different Pathogens | Health is a multidimensional landscape. If we just consider the host, there are many outputs that interest us: evolutionary fitness determining parameters like fecundity, survival and pathogen clearance as well as medically important health parameters like sleep, energy stores and appetite. Hosts use a variety of effector pathways to fight infections and these effectors are brought to bear differentially. Each pathogen causes a different disease as they have distinct virulence factors and niches; they each warp the health landscape in unique ways. Therefore, mutations affecting immunity can have complex phenotypes and distinct effects on each pathogen. Here we describe how two components of the fly's immune response, melanization and phagocytosis, contribute to the health landscape generated by the transcription factor ets21c (CG2914) and its putative effector, the signaling molecule wntD (CG8458). To probe the landscape, we infect with two pathogens: Listeria monocytogenes, which primarily lives intracellularly, and Streptococcus pneumoniae, which is an extracellular pathogen. Using the diversity of phenotypes generated by these mutants, we propose that survival during a L. monocytogenes infection is mediated by a combination of two host mechanisms: phagocytic activity and melanization; while survival during a S. pneumoniae infection is determined by phagocytic activity. In addition, increased phagocytic activity is beneficial during S. pneumoniae infection but detrimental during L. monocytogenes infection, demonstrating an inherent trade-off in the immune response.
| The importance of individual immune responses is incredibly infection dependent, and this paper harnesses the variability in two mutant lines to explain the relative importance of two aspects of fly immunity: melanization and phagocytosis. Increased phagocytic activity is beneficial during S. pneumoniae infection due to increased clearance of the extracellular microbe and detrimental during L. monocytogenes infection as it increases the intracellular niche for L. monocytogenes. Outcomes during L. monocytogenes infection are also dependent on melanization capability, which impacts the ability to control extracellular bacteria.
| Infected fruit flies get sick in ways that human patients would recognize; bacterial infections in Drosophila induce changes in feeding, metabolism and circadian rhythm, and conversely changes in these pathways influence susceptibility to infection [1]–[3]. Many responses affect survival during infection, but this work remains splintered as the field primarily focuses on individual mechanisms in isolation, offering glimpses of the whole picture. Here we use mutations in two genes, ets21c and wntD, to examine their effect on immune responses and survival during infections with two bacteria, Listeria monocytogenes and Streptococcus pneumoniae. Both genes affect multiple arms of the immune system and we wanted to understand how immunity offers protection against pathogens with different lifestyles in Drosophila. We chose these two microbes because they produced dissimilar phenotypes in previous Drosophila immunity assays [4]. Together these mutants and microbes demonstrate how there can be no perfect immune response, as there are responses that are beneficial during one infection and actively detrimental during another.
The Drosophila immune response can be divided into categories based on the speed at which they act following pathogenic challenge. The fast-acting immune responses, which respond within seconds to minutes, are phagocytosis and melanization [5]. Hemocytes are phagocytic cells in the fly and they are concentrated in adherent groups on the dorsal side of the abdomen and the anterior abdominal segment of the heart in adult flies. Inhibition of phagocytosis increases susceptibility to a number of bacteria [6]–[9]. Insects produce melanin from tyrosine using the enzyme phenoloxidase, which is activated by an immune triggered proteolytic cascade. This process is hypothesized to produce reactive oxygen species, which can harm the host in addition to harming the pathogen, and to physically encapsulate the invaders [10], [11]. In Drosophila, some bacterial pathogens (L. monocytogenes, Salmonella typhimurium, and Staphylococcus aureus) induce visible melanization, and flies defective in the melanization activation pathway are less resistant to these infections [4]. Though these relatively quick responses presumably remain active through the whole infection there is at least one response that takes several hours to reach full force. This slow response is the induction of anti-microbial peptides which peak in transcript expression six to 24 hours post infection [12]. We do not know when actual antimicrobial activity peaks as this is seldom assayed directly, but presumably this takes even longer than the increase of transcripts.
While it may be simplest to examine the effect of immune components individually, in order to effectively control immunity clinically we need a better understanding of the full immune network; each response doesn't exist in a vacuum. Knowing which immune responses strongly associate with a positive outcome for a given pathogen and which physiological systems are impacted by infection will allow doctors to more effectively treat disease. Patients normally do not have a single pathway or gene responsible for their entire pathology, and we need to develop the tools to deal with these levels of complexity.
To probe changes in the immune response, we turned to two pathogens that previously exhibited opposing phenotypes: Listeria monocytogenes and Streptococcus pneumoniae. When injected into the hemocoel, L. monocytogenes causes lethal infections in Drosophila melanogaster at doses as low as ten bacteria, and death from infection occurs on the order of one week. L. monocytogenes lives both intracellular and extracellular in the fly and causes robust disseminated melanization [4], [13]–[15]. S. pneumoniae can also cause lethal infections; however, there are sub lethal doses, which prime the fly to become resistant upon subsequent challenges [16]. S. pneumoniae infection kills flies rapidly, within two to four days, and flies surviving past four days have likely cleared the pathogen. S. pneumoniae is an extracellular pathogen and bead inhibition of phagocytosis increases susceptibility to infection [6]. In contrast to L. monocytogenes, flies deficient in melanization are more resistant to S. pneumoniae infection, although the mechanism is unknown [4].
Ets21c (CG2914), a putative transcription factor characterized by its DNA binding ets-domain, was previously implicated in Drosophila immunity. Ayres et al. found that et21c mutants died more rapidly during L. monocytogenes infection with similar bacterial loads compared to wild-type, but were no different from wild type flies when challenged with S. typhimurium or S. aureus [17]. Studies of immune signaling in Drosophila S2 cells and hemocyte cell lines used ets21c transcript as a read out of the early immune response and showed that ets21c induction depends on the imd pathway and one of its transcription factors, basket [18], [19].
WntD (CG8458) is a negative regulator of dorsal signaling in Drosophila, and wntD mutants are more susceptible to L. monocytogenes infection than wild type flies [20]. Previous work measured the signaling and transcriptional effects of wntD on antimicrobial peptides [21]; however, it remains unknown, whether wntD impacts bacterial load during L. monocytogenes infection and how it affects melanization and phagocytosis.
The effect of a given bacterial load has previously been used to categorize genes as either impacting tolerance or resistance [2], [4], [17]. Resistance genes and mechanisms directly impact how well the bacteria grow or are killed, while tolerance genes and mechanisms affect the host's ability to deal with the effect of infection (e.g. energy strain, accumulated damage). While both of these mechanisms are functionally distinct, they way they impact bacterial load cannot be as easily separated and there is a full spectrum of phenotypes possible, from genes that do not impact bacterial load at all to genes that increase bacterial load by hundred-fold in just a day. Determining where in this spectrum our mutants fall helps inform the possible responsible mechanisms.
In this paper, we show that ets21c and wntD mutants are both more susceptible to L. monocytogenes and more resistant to S. pneumoniae, but differ in their ability to control L. monocytogenes bacterial loads. At the levels of specific immune responses, these mutants share an increase in phagocytic activity and a shift in anti-microbial peptide induction, but differ in their melanization capabilities. By examining these differences, we establish the relative contributions of the immune pathways to these outcomes - survival during L. monocytogenes infection depends on multiple factors: melanization and phagocytic ability while phagocytic ability alone predicts survival to S. pneumoniae infection.
Ets21c is a putative transcription factor therefore we assessed the impact of an ets21c mutation on the transcriptome by performing a microarray analysis on infected flies. Complete microarray data is available in the online supplemental materials (Dataset S1). A familiar gene emerged in our list of infection induced genes in the parental line: wntD. WntD is a negative regulator of the dorsal pathway and wntD mutants die more quickly during L. monocytogenes infections [20]. Ets21c mutant flies do not upregulate wntD during L. monocytogenes infection and we confirmed this using real-time qRT-PCR (Figure 1A, p<0.001). S. pneumoniae induces expression of wntD in both the ets21c mutant and its parental line, but only 25-fold, which is lower relative to the hundred-fold induction during L. monocytogenes infection. The reciprocal effect of wntD mutants on ets21c expression was examined both in the microarray published by Gordon et al. and by qRT-PCR, but the levels of ets21c were so low in total fly RNA preparations that results were highly variable and therefore not significant (data not shown). WntD is a good candidate effector for ets21c's immune phenotypes, due to wntD's ability to impact survival to L. monocytogenes.
Ayres and colleagues published that ets21c mutants have increased susceptibility to L. monocytogenes with no significant increase in bacterial load causing them to conclude that the gene affected tolerance [17]. This mutant did, however, show an insignificant increase in L. monocytogenes bacterial load two days post-infection. Upon retesting, we found that these mutants exhibit a small but significant increase in bacterial load at 48 hours post infection (Figure 2A,E). We call this a small effect since there is no change at 24 hours and a nine fold increase in bacterial load at 48 hours whereas a mutation in another gene, gr28b, increases bacterial growth 100 fold at both 24 and 48 hours. This relatively small increase in bacterial growth rates in ets21c mutants was confirmed with flies that had ets21c expression knocked down by RNAi in the fatbody (Figure 3 A,B). Gordon and colleagues showed that wntD mutants were more susceptible to L. monocytogenes, but did not report bacterial loads [20]. We confirmed that wntD mutants die faster than parental flies (Figure 2B), and found that wntD mutants remain able to control bacterial growth to at least 48 hours post-infection (Figure 2F). We stopped measuring CFU at this point as over half the mutant flies die the next day and we worried that we would skew results if we were looking at survivors that may be more resistant or potentially received a smaller infectious dose. Knockdown of wntD in the fatbody confirmed wntD's effect on tolerance to L. monocytogenes (Figure S1). Mutants in ets21c and wntD are both more susceptible to L. monocytogenes infection, but fall on different parts of the tolerance-resistance continuum indicating that there may be mechanistic difference at play in these lines.
