text
stringlengths 87
880k
| pmid
stringlengths 1
8
| accession_id
stringlengths 9
10
| license
stringclasses 2
values | last_updated
stringlengths 19
19
| retracted
stringclasses 2
values | citation
stringlengths 22
94
| decoded_as
stringclasses 2
values | journal
stringlengths 3
48
| year
int32 1.95k
2.02k
| doi
stringlengths 3
61
| oa_subset
stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-501620971010.1186/1471-2148-5-50Research ArticleHeterotachy and long-branch attraction in phylogenetics Philippe Hervé [email protected] Yan [email protected] Henner [email protected] Nicolas [email protected] Frédéric [email protected] Canadian Institute for Advanced Research, Centre Robert-Cedergren, Département de Biochimie, Université de Montréal, Succursale Centre-Ville, Montréal, Québec H3C3J7, Canada2 Laboratoire de Paléontologie, Phylogénie et Paléobiologie, Institut des Sciences de l'Evolution, UMR 5554-CNRS, Université Montpellier II, France2005 6 10 2005 5 50 50 21 7 2005 6 10 2005 Copyright © 2005 Philippe et al; licensee BioMed Central Ltd.2005Philippe et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Probabilistic methods have progressively supplanted the Maximum Parsimony (MP) method for inferring phylogenetic trees. One of the major reasons for this shift was that MP is much more sensitive to the Long Branch Attraction (LBA) artefact than is Maximum Likelihood (ML). However, recent work by Kolaczkowski and Thornton suggested, on the basis of simulations, that MP is less sensitive than ML to tree reconstruction artefacts generated by heterotachy, a phenomenon that corresponds to shifts in site-specific evolutionary rates over time. These results led these authors to recommend that the results of ML and MP analyses should be both reported and interpreted with the same caution. This specific conclusion revived the debate on the choice of the most accurate phylogenetic method for analysing real data in which various types of heterogeneities occur. However, variation of evolutionary rates across species was not explicitly incorporated in the original study of Kolaczkowski and Thornton, and in most of the subsequent heterotachous simulations published to date, where all terminal branch lengths were kept equal, an assumption that is biologically unrealistic.
Results
In this report, we performed more realistic simulations to evaluate the relative performance of MP and ML methods when two kinds of heterogeneities are considered: (i) within-site rate variation (heterotachy), and (ii) rate variation across lineages. Using a similar protocol as Kolaczkowski and Thornton to generate heterotachous datasets, we found that heterotachy, which constitutes a serious violation of existing models, decreases the accuracy of ML whatever the level of rate variation across lineages. In contrast, the accuracy of MP can either increase or decrease when the level of heterotachy increases, depending on the relative branch lengths. This result demonstrates that MP is not insensitive to heterotachy, contrary to the report of Kolaczkowski and Thornton. Finally, in the case of LBA (i.e. when two non-sister lineages evolved faster than the others), ML outperforms MP over a wide range of conditions, except for unrealistic levels of heterotachy.
Conclusion
For realistic combinations of both heterotachy and variation of evolutionary rates across lineages, ML is always more accurate than MP. Therefore, ML should be preferred over MP for analysing real data, all the more so since parametric methods also allow one to handle other types of biological heterogeneities much better, such as among sites rate variation. The confounding effects of heterotachy on tree reconstruction methods do exist, but can be eschewed by the development of mixture models in a probabilistic framework, as proposed by Kolaczkowski and Thornton themselves.
==== Body
Background
The long-branch attraction (LBA) artefact was first demonstrated to affect maximum parsimony (MP) [1,2], and subsequently all main types of tree reconstruction methods [3-5]. In the typical 4-taxa LBA case [1], two unrelated taxa (A and C) evolved significantly faster than their sister-groups (B and D); the inferred tree artefactually groups together the fast evolving taxa, because numerous convergent changes along the two long branches are interpreted as false synapomorphies (Fig. 1a). It should be noted that LBA could be alternatively named short-branch attraction, since the close resemblance of the two slow evolving taxa, due to symplesiomorphies, lead to their artificial attraction. In case of the LBA artefact, tree reconstruction methods are inconsistent, i.e. they converge towards an incorrect solution as more data are considered. Numerous computer simulations have shown that MP is the most sensitive method to the LBA artefact, whereas probabilistic methods, namely Maximum Likelihood (ML) and Bayesian Inference (BI) are more robust [3,4,6-9]. Since rate variation across lineages is almost invariantly observed in real data sets, often very pronounced, LBA artefacts have regularly been found to mislead phylogenetic inference [5,10-13]. As a result, the majority of phylogeneticists consider inferences made with probabilistic methods as the most reliable [8,14-16].
Figure 1 Illustration of the branch length heterogeneity conditions commonly referred as the Felsenstein zone (a) and the Farris zone (b). The Felsenstein zone [3] is characterised by two long branches that are not adjacent in the model topology, a situation where most phylogenetic methods fall into the long-branch attraction artefact [1]. Conversely, in the Farris zone [17], also called the inverse-Felsenstein zone [8], the two long branches are adjacent in the model topology. This last condition strongly favours MP over ML because of the intrinsic bias of parsimony towards interpreting multiple changes that occurred along the two long branches as false synapomorphies [8].
In 1998, Siddall argued that in certain cases MP outperforms ML when lineages evolved at markedly different evolutionary rates [17]. Instead of considering the so-called "Felsenstein zone" [3] where two unrelated taxa have long branches (Fig. 1a), Siddall [17] considered what he called the "Farris zone" where the two fast-evolving taxa are related (Fig. 1b). In this configuration, simulations based on sequences of 1,000 nucleotides demonstrated that MP recovered the correct tree more frequently than ML. The poor performance of ML relative to MP in the Farris zone, and the fact that MP "imposes the fewest assumptions about process", led Siddall to encourage the preferential use of MP over ML [17]. However, it was not demonstrated that ML was inconsistent in the Farris zone, since only short sequences were considered. Indeed, when sufficiently long sequences were used, ML recovered the correct tree [8]. In the Farris zone, ML is simply more cautious than MP for grouping the two long branches together because this method acknowledges the fact that many false synapomorphies uniting these branches are the result of convergence [8]. In contrast, the literal interpretation of substitutions made by MP leads to the grouping of the two long branches even if the internal branch length, i.e. the number of true synapomorphies is zero [8]. Swofford et al. [8] conclude that "most scientists would prefer to use methods that are honest about how strongly a result is [i.e. ML] than to use a method that pretends that a result is strongly supported when the majority of that support is a consequence of bias [i.e. MP]". In addition, since, under various simulation conditions, ML is always more accurate than MP in face of across-lineage rate variation, investigators continued to prefer ML for analysing real data.
It should nevertheless be noted that most early simulations demonstrating the higher accuracy of ML methods were made using a very simple model of evolution, often the Jukes and Cantor model [18]. Substitution properties vary from one position to another, with respect to rates [19] as well as to the type of substitution propensity [20,21]. Simulation studies have therefore been undertaken in order to investigate the effect of across-site rate variation [4,22] and compositional heterogeneity [9]. However, the evolutionary rate of a given position can also vary throughout time [23], a phenomenon called heterotachy (different speed in Greek) [24]. Heterotachy has been shown to be widespread [25,26] and to affect the performance of phylogenetic reconstruction methods in empirical datasets [27-32].
In a recent simulation study, Kolaczkowski and Thornton (hereafter referred as KT) found that, when the level of heterotachy is sufficiently high, MP is more accurate than ML, i.e. recovers the correct tree with infinite sequences under conditions where ML does not [33]. More precisely, KT used a simple but clever approach to simulate heterotachy (Fig. 2a). Two sets of sequences are simulated using the same model topology, but under two totally different sets of branch lengths (e.g. p and q for the branch length leading to A and B, respectively). These two heterogeneous sets of sequences are then combined and analysed using standard tree reconstruction methods (ML and MP). Under this scheme, the level of heterotachy can be modified by changing the values of p and q (Fig. 2a in [33]) or the relative weight (w) of the partitions (Fig. 2b in [33]). The difference in accuracy between two methods can then be evaluated as the value of the internal branch length (r), for which the correct tree is inferred in more than 50% of the simulation replicates (a value called BL50). Even when sequence length is limited (1,000 nucleotides), BL50 provides a good estimate to the boundary value r0 for which tree reconstruction becomes inconsistent when r < r0 (see Fig. 1 and Fig. S2 of [33]). For high levels of heterotachy (w = 0.5 and p/q > 2.2), it appears that ML is less accurate than MP with higher values of BL50 [33]. Consequently, KT "recommend reporting nonparametric analyses along with parametric results and interpreting likelihood-based inferences with the same caution now applied to maximum parsimony trees" [33].
Figure 2 Schematic presentation of the protocol used to simulate heterotachous alignments. Sequences were generated similarly as in ref. [33] under two different sets of branch lengths of equal weight (w = 0.5). In ref. [33], the branch lengths were altered by swapping the values of p and q (a). In our case (b), a single parameter (τ) allows to adjust the level of heterotachy from fully homotachous (τ = 0) to extreme heterotachous (τ = 1) conditions, while keeping the averaged branch length constant. Our branch lengths are (1 + τ) p and (1 - τ) q for the first partition and (1 - τ) p and (1 + τ) q for the second partition. 100 replicates of 5,000 nucleotide positions were simulated for each partition assuming a uniform JC69 model [18] using SeqGen [51] and were concatenated before phylogenetic inference using PAUP* [52].
The simulation results reported by KT and the authors' conclusions on the relative performance of MP and ML [33] prompted the publication of more simulations aimed at exploring heterotachy more widely [34-36]. Spencer et al. [35] performed simulations on all 15 possible combinations of two different edge-length partitions with two long and two short terminal edges and showed that ML performs better or at least as well as MP on the majority of combinations [35]. Moreover, they also demonstrated that when accounting for both substitution and across-site rate heterogeneities, the performance difference between the two methods is largely alleviated [35]. These authors further demonstrated that the correct implementation of a mixture model dealing with heterotachy, first proposed by KT [33], renders ML largely superior to MP under conditions where standard ML was outperformed [35].
In the simulations of KT [33], the terminal branch lengths, averaged over the two partitions, were kept equal to (p + q)/2. Therefore, although heterotachy is accounted for, these simulations largely ignored a major kind of heterogeneity: rate variation across lineages. Neglecting across-lineage rate heterogeneity is problematic because it is the main reason motivating the preference of ML over MP by most investigators. One way of simultaneously altering the level of heterotachy and across-lineage rate variation is to change the relative weight (w) of the two partitions, as in KT's Fig. 2b. In this case however, the averaged terminal branch lengths become heterogeneous in a complex manner and KT reported only the performance of ML [33]. More recently, KT's simulations were expanded by exploring a wider range of w and it was demonstrated that ML in fact outperforms MP over the majority of the parameter space [34,36].
In this report, we define a single parameter controlling the level of heterotachy without modifying the relative weights of the two partitions (w = 0.5). We present computer simulations that simultaneously account for heterotachy and across-lineage rate variation. We show that the known superiority of ML methods over MP when rates vary across lineages still holds in the presence of a realistic level of heterotachy.
Results
First, we introduce a new parameter (τ) that allows for the adjustment of varying levels of heterotachy, while keeping the averaged branch lengths constant. As shown on Figure 2b, terminal branch lengths leading to A and C are equal to (1 + τ) p and (1 - τ) p for the two partitions respectively. Using a weight w of 0.5 allows having a branch length of p, whatever the level of heterotachy. We varied τ from 0 (no heterotachy, homogeneous evolutionary rate) to 0.9 (high level of heterotachy, the evolutionary rate differing by a factor of 19 between the two partitions). Note that a different value of τ could be applied to each branch. For simplicity, we chose the same value of τ for all terminal branches of the model topology and therefore our simulations explore only a specific form of heterotachy.
The first simulations were realised using model topologies belonging to the Felsenstein zone, from severe (q = 0.15 and p = 4.5q) to moderate (q = 0.15 and p = 2q) rate variation among lineages. When p = 4.5q (Fig. 3a), ML (black circles) is much more accurate than MP (red squares), except for extreme heterotachy (τ = 0.9). For example, for τ = 0.5, the internal branch length r for which ML recovers the correct tree in more than 50% of the simulations (BL50) is equal to 0.068 whereas BL50 = 0.146 for MP. Interestingly, the performance of both ML and MP is negatively affected by increasing the level of heterotachy. However, the effect is much more pronounced for ML, going from BL50 ≈ 0 without heterotachy to BL50 ≈ 0.196 when τ = 0.9, whereas MP goes from 0.126 to 0.188. Therefore, for extreme heterotachy, MP is slightly more accurate than ML.
Figure 3 Performance of maximum parsimony (MP) and maximum likelihood (ML) phylogenetic methods for varying levels of heterotachy (τ) and increasing rate variation among species in the Felsenstein zone. For three combinations of p and q (a, b, c), the performance of MP and ML in the Felsenstein zone (i.e. p > q) [8] was evaluated under varying levels of heterotachy. The accuracy was calculated as in ref. [33] with BL50, i.e. the estimated internal branch length that allows recovering the true tree 50% of the time in 100 simulations using PAUP* [52].
The results are very similar when across-lineage rate variation is less extreme with p = 3q (Fig. 3b) or p = 2q (Fig. 3c). With increasing values of τ, the accuracy of both methods decreases, however the decrease is faster for ML than for MP. Since, without heterotachy, the difference in BL50 between MP and ML is lower when the rate heterogeneity is reduced, MP becomes more accurate than ML for lower values of τ (τ > 0.8 when p = 4.5q, τ > 0.7 when p = 3q and τ > 0.5 when p = 2q). Nevertheless, at levels of rate heterogeneity often observed in real data sets (two-fold to four-fold differences) ML is more accurate than MP even in the presence of a significant level of heterotachy (τ = 0.5). In fact, when τ = 0.5, the difference of evolutionary rates between the two partitions is already three-fold.
Finally, we also studied the impact of heterotachy when going from the Felsenstein zone to the Farris zone. We chose a more extreme case of rate heterogeneity (p = 0.75 and q = 0.05). The transition was performed by transferring a part of the length of the branch leading to A to the branch leading to D. For instance, we moved from (A: 0.75, B: 0.05, (C: 0.75, D: 0.05): r) to (A: 0.65, B: 0.05, (C: 0.75, D: 0.15): r). As found previously [3,4,6-9,22], in the Felsenstein zone and in the absence of heterotachy (τ = 0), ML is more accurate than MP until the two longest branches become the adjacent ones (Fig. 4). After entering the Farris zone, the values of BL50 are close to 0 for the two methods because the number of simulated nucleotides used here is large (10,000). As in Fig. 3, the accuracy of ML always decreases with increasing values of τ. In contrast, with increasing levels of heterotachy, the accuracy of MP sometimes increases or is not affected, but generally also decreases, albeit less rapidly than ML. As a result, heterotachy only slightly modifies the relative behaviour of ML and MP. When the two longest branches are not adjacent, ML outperforms MP, except when τ is high. When the two longest branches are adjacent, MP always outperforms ML. The only difference is that when heterotachy is present, the poorest performance of ML is not limited to its efficiency (the number of characters necessary to recover the correct tree) but also to its consistency.
Figure 4 Performance of maximum parsimony (MP) and maximum likelihood (ML) phylogenetic methods for varying levels of heterotachy (τ) while going from the Felsenstein zone to the Farris zone. Nine combinations of p and q (a-i) were explored by realising a morphing from one zone to the other by transferring a part of the length of the branch leading to A to the branch leading to D. The accuracy was calculated as in ref. [33] with BL50, i.e. the estimated internal branch length that allows recovering the true tree 50% of the time in 100 simulations using PAUP* [52]. As in the classical case [8], ML is more accurate than MP in the Felsenstein zone and the situation reverts when entering the Farris zone were MP is less affected than ML by increasing the level of heterotachy. However, the accuracy of ML always decreases with increasing value of τ, whereas the effect of heterotachy on MP is more complex, sometimes it increases but generally it also decreases its accuracy.
Discussion
Our results (Fig. 3 and 4) confirmed previous studies [27,33-36] that heterotachy renders probabilistic methods inconsistent. In contradiction with KT who stated that MP "is not additionally hampered by evolutionary heterogeneity" [33], we found that MP is also affected by heterotachy, its performance being generally degraded, but sometimes also improved depending on the branch length combination considered. In fact, KT's observation of MP being not affected by heterotachy is due to a very specific simulation design. By modifying the relative weight of the two partitions, they simultaneously modified the level of heterotachy and the average terminal branch length. For instance, with w = 0, there is no heterotachy and terminal branch lengths are p and q; with w = 0.2, medium heterotachy and terminal branch lengths are 0.2p + 0.8q and 0.2q + 0.8p; with w = 0.5, strong heterotachy and terminal branch lengths are of equal size, (p + q) / 2 (see also [36]). The lack of sensitivity of MP to heterotachy observed by KT is therefore due to an extremely peculiar combination of branch lengths and heterotachy level. When the effect of heterotachy is explored with a fix set of branch lengths, MP is affected by heterotachy, often to a great extent (BL50 varying from ~0 to 0.238 in Fig. 4e).
Interestingly, the accuracy of MP does not always decrease with increasing heterotachy (Fig. 4a), illustrating a rather complex behaviour over the parameter range here covered (Fig. 4). The explanation is that, with an increasing level of heterotachy, the branch lengths of one or two partitions can shift from the Felsenstein in the direction of the Farris zone, and vice versa. For instance, when the average branch length is well in the Felsenstein zone (Fig. 4a) and τ = 0.9, the first partition is entirely in the Felsenstein zone [model topology (A: 1.425, B: 0.005, (C: 1.425, D: 0.005): r)], whereas the other partition is only on the border of this zone [model topology (A: 0.075, B: 0.095, (C: 0.075, D: 0.095): r)]. Therefore only the first partition contains a large number of convergences that mislead MP, in contrast with the homotachous situation where the two partitions are in the Felsenstein zone. This explains why the accuracy of MP increases in the case of Fig. 4a. In contrast, for the opposite case of Fig. 4e, one starts from (A: 0.4, B: 0.05, (C: 0.75, D: 0.4): r) and goes to (A: 0.76, B: 0.005, (C: 1.425, D: 0.04): r) and (A: 0.04, B: 0.095, (C: 0.075, D: 0.76): r) when τ = 0.9. Here, one of the partitions is clearly in the Felsenstein zone when τ = 0.9, whereas the starting point is exactly in-between the Felsenstein and Farris zones, explaining the decreased accuracy of MP. In summary, contrary to the claim of KT [33], MP is also affected by heterotachy, often to a great extent. However, there is no simple rule to predict whether heterotachy will improve or decrease the accuracy of MP.
Nevertheless, under extreme heterotachy (τ = 0.9), MP almost always outperforms ML whereas ML is generally more accurate when τ < 0.5. But, as noted by Swofford et al. [8], the better performance of MP in the Farris zone (Fig. 4f–i) is due to an intrinsic bias of MP (i.e. misinterpretation of convergences as synapomorphies) and cannot be used as an argument in favour of MP. To guide the choice of investigators in analysing real data, we evaluated the extent of heterotachy in real data sets by developing a Bayesian mixture model that assumes k partitions and estimates the k sets of associated branch lengths and the relative weights of the k partitions, as proposed by KT [33] and corrected in Spencer et al. [35]. For the sake of comparability with our simulations, we assumed two partitions. The values of τ for each branch were calculated for several large alignments of amino acid sequences from various taxonomic groups (133 nuclear proteins from eukaryotes [37], 146 nuclear proteins from animals [38], 45 proteins from Archaea [39], 57 proteins from Bacteria [40], 13 mitochondrial proteins from deuterostomes [41] and 50 proteins from plastids and cyanobacteria [42]). We confirmed that heterotachy exists in real data [25], but the averaged observed value of τ is rather low, 0.17 (Yan Zhou, unpublished results). According to these empirical observations, a realistic level of heterotachy can be considered to fall within the parameter range (0 < τ < 0.4) with evolutionary rate varying between a two to three fold difference across lineages. Under these conditions, ML is always more accurate than MP and we therefore strongly recommend preferential use of ML over MP for inferring phylogenetic trees from real data.
In fact, it is not surprising that the influence of the level of heterotachy on the performance of phylogenetic methods when analysing real data is less important than across-lineage rate variation. Variation of evolutionary rates is indeed widespread and can easily be observed for any gene, with clock-like genes being the exception. In contrast, detecting heterotachy is much more difficult, as demonstrated by a short historical overview of its discovery and characterisation. Fitch recognized early on that invariable sites are not identical in cytochrome c of animals and plants [43]. However, several other heterogeneities such as rate variation across sites [19], across lineages [1], across substitution types [44,45], as well as compositional biases [46], appear to be more prominent in the evolutionary process. Indeed, a larger amount of data is necessary to detect heterotachy [25,28] relative to other evolutionary heterogeneities. All other kinds of evolutionary heterogeneities have been successfully and naturally addressed in a probabilistic framework [47], whereas various attempts to decrease the sensitivity of MP to these problems are far from being efficient and widely accepted. The case study in which MP outperforms ML under heterogeneous conditions [33] is unrealistic in the sense that no evolutionary heterogeneity except a very strong heterotachy (0.36 < τ < 0.75) was considered. We have shown here that taking into account across-lineage rate variation reverses the MP / ML accuracy ratio.
Heterotachy has been proposed as a cause of tree reconstruction artefact in the case of fast evolving lineages such as chloroplasts [48] or microsporidia [30,31]. It was proposed that model violations due to heterotachy render probabilistic methods inaccurate [27]. Contrary to the claims of KT [33], we have found that MP is not a valuable alternative to ML for dealing with heterotachy, as it is too sensitive to LBA. For example, microsporidia represent a phylogenetic problem where the occurrence of both strong evolutionary rate variations and heterotachy have been demonstrated to affect tree reconstruction [30,31]. In agreement with the simulations performed here, we recently showed on a phylogenomic dataset that MP is unable to correctly locate microsporidia among eukaryotes whereas ML can [37].
Conclusion
Phylogenetic reconstruction is rendered difficult by the occurrence of numerous evolutionary heterogeneities in molecular sequence data. KT [33] have judiciously pointed out that heterotachy seriously affects probabilistic methods. The reason is that the averaged branch length, which is fundamental for detecting convergent changes along long branches, no longer represents an accurate estimate when heterotachy is strong. However, from the extremely specific design of their simulations, KT found that MP would be unaffected by heterotachy and therefore suggested to consider with equal caution the results of MP and ML [33]. Here, we have found that MP can be affected by heterotachy and that it is much less efficient than probabilistic methods in dealing with all other evolutionary heterogeneities. We therefore strongly urge the continued preference of probabilistic methods for inferring phylogenies from real sequences (see also [35,36,49]). Indeed, heterotachy, as well as other kinds of heterogeneities [20,21], can be handled properly in a probabilistic framework using mixture models [33,35,50].
Methods
We followed a similar protocol as in [33], with the only difference being in the branch lengths of the model topology. Briefly, DNA sequences of 10,000 nucleotides each were simulated under the Jukes and Cantor [18] model with Seq-Gen version 1.2.7 [51]. Modelling rate heterogeneity across sites using a Gamma distribution (α = 0.5 and 1) gave similar results (data not shown). Considering a transition/transversion ratio greater than 1 (2, 5 or 10) rendered ML more accurate than standard MP (see also [35]), but when a weighted MP is used the same results as with a ratio of 1 were obtained (data not shown). As described in Fig. 2b, a single parameter, τ, allows for the adjustment of the level of heterotachy from fully homotachous (τ = 0) to extreme heterotachous (τ = 1) conditions. We varied τ from 0 to 0.9 by a step of 0.1. The two partitions were always of the same size (w = 0.5). As detailed in the main text, various values of p and q are used. The internal branch r was varied from 0 to 0.4 with a step of 0.01. One hundred simulations were performed for each combination of p, q, r and τ. Phylogenies were inferred by MP and ML (with a Jukes and Cantor model) using PAUP* version 4.0b10 [52]. Finally, to estimate the accuracy for both methods, BL50 (i.e. the value of r for which 50% of the simulations recover the correct tree) was computed through nonlinear regression using the R software version 2.0.0 [53]. When r < BL50, increasing sequence length decreases tree reconstruction method accuracy [33], which corresponds to the definition of inconsistency.
Authors' contributions
HP and FD conceived the study and drew the figures. HP performed the simulations and wrote the first draft of the manuscript. All authors contributed to the analysis of the results and to the writing of the paper. All authors read and approved the final manuscript.
Acknowledgements
We thank David Bryant, Nicolas Lartillot and two anonymous reviewers for helpful comments. This work was financially supported by grants from Génome Québec, the Canadian Research Chair and the Université de Montréal.
==== Refs
Felsenstein J Cases in which parsimony or compatibility methods will be positively misleading Syst Zool 1978 27 401 410
Hendy MD Penny D A framework for the quantitative study of evolutionary trees Syst Zool 1989 38 297 309
Huelsenbeck JP Hillis DM Success of phylogenetic methods in the four-taxon case Syst Biol 1993 42 247 264
Huelsenbeck JP The robustness of two phylogenetic methods: four-taxon simulations reveal a slight superiority of maximum likelihood over neighbor joining Mol Biol Evol 1995 12 843 849 7476130
Philippe H Long branch attraction and protist phylogeny Protist 2000 51 307 316 11212891 10.1078/S1434-4610(04)70029-2
Kuhner MK Felsenstein J A simulation comparison of phylogeny algorithms under equal and unequal evolutionary rates Mol Biol Evol 1994 11 459 468 8015439
Gaut BS Lewis PO Success of maximum likelihood phylogeny inference in the four-taxon case Mol Biol Evol 1995 12 152 162 7877489
Swofford DL Waddell PJ Huelsenbeck JP Foster PG Lewis PO Rogers JS Bias in phylogenetic estimation and its relevance to the choice between parsimony and likelihood methods Syst Biol 2001 50 525 539 12116651 10.1080/106351501750435086
Ho SY Jermiin L Tracing the decay of the historical signal in biological sequence data Syst Biol 2004 53 623 637 15371250 10.1080/10635150490503035
Huelsenbeck JP Is the Felsenstein zone a fly trap? Syst Biol 1997 46 69 74 11975354
Huelsenbeck JP Systematic bias in phylogenetic analysis: is the Strepsiptera problem solved? Syst Biol 1998 47 519 537 12066692
Anderson FE Swofford DL Should we be worried about long-branch attraction in real data sets? Investigations using metazoan 18S rDNA Mol Phylogenet Evol 2004 33 440 451 15336677 10.1016/j.ympev.2004.06.015
Bergsten J A review of long-branch attraction Cladistics 2005 21 163 193 10.1111/j.1096-0031.2005.00059.x
Whelan S Lio P Goldman N Molecular phylogenetics: state-of-the-art methods for looking into the past Trends Genet 2001 17 262 272 11335036 10.1016/S0168-9525(01)02272-7
Holder M Lewis PO Phylogeny estimation: Traditional and Bayesian approaches Nat Rev Genet 2003 4 275 284 12671658 10.1038/nrg1044
Philippe H Delsuc F Brinkmann H Lartillot N Phylogenomics Annu Rev Ecol Evol Syst 2005 in press
Siddall ME Success of parsimony in the four-taxon case: long-branch repulsion by likelihood in the Farris zone Cladistics 1998 14 209 220 10.1111/j.1096-0031.1998.tb00334.x
Jukes TH Cantor CR Munro HN Evolution of protein molecules Mammalian protein metabolism 1969 New York, Academic Press 21 132
Uzzell T Corbin KW Fitting discrete probability distributions to evolutionary events Science 1971 172 1089 1096 5574514
Lartillot N Philippe H A Bayesian mixture model for across-site heterogeneities in the amino-acid replacement process Mol Biol Evol 2004 21 1095 1109 15014145 10.1093/molbev/msh112
Pagel M Meade A A phylogenetic mixture model for detecting pattern-heterogeneity in gene sequence or character-state data Syst Biol 2004 53 571 581 15371247 10.1080/10635150490522232
Sullivan J Swofford DL Should we use model-based methods for phylogenetic inference when we know that assumptions about among-site rate variation and nucleotide substitution pattern are violated? Syst Biol 2001 50 723 729 12116942 10.1080/106351501753328848
Fitch WM Markowitz E An improved method for determining codon variability in a gene and its application to the rate of fixation of mutations in evolution Biochem Genet 1970 4 579 593 5489762 10.1007/BF00486096
Philippe H Lopez P On the conservation of protein sequences in evolution Trends in Biochemical Sciences 2001 26 414 416 11469218 10.1016/S0968-0004(01)01877-1
Lopez P Casane D Philippe H Heterotachy, an important process of protein evolution Mol Biol Evol 2002 19 1 7 11752184
Ane C Burleigh JG McMahon MM Sanderson MJ Covarion structure in plastid genome evolution: a new statistical test Mol Biol Evol 2005 22 914 924 15625184 10.1093/molbev/msi076
Lockhart PJ Larkum AW Steel M Waddell PJ Penny D Evolution of chlorophyll and bacteriochlorophyll: the problem of invariant sites in sequence analysis Proc Natl Acad Sci USA 1996 93 1930 1934 8700861 10.1073/pnas.93.5.1930
Lopez P Forterre P Philippe H The root of the tree of life in the light of the covarion model J Mol Evol 1999 49 496 508 10486007
Lockhart PJ Huson D Maier U Fraunholz MJ Van De Peer Y Barbrook AC Howe CJ Steel MA How molecules evolve in Eubacteria Mol Biol Evol 2000 17 835 838 10779544
Philippe H Germot A Phylogeny of eukaryotes based on ribosomal RNA: long-branch attraction and models of sequence evolution Mol Biol Evol 2000 17 830 834 10917801
Inagaki Y Susko E Fast NM Roger AJ Covarion shifts cause a long-branch attraction artifact that unites Microsporidia and Archaebacteria in EF-1{alpha} phylogenies Mol Biol Evol 2004 21 1340 1349 15034136 10.1093/molbev/msh130
Lockhart PJ Novis P Milligan BG Riden J Rambaut A Larkum AW Heterotachy and tree building: a case study with plastids and eubacteria Mol Biol Evol 2005 Published in Advance Access on September 8, 2005.
Kolaczkowski B Thornton JW Performance of maximum parsimony and likelihood phylogenetics when evolution is heterogeneous Nature 2004 431 980 984 15496922 10.1038/nature02917
Gadagkar SR Kumar S Maximum likelihood outperforms maximum parsimony even when evolutionary rates are heterotachous Mol Biol Evol 2005 22 2139 2141 16014870
Spencer M Susko E Roger AJ Likelihood, parsimony, and heterogeneous evolution Mol Biol Evol 2005 22 1161 1164 15746012 10.1093/molbev/msi123
Gaucher EA Miyamoto MM A call for likelihood phylogenetics even when the process of sequence evolution is heterogeneous Mol Phylogenet Evol 2005 in press
Brinkmann H van der Giezen M Zhou Y Poncelin de Raucourt G Philippe H An empirical assessment of long branch attraction artifacts in phylogenomics Syst Biol 2005 54 743 757 16243762
Philippe H Lartillot N Brinkmann H Multigene analyses of bilaterian animals corroborate the monophyly of Ecdysozoa, Lophotrochozoa, and Protostomia Mol Biol Evol 2005 22 1246 1253 15703236 10.1093/molbev/msi111
Matte-Tailliez O Brochier C Forterre P Philippe H Archaeal phylogeny based on ribosomal proteins Mol Biol Evol 2002 19 631 639 11961097
Brochier C Bapteste E Moreira D Philippe H Eubacterial phylogeny based on translational apparatus proteins Trends Genet 2002 18 1 5 11750686 10.1016/S0168-9525(01)02522-7
Meunier J Lopez P Casane D Philippe H A versatile method for detecting heterotachous sites. Evolutionary Bioinformatics Submitted
Rodriguez-Ezpeleta N Brinkmann H Burey SC Roure B Burger G Loffelhardt W Bohnert HJ Philippe H Lang BF Monophyly of primary photosynthetic eukaryotes: green plants, red algae, and glaucophytes Curr Biol 2005 15 1325 1330 16051178 10.1016/j.cub.2005.06.040
Fitch WM The nonidentity of invariable positions in the cytochromes c of different species Biochem Genet 1971 5 231 241 5088800 10.1007/BF00485794
Dayhoff MO Eck RV Park CM Dayhoff MO A model of evolutionary change in proteins Atlas of protein sequence and structure 1972 5 Washington, DC, National Biomedical Research Fundation 89 99
Kimura M A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences J Mol Evol 1980 16 111 120 7463489 10.1007/BF01731581
Lockhart PJ Howe CJ Bryant DA Beanland TJ Larkum AW Substitutional bias confounds inference of cyanelle origins from sequence data J Mol Evol 1992 34 153 162 1556750 10.1007/BF00182392
Felsenstein J Inferring phylogenies 2004 Sunderland, MA, USA, Sinauer Associates, Inc. 645
Lockhart PJ Steel MA Barbrook AC Huson D Charleston MA Howe CJ A covariotide model explains apparent phylogenetic structure of oxygenic photosynthetic lineages Mol Biol Evol 1998 15 1183 1188 9729882
Steel M Should phylogenetic models be trying to 'fit an elephant'? Trends Genet 2005 21 307 309 15922824 10.1016/j.tig.2005.04.001
Thornton JW Kolaczkowski B No magic pill for phylogenetic error Trends Genet 2005 21 310 311 15922825 10.1016/j.tig.2005.04.002
Rambaut A Grassly NC Seq-Gen: an application for the Monte Carlo simulation of DNA sequence evolution along phylogenetic trees Comput Appl Biosci 1997 13 235 238 9183526
Swofford DL PAUP*: Phylogenetic Analysis Using Parsimony and other methods 2000 4b10 , Sinauer, Sunderland, MA
The R Project for Statistical Computing
|
16209710
|
PMC1274308
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Oct 6; 5:50
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-50
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-531621265810.1186/1471-2148-5-53Research ArticleThe key role for local base order in the generation of multiple forms of China HIV-1 B'/C intersubtype recombinants Zhang Chi-Yu [email protected] Ji-Fu [email protected] Shao-Heng [email protected] Department of Biochemistry and Molecular Biology, Jiangsu University School of Medical Technology, Zhenjiang, Jiangsu 212001, China2 Allergy and Inflammation Research Institute, the Medical College of Shantou University, Shantou, Guangdong, 515031, China2005 7 10 2005 5 53 53 19 2 2005 7 10 2005 Copyright © 2005 Zhang et al; licensee BioMed Central Ltd.2005Zhang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
HIV-1 is a retrovirus with high rate of recombination. Increasing experimental studies in vitro indicated that local hairpin structure of RNA was associated with recombination by favoring RT pausing and promoting strand transfer. A method to estimate the potential to form stem-loop structure by calculating the folding of randomized sequence difference (FORS-D) has been used to investigate the relationship between secondary structure and evolutionary pressure in some genome. It showed that gene regions under strong positive "Darwinian" selection were associated with positive FORS-D values. In the present study, the sequences of HIV-1 subtypes B' and C, both of which represent the parent strains of CRF07_BC, CRF08_BC and China URFs, were selected to investigate the relationship between natural recombination and secondary structure by calculating the FORS-D values.
Results
The apparent higher negative FORS-D value region appeared in the gag-pol gene region (nucleotide 0–3000) of HIV-1 subtypes B' and C. Thirteen (86.7 %) of 15 mosaic fragments and 17 (81 %) of 21 recombination breakpoints occurred in this higher negative FORS-D region. This strongly suggested that natural recombination did not occur randomly throughout the HIV genome, and that there might be preferred (or hot) regions or sites for recombination. The FORS-D analysis of breakpoints showed that most breakpoints of recombinants were located in regions with higher negative FORS-D values (P = 0.0053), and appeared to have a higher negative average FORS-D value than the whole genome (P = 0.0007). The regression analysis also indicated that FORS-D values correlated negatively with breakpoint overlap.
Conclusion
High negative FORS-D values represent high, base order determined stem-loop potentials and influence mainly the formation of stem-loop structures. Therefore, the present results suggested for the first time that occurrence of natural recombination was associated with high base order-determined stem-loop potential, and that local base order might play a key role in the initiation of natural recombination by favoring the formation of stable stem-loop structures.
==== Body
Background
The human immunodeficiency virus type 1 (HIV-1) is a complex retrovirus, which encodes the enzyme reverse transcriptase (RT), and exhibits high mutation rates due to the lack of the DNA proofreading activity of the viral RT. HIV-1 genome is diploid, containing two plus-strand viral RNA copies that can be identical. In the process of viral DNA synthesis, template switching occurs by translocation of RT between two genomic RNAs, and results in both intra-molecular and inter-molecular recombination. If dual infections or superinfection with different strains or subtypes of HIV-1 occurs, two different RNA templates might be co-packaged into one virion, yielding a heterozygous virion. In a subsequent infection cycle, RT may switch from one template (the donor) to the other (the acceptor), producing a mosaic HIV-1 genome [1,2]. HIV-1 has high potential to form recombination variants [3,4]. The high rate of recombination is due to the frequent template switching of RT. At least 2.8 template switching events occur per genome per replication cycle was estimated previously [5]. Genetic recombination and point mutation are both important strategies to increase viral diversity, which allow HIV-1 to escape immune attack and to develop possibly drug-resistant variants [6].
Retroviral recombination generally occurs during minus-strand DNA synthesis [7]. The "Dock and Lock" model had been proposed to shed light on the mechanism of retroviral recombination. This model suggested that RT switches templates when it encounters palindrome (hairpin) structures that can induce RT to pause. RT pausing during synthesis can enhance strand transfer [1,2,8]. RNA secondary structures play an important role in the function of an RNA molecule, such as RNA-protein interactions, transcription, translation, and so on. Previous studies in vitro have indicated that specific RNA secondary structures were associated with strand transfer by favoring RT pausing [9,10]. However, it remains uncertain whether RNA secondary structure is involved in the generation of circulating HIV-1 recombinants.
Currently, some HIV-1 recombination variants have been identified worldwide [6]. Sixteen prevalent inter-subtype recombinants were recognized as circulating recombinant forms (CRFs) from 01 to 16, respectively [11]. Three CRFs, CRF01_AE, CRF07_BC and CRF08_BC were found in China. Of them, CRF07_BC and CRF08_BC possibly arose in Yunnan Province, and had circulated widely among injecting drug users (IDUs) [12-16]. In addition, the unique recombinant forms (URFs), between subtypes B' (Thailand variant of subtype B) and C, are epidemic among IDUs in Dehong Prefecture in western Yunnan, suggesting on-going generation of new HIV-1 intersubtype recombinants [14,15]. Most HIV-1 infected IDUs in China were unemployed, and never received any antiretroviral therapy due to lack of income [16]. Therefore, there is no drug selective pressure associated with generation of recombinants in China, and these recombinants represent the occurrence of natural recombination.
The stem-loop structure is the most important secondary structure of RNA. A method to estimate the potential to form stem-loop structure by calculating FORS-D has been used to investigate the relationship between secondary structure and evolutionary pressure [17,18]. Previous studies by Forsdyke showed that gene regions under strong positive "Darwinian" selection were associated with positive FORS-D values, reflecting the conflict between stem-loop structure potential and specific protein function [17,19-21]. In addition, our previous work found that the FORS-D values correlated negatively with ccr5 gene deletions, indicating that stem-loop structure influences the deletion [22]. These suggested that stem-loop structures might play an important role in mutation strategies and gene evolution. Therefore, in the present study, we selected China CRFs and URFs as a means to investigate the relationship between the secondary structure and natural recombination by analyzing the FORS-D values of HIV-1 genome.
Results and discussion
The distribution of FORS-D values in HIV subtype B' and C genomes
Previous studies have analyzed the local secondary structural information of some HIV-1 strains by calculating the "statistically significant" stem-loop potential, and found that different regions of HIV-1 genome had different potential to form stem-loop structures [23,24]. The regions with high stem-loop potential were generally associated with the interaction between local secondary structures and corresponding protein factors. For example, trans-acting responsive element (TAR) and Rev-responsive element (RRE), both of which are recognized by the Tat protein and Rev protein respectively, have more stable local secondary structures than other regions of HIV-1 genome [23-25]. A negative correlation between "statistically significant" stem-loop potential and sequence variability (substitutions) was observed in the HIV-1 genome. In the regions with higher negative FORS-D values, indicating that base order favors stem-loop potential, the rate of base substitutions tend to be lower. Contrarily, higher positive FORS-D values decrease stem-loop potential and is functionally important, because the rate of base substitutions increases [20,23-25].
Genetic recombination is another important pathway to generate variability for HIV-1. Previous studies found that local stem-loop structure enhanced the occurrence of template switching of RT [2,10,26]. To assess whether local stem-loop structure is involved in the generation of natural HIV-1 recombination, FORS-D analysis was applied to estimate the potential of HIV-1 sequences to form stem-loop structures. The FORS-D value represents a base order-determined stem-loop potential, and provides a measure of the contribution of base order alone to the formation of stem-loop structure [18]. The FONS value determines the trend of total stem-loop potential.
Yunnan Province of China has a high HIV-1 prevalence among IDUs and generates multiple forms of HIV-1 intersubtype recombinants [14,16]. Because most HIV-1 infected IDUs are unemployed, they almost never receive any antiretroviral therapy [16]. Therefore, the recombinant forms circulated among IDUs indicate the occurrence of the natural recombination between HIV-1 subtypes B' and C without drug selective pressure. To investigate the relationship between local stem-loop potential and natural recombination, two closely related HIV-1 strains, 95IN21068 and RL42, which are known to be parent strains of China inter-subtype B' and C recombinants, were selected as objectives to analyze FORS-D.
FONS and FORS-M values found in RL42 and 95IN21068 are shown in part A and B of Figure 1, respectively. Both HIV-1 strains appeared to have similar trends in FONS and FORS-M values. FORS-M values were relatively constant (For subtype B': median value was -35. 79 kcal/mol with a range from -65.78 to -19.6 kcal/mol; for subtype C: median value was -36.42 kcal/mol, with a range from -65.6 to -20.25 kcal/mol). However, a large fluctuation was observed with FONS values. Several higher negative FONS values appeared in 5' and 3' termini, 3' end of the gag gene, and around nucleotide 7250 in the env gene. These results were consistent with the previous observations in other HIV-1 strains [20,23-25]. Figure 1C shows FORS-D values for both 95IN21068 and RL42. Except for a few regions, HIV-1 subtype B' and C genomes appeared to have similar distribution of FORS-D values. The windows with higher negative FONS values accordingly appeared to have higher negative FORS-D values. The fluctuations of FONS values observed in the whole genome of the both HIV-1 subtypes were largely base-order dependent, as reflected in the FORS-D values (Fig. 1A, B and 1C ). The highest negative FORS-D values of both sequences occurred around nucleotide 7250 in RRE region (nucleotide 7081–7432).
Figure 1 FORS-D analysis of HIV-1 subtypes B' and C. A, FONS and FORS-M values of HIV-1 subtypes B'; B, FONS and FORS-M values of HIV-1 subtypes C; C, FORS-D values (± SE) of both HIV-1 subtypes B' and C; D, the location of all breakpoints and mosaic pattern of four full-length HIV-1 recombinants circulated in China. FONS, FORS-M and FORS-D values were calculated in successive 200 nt windows, each of which overlapped the previous window by 150 nt. Vertical dashed lines indicated the location of breakpoints in FORS-D distributions. The red open circles in FORS-D distributions (C) pointed out these recombination breakpoints located in regions with higher negative FORS-D values. The solid triangle indicated the breakpoints shared by more than one HIV-1 recombinant form (D). The inserted recombined fragments were numbered from 1 to 9 (D).
For each HIV-1 subtype, the distribution of FORS-D values differed in different regions of the gene. The apparent higher negative FORS-D value region occurred in the gag gene and at the 5' end of the pol gene (nucleotide 0–3000) (For subtype B': gag-pol region: -5.403 ± 0.8155 kcal/mol; whole genome: -3.165 ± 0.4886 kcal/mol, P = 0.0217. For subtype C: gag-pol region: -6.495 ± 0.6398 kcal/mol; whole genome: -3.775 ± 0.5035 kcal/mol, P = 0.0045). However, an intense fluctuation of positive and negative FORS-D values around the abscissa was observed in the region from the 3' part of pol gene to env gene (nucleotide 3000–8000). This region encodes RT, integrase, envelope glycoproteins gp120 and gp41, other important regulatory (Tat and Rev) and accessory (Vpr, Vif, Vpu, and Nef) proteins. They determine HIV-1 replication and efficient infection, and are exposed immediately to the human immune system and under strong positive "Darwinian" selection [27]. Previous studies on retroviral genes [19], MHC genes [20], snake venom phospholipase A2 [21] and other genes [17], had showed that a region under strong positive selection exhibited generally positive FORS-D values. Our results supported the observation that FORS-D value was associated with evolutionary pressure [17,19-21].
The relationship between the HIV-1 B'/C intersubtype recombination and stem-loop potential
In vitro studies using the HIV-1 derived vector system indicated that HIV-1 genome has high rate of recombination and hot spots for recombination occurrence [3,5]. The hot spots were located in stable hairpin structures [4]. However, the previous observation did not indicate whether the occurrence of HIV-1 CRFs and URFs in worldwide distribution correlates with secondary structures of RNA templates due to sequence difference between vector and circulated strains. To assess whether stem-loop structures are involved in occurrence of natural recombination, we selected subtype B' RL42 and subtype C 95IN21068, both of which represent the parent strains of existing CRFs and URFs in China, to carry out FORS-D analysis. Currently, besides CRF07 and 08, only two other full-length sequences of URFs circulated in China are available. These four existing recombinants, representing four different recombination variants, were selected and analyzed using the Simplot software. Their mosaic patterns are shown in Fig. 1D. The breakpoints of recombination are identified in the FORS-D distribution of RL42 and 95IN21068 (Fig. 1C and 1D) by fine vertical dashed lines. In addition to these four full-length sequences, other URFs were also analyzed despite the availability of only gag-RT region (about 2600 nucleotides) [14]. Figure 2 shows their partial mosaic map and FORS-D distribution in breakpoints.
Figure 2 The location of all breakpoints of seven 2600-nt URFs circulated in China in FORS-D distributions of HIV-1 subtypes B' and C. A, FORS-D values (± SE) of both HIV-1 subtypes B' and C; B, the location of all breakpoints. FORS-D values were calculated in successive 200 nt windows, each of which overlapped the previous window by 150 nt. The positions of recombination were shown as boxes in B. Vertical dashed lines indicated the location of breakpoints in FORS-D distributions. The solid triangle indicated the breakpoints shared by more than one HIV-1 recombinant form (B). The numbers of inserted recombined fragments were continued from 10 to 15 (B). The fragment numbers occurred in Fig. 1D were not shown.
In total, 15 different inserted recombined fragments were identified in China HIV-1 B'/C intersubtype recombinants (Fig. 1D and 2B) [12-15]. Thirteen (86.7 %) of these mosaic fragments occurred in the higher negative FORS-D value region (nucleotide 0–3000) of parent genomes, where the gag gene and 5' end of the pol gene are located. On the other hand, because several shared breakpoints were observed in these mosaic molecules, which were confirmed by our previous reports [14,15], 15 mosaic fragments only contained 21 unique breakpoints. For example, fragment 5, 6, 7 and 13 shared the 3' end breakpoint. Among these breakpoints, 17 (81 %) also located in this higher negative FORS-D value region. This strongly indicates that natural recombination did not occur randomly throughout the HIV genome, and that there might be preferred (or hot) regions or sites for recombination [3,5]. These observations suggest an association between recombination and high negative FORS-D values (Fig. 1C and Fig. 2A).
In order to further confirm the relationship between recombination and high negative FORS-D values, FORS-D values of corresponding breakpoints of parent sequences were calculated as described in the Methods. The FORS-D values of breakpoints of 15 fragments were shown in Table 1. For most fragments, at least one breakpoint of each fragment was found to be located in regions with higher negative FORS-D values. Two exceptions were fragment 11 and 12. They had at least one breakpoint located in the regions of higher negative FORS-D values in one parent subtype, and at least one breakpoint located in the regions of negative FORS-D values (close to the mean of whole genome) in another parent subtype (Table 1). Twenty-one breakpoints appeared to more favor occurring in higher negative FORS-D values region (69 %, P = 0.0053), and negative FORS-D values region (92.9 %, P = 0.0007). In addition, the average FORS-D values of breakpoints (-6.29 ± 0.81 kcal/mol) also appeared to be more negative than whole genome (-3.47 ± 0.35 kcal/mol) (P = 0.0079), suggesting that the values of breakpoints were significantly different from that of whole genomes. The data indicated that recombination preferentially occurred in high negative FORS-D regions.
Table 1 The FORS-D values of breakpoints of inserted recombined fragments occurred in China HIV-1 B'/C intersubtype recombinants.
No. fragment# Recombinant strains Subtype B'* (kcal/mol) Subtype C* (kcal/mol)
Left (5'-) Right (3'-) Left (5'-) Right (3'-)
1 CRF07, HH069, HH086 3.04 -8.12 -7.07 -8.63
2 CRF08 -1.52 -6.38 -7.00 -9.20
3 CRF07 -19.88 -10.43 -14.15 -15.27
4 HH069 3.48 -10.43 -3.76 -15.27
5 CRF08, HH029 -3.97 -13.29 -5.09 -18.91
6 HH069 -7.15 -13.29 -9.62 -18.91
7 CRF07, HH086 -1.07 -13.29 -0.37 -18.91
8 CRF07 -0.92 -5.56 -1.32 -8.10
9 CRF07, CRF08, HH069, HH086 -9.94 -5.19 -5.97 -6.02
10 DH003, DH015 -0.90 -9.30 -2.98 -3.88
11 DH012 -0.90 -10.70 -2.98 -3.67
12 DH012 4.13 -2.93 -5.32 -7.27
13 DH015 -7.15 -13.29 -9.62 -18.91
14 DH016 -19.88 -10.43 -14.15 -15.27
15 DH008 -19.88 -3.97 -14.15 -5.09
# The fragment numbers were consistent with those of Fig. 1D and 2B. * The FORS-D values of breakpoints were calculated as described in the Methods. The average FORS-D values of subtype B' and C HIV-1 genome were -3.17 and -3.78, respectively. The value of < -3.17 (subtype B') or < -3.78 (subtype C) represents the region with higher negative FORS-D value.
In vitro experiments indicated that strand transfer of RT involved RT pausing and triggering retroviral recombination [1,26]. Further evidence showed that secondary structures of RNA template, especially, hairpin or stem-loop structures play a key role in RT pausing and strand transfer [2,10,26]. Two obligatory strand transfers of RT had been observed to occur in terminal sequences of viral genome with stable hairpin structures and high stem-loop potential [9]. The hairpin structure facilitates RT pausing, which stimulates RT-RNase H activity and results in donor template degradation. Pause-induced donor template degradation initiates strand transfer. Then, strand transfer is thought to progress through a two-step mechanism, first acceptor invasion, then primer terminus transfer. Two models, the kissing hairpin interaction model and the "Dock and Lock" model, have been proposed to explain this mechanism [1,26]. Both models emphasize the key role of hairpin structure in strand transfers. High negative FORS-D value, representing high base order-determined stem-loop potential, occurred generally in one or both acceptor and donor sites (Fig. 1C, 2A, and Table 1), which was supported by previous in vitro observation [2].
Further evidence for the location of breakpoints was provided by plotting FONS, FORS-M, and FORS-D values of each window against its percentage of breakpoint overlap. Figure 3 shows the linear regression analysis of the relationships between FONS (A), FORS-M (B), and FORS-D (C) values (kcal/mol) and breakpoint overlap. In form Fig. 3, we observed that plots for FORS-M were horizontal(r = 0.0003), and plots for both FONS and FORS-D slopes were slightly diagonal (r = 0.053 and 0.061, respectively). The slopes of the least-squares regression lines for FONS and FORS-D values were significantly greater than zero (P = 0.006 for FONS and P < 0.0001 for FORS-D). The regression analysis indicated that FONS and FORS-D values correlated negatively with breakpoint overlap, and that the correlation coefficient, as well as P value, also supported these relationships (Fig. 3). FONS value represents the total stem-loop potential, and FORS-D provides a measure of the contribution of base order alone to the stem-loop potential of a sequence. Negative FORS-D values imply that local base order favors the formation of stable stem-loop structures. Therefore, the observation above suggested for the first time that occurrence of natural recombination was associated with high base order-determined stem-loop potential, and local base order was likely to be important for the initiation of natural recombination by favoring the formation of stable hairpin structures [1]. In addition, the present results supported the previous observation that hairpin structures were involved in retroviral recombination [2,10,26]. However, a further study to extend the FORS-D analysis to other intersubtype recombinants (CRFs and URFs) circulated in other countries and regions of the world should be conducted.
Figure 3 Linear regression analysis of the FONS (A), FORS-M (B) and FORS-D (C) value against the degree of the breakpoint overlap (%). Data (FONS, FORS-M and FORS-D) were from the breakpoints of four full-length, HIV-1 representative recombinant forms circulated in China. Parameters of the least-squares line shown in each figure were slope (s), the correlation coefficient (r), and the probability (P value) that the slope of the line is not significantly different from zero (P).
Does the information of local base order expressed by FORS-D values play a key role in the selection of evolutionary strategies for reducing the selective pressure?
Two CRFs and multiple forms of URFs have been detected in Yunnan Province of China, where needle or syringe sharing among IDUs is popular [14,16]. Needle or syringe sharing increase the risk of dual virus infections and subsequent recombination between different subtypes of viruses. To date, however, little is known about whether or not these new recombinants are associated with stronger infectivity and higher replication ability. We do know that the location of natural recombination is not random throughout the HIV genome, and there are some specific hot spots for recombination situated in the genome. Previous experimental observations in vitro revealed that hairpin structures increase the rate of recombination between viruses [2,8,10,26]. Here, our results support that previous notion, and indicated firstly that natural recombination was associated primarily with high base order-determined stem-loop potential (FORS-D values).
FORS-D value tends to fluctuate around zero and determines the trend of FONS value. It contains a large amount of evolutionary information about nucleic acids, and generally appears to be negative number [17]. Negative FORS-D values favor the formation of stem-loop structure and are widely distributed in long genomic segments from a variety of species. Positive FORS-D values, represent the conflict of evolutionary pressures on base order, and occur more frequently in regions under positive Darwinian selection, such as promoter regions, exons and so on [17,19-21]. The regions with positive FORS-D values indicate a tendency for local base order to support protein encoding function rather than formation of stem-loop structure. Therefore, FORS-D value generally appears to correlate positively with base substitution densities and dN/dS ratio [17]. On the other hand, our previous and present studies showed that FORS-D values were associated with the occurrence of deletion [22] and recombination mutations, which suggests that local base order is involved in the occurrence of both recombination and deletion. Because positive FORS-D values correlate with high ratio of base substitution and negative FORS-D values for recombination or deletion, we can deduce that local base order plays a critical role in the selection of gene evolutionary pathways. However, further evidence is required to support this hypothesis.
Conclusion
By analyzing the FORS-D values of HIV-1 subtypes B' and C, both of which represent the parent strains of CRF07_BC, CRF08_BC and China URFs [12-14], we found that most breakpoints of these recombinants were located in regions with higher negative FORS-D values, and appeared to have a higher negative average FORS-D value than for the whole genome. The regression analysis indicated further that FORS-D values correlated negatively with breakpoint overlap. These results suggested for the first time that occurrence of natural recombination was associated with high base order-determined stem-loop potential, and that local base order might play a key role in the initiation of natural recombination of HIV-1 by favoring the formation of stable stem-loop structures. Combining with previous reports that FORS-D correlates positively with the ratio of base substitution [17,19-21], we could deduce that local base order might play a critical role in the selection of gene or genome evolutionary pathways, and determines the evolutionary strategies adopted by gene or genome to reduce the selective pressure.
Methods
Sequences and recombination analysis
In this study, the sequences of RL42, 97CN001, 97CNGX-7F and URFs, were retrieved from GenBank (Table 2). RL42, 97CN001 and 97CNGX-7F are generally used as representative strains of China HIV-1 subtypes B [28], CRF07_BC [12] and CRF08_BC [13], respectively. Sequences of RL42, 97CN001, 97CNGX-7F and URFs were aligned with subtype reference sequences using Clustal X 1.8 [29]. The full-length sequences of subtype reference isolates are available in the Los Alamos database [11]. The recombination breakpoints were analyzed by using Simplot software (version 2. 5) [30]. The parameters were used as follows: Window size: 200 bps; Step size: 10 bps; Tree algorithm: Neighbor; Distance model: Kimura; Bootstrap replicate: 100; Reference type: 50 % consensus.
Table 2 Sequences of HIV-1 strains used in the present study.
Subtype Sequence name GenBank accession number References
B' RL42 U71182 [28]
CRF07_BC 97CN001 AF286226 [12]
CRF08_BC 97CNGX-7F AY008716 [13]
URF HH086 AP005207 [15]
URF HH069 AP005206 [15]
URF HH004 AB090998 [15]
URF HH029 AB090999 [15]
URF DH003 AB078705 [14]
URF DH008 AB078707 [14]
URF DH012 AB078710 [14]
URF DH015 AB078712 [14]
URF DH016 AB078713 [14]
FORS-D analysis of HIV-1 subtypes B'and C
The previous studies had described the application of FORS-D analysis in detail [17,18]. In brief, two factors, base composition and base order, contribute to stem-loop formation of a nucleic acid molecule. So the total stem-loop potential of a sequence can be divided into the contribution of base composition alone and the contribution of base order alone to form stem-loop structure. For a natural sequence, FONS, FORS-M and FORS-D represent total stem-loop potential, base composition-determined stem-loop potential and base order-determined stem-loop potential, respectively. FONS values are calculated by using computer program, RNAstructure (version 3.6) [31], which based on free energy minimization to find a theoretical optimum secondary structure. FORS-M value is the mean minimum free energy value of the 10 randomised sequences generated from the same window. Ten randomised sequences were obtained using the shuffle program included in the on-line software package SMS [32]. FORS-D is the difference between FONS and FORS-M, and closely corresponds to "statistically significant" stem-loop potential developed by Le et al. for analyzing potential RNA folded substructures [33].
Because CRF07_BC, CRF08_BC and most URFs are B'/C inter-subtype recombinants with mostly subtype C and a few small subtype B' segments [12-14], we selected RL42 and 95IN21068 for FORS-D analysis. RL42 and 95IN21068 are related closely to those recombinants in phylogenetic evolution and generally used as a subtypes B' and C sequence reference, respectively. Both sequences lack the 5' long terminal repeats (LTRs) and are available from GenBank under accession numbers AF067155 and U71182, respectively. To analyse the FORS-D, each sequence was divided into 177 successive 200-nucleotide windows. Each overlapped the previous window by 150 nucleotides. For each 200-nucleotide window, both FONS value and FORS-M value were calculated using RNAstructure software. Then, the difference between FONS and FORS-M (FONS less FORS-M) generated FORS-D value.
FORS-D analysis of recombination breakpoints
The position of each breakpoint was obtained using the Simplot software (version 2.5) with 200-nucleotide window size and 10-nucleotide step size. So three successive 200-nucleotide windows were selected to calculate FORS-D values of recombination breakpoints. The position of each breakpoint was used as the center of the middle window, and each window overlapped the previous one by 190 nucleotides. The average of three windows represents the FORS-D value of each recombination breakpoint.
Statistical analysis
All statistical analysis was conducted using GraphPad Prism version 2.00 (Biomedical Sciences, Creighton University). The FORS-D location of breakpoints was analyzed by the χ2 test. The differences of FORS-D values of recombination breakpoints and gag-pol gene region with whole genomes were compared by the t test.
Abbreviations
HIV-1, human immunodeficiency virus type 1; RT, reverse transcriptase; IDUs, injecting drug users; CRFs, circulating recombinant forms; URFs, unique recombinant forms; FORS-D, folding of randomized sequence difference; FONS, folding of natural sequence; FORS-M, folding of randomized sequence mean; TAR, trans-acting responsive element; RRE, Rev-responsive element; dN/dS, nonsynonymous-to-synonymous rate ratio.
Authors' contributions
CYZ and JFW conceived and designed the study, performed the collection and bioinformatic analysis of the data; CYZ drafted the manuscript. SHH supervised and coordinated the whole project. All authors have read and approved the final manuscript.
Acknowledgements
The authors wish to acknowledge the assistance of Dr. Lorna Grant, University of Manitoba, Canada, in the preparation of this manuscript. This work was supported by the Major State Basic Research Program of China (973 Program, No. 2001CB510009), Grants from Li Ka Shing Foundation, Hong Kong, China (No. C0200001), Guiding Programs of Jiangsu Province for Natural Scicence Research in Colleges and Universities (No. 04KJD180044), and the Science Foundation of Jiangsu University for Advanced Scholars (No. 2281270002).
==== Refs
Roda RH Balakrishnan M Kim JK Roques BP Fay PJ Bambara RA Strand transfer occurs in retroviruses by a pause-initiated two-step mechanism J Biol Chem 2002 277 46900 46911 12370183 10.1074/jbc.M208638200
Jetzt AE Yu H Klarmann GJ Ron Y Preston BD Dougherty JP High rate of recombination throughout the human immunodeficiency virus type 1 genome J Virol 2000 74 1234 1240 10627533 10.1128/JVI.74.3.1234-1240.2000
Moumen A Polomack L Unge T Veron M Buc H Negroni M Evidence for a mechanism of recombination during reverse transcription dependent on the structure of the acceptor RNA J Biol Chem 2003 278 15973 15982 12595540 10.1074/jbc.M212306200
Mikkelsen JG Lund AH Duch M Pedersen FS Mutations of the kissing-loop dimerization sequence influence the site specificity of murine leukemia virus recombination in vivo J Virol 2000 74 600 610 10623721 10.1128/JVI.74.2.600-610.2000
Zhuang J Jetzt AE Sun G Yu H Klarmann G Ron Y Preston BD Dougherty JP Human immunodeficiency virus type 1 recombination, rate, fidelity, and putative hot spots J Virol 2002 76 11273 11282 12388687 10.1128/JVI.76.22.11273-11282.2002
Najera R Delgado E Perez-Alvarez L Thomson MM Genetic recombination and its role in the development of the HIV-1 pandemic AIDS 2002 S3 16 12698994
Zhang J Tang LY Li T Ma Y Sapp CM Most retroviral recombinations occur during minus-strand DNA synthesis J Virol 2000 74 2313 2322 10666262 10.1128/JVI.74.5.2313-2322.2000
Roda RH Balakrishnan M Hanson MN Wohrl BM Le Grice SF Roques BP Gorelick RJ Bambara RA Role of the Reverse Transcriptase, Nucleocapsid Protein, and Template Structure in the Two-step Transfer Mechanism in Retroviral Recombination J Biol Chem 2003 278 31536 31546 12801926 10.1074/jbc.M304608200
Kim JK Palaniappan C Wu W Fay PJ Bambara RA Evidence for a unique mechanism of strand transfer from the transactivation response region of HIV-1 J Biol Chem 1997 272 16769 16777 9201981 10.1074/jbc.272.27.16769
Berkhout B Vastenhouw NL Klasens BI Huthoff H Structural features in the HIV-1 repeat region facilitate strand transfer during reverse transcription RNA 2001 7 1097 1114 11497429 10.1017/S1355838201002035
HIV Sequence Database
Rodenburg CM Li Y Trask SA Chen Y Decker J Robertson DL Kalish ML Shaw GM Allen S Hahn BH Gao F UNAIDS and NIAID Networks for HIV Isolation and Characterization Near full-length clones and reference sequences for subtype C isolates of HIV type 1 from three different continents AIDS Res Hum Retroviruses 2001 17 161 168 11177395 10.1089/08892220150217247
Piyasirisilp S McCutchan FE Carr JK Sanders-Buell E Liu W Chen J Wagner R Wolf H Shao Y Lai S Beyrer C Yu XF A recent outbreak of human immunodeficiency virus type 1 infection in southern China was initiated by two highly homogeneous, geographically separated strains, circulating recombinant form AE and a novel BC recombinant J Virol 2000 74 11286 11295 11070028 10.1128/JVI.74.23.11286-11295.2000
Yang R Xia X Kusagawa S Zhang C Ben K Takebe Y On-going generation of multiple forms of HIV-1 intersubtype recombinants in the Yunnan Province of China AIDS 2002 16 1401 1407 12131217 10.1097/00002030-200207050-00012
Yang R Kusagawa S Zhang C Xia X Ben K Takebe Y Identification and characterization of a new class of human immunodeficiency virus type 1 recombinants comprised of two circulating recombinant forms, CRF07_BC and CRF08_BC, in China J Virol 2003 77 685 695 12477871 10.1128/JVI.77.1.685-695.2003
Zhang C Yang R Xia X Qin S Dai J Zhang Z Peng Z Wei T Liu H Pu D Luo J Takebe Y Ben K High prevalence of HIV-1 and hepatitis C virus coinfection among injection drug users in the southeastern region of Yunnan, China J Acquir Immune Defic Syndr 2002 29 191 196 11832691
Forsdyke DR A stem-loop "kissing" model for the initiation of recombination and the origin of introns Mol Biol Evol 1995 12 949 958 7476142
Forsdyke DR An alternative way of thinking about stem-loops in DNA. A case study of the human G0S2 gene J Theor Biol 1998 192 489 504 9680722 10.1006/jtbi.1998.0674
Forsdyke DR Reciprocal relationship between stem-loop potential and substitution density in retroviral quasispecies under positive Darwinian selection J Mol Evol 1995 41 1022 1037 8587101
Forsdyke DR Stem-loop potential in MHC genes, a new way of evaluating positive Darwinian selection? Immunogenetics 1996 43 182 189 8575816
Forsdyke DR Conservation of Stem-Loop Potential in Introns of Snake Venom Phospholipase A2 Genes, An Application of FORS-D Analysis Mol Biol Evol 1995 12 1157 1165
Zhang CY Wei JF He SH Local base order influences the origin of ccr5 deletions mediated by DNA slip-replication Biochem Genet 2005 43 229 237 16144300 10.1007/s10528-005-5214-7
Le SY Chen JH Braun MJ Gonda MA Maizel JV Stability of RNA stem-loop structure and distribution of non-random structure in the human immunodeficiency virus (HIV-I) Nucleic Acids Res 1988 16 5153 5168 3387221
Le SY Chen JH Chatterjee D Maizel JV Sequence divergence and open regions of RNA secondary structures in the envelope regions of the 17 human immunodeficiency virus isolates Nucleic Acids Res 1989 17 3275 3288 2726458
Le SY Malim MH Cullen BR Maizel JV A highly conserved RNA folding region coincident with the Rev response element of primate immunodeficiency viruses Nucleic Acids Res 1990 18 1613 1623 2326200
Balakrishnan M Roques BP Fay PJ Bambara RA Template dimerization promotes an acceptor invasion-induced transfer mechanism during human immunodeficiency virus type 1 minus-strand synthesis J Virol 2003 77 4710 4721 12663778 10.1128/JVI.77.8.4710-4721.2003
Shpaer EG Mullins JI Rates of amino acid change in the envelope protein correlate with pathogenicity of primate lentiviruses J Mol Evol 1993 37 57 65 8395604 10.1007/BF00170462
Graf M Shao Y Zhao Q Seidl T Kostler J Wolf H Wagner R Cloning and characterization of a virtually full-length HIV type 1 genome from a subtype B'-Thai strain representing the most prevalent B-clade isolate in China AIDS Res Hum Retroviruses 1998 14 285 288 9491920
Jeanmougin F Thompson JD Gouy M Higgins DG Gibson TJ Multiple sequence alignment with Clustal X Trends Biochem Sci 1998 23 403 405 9810230 10.1016/S0968-0004(98)01285-7
Salminen MO Carr JK Burke DS McCutchan FE Identification of breakpoints in intergenotypic recombinants of HIV type 1 by bootscanning AIDS Res Hum Retroviruses 1995 11 1423 1425 8573403
Mathews DH Sabina J Zuker M Turner DH Expanded Sequence Dependence of Thermodynamic Parameters Improves Prediction of RNA Secondary Structure J Mol Biol 1999 288 911 940 10329189 10.1006/jmbi.1999.2700
The Sequence Manipulation Suite
Le SY Chen JH Maizel JV Thermodynamic stability and statistical significance of potential stem-loop structures situated at the frameshift sites of retroviruses Nucleic Acids Res 1989 17 6143 6152 2549508
|
16212658
|
PMC1274309
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Oct 7; 5:53
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-53
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-541622130110.1186/1471-2148-5-54Research ArticleMammalian BEX, WEX and GASP genes: Coding and non-coding chimaerism sustained by gene conversion events Winter Eitan E [email protected] Chris P [email protected] MRC Functional Genetics Unit, University of Oxford, Department of Human Anatomy and Genetics, South Parks Road, Oxford OX1 3QX, UK2005 12 10 2005 5 54 54 18 6 2005 12 10 2005 Copyright © 2005 Winter and Ponting; licensee BioMed Central Ltd.2005Winter and Ponting; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The identification of sequence innovations in the genomes of mammals facilitates understanding of human gene function, as well as sheds light on the molecular mechanisms which underlie these changes. Although gene duplication plays a major role in genome evolution, studies regarding concerted evolution events among gene family members have been limited in scope and restricted to protein-coding regions, where high sequence similarity is easily detectable.
Results
We describe a mammalian-specific expansion of more than 20 rapidly-evolving genes on human chromosome Xq22.1. Many of these are highly divergent in their protein-coding regions yet contain a conserved sequence motif in their 5' UTRs which appears to have been maintained by multiple events of concerted evolution. These events have led to the generation of chimaeric genes, each with a 5' UTR and a protein-coding region that possess independent evolutionary histories. We suggest that concerted evolution has occurred via gene conversion independently in different mammalian lineages, and these events have resulted in elevated G+C levels in the encompassing genomic regions. These concerted evolution events occurred within and between genes from three separate protein families ('brain-expressed X-linked' [BEX], WWbp5-like X-linked [WEX] and G-protein-coupled receptor-associated sorting protein [GASP]), which often are expressed in mammalian brains and associated with receptor mediated signalling and apoptosis.
Conclusion
Despite high protein-coding divergence among mammalian-specific genes, we identified a DNA motif common to these genes' 5' UTR exons. The motif has undergone concerted evolution events independently of its neighbouring protein-coding regions, leading to formation of evolutionary chimaeric genes. These findings have implications for the identification of non protein-coding regulatory elements and their lineage-specific evolution in mammals.
==== Body
Background
Discriminating mutations arising during the evolution of mammals which were selectively neutral from those which were adaptive is an important challenge for the current genomic era. In the main, beneficial mutations in mammalian genomes appear to have been gene duplication, rapid sequence divergence, and alteration in gene expression levels [1-4].
An additional lineage-specific mutational process which is also a substrate for selection is concerted evolution, via either unequal crossing-over or gene conversion [5,6]. Non-allelic gene conversion occurs during non-reciprocal homologous recombination when sequence-similar paralogues are misaligned. Converting sequences are often short, in the order of hundreds of basepairs, and when frequent and sustained can lead to the homogenisation of multigene family sequences [7], as observed for mammalian histone [8] and Hsp70 [9] genes. During gene conversion, repair at mismatched positions appears to be biased towards retention of G or C bases which leads to elevation in G+C nucleotide content [10].
Non-allelic gene conversion thus results in phylogenetic trees which display significantly greater proximity between a species' gene paralogues than for gene orthologues of a sister species [11]. Such phylogenetic relationships, however, are also indicative of lineage-specific gene duplication events. Nevertheless, when the affected genes are widely-spread on the genome these relationships are usually indicative of non-allelic gene conversion. This is because gene duplication most frequently results in tandem consecutive genes along the chromosome.
Gene conversion events are expected to occur mostly between protein-coding sequences of genes. This expectation arises from sequence conservation being, in general, highest in protein-coding regions, intermediate in untranslated regions (UTRs), and lowest within introns and intergenic regions [12]. Gene conversion thus is not expected between sequence-dissimilar and non-homologous genes.
Here, we describe genes whose evolution defies these expectations. We present evidence for gene conversion events between mammalian-specific genes which encode sequence-dissimilar and possibly non-homologous, proteins. These conversion events occurred not within their protein-coding or 3' UTR sequences, but rather within their 5' UTRs and upstream regions. We suggest that the occurrence of concerted evolution events during mammalian evolution led to multiple chimaeric genes, with 5' UTR and protein-coding sequences possessing different evolutionary pedigrees.
These proposed events occurred within and between genes from three separate families ('brain-expressed X-linked' [BEX], WWbp5-like X-linked [WEX] and G-protein-coupled receptor-associated sorting protein [GASP]), all of which contain single protein-coding exons. Bex1, -2 and -3 are 'brain-expressed X-linked genes' whose intracellular products bind protein [13]. BEX1 and BEX2 bind the olfactory marker protein (OMP) [14,15] whereas BEX3 binds the p75 neurotrophin receptor (p75NTR) [16], and the second mitochondria-derived activator of caspase (Smac) [17,18], as well as self-associating [16]. WEX proteins include WWbp5 which is a poorly-understood WW domain binding protein [19]. GASP-1 and -2 are G-protein-coupled receptor (GPCR) associated sorting proteins (GASPs) which bind to the COOH-termini of various GPCRs and modulate their endocytic sorting to lysosomes [20,21].
Genes from these three families are all tightly-clustered within a mammal-specific ~2 Mb region of human chromosome Xq22.1-q22.2. We find that these genes all arose during early eutherian evolution and have experienced substantial sequence divergence thereafter. Their localisation to brain tissues, and their unusual and rapid evolution, are thus consistent with their involvement in the evolution of innovative brain cortical structures among eutherian mammals.
Results and discussion
A 2.3 Mb region of human Xq22.1-q22.2 is specific to placental mammals
Our studies focused on a 2.3 Mb region of human and mouse X chromosomes that encode a collinear arrangement of multiple genes from BEX, WEX and GASP families (Figure 1; Table 1). This entire region has a counterpart in neither the chicken nor the metatherian (marsupial) Monodelphis domestica. Orthologues of two genes that flank the region, Gla (encoding α-galactosidase; NM_013463) and Glra4 (encoding glycine receptor subunit α4; NM_010297), are separated by less than 4 Kb on chicken chromosome 4, and 10 Kb on Monodelphis scaffold 13561 (which is expected to lie within the long arm of its X chromosome [22]). These two non-eutherians' intergenic regions are both devoid of sequences homologous to BEX, WEX and GASP genes, and of assembly gaps greater than 100 b.
Figure 1 Protein-coding gene order in human chromosome Xq22.1-q22.2 (~2.3 Mb) and in its mouse orthologous region. Transcriptional orientations are indicated by filled arrow heads. Human- or mouse- specific genes are indicated by short bars. Abbreviations: Cen, Centromere; Ter, Terminal. Gene names are abbreviated as in Table 1.
Table 1 Gene and transcript annotation of protein-coding genes located on human chromosome Xq22.1-q22.2. Human WEX1 has arisen from duplication of WEX2, and human BEX5 is absent from the mouse genome. a Levels of transcript expression in the brain are based on Microarray gene hybridization results [57]. Accession codes and expression tags are shown in parentheses. A gene expression in the brain is absent (-), present (+), or high relative to other tissues (++). Abbreviations: GLA, Galactosidase, alpha; HNRPH2, Heterogeneous nuclear ribonucleoprotein H2; GASP, G protein-coupled receptor-associated sorting; NXF, Nuclear RNA export factor; PRAMEL3, Preferentially expressed in melanoma like 3; WEX, WWbp5-like gene family; BEX, Brain-expressed X-linked; TCP11B, T-complex 11 B; TMSNB, Thymosin, beta, identified in neuroblastoma; RAB40A, Ras oncogene family member; MSP, Microsomal signal peptidase 23 kDa subunit (SPase 22 kDa subunit); KIRL2, Killer immunoglobulin-like receptor-like 2; KIR3DL1, Killer immunoglobulin-like receptor 3DL1; MORF4L2, Mortality factor 4 like 2;
RefSeq Gene RefSeq Protein Synonyms Gene name Gene Start Gene End Transcriptional orientation Expression in the braina (accession/expression-tag) BGW element
NM_000169 NP_000160 GLA 99424654 99434808 - + (GLA/214430_at)
NM_019597 NP_062543 HNRPH2 99435140 99440975 + + (HNRPH2) +
NM_152583 (BC036206) NP_689796 Q8K2R3 GASP4 99512319 99560303 + + (gnf1h06491_at) +
NM_016608 NP_057692 Alex1/Q9P291 GASP7 99577371 99581530 + + (ARMCX1) +
GASP10ψ 99624346 99624903 + N/A +
NM_019007 NP_061880 Q9NWJ3 GASP10 99641972 99644848 - + (FLJ20811) +
NM_016607 NP_057691 Alex3, Q9UH62 GASP6 99649976 99654688 + + (ARMCX3) +
NM_177949 NP_808818 Alex2, O60267 GASP9 99682129 99686692 - + (ARMCX2) +
NM_032946 NP_116564 NXF5 99858942 99884406 - N/A
AK094141 XP_040486 KIAA1789 99910469 99958857 - N/A
NM_080390 (AL540086) NP_525129 my048 WEX1 100152572 100154538 + ++ (AF063606) +
BC071675 N/A WEX2 100167309 100169228 - N/A +
BC042818 N/A BEX5 100180542 100182792 - N/A +
NM_017809 (BX647232) NP060279 NXF2 100242137 100353489 + + (NXF2/220257_x_at)
NM_017809 (BX647232) NP060279 NXF2 100387175 100498589 - + (NXF2/220257_x_at)
NM_021992 NP_068832 TMSNB 100540461 100543504 - ++ Fetal brain (205347_s_at)
NM_207512 NP_997395 NXF4 100576750 100598478 + N/A
NM_022838 NP_073749 Q9H969 GASP5 100626014 100630940 + + (FLJ12969/219335_at)
NM_014710 NP_055525 Q43168 GASP1 100678299 100685235 + ++ (GASP/204793_at)
NM_138437 (AK123832) NP_612446 Q96D09 GASP2 100739148 100744517 + N/A
NM_030639 NP_085142 Q9BE11 GASP3 100747511 100779225 + + (KIAA1701/213709_at)
Rab40A-like 100964038 100965103 + ++ (108_g_at)
NM_018476 (BX098763) NP_060946 BEX1, HBEX2 BEX2 101089445 101090956 - ++ (BEX1/218332_at) +
NM_022052 NP_071335 NXF3 101102606 101119879 - - (NXF3/220110_s_at)
AK000959 N/A Bex4 BEX4 101241922 101243977 + ++ (BEXL1/ 215440_s_at, 40916_at**)
NM_153333 (AL709034) NP_699164 Wex3 WEX3 101279783 101281945 - N/A +
AL550633 N/A Wex4 WEX4 101300507 101303513 - ++ (gnf1h07374_s_at, gnf1h07373_at) +
NM_032621 (AL520932) NP_116010 Bex2, mouseBex1 BEX1 101336132 101337663 - N/A
NM_152278 (H09952) NP_689491 Wex5 WEX5 101357022 101359108 + ++ (gnf1h06733_at) +
NM_016303 (BF028506) NP_057387 Wex6 WEX6 101383280 101385238 + + (WBP5 / 217975_at) +
NM_206915 NP_996798 Ngfrap1 BEX3 101403125 101404856 + ++ (NGFRAP1/217963_s_at) +
NM_080879 NP_543155 RAB40A 101526538 101546274 - N/A (RAB40A, 217589_at, 33477_at)
NM_024863 (AK024827) NP_079139 Wex7 WEX7 101603016 101614300 + ++ (FLJ21174) +
NM_032926 NP_116315 Wex8 WEX8 101634691 101636711 + ++ (MGC15737) +
NM_004780 NP_004771 TCEAL1 WEX9 101655759 101657725 + ++ (TCEAL1) +
NM_012286 NP_036418 MORF4L2 101702290 101713453 - + MORF4L2
Human Xq22.1-q22.2 thus appears to be an innovation of eutherian (placental) mammals. This is surprising since chromosome Xq initially arose after the divergence of the lineages leading to modern birds and mammals, but prior to the metatherian-eutherian split [23]. The observation however, provides an excellent opportunity to investigate issues underlying sequence innovation and functional innovation. We were interested in three interrelated questions: (i) Which evolutionary processes led to the origin of these genes? (ii) Are these genes' functions similar, and are they related to physiological or anatomical innovations in placental mammals? and, (iii) Why are these newly-acquired genes restricted to this one chromosomal region, rather than being dispersed throughout the remainder of the X chromosome or elsewhere?
BEX and WEX proteins are diverged homologues
We investigated the evolutionary origins of BEX, WEX and GASP genes using database searches of known protein and nucleotide sequences. We concluded that mammalian X-linked GASP genes are members of an ancient family that are discernible in early-branching bony vertebrates such as teleost fish (FLJ20811; NM_212853). By contrast, from the results of BLAST and PSI-BLAST [24] searches, BEX or WEX gene homologues appear to be absent outside of eutherian mammals. The order of BEX, WEX and GASP genes is conserved between human, rodent and canine X chromosomes and thus must have been present in their common ancestor.
Nevertheless, despite their sequence divergence we find that BEX and WEX protein sequences are homologous. Using COMPASS [25], significant similarity (Smith-Waterman score 82; E = 4.4 × 10-8) between a multiple sequence alignment of 25 sequences from the BEX protein family, and an alignment of 18 WEX sequences was observed. We thus predict that WEX and BEX proteins arose from a single ancestral gene early in eutherian evolutionary history, but diverged as separate families thereafter by numerous events of gene duplication. This is also consistent with BEX and WEX gene families possessing a common gene structure, namely three exons, with the most 3' of these always containing the entire protein-coding sequence.
5' UTR sequences of BEX, WEX and GASP genes are homologous
Despite the lack of evidence for homology between BEX or WEX, and GASP proteins, we were surprised to observe highly-similar sequences within the 5' UTRs of BEX and GASP genes. For example, a 383 b sequence (human chrX:101090862–101091244) which overlaps and extends the first non-coding exon of BEX2 is 91% identical to a region straddling the first exon of human GASP4.
Using MEME [26] we searched the first 5' UTR exons of human, mouse and rat BEX, WEX and GASP gene sequences for conserved DNA motifs. A highly significant DNA motif (E-value = 5.8 × 10-390, width = 57 bases) was identified by MEME in 37 of these 50 sequences. Using BLASTn searches of genomic sequences, we extended the width and breadth of these motifs, which we call BGW (Bex/GASP/Wex) elements (Figure 2). The program MAST [27] and the position-specific matrix of the MEME BGW motif was then used to search the NCBI non-redundant DNA database for significantly scoring DNA alignments. This search yielded statistically significant hits (E < 10-7) for 18 genomic sequences on human, mouse or rat chromosomes X, all of which represent the 5' UTRs of BEX, WEX and GASP genes. No significant alignments were observed to non-mammalian sequence or to mammalian sequence present outside of the X chromosome.
Figure 2 Multiple sequence alignment of a conserved region of BGW sequence elements. Shaded if columns are conserved in at least 80% of sequences. Abbreviations: Hs, Homo sapiens; Mm, Mus musculus; Rn, Rattus norvegicus; NA, sequences not 5' to an identifiable gene structure.
These BGW motifs are not randomly positioned with respect to coding sequences: 18 of the 23 human homologues occur within the 5' UTRs of BEX or WEX or GASP family genes (Table 1). Of the remaining 5, 2 are homologous to other BGWs that are upstream of BEX or WEX or GASP family genes, as assessed by whole genome BLASTn searches (p < 0.01); 1 is 5' to a neighbouring gene HNRPH2 (and is highly conserved in orthologous sequences in dog, rat and mouse); and, 2, including 1 5' of GASP10ψ, are not upstream of known coding transcripts, so might represent false positives, pseudogenic copies or longer-range regulators. We conclude that the BGW element is a conserved non-coding sequence motif shared by BEX, WEX and GASP genes, which is restricted to eutherian chromosome Xq21-22.
Concerted evolution events within and between the 5' UTRs of BEX, WEX and GASP genes
Further investigation demonstrated that the evolutionary histories of portions of BEX, WEX and GASP genes were not identical to the relationships expected from the species' phylogenetic tree. The first exon of the 5' UTR of BEX2, BEX3 and GASP4 genes encompasses the BGW motif and exhibits greater sequence similarities between paralogues than it does between orthologues (Figure 3A). This finding is indicative of independent concerted evolution events [7,9,11,28,29]. Bootstrap values are sufficiently high to indicate with confidence that concerted evolution events have occurred among these genes in all mammalian lineages examined (human, mouse, rat and dog).
Figure 3 Concerted evolution events among BEX, WEX and GASP genes. Maximum likelihood phylogenetic trees for A) the first exon of the 5'UTR of BEX2, BEX3 and GASP4 genes B) the first exon of the 5' UTR of WEX2, WEX8 and WEX1 genes, and C) the exonic 5' UTR of GASP1 and GASP2 genes. 5' UTR sequences were extracted either from sequence transcripts or from homologous genomic regions. Maximum likelihood tree topologies and branch lengths were obtained using the program BASEML [54], for a given DNA sequence alignment. Each alignment was bootstrapped a 1000 times using the neighbour-joining method, and bootstrap values were overlaid the maximum likelihood tree. Abbreviations: Hs, Homo sapiens; Mm, Mus musculus; Rn, Rattus norvegicus; Cf, Canis familiaris.
Notably, some of these concerted evolution events are lineage-specific. Bootstrap values (Figure 3A) support such events between human or dog BEX2 and GASP4, but not between mouse or rat BEX2 and GASP4, and between mouse, rat or dog BEX2 and BEX3, but not human BEX2 and BEX3.
Similarly, phylogenetic analysis of WEX2 and WEX8 genes indicates that concerted evolution events occurred recently, between the first 5' UTR exons, in the carnivore and primate lineages (Figure 3B). Finally, concerted evolution also occurred between the 5' UTRs of human or mouse GASP1 and GASP2 (Figure 3C).
Some of these concerted evolution events appear to have occurred relatively recently. In particular, concerted evolution between the 5' UTRs of mouse Bex2 and Bex3 (or rat Bex2 and Bex3) genes must have occurred very recently because they exhibit no substitutions of nucleotides within their BGW motifs (Figure 2).
Chimaerism among BEX, WEX and GASP genes
Concerted evolution events between BEX2, BEX3 and GASP4 are restricted to the non-coding regions of their genes. The GASP4 protein exhibits no discernible sequence similarity to either BEX2 or BEX3 proteins; moreover, BEX2 and BEX3 amino acid sequences are relatively divergent (42% identity). Thus, these three genes appear to be chimaeric: their 5' UTRs are highly similar and exhibit a recent ancestry, whereas their protein coding sequences are more distantly related.
This surprising conclusion is supported by a phylogenetic tree of BEX protein coding sequences (Figure 4). With high reliability the protein coding regions of human BEX1 and BEX2 paralogues were found to be more similar than they are to their rodent (mouse and rat), ruminant (cattle) or carnivore (dog) orthologues; similarly mouse BEX1 and BEX2 are most closely related, as are dog BEX1 and BEX2 genes. By contrast, the predicted evolutionary relationships of the three other paralogues, BEX3, BEX4 and BEX5, recapitulate the expected species tree [30].
Figure 4 Concerted evolution events among BEX1 and BEX2 protein-coding sequences. A maximum likelihood phylogenetic tree for the BEX family members' protein-coding regions was constructed using the program BASEML [54] and a DNA sequence alignment. The alignment was bootstrapped a 1000 times using the neighbour-joining method, and provided bootstrap values, which were overlaid onto the maximum-likelihood tree.
These results thus demonstrate distinct and complex evolutionary histories for different regions of genes: concerted evolution events in the 5' UTRs are supported among BEX2, BEX3 and GASP4, while concerted evolution events in protein coding regions are supported only between BEX1 and BEX2. Among WEX proteins, a similar analysis indicates that WEX2, WEX4 and WEX8 coding sequences have experienced concerted evolution events (data not shown).
A pseudogene of human GASP10 (GASP10ψ), which is positioned upstream of GASP10 on chromosome X, appears to be converting with GASP10 in primate (human), rodent (mouse) and carnivore (dog) lineages. Bootstrap analysis supports that GASP10 and GASP10ψ are significantly more similar to each other than they are to their orthologues from dog or mouse (bootstrap value = 100%, data not shown). Therefore, although GASP10ψ does not encode a functional protein, it may function as a redundant copy for facilitating gene conversion events to its neighbouring GASP10 gene, as has occurred elsewhere in the human genome [31,32].
Mechanism and mode of concerted evolution
These and similar inferences of concerted evolution (collated in Additional file 1) could have arisen due to unequal crossing-over, gene duplication or gene conversion. Unequal crossing-over and multiple gene duplications are unlikely mechanisms for concerted evolution of these genes because their orders and transcriptional orientations along their X chromosomes have suffered no rearrangements, inversions (except for one, involving rat WEX2) or large deletions when human, dog and mouse genomes are compared (Table 1; Figure 1; data not shown).