Drosophila has an excellent tool for knocking down gene expression in a tissue specific manner; tissue specific expression of the transcription factor GAL4 can be used to drive gene specific RNAi constructs. A large number of these tissue specific GAL4 lines and RNAi lines are publically available and one simply has to cross the driver line to the RNAi line and test the appropriate offspring. However, these lines are not perfectly tissue specific and can have significant expression levels in a variety of tissues. We tested a panel of GAL4 drivers to determine where ets21c was required during infection and found it difficult to interpret the data because all drivers tested produced a similar phenotype (Figure S2).
We reasoned that the problem was that driver localizations are primarily determined by ability to drive GFP expression in tissues; however, we worried that low expression levels of an RNAi construct might be sufficient to produce a phenotype while registering as background in a fluorescent microscopy assay that measured the induction of GFP. Instead of using the published localization for each driver, we assumed that the driver strength matched the expression data for each driver gene as reported by FlyAtlas, a database of tissue specific gene expression results from both larvae and adults. For each tissue, we used JMP Software (http://www.jmp.com) to determine whether there was a significant correlation between the expression level of the driver gene with the strength of immune phenotype as measured by median time to death (MTD) of RNAi×Driver/MTD of the driver control. As shown in Figure 4, higher expression of the driver genes in both the heart and fatbody correlated with increased sensitivity to L. monocytogenes infection. However, correlation of sensitivity to infection and driver strength in the heart was not as strong and also had a significant p-value for the lack of fit test indicating that the linear model may not be appropriate for this tissue. Driver expression values in the heart correlate with those in the fatbody (data not shown), so it is unsurprising that they would both exhibit a similar relationship with the sensitivity to L. monocytogenes infection. This data does not rule out a role for hemocytes as they are not reported in FlyAtlas and could potentially have adhered to other tissues, particularly the heart, during the isolation used for the FlyAtlas data. The closest available approximation for adult hemocytes was the expression data available for growing S2 cells in culture, which are known to have phagocytic properties, and microarray data published on larval hemocytes. Both of these cultured cells revealed a significant correlation between strength of driver expression and sensitivity to L. monocytogenes infection by ANOVA, but also had significant p-values for the lack of fit test indicating that the model may be inappropriate (data not shown). While it is unsurprising that the fatbody and/or hemocytes are important for our immune phenotype, this technique can be broadly applied to any phenotype that can be quantified and will allow quantitative and methodical use of drivers.
Ets21c mutants had been previously shown to have no impact during S. typhimurium or S. aureus infection, while published studies on wntD mutants only examined the impact during L. monocytogenes infection [17], [20], [21]. We chose to further test the specificity of ets21c with Streptococcus pneumoniae because it behaves uniquely in some other Drosophila immunity mutants [4], [22]. Mutants in both ets21c and wntD survive S. pneumoniae challenge better (Figure 2C–D) and with decreased bacterial loads compared to parental strains (Figure 2G–H). This phenotype was also confirmed using RNAi knockdown in the fatbody as described above (Figure 3B,D and Figure S1B,D). These results indicate that mutants in ets21c and wntD both have an increased resistance to S. pneumoniae infection, whereas they exhibited differences in their type of susceptibility to L. monocytogenes.
To examine which immune responses are responsible for the differences between ets21c and wntD mutants, we performed assays of each immune pathway to assess the strength of both fast acting and slow acting immune responses. Melanization and phagocytosis begin to act within seconds to minutes of infection, while anti-microbial peptide induction takes hours. It can be difficult to measure the strength of immune responses against pathogens; pathogens often evolve mechanisms to beat the immune response. To assess the strength of the fast acting responses we injected flies with Escherichia coli, which is non-pathogenic to D. melanogaster at this dose and is rapidly cleared from circulation. By focusing on the first hour post-injection, we assayed the fast acting responses of phagocytosis and melanization. While this is a very different microbe from both L. monocytogenes and S. pneumoniae, basic immune responses are likely conserved between infections and this is a first approximation of fast-acting responses. Ets21c mutants do not clear E. coli more quickly (Figure 5A). WntD mutants, however, have significantly increased E. coli clearance (Figure 5B). This result does not distinguish whether shifts in melanization or phagocytosis are occurring, and does not rule out the possibility that ets21c mutants have shifts in both melanization and phagocytosis for an overall neutral effect on E. coli clearance.
To determine the potential contribution of each of the fast acting immune responses to the observed phenotypes, we wanted to test both phagocytic ability and melanization capabilities separately. We determined the strength of phagocytic ability by assaying the ability of our mutants to phagocytose dead bacteria. We imaged flies injected with a pHrodo labeled E. coli, which only fluoresces upon encountering a low pH like that found in phagosomes; while pHrodo labeled L. monocytogenes would be the more perfect tool it isn't commercially available and the labeled E. coli provides an initial approximation for phagocytic ability. Quantifying the amount of fluorescence revealed that each mutant phagocytosed more bacteria than their parental line (Figure 6). There was also a dramatic difference in the phagocytic activity of the two parental lines, as evidenced by the fact that we had to use twice as long an exposure to take images of w1118 and ets21c mutants as compared with yw and wntD mutants. No direct comparison between all four lines was done because exposures which can capture w1118 and ets21c mutant differences completely over-exposes the other lines. These four lines offer a spectrum of phagocytic abilities giving us a broad dynamic range to assess the importance of phagocytosis during infection.
L. monocytogenes is a facultative intracellular bacterium that is capable of escaping the phagosome and living within the cytosol of phagocytic cells [15], [23], [24]. While an increase in phagocytosis may help clear extracellular pathogens such as S. pneumoniae, this increase may give additional access to a niche for L. monocytogenes. We tested this by determining how many intracellular L. monocytogenes were found in infected flies. Gentamicin is an antibiotic that is unable to cross cell membranes and injecting this antibiotic into the fly's circulation allows us to measure intracellular and extracellular populations. Figure 7 shows that both wntD and ets21c mutants have increased intracellular bacterial loads compared with their parental line. The water injected control, which reports total bacteria, further confirms that ets21c mutants also have an overall increase in bacterial load at 48 hours post-infection while wntD mutants have no significant change. When comparing the intracellular populations for all four lines at once, we noticed that the intracellular population was positively associated with phagocytic ability (Figure 7C).
We assayed the second fast immune response, melanization, by looking at the capability of the flies to show disseminated melanization after infection. L. monocytogenes causes visible melanotic spots within 3–4 days after infection, and flies defective in melanization are more susceptible to infections that cause this disseminated melanization [4]. When flies were scored as positive or negative for melanization, a significantly lower proportion of ets21c mutants exhibited visible disseminated melanization (Figure 8A), while a higher proportion of wntD mutants exhibited disseminated melanization (Figure 8B).
We examined anti-microbial peptide (AMP) gene induction in these two mutants. Gordon et al. reported that wntD mutants had increased induction of diptericin upon L. monocytogenes infection but saw no change in the induction of drosomycin [20]. We also found that ets21c mutants also have a four-fold increased induction of diptericin (p<0.05)(Figure 9A) and no change in drosomycin induction (data not shown). We also tested attacin, metchnikowin, defensin, drosocin and cecropin for changes during L. monocytogenes infection and found that ets21c only affected cecropin expression in that it was poorly induced during infection, about 10-fold lower than its parental line (p<0.001) (Figure 9B). In the microarray, ets21c mutants also showed up-regulation of most anti-microbial peptides, and did not show significantly different induction than the parental line (Figure S3). We suggest that this is a minor but complex impact on anti-microbial peptide expression as the majority of transcripts do not change and when they do change they can go up or down. While individual anti-microbial peptides have been shown to impact survival to infection in Drosophila, their effect was only visible in an otherwise immune compromised mutant with forced high expression of the anti-microbial peptide [25]. We do not understand the impact of modest changes in AMPs in a background where many AMPs are highly expressed and are unchanging.
The diverse and often opposing strengths and weaknesses of different pathogens leads to inherent trade-offs in immunity. In order to observe these trade-offs, one must infect with a range of pathogens and explore multiple arms of the immune response. This research used mutants in two genes to explore the contribution of two immune components: phagocytosis and melanization to the survival during infections with two bacteria: L. monocytogenes and S. pneumoniae. The line most resistant to S. pneumoniae dies the fastest when faced with L. monocytogenes and the reciprocal also holds true (Figure 10). The differences between these pathogens make them useful tools with which to probe the immune system.
This paper focuses primarily on the immune contribution of the fly's phagocytic ability. Due to differences in pathogen lifestyle, the increased phagocytosis in our mutants has unique consequences for each bacterium. S. pneumoniae is an extracellular bacterium, and phagocytosis is a death knell. L. monocytogenes can harness phagocytosis as an entry way to a protected niche. Our work shows that flies with the most phagocytosis are most susceptible to L. monocytogenes and have the correspondingly highest intracellular populations of L. monocytogenes while also being most resistant to S. pneumoniae. This presents a perplexing dilemma for the design of a robust immune system – what is the most advantageous amount of phagocytosis? This will depend on the frequency of pathogens encountered that will take advantage of this potential niche.
While it might be tempting to explain the range of phenotypes simply by the amount of phagocytic ability, the approximately equivalent survival of the ets21c mutant and yw parental line in spite of a difference in phagocytic ability suggests that additional factors might be at work. A second ingredient, melanization, is potentially that additional factor. The differential impact that ets21c and wntD have on melanization offers an explanation for why ets21c affects resistance while wntD affects tolerance to L. monocytogenes. Too little melanization is detrimental during a L. monocytogenes infection and causes increased extracellular bacteria [4]. Ets21c mutants decrease, but do not obliterate of the ability to melanize and have a corresponding significant but mild difference in their resistance to L. monocytogenes. WntD mutants, however, have an increased proportion of flies showing melanization compared to their parental line. The effects of hyper-melanization during infection are unknown, but production of melanin results in the production of reactive oxygen species (ROS) which can potentially harm the host as well as the microbe. If this increase in melanization in WntD mutants harms the host through ROS production, it could help explain the mutant's defect in tolerance to L. monocytogenes. The flies may become more dependent on this immune response and cause futile damage in their efforts to contain the bacteria.
Our results relegate anti-microbial peptides to a supporting role, primarily because we saw transcriptional changes in very few AMPs. Knockout of a single AMP has never been reported to produce a survival phenotype and we observed both increases and decreases of individual AMP induction. However, our data does not rule out the possibility that there is an AMP which can specifically affect either S. pneumoniae or L. monocytogenes and influence the phenotype. We believe this is unlikely because of the negative association between the two phenotypes and over-expression of a single AMP would have to be capable of producing the opposite effect on the other bacteria.
Another potential factor that this paper does not directly address is energy investment. Implicit in a stronger response is the energetic cost of mounting that response. While we do not measure this cost in our mutants, this is still a factor that could be influencing their survival outcome, especially of wntD mutants. These flies have a hyper-melanization response and increased phagocytic ability which may restrict the fly's access to a crucial amino acid – tyrosine, which is the precursor for melanin production. This could be a contributing factor to why these flies die so quickly compared to the other lines, while having the “strongest” of each of the immune responses.
A robust immune system must have an appropriate balance of immune responses to account for the diversity of pathogens it will encounter; however, even a well designed immune system will contain tradeoffs. A better appreciation of the natural and inevitable antagonism will help us gain a more in-depth appreciation for the evolutionary history behind our immune systems. Encouraging scientists to embrace pathogens which reveal distinct and even opposite phenotypes is necessary to fully explore the robustness and complexities of the immune response.
For ets21c experiments, a piggybac allele (Bloomington 18678) was compared to its parental strain, white1118 (Bloomington 6326) and an RNAi fly line (Vienna 106153) was crossed with GAL4 driver lines to elicit knockdown of ets21c. For wntD experiments, the knock out strain WntDKO1 was compared to its parental line yw and an RNAi fly line (Vienna 15146) was crossed to GAL4 driver lines to elicit knockdown of the WntD. [20] For RNAi experiments the following GAL4 driver lines from Bloomington were used: collagen25c (7011), mef-2 (27390), daughterless (8641), act5c (4414), elav (8765), lsp2 (6357), hemese (8700), and hemolectin (6395), appl (32040). RNAi experiments were conducted by crossing virgins from the RNAi line to males from driver line and collecting the progeny. If driver lines or RNAi lines contained a balancer, the progeny without the balancer were selected. Two control crosses were used for each RNAi experiment; an RNAi control with RNAi line virgins crossed to w1118 and a driver control with w1118 virgins crossed to males from the driver line.
Listeria monocytogenes (strain 10403S) cultures were grown in 4 ml brain heart infusion (BHI) broth at 37°C without shaking after inoculation from L. monocytogenes grown overnight on a Luria Bertani (LB) agar plate. L. monocytogenes stocks were stored at −80°C in BHI broth containing 15% glycerol.
Streptococcus pneumoniae (strain SP1) cultures were grown standing at 37°C 5% CO2 in BHI broth to an OD600 of 0.15, and aliquots were frozen at −80°C in 10% glycerol. For infection, an aliquot of S. pneumoniae was thawed, diluted 1∶3 in fresh BHI broth and allowed to grow at 37°C 5% CO2 for 3–4 hours.
Escherichia coli (strain DH5a) cultures were grown in 4 ml LB broth at 37°C with shaking after inoculation from bacteria grown overnight on a Luria Bertani (LB) agar plate. E. coli stocks were stored at −80°C in BHI broth containing 15% glycerol.
For infection, 50 nL of the bacterial cultures were injected at the following optical densities (OD600): L. monocytogenes, 0.01 (approx. 1,000 CFU/fly); S. pneumoniae, 0.05–0.3 (approx 2,000–10,000 CFU/fly); E. coli, 0.1 (approx. 3,000 CFU/fly).
Five to seven day post-eclosion male flies were used for injection. The flies were raised at 25°C, 65% humidity on yeasted dextrose food in a light cycling incubator (12 hours dark, 12 hours light). Flies were anesthetized with CO2. A picospritzer (Parker Hannin, http://www.parker.com) was used to inject 50 nL of liquid into each fly with pulled glass capillary needles that were individually calibrated by measuring the size of the expelled drop under oil. About 20 flies were placed per vial and then experiments were kept at 29°C, 65% humidity in a light cycling incubator.
Mutant flies and the parental control or RNAi crosses and RNAi/Driver controls were injected with 50 nL of the bacterial culture or medium. About sixty flies were assayed for each condition and placed in three vials of 20 flies each. Death was recorded daily. Survival curves are plotted as Kaplan-Meier plots and statistical significance is tested using log-rank analysis using Prism software (http://www.prism-software.com). All experiments were performed at least three times and yielded similar results.
Colony forming units (CFUs) were determined using both spot-plating and an autoplate spiral plater (Spiral Biotech http://www.aicompanies.com). For spot-plating, eight individual flies were collected at each time point. These flies were homogenized, diluted serially and plated onto the appropriate media (blood agar for S. pneumoniae and LB agar for L. monocytogenes and E. coli) and grown overnight at 37°C (5% CO2 for S. pneumoniae). Some L. monocytogenes experiments were completed using the Spiral Biotech plater and for these six individual flies were homogenized and diluted. 50 µL of liquid was plated exponentially on a LB plate, grown overnight at 37°C and then counted using QCount, which back calculates the original number of CFU per fly. For statistical analysis, if CFU/fly did not approximate a Gaussian distribution we analyzed the log(10) transform of the data. Most CFU experiments were assessed for sources of variation using a two-way ANOVA and followed with Bonferroni post-tests for specific comparisons of interest.
Gentamicin chase assays were performed as described by Ayres et al. 2008. [4] For the zero hour time point, flies were pre-injected with 50 nL of water or gentamicin. Flies were then injected with 50 nL of L. monocytogenes and put at 29°C. The flies were incubated for three hours and then plated to determine CFUs as described above. For 24 and 48 hour time points, flies were injected with 50 nL of water or gentamicin, incubated for three hours at 29°C and similarly plated. The statistical significance of specific comparisons of interest was assessed using a two-tailed t-test.
Flies with either injected with 50 nL of L. monocytogenes or BHI broth, simply stabbed with an empty needle or left unmanipulated. They were placed at 29°C for 6 hours. Groups of 20 flies were flash frozen in a dry ice/ethanol bath and then homogenized in TriZOL. Additional flies were injected and monitored for survival and CFUs to ensure adequate infection. RNA was isolated using a standard TriZOL preparation and then labeled cDNA was generated and hybridized to the Genome Drosophila Array (2.0) as described in the Affymetrix protocol (Affymetrix, http://www.affymetrix.com). Gene lists were assembled using comparisons done with dCHIP (http://biosun1.harvard.edu/complab/dchip). Anti-microbial peptide heatmap (Figure S3) created in Genespring 12.0. Select genes were confirmed by qRT-pCR.
Flies were injected with 50 nL of the indicated microbes or kept unmanipulated. Following injection, the flies were placed in dextrose vials and incubated at 29°C for six hours. Groups of 12 flies were homogenized in TriZOL and stored at −80°C until processed. RNA was isolated using a standard TriZOL preparation, and the samples were treated with DNase (Promega, http://www.promega.com). Quantitative RT-PCR was performed as described previously by Schneider et al. using a Bio-Rad icycler and the following primer sets: WntD, Diptericin, Cecropin, and RpS15Aa (for primer sequences see Table S1) [26].
These assays were performed as described by Shirasu-Hiza et al [6]. Briefly, flies were injected with 50 nL of 1 mg/ml pHrodo labeled E. coli (Molecular Probes, cat# P35361) and allowed to phagocytose at room temperature for 30–60 minutes. The wings of the flies were removed and the flies pinned onto a silicon pad with a minutien pin. Fluorescent images were taken of the dorsal surface using epifluorescent illumindation with Leica MZ3 microscope fitted with an ORCA camera. Images were captured with Openlab (Improvision), and exposures were set so that the brightest images showed no saturated pixels. Each experiment was repeated three times with 6–12 flies with similar results.