The findings also could have arisen from exon shuffling, where exons are duplicated and inserted into different gene contexts [33]. However, inter-allelic gene conversion is distinguished from exon shuffling in that it often occurs within regions possessing high G+C nucleotide compositions [8-10,34]. Indeed, from G+C fractions for BEX, WEX and GASP 5' UTRs and for the third positions of their protein coding regions ("GC3"), we found that sequences with evidence for concerted evolution possess significantly higher G+C fractions than sequences without such evidence (paired T-test, p(5' UTR) < 10-3, p(GC3) = 10-3) (Figure 5, Additional file 1). This indicates strongly that multiple events of interlocus gene conversion, rather than de novo sequence duplication or exon shuffling, have occurred among these genes.
Figure 5 G/C levels are elevated in human genes that undergo concerted evolution events. 5' UTR G+C and protein-coding GC3 properties of human BEX, WEX and GASP genes are indicated. (+) or (-) symbols represent presence or absence of a concerted evolution event in 5' UTR (first/left symbol) or in protein-coding (second/right symbol) regions.
Evidence for sustained gene conversion events
Our findings indicate that the vertical transmission of BEX, WEX and GASP genes has been interrupted frequently by horizontal acquisitions of non-coding, as well as coding, sequences from genes that are closely-linked on human chromosome Xq22.1-q22.2. These genes' sequences appear to have been homogenized by multiple episodes of interlocus gene conversion. Because gene conversion is thought to proceed via formation of heteroduplexes between highly-similar sequences [35], this raises an interesting conundrum: how has recent gene conversion occurred between sequence-dissimilar genes drawn from different gene families?
We resolve this question by proposing that BEX and WEX proteins all arose from a common ancestral GASP-like gene and rapidly diverged in sequence thereafter (Figure 6). We further propose that sequence similarity has been preserved between these homologues' 5' UTRs by recurrent episodes of gene conversion, despite rapid divergence of sequences elsewhere in their genes.
Figure 6 Proposed evolutionary events leading to BEX, WEX and GASP genes on human chromosome Xq22.1-q22.2. BEX and WEX genes arise due to duplication of a GASP-like gene early in eutherian mammalian history. These genes undergo multiple duplication events thereafter, but prior to the divergence of human and dog lineages. Multiple events of gene conversion, either between 5' UTRs or between coding sequences (Additional file 1), occur with the remaining coding sequence diverging rapidly due to relaxed selective constraints and/or adaptive evolution.
This scenario is supported by three observations: that BEX and WEX genes are homologous, that all three families utilize only a single exon to code for protein, and that BEX, WEX and GASP protein-coding sequences evolved rapidly. For the latter observation, we note that KA/KS values of these genes are high (median of 0.34) relative to a median value of 0.10 for all Ensembl human-mouse single orthologues (data not shown). These genes might then have arisen initially from duplication of GASP10, a gene that contains the BGW element and whose orthologues are known in earlier-diverging vertebrates, including fish (hypothetical protein FLJ20811; NM_212853). Unlike other GASP genes, GASP10 contains a three exon structure, similarly to these observed in BEX and WEX genes.
Selection of gene conversion events
To our knowledge there is only a single documented occurrence of gene conversion between paralogues' 3' UTRs [36], and a single observation of gene conversion between paralogues' 5' UTRs [37]. However, multiple conversion events that differentiate between coding and non-coding sequences, as well as sequence conversion between genes that otherwise are not demonstrably homologous, are completely unexpected.
We attribute this singular evolutionary scenario to placental mammal-specific functional innovation. BEX and WEX genes appear, from the absence of homologues among other vertebrates, to have arisen during mammalian evolution by rapid sequence divergence from a common ancestor. In other analogous situations, such as the evolution of the caseins and histatins from a single ancestral gene [38], rapid gene duplication and sequence diversification is causally linked to innovation in physiology and behaviour [2]. In this case, gene function is associated with binding to brain-specific receptors, as seen for BEX and GASP proteins [15,20,21].
On the basis of frameshifts, stop codons, the lack of an initiating methionine codon or introns, all BEX and WEX paralogues outside of the Xq22.1-q22.2 region appear to be retrotransposed processed pseudogenes. This suggests that there is selection for retention of these genes as a closely-linked group on the X chromosome, and perhaps points either to gene conversion being a necessary requirement for long-term sustenance and evolution of their functions, or an X-linked factor that is necessary for their proper gene functions.
Function prediction
The X chromosome contains a disproportionate number of genes related to mental functions which has been linked to the male preponderance of mental retardation cases [39,40]. However, of 9 X-linked genes that, when mutated, lead to mental impairment all possess orthologues in fish or even earlier-branching eukaryotes [40]. When expression information is available, all BEX, WEX and GASP genes are found to be expressed in the brain (Table 1). These eutherian-specific genes are thus possible candidates for the adaptive evolution of the neocortex, a region of the forebrain which is unique to mammals [41].
The presence of a conserved BGW element within the 5' UTRs of BEX, WEX and GASP genes is suggestive of its participation in regulation of translation. This is because translation rates have been shown previously to be affected by regulatory sequences, which include the start site consensus sequence, secondary structures, upstream AUGs, internal ribosome entry sites (IRES) and sequence specific recognition site for regulatory factors, such as protein or RNA [42,43]. Translational control of BEX, WEX and GASP genes might indicate that their proteins are utilized under specific physiological conditions [44], at developmental stages [45] or in subcellular compartments [46,47]. Another possible role for the BGW element might be to regulate alternative splicing. Although these genes possess only single protein coding exons there are several examples of transcripts that exhibit alternative splicing within their 5' UTR exons (e.g. WEX2 (mRNA BQ068054)) and others that exclude the protein-coding exon altogether (e.g. GASP5 (mRNA BC022066)).
Conclusion
We have described the evolutionary history of a large region of human chromosome X, which appears to be an innovation of placental mammals. This region encompasses three previously unrelated protein-coding gene families, BEX, WEX and GASP, which have been the product of multiple gene duplications and large protein-coding sequence diversification since the earliest eutherian mammal. Despite the lack of protein-coding sequence similarity between many genes, we were able to identify a mammalian conserved DNA motif in their exonic 5'UTR, suggesting that they are derived of a common single ancestor, probably a GASP-like gene, found in early-branching bony vertebrates.
We have shown that the evolution of these paralogous genes has been affected by multiple events of gene conversion acting to homogenize among 5'UTR sequences, protein-coding sequences or both. Events of gene conversion in these regions have led to the occurrence of chimaeric genes, where their 5' UTRs are highly similar and exhibit a recent ancestry, but their protein coding sequences are more distantly related. We showed that the composition of sequences undergoing concerted evolution is enriched with G and C nucleotides, suggesting that biased gene conversion has been the underlying mechanism rather than exon shuffling.
BEX, WEX and GASP genes are found to be expressed in the brain (Table 1), suggesting that these eutherian-specific genes are possible candidates for the adaptive evolution of the neocortex, a region of the forebrain which is unique to mammal. The presence of a conserved BGW element within the 5' UTRs of BEX, WEX and GASP genes is suggestive of its participation in regulation of translation, possibly resulting in different spatio-temporal localization of these genes products or in different alternative splicing forms. These findings thus hint at hitherto unappreciated modes of 5' UTR evolution. The identification of such 5' UTRs elsewhere in the genome thus will be required as a contribution to the delineation of all human sequence under selection.
Methods
Genome assemblies and gene models
The July 2003 human (based on NCBI Build 34), the October 2003 mm4 Mus musculus genome assembly (based on NCBI Build 32), the rn3 June 2003 Rattus norvegicus genome assembly (based on version 3.1), the canFam1 July 2004 Canis familiaris whole genome shotgun (WGS) assembly v1.0, the 13 Nov. 2003 chimpanzee (Pan troglodytes) Arachne assembly – NCBI Build 1 Version 1, and the February 2004 chicken (Gallus gallus) draft assembly were analysed. Gene annotations were extracted from the UCSC genome browser [48], or were predicted using Genewise [49] and available transcriptional information.
BGW element
The ENSEMBL web browser blast application [50] and the BlastN program [24] (without optimization for identical hits) were used to perform a sequence similarity search between a region 5' to Alex2 (chrX:99686693–99686992) and the entire human genome assembly. Multiply aligned genomic hits that were predicted (p < 0.01) to be homologous to this region were positioned on human X chromosome, only within an area containing BEX, WEX, and GASP family members. A similar search was done in mouse, rat, zebrafish and chicken genomes. Human, mouse and rat sequences were thereafter multiply aligned, and used to further search the human genomic area (chrX:99850000–100250000) using an HMM [51]. Additional similar sequences (E < 0.1) were added to generate the final multiple alignment (Figure 2).
Tree construction and bootstrap analysis
Protein-coding and exonic 5' UTR phylogenetic trees were both constructed from DNA sequence alignments. Sequences were aligned using ClustalW [52] or MULAN [53], and alignments were subjected to a neighbour-joining bootstrapping process (n = 1000). Non protein-coding branch length estimations were calculated using a maximum likelihood approach (BASEML [54]) as implemented by the molecular evolution package DAMBE [55]. In order to assign bootstrap values for the branch lengths, neighbour-joining bootstrap values were superimposed on the ML tree.
G+C and GC3 content analysis
G+C proportions contained within the exonic-5' UTR or the entire 5' UTR were based on genomic sequences. GC3 content was calculated from examining the GC fraction of the third nucleotide codon positions of protein-coding sequences.
Exonic 5' UTR sequence comparisons
The longest 5'-UTR sequence of each gene was chosen using all available mRNA and EST transcript information (Table 1). In all cases, sequences were further extended by 300 bases, using genomic data, to account for foreshortened transcript evidence. UTR sequences were then aligned using BLASTN and considered to be homologous when p < 0.001.
KA/KS analysis
Ratios of KA (the number of nonsynonymous substitutions per nonsynonymous site) to KS (the number of synonymous substitutions per synonymous site) were calculated using the yn00 method of Yang and Nielsen [56].
Authors' contributions
EEW identified the shared homology between the 5'UTR of BEX and GASP genes and generated much of the analysis. CPP identified the BGW motif, and assisted in data interpretation. EEW and CPP both contributed to the drafting of the manuscript.
Supplementary Material
Additional File 1
Evidence for gene conversion events among mammalian BEX, WEX and GASP genes. Sequences were grouped according to the gene region they encompass: exonic 5' UTR, 5' UTR (exons and introns) and protein-coding regions. Validation method: (a) A phylogenetic tree indicating significantly greater proximity between gene paralogs in a particular genome than for gene orthologous of different organisms' genomes. Neighbour-joining bootstrap values for the tree topology were considered if they were greater than 85% (b) Sequence identity levels between paralogs are substantially higher than sequence identity levels between orthologs (>85% compared with <65%, respectively). Abbreviations: N/D, not determined.
Click here for file
Acknowledgements
We would like to thank the Medical Research Council UK for funding.
==== Refs
Copley RR Goodstadt L Ponting C Eukaryotic domain evolution inferred from genome comparisons Curr Opin Genet Dev 2003 13 623 628 14638325 10.1016/j.gde.2003.10.004
Emes RD Goodstadt L Winter EE Ponting CP Comparison of the genomes of human and mouse lays the foundation of genome zoology Hum Mol Genet 2003 12 701 709 12651866 10.1093/hmg/ddg078
King MC Wilson AC Evolution at two levels in humans and chimpanzees Science 1975 188 107 116 1090005
Ohno S Evolution by gene duplication 1970 London , George Allen and Unwin
Carrington M Cullen M Justified chauvinism: advances in defining meiotic recombination through sperm typing Trends Genet 2004 20 196 205 15041174 10.1016/j.tig.2004.02.006
Dover GA Evolution of genetic redundancy for advanced players Curr Opin Genet Dev 1993 3 902 910 8118216 10.1016/0959-437X(93)90012-E
Li WHGD Fundamentals of Molecular Evolution 1997 Sinauer Associates Inc. Sunderland, MA.
Galtier N Gene conversion drives GC content evolution in mammalian histones Trends Genet 2003 19 65 68 12547511 10.1016/S0168-9525(02)00002-1
Kudla G Helwak A Lipinski L Gene conversion and GC-content evolution in mammalian Hsp70 Mol Biol Evol 2004 21 1438 1444 15084682 10.1093/molbev/msh146
Meunier J Duret L Recombination drives the evolution of GC-content in the human genome Mol Biol Evol 2004 21 984 990 14963104 10.1093/molbev/msh070
Slightom JL Chang LY Koop BF Goodman M Chimpanzee fetal G gamma and A gamma globin gene nucleotide sequences provide further evidence of gene conversions in hominine evolution Mol Biol Evol 1985 2 370 389 3870867
Hillier LW Miller W Birney E Warren W Hardison RC Ponting CP Bork P Burt DW Groenen MA Delany ME Dodgson JB Chinwalla AT Cliften PF Clifton SW Delehaunty KD Fronick C Fulton RS Graves TA Kremitzki C Layman D Magrini V McPherson JD Miner TL Minx P Nash WE Nhan MN Nelson JO Oddy LG Pohl CS Randall-Maher J Smith SM Wallis JW Yang SP Romanov MN Rondelli CM Paton B Smith J Morrice D Daniels L Tempest HG Robertson L Masabanda JS Griffin DK Vignal A Fillon V Jacobbson L Kerje S Andersson L Crooijmans RP Aerts J van der Poel JJ Ellegren H Caldwell RB Hubbard SJ Grafham DV Kierzek AM McLaren SR Overton IM Arakawa H Beattie KJ Bezzubov Y Boardman PE Bonfield JK Croning MD Davies RM Francis MD Humphray SJ Scott CE Taylor RG Tickle C Brown WR Rogers J Buerstedde JM Wilson SA Stubbs L Ovcharenko I Gordon L Lucas S Miller MM Inoko H Shiina T Kaufman J Salomonsen J Skjoedt K Wong GK Wang J Liu B Wang J Yu J Yang H Nefedov M Koriabine M Dejong PJ Goodstadt L Webber C Dickens NJ Letunic I Suyama M Torrents D von Mering C Zdobnov EM Makova K Nekrutenko A Elnitski L Eswara P King DC Yang S Tyekucheva S Radakrishnan A Harris RS Chiaromonte F Taylor J He J Rijnkels M Griffiths-Jones S Ureta-Vidal A Hoffman MM Severin J Searle SM Law AS Speed D Waddington D Cheng Z Tuzun E Eichler E Bao Z Flicek P Shteynberg DD Brent MR Bye JM Huckle EJ Chatterji S Dewey C Pachter L Kouranov A Mourelatos Z Hatzigeorgiou AG Paterson AH Ivarie R Brandstrom M Axelsson E Backstrom N Berlin S Webster MT Pourquie O Reymond A Ucla C Antonarakis SE Long M Emerson JJ Betran E Dupanloup I Kaessmann H Hinrichs AS Bejerano G Furey TS Harte RA Raney B Siepel A Kent WJ Haussler D Eyras E Castelo R Abril JF Castellano S Camara F Parra G Guigo R Bourque G Tesler G Pevzner PA Smit A Fulton LA Mardis ER Wilson RK Sequence and comparative analysis of the chicken genome provide unique perspectives on vertebrate evolution Nature 2004 432 695 716 15592404 10.1038/nature03154
Brown AL Kay GF Bex1, a gene with increased expression in parthenogenetic embryos, is a member of a novel gene family on the mouse X chromosome Hum Mol Genet 1999 8 611 619 10072429 10.1093/hmg/8.4.611
Baldisseri DM Margolis JW Weber DJ Koo JH Margolis FL Olfactory marker protein (OMP) exhibits a beta-clam fold in solution: implications for target peptide interaction and olfactory signal transduction J Mol Biol 2002 319 823 837 12054873 10.1016/S0022-2836(02)00282-6
Behrens M Margolis JW Margolis FL Identification of members of the Bex gene family as olfactory marker protein (OMP) binding partners J Neurochem 2003 86 1289 1296 12911636
Mukai J Shoji S Kimura MT Okubo S Sano H Suvanto P Li Y Irie S Sato TA Structure-function analysis of NADE: identification of regions that mediate nerve growth factor-induced apoptosis J Biol Chem 2002 277 13973 13982 11830582 10.1074/jbc.M106342200
Kimura MT Irie S Shoji-Hoshino S Mukai J Nadano D Oshimura M Sato TA 14-3-3 is involved in p75 neurotrophin receptor-mediated signal transduction J Biol Chem 2001 276 17291 17300 11278287 10.1074/jbc.M005453200
Yoon K Jang HD Lee SY Direct interaction of Smac with NADE promotes TRAIL-induced apoptosis Biochem Biophys Res Commun 2004 319 649 654 15178455 10.1016/j.bbrc.2004.05.043
Jolliffe CN Harvey KF Haines BP Parasivam G Kumar S Identification of multiple proteins expressed in murine embryos as binding partners for the WW domains of the ubiquitin-protein ligase Nedd4 Biochem J 2000 351 Pt 3 557 565 11042109 10.1042/0264-6021:3510557
Simonin F Karcher P Boeuf JJ Matifas A Kieffer BL Identification of a novel family of G protein-coupled receptor associated sorting proteins J Neurochem 2004 89 766 775 15086532
Whistler JL Enquist J Marley A Fong J Gladher F Tsuruda P Murray SR Von Zastrow M Modulation of postendocytic sorting of G protein-coupled receptors Science 2002 297 615 620 12142540 10.1126/science.1073308
Glas R Marshall Graves JA Toder R Ferguson-Smith M O'Brien PC Cross-species chromosome painting between human and marsupial directly demonstrates the ancient region of the mammalian X Mamm Genome 1999 10 1115 1116 10556436 10.1007/s003359901174
Lahn BT Page DC Four evolutionary strata on the human X chromosome Science 1999 286 964 967 10542153 10.1126/science.286.5441.964
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Sadreyev R Grishin N COMPASS: a tool for comparison of multiple protein alignments with assessment of statistical significance J Mol Biol 2003 326 317 336 12547212 10.1016/S0022-2836(02)01371-2
Bailey TL Elkan C Fitting a mixture model by expectation maximization to discover motifs in biopolymers Proc Int Conf Intell Syst Mol Biol 1994 2 28 36 7584402
Bailey TL Gribskov M Combining evidence using p-values: application to sequence homology searches Bioinformatics 1998 14 48 54 9520501 10.1093/bioinformatics/14.1.48
Liao D Concerted evolution: molecular mechanism and biological implications Am J Hum Genet 1999 64 24 30 9915939 10.1086/302221
Nenoi M Mita K Ichimura S Kawano A Higher frequency of concerted evolutionary events in rodents than in man at the polyubiquitin gene VNTR locus Genetics 1998 148 867 876 9504932
Springer MS Murphy WJ Eizirik E O'Brien SJ Placental mammal diversification and the Cretaceous-Tertiary boundary Proc Natl Acad Sci U S A 2003 100 1056 1061 12552136 10.1073/pnas.0334222100
Amor M Parker KL Globerman H New MI White PC Mutation in the CYP21B gene (Ile-172----Asn) causes steroid 21-hydroxylase deficiency Proc Natl Acad Sci U S A 1988 85 1600 1604 3257825
Roesler J Curnutte JT Rae J Barrett D Patino P Chanock SJ Goerlach A Recombination events between the p47-phox gene and its highly homologous pseudogenes are the main cause of autosomal recessive chronic granulomatous disease Blood 2000 95 2150 2156 10706888
Patthy L Genome evolution and the evolution of exon-shuffling--a review Gene 1999 238 103 114 10570989 10.1016/S0378-1119(99)00228-0
Galtier N Piganeau G Mouchiroud D Duret L GC-content evolution in mammalian genomes: the biased gene conversion hypothesis Genetics 2001 159 907 911 11693127
Lamb BC The properties of meiotic gene conversion important in its effects on evolution Heredity 1984 53 ( Pt 1) 113 138 6436195
Bargelloni L Scudiero R Parisi E Carginale V Capasso C Patarnello T Metallothioneins in antarctic fish: evidence for independent duplication and gene conversion Mol Biol Evol 1999 16 885 897 10406107
Bobba A Marra E Lattanzio P Iolascon A Giannattasio S Characterization of the CYP21 gene 5' flanking region in patients affected by 21-OH deficiency Hum Mutat 2000 15 481 10790214 10.1002/(SICI)1098-1004(200005)15:5<481::AID-HUMU14>3.0.CO;2-A
Kawasaki K Weiss KM Mineralized tissue and vertebrate evolution: the secretory calcium-binding phosphoprotein gene cluster Proc Natl Acad Sci U S A 2003 100 4060 4065 12646701 10.1073/pnas.0638023100
Skuse DH X-linked genes and mental functioning Hum Mol Genet 2005 14 Spec No 1 R27 32 15809269 10.1093/hmg/ddi112
Zechner U Wilda M Kehrer-Sawatzki H Vogel W Fundele R Hameister H A high density of X-linked genes for general cognitive ability: a run-away process shaping human evolution? Trends Genet 2001 17 697 701 11718922 10.1016/S0168-9525(01)02446-5
Allman J Evolving Brains 2000 New York , Scientific American Library
Gray NK Wickens M Control of translation initiation in animals Annu Rev Cell Dev Biol 1998 14 399 458 9891789 10.1146/annurev.cellbio.14.1.399
Wilkie GS Dickson KS Gray NK Regulation of mRNA translation by 5'- and 3'-UTR-binding factors Trends Biochem Sci 2003 28 182 188 12713901 10.1016/S0968-0004(03)00051-3
Hentze MW Kuhn LC Molecular control of vertebrate iron metabolism: mRNA-based regulatory circuits operated by iron, nitric oxide, and oxidative stress Proc Natl Acad Sci U S A 1996 93 8175 8182 8710843 10.1073/pnas.93.16.8175
Grosshans H Slack FJ Micro-RNAs: small is plentiful J Cell Biol 2002 156 17 21 11781331 10.1083/jcb.200111033
Nakamura A Amikura R Hanyu K Kobayashi S Me31B silences translation of oocyte-localizing RNAs through the formation of cytoplasmic RNP complex during Drosophila oogenesis Development 2001 128 3233 3242 11546740
Palacios IM St Johnston D Getting the message across: the intracellular localization of mRNAs in higher eukaryotes Annu Rev Cell Dev Biol 2001 17 569 614 11687499 10.1146/annurev.cellbio.17.1.569
UCSC Genome Bioinformatics Site
Birney E Clamp M Durbin R GeneWise and Genomewise Genome Res 2004 14 988 995 15123596 10.1101/gr.1865504
Ensembl Web Browser Blast Application
Eddy SR Profile hidden Markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Ovcharenko I Loots GG Giardine BM Hou M Ma J Hardison RC Stubbs L Miller W Mulan: Multiple-sequence local alignment and visualization for studying function and evolution Genome Res 2005 15 184 194 15590941 10.1101/gr.3007205
Yang Z PAML: a program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 1997 13 555 556 9367129
Xia X Xie Z DAMBE: software package for data analysis in molecular biology and evolution J Hered 2001 92 371 373 11535656 10.1093/jhered/92.4.371
Yang Z Nielsen R Estimating synonymous and nonsynonymous substitution rates under realistic evolutionary models Mol Biol Evol 2000 17 32 43 10666704
GNF Gene Expression Atlas
|
16221301
|
PMC1274310
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Oct 12; 5:54
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-54
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-571624202010.1186/1471-2148-5-57Research ArticleComparative genomics of Thermus thermophilus and Deinococcus radiodurans: divergent routes of adaptation to thermophily and radiation resistance Omelchenko Marina V [email protected] Yuri I [email protected] Elena K [email protected] Vera Y [email protected] Alexander [email protected] Min [email protected] Michael J [email protected] Eugene V [email protected] Kira S [email protected] Department of Pathology, F.E. Hebert School of Medicine, Uniformed Services University of the Health Sciences, Bethesda, MD 20814-4799, USA2 National Center for Biotechnology Information, National Library of Medicine, National Institutes of Health, Bethesda, MD 20894, USA2005 20 10 2005 5 57 57 23 8 2005 20 10 2005 Copyright © 2005 Omelchenko et al; licensee BioMed Central Ltd.2005Omelchenko et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Thermus thermophilus and Deinococcus radiodurans belong to a distinct bacterial clade but have remarkably different phenotypes. T. thermophilus is a thermophile, which is relatively sensitive to ionizing radiation and desiccation, whereas D. radiodurans is a mesophile, which is highly radiation- and desiccation-resistant. Here we present an in-depth comparison of the genomes of these two related but differently adapted bacteria.
Results
By reconstructing the evolution of Thermus and Deinococcus after the divergence from their common ancestor, we demonstrate a high level of post-divergence gene flux in both lineages. Various aspects of the adaptation to high temperature in Thermus can be attributed to horizontal gene transfer from archaea and thermophilic bacteria; many of the horizontally transferred genes are located on the single megaplasmid of Thermus. In addition, the Thermus lineage has lost a set of genes that are still present in Deinococcus and many other mesophilic bacteria but are not common among thermophiles. By contrast, Deinococcus seems to have acquired numerous genes related to stress response systems from various bacteria. A comparison of the distribution of orthologous genes among the four partitions of the Deinococcus genome and the two partitions of the Thermus genome reveals homology between the Thermus megaplasmid (pTT27) and Deinococcus megaplasmid (DR177).
Conclusion
After the radiation from their common ancestor, the Thermus and Deinococcus lineages have taken divergent paths toward their distinct lifestyles. In addition to extensive gene loss, Thermus seems to have acquired numerous genes from thermophiles, which likely was the decisive contribution to its thermophilic adaptation. By contrast, Deinococcus lost few genes but seems to have acquired many bacterial genes that apparently enhanced its ability to survive different kinds of environmental stresses. Notwithstanding the accumulation of horizontally transferred genes, we also show that the single megaplasmid of Thermus and the DR177 megaplasmid of Deinococcus are homologous and probably were inherited from the common ancestor of these bacteria.
==== Body
Background
Deinococcus spp. and Thermus spp. are believed to belong to a distinct branch of bacteria called the Deinococcus-Thermus group [1]. The common origin of these bacteria is supported by the fact that they consistently form a strongly supported clade in phylogenetic trees of ribosomal RNAs and several conserved proteins including ribosomal proteins, RNA polymerase subunits, RecA, and others [1-7]. The Thermus genus currently consists of 14 thermophilic species, whereas the Deinococcus genus includes at least eleven, mostly mesophilic species known for their extreme resistance to γ-irradiation and other agents causing DNA damage, particularly, desiccation [7,8]. Interestingly, two new species, D. frigens sp. nov., D. saxicola, have been isolated recently from Antarctic rock and soil samples [9]. D. geothermalis and D. murrayi, are considered to be thermophilic (topt~45–50°C) [10].
T. thermophilus (TT) (strain HB27, ATCC BAA-163)and D. radiodurans (DR) (strain R1, ATCC BAA-816) have similar general physiology, both being catalase-positive, red-pigmented, non-sporulating, aerobic chemoorganoheterotrophs [11]. However, the two organisms are dramatically different in terms of stress resistance: DR is one of the most resistant to radiation and desiccation among the characterized organisms and, generally, can survive diverse types of oxidative stress, whereas TT is a thermophile that thrives under thermal stress conditions but is relatively sensitive to radiation and other forms of oxidative stress. Recently, the genomes of T. thermophilus HB27 and HB8 [12,13] became available for comparison with the previously sequenced genome of D. raduodurans R1 [14].
Despite extensive research, genetic systems underlying both thermophilic adaptations and radiation resistance remain poorly understood. Attempts have been made to detect "thermophilic determinants" in the proteomes of thermophiles using a comparative-genomic approach. This resulted in the delineation of a set of proteins that might be associated with the thermophilic phenotype, although most of these are significantly enriched in thermophiles but not unique to these organisms [15-17]. Other distinctions between thermophilic and mesophilic proteins might be due to differences in their structural properties, such as different amino acid compositions, loop lengths, number of salt bridges, strength of hydrophobic interaction, number of disulfide bonds, and other features [18-25].
Various hypotheses also have been proposed to explain radiation resistance, some postulating the existence of specialized genetic systems, particularly those for DNA repair and stress response [7,26]. Recently, however, alternative possibilities have been advanced. For example, the post-irradiation adjustment of metabolism of D. radiodurans might prevent production of reactive oxygen species (ROS) by decreasing the number of reactions involving oxygen [27,28], and high intracellular manganese concentrations of Deinococcus spp. might help scavenge ROS generated during irradiation and post-irradiation recovery [28,29]. However, these explanations of radiation resistance have received little direct support from comparative-genomic analyses [7,27,28].
Several evolutionary processes could potentially contribute to the genome differentiation of TT and DR subsequent to the divergence from the common ancestor: (i) differential gene loss and gain, (ii) acquisition of genes via horizontal gene transfer (HGT) which may be followed by loss of the ancestral orthologous gene (xenologous gene displacement (XGD)), (iii) lineage-specific expansion of paralogous gene families by duplication and/or acquisition of paralogs via HGT; (IV) modification of amino acid composition that could affect protein stability. Here, we experimentally characterize radiation and desiccation resistance of TT in comparison to DR, and, assess the contribution of different evolutionary processes to distinct adaptations of TT and DR, using a variety of comparative-genomic approaches and phylogenetic analysis. We identify the unique feature of the gene repertoires of TT and DR that might contribute to these phenotypic differences, which could be the subject of further experimental work. In addition, we describe the results of a detailed analysis of the proteins predicted to be involved in DNA repair and stress response functions, which are particularly relevant for adaptive evolution of resistance phenotypes.
Results and discussion
Experimental characterization of resistance to gamma-radiation and desiccation, and determination of intracellular Mn/Fe ratio for T. thermophilus
While the radiation- and desiccation-resistant phenotypes of DR and the thermal requirements of both TT and DR have been studied extensively ([7,12,30] and references therein), we are unaware of any detailed characterization of the response of TT to irradiation or desiccation. Therefore, we sought to investigate these properties in order to obtain a more complete picture of the differences in the stress response phenotypes of TT and DR. Not unexpectedly, we found that TT was much more sensitive to acute irradiation than DR. The survival curve of TT is similar to that of Esherichia coli K12. For DR, the radiation dose yielding 10% colony forming unit (CFU) survival (D10) is ~16 kGy, whereas for TT and E. coli, the D10 dose is ~0.8 kGy and ~0.7 kGy, respectively (Figure 1). TT is also highly sensitive to desiccation. The 10% CFU desiccation survival frequency of DR is sustained after 30 days, while TT reaches the 10% CFU desiccation survival at ~10 hours (0.4% survival after 24 hours) and, by the 5th day, suffers essentially 100% lethality. The low resistance of TT to desiccation is observed regardless of the temperature and drying rate (see Additional file 1, "Desiccation of TT at 65°C"). The desiccation resistance of E. coli was found to be intermediate between those of TT and DR (Figure 2).
Figure 1 Radiation resistance of T. thermophilus (ATCC BAA-163), D. radiodurans (ATCC BAA-816) and E. coli (K-12 MG1655, provided by M. Cashel, NIH) (60Co irradiation). Standard deviations for the data points are shown.
Figure 2 Desiccation resistance of T. thermophilus (ATCC BAA-163), D. radiodurans (ATCC BAA-816) and E. coli (K-12 MG1655) (room temperature). Note that no surviving cells were obtained for cells desiccated at 65°C. Standard deviations for five data points are shown.
We recently reported a trend of intracellular Mn/Fe concentration ratios in bacterial IR resistance, where very high and very low Mn/Fe ratios correlated with very high and very low resistances, respectively [29]. We have also shown that growing D. radiodurans in conditions which limited Mn(II) accumulation, significantly lowered the cells' IR resistance [29]. These observations led to the hypothesis that the ratio of Mn to Fe in a cell might determine the relative abundance of different ROS induced during exposure to and recovery from IR [28,29]. At high concentrations, Mn(II) can act as true catalyst of the dismutation of superoxide (O2•-), with Mn cycling between the divalent and trivalent states; Mn redox-cycling scavenges both O2•- and hydrogen peroxide [31]. In this context, we determined the intracellular Mn/Fe ratio of TT using an inductively coupled plasma-mass spectrometry method (ICP-MS), as previously described [29]. TT cells contained 0.211 (± 0.0254) nmol Mn/mg protein and 4.54 (± 0.778) nmol Fe/mg protein. The intracellular Mn/Fe ratio of TT (D10, ~0.8 kGy) is 0.047 compared to 0.0072 for E. coli (D10, ~0.7 kGy), 0.24 for D. radiodurans (D10, ~16 kGy), and <0.0001 for Pseudomonas putida (D10, ~0.25 kGy) [29]. Thus, TT appears to be somewhat more sensitive to acute IR and desiccation than predicted by its Mn/Fe ratio, suggesting the possibility of more complex relationships between these two variables. Notably, scavenging of O2•- by Mn(II) is highly dependent on the availability of H2O2, which in TT would be expected to become limiting during recovery at 65°C because of thermal decomposition [32]; inefficient Mn redox-cycling can lead to Mn(III) accumulation, which is cytotoxic. Both DR and TT encode an ABC- type Mn transporter and a transcriptional regulator that probably regulates Mn homeostasis. Additionally, DR has a NRAMP family Mn transporter, for which there is no ortholog in TT.
Reconstruction of the gene-content tree and the gene repertoire of the common ancestor of Deinococcus and Thermus
Information on the presence-absence of orthologous genes in a set of genomes can be used to produce a gene-content tree [33,34]. The topology of a gene-content tree may reflect not only the phylogenetic relationships between the compared species but lifestyle similarities and differences as well [33,35,36]. Given the dramatic differences in the lifestyles and resistance phenotypes of TT and DR, we were interested to determine whether or not the gene content of TT was most similar to that of DR or those of other thermophilic bacteria or, perhaps, even archaea. To this end, we assigned the proteins encoded in the TT genome to the Clusters of Orthologous Groups of proteins (COGs) [37] and, using the patterns of representation of species in COGs to calculate distances between species, reconstructed a gene-content tree as described previously [35]. In the resulting gene-content tree, which included 62 sequenced genomes of prokaryotes and unicellular eukaryotes, TT and DR were confidently recovered as sister species, and the DR-TT lineage was positioned within a subtree that also included Actinobacteria and Cyanobacteria, several of which are known for their extreme radiation and desiccation resistances [38-40] (Figure 3). For this branch, the topology of the gene-content tree mimics the topologies of trees constructed with other approaches based on genome-wide data [34,35], indicating that the gene repertoires of these bacteria, and TT and DR in particular, have been diverging, roughly, in a clock-like fashion. To determine which genes were likely to have been lost and gained in each lineage, we reconstructed a parsimonious scenario of evolution from the last bacterial common ancestor (LBCA) to TT and DR, through their last common ancestor. The reconstruction was performed on the basis of the assignment of TT and DR proteins to COGs, together with COG-based phyletic patterns of 62 other sequenced bacterial and archaeal genomes [37], using a previously developed weighted parsimony method [41] (see Methods and Additional file 2). This approach assigns 1,310 genes (COGs) to the DR-TT common ancestor (Figure 4). Of these, 1,081 (~80%) were retained in both TT and DR and belong to their shared gene core. Since TT (2210 genes) has far fewer predicted protein-coding genes than DR (3191 genes), it seems likely that the divergence of the two involved substantial genome reduction in TT and/or genome expansion in DR. However, the reconstruction results suggest that TT has not experienced massive genome reduction although the total gene flux (i.e., the sum total of genes inferred to have been lost and gained) during the evolution of this lineage was considerable, involving ~25% of the gene complement. In contrast, DR gained 272 COGs, with only 59 lost, which indicates substantial genome growth after the DR-TT divergence (Figure 4).
Figure 3 Gene content tree constructed for 66 species included in the COG database on the basis of the patterns of presence-absence in the COGs. The Thermus-Deinococcus clade is marked by bold type. Black, bacteria; yellow, archaea; blue, eukaryotes.
Figure 4 The reconstructed evolutionary scenario for the Thermus-Deinococcus clade. LBCA – Last Bacterial Common Ancestor; DR-TT – the common ancestor of Thermus-Deinococcus clade; TT – T. thermophilus; DR- D. radiodurans. Total number of COGs is shown in boxes for each node. '+' indicates inferred gene (COG) gain, and '-' indicates inferred loss.
We were further interested in determining whether similarities existed among the gene repertoires of TT and two deeply-branching bacterial hyperthermophiles, Aquifex aeolicus (AA) and Thermotoga maritima (TM). We found that genes that are present in TT but not DR are significantly more likely to be present in AA and TM than genes present in DR but not TT (Table 1). In contrast, among the 170 TT genes inferred to have been lost, only 20 were present in both AA and TM. Thus, the gene repertoire of TT has significantly greater similarity with hyperthermophilic bacteria than the gene repertoire of DR, perhaps resulting from direct or parallel HGT (see also below).
Table 1 Concordant and discordant phyletic patterns between DR, TT, Aquifex aeolicus (AA) and Thermotoga maritima (TM)
AA+TM+ AA-TM-
DR-TT+ 47 (21.7) 110 (135.3)
DR+TT- 23 (48.3) 326 (300.7)
Note: Expected number of COGs under the assumption of independence is shown in parentheses. Probability, associated with the χ2 test, is 2 × 10-12 for the complete 2 × 2 table and 3 × 10-6 for the 2 × 1 table (concordant vs. discordant).
The majority of the genes shared by TT and DR encode house-keeping proteins and are widespread in bacteria. Among those, 14 COGs are unusual in that they are not found in any other bacteria but instead are shared by TT and DR with archaea or eukaryotes. This set includes 8 subunits of the Archaeal/vacuolar-type H+-ATPase and six other COGs that consist of characteristic archaeal genes (phosphoglycerate mutase, COG3635; 2-phosphoglycerate kinase, COG2074; SAR1-like GTPase, COG1100; GTP:adenosylcobinamide-phosphate guanylyltransferase, COG2266; Predicted membrane protein, COG3374 ; Predicted DNA modification methylase, COG1041).
As noticed previously, some DR genes that belong to families well-represented in both bacteria and archaea showed clear archaeal affinity [42]. To assess how many genes of apparent "thermophilic" descent (either archaeal or bacterial) might already have been present already in the common DR-TT ancestor, we performed phylogenetic analysis of genes that were assigned to the DR-TT ancestor and had 3 of the 5 best hits to the genes from thermophiles (both archaeal and bacterial; see Additional file 1: "Phylogenetic analysis of the genes of the reconstructed DR-TT common ancestor"). We found that at least 122 genes (~10% of the predicted gene repertoire of the DR-TT common ancestor) of the originally selected 205 genes showed an affinity to thermophilic species, i.e., either a branch of two orthologous genes from DR and TT, or a DR or TT gene (in cases when the respective ortholog apparently was lost in the other lineage) clustered with thermophiles (see Additional file 1, table 1S; and Additional file 6). Due to the fact that many tree topologies are highly perturbed by multiple HGT events and may be inaccurate due to differences in evolutionary rates between lineages, this is only a rough estimate of the number of "thermophilic" genes in the DR-TT ancestor. In particular, it cannot be ruled out that some of the ancestral "thermophilic" genes, which are currently present in TT but not DR (10 such genes out of 122 tested genes), have been acquired by TT via XGD (see Additional file 1, table 1S). Taken together, these observations suggest the possibility of ecological contacts between the DR-TT ancestor and hyperthermophilic archaea and/or bacteria, leading to substantial acquisition of "thermophilic" genes via HGT.
Gene gain and loss in Thermus
Our reconstruction of the evolutionary events that occurred after the divergence of the TT and DR lineages from the common ancestor delineated the sets of genes that likely have been gained and lost by each lineage (Figure 4). We first consider in greater detail the pattern of gene loss and gain in TT. The absence of certain metabolic genes in TT creates gaps in its metabolic pathways, some of which are essential. However, TT is capable of synthesizing all amino acids, nucleotides and a majority of cofactors, suggesting that the gaps are filled by analogous or at least non-orthologous enzymes. Several such cases have been described. For example, TT and DR encode unrelated thymidylate synthases, DR2630 (COG0207) and TTC0731 (COG1351). The classical, folate-dependent thymidylate synthase (COG0207) present in DR is probably ancestral in bacteria and apparently was displaced via HGT in the Thermus lineage by the flavin-dependent thymidylate synthase, typical of archaea and bacterial thermophiles. In other cases, the substituted analogous enzymes or pathways remain uncharacterized. For instance, the displacement of the folate-dependent thymidylate synthase with the flavin-dependent type in the TT lineage and in other bacteria and archaea correlates with the apparent loss of dihydrofolate reductase (folA, COG0262), which catalyzes the last step of the tetrahydrofolate biosynthesis pathway. Since tetrahydrofolate is an essential cofactor, a displacement appears most likely. Recently, it has been shown that the halobacterial FolP-FolC fusion protein complemented a Haloferax volcanii folA mutant [43]. Thus, it appears likely that FolC and FolP in TT and other organisms complement the activity of FolA although the existence of an unrelated, as yet uncharacterized dihydrofolate reductase cannot be ruled out.
In addition, TT does not encode two enzymes for pyridoxal phosphate biosynthesis (pdxK, COG0259 and pdxH, COG2240), while DR has a complete set of enzymes of this pathway. Our reconstructions suggest that pdxK was likely lost in the TT lineage, whereas DR probably independently acquired pdxH. However, the pathway is likely to be functional in both organisms. Since similar gaps in the pyridoxal phosphate biosynthesis are seen in a variety of prokaryotes [44], it appears that, for at least some steps of this pathway, there exists a set of distinct enzymes which remain unidentified.
Some systems apparently were completely lost in the TT lineage. These include the urease complex, the ramnose metabolism pathway, acyl CoA:acetate/3-ketoacid CoA transferase, fructose transport and utilization, and glycerol metabolism. Notably, most of these systems are also absent in thermophilic bacteria and many thermophilic archaea.
In contrast to DR, the genes that appear to have been acquired by TT show a clear connection to the thermophilic lifestyle. In particular, TT seems to have acquired 23 gene families from the set of putative thermophilic determinants [17], whereas the common DR-TT ancestor had 5 genes from the list, and DR seems to have acquired only one (see Additional file 3). The majority of these proteins (17 of 31) are encoded in the TT megaplasmid and 11 belong to the predicted mobile DNA repair system characteristic of thermophiles [16,45] (Figure 7A; see details below). In addition, TT has acquired 4 "archaeal" genes that are not encoded in any of the genomes of mesophilic bacteria assigned to COGs (peptide chain release factor 1, COG1503; DNA modification methylase, COG1041; and two membrane proteins, COG3462 and COG4645).
Figure 7 A. Organization of genes encoding the putative thermophyle-specific DNA repair system in two Thermus strains and the draft genome of D. geothermalis. The boxes on top of the figure show COG numbers. Genes are shown by block arrows indicating the direction of transcription and identified by their systematic names. For each column of the alignment, the corresponding COG number and predicted function is indicated. D. geothermalis' genes are connected with respective orthologs in both TT strains by a straight line. Generally, orthologous genes are shown by the same color and pattern. The exceptions are the RAMP proteins of COGs 1336, 1367, 1604, 1337 and 1332, which are all shown in pink. Other, more distant RAMPs are also shown in pink, with different patterns [16]. Proteins that do not belong to COGs but have homologs in other species are marked by diamonds. Abbreviations: HEL, predicted helicase, HD nuclease – HD conserved motif containing predicted nuclease conserved region; POL – novel predicted polymerase, RECB – predicted nuclease of RecB family; B. Comparison of gene organization in the region of the megaplasmid coding for reverse gyrase in two TT strains. The shorter gyrase gene in HB27 indicates truncation. Abbreviations: REVGYR, reverse gyrase, MET_PR – predicted metal-dependent protease; REC_DNA – three-domain fusion protein (DnaQ endonuclease, DinG helicase, and RecQ helicase).
The Sox-like sulfur oxidation system is among the group of genes that were apparently acquired in the TT lineage. The TT Sox operon is partly similar to the one identified in AA (see Additional file 1, Figure 2S), and might have been horizontally transferred between the AA and TT lineages with subsequent local rearrangements. The presence of the Sox operon in TT suggests that this bacterium can use reduced sulfur compounds as a source of energy and sulfur. Another system likely acquired by TT is lactose utilization (at least three proteins: LacZ, COG3250; GalK, COG0153; GalT COG1085; GalA, COG3345). This system is also present in TM, which indicates that sugars can be utilized as carbon sources by thermophilic bacteria.
Gene gain and loss in Deinococcus
The DR lineage has apparently acquired many more genes than it has lost (Figure 4). The majority of the genes lost by DR encode enzymes of energy metabolism and biosynthesis of cofactors. One example is the loss of the three subunits of pyruvate:ferredoxin oxidoreductase, which is one of the several known systems for pyruvate oxidation, a key reaction of central metabolism. Another example of gene loss in DR involves the three subunits of NAD/NADP transhydrogenase, which is responsible for energy-dependent reduction of NADP+ [46]. In addition, the DR lineage lost four enzymes of NAD biosynthesis and six enzymes of cobalamine biosynthesis, and consistently, DR is dependent on an exogenous source of NAD for growth [28,47].
A conspicuous number of genes apparently acquired by the DR lineage encode systems of protein degradation and amino acid catabolism (e.g., urease, DRA0311-DRA0319, and a predicted urea transporter, DRA0320-DRA0324; histidine degradation system, DRA0147-DRA0150; monoamine oxidase, DRA0274; lysine 2,3-aminomutase, DRA0027; kynureninase and tryptophan-2,3-dioxygenase, DRA0338-DRA0339; peptidase E, DR1070, and carboxypeptidase C, DR0964; and D-aminopeptidase, DR1843 (see Additional file 4). A similar trend is observed for the expansion of several protein families in DR, such as secreted subtilisin-like proteases (see below). Additionally, DR acquired two three-subunit complexes of aerobic-type carbon monoxide dehydrogenase (DRA0231-DRA0233 and DRA0235-DRA0237); oxidation of CO by this enzyme might be used as an energy source as shown for some bacteria [48]. Acquisition and expansion of these metabolic systems, together with the loss of certain biosynthetic capabilities, supports the possibility that metabolic restructuring could impact oxidative stress resistance in DR by decreasing the need for high-energy-dependent cellular activities. Energy production through the respiratory chain is one major source of free radicals in the cell [49].
DR has many more genes for proteins involved in inorganic ion transport and metabolism than TT. In particular, DR has acquired the multisubunit Na+/H+ antiporter (7 genes), K+-transporting ATPase (3 genes), and the FeoA/FeoB Fe transport system. This abundance of ion transport systems might be indirectly linked to oxidative stress resistance through regulation of membrane ion gradients and Mn/Fe homeostasis (see above).
DR is more dependent than TT on peptide-derived growth substrates [47] and has a more complex stress response circuitry. Consistent with this, the signal transduction systems of DR, as predicted by genome analysis, are considerably more elaborate than those of TT. In particular, DR has at least 33 COGs (15 probably acquired after the divergence from the common ancestor with TT) related to signal-transduction functions that are not represented in TT as compared to 12 (5 acquired) such COGs in TT. Furthermore, although most of the signal-transduction domains are shared by DR and TT, the domain architectures of the respective multidomain proteins are completely different ([7] and KSM, unpublished observations).
Among the genes apparently acquired by DR, two encode multidomain proteins containing distinct periplasmic ligand-binding sensor domains (DRA0202, COG5278 and DR1174, COG3614). Another protein, DRA0204, contains the CHASE3 domain [50] and is located in a predicted operon with superoxide dismutase (DRA0202), indicating a function in oxidative stress response. The protein DR0724 contains the SARP domain which is involved in apoptosis-related signaling pathways in eukaryotes [51] but its function in bacteria is unknown. The roles of the other signal transduction proteins of DR are even less clear, with the notable exception of a phytochrome-like protein (DRA0050) that apparently was acquired by DR from a bacterial source and has been implicated in UV resistance [52].
Additionally, DR has many genes (25 COGs) encoding systems for microbial defense; TT has only 14 COGs in this category, 13 of which are shared with DR. At least 7 genes for restriction-modification system subunits were specifically acquired by the DR lineage, along with several antibiotic-resistance enzymes. This difference might be linked to the reduced metabolic capabilities of DR, which is dependent on nutrient-rich conditions for growth and, perhaps, encounters more microbial species than TT.
The previous analysis of the DR genome revealed 15 genes that appear to have been horizontally transferred from unexpected sources, such as eukaryotes and viruses [7]; only two of these 15 genes are present in TT, the desiccation-related protein of the ferritin family and the Uma2-like family proteins (see discussion of these proteins below). Two desiccation-related proteins have been shown to be involved in desiccation but not radiation resistance [53]; Ro ribonucleoprotein is apparently involved in UV resistance [54], and topoisomerase IB, while active, has no known role in DR [55]. So far, none of these genes has been linked experimentally with radioresistance in DR.
Identification of xenologous gene displacement by phylogenetic analysis
It is well-established that HGT has made major contributions to the gene repertoires of most thermophilic bacteria as supported by the presence of numerous genes with unexpectedly high similarity to and/or phylogenetic affinity with genes typical of hyperthermophilic archaea [56-60]. These cases include even those proteins that have orthologs in mesophilic bacteria but, as shown by phylogenetic analysis, have clear affinity to archaea or thermophilic bacteria from distant bacterial lineages, which is indicative of XGD [57]. To investigate the impact of HGT from thermophiles leading to XGD on the evolution of the gene repertoire of TT, we determined the taxonomic affiliations of the proteins from the common gene core of TT and DR. We used the taxonomic distribution of best hits in BLAST searches for preliminary identification of HGT candidates, followed by a detailed phylogenetic analysis of selected genes.
As expected, compared to DR, TT has a notable excess in the fraction of best hits to thermophilic bacteria and archaea for both core and non-core proteins (Figure 5A; and see Additional file 5). However, it has been reported that the best BLAST hit does not always accurately reflect phylogeny [61]. In particular, artifacts of best hit analysis may be caused by similar biases in the amino acid composition of proteins in TT and other thermophiles, as demonstrated previously [62,63]. To assess these effects systematically, we performed phylogenetic analysis of 112 TT proteins and 21 DR proteins from the common core that had their respective best hits in thermophiles (see Additional file 7). Despite the fact that all these trees were built for families in which TT and DR proteins were not mutual best hits in the non-redundant protein sequence database (National Center for Biotechnology Information, NIH, Bethesda), more than half of the trees (69 trees, 52%) recovered a DR-TT clade, 39 of these grouping this clade with mesophiles and 30 with thermophiles (Figure 5B). Nevertheless, the difference of evolutionary patterns of DR and TT came across clearly in this analysis: a reliable affinity with thermophiles was detected for 18 TT proteins and only one DR protein. The former cases are likely to represent HGT into the TT lineage from other thermophiles, whereas the only "thermophilic" gene of DR may involve the reverse direction of HGT, from the DR lineage to Thermoanaerobacter tencongiensis (data not shown).
Figure 5 Taxonomic affinities of TT and DR proteins. A. Distribution of the numbers of best hits to proteins from thermophiles for core and non-core proteins of DR and TT. B. Distribution of phylogenetic affinities of proteins from the DR-TT core that have best hit to proteins from thermophiles.
Amino acid composition bias is known to affect not only sequence similarity searches but phylogeny reconstruction as well [64]. We tested this effect on our data set by comparing the sequence-based maximum likelihood trees to the neighbor-joining trees reconstructed from the amino acid frequencies of corresponding proteins (see Additional file 1, "Influence of amino acid composition on phylogenetic reconstructions"). We found that, in the majority of cases (>80%), the topology of the sequence-based tree was not congruent with that of the amino acid composition tree; thus, the effect of the amino acid composition on the breakdown shown in Figure 5B is unlikely to be substantial. Taken together, these results suggest that XGD involving genes from thermophiles made a measurable contribution to the evolution of the core gene set of TT after the divergence from the common DR-TT ancestor; no such contribution was detected in the case of DR. While this interpretation seems most plausible given the ecological proximity of TT and other thermophiles, it cannot be ruled out that the observed patterns (Figure 5B) are partially explained by XGD with genes from mesophilic bacteria in the DR lineage. More generally, these results emphasize that the taxonomic distribution of best database hits can be taken only as a rough and preliminary indicator of HGT.
Among the cases of potential XGD supported by phylogenetic analysis, there are two ribosomal proteins, L30 and L15, which are encoded by adjacent genes within a conserved ribosomal operon. The proteins encoded by surrounding genes in these operons showed clear affinity to the corresponding DR orthologs (data not shown). In phylogenetic trees of L30 and L15, the TT proteins reliably clustered with orthologs from bacterial hyperthermophiles and not with the corresponding DR orthologs (Figure 6A and 6B). The congruent evolutionary patterns seen with these two ribosomal proteins encoded by adjacent genes suggest that this gene pair has been replaced via XGD in situ, without disruption of the operon organization [65]. Additional cases of apparent XGD, where DR confidently partitions into the mesophilic clade, whereas TT belongs to the thermophilic clade, are shown in Figure 6C, D (in each of these cases, the thermophilic clade has an admixture of mesophilic species whereas the mesophilic clade includes no thermophiles).
Figure 6 Phylogenetic trees for selected TT genes with apparent XGD involving an ortholog from a thermophilic species. Maximum-likelihood unrooted trees were constructed and bootstrap probabilities were computed using the MOLPHY program. Branches with bootstrap probability >70% are marked by black circles. Each leaf is denoted by the standard gene identifier (for the the complete list of correspondence between genes and species, see Additional file 1: "Gene identifiers and species names for figure 6"). The DR and TT genes are boxed. Genes from thermophilic species are shown in red. A. Ribosomal protein L30. B. Ribosomal protein L15. C. Thiamin biosynthesis protein ThiC. D. tRNA thiolation enzyme TtcA.
The apparent lineage-specific HGT in DR and TT was not limited to XGD or to acquisition of genes from thermophiles. Additional examples of various types of HGT supported by phylogenetic analysis are given in Table 2.
Table 2 Examples of horizontally transferred genes in TT and DR
Selected horizontally transferred genes/operons T. thermophilus Protein ID D. radiodurans Protein ID Comment
Xenologous gene displacement
Ribosomal protein L15 (COG0200) TTC1309 DR2115 Proteobacterial affinity of DR; Affinity to Thermatoga/Aquifex/other Gram-positive bacteria in TT
Ribosomal protein L30/L7E (COG1841) TTC1310 DR2114 Thermatogales/Aquifecales version in TT, affinity to other bacteria in DR
Thiamine biosynthesis protein ThiC (COG0422) TTC0319 DRA0175 Proteobacterial affinity in DR; Cyanobacteria/Actinobacterial affinity in TT
Threonine synthase (COG0498) TTC0117 DRA0360 Proteobacterial version in DR; Cyanobacteria/Actinobacterial affinity in TT
Trk system potassium uptake protein trkG, trkA (COG0168, COG0569) TT0809, TT0810 DR1667, DR1668; DR1666 Archaeal version in TT; Gram-positive version in DR
Ribonucleoside-diphosphate reductase alpha chain and beta chains (COG0208, COG0209) TTP0162, TTP0161 DRB0108, DRB0109 Proteobacterial affinity for TT; affinity with Gram-positive bacteria in DR
Paralogous genes acquisition
5-formyltetrahydrofolate cyclo-ligase (COG0212) TTC1247, TTC1803 DR1815 Gram-positive version in DR and TT; additional pseudoparalog TT1803 appears to be of archaeal origin
Non-orthologous gene displacement/acquisition
Thymidylate synthase TTC0731 (COG1351) DR2630 (COG0207) Represented by two non-homologous genes in DR (protobacterial version) and TT (orthologs in Treponema, Wolbachia)
Purine-nucleoside phosphorylase (purine degradation enzyme) TTC1070 (COG0813) TTC0194 (COG0005) DR2166 (COG0813) TT has two non-homologous enzymes; TTC0194 probably was acquired from a thermopilic source
Expanded families of paralogs
Most bacterial lineages contain unique sets of expanded paralogous gene families [66]. This notion was borne out by the present comparative-genomic analysis of DR and TT. None of the expanded families that have been detected during the previous detailed analysis of the DR genome [7] was expanded in TT, and many were missing altogether. This strongly suggests that extensive gene duplication and acquisition of new pseudoparalogs via HGT, which led to the expansion of these families in DR, occurred after the divergence from the common ancestor with TT, and could contribute to the specific adaptations of DR (Table 3). Expansion of several other families in DR was revealed in the course of the present comparison with TT. One notable example is the family of predicted membrane-associated proteins (DR2080, DR1043, DR1952, DR1953, DR1738), which are encoded adjacent to transcriptional regulators of the PadR-like family (COG1695; 9 paralogs in DR, none in TT) (Table 3). PadR-like regulators are involved in the regulation of the cellular response to chemical stress agents, derivatives of phenolic acid [67]. Another previously unnoticed paralogous family that is expanded in DR includes proteins containing the MOSC (MOCO sulfurase C-terminal) domain (COG2258, four paralogs in DR and one in TT). These proteins have been predicted to function as sulfur-carriers that deliver sulfur for the formation of sulfur-metal clusters associated with various enzymes [68]. One of these genes (DR0273) forms a predicted operon with genes for a Nudix hydrolase and a monooxygenase, suggesting that these proteins might comprise a distinct stress response/house-cleaning complex.
Table 3 Paralogous gene families expanded in DR
Description COG numbers Number of representatives in DR Number of representatives in TT
Widespread families expanded in DR
MutT-like phosphohydrolases (Nudix) COG0494, COG1051 17,5 5,5
Calcineurin-like phosphoesterase COG1768, COG1408, COG1692, COG0639 1; 1; 1; 9 1; 0; 1; 2
Lipase-like alpha/beta hydrolase COG0596, COG0400 10; 1 6; 1
Subtilisin-like protease COG1404 10 2
Sugar deacetylase COG2120 6 4
PadR-like transcriptional regulators (possibly involved in chemical stress response) COG1695 9 0
MOSC sulfur-carrier domains COG2258 4 1
FlaR like kinases - 3 0
LigT phosphatases (may participate in RNA repair or methabolism) - 3 2
McrA endonuclease COG1403 5 1
TerZ family (could confer resistance to a variety of DNA-damaging agents) COG2310 7 0
PR1 family (stress response) COG2340 5 1
DinB family (DNA damage and stress inducible proteins) COG2318; no COG 3; 10 1
Transcriptional regulators - 5 0
Unique DR families
GRXGG repeats containing protein - DR0082, DR2593, DR1748 No
Alpha/beta proteins, tryptophan-rich - DR2532, DR2457 No
Proteins with GXTXXXG and CXPXXXC motifs (DR0871 has duplication of the domain) - DR0871, DR1920, DR2360 No
Secreted alpha/beta proteins with a single conserved domain - DR1251, DR1319, DR1545 No
Predominantly alpha-helical proteins - DR0481, DR1195, DR1301 No
Predominantly alpha-helical proteins - DR0387, DR2038+DR2039 No
Predicted metabolic regulator containing V4R domain - DR2179, DR1611 No
Predicted sirohydrochlorin cobaltochelatase COG2138 DRA0012, DR2241 No
Conserved histidine rich protein (now also found in Caulobacter and Mesorhizobium) COG3798 DR1261, DR1348 No
Several paralogous families are specifically expanded in TT (Table 4). The largest of these (15 paralogs compared to three in DR) is the Uma2 family that is highly expanded in cyanobacteria but otherwise seen in only a few bacteria. The function(s) of these proteins is unknown; the presence of conserved acidic residues suggests that they might be uncharacterized DNA-binding proteins [69]. The expansion of predicted sugar transporters in TT and the paucity of extracellular proteases (including subtilases) is unexpected because it has been shown that TT is predominantly a proteolytic rather than a saccharolytic organism [70]. However, it should be noted that TT, unlike DR, has not been observed to secrete proteases (data not shown).
Table 4 Paralogous gene families expanded in TT
Description COG numbers Number of representatives in DR Number of representatives in TT
Uncharacterized protein conserved in cyanobacteria, Uma2 homolog COG4636 3 15
Rhodanese-related sulfurtransferase COG0607 2 6
ABC-type sugar transport system, periplasmic component COG1653 1 8
ABC-type sugar transport system, permease component COG0395 1 6
ABC-type sugar transport systems, ATPase components COG3839 1 3
ABC-type sugar transport systems, permease components COG1175 2 6
ABC-type Fe3+ transport system, permease component COG1178 1 3
Fe2+/Zn2+ uptake regulation proteins COG0735 1 3
Minimal nucleotidyl transferases COG1708,COG1669 0,3 3,5
PIN-like nucleases COG1487, COG3744, COG4113, COG4374 COG1848 3,0
0,1
1 2,2
1,1
3
Antitoxin of toxin-antitoxin stability system COG4118, COG2161 1,3 3,2
Nucleotide-binding proteins of the UspA family COG0589 2 4
TRAP-type mannitol/chloroaromatic compound transport system, periplasmic component COG4663 0 3
TRAP-type mannitol/chloroaromatic compound transport system, small permease component COG4665 0 2
TRAP-type mannitol/chloroaromatic compound transport system, large permease component COG4664 0 2
Arabinose efflux permease COG2814 0 2
Predicted phosphoesterases, related to the Icc protein COG2129 0 2
HEPN, Nucleotide-binding domain COG2250 0 2
Aldehyde:ferredoxin oxidoreductase COG2414 0 3
S-adenosylmethionine decarboxylase COG1586 0 2
Tfp pilus assembly protein FimT COG4970 0 2
Tfp pilus assembly protein PilE COG4968 0 2
Notably, several protein families that are expanded in TT but are absent in DR belong to the set of potential thermophilic determinants (HEPN nucleotide-binding domain; predicted phosphoesterases related to the Icc protein) [17] or are expanded in thermophylic archaea (PIN-like nuclease domain, minimal nucleotidyl transferases, UspA-like nucleotide-binding domain) [71], Table 4). In particular, TT has three paralogs of the archaea-specific tungsten-containing aldehyde ferredoxin oxidoreductase (TTC0012, TTC1834, TTP0122, TTP0212), which is the first occurrenceof this enzyme in thermophilic bacteria. However, these enzymes are present in several mesophilic bacteria, and have various substrate specificities and might be involved in sugar, amino acids or sulfur metabolism [72,73].
Comparison of DNA repair and stress response systems
Comparative analysis of the well-characterized genetic systems for replication, repair and recombination, and related functions in TT and DR shows that fractions of these genes in the respective genomes are very similar (Table 5; see Additional file 1, Table 2S, 3S). The greatest differences were observed among the proteins associated with direct damage reversal (11 in TT versus 26 in DR), which is due to the extraordinary expansion of the NUDIX (MutT-like) family of hydrolases in DR [7]. It should be noted that the majority of these proteins have other substrates than 7,8-dihydro-8-oxoguanine-triphosphate (or diphosphate), which is cleaved by MutT. Consistently, the majority of the NUDIX proteins appear to be "house cleaning" enzymes rather than bona fide components of repair systems [74]. Other notable differences include the apparent loss of the SOS-response transcriptional repressor LexA [12] and another SOS-response protein, endonuclease VII (XseAB) in the TT lineage; these proteins seem to have been lost also by another thermophilic bacterium, AA. In contrast, DR has two LexA paralogs (DRA0344 and DRA0074), but their functions remain unclear. A genetic disruption of DRA0344, the paralog that shows greater similarity to the canonical bacterial LexA protein, does not result in sensitivity to DNA damage or impairment of RecA expression [75]. Photolyase (PhrB) and endonuclease IV (nfo) are among the few DNA repair proteins that probably were acquired by TT after the divergence from the common ancestor with DR. In addition, the catalytic subunit of DNA polymerase III of TT (DnaE, TTC1806) has two inserted inteins, whereas the orthologous DR0507 has none. In general, it seems that the conventional DNA-repair systems of TT and DR are closely related to each other and to the respective systems of other free-living bacteria. Thus, the unique, shared features of these systems do not explain the very large difference observed in resistance between TT and DR species.
Table 5 Comparison of general repair pathways in DR and TT
DNA repair pathways TT DR
DR – direct damage reversal 11 26
BER – base excision repair 10 15
NER – nucleotide excision repair 9 10
mMM – methylation-dependent mismatch repair 7 6
MM – mismatch repair - 2
MMY – mutY-dependent repair 1 1
VSP – very short path mismatch repair 2 5
RER – recombinational repair 13 15
SOS repair 5 5
MP – multiple pathways 4 4
Total number (Genome fraction) 62 (2.8%) 89 (2.5%)
However, a conclusion that there are no important differences between the repair systems of TT and DR might be premature. Recently, several additional proteins of DR have been implicated in DNA or RNA repair, either in direct experiments or on the basis of up-regulation following irradiation, complemented with protein sequence analysis. These putative repair enzymes include DRB0094, an RNA ligase [76] that is strongly up-regulated in response to irradiation [27] and might be involved in an uncharacterized RNA repair process; a predicted double-strand break repair complex specific for recovery after irradiation, which consists of DRB0100, a DNA ligase, DRB0098, a protein containing an HD family phosphatase and polynucleotide kinase domains [7]; DRB0099, a predicted phosphatase of the H2Macro superfamily ([27] and KSM, unpublished observations); a double-stranded DNA-binding protein PprA (DRA0346), which stimulates the DNA end-joining reaction in vitro [77]; a predicted DNA single-strand annealing protein DdrA (DR0423) [78]; a regulator of radiation response IrrE (DR0167); a metal-dependent protease fused to a helix-turn-helix domain [79,80]; and the uncharacterized protein DR0070 that has been shown to be essential for full resistance to acute irradiation [27]. Among these poorly characterized (predicted) repair proteins of DR, only DdrA has an ortholog in the TT genome. In general, these putative repair genes are sparsely represented in bacteria, and it appears most unlikely that they were present in the common ancestor of TT and DR; most likely, these genes were acquired by the DR lineage via HGT after the divergence from the common ancestor with TT, and might have contributed to the evolution of the resistance phenotype. However, functional relevance of these genes to radiation resistance remains to be confirmed because most of the corresponding knockout mutants showed only relatively small to moderate decreases in radiation resistance [27,78].
Among the unique (predicted) repair enzymes of TT, the most conspicuous ones are the components of the putative thermophile-specific repair system, which are predominantly encoded on the TT megaplasmid (Figure 7A). The functional features of the proteins encoded in this system (COG1203, a DNA helicase, often fused to a predicted HD-superfamily hydrolase; COG1468, a RecB-family exonuclease; COG1353, predicted polymerase) suggest that they are involved in an as yet uncharacterized DNA repair pathway. It has been hypothesized that this novel gene complex might be functionally analogous to the bacterial-eukaryotic system of translesion, mutagenic repair whose central components are DNA polymerases of the UmuC-DinB-Rad30-Rev1 superfamily, which typically are missing in thermophiles [16].
Comparison of proteins comprising various (predicted) systems involved in stress response reveals a greater number and diversity of such proteins in DR, which has 26 COGs with relevant functions that are not represented in TT compared to 3 such COGs in TT (see Additional file 1, Table 4S). Altogether, there are 147 proteins in DR in this category and 86 in TT, suggesting that some of them are additionally expanded in DR (see the section on "Expanded families").
Enzymatic systems of defense against oxidative stress predicted in TT and DR also show important differences. TT has one Mn-dependent superoxide dismutase (SodA) [81] and one Mn-dependent catalase (with no ortholog in DR), whereas DR encodes three superoxide dismutases (one of which is the ortholog of the SodA gene of TT) and three predicted catalases with no TT orthologs [7]. Additionally, DR has a cytochrome C peroxidase (DRA0301) and a predicted iron-dependent peroxidase (DRA0145), enzymes that are likely to provide protection against toxic peroxides [49]. Orthologs of these enzymes are rare among bacteria, suggesting that the Deinococcus lineage acquired them via HGT after the divergence of TT and DR from the common ancestor. Reduction of oxidized methionine residues in proteins is crucial for survival of cells under oxidative stress [82,83]. Consistent with this idea, two peptide methionine sulfoxide reductases (PMSRs), MsrA (DR1849) and MsrB (DR1378), are encoded in the DR genome [84,85], whereas none are present in TT. Interestingly, both PMSRs are also missing in Aquifex, Thermotoga and most thermophilic archaea, suggesting at least two possibilities: either this type of oxidative damage is rare at high temperatures or the known PMSRs are replaced by uncharacterized analogous enzymes due to the inefficiency of the former at high temperatures.
Oxidative stress defense mechanisms also might include control of Mn and Fe partitioning in the cell [29]. Proteins of the Dps/ferritin family are required for the storage of iron in a non-reactive state, which prevents iron-catalyzed formation of hydroxyl radicals, thus protecting the cell from iron toxicity (Fenton-type chemistry) [86]. Two Dps-related proteins are encoded in the DR genome (DR2263, DRB0092, COG1528), and it has been shown that one of them (DR2263) protects DNA from both hydroxyl radical cleavage and from DNase I-mediated cleavage [87]. Some proteins homologous to DPS can non-specifically bind DNA and therefore are viewed as DNA-specific protectors [88]. Like most thermophiles, TT has no proteins of this family but encodes a ferritin from another family (TTC1122).
Since desiccation also causes oxidative stress, proteins involved in desiccation resistance belong to the general cellular defense category [89]. Desiccation-related proteins from at least two distinct families have been detected in DR [7]. These desiccation resistance protein families (Lea 76 family, DR0105 and DR1172; and Lea14 family, DR1372) are not represented in TT. However, three TT proteins (TTP0170, TTP0166, TTP0169) are homologs of another desiccation resistance protein that was originally characterized in a plant, Craterostigma plantagineum [90]; DR also has two proteins of this family, DRB0118 and DRA0258. These proteins are distantly related to COG1633 and belong to the ferritin family of iron storage proteins (KSM, unpublished). Highly conserved homologs of these proteins are also present in thermophilic bacteria and archaea. Two desiccation-related proteins (DR1172 and DRB0118) appear to be essential for desiccation resistance but not for radiation resistance in DR [53].
Comparison of the genome partitions of TT and DR
Both TT and DR have multipartite genomes. To examine possible evolutionary relationships between the genome partitions of TT and DR, we analyzed the distribution of symmetrical best hits (putative orthologs) in the single extra-chromosomal element of TT, the pTT27 megaplasmid, and the three smaller genome partitions of DR (small chromosome, DR412; megaplasmid, DR177 and plasmid, CP1; Table 6). The results of this analysis show that pTT27 has a highly significant excess of orthologs on DR177, suggesting that these two megaplasmids are homologous, i.e., probably evolved from a distinct genome partition of the common DR-TT ancestor. Apparently, however, the genomes of the megaplasmids have undergone extensive rearrangements since the divergence from the common ancestor because no conservation of gene order could be identified (data not shown).
Table 6 Homology between the DR and TT megaplasmids
A. Number of orthologs of TT proteins, encoded in DR genome partitions
DR chromosome DR412 DR177+plasmid CP1 P(χ2)
pTT27 65 (84.2) 12 (11.9) 25 (6.0) 7 × 10-15
B. Number of orthologs of DR proteins encoded in TT genome partitions
TT chromosome pTT27 P(χ2)
DR412 89 (87. 9) 9 (10.1) 0.7
DR177 10 (26.9) 20 (3.1) 3 × 10-24
Note: Expected number of orthologs under the assumption of independent distribution is shown in parentheses.
Notably, among the putative thermophilic determinants of TT, ~50% are encoded on the megaplasmid (18 out of 36). Of these, 11 belong to the putative mobile thermophile-specific DNA repair system [16,45] (Figure 7A). Additionally, the megaplasmid carries at least four other genes associated with this system, which have not yet been assigned to COGs (Figure 7A). Moreover, the TT megaplasmid also carries a pseudogene for reverse gyrase, the most conspicuous signature protein of hyperthermophiles [15,17].
Recently, the genome of another strain of TT (HB8) has been completely sequenced and became available in public databases [13]. A preliminary comparison of the two TT strains (HB27 and HB8) revealed considerable differences in the gene orders and contents of the megaplasmids (see Additional file 1, Figure 3S). Interestingly, these differences in gene content are derived mostly from genes that appear to be associated with the thermophylic lifestyle. In particular, strain HB8 encodes an intact reverse gyrase. Thus, it appears most likely that the gene for reverse gyrase was acquired from a hyperthermophilic source by the TT lineage and was present in the common ancestor of HB27 and HB8 but decayed in the former. Conversely, HB27 encodes a unique, three-domain fusion protein (DnaQ endonuclease, DinG helicase and RecQ helicase), whereas HB8 lacks the DinG and RecQ orthologs (Figure 7B). Furthermore, there are unexpected differences in the organization of predicted thermophile-specific repair systems between the two strains of TT. Specifically, HB27 contains a "gram-positive version" (TTP0132-TTP0136), whereas HB8 has a "proteobacterial version" of these genes (TTHB186-TTHB194) (Figure 7A).
Furthermore, a nearly complete draft genome sequence of Deinococcus geothermalis (DG) has recently become publicly available [91]. Since DG is closely related to DR but is moderately thermophilic, we searched for "thermophilic" genes in DG genome. Using the "thermophilic" protein sequences ([17] and see above) of TT and DR as queries, we identified orthologs of 5 of the 6 DR proteins from this set (four of these are also present in both TT strains) and orthologs of 7 of the remaining 23 "thermophilic" proteins of TT (all components of the predicted thermophilic DNA repair system). These 7 proteins had the respective TT proteins as the best hits, and their monophyly was supported by phylogenetic analysis (data not shown), suggesting that these genes were already present in the genome of the DR-TT common ancestor.
These observations give rise to two hypotheses: (i) the TT megaplasmid is essential for the survival of the organism at high temperatures. Consistent with this idea, we detected an expansion of a two-component toxin-antitoxin system, which consists of a PIN-like nuclease (toxin) and a MazE family transcriptional regulator (antitoxin) [69]. Such a system is known to be responsible for the segregational stability of antibiotic resistance plasmids and other plasmids via selective elimination of cells that have failed to acquire a plasmid copy [92], and/or exclusion of competing plasmids [93]; and (ii) the TT megaplasmid is a dynamic genome compartment and a veritable sink for horizontally transferred genes, some of which might affect the thermophilic phenotype of this bacterium. This is compatible with the considerable differences in gene content observed between the two TT strains.
Specific roles of plasmid-borne genes in recovery from DNA damage have been proposed previously, including class Ib ribonucleotide reductase, periplasmic alkaline phosphatase, and extracellular nuclease, and subsequent analyses revealed at least five other genes implicated in this process (two operons, DRB0098-DRB0100, DRB0094-DRB0084, see above in "Comparison of DNA repair and stress response systems") [27,76]. There is also a toxin-antitoxin system operon in the DR177 megaplasmid (DRB0012a and another gene located immediately upstream of DRB0012a and currently absent from the genome annotation). Additionally, there are five other toxin-antitoxin systems encoded on DR412, the smaller chromosome of DR, which might be related to maintenance of DR177 megaplasmid in DR. The DR412 chromosome appears to have some special features as well. Numerous genes that apparently have been acquired by DR via HGT after the divergence from the common ancestor with TT and are implicated in various processes of amino acid and nucleotide degradation, map to this genome partition. Thus, the megaplasmids of TT and DR (and the smaller chromosome of DR, DR412) appear to have participated in extensive HGT, which might have been important for the evolution of thermophily and radioresistance, although the repertoires of the respective acquired genes are completely different.
Conclusion
TT and DR share a large core of genes and form a clade in the gene-content tree, which supports the idea that these bacteria form a distinct clade, as indicated previously by phylogenetic analysis of rRNA and various proteins, and that the evolution of their gene complements was, roughly, clock-like. However, major differences between the gene repertoires of TT and DR were observed, indicating that both genomes lost numerous ancestral genes and acquired distinct sets of new genes primarily via HGT. In addition, numerous lineage-specific expansions of paralogous gene families were identified, particularly, in DR.
Some of the differences in the gene repertoires of TT and DR can be linked to the distinctive adaptive strategies of these bacteria. For example, TT appears to have acquired many genes from (hyper)thermophilic bacteria and archaea, whereas DR apparently acquired various genes involved in oxidative stress response and other "house-cleaning" functions from diverse bacterial sources.
The gene content of the TT megaplasmid (pTT27) and the DR megaplasmid (DR177) are sufficiently similar to conclude that they evolved from a common ancestor. To our knowledge, this is the first evidence of persistence of a megaplasmid beyond the genus level. However, the TT megaplasmid also carries many genes whose functions are implicated in the thermophylic phenotype; in particular, components of the predicted thermophile-specific repair system. These megaplasmids are likely to be essential for the survival of both TT and DR, with their maintenance controlled by toxin-antitoxin systems. Furthermore, the substantial differences between the gene repertoires of the megaplasmids of TT strains HB27 and HB8 indicate that this genome partition has been highly dynamic, with high rates of gene loss and HGT events occurring during evolution.
The evolutionary reconstruction based on the parsimony principle is generally compatible with the idea that the common ancestor of TT and DR was a mesophilic bacterium, whereas the thermophylic phenotype of TT evolved gradually via HGT of genes from thermophiles. Conversely, the radiation-desiccation resistance phenotype of DR might have gradually evolved via HGT of genes from other mesophiles, particularly, with highly developed oxidative stress response systems. However, it should be noted that the TT-DR common gene core includes dozens of genes of apparent archaeal origin or, at least, genes with thermophilic affiliation. Moreover, the DG genome encodes a few additional "thermophilic determinants" that are missing in DR but are unlikely to have been transferred from a thermophilic source independently of TT, as shown by comparative-genomic and phylogenetic analyses described here. Thus, acquisition of a considerable number of archaeal genes might have occurred along the evolutionary branch leading to the common ancestor of TT and DR. Accordingly, at this stage, we cannot rule out the possibility that this ancestor was a moderate thermophile rather than a mesophile. Further sequencing of bacterial genomes of the Thermus-Deinococcus clade should allow more definitive comparative-genomic analysis to elucidate the nature of the common ancestor of these bacteria.
Methods
Irradiation and desiccation
Irradiations. Three TT colony-isolates were inoculated individually in liquid TGY (10 g/L Bactotryptone, 1 g/L glucose, 5 g/L yeast extract) and incubated at 70°C. Cells were harvested at OD600 ~0.9, which corresponds to 107 – 108 colony forming units (CFU)/ml; 1 TT cell/CFU. TT cells grown in TGY were examined for their total Mn and Fe content by ICP-MS (see main text). For radiation resistance assays, cells were irradiated without change of broth on ice with 60Co at 6.8 kGy/hour (60Co Gammacell irradiation unit [J. L. Shepard and Associates, Model 109]). At the indicated doses, cultures (3 biological replicates) were appropriately diluted and plated on solid medium (8 g/L Bactotryptone, 4 g/L yeast extract, 3 g/L NaCl, pH 7.3, 2.8% Bactoagar), and CFU counts were determined after 2 days' incubation at 65°C.
Desiccation. Five separate colony-isolates were pre-grown in TGY as for irradiation trials. Cell samples 106–107 cells (25 μl) were transferred to microtiter plates, which were transferred to desiccation chambers containing anhydrous calcium sulfate (drierite) and incubated at room temperature or 65°C. At the indicated times, cells were re-suspended in TGY, and CFU-survival frequencies were determined by dilution-plating on solid medium (65°C).
Genome analysis
The sets of predicted proteins of TT and DR were searched against each other for symmetrical and non-symmetrical hits using PSI-BLAST with expectation (E) value threshold of 10-5. Taxonomic affiliation was determined by best hits in non-redundant database of protein sequences at the National Center of Biotechnology Information (NIH, Bethesda) using BLASTP program [94] with default expectation value (0.01). Assignments to COGs were performed using the COGNITOR program [95] and CDD-search against COG-based profiles [96]. Contradictory assignments were resolved manually. Lineage-specific expansions (LSE) were identified as described previously [66,97]. The common genomic core was determined as follows: among genes that were not assigned to any COG, orthology relationships between TT and DR were determined via symmetrical best hits. Genes belonging to COGs, shared between TT and DR and having only one ortholog from each of the two genomes were directly assigned to the core. For multiple-paralog COGs, symmetrical best hits between GOG members were used to refine the relationships between TT and DR proteins. Members of the corresponding lineage-specific expansions were added to SymBeT pairs to form many-to-many core clusters. LBCA gene set was determined using an empirical parsimony procedure based on COG phyletic patterns (See Additional file 1: "Reconstruction of the gene set of the Last Bacterial Common Ancestor") which assigned a COG to LBCA if it was present in several diverse bacterial clades. All COGs that were present in LBCA and in TT and/or DR were assigned to the DR-TT ancestor.
Phylogenetic analysis
Multiple alignments for phylogenetic analysis were constructed using the MUSCLE program [98]; columns containing gaps in >30% of the sequences were discarded. Maximum likelihood trees were constructed using the ProtML program of the MOLPHY package by optimizing the least-square trees with local rearrangements [99]. Trees based on amino acid content were constructed from the matrix of Euclidean distances between frequency vectors using the NEIGHBOR program of the PHYLIP package [100]. Support for particular arrangements of species (relationships between DR, TT, thermophiles and mesophiles) was calculated using the bipartition analysis of bootstrap samples from original sequences (see Additional file 1, "Influence of amino acid composition on phylogenetic reconstructions"). The gene-content tree based on COG patterns was constructed using the NEIGHBOR program of the PHYLIP package [100]; the number of COGs shared between two genomes was normalized by the smaller genome size [33].
Authors' contributions
MVO performed the majority of genomic comparisons, phylogenetic analysis and helped to draft the manuscript. YIW and KSM performed some genomic comparisons and statistical analysis of the data. EKG, VYM, AV and MZ performed the experiments on the comparison of the Thermus and Deinococcus response to irradiation and desiccation and the analysis of Mn/Fe ratio of Thermus. MJD, EVK, and KSM participated in design of experiments and project coordination, and contributed to the writing of the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
(contains supplementary text and supplementary tables 1S, 2S, 3S, 4S and supplementary figures 1S, 2S, 3S)
Click here for file
Additional File 2
contains the assignments of DR and TT proteins to the COG database.
Click here for file
Additional File 6
contains compressed file for all the trees (in PHYLIP format) for 122 proteins from the reconstructed gene set of the DR-TT common ancestor that showed affinity to thermophiles (see also Additional file 1: "Phylogenetic analysis of the genes of the reconstructed DR-TT common ancestor").
Click here for file
Additional File 3
contains DR and TT protein comparison with COGs predicted to be associated with the thermophilic phenotype.
Click here for file
Additional File 4
contains the results of the reconstruction of the gene repertoire of the common ancestor of TT and DR.
Click here for file
Additional File 5
contains information on taxonomic assignments for the best hits of the TT and DR proteins.
Click here for file
Additional File 7
contains compressed file for all the trees (in PHYLIP format) for proteins from DR-TT core that had best hits to thermophiles.
Click here for file
Acknowledgements
This research was partially supported by the Office of Science (Biological and Environmental Research), U. S. Department of Energy (DOE) grant number DE-FG02-04ER63918 awarded to MJD. We are grateful to James K. Fredrickson and Heather M. Kostandarithes at Pacific Northwest National Laboratory for Fe and Mn ICP-MS analyses of T. thermophilus.