These assays were performed as described by Ayres et al [2]. Briefly, four days after injection flies were visualized by light microscopy and scored for a disseminated melanization response. Flies that melanized beyond the initial site of injection were scored positive for melanization response. Flies that only melanized at the site of injection were scored as negative for a melanization response.
|
10.1371/journal.ppat.1007849 | Multidimensional analysis of Gammaherpesvirus RNA expression reveals unexpected heterogeneity of gene expression | Virus-host interactions are frequently studied in bulk cell populations, obscuring cell-to-cell variation. Here we investigate endogenous herpesvirus gene expression at the single-cell level, combining a sensitive and robust fluorescent in situ hybridization platform with multiparameter flow cytometry, to study the expression of gammaherpesvirus non-coding RNAs (ncRNAs) during lytic replication, latent infection and reactivation in vitro. This method allowed robust detection of viral ncRNAs of murine gammaherpesvirus 68 (γHV68), Kaposi’s sarcoma associated herpesvirus and Epstein-Barr virus, revealing variable expression at the single-cell level. By quantifying the inter-relationship of viral ncRNA, viral mRNA, viral protein and host mRNA regulation during γHV68 infection, we find heterogeneous and asynchronous gene expression during latency and reactivation, with reactivation from latency identified by a distinct gene expression profile within rare cells. Further, during lytic replication with γHV68, we find many cells have limited viral gene expression, with only a fraction of cells showing robust gene expression, dynamic RNA localization, and progressive infection. Lytic viral gene expression was enhanced in primary fibroblasts and by conditions associated with enhanced viral replication, with multiple subpopulations of cells present in even highly permissive infection conditions. These findings, powered by single-cell analysis integrated with automated clustering algorithms, suggest inefficient or abortive γHV infection in many cells, and identify substantial heterogeneity in viral gene expression at the single-cell level.
| The gammaherpesviruses are a group of DNA tumor viruses that establish lifelong infection. How these viruses infect and manipulate cells has frequently been studied in bulk populations of cells. While these studies have been incredibly insightful, there is limited understanding of how virus infection proceeds within a single cell. Here we present a new approach to quantify gammaherpesvirus gene expression at the single-cell level. This method allows us to detect cell-to-cell variation in the expression of virus non-coding RNAs, an important and understudied class of RNAs which do not encode for proteins. By examining multiple features of virus gene expression, this method further reveals significant variation in infection between cells across multiple stages of infection, even in conditions generally thought to be highly uniform. These studies emphasize that gammaherpesvirus infection can be surprisingly heterogeneous when viewed at the level of the individual cell. Because this approach can be broadly applied across diverse viruses, this study affords new opportunities to understand the complexity of virus infection within single cells.
| The Herpesviridae are a family of large dsDNA viruses that include multiple prominent human and animal pathogens [1]. Although these viruses infect different cell types, and are associated with diverse pathologies, they share conserved genes and two fundamental phases of infection: lytic replication and latent infection [1]. Lytic replication is characterized by a cascade of viral gene expression, active viral DNA replication and the production of infectious virions. Conversely, latency is characterized by limited viral gene expression and the absence of de novo viral replication. While latent infection is a relatively quiescent form of infection, the herpesviruses can reactivate from latency, to reinitiate lytic replication.
Among the herpesviruses, the gammaherpesviruses (γHV) are lymphotropic viruses that include the human pathogens Epstein-Barr virus (EBV) [2] and Kaposi’s sarcoma associated herpesvirus (KSHV) [3]. Murine gammaherpesvirus 68 (γHV68, or MHV-68; ICTV nomenclature Murid herpesvirus 4, MuHV-4), is a well-described small animal model for the γHVs [4]. While these viruses establish a lifelong infection that is often clinically inapparent, immune-suppressed individuals are particularly at risk for γHV-associated malignancies [5].
Herpesvirus gene expression is extremely well-characterized in bulk populations. Despite increasing evidence for single-cell heterogeneity in gene expression [6–8], there remains limited understanding of herpesvirus infection at the single-cell level [9–12]. Here, we tracked endogenous viral and host RNAs using a sensitive, robust fluorescent in situ hybridization assay combined with multiparameter flow cytometry (PrimeFlow) [13] to analyze the expression and inter-relationships of viral ncRNA, viral mRNA and cellular mRNA at the single-cell level during γHV latency, reactivation and lytic replication. These studies revealed unanticipated heterogeneity of infection, emphasizing how single-cell analysis of virus infection can afford significant new insights into the complexity of γHV infection.
Traditional measurements of gene expression frequently rely on pooled cellular material, obscuring intercellular variation in gene expression. To better define expression of γHV RNAs at the single cell level, we employed the PrimeFlow RNA assay [13] to study viral gene expression during murine gammaherpesvirus 68 (γHV68) infection, a small animal γHV [4, 13]. This method is a highly sensitive, extremely specific in situ hybridization assay, integrating Affymetrix-designed branched DNA technology with single-cell analysis powered by multiparameter flow cytometry. This method has been successfully used to detect both virus and host RNAs (e.g. in the context of HIV infected individuals [13, 14]).
We first tested the ability of PrimeFlow to measure multiple viral RNAs during lytic infection with γHV68, including small non-coding RNAs (tRNA-miRNA encoding RNAs or TMERs [15]) and mRNAs. Mouse 3T12 fibroblasts were infected with a multiplicity of infection (MOI = 5 plaque forming units (PFU) of virus/cell). Under these conditions, TMER-5, one of the eight γHV68 TMERs, and the γHV68 ORF73, were readily detectable by conventional real-time PCR in γHV68-infected, but not mock-infected, cultures (Fig 1A and 1C). Parallel cultures were analyzed for RNA expression by PrimeFlow. Whereas mock-infected cells had no detectable expression of either the γHV68 TMERs or ORF73, WT γHV68-infected fibroblasts had a prominent population of TMER+ and ORF73+ cells, respectively (Fig 1B and 1D). Infection of cells with a TMER-deficient γHV68 (TMER-TKO [16]), in which TMER expression is ablated through promoter disruption, revealed no detectable TMER expression (Fig 1B), yet robust ORF73 expression (Fig 1D). Parallel studies revealed ready detection of ORF18, another γHV68 gene product (Fig 1E). These studies show that PrimeFlow is a sensitive, robust and specific method to detect both viral non-coding and messenger RNAs during lytic infection, quantifying both the frequency of gene expression and expression on a per cell basis.
γHV latency is characterized by limited gene expression. We next measured viral RNAs during latency and reactivation using the γHV68-infected A20 HE2.1 cell line (A20.γHV68), a drug-selected latency model with restricted viral gene expression that can reactivate following stimulation [17]. A20.γHV68 cells are characterized by restricted viral gene expression, yet remain competent for reactivation from latency and the production of infectious virions following chemical stimulation with the phorbol ester, TPA [17, 18].
When we compared TMER expression between uninfected (parental, virus-negative A20) and infected (A20.γHV68) cells by qRT-PCR, the viral ncRNA TMER-5 was readily detectable in A20.γHV68 cells above background signals in parental A20 cells, with minimal changes between untreated and chemically-stimulated conditions (Fig 2A). PrimeFlow analysis of TMER expression in untreated A20.γHV68 cells revealed that a majority of these cells expressed the TMERs, as defined by a positive signal in samples subjected to the TMER probe relative to unstained cells (Fig 2B). Untreated A20.γHV68 cells contained a high frequency of cells expressing intermediate levels of TMERs (i.e. TMERmid cells), with a significant signal enrichment above parental, virus-negative A20 cells (Fig 2C). While the frequency of TMERmid cells remained relatively constant following treatment with TPA (compare “Untreated” versus “TPA stimulated”, Fig 2C), TPA stimulated A20.γHV68 cultures also contained a small fraction of cells with high levels of TMERs (i.e. TMERhigh cells), not present in untreated cultures (Fig 2C and 2D). Chemical stimulation is known to result in variable penetrance of reactivation in latently infected cell lines [17]. Based on this, we hypothesized that these rare, TMERhigh cells may represent a subset of cells that are undergoing reactivation from latency.
To test this, we analyzed the properties of TMERmid and TMERhigh cells, comparing viral protein expression in untreated and stimulated A20.γHV68 cells. We analyzed: 1) a γHV68 expressed GFP-hygromycin resistance fusion protein (HygroGFP), under the control of a heterologous viral promoter (the CMV immediate early promoter) [17], and 2) the γHV68 regulator of complement activation (RCA), a viral protein encoded by the γHV68 ORF4, an early-late transcript [19]. The vast majority of TMERmid cells were negative for HygroGFP and RCA (i.e. HygroGFP- RCA-), regardless of whether the cells were present in untreated or stimulated cultures (Fig 2E and 2F). Conversely, TMERhigh cells, which were present at an increased frequency in stimulated cultures, had a significantly increased frequency of HygroGFP+ cells with induction of RCA protein+ cells in a subset of cells when compared to TMERmid cells present in either untreated or stimulated cultures (Fig 2E and 2F). By using imaging flow cytometry, we further analyzed the subcellular localization of TMERs in TMERmid cells compared with TMERhigh RCA+ cells. TMERs were predominantly nuclear in both TMERmid and TMERhigh RCA+ cells, as defined by co-localization with DAPI fluorescence (Fig 2G). These data demonstrate that the TMERs are expressed during latency, and that following reactivation-inducing stimulation, TMERs are further induced in a rare subset of cells which are characterized by increased viral transcription and translation.
To extend these findings, we analyzed viral gene expression in the KSHV infected B cell tumor line, BCBL-1, focused on detection of an abundant viral ncRNA, the KSHV polyadenylated nuclear RNA (PAN, nut1, or T1.1) [20]. PAN RNA is known to be highly inducible upon induction of reactivation in KSHV latently infected B cell lymphoma cell lines [10, 20]. The frequency of PAN RNA+ cells was low in untreated BCBL-1 cells, with ~1% of cells spontaneously expressing this ncNRA (Fig 3A and 3B). Despite the low frequency, this hybridization was clearly above background, as defined on the KSHV- and EBV-negative B cell lymphoma cell line BL41 [21, 22] (Fig 3A and 3B). Upon stimulation of BCBL-1 cells with the reactivation-inducing stimuli TPA and sodium butyrate, the frequency of PAN RNA+ cells significantly increased with expression in ~25% of cells (Fig 3A and 3B). Although stimulation of BCBL-1 cells significantly increased the frequency of PAN RNA+ events compared to untreated cultures, PAN RNA expression on an individual cell basis was comparable between cells from untreated or stimulated cultures (Fig 3C). As anticipated, stimulation of BCBL-1 cells was associated with increased viral DNA, consistent with stimulated cultures undergoing reactivation from latency (Supplemental Fig 1).
We next analyzed the properties of BCBL-1 cells as a function of PAN RNA expression. In untreated cells, PAN RNA+ or RNA- cells had comparable cell size (define by forward scatter, FSC) and granularity (defined by side scatter, SSC). ORF73 RNA expression was low in untreated BCBL-1 samples, with signal intensity in PAN RNA- cells close to the background fluorescence observed in unstained samples. PAN RNA+ cells in untreated cultures had a modest increase in ORF73 RNA expression relative to PAN RNA- cells (Fig 3D and 3E). In stimulated BCBL-1 cultures, PAN RNA+ cells had a modest decrease in cell size (defined by forward scatter) and a trend towards reduced granularity (defined by side scatter) compared to PAN RNA- cells (Fig 3F and 3G). Stimulated BCBL-1 cultures also had an increased ORF73 signal when compared to unstained samples (Fig 3F), with PAN RNA+ cells again showing ~2-fold increase compared to PAN RNA- cells (Fig 3F and 3G). These data demonstrate robust detection of PAN RNA by PrimeFlow, and further identify PAN RNA expression in a subset of both untreated and reactivation-induced BCBL-1 cells.
EBV encodes two abundant non-coding RNAs, the EBV-encoded RNAs (EBERs) EBER1 and EBER2. We tested the ability of the PrimeFlow method to detect EBER in an EBV positive, Burkitt lymphoma type I latency cell line, Mutu I [23]. EBER expression was detected in ~45% of Mutu I cells in either untreated or TPA stimulated conditions, with EBER+ cells defined relative to background probe hybridization in the KSHV- and EBV-negative BL41 cell line, a conservative measurement (Fig 4A). TPA stimulated Mutu I cells showed a modest, 2-fold increase in EBER expression on an individual cell basis, relative to untreated EBER+ cells (Fig 4B). Based on these data, EBER expression in Mutu I cells appears to be constitutive, with stimulation under these conditions resulting in minimal consequences on either the frequency or per-cell expression of the EBERs.
To extend these findings, we further analyzed EBER expression in a panel of LCLs and during in vitro infection of primary human B cells. EBER expression was readily detected in LCL cultures, with EBER+ cells also found in a subset of human primary B cells following in vitro EBV infection (S2 and S3 Figs). During EBV infection of human primary B cells, EBER+ B cells had an increased cell size and granularity relative to EBER- B cells, with increased expression of both CD69 and actin mRNA, an activated cell phenotype (S3D Fig). These data demonstrate cell to cell variation in EBER expression and suggest EBER expression as a potential discriminator to investigate variability during EBV infection.
Many herpesviruses, including γHV68, EBV and KSHV, induce host shutoff during lytic replication and reactivation from latency, a process characterized by dramatic decreases in host mRNAs [24, 25]. Consistent with published reports [24], qRT-PCR analysis of a cellular housekeeping gene, β-actin (Actb), showed reduced actin mRNA in γHV68 lytically-infected fibroblasts by 18 hours pi (Fig 5A). While mock-infected samples had a uniformly positive population of actin RNAhigh cells detectable by PrimeFlow, γHV68-infected fibroblast cultures demonstrated a bimodal distribution of actin RNAhigh and actin RNAlow cells (Fig 5B). The actin RNAlow population had a fluorescent signal that was only modestly above background fluorescence (defined by the “No probe” sample), suggesting an all-or-none phenomenon in which cells either had no change in actin RNA levels or had pronounced actin RNA degradation. Simultaneous analysis of TMER and actin RNA expression revealed that actin RNAlow cells were frequently TMERhigh, with actinhigh cells frequently TMERnegative at this time (Fig 5C).
To determine whether actin RNA regulation could also be observed during γHV68 latency and reactivation, we measured actin RNA levels in A20.γHV68 cells. Parental, virus-negative A20 cells and A20.γHV68 cells had relatively comparable actin RNA levels by qRT-PCR, in both untreated and stimulated cells (Fig 5D). Given that host shutoff is expected to primarily occur in rare, reactivating cells, we measured actin RNA degradation relative to TMER expression by the PrimeFlow method. Untreated A20.γHV68 cultures had no discernable population of TMER+ actin RNAlow events, whereas stimulated cultures were characterized by a rare population of TMERhigh actin RNAlow cells (Fig 5E). We further compared actin RNA expression between TMERmid and TMERhigh cells, in untreated versus stimulated cultures using our previously defined subpopulations (Fig 2). While TMERmid cells from either untreated or stimulated cultures were predominantly actin RNA+, TMERhigh cells from stimulated cultures showed a significant increased frequency of actin RNAlow events (Fig 5F and 5G). These studies reveal actin RNA as a sensitive indicator of virus-induced host shutoff, and demonstrate this as an all-or-none phenomenon that can be readily queried at the single-cell level.
Next, we revisited our analysis of gene expression during de novo lytic infection of fibroblasts, to examine co-expression relationships between viral ncRNA (TMERs), viral mRNA (the γHV68 ORF73), viral protein (RCA protein) and cellular actin mRNA degradation [19, 24]. Mouse 3T12 fibroblasts were infected with a multiplicity of infection (MOI = 5 plaque forming units of virus/cell), harvested 16 hpi and then subjected to the PrimeFlow method.
To enable an unbiased, automated analysis of gene expression profiles in γHV68 lytically infected cells relative to mock infected cells, data were subjected to the automated clustering algorithm X-shift [26], to identify potential subpopulations of cells with heterogeneous gene expression in these cultures. By sampling 1,000 cells from multiple mock- and virus-infected cultures, the X-shift algorithm consistently identified 7 major clusters of cells (Fig 6A) defined by varying gene expression patterns. While some of the clusters were exclusively found in mock-infected cultures, virus-infected cultures contained three broad types of cell clusters: 1) cells, with no detectable expression of either the TMERs or ORF73 and normal actin RNA, 2) fully infected cells, with robust expression of the TMERs, ORF73, actin RNA downregulation and frequent expression of the RCA protein, and 3) intermediate populations characterized by variable expression of the TMERs and ORF73 (Fig 6A).
To validate these findings using a more conventional method, we compared TMER and ORF73 RNA co-expression on a biaxial plot. By comparing mock-infected, WT-infected and TMER-TKO-infected cultures, this analysis revealed five populations of gene expression (Fig 6B), including cells with: 1) no detectable expression of either viral RNA (TMER- ORF73-), 2) TMER+ ORF73- cells (bottom right quadrant), 3) TMER- ORF73+ cells (upper left quadrant), 4) TMERlow ORF73low cells (lower left edge of the upper right quadrant), and 5) TMERhigh ORF73high cells (upper right quadrant). The definition of TMER positive events was defined based on background fluorescent levels observed in TMER-TKO infected cultures (Fig 6B). These 5 populations were each assigned a unique color for subsequent analysis (Fig 6C).
We compared cellular phenotype and gene expression across these 5 populations. Analysis of TMERs, ORF73, actin RNA, RCA protein, cell size (forward scatter), and granularity (side scatter) revealed multiple types of viral gene expression. TMER- ORF73- cells (in black) had no evidence of viral gene expression, with no detectable viral protein (RCA) or actin downregulation (Fig 6D and 6E). Cells with low expression of either the TMERs and/or ORF73 contained viral RNAs, but had minimal expression of either viral protein or actin downregulation (Fig 6D and 6E). In stark contrast, cells that were TMERhigh ORF73high (in red, Fig 6D and 6E) had multiple characteristics of progressive virus infection including a prominent fraction of cells that expressed RCA and/or had actin RNA downregulation. Further, TMERhigh ORF73high cells were consistently smaller in cell size (defined by forward scatter, FSC) and higher in granularity (defined by side scatter, SSC), a feature that was unique to this phenotype (Fig 6D and 6E).
TMERhigh ORF73high also demonstrated robust expression of an early gene (ORF64) and a late gene (ORF18) (Fig 7A). While some TMERlow ORF73low cells expressed early and late genes, cells defined as TMER- 73- had no discernable expression of either early or late genes (Fig 7A). These data emphasize that a subset of 3T12 fibroblasts subjected to infection with an MOI = 5 and harvested at 16 hpi have negligible signs of viral gene expression using this method. This is further illustrated by comparing mock and γHV68 infected cultures for TMER, ORF73, ORF64 and ORF18 expression which demonstrates a subset of cells with minimal expression across each of these viral genes (Fig 7B). One potential explanation for this could be that TMER- ORF73- cells are not infected. To test this, we sort purified cells based on TMER and ORF73 expression, followed by quantitation of viral DNA. TMERhigh ORF73high cells and TMER+ 73- cells had comparable levels of viral DNA, which were approximately ten-fold higher than viral DNA levels in TMER- 73- cells (Fig 7C). Despite this moderate difference in viral DNA, however, viral DNA levels in TMER- 73- cells were still significantly above background signal present in mock infected cultures (Fig 7C). Another potential explanation for this limited gene expression in TMER- 73- cells is that perhaps they are in a distinct phase of the cell cycle relative to other cells. We found that there was no sizable impact of cell cycle stage on whether cells were TMER- ORF73- or TMERhigh ORF73high (Fig 7D). These data identify unexpected heterogeneity during in vitro lytic infection, with TMERhigh ORF73high cells characterized by robust viral transcription, TMERlow ORF73low cells characterized by a lower frequency of cells expressing early and late genes, and TMER- ORF73- cells containing viral DNA with little to no detectable viral gene expression.
Given the heterogeneous patterns of RNA and protein expression among lytically-infected cells, we next queried TMER subcellular localization as a function of viral gene expression using imaging flow cytometry. While the majority of TMER+ cells had a primarily nuclear TMER localization (defined by DAPI co-localization, as in [27]), the frequency of cells with nuclear TMER localization was highest among TMER+ ORF73- cells and lowest among TMER+ ORF73+ RCA+ cells (Fig 8A and 8B). These data suggest that the TMERs can be localized in either the nucleus or cytoplasm during γHV68 lytic replication, and that this localization is not strictly a function of magnitude of gene expression.