==== Refs
Weisburg WG Giovannoni SJ Woese CR The Deinococcus-Thermus phylum and the effect of rRNA composition on phylogenetic tree construction Syst Appl Microbiol 1989 11 128 134 11542160
Eisen JA The RecA protein as a model molecule for molecular systematic studies of bacteria: comparison of trees of RecAs and 16S rRNAs from the same species J Mol Evol 1995 41 1105 1123 8587109 10.1007/BF00173192
Rainey FA Nobre MF Schumann P Stackebrandt E da Costa MS Phylogenetic diversity of the deinococci as determined by 16S ribosomal DNA sequence comparison Int J Syst Bacteriol 1997 47 510 514 9103641
Gupta RS Bustard K Falah M Singh D Sequencing of heat shock protein 70 (DnaK) homologs from Deinococcus proteolyticus and Thermomicrobium roseum and their integration in a protein-based phylogeny of prokaryotes J Bacteriol 1997 179 345 357 8990285
Gupta RS Protein phylogenies and signature sequences: A reappraisal of evolutionary relationships among archaebacteria, eubacteria, and eukaryotes Microbiol Mol Biol Rev 1998 62 1435 1491 9841678
Huang YP Ito J DNA polymerase C of the thermophilic bacterium Thermus aquaticus: classification and phylogenetic analysis of the family C DNA polymerases J Mol Evol 1999 48 756 769 10229580
Makarova KS Aravind L Wolf YI Tatusov RL Minton KW Koonin EV Daly MJ Genome of the extremely radiation-resistant bacterium Deinococcus radiodurans viewed from the perspective of comparative genomics Microbiol Mol Biol Rev 2001 65 44 79 11238985 10.1128/MMBR.65.1.44-79.2001
Battista JR Earl AM Park MJ Why is Deinococcus radiodurans so resistant to ionizing radiation? Trends Microbiol 1999 7 362 365 10470044 10.1016/S0966-842X(99)01566-8
Hirsch P Gallikowski CA Siebert J Peissl K Kroppenstedt R Schumann P Stackebrandt E Anderson R Deinococcus frigens sp. nov., Deinococcus saxicola sp. nov., and Deinococcus marmoris sp. nov., low temperature and draught-tolerating, UV-resistant bacteria from continental Antarctica Syst Appl Microbiol 2004 27 636 645 15612620 10.1078/0723202042370008
Ferreira AC Nobre MF Rainey FA Silva MT Wait R Burghardt J Chung AP da Costa MS Deinococcus geothermalis sp. nov. and Deinococcus murrayi sp. nov., two extremely radiation-resistant and slightly thermophilic species from hot springs Int J Syst Bacteriol 1997 47 939 947 9336890
Holt JG Krieg NR Sneath PH Staley JT Williams ST Bergy's Manual of Determinative Bacteriology 1997 9 London , Williams@Wilkins
Henne A Bruggemann H Raasch C Wiezer A Hartsch T Liesegang H Johann A Lienard T Gohl O Martinez-Arias R Jacobi C Starkuviene V Schlenczeck S Dencker S Huber R Klenk HP Kramer W Merkl R Gottschalk G Fritz HJ The genome sequence of the extreme thermophile Thermus thermophilus Nat Biotechnol 2004 22 547 553 15064768 10.1038/nbt956
NCBI, Thermus thermophilus HB8, complete genome
White O Eisen JA Heidelberg JF Hickey EK Peterson JD Dodson RJ Haft DH Gwinn ML Nelson WC Richardson DL Moffat KS Qin H Jiang L Pamphile W Crosby M Shen M Vamathevan JJ Lam P McDonald L Utterback T Zalewski C Makarova KS Aravind L Daly MJ Fraser CM Genome sequence of the radioresistant bacterium Deinococcus radiodurans R1 Science 1999 286 1571 1577 10567266 10.1126/science.286.5444.1571
Forterre P A hot story from comparative genomics: reverse gyrase is the only hyperthermophile-specific protein Trends Genet 2002 18 236 237 12047940 10.1016/S0168-9525(02)02650-1
Makarova KS Aravind L Grishin NV Rogozin IB Koonin EV A DNA repair system specific for thermophilic Archaea and bacteria predicted by genomic context analysis Nucleic Acids Res 2002 30 482 496 11788711 10.1093/nar/30.2.482
Makarova KS Wolf YI Koonin EV Potential genomic determinants of hyperthermophily Trends Genet 2003 19 172 176 12683966 10.1016/S0168-9525(03)00047-7
McDonald JH Grasso AM Rejto LK Patterns of temperature adaptation in proteins from Methanococcus and Bacillus Mol Biol Evol 1999 16 1785 1790 10605119
Haney PJ Badger JH Buldak GL Reich CI Woese CR Olsen GJ Thermal adaptation analyzed by comparison of protein sequences from mesophilic and extremely thermophilic Methanococcus species Proc Natl Acad Sci U S A 1999 96 3578 3583 10097079 10.1073/pnas.96.7.3578
Thompson MJ Eisenberg D Transproteomic evidence of a loop-deletion mechanism for enhancing protein thermostability J Mol Biol 1999 290 595 604 10390356 10.1006/jmbi.1999.2889
Sterner R Liebl W Thermophilic adaptation of proteins Crit Rev Biochem Mol Biol 2001 36 39 106 11256505 10.1080/20014091074174
Chakravarty S Varadarajan R Elucidation of factors responsible for enhanced thermal stability of proteins: a structural genomics based study Biochemistry 2002 41 8152 8161 12069608 10.1021/bi025523t
Vieille C Zeikus GJ Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability Microbiol Mol Biol Rev 2001 65 1 43 11238984 10.1128/MMBR.65.1.1-43.2001
Mallick P Boutz DR Eisenberg D Yeates TO Genomic evidence that the intracellular proteins of archaeal microbes contain disulfide bonds Proc Natl Acad Sci U S A 2002 99 9679 9684 12107280 10.1073/pnas.142310499
Hickey DA Singer GA Genomic and proteomic adaptations to growth at high temperature Genome Biol 2004 5 117 15461805 10.1186/gb-2004-5-10-117
Tanaka M Earl AM Howell HA Park MJ Eisen JA Peterson SN Battista JR Analysis of Deinococcus radiodurans's transcriptional response to ionizing radiation and desiccation reveals novel proteins that contribute to extreme radioresistance Genetics 2004 168 21 33 15454524 10.1534/genetics.104.029249
Liu Y Zhou J Omelchenko MV Beliaev AS Venkateswaran A Stair J Wu L Thompson DK Xu D Rogozin IB Gaidamakova EK Zhai M Makarova KS Koonin EV Daly MJ Transcriptome dynamics of Deinococcus radiodurans recovering from ionizing radiation Proc Natl Acad Sci U S A 2003 100 4191 4196 12651953 10.1073/pnas.0630387100
Ghosal D Omelchenko MV Gaidamakova EK Matrosova VY Vasilenko A Venkateswaran A Zhai M Kostandarithes HM Brim H Makarova K Wackett LP Fredrickson JK Daly MJ How radiation kills cells: Survival of Deinococcus radiodurans and Shewanella oneidensis under oxidative stress FEMS Microbiol Lett 2005
Daly MJ Gaidamakova EK Matrosova VY Vasilenko A Zhai M Venkateswaran A Hess M Omelchenko MV Kostandarithes HM Makarova KS Wackett LP Fredrickson JK Ghosal D Accumulation of Mn(II) in Deinococcus radiodurans facilitates gamma-radiation resistance Science 2004 306 1025 1028 15459345 10.1126/science.1103185
Battista JR Against all odds: the survival strategies of Deinococcus radiodurans Annu Rev Microbiol 1997 51 203 224 9343349 10.1146/annurev.micro.51.1.203
Archibald FS Fridovich I Arch Biochem Biophys 1982 214 452 6284026 10.1016/0003-9861(82)90049-2
Lee Y Lee C Yoon J High temperature dependence of 2,4-dichlorophenoxyacetic acid degradation by Fe3+/H(2)O(2) system Chemosphere 2003 51 963 971 12697187 10.1016/S0045-6535(03)00043-2
Snel B Bork P Huynen MA Genome phylogeny based on gene content Nat Genet 1999 21 108 110 9916801 10.1038/5052
Wolf YI Rogozin IB Grishin NV Koonin EV Genome trees and the tree of life Trends Genet 2002 18 472 479 12175808 10.1016/S0168-9525(02)02744-0
Wolf YI Rogozin IB Grishin NV Tatusov RL Koonin EV Genome trees constructed using five different approaches suggest new major bacterial clades BMC Evol Biol 2001 1 8 11734060 10.1186/1471-2148-1-8
Korbel JO Snel B Huynen MA Bork P SHOT: a web server for the construction of genome phylogenies Trends Genet 2002 18 158 162 11858840 10.1016/S0168-9525(01)02597-5
Tatusov RL Fedorova ND Jackson JD Jacobs AR Kiryutin B Koonin EV Krylov DM Mazumder R Mekhedov SL Nikolskaya AN Rao BS Smirnov S Sverdlov AV Vasudevan S Wolf YI Yin JJ Natale DA The COG database: an updated version includes eukaryotes BMC Bioinformatics 2003 4 41 12969510 10.1186/1471-2105-4-41
Phillips RW Wiegel J Berry CJ Fliermans C Peacock AD White DC Shimkets LJ Kineococcus radiotolerans sp. nov., a radiation-resistant, gram-positive bacterium Int J Syst Evol Microbiol 2002 52 933 938 12054260 10.1099/ijs.0.02029-0
Ferreira AC Nobre MF Moore E Rainey FA Battista JR da Costa MS Characterization and radiation resistance of new isolates of Rubrobacter radiotolerans and Rubrobacter xylanophilus Extremophiles 1999 3 235 238 10591012 10.1007/s007920050121
Billi D Friedmann EI Hofer KG Caiola MG Ocampo-Friedmann R Ionizing-radiation resistance in the desiccation-tolerant cyanobacterium Chroococcidiopsis Appl Environ Microbiol 2000 66 1489 1492 10742231 10.1128/AEM.66.4.1489-1492.2000
Mirkin BG Fenner TI Galperin MY Koonin EV Algorithms for computing parsimonious evolutionary scenarios for genome evolution, the last universal common ancestor and dominance of horizontal gene transfer in the evolution of prokaryotes BMC Evol Biol 2003 3 2 12515582 10.1186/1471-2148-3-2
Olendzenski L Liu L Zhaxybayeva O Murphey R Shin DG Gogarten JP Horizontal transfer of archaeal genes into the deinococcaceae: detection by molecular and computer-based approaches J Mol Evol 2000 51 587 599 11116332
Levin I Giladi M Altman-Price N Ortenberg R Mevarech M An alternative pathway for reduced folate biosynthesis in bacteria and halophilic archaea Mol Microbiol 2004 54 1307 1318 15554970 10.1111/j.1365-2958.2004.04339.x
Mittenhuber G Phylogenetic analyses and comparative genomics of vitamin B6 (pyridoxine) and pyridoxal phosphate biosynthesis pathways J Mol Microbiol Biotechnol 2001 3 1 20 11200221
Jansen R Embden JD Gaastra W Schouls LM Identification of genes that are associated with DNA repeats in prokaryotes Mol Microbiol 2002 43 1565 1575 11952905 10.1046/j.1365-2958.2002.02839.x
Sauer U Canonaco F Heri S Perrenoud A Fischer E The soluble and membrane-bound transhydrogenases UdhA and PntAB have divergent functions in NADPH metabolism of Escherichia coli J Biol Chem 2004 279 6613 6619 14660605 10.1074/jbc.M311657200
Venkateswaran A McFarlan SC Ghosal D Minton KW Vasilenko A Makarova K Wackett LP Daly MJ Physiologic determinants of radiation resistance in Deinococcus radiodurans Appl Environ Microbiol 2000 66 2620 2626 10831446 10.1128/AEM.66.6.2620-2626.2000
Ferry JG CO dehydrogenase Annu Rev Microbiol 1995 49 305 333 8561463 10.1146/annurev.mi.49.100195.001513
Imlay JA Pathways of oxidative damage Annu Rev Microbiol 2003 57 395 418 14527285 10.1146/annurev.micro.57.030502.090938
Zhulin IB Nikolskaya AN Galperin MY Common extracellular sensory domains in transmembrane receptors for diverse signal transduction pathways in bacteria and archaea J Bacteriol 2003 185 285 294 12486065 10.1128/JB.185.1.285-294.2003
Melkonyan HS Chang WC Shapiro JP Mahadevappa M Fitzpatrick PA Kiefer MC Tomei LD Umansky SR SARPs: a family of secreted apoptosis-related proteins Proc Natl Acad Sci U S A 1997 94 13636 13641 9391078 10.1073/pnas.94.25.13636
Davis SJ Vener AV Vierstra RD Bacteriophytochromes: phytochrome-like photoreceptors from nonphotosynthetic eubacteria Science 1999 286 2517 2520 10617469 10.1126/science.286.5449.2517
Battista JR Park MJ McLemore AE Inactivation of two homologues of proteins presumed to be involved in the desiccation tolerance of plants sensitizes Deinococcus radiodurans R1 to desiccation Cryobiology 2001 43 133 139 11846468 10.1006/cryo.2001.2357
Chen X Quinn AM Wolin SL Ro ribonucleoproteins contribute to the resistance of Deinococcus radiodurans to ultraviolet irradiation Genes Dev 2000 14 777 782 10766734
Krogh BO Shuman S A poxvirus-like type IB topoisomerase family in bacteria Proc Natl Acad Sci U S A 2002 99 1853 1858 11830640 10.1073/pnas.032613199
Koonin EV Makarova KS Archaeal genomics: how much have we learned in six years and what's next? Genome Biol 2003 in press
Koonin EV Makarova KS Aravind L. Horizontal gene transfer in prokaryotes - quantification and classification AnnuRevMicrobiol 2001
Nelson KE Clayton RA Gill SR Gwinn ML Dodson RJ Haft DH Hickey EK Peterson JD Nelson WC Ketchum KA McDonald L Utterback TR Malek JA Linher KD Garrett MM Stewart AM Cotton MD Pratt MS Phillips CA Richardson D Heidelberg J Sutton GG Fleischmann RD Eisen JA Fraser CM Evidence for lateral gene transfer between Archaea and bacteria from genome sequence of Thermotoga maritima Nature 1999 399 323 329 10360571 10.1038/20601
Aravind L Tatusov RL Wolf YI Walker DR Koonin EV Evidence for massive gene exchange between archaeal and bacterial hyperthermophiles Trends Genet 1998 14 442 444 9825671 10.1016/S0168-9525(98)01553-4
Worning P Jensen LJ Nelson KE Brunak S Ussery DW Structural analysis of DNA sequence: evidence for lateral gene transfer in Thermotoga maritima Nucleic Acids Res 2000 28 706 709 10637321 10.1093/nar/28.3.706
Koski LB Golding GB The closest BLAST hit is often not the nearest neighbor J Mol Evol 2001 52 540 542 11443357
Cambillau C Claverie JM Structural and genomic correlates of hyperthermostability J Biol Chem 2000 275 32383 32386 10940293 10.1074/jbc.C000497200
McDonald JH Patterns of temperature adaptation in proteins from the bacteria Deinococcus radiodurans and Thermus thermophilus Mol Biol Evol 2001 18 741 749 11319258
Felsenstein J Inferring phylogenies 2004 Sunderland, MA , Sinauer Associates, Inc.
Omelchenko MV Makarova KS Wolf YI Rogozin IB Koonin EV Evolution of mosaic operons by horizontal gene transfer and gene displacement in situ Genome Biol 2003 4 R55 12952534 10.1186/gb-2003-4-9-r55
Jordan IK Makarova KS Wolf YI Koonin EV Gene conversions in genes encoding outer-membrane proteins in H. pylori and C. pneumoniae Trends Genet 2001 17 7 10 11163905 10.1016/S0168-9525(00)02151-X
Gury J Barthelmebs L Tran NP Divies C Cavin JF Cloning, deletion, and characterization of PadR, the transcriptional repressor of the phenolic acid decarboxylase-encoding padA gene of Lactobacillus plantarum Appl Environ Microbiol 2004 70 2146 2153 15066807 10.1128/AEM.70.4.2146-2153.2004
Anantharaman V Aravind L MOSC domains: ancient, predicted sulfur-carrier domains, present in diverse metal-sulfur cluster biosynthesis proteins including Molybdenum cofactor sulfurases FEMS Microbiol Lett 2002 207 55 61 11886751 10.1016/S0378-1097(01)00515-8
Anantharaman V Aravind L New connections in the prokaryotic toxin-antitoxin network: relationship with the eukaryotic nonsense-mediated RNA decay system Genome Biol 2003 4 R81 14659018 10.1186/gb-2003-4-12-r81
Oshima T [Comparative studies on biochemical properties of an extreme thermophile, Thermus thermophilus HB 8 (author's transl)] Seikagaku 1974 46 887 907 4280408
Makarova KS Aravind L Galperin MY Grishin NV Tatusov RL Wolf YI Koonin EV Comparative genomics of the Archaea (Euryarchaeota): evolution of conserved protein families, the stable core, and the variable shell Genome Res 1999 9 608 628 10413400
Roy R Menon AL Adams MW Aldehyde oxidoreductases from Pyrococcus furiosus Methods Enzymol 2001 331 132 144 11265456
Roy R Adams MW Characterization of a fourth tungsten-containing enzyme from the hyperthermophilic archaeon Pyrococcus furiosus J Bacteriol 2002 184 6952 6956 12446645 10.1128/JB.184.24.6952-6956.2002
Xu W Shen J Dunn CA Desai S Bessman MJ The Nudix hydrolases of Deinococcus radiodurans Mol Microbiol 2001 39 286 290 11136450 10.1046/j.1365-2958.2001.02267.x
Narumi I Satoh K Kikuchi M Funayama T Yanagisawa T Kobayashi Y Watanabe H Yamamoto K The LexA protein from Deinococcus radiodurans is not involved in RecA induction following gamma irradiation J Bacteriol 2001 183 6951 6956 11698386 10.1128/JB.183.23.6951-6956.2001
Martins A Shuman S An RNA ligase from Deinococcus radiodurans J Biol Chem 2004 279 50654 50661 15333634 10.1074/jbc.M407657200
Narumi I Satoh K Cui S Funayama T Kitayama S Watanabe H PprA: a novel protein from Deinococcus radiodurans that stimulates DNA ligation Mol Microbiol 2004 54 278 285 15458422 10.1111/j.1365-2958.2004.04272.x
Harris DR Tanaka M Saveliev SV Jolivet E Earl AM Cox MM Battista JR Preserving genome integrity: the DdrA protein of Deinococcus radiodurans R1 PLoS Biol 2004 2 e304 15361932 10.1371/journal.pbio.0020304
Earl AM Mohundro MM Mian IS Battista JR The IrrE protein of Deinococcus radiodurans R1 is a novel regulator of recA expression J Bacteriol 2002 184 6216 6224 12399492 10.1128/JB.184.22.6216-6224.2002
Hua Y Narumi I Gao G Tian B Satoh K Kitayama S Shen B PprI: a general switch responsible for extreme radioresistance of Deinococcus radiodurans Biochem Biophys Res Commun 2003 306 354 360 12804570 10.1016/S0006-291X(03)00965-3
Ludwig ML Metzger AL Pattridge KA Stallings WC Manganese superoxide dismutase from Thermus thermophilus. A structural model refined at 1.8 A resolution J Mol Biol 1991 219 335 358 2038060 10.1016/0022-2836(91)90569-R
Moskovitz J Methionine sulfoxide reductases: ubiquitous enzymes involved in antioxidant defense, protein regulation, and prevention of aging-associated diseases Biochim Biophys Acta 2005 1703 213 219 15680229
Morgan PE Dean RT Davies MJ Protective mechanisms against peptide and protein peroxides generated by singlet oxygen Free Radic Biol Med 2004 36 484 496 14975451 10.1016/j.freeradbiomed.2003.11.021
Moskovitz J Weissbach H Brot N Cloning the expression of a mammalian gene involved in the reduction of methionine sulfoxide residues in proteins Proc Natl Acad Sci U S A 1996 93 2095 2099 8700890 10.1073/pnas.93.5.2095
Grimaud R Ezraty B Mitchell JK Lafitte D Briand C Derrick PJ Barras F Repair of oxidized proteins. Identification of a new methionine sulfoxide reductase J Biol Chem 2001 276 48915 48920 11677230 10.1074/jbc.M105509200
van Vliet AH Ketley JM Park SF Penn CW The role of iron in Campylobacter gene regulation, metabolism and oxidative stress defense FEMS Microbiol Rev 2002 26 173 186 12069882 10.1016/S0168-6445(02)00095-5
Grove A Wilkinson SP Differential DNA Binding and Protection by Dimeric and Dodecameric forms of the Ferritin Homolog Dps from Deinococcus radiodurans J Mol Biol 2005 347 495 508 15755446 10.1016/j.jmb.2005.01.055
Grant RA Filman DJ Finkel SE Kolter R Hogle JM The crystal structure of Dps, a ferritin homolog that binds and protects DNA Nat Struct Biol 1998 5 294 303 9546221 10.1038/nsb0498-294
Mattimore V Battista JR Radioresistance of Deinococcus radiodurans: functions necessary to survive ionizing radiation are also necessary to survive prolonged desiccation J Bacteriol 1996 178 633 637 8550493
Piatkowski D Schneider K Salamini F Bartels D Characterization of five abscisic acid-responsive cDNA clones isolated from the desiccation-tolerant plant Craterostigma plantagineum and their relationship to other water-stress genes Plant Physiol 1990 94
JGI, Deinococcus geothermalis draft genome [http://genome.jgi psf.org/draft_microbes/deige/deige.home.html]
Grady R Hayes F Axe-Txe, a broad-spectrum proteic toxin-antitoxin system specified by a multidrug-resistant, clinical isolate of Enterococcus faecium Mol Microbiol 2003 47 1419 1432 12603745 10.1046/j.1365-2958.2003.03387.x
Cooper TF Heinemann JA Postsegregational killing does not increase plasmid stability but acts to mediate the exclusion of competing plasmids Proc Natl Acad Sci U S A 2000 97 12643 12648 11058151 10.1073/pnas.220077897
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Tatusov RL Natale DA Garkavtsev IV Tatusova TA Shankavaram UT Rao BS Kiryutin B Galperin MY Fedorova ND Koonin EV The COG database: new developments in phylogenetic classification of proteins from complete genomes Nucleic Acids Res 2001 29 22 28 11125040 10.1093/nar/29.1.22
Marchler-Bauer A Anderson JB Cherukuri PF DeWeese-Scott C Geer LY Gwadz M He S Hurwitz DI Jackson JD Ke Z Lanczycki CJ Liebert CA Liu C Lu F Marchler GH Mullokandov M Shoemaker BA Simonyan V Song JS Thiessen PA Yamashita RA Yin JJ Zhang D Bryant SH CDD: a Conserved Domain Database for protein classification Nucleic Acids Res 2005 33 Database Issue D192 6 15608175
Lespinet O Wolf YI Koonin EV Aravind L The role of lineage-specific gene family expansion in the evolution of eukaryotes Genome Res 2002 12 1048 1059 12097341 10.1101/gr.174302
Edgar RC MUSCLE: a multiple sequence alignment method with reduced time and space complexity BMC Bioinformatics 2004 5 113 15318951 10.1186/1471-2105-5-113
Hasegawa M Kishino H Saitou N On the maximum likelihood method in molecular phylogenetics J Mol Evol 1991 32 443 445 1904100 10.1007/BF02101285
Felsenstein J Inferring phylogenies from protein sequences by parsimony, distance, and likelihood methods Methods Enzymol 1996 266 418 427 8743697
|
16242020
|
PMC1274311
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Oct 20; 5:57
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-57
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-511623617510.1186/1471-2156-6-51Research ArticleAssessing the power of tag SNPs in the mapping of quantitative trait loci (QTL) with extremal and random samples Zhang Kui [email protected] Fengzhu [email protected] Section on Statistical Genetics, Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, Birmingham, AL 35294, USA2 Molecular and Computational Biology Program, Department of Biological Sciences, University of Southern California, Los Angeles, CA 90089, USA2005 19 10 2005 6 51 51 8 6 2005 19 10 2005 Copyright © 2005 Zhang and Sun; licensee BioMed Central Ltd.2005Zhang and Sun; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent studies have indicated that the human genome could be divided into regions with low haplotype diversity interspersed with regions of high haplotype diversity. In regions of low haplotype diversity, a small fraction of SNPs (tag SNPs) are sufficient to account for most of the haplotype diversity of the human genome. These tag SNPs can be extremely useful for testing the association of a marker locus with a qualitative or quantitative trait locus in that it may not be necessary to genotype all the SNPs. When tag SNPs are used to reduce the genotyping effort in association studies, it is important to know how much power is lost. It is also important to know how much power is gained when tag SNPs instead of the same number of randomly chosen SNPs are used.
Results
We design a simulation study to tackle these problems for a variety of quantitative association tests using either case-parent samples or unrelated population samples. First, the samples are generated based on the quantitative trait model with the assumption of either an extremal sampling scheme or a random sampling scheme. Second, a small number of samples are selected to determine the haplotype blocks and the tag SNPs. Third, the statistical power of the tests is evaluated using four kinds of data: (1) all the SNPs and the corresponding haplotypes, (2) the tag SNPs and the corresponding haplotypes, (3) the same number of evenly spaced SNPs with minor allele frequency greater than a threshold and the corresponding haplotypes, (4) the same number of randomly chosen SNPs and their corresponding haplotypes.
Conclusion
Our results suggest that in most situations genotyping efforts can be significantly reduced by using tag SNPs for mapping the QTL in association studies without much loss of power, which is consistent with previous studies on association mapping of qualitative traits. For all situations considered, two-locus haplotype analysis using tag SNPs are more powerful than those using the same number of randomly selected SNPs, but the degree of such power differences depends upon the sampling scheme and the population history.
==== Body
Background
Single-nucleotide polymorphism (SNP) markers are preferred over microsatellite markers in association studies because of their high abundance, low mutation rate, and suitability for high-throughput genotyping. The genome-wide association studies on dissection of human complex traits need to screen a large number of SNPs. However, it is prohibitively expensive to genotype all SNPs in an association study with the throughput of current technologies. Judicial selection of SNPs for association studies is therefore of paramount importance. The observation that the human genome can be divided into regions of high linkage disequilibrium (LD) with limited haplotype diversity interspersed with regions of low LD suggests one way of doing this. The regions with high LD are referred to as blocks in the literature. One of the objectives in the Human HapMap project is to describe the set of haplotype blocks and the SNPs that tag them.
Many methods have recently been developed for haplotype block partitioning and tag SNP selection. Available methods can be classified into two groups – block-dependent methods and block-free methods – although all of them are based on LD patterns of the human genome. The first group of methods relies on haplotype diversity or pair-wise LD measures such as D' to first partition the haplotypes into blocks and then select tag SNPs in each resulting block (e.g., [1-5]). The other methods select tag SNPs directly in accordance with LD patterns (e.g., [6-8]) or through comprehensive power computations (e.g., [9-11]) across the human genome. However, it is still not clear which method should be used in tag SNP identification. Here, we concentrate on two different methods. One is a variant of the first group of methods, which involves partitioning the haplotypes into blocks to minimize the total number of tag SNPs over a region of interest or the whole genome [4,5,12,13]. With this method, we expect to reduce the genotyping effort as much as possible. The other method selects tag SNPs based on pair-wise LD measure r2 [6], where for each SNP the maximum r2 between this SNP and tag SNPs must be greater than a pre-specified threshold.
The general procedure for using tag SNPs in association studies can be described as follows. First, a small number of samples (e.g., 40~50 individuals) are genotyped using a very dense SNP map. Second, a method or an algorithm is applied to obtain the set of tag SNPs. Third, a large number of samples is genotyped at only the tag SNP marker loci. Fourth, association tests of the SNPs with a qualitative or quantitative trait of interest are conducted using all the genotyped samples at tag SNP marker loci. The above approach can significantly reduce the genotyping effort [14], but it also causes loss in statistical power for association studies. There are two key questions. First, how much power will be lost when tag SNPs instead of all the SNPs are used. Second, how much power will be gained when tag SNPs instead of the same number of randomly chosen SNPs are used if only a given number of SNPs can be genotyped due to limited resources.
Several studies have examined these problems for association mapping of qualitative traits. Zhang et al. [12,13] studied this problem for qualitative traits with simulations and found that the loss of power was moderate – certainly much smaller than if an equivalent number of markers had been chosen randomly. Thompson et al. [15] used similar simulation procedures and observed similar results. Zhai et al. [16] did a similar simulation study but chose a set of evenly spaced SNPs with minor allele frequency greater than a threshold. They found that the power to detect association with evenly distributed SNPs can be almost the same as the power to detect association with tag SNPs, and sometimes even better. However, this study has several limitations. Zhai et al. [16] did not use haplotype based methods, which can be more powerful than methods based on individual marker when the disease susceptibility SNP is not included in the set of SNPs used for mapping. The number of tag SNPs is usually much smaller than the number of all SNPs, and the tag SNPs have a much sparse map, which can be favorable for haplotype analysis. This was clearly shown in the power comparison of Zhang et al. [12,13]. Also, the highest power occurred when the markers and the disease gene had similar allele frequencies [17,18]. However, it is very difficult to know the disease allele frequency in advance. Thus, the threshold that should be used to select SNP markers is problematic. Zhang et al. [18] studied power using tag SNPs identified by haplotype diversity and found that tag SNPs may not be efficient when the allele frequency at the marker locus is much different from the allele frequency at the disease locus. Other studies have investigated the power issue theoretically, which may not account for the complexities and heterogeneities in LD mapping of disease genes, and thus their conclusions cannot be readily extended to comprehensive haplotype-based methods [6,10,19]. Furthermore, most studies involving assessment of such power loss with tag SNPs have focused on qualitative traits using either simulations or real data sets [6,12,13,15,16]. The exception is Zhang et al. [18], who studied quantitative traits. However, they did not consider the analysis based on haplotypes. In this paper, we assess the loss of power to detect quantitative trait loci using tag SNPs through extensive simulations. One major difference between this study and previous studies is the use of haplotypes to study quantitative traits.
Methods
The coalescent with recombination
To carry out our study, we first simulated a large number of haplotypes consisting of a large number of consecutive SNPs across a genomic region. The simulation procedure and corresponding parameters are similar to simulations conducted in our previous studies [12,13]. Specifically, we used the coalescent process with recombination [20-22] to construct haplotype populations. The genealogies of haplotypes were generated with a population recombination rate r over the region of interest, and they are denoted by [0, 1] for easy presentation. Once the ancestral relationship between haplotypes has been simulated, mutations are added onto the genealogies to generate SNPs with a population mutation rate θ according to the infinite-many-site mutation model. The infinite-many-site mutation model assumes that mutations occur uniformly in [0, 1] and that a new mutation creates a new SNP that does not already exist in the population – i.e., recurrent mutations are not allowed.
It is well known that recombination hot and cold spots can give rise to discrete haplotype block-like structure [23,24]. Studies from both empirical data and simulations also suggested that haplotype blocks may be created due to genetic drift [25,26]. To accommodate such features of human evolution in our simulations, we employed four different population models to generate haplotypes. For the first population model, we assumed a constant population size with a uniform recombination rate. We set both r and θ to 200, which correspond to simulating a genomic region of about 200 kb [27]. We simulated the second haplotype population with recent population expansion but with the assumption of a uniformly distributed recombination rate. Again, both the recombination rate and the mutation rate were set at 200. We assumed that the population was constant at size 10,000 for a very long time and that it began growing exponentially until it reached the present population size of 107 from 1,500 generations ago. We also generated two haplotype populations with varied schemes of recombination hot spots. Both the mutation rate and the background recombination rate were set at 200 in this situation. For the third population model, two regions [0.20, 0.30] and [0.70, 0.80] were selected, with recombination rates 15 times higher than the background recombination rate. For the fourth population, one region [0.40, 0.60] was selected, with a recombination rate 15 times higher than the background recombination rate. The first two population models have been used in previous studies [12,13,16]. As we will describe below, the simulated disease susceptibility loci were positioned within the region [0.40, 0.60]. The two additional populations allow us to thoroughly assess the effect of recombination hot spots on the power loss because the disease susceptibility gene is not in recombination hot spot regions for the third population, while it is in a recombination hot spot region for the fourth population. For simplicity, we refer to these population models as P1, P2, P3, and P4, respectively, in the rest of this paper.
We simulated a quantitative trait locus with the frequency of the allele corresponding to the high trait value in a designated range. Here, we considered three different scenarios corresponding to rare, moderate, and common alleles for the high trait values, respectively. For each set of haplotypes, we chose a marker locus as the candidate trait locus if it satisfied two conditions: (1) the frequency of the minor allele is in a designated range, and (2) the position of the trait locus is between 0.40 and 0.60. The first condition restricts the variant allele frequency, and the second condition ensures that the candidate trait locus is approximately in the middle of the region of interest. If no such marker loci exist, this data set was discarded. If several marker loci satisfy these conditions in a data set, the marker locus closest to 0.50 was chosen as the quantitative trait locus. Once the candidate trait locus has been determined, the marker loci were selected sequentially from the left to the right along the chromosome based on the following conditions. (1) The frequency of the minor allele is at least 5%. (2) The distance between any two adjacent marker loci, including the candidate trait locus, is not less than a threshold. In this study, we set the threshold at 0.005. Because the length of the simulated genomic region is about 200 kb, the distance between two adjacent markers is at least 200*0.005 = 1 kb, resulting in at most 200 markers. In addition, the trait locus is not one of the marker loci used in mapping and is away at least 0.005 from the closest marker locus. The haplotypes at these marker loci and the trait locus were retained for further analysis.
The quantitative trait models
Based on the set of haplotypes generated above and a given quantitative trait model, we generated parents-offspring samples or population samples using either random sampling or extremal sampling. We considered the following widely used quantitative trait model at the candidate trait locus:
Qi = μ + aAi + dDi + εi, (1)
where Qi is the trait value; Ai and Di are the additive and dominant genotypic scores, respectively; and εi is a normal random variable with mean 0 and variance 1 and is independent of the genotype.Ai takes the values 1, 0, and -1, and Di takes the values 0, 1, and 0 for genotypes MM, Mm and mm, respectively, in which M is the allele corresponding to the high trait value. The additive genetic variance attributable to the locus is , the dominant genetic variance is , and the total genetic variance is , where p is the frequency of the allele corresponding to the high trait value at the trait locus and q = 1 - p. The broad-sense heritability attributable to the locus is computed by [28,29].
Here, we only considered the additive model (d = 0). For a given frequency of the allele corresponding to high trait value p, and the broad-sense heritability H2, we calculated the value of a. In this paper, we let μ = 0 and H2 = 20%. For random sampling, we generated 600 family samples consisting of unrelated individuals with their parents and 300 population samples of unrelated individuals based on the given quantitative trait model. For the family samples in extremal sampling, we chose 125 individuals with trait values in the top 20% of the population distribution of the trait values together with their parents and 125 individuals with trait values in the lower 20% of the population distribution of the trait values together with their parents. For the population samples in the extremal sampling, we selected 75 individuals with trait values in the upper 20% as cases and 75 individuals with trait values in the lower 20% as controls. We found that the power using these sample sizes for most methods is in an appropriate range under all situations considered, and thus meaningful comparisons can be made.
Algorithms for tag SNP selection and random SNP selection
Many methods have recently been developed for haplotype block partitioning and tag SNP selection. Here, we mainly focus on two methods. The first method is block-dependent, in which a dynamic programming algorithm is used to find the optimal block partition to minimize the total number of tag SNPs [5,12]. We followed commonly used definitions of blocks and tag SNPs [4,5] in our simulation. We defined blocks as in Patil et al. [4], where at least α percentage of observed or inferred haplotypes must be common haplotypes. Common haplotypes are those with frequencies greater than a threshold β. We defined tag SNPs within a block as the minimal set of SNPs that can distinguish α percentage of all observed or inferred haplotypes. We fixed α and β as 0.80 and 0.05, respectively. The second method is a block-free method based on LD measure r2 [6]. Because the statistical power of association studies is proportional to the value of r2 [30], this method has become popular in recent studies. Here, for any given subset of SNPs, all pair-wise r2 values between the SNPs in this subset and the SNPs not in this subset were calculated. For a given SNP not in the subset, we took the maximum value of r2 as its individual prediction power. The minimum value over all of the SNPs was taken as the overall prediction power. The minimum set of SNPs with prediction power exceeding a pre-specified threshold, γ, was considered as a set of tag SNPs. We adapted the greedy algorithm developed by Carlson et al. [6] to select the set of tag SNPs. For comparison, we choose an appropriate γ that enables the same number of tag SNPs for two different algorithms.
For power comparison, Zhang et al. [12,13] used a set of SNPs chosen uniformly at random among all SNPs. Zhai et al. [16] argued that researchers would prefer a set of evenly spaced SNPs with minor allele frequencies greater than a threshold if no prior knowledge of these SNPs was available in real studies, and they conducted the power comparison of association studies between them and tag SNPs. Here, we chose the same number of SNPs using both methods. The threshold for the minor allele frequency, t, was set to 0.05, 0.10, 0.15, and 0.20, respectively.
Tests of association of quantitative trait locus (QTL) by linkage disequilibrium
Linkage disequilibrium mapping studies for QTL typically use either family samples or unrelated individuals. When family data are used, the transmission/disequilibrium test (TDT) for quantitative traits can be used to test for linkage or association [32-35]. In this study, we assumed that we had n families with two parents and one offspring in each family, and we used the statistic TDTQ [33,34]. Many methods have been developed for mapping quantitative trait loci using unrelated population individuals [28,29,36]. Here, we employed the regression method to test whether there is any association between a marker locus and a QTL. Suppose we have the genotypes and the trait value of n individuals. The test of association can be implemented based on the standard linear regression model, as in equation (1). The null hypothesis is a = d = 0.
The QTL mapping using haplotype data is of great interest. Thus, we also implemented the haplotype-based method developed by Dudbridge [37] and Zaykin et al. [38]. The first method is an extension of classical TDT, and the second is based on regression analysis. Both methods estimate haplotypes and their frequencies using the EM algorithm and can account for the uncertainty of haplotype frequencies.
Results
We generated 20 sets of 2,000 haplotypes using the coalescent program with population models from P1 to P4, respectively. The QTL and the marker loci used for mapping were determined by the approach described in the Methods section. In summary, the number of markers used for mapping varies from 131 to 156. For the rare QTL, the frequency of the allele corresponding to high trait values varies from 0.040 to 0.060. For the moderate QTL, the frequency of the allele corresponding to high trait values varies from 0.125 to 0.175. For the common QTL, The frequency of the allele corresponding to high trait values varies from 0.270 to 0.329. The position of the QTL varies from 0.443 to 0.583.
These sets of haplotypes were then used to construct samples with quantitative traits. For each set of haplotypes, we generated 50 replicates of parent-offspring samples or population samples using either the random sampling scheme or the extremal sampling scheme described in the Methods section. We thus had a total of 1,000 replicates for each sampling scheme and each population. For each set of family samples, 20 pairs of parents (i.e., 80 haplotypes) were randomly selected to obtain the haplotype block partitions and tag SNPs. For each set of population samples, we chose 40 individuals (i.e., 80 haplotypes) as tagged samples. Several studies have shown that such a number of individuals can give similar block partitions and tag SNPs as a larger number of samples [12,15,26]. We calculated the test statistics based on individual SNPs and two-locus haplotypes, and we adjusted the p-values over all the markers using the Bonferroni correction. Table 1 summarizes the test methods compared in this study. The power of each test was conducted using several different kinds of data with type I error of 0.05: (1) all the SNPs, (2) the tag SNPs, (3) the same number of evenly spaced SNPs with minor allele frequency greater than a threshold, and (4) the same number of randomly chosen SNPs as the number of tag SNPs.
Table 1 Summary of methods used for power comparison.
Test Methods Description
TDT-R-SNP Random sampling, marker-by-marker analysis for family samples
TDT-R-HAP Random sampling, two-locus haplotype analysis for family samples
POP-R-SNP Random sampling, marker-by-marker analysis for population samples
POP-R-HAP Random sampling, two-locus haplotype analysis for population samples
TDT-E-SNP Extremal sampling, marker-by-marker analysis for family samples
TDT-E-HAP Extremal sampling, two-locus haplotype analysis for family samples
POP-E-SNP Extremal sampling, marker-by-marker analysis for population samples
POP-E-HAP Extremal sampling, two-locus haplotype analysis for population samples
Power comparisons
Here, we describe the results from our power study using the above methods with a significance level of 0.05. On average, 36, 44, 43, and 42 tag SNPs were selected for population models from P1 to P4, respectively. As expected, more tag SNPs were needed with the inclusion of the recombination hot spots and the population expansion. For the moderate QTL, the power results for the different testing methods and populations with the random sampling scheme and the extremal sampling scheme are shown in Figure 1 and Figure 2, respectively. Several general conclusions emerge from these figures. First, except for the degree of the power differences among the tests, Figure 1 and Figure 2 show very similar patterns, indicating that the conclusions drawn based on the random sampling scheme can be generally applicable to the situation involving the extremal sampling scheme. Second, the tests based on family samples and population samples also have similar power patterns. Third, although the power to detect the QTL based on two-locus haplotypes is generally higher than the power based on marker-by-marker analysis because of the exclusion of the QTL in the analysis. Such a gain in power depends not only upon population models but also on the methods for tag SNPs selection. Fourth, population models used in the simulation substantially affect the power patterns using the tag SNPs, the evenly spaced SNPs, and the randomly selected SNPs. Population models also affect the performance of tag SNPs. Because the power for detecting association is generally very low in population P4, we will only compare our results in the first three populations (P1, P2, and P3).
Figure 1 The power using SNPs with a random sampling scheme for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a moderate QTL. In each bin, the figure shows the power based on marker-by-marker analysis and two-locus haplotype analysis (from left to right). Between bins, it shows the power using (from left to right): (1) all SNPs; (2) the tag SNPs identified using the haplotype diversity [4]; (3) the tag SNPs identified using r2 [6]; (4) the evenly spaced SNPs with minor allele frequencies greater than 0.05; (5) the evenly spaced SNPs with minor allele frequencies greater than 0.10; (6) the evenly spaced SNPs with minor allele frequencies greater than 0.15; (7) the evenly spaced SNPs with minor allele frequencies greater than 0.05; and (8) the randomly selected SNPs. In each graph, the method having the highest power based on two-locus haplotype analysis is indicated with the "+" sign. The methods having power significantly lower than the highest one (one-sided chi-square test with 0.05 type I error rate) are indicated with the "*" sign.
Figure 2 The power using SNPs with an extremal sampling scheme for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a moderate QTL. The bars in each bin and the symbols ("+" and "*") have the same meaning with those in Figure 1.
For individual marker analysis, there are no clear patterns to indicate which approach for tag SNP selection performs the best. In population P1, the tag SNPs identified using r2 perform the best. In population P2, the tag SNPs identified using haplotype diversity outperform the tag SNPs identified using other methods. In population P3, the maximum power among the evenly spaced SNPs is the highest. These differences can be quite significant. The findings in this study are consistent with previous studies [16,18].
We emphasize that the idea behind haplotype tagging is to account for most haplotype diversity using the smallest number of tag SNPs and then to do haplotype-based analysis, not individual marker analysis. Indeed, haplotype-based analysis performs similarly in high-LD regions and outperforms in low-LD regions compared with individual marker analysis. Note that the QTL is excluded from our analysis. If the QTL is one of the marker loci analyzed, individual marker analysis can be more powerful than haplotype analysis. We envision that the chance of having the QTL in the marker set is low, and thus we suggest using haplotype analysis in real studies. Next, we concentrate on haplotype analysis. In population P1, with high LD, the performances of the three methods are similar. In population P2, with medium LD, the tag SNPs identified based on haplotype diversity perform significantly better than those selected using the other two approaches. In population P3, with regions of low LD and regions of high LD, the tag SNPs identified based on haplotype diversity perform similarly to the evenly spaced SNPs and perform significantly better than the tag SNPs selected based on r2. In all but one case, the tag SNPs identified based on haplotype diversity perform the best or are not significantly different from the best-performing ones.
For rare and common QTLs, the power patterns for random or extremal sampling, population or family sampling are also similar. Therefore, we only present the results based on random sampling of population samples in Figure 3. For rare QTLs and individual marker analysis, the tag SNPs identified based on haplotype diversity perform poorly and have smaller power even than the randomly selected SNPs in populations P1, P2, and P3. This is expected because the method based on haplotype diversity tends to select more common SNPs, while the rare SNPs have more power to detect the QTL at this situation. The tag SNPs selected based on r2 have the highest power in populations P1 and P2 and have power comparable with the evenly spaced SNPs in population P3. The evenly spaced SNPs, with minor allele frequencies close to the frequency of the high-risk allele at QTLs, have the highest power in population P3 and have power comparable with the tag SNPs selected based on r2 in populations P1 and P2. However, the evenly spaced SNPs with minor allele frequencies greater than 0.15 generally perform poorly and have the least power in all populations.
Figure 3 The power using SNPs with random sampling of population samples for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a rare and a common QTL, respectively. The bars in each bin and the symbols ("+" and "*") have the same meaning with those in Figure 1.
For common QTLs and individual marker analysis, the performances of the three methods are similar in population P1. The tag SNPs identified based on haplotype diversity performs significantly better than the other two approaches in population P2. In population P3, the tag SNPs identified based on haplotype diversity perform similarly to the evenly spaced SNPs and significantly better than the tag SNPs selected based on r2.
For rare QTLs and two-locus haplotype analysis, the tag SNPs identified based on haplotype diversity are significantly less powerful than the tag SNPs selected based on r2 or the evenly spaced SNPs in populations P1 and P2 (a one-sided chi-square test with a 0.05 type I error rate) but have power comparable to the evenly spaced SNPs in population P3. The tag SNPs selected based on r2 perform similarly to the evenly spaced SNPs in populations P1 and P2 but are significantly less powerful than the other two methods in population P3. For common QTLs and two-locus haplotype analysis, the power of tag SNPs identified based on haplotype diversity is the highest and is significantly greater than the power of the tag SNPs identified based on r2 and the evenly spaced SNPs in populations P1, P2, and P3. Again, the tag SNPs identified based on r2 have less power than the evenly spaced SNPs in populations P1, P2, and P3.
These power patterns can arise for several possible reasons. First, SNPs close to the QTL generally have more power. Therefore, one method will be more powerful than the other methods if it can choose more SNPs around the QTL than the other method methods. As an example, the QTL was positioned far away from two designated recombination hot spot regions for population model P3. The methods based on the haplotype diversity [4] and r2 [6] select more SNPs in two recombination regions but fewer SNPs around the QTL than the number of evenly spaced SNPs, resulting in less power to detect the QTL. In contrast, more tag SNPs were selected around the QTL for population model P4, and in this situation the power to detect the QTL using the tag SNPs is higher than the power when the same number of evenly spaced SNPs is used. Second, it has been suggested that both the allele frequencies of marker loci and the QTL affect the power in association studies. The power generally achieves its maximum when the minor allele frequency of SNPs is close to the frequency of the disease allele for individual marker analysis [17]. This is very clear when we compare the power of individual marker analysis based on rare QTLs and common QTLs. Here, we recorded the minor allele frequency in each of 1,000 replicates.
Figure 4 shows the histogram of the minor allele frequencies for those selected SNPs for population models P1 and P2 with random sampling of family samples and moderate QTLs. It can be seen that the distributions of minor allele frequencies for the tag SNPs identified by r2 and the evenly-spaced SNPs with minor allele frequencies greater than 0.05 are similar in Figure 4, and that they have a shape similar to the distribution of minor allele frequencies for all of the SNPs. On the other hand, the distribution of minor allele frequencies for the tag SNPs identified using haplotype diversity is different from others. This tag SNP selection method tends to choose more common SNPs in order to characterize the common haplotypes using as few tag SNPs as possible within each block.
Figure 4 The histogram of the minor allele frequencies for those selected SNPs for population models P1 and P2 with random sampling of family samples and moderate QTLs. In each column, (a) represents all SNPs (b) the tag SNPs identified using the haplotype diversity [4], (c) the tag SNPs identified using r2 [6], and (d) the evenly spaced SNPs with minor allele frequencies greater than 0.05, respectively.
Discussion
Genome-wide association studies based on linkage disequilibrium patterns play a central role in localizing genetic variation responsible for common human diseases and traits. Recently, several studies have revealed a block-like structure across the human genome. Understanding this block-like structure is essential for the current effort. It is important to develop methods for locating the haplotype block structure and the corresponding tag SNPs as well as to understand the usefulness and limitations of tag SNPs in association studies for QTL mapping. In this paper, we used Monte Carlo simulations to assess the power loss when the tag SNPs instead of all of the SNPs are used to detect the QTL in association studies. We also compared the power using tag SNPs and the power using the same number of evenly-spaced SNPs and randomly chosen SNPs. This is one of few studies to assess the power using tag SNPs to detect the QTL. Our results confirmed some conclusions from previous studies with qualitative traits and produced some novel findings. We showed that there are no clear winners for the three tag SNP selection methods studied in this paper based on individual marker analysis. For two-locus haplotype analysis and QTLs with moderate to high minor allele frequency, the tag SNPs identified based on haplotype diversity are comparable to the best approach in almost all of the situations examined. However, the power of the tag SNPs selected based on r2 can be significantly lower than the power of the best methods. On the other hand, the evenly spaced SNPs perform quite well in most situations if we know the allele frequency of the QTL, but they can express relatively low power when the allele of the QTL is incorrectly specified. For QTLs with low minor allele frequency, the power using tag SNPs identified based on haplotype diversity can be much lower than the power using the other two approaches.
Several possible improvements in future studies can be carried out using more sophisticated simulation strategies. In this paper, we used the coalescent process to simulate the haplotypes because the coalescent theory captures the essentials of the population genetic data. It also allows us to explore the effects of some key factors, such as the mutation rate, the recombination rate, and the population expansion. In addition to the populations simulated with constant population size and uniformly distributed mutation and recombination rates, we also generated haplotypes with the inclusion of recombination hot spots and rapid population expansion. Our results show that the population history has a substantial effect on the usage of tag SNPs in association studies. However, simulations based on the coalescent model may fail to capture some features of human evolution as found in real data sets. Therefore, simulations based on real data sets would be desirable. In this paper, we considered QTLs with low, moderate, and high minor allele frequencies and it was assumed that they are positioned at the center of region of interest. However, this information generally remains unknown in real studies. A possible approach would be to consider each SNP as a potential QTL, then compare the average power for all the SNPs or the average power based on different ranges of minor allele frequencies. In this paper, the samples used for tag SNP selection are a subset of the samples used to detect QTL. This strategy is different from the HapMap project, in which tag SNPs are identified by a set of samples and then can be used in virtually any studies based on the same population. However, it is far from obvious that tag SNPs chosen in this way will be the best ones for mapping genes in another sample, and there is no assessment of such consequences. Given that the Human HapMap project is nearly complete and that many researchers have attempted to use tag SNPs for their disease-mapping studies, it is clearly important to develop sophisticated simulation plans to evaluate the effectiveness of such a design.
In this paper, we presented our simulation results based on relatively high broad-sense heritability (H2 = 20%). It maybe is too high for many QTLs. We also simulated QTLs with broad-sense heritability H2 = 5%. In order that the power is relatively high (about 0.70–0.80 using all the SNPs) for meaningful comparisons, a larger sample size is required. Detailed results are provided as supporting materials (Supplemental Figure 1 and 2 [see additional file 1]). The power patterns are similar to those presented in Figure 1 to 4. In our simulations, we used the simple QTL model (a single main QTL with the additive effect) and the simple statistical methods for detecting associations (e.g., TDT and the regression analysis). Nonetheless, our simulations are still valid for many QTLs as long as each of them has the detectable marginal effect. New simulation designs are needed to investigate the effectiveness of tag SNPs in detecting those QTLs with only interactions but no marginal effects. In this study, we also constrained our simulations on a homologous population. The presence of sub-population structures in studied samples can greatly complicate the analysis, not only because it can result false association but also the blocks and tag SNPs depend on the specific populations [39,40]. A possible way around this problem is to first use unrelated SNPs to divide a general population into several homogeneous populations [41], and then obtain the haplotype block partitions and the tag SNPs and conduct the association analysis for each population.
In this study, we concentrated on two tag SNP selection methods. They can represent two distinct groups of methods. The first method is block-dependent and is based on haplotype diversity [4]. The second method is block-free and is solely based on pair-wise LD measure r2 [6]. Our results have important implications for association studies in that we found that these two methods perform differently for different population models. There are also many other methods for tag SNP selection, but only a few of them have been evaluated in previous studies. In addition, many factors, including population structure and history [39,40], marker allele frequency [12,42], marker density [12,40,42], and number of samples [13,15,26], can affect the selection of tag SNPs and their performances in association studies. It still remains unclear which method should be used in tag SNP selection for association studies, but we believe no method will perform best under all situations. Thus, it is important to determine which method is better under certain conditions in future studies.
Finally, we would like to emphasize that any method for tag SNP selection must be combined with existing biological knowledge. For example, if two adjacent SNPs are in complete LD with similar minor allele frequencies, the methods based on pair-wise LD may only choose one of them as a tag SNP. If both of them have been suggested by the previous biological knowledge to be important, there is no reason both of them should not be included in the set of tag SNPs. The best way may be to combine several methods to come up with a "consensus set" of tag SNPs with biologically important SNPs.
Conclusion
In this paper, we studied the power of tag SNPs to detect the QTL using extensive Monte Carlo simulations. Our study confirmed some conclusions from previous studies with qualitative traits and produced some novel findings, which have important implications in designing optimal association studies using tag SNPs. First, the use of tag SNPs can significantly reduces the genotyping effort without much loss of power in most situations. Second, two-locus haplotype analysis using tag SNPs are more powerful than those using the same number of randomly selected SNPs. Third, among several methods for tag SNP selection compared in this paper, there is no single method that outperforms the others in all situations. Fourth, the population structure and history and the allele frequency at the disease locus have substantial effects on the usage of tag SNPs in association studies. The effect of other factors, such as marker allele frequency and marker density, on the power of tag SNP selected by many other methods still remains unclear and needs further investigation.
Authors' contributions
Kui Zhang participated in the design of the simulation studies, conducted the simulations, performed the statistical analysis, interoperated the results, and drafted the paper. Fengzhu Sun participated in the design of the simulation studies, interpreted the results, and helped to draft the paper. All authors read and approved the final manuscript.
Supplementary Material
Additional file 1
this Microsoft Word file contains two supplemental figures (supplemental figure 1 and 2) and their legends. These figures present the power results based on the heritability of 5%.
Click here for file
Acknowledgements
The work is partially supported by NIH grant R01ES09912 (Kui Zhang) and NIH grant P50 HG 002790 (Fengzhu Sun). The authors wish to thank Peter Calabrese for providing his program to simulate the haplotype data with recombination hot spots. The authors also thank two anonymous reviewers for their thoughtful comments.
==== Refs
Daly MJ Rioux JD Schaffner SF Hudson TJ Lander ES High-resolution haplotype structure in the human genome Nat Genet 2001 29 229 232 11586305 10.1038/ng1001-229
Dawson E Abecasis GR Bumpstead S Chen Y Hunt S Beare DM Pabilal J Dibling T Tinsley E Kirby S Carter D Papaspyidonos M Livingstone S Ganske R Lõhmussaar E Zernant J Tõnisson N Remm M Mägi R Puurand T Vilo J Kurg A Rice K Deloukas P Mott R Metspalu A Bentley DR Cardon LR Dunham I A first-generation linkage disequilibrium map of human chromosome 22 Nature 2002 418 544 548 12110843 10.1038/nature00864
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J Blumenstiel B Higgins J DeFelice M Lochner A Faggart M Liu-Cordero SN Rotimi C Adeyemo A Cooper R Ward R Lander ES Altshuler D The structure of haplotype blocks in the human genome Science 2002 296 2225 2229 12029063 10.1126/science.1069424
Patil N Berno AJ Hinds DA Barrett WA Doshi JM Hacker CR Kautzer CR Lee DH Marjoribanks C McDonough DP Nguyen BTN Norris MC Sheehan JB Shen N Stern D Stokowski RP Thomas DJ Trulson MO Vyas KR Frazer KA Fodor SPA Cox DR Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21 Science 2001 294 1719 1723 11721056 10.1126/science.1065573
Zhang K Deng M Chen T Waterman MS Sun F A dynamic programming algorithm for haplotype block partitioning Proc Natl Acad Sci U S A 2002 99 7335 7339 12032283 10.1073/pnas.102186799
Carlson CS Eberle MA Rieder MJ Yi Q Kruglyak L Nickerson DA Selecting a maximally informative set of single nucleotide polymorphisms for association analyses using linkage disequilibrium Am J Hum Genet 2004 74 106 120 14681826 10.1086/381000
Halldorsson BV Bafna V Lippert R Schwartz R De La Vega FM Clark AG Istrail S Optimal haplotype block-free selection of tagging SNPs for genome-wide association studies Genome Res 2004 14 1633 1640 15289481 10.1101/gr.2570004
Meng Z Zaykin DV Xu C Wagner M Ehm MG Selection of genetic markers for association analyses using linkage disequilibrium and haplotypes Am J Hum Genet 2003 73 115 130 12796855 10.1086/376561
Chapman JM Cooper JD Todd JA Clayton DG Detecting disease associations due to linkage disequilibrium using haplotype tags: a class of tests and the determinants of statistical power Hum Hered 2003 56 18 31 14614235 10.1159/000073729
Cousin E Genin E Macce S Ricard S Chansac C Del Zompo M Deleuze JF Association studies candidate genes: strategies to select SNPs to be tested Hum Hered 2003 56 151 159 15031617 10.1159/000073200
Hu X Schrodi SJ Ross DA Cargill M Selecting tagging SNPs for association studies using power calculations from genotype data Hum Hered 2004 57 156 170 15297809 10.1159/000079246
Zhang K Qin ZS Liu JA Chen T Waterman MS Sun F Haplotype block partitioning and tag SNP selection using genotype data and their applications to association studies Genome Res 2004 14 908 916 15078859 10.1101/gr.1837404
Zhang K Calabrese P Nordborg M Sun F Haplotype structure and its applications to association studies: power and study design Am J Hum Genet 2002 71 1386 1394 12439824 10.1086/344780
Johnson GCL Esposito L Barratt BJ Smith AN Heward J Genova GD Ueda H Cordell HJ Eaves IA Dudbridge F Twells RCJ Payne F Hughes W Nutland S Stevens H Phillipa C Tuomilehto-Wolf E Tuomilehto J Gough SCL Clayton DG Todd JA Haplotype tagging for the identification of common disease genes Nat Genet 2001 29 233 237 11586306 10.1038/ng1001-233
Thompson D Stram D Goldgar D Witte JS Haplotype tagging single nucleotide polymorphisms and association studies Hum Hered 2003 56 48 55 14614238 10.1159/000073732
Zhai W Todd MJ Nielsen R Is haplotype block identification useful for association mapping studies? Genet Epidemiol 2004 27 80 83 15185406 10.1002/gepi.20014
Kaplan N Morris R Issues concerning association studies for fine mapping a susceptibility gene for a complex disease Genet Epidemiol 2001 20 432 457 11319784 10.1002/gepi.1012
Zhang W Collins A Morton N Does haplotype diversity predict power for association mapping of disease susceptibility? Hum Genet 2004 115 157 164 15221450 10.1007/s00439-004-1122-x
Weale ME Depondt C Macdonald SJ Smith A Lai PS Shorvon SD Wood NW Goldstein DB Selection and evaluation of tagging SNPs in the neuronal-sodium-channel gene SCN1A: Implications for linkage-disequilibrium gene mapping Am J Hum Genet 2003 73 551 565 12900796 10.1086/378098
Hudson RR Properties of a neutral-allele model with intergenic recombination Theor Popul Biol 1983 23 183 201 6612631 10.1016/0040-5809(83)90013-8
Kaplan NL Hudson RR The use of sample genealogies for studying a selectively neutral m-loci model with recombination Theor Popul Biol 1985 28 382 396 4071443 10.1016/0040-5809(85)90036-X
Griffiths RC Marjoram P Donnelly P, Tavaré S An ancestral recombination graph Progress in Population Genetics and Human Evolution 1997 Springer-Verlag, New York
Jeffreys AJ Kauppi L Neumann R Intensely punctuate meiotic recombination in the class II region of the major histocompatibility complex Nat Genet 2001 29 217 222 11586303 10.1038/ng1001-217
Schneider JA Peto TE Boone RA Boyce AJ Clegg JB Direct measurement of the male recombination fraction in the human β-globin hot spot Hum Mol Genet 2002 11 207 215 11823440 10.1093/hmg/11.3.207
Phillips MS Lawrence R Sachidanandam R Morris AP Balding DJ Donaldson MA Studebaker JF Ankener WM Alfisi SV Kuo FS Camisa AL Pazorov V Scott KE Carey BJ Faith J Katari G Bhatti HA Cyr JM Derohannessian V Elosua C Forman AM Grecco NM Hock CR Kuebler JM Lathrop JA Mockler MA Nachtman EP Restine SL Varde SA Hozza MJ Gelfand CA Broxholme J Abecasis GR Boyce-Jacino MT Cardon LR Chromosome-wide distribution of haplotype blocks and the role of recombination hot spots Nat Genet 2003 33 382 387 12590262 10.1038/ng1100
Wang N Akey JM Zhang K Chakraborty K Jin L Distribution of recombination crossovers and the origin of haplotype blocks: the interplay of population history, recombination, and mutation Am J Hum Genet 2002 71 1227 1334 12384857 10.1086/344398
Nordborg M Tavaré S Linkage disequilibrium: what history has to tell us? Trends Genet 2002 18 83 90 11818140 10.1016/S0168-9525(02)02557-X
Fisher RA The correlation between relatives on the supposition of Mendelian inheritance Trans Royal Soc Edinburgh 1918 52 399 433
Schork NJ Nath SK Fallin D Chakravarti A Linkage disequilibrium analysis of biallelic DNA markers, human quantitative trait loci, and threshold-defined case and control subjects Am J Hum Genet 2000 67 1208 1218 11032785
Pritchard JK Przeworski M Linkage disequilibrium in humans: models and data Am J Hum Genet 2001 69 1 14 11410837 10.1086/321275
Abecasis GR Cardon LR Cookson WOC A general test of association for quantitative traits in nuclear families Am J Hum Genet 2000 66 279 292 10631157 10.1086/302698
Monks SA Kaplan NL Removing the sampling restrictions from family-based tests of association for a quantitative-trait locus Am J Hum Genet 2000 66 576 592 10677318 10.1086/302745
Rabinowitz D A transmission disequilibrium test for quantitative trait loci Hum Hered 1997 47 342 350 9391826
Spielman RS Ewens WJ The TDT and other family based tests for linkage disequilibrium and association Am J Hum Genet 1996 59 983 9 8900224
Sun FZ Flanders WD Yang QH Zhao HY Transmission/disequilibrium tests for quantitative traits Ann Hum Genet 2000 64 555 565 11281218 10.1017/S000348000000840X
Slatkin M Disequilibrium mapping of a quantitative-trait locus in an expanding population Am J Hum Genet 1999 64 1764 72 10330364 10.1086/302413
Dudbridge F Pedigree disequilibrium tests for multilocus haplotypes Genet Epidemiol 2003 25 115 121 12916020 10.1002/gepi.10252
Zaykin DV Westfall PH Young SS Karnoub MA Wagner MJ Ehm MG Testing association of statistically inferred haplotypes with discrete and continuous traits in samples of unrelated individuals Hum Hered 2002 53 79 91 12037407 10.1159/000057986
Ke XY Durrant C Morris AP Hunt S Bentley DR Deloukas P Cardon LR Efficiency and consistency of haplotype tagging of dense SNP maps in multiple samples Hum Mol Genet 2004 13 2557 2565 15367493 10.1093/hmg/ddh294
Liu N Sawyer SL Mukherjee N Pakstis AJ Kidd JR Kidd KK Brooks AJ Zhao H Haplotype block structures show significant variation among populations Genet Epidemiol 2004 27 385 400 15389924 10.1002/gepi.20026
Pritchard JK Rosenberg NA Use of unlinked genetic markers to detect population stratification in association studies Am J Hum Genet 1999 65 220 228 10364535 10.1086/302449
Schulze TG Zhang K Chen YS Akula N Sun FZ Mcmahon FJ Defining haplotype block and tag single-nucleotide polymorphisms in the human genome Hum Mol Genet 2004 13 335 342 14681300 10.1093/hmg/ddh035
|
16236175
|
PMC1274312
|
CC BY
|
2021-01-04 16:30:39
|
no
|
BMC Genet. 2005 Oct 19; 6:51
|
utf-8
|
BMC Genet
| 2,005 |
10.1186/1471-2156-6-51
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-511623617510.1186/1471-2156-6-51Research ArticleAssessing the power of tag SNPs in the mapping of quantitative trait loci (QTL) with extremal and random samples Zhang Kui [email protected] Fengzhu [email protected] Section on Statistical Genetics, Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, Birmingham, AL 35294, USA2 Molecular and Computational Biology Program, Department of Biological Sciences, University of Southern California, Los Angeles, CA 90089, USA2005 19 10 2005 6 51 51 8 6 2005 19 10 2005 Copyright © 2005 Zhang and Sun; licensee BioMed Central Ltd.2005Zhang and Sun; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent studies have indicated that the human genome could be divided into regions with low haplotype diversity interspersed with regions of high haplotype diversity. In regions of low haplotype diversity, a small fraction of SNPs (tag SNPs) are sufficient to account for most of the haplotype diversity of the human genome. These tag SNPs can be extremely useful for testing the association of a marker locus with a qualitative or quantitative trait locus in that it may not be necessary to genotype all the SNPs. When tag SNPs are used to reduce the genotyping effort in association studies, it is important to know how much power is lost. It is also important to know how much power is gained when tag SNPs instead of the same number of randomly chosen SNPs are used.
Results
We design a simulation study to tackle these problems for a variety of quantitative association tests using either case-parent samples or unrelated population samples. First, the samples are generated based on the quantitative trait model with the assumption of either an extremal sampling scheme or a random sampling scheme. Second, a small number of samples are selected to determine the haplotype blocks and the tag SNPs. Third, the statistical power of the tests is evaluated using four kinds of data: (1) all the SNPs and the corresponding haplotypes, (2) the tag SNPs and the corresponding haplotypes, (3) the same number of evenly spaced SNPs with minor allele frequency greater than a threshold and the corresponding haplotypes, (4) the same number of randomly chosen SNPs and their corresponding haplotypes.
Conclusion
Our results suggest that in most situations genotyping efforts can be significantly reduced by using tag SNPs for mapping the QTL in association studies without much loss of power, which is consistent with previous studies on association mapping of qualitative traits. For all situations considered, two-locus haplotype analysis using tag SNPs are more powerful than those using the same number of randomly selected SNPs, but the degree of such power differences depends upon the sampling scheme and the population history.
==== Body
Background
Single-nucleotide polymorphism (SNP) markers are preferred over microsatellite markers in association studies because of their high abundance, low mutation rate, and suitability for high-throughput genotyping. The genome-wide association studies on dissection of human complex traits need to screen a large number of SNPs. However, it is prohibitively expensive to genotype all SNPs in an association study with the throughput of current technologies. Judicial selection of SNPs for association studies is therefore of paramount importance. The observation that the human genome can be divided into regions of high linkage disequilibrium (LD) with limited haplotype diversity interspersed with regions of low LD suggests one way of doing this. The regions with high LD are referred to as blocks in the literature. One of the objectives in the Human HapMap project is to describe the set of haplotype blocks and the SNPs that tag them.
Many methods have recently been developed for haplotype block partitioning and tag SNP selection. Available methods can be classified into two groups – block-dependent methods and block-free methods – although all of them are based on LD patterns of the human genome. The first group of methods relies on haplotype diversity or pair-wise LD measures such as D' to first partition the haplotypes into blocks and then select tag SNPs in each resulting block (e.g., [1-5]). The other methods select tag SNPs directly in accordance with LD patterns (e.g., [6-8]) or through comprehensive power computations (e.g., [9-11]) across the human genome. However, it is still not clear which method should be used in tag SNP identification. Here, we concentrate on two different methods. One is a variant of the first group of methods, which involves partitioning the haplotypes into blocks to minimize the total number of tag SNPs over a region of interest or the whole genome [4,5,12,13]. With this method, we expect to reduce the genotyping effort as much as possible. The other method selects tag SNPs based on pair-wise LD measure r2 [6], where for each SNP the maximum r2 between this SNP and tag SNPs must be greater than a pre-specified threshold.
The general procedure for using tag SNPs in association studies can be described as follows. First, a small number of samples (e.g., 40~50 individuals) are genotyped using a very dense SNP map. Second, a method or an algorithm is applied to obtain the set of tag SNPs. Third, a large number of samples is genotyped at only the tag SNP marker loci. Fourth, association tests of the SNPs with a qualitative or quantitative trait of interest are conducted using all the genotyped samples at tag SNP marker loci. The above approach can significantly reduce the genotyping effort [14], but it also causes loss in statistical power for association studies. There are two key questions. First, how much power will be lost when tag SNPs instead of all the SNPs are used. Second, how much power will be gained when tag SNPs instead of the same number of randomly chosen SNPs are used if only a given number of SNPs can be genotyped due to limited resources.
Several studies have examined these problems for association mapping of qualitative traits. Zhang et al. [12,13] studied this problem for qualitative traits with simulations and found that the loss of power was moderate – certainly much smaller than if an equivalent number of markers had been chosen randomly. Thompson et al. [15] used similar simulation procedures and observed similar results. Zhai et al. [16] did a similar simulation study but chose a set of evenly spaced SNPs with minor allele frequency greater than a threshold. They found that the power to detect association with evenly distributed SNPs can be almost the same as the power to detect association with tag SNPs, and sometimes even better. However, this study has several limitations. Zhai et al. [16] did not use haplotype based methods, which can be more powerful than methods based on individual marker when the disease susceptibility SNP is not included in the set of SNPs used for mapping. The number of tag SNPs is usually much smaller than the number of all SNPs, and the tag SNPs have a much sparse map, which can be favorable for haplotype analysis. This was clearly shown in the power comparison of Zhang et al. [12,13]. Also, the highest power occurred when the markers and the disease gene had similar allele frequencies [17,18]. However, it is very difficult to know the disease allele frequency in advance. Thus, the threshold that should be used to select SNP markers is problematic. Zhang et al. [18] studied power using tag SNPs identified by haplotype diversity and found that tag SNPs may not be efficient when the allele frequency at the marker locus is much different from the allele frequency at the disease locus. Other studies have investigated the power issue theoretically, which may not account for the complexities and heterogeneities in LD mapping of disease genes, and thus their conclusions cannot be readily extended to comprehensive haplotype-based methods [6,10,19]. Furthermore, most studies involving assessment of such power loss with tag SNPs have focused on qualitative traits using either simulations or real data sets [6,12,13,15,16]. The exception is Zhang et al. [18], who studied quantitative traits. However, they did not consider the analysis based on haplotypes. In this paper, we assess the loss of power to detect quantitative trait loci using tag SNPs through extensive simulations. One major difference between this study and previous studies is the use of haplotypes to study quantitative traits.
Methods
The coalescent with recombination
To carry out our study, we first simulated a large number of haplotypes consisting of a large number of consecutive SNPs across a genomic region. The simulation procedure and corresponding parameters are similar to simulations conducted in our previous studies [12,13]. Specifically, we used the coalescent process with recombination [20-22] to construct haplotype populations. The genealogies of haplotypes were generated with a population recombination rate r over the region of interest, and they are denoted by [0, 1] for easy presentation. Once the ancestral relationship between haplotypes has been simulated, mutations are added onto the genealogies to generate SNPs with a population mutation rate θ according to the infinite-many-site mutation model. The infinite-many-site mutation model assumes that mutations occur uniformly in [0, 1] and that a new mutation creates a new SNP that does not already exist in the population – i.e., recurrent mutations are not allowed.
It is well known that recombination hot and cold spots can give rise to discrete haplotype block-like structure [23,24]. Studies from both empirical data and simulations also suggested that haplotype blocks may be created due to genetic drift [25,26]. To accommodate such features of human evolution in our simulations, we employed four different population models to generate haplotypes. For the first population model, we assumed a constant population size with a uniform recombination rate. We set both r and θ to 200, which correspond to simulating a genomic region of about 200 kb [27]. We simulated the second haplotype population with recent population expansion but with the assumption of a uniformly distributed recombination rate. Again, both the recombination rate and the mutation rate were set at 200. We assumed that the population was constant at size 10,000 for a very long time and that it began growing exponentially until it reached the present population size of 107 from 1,500 generations ago. We also generated two haplotype populations with varied schemes of recombination hot spots. Both the mutation rate and the background recombination rate were set at 200 in this situation. For the third population model, two regions [0.20, 0.30] and [0.70, 0.80] were selected, with recombination rates 15 times higher than the background recombination rate. For the fourth population, one region [0.40, 0.60] was selected, with a recombination rate 15 times higher than the background recombination rate. The first two population models have been used in previous studies [12,13,16]. As we will describe below, the simulated disease susceptibility loci were positioned within the region [0.40, 0.60]. The two additional populations allow us to thoroughly assess the effect of recombination hot spots on the power loss because the disease susceptibility gene is not in recombination hot spot regions for the third population, while it is in a recombination hot spot region for the fourth population. For simplicity, we refer to these population models as P1, P2, P3, and P4, respectively, in the rest of this paper.
We simulated a quantitative trait locus with the frequency of the allele corresponding to the high trait value in a designated range. Here, we considered three different scenarios corresponding to rare, moderate, and common alleles for the high trait values, respectively. For each set of haplotypes, we chose a marker locus as the candidate trait locus if it satisfied two conditions: (1) the frequency of the minor allele is in a designated range, and (2) the position of the trait locus is between 0.40 and 0.60. The first condition restricts the variant allele frequency, and the second condition ensures that the candidate trait locus is approximately in the middle of the region of interest. If no such marker loci exist, this data set was discarded. If several marker loci satisfy these conditions in a data set, the marker locus closest to 0.50 was chosen as the quantitative trait locus. Once the candidate trait locus has been determined, the marker loci were selected sequentially from the left to the right along the chromosome based on the following conditions. (1) The frequency of the minor allele is at least 5%. (2) The distance between any two adjacent marker loci, including the candidate trait locus, is not less than a threshold. In this study, we set the threshold at 0.005. Because the length of the simulated genomic region is about 200 kb, the distance between two adjacent markers is at least 200*0.005 = 1 kb, resulting in at most 200 markers. In addition, the trait locus is not one of the marker loci used in mapping and is away at least 0.005 from the closest marker locus. The haplotypes at these marker loci and the trait locus were retained for further analysis.
The quantitative trait models
Based on the set of haplotypes generated above and a given quantitative trait model, we generated parents-offspring samples or population samples using either random sampling or extremal sampling. We considered the following widely used quantitative trait model at the candidate trait locus:
Qi = μ + aAi + dDi + εi, (1)
where Qi is the trait value; Ai and Di are the additive and dominant genotypic scores, respectively; and εi is a normal random variable with mean 0 and variance 1 and is independent of the genotype.Ai takes the values 1, 0, and -1, and Di takes the values 0, 1, and 0 for genotypes MM, Mm and mm, respectively, in which M is the allele corresponding to the high trait value. The additive genetic variance attributable to the locus is , the dominant genetic variance is , and the total genetic variance is , where p is the frequency of the allele corresponding to the high trait value at the trait locus and q = 1 - p. The broad-sense heritability attributable to the locus is computed by [28,29].
Here, we only considered the additive model (d = 0). For a given frequency of the allele corresponding to high trait value p, and the broad-sense heritability H2, we calculated the value of a. In this paper, we let μ = 0 and H2 = 20%. For random sampling, we generated 600 family samples consisting of unrelated individuals with their parents and 300 population samples of unrelated individuals based on the given quantitative trait model. For the family samples in extremal sampling, we chose 125 individuals with trait values in the top 20% of the population distribution of the trait values together with their parents and 125 individuals with trait values in the lower 20% of the population distribution of the trait values together with their parents. For the population samples in the extremal sampling, we selected 75 individuals with trait values in the upper 20% as cases and 75 individuals with trait values in the lower 20% as controls. We found that the power using these sample sizes for most methods is in an appropriate range under all situations considered, and thus meaningful comparisons can be made.
Algorithms for tag SNP selection and random SNP selection
Many methods have recently been developed for haplotype block partitioning and tag SNP selection. Here, we mainly focus on two methods. The first method is block-dependent, in which a dynamic programming algorithm is used to find the optimal block partition to minimize the total number of tag SNPs [5,12]. We followed commonly used definitions of blocks and tag SNPs [4,5] in our simulation. We defined blocks as in Patil et al. [4], where at least α percentage of observed or inferred haplotypes must be common haplotypes. Common haplotypes are those with frequencies greater than a threshold β. We defined tag SNPs within a block as the minimal set of SNPs that can distinguish α percentage of all observed or inferred haplotypes. We fixed α and β as 0.80 and 0.05, respectively. The second method is a block-free method based on LD measure r2 [6]. Because the statistical power of association studies is proportional to the value of r2 [30], this method has become popular in recent studies. Here, for any given subset of SNPs, all pair-wise r2 values between the SNPs in this subset and the SNPs not in this subset were calculated. For a given SNP not in the subset, we took the maximum value of r2 as its individual prediction power. The minimum value over all of the SNPs was taken as the overall prediction power. The minimum set of SNPs with prediction power exceeding a pre-specified threshold, γ, was considered as a set of tag SNPs. We adapted the greedy algorithm developed by Carlson et al. [6] to select the set of tag SNPs. For comparison, we choose an appropriate γ that enables the same number of tag SNPs for two different algorithms.
For power comparison, Zhang et al. [12,13] used a set of SNPs chosen uniformly at random among all SNPs. Zhai et al. [16] argued that researchers would prefer a set of evenly spaced SNPs with minor allele frequencies greater than a threshold if no prior knowledge of these SNPs was available in real studies, and they conducted the power comparison of association studies between them and tag SNPs. Here, we chose the same number of SNPs using both methods. The threshold for the minor allele frequency, t, was set to 0.05, 0.10, 0.15, and 0.20, respectively.
Tests of association of quantitative trait locus (QTL) by linkage disequilibrium
Linkage disequilibrium mapping studies for QTL typically use either family samples or unrelated individuals. When family data are used, the transmission/disequilibrium test (TDT) for quantitative traits can be used to test for linkage or association [32-35]. In this study, we assumed that we had n families with two parents and one offspring in each family, and we used the statistic TDTQ [33,34]. Many methods have been developed for mapping quantitative trait loci using unrelated population individuals [28,29,36]. Here, we employed the regression method to test whether there is any association between a marker locus and a QTL. Suppose we have the genotypes and the trait value of n individuals. The test of association can be implemented based on the standard linear regression model, as in equation (1). The null hypothesis is a = d = 0.
The QTL mapping using haplotype data is of great interest. Thus, we also implemented the haplotype-based method developed by Dudbridge [37] and Zaykin et al. [38]. The first method is an extension of classical TDT, and the second is based on regression analysis. Both methods estimate haplotypes and their frequencies using the EM algorithm and can account for the uncertainty of haplotype frequencies.
Results
We generated 20 sets of 2,000 haplotypes using the coalescent program with population models from P1 to P4, respectively. The QTL and the marker loci used for mapping were determined by the approach described in the Methods section. In summary, the number of markers used for mapping varies from 131 to 156. For the rare QTL, the frequency of the allele corresponding to high trait values varies from 0.040 to 0.060. For the moderate QTL, the frequency of the allele corresponding to high trait values varies from 0.125 to 0.175. For the common QTL, The frequency of the allele corresponding to high trait values varies from 0.270 to 0.329. The position of the QTL varies from 0.443 to 0.583.
These sets of haplotypes were then used to construct samples with quantitative traits. For each set of haplotypes, we generated 50 replicates of parent-offspring samples or population samples using either the random sampling scheme or the extremal sampling scheme described in the Methods section. We thus had a total of 1,000 replicates for each sampling scheme and each population. For each set of family samples, 20 pairs of parents (i.e., 80 haplotypes) were randomly selected to obtain the haplotype block partitions and tag SNPs. For each set of population samples, we chose 40 individuals (i.e., 80 haplotypes) as tagged samples. Several studies have shown that such a number of individuals can give similar block partitions and tag SNPs as a larger number of samples [12,15,26]. We calculated the test statistics based on individual SNPs and two-locus haplotypes, and we adjusted the p-values over all the markers using the Bonferroni correction. Table 1 summarizes the test methods compared in this study. The power of each test was conducted using several different kinds of data with type I error of 0.05: (1) all the SNPs, (2) the tag SNPs, (3) the same number of evenly spaced SNPs with minor allele frequency greater than a threshold, and (4) the same number of randomly chosen SNPs as the number of tag SNPs.
Table 1 Summary of methods used for power comparison.
Test Methods Description
TDT-R-SNP Random sampling, marker-by-marker analysis for family samples
TDT-R-HAP Random sampling, two-locus haplotype analysis for family samples
POP-R-SNP Random sampling, marker-by-marker analysis for population samples
POP-R-HAP Random sampling, two-locus haplotype analysis for population samples
TDT-E-SNP Extremal sampling, marker-by-marker analysis for family samples
TDT-E-HAP Extremal sampling, two-locus haplotype analysis for family samples
POP-E-SNP Extremal sampling, marker-by-marker analysis for population samples
POP-E-HAP Extremal sampling, two-locus haplotype analysis for population samples
Power comparisons
Here, we describe the results from our power study using the above methods with a significance level of 0.05. On average, 36, 44, 43, and 42 tag SNPs were selected for population models from P1 to P4, respectively. As expected, more tag SNPs were needed with the inclusion of the recombination hot spots and the population expansion. For the moderate QTL, the power results for the different testing methods and populations with the random sampling scheme and the extremal sampling scheme are shown in Figure 1 and Figure 2, respectively. Several general conclusions emerge from these figures. First, except for the degree of the power differences among the tests, Figure 1 and Figure 2 show very similar patterns, indicating that the conclusions drawn based on the random sampling scheme can be generally applicable to the situation involving the extremal sampling scheme. Second, the tests based on family samples and population samples also have similar power patterns. Third, although the power to detect the QTL based on two-locus haplotypes is generally higher than the power based on marker-by-marker analysis because of the exclusion of the QTL in the analysis. Such a gain in power depends not only upon population models but also on the methods for tag SNPs selection. Fourth, population models used in the simulation substantially affect the power patterns using the tag SNPs, the evenly spaced SNPs, and the randomly selected SNPs. Population models also affect the performance of tag SNPs. Because the power for detecting association is generally very low in population P4, we will only compare our results in the first three populations (P1, P2, and P3).
Figure 1 The power using SNPs with a random sampling scheme for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a moderate QTL. In each bin, the figure shows the power based on marker-by-marker analysis and two-locus haplotype analysis (from left to right). Between bins, it shows the power using (from left to right): (1) all SNPs; (2) the tag SNPs identified using the haplotype diversity [4]; (3) the tag SNPs identified using r2 [6]; (4) the evenly spaced SNPs with minor allele frequencies greater than 0.05; (5) the evenly spaced SNPs with minor allele frequencies greater than 0.10; (6) the evenly spaced SNPs with minor allele frequencies greater than 0.15; (7) the evenly spaced SNPs with minor allele frequencies greater than 0.05; and (8) the randomly selected SNPs. In each graph, the method having the highest power based on two-locus haplotype analysis is indicated with the "+" sign. The methods having power significantly lower than the highest one (one-sided chi-square test with 0.05 type I error rate) are indicated with the "*" sign.
Figure 2 The power using SNPs with an extremal sampling scheme for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a moderate QTL. The bars in each bin and the symbols ("+" and "*") have the same meaning with those in Figure 1.
For individual marker analysis, there are no clear patterns to indicate which approach for tag SNP selection performs the best. In population P1, the tag SNPs identified using r2 perform the best. In population P2, the tag SNPs identified using haplotype diversity outperform the tag SNPs identified using other methods. In population P3, the maximum power among the evenly spaced SNPs is the highest. These differences can be quite significant. The findings in this study are consistent with previous studies [16,18].
We emphasize that the idea behind haplotype tagging is to account for most haplotype diversity using the smallest number of tag SNPs and then to do haplotype-based analysis, not individual marker analysis. Indeed, haplotype-based analysis performs similarly in high-LD regions and outperforms in low-LD regions compared with individual marker analysis. Note that the QTL is excluded from our analysis. If the QTL is one of the marker loci analyzed, individual marker analysis can be more powerful than haplotype analysis. We envision that the chance of having the QTL in the marker set is low, and thus we suggest using haplotype analysis in real studies. Next, we concentrate on haplotype analysis. In population P1, with high LD, the performances of the three methods are similar. In population P2, with medium LD, the tag SNPs identified based on haplotype diversity perform significantly better than those selected using the other two approaches. In population P3, with regions of low LD and regions of high LD, the tag SNPs identified based on haplotype diversity perform similarly to the evenly spaced SNPs and perform significantly better than the tag SNPs selected based on r2. In all but one case, the tag SNPs identified based on haplotype diversity perform the best or are not significantly different from the best-performing ones.
For rare and common QTLs, the power patterns for random or extremal sampling, population or family sampling are also similar. Therefore, we only present the results based on random sampling of population samples in Figure 3. For rare QTLs and individual marker analysis, the tag SNPs identified based on haplotype diversity perform poorly and have smaller power even than the randomly selected SNPs in populations P1, P2, and P3. This is expected because the method based on haplotype diversity tends to select more common SNPs, while the rare SNPs have more power to detect the QTL at this situation. The tag SNPs selected based on r2 have the highest power in populations P1 and P2 and have power comparable with the evenly spaced SNPs in population P3. The evenly spaced SNPs, with minor allele frequencies close to the frequency of the high-risk allele at QTLs, have the highest power in population P3 and have power comparable with the tag SNPs selected based on r2 in populations P1 and P2. However, the evenly spaced SNPs with minor allele frequencies greater than 0.15 generally perform poorly and have the least power in all populations.
Figure 3 The power using SNPs with random sampling of population samples for different population models. The power is obtained using single-SNP and two-locus haplotype data based on 1,000 simulations with a rare and a common QTL, respectively. The bars in each bin and the symbols ("+" and "*") have the same meaning with those in Figure 1.
For common QTLs and individual marker analysis, the performances of the three methods are similar in population P1. The tag SNPs identified based on haplotype diversity performs significantly better than the other two approaches in population P2. In population P3, the tag SNPs identified based on haplotype diversity perform similarly to the evenly spaced SNPs and significantly better than the tag SNPs selected based on r2.
For rare QTLs and two-locus haplotype analysis, the tag SNPs identified based on haplotype diversity are significantly less powerful than the tag SNPs selected based on r2 or the evenly spaced SNPs in populations P1 and P2 (a one-sided chi-square test with a 0.05 type I error rate) but have power comparable to the evenly spaced SNPs in population P3. The tag SNPs selected based on r2 perform similarly to the evenly spaced SNPs in populations P1 and P2 but are significantly less powerful than the other two methods in population P3. For common QTLs and two-locus haplotype analysis, the power of tag SNPs identified based on haplotype diversity is the highest and is significantly greater than the power of the tag SNPs identified based on r2 and the evenly spaced SNPs in populations P1, P2, and P3. Again, the tag SNPs identified based on r2 have less power than the evenly spaced SNPs in populations P1, P2, and P3.
These power patterns can arise for several possible reasons. First, SNPs close to the QTL generally have more power. Therefore, one method will be more powerful than the other methods if it can choose more SNPs around the QTL than the other method methods. As an example, the QTL was positioned far away from two designated recombination hot spot regions for population model P3. The methods based on the haplotype diversity [4] and r2 [6] select more SNPs in two recombination regions but fewer SNPs around the QTL than the number of evenly spaced SNPs, resulting in less power to detect the QTL. In contrast, more tag SNPs were selected around the QTL for population model P4, and in this situation the power to detect the QTL using the tag SNPs is higher than the power when the same number of evenly spaced SNPs is used. Second, it has been suggested that both the allele frequencies of marker loci and the QTL affect the power in association studies. The power generally achieves its maximum when the minor allele frequency of SNPs is close to the frequency of the disease allele for individual marker analysis [17]. This is very clear when we compare the power of individual marker analysis based on rare QTLs and common QTLs. Here, we recorded the minor allele frequency in each of 1,000 replicates.
Figure 4 shows the histogram of the minor allele frequencies for those selected SNPs for population models P1 and P2 with random sampling of family samples and moderate QTLs. It can be seen that the distributions of minor allele frequencies for the tag SNPs identified by r2 and the evenly-spaced SNPs with minor allele frequencies greater than 0.05 are similar in Figure 4, and that they have a shape similar to the distribution of minor allele frequencies for all of the SNPs. On the other hand, the distribution of minor allele frequencies for the tag SNPs identified using haplotype diversity is different from others. This tag SNP selection method tends to choose more common SNPs in order to characterize the common haplotypes using as few tag SNPs as possible within each block.
Figure 4 The histogram of the minor allele frequencies for those selected SNPs for population models P1 and P2 with random sampling of family samples and moderate QTLs. In each column, (a) represents all SNPs (b) the tag SNPs identified using the haplotype diversity [4], (c) the tag SNPs identified using r2 [6], and (d) the evenly spaced SNPs with minor allele frequencies greater than 0.05, respectively.
Discussion
Genome-wide association studies based on linkage disequilibrium patterns play a central role in localizing genetic variation responsible for common human diseases and traits. Recently, several studies have revealed a block-like structure across the human genome. Understanding this block-like structure is essential for the current effort. It is important to develop methods for locating the haplotype block structure and the corresponding tag SNPs as well as to understand the usefulness and limitations of tag SNPs in association studies for QTL mapping. In this paper, we used Monte Carlo simulations to assess the power loss when the tag SNPs instead of all of the SNPs are used to detect the QTL in association studies. We also compared the power using tag SNPs and the power using the same number of evenly-spaced SNPs and randomly chosen SNPs. This is one of few studies to assess the power using tag SNPs to detect the QTL. Our results confirmed some conclusions from previous studies with qualitative traits and produced some novel findings. We showed that there are no clear winners for the three tag SNP selection methods studied in this paper based on individual marker analysis. For two-locus haplotype analysis and QTLs with moderate to high minor allele frequency, the tag SNPs identified based on haplotype diversity are comparable to the best approach in almost all of the situations examined. However, the power of the tag SNPs selected based on r2 can be significantly lower than the power of the best methods. On the other hand, the evenly spaced SNPs perform quite well in most situations if we know the allele frequency of the QTL, but they can express relatively low power when the allele of the QTL is incorrectly specified. For QTLs with low minor allele frequency, the power using tag SNPs identified based on haplotype diversity can be much lower than the power using the other two approaches.
Several possible improvements in future studies can be carried out using more sophisticated simulation strategies. In this paper, we used the coalescent process to simulate the haplotypes because the coalescent theory captures the essentials of the population genetic data. It also allows us to explore the effects of some key factors, such as the mutation rate, the recombination rate, and the population expansion. In addition to the populations simulated with constant population size and uniformly distributed mutation and recombination rates, we also generated haplotypes with the inclusion of recombination hot spots and rapid population expansion. Our results show that the population history has a substantial effect on the usage of tag SNPs in association studies. However, simulations based on the coalescent model may fail to capture some features of human evolution as found in real data sets. Therefore, simulations based on real data sets would be desirable. In this paper, we considered QTLs with low, moderate, and high minor allele frequencies and it was assumed that they are positioned at the center of region of interest. However, this information generally remains unknown in real studies. A possible approach would be to consider each SNP as a potential QTL, then compare the average power for all the SNPs or the average power based on different ranges of minor allele frequencies. In this paper, the samples used for tag SNP selection are a subset of the samples used to detect QTL. This strategy is different from the HapMap project, in which tag SNPs are identified by a set of samples and then can be used in virtually any studies based on the same population. However, it is far from obvious that tag SNPs chosen in this way will be the best ones for mapping genes in another sample, and there is no assessment of such consequences. Given that the Human HapMap project is nearly complete and that many researchers have attempted to use tag SNPs for their disease-mapping studies, it is clearly important to develop sophisticated simulation plans to evaluate the effectiveness of such a design.
In this paper, we presented our simulation results based on relatively high broad-sense heritability (H2 = 20%). It maybe is too high for many QTLs. We also simulated QTLs with broad-sense heritability H2 = 5%. In order that the power is relatively high (about 0.70–0.80 using all the SNPs) for meaningful comparisons, a larger sample size is required. Detailed results are provided as supporting materials (Supplemental Figure 1 and 2 [see additional file 1]). The power patterns are similar to those presented in Figure 1 to 4. In our simulations, we used the simple QTL model (a single main QTL with the additive effect) and the simple statistical methods for detecting associations (e.g., TDT and the regression analysis). Nonetheless, our simulations are still valid for many QTLs as long as each of them has the detectable marginal effect. New simulation designs are needed to investigate the effectiveness of tag SNPs in detecting those QTLs with only interactions but no marginal effects. In this study, we also constrained our simulations on a homologous population. The presence of sub-population structures in studied samples can greatly complicate the analysis, not only because it can result false association but also the blocks and tag SNPs depend on the specific populations [39,40]. A possible way around this problem is to first use unrelated SNPs to divide a general population into several homogeneous populations [41], and then obtain the haplotype block partitions and the tag SNPs and conduct the association analysis for each population.
In this study, we concentrated on two tag SNP selection methods. They can represent two distinct groups of methods. The first method is block-dependent and is based on haplotype diversity [4]. The second method is block-free and is solely based on pair-wise LD measure r2 [6]. Our results have important implications for association studies in that we found that these two methods perform differently for different population models. There are also many other methods for tag SNP selection, but only a few of them have been evaluated in previous studies. In addition, many factors, including population structure and history [39,40], marker allele frequency [12,42], marker density [12,40,42], and number of samples [13,15,26], can affect the selection of tag SNPs and their performances in association studies. It still remains unclear which method should be used in tag SNP selection for association studies, but we believe no method will perform best under all situations. Thus, it is important to determine which method is better under certain conditions in future studies.
Finally, we would like to emphasize that any method for tag SNP selection must be combined with existing biological knowledge. For example, if two adjacent SNPs are in complete LD with similar minor allele frequencies, the methods based on pair-wise LD may only choose one of them as a tag SNP. If both of them have been suggested by the previous biological knowledge to be important, there is no reason both of them should not be included in the set of tag SNPs. The best way may be to combine several methods to come up with a "consensus set" of tag SNPs with biologically important SNPs.
Conclusion
In this paper, we studied the power of tag SNPs to detect the QTL using extensive Monte Carlo simulations. Our study confirmed some conclusions from previous studies with qualitative traits and produced some novel findings, which have important implications in designing optimal association studies using tag SNPs. First, the use of tag SNPs can significantly reduces the genotyping effort without much loss of power in most situations. Second, two-locus haplotype analysis using tag SNPs are more powerful than those using the same number of randomly selected SNPs. Third, among several methods for tag SNP selection compared in this paper, there is no single method that outperforms the others in all situations. Fourth, the population structure and history and the allele frequency at the disease locus have substantial effects on the usage of tag SNPs in association studies. The effect of other factors, such as marker allele frequency and marker density, on the power of tag SNP selected by many other methods still remains unclear and needs further investigation.
Authors' contributions
Kui Zhang participated in the design of the simulation studies, conducted the simulations, performed the statistical analysis, interoperated the results, and drafted the paper. Fengzhu Sun participated in the design of the simulation studies, interpreted the results, and helped to draft the paper. All authors read and approved the final manuscript.
Supplementary Material
Additional file 1
this Microsoft Word file contains two supplemental figures (supplemental figure 1 and 2) and their legends. These figures present the power results based on the heritability of 5%.
Click here for file
Acknowledgements
The work is partially supported by NIH grant R01ES09912 (Kui Zhang) and NIH grant P50 HG 002790 (Fengzhu Sun). The authors wish to thank Peter Calabrese for providing his program to simulate the haplotype data with recombination hot spots. The authors also thank two anonymous reviewers for their thoughtful comments.
==== Refs
Daly MJ Rioux JD Schaffner SF Hudson TJ Lander ES High-resolution haplotype structure in the human genome Nat Genet 2001 29 229 232 11586305 10.1038/ng1001-229
Dawson E Abecasis GR Bumpstead S Chen Y Hunt S Beare DM Pabilal J Dibling T Tinsley E Kirby S Carter D Papaspyidonos M Livingstone S Ganske R Lõhmussaar E Zernant J Tõnisson N Remm M Mägi R Puurand T Vilo J Kurg A Rice K Deloukas P Mott R Metspalu A Bentley DR Cardon LR Dunham I A first-generation linkage disequilibrium map of human chromosome 22 Nature 2002 418 544 548 12110843 10.1038/nature00864
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J Blumenstiel B Higgins J DeFelice M Lochner A Faggart M Liu-Cordero SN Rotimi C Adeyemo A Cooper R Ward R Lander ES Altshuler D The structure of haplotype blocks in the human genome Science 2002 296 2225 2229 12029063 10.1126/science.1069424
Patil N Berno AJ Hinds DA Barrett WA Doshi JM Hacker CR Kautzer CR Lee DH Marjoribanks C McDonough DP Nguyen BTN Norris MC Sheehan JB Shen N Stern D Stokowski RP Thomas DJ Trulson MO Vyas KR Frazer KA Fodor SPA Cox DR Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21 Science 2001 294 1719 1723 11721056 10.1126/science.1065573
Zhang K Deng M Chen T Waterman MS Sun F A dynamic programming algorithm for haplotype block partitioning Proc Natl Acad Sci U S A 2002 99 7335 7339 12032283 10.1073/pnas.102186799
Carlson CS Eberle MA Rieder MJ Yi Q Kruglyak L Nickerson DA Selecting a maximally informative set of single nucleotide polymorphisms for association analyses using linkage disequilibrium Am J Hum Genet 2004 74 106 120 14681826 10.1086/381000
Halldorsson BV Bafna V Lippert R Schwartz R De La Vega FM Clark AG Istrail S Optimal haplotype block-free selection of tagging SNPs for genome-wide association studies Genome Res 2004 14 1633 1640 15289481 10.1101/gr.2570004
Meng Z Zaykin DV Xu C Wagner M Ehm MG Selection of genetic markers for association analyses using linkage disequilibrium and haplotypes Am J Hum Genet 2003 73 115 130 12796855 10.1086/376561
Chapman JM Cooper JD Todd JA Clayton DG Detecting disease associations due to linkage disequilibrium using haplotype tags: a class of tests and the determinants of statistical power Hum Hered 2003 56 18 31 14614235 10.1159/000073729
Cousin E Genin E Macce S Ricard S Chansac C Del Zompo M Deleuze JF Association studies candidate genes: strategies to select SNPs to be tested Hum Hered 2003 56 151 159 15031617 10.1159/000073200
Hu X Schrodi SJ Ross DA Cargill M Selecting tagging SNPs for association studies using power calculations from genotype data Hum Hered 2004 57 156 170 15297809 10.1159/000079246
Zhang K Qin ZS Liu JA Chen T Waterman MS Sun F Haplotype block partitioning and tag SNP selection using genotype data and their applications to association studies Genome Res 2004 14 908 916 15078859 10.1101/gr.1837404
Zhang K Calabrese P Nordborg M Sun F Haplotype structure and its applications to association studies: power and study design Am J Hum Genet 2002 71 1386 1394 12439824 10.1086/344780
Johnson GCL Esposito L Barratt BJ Smith AN Heward J Genova GD Ueda H Cordell HJ Eaves IA Dudbridge F Twells RCJ Payne F Hughes W Nutland S Stevens H Phillipa C Tuomilehto-Wolf E Tuomilehto J Gough SCL Clayton DG Todd JA Haplotype tagging for the identification of common disease genes Nat Genet 2001 29 233 237 11586306 10.1038/ng1001-233
Thompson D Stram D Goldgar D Witte JS Haplotype tagging single nucleotide polymorphisms and association studies Hum Hered 2003 56 48 55 14614238 10.1159/000073732
Zhai W Todd MJ Nielsen R Is haplotype block identification useful for association mapping studies? Genet Epidemiol 2004 27 80 83 15185406 10.1002/gepi.20014
Kaplan N Morris R Issues concerning association studies for fine mapping a susceptibility gene for a complex disease Genet Epidemiol 2001 20 432 457 11319784 10.1002/gepi.1012
Zhang W Collins A Morton N Does haplotype diversity predict power for association mapping of disease susceptibility? Hum Genet 2004 115 157 164 15221450 10.1007/s00439-004-1122-x
Weale ME Depondt C Macdonald SJ Smith A Lai PS Shorvon SD Wood NW Goldstein DB Selection and evaluation of tagging SNPs in the neuronal-sodium-channel gene SCN1A: Implications for linkage-disequilibrium gene mapping Am J Hum Genet 2003 73 551 565 12900796 10.1086/378098
Hudson RR Properties of a neutral-allele model with intergenic recombination Theor Popul Biol 1983 23 183 201 6612631 10.1016/0040-5809(83)90013-8
Kaplan NL Hudson RR The use of sample genealogies for studying a selectively neutral m-loci model with recombination Theor Popul Biol 1985 28 382 396 4071443 10.1016/0040-5809(85)90036-X
Griffiths RC Marjoram P Donnelly P, Tavaré S An ancestral recombination graph Progress in Population Genetics and Human Evolution 1997 Springer-Verlag, New York
Jeffreys AJ Kauppi L Neumann R Intensely punctuate meiotic recombination in the class II region of the major histocompatibility complex Nat Genet 2001 29 217 222 11586303 10.1038/ng1001-217
Schneider JA Peto TE Boone RA Boyce AJ Clegg JB Direct measurement of the male recombination fraction in the human β-globin hot spot Hum Mol Genet 2002 11 207 215 11823440 10.1093/hmg/11.3.207
Phillips MS Lawrence R Sachidanandam R Morris AP Balding DJ Donaldson MA Studebaker JF Ankener WM Alfisi SV Kuo FS Camisa AL Pazorov V Scott KE Carey BJ Faith J Katari G Bhatti HA Cyr JM Derohannessian V Elosua C Forman AM Grecco NM Hock CR Kuebler JM Lathrop JA Mockler MA Nachtman EP Restine SL Varde SA Hozza MJ Gelfand CA Broxholme J Abecasis GR Boyce-Jacino MT Cardon LR Chromosome-wide distribution of haplotype blocks and the role of recombination hot spots Nat Genet 2003 33 382 387 12590262 10.1038/ng1100
Wang N Akey JM Zhang K Chakraborty K Jin L Distribution of recombination crossovers and the origin of haplotype blocks: the interplay of population history, recombination, and mutation Am J Hum Genet 2002 71 1227 1334 12384857 10.1086/344398
Nordborg M Tavaré S Linkage disequilibrium: what history has to tell us? Trends Genet 2002 18 83 90 11818140 10.1016/S0168-9525(02)02557-X
Fisher RA The correlation between relatives on the supposition of Mendelian inheritance Trans Royal Soc Edinburgh 1918 52 399 433
Schork NJ Nath SK Fallin D Chakravarti A Linkage disequilibrium analysis of biallelic DNA markers, human quantitative trait loci, and threshold-defined case and control subjects Am J Hum Genet 2000 67 1208 1218 11032785
Pritchard JK Przeworski M Linkage disequilibrium in humans: models and data Am J Hum Genet 2001 69 1 14 11410837 10.1086/321275
Abecasis GR Cardon LR Cookson WOC A general test of association for quantitative traits in nuclear families Am J Hum Genet 2000 66 279 292 10631157 10.1086/302698
Monks SA Kaplan NL Removing the sampling restrictions from family-based tests of association for a quantitative-trait locus Am J Hum Genet 2000 66 576 592 10677318 10.1086/302745
Rabinowitz D A transmission disequilibrium test for quantitative trait loci Hum Hered 1997 47 342 350 9391826
Spielman RS Ewens WJ The TDT and other family based tests for linkage disequilibrium and association Am J Hum Genet 1996 59 983 9 8900224
Sun FZ Flanders WD Yang QH Zhao HY Transmission/disequilibrium tests for quantitative traits Ann Hum Genet 2000 64 555 565 11281218 10.1017/S000348000000840X
Slatkin M Disequilibrium mapping of a quantitative-trait locus in an expanding population Am J Hum Genet 1999 64 1764 72 10330364 10.1086/302413
Dudbridge F Pedigree disequilibrium tests for multilocus haplotypes Genet Epidemiol 2003 25 115 121 12916020 10.1002/gepi.10252
Zaykin DV Westfall PH Young SS Karnoub MA Wagner MJ Ehm MG Testing association of statistically inferred haplotypes with discrete and continuous traits in samples of unrelated individuals Hum Hered 2002 53 79 91 12037407 10.1159/000057986
Ke XY Durrant C Morris AP Hunt S Bentley DR Deloukas P Cardon LR Efficiency and consistency of haplotype tagging of dense SNP maps in multiple samples Hum Mol Genet 2004 13 2557 2565 15367493 10.1093/hmg/ddh294
Liu N Sawyer SL Mukherjee N Pakstis AJ Kidd JR Kidd KK Brooks AJ Zhao H Haplotype block structures show significant variation among populations Genet Epidemiol 2004 27 385 400 15389924 10.1002/gepi.20026
Pritchard JK Rosenberg NA Use of unlinked genetic markers to detect population stratification in association studies Am J Hum Genet 1999 65 220 228 10364535 10.1086/302449
Schulze TG Zhang K Chen YS Akula N Sun FZ Mcmahon FJ Defining haplotype block and tag single-nucleotide polymorphisms in the human genome Hum Mol Genet 2004 13 335 342 14681300 10.1093/hmg/ddh035
|
16221303
|
PMC1274313
|
CC BY
|
2021-01-04 16:28:16
|
no
|
BMC Infect Dis. 2005 Oct 12; 5:82
|
latin-1
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-82
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-861622570110.1186/1471-2334-5-86Research ArticleExtended Spectrum β-Lactamases among Gram-negative bacteria of nosocomial origin from an Intensive Care Unit of a tertiary health facility in Tanzania Ndugulile Faustine [email protected] Roland [email protected] Stig [email protected] Willy [email protected] Nina [email protected] Institute of Internal Medicine, University of Bergen, N-5021, Bergen, Norway2 Centre for International Health, University of Bergen, N-5021 Bergen, Norway3 Department of Internal Medicine, Haukeland University Hospital, N-5021, Bergen, Norway4 Department of Infection Control, Haukeland University Hospital, N-5021 Bergen, Norway5 Department of Microbiology and Immunology, Muhimbili University College of Health Sciences, Dar es Salaam, Tanzania6 Department of Laboratory Medicine, Alexandra Hospital, Singapore2005 15 10 2005 5 86 86 27 5 2005 15 10 2005 Copyright © 2005 Ndugulile et al; licensee BioMed Central Ltd.2005Ndugulile et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Resistance to third generation cephalosporins due to acquisition and expression of extended spectrum β-lactamase (ESBL) enzymes among Gram-negative bacteria is on the increase. Presence of ESBL producing organisms has been reported to significantly affect the course and outcome of an infection. Therefore infections due to ESBL isolates continue to pose a challenge to infection management worldwide. The aim of this study was to determine the existence and to describe phenotypic and genotypic characteristics of ESBLs in an Intensive Care Unit (ICU) setting in Tanzania.
Methods
Between October 2002 and April 2003, clinical information and samples were collected from patients suspected to have nosocomial infections in an Intensive Care Unit of a tertiary hospital in Tanzania. The isolates were identified, tested for antimicrobial susceptibility and analysed for presence of ESBL genes.
Results
Thirty-nine Gram-negative bacteria were isolated from clinical samples of 39 patients. These isolates included 13 Escherichia coli, 12 Enterobacter spp, 5 Pseudomonas spp, 4 Proteus spp, 2 Klebsiella. pneumoniae, 2 Citrobacter freundii and 1 Chryseomonas luteola. Eleven (28.2%) of these isolates were ESBL producing. The ESBL genes characterised were SHV-12, SHV-28 and CTX-M-15. The ESBL producing isolates were more resistant to gentamicin and ciprofloxacin than non-ESBL producing isolates.
Conclusion
This study shows the presence of ESBL genes among Gram-negative bacteria in the ICU setting in Tanzania. There is a need to institute strict hospital infection control policy and a regular surveillance of resistance to antimicrobial agents.
==== Body
Background
Extended spectrum β-lactamase (ESBL) enzymes have been reported in a number of species in Gram-negative bacteria. The ESBL enzymes are usually plasmid mediated and are capable of hydrolysing and inactivating a wide variety of β-lactams, including third-generation cephalosporins, penicillins and aztreonam, but are susceptible to β-lactamase inhibitors such as clavulanate, sulbactam and tazobactam [1,2]. Many ESBL producers also carry other genes that confer resistance to other antimicrobial agents such as aminoglycosides and fluoroquinolones [3,4].
Extensive use of broad spectrum antibiotics, prolonged hospitalisation, indwelling devices and severe underlying diseases have been reported as factors that have led to spread of ESBL and difficulties in managing severe infections in many parts of the world [5,6].
Treatment of ESBL-producing bacterial infections can place an added constraint on already overburden health systems in developing countries. ESBL producing organisms are reported to account for a significant proportion of intensive care infections [7]. Problems of ESBLs have led to limited, as well as expensive treatment options, and have impacted negatively on clinical outcomes [8]. Nosocomial infections due to ESBL producing organisms have been known to cause high mortality [9].
Very few studies have reported on the problem of ESBLs in Africa in general and Tanzania in particular. There have been reports of CTX -M-12 ESBL in Klebsiella pneumoniae in Kenya [10], TEM-53, TEM-63, SHV-2, SHV-5, SHV-19, SHV-20, SHV-21 and SHV-22 in K. pneumoniae in South Africa [11], CTX-M 15 and SHV-12 in Tanzania [12], SHV-12 in Salmonella enterica serotype Babelsberg and Enteretidis from Mali orphans [13] and TEM-3 in S. typhimurium in Morocco [14].
The aim of this study was to determine the existence and to describe phenotypic and genotypic characteristics of ESBLs in an Intensive Care Unit (ICU) setting in Tanzania.
Methods
Study design
Muhimbili National Hospital is a 1000 bed tertiary health facility with an eight-bed ICU that caters for surgical, medical and trauma emergencies. During the study period October 2002 to April 2003, 206 patients were admitted to the ICU and of these 50 specimens were collected from patients with suspected nosocomial infections, according to definitions as described by the Centers for Disease Control (CDC) [15]. In particular, infections were considered to be nosocomial if symptoms and signs appeared >48 hours following admission to the ICU.
Bacterial strains and susceptibility testing
Specimen from 50 patients with suspected nosocomial infections were collected and cultured on blood agar (Oxoid Ltd, Basingstoke, UK) and MacConkey agar (Difco/BD Diagnostics Systems, Sparks, MI, USA) except for urine samples which were plated on Cysteine Lactose Electrolytes Deficient (CLED) agar. Isolated strains were stored in tubes containing 1.5 ml Brain Heart Infusion broth with 10% v/v glycerol at -70°C until the time of analysis. The isolates were identified using biochemical tests and confirmed using the API 20E and API 20 NE identification systems (bioMerieux, Marcy-l'Etoile, France).
The susceptibility of the Enterobacteriaceae to antimicrobial agents was examined by an agar diffusion method using paper disks (AB Biodisk, Solna, Sweden) containing the following antibiotic concentration [14]: amoxicillin/clavulanic acid (20/10 μg), ampicillin (10 μg), ceftriaxone (30 μg), cefuroxime (30 μg), chloramphenicol (30 μg), ciprofloxacin (5 μg), doxycycline (30 μg), gentamicin (10 μg) and trimethoprim/sulfamethoxazole (1.25/23.75 μg). The susceptibility of the non-Enterobacteriaceae was examined using ceftriaxone, ciprofloxacin, ceftazidime (30 μg), gentamicin, imipenem (10 μg) and tobramycin (10 μg). E. coli ATCC 25922 and P. aeruginosa ATCC 27853 were used as quality control strains.
Testing for presence of ESBL
Gram negative bacteria with reduced susceptibility to third generation cephalosporins according to the NCCLS criteria [16], that is, zones of inhibition of = 27 mm for cefotaxime and = 22 mm for ceftazidime, were tested by three ESBL Etest strips (AB Biodisk, Solna, Sweden) on Mueller Hinton agar. Minimum Inhibitory Concentrations of cefotaxime (CT), ceftazidime (TZ) and cefepime (PM) with and without clavulanic acid were used. The inoculated plates were incubated for 16–18 hours at 37°C.
ESBL results were read either as MIC values or observation of 'phantom' zones or deformation of inhibition ellipses. Reduction of MIC by = 3 two-fold dilutions in the presence of clavulanic acid was indicative of ESBL production. Deformation of ellipses or the presence of a 'phantom' zone was also indicative of ESBL production even if the MIC ratio is <8 or cannot be read. Isolates were reported as having ESBL phenotype if one or more of the three ESBL Etests were positive.
The outcome of the test was indeterminate when MICs were outside the test range of the test device. Those strains found to be ESBL producing phenotypically by Etest were examined for the presence of the TEM, SHV and CTX-M genes by Polymerase Chain Reaction (PCR).
Molecular analysis techniques
Isolates with ESBL phenotype were examined for the presence of blaTEM, blaSHV and blaCTX-M by PCR [17,18]. The PCR products were purified using QIAquick PCR Purification Kits (Qiagen, Hilden, Germany). The templates were sequenced using the same primers as used in the PCR amplification. The cycle sequencing parameters were 25 cycles at 96°C for 10 seconds, 58°C for 5 seconds (50°C for blaCTX-M) and 60°C for 4 minutes. The products were analysed using an ABI PRISM 3700 DNA sequencer (PE Biosystems, Warington, UK). Point mutations were accepted if present in both the forward and reverse sequences.
Analysis of chromosomal DNA by PFGE
Pulsed-field gel electrophoresis (PFGE) was performed as described previously by Gautom et al [19]. DNA of all isolates was digested using XbaI (New England BioLabs, Beverly, Mass., USA) at 37°C for 4 hours, according to supplier's instructions. The slices of the digested DNA were loaded into the wells of the poured gels. Electrophoresis agarose gel (Promega, Madison, USA) with a concentration of 1% was used for different organisms in 0.5 × Tris-Borate -EDTA buffer using contour -clamped homogenous electric field apparatus (CHEF-DR III; Bio-Rad, Richmond, Calif., USA). A 48.5 kb lambda PFG marker 50 ug/ml (New England BioLabs) was used as a marker. The conditions for electrophoresis were angle, 120°; gradient, 6 V/cm; temperature, 14°C; pulse times ranging from 10 to 45 s and running time was 20 hours. The fragment patterns were interpreted as described by Tenover et al [20].
Statistical analysis
All statistics were performed by SPSS software 11.5. Contingency table analysis was done by χ2 test or two-tailed Fisher's exact test for categorical variables. P-value of <0.05 was considered statistically significant.
Results
Bacteriology and antimicrobial susceptibility pattern
A total of 50 bacteria were isolated from 50 patients suspected of having nosocomial infections. These consisted of 30 Urinary Tract Infections, 15 wound infections, 3 blood stream infections and 2 cases of pneumonia. Out of these 50 bacterial isolates, 39 were Gram-negative bacteria, The remaining eleven isolates were S. aureus. These Gram-negative isolates were identified as Escherichia coli (N = 13), Enterobacter cloacae (N = 8), Enterobacter aerogenes (N = 3), Pantoea agglomerans (N = 1), Pseudomonas aeruginosa (N = 5), Proteus mirabilis (N = 3), Proteus vulgaris (N = 1), Klebsiella pneumoniae (N = 2), Citrobacter freundii (N = 2) and Chryseomonas luteola (N = 1).
Among the ESBL producing strains a significant proportion were found to be resistant to antimicrobial agents including amoxicillin/ clavulanic acid (90.9%), doxycycline (81.5%), gentamicin (72.7%) and trimethoprim/ sulfamethoxazole (90.9%), ceftriaxone (100%) and cefuroxime (100%). The lowest levels of resistance were seen for ciprofloxacin and chloramphenicol with 45.5% each (Table 1). One E. cloacae and one E. coli were resistant to all antimicrobial agents tested. ESBL producing isolates were resistant to more antimicrobial agents than non-ESBL producing isolates. The highest rates of resistance in ESBL negative isolates were seen against ampicillin (86.4%), doxycycline (77.3%) and trimethoprim/sulfamethoxazole (63.6%). Pseudomonas spp were fully susceptible to imipenem, ceftazidime and tobramycin. The difference in resistance levels between ESBL and non-ESBL producing isolates for ciprofloxacin (p = 0.017), ceftriaxone (p = 0.000) and gentamicin (p = 0.003) were statistically significant (Table 1).
Table 1 Antimicrobial susceptibility pattern of ESBL and non-ESBL producing isolates from the ICU, Dar es Salaam, Tanzania§.
Antimicrobial ESBL (n = 11) Non-ESBL (n = 22) p-value
% resistant % resistant
Amoxicillin/Clavulanic 90.9 72.8 0.451
Ampicillin 100 86.4 0.521
Ceftriaxone 100 13.6 0.000
Cefuroxime 100 72.7 0.151
Chloramphenicol 45.5 27.3 0.514
Ciprofloxacin 45.5 4.5 0.017
Doxycycline 81.8 77.3 0.880
Gentamicin 72.7 13.6 0.003
Trimethoprim/Sulfa. 90.9 63.6 0.214
Pseudomonas spp and C. luteola were excluded in this analysis§.
PCR and sequencing of ESBL genes
All Gram-negative isolates except for Pseudomonas spp were screened for ESBL. The eleven isolates that were phenotypically positive for ESBL were tested by PCR. Six, seven and five isolates were positive for blaTEM, blaSHV and blaCTX-Mrespectively (Table 2). One isolate each of E. coli, K. pneumoniae and E. cloacae had both blaTEM and blaSHV>. One K. pneumoniae isolate had only blaSHV, three isolates of E. coli had both blaTEM and blaCTX-M and two isolates of E. cloacae had both blaSHV and blaCTX-M. One E. cloacae isolate was phenotypically ESBL positive on Etest but was negative for blaTEM, blaSHV and blaCTX-M on PCR.
Table 2 Characteristics of ESBL producing isolates from ICU, Dar es Salaam, Tanzania.
Organism ESBL enzyme Date of Isolation Specimen
E. cloacae SHV-12, CTX-M-15 13 December 2002 pus
" SHV-12, CTX-M-15 27 October 2002 urine
" SHV-12 30 December 2002 urine
" -* 16 February 2002 urine
E. aerogenes SHV-12 18 October 2002 urine
E. coli SHV 12 22 October 2002 urine
" CTX-M-15 28 November 2002 urine
" CTX-M-15 05 December 2002 urine
" CTX-M-15 30 November 2002 pus
K. pneumoniae SHV-12 30 December 2002 pus
" SHV-28 20 February 2003 urine
*Isolate no.5 (E. cloacae) was phenotypically positive on E-test but negative on PCR.
Nucleotide sequence analysis of the isolates carrying blaCTX-M revealed that all isolates carried the CTX-M-15 gene. Of the seven isolates that had bla SHV, six had SHV-12 and one had SHV-28 (Table 2). TEM-1 was identified in TEM producing isolates.
Genotypic relationship of the ESBL isolates
Two of the 13 E. coli isolated had indistinguishable restriction patterns on PFGE. These were isolates from two different patients and from different specimen (urine and pus). The two patients had at one time been together in the ICU. Other isolates tested were unrelated.
Discussion
This study demonstrates the presence of ESBL-mediated resistance in bacteria causing infections in the ICU of a tertiary hospital in Tanzania. Despite the rise in the prevalence of ESBL in some countries, there are very few reports from Africa [10-12,14]. ESBL detection is not commonly carried out in many microbiology units in developing countries, Tanzania included. This could be attributed to lack of awareness and lack of resources and facilities to conduct ESBL identification. The high rate of resistance noted among the isolates in the present study, although few in numbers, is of serious concern. Eleven of the 39 (28.2%) of the Enterobacteriaceae were ESBL producing.
In this study, ESBL producing isolates were significantly more resistant to ciprofloxacin (p = 0.017), ceftriaxone (p = 0.000) and gentamicin (p = 0.003) as compared to non-ESBL producing Gram-negative isolates. Other studies have reported on cross-resistance to aminoglycosides, fluoroquinolones and trimethoprim/sulfamethoxazole in ESBL producing organisms [21,22]. Mechanisms of co-resistance are not clear, but one possible mechanism is the co-transmission of ESBL and resistance to other antimicrobials within the same conjugative plasmids [23].
This study reports on the presence of CTX-M-15 producing isolates in the ICU setting in Tanzania. CTX-M-15 gene was also found among bloodstream infection isolates from children within the same hospital [12]. This gene is widespread in European and Asian countries including Northern Italy [18], United Kingdom [24] and India [25]. The other reported gene is CTX-M-12 that was identified amongst K. pneumoniae isolates from cerebrospinal fluid and blood in Kenya [10]. In the present study, the CTX-M-15 gene was found in two different genera of the bacterial isolates in the ICU (E. coli and E. cloacae). Two of the CTX-M-15 producing isolates were also found to possess the SHV-12 gene. The two E. coli with indistinguishable PGFE restriction patterns were found to be resistant to all antimicrobials tested except for chloramphenicol.
SHV-12 was found in the two strains of E. coli with indistinguishable PFGE restriction patterns. SHV-12 was also found among bloodstream infection isolates from children within the same hospital [12]. This gene has also been reported in several European and Asian countries including Switzerland [26] and Thailand [27].
In the present study, this gene was found across different genera of bacterial isolates. The other SHV related genes that have been reported in Africa include SHV-2, SHV-5, SHV-19, SHV-20, SHV-21 and SHV-22 that were reported in South Africa [11].
In this study, one K. pneumoniae isolate was found to possess SHV-28. Initial reports of this gene came from China (GenBank accession no. AF299299). To the best of our knowledge, there are no other reports of this gene in the literature. This isolate was obtained from a urine specimen from a patient with urinary tract infection. This isolate was resistant to amoxicillin/clavulanic acid, doxycycline and trimethoprim/sulfamethoxazole and cephalosporins but sensitive to chloramphenicol, ciprofloxacin and gentamicin.
Presence of similar SHV, TEM and CTX-M genes in various non-genetically related strains suggests a possibility of horizontal transfer of these genes in the ICU. On the other hand, the two E. coli that produced indistinguishable restriction patterns on PFGE suggests that clonal spread of the resistant bacteria is also a potential factor in the spread of resistance in the ICU. These finding of similar ESBL genes in the ICU and in paediatric wards [12] within the same hospital could mean the ESBL genes are widespread in this hospital.
Conclusion
This study shows that there is a presence of ESBL producing isolates in the ICU of a major hospital in Tanzania. This is also the first report of presence of SHV-28 in Africa. This study also shows that the ESBL phenotype spread in this hospital is due to multiple enzymes found in different genera of bacteria. There is a need to institute a strict hospital infection control policy and a regular surveillance of resistance to antimicrobial agents.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FN was the principal investigator, participated in the planning and execution of the study, performed data entry and data analysis, laboratory work and was the main responsible author. RJ participated in the planning of the study, contributed to the writing process and provided advice with the laboratory work. SH and WU participated in the planning of the study and contributed to the writing process. NL was the project coordinator and participated in planning, data analysis and writing.
All authors have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We would like to thank the staff at the ICU in Dar es Salaam for their assistance and to Prof Victor Mwafongo for his advice and assistance in Dar es Salaam, Tanzania. We would also like to thank Dr Bjorn Blomberg for reviewing the manuscript and for the provision of laboratory assistance in Bergen, Johanna Sollid and Mekonnen Kurabachew for their assistance in interpreting sequence results and to Marit Gjerde Tellevik for her assistance with the laboratory methods and supplies. This study was supported by the grant from the Norwegian State Education Loan Fund.
==== Refs
Sturenburg E Mack D Extended-spectrum beta-lactamases: implications for the clinical microbiology laboratory, therapy, and infection control J Infect 2003 47 273 295 14556752 10.1016/S0163-4453(03)00096-3
Gniadkowski M Evolution and epidemiology of extended-spectrum beta-lactamases (ESBLs) and ESBL-producing microorganisms Clin Microbiol Infect 2001 7 597 608 11737084 10.1046/j.1198-743x.2001.00330.x
Spanu T Luzzaro F Perilli M Amicosante G Toniolo A Fadda G Occurrence of extended-spectrum beta-lactamases in members of the family Enterobacteriaceae in Italy: implications for resistance to beta-lactams and other antimicrobial drugs Antimicrob Agents Chemother 2002 46 196 202 11751134 10.1128/AAC.46.1.196-202.2002
Sekowska A Janicka G Klyszejko C Wojda M Wroblewski M Szymankiewicz M Resistance of Klebsiella pneumoniae strains producing and not producing ESBL (extended-spectrum beta-lactamase) type enzymes to selected non-beta-lactam antibiotics Med Sci Monit 2002 8 BR100 4 11889457
Nathisuwan S Burgess DS Lewis JS Extended-spectrum beta-lactamases: epidemiology, detection, and treatment Pharmacotherapy 2001 21 920 928 11718498 10.1592/phco.21.11.920.34529
Lin MF Huang ML Lai SH Risk factors in the acquisition of extended-spectrum beta-lactamase Klebsiella pneumoniae: a case-control study in a district teaching hospital in Taiwan J Hosp Infect 2003 53 39 45 12495684 10.1053/jhin.2002.1331
Paterson DL Ko WC Von Gottberg A Mohapatra S Casellas JM Goossens H Mulazimoglu L Trenholme G Klugman KP Bonomo RA Rice LB Wagener MM McCormack JG Yu VL International prospective study of Klebsiella pneumoniae bacteremia: implications of extended-spectrum beta-lactamase production in nosocomial Infections Ann Intern Med 2004 140 26 32 14706969
Menashe G BA Yagupsky P Peled N Gilad J Fraser D Riesenberg K Schlaeffer F Clinical significance and impact on mortality of extended-spectrum beta lactamase-producing Enterobacteriaceae isolates in nosocomial bacteremia. Scand J Infect Dis 2001 33 188 193 11303808 10.1080/00365540151060806
Ho PL Chan WM Tsang KW Wong SS Young K Bacteremia caused by Escherichia coli producing extended-spectrum beta-lactamase: a case-control study of risk factors and outcomes Scand J Infect Dis 2002 34 567 573 12238570 10.1080/00365540210147516
Kariuki S Corkill JE Revathi G Musoke R Hart CA Molecular characterization of a novel plasmid-encoded cefotaximase (CTX-M-12) found in clinical Klebsiella pneumoniae isolates from Kenya Antimicrob Agents Chemother 2001 45 2141 2143 11408239 10.1128/AAC.45.7.2141-2143.2001
Essack SY Hall LM Pillay DG McFadyen ML Livermore DM Complexity and diversity of Klebsiella pneumoniae strains with extended-spectrum beta-lactamases isolated in 1994 and 1996 at a teaching hospital in Durban, South Africa Antimicrob Agents Chemother 2001 45 88 95 11120950 10.1128/AAC.45.1.88-95.2001
Blomberg B Jureen R Manji KP Tamim BS Mwakagile DS Urassa WK Fataki M Msangi V Tellevik MG Maselle SY Langeland N High rate of fatal cases of pediatric septicemia caused by gram-negative bacteria with extended-spectrum beta-lactamases in Dar es Salaam, Tanzania J Clin Microbiol 2005 43 745 749 15695674 10.1128/JCM.43.2.745-749.2005
Weill FX Demartin M Tande D Espie E Rakotoarivony I Grimont PA SHV-12-like extended-spectrum-beta-lactamase-producing strains of Salmonella enterica serotypes Babelsberg and Enteritidis isolated in France among infants adopted from Mali J Clin Microbiol 2004 42 2432 2437 15184415 10.1128/JCM.42.6.2432-2437.2004
AitMhand R Soukri A Moustaoui N Amarouch H ElMdaghri N Sirot D Benbachir M Plasmid-mediated TEM-3 extended-spectrum beta-lactamase production in Salmonella typhimurium in Casablanca J Antimicrob Chemother 2002 49 169 172 11751783 10.1093/jac/49.1.169
Garner JS JW Emori TG Horan TC Hughes JM RN O CDC definitions for nosocomial infections Infection Control and Applied Epidemiology:Principles and Practice 1996 St. Louis, Mosby A 1-A-20
National Committee for Clinical Laboratory Standards Performance standards for antimicrobial susceptibility testing; ninth informational supplement. M100-S9. 1999 Wayne, PA,
Rasheed JK Jay C Metchock B Berkowitz F Weigel L Crellin J Steward C Hill B Medeiros AA Tenover FC Evolution of extended-spectrum beta-lactam resistance (SHV-8) in a strain of Escherichia coli during multiple episodes of bacteremia Antimicrob Agents Chemother 1997 41 647 653 9056008
Pagani L Dell'Amico E Migliavacca R D'Andrea MM Giacobone E Amicosante G Romero E Rossolini GM Multiple CTX-M-type extended-spectrum beta-lactamases in nosocomial isolates of Enterobacteriaceae from a hospital in northern Italy J Clin Microbiol 2003 41 4264 4269 12958255 10.1128/JCM.41.9.4264-4269.2003
Gautom RK Rapid pulsed-field gel electrophoresis protocol for typing of Escherichia coli O157:H7 and other gram-negative organisms in 1 day. J Clin Microbiol 1997 35 2977 2980 9350772
Tenover FC Arbeit RD Goering RV Mickelsen PA Murray BE Persing DH Swaminathan B Interpreting chromosomal DNA restriction patterns produced by pulsed-field gel electrophoresis: criteria for bacterial strain typing J Clin Microbiol 1995 33 2233 2239 7494007
Alhambra A Cuadros JA Cacho J Gomez-Garces JL Alos JI In vitro susceptibility of recent antibiotic-resistant urinary pathogens to ertapenem and 12 other antibiotics J Antimicrob Chemother 2004 53 1090 1094 15117925 10.1093/jac/dkh218
Procop GW Tuohy MJ Wilson DA Williams D Hadziyannis E Hall GS Cross-class resistance to non-beta-lactam antimicrobials in extended-spectrum beta-lactamase-producing Klebsiella pneumoniae Am J Clin Pathol 2003 120 265 267 12931557 10.1309/BWQK-WB2W-N6W5-X1CC
Martinez-Martinez L Pascual A Jacoby GA Quinolone resistance from a transferable plasmid Lancet 1998 351 797 799 9519952 10.1016/S0140-6736(97)07322-4
Mushtaq S Woodford N Potz N Livermore DM Detection of CTX-M-15 extended-spectrum beta-lactamase in the United Kingdom J Antimicrob Chemother 2003 52 528 529 12888591 10.1093/jac/dkg353
Karim A Poirel L Nagarajan S Nordmann P Plasmid-mediated extended-spectrum beta-lactamase (CTX-M-3 like) from India and gene association with insertion sequence ISEcp1 FEMS Microbiol Lett 2001 201 237 241 11470367 10.1016/S0378-1097(01)00276-2
Nuesch-Inderbinen MT Kayser FH Hachler H Survey and molecular genetics of SHV beta-lactamases in Enterobacteriaceae in Switzerland: two novel enzymes, SHV-11 and SHV-12 Antimicrob Agents Chemother 1997 41 943 949 9145849
Chanawong A M'Zali FH Heritage J Lulitanond A Hawkey PM SHV-12, SHV-5, SHV-2a and VEB-1 extended-spectrum beta-lactamases in Gram-negative bacteria isolated in a university hospital in Thailand J Antimicrob Chemother 2001 48 839 852 11733468 10.1093/jac/48.6.839
|
16225701
|
PMC1274314
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Oct 15; 5:86
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-86
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-901624201610.1186/1471-2334-5-90Case ReportCardiac involvement in a patient with clinical and serological evidence of African tick-bite fever Bellini Cristina [email protected] Matteo [email protected] Mathieu [email protected] Anne Dalle [email protected] Jacques [email protected] Gilbert [email protected] Infectious Diseases Unit, University Hospital of Lausanne, Lausanne, Switzerland2 Internal Medicine Unit, University Hospital of Lausanne, Lausanne, Switzerland3 Center for Research on Intracellular Bacteria, Institute of Microbiology, University of Lausanne, Lausanne, Switzerland2005 20 10 2005 5 90 90 29 4 2005 20 10 2005 Copyright © 2005 Bellini et al; licensee BioMed Central Ltd.2005Bellini et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Myocarditis and pericarditis are rare complications of rickettsiosis, usually associated with Rickettsia rickettsii and R. conorii. African tick-bite fever (ATBF) is generally considered as a benign disease and no cases of myocardial involvement due to Rickettsia africae, the agent of ATBF, have yet been described.
Case presentation
The patient, that travelled in an endemic area, presented typical inoculation eschars, and a seroconversion against R. africae, was admitted for chest pains and increased cardiac enzymes in the context of an acute myocarditis.
Conclusion
Our findings suggest that ATBF, that usually presents a benign course, may be complicated by an acute myocarditis.
==== Body
Background
Myocarditis and pericarditis are rare complications of rickettsiosis, usually occuring in the setting of an acute disseminated infection due to Rickettsia rickettsii and R. conorii [1-3].
Rickettsia africae, the causative agent of African tick-bite fever, an emerging disease transmitted by Amblyomma ticks in rural sub-Saharan Africa, has been recently described [4]. Symptoms usually includes abrupt appearance of fever (59–100% of cases), headache (62–83%), myalgia (63–87%), prominent neck muscle myalgia (81%), regional lymphadenitis (43–100%), cutaneous rash (15–46%) and inoculation eschar (53–100%), typically present in multiple sites (21–54%) [1]. The time lag from tick bite to symptom onset is usually 5 to 7 days but may be as long as 10 days [2-4].
Several case reports of ATBF in travellers from Europe and elsewhere have been published [1-5] and recently, ATBF have also been reported in autochthonous Africans [6]. However, no cases of myocardial involvement have yet been described. Here, we report the first evidence that ATBF may be complicated by an acute myocarditis.
Case presentation
Patient A and his wife, two healthy 35 years old adults, did a 4-weeks camping's holidays in South Africa in July 2004. They travelled along the southern coast (Figure 1), and were daily exposed to insect and tick bites. Four days after their arrival in Swaziland and 19 days after the beginning of the trip, patient A suddenly presented fever and a vesicular erythematous rash located on both legs. His condition improved with acetylsalicylic acid. However, he developed in the following week several inguinal lymphadenitis and severe asthenia. His wife presented – with a two days delay – an acute febrile illness, associated with a macular rash on both legs, myalgias, arthralgias and headache. Her fever resolved within 24 hours. However, skin lesions were still present four days later, when she observed a new inguinal lymphadenopathy. Seven days after the symptoms onset, they returned to Switzerland. Two days later, patient A presented an oppressive chest pain and profuse perspiration, which spontaneously resolved in 3 hours. Because of this acute episode, he went to the emergency unit of Lausanne's University hospital.
Figure 1 Travel of patient A and his wife in South Africa and Swaziland and description of key events, signs and symptoms.
At admission, all symptoms including chest pain, were resolved. He was afebrile. Multiple reddish vesicles, partially crusted, measuring 1 to 2 mm, were present on the legs and on the abdomen (Figure 2A). We also observed a 10 mm infiltrated abdominal lesion with a small central necrosis similar to typical tick-bite inoculation eschars (Figure 2B). There were no signs of heart failure and cardiac auscultation was physiological, without friction rub. To investigate the thoracic pain, an electrocardiogram was performed, which was normal, without signs of ischemic heart disease. Cardiac enzymes were slightly increased, with CK, CK-MB, and troponin I level in blood of 330 UI/l (Normal Values [NV] 25–190 UI/l), 40% (NV <6%), and 2,12 mg/ml (NV < 0.04 mg/ml), respectively. The other blood analyses revealed a moderate leucopenia (3.7 G/l, NV 4–10 G/l), and a slight increase of hepatic enzymes. A normal thoraco-abdominal Computer Tomography excluded aortic dissection. About 6 hours after the first electrocardiogram, repolarisation abnormalities were observed in the inferior leads (Figure 3). This was associated with an increase of cardiac enzymes (peak of CK and Troponin I of 400 UI/l and 3.62 mg/ml respectively). A coronarography showed normal vessels. Both cardiac echocardiographies performed respectively at admission and two weeks later, showed no pericardial effusion and a normal ejection fraction of about 65%. A rickettsial disease was suspected based on the presence of a febrile illness with a rash and inoculation eschars. The history of multiples tick bites and the simultaneous occurrence of the disease in both patients were suggestive of ATBF. This diagnosis was confirmed by a R. africae seroconversion of patient A (titres of 1/64 in IgG and 1/16 in IgM on the convalescent serum taken one month after the symptom onset) and by a sustained positive R. africae serology for his wife (IgG titers of 1:64 and IgM titres of 1:32 on both acute and convalescent sera; IgG titers >= 1:64 and IgM titers >= 1:32 are considered indicative of infection by R. africae). As frequently observed, similar antibody titres were obtained for other spotted fever group rickettsia, such as R.conorii and R. massiliae. Western-blot and cross-adsorption that may help to precise the etiological agent [11] was not feasible due to the low antibody titres. Both patients were treated with doxycycline 200 mg daily with rapid recovery.
Figure 2 Skin lesions (patient A): 2A. vesicular erythematous rash of both legs; 2B. one inoculation eschar on the abdomen.
Figure 3 Electrocardiograms performed (A) at admission and (B) 6 hours later: repolarisation abnormalities not present initially are present in the inferior leads of the 2nd electrocardiogram.
Conclusion
Contrarily to R. rickettsii and R. conorii that are considered potential agents of myocarditis and pericarditis [1-3], R. africae usually present a benign, uncomplicated course, and has never been yet associated with cardiac complications. Our findings suggest that R. africae, the agent of ATBF, may lead to myocarditis.
In this report, the grouped cases, the presence of multiple inoculation eschars and the serologic seroconversion for R. africae, strongly supported the diagnosis of ATBF. To confirm the diagnosis, we could also perform PCR and/or culture on the biopsy of the eschar bite, since both techniques are good tools to diagnose acute rickettsial infection [4]. However, no skin biopsy was performed for ethical reasons since the diagnosis of ATBF was evident.
In presence of highly specific pathologic cardiac enzymes (troponin I) and ECG alteration, and in absence of coronary stenosis at coronarography, and despite the absence of morphological abnormalities at echocardiography, the more likely cause of the chest pain is a myocarditis with or without pericardial involvement. Indeed, myocarditis that is defined as an acute inflammatory syndrome involving the heart and related structures is typically characterized by increased troponin and normal coronary arteries [13-15]. Since the pathogenesis of rickettsial disease is generally associated with endothelium damage [1], unstable angina might occur in patients with spotted fever Rickettsiosis. However, in this case, the normal coronarography does not support the occurrence of a peripheric, transitory thrombotic event. A myocarditis is much more likely in this setting.
To our knowledge, there are no described case of cardiac involvement associated with a R. africae infection. Although a serological cross reaction with another rickettsial infection can not be formally excluded, the endemic presence of R.africae in Swaziland and South Africa [2-4], the presence of multiple inoculation eschars and the simultaneous infection of both travellers [5], strongly support the diagnosis of ATBF.
In conclusion, if ATBF usually presents a benign course, rare complications such as myocardial involvement may occur. Travellers to endemic areas should be informed of the risk of contracting ATBF and be encouraged to take personal protective measures against tick bites [5,7].
List of abbreviations
ATBF = african tick bite fever
CK = creatine phosphokinase
CK-MB = creatine phosphokinase muscle-brain isoform
NV = normal value
UI= international unit
PCR = polymerase chain reaction
ECG = electrocardiogram
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors were involved in patient care. CB wrote the first draft of the paper. All authors improved the manuscript and approved its final version.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank the "Unité des Rickettsies, Marseille, France" for performing the serologies.
==== Refs
Shah SS McGowan JP Rickettsial, ehrlichial and Bartonella infections of the myocardium and pericardium Front Biosci 2003 1 197 201
Marin-Garcia J Gooch WM Coury DL Cardiac manifestations of Rocky Mountain spotted fever Pediatrics 1981 67 358 61 6454108
Drancourt M Brouqui P Chiche G Raoult D Acute pericarditis in Mediterranean Spotted Fever Trans R Soc Trop Med Hyg 1991 85 799 1801359 10.1016/0035-9203(91)90461-7
Kelly PJ Beati L Mason PR Matthewman LA Roux V Raoult D Rickettsia africae sp nov, the etiological agent of African tick bite fever Int J Syst Bact 1996 46 611 4 8934912
Jensenius M Fournier PE Vene S Hoel T Hasle G Henriksen AZ Hellum KB Raoult D Myrvang B African tick bite fever in travelers to rural sub-Equatorial Africa Clin Infect Dis 2003 36 1411 7 12766836 10.1086/375083
Ndip LM Fokam EB Bouyer DH Ndip RN Titanji VP Walker DH McBride JW Detection of Rickettsia africae in patients and ticks along the coastal region of Cameroon Am J Trop Med Hyg 2004 71 363 6 15381820
Fournier PE Roux V Caumes E Donzel M Raoult D Outbreak of Rickettsia africae infections in participants of an adventure race in South Africa Clin Infect Dis 1998 27 316 23 9709882
Raoult D Fournier PE Fenollar F Jensenius M Prioe T de Pina JJ Caruso G Jones N Laferl H Rosenblatt JE Marrie TJ Rickettsia africae, a tick-borne pathogen in travelers to sub-Saharan Africa N Engl J Med 2001 344 1504 10 11357153 10.1056/NEJM200105173442003
Jackson Y Chappuis F Loutan L African tick-bite fever: four cases among Swiss travelers returning from South Africa J Travel Med 2004 11 225 8 15541225
Jensenius M Fournier PE Raoult D Tick-borne rickettsioses in international travellers Int J Infect Dis 2004 8 139 46 15109588 10.1016/j.ijid.2003.06.004
Fournier PE Jensenius M Laferl H Vene S Raoult D Kinetics of antibody responses in Rickettsia africae and Rickettsia conorii infections Clin Diagn Lab Immunol 2002 9 324 8 11874871 10.1128/CDLI.9.2.324-328.2002
Walker DH Valbuena GA Olano JP Pathogenic mechanisms of diseases caused by Rickettsia Ann N Y Acad Sci 2003 990 1 11 12860594
Feldman AM and McNamara D Myocarditis N Engl J Med 2000 343 1388 1398 11070105 10.1056/NEJM200011093431908
Smith SC Ladenson JH Mason JW Jaffe AS Elevations of cardiac troponin I associated with myocarditis. Experimental and clinical correlates Circulation 1997 95 163 8 8994432
Hamm CW Giannitsis E Katus HA Cardiac troponin elevations in patients without acute coronary syndrome Circulation 2002 106 2871 2 12460862 10.1161/01.CIR.0000044342.50593.63
|
16242016
|
PMC1274315
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Oct 20; 5:90
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-90
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-911624202710.1186/1471-2334-5-91Research ArticleSystemic and local antibiotic prophylaxis in the prevention of Staphylococcus epidermidis graft infection Turgut Huseyin [email protected] Suzan [email protected] Ilknur [email protected] Mustafa [email protected] Ibrahim [email protected] Semra [email protected] Ali [email protected] Nural [email protected] Koray [email protected] Ahmet [email protected] Department of Infectious Diseases and Clinical Microbiology, Pamukkale University, Faculty of Medicine, Denizli, Turkey2 Department of Microbiology and Clinical Microbiology, Pamukkale University, Faculty of Medicine, Denizli, Turkey3 Department of Cardiovascular Surgery, Pamukkale University, Faculty of Medicine, Denizli, Turkey4 Department of General Surgery, Pamukkale University, Faculty of Medicine, Denizli, Turkey2005 21 10 2005 5 91 91 11 6 2005 21 10 2005 Copyright © 2005 Turgut et al; licensee BioMed Central Ltd.2005Turgut et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The aim of the study was to investigate the in vivo efficacy of local and systemic antibiotic prophylaxis in the prevention of Staphylococcus (S.) epidermidis graft infection in a rat model and to evaluate the bacterial adherence to frequently used prosthetic graft materials.
Methods
Graft infections were established in the subcutaneous tissue of 120 male Wistar rats by implantation of Dacron/ePTFE grafts followed by topical inoculation with 2 × 107 CFUs of clinical isolate of methicillin-resistant S. epidermidis. Each of the graft series included a control group, one contaminated group that did not receive any antibiotic prophylaxis, two contaminated groups that received systemic prophylaxis with teicoplanin or levofloxacin and two contaminated groups that received teicoplanin-soaked or levofloxacin-soaked grafts. The grafts were removed 7 days after implantation and evaluated by quantitative culture.
Results
There was significant bacterial growth inhibition in the groups given systemic or local prophylaxis (P < 0.05). Methicillin-resistant S. epidermidis had greater affinity to Dacron graft when compared with ePTFE graft in the untreated contaminated groups (P < 0.05).
Conclusion
The study demonstrated that the usage of systemic or local prophylaxis and preference of ePTFE graft can be useful in reducing the risk of vascular graft infections caused by staphylococcal strains with high levels of resistance.
==== Body
Background
One of the most feared complications of the use of a prosthetic material is the appearance of infection after implant [1-5]. Graft infection often results in prolonged hospitalization, organ failure, amputation and death [4,6,7]. The causative organisms are predominantly S. auerus and S. epidermidis [4,8]. Most commonly contamination occurs at the time of graft insertion and the most frequent source of infection is from staphylococci from the patient's skin [3,4,9]. The most important strategies for the prevention of prosthetic infection are asepsis and the perioperative administration of systemic antibiotics [10-12]. Moreover, in the case of vascular grafts, alternative methods such as antimicrobials bound in high concentrations to prosthetic grafts have been proposed [2,10,11,13].
However, the antibiotic usage should be guided by local bacterial prevalence and sensitivities, remembering that most infections are caused by staphylococci. Levofloxacin is a fluoroquinolone with an enhanced activity against gram-positive cocci [14,15]. Teicoplanin is a glycopeptide antibiotic and has an excellent bactericidal activity against penicillinase-producing and methicillin-resistant S. epidermidis [16,17].
Synthetic vascular prostheses have been developed to supplement the limited supply of native graft materials. But they may act as a foreign body in the patient and may harbour bacteria which results in graft infection [18]. Vascular graft composition and construction have been shown to influence bacterial adherence [3,19].
The aim of the present study was to investigate the in vivo efficacy of local and systemic antibiotic prophylaxis in the prevention of S. epidermidis graft infection in a rat model and to evaluate the bacterial adherence to frequently used prosthetic graft materials.
Methods
This study was carried out in the animal laboratory of our institution. All animals received humane care in compliance with Principles of Laboratory Animal Care, formulated by the Guide for the Care and Use of Laboratory Animals, prepared by the National Academy of Sciences [20]. This study was also approved by the Pamukkale University Animal Research Ethics Committee, Denizli, Turkey.
The strain of methicillin-resistant S. epidermidis used in the present study was isolated from a clinical specimen (graft infection) submitted for routine bacteriological investigation. Identification of the clinical isolate was determined by Gram staining, catalase reaction, tube coagulation test and Api-staph test (biomérieux, Lyon, France). Methicillin sensitivity was investigated by oxacillin disk diffusion test [21]. Levofloxacin and teicoplanin (both from Aventis Pharma) were diluted in accordance with manifacturers' recommendations yielding 1 mg/ml stock solution. Solutions of drugs were made fresh on the day of assay.
One-hundred and twenty adult male Wistar rats (weight range 300 to 350 g) were used in this study. The study included two series composed of 6 groups for each of the woven, gelatin-imregnated polyethyleneterephthalate (Dacron) (Gelwave, Sulzer Vascutek) (D1–6) and expanded polytetrafluoroethylene (ePTFE) (Alpha Graft®, Alpha Research) (P1–6) grafts. Each of the series included one group with no graft contamination and no antibiotic prophylaxis (uncontaminated control, Dacron1, ePTFE1), one contaminated group that did not receive any antibiotic prophylaxis (untreated control, Dacron2, ePTFE2), one contaminated group that recieved teicoplanin-soaked grafts (Dacron3, ePTFE3), one contaminated group in which perioperative intraperitoneal prophylaxis with teicoplanin (10 mg/kg) was administered (Dacron4, ePTFE4), one contaminated group that recieved levofloxacin-soked grafts (Dacron5, ePTFE5) and one contaminated group in wich perioperative intraperitoneal prophylaxis with levofloxacin (10 mg/kg) (Dacron6, ePTFE6) was administered. Each rat was anesthetized with a 2:1 mixture of kethamine hydrochloride (100 mg/ml) (Pfizer):xylazine hydrochloride (20 mg/ml) (Bayer) at a dose of 0.75 ml/kg intramuscularly. Rats' hair of the back was shaved and the skin was cleaned with 10% povidone-iodine solution. One subcutaneous pocket was made on each side of the median line by a 1.5 cm incision. Aseptically, 1-cm2 sterile collagen-sealed Dacron or ePTFE grafts were implanted into the pockets. Prior to implantation, in the groups Dacron3, ePTFE3 and Dacron5, ePTFE5 the Dacron and ePTFE graft segments were impregnated with 1 mg/ml teicoplanin and levofloxacin, respectively. Antibiotic bonding was obtained immediately before implantation by soaking grafts for 20 minutes in a sterile solution of antibiotic. The effect of preoperative intraperitoneal teicoplanin and levofloxacin administered 30 minutes before implantation at the standard dose of 10 mg/kg was evalueted in the groups Dacron4, ePTFE4 and Dacron6, ePTFE6, respectively. The pockets were closed by means of skin clips and sterile saline solution (1 ml) containing methicillin-resistant strain S.epidermidis at a concentration of 2 × 107 CFUs/ml was inoculated onto the graft surface by using a tuberculin syringe to create a subcutaneous fluid-filled pocket. The animals were returned to individual cages and thoroughly examined daily. All grafts were removed 7 days following implantation.
The explanted grafts were placed in sterile tubes, washed in sterile saline solution, placed in tubes containing 10 ml of phosphate-buffered saline solution and sonicated for 5 minutes to remove the adherent bacteria from the grafts. Quantitation of viable bacteria was performed by culturing serial dilutions (0.1 ml) of the bacterial suspension on blood agar plates. All plates were incubated at 37°C for 48 hours and evaluated for the presence of the staphylococcal strains. The organisms were quantitated by counting the number of colony-forming units (CFUs) per plate.
Quantitative culture results were presented as arithmetic mean ± standard deviation (S.D.). Differences among the groups were evaluated using one-way analysis of variance (ANOVA), and multiple comparisons between the groups were performed with a posthoc test (Tukey's HSD test). Differences were considered statistically significant when P < 0.05. Data were analyzed by a statistical software (SPSS for Windows 11.0; SPSS, Chicago, Illinois).
Results
None of the animals included in any group died or had clinical evidence of drug related adverse effects, such as local signs of perigraft inflammation, anorexia, vomiting, diarrhoea, and behavioural alterations.
There was no anatomic and microbiological evidence of graft infection in the animals included in the uncontaminated control groups. In contrast, all 20 rats included in the untreated contaminated control groups (Dacron2 and ePTFE2) demonstrated evidence of graft infection, with quantitative culture results showing 3.7 × 107 ± 1.1 × 107 CFU/ml and 5.3 × 106 ± 2.4 × 106 CFU/ml, respectively. The quantitative graft cultures of the other groups demonstrated bacterial growth in different counts. The results are summarised in Table 1.
Table 1 Quantitative microbiological results of the in vivo experiments.
Groupa Graft-bonded drugb Intraperitoneal preoperative drugc Quantitative graft culture (CFUs/ml)
Dacron 1d,e -- -- 0.0
Dacron 2 -- -- 3.7 × 107 ± 1.1 × 107
Dacron 3d,e Teicoplanin -- 5.6 × 103 ± 1.2 × 103
Dacron 4d,e -- Teicoplanin 5.3 × 102 ± 1.2 × 102
Dacron 5d Levofloxacin -- 4.0 × 105 ± 5.5 × 104
Dacron 6d,e -- Levofloxacin 4.9 × 104 ± 4.7 × 103
ePTFE 1d,e -- -- 0.0
ePTFE 2d -- -- 5.3 × 106 ± 2.4 × 106
ePTFE 3d,e Teicoplanin -- 4.7 × 103 ± 1.2 × 103
ePTFE 4d,e -- Teicoplanin 4.2 × 102 ± 1.4 × 102
ePTFE 5d Levofloxacin -- 3.7 × 105 ± 2.8 × 104
ePTFE 6d,e -- Levofloxacin 4.7 × 104 ± 4.1 × 103
a Each group was performed by 10 animals; Dacron1–6, groups of animals by implantation of Dacron protheses; ePTFE1–6, groups of animals by implantation of ePTFE protheses.
b Graft segments were impregnated with 1 mg/ml teicoplanin; 1 mg/ml levofloxacin.
c Teicoplanin 10 mg/kg, levofloxacin 10 mg/kg.
d Statistically significant when compared with group Dacron2
e Statistically significant when compared with group ePTFE2
The amount of bacterial growth was statistically significantly higher in untreated contaminated Dacron-group (Dacron2) when compared to the other groups (P < 0.05). The second highest bacterial growth was observed in untreated contaminated ePTFE-group (ePTFE2) and it was also significantly different from all antibiotic-treated groups, except levofloxacin-bonded Dacron (Dacron5) and ePTFE (ePTFE5) groups (P < 0.05). Although the highest reduction in bacterial growth number was observed in the intraperitoneal teicoplanin groups (Dacron4 and ePTFE4), there was no statistically significant difference among the treated contaminated groups (P > 0.05).
Discussion
Various studies have demonstrated that systemic antibiotic prophylaxis reduces the incidence of prosthetic vascular graft infections, but not completely prevent them [22-24]. That's because antibiotic impregnated grafts are arousing interest as they can deliver antibiotic at the time that the graft is at the greatest risk of contamination [22].
In vascular surgery, S. epidermidis has been shown to be the leading isolate with infection appearing late after implantation [3,25,26]. So, in order to simulate the clinical setting, we preferred to use an isolate of S. epidermidis which was obtained from an infected vascular conduit in our hospital.
Teicoplanin and levofloxacin are attractive options for local and systemic antibiotic prophylaxis in preventing S. epidermidis graft infections, because they are effective antibiotics against coagulase positive and negative staphylococci [23,27]. Teicoplanin is used parenterally to treat infections caused by staphylococcal infections since the emergence of methicillin-resistant staphylococci [28]. Recently, teicoplanin has been administered as perioperative antibiotic prophylaxis [29-31]. Teicoplanin has a long half-life and good tissue and bone penetration [32]. In an earlier study at St James's University Hospital, teicoplanin exhibited good penetration into ischaemic tissue, which may be desirable for prophylaxis in vascular surgery [33]. Kester et al. [34] concluded from a two-centre study that a single dose of teicoplanin showed similar efficacy to a three-dose regimen of cephradine plus metronidazole as prophylaxis for wound infection in vascular surgery. It was showed that levofloxacin had greater in vitro and in vivo anti-staphylococcal activity than the other fluoroquinolones such as ofloxacin or ciprofloxacin [35-37]. In previous studies levofloxacin was reported to reach high concentrations in serum and various tissues, following single dose levofloxacin administration [38,39]. In the present study it was found that both local and systemic levofloxacin usage reduced bacterial count in the graft segments when compared to untreated contaminated controls (Dacron2 and ePTFE2) . Although it was not statistically significant, systemic usage of levofloxacin was more effective than the local one. However, interestingly, the reducing efficacy of local usage of levofloxacin on bacterial count was also inadequate to reach a statistical difference when compared to untreated contaminated ePTFE graft (ePTFE2) (P > 0.05).
The question remains whether local or systemic antibiotic prophylaxis is the best choice for reducing the risk of prosthetic vascular graft infection. When a local prophylaxis is used, the dose of antibiotic delivered to the operative site could be important in eliminating infections, as most of them are due to contamination at the time of implantation. Simple soaking in an antibiotic solution immediately prior to implantation is an easy way of impregnating the prosthetic graft that can be done extemporaneously by the surgeon himself. But during this procedure the risk of contamination of the graft increases.
Antibiotics having good activity against gram-positive bacteria were used in bonding vascular grafts in experimental models. Hernandez-Richter et al. [40] found that rifampicin and triclosan but not silver was effective in preventing bacterial infection of vascular Dacron graft material. Giacometti et al. [41] confirmed the efficacy of mupirocin-soaked grafts against methicillin-susceptible, methicillin-resistant and vancomycin-intermediate S. epidemidis. Ghiselli et al. [42] showed that Temporin A had an antibacterial in vitro activity against methicillin-susceptible and methicillin-resistant S. epidermidis. Efficacy of polycationic peptides in preventing vascular graft infection due to methicillin-resistant S. aureus with intermediate resistance to glycopeptides was demonstrated to be very good [11]. Dell'Acqua et al. [43] suggested that RNAIII-inhibiting peptide (YSPWTNF-NH2), applied locally and systemically, can completely inhibit drug-resistant S. aureus and S. epidermidis biofilms. Osada et al. [44] demostrated that levofloxacin incorporated into albumin-sealed Dacron graft had a bactericidal action and adhesive prevention against inoculated S. aureus in a graft model.
Soaking of the grafts in solutions of 1 mg/ml of different antibiotics has been shown to be effective in making antibiotic-bonding drugs [45]. Previously, Levofloxacin and teicoplanin, used in 10 mg/kg consantration have been shown to be successful in reducing the risk of prosthetic graft infection [23]. We chosed the same concentrations for the each antibiotic in our model according to these studies' findings.
Our data indicate that, although the difference was not statistically significant, local usage of teicoplanin more effectively reduces the bacterial count in graft segments than the local levofloxacin application and it seems to be the most appropriate antibiotic for local vascular graft prophylaxis. The finding that antibiotic-impregnated grafts alone can not prevent prosthetic vascular graft infection is similar to the results found by other groups [11,22]. The combined usage of systemic antibiotic prophylaxis and drug-bonded grafts has been shown to be more effective in decreasing the incidence of prosthetic vascular graft infections [10,23]. But this combination may have disadvantages like increased intraoperatif contamination risk and cost.
The results of this study demonstrated that both systemic and local prophylactic antibiotic treatment was useful. Although, the difference was not significant, the highest reduction in bacterial number was in the intraperitoneal teicoplanin groups in our study and this finding is parallel with previous studies, [23,37]. Second reason for choosing teicoplanin was its advantage as single dose application. This is certainly more desirable than frequent dosing required by conventional systemic antibiotics or other glycopeptides.
All kinds of prosthetic vascular grafts are susceptible in varying degrees to infection via direct contamination during implantation or bacteremia after operation. Dacron and ePTFE are the most frequently used materials. Surface area and molecular structure differs between the two types of grafts. Graft material of ePTFE is relatively nonporous when it is compared with multifilamented Dacron grafts. EPTFE is more hydrophobic than Dacron, perhaps that's why it is less likely to form bonds with those bacteria in which the cell walls have hydrophobic properties [19,46]. The findings of the present study was similar with previous studies reporting that S. epidermidis, S. aureus and Escherichia coli had greater affinity to Dacron graft when compared with ePTFE [19,47,48]. The bacterial count in the untreated contaminated ePTFE grafts was very low than the untreated contaminated Dacron groups. Untreated contaminated ePTFE graft was found to be almost as effective as levofloxacin-bonded Dacron and ePTFE grafts. Although there were not statistical signifficant differences among the treated contaminated groups, it was observed that bacterial growth was more in Dacron grafts. This finding may be of a clinical importance and may influence the choice of a surgeon when he has to prefer one of these grafts.
Further animal studies are needed to assess the efficacy of commercially available grafts soaked in various antibiotic solutions, against infections after sequental bacterial seeding for up 7 days. Additionally, these results have to be compared with the real situation of an implanted graft in a living human being. It should not be forgotten that antibiotic/antiseptic impregnation is not the only way of protecting synthetic grafts. Modifying the surface characteristics of prosthetic graft to minimise bacterial adherence needs to be investigated further.
Conclusion
The study demonstrated that the usage of systemic or local prophylaxis and preference of ePTFE graft can be useful in reducing the risk of vascular graft infections caused by staphylococcal strains with high levels of resistance.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HT and SS jointly conceived of the study, secured funding, participated in its design and coordination and co-drafted the manuscript. MS and AB participated in the design and coordination of the study, data analysis and writing the paper. IK, NC and ST participated in the microbiological studies and data analysis. IG and AA participated in the administration of the drugs and in the conduct of the animals' experiments. KT performed the statistical analysis and revised the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Soetevent C Klemm PL Stalenhoef AF Bredie SJ Vascular graft infection in aortoiliac and aortofemoral bypass surgery: clinical presentation, diagnostic strategies and results of surgical treatment Neth J Med 2004 62 446 52 15685895
Bunt TJ Vascular graft infections: an update J Cardiovasc Surg 2001 9 225 233 10.1016/S0967-2109(00)00104-6
Jones L Braithwaite BD Davies B Heather BP Earnshaw JJ Mechanism of late prosthetic vascular graft infection J Cardiovasc Surg 1997 5 486 489 10.1016/S0967-2109(97)00056-2
Bergamini TM Corpus RA Brittian KR Peyton JC Cheadle WG The natural history of bacterial biofilm graft infection J Surg Res 1994 56 393 396 8170137 10.1006/jsre.1994.1062
Taylor G Gravel D Johnston L Embil J Holton D Paton S Prospective surveillance for primary bloodstream infections occurring in Canadian hemodialysis units Infect Control Hosp Epidemiol 2002 23 716 20 12517012
Mingoli A Sapienza P Di Marzo L Sgarzini G Burchi C Modini C Cavallaro A Management of abdominal aortic prosthetic graft infection requiring emergent treatment Angiology 1997 48 491 495 9194534
Seeger JM Pretus HA Welborn MB Ozaki CK Flynn TC Huber TS Long-term outcome after treatment of aortic graft infection with staged extra-anatomic bypass grafting and aortic graft removal J Vasc Surg 2000 32 451 461 10957651 10.1067/mva.2000.109471
Abdulrahman IS Al-Mueilo SH Bokhary HA Ladipo GOA Al-Rubaish A A prospective study of hemodialysis access-related bacterial infections J Infect Chemother 2002 8 242 246 12373488 10.1007/s10156-002-0184-8
Liekwig WG Greenfield LJ Vascular graft infection: collected experience and results of treatment Surgery 1977 81 335 342 841471
Ghiselli R Giacometti A Cirioni O Mocchegiani F Orlando F Del Prete MS D'Amato G Scalise G Saba V Quinopristin/dalfopristin bonding in combination with intraperitoneal antibiotics prevent infection of knitted plyester graft material in a subcutaneous rat pouch model infected with resistant Staphylococcus epidermidis Eur J Vasc Endovasc Surg 2002 24 230 234 12217284 10.1053/ejvs.2002.1713
Giacometti A Cirioni O Ghiselli R Goffi L Moccheigiani F Riva A Scalise G Saba V Efficacy of polycationic peptides in preventing vascular graft infection due to Staphylococcus epidermidis J Antimicrob Chemother 2000 46 751 756 11062194 10.1093/jac/46.5.751
Hasselgren PO Ivarsson L Risberg B Sheeman T Effects of prophylactic antibiotics in vascular surgery. A prospective randomised double blind study Ann Surg 1984 200 86 92 6375599
Hennes N Sandmann W Torsello G Kniemeyer HW Grabitz K Infection of a vascular prosthesis-a retrospective analysis of 99 cases Chirurg 1996 67 37 43 8851674
Casellas JM Gilardoni M Tomè G Goldberg M Ivanovich S Orduna M Dolmann A Ascoli M Ariza H Montero JM Comparative in-vitro activity of levofloxacin against isolates of bacteria from adult patients with community-acquired lower respiratory tract infections J Antimicrob Chemother 1999 43 37 42 10404336 10.1093/jac/43.suppl_3.37
Soussy CJ Cluzel M Ploy MC Kitzis MD Morel C Bryskier A Courvalin P In-vitro antibacterial activity of levofloxacin against hospital isolates: a multicentre study J Antimicrob Chemother 1999 43 43 50 10404337 10.1093/jac/43.suppl_3.43
Pauluzzi S Del Favero A Menichetti F Barrata E Moretti VM Di Fillipo F Pasticci MB Guerciolini R Patoia L Frongillo RF Treatment of infections by staphylococci and other gram-positive bacteria with teicoplanin: an open study J Antimicrob Chemother 1987 20 431 438 2960643
De Lalla F Santoro D Rinaldi E Suter F Cruciani M Guaglianone MH Rizzardini G Pellegata G Teicoplanin in the treatment of infections by staphylococci, Clostridium difficil and other gram-positive bacteria J Antimicrob Chemother 1989 23 131 142 2526109
Thomas A Campbell GR Campbell JH Advances in vascular tissue engineering Cardiovasc Pathol 2003 12 271 276 14507577 10.1016/S1054-8807(03)00086-3
Schmitt DD Bandyk DF Pequet AJ Towne JB Bacterial adherence to vascular protheses J Vasc Surg 1986 3 732 40 2939263 10.1067/mva.1986.avs0030732
Institute of Laboratory Animal Resources National Research Council Guide for the care and use of laboratory animals 1996 Washington, DC: National Academy Press
Kloos WE Barnerman TL Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolken RH Staphylococcus and Micrococcus Manual of Clinical Microbiology 1999 7 Washington: ASM Press 264 277
Lenhgharardit FJ Torsello G Claeys LGY Pfeiffer M Wachol-Drewek Z Grundmann RT Sandmann W Systemic and local antibiotic prophylaxis in the prevention of prosthetic vascular graft infection: an experimental study Eur J Vasc Surg 2002 23 127 133 10.1053/ejvs.2001.1571
Ghiselli R Giacometti A Goffi L Cirioni O Boccoli G Mocchegiani F Riva A Scalise G Saba V Efficacy of rifampin-levofloxacin as a prophylactic agent in preventing Staphylococcus epidermidis graft infection Eur J Vasc Endovasc Surg 2000 20 508 511 11136585 10.1053/ejvs.2000.1239
Ghiselli R Giacometti A Goffi L Cirioni O Mocchegaiani F Orlando F Paggi AM Petrelli E Scalise G Saba V Prophylaxis against Staphylococcus aureus vascular graft infection with mupirocin-soaked, collagen-sealed Dacron J Surg Res 2001 99 316 320 11469904 10.1006/jsre.2001.6138
Sladen JG Chen JC Reid JDS An aggresive local approach to vascular graft infection Am J Surg 1998 176 222 225 9737638 10.1016/S0002-9610(98)00130-5
Bandyk DF Berni GA Thiele BL Towne JB Aortofemoral graft infection due to Staphylococcus epidermidis Arch Surg 1984 119 102 108 6689867
Murase K Hirose H Mori Y Takagi H Iwata H Sago T Kawamura Y Graft-preserving treatment for vascular graft infected with Staphylococcus aureus with antibiotic-releasing porous apatite ceramic in the rabbit J Vasc Surg 2003 38 368 73 12891122 10.1016/S0741-5214(03)00120-4
Woodford N Johnson AP Mirrison D Speller DCE Current perspectives on glycopeptide resistance Clin Microbiol Rev 1995 8 585 615 8665471
Lim SH Smith MP Salooja N Machin SJ Goldstone AH A prospective randomized study of prophylactic teicoplanin to prevent early Hickman catheter related sepsis in patients receiving intensive chemotherapy for haematological malignancies J Antimicrob Chemother 1991 28 109 116 1837550
De Lalla F Viola R Pellizzer G Lazzarini L Tramarin A Fabris P Regional prophylaxis with teicoplanin in monolateral or bilateral total knee replacement: an open study Antimicrob Agents Chemother 2000 44 316 319 10639356 10.1128/AAC.44.2.316-319.2000
Marroni M Cao P Fiorio M Maghini M Lenti M Repeto A Menichetti F Prospective, randomized, double-blind trial comparing teicoplanin and cefazolin as antibiotic prophylaxis in prosthetic vascular surgery Eur J Clin Microbiol Infect Dis 1999 18 175 178 10357049 10.1007/s100960050253
Suter F Avai A Bottura P Fusco U Gerundini M Caprioli S Maggiolo F Teicoplanin versus cefamandole in the prevention of infection in total hip replacement Eur J Clin Microbiol Infect Dis 1994 13 793 796 7889947 10.1007/BF02111338
Antrum RM Bibby SR Ramsden CH Kester RC Teicoplanin: part I. An evaluation of the concentrations seen in serum and the subcutaneous fat of the relatively ischaemic limb following a single intravenous bolus Drugs Exp Clin Res 1989 15 21 23 2526008
Kester RC Antrum R Thornton CA Ramsdenf CH Harding I A comparison of teicoplanin versus cephradine plus metronidazole in the prophylaxis of post-operative infection in vascular surgery J Hosp Infect 1999 41 233 243 10204127 10.1016/S0195-6701(99)90022-1
Von Eiff C Peters G In-vitro activity of ofloxacin, levofloxacin and D-ofloxacin against staphylococci J Antimicrob Chemother 1996 38 259 263 8877540
Gesu GP Marchetti F Piccoli L Cavallero A Levofloxacin and ciprofloxacin in vitro activities against 4,003 clinical bacterial isolates collected in 24 Italian laboratories Antimicrob Agents Chemother 2003 47 816 819 12543701 10.1128/AAC.47.2.816-819.2003
Chambers HF Xiang Liu Q Liuxin Chow L Hackbarth C Efficacy of levofloxacin for experimental aortic-valve endocarditis in rabbits infected with viridans group streptococcus or Staphylococcus aureus Antimicrob Agents Chemother 1999 43 2742 2746 10543757
Von Baum H Böttcher S Hoffmann H Sonntag HG Tissue penetration of a single dose of levofloxacin intravenously for antibiotic prophylaxis in lung surgery JAC 2001 13 729 730 11328803
Swoboda S Oberdorfer K Klee F Hoppe-Tichy T von Baum H Geiss HK Tissue and serum concentrations of levofloxacin 500 mg administered intravenously or orally for antibiotic prophylaxis in biliary surgery JAC 2003 51 459 462 12562723
Hernandez-Richter T Schardey HM Wittmann F Mayr S Schmitt-Sody M Blasenbreu S Heiss MM Gabka C Angele MK Rifampin and triclosan but not silver is effective in preventing bacterial infection of vascularDacron graft material Eur J Vasc Surg 2003 26 550 557 10.1016/S1078-5884(03)00344-7
Giacometti A Cirioni O Ghiselli R Goffi L Viticchi C Mocchegiani F Riva A Orlando F Saba V Scalise G Mupirocin prophylaxis against methicillin susceptible, methicillin-resistant or vancomycin-intermediate Staphylococcus epidermidis vascular-graft infection Antimicrob Agents Chemother 2000 44 2842 2844 10991868 10.1128/AAC.44.10.2842-2844.2000
Ghiselli R Giacometti A Cirioni O Mocchegiani F Orlando F Kamysz W Del Prete MS Lukasiak J Scalise G Saba V Temporin A as a prophylactic agent against methicillin sodium-susceptible and methicillin sodium-resistant Staphylococcus epidermidis vascular graft infection J Vasc Surg 2002 36 1027 30 12422090 10.1067/mva.2002.127530
Dell'Acqua G Giacometti A Cirioni O Ghiselli R Saba V Scalise G Gov Y Balaban N Suppression of drug-resistant staphylococcal infections by the quorum-sensing inhibitor RNAIII-inhibiting peptide J Infect Dis 2004 190 318 320 15216467 10.1086/386546
Osada T Yamamura K Fujimoto K Mizuno K Sakurai T Ohta M Nabeshima T Prophylaxis of local vascular graft infection with levofloxacin incorporated into albumin-sealed dacron graft (LVFX-ALB Graft) Microbiol Immunol 1999 43 317 21 10385197
Galdbart JO Branger C Andreassian B Lambert-Zechovsky N Kitzis M Elution of six antibiotics bonded to polyeyhylene vascular grafts sealed with three proteins J Surg Res 1996 66 174 178 9024831 10.1006/jsre.1996.0391
Falkenback D Lundberg F Ribbe E Ljungh A Exposure of plasma proteins on Dacron and ePTFE vascular material in a perfusion model Eur J Vasc Surg 2000 19 468 475 10.1053/ejvs.1999.1075
Malangoni MA Livingston DH Peyton JC Of protein Binding on the adherence of staphylococci to prosthetic vascular grafts J Surg Res 1993 54 168 172 8479177 10.1006/jsre.1993.1027
Camiade C Goldschmidt P Koskas F Ricco JB Jarraya M Gerota J Kieffer E Optimization of the resistance of arterial allografts to infection: comparative study with synthetic prostheses Ann Vasc Surg 2001 15 186 196 11265083 10.1007/s100160010051
|
16242027
|
PMC1274316
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Oct 21; 5:91
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-91
|
oa_comm
|
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-6-371622974710.1186/1471-2350-6-37Research ArticleInsulin Promoter Factor 1 variation is associated with type 2 diabetes in African Americans Karim Mohammad A [email protected] Xiaoqin [email protected] Terri C [email protected] Steven C [email protected] Endocrinology Section, Medical Service, Central Arkansas Veterans Healthcare System, College of Medicine, University of Arkansas for Medical Sciences, Little Rock, Arkansas 72205 USA2 Division of Endocrinology and Metabolism, Department of Internal Medicine, College of Medicine, University of Arkansas for Medical Sciences, Little Rock, Arkansas 72205 USA2005 17 10 2005 6 37 37 23 5 2005 17 10 2005 Copyright © 2005 Karim et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Defective insulin secretion is a key defect in the pathogenesis of type 2 diabetes (T2DM). The β-cell specific transcription factor, insulin promoter factor 1 gene (IPF1), is essential to pancreatic development and the maintenance of β-cell mass. We hypothesized that regulatory or coding variants in IPF1 contribute to defective insulin secretion and thus T2DM.
Methods
We screened 71 Caucasian and 69 African American individuals for genetic variants in the promoter region, three highly conserved upstream regulatory sequences (PH1, PH2 and PH3), the human β-cell specific enhancer, and the two exons with adjacent introns. We tested for an association of each variant with T2DM Caucasians (192 cases and 192 controls) and African Americans (341 cases and 186 controls).
Results
We identified 8 variants in the two populations, including a 3 bp insertion in exon 2 (InsCCG243) in African Americans that resulted in an in-frame proline insertion in the transactivation domain. No variant was associated with T2DM in Caucasians, but polymorphisms at -3766 in the human β-cell enhancer, at -2877 bp in the PH1 domain, and at -108 bp in the promoter region were associated with T2DM in African American subjects (p < 0.01), both individually and as haplotypes (p = 0.01 correcting by permutation test). No SNP altered a binding site for the expected β-cell transcription factors. The rare alleles of InsCCG243 in exon 2 showed a trend to over-representation among African American diabetic subjects (p < 0.1), but this trend was not significant on permutation test.
Conculsion
The common alleles of regulatory variants in the 5' enhancer and promoter regions of the IPF1 gene increase susceptibility to type 2 diabetes among African American individuals, likely as a result of gene-gene or gene-environment interactions. In contrast, IPF1 is not a cause of type 2 diabetes in Caucasians. A previously described InsCCG243 variant may contribute to diabetes susceptibility in African American individuals, but is of low penetrance.
==== Body
Background
Type 2 diabetes (T2DM) has a substantial genetic component, but genetic heterogeneity, gene-environment interactions, and a large number of loci with small effect have combined to confound the identification of susceptibility genes. The pathogenesis of T2DM is characterized by early resistance to insulin-mediated glucose uptake and β-cell dysfunction followed by the further inexorable decline in function and possibly mass [1]. To maintain insulin secretion and glucose homeostasis in the face of resistance to insulin mediated glucose uptake, the β-cell must increase insulin secretion, either by increased function or increased β-cell mass, a concept known as β-cell compensation [2]. Impaired β-cell function predicts future diabetes [3], and work from our laboratory [4] and others [5,6] suggest that the ability of the pancreatic β-cell to compensate for prevailing insulin sensitivity (ie, β-cell compensation) is highly heritable. Nonetheless, the genetic controls over β-cell failure are largely unknown.
The islet transcription factor, insulin promoter factor 1 (IPF1) gene (also known as the pancreatic duodenal homeobox 1, PDX1, and insulin upstream factor 1, IUF1), is required for both the differentiation and maintenance of the β-cell phenotype [7]. The importance of IPF1 in pancreatic β-cell development and function is demonstrated in naturally occurring human mutations and in experimental mouse models. Humans lacking a functional IPF1 allele have pancreatic agenesis [8], whereas humans heterozygous for the same variant develop early onset, insulin deficient diabetes (Maturity Onset Diabetes of the Young, MODY4) [9]. Similarly, mice homozygous for targeted disruption of IPF1 (PDX1) fail to develop a pancreas, whereas haploinsufficient mice have impaired glucose-stimulated insulin secretion and develop T2DM with aging [7]. Impaired glucose homeostasis with reduced IPF1 activity likely derives from an influence on both β-cell mass and function. Both isolated mouse islets and dispersed β-cells with haploinsufficiency for IPF1 showed increased apoptosis even at basal glucose levels, but were functionally normal [10]. However, IPF1 also transactivates the promoters of multiple islet-specific genes, including insulin, the GLUT2 islet glucose transporter, islet amyloid polypeptide (IAPP), and somatostatin [11]. Thus, IPF1 sequence variants might be expected to influence both β-cell mass and the expression of key β-cell genes.
Regulation of IPF1 gene expression is complex, with multiple upstream regulatory elements that bind other key β-cell genes including HNF3β, HNF1α, and SP transcription factors [12,13]. Upstream sequences include a human and β-cell specific enhancer region at -3.7 kb to -3.4 kb that binds HNF1α, HNF3β, SP1, and SP3, and 3 additional regions that are highly conserved between mouse and human: PH1 (-2.6 kb to -2.8 kb), PH2 (-2.1 kb to -2.2 kb), and PH3 (-1.6 kb to -1.9 kb) [14]. PH1 and PH2 bind HNF3β, and PH1 also binds IPF1 [12] and HNF1α [15].
Multiple studies have examined the IPF1 gene for mutations in early onset, autosomal dominant diabetes and a few studies have searched for mutations in T2DM in Caucasians [16-19]. Only rare coding variants have been identified. However, neither the far upstream regulatory regions nor other ethnic groups including African Americans have been examined. We hypothesized that common variants in coding or regulatory regions of the IPF1 gene contribute to the failure of β-cell compensation and to the susceptibility to common T2DM. To address this hypothesis, we screened the coding and upstream regulatory regions of the IPF1 gene in both African American and Caucasian diabetic individuals with a family history of diabetes. We then tested both individual variants and haplotypes for diabetes susceptibility using a case-control design for each population.
Methods
Experimental subjects
We examined two populations: Caucasian individuals ascertained primarily for Northern European Ancestry, and African American individuals. Screening for new sequence variants was conducted in two stages. Initial studies of coding, 5' and 3' untranslated regions, 5' flanking region, and far upstream enhancer and regulatory elements upstream were conducted in 48 individuals: 12 African American subjects with T2DM, 12 African American control subjects, 12 Caucasian subjects with T2DM, and 12 nondiabetic Caucasian subjects. To better detect uncommon coding variants, we subsequently screened an additional 45 African American subjects with T2DM and an additional 47 Caucasian subjects with T2DM for exonic regions. Thus, exons were screened in total for 12 African American control subjects and 57 African American subjects with T2DM, and for 12 Caucasian control subjects and 59 Caucasian subjects with T2DM. To further improve sensitivity to detect coding variants, we selected affected subjects with early onset of T2DM: ages 25 – 40 years in African American subjects, and ages 30 – 45 years in Caucasian subjects.
Case control studies were conducted similarly in both Caucasian and African American populations. Our Caucasian study comprised 188 unrelated nondiabetic control individuals (73 male, 115 female) and 190 individuals with T2DM (133 male, 57 female), and has been described previously [20]. This population has 80% power to detect an absolute difference in allele frequencies of 10% difference between cases and controls for minor allele frequencies in controls of 10% to 50%. Initial case-control studies in African Americans were conducted on 165 control individuals (82 male, 83 female) and 255 diabetic cases (142 male, 113 female). This population likewise has at least 80% power to detect a difference between case and control allele frequencies of 10% over the range of allele frequencies from 10% to 50%. For both case-control studies, all diabetic individuals had at least one diabetic first degree relative. Control individuals had a normal 75 g oral glucose tolerance test or a fasting or random glucose below 5.6 mmol/l, and no diabetic first degree relative. No individual with known impaired glucose tolerance was included in either group, but because some subjects were ascertained at health fairs, not all subjects underwent glucose tolerance testing and thus impaired glucose tolerance could not be excluded.
During the course of this study, additional African American samples became available. Given the evidence for association and newly available samples, we subsequently expanded typing for SNPs 1, 4, 11, and the INSCCG243 variant by an additional 21 African American cases and 85 African American controls. Hence, for these markers we present data for on 186 African American controls (95 male, 91 female) and 341 African American diabetic individuals (186 male, 155 female). The African American control population had a BMI of 30.2 ± 7.1 kg/m2 and an age of 42.7 ± 13.0 years. The African American T2DM population had a BMI of 32.4 ± 7.3 kg/m2, and age of diabetes diagnosis of 42.6 ± 11.9 years, and an age at testing of 55.0 ± 12.6 years. All subjects provided informed consent under protocols approved by the University of Utah or University of Arkansas for Medical Sciences Institutional Review Boards.
Mutation detection and genotyping
We designed primers from the human genome sequence (AL353195) and alignment with the human IPF1 mRNA sequence (NM_000209) to cover exons 1 and 2, the 5' and 3' untranslated regions, 1.5 kb of 5' flanking and proximal promoter sequence, and the reported enhancer and regulatory elements PH1, PH2, and PH3 at positions -3.6 kb, -2.76 kb, -2.2 kb, and -1.76 kb from the ATG start site [12,15] (Figure 1). Initial screening was by denaturing high pressure liquid chromatography (DHPLC) using a Transgenomic WAVE HT DNA Fragment Analysis System (Transgenomic, Inc, Omaha, NE). Altered migration was confirmed and characterized by bidirectional sequence analysis [21] using infrared dye-labeled primers and GR4200 Sequencers (LI-COR Biotech, Lincoln, NE).
The proline insertion variant (InsCCG243) was typed using infrared dyes with detection on a LICOR GR4200 sequencer and scored using SAGA GT fragment analysis software (LICOR Biotech). Because the sequences in the exon 2 around the InsCCG243 variant are highly G-C rich, we confirmed our results using the Advantage GC-2 PCR kit (BD Biosciences Clontech, Palo Alto, CA). The remaining 8 SNPs were genotyped by Pyrosequencing on a PSQ-96 machine according to manufacturer methods (Biotage AB, Uppsula, Sweden). Primer sequences are available in Table 1.
Statistical and binding factor analyses
Our primary analysis was allelic association. Allelic frequencies in cases and controls were compared using the Fisher exact test. We report both the uncorrected p values for allelic association and the simulated p values which correct for the number of tests using HaploView version 3.2 [22]. We considered p < 0.05 to be significant without correction for multiple testing. In exploratory analyses, we also examined SNPs with an allelic association of p < 0.10 using several analyses. First, we tested for a genotypic association using the Fisher Exact Test under dominant and recessive models. Second, we tested for association using logistic regression analysis under additive, dominant and recessive models. For uncommon SNPs in which few recessive individuals were observed, recessive and additive models were not tested. Logistic regression included age (testing age for controls, age of diagnosis for cases), ln-transformed body mass index (BMI), and gender as covariates.