Finally, we used tSNE, a dimensionality reduction algorithm, to better delineate the relationship between TMER, ORF73, RCA protein and actin downregulation across populations defined by variable TMER and ORF73 expression. In tSNE-based visualization, events are subjected to dimensionality reduction, with all events plotted according to the composite parameters tSNE1 and tSNE2. In each tSNE-based plot, values from individual cells are depicted as individual dots on the plot. Cells with similar expression profiles are visually clustered together using this algorithm (e.g. [28]). Consistent with our histogram analysis (Fig 6D and 6E), cells that were negative for TMER and ORF73 and cells that expressed either TMERs or ORF73 were relatively uniform in gene expression (Fig 8C). In contrast, TMERhigh ORF73high cells expressed a wider array of phenotypes, including both a predominant fraction of cells that were actin RNAlow RCA+, and a distinct group of cells that were actin RNA+ RCA- (Fig 8C). Notably, RCA expression and actin degradation were inversely correlated, with very few cells that expressed RCA also high for actin RNA. Actin RNA+ populations among TMERhigh ORF73high cells were associated with larger cell size (S4 Fig). The diversity of phenotypes among TMERhigh ORF73high cells was confirmed by biaxial gating of actin RNA versus RCA protein expression (Fig 8D). In total, these data indicate heterogeneous progression of lytic replication in vitro. While some cells have robust viral mRNA and protein expression, additional cell subsets are characterized by limited or divergent gene expression.
Our previous studies of γHV68 lytic replication used 3T12 fibroblasts, an immortalized cell line. Mouse embryonic fibroblasts (MEFs) are primary cells that are highly permissive for γHV68 infection, with 5 to 10-fold greater sensitivity to virus infection than 3T12 fibroblasts [29]. We therefore compared viral gene expression between 3T12 fibroblasts and MEFs, using MEFs derived either from wild-type or IFNAR1 KO mice. All cells were infected with WT γHV68 (MOI = 5) and analyzed at 16 hpi. Both WT and IFNAR KO MEFs showed enhanced viral gene expression relative to infected 3T12 cells at 16 hpi, with infected MEF cultures characterized by an increased frequency of events that expressed ORF73, TMERs, RCA and were actinlow (Fig 9A). Infected MEFs also had an increased frequency of ORF73+TMER+ cells and a decreased frequency of ORF73-TMER- cells relative to 3T12 cells (Fig 9B). When we analyzed the cellular distribution of viral gene expression across all 16 possible combinations of viral gene expression, resulting from different combinations of ORF73, TMER, RCA and actinlow phenotypes, infected MEF cultures compared to infected 3T12 cultures had: i) fewer ORF73-TMER- cells, ii) an increased frequency of ORF73+TMER+actinlow cells, iii) with the most frequent subset defined as cells with full viral gene expression (i.e. ORF73+TMER+RCA+actinlow) (Fig 9C). Viral gene expression distribution between WT and IFNAR1 KO MEFs were generally comparable, suggesting little to no contribution of type I IFN receptor mediated signaling as a regulator of viral gene expression in MEFs. These data demonstrate that the diversity and relative abundance of viral gene expression can be influenced by target cell type.
The increased viral gene expression observed in MEFs raised the possibility that viral gene expression could be further modulated in 3T12 cells, either as a function of time, multiplicity of infection or viral genotype. We therefore compared viral gene expression in 3T12 fibroblasts, comparing: i) WT γHV68 infection with 5 PFU/cell, ii) C-RTA γHV68 (a recombinant virus engineered to overexpress Rta, the immediate early viral transactivator [30]) infection with 5 PFU/cell, and iii) WT infection using 100 PFU/cell. Cells were harvested at 6 and 16 hpi. When comparing WT infection at MOI = 5 between 6 and 16 hpi, ORF73 expression peaked at 6 hpi with decreased expression by 16 hpi, in contrast to TMER, RCA and actinlow phenotypes which were infrequent at 6 hpi and increased by 16 hpi (Fig 9D). 3T12 cells subjected to WT infection at an MOI = 100 showed an increased expression of ORF73, TMER and RCA by 6 hpi, above WT infection with an MOI = 5. By 16 hpi, cultures infected with C-RTA and WT MOI = 100 had increased frequencies of TMER+ and ORF73+TMER+ events relative to WT MOI = 5 cultures (Fig 9E).
The diversity and progression of viral gene expression was further illustrated by analyzing the frequency of all 16 different combinations of possible gene expression between these different virus conditions. While WT MOI = 5 cultures showed a time-dependent switch from ORF73+ events to TMER+, ORF73+TMER+, and ORF73+TMER+RCA+actinlow events, the frequency of each of these populations was typically below 10% of cultures (Fig 9F). In contrast, cultures infected with C-RTA had limited viral gene expression at 6 hpi, but showed enhanced progression to both TMER+RCA+actinlow and ORF73+TMER+RCA+actinlow phenotypes relative to WT infection MOI = 5 (Fig 9F). Cultures infected with WT virus at an MOI = 100 showed rapid induction of multiple phenotypes, including a sizable portion of cultures which were ORF73+RCA+ and ORF73+TMER+RCA+ by 6 hpi, both populations which retained actin RNA expression. By 16 hpi, cultures infected with WT virus at an MOI = 100 had sizable frequencies of TMER+RCA+actinlow and ORF73+TMER+RCA+actinlow events (Fig 9F). While C-RTA and MOI = 100 infected cultures showed enhanced viral gene expression, WT cultures treated with phosphonoacetic acid (PAA), an inhibitor of viral DNA synthesis, showed impaired expression of TMER with a complete absence of TMERhigh events, negligible ORF73 expression and a pronounced inability to induce an actinlow phenotype (S5 Fig). These data demonstrate diversity of viral gene expression that can occur during lytic cycle and provide direct demonstration that these different gene expression profiles can be experimentally manipulated by either enhancing or restraining lytic replication.
Herpesvirus gene expression has been historically analyzed in bulk cell populations. These studies have provided an essential cornerstone to understanding the transcriptional and translational capacity of the herpesviruses. Despite this, recent studies on cellular and viral transcription from other systems have emphasized a high degree of cell-to-cell variation in gene expression [6–11, 13, 14], something we have further investigated here. By applying the PrimeFlow methodology to measure endogenous viral gene expression across multiple gammaherpesviruses, and multiple stages of infection, we have gained critical new insights into the inter-relationships of gene expression at the single-cell level.
A primary focus of the current study has been to analyze expression of γHV ncRNAs. Although the TMERs, EBERs and PAN RNA all represent abundant γHV ncRNAs, these ncRNAs are transcribed by distinct mechanisms: KSHV PAN is a highly-inducible, RNA pol II-transcribed ncRNA [20], in contrast to the RNA pol III-transcribed TMERs and EBERs [15, 31]. This differential regulation was mirrored in the expression patterns we observed. Whereas TMERs and EBERs were detected in a large fraction of latently infected cells, PAN RNA was expressed in a low frequency of latently infected cells, with prominent induction following cell stimulation and the induction of reactivation. The viral ncRNAs were efficiently detected, as might be predicted due to their abundance. The viral ORF73 encodes a transcription factor that is expressed at a far lower level and are also efficiently detected, demonstrating that rare mRNAs can be measured coincidently with abundant RNAs and with proteins, with no modifications required. A unique advantage of our current approach is the ability to measure the frequency of ncRNA expressing cells and changes in expression within individual cells. This has been particularly insightful for the identification of rare PAN RNA+ cells in untreated BCBL-1 cells and a TMERhigh subpopulation of cells in reactivating A20.γHV68 cells. Integrating this method with cell sorting will afford future opportunities to investigate unique properties of these rare cell populations.
Among the viral ncRNAs measured, in-depth analysis of TMER expression during γHV68 infection has revealed new insights into infection. In the context of latency, the TMERs are constitutively expressed in many, but not all, latently infected cells using the A20.γHV68 model. Further, stimulating these cells to undergo reactivation has a minimal effect on the frequency of cells expressing intermediate levels of TMERs (i.e. TMERmid cells), instead resulting in the appearance of a minor population of TMERhigh cells. Notably, TMERhigh cells show additional features of lytic cycle progression, including actin RNA degradation and RCA protein expression. Why only some latently infected cells show the TMERhigh phenotype, and what regulates the inducible expression of the RNA pol III-transcribed TMERs remain important questions raised by this analysis.
Of the γHVs studied here, only γHV68 has a robust in vitro lytic replication system. Our studies on γHV68 lytic replication revealed multiple unanticipated results. First, our analysis identified heterogeneity of viral gene expression, stratified by differential viral gene expression of the TMERs and ORF73. Strikingly, in cultures subjected to infection with 5 PFU per cell, conditions generally considered to induce synchronous lytic replication, there were many cells with limited viral gene expression, expressing low levels of either the TMERs and/or ORF73, but lacking additional signs of virus gene expression (i.e. actin RNA degradation or RCA protein expression). This analysis further identified a reproducible subset of cells with negligible viral gene expression, defined by a TMER- ORF73- phenotype. Curiously, these cells contain viral DNA but did not have further evidence of viral gene expression. Whether these cells represent an abortive state of infection, or if viral genes are expressed at levels below our current limit of detection remains to be determined. Our studies further revealed that only some viral RNA+ cells showed full progression of virus infection characterized by robust viral gene expression, defined as TMERhigh ORF73high RCA+ actin RNAlow. Though progression to full viral gene expression was enhanced by infection of primary fibroblasts, infection with the C-RTA virus, or a significantly higher MOI, even in these conditions there remained heterogeneity in viral gene expression. Whether heterogeneity of viral gene expression persists when cells manifest cytopathic effect, a process which occurs after 16 hpi, is not yet known. While there is precedence that reactivation from latency in KSHV infection can be asynchronous [9], this heterogeneity of viral gene expression during in vitro lytic replication was unanticipated and suggests that lytic infection under these reductionist conditions is either asynchronous, abortive, or inefficient. This heterogeneity of gene expression raises important questions regarding the universality of the prototypical cascade of immediate early, early and late gene expression that is widely accepted in the herpesvirus field and suggests additional levels of complexity that may be obscured by bulk cell analysis. The molecular mechanisms that are responsible for this heterogeneity still remain to be elucidated but appear to be independent of cell cycle stage with little to no contribution of type I interferon receptor mediated signaling based on relatively comparable gene expression profiles in wild-type and IFNAR1 KO MEFs.