Pair-wise linkage disequilibrium coefficients were calculated by allele counting from the combined case and control data. Phase was estimated using the Expectation Maximum algorithm. Haplotype distribution between cases and controls was tested using Phase v2.1.1 [23], Arlequin [24], or HaploView v3.2 [22]. TagSNPs were selected using the LDSelect program based on the correlation between SNPs (r2) set at 0.8 [25]. Altered transcription factor binding sites were identified using the TFSEARCH program based on the TRANSFAC database [26].
Results
We detected a total of 9 sequence variants, including 8 SNPs in noncoding regions and a single coding variant observed only in African American subjects and comprising a 3 bp CCG/proline insertion (InsCCG243) in exon 2 (Table 2 and Figure 1). No other common or rare coding variants were detected among a total of 138 African American or 142 Caucasian alleles, including the previously reported D76N variant [17,19]. We identified 3 SNPs in the far 5' regulatory sequences, including one SNP in the human-specific enhancer region (SNP1) and two SNPs (SNP3, SNP4) in the PH1 region. SNP3 was common among Caucasians but not observed in African Americans. Two additional SNPs (SNP2, SNP11) were located in the proximal 5' flanking region (Figure 1).
We tested each of 7 SNPs that had minor allele frequencies over 10% in 190 Caucasian individuals with T2DM and 188 Caucasian control individuals. No individual SNP was associated with T2DM (p > 0.9 on permutation p value for all; Table 2 shows allelic frequencies, Table 3 provides raw numbers). The complete variation in the 7 SNPs could be captured with only 4 tagSNPs (SNPs 1, 2, 3, and 4 at positions -3766, -2890, -2877, and -279). All 7 SNPs fell into a single haplotype block with pairwise D' values of 1.0. Consistent with these observation, only 4 haplotypes were observed at over 1% frequency (Table 4). The 4 haplotypes could be distinguished by typing SNP 2, SNP 3, and SNP 4. Neither the distribution of haplotypes in Caucasians (p = 0.98 by Phase v2.1.1), nor any individual haplotype (p > 0.44; Table 4) was associated with T2DM.
In contrast to Caucasians, SNP1 in the human β-cell specific enhancer, SNP4 in the PH1 region, and SNP11, a G insertion in the proximal 5' flanking region, were significantly associated with T2DM (p = 0.007, p = 0.008, and p = 0.0008, respectively), with predicted odds ratios for the major allele of 1.59, 1.45, and 1.79, respectively. In contrast, the proline insertion in exon 2 (InsCCG243) showed a trend to an association, but did not reach statistical significance even without correction for multiple testing (p = 0.088, OR 1.58). The common alleles of SNPs 1, 4, and 11 were over-represented in subjects with T2DM, whereas the insertion allele of InsCCG243 was increased in T2DM subjects. SNPs 1 and 11 were in strong linkage disequilibrium (r2 = 0.927). Based on r2>0.8, we could capture the full diversity among African American subjects with 6 tagSNPs: SNPs 1, 2, 4, 5, 6, and InsCCG243. Using all observed variants, we identified only 7 haplotypes with over 1% frequency. Only Ins243CCG fell outside of the block defined using confidence interval definitions (Figure 2). Although the overall distribution of haplotypes was different between cases and controls (permuted p = 0.01), no single haplotype was over-represented in cases compared with controls (Table 5). In contrast, when the 3 individually associated variants were examined together, 2/3 haplotypes showed 8% differences between cases and controls (Table 5). The Ins243CCG proline insertion split the most common haplotype, and occurred on a single haplotype that showed a similar distribution between cases and controls as the Ins243CCG SNP (Table 5). When Ins243CCG was included in the analysis, neither major haplotype (CAID or CAII, where I is the insertion of the G at SNP 11 or the proline at Ins243CCG and D is the absence of the extra bases) was associated with T2DM when the proline insertion was included, but was associated when not split. Hence, the proline insertion was not driving the observed association.
In exploratory analyses, we sought to determine the most likely mode of inheritance and to determine whether the observed allelic associations were modulated by age, age of onset, or obesity. SNPs 1 and 11 acted as a recessive trait for the major allele (p = 0.017 and 0.003, respectively), whereas SNP 4 acted as a dominant trait for the major allele (p = 0.0016). No other SNP was associated with T2DM on exploratory analyses. Logistic regression confirmed the allelic association tests. Only SNPs 1, 4, and 11 and BMI were significant factors in the model. SNP 4 showed a stronger dominant effect (p = 0.0002, OR 3.23) with correction for age, BMI, and gender, whereas SNPs 1 and 11 were again consistent with a recessive effect of the major allele (p = 0.017 and 0.003, respectively, and OR 1.713 and 1.916, respectively).
The haplotype analysis and association analyses did not suggest which of the 3 SNPs was driving the observed association. No SNP altered the binding sites for known β-cell regulatory factors, including HNF1α, HNF1β, HNF3β, SP1/3, or auto-regulatory IPF1/PDX1 binding [13]. However, the minor allele of SNP1 abolished the predicted binding of heat shock factors 1 and 2 (HSF1 and HSF2), which are involved in cellular stress responses [27].
Discussion
As a key transcription factor in the pathways controlling both β-cell mass and essential genes for insulin biosynthesis and secretion, IPF1 is a strong candidate for the inherited defect in insulin secretion that characterizes T2DM and the prediabetic state. Mutations in IPF1 are a rare cause of early onset T2DM (MODY4)[16,18,28]. Several previous studies have searched for mutations in late onset T2DM among Caucasians [17,19,28] with variable results, but these studies have not focused on the well described conserved elements that extend 5 kb upstream of the ATG translation start site. Furthermore, no published study has examined a non-Caucasian population. The role of previously reported, rare nonsynonymous SNPs in typical T2DM [17,19] is unclear. We recently were unable to demonstrate a major role in T2DM susceptibility or in reduced insulin secretion for the most common of these missense variants, D76N, among Caucasians [29]. In screening 282 African American diabetic and 96 African American control subjects, we observed the D76N variant only in 3 individuals with T2DM (allelic frequency 0.005), and thus lacked the power to evaluate this variant in African American subjects. In the present study, we found no new coding variants among Caucasian samples, nor were any of 6 SNPs in the 5' flanking region, including those in the enhancer and PH1 domains, associated with T2DM in Caucasians. The lack of involvement of IPF1 in Caucasians was supported by the haplotype analysis.
In contrast, 3 of 8 sequence variants in African Americans were associated with T2DM, and two variants that were not seen in Caucasians showed a trend to an association. SNPs 1 and 11 (G insertion at -108 bp) were in strong linkage disequilibrium. SNP11 was previously reported in Japanese, where the (G)4 allele was less common than in African Americans and was of similar frequency in 88 cases and 67 controls [30]. Genetic studies likely cannot distinguish the impact of SNP1 in the enhancer and SNP11 in the proximal promoter on IPF1 transcription. The association of SNP11 was statistically the strongest. The haplotypes constructed from SNPs 1, 4, and 11 together confirmed the individual SNP results, but the association was not stronger using either PHASE or HaploView 3.2 than observed for individual SNPs. Hence, we cannot determine whether SNPs 1, 4, and 11 might interact to increase the risk of T2DM. Examination of the two haplotypes that were associated with T2DM showed that the risk and protective haplotypes differed at all three positions (CAG4 vs TTG3). Only the G3 allele uniquely distinguished a haplotype that differed in frequency between cases and controls. Notably, this haplotype is protective with regard to T2DM susceptibility. The "risk" haplotype (GAG4) is very common (71% of T2DM, 62.5% of controls). The contribution of the high prevalence allele to T2DM susceptibility has also been seen in other T2DM susceptibility genes, such as the PPARγ Pro12Ala variant [31].
The SNPs detected in this study are not predicted to alter the binding of known regulators of IPF1 gene expression, including HNF1α, HNF3β, or IFP1/PDX1, which have been shown to bind to these two regions [14]. SNPs 3 and 4 lie only 33 bp apart in the highly conserved PH1 element and approximately 50 bp upstream of the binding sites for known β-cell transcription factors NKX2.2, PBX1, and HNF3β. Several predicted binding sites for other transcription factors are altered by these associated variants. As noted above, the minor (T) allele of SNP1 is predicted to abolish the binding of heat shock proteins HSF1 and HSF2, although the role in insulin secretion or β-cell function is speculative. The common (A), T2DM-associated allele of SNP4 was predicted to abolish binding of the transcription factor Ets1. Ets1 has been described primarily in oncogenesis and angiogenesis, and is expressed in endothelial and lymphoid cells [32]. A role in the pancreatic β-cell has not been described previously. For SNP11, the T2DM-associated (G)4 was predicted to bind basic helix-loop-helix factor E47, whereas this binding was not present for the minor (G)3 allele. E47 is widely distributed, including pancreatic β-cells where it is well described as a regulator of insulin gene transcription [33]. Binding of E47 to the (G)4 allele might block activation by other transcription factors, and thus explain the association.
SNPs 1, 4, and 11 were not associated with T2DM in Caucasians. Thus, the association in African Americans may result from gene-gene or gene-environment interactions that are unique to this population. Alternatively, African Americans are known to be an admixed population, and concerns have been raised regarding spurious associations as a result of this admixture [34,35]. Several factors argue against a spurious association based on population structure, however. First, the 3 associated SNPs have similar frequencies in Caucasians and African Americans, such that population structure from admixutre would be less likely to lead to a spurious association. Second, of 87 SNPs previously typed, including 16 randomly chosen for large differences in African American and Caucasian allele frequencies and 71 chosen from candidate genes, only 3 have shown differences in allele frequencies that were significant at the p < 0.05 level. Thus, the findings in the current study have not been observed for multiple other genes tested.
Recently, lack of power has been raised as a reason for the inconsistent replication of associations [31,36,37], and very large sample sizes have been proposed to detect the small effects of variants such as the P12A polymorphism of the PPARγ gene or the E23K variant of the β-cell potassium channel, KCNJ11 [38]. By these standards, our study is small and likely would not have detected effects in either Caucasian or African American populations with a relative risk of below 1.4. However, among Caucasians we found no trend to an association. Indeed, for SNP 3, which showed the largest difference between cases and controls for any IPF1 SNP in Caucasians, achievement of 80% power to detect a difference significant at p < 0.05 would require 1700 cases and 1700 controls, based on a test of allelic association (3400 alleles for each group). Thus, although we cannot exclude an effect of these variants on T2DM risk that is comparable to that of PPARγ, the likelihood that a large enough study will be performed is small. The tagSNPs derived from our study will be useful should other investigators choose to undertake such a study.
The most intriguing of the unique African American variants is InsCCG243, which results in the inframe insertion of a proline in the carboxy-terminal polyproline tail, a region that is predicted to be involved in transactivation. We found this variant exclusively among African American subjects. We did not observe InsCCG243 among 142 Caucasian haplotypes, nor have other authors reported this variant in Caucasian populations [19,28]. InsCCG243 was reported previously in two French families, where it appeared to segregate in an autosomal dominant fashion and was associated with progressive insulin impairment [17]. The ethnic origin of these French families was not reported [17] and may have been of African or Afro-Caribbean. Expression of the InsCCG243 allele inhibited the endogenous IPF1 activation of the insulin gene by over 50% [17]. In the current study, InsCCG243 had an allele frequency of nearly 10% among diabetic subjects and 6.3% among controls. However, this difference did not reach statistical significance under any model in our studies. Hence, although InsCCG243 could contribute to 18% of T2DM among African Americans, the high prevalence observed in control individuals who had normal glucose tolerance and no family history of T2DM is inconsistent with the very high penetrance suggested by Hani et al [17]. Furthermore, among our African American families, InsCCG243 did not segregate in an autosomal dominant fashion (unpublished observations).
Conclusion
We have carefully examined the IPF1 gene in two ethnic groups. We have extended earlier studies to the highly conserved upstream regulatory regions. Although we find no evidence for an association with T2DM among Caucasians, three putative regulatory variants are associated with T2DM in African Americans. Furthermore, a proline insertion in the transactivation domain was unique to African Americans and showed a trend to an association. These variants thus may explain part of the increased diabetes prevalence among African Americans. However, the lack of association of the same variants in Caucasians suggests gene-gene or gene-environment interactions, or perhaps a spurious association due to population structure. Additional population association and physiologic studies will be needed to confirm and extend these findings.
Abbreviations
T2DM, type 2 diabetes
BMI, body mass index
IPF1, insulin promoter factor 1
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MAK was responsible for the direct conduct of the study including screening for sequence variation, assay design and data collection. XW assisted with genotyping, particularly of the InsCCG243 variant. TCH assisted with recruitment of all African American subjects and assisted in preparing data for analysis. SCE was responsible for study conception, oversight, all data analyses, and manuscript preparation. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by grants DK39311 and DK54636 from the National Institutes of Health/NIDDK, by the Research Service of the Department of Veterans Affairs, and by grants from the American Diabetes Association. Subject ascertainment, DNA preparation, data management and statistical assistance were supported in part by grant M01RR14288 from National Institutes of Health/National Center for Research Resources to the General Clinical Research Center (GCRC) of the University of Arkansas for Medical Sciences, College of Medicine. We thank the GCRC nursing staff for the support of this study, Judith Johnson Cooper for assistance with ascertainment of African American subjects in Arkansas, and Zhengxian Zhang for assistance with haplotype analyses in Arlequin.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Map of the IFP1 gene and upstream regulatory regions. Upstream regulatory regions are shown in white boxes. Exons are shown in boxes; translated regions are black and the untranslated regions are grey. The approximate locations of each SNP and the CCG (proline) insertion polymorphism are shown. Lines with arrow heads show the areas covered for PCR/DHPLC screening.
Figure 2 Linkage disequilibrium (D') plot of the IPF1 gene in African American subjects. SNPs are shown by location as in Table 1 and Figure 1. Squares without numbers represent D' values of 1.0; all numbers represent the D' value expressed as a percentile. Red squares represent pairs with LOD score for linkage disequilibrium of ≥ 2, blue squares represent D' = 1 but LOD<2, and white squares represent LOD<2 and D'<1.0. Plots were generated using HaploView v3.2.
Figure 3 Linkage disequilibrium (D') plot of IPF1 gene in Caucasians. SNPs are shown by location as in Table 1 and Figure 1. As in Figure 2, plots were generated in HaploView v3.2. Squares without numbers represent D' values of 1.0; all numbers represent the D' value expressed as a percentile. Red squares represent pairs with LOD score for linkage disequilibrium of ≥ 2, blue squares represent D' = 1 but LOD<2, and white squares represent LOD<2 and D'<1.0. plots were generated using HaploView v3.2.
Figure 4 Linkage disequilibrium (r2) plot for in African Americans. Figure 4 is analogous to the D' plots shown in Figure 2. The r2 value is shown on a grey scale, where white represents r2 = 0, black represents r2 = 1, and shades of grey represent 0<r2<1. All plots were generated using HaploView v3.2. SNP names and locations are as in Figure 2.
Figure 5 Linkage disequilibrium (r2) plot for Caucasians. Figure 5 is analogous to the D' plots shown in Figure 3. The value of r2 is shown on a grey scale, where white represents r2 = 0, black represents r2 = 1, and shades of grey represent 0<r2<1. All plots were generated using HaploView v3.2. SNP names and locations are as in Figure 3.
Table 1 Primer sequences for IPF1 SNPs
SNP NAME FORWARD PRIMER 5' to 3' REVERSE PRIMER 5' to 3' SEQUENCE PRIMER 5' to 3' Anneal Temp.
SNP1 *ATTGCTTAGCCCTAGGAATAT AGAGGGGCCAGGGAAACCCAG GGATTGGAGAGAGGAAA 55°C
SNP2 GACGCCAGCTGCCCGTTCA *CTGGCTGGCCGCACTAAGAG AATTGGAACAAAAGCAG 55°C
SNP3 rs2293942 GGCAAGGACCTCCAGTATCAG *CCCGAGCCATTTAACAG CCTCCAGTATCAGCGAGGAC 55°C
SNP4 rs2293943 *GGCAAGGACCTCCAGTATCAG CCCGAGCCATTTAACAG TGAAAAAGTCGTTTATTAGC 55°C
SNP5 rs4002827 GATATCATGGAAAATGCAGCG *GCTTCCCAATACAGCGAGG GCAGAAGAGAGTGAGTGTT 55°C
SNP6 GTTTCGAGAAACGTCCTCATTT *GCTTCTGGGGTCCTGACT CAGTCAGAGGCTGGTCA 55°C
SNP8 rs4430606 Acttcccgcgcttcgtta *CCAGCCCCTTCCTCTTTACT CCAGGTAGGTGCAGAAAG 52°C
SNP11 GTCGTGCGGAGCTGTCAAAGCGAG *CTGGAGCCGGGGATTT AGCTGTCAAAGCGAGCAGGG 55°C
InsCCG243 CACGACGTTGTAAAACGACGAGACACATCAAGATCTGGTTCCAA GGATAACAATTTCACACAGGGCAGCGGGCGGCACA
* Denotes additional of universal primer sequence to the primer. Universal primer sequence is as follows: TCTGCTGCTCCGGTTCATAGATT-3'
Table 2 IFP1 Variation and Association with Type 2 Diabetes
Name Variat dbSNP Posit Popul Frequency, Caucasian Frequency, Af. American
Cases Controls Cases Controls
SNP1 C/T ------ -3766 AA/Cauc 0.183 (0.143, 0.223) 0.192 (0.152, 0.232) 0.144 (0.118,0.170) 0.2101 (0.167,0.253)
SNP3 T/C rs2293942 -2890 Cauc 0.278 (0.232, 0.324) 0.249 (0.206, 0.292) ------ ------
SNP4 A/T rs2293943 -2877 AA/Cauc 0.339 (0.291, 0.387) 0.369 (0.321,0.417) 0.289 (0.255,0.323) 0.3702 (0.321,0.419)
SNP6 G/T ------ -1263 AA/Cauc 0.196 (0.156, 0.236) 0.189 (0.151, 0.229) 0.119 (0.091,0.147) 0.129 (0.093, 0.165)
SNP5 C/T rs4002827 -992 AA ------ ------ 0.071 (0.048, 0.094) 0.044 (0.021, 0.067)
SNP2 G/A ------ -279 AA/Cauc 0.458 (0.408, 0.508) 0.429 0.379, 0.479) 0.114 (0.086, 0.142) 0.139 (0.101, 0.177)
SNP11 (G)4/(G)3 ------ -108 AA/Cauc 0.182 (0.143, 0.221) 0.177 (0.138, 0.216) 0.134 (0.108, 0.160) 0.2173 (0.175, 0.259)
SNP8 G/T rs4430606 +918 AA/Cauc 0.197 (0.157, 0.237) 0.191 (0.152, 0.232) 0.115 (0.087, 0.143) 0.133 (0.096, 0.170)
InsCCG243 InsCCG ------ +4437 AA ----- ----- 0.091 (0.069, 0.113) 0.0604 (0.035, 0.085)
Name, name from Figure 1 and text; variant, major/minor allele except for InsCCG243, in which the insertion is the minor allele; dbSNP, catalog number in public database if available; position, location relative to ATG start; Populat, population in which variant was detected, AA is African American, Cauc is Caucasian; frequency, minor allele frequency. Frequencies are shown with 95% confidence intervals in parentheses. Significance by Fisher Exact test: 1p = 0.007;2p = 0.008; 3p = 0.0008; 4p = 0.088; simulated p values based on 10,000 replicates were 0.027, 0.029, 0.002, and 0.28, for SNP1, SNP4, SNP11, and InsCCG243, respectively. No other simulated p values approached significance.
Table 3 Raw Counts for Caucasian and African American Case Control Studies
SNP Name Caucasian African American
Major/Major Major/Minor Minor/Minor Major/Major Major/Minor Minor/Minor
DM Cnt DM CNT DM CNT DM CNT DM CNT DM CNT
SNP 1 120 124 54 59 6 7 249 117 84 60 7 9
SNP 2 54 61 97 95 38 34 197 124 53 35 2 3
SNP 3 97 73 15 108 71 12 --- --- --- --- --- ---
SNP 4 84 82 43 76 89 26 169 147 25 79 75 31
SNP 5 --- --- --- --- --- --- 210 143 34 14 1 0
SNP 6 124 125 56 58 7 9 200 119 53 43 2 3
SNP 8 123 123 56 58 9 7 198 115 50 41 2 3
SNP 11 125 8 51 125 53 6 253 114 76 57 7 11
ProIns --- --- --- --- --- --- 265 57 1 155 21 0
Numbers of individuals with each genotype are shown for the SNPs in Table 1. Significance by allelic association is shown in Table 1 with confidence intervals for allele frequencies. Note that allelic association was the primary test performed. Data not shown (--) was not typed in the full case control set because of low frequency. Counts differ slightly due to genotypes that were not called, and because additional African American samples were typed for SNPs 1, 4, 11, and ProIns (proline insertion) based on initial data showing an association.
Table 4 Haplotypes observed in the Caucasian Population
Haplotype Case Freq Control Freq p value
CTAGGIG 0.337 0.364 0.4406
CCTGAIG 0.269 0.247 0.4939
TTTTGDT 0.192 0.187 0.8625
CTTGAIG 0.184 0.186 0.936
Haplotypes observed at over 1% frequency are shown for SNPs observed in the Caucasian population (SNPs 1, 3, 4, 6, 2, 11, and 8). SNP 11 is shown as I (insertion, G4 or 4 G's) or D (deletion, G3 or 3 G's). All other SNPs are shown as listed in Table 2.
Table 5 Haplotypes for 8, 4, and 3 Markers Observed in the African American Population
Variants No Haplotype Case Freq Control Freq ChiSq P value Simulated p value Global p
1,4,6,5,2,11,8,P
1 CAGCGIGD 0.531 0.511 0.326 0.5683 1 0.01
2 TTTCGDTD 0.103 0.145 3.336 0.0678 0.465
3 CTGCAIGD 0.112 0.124 0.247 0.6193 1
4 CAGCGIGI 0.108 0.064 4.676 0.0306 0.239
5 CAGTGIGD 0.072 0.046 2.389 0.1222 0.693
6 TTGCGDGD 0.031 0.067 5.873 0.0154 0.131
7 CTGCGIGD 0.026 0.024 0.014 0.9057 1
1,4,11,P
1 CAID 0.622 0.565 3.197 0.0738 0.3053 0.01
2 TTDD 0.138 0.209 9.084 0.0026 0.0094
3 CTID 0.143 0.151 0.134 0.714 0.9995
4 CAII 0.088 0.059 2.735 0.098 0.3809
1,4,11
1 CAI 0.71 0.625 7.96 0.0048 0.02 0.01
2 TTD 0.138 0.209 9.085 0.0026 0.011
3 CTI 0.145 0.152 0.091 0.763 1
Haplotypes observed at over 1% frequency for all 8 variants, the 4 variants typed in additional individuals, and the three variants that showed an association with T2DM. Simulated p values are based on 10,000 simulations in HaploView 3.2; global p values are based on simulations conducted in Phase 2.1.1.
==== Refs
DeFronzo RA Pathogenesis of type 2 diabetes: metabolic and molecular implications for identifying diabetes genes Diabetes Reviews 1997 5 177 270
Kahn SE Clinical review 135: The importance of beta-cell failure in the development and progression of type 2 diabetes J Clin Endocrinol Metab 2001 86 4047 4058 11549624 10.1210/jc.86.9.4047
Weyer C Bogardus C Mott DM Pratley RE The natural history of insulin secretory dysfunction and insulin resistance in the pathogenesis of type 2 diabetes mellitus J Clin Invest 1999 104 787 794 10491414
Elbein SC Hasstedt SJ Wegner K Kahn SE Heritability of pancreatic beta-cell function among nondiabetic members of Caucasian familial type 2 diabetic kindreds J Clin Endocrinol Metab 1999 84 1398 1403 10199785 10.1210/jc.84.4.1398
Watanabe RM Valle T Hauser ER Ghosh S Eriksson J Kohtamaki K Ehnholm C Tuomilehto J Collins FS Bergman RN Boehnke M Familiarity of quantitative metabolic traits in Finnish families with non-insulin-dependent diabetes mellitus. Finland-United States Investigation of NIDDM Genetics (FUSION) Study investigators Hum Hered 1999 49 159 168 10364681 10.1159/000022865
Poulsen P Levin K Petersen I Christensen K Beck-Nielsen H Vaag A Heritability of insulin secretion, peripheral and hepatic insulin action, and intracellular glucose partitioning in young and old danish twins Diabetes 2005 54 275 283 15616039
Melloul D Transcription factors in islet development and physiology: role of PDX-1 in beta-cell function Ann N Y Acad Sci 2004 1014 28 37 15153417 10.1196/annals.1294.003
Stoffers DA Zinkin NT Stanojevic V Clarke WL Habener JF Pancreatic agenesis attributable to a single nucleotide deletion in the human IPF1 gene coding sequence Nat Genet 1997 15 106 110 8988180 10.1038/ng0197-106
Stoffers DA Ferrer J Clarke WL Habener JF Early-onset type-II diabetes mellitus (MODY4) linked to IPF1 Nat Genet 1997 17 138 139 9326926 10.1038/ng1097-138
Johnson JD Ahmed NT Luciani DS Han Z Tran H Fujita J Misler S Edlund H Polonsky KS Increased islet apoptosis in Pdx1+/- mice J Clin Invest 2003 111 1147 1160 12697734 10.1172/JCI200316537
McKinnon CM Docherty K Pancreatic duodenal homeobox-1, PDX-1, a major regulator of beta cell identity and function Diabetologia 2001 44 1203 1214 11692168 10.1007/s001250100628
Melloul D Marshak S Cerasi E Regulation of pdx-1 gene expression Diabetes 2002 51 S320 S325 12475770
Ben Shushan E Marshak S Shoshkes M Cerasi E Melloul D A pancreatic beta -cell-specific enhancer in the human PDX-1 gene is regulated by hepatocyte nuclear factor 3beta (HNF-3beta), HNF-1alpha, and SPs transcription factors J Biol Chem 2001 276 17533 17540 11278466 10.1074/jbc.M009088200
Marshak S Ben Shushan E Shoshkes M Havin L Cerasi E Melloul D Regulatory elements involved in human pdx-1 gene expression Diabetes 2001 50 S37 S38 11272196
Gerrish K Van Velkinburgh JC Stein R Conserved transcriptional regulatory domains of the pdx-1 gene Mol Endocrinol 2004 18 533 548 14701942 10.1210/me.2003-0371
Chevre JC Hani EH Stoffers DA Habener JF Froguel P Insulin promoter factor 1 gene is not a major cause of maturity- onset diabetes of the young in French Caucasians Diabetes 1998 47 843 844 9588460
Hani EH Stoffers DA Chevre JC Durand E Stanojevic V Dina C Habener JF Froguel P Defective mutations in the insulin promoter factor-1 (IPF-1) gene in late-onset type 2 diabetes mellitus J Clin Invest 1999 104 R41 R48 10545531
Frayling TM Evans JC Bulman MP Pearson E Allen L Owen K Bingham C Hannemann M Shepherd M Ellard S Hattersley AT beta-cell genes and diabetes: molecular and clinical characterization of mutations in transcription factors Diabetes 2001 50 S94 100 11272211
Macfarlane WM Frayling TM Ellard S Evans JC Allen LI Bulman MP Ayres S Shepherd M Clark P Millward A Demaine A Wilkin T Docherty K Hattersley AT Missense mutations in the insulin promoter factor-1 gene predispose to type 2 diabetes J Clin Invest 1999 104 R33 R39 10545530
Das SK Chu W Zhang Z Hasstedt SJ Elbein SC Calsquestrin 1 (CASQ1) gene polymorphisms under chromosome 1q21 linkage peak are associated with type 2 diabetes in Northern European Caucasians Diabetes 2004 53 3300 3306 15561963
Shevchenko YO Bale SJ Compton JG Mutation screening using automated bidirectional dideoxy fingerprinting BioTechniques 2000 28 134 138 10649784
Barrett JC Fry B Maller J Daly MJ Haploview: analysis and visualization of LD and haplotype maps Bioinformatics 2005 21 263 265 15297300 10.1093/bioinformatics/bth457
Stephens M Donnelly P A comparison of bayesian methods for haplotype reconstruction from population genotype data Am J Hum Genet 2003 73 1162 1169 14574645 10.1086/379378
Schneider S Roessli D Excoffier L Arlequin: A software for population genetics data analysis. [2.000] 2000 Genetics and Biometry Lab, Dept. of Anthropology, University of Geneva. Ref Type: Computer Program
Carlson CS Eberle MA Rieder MJ Yi Q Kruglyak L Nickerson DA Selecting a maximally informative set of single-nucleotide polymorphisms for association analyses using linkage disequilibrium Am J Hum Genet 2004 74 106 120 14681826 10.1086/381000
Heinemeyer T Wingender E Reuter I Hermjakob H Kel AE Kel OV Ignatieva EV Ananko EA Podkolodnaya OA Kolpakov FA Podkolodny NL Kolchanov NA Databases on transcriptional regulation: TRANSFAC, TRRD and COMPEL Nucleic Acids Res 1998 26 362 367 9399875 10.1093/nar/26.1.362
Morimoto RI Kroeger PE Cotto JJ The transcriptional regulation of heat shock genes: a plethora of heat shock factors and regulatory conditions EXS 1996 77 139 163 8856973
Hansen L Urioste S Petersen HV Jensen JN Eiberg H Barbetti F Serup P Hansen T Pedersen O Missense mutations in the human insulin promoter factor-1 gene and their relation to maturity-onset diabetes of the young and late-onset type 2 diabetes mellitus in caucasians J Clin Endocrinol Metab 2000 85 1323 1326 10720084 10.1210/jc.85.3.1323
Elbein SC Karim MA Does the Aspartic Acid to Asparagine Substitution at Position 76 in the Pancreas Duodenum Homeobox Gene (PDX1) Cause Late-Onset Type 2 Diabetes? Diabetes Care 2004 27 1968 1973 15277425
Yamada K Yuan X Ishiyama S Ichikawa F Kohno S Shoji S Hayashi H Nonaka K Identification of a single nucleotide insertion polymorphism in the upstream region of the insulin promoter factor-1 gene: an association study with diabetes mellitus Diabetologia 1998 41 603 605 9628281 10.1007/s001250050954
Altshuler D Hirschhorn JN Klannemark M Lindgren CM Vohl MC Nemesh J Lane CR Schaffner SF Bolk S Brewer C Tuomi T Gaudet D Hudson TJ Daly M Groop L Lander ES The common PPARgamma Pro12Ala polymorphism is associated with decreased risk of type 2 diabetes Nat Genet 2000 26 76 80 10973253 10.1038/79839
Lelievre E Lionneton F Soncin F Vandenbunder B The Ets family contains transcriptional activators and repressors involved in angiogenesis Int J Biochem Cell Biol 2001 33 391 407 11312108 10.1016/S1357-2725(01)00025-5
Lu M Seufert J Habener JF Pancreatic beta-cell-specific repression of insulin gene transcription by CCAAT/enhancer-binding protein beta. Inhibitory interactions with basic helix-loop-helix transcription factor E47 J Biol Chem 1997 272 28349 28359 9353292 10.1074/jbc.272.45.28349
Reiner AP Ziv E Lind DL Nievergelt CM Schork NJ Cummings SR Phong A Burchard EG Harris TB Psaty BM Kwok PY Population structure, admixture, and aging-related phenotypes in African American adults: the Cardiovascular Health Study Am J Hum Genet 2005 76 463 477 15660291 10.1086/428654
Tang H Quertermous T Rodriguez B Kardia SL Zhu X Brown A Pankow JS Province MA Hunt SC Boerwinkle E Schork NJ Risch NJ Genetic structure, self-identified race/ethnicity, and confounding in case-control association studies Am J Hum Genet 2005 76 268 275 15625622 10.1086/427888
Florez JC Sjogren M Burtt N Orho-Melander M Schayer S Sun M Almgren P Lindblad U Tuomi T Gaudet D Hudson TJ Daly MJ Ardlie KG Hirschhorn JN Altshuler D Groop L Association testing in 9,000 people fails to confirm the association of the insulin receptor substrate-1 G972R polymorphism with type 2 diabetes Diabetes 2004 53 3313 3318 15561965
Florez JC Agapakis CM Burtt NP Sun M Almgren P Rastam L Tuomi T Gaudet D Hudson TJ Daly MJ Ardlie KG Hirschhorn JN Groop L Altshuler D Association testing of the protein tyrosine phosphatase 1B gene (PTPN1) with type 2 diabetes in 7,883 people Diabetes 2005 54 1884 1891 15919813
Florez JC Burtt N de Bakker PI Almgren P Tuomi T Holmkvist J Gaudet D Hudson TJ Schaffner SF Daly MJ Hirschhorn JN Groop L Altshuler D Haplotype structure and genotype-phenotype correlations of the sulfonylurea receptor and the islet ATP-sensitive potassium channel gene region Diabetes 2004 53 1360 1368 15111507
|
16229747
|
PMC1274317
|
CC BY
|
2021-01-04 16:03:33
|
no
|
BMC Med Genet. 2005 Oct 17; 6:37
|
utf-8
|
BMC Med Genet
| 2,005 |
10.1186/1471-2350-6-37
|
oa_comm
|
==== Front
BMC MedBMC Medicine1741-7015BioMed Central London 1741-7015-3-151622974210.1186/1741-7015-3-15Research ArticleChlorpromazine for schizophrenia: a Cochrane systematic review of 50 years of randomised controlled trials Adams Clive Elliott [email protected] John [email protected] Ben [email protected] Mike [email protected] Jo [email protected] Kristian [email protected] A George [email protected] Cochrane Schizophrenia Group, Academic Department of Psychiatry and Behavioural Sciences, University of Leeds, 15 Hyde Terrace, Leeds, LS2 9LT, UK2 UK Cochrane Centre, Summertown Pavilion, Middle Way, Summertown, Oxford, OX2 7LG, UK3 Safer Custody Group, HM Prison Service, Abell House, John Islip Street, London, SW1P 4LH, UK4 STAKES/Vasa Central Hospital, Department of Psychiatry, FIN-65130 Vaasa, Finland5 University of Toronto, Humber River Regional Hospital, Keele Street Site, 2175 Keele Street, Toronto, Ontario, M6M 3Z4, Canada2005 17 10 2005 3 15 15 16 5 2005 17 10 2005 Copyright © 2005 Adams et al; licensee BioMed Central Ltd.2005Adams et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Chlorpromazine (CPZ) remains one of the most common drugs used for people with schizophrenia worldwide, and a benchmark against which other treatments can be evaluated. Quantitative reviews are rare; this one evaluates the effects of chlorpromazine in the treatment of schizophrenia in comparison with placebo.
Methods
We sought all relevant randomised controlled trials (RCT) comparing chlorpromazine to placebo by electronic and reference searching, and by contacting trial authors and the pharmaceutical industry. Data were extracted from selected trials and, where possible, synthesised and random effects relative risk (RR), the number needed to treat (NNT) and their 95% confidence intervals (CI) calculated.
Results
Fifty RCTs from 1955–2000 were included with 5276 people randomised to CPZ or placebo. They constitute 2008 person-years spent in trials. Meta-analysis of these trials showed that chlorpromazine promotes a global improvement (n = 1121, 13 RCTs, RR 0.76 CI 0.7 to 0.9, NNT 7 CI 5 to 10), although a considerable placebo response is also seen. People allocated to chlorpromazine tended not to leave trials early in both the short (n = 945, 16 RCTs, RR 0.74 CI 0.5 to 1.1) and medium term (n = 1861, 25 RCTs, RR 0.79 CI 0.6 to 1.1). There were, however, many adverse effects. Chlorpromazine is sedating (n = 1242, 18 RCTs, RR 2.3 CI 1.7 to 3.1, NNH 6 CI 5 to 8), increases a person's chances of experiencing acute movement disorders, Parkinsonism and causes low blood pressure with dizziness and dry mouth.
Conclusion
It is understandable why the World Health Organization (WHO) have endorsed and included chlorpromazine in their list of essential drugs for use in schizophrenia. Low- and middle-income countries may have more complete evidence upon which to base their practice compared with richer nations using recent innovations.
==== Body
Background
Chlorpromazine is in the World Health Organization (WHO) list of essential drugs [1]. It is estimated that 24 million people currently suffer from schizophrenia [2], the majority of whom live in low or middle-income countries. Until recently, it would have been common practice for anyone with schizophrenia to have been treated with chlorpromazine at some point [3,4]. Despite well-documented adverse effects, and the advent of a new generation of antipsychotic drugs, chlorpromazine remains one of the most commonly used and inexpensive treatments for people with schizophrenia [5]. In Africa, chlorpromazine was widely used [6], although we have failed to identify any more recent surveys. In India chlorpromazine is commonly prescribed, and in South East Asia the older generation of antipsychotics are used to treat the majority of people with schizophrenia [7]. In 2003, in the UK, chlorpromazine was the most frequently prescribed of the first generation 'typical' antipsychotic drugs, where, at that time, the 'typical' group of antipsychotics accounted for 44% of all anti-psychotic prescriptions [8].
As well as its almost universal use in clinical practice, chlorpromazine is a benchmark by which other treatments are evaluated [9]. There are many qualitative reviews of chlorpromazine but few attempts have been made to quantify data from randomised controlled trials (RCTs) [10,9,11]. An up-to-date quantitative review of the effects of this old, highly prevalent treatment is long overdue. 50 years after its formulation, the evidence should be more complete than when the drug was under patent.
Methods
Inclusion criteria
The inclusion criteria were defined and disseminated for peer review within a Cochrane protocol first published 1998 [12]. Articles were included if they reported RCTs where the participants had schizophrenia or non-affective serious/chronic mental illness, and where the interventions included chlorpromazine (any dose or mode of administration) versus placebo or no treatment.
Identification of relevant trials
We identified relevant randomised trials by searching the Cochrane Schizophrenia Group's register of trials (June 2002), with a phrase designed to identify the many ways of naming chlorpromazine [see Additional file 1]. Citations in all identified articles were inspected for further trials. Rhône-Poulenc Rorer (the original distributors of chlorpromazine) was contacted to request access to archive material, and Dr RA Pargiter (Hobart, Tasmania) donated a large series of May and Baker chlorpromazine reports from 1955 to 1973.
Data extraction and study appraisal
All electronic records identified were independently inspected by BT, CA and JR. The reliability of selection processes and data extraction was checked using a 10% random sample. Full reports of studies of agreed relevance were obtained, quality rated [13], and data relating to methods, participants, interventions and outcomes, extracted. Any disagreement was discussed and decisions documented. If there were outstanding issues, the authors of the studies were contacted where possible to help resolve problems.
Statistical methods
Dichotomous and continuous data were not used if over half of those randomised did not contribute to the outcome due to early attrition from the study or non-compliance. Dichotomous data were combined using a random effects Relative Risk (RR) [14]. Numbers needed to treat/harm (NNT/H) [10] were also calculated, and χ2 tests for heterogeneity were performed. Where <50% of people were lost to follow-up at the end of a trial, 'worst case' intention-to-treat analyses were undertaken by assuming that those who had left a trial early had had a poor outcome. The sensitivity of the final results to this assumption was tested. Continuous data were excluded if derived from scales of unknown validity and if totals or measures of variance were not reported. Summation was not attempted where continuous data were too skewed [15]. All estimates of effect are presented with their 95% confidence intervals (CI).
Results
Electronic searches identified over 1000 records, most of which were ineligible. Full copies of 351 citations were obtained for detailed scrutiny, including a further 50 papers identified from citations. Of these, 302 papers were excluded and 99 reports of the 50 RCTs included (Table 1). Studies were mainly excluded due to lack of random allocation (68%). However, 43 randomised trials (30%) reported irrelevant outcomes, such as serum levels of chlorpromazine breakdown products, or presented data in such a way as to make the outcomes unintelligible or impossible to use.
Table 1 Included studies.
METHODS PARTICIPANTS INTERVENTIONS OUTCOMES
INCLUDED STUDIES (date of publication) Randomised Double-blind Three+ arm study Duration (weeks) Only Schizophrenia History Total number of participants Age (years) Sex CPZ dose (mg/day) Number allocated CPZ Number allocated placebo Leaving the study early Global improvement Mental State Side-effects Global clinical state Behaviour Relapse
1955 Hall ● ● 9 ● C 175 20–59 M+F 750 max 87 88 ● ● ● ●
Vaughan ● ● U/K C 48 M = 43 F 75–450 24 24 ●
1956 Shepherd ● U/K ● 6 ● C 24 27–52 F 300 8 8 ● ●
1958 Abrams ● 4 ● C 40 20–55 F 200–600 20 20 ●
Grygier ● U/K 24 ● C 30 m = 50 F 150 15 15 ●
Hine ● ● 20 ● C 22 30–50 F 750 max 11 11 ● ● ●
Simon ● ● 4 U/K 80 m = 31 U/K 200–1200 20 0 ●
1959 Baker ● ● ● 5 C 25 33–79 F 150–300 7 7 ● ● ● ●
Flemming ● ● ● 26 ● C 63 m = 58 F 75–300 21 21 ● ● ●
Walsh ● U/K ● 8 ● C 66 27–50 F 75–300 22 22 ● ●
1960 Englhardt ● ● ● 78 ● U/K 173 18–40 U/K 50–800 62 56 ● ●
Hamilton ● ● ● 8 ● C 54 m = 38 M 300 18 18 ● ●
Payne ● ● ● 6 ● C 21 23–73 M 25–100 7 7 ●
Somerville ● ● ● 6 C+A 60 24–58 F 200–800 15 30 ● ● ● ●
1961 Clark ● ● ● 24 ● C 60 26–52 F 200–800 20 20 ● ●
Lorr ● ● ● 12 A 308 <50 M 50–100 63 61 ●
Kurland ● ● ● 6 A 277 18–61 M+F 300 33 72 ● ● ●
Schiele ● ● ● 16 ● C 80 m = 80 M 200–1000 20 20 ● ● ● ●
Smith ● ● ● 14 ● C 30 m = 42 M+F 150–600 13 15 ●
1963 Bishop ● ● ● 10 ● C 30 U/N M+F 800 10 10 ● ●
Fink ● U/K ● 6 S 311 m = 31 M+F 1200 51 44 ●
1964 NIMH ● ● ● 6 ● A 463 16–45 M+F 200–1600 112 125 ●
1966 Reardon ● ● ● 4 ● A 34 U/K M+F 300–600 11 12 ● ●
Saretsky ● ● 12 ● A 40 <55 M 400 20 20 ●
1967 Clark ● ● 10 ● C 72 25–55 F 678 m 51 21 ●
Letemendia ● ● 39 C 28 <65 M 300 14 14 ●
1968 Clark a ● ● ● 14 ● C 72 20–60 F 1000 max 18 36 ● ● ●
Clark b ● ● ● 16 ● C 69 20–60 F 1000 max 23 23 ● ●
Cohen ● ● ● 60 ● C 126 18–42 M+F 180 42 42 ●
Prien ● ● ● 24 ● C 838 19–55 M+F 2000 208 212 ● ● ● ● ●
300 208
1969 Tetreault ● ● ● 12 ● C 45 m = 50 F 300–600 15 15 ● ● ● ●
1970 Clark a ● ● ● 12 ● C 44 22–55 M+F 200–1000 15 14 ● ● ● ● ●
Clark b ● ● 24 ● C 71 21–60 F 150–600 54 18 ● ● ●
1971 Clark ● ● ● 4 ● C 86 21–45 M+F 200–1000 23 21 ● ● ●
1972 Clark ● ● ● 12 ● C 55 21–60 M+F 1000 19 18 ● ● ● ●
Serafetinedes ● ● ● 12 ● C 57 21–61 M+F 1000 max 14 13 ● ● ●
1973 Hogarty ● ● 156 ● S 374 18–53 M+F 270 m 192 182 ● ●
Klein ● ● 6 ● 88 17–61 M+F 300–1200 46 42 ●
1974 Reschke ● ● ● 0.1 ● A 50 19–57 M+F 25 im 10 11 ● ● ●
1975 Ban ● ● ● 12 ● C+A 30 17–46 M+F 200–800 10 10 ● ●
Hamill ● 0.7 ● A 44 18–55 M+F 306–475 22 22 ●
1977 Clark ● ● ● 12 ● C 27 23–61 M+F 1000 9 9 ● ● ●
Spohn ● ● 6+ ● U/K 40 18–55 M+F 200 min 20 20 ●
1978 Rappaport ● U/K ● A 127 16–40 M 300–900 53 74 ● ●
1981 Peet ● ● ● 12 ● U/K 53 m = 51 M+F 400 max 16 18 ● ● ●
1982 Nishikawa ● ● ● 156 ● S 55 m = 33 M+F 75 10 10 ●
1986 Zuoze ● ● 4 ● C 60 m = 36 U/K 450 m 20 20 ●
1990 Chouinard ● ● ● 4 ● A 62 19–62 M+F 300–1200 21 21 ● ● ● ●
1991 Borison ● ● ● 4 ● A 30 22–58 M 400–1600 9 10 ● ●
2000 Cooper ● ● ● 8 ● C 159 18–42 M+F 600 53 53 ● ● ●
Key: ● = Yes; U/K = Unknown; C = Chronic; A = Acute; M = Male; F = Female; m = Mean; im = Intramuscular injection.
Study quality
All 50 included studies reported the use of random allocation; only 4 were explicit about the process used. Two used the toss of a coin [16,17], and 2 used random number tables [18,19]. Citations to all included and excluded studies are available in the full Cochrane Review [12], otherwise the names and dates cited in this text relate to Table 1. A further 2 trials [20,21] described some form of allocation concealment (sealed envelopes in both cases). The other 44 studies gave little assurance that bias was minimised during the allocation procedure and this may mean that this review overestimates the effect of chlorpromazine [22]. Twenty-eight (56%) of the trials adequately described their attempts to be double-blind, with two [20] and [23] reporting how successful these attempts were. Two studies [18] and [24] gave no indication that blinding had been attempted. Other trials indicated that an attempt at blinding had been made, but they gave no description of how this had been done. The description of participants who left studies early was poor; 12 of the 50 included studies providing no details of treatment withdrawals. Presentation of data was also poor. Trials frequently presented both dichotomous and continuous data in graphs, or reported inexact statistical measures of probability, for example p > 0.05. This often made it impossible to extract raw data for synthesis. Continuous scale data were frequently collected in the trials, but were often poorly reported; 30/50 trials did not report standard deviations and 9/42 did not present any data from the scales they had used.
Study designs
The studies were mostly either 6 or 12 weeks long, but the range was large (24 h to 3 years). The great majority of participants in nearly all of the trials were diagnosed as suffering from schizophrenia. These studies reported on >5276 people, 3318 of whom were allocated to chlorpromazine-placebo comparison. Eleven of the 50 trials described the diagnostic criteria used, or the symptoms required for people to be included. Otherwise entry to most of the included studies was based on a pragmatic diagnosis of schizophrenia. The trials ranged in size from 21 [25] to 838 participants [26]. Most people were hospitalised at the time of the study. The lowest dose of chlorpromazine tested was 25 mg/day [27] and the highest 2000 mg/day [26]. One trial [28] included both a placebo and a no-drug group, which we combined. Another study included both a placebo group and a "routine conventional hospital treatment" group [26]. Data from the latter were not used in this review, as people in this group will probably have been given antipsychotic drugs.
Outcomes
Table 2 presents the main results of this review. These intention-to-treat data are derived by synthesising homogeneous trial findings. The results remain essentially unchanged when we only used data from participants who completed the studies. The data show no clear pattern indicative of publication bias when sorted by study size and effect [29].
Table 2 Results relating to clinical change and study attrition.
Months Number of trials Chlorpromazine Placebo RR (95% CI) Test for heterogeneity
events/total participants
Relapse 6–24 3 108/202 159/192 0.65 (0.5–9.0) Chi2 7.83, df 2, p = 0.02 I2 = 74.5%*
No global improvement 2–6 13 470/654 406/467 0.76 (0.7–0.9) Chi2 25.4, df 12, p = 0.01 I2 = 52.8%
Leaving the study early 6–24 2 38/254 33/238 1.09 (0.7–1.6) Chi2 0.47, df 1, p = 0.49 I2 = 0%
* no clear cause of heterogeneity found on close re-inspection of trials
Data on global improvement (a dichotomised impression of change), in the period up to 6 months favours chlorpromazine (n = 1121, 13 RCTs, RR No global improvement 0.76 CI 0.7 to 0.9) but is moderately heterogeneous (I2 = 52.8%). Global severity of illness at study end (a dichotomised impression of clinical state) also favours chlorpromazine (n = 778, 5 RCTs, RR severely ill 0.67 CI 0.5 to 0.8, NNT 4 CI 3 to 10; Figure 1). Very few studies present usable data directly relating to end point mental state. The continuous data that are available (Brief Psychiatric Rating Scale [30] are equivocal (n = 49, 2 RCTs, RR -4.82 CI -8.5 to 1.2). Most information on behaviour relates to a dichotomous outcome of 'behaviour deteriorated/disturbed/uncooperative' (n = 1127, 10 RCTs, RR 0.53 CI 0.3 to 0.9) but these data are heterogeneous (χ2 73, df 9, p < 0.00001).
Figure 1 Chlorpromazine versus placebo – global outcomes.
Chlorpromazine has many adverse effects (Table 3). It is a sedative (n = 1242, 18 RCTs, RR 2.3 CI 1.7 to 3.1, NNH 6 CI 5 to 8) that may cause weight gain (n = 165, 5 RCTs, RR 4.44 CI 2.1 to 9.3, NNH 3 CI 2 to 5). Extrapyramidal symptoms are common and include acute dystonias (n = 780, 4 RCTs, RR 3.1 CI 1.3 to 7.7, NNH 24 CI 15–57) and Parkinsonism (n = 1265, 12 RCTs, RR 2.6 CI 1.2 to 5.4, NNH 10 CI 8 to 16). Data on chronic movement disorders such as tardive dyskinesia, however, are not available from this review as this requires longer follow-up than was attempted for nearly all the trials. Occurrence of akathisia is similar in the chlorpromazine and placebo groups. For every 7 people given chlorpromazine, one will experience some form of photosensitive reaction (n = 799, 6 RCTs, RR 5.19 CI 3 to 10, NNH 7 CI 6 to 10); hypotension and dizziness are common (n = 1232, 15 RCTs, RR 1.9 CI 1.4 to 27, NNH 12 CI 8 to 22); and dry mouth is considerably increased (n = 756, 5 RCTs, RR 4.00 CI 1.6 to 10, NNH 18 CI 13 to 37). Eye opacities, as identified by slit-lamp examination within one large trial using high dose chlorpromazine 2 gms/day [26], were increased within the drug group (n = 657, RR 3.09 CI 1.9 to 5.1, NNH 7 CI 5 to 10). There were no significant differences between people given placebo and those allocated chlorpromazine in the frequency of complaints of constipation, urinary retention and blurred vision.
Table 3 Adverse effects.
No. of trials Chlorpromazine Placebo RR (95% CI) Random
events/total participants
General symptoms
Sedation 18 224/725 68/517 2.30 (1.7–3.1)
Weight gain > 10 lb; 4.5 Kg 5 31/75 7/90 4.44 (2.1–9.3)
Extrapyramidal symptoms
Acute dystonia 4 28/472 5/306 3.10 (1.3–7.7)
Parkinsonism 12 123/723 40/542 2.60 (1.2–5.4)
Fits 3 19/450 4/245 2.41 (0.4–16.4)
Akathisia 8 53/602 40/400 0.95 (0.5–1.9)
Allergic-type symptoms
Agranulocytosis/leucopenia 7 10/207 2/187 2.02 (0.7–5.6)
Rashes/itching 11 42/658 21/475 1.43 (0.9–2.4)
Jaundice 3 8/116 1/115 4.04 (0.9–17.9)
Photosensitivity 6 81/496 9/303 5.19 (2.7–9.8)
Eye opacity 2 97/431 16/226 3.09 (1.9–5.1)
Anti-cholinergic/nor-adrenergic symptoms
Hypotension + dizziness 15 113/708 38/524 1.90 (1.4–2.7)
Constipation 9 40/590 16/365 1.68 (0.9–2.9)
Urinary retention 3 11/459 5/253 1.49 (0.5–4.3)
Dry mouth 5 32/473 4/283 4.00 (1.6–9.8)
Blurred vision 6 10/529 9/381 1.10 (0.5–2.9)
There were no reports of deaths occurring during any of the studies. Any data relating to violent incidents, hospital discharge or admissions, presence of delusions or hallucinations were either absent or impossible to use. Not one of the studies, even in recent years, reported levels of satisfaction and quality of life, nor could we identify any direct economic evaluation of chlorpromazine.
Discussion
These 50 studies amounted to a total of >2000 person-years of exposure to chlorpromazine or placebo. For people with this serious mental illness, and certainly in situations were resources are limited, chlorpromazine remains a first line treatment. The medium term data on improvement suggest that about 7 people have to be treated for one to have what the trialists would describe as 'global improvement' (n = 1121, 13 RCTs, RR 0.76 CI 0.7 to 0.9, NNT 7 CI 5 to 10). This outcome relates to a simple dichotomised impression of a person's mental state, behaviour and functioning. Given the limited quality of reporting and the fact that we may be only able to pool data from a subset of the included trials, even this finding may be an over estimate of the positive and an underestimate of the negative effects of giving chlorpromazine.
The increased likelihood that people given chlorpromazine continued in their trial may be heartening. It could indicate a genuine decrease in the distressing symptoms of schizophrenia that led to an increased compliance with medication, despite common and unpleasant adverse effects such as sedation and hypotension. Doctors and nurses may, at times of acute disturbance, welcome this sedative effect but people with schizophrenia may not.
Despite limitations, this review provides quantitative evidence to confirm many of the impressions held by clinicians and recipients of care about the effects of chlorpromazine. Chlorpromazine is a sedating drug, prone to cause movement problems. Reliable evidence about its short-term effects is surprisingly weak, but information from studies that are >6 months does suggest that chlorpromazine facilitates a global improvement and may decrease the likelihood of behaving in a disturbed manner, at least within the confines of hospital.
Conclusion
Chlorpromazine represents a low-cost choice for clinicians world-wide and merits its position as a benchmark treatment for psychotic symptoms. Until large, high quality, clinically relevant trials show equally inexpensive treatments to be both more effective and safe, chlorpromazine is likely to continue to be one of the most widely used treatments for the millions of people who suffer with schizophrenia.
Although the NNTs may seem high, and the NNHs low, these estimates are likely to be more realistic than those for new drugs, for which all evidence has not been made available. As time passes, studies not seen in the early years of marketing tend to become apparent. These studies may be systematically different from those initially used to sell the drug. As a result, clinical practice of low and middle-income countries, often having to use older generations of drug, may, nevertheless, have more chance of being based on all evidence than that of high income nations. The latter are prone to purchase new expensive innovations the evidence for which is treated with sensitivity by researchers, marketers, and licensing agencies mindful of pecuniary influences.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BT participated in the protocol development, trial searching, data extraction, analysis, data interpretation and writing the manuscript.
JR participated in the trial searching, data extraction, analysis, data interpretation, writing the manuscript and maintaining the review.
CEA participated in the protocol development, trial searching, data extraction, analysis, data interpretation, writing the final manuscript and maintaining the review.
GA participated in the protocol development and data interpretation.
MC participated in the calculation and understanding of the results and production of the final manuscript.
JB participated in the calculation and understanding of results, and production of the final manuscript.
KW participated in the protocol development, calculation and understanding of the results and writing the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Search strategy for identification of studies. The Schizophrenia Group's register is based on regular searches of BIOSIS Inside; CENTRAL; CINAHL; EMBASE; MEDLINE and PsycINFO; the hand searching of relevant journals and conference proceedings, and searches of several key grey literature sources. A full description is given in the Group's module on the Cochrane Library.
Click here for file
Acknowledgements
Thanks to Dr Pargiter (Hobart, Tasmania), Doreen Ledgard, Marie Montague and Sarah Old (Warneford Hospital library) and Ms Karen Butter (Medical Information Pharmacist, Rhone-Poulenc Rorer, UK) for helping track down useful data for this study. Gill Rizzello for her comments on the final manuscript.
==== Refs
WHO Essential Medicines 13th edition. WHO Model List 2003
WHO Global Health Statistics Online, 1998
Wysowski DK Baum C Antipsychotic drug use in the United States, 1976–1985 Arch Gen Psychiatry 1989 46 929 932 2572207
Burgess S Emergency drug treatment of disturbed psychotic patients Auditorium 1997 6 28 30
Norquist G Hyman SE Advances in understanding and treating mental illness: implications for policy Health Aff 1999 18 32 47 10.1377/hlthaff.18.5.32
Odejide AO Ban TA Psychotropic drug prescription pattern in a developing country (Nigeria). The need for an essential psychotherapeutic drug list Int Pharmacopsychiatry 1982 17 163 169 6128330
Chong MY Tan CH Fujii S S Yang Ungvari GS Si T Chung EK Sim K Tsang HY Shinfuku N Antipsychotic drug prescription for schizophrenia in East Asia: rationale for change Psychiatry Clin Neurosci 2004 58 61 67 14678459 10.1111/j.1440-1819.2004.01194.x
Prescription Pricing Authority: NHS 2003
Adams CE Bartsch S Fenton M Publishing patterns of the four drugs most commonly reported in schizophrenia trials Schizophr Res 2004 67 268
Cook RJ Sackett DL The number needed to treat: a clinically useful measure of treatment effect BMJ 1995 310 452 454 7873954
Cole JO Davis JM Antipsychotic drugs The Schizophrenic Syndrome 1969 New York: Grune & Stratton 478 568
Thornley B Rathbone J Adams CE Awad G Chlorpromazine versus placebo for schizophrenia Cochrane Database Systematic Review 2004 CD000284
Clarke M Oxman AD The Cochrane Reviewers Handbook 4.2.0 The Cochrane Library 2003 2
DerSimonian R Laird N Meta-analysis in clinical trials Control Clin Trials 1986 7 177 188 3802833 10.1016/0197-2456(86)90046-2
Altman DG Bland JM Detecting skewness from summary information BMJ 1996 313 1200 8916759
Hine FR Chlorpromazine in schizophrenic withdrawal and in the withdrawn schizophrenic J Nerv Ment Dis 1958 127 220 227 13588376
Letemendia F Harris AD Chlorpromazine and the untreated chronic schizophrenic: a long-term trial Br J Psychiatry 1967 113 950 958 4862704
Hamill WT Fontana AF The immediate effects of chlorpromazine in newly admitted schizophrenic patients Am J Psychiatry 1975 132 1023 1026 1101707
Tetreault L Bordeleau JM Gauthier R Vulpe M Lapointe L Comparative study of TPS-23, chlorpromazine and placebo in chronic schizophrenic patients Dis Nerv Syst 1969 30 74 84 4886994
Hall RA Dunlap DJ A study of chlorpromazine: methodology and results with chronic semi-disturbed schizophrenics J Nerv Ment Dis 1955 122 301 314 13307264
Cole JO Goldberg SC Klerman GL National Institute of Mental Health Psychopharmacology Service Center Collaborative Study Group Phenothiazine treatment in acute schizophrenia Arch Gen Psychiatry 1964 10 246 261 14089354
Schulz KF Chalmers I Hayes RJ Altman DG Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials JAMA 1995 273 408 412 7823387 10.1001/jama.273.5.408
Grygier P Waters MA Chlorpromazine used with an intensive occupational therapy program Arch Neuro Psychiatr 1958 79 697 705
Clark M Dubowski K Colmore J The effect of chlorpromazine on serum cholesterol in chronic schizophrenic patients Clin Pharmacol Ther 1970 11 883 889 5481574
Payne P A comparison of trifluopromazine, chlorpromazine and a placebo in twenty-one chronic schizophrenic patients Manit Med Rev 1960 196 198
Prien R Cole JO High dose chlorpromazine therapy in chronic schizophrenia Arch Gen Psychiatry 1968 18 482 495 4868797
Reschke RW Parenteral haloperidol for rapid control of severe, disruptive symptoms of acute schizophrenia Dis Nerv Syst 1974 35 112 115
Clark ML Huber WK Kyriakopoulos AA Ray TS Colmore JP Ramsey HR Evaluation of trifluperidol in chronic schizophrenia Psychopharmacologia 1968 12 193 203 10.1007/BF00403773
Egger M Davey Smith G Schneider M Minder C Bias in meta-analysis detected by a simple, graphical test BMJ 1997 315 629 634 9310563
Overall JE Gorham DR The Brief Psychiatric Rating Scale Psychol Rep 1962 10 799 812
|
16229742
|
PMC1274318
|
CC BY
|
2021-01-04 16:27:56
|
no
|
BMC Med. 2005 Oct 17; 3:15
|
utf-8
|
BMC Med
| 2,005 |
10.1186/1741-7015-3-15
|
oa_comm
|
==== Front
BMC Med EthicsBMC Medical Ethics1472-6939BioMed Central London 1472-6939-6-101624201410.1186/1472-6939-6-10Research ArticleStatus of national research bioethics committees in the WHO African region Kirigia Joses M [email protected] Charles [email protected] Amido [email protected] World Health Organization, Regional Office for Africa, Brazzaville, Congo2 International Biomedical Research in Africa, Abuja, Nigeria2005 20 10 2005 6 10 10 6 6 2005 20 10 2005 Copyright © 2005 Kirigia et al; licensee BioMed Central Ltd.2005Kirigia et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The Regional Committee for Africa of the World Health Organization (WHO) in 2001 expressed concern that some health-related studies undertaken in the Region were not subjected to any form of ethics review. In 2003, the study reported in this paper was conducted to determine which Member country did not have a national research ethics committee (REC) with a view to guiding the WHO Regional Office in developing practical strategies for supporting those countries.
Methods
This is a descriptive study. The questionnaire was prepared and sent by diplomatic pouch to all the 46 Member States in the WHO African Region, through the WHO country representatives, for facilitation and follow up. The data were entered in Excel spreadsheet and subsequently exported to STATA for analysis. A Chi-Squared test (χ2) for independence was undertaken to test the relationship between presence/absence of Research Ethics Committee (REC) and selected individual socioeconomic and health variables.
Results
The main findings were as follows: the response rate was 61% (28/46); 64% (18/28) confirmed the existence of RECs; 36% (10/28) of the respondent countries did not have a REC (although 80% of them reported that they had in place an ad hoc ethical review mechanism); 85% (22/26) of the countries that responded to this question indicated that ethical approval of research proposals was, in principle, required; and although 59% of the countries that had a REC expected it to meet every month, only 44% of them reported that the REC actually met on a monthly basis. In the Chi-Squared test, only the average population in the group of countries with a REC was statistically different (at 5% level of significance) from that of the group of countries without a REC.
Conclusion
In the current era of globalized biomedical research, good ethics stewardship demands that every country, irrespective of its level of economic development, should have in place a functional research ethics review system in order to protect the dignity, integrity and safety of its citizens who participate in research.
==== Body
Background
"Regrettably, (over) 50 years after the Nuremberg trials and the Nuremberg code, unethical (bio)medical research on humans continues, even in highly privileged countries" [1]
Biomedical research involves research on pharmaceuticals, medical devices, medical radiation and imaging, surgical procedures, medical records and biomedical samples, as well as epidemiological, social, and psychological investigations [2]. It often entails collection, analysis, and interpretation of information obtained from human beings. Research must be undertaken in an ethical manner so that it assures protection of the dignity, integrity, and safety of all actual or potential research participants [3].
A recent revelation that about 25% of health-related studies in developing countries were not subjected to some form of ethics review by an international review board, national ethics board, or ministry/department of health is worrisome [4]. A similar concern was expressed by the Regional Committee for Africa of the World Health Organization (WHO) in 2001.
Since the end of World War II, ethical and scientific standards for conducting biomedical research on human subjects have been enshrined in international guidelines, including the Nuremberg Code [5], the Declaration of Geneva [6], the International Covenant on Civil and Political Rights [7], the International Code of Medical Ethics of the World Medical Association [8], the Declaration of Helsinki [9], the CIOMS International Ethical Guidelines for Biomedical Research Involving Human Subjects [3], the WHO Operational Guidelines for Ethics Committees that Review Biomedical Research [2] and the ICH Guidelines for Good Clinical Practice [10].
Although there is variance in the scope and emphasis of the above-mentioned international instruments on the ethics of medical research, they all require ethical justification and scientific validity of research; ethical review; informed consent; vulnerability of individuals, groups, communities and populations; women as research subjects; equity regarding burdens and benefits; choice of control in clinical trials; confidentiality; compensation for injury; strengthening of national or local capacity for ethical review; and obligations of sponsors to provide health care [3].
The growing volume of collaborative biomedical studies involving national, multinational and transnational partners developing various interventions targeted against health conditions such as HIV/AIDS, malaria, tuberculosis, childhood illnesses, and causes of maternal morbidity and mortality contained in the Millennium Development Goals, and the potential for exploitation in such research, make it essential for every country to have a functional Research Ethics Committee (REC). The purpose of national REC is to contribute independently, competently, and efficiently to safeguard the dignity, rights, safety, and well-being of all actual or potential research participants; and ensuring the highest attainable quality in the science and ethics of biomedical research [2] in the country.
The aim of the study reported here was to ascertain which Member State in the WHO African Region had a national bioethics committee and which did not (as at 2003), in order to guide the Organization in its support to the establishment of RECs, as well as their strengthening wherever they existed.
Methods
The questionnaire was deliberately kept short and simple in order to ensure quick response. It consisted of questions aimed at obtaining the following information: existence of a national ethics committee dealing with health issues; its composition and main functions; frequency of scheduled meetings and number of times the committee actually met last year; if a REC did not exist, were there mechanisms for clearing ethical issues in research; and whether a national ethical approval was required for implementation of research proposals.
The questionnaire was developed in French and subsequently translated into English and Portuguese. Of the 46 Member countries in the WHO African Region, 21 speak French, 20 English and 5 Portuguese. It was sent by WHO diplomatic pouch to each of the 46 countries through the WHO country representatives for facilitation and follow up. The data were entered in Excel spreadsheet and subsequently exported to STATA for analysis.
A Chi-Squared test (χ2) for independence was undertaken to test the relationship between presence/absence of Research Ethics Committee (REC) and selected individual socioeconomic and health variables. The null hypothesis (H0) is that there is no difference between the mean of a given socioeconomic or health variable among the group of countries with REC and the group without REC. The alternative hypothesis (HA) is that there is difference between the mean of a given socioeconomic or health variable among the group of countries with REC and the group without REC. Thus, for maternal mortality per 100000 live births, the hypotheses are as follows: (i) H0: There is no difference in average maternal mortality ratio between the group of countries with REC and those without; and (ii) HA: There is difference in average maternal mortality ratio between the group of countries with REC and those without. If the computed Chi-square () is greater than critical Chi-square (), then we reject the null hypothesis and accept the alternative hypothesis at a given level of significance, e.g. 95% confidence level or 5% level of significance. On the other hand, if the computed Chi-square () is less than critical Chi-square (), then we accept the null hypothesis, i.e. there is no reasonable evidence to reject the H0.
The data on maternal mortality per 100000 live births, population in a country, life expectancy at birth, probability of dying (per 1000) below age 5 years, probability of dying (per 1000) between age 15–60 years and infant mortality per 1000 live births were obtained from the World Health Report 2005 [19]. While the data on gross national income per capita (US$), percentage of population aged 15 years and above that is illiterate, and gross primary enrolment (% of school age population) were taken from the World Bank website [20].
Results
A total of 28 (60.9%) of the 46 countries completed the questionnaire and returned it to the investigators. Response rates of 50% (10/20), 67% (14/21) and 80% (4/5) were recorded for the English, French and Portuguese speaking countries respectively. An average of three remainders were sent through WHO Country Representatives to non-responding countries.
Of the 28 countries that responded, 64% (18/28) confirmed the existence of a national REC dealing with bioethics research issues (Table 1). This meant that 36% (10/28) of the respondent countries did not have an ethics committee. Sixty-seven per cent (12/18) of the countries that did have an ethics committee had it named as national bioethics committee; the remaining 33% (6/18) had alternative names for their national entity that performed the functions of REC.
Table 1 Presence/absence of RECs
Country Presence of REC
Algeria Yes
Angola Yes
Botswana Yes
Burkina Faso Yes
Cameroon Yes
Cape Verde No
Chad No
Congo No
Democratic Republic of Congo Yes
Ethiopia Yes
Gambia Yes
Guinea Yes
Guinea Bissau No
Guinea Equatorial No
Kenya Yes
Malawi No
Mali Yes
Mauritania No
Mauritius Yes
Mozambique Yes
Niger Yes
Rwanda Yes
Sao Tome et Principe No
Seychelles Yes
Swaziland No
Togo No
Zambia Yes
Zimbabwe Yes
Source: Survey data
As indicated in Table 2, the reported composition of REC membership varied widely. Over 50% of the countries had specialists from national medical research council/institute, universities and social sciences, and public health professionals from ministry of health.
Table 2 Composition of research ethics committees in countries that reported their existence
Cadre Percentage (Frequency)
National medical research institute/council 72% (13/18)
University 61% (11/18)
National medical association 44% (8/18)
Renowned researcher 33% (6/18)
Social scientist 50% (9/18)
Public health professional from ministry of health 89% (16/18)
National bureau of standards 11% (2/18)
WHO 5% (1/18)
Attorney-general 39% (7/18)
Others 83% (15/18)
The number of members on the national committee ranged from 4 to 37. The average membership was 11, with a standard deviation of 8.
The countries that confirmed the existence of a REC were requested to name its main functions. Their responses are summarized in Table 3. All these countries (18/18) indicated 'Review and approve all research protocols on human subjects' as one of the main functions of REC.
Table 3 Main functions of research ethics committees in countries that reported their existence
Functions Percentage (Frequency)
Review and approve all research protocols on human subjects 100% (18/18)
Ensure that projects sponsored by external donors are submitted for approval to ethical clearance committee of the initiating country 5% (1/18)
Build the capacity of institutional and regional ethical clearance committees 11% (2/18)
Give clearance for export of samples or specimens for research related to human health 5% (1/18)
Coordinating and promoting research 22% (4/18)
Eighty per cent (8/10) of the countries that did not have a REC reported that they had ad hoc mechanisms for ethical review of bioethics research protocols. These consisted of mainly hand-picking a few colleagues at the ministry of health to review proposals whenever they were submitted.
The countries were asked whether REC was supposed to meet monthly, quarterly, twice a year, once a year, or on demand. Seventeen countries responded to this question; of them, ten indicated the frequency as monthly, three quarterly, one twice a year, and three on demand.
The countries were also requested to indicate the actual frequency of REC meetings during the last year. A total of 16 countries responded to this question; of them, seven reported as monthly, five quarterly, three twice a year, zero annually, and one indicated that their REC had never met.
Lastly, all countries were asked whether ethical approval for research proposals was actually required. Eighty-five per cent (22/26) of the countries that responded to this question indicated that an approval was required.
Table 4 presents the Kruskal-Wallis test of equality of populations means of socio-economic and health variables. This test was undertaken to test the relationship between presence/absence of REC and the individual socio-economic and variables. The p value reported in the third column of Table 4 is the estimated probability of rejecting the null hypothesis (H0) when the hypothesis is true. The computed Chi-squared value of the population variable was greater than the critical Chi-squared value; thus we can conclude that the average population in the group of countries with a REC was statistically different from that of the group of countries without a REC. On the other hand, since the computed values for all the other socio-economic and health variables were less than their respective critical Chi-squared values (at 5% level of significance) we accept the H0.
Table 4 Kruskal-Wallis pairwise comparisons between variables of the group of countries with a research ethics committee and the group without
Variables Chi-squared statistic p-value Decision at 5% significance level
Maternal mortality per 100000 live births 0.026144 P = 0.8712 not significant, accept H0
Population in a country 7.468966 P = 0.0063 Significant, reject H0
Life expectancy at birth 0.7468966 P = 0.3876 not significant, accept H0
Probability of dying (per 1000) below age 5 years 0.331125 P = 0.565 not significant, accept H0
Probability of dying (per 1000) between age 15–60 years 0.186207 P = 0.6661 not significant, accept H0
Gross national income per capita (US$) 0.111875 P = 0.738 not significant, accept H0
Infant mortality per 1000 live births 0.028161 P = 0.8667 not significant, accept H0
Percentage of population aged 15+ years that is illiterate 0.22623 P = 0.6343 not significant, accept H0
Gross primary enrolment (% of school age population) 2.712306 P = 0.0996 not significant, accept H0
Note: Total number of observations = 28.
Discussion
Key findings
The purpose of this study was to ascertain which country in the WHO African Region had a national bioethics committee and which did not. The key findings were as follows: (i) the response rate was 61% (28/46); (ii) 64% (18/28) confirmed the existence of a national ethics committee; (iii) 36% (10/28) of the respondent countries did not have an ethics committee (although 80% of them reported that they used an ad hoc ethical review mechanism); (iv) 85% (22/26) of the countries that responded to this question indicated that ethical approval of research proposals was, in principle, required; and (v) although 59% of the countries expected their REC to meet every month, only 44% reported that it actually met on a monthly basis.
Implications
Response rate
It is very sad that only 28 out of 46 countries responded to the questionnaire. It would have been preferable to have responses from all the 46 countries. This would have provided WHO and the Special Programme for Research and Training in Tropical Diseases Research (TDR) with a strong basis for planning support to the countries in the African Region that needed it to develop or strengthen their ethical review systems.
Thus, the analysis reported in this paper was confined to the 28 countries that responded and mainly to the 18 who reported to have had a REC. We are not certain why 18 countries did not respond to the questionnaire. However, it could partly be attributed to lack of a culture for generating and utilizing evidence in decision-making in the Region.
Existence of REC
Although from this study we cannot assess the functional status of RECs in the 18 countries that reported their existence, we are more concerned about the remaining ten countries that did not have a national ethics committee in an era in which biomedical research (and science in general) has become increasingly globalized. This is quite disconcerting in a Region where 39% of the adult population is illiterate and 44% live below the international poverty line of US$1 per day [11]. They thus are largely unaware of their basic human rights, including the right to refuse to participate in health research. In the absence of any checks and balances due to lack of functional RECs, the populations in these countries run the risk of being abused by unscrupulous researchers [1,2,12].
WHO is currently supporting Member countries without RECs to establish them and those with RECs to improve their performance. Given the critical importance of RECs, we do hope that the ongoing support to countries in the African Region would be accelerated and sustained.
Functions/roles of REC
The roles of ethical review committees are to: ensure that all proposed interventions (drugs, vaccines, medical devices or procedures) are safe; ensure that proposed research is scientifically sound; ensure that all other ethical concerns arising from a protocol are satisfactorily resolved both in principle and in practice; and ensure competence of investigators; keep records of decisions; monitor and audit the conduct of ongoing research projects [3]; "... evaluate research proposals with special attention to risk/benefit ratios, equity in distribution of benefits and burdens, potential conflicts of interests, the adequacy of information provided for subjects, and the protection of freedom: within the consent process; enable study subjects to withdraw without prejudice to care; and persuade investigators to publish, to educate and assist faculty, researchers and community in understanding and appreciating the ethics of research" [1].
It was beyond the scope of the current study to assess the extent to which RECs in the Region performed the above-mentioned roles. However, it appeared that all the respondent countries regarded the main function of REC to be to review and approve research protocols on human subjects. There was no mention during the survey that RECs were expected to also monitor the actual conduct of research to ensure justice in the distribution of costs and benefits. We concur with Benatar [1] that "lack of attention to how research is actually being conducted is a serious shortcoming, requiring critical attention in an era of expanding research, growing links with industry and commercial organizations, documented inadequacies in the protection of research subjects and with growing recognition of the need to avoid exploitation." In short, it is vitally important for RECs to ensure that they competently implement all the guidelines stipulated by the Council for International Organizations of Medical Sciences (CIOMS) [3].
Composition of REC
The CIOMS guidelines [3] recommend that the membership of REC "... should include physicians, scientists and other professionals such as nurses, lawyers, ethicists and clergy as well as lay persons qualified to represent the cultural and moral values of the community." The survey undertaken in the current study did not reveal the involvement of nurses, ethicists and qualified laypersons in the work of RECs. Since nurses constitute the largest group of health personnel in health systems delivery and play a critical role as members of the health team, it is vitally important that they are involved in all ethical review processes. Also, non-inclusion of qualified laypersons in RECs may compromise cultural and moral values [1].
Ethical approval
About 15% of the respondent countries indicated that ethical approval of research proposals was not required. In such countries, the health and safety of persons participating in research was not assured. Even among the countries that indicated that ethical approval of research proposals was, in principle, required, we were not able to determine: (i) what proportion of research protocols were actually approved before implementation; and (ii) what proportion of the approved studies were actually monitored by REC throughout the research project cycle, i.e. protocol design, data collection and archival, data analysis, dissemination of results, etc.
Frequency of REC meetings
In the current study, the frequency of REC meetings was used both as an indicator of their functionality and as an indirect proxy of performance. Half of the countries that responded to this issue reported that their RECs met either quarterly (31%) or twice a year (19%). In our view, the 3 to 6-month interval for REC meetings was too long as this may result in unnecessary delays in processing (approval or rejection) of research protocols. Where the sponsors of research projects have stringent deadlines, such delays may: (i) lead to withdrawal of research funding; (ii) reduce the probability of researchers from concerned countries getting funding in the future; (iii) increase the cost of research; (iv) hamper the development and availability of public health interventions; and (v) provide adverse incentives for some researchers to undertake research covertly, without ethical review, especially in settings where appropriate legislations either do not exist or are not policed.
One of the countries indicated that its REC had never met. On hindsight, we feel we should have included questions to probe the reasons for not holding REC meetings. At the moment, it is not possible to know whether the absence of meetings was due to lack of protocols to review, lack of resources to organize the meetings, lack of quorum, incompetent leadership, or due to some other logistical reasons.
Agenda for action
The following actions need to be taken by countries in the African Region:
• Countries should adapt appropriately international guidelines for biomedical research involving human subjects and make them available to all national health and health-related research institutions and health facilities.
• As a good national ethics steward, every country should ensure that it has an operational bioethics research review system in place, which includes national, regional, district and institutional (health facility) ethics committees.
• All countries should make sure that there are appropriate policies and legislations to guide and reinforce national bioethics research review systems.
• All countries should ensure that the composition and mandates of RECs concur with international standards, while providing institutional and financial resources to ensure independent and competent execution of all their roles. In this regard, we agree with Benatar [1] that all countries should 'develop the expertise and infrastructure required to (i) evaluate ethical problems; and (ii) educate practitioners and researchers and facilitate development of policy'.
• Each country should champion 'institutionalization of training in ethics and human rights in relation to health at all stages of the education and training of all health workers, including medical, public health and nursing schools' [13].
• All countries with RECs should develop mechanisms for monitoring and auditing their work to ensure that they are guaranteeing research adherence to all the CIOMS guidelines [3].
Further research
Our suggestions for further research in countries in the African Region are as follows:
Situation analysis
A detailed survey and evaluation should be undertaken of the effectiveness and efficiency of the entire national ethics review system (including regional, district, institutional and community-based RECs) in implementing the CIOMS guidelines [3]. The evaluation can be accomplished using the WHO Guidelines [14]. It would be important to evaluate: (i) the organization, financing and functionality of the whole ethics review system; (ii) the extent to which RECs are involved in monitoring every stage/step in a research project cycle, including protocol design, implementation, archival of data, analysis and public dissemination of results; (iii) the effects of recent information and technology developments on RECs' modus operandi; and (iv) challenges faced by RECs.
Best practices
A detailed study and documentation of best-performing research ethical review systems (including RECs) in the Region should be conducted with a view to drawing lessons that other countries can emulate.
Principal-agency relationship
The principal-agency relationship between ethics committees and human research subjects should be established.
Institutionalisation of ethics education and training
A review of the existing international ethics review guidelines should be undertaken with a view to designing appropriate undergraduate and postgraduate curricula on research ethics. We concur with Fischer and Zigmund [15] that research ethics should be taught throughout the graduate curriculum. However, in addition, we are of the opinion that in Africa where a majority of the health and allied sciences undergraduates do not proceed to postgraduate studies, it is critically important to introduce undergraduates also to research ethics. After obtaining their degrees, most undergraduates are normally deployed in rural areas where, by virtue of being the most educated, they often bear the burden of assuring that human rights of their actual and potential clients are respected and protected in the course of their clinical work and research carried out by others.
Ways of improving REC performance
There is need for studies that explore the cost and benefits (effectiveness) of alternative ways of leveraging the recent advances in technology (teleconferencing, video conferencing, e-mail) to boost the work of RECs. Where these technologies exist, they would not only reduce the cost of face-to-face meetings but will also ensure timely review of research protocols.
Partnerships
An exploration of the modalities of South-South and North-South cooperation to strengthen the capacities of bioethics review systems in the Region should be made. For example, the WHO Regional Committee for Africa [16] identified the need for effective inter-country mechanisms to monitor health research in order to ensure that existing national and international bioethics guidelines were adhered to. In addition, countries with limited bioethics capacities could easily tap into the internationally available bioethics expertise through the Internet. We are encouraged by the emphasis laid on capacity-building for national ethics committees by the European and Developing Countries Clinical Trials Partnership (EDCTP).
Financing of REC work
Currently, the effectiveness of RECs in many countries is greatly constrained by lack of resources [17]. Thus, there is urgent need for research into finding innovative mechanisms for ethically financing REC activities.
Conclusion
In the current era of globalized research, good ethics stewardship demands that every country, whatever its level of economic development, should have a functional research ethics review system for protecting the dignity, integrity and health safety of all its citizens participating in research. Those countries that do not currently have such systems should urgently leverage the services provided by the WHO Strategic Initiative for Developing Capacity in Ethical Review (SIDCER) [18] to develop capacities for ensuring ethical research practices.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JMK recoded the raw data, did the analysis and participated in drafting of all sections of the document. CW participated in the analysis and drafting of all sections of the document. AB sent out the questionnaires and made follow-up through WRs. He also participated in drafting the Background and Methods sections. The WHO Country Representatives of the 28 countries that responded ensured that the relevant persons in their countries of duty completed the questionnaires. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Dr Antonio Filipe, Dr Andrew Kosia and Professor Rose Leke made useful suggestions in the design of the questionnaire. We are grateful to Dr Luis Sambo of WHO/AFRO for encouraging us to undertaking this study. We owe profound gratitude to: the national authorities in the countries that completed the questionnaires; and the WHO Country Representatives and WHO Country Teams from the 28 countries that responded for facilitating data collection. We are grateful to A Kochar for editorial help. We are very grateful to the two peer reviewers, Professor Richard H Morrow and Professor Solomon R Benatar, for their suggestions that helped in improving the quality of this paper. Finally, we are immensely indebted to Adonai Sabbaoth-Shamah-Mekodeshkum for multifaceted support.
This article contains the views of the authors only and does not represent the decisions or the stated policies of the Bingham University or the World Health Organization.
==== Refs
Benatar SR Reflections and recommendations on research ethics in developing countries Social Science and Medicine 2002 54 1131 1141 11999507 10.1016/S0277-9536(01)00327-6
World Health Organization Operational guidelines for ethics committees that review biomedical research Geneva 2000
CIOMS International ethical guidelines for biomedical research involving human subjects Geneva 2002
Hyder AA Wali SA Khan AN Teoh NB Kass NE Dawson L Ethical review of health research: a perspective from developing country researchers Journal of Medical Ethics 2004 30 68 72 14872079 10.1136/jme.2002.001933
United States Government Printing Office Trials of War Criminals before the Nuremberg Military Tribunals under Control Council Law No 10, Washington, DC 1949 2 181 182
World Medical Association Declaration of Geneva – physician's oath Geneva 1948
United Nations International covenant on civil and political rights New York 1966
World Medical Association International code of medical ethics of the World Medical Association World Medical Association Bulletin 1949 1 109 111
World Medical Organization Declaration of Helsinki British Medical Journal 1996 313 1448 1449
International Conference on Harmonization (ICH) Guidelines for good clinical practice Toronto 2002
United Nations Development Programme Human Development Report 2003: Millennium Development Goals: a compact among nations to end human poverty New York 2003
Weindling P Human guinea pigs and the ethics of experimentation: the BMJ's correspondent at the Nuremberg medical trail British Medical Journal 1996 313 1467 1470 8973237
CIOMS CIOMS-WHO Initiative on ethics, equity and health for all Geneva 1997
World Health Organization Surveying and evaluating ethical review practices Geneva 2002
Fischer BA Zigmund MJ Teaching ethics: resources for researchers TINS 1996 19 523 524
WHO/AFRO Fifty-first session of the WHO Regional Committee for Africa: final report Brazzaville 2001
Dickens BM Cook RJ Challenges of ethical research in resource-poor settings International Journal of Gynaecology and Obstetrics 2003 80 79 86 12527468 10.1016/S0020-7292(02)00349-1
World Health Organization Strategic Initiative for Developing Capacity in Ethical Review (SIDCER): terms of reference and strategic plan Geneva 2002
World Health Organization World Health Report 2005: making every mother and child count Geneva 2005
World Bank
|
16242014
|
PMC1274319
|
CC BY
|
2021-01-04 16:31:58
|
no
|
BMC Med Ethics. 2005 Oct 20; 6:10
|
utf-8
|
BMC Med Ethics
| 2,005 |
10.1186/1472-6939-6-10
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-591622344310.1186/1471-2180-5-59Research ArticlePolyamine stress at high pH in Escherichia coli K-12 Yohannes Elizabeth [email protected] Amy E [email protected] Jessica C [email protected] Daniel P [email protected] Joan L [email protected] Department of Biology, Kenyon College, Gambier, OH 430222005 13 10 2005 5 59 59 8 7 2005 13 10 2005 Copyright © 2005 Yohannes et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Polyamines such as spermine and spermidine are required for growth of Escherichia coli; they interact with nucleic acids, and they bind to ribosomes. Polyamines block porins and decrease membrane permeability, activities that may protect cells in acid. At high concentrations, however, polyamines impair growth. They impair growth more severely at high pH, probably due to their increased uptake as membrane-permeant weak bases. The role of pH is critical in understanding polyamine stress.
Results
The effect of polyamines was tested on survival of Escherichia coli K-12 W3110 in extreme acid or base (pH conditions outside the growth range). At pH 2, 10 mM spermine increased survival by 2-fold, and putrescine increased survival by 30%. At pH 9.8, however, E. coli survival was decreased 100-fold by 10 mM spermine, putrescine, cadaverine, or spermidine. At pH 8.5, spermine decreased the growth rate substantially, whereas little effect was seen at pH 5.5. Spermidine required ten-fold higher concentrations to impair growth. On proteomic 2-D gels, spermine and spermidine caused differential expression of 31 different proteins. During log-phase growth at pH 7.0, 1 mM spermine induced eight proteins, including PykF, GlpK, SerS, DeaD, OmpC and OmpF. Proteins repressed included acetate-inducible enzymes (YfiD, Pta, Lpd) as well as RapA (HepA), and FabB. At pH 8.5, spermine induced additional proteins: TnaA, OmpA, YrdA and NanA (YhcJ) and also repressed 17 proteins. Four of the proteins that spermine induced (GlpK, OmpA, OmpF, TnaA) and five that were repressed (Lpd, Pta, SucB, TpiA, YfiD) show similar induction or repression, respectively, in base compared to acid. Most of these base stress proteins were also regulated by spermidine, but only at ten-fold higher concentration (10 mM) at high pH (pH 8.5).
Conclusion
Polyamines increase survival in extreme acid, but decrease E. coli survival in extreme base. Growth inhibition by spermine and spermidine requires neutral or higher pH. At or above pH 7, spermine and spermidine regulate specific proteins, many of which are known to be regulated by base stress. High pH amplifies polyamine stress; and naturally occurring polyamines may play an important role in base stress.
==== Body
Background
Polyamines are required for the normal cell growth of Escherichia coli, although their functions are poorly understood [1-3]. Polyamines bind nucleic acids and ribosomes, where they are needed for optimal function [4,5]. Excessive intracellular concentrations, however, retard protein synthesis and cell growth [6]. Polyamine metabolism is stimulated by a variety of environmental stresses such as heat shock [7].
The major polyamine of bacteria, putrescine [NH2(CH2)4NH2], is synthesized by biosynthetic decarboxylation of arginine and/or ornithine [8,9]. Putrescine is metabolized to spermidine [NH2(CH2)3NH(CH2)4NH2]. Spermine [NH2(CH2)3NH(CH2)4NH(CH2)3NH2], a longer polyamine commonly produced by eukaryotes, is not produced by E. coli. Nevertheless, uptake of exogenous spermine fullfills the bacterial requirement for polyamines [4,5]. Spermine and spermidine do not undergo catabolism by E. coli, although excess concentrations are acetylated by polyamine acetyltransferase [10,11].