This method allows multiplexed analysis of single-cell gene expression, to both directly measure viral RNAs and downstream consequences of gene expression including viral protein production and host RNA degradation, secondary to protein translation. This approach has notable advantages to conventional analyses of gene expression: 1) it can measure endogenous viral gene expression (both mRNA and ncRNA) in the absence of recombinant viruses or marker genes, and 2) it can rapidly analyze gene and protein expression inter-relationships, across millions of cells, providing unique complementary strengths to other single-cell methodologies (e.g. single-cell RNA-seq). In the future, this method can be further integrated with additional antibody-based reagents, to simultaneously query post-translational modifications (e.g. protein phosphorylation) as a function of cell cycle stage. It is also notable that through the use of imaging flow cytometry, it is possible to interrogate subcellular RNA and protein localization throughout distinct stages of virus infection, studies revealing single cell variability in subcellular RNA localization. We anticipate that this approach will have widespread utility, from addressing fundamental questions about herpesvirus gene expression using reductionist approaches, to a better delineation of replication or reactivation defects in viral mutants, to the in vivo identification of viral gene expression in primary infected samples.
The approach presented here provides a powerful complementary method to other single cell methods, affording the opportunity to query a diverse set of experimental manipulations in a relatively rapid manner. It is worthwhile to note, however, some important considerations with this approach. First, these studies rely on fluorescent measurements of probe hybridization using a flow cytometer. Though this approach detects a range of viral and host RNAs, sensitivity of this method is influenced by target RNA expression level, fluorophores used for the analysis and controls to define probe specificity and limit of detection. Ideally, comparisons can be strengthened by comparing isogenic conditions (e.g. comparing WT virus with a genetically deficient virus, as done with the TMER deficient virus). In cases where non-isogenic conditions are used (e.g. comparing across cell lines), variable background fluorescence may limit the sensitivity of this method. For new users of this technology, we strongly suggest the use of multiple controls, including a full-minus-one (i.e. FMO) control to accurately define background and signal to noise ratio for targets of interest. We consider multiparameter flow cytometry, as well as mass cytometry [32, 33], two increasingly useful technologies to afford new insight into heterogeneity of virus infection and gene expression that provide a complementary approach to other single cell technologies.
In total, these studies demonstrate the power of single-cell analysis of herpesvirus gene expression. Our data emphasize the heterogeneity of γHV gene expression at the single-cell level, even in conditions considered to result in uniform infection. The factors that underlie this heterogeneity are currently unknown, but could reflect either asynchronous or inefficient infection in many infected cells (e.g. in the context of lytic infection). The existence of specific infected cell subsets, based on heterogeneous gene expression, may identify new susceptibilities and points for intervention during the course of virus infection. Whether this variation arises from viral or cellular heterogeneity is a fundamental question for future research.
γHV68 viruses were derived from the γHV68 strain WUMS (ATCC VR-1465) [34], using either bacterial artificial chromosome-derived wild-type (WT) γHV68 or γHV68.TMER-Total KnockOut (TMER-TKO) [16], or virus derived by homologous recombination, γHV68.C-RTA [30]. Virus stocks were passaged, grown, and titered as previously described [16]. Mouse 3T12 fibroblasts (ATCC CCL-164) were infected with a multiplicity of infection (MOI) of either 5 or 100 plaque forming units/cell, analyzed 6–18 hpi. Primary mouse embryonic fibroblasts isolated from C57BL/6 mice (B6 MEFs) or from B6 IFNAR1 KO mice (IFNAR KO MEFs, kindly provided by Dr. Thomas E. Morrison at the University of Colorado) were cultured in DMEM with 10% FBS, 1% Penicillin/Streptomycin and then either mock- or virus-infected (MOI = 5) harvested at 16 hpi. For experiments using PAA, PAA (Sigma-Aldrich, catalog #284270) was added at a concentration of 200 μg/mL at time of virus infection and left on cultures until the time of harvest. The parental, virus-negative A20 B cell lymphoma cell line was obtained from ATCC (ATCC TIB-208) and cultured in RPMI 1640 with 10% FBS, 1% Penicillin/Streptomycin, L-glutamine and 50 μM β-mercaptoethanol (ME). γHV68 infected, and hygromycin selected A20.γHV68 (HE2.1) B cells [17] were obtained from Dr. Sam Speck (Emory University) and cultured in RPMI 1640 with 10% FBS, 1% Penicillin/Streptomycin + L-glutamine, 50 μM βME and 300 μg/mL Hygromycin B. A20 and A20.γHV68 B cells were treated with vehicle (untreated) or stimulated with 12-O-tetradecanolphrobol-13-actate (TPA) 20 ng/ml (Sigma) (in DMSO) harvested 24 hr later. BCBL-1 B cells, a body-cavity based B cell lymphoma cell line that is latently infected with KSHV (HHV-8), were obtained from the NIH AIDS reagent program (catalog # 3233). BCBL-1 cells were cultured in RPMI containing 20% FBS, 1% Penicillin/Streptomycin with L-glutamine, 1% HEPES and 50 μM βME. BCBL-1 B cells were treated with vehicle (untreated) or stimulated with 20 ng/ml TPA (in DMSO) and Sodium Butyrate (NaB) 0.3 mM (Calbiochem) (in water) and then harvested 72 hr later. BL41 B cells (negative for KSHV and EBV), were cultured in RPMI with 10% FBS, 1% Penicillin/Streptomycin with L-glutamine, and 50 μM βME. Mutu I cells, an EBV-infected, type I latency Burkitt’s lymphoma cell line [23] were obtained from Dr. Shannon Kenney (University of Wisconsin), and cultured in RPMI with 10% FBS, 1% Penicillin/Streptomycin and L-glutamine. Mutu I or BL41 B cells were either treated with vehicle (DMSO) or stimulated with 20 ng/ml TPA (in DMSO) and then harvested 48 hr later. EBV-immortalized B-lymphoblastoid cell lines (LCLs) were cultured in RPMI, 10% FBS, 1% Penicillin/Streptomycin with L-glutamine. LCL3, LCL9 and LCL209 BM were generated from Kenyan samples as previously described [35, 36].
Peripheral blood was obtained from consenting healthy adult donors and layered over Ficoll-Paque to isolate peripheral blood mononuclear cells (PBMCs). B cells were isolated from PBMCs through negative enrichment using EasySep Human B cell isolation kit (Stem Cell Technologies) following manufacturer’s protocol. B cells were plated at 1x106 cells per mL in RPMI, 10% FBS, 1% Penicillin/Streptomycin with L-glutamine, then infected with EBV at a MOI of 10 genome copies per cell or mock infected. 5 days post infection cells were harvested for flow cytometry analysis. EBV virus stocks were generated from the EBV+ cell line B95.8 which was reactivated with TPA (50:50 EtOH:Acetone) and Sodium Butyrate (NaB) for 5 days. Supernatant was collected and centrifuged at 4,000xg for 10 min then passed over a 0.7 micron filter. The supernatant was then ultra-centrifuged at 16,000x g for 90 mins and resuspended in 1/200th the initial volume using RPMI, 10% FBS, 1% Penicillin/Streptomycin with L-glutamine. Viral stocks were quantified following DNase treatment with qPCR analysis of the EBV BALF 5 gene done as previously described [37].
Cells were harvested at the indicated time points and processed for flow cytometry using the PrimeFlow RNA Assay (Thermo Fisher). Mouse cells were incubated with an Fc receptor blocking antibody (2.4G2) for 10 min and then fixed with 2% PFA (Fisher), washed with PBS (Life Technology). Cells were stained with a rabbit antibody against the γHV68 ORF4 protein, regulator of complement activation (RCA) [19], labeled with Zenon R-phycoerythrin rabbit IgG label reagent (Life Technologies) following manufacturer’s protocol. Human B cells were incubated in human Fc receptor blocking antibody then stained with Zombie Aqua fixable viability dye (1:500 dilution) (BioLegend), according to manufacturer’s protocol. Primary human B cells were subsequently stained with anti-CD19-FITC (Clone HIB19, dilution 1:25), and anti-CD69-PE antibody (clone FN50, dilution 1:20). Samples were subjected to the PrimeFlow RNA Assay following manufacturer’s protocols, using viral and host target probes conjugated to fluorescent molecules (Table A in S1 Text). DAPI (BioLegend) was used on a subset of samples following manufacturer’s protocol, prior to PrimeFlow probe hybridization. Flow cytometric analysis was done on LSR II (BD Biosciences), Fortessa (BD Biosciences), and ZE5 (Bio-Rad) flow cytometers, with compensation values based on antibody-stained beads (BD Biosciences) and cross-validated using cell samples stained with individual antibody conjugates, with compensation modified as needed post-collection using FlowJo.