In the human colon, bacteria excrete putrescine and cadaverine during digestion of high-protein foods. Exposure of colonic epithelium to these polyamines stimulates human cell proliferation and leads to colonic tumors [12], which can be treated by drugs that deplete polyamine content [13]. Thus the modulation of polyamine metabolism under conditions of the gut is an important medical concern.
The inhibition of growth by excess polyamines is amplified at high pH [14]. Polyamine stress enhances translation of the growth phase sigma RpoS [15], whose role in stationary-phase survival involves high pH [16]. A possible explanation for the amplification of polyamine stress at high pH is that polyamines become deprotonated and neutralized, thus capable of crossing the cell membane as membrane-permeant weak bases. Base-dependent uptake could augment the uptake through transporters [17]. The uptake of amines leads to their accumulation proportional to the transmembrane pH difference (ten-fold for each pH unit). Only a small fraction of an amine needs to be unprotonated (less than 1%) in order for significant membrane passage to occur. The pKa values of linear polyamines range from pKa = 8.3 to 11.6; for example, Ref [18] reports for 30 mM spermidine values of pKa1 = 8.6, pKa2 = 10.0, pKa3 = 11.1. However, literature reports vary for different conditions, and polyamine protonation levels under biological conditions are further influenced by complexing with fatty acids and phospholipids.
At low pH, on the other hand, the transmembrane pH difference (ΔpH) limits the entry of exogenous amines; and cytoplasmic production of membrane-permeant amines can enhance bacterial growth. A pH-dependent source of amines in E. coli is the degradative arginine and ornithine decarboxylases (AdiA and SpeF respectively) and lysine decarboxylase (CadA), which are induced anaerobically at low pH [19]. The generation of putrescine (by AdiA) or cadaverine (by CadA), followed by excretion via cotranscribed transporters, neutralizes the acidic external environment [20]; for review, see [21,22].
At low-to-neutral pH, E. coli polyamines block the porins OmpF and OmpC, decreasing membrane permeability [23-25]. It was proposed that polyamines contribute to E. coli survival in extreme acid, below the growth range [24]; a phenomenon termed acid resistance or acid survival [21,22]. An OmpC mutant in which the porin fails to be blocked by cadaverine shows decreased survival at low pH (acid resistance) in the presence of cadaverine [24]. To our knowledge, it has not been shown directly that exogenous polyamines enhance acid resistance of wild-type cells. If polyamines do enhance survival in extreme acid, they could assist E. coli and other enteric pathogens during their passage through the stomach [26].
The interactions between pH and polyamines however remain poorly characterized, and are often discounted in studies of polyamine stress. For example, a recent major study of polyamine-mediated gene regulation does not address pH [3]. We report the effect of exogenous spermine and other polyamines on E. coli survival in extreme acid or extreme base. We also present protein profiles of E. coli exposed to exogenous polyamines under neutral and alkaline pH conditions, in the context of known pH-dependent expression profiles [27,28].
Results
Survival at extreme pH
E. coli grown at moderate pH values possess mechanisms of protection against more extreme pH; these mechanisms are typically induced during stationary phase, or during growth near the acid or base end of their pH range [21,22]. We tested the effects of spermine, putrescine, spermidine and cadaverine on survival rates in extreme acid or base, outside the range of pH permitting growth. Stationary-phase cultures of E. coli K-12 were diluted 1000-fold in LBK pH 2.0, or in LBK 100 mM CAPS pH 9.8, and incubated for 2 h at 37°C, then dilutions were plated on LBK agar. The pH of the incubation medium was tested before and after culture inoculation, and no change in pH was seen. Under these conditions of extreme acid or base, overall survival (based on LBK plate count) was about 20% compared to that of a control culture diluted in LBK.
At pH 2.0, the presence of 10 mM spermine doubled the proportion of E. coli survivors (Figure 1). Putrescine and cadaverine increased survival by smaller increments, and spermidine had no effect. At pH 9.8, however, all four polyamines tested decreased survival by more than 100-fold. Thus, the presence of polyamines enhanced survival in acid but drastically diminished survival in extreme base.
Growth rates with spermine or spermidine
In order to observe polyamine stress and protein expression during growth, we treated cultures with exogenous spermine and spermidine. These polyamines were selected because they do not undergo the rapid catabolism that occurs to putrescine, the major endogenous polyamine of E. coli. Spermine interacts with cell components in similar ways as other polyamines found in E. coli [5], and it effectively enters cells [17].
Growth rates were observed in the presence and absence of spermine at pH 5.5, 7.0, or 8.5 (Figure 2). At pH 7.0 or higher, 3 mM spermine eliminated growth, and at pH 8.5 as little as 1 mM spermine substantially decreased growth. At pH 5.5, however, spermine had no effect up to concentrations as high as 10 mM, and spermidine had no effect at 20 mM (data not shown). Spermidine required larger concentrations to affect growth at high pH. At pH 7.0, spermidine concentrations up to 15 mM did not affect the generation time, but at pH 8.5, 15 mM spermidine prevented growth. Thus both spermine and spermidine showed base-dependent depression of E. coli growth, although spermidine required higher concentrations.
2-D protein gels
The protein profiles of E. coli in the presence and absence of exogenous spermine or spermidine were observed at pH 7.0 and at pH 8.5, using methods that previously revealed pH-dependent protein profiles [28,29]. The concentrations of each polyamine were chosen based on Figure 2 so as to cause significant stress at high pH without preventing growth.
The composite 2-D protein profiles are shown in Figures 3 and 4, and the results of quantitative analysis of pairwise comparisons are shown in Table 1. At pH 7.0, 1 mM spermine increased the expression of eight proteins, including PykF, GlpK, SerS, DeaD, OmpC and OmpF. On the other hand, 19 proteins were repressed, including acetate-inducible proteins (YfiD, Pta, Lpd), the RNA polymerase-binding protein RapA (HepA), and the fatty acid synthesis proteins FabB, FabE. At pH 8.5, spermine induced all the proteins induced at pH 7.0, plus four additional proteins: TnaA, OmpA, NanA, and YrdA. Spermidine regulated most of the same proteins as spermine, although 10-fold higher concentration was required.
The proteins induced by spermine and spermidine include known base-inducible proteins such as TnaA, which deaminates several amino acids [29,30], as well as base-inducible GlpK, OmpA, and OmpF [27]. Proteins inducible at low pH or by acetate, however, were repressed by spermine, including Lpd, Pta, SucB, TpiA, and YfiD [27,31]. By contrast, at pH 5.5, spermine and spermidine had no effect on protein profiles (data not shown). The lack of effect of polyamines at low pH is consistent with their exclusion from the cell as permeant bases, in equilibrium with the trans-membrane ΔpH.
Several porins (OmpC, OmpF, and OmpA) showed increased expression in the presence of spermine. While blockage of porin function by polyamines is eliminated at pH 9.5 [32], some porin blockage may occur as high as pH 8.5. A possible interpretation could be that cells respond to porin blockage by overexpressing the porins.
The polyamines consistently repressed RapA, a general transcriptional activator that stimulates recycling of RNA polymerase [33]; its expression is growth phase dependent, peaking in early log phase [34]. The repression of RapA could contribute to the growth inhibition by polyamines at high pH.
Discussion
We show that the presence of spermine enhances bacterial survival in extreme acid, as well as diminishing survival in extreme base. This finding is of medical significance, as it suggests that exogenous polyamines enhance survival during passage through the stomach to the colon. The lumen of the colon, at pH 6–8, normally contains millimolar concentrations of polyamines [12], and levels may reach 5–10 mM in children with diseases of nutrient malabsorption such as cystic fibrosis [35]. Polyamines are implicated in colonic hyperproliferation and cell migration, leading to tumorigenesis.
The polyamine enhancement of extreme-acid survival could involve several possible mechanisms. Amines produced by amino-acid decarboxylases can neutralize acidity, a factor in acid resistance [19]. The OmpC porin is blocked by the polyamine cadaverine at low to neutral pH; the blockage of porins could protect cells from acidification [23-25]. Polyamines could also protect the cell's DNA and RNA from effects of internal acidification during extreme-acid exposure [36].
In protein profiles, spermine and spermidine showed base-dependent regulation of 31 proteins, including nine proteins known to respond to base stress. Polyamines have numerous effects upon cell function; their effects could be enhanced at high pH by ΔpH-driven uptake of polyamines. Alternatively, polyamines could amplify base stress by taking up protons within the cytoplasm, thus requiring the cell to spend more energy maintaining pH homeostasis.
Our results of protein expression differ significantly from the report of putrescine-dependent genes by Yoshida and colleagues [3]. For example, Ref. [3] reports putrescine induction of acid stress genes such as dps, hdeA, hdeB, gatABD, and the flagellar regulon. We however find mainly induction of base stress proteins and repression of acid stress proteins. The difference may relate to the fact that in Ref. [3], the growth medium used was glucose-minimal Medium A, with pH unspecified but presumably buffered with phosphate at pH 7, with cells grown to late log phase. These growth conditions involve substantial acidification and acetate production. Furthermore, the putrescine microarray study was performed using a mutant deficient for putrescine conversion to spermidine, which requires exogenous polyamines for optimal growth. The effects of acidification could have been amplified by the growth defect of the polyamine-deficient strain; thus, culture acidification could explain why Ref. [3] found acid stress induction. Our proteomic experiments, however, used an E. coli K-12 strain with no known polyamine defects, and growth pH was maintained by buffering at pH 7.0 or 8.5.
Conclusion
It has been shown that polyamines increase survival in extreme acid, but decrease E. coli survival in extreme base. At or above pH 7, spermine and spermidine induce specific proteins, including those associated with base stress. Base stress and base-enhanced polyamine uptake should be considered during all studies of polyamine stress.
Methods
Growth conditions
E. coli K-12 strain W3110 was grown overnight in potassium-modified Luria broth (LBK) (10 g/l of tryptone, 5 g/l of yeast extract, 7.45 g/l of KCl). The overnight cultures were diluted 500-fold into 20 ml of buffered medium supplemented with 1 mM spermine or 10 mM spermidine. The buffers used included 3-(N- morpholino)propanesulfonic acid (MOPS) (pKa 7.01), 3- [N-tris(hydroxymethyl)methyl]- 3-aminopropanesulfonic acid (TAPS) (pKa 8.11), and cyclohexyl-3-amino-1-propanesulfonic acid (CAPS) (pKa 10.4). Buffers and polyamines were purchased from Research Organics and from Sigma. The pH values of media were adjusted by using KOH to avoid extra sodium ions, which stress cells at high pH [37]. The pH was tested after culture growth to ensure that the values were maintained at ± 0.1 pH unit of the pH of the original uninoculated medium. For 2-D gels, all cultures were grown aerobically in baffled flasks rotated at 240 cycles/sec at 37°C until the OD600 reached 0.2.
Survival assays
To assay survival in extreme acid, overnight cultures of bacteria in LBK 50 mM MOPS pH 7.0 were diluted 1000-fold in LBK adjusted to pH 2.0. After 2 h incubation at 37°C, the pH of the culture was tested to confirm maintenance at pH 2.0. Cells were plated on LBK agar at 37°C. For survival in extreme base, overnight cultures in LBK 50 mM TAPS pH 8.5 were diluted 1000-fold in LBK 100 mM CAPS pH 9.8, incubated for 2 h at 37°C, and the pH tested to confirm maintenance at pH 9.8. Dilutions were plated on LBK. For either acid or base experiments, survival was normalized to that of control samples diluted in LBK 50 mM MOPS pH 7.0 and plated directly without incubation.
2-D protein gel electrophoresis
2-D gel electrophoresis of proteins was performed as described [28,31,38]. The protein mixtures were first separated by isoelectric focusing using 18-cm polyacrylamide gel strips with an immobilized pH 4 to 7 gradient (AP Biotech). The second dimension was an electrophoretic gel slab containing 11.5% acrylamide, and the proteins were silver-stained. Certain stained proteins commonly appear as isoforms in multiple spot positions [28].
Protein spot densities were quantified using the Compugen Z3 v.3.0 software (Compugen, Tel Aviv, Israel). The differential expression ratio (DE) of the spot densities for two growth conditions (experimental and control) was computed by pairwise comparisons of a set of three gels from each of two growth conditions. A protein spot was considered a candidate for significant induction if seven of nine pairwise comparisons produced a DE greater than or equal to 1.5 or less than 0.67. For each protein, the log10 of all nine DE values was computed, and the mean log10 DE (LDE) was considered a measure of induction (positive values) or repression (negative values) (Table 1).
Protein identification
Proteins having a significant LDE were identified either by MALDI-TOF analysis at the Proteomic Mass Spectrometry Laboratory at the University of Massachusetts [39] or by positional comparison with previous gels in which proteins had been identified by MALDI-TOF analysis or N-terminal sequencing [28,31]. For database searches of MALDI-TOF masses the Protein Prospector site was used [40] at the University of California at San Francisco.
Authors' contributions
EY designed the study, conducted the 2-D gels, and drafted the manuscript. AT and JW designed and conducted the survival assays. DPT conducted growth curves. JLS conceived of the study, contributed to its design, and revised the manuscript.
Acknowledgements
This work was supported by grant MCB-0234732 from the National Science Foundation.
Figures and Tables
Figure 1 Extreme-acid survival in the presence of polyamines. The mean survival at pH 2 after 2 h in the presence of each polyamine (10 mM) was divided by the mean survival of a sample diluted from the same overnight culture without exogenous polyamines. The mean of six independent cultures is shown, +/- SEM.
Figure 2 Generation time with polyamines, as a function of pH. E. coli was grown aerobically at 37°C in LBK buffered at pH 5.5, 7.0, or 8.5, with or without 1 mM spermine or 10 mM spermidine. Overnight cultures were diluted 1000-fold, and growth rates were measured between OD600 values of 0.1 and 0.2. Error bars represent SEM (N = 3).
Figure 3 Protein profiles in the presence or absence of 1 mM spermine. The horizontal axis represents the approximate pH range of the isoelectric focusing first dimension, and the vertical axis represents the molecular weight (Mw). The "layered view" superimposes two composite images, each composed from three 2-D gels of independently grown replicate cultures. All E. coli cultures were diluted 500-fold and grown at 37°C to an OD600 of 0.2 in LBK with (pink) or without (green) 1 mM spermine. Culture media were buffered as described under Methods. (A) 100 mM MOPS, pH 7.0. (B) 100 mM TAPS, pH 8.5.
Figure 4 Protein profiles in the presence or absence of 10 mM spermidine. Cultures were grown as for Fig. 2B in LBK 100 mM TAPS, pH 8.5, with (pink) or without (green) 10 mM spermidine.
Table 1 Proteins showing differential expression on 2-D gels.
Differential expression1
Spot no. Protein Spermine (1 mM) Spermidine (10 mM) pH 8.5 Known or predicted function
pH 7.0 pH 8.5
1 PykF 0.34 ± 0.09 (+) (+) Pyruvate kinase
2 PykF 0.39 ± 0.06 (+) (+) Pyruvate kinase
3 GlpK 0.68 ± 0.09 0.63 ± 0.12 0.50 ± 0.09 Glycerol kinase
4 SerS 0.83± 0.08 0.60 ± 0.1 0.60 ± 0.1 Serine-tRNA ligase
5 ThrC -0.34 ± 0.06 Threonine synthase
6 0.51 ± 0.03 0.81 ± 0.07
7 TnaA 0.92 ± 0.06 0.87 ± 0.09 Tryptophanase
8 TnaA 0.53 ± 0.07 0.77 ± 0.07 Tryptophanase
9 MalE -1.04 ± 0.05 Maltose-binding, periplasmic
10 MalE -1.0 ± 0.00 Maltose-binding, periplasmic
11 MalE Maltose-binding, periplasmic
12 MalE -0.74 ± 0.08 Maltose-binding, periplasmic
13 Asd -0.54 ± 0.10 -0.42 ± 0.07 Aspartate semialdehyde dehydrogenase
14 FabB (-) (-) -0.64 ± 0.11 Beta-Ketoacyl-ACP synthase I
15 DeaD 0.27 ± 0.03 0.41 ± 0.08 ATP-dependent RNA helicase
16 TktA -0.46 ± 0.11 (-) -0.48 ± 0.07 Transketolase
17 OmpC (+) 0.56 ± 0.05 Outer membrane protein C
18 OmpF 0.96 ± 0.03 0.74 ± 0.15 Outer membrane porin
19 OmpF 0.56 ± 0.05 0.90 ± 0.08 Outer membrane porin
20 Pta (-) -0.50 ± 0.15 (-) Phosphate aceyltransferase
21 -0.47 ± 0.06 (-) (-)
22 MglB 0.23 ± 0.02 Galactose-binding protein
23 0.69 ± 0.11 (+)
24 MalM -59 ± 0.10 Maltose periplasmic protein
25 SucB -0.25 ± 0.02 -0.89 ± 0.08 Dihydrolipoamide Succinyltransferase
26 GlnS (-) -0.59 ± 0.11 -0.60 ± 0.1 Glutaminyl-tRNA synthetase
27 NanA 0.49 ± 0.12 Putative N-acetylmanosamine-6-phosphate 2-epimerase
28 Lpd (-) -0.79 ± 0.1 -0.69 ± 0.1 Dihydrolipoamide dehydrogenase
29 Lpd (-) -0.55 ± 0.1 -0.72 ± 0.17 Dihydrolipoamide dehydrogenase
30 Lpd (-) -0.58 ± 0.1 -0.27 ± 0.05 Dihydrolipoamide dehydrogenase
31 FabE -0.24 ± 0.04 -0.82 ± 0.07 Acetyl-CoA carboxylase
32 RpsB (-) -0.26 ± 0.05 30S ribosomal subunit protein S2
33 YrdA 0.26 ± 0.04 0.44 ± 0.12
34 OmpA 0.27 ± 0.04 0.43 ± 0.09 Outer membrane protein A
35 RpsF -0.22 ± 0.01 -0.85 ± 0.09 -0.50 ± 0.12 30S ribosomal subunit protein S6
36 RpsF -0.22 ± 0.01 30S ribosomal subunit protein S6
37 Gpt 0.18 ± 0.03 0.40 ± 0.08 0.58 ± 0.1
38 TpiA -0.28 ± 0.03 -0.57 ± 0.06 (-) Triosephosphate isomerase
39 RapA (HepA) -0.61 ± 0.02 -0.76 ± 0.09 -0.63 ± 0.11 RNA polymerase binding protein
40 YfiD -0.27 ± 0.02 -0.21± 0.02 -0.37 ± 0.03 Pyruvate formate-lyase homolog
41 GapA -0.51 ± 0.1 -0.87 ± 0.06 -0.50 ± 0.13 Glyceraldehyde 3-phosphate dehydrogenase A
42 AtpA -0.27 ± 0.07 ATP synthase subunit alpha
43 -0.74 ± 0.02 (-) (-)
44 GuaB -0.75 ± 0.1 Inosine-5'-monophosphate dehydrogenase
1Relative differential expression of protein, shown as LDE ± standard error (n = 9), was determined as described under Materials and Methods. Symbols: (+), induced; (-) repressed, representing spots that showed differential expression based on visual inspection, although not quantifiable by the Z3 software.
==== Refs
Chattopadhyay MK Tabor CW Tabor H Polyamines protect Escherichia coli cells from the toxic effect of oxygen Proc Natl Acad Sci 2003 100 USA 2261 2265 12591940 10.1073/pnas.2627990100
Minton KW Tabor H Tabor CW Paraquat toxicity is increased in Escherichia coli defective in the synthesis of polyamines Proc Natl Acad Sci 1990 87 USA 2851 2855 2181453
Yoshida M Kashiwagi K Shigemasa A Taniguchi S Yamamoto K Makinoshima H Ishihama A Igarashi K A unifying model for the role of polyamines in bacterial cell growth, the polyamine modulon J Biol Chem 2004 279 46008 46013 15326188 10.1074/jbc.M404393200
Kakegawa T Guo YCY Miyazaki T Nakamura M Hirose S Canellakis ZN Igarashi K Effect of acetylpolyamines on in vitro protein synthesis and on the growth of a polyamine-requiring mutant of Escherichia coli. J Biochem 1991 109 627–631 1869516
Xaplanteri MA Petropoulos AD Dinos GP Kalpaxis DL Localization of spermine binding sites in 23S rRNA by photoaffinity labeling: parsing the spermine contribution to ribosomal 50S subunit functions. Nucleic Acids Res 2005 33 2792 2805 15897324 10.1093/nar/gki557
He Y K. Kashiwagi J. Fukuchi Terao K Shirahata A Igarashi K Correlation between the inhibition of cell growth by accumulated polyamines and the decrease of magnesium and ATP Eur J Biochem 1993 208 511 517
Carper SW Willis DG Manning KA Gerner. EW Spermidine acetylation in response to a variety of stresses in Escherichia coli. J Biol Chem 1991 266 12439 12441 2061318
Moore RC Boyle SM Nucleotide sequence and analysis of the speA gene encoding biosynthetic arginine decarboxylase in Escherichia coli J Bacteriol 1990 172 4631 4640 2198270
Tabor CW Tabor H The speEspeD Operon of Escherichia coli: formation and processing of a proenzyme form of s-adenosylmenthionine decarboxylase J Biol Chem 1987 262 16037 16040 3316212
Fukuchi J Kashiwagi K Yamagishi M Ishihama A Decrease in cell viability due to the accumulation of spermidine in the spermidine acetyltransferase-deficient mutant of Escherichia coli J Biol Chem 1995 270 18831 18835 7642535 10.1074/jbc.270.23.13733
Limsuwun K Jones PG Spermidine acetyltransferase is required to prevent spermindine toxicity at low temperatures in Escherichia coli J Bacteriol 2000 182 5373 5380 10986239 10.1128/JB.182.19.5373-5380.2000
Milovic V Turchanowa L Polyamines and colon cancer Biochem Soc Trans 2003 31 381 383 12653643 10.1042/BST0310381
Turchanowa L Dauletbaev N Milovic V Stein J Nonsteroidal anti-inflammatory drugs stimulate spermidine/spermine acetyltransferase and deplete polyamine content in colon cancer cells. Eur J of Clin Invest 2001 31 887 893 11737227 10.1046/j.1365-2362.2001.00901.x
Dubin DT Rosenthal SM The acetylation of polyamines in Escherichia coli J Biol Chem 1960 235 776 782 13818299
Yoshida M K. Kashiwagi G.Kawai Ishihama A Igarashi K Polyamines enhance synthesis of the RNA polymerase sigma 38 subunit by suppression of an amber termination codon in the open reading frame J Biol Chem 2002 277 37139 37146 12147703 10.1074/jbc.M206668200
Farrell MJ Finkel SE The growth advantage in stationary-phase phenotype conferred by rpoS mutations is dependent on the pH and nutrient environment J Bacteriol 2003 185 7044 7052 14645263 10.1128/JB.185.24.7044-7052.2003
Tabor CW Tabor H Transport systems for 1,4-diaminobutane, spermidine and spermine in Escherichia coli. J Biol Chem 1966 241 3714 3723 5330964
Kimberly MM Goldstein JH Determination of pKa values and total proton distribution pattern of spermidine by carbon-13 nuclear magnetic resonance titrations Anal Chem 1981 53 789 793
Meng S Bennett GN Nucleotide sequence of the Escherichia coli cad operon: a system for neutralization of low extracellular pH J Bacteriol 1992 174 2659 2669 1556085
Watson N Dunyak DS Rosey EL Slonczewski JL Olson. ER Identification of elements involved in transcriptional regulation of the Escherichia coli cad operon by external pH. J Bacteriol 1992 174 530 540 1370290
Foster JW Escherichia coli acid resistance: tales of an amateur acidophile Nature Reviews Microbiology 2004 2 898 907 15494746 10.1038/nrmicro1021
Slonczewski JL Foster JW Neidhardt FC, III RC, Ingraham JL, Lin ECC, Low KB, Magasanik B, Reznikoff WS, Riley M, Schaechter M, Umbarger HE pH-regulated genes and survival at extreme pH. Escherichia coli and Salmonella: cellular and molecular biology 1996 1 2 Washington, DC , ASM Press 1539 1549
Iyer R Delcour AH Complex inhibition of OmpF and OmpC bacterial porins by polyamines J Biol Chem 1997 272 18595 18601 9228026 10.1074/jbc.272.30.18595
Samartzidou H Mehrazin M Xu Z Benedik MJ Delcour AH Cadaverine inhibition of porin plays a role in cell survival at acidic pH J Bacteriol 2003 185 13 19 12486035 10.1128/JB.185.1.13-19.2003
Samartzidou H Delcour AH Excretion of endogenous cadaverine leads to a decrease in porin-mediated outer membrane permeability. J Bacteriol 1999 181 791 798 9922241
Gorden J Small PL Acid resistance in enteric bacteria Infect Immun 1993 61 8418063
Maurer LM Yohannes E BonDurant SS Radmacher M Slonczewski JL pH regulates genes for flagellar motility, catabolism, and oxidative stress in Escherichia coli K-12 J Bacteriol 2005 187 304 319 15601715 10.1128/JB.187.1.304-319.2005
Stancik LM Stancik DM Schimidt B Barnhart DM Yoncheva YN Slonczewski. JL pH-dependent expression of periplasmic proteins and amino acid catabolism in Escherichia coli J Bacteriol 2002 184 4246 4258 12107143 10.1128/JB.184.15.4246-4258.2002
Blankenhorn DJ Phillips J Slonczewski JL Acid- and base-induced proteins during aerobic and anaerobic growth of Escherichia coli revealed by two-dimensional gel electrophoresis. J Bacteriol 1999 181 2209 2216 10094700
Newton WA Snell EE Catalytic properties of tryptophanase, a multifunctional pyridoxal phosphate enzyme. Proc Natl Acad Sci USA 1964 51 382 389 14171448
Kirkpatrick C Maurer LM Oyelakin NE Yoncheva YN Maurer R Slonczewski JL Acetate and formate stress: opposite responses in the proteome of Escherichia coli J Bacteriol 2001 183 6466 6477 11591692 10.1128/JB.183.21.6466-6477.2001
Dela Vega AL Delcour AH Cadaverine induces closing of E. coli porins. EMBO J 1995 14 6058 6065 8846798
Sukhodolets MV Cabrera JE Zhi H Jin DJ RapA, a bacterial homolog of SWI2/SNF2, stimulates RNA polymerase recycling in transcription Genes Dev 2001 15 3330 3341 11751638 10.1101/gad.936701
Cabrera JE Jin DJ Growth phase and growth rate regulation of the rapA gene, encoding the RNA polymerase-associated protein RapA in Escherichia coli. J Bacteriol 2001 183 6126 6134 11567013 10.1128/JB.183.20.6126-6134.2001
Forget P Sinaasappel M Bouquet J Dentz NE Smeets C Fecal polyamine concentration in children with and without nutrient malabsorption. J Pediatr Gastroenterol Nutr 1997 24 285 288 9138174 10.1097/00005176-199703000-00010
Small P Blankenhorn D Welty D Zinser E Slonczewski JL Acid and base resistance in Escherichia coli and Shigella flexneri: role of rpoS and growth pH. J Bacteriol 1994 176 1729 1737 8132468
Padan E Shuldiner S Molecular physiology of Na+/H+ antiporters, key transporters in circulation of Na+ and H+ in cells. Biochim Biophys Acta 1994 1185 129 151 8167133
Yohannes E Barnhart DM Slonczewski JL pH-dependent catabolic protein expression during anaerobic growth of Escherichia coli K-12 J Bacteriol 2004 186 192 199 14679238 10.1128/JB.186.1.192-199.2004
|
16223443
|
PMC1274320
|
CC BY
|
2021-01-04 16:03:40
|
no
|
BMC Microbiol. 2005 Oct 13; 5:59
|
utf-8
|
BMC Microbiol
| 2,005 |
10.1186/1471-2180-5-59
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-631622570210.1186/1471-2180-5-63Research ArticleM protein typing of Thai group A streptococcal isolates by PCR-Restriction fragment length polymorphism analysis Yoonim Nonglak [email protected] Colleen [email protected] Chulabhorn [email protected] Michael F [email protected] Sumalee [email protected] Department of Microbiology, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand2 Department of Pediatrics, Faculty of Medicine, Chiang Mai University, Chiang Mai, Thailand3 Queensland Institute of Medical Research, 300 Herston Road, Herston, Brisbane, QLD, Australia2005 16 10 2005 5 63 63 8 6 2005 16 10 2005 Copyright © 2005 Yoonim et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Group A streptococcal (GAS) infections can lead to the development of severe post-infectious sequelae, such as rheumatic fever (RF) and rheumatic heart disease (RHD). RF and RHD are a major health concern in developing countries, and in indigenous populations of developed nations. The majority of GAS isolates are M protein-nontypeable (MNT) by standard serotyping. However, GAS typing is a necessary tool in the epidemiologically analysis of GAS and provides useful information for vaccine development. Although DNA sequencing is the most conclusive method for M protein typing, this is not a feasible approach especially in developing countries. To overcome this problem, we have developed a polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP)-based assay for molecular typing the M protein gene (emm) of GAS.
Results
Using one pair of primers, 13 known GAS M types showed one to four bands of PCR products and after digestion with Alu I, they gave different RFLP patterns. Of 106 GAS isolates examined from the normal Thai population and from patients with GAS-associated complications including RHD, 95 isolates gave RFLP patterns that corresponded to the 13 known M types. Only 11 isolates gave RFLP patterns that differed from the 13 known M types. These were then analyzed by DNA sequencing and six additional M types were identified. In addition, we found that M93 GAS was the most common M type in the population studied, and is consistent with a previous study of Thai GAS isolates.
Conclusion
PCR-RFLP analysis has the potential for the rapid screening of different GAS M types and is therefore considerably advantageous as an alternative M typing approach in developing countries in which GAS is endemic.
==== Body
Background
Streptococcus pyogenes or group A streptococcus (GAS) causes a number of clinical manifestations and diseases including sore throat, pyoderma, necrotizing fasciitis, toxic-shock syndrome, and the post-infectious sequelae – rheumatic fever (RF) and rheumatic heart disease (RHD) [1]. RF and RHD are a major health concern worldwide but especially in indigenous communities within developed countries, and in populations of developing countries [2]. The GAS surface M protein is known to prevent opsonophagocytosis and is a major virulence factor in GAS infection [1]. The N-terminal region of the M protein is highly variable between different GAS strains and contains a type-specific moiety, the antigenic variation of which forms the basis for the classical M protein serological typing of GAS. However, there are disadvantages with this method of typing including ambiguities in the results, the emergence of new M types, and the high rates of M protein-nontypeable (MNT) strains [3] largely due to the unavailability of specific typing antisera. As a consequence, there has been a surge of interest in the development of alternative methods for M typing utilizing molecular technologies.
Several methods have been developed for GAS typing such as enzyme electrophoretic polymorphism [4-6], genomic typing methods such as RAPD [7], 16s rDNA typing [8], RFLP analysis [9-12], vir typing [13,14], DNA hybridization using N-terminal sequences of the M protein gene (emm) as oligonucleotide probes [15,16], polymerase chain reaction (PCR)-enzyme linked immunosorbant assay [17], and PCR M typing using type-specific oligonucleotide primers for PCR amplification of the N-terminal region of the emm gene [18]. PCR-RFLP analysis which utilizes PCR to amplify the emm gene amplicons encoding the M protein prior to digestion with restriction endonucleases has been used for specific molecular M typing methods [19-22], as well as multilocus sequence typing (MLST) [23]. N-terminal sequencing of the M protein gene, however, is the most conclusive method for typing of GAS [24,25], This method of typing is not an option in most laboratories in developing countries worldwide, due to limited resources, and therefore an alternative approach is required for the identification of GAS types. The purpose of this study was therefore to analyze the N-terminal regions of the emm gene of Thai GAS isolates using PCR-RFLP analysis. PCR products were digested with an appropriate restriction enzyme and the fragments analyzed by polyacrylamide gel electrophoresis (PAGE).
Results and discussion
The N- and C-terminal region of the emm gene was amplified by PCR from all of the 119 GAS isolates. The amplicons consisted of one to four bands which varied from approximately 450 to 1200 bp depending on the M types and GAS strains (Fig. 2A). However, some M types had similar sizes of their emm amplicons, but gave distinct bands after digestion with the restriction enzyme, Alu I. This revealed RFLP patterns consisting of two to nine distinct fragments, which ranged from approximately 25 to 700 bp depending on the variability of DNA sequences in the emm amplicons of different M types (Fig. 2B). The RFLP patterns of 95 unknown GAS M types corresponded to 13 known M types (Table 1, Fig. 2B). The M types of all isolates represented by the PCR-RFLP patterns were confirmed by DNA sequencing analysis. All isolates were sequenced independently of the PCR-RFLP analysis which was carried out blinded. The two sets of data were then compared. Only 11 isolates gave patterns that did not corresponded to the 13 known M types. The isolates that could not be identified by PCR-RFLP analysis were also analyzed by DNA sequencing. Six M types that differed from the 13 known M types were obtained from the DNA sequencing analysis (M4, M33, emm58.5, M102, M109 and M89). Although these M protein DNA sequences were ≥95% similar to the data in GenBank, we found both point mutations and deletions in these isolates. M89 has a single-base substitution (T→C) that resulted in a new Alu I restriction site whereas M4 and M102 have 21-base and 33-base deletions, respectively. Alignment of the amino acid sequences of M4 and M102 with reference strains in GenBank, demonstrated ≥95% similarity (Fig. 3).
For decades serotyping has been the method of choice for GAS M typing. However, serotyping is time consuming and it is often difficult to produce high-titer M type-specific antisera, and therefore this technique is limited to a few laboratories in the world. To overcome these limitations, genomic typing methods have been developed [7-12], as well as specific M typing methods [15-18]. RFLP analysis of the emm gene revealed different RFLP patterns among GAS isolates with the same M type [19-22]. Therefore, it could potentially be used to differentiate among isolates with the same M type, but which are not clonal, and may originate from different geographical locations or populations.
In this study, we applied PCR-RFLP analysis for GAS M typing. PCR products derived from the amplification of the N- and C-terminal region of the emm gene and digested with the restriction enzyme, Alu I, produced different RFLP patterns among different M types. Ninety five clinical isolates could be compared with known M types by their RFLP patterns. Only 11 isolates had novel RFLP patterns and required sequencing to confirm the M types. Our results showed that M93 GAS was the most common GAS strain isolated from the population studied with 22 isolates (20.75%) having this type, and is consistent with a previous study of Thai GAS isolates [3]. Other M types that represented more than 10% of the Thai GAS isolates were M1 and M44/61, whereas M66, emm58.5, M89 and M102 were quite rare (0.94%). Furthermore, the specificity of the PCR-RFLP analysis was confirmed by sequencing of the N-terminal region of the emm gene. The majority of isolates were sequenced for each M type. M4, M33, M48, emm58.5, M66, M81, M89, M102, M109, M112 and st11014 were sequenced in all isolates. M1, M11, M12, M25, M44, M49, M93 and M104 were sequenced in 8 of 16, 7 of 8, 2 of 3, 3 of 10, 3 of 11, 1 of 2, 11 of 22 and 2 of 3 isolates, respectively. All isolates were sequenced independently of the PCR-RFLP analysis which was carried out blinded. The two sets of data were then compared. The results obtained from the PCR-RFLP analysis were in agreement with the sequencing data.
In comparison with other genotyping methods based on RFLP analysis [8-10,19-22], the PCR-RFLP analysis used here has some technical advantages. Only one pair of PCR primers is required for the amplification of all of the GAS isolates, and the digested PCR products can be discriminated using standard PAGE which is easy to perform, interpret and is less time consuming. In addition, compared with sequencing analysis [24,25], the PCR-RFLP analysis is technically less demanding and more economical. Therefore, with this simple and rapid protocol, GAS typing can be used in any laboratory in which PCR is routinely used, and is particularly useful for typing large numbers of GAS isolates. However, our findings are based on a relatively small number of emm types isolated in a relatively limited geographic region over a relatively short period of time. Greater geographic and temporal diversity may result in greater clonal diversity thus complicating interpretation of RFLP patterns. Therefore, we suggest that each laboratory should make their own reference RFLP patterns from M types that circulate in their region. However, there are several problems that may occur such as point mutation and deletion. These result in the strains that can not be identified. In this case, it should then be typed by sequence analysis and added these variants into their reference RFLP patterns. In addition, to confirm validation of their results, sequencing a subset of their analyzed isolated should be done periodically.
Conclusion
PCR-RFLP analysis is a rapid, economical and practical genotyping method for determining the M type of GAS. It is therefore particularly suited as a method of choice in developing countries in which GAS is endemic and the majority of GAS isolates are M protein-nontypeable by conventional serological typing.
Methods
Bacteria
One hundred and six strains of GAS isolated from the normal Thai population, patients with sore throat, RHD or impetigo from 1985, 1990, 1995, 2000, 2003 and 2004, and 13 known GAS M serotypes, were included in the study.
DNA isolation and PCR
DNA was isolated from GAS based on the method previously described [3]. The sense primer (CAGTATTCGCTTAGAAAATTAAAA) was derived from the conserved leader sequence of emm gene [26]. The antisense primer (CCCTTACGGCTTGC TTCTGA) was derived from the C-repeat region of the emm gene [3]. The PCR conditions were as follows: denaturation at 94°C for 30 s, annealing at 45°C for 30 s, and extension at 72°C for 2 min for 35 cycles. These primers were also used for sequencing analysis. Using these primers, up to four emm gene amplicons can be generated for a particular GAS strain following PCR amplification depending on the number of C-repeat regions in the M protein (Fig. 1).
PCR-RFLP analysis
PCR-RFLP analysis can be used to type different emm genes. PCR amplification and digestion with appropriate restriction enzymes is expected to give different RFLP patterns among GAS. Based on the emm gene sequences in the GenBank database and analysis using the restriction mapper program [27], we chose the restriction enzyme, Alu I, because almost all sequences contained at least one Alu I site in different positions in their sequences. Therefore, digestion with Alu I is likely to give different RFLP patterns among GAS M types. Following PCR, the products (emm amplicons) were analyzed by electrophoresis on a 1% agarose gel in TBE buffer. The sizes of PCR products were compared with the 100 bp standard marker and recorded following staining with ethidium bromide. PCR products were then partially purified by ethanol precipitation prior to quantitation and digestion with the Alu I restriction enzyme according to the manufacturer's specifications. Digested PCR products were separated on a 15% polyacrylamide gel in TBE buffer and compared with the 100 bp standard marker following staining with ethidium bromide. A log-linear standard curve was initially generated in the Excel program using the known marker sizes versus gel running distance. The equation from the standard curve was then used to estimate the sizes of experimental bands.
DNA sequencing analysis
PCR products were sequenced using the ABI Dye Terminator Cycle Sequencing Ready Reaction Kit according to the manufacturer's instructions (The Perkin-Elmer Corporation) and analyzed using an ABI 310 automated sequencer (The Perkin-Elmer Corporation). DNA sequences were transferred to the DNASIS program for sequence comparison between isolates. The BLAST 2 program (NCBI) was used to determine the levels of DNA homology with published sequences in the GenBank database.
Authors' contributions
NY carried out the microbiologic experiments, performed the molecular genetic analysis, participated in the sequence alignment, interpretation of data and drafted the manuscript. CO contributed to the editing of the manuscript. CP provided clinical specimens and clinical support. MFG provided PCR primers and molecular reagents. SP conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by the Royal Golden Jubilee Ph.D. Program, the Thailand Research Fund, grant No. PHD/00100/2541.
Figures and Tables
Figure 1 Diagrammatic representation of the emm1 and emm25 genes of M1 and M25 proteins, respectively, showing the different A-, B- and C-repeat regions encoding the M protein. The location of the sense and antisense primers used for PCR amplification of the emm gene are indicated as well as the different emm gene amplicons (PCR products) that arise depending upon the number of C-repeat regions within the emm gene.
Figure 2 Agarose gel electrophoresis of emm gene amplicon PCR products (A), and polyacrylamide gel electrophoresis RFLP patterns using Alu I restriction enzyme (B). The size differences of PCR products and RFLP patterns of M1, M11, M49, M12, M25, M48, M112, st11014, M44/61, M93, M81, M104 and M66 are shown in lanes 1–13, respectively. Lane M represents a 100 bp ladder.
Figure 3 Alignment of amino acid sequences from U768 with M4 (emm4 gene) and P354 with M102 (emm102 gene) in GenBank.
Table 1 PCR products and Alu I restriction fragment sizes and distribution in the Thai GAS isolates studied.
M type Size of PCR product (bp) Size of restriction fragment (bp) No. (%) of isolates from PCR-RFLP analysis M type Size of PCR product (bp) Size of restriction fragment (bp) No. (%) of isolates from PCR-RFLP analysis
M1 1071 354 16 (15.09%) M49 450 336 2 (1.89%)
892 257 252
744 218 196
592 147 135
110 89
87 Emm58.5 545 370 1 (0.94%)
47 185
37 M66 566 252 1 (0.94%)
29 450 206
M4 747 314 2 (1.89%) 148
609 174 123
476 57 M81 649 322 7 (6.60%)
47 493 185
M11 710 593 8 (7.55%) 375 127
592 434 84
450 147 62
117 M89 404 350 1 (0.94%)
M12 1121 349 3 (2.83%) 77
116 M93 679 254 22 (20.75%)
93 540 227
78 411 168
59 130
51 93
M25 649 247 10 (9.43%) 53
540 203 M102 420 332 1 (0.94%)
148 92
51 M104 450 210 3 (2.83%)
37 146
M33 812 347 3 (2.83%) 118
670 168 M109 509 386 3 (2.83%)
541 94 168
74 M112 450 435 5 (4.72%)
M44/61 450 212 11 (10.38%) 104
133 st11014 744 596 5 (4.72%)
M48 710 162 2 (1.89%) 620 87
592 121 493 29
471 81
==== Refs
Fischetti VA Streptococcal M protein: molecular design and biological behavior Clin Microbiol Rev 1989 2 285 314 2670192
Bisno AL Group A streptococcal infections and acute rheumatic fever N Engl J Med 1991 325 783 793 1870652
Pruksakorn S Sittisombut N Phornphutkul C Pruksachatkunakorn C Good MF Brandt E Epidemiological analysis of non-M-typeable group A Streptococcus isolates from a Thai population in northern Thailand J Clin Microbiol 2000 38 1250 1254 10699034
Musser JM Gray BM Schlievert PM Pichichero ME Streptococcus pyogenes pharyngitis: characterization of strains by multilocus enzyme genotype, M and T protein serotype, and pyrogenic exotoxin gene probing J Clin Microbiol 1992 30 600 603 1551976
Haase AM Melder A Mathews JD Kemp DJ Adams M Clonal diversity of Streptococcus pyogenes within some M-types revealed by multilocus enqyme electrophoresis Epidemiol Infect 1994 113 455 62 7995355
Bert F Picard B Lambert-Zechovsky N Goullet P Identification and typing of pyogenic streptococci by enzyme electrophoretic polymorphism J Med Microbiol 1995 42 442 51 7791210
Seppala H Qiushui H Osterblad M Huovinen P Typing of group A streptococci by random amplified polymorphic DNA analysis J Clin Microbiol 1994 32 1945 1948 7989547
Stanley J Linton D Desai M Efstratiou A George R Molecular subtyping of prevalent M serotypes of Streptococcus pyogenes causing invasive disease J Clin Microbiol 1995 33 2850 2855 8576332
Patric Cleary P Kaplan EL Livdahl C Skjold S DNA fingerprints of Streptococcus pyogenes are M type specific J Infect Dis 1988 158 1317 1323 3058817
Bingen E Denamur E Lambert-Zechovsky N Boissinot C Brahimi N Aujard Y Blot P Elion J Mother-to-infant vertical transmission and cross-colonization of Streptococcus pyogenes confirmed by DNA restriction fragment length polymorphism analysis J Infect Dis 1992 165 147 50 1345793
Desai M Tanna A Wall R Efstratiou A George R Stanley J Fluorescent amplified-fragment length polynorphism analysis of an outbreak of group A streptococcal invasive disease J Clin Microbiol 1998 36 3133 3137 9774552
Desai M Efstratiou A George R Stanley J High-resolution genotyping of Streptococcus pyogenes serotype M1 isolates by fluorescent amplified-fragment length polymorphism analysis J Clin Microbiol 1999 37 1948 1952 10325352
Gardiner D Hartas J Currie B Mathews JD Kemp DJ Sriprakash KS Vir typing: a long-PCR typing method for group A streptococci PCR Methods Appl 1995 4 288 293 7580916
Gardiner DL Sriprakash KS Molecular epidemiology of impetiginous group A streptococcal infections in aboriginal communities of northern Australia J Clin Microbiol 1996 34 1448 1452 8735096
Kaufhold A Podbielski A Baumgarten G Blokpoel M Top J Schouls L Rapid typing of group A streptococci by the use of DNA amplification and non-radioactive allele-specific oligonucleotide probes FEMS Microbiol Lett 1994 119 19 25 8039659 10.1016/0378-1097(94)90386-7
Penney TJ Martin DR Williams LC de Malmanche SA Bergquist PL A single emm gene-specific oligonycleotide probe does not recognise all members of the Streptococcus pyogenes M type 1 FEMS Microbiol Lett 1995 130 145 9 7649435 10.1016/0378-1097(95)00195-B
Saunders NA Hallas G Gaworzewska ET Metherell L Efstratiou A Hookey JV George RC PCR-enzyme-linked immunosorbent assay and sequencing as an alternative to serology for M-antigen typing of Streptococcus pyogenes J Clin Microbiol 1997 35 2689 2691 9316938
Vitali LA Zampaloni C Prenna M Ripa S PCR m typing: a new method for rapid typing of group a streptococci J Clin Microbiol 2002 40 679 681 11825995 10.1128/JCM.40.2.679-681.2002
Stanley J Desai M Xerry J Tanna A Efstratiou A George R High-resolution genotyping elucidates the epidemiology of group A streptococcus outbreaks J Infect Dis 1996 174 500 506 8769606
Dicuonzo G Gherardi G Lorino G Angeletti S De Cesaris M Fiscarelli E Bessen DE Beall B Group A streptococcal genotypes from pediatric throat isolates in Rome, Italy J Clin Microbiol 2001 39 1687 1690 11325974 10.1128/JCM.39.5.1687-1690.2001
Perea-Mejia LM Inzunza-Montiel AE Cravioto A Molecular characterization of group A Streptococcus strains isolated during a scarlet fever outbreak J Clin Microbiol 2002 40 278 280 11773132 10.1128/JCM.40.1.278-280.2002
Beall B Facklam RR Elliott JA Franklin AR Hoenes T Jackson D Laclaire L Thomson T Viswanathan R Streptococcal emm types associated with T-agglutination types and the use of conserved emm gene restriction fragment patterns for subtyping group A streptococci J Clin Microbiol 2002 40 278 280 11773132 10.1128/JCM.40.1.278-280.2002
Enright MC Spratt BG Kalia A Cross JH Bessen D Multilocus sequence typing of Streptococcus pyogenes and the relationships between emm type and clone Infection and Immunity 2001 69 2416 2427 11254602 10.1128/IAI.69.4.2416-2427.2001
Beall B Facklam R Thompson T Sequencing emm-specific PCR products for routine and accurate typing of group A streptococci J Clin Microbiol 1996 34 953 958 8815115
Brandt CM Spellerberg B Honscha M Truong ND Hoevener B Lutticken R Typing of Streptococcus pyogenes strains isolated from throat infections in the region of Aachen, Germany Infection 2001 29 163 165 11440388 10.1007/s15010-001-1106-x
Manjula BN Khandke KM Fairwell T Relf WA Sriprakash KS Heptad motifs within the distal subdomain of the coiled-coil rod region of M protein from rheumatic fever and nephritis associated serotypes of group A streptococci are distinct from each other: nucleotide sequence of the M57 gene and relation of the deduced amino acid sequence to other M proteins J Protein Chem 1991 10 369 384 1781883 10.1007/BF01025251
RestrictionMapper
|
16225702
|
PMC1274321
|
CC BY
|
2021-01-04 16:03:40
|
no
|
BMC Microbiol. 2005 Oct 16; 5:63
|
utf-8
|
BMC Microbiol
| 2,005 |
10.1186/1471-2180-5-63
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-321621910210.1186/1472-6947-5-32Research ArticleMedical record linkage in health information systems by approximate string matching and clustering Sauleau Erik A [email protected] Jean-Philippe [email protected] Antoine [email protected] Service des études et applications de l'information médicale (SEAIM), Hospital, 87 Ave d'Altkirch, F68051 Mulhouse, France2005 11 10 2005 5 32 32 6 5 2005 11 10 2005 Copyright © 2005 Sauleau et al; licensee BioMed Central Ltd.2005Sauleau et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Multiplication of data sources within heterogeneous healthcare information systems always results in redundant information, split among multiple databases. Our objective is to detect exact and approximate duplicates within identity records, in order to attain a better quality of information and to permit cross-linkage among stand-alone and clustered databases. Furthermore, we need to assist human decision making, by computing a value reflecting identity proximity.
Methods
The proposed method is in three steps. The first step is to standardise and to index elementary identity fields, using blocking variables, in order to speed up information analysis. The second is to match similar pair records, relying on a global similarity value taken from the Porter-Jaro-Winkler algorithm. And the third is to create clusters of coherent related records, using graph drawing, agglomerative clustering methods and partitioning methods.
Results
The batch analysis of 300,000 "supposedly" distinct identities isolates 240,000 true unique records, 24,000 duplicates (clusters composed of 2 records) and 3,000 clusters whose size is greater than or equal to 3 records.
Conclusion
Duplicate-free databases, used in conjunction with relevant indexes and similarity values, allow immediate (i.e.: real-time) proximity detection when inserting a new identity.
==== Body
Background
Because of the fast growth of communication protocols (Internet technologies amongst them), health services are undergoing a large paradigm change: a shift from institution-centred care to consumer-centred care. Unambiguous identification of patients is a critical success factor for health care reform and for the provision of speedy, safe, high quality, comprehensive and efficient health care. We may consider more complete information on which to base potentially life-critical clinical decisions and less wasted time and less inconvenience as a result of hunting for information and/or re-gathering as being amongst the benefits of positive identification. However, client information flows are often substantially limited by an inability to positively identify the subjects of care and to locate their relevant details amongst an extensive array of data repositories which may be unlinked, duplicated and catalogued in different ways. Health workers are now confronted with "health records" for each patient dealing with contacts in a hospital, with a general practitioner, etc. If every contact with the patient is an opportunity to collect information, the information noted must, within an identification domain, be perfectly linked to the patient and identified within the domain by a unique identification number (UID) and a minimum profile (i.e. a set of features such as name, first name, gender and date of birth). This UID and profile then constitute the identity of the subject. When matching different identification domains, profiles are compared in order to track the same individuals. The ultimate aim is a complete "data reconciliation", meaning the coherent association between a unique identity and several other features like medical data, demographic data, ... The expression "record linkage" refers to the use of an algorithm technique to match records from different datasets which correspond to the same statistical unit [1,2]. But whatever the chosen method is, the final decision remains in the hands of a human specialist. Computer logic, based upon Boolean operations can basically tell us if two values are equal ("True or False", "1 or 0"), but it is limited if the answer has to be a bit more fuzzy. The decision-making ability of the human mind is still difficult and time consuming to replicate.
Medical information, and by extension identity records, has to be over-time stable, mainly due to legal requirements and for reasons of traceability. So we must be sure, until a final decision is made by the specialist, not to alter the original data in any way during the linkage process. It would be useful, therefore, when comparing records, to process atomic parts of them (such as name, first name, date of birth ...) in order to avoid the most common mistakes made during data entry. This step, called standardization, will ensure a greater effectiveness of the comparison algorithm. Most of the mistakes are a result of not adhering to data entry guidelines (i.e. abbreviations, accented letters, date format...), which are too often the responsibility of the operator. Some others appear when attempting to match records from different data sources (or different identification domains), ruled by heterogeneous and non-compatible data-entry guidelines. This standardization step is mandatory and its lifetime should not exceed the comparison operation.
In a database containing n records, record linkage is often described as an O(n2) complexity problem, due to its Cartesian aspect. The brute force approach, comparing each record to every other record, requires (n2-n)/2 comparisons. This approach, while being the most reliable (as no record is missed during the comparison process) is also the most time consuming and least effective (CPU load speaking). To reduce the number of comparisons, indexation of databases by blocking techniques is used. The data sets are split into smaller blocks and only records within the same blocks are compared.
Whichever blocking technique is chosen, the principle of a record linkage process is to try to pair similar records. But how is the decision which qualifies two different records as "duplicates" made? A basic way is to compare each atomic part of a sample record (i.e. the fields) to its counterpart in the reference record. If a binary equality is obtained, we can then assume that the entities are the same. The main drawback with such a method is that keystroke mistakes or subtle spelling changes are ignored, even after standardization. To be more efficient, we must enhance the comparison principle. While a binary operator returns a Boolean value (1 if the fields are strictly the same, 0 if they are different), a similarity operator will return a score ranging from 0 to 1, showing fields proximity and quantifying their differences. The higher the result, the nearer the field values are. Such a similarity operator is based on approximate string matching methods, so non-literal data fields such as dates or numbers have to be converted before processing. The main approach of approximate string matching has always, until now, been based on the edit-distance [3,4], the oldest kind of algorithm according to Navarro [5,6]. For example, Levenhstein's distance [7] is the minimum number of operations on individual characters like substitutions, insertions and deletions, needed to transform one string to another. It remains the most flexible although it is no longer the most effective, at least in "text retrieval" cases. The Smith-Waterman algorithm [8] is one of the most popular algorithms for edit-distance. Much of its power is due to its ability to introduce gaps in the records (sequence of non-matching symbols). Hence, the use of such an algorithm on each field of records allows us to calculate the same number of atomic similarities (noted aS in the sequel) linking the fields of the sample and the reference records. Such atomic values are then combined in a weighted mean to obtain a record to record similarity or global similarity (gS) value. If one uses k fields for comparing records then where wj are the weights (corrected so that the sum of all of them is 1) in the mean. Indeed, not all the atomic similarities have the same discriminative power when comparing two records. For example, gender is less reliable than date of birth: two records with the same gender have less chance of representing the same identity than two records with the same date of birth. The weights are obtained, for example, by the Expectation-Maximization algorithm [9,10]. Several general algorithms for record linkage have been written since the 70's, most of them in biomedical papers in order to perform epidemiological studies [11]. The seminal theoretical paper on record linkage by Felligi and Sunter [12] has, as its goal, the division of record pairs into linked pairs (designated matches), possible linked pairs (pairs for which human oversight, also known as clerical review, is needed) and non-linked pairs (non-matches). Their classification rule is based on the comparison of the common fields of the two sets of records. To be a little more specific, let us suppose that all the records of a dataset are distributed into two equal sets, say A and B, and that furthermore we create three more sets: for matched pairs, for non-matched pairs and a set of possible matched pairs. If each record has k fields, it is possible to define an agreement k-vector γ (all the atomic similarities for example) between a record of the set A and a record of the set B. The main issue of the Felligi-Sunter theory is to define an optimal linkage rule (for given levels of type I and type II errors), where optimality is defined as minimising the probability of classifying a pair in the set of possible matched pairs. Assuming conditional independence of the k components of γ, the decision rule is a function of where those weights wj are . Here m(γj) and u(γj) are the conditional probability of observing γj given that the pair is respectively a true matched pair and a true non-matched pair. For example, Jaro [13,14] uses the EM algorithm to estimate these weights (or at least the m(γj)s).
Results provided by those techniques of record linkage consist of a list of paired records linked by a global similarity score. Whilst being a good starting point, such a presentation is far from being the most efficient for clerical review. Specialists have to deal with complex patterns of similarities between several records split-down into a list of similarities between couples of records. We can then try to build clusters of duplicates, "n-plicates". A weak definition of a cluster could be "a set of entities which are alike and entities from different clusters are not alike". Methods for clustering can be divided into hierarchical, graph-based (equivalent to the graph partitioning), model-based or mixed methods [15,16]. An additional stage in the representation of the records with their similarities may be necessary. From a representation of records in clusters, we can switch to a representation of graphs. Following the graph theory, the clusters become undirected sub-graphs with vertices (records) and edges (global similarity value). It would be useful, for the visual comfort of the user, to represent these clusters in a reliable manner with vertices and edges, whose lengths are proportional to the strength of the relation between the vertices (global similarity). The most used methods for graph drawing are force-directed methods [17-19]. Generally, these methods view a graph as a system of particles (vertices) with forces acting between them (edges). They seek a state of balance where the sum of the forces acting on each particle is zero. All these methods are relatively simple to implement, heuristic improvements can easily be added, but they can be time consuming.
But specialists also have to deal with the possibility that some of the relations between some record pairs may be unreliable. For example, two different paths between two records can exist: one path with only edges labelled "duplicate" and one with at least one edge labelled "non-duplicate". Hence, some of the sub-graphs are complete (each of the vertices is linked with each other in the sub-graph by an edge), like the left panel on Figure 1. But some are incomplete (n vertices but fewer than n(n-1)/2 edges), like the right panel on Figure 1. If we choose not to enforce transitivity of the relation "is a duplicate of", unlike some other authors [20,21], the vision of this sub-graph as a "n-plicate" is not so straightforward. Thus, it would be interesting to consider complete sub-graphs within this incomplete one, and then to consider each of these last sub-graphs as "n-plicate". Methods for this result are derived from graphs partitioning algorithms [22], in an unsupervised situation (the number of sub-graphs is, a priori, unknown). They go on to delete several edges in order to isolate groups of vertices. Such a cut is said to be of minimum weight, if the sum of the weights on the edges is the minimum necessary to isolate such groups of vertices. The gain of a vertex is the difference between the sum of the similarities with vertices of other sub-graphs and the sum of the similarities with vertices of its own sub-graph. Notably, the gain is the cost when a vertex changes its sub-graph.
Figure 1 Example of a complete graph (left panel) and of an incomplete graph (right panel).
Our very general aim is to provide specialists with a reliable presentation of a set of several potentially duplicated records in a data set with a score showing how similar records in this homogeneous set are. These specialists may be responsible for merging similar records or for avoiding duplicate entry. This method, then, deals with two different situations: batch and real-time data browsing.
Methods
The method used for achieving our goal comprises three steps: the first one is a pre-processing of data in order to eliminate the most common errors and to present this data in a convenient way; the second is to compare records and the third is to organize the records. We will focus on showing its application upon a single data source, even if the proposed technique can be applied to link foreign domain records (i.e. a sample data source to a reference one).
A particularity in Anglo-Saxon countries and in France (but not in all Latin countries) is the use by a married woman of her husband's name: her birth name is no longer used but "replaced" by her married name. Because of the instability of this married name (divorced persons), the birth name is always recorded in a patient database to guarantee an over-time reference. But a high risk of errors exists while encoding names between birth name and married name. First of all, we define a typical patient-identity record as constituting a unique identification number and a minimum profile: birth name, married name, first name, gender and date of birth. We are working, here, on the patient database of our hospital as it was on the 1 st July 2003. This paper reports the use of our proposed method on this database from which we want to eradicate duplicates. Each of the variables belonging to the patient profile is already encoded in formatted fields within the database and thus does not need any parsing or re-allocating.
Accents, hyphens and apostrophes may be uncommon in English, but are very important and may appear often in French or in other Latin languages. Data entry of patient identities is, for the most part, under human control and subject to missing data or data entry errors (such as typographical or keystroke mistakes), non standard abbreviations or differences in detailed schemas of records from multiple databases. We first of all try to standardize the strings before comparing them with the algorithm: replacing all accented letters with the same unaccented letter, converting all strings in uppercase, replacing punctuation signs with a space, discarding all non-informative spaces (double spaces, spaces at the end of strings...), discarding all spaces in names and, in double first names (quite different from first name and initial in English), replacing the first part by its first letter (for example "Jean-Philippe" becomes "J PHILIPPE").
Baxter et al [23] describe four different blocking methods. The standard blocking method clusters records that share an identical blocking key, composed of one or more attributes of each record, like the postcode or phonetic encoding of names. The sorted neighbourhood method [24] sorts the records based on a sorting key and then moves a window of fixed size sequentially over the sorted records. Records within the window are then compared with each other. In the bigram indexing method [25], the blocking key values are converted into a list of bigrams, and sub-lists of all possible permutations are built using a threshold. The canopy clustering method [26] creates overlapping subsets, called "canopies" composed, for each record, of all the records within a certain loose threshold distance. For indexing the database, we chose a blocking method borrowed from this last technique. For each record a certain number of blocking keys is calculated, each one being three bytes in length which allow us to scan, for a given record, not the entire database but only the records whose blocking keys correspond. The structure of these keys is very simple. They are comprised of the first three characters of the standardized string (normal version) and the last three characters (inverted version). These two keys are computed for birth name (i.e. maiden name for a married woman), first name and, if it exists, married name. The use of inverted keys allows us to compare strings with an error on the first few characters, which seems to be common in patient databases. Furthermore, the speed of the algorithm remains high and the window screening remains feasible. Two records sharing the same canopy are compared if they meet one of the following two conditions:
• Their dates of birth and their gender are the same;
• Their gender are the same and their blocking keys on birth name and married name (if necessary) meet at least one of these two conditions:
1. their normal or reversed blocking keys for birth names or for married names are the same
2. the normal (respectively the reversed) blocking key for birth name in one of the two records is the same as the normal (respectively the reversed) blocking key for married name in the other record.
A comparison is made for each atomic part (i.e. field) of the identity record, giving an atomic similarity value for each. In order to compare two records, we run the approximate string matching algorithm on the different matching variables and write out six atomic similarities: three for the comparisons of the two dates of birth, the two first names and the two birth names, one, if necessary, between the married names (otherwise the atomic similarity is considered as missing), one if necessary between the married name of the first record and the birth name of the second (otherwise missing) and one, if necessary, between the birth name of the first record and the married name of the second (otherwise missing). From a series of different algorithms [27], we chose to use an algorithm we call the Porter-Jaro-Winkler algorithm [28], an improvement of an initial algorithm [13,14]. Its basic principle is to compare two strings C1 and C2, and to compute a similarity rate. It considers letters in the first string which are within half the length of the second string. In addition, some errors are penalized less harshly (visual scanning errors, keypunch errors and errors at the end of the string with respect to the errors at the beginning of the string). It is a looser matching criterion than edit-distance and, unlike the other techniques, it is not an "all or nothing" computation, but gives a degree of match. Noting L1 and L2, C1 and C2's respective lengths, the similarity rate is obtained by , where Nc is the number of common characters in the two strings and Nt the number of transpositions. From this basic formula, several improvements are successively obtained, but we use only two of them:
• The number of "similar" characters Ns replaces the number of common characters Nc. A character is decreed "similar" to another if it belongs to one of 36 couples defined in the algorithm, which are for example (X, K), (5, S), (O, Q);
• Additional weight is given to the similarity when the first four characters of C1 and C2 are similar or identical. Typographical errors are often not located at the beginning of the string, but rather in its body;
According to the authors, a similarity of less than 0.7 means two different strings and more than 0.95 means two similar strings (common strings exhibit a rate of 1). We then add different refinements to this algorithm. We weight the atomic similarities of names (birth names and married names) depending on their frequency in the database: the more frequent a name is, the less the similarity seems to be credible. In this early report of our work, we have created just two categories according to the name frequency (approximately 5 per 10,000) and under-weight if necessary the similarity of 0.05. This 5 per 10,000 cut-off was decided from the graph of the names frequency, which has an inverse function aspect with the end of the strong decrease about 5 per 10,000. We also take into account the difficulty of comparing two strings of different lengths which present a strong similarity on the shorter length. For instance, the names "ABC" and "ABCDEFG" are quite different but have an atomic similarity according to the initial algorithm of 1. In this case, we calculate a similarity like the initial one minus 0.01 times the number of additional characters (4 in the above example, yielding an atomic similarity of 0.96). A global similarity is then computed as a weighted mean of atomic similarities. Because of the particular configurations of our records (with or without married names) we use a pragmatic weighting based on common weights used in literature and on our prior knowledge. Once again, we define a few rules:
1. If the two birth names and the two first names are almost the same (i.e.: aS > 0.7) then the date of birth should have a high discriminative weight;
2. If the similarity between the married name of a record and the birth name of a second record is more than the similarity between the two birth names then we suppose that an inversion between the married name and the birth name has occurred.
The weights are standardized to 1. A complete description of this weighting step appears in the Appendix. For example, the simplest situation (without any married name to compare) can be split into two sub-cases: if the birth name's atomic similarity is high (i.e. > 0.7) then the weight for birth names is 1/3, the weight for first names is 1/6, the weight for dates of birth is 1/2, otherwise the three weights are respectively set to 1/2, 1/4 and 1/4. Unlike the Felligi-Sunter probabilistic theory, we chose not to use two thresholds which would classify pairs into linked, non-linked and possibly-linked, but to determine a unique threshold with a set of non-linked pairs and a set of possibly-linked pairs. Before analyzing the database and after multiple experiments, this threshold was set to 0.85.
To achieve the clustering step, because of our confidence in the similarity algorithm, we need a simple, unsupervised (the number of clusters is, a priori, unknown) hard-clustering method (not a probability of classification). Our choice is a greedy agglomerative algorithm with complete linkage to avoid clusters in a chain. For graph drawing, we will choose a very simple version of a force directed algorithm, but in the early version of our work, this functionality is not yet implemented. Then, in order to automate the graph partitioning, several algorithms have been proposed [29-33], they are often heuristics based on seminal algorithms like Kerninghan-Lin or Fiduccia-Mattheyses. Because of our confidence in the previous steps and the use of the Porter-Jaro-Winkler algorithm, a simple algorithm based on the gain is then used to partition incomplete clusters in complete objects (clusters or isolated records).
Results
With 300,859 records in our patient database, the brute force approach would need 45 109 comparisons. By using blocking variables, only 24.8 106 comparisons are computed, which means a 1,825-fold gain. In the database, there are 287,850 different canopies with a maximum canopy size of 11,550 records. In fact, the distribution of the size of the canopies is very left-skewed. If the mean size is 82 and the standard deviation 195, the median is 18 and the 95th percentile is 398. The most frequent size is 7. Each record of the database resides (in average) in 8.9 canopies.
Among 300,859 records, our method detects 38,083 couples, whose global similarity equals or exceeds the decision threshold set to 0.85. The Figure 3 summarises the results of our entire method applied to this data set. Table 1 shows the number among these 38,083 couples with exact concordance in global similarity and in atomic similarities. This table shows that 9,566 couples exhibit a unit value for their global similarity. Even if they have different UIDs, these couples seem to correspond to 9,566 different individuals. The value of the atomic similarities for the fields birth name, first name and date of birth is exactly one in about 70% of the couples but less for married name. For the comparisons, when accurate, between birth name and married name, 3,065 couples match exactly (4%). This means that in 3,065 pairs the married name of a record is exactly the same as the birth name of the other record of the pair, probably indicating an inversion between these names. The Table 2 describes the characteristics of global similarity and atomic similarities among the remaining couples, without perfect concordance. About 29,000 couples exhibit a value of global similarity which is not exactly the unit 1 but a value with a mean (and a median) of 0.92. The distribution is approximately uniform on the interval [0.85 – 1.00], except for a small peak at 0.95 corresponding to the underweight of 0.05 concerning exact matching of records with frequent names. Concerning birth name, first name and date of birth, among about 13,000 remaining couples without a unit value of 1, the mean is about 0.80 but the quartiles show that the similarity on first name is more left-skewed than the similarity on birth name. The similarities between married names are globally higher than the similarities between birth names. Among the 16,013 comparisons remaining in the comparisons between birth name and married name, only about 800 exhibit a similarity higher than 0.75 and the mean of these 16,013 is 0.36.
Figure 2 Percentage of true positive couples by global similarity threshold value.
Figure 3 Summary of the entire linkage procedure.
Table 1 Number of exact concordance in global similarity (gS) and atomic similarities (aS) in the 38,083 identified pairs in a database of 300,859 patients
Atomic similarities
gS BN* MN* BN/MN* First name Date of birth
Pairs 38,083 38,083 38,083 76,166 ** 38,083 38,083
Missing 0 0 31,747 57,088 0 0
Values at 1 9,566 25,990 5,348 3,065 26,017 26,505
in % 25.1 68.2 14.0 4.0 68.3 69.6
* BN = birth name, MN = married name
** Theoretical number of comparisons between BN and MN
Table 2 Characteristics of global similarity (gS) and atomic similarities (aS) in the identified pairs without exact concordance
Atomic similarities
gS BN* MN* BN/MN* First name Date of birth
Pairs 28,517 12,093 988 16,013 12,066 11,578
Mean 0.92 0.79 0.78 0.36 0.77 0.82
Stand. error 0.48 0.20 0.23 0.25 0.17 0.07
Minimum 0.85 0.00 0.00 0.00 0.00 0.65
Percentiles
25th 0.87 0.76 0.62 0.22 0.63 0.77
50th 0.92 0.88 0.90 0.41 0.82 0.85
75th 0.97 0.92 0.94 0.52 0.92 0.88
90th 0.99 0.95 0.96 0.63 0.95 0.88
95th 0.99 0.96 0.96 0.75 0.96 0.88
* BN = birth name, MN = married name
Among those 38,083 couples (T), 19,882 (C) may be found by classical techniques: first three last-name letters exact matching and first three first-name letters exact matching and date of birth matching and gender matching, while detection of the last 18,201 is made possible by using a string similarity algorithm. An in-depth analysis of those 18,201 paired records shows that 9,690 couples (P) are true positives, while 8,511 are false ones. Effectiveness over classical techniques can then be calculated: a minimum of P/(C+P) = 33% more true positive couples are detected than by a classical method (for reasons of convenience, the C couples are all declared to be true positives, while in fact, a few false positives may be present). The proportion of true positives according to this clerical review is shown on Figure 2, by different levels of global similarity (gS by unit of one). As a result, we can now assume that the positive predictive value defined as the number of true positives divided by the total number of linked pairs is (P+C)/T = 78%, which represents a good indicator of accuracy. In Figure 2, we see that the drops, at 0.96, 0.91 and 0.88, reflect name frequency adjustments.
Once the couples have been constituted and their global similarity values calculated, we can now define the relevant graphs. The 38,083 couples represent 27,184 clusters. This final step creates two kinds of clusters: 25,642 complete and 1,542 incomplete graphs. The first ones have all their UIDs (vertices) linked by an edge (global similarity above the threshold). The second ones present some UIDs that are not linked to some of the other cluster components. In the complete graphs there are 23,004 clusters of a size of 2 records, 2,237 of size 3 and 401 of a size greater than 3. In the incomplete graphs there are 1,012 clusters of size 3 and 530 of a size greater than 3. The complete graphs represent 54,443 initial distinct UIDs while the incomplete represent 5,760 UIDs. But before clerical review, our method allows us to state that the 25,642 complete graphs potentially represent 25,642 different individuals. Whatever the threshold within the range 0.85 to 1 is, the mean size for the graphs is constant; about 2.1 for complete graphs and 3.5 for incomplete graphs but the relative number of each kind of graph changes: 200-fold more complete graphs than incomplete for a threshold set to 1, 60-fold for 0.95 and 15-fold for 0.85. The use of the partition algorithm allows us to merge the 1,542 incomplete graphs into several complete graphs. The 1,012 incomplete clusters of size 3 become 1,012 single records and 1,012 couples. The other 530 incomplete graphs become 532 single records, 329 couples, 310 complete graphs of size 3 and 111 complete graphs of a size greater than 3. The 38,083 couples correspond to 60,203 initial UIDs. Out of a total number of UIDs of 300,859, 240,656 UIDs do not form part of a couple, corresponding to 240,656 different individuals. 25,642 individuals come from complete graphs and 3,306 from incomplete graphs. Finally, there are 269,604 different individuals in our database, corresponding to a 90% rate of uniqueness. This means that 1 record in 10 is involved in at least one cluster of "n-plicate" identities.
Discussion
Our method gives those specialists responsible for merging similar records a representation showing how close records in some homogeneous sets are. But real-time detection is also one of the main goals of our proposal: avoiding duplicate entry by alerting the user that several neighbour records already exist, and furthermore, they help in decision making whilst providing proximity values between these similar records. This real-time use would be involved in multi-criteria searches for identities or for the creation of identities. But only a simple and easy to use front-end algorithm and short response time allow this possibility. Response times are closely related to an optimization of the algorithm and especially the blocking part. Its improvement allows the reduction in the number of potential duplicates to be tested by the main algorithm.
According to Porter and Winkler, the use of their algorithm improves the number of duplicates by 30% compared to a binary or semi-binary method [28]. Our results are in accordance with this number. Furthermore, we consider only the true positive identified couples in excess of those identified by the classical methods. Nevertheless, methods other than an approximate string matching can also be considered. For example, the CART method (classification tree-based models) could be used for mapping duplicates and defining clusters of records, which have to be seen not as duplicates but "n-plicates". A logistic model could also be proposed, using for its dependent variable a dummy variable indicating a duplicate or not, and the edit-distance as the independent variable. Some other independent variables can be also used.
Even if our approach is feasible and seems to be useful and reliable, we have to improve our process. The formal Jaro-Winkler algorithm is a domain-independent algorithm, which can be used, without any modification, for a wide range of applications. Our algorithm is no more domain-independent because of our weighting procedure of matching variables, and because of our names frequency based weighting. The classical approach in this context, to solve both problems, is to use the EM algorithm [9,10,34] to retrieve weights driven by the dataset. But the formalisation of using the EM algorithm was developed in the probabilistic Felligi-Sunter theory framework. Our approach is no longer within this framework since our weighted average of atomic similarity has no probabilistic interpretation. Furthermore, the conditional independence assumption of the components of the agreement vector is probably not fulfilled. One main basis of our approach is that all records do not have the same fields because of the potential absence of the married name. Hence the theoretical considerations for the extension of the EM algorithm to our approach are not so straightforward. Instead, other attractive methods seem to be machine-learning in a Bayesian classification framework [35,36]. The aim is to improve a partial automation of the linkage decision process, involving the use of a training data set, in order to instruct a learning algorithm to classify data in link or non-link. Moreover, our method essentially deals with real-time data browsing. In this issue the data set has to be viewed essentially as an evolving database and the machine learning seems to be very accurate: each decision taken by the user can become part of the learning process of the algorithm, instead of a new calculation of the weights by the EM algorithm.
The similarities we use (atomic and moreover global) do not fulfil the conditions necessary in order to qualify as a distance. An interesting condition on distance is the transitivity one, for example between records A, B and C: d(A,B) ≤ d(A,C) + d(C,B). A weighted mean of several distances is indeed a distance unless the weights differ for at least one distance between records. This is the case in our approach (notably due to the difference of treatment between records with married name and records without) but that is also the case with the Porter-Jaro-Winkler algorithm due to the improvement involved when the four first characters of two strings are exactly matched: it can be the case between, for example, record A and C but not the case between records B and C. The condition of transitivity is not sufficient to be really useful in the case of string matching for sparing the number of comparisons between strings. A really interesting property would be the exact knowledge of the relation between d(A,B), d(A,C) and d(C,B). But this knowledge is probably not compatible with the complexity of building an indicator to evaluate string proximity and an indicator of neighbouring records. We noticed, empirically, that in our data set, the transitivity property seems to be fulfilled and hence increases the confidence it is possible to have in the Porter-Jaro-Winkler algorithm.
Here we use, a posteriori, an improvement of window algorithms already mentioned in literature [24,37,38]: records are ordered according to a given criterion and the algorithm, for a given record, is not used on the whole reference dataset, but only on records in a window of a given length. The standard method of detecting exact duplicates in a table is to sort the table and then to check if neighboring tuples are identical. Exact duplicates are guaranteed to be next to each other in the sorted order, regardless of which part of a record the sort is performed on. The idea is to do sorting to achieve preliminary clustering and then to do pairwise comparisons of nearby records. But in this case there is no guarantee that approximate duplicates are all next to each other in the sorted order. In the worst case, they will be found at opposite extremes of the sorted heap. In our case, the criterion is the blocking key and we used the algorithm on all the records sharing the same blocking keys. Much more than a sorted list, the canopy technique we use creates overlapping subsets in which records are compared. In the case of a dataset of size n, divided into b blocks by standard blocking or bigram indexing, the complexity of the record linkage decreases from O(n2) to O(n/b) but with a much larger b with bigram indexing. The complexity is O(wn) with the sorted neighbourhood with a w-size window. For canopy clustering, the number of record pair comparisons is O(f2n2/c) where c is the number of canopies and f the average number of canopies a record belongs to. In our data set, we find that the number of canopies has the same order as the number of records and that each record belongs to about 10 canopies (in mean). Our complexity decreases, then, from O(n2) to O(n), which is less accurate than a bigram indexing complexity. However, the calculation of canopies technique complexity over-estimates this complexity [26]. Actually, it seems to be clear that the most efficient improvement would be to consider not a window but a priority queue. For example Monge [20] proposes a three-step procedure. First, the Smith-Waterman algorithm is used to recognize pairs of approximate duplicates, then the union-find algorithm to keep track of clusters of duplicate records incrementally, as pairwise duplicate relationships are discovered. Thirdly, a priority queue of cluster subsets responds adaptively to the size and homogeneity of the clusters discovered as the database is scanned. All these improvements are nevertheless based on identification algorithms using edit-distance.
Our method adds to the calculation of similarity between couples, a step of clustering and partitioning. To our knowledge, no other published method in the field of record linkage offers these last steps. Hence, for comparing our method with the others, we have to rely on the similarity step even if this is not our "final product". The measurement of the quality of record linkage relies on 4 different values: the number of record pairs linked correctly (true positives), the number of record pairs linked incorrectly (false positives), the number of record pairs unlinked correctly (true negatives) and the number of record pairs unlinked incorrectly (false negatives). These values allow the calculation of different estimates of the performance of an algorithm and notably: the specificity (true negatives divided by the number of true non-match pairs), the negative predictive value (true negatives divided by the total number of non-linked pairs), the sensitivity (true positives divided by the total number of true match pairs, which is the sum of the true positives and the false negatives) and the positive predictive value (true positives divided by the total number of linked pairs). The goal of methods for approximate records matching is to retrieve the truest duplicates possible, even if the price to pay is to get some false positives as well. Thus, the interest is mostly in the sensitivity and in the positive predictive value of methods, corresponding with, respectively, the recall and the precision in the text retrieval field. Furthermore, the number of true negatives represents about 80 or 90 % of the total number of the pairs and any comparisons based on a quality indicator involving this number will be difficult to interpret. Indeed, for example, the specificity of different methods will always (except with bad methods) be close to 1 because the differences in the number of false positives will play a very minor role in the division by the number of true negatives. With respect to the tremendous number of records it is very time consuming, and almost impossible, to review all pairs non-linked by the methods we want to compare with, in order to detect false negatives and true negatives. The sensitivity (and the negative predictive value) is hence very difficult to calculate. Only the positive predictive value is a reliable and easy to calculate indicator for measuring the performance of a method for record linkage. As it is quite impossible to review all the pairs, a solution would be to sample some pairs and just review these ones. But this solution does not seem to be completely accurate as the main quality of sampling is to respect the "data generating process". In the case of duplicate identities generating, this process seems to be much too complicated to be reproduced. The risk then, is to have a biased validation sample and to get a wrong quality indicator. Another solution is to have an efficient external data set with several known characteristics. The underlying idea is to compare an exhaustive and clean (meaning without duplicates, after clerical review) data set with these characteristics, with the subset of the general hospital data set with these same characteristics. This procedure was, for example, (with people remaining anonymous) used with a digestive cancer registry [39].
The most common used final indicator, in the framework of record matching, is the percentage of duplicate identities. But we cannot rely on this indicator as our method builds several clusters of different sizes; each cluster corresponding, after clerical review, to one identity. Hence we use an indicator of uniqueness in the database calculated as the sum of the number of UIDs not involved in any cluster and the number of clusters, divided by the number of initial UIDs. As some clusters are not size 2, this uniqueness is not directly related to the number of duplicate identities (meaning couples whose similarity is above the threshold).
When we identify clusters of n-plicates, we have to decide, furthermore, whether they are real n-plicates. The last step is then to merge, not only on the identities – which is already a problem – but also the medical records relating to these identities. Even if all the global similarities are 1 in a given cluster, identified n-plicates stay "possible n-plicates" and only a very close examination can allow us to discard homonymic identities (in our case, same names, same first name, same date of birth, same gender and same address but different subjects). If merging clients' addresses is not so risky, the problem is quite different in the case of medical records. Complete traceability should be possible for "un-merging" clusters if necessary. The necessity of a human decision for merging is the most important limitation for a completely automatic process.
Using matching variables other than the minimum set we defined would dramatically improve the efficiency of the records matching. For example, parents name or the city of birth is a very discriminatory matching variable. Furthermore, the first name seems to be less reliable than the birth name for comparing records, and difficulties with married names yield a complex algorithm for the calculation of global similarity. Hence, perhaps the minimal set for identifying individuals is neither really reliable nor sufficient.
Appendix
The goal of this section is to detail the weighting procedure of the atomic similarities for building the global similarity as a weighted mean of those same atomic similarities. The Table 3 summarizes this procedure. We call A the first record (reference record) and B the second record (sample record). Each record contains four different fields (birth name, married name, first name and date of birth) whose value cannot be the null string (noted {0}) unless it is the married name. These fields, for example in record A, are respectively noted A(BN), A(MN), A(FN) and A(DB). We call aS the vector of atomic similarities and gS the global similarity. The components of aS are the atomic similarities between two fields: which are noted aS(BN) between the two birth names, aS(MN) between the married names, aS(BN/MN) between the birth name of the first record and the married name of the second, aS(MN/BN) between the married name of the first record and the birth name of the second, aS(FN) between the first names and aS(DB) between the dates of birth. T is the threshold under which two atomic similarities are acknowledged to correspond to two different values of the fields. With reference to Porter-Jaro-Winkler, this threshold is fixed to 0.7. In the procedure there are four main cases, each one is further divided in several sub-cases.
Table 3 Weighting procedure of the atomic similarities (Appendix)
Weights BN MN MN/BN BN/MN FN DB
Case 1
A(MN) = {0} and B(MN) = {0} then
If aS (BN) <T or aS (FN) <T then 2/4 - - - 1/4 1/4
Else 2/6 - - - 1/6 3/6
Case 2
A(MN)! = {0} and B(MN) = {0}
(1) If aS (MN/BN) <aS (BN) then
If aS (BN) <T or aS (FN) <T then 2/4 - - - 1/4 1/4
Else 2/6 - - - 1/6 3/6
(*) Else 1/6 - 1/6 - 1/6 3/6
Case 3
B(MN)! = {0} and A(MN) = {0}, switch A and B and go to (1)
Case 4
A(MN)! = {0} and B(MN)! = {0}
If aS (BN/MN) <aS (BN) and aS (MN/BN) <aS (MN) then 2/7 1/7 - - 1/7 3/7
If aS (BN/MN) >aS (BN) and aS (MN/BN) <aS (MN) then 1/7 1/7 - 1/7 1/7 3/7
If aS (BN/MN) < aS(BN) and aS (MN/BN) >aS (MN) then 1/7 1/7 1/7 - 1/7 3/7
If aS (BN/MN) >aS (BN) and aS (MN/BN) >aS (MN) then 1/8 1/8 1/8 1/8 1/8 3/8
For example, suppose that the atomic similarities for two records A and B are aS(BN) = 0.80, aS(MN/BN) = 0.97, aS(FN) = 0.90 and aS(DB) = 0.92. This case corresponds to a first record with a married name and a second record without a married name but with a birth name which is nearer the married name of the first record than the birth name of this first record (probably an inversion between the two names). In the Table 3, these atomic similarities go with the row highlighted with (*). Hence, the global similarity is calculated as 1/6.0.80 + 1/6.0.97 + 1/6.0.90 + 3/6.0.92 or 0.905. Even if there is a further error with the dates of birth, the global similarity indicates a relatively strong proximity between the two records. Of course, if the errors on dates of birth were more pronounced, the global similarity would quickly decrease.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EAS reviewed papers on the subject and built, with JPP, the methodology. JPP is in charge of all methods implementation. AB supervised this work. All authors read and improved successive drafts.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Mr W.E. Winkler for kindly providing his strings comparator algorithm and for his helpful comments. The revision of this paper has also benefited from comments by the three reviewers for its improvement. The authors also wish to thank Mr Darryl Moran for his helpful reading of the paper.
==== Refs
Belin TR Rubin DB A method for calibrating false match rates in record linkage Journal of the American Statistical Association 1995 90 697 707
Newcombe HB Kennedy JM Record linkage: making maximum use of the discriminating power of identifying information Communications of the ACM 1962 5 563 566 10.1145/368996.369026
Vintsyuk T Speech discrimination by dynamic programming Cybernetics 1968 4 52 58 10.1007/BF01074755
Sellers P The theory and computation of evolutionary distances: pattern recognition Journal of Algorithms 1980 1 359 373 10.1016/0196-6774(80)90016-4
Navarro G Raffinot M Flexible pattern matching in strings 2002 Cambridge, Cambridge University Press
Navarro G A guided tour to approximate string matching Association of Computing Machinery Computing Surveys 2001 33 31 88
Levenhstein V Binary code capable of correcting deletions, insertions and reversals Soviet Physics Doklady 1966 10 707 710
Smith TF Waterman MS Identification of common molecular subsequences Journal of Molecular Biology 1981 14 195 197 7265238 10.1016/0022-2836(81)90087-5
Dempster AP Laird N Rubin DB Maximum likelihood from incomplete data via the EM algorithm (with discussion) Journal of the Royal Statistical Society, series B 1977 39 1 38
Winkler WE Using the EM algorithm for weight computation in the Fellegi-Sunter model of record linkage Statistical research report series N°05 2000 US bureau of census, Washington DC
Newcombe HB Handbook of record linkage: methods for health and statistical studies, administration and business 1988 Oxford, Oxford University Press
Fellegi I Sunter A A theory for record linkage Journal of the American Statistical Association 1969 64 1183 1210
Jaro MA Advances in record linkage methodology as applied to matching the 1985 Census of Tempa, Florida Journal of the American Statistical Association 1989 84 414 420
Jaro MA Probabilistic linkage of large public health data files Statistics in Medicine 1995 14 491 498 7792443
Hartigan J Clustering algorithms 1975 New York, John Wiley and Sons
Everitt B Cluster analysis 1993 3 London, Edward Arnold
Eades P A heuristique for graph drawing Congressus Numerantium 1984 42 149 160
Fruchterman T Reingold E Graph drawing by force-directed placement Software-Practice and Experience 1991 21 1129 1164
Kamada T Kawai S An algorithm for drawing general undirected graphs Information Processing Letters 1989 31 7 15 10.1016/0020-0190(89)90102-6
Monge AE Elkan CP An efficient domain-independent algorithm for detecting approximately duplicate database records SIGMOD Workshop on Research Issues on Data Mining and Knowledge Discovery, Tuscon 1997
Hylthon JA Identifying and merging related bibliographic records 1996 Master of Engineering in Electrical and Computer Sciences. MIT
Sharan R Shamir R Jiang T Algorithmic approaches to clustering gene expression data Current Topics in Computational Molecular Biology 2002 The MIT Press 269 300
Baxter A Christen P Churches T A comparison of fast blocking methods for record linkage ACM SIGKDD '03 Workshop on Data Cleaning, Record Linkage and Object consolidation 2003 Washington DC
Hernandez M Stolfo S The merge/purge problem for large databases ACM SIGMOD International Conference on Management of Data, San Jose 1995
Baeza-Yates R Frakes WB Information Retrieval: Algorithms and Data Structures 1992 Englewood Cliffs, Prentice-Hall
McCallum AK Nigam K Ungar LH Efficient clustering of high-dimensional datasets with application to reference matching Sixth International Conference on Knowledge Discovery and Data Mining, Boston 2000
Cohen W Ravikumar P Fienberg S A comparison of string metrics for matching names and records ACM SIGKDD '03 Workshop on Data Cleaning, Record Linkage and Object consolidation 2003 Washington DC
Porter EH Winkler WE Approximate string comparison and its effect on an advanced record linkage system 1997 Washington DC, US Census Bureau
Hartuv E Shamir R A clustering algorithm based on graph connectivity Information Processing Letters 2000 76 175 181 10.1016/S0020-0190(00)00142-3
Sharan R Shamir R CLICK: a clustering algorithm with application to gene expression analysis Eighth International Conference on Intelligent Systems for Molecular Biology, La Jolla 2000
Ben-Dor A Shamir R Yakhini Z Clustering gene expression patterns Journal of Computational Biology 1999 6 281 297 10582567 10.1089/106652799318274
Pavan M Pellilo M A new graph-theoretic approach to clustering and segmentation IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Madison 2003
Kawaji H Yamaguchi Y Matsuda H Hashimoto A A graph-based clustering method for a large set of sequences using a graph partitioning algorithm Genome Informatics 2001 12 93 102 11791228
Yancey WE Frequency-dependent probability measures for record linkage Statistical research report series N°07 2000 US bureau of census, Washington DC
Winkler WE Machine learning, information retrieval and record linkage Proceedings of the Survey Research Methods Section 2000 American Statistical Association
Fortini M Liseo B Nuccitelli A Scanu M On bayesian record linkage Sixth World Meeting of the International Society for Bayesian Analysis, Hersonissos, Greece 2000
Kilss B Alvey W Eds Record linkage techniques – 1985 Proceedings of the Workshop on Exact Matching Methodologiese 1985 Arlington, US Internal Revenue Service
Monge AE Matching algortihm within a duplicate detection system Bulletin of the IEEE Computer Society Technical Committee on Data Engineering 2000 23 14 20
Quantin C Bouzelat H Allaert FA Benhamiche AM Faivre J Dusserre L Automatic record hash coding and linkage for epidemiological follow-up data confidentiality Methods on Information in Medicine 1998 37 271 277
|
16219102
|
PMC1274322
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Oct 11; 5:32
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-32
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-321621910210.1186/1472-6947-5-32Research ArticleMedical record linkage in health information systems by approximate string matching and clustering Sauleau Erik A [email protected] Jean-Philippe [email protected] Antoine [email protected] Service des études et applications de l'information médicale (SEAIM), Hospital, 87 Ave d'Altkirch, F68051 Mulhouse, France2005 11 10 2005 5 32 32 6 5 2005 11 10 2005 Copyright © 2005 Sauleau et al; licensee BioMed Central Ltd.2005Sauleau et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Multiplication of data sources within heterogeneous healthcare information systems always results in redundant information, split among multiple databases. Our objective is to detect exact and approximate duplicates within identity records, in order to attain a better quality of information and to permit cross-linkage among stand-alone and clustered databases. Furthermore, we need to assist human decision making, by computing a value reflecting identity proximity.