Cells were harvested at the indicated time points and processed for flow cytometry using the PrimeFlow RNA Assay (Thermo Fisher). Cells were incubated with an Fc receptor blocking antibody (2.4G2), then fixed and permeabilized using reagents from the PrimeFlow RNA Assay (Fixation Buffer 1 and Permeabilization Buffer). Samples from different experimental conditions were fluorescently barcoded, with cells treated with either: no fluorescent dye, Ghost Dye Violet 450 or Ghost Dye Violet 510 (Tonbo Biosciences, 1:200 dilution) according to manufacturer’s protocol. After washing, the three labeled samples were pooled together into a single sample and stained with Zenon-labeled polyclonal rabbit antisera against viral RCA as desribed above. Cells were fixed to cross-link antibody stain with Fixation Buffer 2 from PrimeFlow RNA Assay, then subjected to the PrimeFlow RNA Assay following manufacturer’s protocols. For the analysis of barcoded samples, singlet cells from barcoded samples were analyzed for fluorescence of either Ghost Dye Violet 450 or Ghost Dye Violet 510, to identify the three input populations: cells negative for Ghost 450 and Ghost 510, cells singly positive for Ghost 450, and cell singly positive for Ghost 510. To ensure that no artifacts were introduced due to assignment of one barcode to a specific experimental condition, barcodes used for each experimental condition were shuffled across independent samples.
Cells were treated as described above then harvested, and split into two aliquots: one for conventional flow cytometry, and one for imaging flow cytometry, acquired on an Amnis ImageStreamX Mark II imaging flow cytometer (MilliporeSigma) with a 60X objective and low flow rate/high sensitivity using INSPIRE software. Brightfield (BF) and side scatter (SSC) images were illuminated by LED light and a 785nm laser respectively. Fluorescent probes were excited off 405nm, 488nm, and 642nm lasers with the power adjusted properly to avoid intensity saturation of the camera. Single color controls for compensation were acquired by keeping the same acquisition setting for samples, with the difference of turning the BF LED light and 785nm (SSC) laser off.
The acquired data were analyzed using IDEAS software (MilliporeSigma). Single cells that were in focus were defined as a population with a high “gradient RMS” value, an intermediate “Area” value, and a medium to high “Aspect ratio” value for subsequent analysis. Positive and negative events for each fluorescent marker were determined using the “Intensity” feature. TMER nuclear localization was quantified using “Similarity” feature, the log-transformed Pearson’s correlation coefficient by analyzing the pixel values of two image pairs [27]. The degree of nuclear localization of TMER was measured by correlating the pixel intensity of two images with the same spatial registry. The paired TMER and DAPI images were quantified by measuring the “Similarity Score” which cells with high similarity scores display high TMER nuclear localization with similar image pairs. By contrast cells with low similarity scores show low TMER nuclear localization with dissimilar image pairs. Cell cycle stage analysis was performed on data obtained from the imaging flow cytometer, stratifying cells based on DAPI content to identify cells in the G0/G1 phase, S phase or G2/M phase of the cell cycle.
Mouse fibroblasts (3T12) were infected with WT γHV68 (MOI = 5), harvested and processed for PrimeFlow at 16 hpi, followed by cell sorting using a BD FACSAria to purify cells based on relative TMER and ORF73 expression. Post-sort purity checks on sorted populations indicated that cell purities were 100% for TMER-ORF73- cells, 94.1% for TMER+ ORF73- and 98% for TMER+ORF73+ cells. DNA was isolated from samples using the DNeasy Blood and Tissue Kit (Qiagen), with an overnight proteinase K incubation followed by heat inactivation (95°C for 10 min). DNA was precipitated using ammonium acetate alcohol. 40 ng of DNA per sample was subjected to qPCR analysis using LightCycler 480 Probe Master-Mix kit (Roche) and primer sets for γHV68 gB and host NFAT5 (Table B in S1 Text). gB standard curve was generated using a gB plasmid dilution series ranging from 1010 to 101 copies diluted in background DNA, with a limit of detection (LOD) of 100 copies [38]. Host NFAT5 standard curve was generated using HEK293 cells, with 2x105 cell equivalents subjected to serial 10-fold dilutions in background, salmon sperm DNA, from 105 to 101 copies (LOD = 10 copies). Quantitation of viral gB copy number was standardized relative to input material using the formula (gB copy number / NFAT5 copy number) / 2, based on the assumption that cells have 2 copies of NFAT5 gene.
BCBL-1 or BL41 cells were plated at 7.5e5 cells/well in a 6 well plate with 20 ng/ml TPA and 0.3 mM NaB or vehicle only (DMSO and H2O). Cells and supernatant were harvested at 72 hrs post-treatment, hard-spun for 30 min at 4° C and DNA was isolated using the DNeasy Blood and Tissue kit following manufacturer’s protocol, except for sample digestion for 1 hour instead of 10 min. 100 ng of DNA per sample was used for qPCR analysis via SYBR green detection using KSHV ORF50 primers (5' -TCC GGC GGA TAT ACC GTC AC- 3' and 5'- GGT GCA GCT GGT ACA GTG TG-3') [39]. qPCR was analyzed using relative quantification normalized against unit mass calculation, ratio = EdeltaCt (Real-Time PCR Application Guide, Bio-Rad Laboratories Inc. 2006).
RNA was isolated using Trizol (Life Technologies) per manufacturer’s protocol and re-suspended in DEPC treated water. 3 μg of RNA was treated with DNase 1 (Promega) for 2 hours at 37°C, heat inactivated for 10 min at 65°C. 500 ng of RNA was then subjected to reverse transcription using SuperScript II (Life Technologies) following manufacturer’s protocol for gene specific, oligo(dT), or random primers (Life Technologies). Quantitative PCR (qPCR) was performed using iQ SYBR Green super mix (Bio-Rad) follow manufacturer’s protocol using host and viral primer sets (Table C in S1 Text) or using QuantiTech Primer Assay (Qiagen) for 18s (Hs-RRN18S_1_SG). qPCR conditions: 3 min at 95°C, amplification cycles for 40 cycles of 15 sec at 95°C, annealing/ extension at temperature for specific primer set for 1 min ending with a melt curve which started at 50°C or 55°C to 95°C increasing 0.5°C for 0:05 sec. A standard curve for each primer set was generated by pooling a portion of each sample together and doing a 1:3 serial dilution. 75 ng of cDNA of the unknown samples was loaded per qPCR reaction/primer set, with reactions run on a Bio-Rad 384 CFX LightCycler and data analyzed using Bio-Rad CFX manager software. Data analysis was done using the 1:3 standard curve as the control Ct value to calculate the delta ct, and the Pfaffl equation was used to define the fold difference between the gene of interest and 18s (reference gene) [40]. qPCR products were analyzed by melt curve analysis, with all reactions having a prominent, uniform product. In the case of primers with an aberrant melt curve product (e.g. that arose at late cycles), products were clearly a different product as defined by melt curve analysis.
All flow cytometry data were analyzed in FlowJo (version 8.8.7, 10.5.0, and 10.5.3), with flow cytometry data shown either as histogram overlays or pseudo-color dot plots (with or without smoothing), showing outliers (low or high resolution) on log10 scales. Statistical analysis and graphing were done in GraphPad Prism (Version 6.0d and 7.0d). Statistical significance was tested by unpaired t test (when comparing two conditions) or by one-way ANOVA (when comparing three or more samples) subjected to multiple corrections tests using recommended settings in Prism. X-shift analysis: For automated mapping of flow cytometry data using X-shift, data were obtained from compensated flow cytometry files, exported from FlowJo, using singlets that were live (defined by sequential gating on single cells by FSC-H vs. FSC-W and SSC-H vs. SSC- W, that were DAPI bright vs. SSC-A). These events were imported into the Java based program VorteX (http://web.stanford.edu/~samusik/vortex/) [26]. Four parameters [TMER (AlexaFluor (AF) 488), RCA (PE), ORF73 (AF647), and Actin (AF750)] were selected for clustering analysis using the X-shift algorithm. The following settings were used when importing the data set into VorteX: i) Numerical transformation: arcsinh(x/f), f = 150, ii) noise threshold: apply noise threshold of 1.0 (automatic and recommended setting), iii) feature rescaling: none, and iv) normalization: none, v) a Euclidean noise filter was used with a Minimal Euclidean length of the profile of 1.0, and vi) an import max of 1,000 rows from each file after filtering was selected. The following settings were used when preparing the data set for clustering analysis: i) distance measure: angular distance, ii) clustering algorithm: X-shift (gradient assignment), iii) density estimate: N nearest neighbors (fast), iv) number of neighbors for density estimate (K): from 150 to 5, with 30 steps, and v) number of neighbors for mode finding (N): determine automatically. After the cluster analysis was completed, all results were selected and the K value that corresponded with optimal clustering (the elbow point) was calculated, in this case K = 50. All clusters (seven clusters total) for the optimal K value were selected and a Force-Directed Layout was created. The maximum number of events sampled from each cluster was 20, and the number of nearest neighbors was 10. All settings used for this analysis were automated or explicitly recommended (https://github.com/nolanlab/vortex/wiki). Force-Directed layouts in Fig 6A were saved as graphml files from VorteX, opened in the application Gephi v 0.9.1, and colored by different variables (Cluster ID, experimental group, Actin mRNA, RCA, ORF73, and TMERs respectively) in Adobe Illustrator CC 2017. Full details on use of the X-shfit algorithm and analysis pipeline can be found in [41]. tSNE analysis: Gated events for each of the six identified populations were exported from FlowJo, and then imported into Cytobank (www.cytobank.org) for analysis using the viSNE algorithm. Each file was used for a separate viSNE analysis (six total runs), where all available events were selected for clustering (202,669, 128,028, 29,096, 5,610, 8,956, 35,850 respectively) and four parameters were selected for clustering (Actin mRNA, RCA, ORF73, and TMERs). The resulting tSNE plots were colored according to expression using the "rainbow" color option, with individual events shown using the stacked dot option. The channel range was user-defined for each marker according to the range in expression established in Fig 6E.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.