Methods
The proposed method is in three steps. The first step is to standardise and to index elementary identity fields, using blocking variables, in order to speed up information analysis. The second is to match similar pair records, relying on a global similarity value taken from the Porter-Jaro-Winkler algorithm. And the third is to create clusters of coherent related records, using graph drawing, agglomerative clustering methods and partitioning methods.
Results
The batch analysis of 300,000 "supposedly" distinct identities isolates 240,000 true unique records, 24,000 duplicates (clusters composed of 2 records) and 3,000 clusters whose size is greater than or equal to 3 records.
Conclusion
Duplicate-free databases, used in conjunction with relevant indexes and similarity values, allow immediate (i.e.: real-time) proximity detection when inserting a new identity.
==== Body
Background
Because of the fast growth of communication protocols (Internet technologies amongst them), health services are undergoing a large paradigm change: a shift from institution-centred care to consumer-centred care. Unambiguous identification of patients is a critical success factor for health care reform and for the provision of speedy, safe, high quality, comprehensive and efficient health care. We may consider more complete information on which to base potentially life-critical clinical decisions and less wasted time and less inconvenience as a result of hunting for information and/or re-gathering as being amongst the benefits of positive identification. However, client information flows are often substantially limited by an inability to positively identify the subjects of care and to locate their relevant details amongst an extensive array of data repositories which may be unlinked, duplicated and catalogued in different ways. Health workers are now confronted with "health records" for each patient dealing with contacts in a hospital, with a general practitioner, etc. If every contact with the patient is an opportunity to collect information, the information noted must, within an identification domain, be perfectly linked to the patient and identified within the domain by a unique identification number (UID) and a minimum profile (i.e. a set of features such as name, first name, gender and date of birth). This UID and profile then constitute the identity of the subject. When matching different identification domains, profiles are compared in order to track the same individuals. The ultimate aim is a complete "data reconciliation", meaning the coherent association between a unique identity and several other features like medical data, demographic data, ... The expression "record linkage" refers to the use of an algorithm technique to match records from different datasets which correspond to the same statistical unit [1,2]. But whatever the chosen method is, the final decision remains in the hands of a human specialist. Computer logic, based upon Boolean operations can basically tell us if two values are equal ("True or False", "1 or 0"), but it is limited if the answer has to be a bit more fuzzy. The decision-making ability of the human mind is still difficult and time consuming to replicate.
Medical information, and by extension identity records, has to be over-time stable, mainly due to legal requirements and for reasons of traceability. So we must be sure, until a final decision is made by the specialist, not to alter the original data in any way during the linkage process. It would be useful, therefore, when comparing records, to process atomic parts of them (such as name, first name, date of birth ...) in order to avoid the most common mistakes made during data entry. This step, called standardization, will ensure a greater effectiveness of the comparison algorithm. Most of the mistakes are a result of not adhering to data entry guidelines (i.e. abbreviations, accented letters, date format...), which are too often the responsibility of the operator. Some others appear when attempting to match records from different data sources (or different identification domains), ruled by heterogeneous and non-compatible data-entry guidelines. This standardization step is mandatory and its lifetime should not exceed the comparison operation.
In a database containing n records, record linkage is often described as an O(n2) complexity problem, due to its Cartesian aspect. The brute force approach, comparing each record to every other record, requires (n2-n)/2 comparisons. This approach, while being the most reliable (as no record is missed during the comparison process) is also the most time consuming and least effective (CPU load speaking). To reduce the number of comparisons, indexation of databases by blocking techniques is used. The data sets are split into smaller blocks and only records within the same blocks are compared.
Whichever blocking technique is chosen, the principle of a record linkage process is to try to pair similar records. But how is the decision which qualifies two different records as "duplicates" made? A basic way is to compare each atomic part of a sample record (i.e. the fields) to its counterpart in the reference record. If a binary equality is obtained, we can then assume that the entities are the same. The main drawback with such a method is that keystroke mistakes or subtle spelling changes are ignored, even after standardization. To be more efficient, we must enhance the comparison principle. While a binary operator returns a Boolean value (1 if the fields are strictly the same, 0 if they are different), a similarity operator will return a score ranging from 0 to 1, showing fields proximity and quantifying their differences. The higher the result, the nearer the field values are. Such a similarity operator is based on approximate string matching methods, so non-literal data fields such as dates or numbers have to be converted before processing. The main approach of approximate string matching has always, until now, been based on the edit-distance [3,4], the oldest kind of algorithm according to Navarro [5,6]. For example, Levenhstein's distance [7] is the minimum number of operations on individual characters like substitutions, insertions and deletions, needed to transform one string to another. It remains the most flexible although it is no longer the most effective, at least in "text retrieval" cases. The Smith-Waterman algorithm [8] is one of the most popular algorithms for edit-distance. Much of its power is due to its ability to introduce gaps in the records (sequence of non-matching symbols). Hence, the use of such an algorithm on each field of records allows us to calculate the same number of atomic similarities (noted aS in the sequel) linking the fields of the sample and the reference records. Such atomic values are then combined in a weighted mean to obtain a record to record similarity or global similarity (gS) value. If one uses k fields for comparing records then where wj are the weights (corrected so that the sum of all of them is 1) in the mean. Indeed, not all the atomic similarities have the same discriminative power when comparing two records. For example, gender is less reliable than date of birth: two records with the same gender have less chance of representing the same identity than two records with the same date of birth. The weights are obtained, for example, by the Expectation-Maximization algorithm [9,10]. Several general algorithms for record linkage have been written since the 70's, most of them in biomedical papers in order to perform epidemiological studies [11]. The seminal theoretical paper on record linkage by Felligi and Sunter [12] has, as its goal, the division of record pairs into linked pairs (designated matches), possible linked pairs (pairs for which human oversight, also known as clerical review, is needed) and non-linked pairs (non-matches). Their classification rule is based on the comparison of the common fields of the two sets of records. To be a little more specific, let us suppose that all the records of a dataset are distributed into two equal sets, say A and B, and that furthermore we create three more sets: for matched pairs, for non-matched pairs and a set of possible matched pairs. If each record has k fields, it is possible to define an agreement k-vector γ (all the atomic similarities for example) between a record of the set A and a record of the set B. The main issue of the Felligi-Sunter theory is to define an optimal linkage rule (for given levels of type I and type II errors), where optimality is defined as minimising the probability of classifying a pair in the set of possible matched pairs. Assuming conditional independence of the k components of γ, the decision rule is a function of where those weights wj are . Here m(γj) and u(γj) are the conditional probability of observing γj given that the pair is respectively a true matched pair and a true non-matched pair. For example, Jaro [13,14] uses the EM algorithm to estimate these weights (or at least the m(γj)s).
Results provided by those techniques of record linkage consist of a list of paired records linked by a global similarity score. Whilst being a good starting point, such a presentation is far from being the most efficient for clerical review. Specialists have to deal with complex patterns of similarities between several records split-down into a list of similarities between couples of records. We can then try to build clusters of duplicates, "n-plicates". A weak definition of a cluster could be "a set of entities which are alike and entities from different clusters are not alike". Methods for clustering can be divided into hierarchical, graph-based (equivalent to the graph partitioning), model-based or mixed methods [15,16]. An additional stage in the representation of the records with their similarities may be necessary. From a representation of records in clusters, we can switch to a representation of graphs. Following the graph theory, the clusters become undirected sub-graphs with vertices (records) and edges (global similarity value). It would be useful, for the visual comfort of the user, to represent these clusters in a reliable manner with vertices and edges, whose lengths are proportional to the strength of the relation between the vertices (global similarity). The most used methods for graph drawing are force-directed methods [17-19]. Generally, these methods view a graph as a system of particles (vertices) with forces acting between them (edges). They seek a state of balance where the sum of the forces acting on each particle is zero. All these methods are relatively simple to implement, heuristic improvements can easily be added, but they can be time consuming.
But specialists also have to deal with the possibility that some of the relations between some record pairs may be unreliable. For example, two different paths between two records can exist: one path with only edges labelled "duplicate" and one with at least one edge labelled "non-duplicate". Hence, some of the sub-graphs are complete (each of the vertices is linked with each other in the sub-graph by an edge), like the left panel on Figure 1. But some are incomplete (n vertices but fewer than n(n-1)/2 edges), like the right panel on Figure 1. If we choose not to enforce transitivity of the relation "is a duplicate of", unlike some other authors [20,21], the vision of this sub-graph as a "n-plicate" is not so straightforward. Thus, it would be interesting to consider complete sub-graphs within this incomplete one, and then to consider each of these last sub-graphs as "n-plicate". Methods for this result are derived from graphs partitioning algorithms [22], in an unsupervised situation (the number of sub-graphs is, a priori, unknown). They go on to delete several edges in order to isolate groups of vertices. Such a cut is said to be of minimum weight, if the sum of the weights on the edges is the minimum necessary to isolate such groups of vertices. The gain of a vertex is the difference between the sum of the similarities with vertices of other sub-graphs and the sum of the similarities with vertices of its own sub-graph. Notably, the gain is the cost when a vertex changes its sub-graph.
Figure 1 Example of a complete graph (left panel) and of an incomplete graph (right panel).
Our very general aim is to provide specialists with a reliable presentation of a set of several potentially duplicated records in a data set with a score showing how similar records in this homogeneous set are. These specialists may be responsible for merging similar records or for avoiding duplicate entry. This method, then, deals with two different situations: batch and real-time data browsing.
Methods
The method used for achieving our goal comprises three steps: the first one is a pre-processing of data in order to eliminate the most common errors and to present this data in a convenient way; the second is to compare records and the third is to organize the records. We will focus on showing its application upon a single data source, even if the proposed technique can be applied to link foreign domain records (i.e. a sample data source to a reference one).
A particularity in Anglo-Saxon countries and in France (but not in all Latin countries) is the use by a married woman of her husband's name: her birth name is no longer used but "replaced" by her married name. Because of the instability of this married name (divorced persons), the birth name is always recorded in a patient database to guarantee an over-time reference. But a high risk of errors exists while encoding names between birth name and married name. First of all, we define a typical patient-identity record as constituting a unique identification number and a minimum profile: birth name, married name, first name, gender and date of birth. We are working, here, on the patient database of our hospital as it was on the 1 st July 2003. This paper reports the use of our proposed method on this database from which we want to eradicate duplicates. Each of the variables belonging to the patient profile is already encoded in formatted fields within the database and thus does not need any parsing or re-allocating.
Accents, hyphens and apostrophes may be uncommon in English, but are very important and may appear often in French or in other Latin languages. Data entry of patient identities is, for the most part, under human control and subject to missing data or data entry errors (such as typographical or keystroke mistakes), non standard abbreviations or differences in detailed schemas of records from multiple databases. We first of all try to standardize the strings before comparing them with the algorithm: replacing all accented letters with the same unaccented letter, converting all strings in uppercase, replacing punctuation signs with a space, discarding all non-informative spaces (double spaces, spaces at the end of strings...), discarding all spaces in names and, in double first names (quite different from first name and initial in English), replacing the first part by its first letter (for example "Jean-Philippe" becomes "J PHILIPPE").
Baxter et al [23] describe four different blocking methods. The standard blocking method clusters records that share an identical blocking key, composed of one or more attributes of each record, like the postcode or phonetic encoding of names. The sorted neighbourhood method [24] sorts the records based on a sorting key and then moves a window of fixed size sequentially over the sorted records. Records within the window are then compared with each other. In the bigram indexing method [25], the blocking key values are converted into a list of bigrams, and sub-lists of all possible permutations are built using a threshold. The canopy clustering method [26] creates overlapping subsets, called "canopies" composed, for each record, of all the records within a certain loose threshold distance. For indexing the database, we chose a blocking method borrowed from this last technique. For each record a certain number of blocking keys is calculated, each one being three bytes in length which allow us to scan, for a given record, not the entire database but only the records whose blocking keys correspond. The structure of these keys is very simple. They are comprised of the first three characters of the standardized string (normal version) and the last three characters (inverted version). These two keys are computed for birth name (i.e. maiden name for a married woman), first name and, if it exists, married name. The use of inverted keys allows us to compare strings with an error on the first few characters, which seems to be common in patient databases. Furthermore, the speed of the algorithm remains high and the window screening remains feasible. Two records sharing the same canopy are compared if they meet one of the following two conditions:
• Their dates of birth and their gender are the same;
• Their gender are the same and their blocking keys on birth name and married name (if necessary) meet at least one of these two conditions:
1. their normal or reversed blocking keys for birth names or for married names are the same
2. the normal (respectively the reversed) blocking key for birth name in one of the two records is the same as the normal (respectively the reversed) blocking key for married name in the other record.
A comparison is made for each atomic part (i.e. field) of the identity record, giving an atomic similarity value for each. In order to compare two records, we run the approximate string matching algorithm on the different matching variables and write out six atomic similarities: three for the comparisons of the two dates of birth, the two first names and the two birth names, one, if necessary, between the married names (otherwise the atomic similarity is considered as missing), one if necessary between the married name of the first record and the birth name of the second (otherwise missing) and one, if necessary, between the birth name of the first record and the married name of the second (otherwise missing). From a series of different algorithms [27], we chose to use an algorithm we call the Porter-Jaro-Winkler algorithm [28], an improvement of an initial algorithm [13,14]. Its basic principle is to compare two strings C1 and C2, and to compute a similarity rate. It considers letters in the first string which are within half the length of the second string. In addition, some errors are penalized less harshly (visual scanning errors, keypunch errors and errors at the end of the string with respect to the errors at the beginning of the string). It is a looser matching criterion than edit-distance and, unlike the other techniques, it is not an "all or nothing" computation, but gives a degree of match. Noting L1 and L2, C1 and C2's respective lengths, the similarity rate is obtained by , where Nc is the number of common characters in the two strings and Nt the number of transpositions. From this basic formula, several improvements are successively obtained, but we use only two of them:
• The number of "similar" characters Ns replaces the number of common characters Nc. A character is decreed "similar" to another if it belongs to one of 36 couples defined in the algorithm, which are for example (X, K), (5, S), (O, Q);
• Additional weight is given to the similarity when the first four characters of C1 and C2 are similar or identical. Typographical errors are often not located at the beginning of the string, but rather in its body;
According to the authors, a similarity of less than 0.7 means two different strings and more than 0.95 means two similar strings (common strings exhibit a rate of 1). We then add different refinements to this algorithm. We weight the atomic similarities of names (birth names and married names) depending on their frequency in the database: the more frequent a name is, the less the similarity seems to be credible. In this early report of our work, we have created just two categories according to the name frequency (approximately 5 per 10,000) and under-weight if necessary the similarity of 0.05. This 5 per 10,000 cut-off was decided from the graph of the names frequency, which has an inverse function aspect with the end of the strong decrease about 5 per 10,000. We also take into account the difficulty of comparing two strings of different lengths which present a strong similarity on the shorter length. For instance, the names "ABC" and "ABCDEFG" are quite different but have an atomic similarity according to the initial algorithm of 1. In this case, we calculate a similarity like the initial one minus 0.01 times the number of additional characters (4 in the above example, yielding an atomic similarity of 0.96). A global similarity is then computed as a weighted mean of atomic similarities. Because of the particular configurations of our records (with or without married names) we use a pragmatic weighting based on common weights used in literature and on our prior knowledge. Once again, we define a few rules:
1. If the two birth names and the two first names are almost the same (i.e.: aS > 0.7) then the date of birth should have a high discriminative weight;
2. If the similarity between the married name of a record and the birth name of a second record is more than the similarity between the two birth names then we suppose that an inversion between the married name and the birth name has occurred.
The weights are standardized to 1. A complete description of this weighting step appears in the Appendix. For example, the simplest situation (without any married name to compare) can be split into two sub-cases: if the birth name's atomic similarity is high (i.e. > 0.7) then the weight for birth names is 1/3, the weight for first names is 1/6, the weight for dates of birth is 1/2, otherwise the three weights are respectively set to 1/2, 1/4 and 1/4. Unlike the Felligi-Sunter probabilistic theory, we chose not to use two thresholds which would classify pairs into linked, non-linked and possibly-linked, but to determine a unique threshold with a set of non-linked pairs and a set of possibly-linked pairs. Before analyzing the database and after multiple experiments, this threshold was set to 0.85.
To achieve the clustering step, because of our confidence in the similarity algorithm, we need a simple, unsupervised (the number of clusters is, a priori, unknown) hard-clustering method (not a probability of classification). Our choice is a greedy agglomerative algorithm with complete linkage to avoid clusters in a chain. For graph drawing, we will choose a very simple version of a force directed algorithm, but in the early version of our work, this functionality is not yet implemented. Then, in order to automate the graph partitioning, several algorithms have been proposed [29-33], they are often heuristics based on seminal algorithms like Kerninghan-Lin or Fiduccia-Mattheyses. Because of our confidence in the previous steps and the use of the Porter-Jaro-Winkler algorithm, a simple algorithm based on the gain is then used to partition incomplete clusters in complete objects (clusters or isolated records).
Results
With 300,859 records in our patient database, the brute force approach would need 45 109 comparisons. By using blocking variables, only 24.8 106 comparisons are computed, which means a 1,825-fold gain. In the database, there are 287,850 different canopies with a maximum canopy size of 11,550 records. In fact, the distribution of the size of the canopies is very left-skewed. If the mean size is 82 and the standard deviation 195, the median is 18 and the 95th percentile is 398. The most frequent size is 7. Each record of the database resides (in average) in 8.9 canopies.
Among 300,859 records, our method detects 38,083 couples, whose global similarity equals or exceeds the decision threshold set to 0.85. The Figure 3 summarises the results of our entire method applied to this data set. Table 1 shows the number among these 38,083 couples with exact concordance in global similarity and in atomic similarities. This table shows that 9,566 couples exhibit a unit value for their global similarity. Even if they have different UIDs, these couples seem to correspond to 9,566 different individuals. The value of the atomic similarities for the fields birth name, first name and date of birth is exactly one in about 70% of the couples but less for married name. For the comparisons, when accurate, between birth name and married name, 3,065 couples match exactly (4%). This means that in 3,065 pairs the married name of a record is exactly the same as the birth name of the other record of the pair, probably indicating an inversion between these names. The Table 2 describes the characteristics of global similarity and atomic similarities among the remaining couples, without perfect concordance. About 29,000 couples exhibit a value of global similarity which is not exactly the unit 1 but a value with a mean (and a median) of 0.92. The distribution is approximately uniform on the interval [0.85 – 1.00], except for a small peak at 0.95 corresponding to the underweight of 0.05 concerning exact matching of records with frequent names. Concerning birth name, first name and date of birth, among about 13,000 remaining couples without a unit value of 1, the mean is about 0.80 but the quartiles show that the similarity on first name is more left-skewed than the similarity on birth name. The similarities between married names are globally higher than the similarities between birth names. Among the 16,013 comparisons remaining in the comparisons between birth name and married name, only about 800 exhibit a similarity higher than 0.75 and the mean of these 16,013 is 0.36.
Figure 2 Percentage of true positive couples by global similarity threshold value.
Figure 3 Summary of the entire linkage procedure.
Table 1 Number of exact concordance in global similarity (gS) and atomic similarities (aS) in the 38,083 identified pairs in a database of 300,859 patients
Atomic similarities
gS BN* MN* BN/MN* First name Date of birth
Pairs 38,083 38,083 38,083 76,166 ** 38,083 38,083
Missing 0 0 31,747 57,088 0 0
Values at 1 9,566 25,990 5,348 3,065 26,017 26,505
in % 25.1 68.2 14.0 4.0 68.3 69.6
* BN = birth name, MN = married name
** Theoretical number of comparisons between BN and MN
Table 2 Characteristics of global similarity (gS) and atomic similarities (aS) in the identified pairs without exact concordance
Atomic similarities
gS BN* MN* BN/MN* First name Date of birth
Pairs 28,517 12,093 988 16,013 12,066 11,578
Mean 0.92 0.79 0.78 0.36 0.77 0.82
Stand. error 0.48 0.20 0.23 0.25 0.17 0.07
Minimum 0.85 0.00 0.00 0.00 0.00 0.65
Percentiles
25th 0.87 0.76 0.62 0.22 0.63 0.77
50th 0.92 0.88 0.90 0.41 0.82 0.85
75th 0.97 0.92 0.94 0.52 0.92 0.88
90th 0.99 0.95 0.96 0.63 0.95 0.88
95th 0.99 0.96 0.96 0.75 0.96 0.88
* BN = birth name, MN = married name
Among those 38,083 couples (T), 19,882 (C) may be found by classical techniques: first three last-name letters exact matching and first three first-name letters exact matching and date of birth matching and gender matching, while detection of the last 18,201 is made possible by using a string similarity algorithm. An in-depth analysis of those 18,201 paired records shows that 9,690 couples (P) are true positives, while 8,511 are false ones. Effectiveness over classical techniques can then be calculated: a minimum of P/(C+P) = 33% more true positive couples are detected than by a classical method (for reasons of convenience, the C couples are all declared to be true positives, while in fact, a few false positives may be present). The proportion of true positives according to this clerical review is shown on Figure 2, by different levels of global similarity (gS by unit of one). As a result, we can now assume that the positive predictive value defined as the number of true positives divided by the total number of linked pairs is (P+C)/T = 78%, which represents a good indicator of accuracy. In Figure 2, we see that the drops, at 0.96, 0.91 and 0.88, reflect name frequency adjustments.
Once the couples have been constituted and their global similarity values calculated, we can now define the relevant graphs. The 38,083 couples represent 27,184 clusters. This final step creates two kinds of clusters: 25,642 complete and 1,542 incomplete graphs. The first ones have all their UIDs (vertices) linked by an edge (global similarity above the threshold). The second ones present some UIDs that are not linked to some of the other cluster components. In the complete graphs there are 23,004 clusters of a size of 2 records, 2,237 of size 3 and 401 of a size greater than 3. In the incomplete graphs there are 1,012 clusters of size 3 and 530 of a size greater than 3. The complete graphs represent 54,443 initial distinct UIDs while the incomplete represent 5,760 UIDs. But before clerical review, our method allows us to state that the 25,642 complete graphs potentially represent 25,642 different individuals. Whatever the threshold within the range 0.85 to 1 is, the mean size for the graphs is constant; about 2.1 for complete graphs and 3.5 for incomplete graphs but the relative number of each kind of graph changes: 200-fold more complete graphs than incomplete for a threshold set to 1, 60-fold for 0.95 and 15-fold for 0.85. The use of the partition algorithm allows us to merge the 1,542 incomplete graphs into several complete graphs. The 1,012 incomplete clusters of size 3 become 1,012 single records and 1,012 couples. The other 530 incomplete graphs become 532 single records, 329 couples, 310 complete graphs of size 3 and 111 complete graphs of a size greater than 3. The 38,083 couples correspond to 60,203 initial UIDs. Out of a total number of UIDs of 300,859, 240,656 UIDs do not form part of a couple, corresponding to 240,656 different individuals. 25,642 individuals come from complete graphs and 3,306 from incomplete graphs. Finally, there are 269,604 different individuals in our database, corresponding to a 90% rate of uniqueness. This means that 1 record in 10 is involved in at least one cluster of "n-plicate" identities.
Discussion
Our method gives those specialists responsible for merging similar records a representation showing how close records in some homogeneous sets are. But real-time detection is also one of the main goals of our proposal: avoiding duplicate entry by alerting the user that several neighbour records already exist, and furthermore, they help in decision making whilst providing proximity values between these similar records. This real-time use would be involved in multi-criteria searches for identities or for the creation of identities. But only a simple and easy to use front-end algorithm and short response time allow this possibility. Response times are closely related to an optimization of the algorithm and especially the blocking part. Its improvement allows the reduction in the number of potential duplicates to be tested by the main algorithm.
According to Porter and Winkler, the use of their algorithm improves the number of duplicates by 30% compared to a binary or semi-binary method [28]. Our results are in accordance with this number. Furthermore, we consider only the true positive identified couples in excess of those identified by the classical methods. Nevertheless, methods other than an approximate string matching can also be considered. For example, the CART method (classification tree-based models) could be used for mapping duplicates and defining clusters of records, which have to be seen not as duplicates but "n-plicates". A logistic model could also be proposed, using for its dependent variable a dummy variable indicating a duplicate or not, and the edit-distance as the independent variable. Some other independent variables can be also used.
Even if our approach is feasible and seems to be useful and reliable, we have to improve our process. The formal Jaro-Winkler algorithm is a domain-independent algorithm, which can be used, without any modification, for a wide range of applications. Our algorithm is no more domain-independent because of our weighting procedure of matching variables, and because of our names frequency based weighting. The classical approach in this context, to solve both problems, is to use the EM algorithm [9,10,34] to retrieve weights driven by the dataset. But the formalisation of using the EM algorithm was developed in the probabilistic Felligi-Sunter theory framework. Our approach is no longer within this framework since our weighted average of atomic similarity has no probabilistic interpretation. Furthermore, the conditional independence assumption of the components of the agreement vector is probably not fulfilled. One main basis of our approach is that all records do not have the same fields because of the potential absence of the married name. Hence the theoretical considerations for the extension of the EM algorithm to our approach are not so straightforward. Instead, other attractive methods seem to be machine-learning in a Bayesian classification framework [35,36]. The aim is to improve a partial automation of the linkage decision process, involving the use of a training data set, in order to instruct a learning algorithm to classify data in link or non-link. Moreover, our method essentially deals with real-time data browsing. In this issue the data set has to be viewed essentially as an evolving database and the machine learning seems to be very accurate: each decision taken by the user can become part of the learning process of the algorithm, instead of a new calculation of the weights by the EM algorithm.
The similarities we use (atomic and moreover global) do not fulfil the conditions necessary in order to qualify as a distance. An interesting condition on distance is the transitivity one, for example between records A, B and C: d(A,B) ≤ d(A,C) + d(C,B). A weighted mean of several distances is indeed a distance unless the weights differ for at least one distance between records. This is the case in our approach (notably due to the difference of treatment between records with married name and records without) but that is also the case with the Porter-Jaro-Winkler algorithm due to the improvement involved when the four first characters of two strings are exactly matched: it can be the case between, for example, record A and C but not the case between records B and C. The condition of transitivity is not sufficient to be really useful in the case of string matching for sparing the number of comparisons between strings. A really interesting property would be the exact knowledge of the relation between d(A,B), d(A,C) and d(C,B). But this knowledge is probably not compatible with the complexity of building an indicator to evaluate string proximity and an indicator of neighbouring records. We noticed, empirically, that in our data set, the transitivity property seems to be fulfilled and hence increases the confidence it is possible to have in the Porter-Jaro-Winkler algorithm.
Here we use, a posteriori, an improvement of window algorithms already mentioned in literature [24,37,38]: records are ordered according to a given criterion and the algorithm, for a given record, is not used on the whole reference dataset, but only on records in a window of a given length. The standard method of detecting exact duplicates in a table is to sort the table and then to check if neighboring tuples are identical. Exact duplicates are guaranteed to be next to each other in the sorted order, regardless of which part of a record the sort is performed on. The idea is to do sorting to achieve preliminary clustering and then to do pairwise comparisons of nearby records. But in this case there is no guarantee that approximate duplicates are all next to each other in the sorted order. In the worst case, they will be found at opposite extremes of the sorted heap. In our case, the criterion is the blocking key and we used the algorithm on all the records sharing the same blocking keys. Much more than a sorted list, the canopy technique we use creates overlapping subsets in which records are compared. In the case of a dataset of size n, divided into b blocks by standard blocking or bigram indexing, the complexity of the record linkage decreases from O(n2) to O(n/b) but with a much larger b with bigram indexing. The complexity is O(wn) with the sorted neighbourhood with a w-size window. For canopy clustering, the number of record pair comparisons is O(f2n2/c) where c is the number of canopies and f the average number of canopies a record belongs to. In our data set, we find that the number of canopies has the same order as the number of records and that each record belongs to about 10 canopies (in mean). Our complexity decreases, then, from O(n2) to O(n), which is less accurate than a bigram indexing complexity. However, the calculation of canopies technique complexity over-estimates this complexity [26]. Actually, it seems to be clear that the most efficient improvement would be to consider not a window but a priority queue. For example Monge [20] proposes a three-step procedure. First, the Smith-Waterman algorithm is used to recognize pairs of approximate duplicates, then the union-find algorithm to keep track of clusters of duplicate records incrementally, as pairwise duplicate relationships are discovered. Thirdly, a priority queue of cluster subsets responds adaptively to the size and homogeneity of the clusters discovered as the database is scanned. All these improvements are nevertheless based on identification algorithms using edit-distance.
Our method adds to the calculation of similarity between couples, a step of clustering and partitioning. To our knowledge, no other published method in the field of record linkage offers these last steps. Hence, for comparing our method with the others, we have to rely on the similarity step even if this is not our "final product". The measurement of the quality of record linkage relies on 4 different values: the number of record pairs linked correctly (true positives), the number of record pairs linked incorrectly (false positives), the number of record pairs unlinked correctly (true negatives) and the number of record pairs unlinked incorrectly (false negatives). These values allow the calculation of different estimates of the performance of an algorithm and notably: the specificity (true negatives divided by the number of true non-match pairs), the negative predictive value (true negatives divided by the total number of non-linked pairs), the sensitivity (true positives divided by the total number of true match pairs, which is the sum of the true positives and the false negatives) and the positive predictive value (true positives divided by the total number of linked pairs). The goal of methods for approximate records matching is to retrieve the truest duplicates possible, even if the price to pay is to get some false positives as well. Thus, the interest is mostly in the sensitivity and in the positive predictive value of methods, corresponding with, respectively, the recall and the precision in the text retrieval field. Furthermore, the number of true negatives represents about 80 or 90 % of the total number of the pairs and any comparisons based on a quality indicator involving this number will be difficult to interpret. Indeed, for example, the specificity of different methods will always (except with bad methods) be close to 1 because the differences in the number of false positives will play a very minor role in the division by the number of true negatives. With respect to the tremendous number of records it is very time consuming, and almost impossible, to review all pairs non-linked by the methods we want to compare with, in order to detect false negatives and true negatives. The sensitivity (and the negative predictive value) is hence very difficult to calculate. Only the positive predictive value is a reliable and easy to calculate indicator for measuring the performance of a method for record linkage. As it is quite impossible to review all the pairs, a solution would be to sample some pairs and just review these ones. But this solution does not seem to be completely accurate as the main quality of sampling is to respect the "data generating process". In the case of duplicate identities generating, this process seems to be much too complicated to be reproduced. The risk then, is to have a biased validation sample and to get a wrong quality indicator. Another solution is to have an efficient external data set with several known characteristics. The underlying idea is to compare an exhaustive and clean (meaning without duplicates, after clerical review) data set with these characteristics, with the subset of the general hospital data set with these same characteristics. This procedure was, for example, (with people remaining anonymous) used with a digestive cancer registry [39].
The most common used final indicator, in the framework of record matching, is the percentage of duplicate identities. But we cannot rely on this indicator as our method builds several clusters of different sizes; each cluster corresponding, after clerical review, to one identity. Hence we use an indicator of uniqueness in the database calculated as the sum of the number of UIDs not involved in any cluster and the number of clusters, divided by the number of initial UIDs. As some clusters are not size 2, this uniqueness is not directly related to the number of duplicate identities (meaning couples whose similarity is above the threshold).
When we identify clusters of n-plicates, we have to decide, furthermore, whether they are real n-plicates. The last step is then to merge, not only on the identities – which is already a problem – but also the medical records relating to these identities. Even if all the global similarities are 1 in a given cluster, identified n-plicates stay "possible n-plicates" and only a very close examination can allow us to discard homonymic identities (in our case, same names, same first name, same date of birth, same gender and same address but different subjects). If merging clients' addresses is not so risky, the problem is quite different in the case of medical records. Complete traceability should be possible for "un-merging" clusters if necessary. The necessity of a human decision for merging is the most important limitation for a completely automatic process.
Using matching variables other than the minimum set we defined would dramatically improve the efficiency of the records matching. For example, parents name or the city of birth is a very discriminatory matching variable. Furthermore, the first name seems to be less reliable than the birth name for comparing records, and difficulties with married names yield a complex algorithm for the calculation of global similarity. Hence, perhaps the minimal set for identifying individuals is neither really reliable nor sufficient.
Appendix
The goal of this section is to detail the weighting procedure of the atomic similarities for building the global similarity as a weighted mean of those same atomic similarities. The Table 3 summarizes this procedure. We call A the first record (reference record) and B the second record (sample record). Each record contains four different fields (birth name, married name, first name and date of birth) whose value cannot be the null string (noted {0}) unless it is the married name. These fields, for example in record A, are respectively noted A(BN), A(MN), A(FN) and A(DB). We call aS the vector of atomic similarities and gS the global similarity. The components of aS are the atomic similarities between two fields: which are noted aS(BN) between the two birth names, aS(MN) between the married names, aS(BN/MN) between the birth name of the first record and the married name of the second, aS(MN/BN) between the married name of the first record and the birth name of the second, aS(FN) between the first names and aS(DB) between the dates of birth. T is the threshold under which two atomic similarities are acknowledged to correspond to two different values of the fields. With reference to Porter-Jaro-Winkler, this threshold is fixed to 0.7. In the procedure there are four main cases, each one is further divided in several sub-cases.
Table 3 Weighting procedure of the atomic similarities (Appendix)
Weights BN MN MN/BN BN/MN FN DB
Case 1
A(MN) = {0} and B(MN) = {0} then
If aS (BN) <T or aS (FN) <T then 2/4 - - - 1/4 1/4
Else 2/6 - - - 1/6 3/6
Case 2
A(MN)! = {0} and B(MN) = {0}
(1) If aS (MN/BN) <aS (BN) then
If aS (BN) <T or aS (FN) <T then 2/4 - - - 1/4 1/4
Else 2/6 - - - 1/6 3/6
(*) Else 1/6 - 1/6 - 1/6 3/6
Case 3
B(MN)! = {0} and A(MN) = {0}, switch A and B and go to (1)
Case 4
A(MN)! = {0} and B(MN)! = {0}
If aS (BN/MN) <aS (BN) and aS (MN/BN) <aS (MN) then 2/7 1/7 - - 1/7 3/7
If aS (BN/MN) >aS (BN) and aS (MN/BN) <aS (MN) then 1/7 1/7 - 1/7 1/7 3/7
If aS (BN/MN) < aS(BN) and aS (MN/BN) >aS (MN) then 1/7 1/7 1/7 - 1/7 3/7
If aS (BN/MN) >aS (BN) and aS (MN/BN) >aS (MN) then 1/8 1/8 1/8 1/8 1/8 3/8
For example, suppose that the atomic similarities for two records A and B are aS(BN) = 0.80, aS(MN/BN) = 0.97, aS(FN) = 0.90 and aS(DB) = 0.92. This case corresponds to a first record with a married name and a second record without a married name but with a birth name which is nearer the married name of the first record than the birth name of this first record (probably an inversion between the two names). In the Table 3, these atomic similarities go with the row highlighted with (*). Hence, the global similarity is calculated as 1/6.0.80 + 1/6.0.97 + 1/6.0.90 + 3/6.0.92 or 0.905. Even if there is a further error with the dates of birth, the global similarity indicates a relatively strong proximity between the two records. Of course, if the errors on dates of birth were more pronounced, the global similarity would quickly decrease.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EAS reviewed papers on the subject and built, with JPP, the methodology. JPP is in charge of all methods implementation. AB supervised this work. All authors read and improved successive drafts.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Mr W.E. Winkler for kindly providing his strings comparator algorithm and for his helpful comments. The revision of this paper has also benefited from comments by the three reviewers for its improvement. The authors also wish to thank Mr Darryl Moran for his helpful reading of the paper.
==== Refs
Belin TR Rubin DB A method for calibrating false match rates in record linkage Journal of the American Statistical Association 1995 90 697 707
Newcombe HB Kennedy JM Record linkage: making maximum use of the discriminating power of identifying information Communications of the ACM 1962 5 563 566 10.1145/368996.369026
Vintsyuk T Speech discrimination by dynamic programming Cybernetics 1968 4 52 58 10.1007/BF01074755
Sellers P The theory and computation of evolutionary distances: pattern recognition Journal of Algorithms 1980 1 359 373 10.1016/0196-6774(80)90016-4
Navarro G Raffinot M Flexible pattern matching in strings 2002 Cambridge, Cambridge University Press
Navarro G A guided tour to approximate string matching Association of Computing Machinery Computing Surveys 2001 33 31 88
Levenhstein V Binary code capable of correcting deletions, insertions and reversals Soviet Physics Doklady 1966 10 707 710
Smith TF Waterman MS Identification of common molecular subsequences Journal of Molecular Biology 1981 14 195 197 7265238 10.1016/0022-2836(81)90087-5
Dempster AP Laird N Rubin DB Maximum likelihood from incomplete data via the EM algorithm (with discussion) Journal of the Royal Statistical Society, series B 1977 39 1 38
Winkler WE Using the EM algorithm for weight computation in the Fellegi-Sunter model of record linkage Statistical research report series N°05 2000 US bureau of census, Washington DC
Newcombe HB Handbook of record linkage: methods for health and statistical studies, administration and business 1988 Oxford, Oxford University Press
Fellegi I Sunter A A theory for record linkage Journal of the American Statistical Association 1969 64 1183 1210
Jaro MA Advances in record linkage methodology as applied to matching the 1985 Census of Tempa, Florida Journal of the American Statistical Association 1989 84 414 420
Jaro MA Probabilistic linkage of large public health data files Statistics in Medicine 1995 14 491 498 7792443
Hartigan J Clustering algorithms 1975 New York, John Wiley and Sons
Everitt B Cluster analysis 1993 3 London, Edward Arnold
Eades P A heuristique for graph drawing Congressus Numerantium 1984 42 149 160
Fruchterman T Reingold E Graph drawing by force-directed placement Software-Practice and Experience 1991 21 1129 1164
Kamada T Kawai S An algorithm for drawing general undirected graphs Information Processing Letters 1989 31 7 15 10.1016/0020-0190(89)90102-6
Monge AE Elkan CP An efficient domain-independent algorithm for detecting approximately duplicate database records SIGMOD Workshop on Research Issues on Data Mining and Knowledge Discovery, Tuscon 1997
Hylthon JA Identifying and merging related bibliographic records 1996 Master of Engineering in Electrical and Computer Sciences. MIT
Sharan R Shamir R Jiang T Algorithmic approaches to clustering gene expression data Current Topics in Computational Molecular Biology 2002 The MIT Press 269 300
Baxter A Christen P Churches T A comparison of fast blocking methods for record linkage ACM SIGKDD '03 Workshop on Data Cleaning, Record Linkage and Object consolidation 2003 Washington DC
Hernandez M Stolfo S The merge/purge problem for large databases ACM SIGMOD International Conference on Management of Data, San Jose 1995
Baeza-Yates R Frakes WB Information Retrieval: Algorithms and Data Structures 1992 Englewood Cliffs, Prentice-Hall
McCallum AK Nigam K Ungar LH Efficient clustering of high-dimensional datasets with application to reference matching Sixth International Conference on Knowledge Discovery and Data Mining, Boston 2000
Cohen W Ravikumar P Fienberg S A comparison of string metrics for matching names and records ACM SIGKDD '03 Workshop on Data Cleaning, Record Linkage and Object consolidation 2003 Washington DC
Porter EH Winkler WE Approximate string comparison and its effect on an advanced record linkage system 1997 Washington DC, US Census Bureau
Hartuv E Shamir R A clustering algorithm based on graph connectivity Information Processing Letters 2000 76 175 181 10.1016/S0020-0190(00)00142-3
Sharan R Shamir R CLICK: a clustering algorithm with application to gene expression analysis Eighth International Conference on Intelligent Systems for Molecular Biology, La Jolla 2000
Ben-Dor A Shamir R Yakhini Z Clustering gene expression patterns Journal of Computational Biology 1999 6 281 297 10582567 10.1089/106652799318274
Pavan M Pellilo M A new graph-theoretic approach to clustering and segmentation IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Madison 2003
Kawaji H Yamaguchi Y Matsuda H Hashimoto A A graph-based clustering method for a large set of sequences using a graph partitioning algorithm Genome Informatics 2001 12 93 102 11791228
Yancey WE Frequency-dependent probability measures for record linkage Statistical research report series N°07 2000 US bureau of census, Washington DC
Winkler WE Machine learning, information retrieval and record linkage Proceedings of the Survey Research Methods Section 2000 American Statistical Association
Fortini M Liseo B Nuccitelli A Scanu M On bayesian record linkage Sixth World Meeting of the International Society for Bayesian Analysis, Hersonissos, Greece 2000
Kilss B Alvey W Eds Record linkage techniques – 1985 Proceedings of the Workshop on Exact Matching Methodologiese 1985 Arlington, US Internal Revenue Service
Monge AE Matching algortihm within a duplicate detection system Bulletin of the IEEE Computer Society Technical Committee on Data Engineering 2000 23 14 20
Quantin C Bouzelat H Allaert FA Benhamiche AM Faivre J Dusserre L Automatic record hash coding and linkage for epidemiological follow-up data confidentiality Methods on Information in Medicine 1998 37 271 277
|
16232314
|
PMC1274323
|
CC BY
|
2021-01-04 16:35:47
|
no
|
BMC Med Inform Decis Mak. 2005 Oct 18; 5:35
|
latin-1
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-35
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-311620214910.1186/1471-2288-5-31Research ArticleLack of interchangeability between visual analogue and verbal rating pain scales: a cross sectional description of pain etiology groups Lund Iréne [email protected] Thomas [email protected] Louise [email protected] Cecilia Norrbrink [email protected] Jan [email protected] Elisabeth [email protected] Department of Physiology and Pharmacology, Karolinska Institutet, Stockholm, SE-171 77 Sweden.2 Rehabilitation Medicine Clinic, Danderyds Hospital AB, Stockholm, SE-182 88 Sweden.3 Spinalis SCI unit, Karolinska University Hospital, Stockholm, SE-169 89 Sweden.4 Department of Statistics (ESI), Örebro University, Örebro, SE 701-81 Sweden.2005 4 10 2005 5 31 31 21 6 2005 4 10 2005 Copyright © 2005 Lund et al; licensee BioMed Central Ltd.2005Lund et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background:
Rating scales like the visual analogue scale, VAS, and the verbal rating scale, VRS, are often used for pain assessments both in clinical work and in research, despite the lack of a gold standard. Interchangeability of recorded pain intensity captured in the two scales has been discussed earlier, but not in conjunction with taking the influence of pain etiology into consideration.
Methods:
In this cross-sectional study, patients with their pain classified according to its etiology (chronic/idiopathic, nociceptive and neuropathic pain) were consecutively recruited for self-assessment of their actual pain intensity using a continuous VAS, 0–100, and a discrete five-category VRS. The data were analyzed with a non-parametric statistical method, suitable for comparison of scales with different numbers of response alternatives.
Results:
An overlapping of the VAS records relative the VRS categories was seen in all pain groups. Cut-off positions for the VAS records related to the VRS categories were found lower in patients with nociceptive pain relative patients suffering from chronic/idiopathic and neuropathic pain. When comparing the VAS records transformed into an equidistant five-category scale with the VRS records, systematic disagreements between the scales was shown in all groups. Furthermore, in the test-retest a low percentage of the patients agreed to the same pain level on the VAS while the opposite hold for the VRS.
Conclusion:
The pain intensity assessments on VAS and VRS are in this study, not interchangeable due to overlap of pain records between the two scales, systematic disagreements when comparing the two scales and a low percentage intra-scale agreement. Furthermore, the lower VAS cut-off positions relative the VRS labels indicate different meaning of the rated pain intensity depending on pain etiology. It is also indicated that the scales have non-linear properties and that the two scales probably have different interpretation. Our findings are in favor of using the VRS in pain intensity assessments but if still the VAS is preferred, the VAS data should be analyzed as continuous using statistical methods suitable for ordinal data. Furthermore, our findings indicate a risk to over or under estimate the patient's perceived pain when interpreting condensed VAS data.
==== Body
Background
The assessment of perceived pain is necessary in the clinical setting for diagnosis and choice of treatment but also for the evaluation of treatment efficacy in a research context. The multidimensional pain sensation involves the subjective evaluations of the sensory aspect like intensity, the affective component such as unpleasantness and the cognitive aspect like thoughts related to the condition. The pain intensity, also mentioned as the severity of pain, is probably the most commonly assessed dimension of pain [1].
The level of personal pain experience is only possible to determine indirectly by self-reported ratings often by using uni-dimensional pain rating scales that may be used for various dimensions of pain. The most commonly used scales, both in ordinary clinical work and in research, are the continuous visual analogue scale, VAS, and discrete categorical scales like the verbal rating scale, VRS, and the numerical rating scale, NRS. Although widely used, there is so far no support for a rational choice of one of these scales [2] even though NRS has previously been recommended as an outcome measure for chronic/idiopathic pain clinical trials [3]. In the absence of gold standard there is a need to study to what extent the individual scores captured on one pain scale are interchangeable with the individual scoring on another pain scale, i.e. the quality of the intra-individual assessments.
On group level, the pain assessments on VAS and VRS have been variably reported as highly inter-correlated [4,5] but also as not being interchangeable [6,7] for example due to overlapping VAS records when related to the categories of the VRS. This overlap is obvious, albeit not highlighted, from the results of several studies related to various clinical conditions [8-10] though not in conjunction with taking the etiology or mechanism of pain classification into consideration. A similar overlap was also demonstrated when comparing VAS and NRS of pain in rest and during activity in different pain conditions [11].
To provide a rational treatment approach, classification of pain is also recommended according to its etiology [12,13] or, if possible, to its mechanism [14]. Since the pain experience is uncertainly related to the extent of injury or stimulation [15], the perceived pain may have linear or non-linear properties [16].
The purpose of this study was to evaluate the quality of the intra-individual assessments of self-reported pain intensity on a continuous VAS (0–100) and a discrete five-category VRS, in patients with pain. The patients were separately described in groups of pain etiology. The evaluation includes inter-scale concordance, implying to which extent the assessment on one scale can be replaced by the assessment on the other, without change of the result. The consistency between the scales were also evaluated when continuous VAS assessments were transformed into discrete scales defined by equidistant cut-off positions as well as by unbiased cut-off positions relative the VRS data. The intra-individual assessment stability of both scales is evaluated by test-retest reliability. A statistical approach will be applied that is suitable for all types of data having at least an ordered structure, though distances and magnitude are unknown [17].
Methods
Subjects
Outpatients with diagnosed pain conditions were consecutively recruited from the rehabilitation medicine clinic and the spinal cord injury out patient department at the Karolinska University Hospital, in Stockholm. The assessments were conducted in accordance with the declaration of Helsinki and the patients gave their informed consent to participate. The study was approved by the Ethics Committee of Karolinska University Hospital (dnr 03–162).
The patient's pains, in general located to, and/or projected to the musculoskeletal system, were previously classified according to its etiology by their physicians into – chronic/idiopathic pain, nociceptive or neuropathic pain [12,13]. The chronic/idiopathic pain was described as generally persistent, distributed without neuro-anatomical distribution and present without noxious stimulus which could result from abnormal processing of normal input in the central nervous system. The criteria of nociceptive pain can be described as a response to activation of damaged tissue where the local pain intensity increases during movement or loading of the affected tissue. The characteristic features of neuropathic pain were among the patients in this study, pain located at and/or below the level of the damaged neural structure, i.e. in this case the spinal cord injury, in an area with altered sensibility and persistent or spontaneous pain unrelated to loading.
All patients were also asked about their prescription of analgesics and whether they had consumed any pain killing drugs on the day of assessment.
Study design and pain rating scales
This is a cross-sectional study in the sense that the three pain etiology groups will be described separately. In order to avoid assessment bias the two scales for self-rated pain intensity were administrated to the patients in random order 30 minutes prior to their appointment, scheduled in advance, with their physician. The scales were a continuous, horizontal, visual analogue scale, VAS, (0–100) with the anchor points, "no pain" and "worst possible pain " respectively and, a discrete, five-category, verbal rating scale, VRS, with the eligible alternatives – no pain (0), mild (1), moderate (2), severe (3), worst possible pain (4), [see Additional file 1].
Although the pain rating scales were, per se, familiar to almost all patients, they were again informed about their use and encouraged to try them out prior to the real assessments. Thereafter, the patients were asked to rate their actual pain intensity by marking a level on the scales corresponding to their experienced pain intensity level. In case of not perceiving pain in rest, which was the case among some of the patients with nociceptive pain, the engaged tissue was loaded by isometric muscle contractions or by testing the respective joints active/passive range of movement in order to provoke the pain and thereby be able to rate any actual pain. The VAS was presented on paper sheets and the VRS on an electronic diary (Clinitrac®). The assessment on the electronic diary was transformed to a code-locked data base.
The assessment procedure was repeated for the intra-individual stability evaluation. In the analyses the pain assessments on the VAS were assigned the numeric values 0 through 100 yielding 101 ordered positions.
Statistical methods
The mean value and standard deviation (SD) were calculated for age. Frequency distributions were shown for patients' duration of pain and the use of different analgesics. The median and range (minimum to maximum) were used to describe the ordinal data of self-rated pain.
The statistical method used is designed for comparing scales with different numbers of possible response alternatives [6,7]. As each individual assessed their perceived pain on two scales the data set consists of paired data, (VAS, VRS). Interchangeability between scales with different numbers of response categories requires a high level of order-consistency, i.e. lack of overlapping of the records on one scale relative the other. A possible presence of overlapping was described and evaluated from scatter and line plots. For example, the pairs (34, no pain), (34, mild pain) and (34, moderate pain) are overlapping. The two pairs (43, mild pain) and (48, moderate pain) represent ordered pairs and the two pairs (43, severe pain) and (48, moderate pain) exemplify disordered pairs. The number of disordered pairs, out of all possible different pairs, was calculated and defines the measure of disorder, D [6,7]. The level of order-consistency is defined by the coefficient of monotonic agreement, MA, which can be calculated by MA = 1-2D and ranges from -1 to 1.
In order to describe the correspondence between condensed VAS data and the VRS categories, the continuous VAS assessments were transformed to a discrete five-category scale in two ways; the cut-off positions being defined unbiased relative the VRS assessments, and being defined equidistantly, respectively.
The cut-off positions of the visual analogue line, which define a discrete VAS that is unbiased to the VRS data, are constructed by pairing off the two sets of frequency distribution to each other and by identifying the cut-off positions in VAS that corresponds to the change in category of the VRS. This procedure creates pairs that are in complete order, MA = 1. Thus the condensed discrete scale based on the continuous VAS records will, under this circumstance, show a total order consistency and no systematic disagreement (be unbiased) relative the VRS. Another approach is to condense the continuous VAS records into an equidistant five-category scale that is to be compared with the five-category VRS.
A high level of order consistency between scales with the same number of categories, in our case the condensed VAS and the VRS, requires a high percentage agreement (PA, %) and a lack of systematic disagreement (bias) of the pairs of data. The frequency distribution of the pairs of data was evaluated by means of square (5 × 5)-contingency tables. The proportion of identical pairs defines the PA. A presence of different frequency distributions, also called marginal distributions, indicate a presence of systematic disagreement (bias), which means that the categories of the two scales have different interpretations and are, thereby, not regarded as interchangeable. Two measures of systematic disagreement were calculated; the relative position, RP, and the relative concentration, RC, with possible values ranging from -1 to 1 [16]. The RP estimates the difference between the probability of the pain assessments on one scale being shifted towards higher categories relative to the other scale and the probability of the assessments on one scale being shifted towards lower categories relative to the other. The RC estimates the difference between the probability of the pain assessments on one scale being concentrated relative to the other and vice versa.
The stability of intra-individual assessments was calculated from test-retest pairs of data and a high level of stability requires high level of intra-individual agreement, PA, and a lack of systematic disagreement, which means zero or negligible RP and RC values.
The software package of Statistica, 6.0 was used for descriptive statistics and SYSRAN 1.0 for Matlab 6 was used to calculate D, MA, RP, RC and the corresponding 95% confidence intervals for the RP and RC.
Results
Eighty patients, (mean age 42.8; SD 12.7 years), recruited from the three pain groups, participated in the study and rated their actual pain intensity. All were capable to independently managing the pain assessment instruments. Analgesic drugs were most frequently prescribed to the patients in the chronic/idiopathic and neuropathic pain groups. On the day for pain intensity assessments, fewer patients had consumed analgesic drugs than what was prescribed, table 1.
Table 1 Demographic data of pain patients.
Pain etiology group
Chronic/idiopathic, n = 30 (women, n = 13) Nociceptive, n = 31 (women, n = 15) Neuropathic, n = 19 (women, n = 8)
Age, mean (SD), years 42.8 (10.6) 40.0 (14.2) 47.3 (12.7)
Duration of pain, months, n (%)
0–3 8 (26) 1 (5)
4–6 6 (19)
7–12 5 (16) 3 (16)
> 12 30 (100) 12 (39) 15 (79)
Patients prescribed with analgesics, n (%) 26 (87) 3 (10) 17 (89)
Patients consuming analgesics the day of assessments, n (%) 11 (37) 3 (10) 14 (74)
The results of the first assessments were chosen for inter-scale comparison and both assessments were used for the test of intra-scale stability.
The median levels of rated pain intensity on the VAS were: chronic/idiopathic pain, 59 (range, 12 to 96); nociceptive pain, 25 (range, 4 to 76); neuropathic pain 64 (range, 18 to 100), fig 2. The corresponding median levels of VRS ratings were moderate (2) for all subgroups but with different ranges: chronic pain ranged from mild pain (1) to worst possible pain (4); nociceptive pain ranged from mild pain (1) to severe pain (3); neuropathic pain ranged from mild pain (1) to worst possible pain (4), figure 1.
Figure 1 Joint distribution of rated pain intensity on the continuous VAS versus the discrete VRS in patients with chronic, nociceptive and neuropathic pain, respectively.
Figure 2 Line plots of recorded rated pain intensity on continuous VAS, 0–100 and on the VRS relative the VAS, for the three pain etiology groups respectively.
Inter-scale comparison, continuous VAS versus VRS
Overlapping VAS records relative the VRS categories mild, moderate, and severe pain were seen in all groups in this study, figures 1, 2, indicating that rated pain intensity labeled as e.g. moderate and severe according to the VRS corresponds to any possible value from 26 to 66, and from 41 to 92 respectively on the VAS in chronic/idiopathic pain patients.
The measured level of concordance, monotonic agreement (MA), was found to be similar in all groups of etiology (chronic/idiopathic pain, MA = 0.89; nociceptive pain, MA = 0.87; neuropathic pain, MA = 0.88), revealing a difference between the ordered and disordered pairs of assessments.
Inter-scale comparison, discrete VAS versus VRS
The cut-off positions of the discrete five-category VAS, unbiased to the VRS, were similar in the chronic/idiopathic and neuropathic pain groups (12, 31, 66, 90 and 18, 19, 65, 99 respectively) while cut-off positions in the nociceptive pain group were lower (4, 22, 51, 77), figure 3. The different cut-off positions indicate that the rated pain intensity could have different meaning depending on pain etiology. Figure 3 also shows the inconsistencies between the scales when the VAS records were divided into a five equidistant category scale. For example a patient rating the perceived pain as mild, could be labeled as no pain in the equidistant VAS. This phenomenon was seen in all groups.
Figure 3 Line plots of VAS records condensed into discrete five-category scales relative the VRS – totally ordered (unbiased) and equidistant for the three pain etiology groups respectively.
The observed inconsistencies between the scales imply lack of interchangeability which were confirmed by the PA (ranging from 29% to 60%) and the measures of systematic disagreement, especially in concentration (RC), figure 4a–b, table 2.
Figure 4 a–b Contingency tables of frequency distribution of discrete VAS records relative the VRS on a) the unbiased five-category VAS (v0–v4) relative the VRS (0–4) and b) the equidistant five category VAS (v0–v4) versus the discrete five category VRS (0–4) in patients with nociceptive pain. Agreeing pairs of data are shown in the grey shaded main diagonal.
Table 2 Inter-scale comparisons of five categories VAS versus VRS.
Unbiased VAS vs VRS Equidistant VAS vs VRS
Pain etiology group PA (%) MA PA (%) MA RP (95% CI) RC (95% CI)
All, n = 81 73 0.93 44 0.96 0.18
(0.07 to 0.28) 0.43
(0.33 to 0.53)
Chron/idiop,
n = 30 67 0.95 60 0.99 0.06
(-0.07 to 0.20) 0.27
(0.12 to 0.42)
Nociceptive,
n = 31 77 0.90 29 0.91 0.44
(0.27 to 0.61) 0.56
(0.29 to 0.83)
Neuropathic,
n = 19 58 0.86 42 0.96 -0.02
(-0.25 to 0.21) 0.36
(0.08 to 0.64)
Chron/idiop = Chronic/Idiopathic; PA = Percentage Agreement; MA = Monotonic Agreement; RP = Relative Position; RC = Relative Concentration; CI = Confidence Interval
Test-retest reliability, intra-scale stability
In the two repeated VAS assessments a low proportion of the patients, 11% to 26%, in the three groups recorded the same pain level, and 87% to 100% of the patients recorded the same level in the repeated ratings on the VRS, table 3.
Table 3 Test of intra-scale stability in VAS (0–100) and VRS.
VAS VRS
Pain etiology group PA (%) RP (95% CI) RC (95% CI) PA (%) RP (95% CI) RC (95% CI)
All, n = 81 20 0.01
(-0.03 to 0.04) -0.02
(-0.08 to 0.05) 94 0.03
(-0.01 to 0.06) 0.04
(-0.01 to 0.08)
Chron/idiop,
n = 30 20 -0.05
(-0.12 to 0.01) -0.07
(-0.18 to 0.03) 97 0.02
(-0.02 to 0.07) -0.01
(-0.04 to 0.02)
Nociceptive,
n = 31 26 0.07
(-0.004 to 0.15) 0.03
(-0.05 to 0.12) 87 0.06
(-0.04 to 0.16) 0.11
(-0.01 to 0.22)
Neuropathic,
n = 19 11 0.04
(-0.06 to 0.14) -0.05
(-0.19 to 0.10) 100 0.00 0.00
PA = Percentage Agreement; RP = Relative Position; RC = Relative Concentration; CI = Confidence Interval
Discussion
The results of this study showed overlapping records between the two scales and a comparable level of inter-scale discordance in all pain etiology groups. For the VAS data condensed into a discrete scale unbiased the VRS, the cut-off positions corresponding to the labels – no pain, mild, moderate, severe and worst possible pain – were similar in patients with chronic/idiopathic and neuropathic pain but lower in the patients with nociceptive pain, indicating influence depending on pain etiology. For the equidistant discrete VAS data a systematic disagreement especially in concentration was found relative the VRS levels which means lack of interchangeability. Similar consequences of condensing continuous VAS data into discrete levels have been found elsewhere [6,7] but is not discussed in the findings of Jensen et al. [18].
In the test-retest of the two scales, a low percentage agreement were seen in assessments on VAS through all pain categories, where only 11 to 26% of the patients agreed to the same level, while a high percentage agreement were found in assessments on VRS where 87 to 100% of the patients agreed to the same level. No systematic disagreement was found in test-retest of either scale. The results of this study therefore imply that the records of self-assessed pain intensity on the VAS and the VRS, performed by the same individuals, are not interchangeable, possibly requiring different interpretation, and that the pain intensity assessments on the VAS do not have linear properties. This confirms the results of Svensson and Svensson et Berndtson [6,7,9] in evaluating the use of rating scales for the assessment of subjective variables.
The lack of operational definition of the VAS can possibly induce insecurity on how to relate to the continuous VAS line, thereby contributing to the low percentage agreement of the individual repeated records. The principles of pain classification, that are continuously discussed, could also contribute to the variable results of the present study. Due to its complexity, the pain classifications are not easily executed and there may be unidentified differences between the different pain etiologies, i.e. chronic/idiopathic, nociceptive and neuropathic pain. For instance, the chronic/idiopathic pain is not recommended to be regarded as a single entity [13] since it may include several etiologies and, furthermore, may be referred to as a disease on its own rights [19]. The associated chronic pain is probably not directly related to their initial injury or disease condition, but rather to secondary changes, including ones that occur in the pain detection system itself [14,19]. Also the diversity in neuropathic pain conditions is discussed in terms of its appearance as "definite, possible or unlikely" [20] and, besides the existence of varying degrees of 'neuropathic' components in chronic pain conditions [20,21]. According to the classification by Rasmussen et al. [20], the patients in this study that were classified as neuropathic, could with most certainty be considered as definite since the etiology is spinal cord injury. The patients suffering from chronic pain may, on the other hand, include some degree of possible neuropathic pain. The pain experience may also be influenced by multiple other factors such as gender, cultural conditioning, expectations, social contingencies, mood state, and perceptions of control. In a future however, the principles of pain categorization is hypothesized to be based on the pain mechanism [14].
There is a controversy in the literature regarding which rating scale being most sensitive to change. Because verbal scales usually have few steps, they are considered to be less sensitive than VAS. Breivik and collaborators [8] reported that assessments of acute pain with a four category VRS, was less sensitive than VAS, 0–100, while VAS and an eleven category NRS, showed similar sensitivity and was recommended to be adopted based on subjective preference. Interestingly the VAS scores were, in the same study, reported as being possible to be classified into any of the four VRS categories. Furthermore, the shortcomings of using the VRS has been described as that the patient is forced to translate a feeling into a predefined word that possibly not fit exactly to the patient's experience and, also, that the same word does not necessarily mean the same thing to each patient [4]. On the other hand, a recent study showed that a VRS was superior to the VAS, NRS, verbal numerical rating scale and a faces pain scale considering internal consistency reliability, sensitivity, and preferred by adults [22]. Furthermore, a preference for VRS over VAS was found by Clark and collaborators [23] when 113 patients were asked, and the VRS is recommended for clinical trials due to it easiness to learn how to handle and to interpret its changed score [24].
Consistent with the findings of Ponce de Leon et al. [25], we found a greater intra-individual agreement using the VRS than using the VAS for assessment of subjective phenomena such as pain. The reason for this response may be due to the use of verbal descriptors or the use of only five categories, but also possibly due to that subjective perceptions, such as pain, could be more easily expressed in words than by a mark on a continuous line without operational definition or by numbers. Different expressions such as faces and images could also serve as response alternative of perceived pain level.
Based on the results of this study, numerals in pain rating seem meaningless since rated moderate pain intensity could be presented on the VAS with a range from 22–65 though there are suggestions of regarding ratings more than 30 mm on VAS as probable moderate and ratings more than 54 mm as probable severe when using a 4-point categorical scale [26].
Limitation of this study
One limitation of our study could be the small number of patients and the possible presence of various pain etiologies in some individuals. Our results refer to rated, individual actual pain intensity of patients suffering from pain of different etiologies and cannot be generalized to other situations.
Conclusion
The records of actual pain intensity on the VAS and the VRS are, in this study, not interchangeable in any of the pain etiology groups due to overlap of pain records between the two scales, systematic disagreements when comparing the two scales and a low percentage intra-scale agreement. Furthermore, the lower VAS cut-off positions relative the VRS labels indicate different meaning of the rated pain intensity depending on pain etiology. The results also indicate that the scales have non-linear properties and that the two scales probably have different interpretation. Our findings are in favor of using the VRS in pain intensity assessments but if still the VAS is preferred, the VAS data should be analyzed as continuous using statistical methods suitable for ordinal data. Furthermore, our findings indicate a risk to over or under estimate the patient's perceived pain when interpreting condensed VAS data.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IL, TL, JK, CNB and LS designed the study. CNB and LS collected the data. IL extracted and analyzed the data. JK designed and supplied the electronic diaries. IL wrote the manuscript and TL and ES critically revised different sections of the manuscript. All authors contributed to commenting on drafts of the manuscript and have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Supplementary Figure 1a-b. The two rating scales used for self-assessed actual pain intensity. In the analysis, the VAS and the VRS assessments were assigned the numeric values 0 through 100 and 0 through 4 respectively, each with the anchor points "no pain" and "worst possible pain" respectively.
Click here for file
==== Refs
Von Korff M Jensen MP Karoly P Assessing global pain severity by self-report in clinical and health services research Spine 2000 25 3140 3151 11124730 10.1097/00007632-200012150-00009
McQuay H Consensus on outcome measures for chronic pain trials Pain 2005 113 1 2 15621355 10.1016/j.pain.2004.10.019
Dworkin RH Turk DC Farrar JT Haythornthwaite JA Jensen MP Katz NP Kerns RD Stucki G Allen RR Bellamy N Carr DB Chandler J Cowan P Dionne R Galer BS Hertz S Jadad AR Kramer LD Manning DC Martin S McCormick CG McDermott MP McGrath P Quessy S Rappaport BA Robbins W Robinson JP Rothman M Royal MA Simon L Stauffer JW Stein W Tollett J Wernicke J Witter J Core outcome measures for chronic pain clinical trials: IMMPACT recommendations Pain 2005 113 9 19 15621359 10.1016/j.pain.2004.09.012
Ohnhaus EE Adler R Methodological problems in the measurement of pain: a comparison between the verbal rating scale and the visual analogue scale Pain 1975 1 379 384 800639 10.1016/0304-3959(75)90075-5
Bolognese JA Schnitzer TJ Ehrich EW Response relationship of VAS and Likert scales in osteoarthritis efficacy measurement Osteoarthritis Cartilage 2003 11 499 507 12814613 10.1016/S1063-4584(03)00082-7
Svensson E Comparison of the quality of assessments using continuous and discrete ordinal rating scales Biomet J 2000 42 417 434 10.1002/1521-4036(200008)42:4<417::AID-BIMJ417>3.0.CO;2-Z
Svensson E Concordance between ratings using different scales for the same variable Statist Med 2000 19 3483 3496 10.1002/1097-0258(20001230)19:24<3483::AID-SIM786>3.0.CO;2-A
Breivik EK Bjornsson GA Skovlund E A comparison of pain rating scales by sampling from clinical trial data Clin J Pain 2000 16 22 28 10741815 10.1097/00002508-200003000-00005
Berntson L Svensson E Pain assessment in children with juvenile chronic arthritis: a matter of scaling and rater Acta Paediatr 2001 90 1131 1136 11697423 10.1080/080352501317061521
Averbuch M Katzper M Assessment of visual analog versus categorical scale for measurement of osteoarthritis pain J Clin Pharmacol 2004 44 368 372 15051743 10.1177/0091270004263995
Hartrick CT Kovan JP Shapiro S The numeric rating scale for clinical pain measurement: a ratio measure? Pain Practice 2003 3 310 316 17166126 10.1111/j.1530-7085.2003.03034.x
Lundeberg T Ekholm J Pain from periphery to brain Disabil Rehabil 2002 24 402 406 12033994 10.1080/09638280110108823
Wincent A Lidén Y Arnér S Pain questionnaires in the analysis of long lasting (chronic) pain conditions Eur J Pain 2003 7 311 321 12821401 10.1016/S1090-3801(03)00044-2
Woolf CJ Pain: moving from symptom control toward mechanism-specific pharmacologic management Ann Intern Med 2004 140 441 451 15023710
Melzack R Coderre TJ Katz J Vaccarino AL Central neuroplasticity and pathological pain Ann N Y Acad Sci 2001 933 157 174 12000018
Donaldson GW Chapman CR Nakamura Y Bradshaw DH Jacobson RC Chapman CN Pain and the defense response: structural equation modeling reveals a coordinated psychophysiological response to increasing painful stimulation Pain 2003 102 97 108 12620601 10.1016/s0304-3959(02)00351-2
Svensson E Analysis of systematic and random differences between paired ordinal categorical data 1993 Stocholm, Almqvist & Wiksell International, Sweden
Jensen MP Chen C Brugger AM Interpretation of visual analog scale ratings and change scores: a reanalysis of two clinical trials of postoperative pain J Pain 2003 4 407 414 14622683 10.1016/S1526-5900(03)00716-8
Niv D Devor M Chronic pain as a disease in its own right Pain Practice 2004 4 179 181 17173599 10.1111/j.1533-2500.2004.04301.x
Rasmussen PV Sindrup SH Jensen TS Bach FW Symptoms and signs in patients with suspected neuropathic pain Pain 2004 110 461 469 15275799 10.1016/j.pain.2004.04.034
Attal N Bouhassira D Can pain be more of less neuropathic? Pain 2004 112 223 224 15702512 10.1016/j.pain.2004.08.024
Herr KA Spratt K Mobily PR Richardson G Pain intensity assessment in older adults: use of experimental pain to compare psychometric properties and usability of selected pain scales with younger adults Clin J Pain 2004 20 207 219 15218405
Clark P Lavielle P Martínez H Learning from pain scales: patient perspective J Rheumatol 2003 30 1584 1588 12858463
Jaeschke R Singer J Guyatt GH A comparison of seven-point and visual analogue scales. Data from a randomized trial. Control Clin Trials 1990 11 43 51 2157581 10.1016/0197-2456(90)90031-V
Ponce de Leon S Lara-Muñoz C Feinstein AR Wells CK A comparison of three rating scales for measuring subjective phenomena in clinical research. II. Use of experimentally controlled visual stimuli Arch Med Res 2004 35 157 162 15010197 10.1016/j.arcmed.2003.07.009
Collins SL Moore RA McQuay HJ The visual analogue pain intensity scale: what is moderate pain in millimetres? Pain 1997 72 95 97 9272792 10.1016/S0304-3959(97)00005-5
|
16202149
|
PMC1274324
|
CC BY
|
2021-01-04 16:32:52
|
no
|
BMC Med Res Methodol. 2005 Oct 4; 5:31
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-31
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-321622130610.1186/1471-2288-5-32Research ArticleMethods for confidence interval estimation of a ratio parameter with application to location quotients Beyene Joseph [email protected] Rahim [email protected] Department of Public Health Science, University of Toronto, Toronto, Ontario, Canada2 The Research Institute, Hospital for Sick Children, Toronto, Ontario, Canada3 Department of Family and Community Medicine, Research Program, University of Toronto, Toronto, Ontario, Canada2005 12 10 2005 5 32 32 17 6 2005 12 10 2005 Copyright © 2005 Beyene and Moineddin; licensee BioMed Central Ltd.2005Beyene and Moineddin; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The location quotient (LQ) ratio, a measure designed to quantify and benchmark the degree of relative concentration of an activity in the analysis of area localization, has received considerable attention in the geographic and economics literature. This index can also naturally be applied in the context of population health to quantify and compare health outcomes across spatial domains. However, one commonly observed limitation of LQ is its widespread use as only a point estimate without an accompanying confidence interval.
Methods
In this paper we present statistical methods that can be used to construct confidence intervals for location quotients. The delta and Fieller's methods are generic approaches for a ratio parameter and the generalized linear modelling framework is a useful re-parameterization particularly helpful for generating profile-likelihood based confidence intervals for the location quotient. A simulation experiment is carried out to assess the performance of each of the analytic approaches and a health utilization data set is used for illustration.
Results
Both the simulation results as well as the findings from the empirical data show that the different analytical methods produce very similar confidence limits for location quotients. When incidence of outcome is not rare and sample sizes are large, the confidence limits are almost indistinguishable. The confidence limits from the generalized linear model approach might be preferable in small sample situations.
Conclusion
LQ is a useful measure which allows quantification and comparison of health and other outcomes across defined geographical regions. It is a very simple index to compute and has a straightforward interpretation. Reporting this estimate with appropriate confidence limits using methods presented in this paper will make the measure particularly attractive for policy and decision makers.
==== Body
Background
Effects in comparative analysis are commonly expressed as ratios. One such example is the Location Quotient (LQ), a ratio statistic widely used by geographers, economists and regional planners to measure the degree of relative concentration of an activity on a map [1,2]. The LQ, which is sometimes referred to as concentration ratio, allows the comparison of an area's share of a specific activity with the share of a base aggregate. Furthermore, LQ can produce a rough benchmark in the analysis of localization in an area [3].
In general, statistical inference is more complicated for a ratio of parameters than measures that are expressed as linear combinations. In epidemiological studies, for example, association of risk factors with occurrence of disease in a given study population can be quantified using absolute measures such as the risk difference or by applying relative measures such as the relative risk or odds ratio [4]. Both the relative risk and the odds ratio require more caution from an inferential point of view than the simple risk difference. One of the difficulties in dealing with ratios arises in computing variance estimators.
Despite its popularity as a relative measure, the location quotient is often interpreted and reported primarily as a point estimate without an accompanying measure of precision. However, statistical reasoning and any inferential conclusion one may draw from a sample statistics should reflect uncertainty inherent in the estimation procedure, and appropriate methods that allow proper interpretation of study findings should be used. One way of proper analysis is to construct confidence limits around the sample estimates.
The objective of this paper is to present a number of alternative approaches that can be used to construct confidence limits for measures involving ratios quantities in general and the location quotient in particular.
Methods
A location quotient is a way of measuring the relative contribution of one specific area to the whole for a given outcome. Let xi and ni denote the outcome and population size of the ith area, respectively. Similarly, let x = ∑xi and n = ∑ni be the outcome and population size of the whole, respectively. The location quotient for the ith area is defined as
Depending upon the health outcome under study, the random variables in equation (1) may have different scales of measurements including continuous, binary or counts.
The interpretation of the LQ is as follows: (1) LQi = 1 which indicates that the outcome in the specific region is at the same level as the aggregate, (2) LQi > 1 indicating that the specific region is at a level greater than expected, and (3) LQi < 1 which would indicate that the regional measure is at a level that is less than expected.
To fix notations, suppose interest lies in making inference about a ratio parameter . In this paper, our focus is on confidence interval estimation of θ. Let be an estimate of θ, where the mean parameters for the estimates are given by E() = α and E() = β, respectively. Furthermore, let the estimated variance-covariance matrix of the estimators (, ) be given by
where V11 and V22 represent the variance of and , respectively, and V12 = V21denote the covariance between and .
For the location quotient described in equation (1), , .
It can easily be shown that
where the parameter pi denotes the true incidence rate in the ith area [5].
Using the notation introduced above, we now describe three analytical and computational approaches that can be used to construct confidence intervals for ratio parameters, namely: (1) Delta method (2) Fieller's method and (3) profile-likelihood based interval on generalized linear model (GLM) technique.
The delta method
The delta method is a classic technique in statistics that is based on a truncated Taylor series expansion [6]. According to the delta method, the variance of is estimated by
For sufficiently large sample size, one may assume that has a Gaussian distribution with mean θ and variance σ2 from which a (1 - α)% delta-method based confidence interval can be obtained as
where zα/2 is the (1 - α/2)% quintile of the standard normal distribution (for instance, for a 95% confidence interval α = 0.05 and zα/2 = 1.96) and is the square-root of the expression in equation (3).
The Fieller method
Fieller [7] introduced a novel way of expressing ratios as linear combination of random variables which made computation of confidence intervals of ratios relatively simple.
The justification for Fieller's method proceeds as follows. Suppose and have a bivariate normal distribution with mean vector (α, β)' and variance-covariance matrix as given in equation (2). If we let , then it follows that α + θβ = 0. Now consider the linear combination + θ = 0. It is a well known fact of mathematical statistics that the distribution of a linear combination of normally distributed random variables is itself normal. In particular, it can be shown that
where σ2 = (V11 + 2θV12 + θ2V22). This result implies that is a standard normal random variable and its square is a chi-squared variable with 1 degree of freedom, χ12.
A (1 - α)% Fieller confidence interval is then obtained by finding the set of θ values satisfying the inequality
Equation (6) is a quadratic function in the parameter of interest θ and solving for θ leads to the confidence limits
where .
Both the delta method and Fieller's approach are quite generic and have been used in a wide range of applications [8-10]. The implementation of these two approaches (using equations (4) and (7) respectively neither require sophisticated programming nor specialized software.
Generalized linear modelling
A model that is widely applicable in a number of different distributional scenarios is generalized linear model (GLM) [11]. Among others, the normal, binomial and Poisson distributions are included in this rich family of models.
We consider a situation where we have k regions and need to estimate k location quotients along with the corresponding confidence limits. We formulate the generalized linear framework by re-expressing equation (1) as
log(pi) = log(x/n) + β1I1 + … + βkIk, (8)
where pi is estimated by and the link function relating the outcome to the independent variables is a logarithmic transformation. The indicator variables Ij, j = 1,…,k take on the value 1 if the region is j and 0 otherwise.
The estimated regression coefficients in the above model provide a point estimate of the location quotient in each area in a logarithmic scale. Exponentiating these estimates give the location quotients in their natural scale. For example, if the region of interest is region 1, then the indicator variable I1 will take the value 1 and the remaining indicator variables will be zero. In this case, equation (8) becomes log(x1/n1) = log(x/n) + β1, and a simple re-arrangement shows that the estimated regression coefficient is expressed as . Exponentiating this result leads to , which is the location quotient for region-1.
There are a number of attractive features with this formulation. Firstly, the model can be fitted using standard statistical software such as using the GENMOD procedure in the SAS statistical package [12]. The resulting estimators are maximum likelihood estimators that are well known to have desirable optimality properties. In fact the reason we were able to take the anti-logarithm (using exponential) to get back to the natural scale for the location quotients from the logarithmic scale was due to the invariance property of the maximum likelihood estimates. The invariance property ensures that, if is the maximum likelihood estimator of θ, then g() is a maximum likelihood estimator of g(θ).
Secondly, confidence intervals for model parameters are by products of the modelling procedure. One of the intervals that can be extracted from fitting the generalized linear model is the profile-likelihood based interval. This approach is an iterative procedure which in general gives more accurate confidence limits, especially for small sample sizes [13]. In addition, significance levels are generated automatically that can be used along side the confidence intervals in order to test whether or not the LQ for a given region is significantly different from the null value of one (H0: LQ = 1) or to make comparisons across different regions.
Thirdly, as mentioned earlier, the GLM family encompasses a large number of commonly used statistical distributions. Thus one can use this framework for modelling indices that are based on health outcome measurements with different scales including continuous scale and categorical outcomes.
Results
Simulation results
A simulation study was carried out to investigate the performance of the methods for calculating confidence limits for the location quotient. Three areas with varying population sizes and incidence rates were considered. Table 1 summarizes confidence limits for the resulting three location quotients based on 1000 simulated data sets within each configuration. The average 2.5%-ile and 97.5%-ile values, shown in Table 1 along the rows designated by method "S", were used as "benchmarks" to compare the performance of the delta (D), Fieller (F), and profile-likelihood (P) methods.
Table 1 Comparison of 95% confidence intervals for three location quotients (LQ1, LQ2, LQ3) using three methods (D = delta, F = Fieller, P = Profile-likelihood). Varying incidence rates (p1, p2, p3) were used along with 3 sets of population size configurations (a) n1 = 50, n2 = 80, n3 = 60 (b) n1 = 500, n2 = 900, n3 = 100 (c) n1 = 2000, n2 = 2500, n3 = 1500. A total of 1000 simulated data sets were used to generate "benchmark" limits (designated as "S" under method)
p1 p2 p3 Method
(a) 0.25 0.3 0.2 S 0.5903 1.3959 0.8787 1.4844 0.4332 1.1237
D 0.5789 1.3905 0.8860 1.4679 0.4380 1.1155
F 0.5664 1.4043 0.8815 1.4824 0.4239 1.1234
P 0.5715 1.4995 0.8109 1.5950 0.4375 1.2183
0.2 0.4 0.7 S 0.2249 0.6926 0.7125 1.0744 1.3548 1.8404
D 0.2235 0.6776 0.7168 1.0909 1.3309 1.8413
F 0.2156 0.6758 0.7126 1.0918 1.3464 1.8651
P 0.2409 0.7287 0.6719 1.1505 1.3096 1.8251
0.9 0.1 0.5 S 1.8387 2.3249 0.0956 0.3765 0.9383 1.3492
D 1.7335 2.3942 0.0901 0.3640 0.9040 1.3841
F 1.7718 2.4478 0.0846 0.3624 0.9044 1.3910
P 1.8297 2.2054 0.1088 0.4035 0.8607 1.4280
0.02 0.01 0.1 S 0.0000 1.6889 0.0000 1.0037 1.2667 3.1667
D -0.1536 1.1816 -0.0924 0.6015 1.4479 3.3499
F -0.5404 1.5539 -0.3073 0.7883 1.0254 4.2516
P 0.3015 1.0973 0.1272 0.6214 1.9683 3.1071
(b) 0.25 0.3 0.2 S 0.7924 1.0245 1.0156 1.1439 0.4483 1.0191
D 0.7917 1.0197 1.0167 1.1489 0.4517 1.0012
F 0.7911 1.0199 1.0167 1.1493 0.4504 1.0017
P 0.7731 1.0476 0.9765 1.1932 0.4736 1.0343
0.2 0.4 0.7 S 0.4766 0.6536 1.0807 1.1854 1.7154 2.2426
D 0.4789 0.6537 1.0755 1.1865 1.7326 2.2467
F 0.4781 0.6533 1.0757 1.1870 1.7370 2.2525
P 0.4715 0.6698 1.0411 1.2225 1.7240 2.2265
0.9 0.1 0.5 S 2.2140 2.3684 0.2129 0.2938 1.0390 1.5164
D 2.1557 2.4273 0.2084 0.2971 1.0277 1.5079
F 2.1629 2.4354 0.2078 0.2967 1.0281 1.5093
P 2.2196 2.3530 0.2060 0.3053 1.022 1.5138
0.02 0.01 0.1 S 0.5071 1.5517 0.2381 0.8333 2.5862 7.9550
D 0.5212 1.5436 0.2396 0.7928 2.6717 7.7130
F 0.4819 1.5835 0.2174 0.8136 2.5448 7.9803
P 0.5205 1.8078 0.2494 0.9322 2.6765 8.8046
(c) 0.25 0.3 0.2 S 0.9113 1.0322 1.1066 1.2119 0.7077 0.8480
D 0.9089 1.0296 1.1088 1.2115 0.7043 0.8439
F 0.9088 1.0296 1.1089 1.2117 0.7041 0.8438
P 0.8970 1.0440 1.0914 1.2305 0.6977 0.8544
0.2 0.4 0.7 S 0.4518 0.5288 0.9424 1.0115 1.6636 1.7688
D 0.4519 0.5285 0.9427 1.0147 1.6588 1.7717
F 0.4518 0.5283 0.9427 1.0147 1.6595 1.7726
P 0.4482 0.5341 0.9319 1.026 1.6577 1.7712
0.9 0.1 0.5 S 1.8962 1.9663 0.1923 0.2365 1.0271 1.1150
D 1.8797 1.9797 0.1903 0.2374 1.0237 1.1176
F 1.8808 1.9808 0.1902 0.2373 1.0237 1.1177
P 1.9004 1.9567 0.1896 0.2399 1.0165 1.1249
0.02 0.01 0.1 S 0.4000 0.7184 0.1798 0.3847 2.5421 3.0220
D 0.4020 0.7094 0.1776 0.3813 2.5418 3.0449
F 0.4005 0.7106 0.1766 0.3821 2.5428 3.0504
P 0.4018 0.7441 0.1844 0.4026 2.3876 3.2374
Panel (a) in Table 1 shows results when area population sizes are relatively small with n1 = 50, n2 = 80, n3 = 60, For this scenario, the results from the different approaches can differ, specially when the incidence rates are also small. For instance, when the three incidence rates are set to p1 = 0.02, p2 = 0.01, and p3 = 0.10, both the delta and Fieller intervals resulted in negative lower limits, which would obviously be inappropriate for location quotients. In such cases, the profile-likelihood method may be preferable. On the other hand, we observe a remarkable agreement among the different methods when population sizes are relatively large with n1 = 2000, n2 = 2500, n3 = 1500. In this case, the accuracy of the results is quite good even when the incidence rates are small, i.e., the last 4 rows of Table 1. Panel (b) provides results for moderate population sizes with n1 = 500, n2 = 900, n3 = 100. Overall, the three methods lead to quite similar confidence intervals in this situation, with the delta method and Fieller's intervals being more close to each other.
Application to health utilization data
In this section, the different methods of estimating confidence interval for the location quotient are illustrated using a health utilization data set from Ontario, Canada. Data were extracted from the Ontario Health Insurance Plan (OHIP) database for all ambulatory specialist visits due to rheumatoid arthritis in the fiscal year 1996. The visits were assigned to census divisions based on where the patient was registered and not where services were received. Forty four censuses divisions were used for analysis. For each county, the LQ is defined as the ratio of two proportions with the numerator representing the number of visits to rheumatologists in the census division divided by the total number of specialist visits in the census division, and the denominator defined as the number of visits to rheumatologists for the province divided by the total number of specialist visits in the province.
For a given county, a location quotient of less than one indicates that the utilization of health care services is under represented compared to the provincial utilization rate. On the other hand, a location quotient of greater than one suggests that health care services utilization is greater than expected. A location quotient of one indicates lack of under or over concentration of utilization in the census division.
Table 2 shows the data, along with the estimated location quotients and confidence intervals. The intervals based on the delta and Fieller's methods were identical to three decimal places. Therefore only the Fieller's lower and upper limits are shown in Table 2 and compared with the intervals based on the profile-likelihood results in the generalized linear models.
Table 2 Location quotients for health utilization data, with 95% lower and upper confidence limits using (1) Fieller's and (2) Profile-likelihood methods (see text for details about the data and methods)
County ni xi LQi Fieller Lower Fieller Upper Profile Lower Profile Upper
1 975 365 0.779 0.716 0.842 0.717 0.843
2 370 115 0.647 0.549 0.745 0.552 0.747
3 275 155 1.173 1.051 1.295 1.050 1.293
4 635 170 0.557 0.486 0.629 0.487 0.631
5 2405 1835 1.588 1.552 1.623 1.552 1.622
6 655 175 0.556 0.486 0.626 0.487 0.628
7 730 335 0.955 0.880 1.030 0.880 1.030
8 115 60 1.086 0.896 1.276 0.896 1.273
9 3205 1545 1.003 0.968 1.038 0.967 1.039
10 500 220 0.916 0.825 1.006 0.826 1.007
11 365 125 0.713 0.611 0.814 0.614 0.816
12 915 435 0.989 0.922 1.056 0.922 1.057
13 4920 225 0.095 0.083 0.107 0.084 0.108
14 500 315 1.311 1.223 1.399 1.222 1.397
15 525 215 0.852 0.765 0.939 0.766 0.940
16 20770 11035 1.106 1.093 1.118 1.091 1.120
17 3025 1595 1.097 1.061 1.134 1.060 1.134
18 350 155 0.922 0.813 1.030 0.814 1.030
19 4500 1675 0.775 0.746 0.803 0.745 0.804
20 610 460 1.569 1.498 1.640 1.496 1.638
21 3915 2780 1.478 1.448 1.507 1.448 1.507
22 10550 6710 1.323 1.305 1.342 1.304 1.343
23 720 250 0.723 0.650 0.795 0.651 0.796
24 6080 3130 1.071 1.046 1.096 1.045 1.097
25 350 140 0.832 0.726 0.939 0.727 0.940
26 1840 1140 1.289 1.244 1.335 1.243 1.335
27 685 500 1.519 1.450 1.588 1.448 1.586
28 920 405 0.916 0.850 0.982 0.850 0.983
29 2585 1395 1.123 1.084 1.162 1.083 1.163
30 1020 345 0.704 0.644 0.764 0.644 0.765
31 770 470 1.270 1.199 1.342 1.198 1.341
32 4240 1515 0.744 0.714 0.773 0.714 0.774
33 1580 205 0.270 0.236 0.304 0.237 0.306
34 5505 2475 0.936 0.909 0.962 0.908 0.963
35 3665 2420 1.374 1.343 1.405 1.342 1.406
36 1145 310 0.563 0.510 0.617 0.511 0.618
37 485 100 0.429 0.354 0.504 0.357 0.507
38 400 210 1.092 0.991 1.194 0.991 1.194
39 170 50 0.612 0.470 0.754 0.477 0.760
40 610 115 0.392 0.328 0.457 0.330 0.460
41 380 115 0.630 0.534 0.726 0.537 0.728
42 2695 865 0.668 0.632 0.704 0.632 0.705
43 1865 540 0.603 0.560 0.645 0.560 0.646
44 165 30 0.378 0.256 0.501 0.267 0.511
Total 98685 47425
There is a remarkable agreement in the confidence intervals from both the Fieller and profile-likelihood approaches due to the fact that the denominator of the ratio is estimated with sufficiently high precision, which would be the case for applications with large sample sizes. Small relative differences are observed when sample sizes are small, as in counties 39 and 44 in Table 2.
The LQ was significantly greater than 1, which is the entire confidence interval falls above 1, for 32% (14/44) of the census divisions, indicating significantly higher utilization of health care services for rheumatoid arthritis than the provincial rate. Similarly, 52% (23/44) of the census divisions showed a significantly lower utilization rate. The remaining 16% experienced a utilization rate compatible with the provincial rate.
Discussion
The location quotient (LQ) is one of several spatial measures that is widely used to examine spatial variation of area characteristics [14]. However, a frequently occurring 'gap' is the widespread use of LQ as a point estimate without an accompanying confidence interval. In this paper, we have demonstrated that confidence intervals for ratio parameters in general and location quotients in particular can be obtained using a number of complimentary approaches. Three techniques – the delta method, Fieller's interval, and profile-likelihood based interval from a generalized linear model – are presented and illustrated. We also demonstrated that if the denominator of the ratio is estimated with sufficiently high precision, then the methods introduced in this paper will produce very similar confidence intervals. The normal approximation to the binomial is used for the variance estimate for the calculation by the delta and Fieller methods. Hence it is not surprising that these methods do not perform well for small sample sizes and extreme proportions.
The techniques we described are generic and can be applied to a wide range of settings where ratio parameters are in use. The generalized linear model (GLM) approach is particularly appealing since the parameters of interest (in our application the location quotients) are estimated directly along with confidence intervals and significance levels. These considerations can be important in practical applications where there are several parameters to be estimated. For instance, for the health utilization data set we analyzed in this paper, 44 location quotients and their confidence intervals had to be generated and the modelling approach was the preferred method over the delta and Fieller's techniques.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JB initiated, designed, and drafted the study. Both JB and RM participated in simulating, analyzing, and discussing the results and in writing the paper. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to acknowledge helpful comments by the reviewers. JB acknowledges the support of the Research Institute at the Hospital for Sick Children.
==== Refs
Thrall GI Borden E Thrall S Delineating Hospital Trade Areas GeoSpatial Solution 2002 12 46 51
Cortese CF Leftwich JE A technique for measuring the effect of economic base on opportunity for blacks Demography 1975 12 325 329 1157992
Robinson GM Methods and Techniques in Human Geography 1998 Toronto: John Wiley & Sons
Fleiss JL Statistical Methods for Rates and Proportions 1981 New York: John Wiley & Sons
Moineddin R Beyene J Boyle E On the location quotient confidence interval Geographical Analysis 2003 35 249 256
Oehlert GW A note on the delta method American Statistician 1992 46 27 29
Fieller EC The biological standardization of Insulin Suppl to J R Statist Soc 1940 7 1 64
Cordell HJ Elston RC Fieller's theorem and linkage disequilibrium mapping Genet Epidemiol 1999 17 237 252 10520208 10.1002/(SICI)1098-2272(199911)17:4<237::AID-GEPI1>3.0.CO;2-P
Polsky D Glick HA Willke R Schulman K Confidence intervals for cost-effectiveness ratios: a comparison of four methods Health Econ 1997 6 243 252 9226142 10.1002/(SICI)1099-1050(199705)6:3<243::AID-HEC269>3.0.CO;2-Z
Silcocks P Estimating confidence limits on a standardized mortality ratio when the expected number is not error free J Epidemiol Community Health 1994 48 313 317 8051534
McCullagh P Nelder JA Generalized Linear Models 1989 2 New York: Chapman and Hall
SAS Institute Inc SAS/STAT User's Guide, Version 8 1999 SAS Institute, Cary, NC
Knight K Mathematical statistics 2000 New York: Chapman and Hall/CRC Press
Thrall GI Fandrich J Elshaw-Thrall S Location quotient: Descriptive geography for the community reinvestment act Geo Info Systems 1995 5 18 22
|
16221306
|
PMC1274325
|
CC BY
|
2021-01-04 16:32:52
|
no
|
BMC Med Res Methodol. 2005 Oct 12; 5:32
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-32
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-331622569210.1186/1471-2288-5-33Research ArticleInvesting in updating: how do conclusions change when Cochrane systematic reviews are updated? French Simon D [email protected] Steve [email protected] Joanne E [email protected] Sally E [email protected] Australasian Cochrane Centre, Institute of Health Services Research, Monash University, Monash Medical Centre, Locked Bag 29, Clayton, Victoria, 3168, Australia2005 14 10 2005 5 33 33 5 8 2005 14 10 2005 Copyright © 2005 French et al; licensee BioMed Central Ltd.2005French et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cochrane systematic reviews aim to provide readers with the most up-to-date evidence on the effects of healthcare interventions. The policy of updating Cochrane reviews every two years consumes valuable time and resources and may not be appropriate for all reviews. The objective of this study was to examine the effect of updating Cochrane systematic reviews over a four year period.
Methods
This descriptive study examined all completed systematic reviews in the Cochrane Database of Systematic Reviews (CDSR) Issue 2, 1998. The latest version of each of these reviews was then identified in CDSR Issue 2, 2002 and changes in the review were described. For reviews that were updated within this time period and had additional studies, we determined whether their conclusion had changed and if there were factors that were predictive of this change.
Results
A total of 377 complete reviews were published in CDSR Issue 2, 1998. In Issue 2, 2002, 14 of these reviews were withdrawn and one was split, leaving 362 reviews to examine for the purpose of this study. Of these reviews, 254 (70%) were updated. Of these updated reviews, 23 (9%) had a change in conclusion. Both an increase in precision and a change in statistical significance of the primary outcome were predictive of a change in conclusion of the review.
Conclusion
The concerns around a lack of updating for some reviews may not be justified considering the small proportion of updated reviews that resulted in a changed conclusion. A priority-setting approach to the updating of Cochrane systematic reviews may be more appropriate than a time-based approach. Updating all reviews as frequently as every two years may not be necessary, however some reviews may need to be updated more often than every two years.
==== Body
Background
When people make decisions about health care they should have access to the most up-to-date and reliable evidence. Cochrane systematic reviews aim to provide healthcare professionals, consumers and policy makers with the 'best available' and most up-to-date evidence on the effects of healthcare interventions. One of the advantages of an electronic publication such as The Cochrane Library is that reviews are replaced with an updated version as this new evidence becomes available or mistakes are identified [1]. This differs from print journals where readers do not necessarily know if they are accessing the most up-to-date systematic review available. A reader of a Cochrane review will therefore expect that the information is up-to-date.
Since evidence used to inform decision-making is continually evolving, it is assumed that new research should be incorporated into reviews. In addition, other aspects relevant to the review might change, such as ideas about the cause of the illness, ways of dealing with any adverse effects of the intervention, changes to the methodology used to combine intervention effects, or other aspects of health care relevant to people making decisions about the intervention. To prevent the evidence, and other information in the review, from becoming out-of-date or misleading, recommendations are sometimes made about the frequency with which the evidence base needs to be updated. It is the policy of The Cochrane Collaboration that reviews should be assessed, and if necessary updated, every two years, or should have a commentary added to explain why this is done less frequently [2].
However, the decision about when to update a Cochrane review must account for various factors. Failure to update reviews soon enough may cause decision makers to act on out-of-date information. On the other hand, reviews that are updated too soon may represent a waste of effort and resources [3] or introduce bias [4]. For example, systematic reviews with very few studies are particularly susceptible to 'time lag bias', which occurs when trials with positive results are published more quickly than those with negative or null results. A further danger of updating too frequently is that repeated significance tests can lead to inflated Type I Error. Constant updating could see positive results of meta-analyses purely by chance [5].
Several investigations of Cochrane reviews have examined the updating process. Chapman and colleagues [3] suggested that over three years, only a small percentage of updated reviews actually resulted in a change in conclusion. Koch [6] extracted and analysed the date fields of all reviews from The Cochrane Library Issue 1, 2002, to determine whether or not Cochrane reviews are being updated. He found that 68% of reviews (867 of the 1268) examined did not provide the date when new studies were found or the date of any amendments to the review authors' conclusions, and was unable to determine what proportion of reviews had actually been updated as opposed to simply edited. Higgins [7] examined all reviews in The Cochrane Library Issue 4, 1998. He found that 65 out of 481 reviews had gained at least one additional study since first appearing. Examination of the summary statistic of the primary outcome measure of these updated reviews revealed that statistical significance changed over time in just five reviews. Bastian and Doust [8] examined all reviews tagged as 'updated' in The Cochrane Library in 2003. Although this should be an indication that the review has changed substantively enough to warrant rereading [2], they found that it was often difficult to identify what had changed in the review or even which reviews had new data and/or changed conclusions.
The objective of this descriptive study is to describe the changes that occurred in a cohort of Cochrane systematic reviews over a four year period.
Methods
Development of a cohort of updated systematic reviews
All completed systematic reviews in the Cochrane Database of Systematic Reviews (CDSR) Issue 2, 1998 formed the original cohort of reviews for this project. The latest version of each of these reviews was then identified in CDSR Issue 2, 2002. The Cochrane Collaboration policy is that each review should be updated every two years, thus we chose a four year period to allow for as many reviews as possible to be updated.
Cochrane reviews give the date when the search for studies for inclusion was carried out. A review was defined as updated if a new search had been carried out between the two issues of CDSR examined, or the number of studies included in the meta-analysis of the primary outcome had changed.
Descriptive information was extracted from these pairs of reviews to determine the changes that had occurred during the four years. The information extracted included a determination of whether the review had been updated, whether it included new studies, or whether the original review had been withdrawn, replaced, merged with another review or split into multiple reviews. We also extracted the intervention summary statistic and its confidence interval for the primary outcome from each review. One member of the project team completed the data extraction. For validation purposes, a second member completed a quality assurance procedure on a 10% random sample of the data.
For the rest of the project, we examined only reviews that had been updated with the inclusion of additional studies.
Determination of a changed conclusion
Cochrane reviews contain a section called 'Reviewers' Conclusions' where the authors discuss the implications of the review for practice and the implications for research. Both these sections of the conclusions were examined for change. The updated version of each review was compared to the original and any changes to conclusions were categorised by two investigators independently. Changes were classified as follows: no change; minor change (changes in style or wording that do not alter the substance or meaning of a section); and major change (changes that alter the substance or meaning of a section or alter the interpretation) [9]. Reviews that were judged as having a major change were categorised as having a changed conclusion. We did not independently assess or verify the review authors' conclusions, but instead relied on their interpretation of the results.
Selection of the primary outcome measure
A primary outcome measure was identified in each of the reviews in Issue 2, 1998 using a pre-specified rule adapted from Higgins [7]. The primary outcome was determined from the review as either that stated by the authors as the primary outcome of interest or the first one listed under the 'Objectives' section of the review. If none was mentioned then we used the first outcome listed under the 'Types of outcome measures' subheading. If these approaches failed we used mortality.
Statistical methods
Ratio of confidence intervals
For each review we calculated the width of the CI for the primary outcome in 1998 and 2002. When the summary statistic was an odds ratio or relative risk, the width of the CI was calculated on the natural log scale.
Relative precision of the original and updated review was calculated as the ratio of the width of the CI in 2002 to the width of the CI in 1998. To estimate the mean ratio and its precision, ratios were natural log transformed. This distribution was very skewed, and it was decided that bootstrapping would be a more appropriate method to calculate CIs rather than relying on large sample assumptions. Five thousand bootstrapped data sets were created using simple random sampling with replacement. For each of these data sets the mean of the logged ratios was calculated. The 2.5th and 97.5th percentiles of the distribution of estimated means were used to produce a 95% CI. These estimates were then back transformed to the original scale.
Logistic regression was used to examine if an increase in the precision of the CI in 2002, compared to 1998, was associated with a change in conclusion of the systematic review. An increase in precision being defined as a narrowing of the width of the CI in 2002 compared to 1998.
Determination of significance change
We examined whether a change in statistical significance of the summary statistic of the primary outcome was associated with a change in conclusion of the systematic review, using an exact 95% CI [10] for a difference in proportions.
All statistical analyses were performed using Stata 8.1 (StataCorp 2003. Stata Statistical Software: Release 8.1. College Station, TX: Stata Corporation.).
Results
A total of 377 complete reviews were published in CDSR Issue 2, 1998. Thirty one Cochrane Collaborative Review Groups contributed to these reviews, with the greatest representation (33%) from the Cochrane Pregnancy and Childbirth Group.
Figure 1 outlines the status in CDSR Issue 2, 2002 of all the original reviews from CDSR Issue 2, 1998. During this time period, 14 reviews were withdrawn and one was split, leaving 362 reviews that were present in both issues of CDSR. Of these 362 reviews, approximately one third (38%) had been updated with new included studies, one third (32%) had re-run searches but included no new studies, and 30% had not been updated at all. The median number of studies per review increased from 5 (range 0 to 72) in 1998 to 6 (range 0 to 108) in 2002. Of the 254 updated reviews with and without new studies included, only 23 (9%) had a changed conclusion.
Figure 1 The status of all the reviews from CDSR Issue 2, 1998 in CDSR Issue 2, 2002.
Of the 137 reviews updated with new studies, 18 reviews were excluded from our analysis because a summary statistic for the primary outcome was not available in one or both versions of the review, for example, if the review authors decided that the results of the included studies were not suitable for meta-analysis. Nine of these reviews had a change of conclusion and nine were unchanged. For the reviews with an unchanged conclusion, seven of the nine did not have a summary analysis available in either versions of the review, and two changed their research question resulting in a change in primary outcome. In the nine reviews with a changed conclusion, three did not have a summary analysis available in either version of the review and in four reviews either the research question changed or the outcomes examined changed resulting in a change in conclusion. In the remaining two reviews with a changed conclusion, a meta-analysis was not possible in the 1998 version and was then subsequently possible in the 2002 version.
This left 119 reviews in which new studies had been added to the updated review and for which data were available from the meta-analysis of the primary outcome from both 1998 and 2002. Further analysis was conducted only on these reviews to determine the effects of updating.
Ratio of confidence intervals
Relative precision of the review pairs has been determined as a ratio of the width of the CI in 2002 to the width of the CI in 1998. For 85 of the 119 reviews (71%), the width of the CI around the primary outcome changed by less than 20% as a result of adding new studies (Figure 2). Of these, the width of the CI increased, remained the same, or decreased in five, 36, and 44 of the updated reviews respectively. For the five reviews with widened CIs, two had the same number of studies but had re-extracted the data from their original studies and in the other three reviews the updated meta-analysis led to only a small change in the precision (ratio range 1.04–1.11). The mean ratio of the width of the CI in 2002 to the width of the CI in 1998 was 0.81 (95% CI; 0.75, 0.86). For reviews with unchanged conclusions (n = 105) and changed conclusions (n = 14), this ratio was 0.85 (0.81, 0.89) and 0.56 (0.36, 0.81), respectively.
Figure 2 Ratios of confidence intervals of the summary statistic of the primary outcome of Cochrane reviews from Issue 2, 2002 to Issue 2, 1998.
An increase in the precision of the CI in 2002 compared to 1998 (that is, a decrease in the width of the CI in 2002), increased the odds of a change in conclusion of the systematic review. In particular, for each percentage increase in the precision of the CI in 2002, the odds of a change in conclusion were 3.3% (95% CI; 1.0%, 5.6%) higher than the previous odds. Therefore, for a 19.1% increase in precision, as was observed on average between 1998 and 2002, the odds of a change in conclusion were 85.7% (95% CI; 21.1%, 184.6%) higher than those of no change in precision.
Determination of significance change
Of the 119 reviews, 11 summary statistics of the primary outcome changed statistical significance between 1998 and 2002. Five changed from significant (p < 0.05) to non-significant (p ≥ 0.05), while the remaining six changed from non-significant to significant.
For the 11 reviews where there was a change in statistical significance of the summary statistic for the primary outcome, four reviews changed conclusion (36.4%). Of the 108 reviews where there was no change in the statistical significance, 10 reviews changed conclusion (9.3%). The difference in these percentages was 27.1% (95% CI; 0.7%, 60.3%).
Ten of the 14 reviews that changed conclusion did not have a change in significance of their summary statistic for the primary outcome. For eight of these reviews, the change in conclusion was related to a change for an outcome other than the primary outcome. Of the remaining two reviews, the change in conclusion in one was related to a sub-group analysis, and for the other review, the 1998 analysis was significant but based on only one small trial, and a further trial confirmed this in the 2002 review leading the review authors to be firmer in their conclusions.
Discussion
The updating of Cochrane reviews has become the focus of some research in recent years [3,5-8,11]. Our study has examined in detail the impact of the updating process on the conclusions of a cohort of Cochrane systematic reviews over four years.
After this four year period, most (70%) reviews had been updated and of these, over half (54%) included new studies. A third of reviews (30%) had not been updated, with no indication that any new search for studies had been carried out. From the updated reviews, we estimated that 9% had changed conclusion. Concerns around a lack of updating for some reviews may not be justified considering the small percentage of updated reviews that resulted in a changed conclusion.
It is possible that the percentage of reviews that changed conclusion may differ in the reviews that had not been updated. The percentage would be overestimated, if for example, review authors were more likely to update their review given they had knowledge of new studies with conclusions that may alter the review conclusion. Conversely, the percentage would be underestimated, if under the above scenario, review authors were less likely to update their review because, for example, of workload issues. However, we believe this estimate is reasonable since the percentage of reviews with changed conclusions in those not updated would have to be quite different to that of the reviews that were updated to substantially change the overall estimate.
We compared reviews where the conclusion remained the same to those where the conclusion had changed. Factors predictive of a change in review conclusions were a decrease in the width of the CI in the updated review and a change in significance of the summary statistic of the primary outcome. These factors are of limited value in predicting which reviews should be updated, since they are only known after the update has taken place. A large percentage (50%) of the eighteen reviews excluded from our analysis had a change in conclusion. In the majority of these reviews, the change in the conclusion was due to a change of the research question.
The cohort of reviews selected for inclusion in this study was from the 1998 CDSR. It is possible that examination of a more recent cohort of reviews may provide different results. This may occur if, for example, the percentage of updated reviews was different, or the sizes and types of randomised controlled trials included have changed over time.
The process of updating systematic reviews is not unique to Cochrane reviews. The CDSR does, however, provide a unique opportunity to investigate this process due to the Cochrane Collaboration's policy to update reviews every two years. We are not aware of another study that has addressed the issue of updating in non-Cochrane reviews. The results we have found may also apply to non-Cochrane reviews, but we cannot be sure of this without further investigation. There may be reasons that non-Cochrane reviews are more likely to be updated and their conclusions change than Cochrane reviews. For example, non-Cochrane reviews may be carried out in clinical areas that are rapidly changing, and researchers may not want to go through the full Cochrane review process in order to get their review published more quickly. Systematic reviews in these clinical areas may be more likely to change conclusion.
Future research in this area should aim to establish the suitability of the current policy of The Cochrane Collaboration of updating every review every two years. This should provide a platform for a more empirically based set of recommendations about the updating of evidence in systematic reviews. In the meantime, there may be valid reasons to update a review more or less often than the current two year policy. Consequently, an approach that assesses the priority for updating individual reviews may be more appropriate in determining when reviews should be updated than a blanket time-bound policy. Other reasons for updating, such as the inclusion of a new intervention or outcome, and/or new methods for systematic reviews, will also need to be considered.
Conclusion
The updating of Cochrane reviews consumes considerable time and resources and in many cases may not change the conclusion or lead to a more precise conclusion. Our study does not provide evidence to support or refute the current policy of The Cochrane Collaboration of updating reviews every two years. However, concerns around less frequent updating may not be justified in light of our results. A priority-setting approach for all reviews may be more appropriate than a rigid time-based approach.
Competing interests
The Australasian Cochrane Centre is funded by the Australian Commonwealth Department of Health and Ageing and supported by Monash University. All authors are employed by the Australasian Cochrane Centre. SG is a member of the Cochrane Collaboration Steering Group.
The views expressed in this paper represent those of the authors and are not necessarily the views or the official policy of The Cochrane Collaboration (unless otherwise stated and referenced).
Authors' contributions
SDF participated in the design of the study and extracted data. SM participated in the design of the study and extracted data. JEM performed the statistical analysis. SEG oversaw the study and participated in its design and coordination. All authors contributed to drafting of the manuscript, approved the final manuscript and are guarantors.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was funded by a grant from the Department of Health and Ageing, Commonwealth of Australia. This study was based on a project originally conceived by Chris Silagy. The authors appreciate the data extraction performed by Jason Wasiak and input and comments at different stages of the project from Mike Clarke, Jon Deeks, Julian Higgins Philippa Middleton and Denise O'Connor.
==== Refs
Chalmers I Haynes B Reporting, updating, and correcting systematic reviews of the effects of health care BMJ 1994 309 862 865 7950620
Higgins J Green S editors Cochrane Handbook for Systematic Reviews of Interventions 4.2.5 [updated May 2005] The Cochrane Library 2005 Chichester, UK: John Wiley & Sons, Ltd
Chapman A Middleton P Maddern G Early updates of systematic reviews – a waste of resources? Proceedings of the 4th Symposium on Systematic Reviews: Pushing the Boundaries: Oxford 2002
Hopewell S Clarke M Stewart L Tierney J Time to publication for results of clinical trials The Cochrane Database of Methodology Reviews 2001 Issue 3
Higgins J Prevalence and problems of updated reviews: a survey and discussion Proceedings of the 2nd Symposium on Systematic Reviews: Beyond the Basics: Oxford 1999
Koch G Are Cochrane Reviews updated regularly or not? Proceedings of the 10th Cochrane Colloquium: 31 July – 3 August 2002; Stavanger
Higgins J How should we interpret updated meta-analyses? Proceedings of the 7th Annual Cochrane Colloquium: Rome 1999
Bastian H Doust J When does an updated meta-analysis have enough content to justify re-reading? Proceedings of the XI Cochrane Colloquium: Evidence, Health Care and Culture: Barcelona, Spain 2003
Silagy CA Middleton P Hopewell S Publishing protocols of systematic reviews: comparing what was done to what was planned JAMA 2002 287 2831 2834 12038926 10.1001/jama.287.21.2831
Newcombe R Interval estimation for the difference between independent proportions: comparison of eleven methods Statistics in Medicine 1998 17 873 890 9595617 10.1002/(SICI)1097-0258(19980430)17:8<873::AID-SIM779>3.0.CO;2-I
Barrowman NJ Fang M Sampson M Moher D Identifying null meta-analyses that are ripe for updating BMC Med Res Methodol 2003 3 13 12877755 10.1186/1471-2288-3-13
|
16225692
|
PMC1274326
|
CC BY
|
2021-01-04 16:32:51
|
no
|
BMC Med Res Methodol. 2005 Oct 14; 5:33
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-33
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-501622131110.1186/1471-2474-6-50Study ProtocolThe McKenzie method for the management of acute non-specific low back pain: design of a randomised controlled trial [ACTRN012605000032651] Machado Luciana AC [email protected] Chris G [email protected] Rob D [email protected] Helen [email protected] James [email protected] Back Pain Research Group, School of Physiotherapy, The University of Sydney, PO Box 170, Lidcombe, NSW, 1825, Australia2 Private Practice, 16 Ayres Road, St Ives, NSW, 2075, Australia2005 13 10 2005 6 50 50 28 7 2005 13 10 2005 Copyright © 2005 Machado et al; licensee BioMed Central Ltd.2005Machado et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Low back pain (LBP) is a major health problem. Effective treatment of acute LBP is important because it prevents patients from developing chronic LBP, the stage of LBP that requires costly and more complex treatment.
Physiotherapists commonly use a system of diagnosis and exercise prescription called the McKenzie Method to manage patients with LBP. However, there is insufficient evidence to support the use of the McKenzie Method for these patients. We have designed a randomised controlled trial to evaluate whether the addition of the McKenzie Method to general practitioner care results in better outcomes than general practitioner care alone for patients with acute LBP.
Methods/design
This paper describes the protocol for a trial examining the effects of the McKenzie Method in the treatment of acute non-specific LBP. One hundred and forty eight participants who present to general medical practitioners with a new episode of acute non-specific LBP will be randomised to receive general practitioner care or general practitioner care plus a program of care based on the McKenzie Method. The primary outcomes are average pain during week 1, pain at week 1 and 3 and global perceived effect at week 3.
Discussion
This trial will provide the first rigorous test of the effectiveness of the McKenzie Method for acute non-specific LBP.
==== Body
Background
In Australia, low back pain (LBP) is the most frequently seen musculoskeletal condition in general practice and the seventh most frequent reason for consulting a physician[1,2]. According to the Australian National Health Survey, 21% of Australians reported back pain in 2001; additionally, the Australian Bureau of Statistic's 1998 Survey of Disability, Ageing and Carers estimated that over one million Australians suffer from some form of disability associated with back problems[1].
LBP poses an enormous economic burden to society in countries such as the USA, UK and The Netherlands[3]. In the largest state in Australia, New South Wales, back injuries account for 30% of the cost of workplace injuries, with a gross incurred cost of $229 million in 2002/03[4]. It is expected that most people with an acute episode of LBP will improve rapidly, but a proportion of patients will develop persistent lower levels of pain and disability[5,6]. Those patients with chronic complaints are responsible for most of the costs[6]. Effective treatment of acute LBP is important because it prevents patients from developing chronic LBP, the stage of LBP that requires costly and more complex treatment.
There is a growing concern about effectiveness of treatments for LBP, as reflected in the large number of systematic reviews published in the last 5 years addressing this issue. [7-12]. Despite the large amount of evidence regarding LBP management, a definitive conclusion on which is the most appropriate intervention is not yet available. A comparison of 11 international clinical practice guidelines for the management of LBP showed that the provision of advice and information, together with analgesics and NSAIDs, is the approach consistently recommended for patients with an acute episode[13]. Most guidelines do not recommend specific exercises for acute LBP because trials to date have concluded that it is not more effective than other active treatments, or than inactive or placebo treatments[8]. However, some authors have suggested that the negative results observed in trials of exercises are a consequence of applying the same exercise therapy to heterogeneous groups of patients. [14-16]. This hypothesis has some support from a recent high-quality randomised trial in which treatment based on a diagnostic classification system led to larger reductions in disability and promoted faster return to work in patients with acute LBP than the therapy recommended by the clinical guidelines[17].
In 1981, McKenzie proposed a classification system and a classification-based treatment for LBP labelled Mechanical Diagnosis and Treatment (MDT), or simply McKenzie Method[18]. Of the large number of classification schemes developed in the last 20 years [19-26], the McKenzie Method has the greatest empirical support (e.g. validity, reliability and generalisability) among the systems based on clinical features[27] and therefore seems to be the most promising classification system for implementation in clinical practice.
Physiotherapists commonly adopt the McKenzie Method for treating patients with LBP[28,29]. A survey of 293 physiotherapists in 1994 found that 85% of them perceived the McKenzie Method as moderately to very effective[28]. Nevertheless, a recent systematic review concluded that there is insufficient evidence to evaluate the effectiveness of the McKenzie Method for patients with LBP [30]. A critical concern is that most trials to date have not implemented the McKenzie Method appropriately. The most common flaw is that all trial participants are given the same intervention regardless of classification, an approach contradictory to the principles of McKenzie therapy.
The primary aim of this trial is to evaluate whether the addition of the McKenzie Method to general practitioner (GP) care results in better outcomes than GP care alone for patients with acute non-specific LBP when effect is measured in terms pain, disability, global perceived effect, and persistent symptoms.
Methods
The University of Sydney Human Research Ethics Committee granted approval for this study.
Study sample
One hundred and forty eight participants with a new episode of acute non-specific LBP who present to GPs will be recruited for the study. A new episode of LBP will be defined as an episode of pain lasting longer than 24 hours, preceded by a period of at least one month without LBP and in which the patient did not consult a health care practitioner[31]. Participants will be screened for eligibility at their first appointment with the GP according to the inclusion and exclusion criteria.
Inclusion criteria
To be eligible for inclusion, participants must have pain extending in an area between the twelfth rib and buttock crease (this may or may not be accompanied by leg pain); pain of at least 24 hours duration; pain of less than 6 weeks duration; and they need to be eligible for referral to private physiotherapy practice within 48 hours.
Exclusion criteria
Participants will be excluded if they have one of the following conditions: nerve root compromise (defined as 2 positive tests out of sensation, power and reflexes for the same spinal nerve root); known or suspected serious spinal pathology; spinal surgery within the preceding 6 months; pregnancy; severe cardiovascular or metabolic disease; or inability to read and understand English.
Recruiting GPs will record the number of patients who are invited to participate, the number who decline to participate, and the number of screened patients who are ineligible and their reasons for declining participation or ineligibility. Written consent will be obtained for each participant.
Subjects who volunteer to participate and satisfy the eligibility criteria will receive baseline treatment and then be randomly allocated to one of the study groups. To ensure equal-sized treatment groups, random permuted blocks of 4–8 participants will be used[32]. Randomisation will be stratified by Workcover compensation status. The stratified random allocation schedule will be generated by a person not otherwise involved in recruitment, assessment or treatment of subjects and the randomisation sequence will be placed in sequentially numbered, sealed envelopes. The flow of participants through the study is detailed in Figure 1.
Figure 1 Flow of participants through the study. Legend: GP – General practitioner; NRS – Numeric pain rating scale; PSFS – Patient-specific functional scale; RMQ – Roland-Morris questionnaire; GPE – Global perceived effect; LBP – Low back pain.
Outcome measures
The McKenzie protocol is thought to promote rapid symptom improvement in patients with LBP[33,34] and this is one of the reasons that therapists choose this therapy. Therefore it is important to focus assessment on short-term outcomes. The primary outcomes will be:
1. Usual pain intensity over last 24 hours recorded each morning in a pain diary over the first week. Pain will be measured on a 0–10 numerical rating scale (NRS). The unit of analysis will be the mean of the 7 measures[35];
2. Usual pain intensity over last 24 hours (0–10 NRS) recorded at 1 and 3 weeks[35];
3. Global perceived effect (0–10 GPE) recorded at 3 weeks.
The secondary outcomes will be:
1. Global perceived effect (0–10 GPE) recorded at 1 week;
2. Patient-generated measure of disability (Patient-Specific Functional Scale; PSFS) recorded at 1 and 3 weeks[36];
3. Condition-specific measure of disability (Roland Morris Questionnaire; RMQ) recorded at 1 and 3 weeks[37];
4. Number of patients reporting persistent back pain at 3 months.
Following the screening consultation in which the inclusion and exclusion criteria are assessed, the GP will supervise the baseline measurement of pain. All patients will then receive an assessment booklet and a pre-paid envelope in which all other self-assessed outcome measures are to be recorded and sealed. One member of the research team will contact patients by telephone within 24 hours of the consultation with the GP in order to give explanations regarding the appropriate form of filling in the assessment booklet. At this time, other baseline outcomes will be recorded and then the patient will be randomised to study groups. The patient will be advised to keep the booklet at home, to seal it into the pre-paid envelope after the final assessment and mail the sealed envelope to the research team. To ensure the proper use of the assessment booklet and to avoid loss of data due to non-returned booklets, a blinded assessor will contact all patients by telephone 9 and 22 days after the consultation with the GP to collect patient's answers from the 1st week and 3rd week assessments, respectively.
The procedure for obtaining outcome data will be followed for all participants, regardless of compliance with trial protocols. At 3 months, data regarding the presence of persistent (chronic) symptoms will be collected by telephone. Participants will be asked to answer the following yes-no question: "During the past 3 months have you ever been completely free of low back pain? By this I mean no low back pain at all and would this pain-free period have lasted for a whole month". Those answering no will be considered to have persistent LBP. Information on additional treatment and the direct costs with low back pain management will also be collected at 3 months.
A secondary analysis will be performed on predictors of response to McKenzie treatment and prediction of chronicity. This will involve the measurement of participants' expectation about the helpfulness of both treatments under investigation as well as information on the occurrence of the centralisation phenomenon. Expectation will be recorded prior to randomisation according to the procedures described by Kalauokalani et al[38].
Treatments
All participants will receive GP care as advocated by the NHMRC guideline for the management of acute musculoskeletal pain[2]. Guideline-based GP care consists of providing information on a favourable prognosis of acute LBP and advising patients to stay active, together with the prescription of paracetamol. Patients randomised to the experimental group will be referred to physiotherapy to receive the McKenzie Method. A research assistant not involved in the assessment or treatment of subjects will be responsible for the randomisation process and will contact therapists and patients to arrange the first physiotherapy session. The McKenzie treatment will be delivered by credentialed physiotherapists who will follow the treatment principles described in McKenzie's text book[18]. All therapists will have completed the four basic courses taught by the McKenzie Institute International. To ensure the appropriate implementation of the McKenzie's classification algorithm, a training session with a member of McKenzie's educational program will be conducted prior to the commencement of the study. The treatment frequency will be at the discretion of the therapist with a maximum of 7 sessions over 3 weeks. We chose to restrict the McKenzie treatment to a maximum of 7 sessions based on the study of Werneke and colleagues[39], which concluded that further reductions in pain and function are not expected if favourable changes in pain location are not present until the seventh treatment visit. Treatment procedures from the McKenzie Method are summarised in the Appendix.
Participants randomised to the control group will continue their GP care as usual. All participants regardless of intervention group will be advised not to seek other treatments for their low back pain during the treatment period. Physiotherapists will be asked to withhold co-interventions during the course of the trial.
Several mechanisms will be used to ensure that the trial protocol is applied consistently. Protocol manuals will be developed and all involved researchers (GPs, physiotherapists, assessor, and statistician) will be trained to ensure that screening, assessment, random allocation and treatment procedures are conducted according to the protocol. A random sample of treatment sessions will be audited to check that treatment is being administered according to the protocol.
Data analysis
Power was calculated based on the primary outcome measures (pain intensity and global perceived effect). A sample size of 148 participants will provide 80% power to detect a difference of 1 unit (15%) on a 0–10 pain scale (SD = 2.0) between the experimental and control groups, assuming alpha of 0.05. This allows for loss to follow-up of 15%. This sample size also allows the detection of a difference of 1.2 units (12%) on a 0–10 global perceived effect scale (SD = 2.4).
Data will be analysed by a research member blinded to group status. The primary analysis will be by intention-to-treat. In order to estimate treatment effects, between-group mean differences (95%CI) will be calculated for all outcome measures. In the primary analysis these will be calculated using linear models that include baseline values of outcome variables as covariates to maximise precision.
Discussion
We have presented the rationale and design of an RCT evaluating the effects of the McKenzie Method in the treatment of acute non-specific LBP. The results of this trial will be presented as soon as they are available.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LACM, CGM and RDH were responsible for the design of the study. HC was responsible for recruiting McKenzie therapists and she will also participate as a clinician in the trial. LACM and JMc will act as trial coordinators. All authors have read and approved the final manuscript.
Appendix
Clinical picture and treatment principles according to the McKenzie Method
This table summarises the procedures involved in the McKenzie Method (Table 1). For detailed description of all procedures and progressions, refer to McKenzie's text book. This is particularly important for Derangement syndrome since the treatment is extremely variable and complex and the full description of procedures would not be appropriate for the purposes of this paper.
Table 1 Postural Syndrome
Clinical picture Intermittent back pain under prolonged, static end-range postures (usually flexion); no loss of movement, absence of deformity.
Treatment Patient education and postural correction.
Procedures Patient adopts the posture that produces their symptoms. Physiotherapist instructs patient how to abolish symptoms by correcting the posture and provides explanation on the mechanism that produces pain of postural origin. Attainment of the corrected posture is taught through the use of the "slouch-overcorrect" exercise. Patients are taught how to maintain the corrected posture through the use of a Lumbar roll and actively when a lumbar roll can-not be used. Consequences of postural neglect are discussed.
Dysfunction Syndrome
Clinical picture Intermittent back pain at premature end-range; radiation only in the case of the dysfunction of an adherent nerve root; partial loss of movement.
Treatment Patient education, postural correction, and stretching of contracted structures.
Procedures Posture correction and repeated end-range movements towards the direction of dysfunction (e.g. extension exercises for extension dysfunction). Ten to 15 stretches are repeated at 2/3-hourly intervals, until the movement loss is restored. Treatment progression may include clinician overpressure and/or mobilisation.
Derangement Syndrome
Clinical picture Constant or intermittent back pain and/or leg pain that moves proximally or distally during repeated movements; variable degree of loss of movement; deformity, paraesthesia, numbness and myotomal weakness may be present. A rapid change in the location of symptoms and in the range of movement is seen.
Treatment Reduction of derangement and maintenance of reduction, recovery of function and prophylaxis.
Procedures Reduction of derangement is achieved with sustained positions and/or repeated end-range movements. The treatment principle (extension, flexion or lateral) is selected according to the movements that abolish, decrease or centralise symptoms, as well as those that restore mobility and function (e.g. extension principle is adopted when extension centralises symptoms). Patient generated forces are used as the procedure of first choice. The exercises are repeated at home at 2-hourly intervals or as necessary for pain relief. Forces are progressed when the progress plateaus including over-pressures and therapist mobilisation. To ensure the maintenance of the reduction the patient is instructed to avoid aggravating postures or movements. Lumbar supports are used where necessary for the maintenance of lumbar lordosis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors thank the physiotherapists credentialed in the McKenzie Method for their participation in this project.
==== Refs
Australian Institute of Health and Welfare Australia's health 2004 2004 1st Camberra , AIHW
Australian Acute Musculoskeletal Pain Guidelines Group Evidence-based management of acute musculoskeletal pain. 2003
Maetzel A Li L The economic burden of low back pain: a review of studies published between 1996 and 2001 Best Pract Res Clin Rheumatol 2002 16 23 30 11987929 10.1053/berh.2001.0204
WorkCover Authority NSW Statistical Bulletin. NSW Workers Compensation 2002/03 2003 Sydney , The WorkCover Authority NSW
Pengel LH Herbert RD Maher CG Kathryn RM Acute low back pain: Systematic review of its prognosis BMJ 2003 327 1 5 12842922
Thomas E Silman AJ Croft PR Papageorgiou AC Jayson M Macfarlane GJ Predicting who develops chronic low back pain in primary care: a prospective study BMJ 1999 318 1662 1667 10373170
Guzmán J Esmail R Karjalainen K Malmivaara A Irvin E Bombardier C Multidisciplinary rehabilitation for chronic low back pain: systematic review BMJ 2001 322 1511 1516 11420271 10.1136/bmj.322.7301.1511
van Tulder M Malmivaara A Esmail R Koes B Exercise therapy for low back pain. A systematic review within the framework of the Cochrane Collaboration Back Review Group Spine 2000 25 2784 2796 11064524 10.1097/00007632-200011010-00011
van Tulder M Ostelo R Vlaeyen JWS Linton SJ Morley SJ Assendelft WJJ Behavioral treatment for chronic low back pain. A systematic review within the framework of the Cochrane Back Review Group Spine 2000 25 2688 2699 11034658 10.1097/00007632-200010150-00024
Jellema P van Tulder MW van Poppel MN Nachemson AL Bouter LM Lumbar supports for prevention and treatment of low back pain. A systematic review within the framework of the Cochrane Back Review Group Spine 2001 26 377 386 11224885 10.1097/00007632-200102150-00014
Ferreira ML Ferreira PH Latimer J Herbert RD Maher CG Does spinal manipulative therapy help people with chronic low back pain? Aust J Physiother 2002 48 277 284 12443522
Pengel HM Maher CG Refshauge KM Systematic review of conservative interventions for subacute low back pain Clin Rehabil 2002 16 811 820 12501942 10.1191/0269215502cr562oa
Koes BW van Tulder MW Ostelo R Burton K Waddell G Clinical guidelines for the management of low back pain in primary care: an international comparison Spine 2001 26 2504 2514 11707719 10.1097/00007632-200111150-00022
Borkan J Koes B Reis S Cherkin DC A report from the Second International Forum for Primary Care Research on low back pain: reexamining priorities Spine 1998 23 1992 1996 9779533 10.1097/00007632-199809150-00016
Bouter LM van Tulder MW Koes BW Methodologic issues in low back pain research in primary care Spine 1998 23 2014 2020 9779536 10.1097/00007632-199809150-00019
Leboeuf-Yde C Lauritsen JM Lauritzen T Why has the search for causes of low back pain largely been nonconclusive? Spine 1997 22 877 881 9127921 10.1097/00007632-199704150-00010
Fritz JM Delitto A Erhard RE Comparison of classification-based physical therapy with therapy based on clinical practice guidelines for patients with acute low back pain Spine 2003 28 1363 1372 12838091 10.1097/00007632-200307010-00003
McKenzie R May S The lumbar spine. Mechanical diagnosis & therapy 2003 1 2nd Waikanae , Spinal Publications New Zealand Ltd 374
van Dillen LR Sahrmann SA Norton BJ Caldwell CA McDonnell MK Bloom NJ Movement system impairment-based categories for low back pain: stage 1 validation J Orthop Sports Phys Ther 2003 33 126 142 12683688
BenDebba M Torgerson WS Long DM A validated, practical classification procedure for many persistent low back pain patients Pain 2000 87 89 97 10863049 10.1016/S0304-3959(00)00278-5
Delitto A Erhard RE Bowling RW DeRosa CP Greathouse DG A treatment-based classification approach to low back syndrome: identifying and staging patients for conservative treatment Phys Ther 1995 75 470 485 7770494
Klapow JC Slater MA Patterson TL Doctor JN Atkinson JH Garfin SR An empirical evaluation of multidimensional clinical outcome in chronic low back pain patients Pain 1993 55 107 118 8278203 10.1016/0304-3959(93)90190-Z
Laslett M van Wijmen P Low back and referred pain: diagnosis and proposed new system of classification N Z J Physiother 1999 27 5 14
Maluf KS Sahrmann SA van Dillen LR Use of a classification system to guide nonsurgical management of a patient with chronic low back pain Phys Ther 2000 80 1097 1111 11046197
Petersen T Laslett M Thorsen H Manniche C Ekdahl C Jacobsen S Diagnostic classification of non-specific low back pain. A new system integrating patho-anatomic and clinical categories Physiother Theory Pract 2003 19 213 237
Stiefel F deJonge P Huyse F al INTERMED - An assessment and classification system for case complexity: Results in patients with low back pain Spine 1999 24 378 384 10065523 10.1097/00007632-199902150-00017
McCarthy CJ Arnall FA Strimpakos N Freemont A Oldham JA The biopsychosocial classification of non-specific low back pain: a systematic review Phys Ther Rev 2004 9 17 30 10.1179/108331904225003955
Battié MC Cherkin DC Dunn R Ciol MA Wheeler KJ Managing low back pain: attitudes and treatment preferences of physical therapists Phys Ther 1994 74 219 226 8115455
Li LC Bombardier C Physical therapy management of low back pain: An exploratory survey of therapist approaches Phys Ther 2001 81 1018 1028 11276184
Machado LAC de Souza MS Ferreira PH Ferreira ML The McKenzie protocol for low back pain: a systematic review of the literature with a meta-analysis approach Spine (in press) 2005
de Vet HCWPD Heymans MWMS Dunn KMMP Pope DPPD van der Beek AJPD Macfarlane GJPD Bouter LMPD Croft PRPD Episodes of Low Back Pain: A Proposal for Uniform Definitions to Be Used in Research Spine 2002 27 2409 2416 12438991 10.1097/00007632-200211010-00016
Pocock SJ Clinical trials. A practical approach 1984 1st Chichester , John Wiley & Sons
Delitto A Cibulka MT Erhard RE Bowling RW Tenhula JA Evidence for use of an extension-mobilization category in acute low back syndrome: A prescriptive validation pilot study Phys Ther 1993 73 216 228 8456141
Schenk RJ Jozefczyk C Kopf A A randomized trial comparing interventions in patients with lumbar posterior derangement J Manual Manip Ther 2003 11 95 102
Farrar J Young J LaMoreaux L al Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale Pain 2001 94 149 158 11690728 10.1016/S0304-3959(01)00349-9
Stratford P Gill C Westaway M Binkley J Assessing disability and change on individual patients: a report of a patient specific measure Physiother Can 1995 47 258 263
Roland M Morris R A study of the natural history of back pain. Part I: development of a reliable and sensitive measure of disability in low-back pain Spine 1983 8 141 144 6222486
Kalauokalani D Cherkin D Sherman K Koepsell T R D Lessons from a trial of acupuncture and massage for low back pain Spine 2001 26 1418 1424 11458142 10.1097/00007632-200107010-00005
Werneke M Hart DL Cook D A descriptive study of the centralization phenomenon. A prospective analysis Spine 1999 24 676 683 10209797 10.1097/00007632-199904010-00012
|
16221311
|
PMC1274327
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Oct 13; 6:50
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-50
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-511622974810.1186/1471-2474-6-51Technical AdvanceA sonography assisted technique for the removal of a femoral interlocking nail – a technical note Tsai Kai-Jow [email protected] Po-Wen [email protected] William C [email protected] Department of Orthopaedic Surgery Cathay General Hospital, Taipei, Taiwan2 Emory Orthopaedic and Spine Center, Department of Orthopaedic Surgery, Emory University, Atlanta, GA, USA2005 17 10 2005 6 51 51 19 5 2005 17 10 2005 Copyright © 2005 Tsai et al; licensee BioMed Central Ltd.2005Tsai et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Open methods for removal of femoral interlocking nails involve an incision (up to 10 cm) over the trochanter to find the tip of the nail. The distal locking screws are some times difficult to palpate and an incision (up to about 5 cm) is often needed for exposure. Intra-operative fluoroscopy is often used as an adjunct technique to minimize the surgical wound. However, patients and surgeons are exposed to a radiation hazard. Sonography can provide a real-time and efficient alternative to fluoroscopy.
Methods
Sonography of soft tissue has been established to identify a foreign body. A metallic implant has a hyperechoic image; therefore, we can identify the correct position of the screws preoperatively and intraoperatively.
Results
We have developed a technique using sonography and minimal incisions for the removal of a femoral interlocking nail. The proximal wound is 2.5 cm in length and the wound is 0.5 cm in length for each distal locking screw.
Conclusion
The sonography can be used to minimize the length of incision and prevent radiation exposure in the removal of intramedullary femoral nails.
==== Body
Background
The development of closed interlocking intramedullary nailing has allowed the treatment of femoral diaphyseal fractures to become safer and more effective [1,2]. The nail is usually inserted under fluoroscopic control which brings concern over the radiation exposure [3]. There have been efforts to minimize the fluoroscopic radiation [4]. Ultrasound, on the other hand, is cheaper and more easily available and can be used to monitor alignment during closed femoral nailing [5]. Thus, using ultrasound can reduce the fluoroscopic monitoring time and reduce the radiation exposure to the patient and the surgeon.
It is often necessary to remove femoral nails after bony union. Conventional open methods require up to a 10 cm incision over the trochanter. The distal locking screws are difficult to palpate, and open distal incisions are often needed. Fluoroscopy is frequently used in an attempt to decrease the size of the wound.
Sonography for evaluation of soft tissues has been in use for years. The sonographic signal is reflected by cortical bone [6], and any metallic implant has a hyperechoic image. Therefore sonography can identify the position of locking screws. We applied sonography for the wound of the removal of distal locking screws using a minimal incision. We report on this technique that was used successfully in three patients.
Methods
Preoperative localization by sonography
After bony union was achieved, the patient had surgery for the removal of the intramedullary nail. At this time the swelling had subsided and the scar contracture was not aligned to indicate the position of the distal locking screw head. The conventional method to find the screw is to open the skin, fascia, and muscles and expose the bony cortex; the screw is then extracted under direct vision. Using this conventional method there is extensive invasion of the soft tissues. Although by using palpation or intra-operative fluoroscopy, the surgeon can sometimes locate and remove the screws without direct vision. We used Sonosite® 180 plus (Bothell Washington) with L38/10.5 MHz linear array for the purpose. The metal implants have a characteristic sonographic appearance. The wide difference in acoustic impedance between soft tissue and metal results in an extremely bright interface, with a posterior "comet-tail" reverberation artefact [7]. The image of the two hyperechoic reflection and comet-tail indicates the distal locking screw heads (Figure 1). A 0.5-cm incision is made over the screw head and the screwdriver is inserted under real-time sonographic assistance.
Figure 1 Sonography of screws. Two distal locking screws can be visualized by sonography. The image of two extremely brightness with "comet-tails" indicate the distal locking screw heads.
Locking screws removal
There were two distal locking screws and one proximal locking screw. These three screws form the long axis of femoral canal. The long axis is guidance for removal of the nail. The screws are not removed completely but left protruding from the skin to indicate the long axis of femoral shaft (Figure 2).
Figure 2 Long axis of femur. The three points of the locking screws form a line that indicates the long axis of femur.
The 2.5-cm incision on the tip of the long axis over the buttock
The skin incision for the removal of the intramedullary nail is at the tip of the long axis over the buttock. The muscles over piriformis fossa were dissected with fingers and the guide pin was inserted through the fascia. The bony structure can be palpated by guide pin. While these three locking screws revealed the long axis of intramedulary nail, the guide pin is inserted to the nail in the medullary cavity. The custom-made tube sleeve is used in the minimally invasive technique for soft tissue protection (Figure 3). The sleeve maintains the direction and prevent any injury to muscles, nerve and vessels [8].
Figure 3 Tube sleeve. The custom-made tube sleeve is used in the minimal invasive technique for soft tissue protection.
Results
The authors have used this technique successfully in three patients with no failures. We present a typical patient who suffered from a fall that induced a fracture of his left femur. He was treated with a femoral interlocking nail and the fracture eventually healed one year later. The removal of the interlocking nail was performed using this minimally invasive technique. The proximal wound was about 2.5 cm in length; the previous wound had been about 10 cm in length (Figure 4).
Figure 4 2.5 cm proximal wound. The proximal wound is 2.5 cm in length, which is smaller than previous wound that is 10 cm in length.
Discussion
The applications of sonography to removal of surgical implant have been documented in gynecology literature. Nelson et al reported on real-time sonographic localization and guidance could enable safe removal of deeply placed, nonpalable and intramuscular contraceptive capsules [9]. High resolution sonography allowed accurate localization of a foreign body in the soft tissue in spite of radio-lucent or radio-opaque [10].
Gynaecologists are familiar with sonography, while orthopaedic surgeons are familiar with fluoroscopy. Intra-operative fluoroscopy has been widely used for many procedures, such as closed reduction, internal fixation and removal of implant. However, the removal of implant can also be achieved by sonography, because the metal implants located on the surface of bone and have high echogenecity and are well distinguished from the other tissues [11].
The sonography of musculoskeletal system has various applications including detection of an occult fracture, reduction of fracture, assessment of joint fluid, and identification of a foreign body[5,6,12,13]. During the removal of a metal implant, the sonography can provide real-time guidance to apply the screw driver to the screw head and thus assist in the removal of screws.
The authors have used this procedure to remove three femoral interlocking nails. However, most high frequency probes have a limited depth of view, typically 3–4 cm at 12 MHz. When assessing deeper structures, 5-MHz curvilinear probes can give a deeper and wider view [7]. Therefore, the morbidity obesity, heterotopic bone growth or deep seated implant in bone may be not feasible for this technique. The surgeons should consider another modality such as intra-operative fluoroscopy or conventional open procedures.
Conclusion
The sonography can be used to minimize the incision length and radiation exposure in the removal of intramedullary femoral nails.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KJT: Original idea for this procedure, organize and write the text.
PWS: Advice and support.
WCH: Advice and write the text.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Nil.
==== Refs
Kempf I Grosse A Beck G Closed locked intramedullary nailing. Its application to comminuted fractures of the femur J Bone Joint Surg Am 1985 67 709 720 3997923
Winquist RA Hansen STJ Clawson DK Closed intramedullary nailing of femoral fractures. A report of five hundred and twenty cases J Bone Joint Surg Am 1984 66 529 539 6707031
Levin PE Schoen RWJ Browner BD Radiation exposure to the surgeon during closed interlocking intramedullary nailing J Bone Joint Surg Am 1987 69 761 766 3597477
Tremains MR Georgiadis GM Dennis MJ Radiation exposure with use of the inverted-c-arm technique in upper-extremity surgery J Bone Joint Surg Am 2001 83-A 674 678 11379736
Mahaisavariya B Suibnugarn C Mairiang E Saengnipanthkul S Laupattarakasem W Kosuwon W Ultrasound for closed femoral nailing J Clin Ultrasound 1991 19 393 397 1658064
Hubner U Schlicht W Outzen S Barthel M Halsband H Ultrasound in the diagnosis of fractures in children J Bone Joint Surg Br 2000 82 1170 1173 11132281 10.1302/0301-620X.82B8.10087
Gibbon WW Long G Barron DA O'Connor PJ Complications of orthopedic implants: sonographic evaluation J Clin Ultrasound 2002 30 288 299 12116109 10.1002/jcu.10065
Tsai KJ Liaw JK Lin CC Hou SM Minimally invasive technique in compression hip screw insertion. Journal of orthopedic surgery, Taiwan, ROC 2001 18 130 135
Nelson AL Sinow RM Real-time ultrasonographically guided removal of nonpalpable and intramuscular Norplant capsules Am J Obstet Gynecol 1998 178 1185 1193 9662300
Amann P Botta U Montet X Bianchi S Sonographic detection and localization of a clinically nondetectable subcutaneous contraceptive implant J Ultrasound Med 2003 22 855 859 12901417
Cardinal E Chhem RK Beauregard CG Ultrasound-guided interventional procedures in the musculoskeletal system Radiol Clin North Am 1998 36 597 604 9597077 10.1016/S0033-8389(05)70048-8
Grechenig W Peicha G Clement H Fellinger M Mayr J [Ultrasonography in trauma] Orthopade 2002 31 143 153 11963479 10.1007/s00132-001-0235-3
Graif M Stahl-Kent V Ben-Ami T Strauss S Amit Y Itzchak Y Sonographic detection of occult bone fractures Pediatr Radiol 1988 18 383 385 3050844
|
16229748
|
PMC1274328
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Oct 17; 6:51
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-51
|
oa_comm
|
==== Front
BMC Nucl MedBMC Nuclear Medicine1471-2385BioMed Central London 1471-2385-5-51623232310.1186/1471-2385-5-5Research ArticleHydrophilic and lipophilic radiopharmaceuticals as tracers in pharmaceutical development: In vitro – In vivo studies Terán Mariella [email protected] Eduardo [email protected] Andrea [email protected] Malcolm [email protected] Cátedra de Radioquímica – Facultad de Química – Universidad de la República. Montevideo, Uruguay2 Radiopharmacy Unit. Queen's Medical Centre, Nottingham University. Nottingham, UK2005 18 10 2005 5 5 5 21 3 2005 18 10 2005 Copyright © 2005 Terán et al; licensee BioMed Central Ltd.2005Terán et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Scintigraphic studies have been performed to assess the release, both in vitro and in vivo, of radiotracers from tablet formulations. Four different tracers with differing physicochemical characteristics have been evaluated to assess their suitability as models for drug delivery.
Methods
In-vitro disintegration and dissolution studies have been performed at pH 1, 4 and 7. In-vivo studies have been performed by scintigraphic imaging in healthy volunteers. Two hydrophilic tracers, (99mTc-DTPA) and (99mTc-MDP), and two lipophilic tracers, (99mTc-ECD) and (99mTc-MIBI), were used as drug models.
Results
Dissolution and disintegration profiles, differed depending on the drug model chosen. In vitro dissolution velocity constants indicated a probable retention of the radiotracer in the formulation. In vivo disintegration velocity constants showed important variability for each radiopharmaceutical. Pearson statistical test showed no correlation between in vitro drug release, and in vivo behaviour, for 99mTc-DTPA, 99mTc-ECD and 99mTc-MIBI. High correlation coefficients were found for 99mTc-MDP not only for in vitro dissolution and disintegration studies but also for in vivo scintigraphic studies.
Conclusion
Scintigraphic studies have made a significant contribution to the development of drug delivery systems. It is essential, however, to choose the appropriate radiotracers as models of drug behaviour. This study has demonstrated significant differences in release patterns, depending on the model chosen. It is likely that each formulation would require the development of a specific model, rather than being able to use a generic drug model on the basis of its physicochemical characteristics.
==== Body
Background
Development of new drug formulations requires the performance of extensive studies, both in the laboratory, and in vivo, in animals and in volunteers. In vitro studies can be very expensive but costs are even higher when in vivo stages are reached. Methodology that can generate relevant information but shorten the preformulation phases means important savings in economic, human and time terms [1-9]. Gamma scintigraphy provides rapid, complementary information that often cannot be obtained by other methodologies. It has been successfully used during development stages of feasibility studies and in determining specific parameters of the final product. The information obtained by scintigraphy gives support during investigation and development, and also complements the development of registration dossiers and marketing publicity. The incorporation of a radiopharmaceutical into a drug formulation allows determination of the biodistribution kinetics and the release sites [10-14]. It is very important to choose the proper radionuclide, often 99mTc (technetium), as this has optimal characteristics of half-life and energy, allowing images to obtain with high efficiency and low doses [15,16].
Many studies to validate the methodology mentioned above have already been undertaken [17,18], but, in general within these studies, radiopharmaceuticals are only rarely used as drug surrogates. Radiopharmaceuticals were primarily developed for diagnostic purposes in nuclear medicine, most of them being intended for intravenous administration. In order to optimise the usage of radiopharmaceuticals in the development of pharmaceutical drug-delivery systems, the behaviour of these tracers following administration by other routes must be validated. Stability and in vitro studies of different radiopharmaceuticals have been previously reported by our group [17], with the aim of creating a database of radiotracers or model drugs with known physicochemical properties to be used as in pharmaceutical dosage development, particularly of tablet formulations.
Provided that the tablet formulation can be radiolabelled with a suitable gamma emitter without altering its characteristics, imaging with a gamma camera can be used to monitor in vivo transit and dispersion of the tablet, and give some indication of deposition and absorption of the drug. The aim of this work is to assess in vivo and in vitro behaviour of a tablet formulation using scintigraphic studies, to examine the way in which radiopharmaceuticals model the release of drugs, using four different tracers with different physicochemical characteristics. Based on this assessment, the correlation between in-vitro and in-vivo behaviour could be investigated.
The characterised radiopharmaceuticals were 99mTc-diethylenetriamine-pentaacetic acid (99mTc-DTPA), 99mTc-ethyl cysteinate dimer (99mTc-ECD), 99mTc-methylene diphosphonate (99mTc-MDP) and 99mTc-sestamibi (99mTc-MIBI) [17]. They were incorporated into the tablets during wet granulation. Disintegration profiles were assessed in vitro in dissolution vessels and in vivo by scintigraphic imaging in healthy volunteers. In vitro dissolution studies were performed for tablets containing the above-mentioned radiotracers
Methods
Radiopharmaceuticals labelling and control
The radiopharmaceuticals 99mTc-DTPA, 99mTc-ECD, 99mTc-MDP and 99mTc-MIBI were obtained by labelling commercial kits with 99mTcO4- from a 99Mo /99mTc generator (Technonuclear). Radiochemical purity testing of the 99mTc-DTPA complex was performed by chromatographic analysis on Whatman N° 1 paper with propanone and sodium chloride solution (0.9%) as developing solvents.
99mTc-ECD radiochemical purity was studied by chromatography on Whatman N°1 paper with a methanol/water (85/15) solvent and by HPLC (Shimadzu LC-10 AS) using a Partisphere C18 (Whatman) column as the stationary phase. The mobile phases were phosphate buffer 0.0125 M pH 2.5 (A), and ethanol 100% (B). Solvent program was, at time = 0 min, 100 % A, and at time = 10 min, 70% A, 30% B. The flow rate was set at 2 mL min-1.
The radiochemical purity of 99mTc-MDP was determined by chromatography on Whatman N°1 paper using propanone and sodium chloride 0.9 % as developing solvents.
The 99mTc-MIBI complex was assessed by HPLC (Shimadzu LC-10 AS) using a Partisphere C18 (Whatman) column 12.5 cm as stationary phase. The mobile phases were methanol (A), and ammonium sulphate 0.05 M (B). Solvents program was at time = 0 min, 20 % A, and at time = 5 min, 95% A. The flow rate was set at 2 mL min-1.
Tablet preparation
Tablets were prepared by wet granulation using lactose and starch as diluents, PVP K30 (ISP Corp) as granulating agent, Ac-Di-Sol (FMC Biopolymer) as disintigrant and magnesium stearate as lubricant. The exipients were passed through a No. 16 sieve (16 meshes per centimetre) and mixed by a geometric method. After compression, tablets were assessed for weight (target weight 400 mg) and radioactivity content (target activity 18.5 MBq). Activity was measured in an ionisation chamber dose calibrator (Capintec CRC 25).
Tracer stability was verified as in a previous communication [18] both during tablet preparation and in dissolution and disintegration studies.
In vitro studies
Dissolution tests were carried out in a Vankel USP Type II apparatus, 900 mL of dissolution medium was used at three different pH values comprising pH 1 (0.1 M hydrochloric acid), pH4 (0.016 M acetic acid/sodium acetate) and pH 7 (water). Temperature was maintained at 37°C, and a rotation speed of 50 r.p.m. used. Samples of 2 mL were withdrawn through cellulose acetate membrane filters (0.22 μm) at 1, 2, 3, 5, 10, 15 and 30 minutes and assessed for radioactivity in a solid scintillation counter (Ortec-Maestro MCB1). Corrections were applied for volume, decay and counting efficiency. Curves of log % non-dissolved vs. time were plotted.
Disintegration profiles were recorded concurrently with the dissolution procedure by placing the vessels in front of a circular field of view gamma camera (Dyna (Pyker 4/15) equipped with a low energy, high resolution collimator, operating with a 20% window centred on 140 Kev. Serial images were recorded at 1 frame/min with a matrix of 128 × 128 without acquisition zoom. Three regions of interest were delineated, one on the tablet, one on the dissolution medium and the third on the external field to determine the background [19]. Net activity was calculated by subtracting mean background pixel counts from each pixel in the selected region of interest. No attenuation correction was necessary.
In vivo studies
Studies were performed in four healthy volunteers (2 men and 2 women) mean weight 60 Kg, mean age 38 years. They were non smokers and were not receiving any medication. Written informed consent was obtained from all subjects. They were not allowed to consume alcohol during the study period or in the preceding 24 hours. The protocol was in accordance with the Declaration of Helsinki guidelines for ethics in research and with resolutions XXIX and XXXV of the World Medical Assembly.
Volunteers fasted for 6 hours prior to the study. Tablets were administered with 200 mL of still mineral water. Volunteers stood erect during the acquisition process in front of the detector. Urine samples were collected at 30 and 120 minutes post ingestion [22,23].
Serial images were recorded during the 30 minute period after administration at 2 frames / min using a rectangular field view gamma camera (Sophy). Later static images were recorded 2 hours after tablet intake.
Three regions of interest were delineated, one on the stomach, one on the intestines and a third on a region outside the body perimeter to determine the background. Net activity was quantified by subtracting mean background pixel counts from each pixel in the selected region of interest. Gamma ray attenuation was quantified by calculating the geometric means of count rates in paired anterior and posterior planar views.
Scintigraphic studies enabled the determination of tablet transit characteristics within the gastrointestinal tract specifically in the stomach. Curves of Log % non-disintegrated vs. time were plotted [24,25].
Data processing
Disintegration velocity constant of the tablets was kd, which corresponded to the slope of the linear regression in first order kinetics [26]:
log [1- Qt/Q∞] = kdt / 2.303
Where
Qt the amount of activity of the tracer disintegrated at time t present in the region of interest.
Q∞ the maximum amount of activity measured in the region of interest.
Disintegration velocity constants in the stomach and the constant of appearance in small intestine were determined in a similar way, considering Qt the amount of activity of the tracer disintegrated at time t present in the region of interest and Q∞ the total amount of activity of the region at the end of the study [27].
Results and Discussion
Radiotracers were incorporated during the wet granulation process of tablet preparation and appropriate controls of quality were applied as previously described [17]. All the radiopharmaceuticals were produced with radiochemical purities higher than 95 %, in accordance with manufacturers' specifications.
Disintegration velocity constants in the gastrointestinal tract were determined by in vivo scintigraphic studies, specifically in stomach and small intestine for the different radiopharmaceuticals evaluated as tracers.
In vitro results
99mTc-MIBI
Dissolution velocity constants for tablets containing this lipophilic tracer showed no significant variation across the whole pH range studied and the values were very similar to disintegration velocity constants determined under the same conditions (Table 1).
Table 1 In vitro dissolution and disintegration constants (p 0.05)
Radiopharmaceutical Dissolution velocity Constant (min-1) (n = 6) Disintegration velocity Constant (min-1) (n = 6)
99mTc-MIBI (0.05 ± 0.02) all pH range (0.04 ± 0.01) all pH range
99mTc-ECD (0.040 ± 0.005) all pH range (0.05 ± 0.01) (pH 1)
(0.09 ± 0.01) (pH4 and 7)
99mTc-MDP (0.020 ± 0.005) all pH range (0.20 ± 0.04) (pH 1 and 7)
(0.040 ± 0.005) (pH 4)
99mTc-DTPA (0.06 ± 0.01) (pH1and 7)
(0.11 ± 0.02) (pH 4) 0.011 ± 0.002 (pH1and 7)
0.080 ± 0.006 (pH 4)
The Pearson statistical test was used to quantify correlation within both series of data. Pearson's correlation coefficient r is always between -1 and 1 indicating that the points are near a negative or a positive slope respectively. The closer the r value is to -1 or 1 the higher the correlation of the series is. [27]
Dissolution and disintegration velocity constants (r values) were compared using this test. Correlation coefficients are shown in Table 3.
Table 3 In vitro dissolution – disintegration Pearson correlation coefficient
pH 99m Tc MIBI 99m Tc ECD 99m Tc MDP 99m Tc DTPA
1 0.75 0.99 0.68 0.99
4 0.77 0.78 0.90 0.99
7 0.50 0.94 0.68 0.99
They were not so high indicating a probable retention of the radiotracer in the formulation.
99mTc-ECD
This lipophilic radiopharmaceutical demonstrated identical dissolution velocity constants across the whole pH range studied, while disintegration velocity constants increased with pH (Table 1).
Pearson correlation coefficients are shown in Table 3. A high correlation was found.
99mTc-MDP
In vitro dissolution velocity constants did not show significant variation across the whole pH range while in vitro disintegration velocity constants increased with pH (Table 1). In vitro correlation coefficients were high at all pH values. (Table 3)
99mTc-DTPA
In vitro dissolution and disintegration data showed similar patterns but the values were different (Table 1). In this case at pH 1 and 7 both disintegration and dissolution velocity constants were lower than those at pH 4. The profiles were similar in both series of data and Pearson correlation was very high at all pH ranges studied. (Table 3)
In vivo results
99mTc- MIBI
No significant variability within volunteers was observed in stomach disintegration velocity constants. Small intestine appearance constant showed similar behaviour but different values and did not represent significant differences within volunteers (Table 2). Two hours after administration there was no significant uptake in any organ except urinary bladder (Figure 1). Pearson Test was used to compare in vitro – in vivo disintegration velocity constants (Table 4) assuming pH 1 for stomach and pH 4 for small intestine no correlation was found. When Pearson test was used for in vivo disintegration velocity constants and in vitro dissolution ones, (Table 5) correlation coefficient was again low for stomach and a bit higher for small intestine, but in neither case was correlation considered to be established.
Table 2 In vivo gastric transit constants (p 0.05)
Radio pharmaceutical Stomach disintegration velocity Constant (min-1) (n = 4) Small intestine appearance Constant (min-1) (n = 4)
9mTc-MIBI (0.016 ± 0.006) (0.13 ± 0.02)
99mTc-ECD From (0.7 ± 0.1) to (0.010 ± 0.002) (0.02 ± 0.01)
99mTc-MDP (0.007 ± 0.002) (0.037 ± 0.006)
99mTc-DTPA From (0.13 ± 0.03) to (0.07 ± 0.01) From (0.004 ± 0.001) to (0.08 ± 0.01)
Figure 1 99mTc-MIBI placebo tablet scintigraphic image 2 hours after drug intake for one volunteer.
Table 4 In vitro-In vivo disintegration velocity constants Pearson correlation coefficient.
pH 99m Tc MIBI 99m Tc ECD 99m Tc MDP 99m Tc DTPA
1 0.29 0.44 0.29 0.8
4 0.21 0.29 0.08 0.12
Table 5 In vitro dissolution – in vivo disintegration Pearson correlation coefficient
pH 99m Tc MIBI 99m Tc ECD 99m Tc MDP 99m Tc DTPA
1 0.13 0.08 0.68 0.07
4 0.51 0.69 0.59 0.12
99m Tc ECD
In this particular case there was a substantial inter subject variability with coefficient variations during the 30-minute study period ranging from 14 to 95 % (Table 2). Small intestine appearance velocity constants were more homogeneous and late images 2 hours after administration showed liver uptake (Figure 2).
Figure 2 Scintigraphic image 2 hours after drug intake for one volunteer 99mTc-ECD placebo tablets.
Correlation coefficient was not significant when in vitro disintegration – in vivo appearance velocity constants were compared (Table 4).
When Pearson correlation was quantified between in vivo disintegration velocity constants and in vitro dissolution ones (Table 5) as indicated for 99mTc-MIBI, there was no correlation at pH 1 (stomach) and it was low at pH 4 (small intestine).
99m Tc- MDP
No significant variations within volunteers were observed in stomach disintegration velocity constants. Small intestine appearance constant showed similar behaviour but different values and did not represent significant differences within volunteers (Table 2). Two hours after administration there was no significant uptake in any organ except urinary bladder (Figure 3). Pearson Test was used to compare in vitro – in vivo disintegration velocity constants (Table 4). In this case correlation was negligible. When Pearson Test was used for in vivo disintegration velocity constants and in vitro dissolution ones, (Table 5) correlation coefficients had the highest values found in this study.
Figure 3 Scintigraphic image 2 hours after drug intake for one 99mTc-MDP placebo tablets in one volunteer.
99m Tc- DTPA
Significant variations within volunteers were observed in stomach disintegration velocity constants and small intestine appearance constants (Table 2). Two hours after administration there was no significant uptake in any organ except urinary bladder (Figure 4). Pearson Test was used to compare in vitro – in vivo disintegration velocity constants (Table 4).
Figure 4 Scintigraphic image 2 hours after drug intake for one volunteer 99mTc-DTPA placebo tablets.
In this case correlation was negligible for pH 4 and higher for pH 1. When Pearson Test was used for in vivo disintegration velocity constants and in vitro dissolution ones, (Table 5) correlation was negligible for the entire gastrointestinal tract.
A good correlation was found between in vitro disintegration and dissolution velocity constants of tablets containing each of the radiopharmaceuticals used as radiotracer.
As scintigraphic studies give information of a physical process, tablets containing the same components, but different tracers, might be expected to have a similar behavior in all cases. As tracers are present in low concentrations (lower than 10-9 M), it is unlikely that they are responsible for the observed differences in the disintegration constants. It is possible that the tracers are bound to different components within the tablet, which disperse at different rates, which would be a likely explanation. Although the dissolution profiles might be expected to differ depending on the model drug used, the same would not be expected of the disintegration profile. This suggests that the measured disintegration profile can depend on the model chosen.
The radiopharmaceutical 99mTc ECD showed absorption in the gastrointestinal tract. This is not a desirable characteristic and makes it an unsuitable tracer for this type of study. It was probably the main factor giving rise to the greater inter individual variability disintegration velocity constants observed when in vivo studies were performed for 99mTc-ECD formulations. This fact was not observed with the other radiotracers under the same conditions of study.
Despite the physicochemical differences between 99mTc-MIBI and 99mTc-DTPA, both tracers presented very low correlation coefficients when in vitro dissolution – in vivo disintegration velocity constants were compared throughout the whole gastrointestinal tract (Table 3).
Only 99m Tc-MDP showed high correlation coefficients both in vitro between dissolution and disintegration, and between in vitro dissolution and in vivo disintegration.
Conclusion
Even though 99mTc-ECD, 99mTc-MIBI and 99mTc-DTPA showed high in vitro correlation between dissolution and disintegration, statistical tests revealed that none of them was an adequate predictor of in vivo performance for this particular tablet formulation, for any region of the gastrointestinal tract.
The hydrophilic radiotracer 99mTc-MDP was the only radiopharmaceutical suitable as a drug model for this particular tablet formulation in the prediction of its in vivo behavior.
The method is an interesting tool especially at early stages of pharmaceutical formulation development when different formulations are being chosen, but may not have adequate sensitivity to discriminate between formulation variables.
Careful choice of drug model, together with substantial in vitro validation is essential in order to reduce in vivo studies and make significant savings of human and financial resources.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Mrs. Mariella Terán carried out all the experimental work and together with Dr. Eduardo Savio made substantial contributions to conception, design, analysis and data interpretation. They also gave their final approval of the version to be published. Mrs. Andrea Paolino contributed to data acquisition, analysis and interpretation. Dr. Malcolm Frier was involved in revising it critically for important intellectual content.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors thank TECHI S.A. for supplying radiopharmaceutical kits and Centro de Medicina Nuclear, Hospital de Clínicas for Nuclear Medicine equipment facilities.
==== Refs
Wilson CG Perkins AC Gamma Scintigraphy and the study of drug deposition. Advances in pharmaceutical sciences 1992 London Academic Press
Alan Perkins Malcolm Frier Nuclear Medicine in Pharmaceutical Research 1999 London Taylor and Francis 10614836
Wilson CG Washington N. Chichester Physiological Pharmaceutics: Biological Barriers to Drug Absorption 1989 London Ellis Horwood
Budisantoso N Frier M Wilson CG Lipophilic models for scintigraphic evaluation of drug delivery systems J Pharmacy and Pharmacology 1997 49 49
Adkin DA Davis SS Sparrow RA Huckle PD Phillips AJ Wilding IR The effects of pharmaceutical excipients on small intestinal transit Br J Clin Pharmac 1995 39 381 387
Wood E Wilson CG Hardy JG The spreading of foam and solution enemas Int J Pharm 1985 25 191 197 10.1016/0378-5173(85)90092-4
Newman SP Woodman G Clarke SW Deposition of carbenicillin aerosols in cystic fibrosis; efects of nebuliser system and breathing pattern Thorax 1988 43 318 322 3406919
Hak-Kim Chan Evangelina Daviskas Stefan Eberl Michael Robinson George Bautovich Iven Young Deposition of aqueous aerosol of 99mTc-DTPA and delivered by novel system (AER) in healthy subjects Eur J Nucl Med 1999 26 4
Bartlett RJV Perkins A Ware FW Riley A Robinson PJA Reproducibility of oesophageal transit studies: several single swallows must be performed Nucl Med Commun 1987 8 317 326 3317159
Hardy JG Radiopharmaceuticals: using radioactive compounds in pharmaceutics and medicine. Radionucleide imaging in drug formulation 1989 London Ellis Horwood Limited
Dams TTM Cortens H Lessons for medicine and nuclear medicine research Eur J Nucl Med 1999 26 311 313 10199934 10.1007/s002590050391
Newman SP Wilding IR Imaging techniques for assessing drug delivery in man Pharm Sci Technol Today 1999 2 181 189 10322380 10.1016/S1461-5347(99)00152-2
Davies SS Hardy JG Newman SP Wilding IR Gamma scintigraphy in the evaluation of pharmaceutical dosage forms. Review article European Journal of Nuclear Medicine 1992 19 971 986 1425786
Perkins AC Frier M Nuclear medicine techniques in the evaluation of pharmaceutical formulations Pharmacy World and Science 1996 18 97 104 8826534 10.1007/BF00417757
Early PJ Sodee DB Principles and Practice of Nuclear Medicine 1995 Second USA. Mosby -Year Book Inc 94 117
Jurisson S Berning D Wei Jia Dangshe Ma Coordination compounds in Nuclear Medicine Chemical Reviews 1993 3 1137 1156 l 93 10.1021/cr00019a013
Terán M Paolino A Savio E Gaudiano JP León AS Frier M Centellografía en el desarrollo farmacéutico: Estudios in vitro e in vivo de comprimidos de ranitidina Acta Farmacéutica Bonaerense 2000 19 217 224
Terán M Savio E Paolino A Frier M Usage of radiopharmaceuticals in the development of pharmaceutical drug delivery systems: validation of 99mTc-DTPA and 99mTc-ECD European Journal of Pharmaceutics and Biopharmaceutics 2004 57 347 352 15018995 10.1016/j.ejpb.2003.11.003
Donald Burns H Raymond Gibson E Robert Dannals Peter KS Nuclear Imaging in Drug Discovery, Development and Approval 1993 Siegl Birkhäuser-Boston
David Ganderton Advances in Pharmaceutical Sciences 1992 Academic Press London, England
Wilding IR Using product visualization techniques to characterize dosage form performance in man British Pharmaceutical Conference Abstract Book 2001 306
Wilson CG Mc Jury M O'Mahoy B Frier M Perkins AC Imaging of oily formulations in the gastrointestinal tract Advanced Drug Delivery Reviews 1997 25 91 101 10.1016/S0169-409X(96)00493-0
Perkins AC Mann C Wilson CG Three dimensional visualisation of the large bowel: A potential tool for assessing targeted drug delivery and colonic pathology Eur J Nucl Med 1995 22 1035 1038 7588941 10.1007/BF00808416
Perkis AC Oesophageal transit, disintegration and gastric empting of a film – coated resindronate placebo tablet in gastro – oesophageal reflux disease and normal control subjects Aliment Pharmacol Ther 2001 15 115 121 11136284 10.1046/j.1365-2036.2001.00865.x
Perkins AC Esophageal transit of risedronate cellulose – coated tablet and gelatin capsule formulations International Journal of Pharmaceutics 1999 186 169 175 10486435 10.1016/S0378-5173(99)00172-6
Cid Cárcamo E Control de calidad biofarmacéutico de medicamentos 1993 Ed Churcill Livingstone Spain
Donald L Smith Probability, Statistics and Data Uncertainties in Nuclear Science and Technology American Nuclear Society 1991 4 LaGrange Park, Illinois, USA
|
16232323
|
PMC1274329
|
CC BY
|
2021-01-04 16:30:51
|
no
|
BMC Nucl Med. 2005 Oct 18; 5:5
|
utf-8
|
BMC Nucl Med
| 2,005 |
10.1186/1471-2385-5-5
|
oa_comm
|
==== Front
BMC NursBMC Nursing1472-6955BioMed Central London 1472-6955-4-51622570410.1186/1472-6955-4-5Research ArticleAre patient falls in the hospital associated with lunar cycles? A retrospective observational study Schwendimann René [email protected] Franco [email protected] Sabina De [email protected] Koen [email protected] Institute of Nursing Science, University of Basel, Bernoullistrasse 28, 4056 Basel, Switzerland2 Stadtspital Waid Zurich, Switzerland3 Institute of Astronomy, Swiss Federal Institute of Technology, Zurich, Switzerland4 Center for Health Services and Nursing Research, Catholic University of Leuven, Leuven, Belgium2005 17 10 2005 4 5 5 25 7 2005 17 10 2005 Copyright © 2005 Schwendimann et al; licensee BioMed Central Ltd.2005Schwendimann et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Falls and associated negative outcomes in hospitalized patients are of significant concerns. The etiology of hospital inpatient falls is multifactorial, including both intrinsic and extrinsic factors. Anecdotes from clinical practice exist in which health care professionals express the idea that the number of patient falls increases during times of full moon. The aim of this study was to examine in-hospital patient fall rates and their associations with days of the week, months, seasons and lunar cycles.
Methods
3,842 fall incident reports of adult in-patients who fell while hospitalized in a 300-bed urban public hospital in Zurich, Switzerland were included. Adjusted fall rates per 1'000 patient days were compared with days of the week, months, and 62 complete lunar cycles from 1999 to 2003.
Results
The fall rate per 1000 patient days fluctuated slightly over the entire observation time, ranging from 8.4 falls to 9.7 falls per month (P = 0.757), and from 8.3 falls on Mondays to 9.3 falls on Saturdays (P = 0.587). The fall rate per 1000 patient days within the lunar days ranged from 7.2 falls on lunar day 17 to 10.6 falls on lunar day 20 (P = 0.575).
Conclusion
The inpatient fall rates in this hospital were neither associated with days of the week, months, or seasons nor with lunar cycles such as full moon or new moon. Preventive strategies should be focused on patients' modifiable fall risk factors and the provision of organizational conditions which support a safe hospital environment.
==== Body
Background
Falls occur frequently in hospitalized patients. Patient fall rates in hospital settings vary from 2.2 to 9.1 falls per 1000 patient days depending on patient populations and disease groups [1-7]. The etiology of falls in hospitalized patients is multifactorial consisting of both intrinsic and extrinsic risk factors [8-10]. Studies on hospital falls that focus on occurrences over time are limited to the frequencies of falls during the hours of the day [1,5-7,11,12], and to specific time spans e.g. number of falls within the first week of hospitalization [2,4,7,13].
Reasons for the fluctuation in fall-rates over time have been debated, but never scientifically researched. There exist anecdotes from health care professionals in our clinical practice that express the idea that the number of patient falls increasing during times of full moon. One survey indicated that specifically mental health professionals including psychologists, nurses and others held the personal belief that lunar phases affect patient's behavior [14]. However, only one study could be found which reports an increased frequency in patient accidents in a hospital, of which 90% were patient falls, during times of full moon and new moon [15].
Associations between lunar cycles and health conditions, however, such as increased phone call rates by females to a crisis-call centre, higher frequency in misbehaviors in institutionalized patients, greater behavioral deterioration in patients with schizophrenia, increased occurrence in gout attacks, and higher frequencies in the number of appointment requests in thyroid outpatients; rates of gastrointestinal bleeding; multiparae delivery rates; and numbers of births, have been reported [16-23].
Several beliefs, theories and hypotheses regarding lunar impact on the human body have been generated throughout the history of human kind. Assumptions such as the "Gravitational pull hypothesis" or the "Tidal force hypothesis" were extensively analyzed but their impact on the human organism could not be empirically substantiated [24]. A series of studies have rejected the hypothesis of a lunar influence on human health in view of the following: suicide rates [25,26]; violent behavior and aggression [27,28]; agitation in nursing home residents [29]; use of psychiatric community services [30]; psychiatric hospital admissions [31]; frequency of admissions toemergency care [32]; volume of patients admitted to emergency departments [33]; cardiopulmonary arrests in emergency departments [34]; incidence of myocardial infarction and sudden cardiac death [35]; onset of spontaneous pneumothorax [36]; survival time for breast cancer patients [37]; number of surgical complications [38]; postoperative nausea and vomiting [39]; workload on labor and delivery wards [40]; and number of deliveries [41].
There is evidence stating that professionals believe there are correlations between falls and times of the full moon, although an association between patient falls during hospitalization and lunar cycles, especially the influence of the full moon, has not yet been scientifically explored. We hypothesized that no relationship exists between falls in hospitalized patients and lunar cycles. The aim of this study was therefore to examine in-hospital patient fall rates and their associations with days of the week, months, seasons and lunar cycles.
Methods
Study sample and setting
We conducted a retrospective analysis of all registered in-patient falls amongst the adult patients hospitalized on the general internal medicine, surgery and geriatric rehabilitation wards of a 300-bed public hospital, which provides medical services for the inhabitants of the Northern part of the city of Zurich, Switzerland. The observation period was from January 1, 1999 to December 31, 2003. Ethical approval was granted by the Ethics Committee of the City hospitals of the City of Zurich.
Variables and measurements
Patient falls were defined as "an incident in which a patient suddenly and involuntary came to rest upon the ground or surface" and were registered regularly by the nurses discovering the fall. We retrieved the number of registered patient falls occurring during hospital stay from the incident report data system of the quality management department, and screened administrative patient data to determine daily number of hospitalized patients, individual length of patient stay, and daily bed occupancy rates. We identified the dates of the synodic lunar months within the study period, based on the European Southern Observatory Munich Image Data Analysis System (ESO-MIDAS). One synodic lunar month counts 29.53 days (29 d. 12 h. 44 m.) which is the period of time required for the moon to travel from one position relative to the sun as seen from the Earth (e.g. full moon) and return to the same position. The day counts started with the new moon at day 0 until the full moon between day 14 and 15 and ended before the next new moon on day 28 or day 29.
Data analysis
We calculated fall rates per 1000 patient days to adjust for number of falls per day and number of hospitalized patients per day. To examine the pattern of fall rates over time, we calculated mean (including standard deviations (SD), and 95% confidence intervals (CI)) fall rates per day of the week, month and season throughout the study period. To model the rate of falls per 1000 patient days with lunar days, days of the week, and months as predictor variables, we used a general linear model. Statistical tests and confidence intervals were calculated two sided, and p-values <0.05 were considered statistically significant. All analyses were performed using SPSS (12.0).
Results
The 5 year study period included 1,826 observation days. During this time a total of 34,970 patients were hospitalized (mean age: 67.3 (SD 19.3) years, female: 53.6%), accounting for 431,149 patient days. Mean length of stay was 12.3 (SD 14.4) days. The hospital bed occupancy rate was 86.2% (Median: 86.6%). Overall, a total of 3,843 falls were registered, affecting 2,512 (7.2%) patients.
Number of hospitalized patients
The number of hospitalized patients per day ranged over the entire study period from 182 to 279 with a mean of 236 patients (SD 17, median 237). The mean number of hospitalized patients per day of the week varied significantly from 221 (SD 14) on Sundays to 244 (SD 16) on Thursdays (p < 0.001). The mean number of hospitalized patients per month varied significantly from 220 (SD 17) per day in August to 247 (SD 16) per day in February (p < 0.001).
Incidence of patient falls over time
Throughout the study period, the frequency of daily falls ranged from zero to eight falls. The overall mean fall rate was 8.9 (SD 6.4) falls per 1000 patient days. Per day of the week, the mean fall rate ranged from 8.3 (SD 6.9) falls per 1000 patient days on Mondays to 9.3 (SD 6.7) falls per 1000 patient days on Saturdays (df 6; F = .778; p = .587). Per month, the mean fall rate ranged from a low of 8.4 (SD 6.1) falls per 1000 patient days in December to a high of 9.7 (SD 6.8) falls per 1000 patient days in November (df 11; F = .682; p = .757) (Fig. 1). The mean fall rate per 1000 patient days per season of the year varied although not significantly: The lowest rate was in Autumn, with 8.7 (SD 6.2) falls/1000 patient days; In Winter there were 9.0 (SD 6.2) falls; the highest rate of falls was in Spring with 9.1 (SD 6.8), and in Summer there were 9.0 (SD 6.2) (df = 3: F = 0.213; p = 0.887).
Figure 1 Mean fall rates per month (1999–2003).
Falls, lunar cycle, and variation in time
Sixty two complete synodic lunar cycles were observed during the study period. The first full moon was observed on January 2, 1999 (first new moon: January 17, 1999) and the last full moon was seen on December 8, 2003 (last new moon: December 23, 2003). Within the days of the lunar cycle, the variation in mean fall rates per 1000 patient days was not significant. The lowest rate was 7.2 (SD 6.0) falls on lunar day 17, and the highest rate was 10.6 (SD 6.3) falls on lunar day 20 (df 29; F = .929; p = 0.575) (Fig. 2).
Figure 2 Mean fall rates per lunar day (1999–2003).
The fall rates per 1000 patient days, lunar days, and variation in time including days of the week, and months of the year, showed neither a statistically significant main effect, nor a statistically significant interaction between the variables under study (Table 1).
Table 1 Associations between falls/1000 patient days, lunar cycle, days of the week & month
df F-value P-value
Corrected modela) 1503 0.989 0.560
Lunar day 29 0.973 0.509
Days of the week 6 0.545 0.773
Month 11 0.368 0.967
Day of the week & month 66 1.040 0.403
Lunar day & days of the week 174 1.077 0.283
Lunar day & month 318 1.046 0.345
Lunar day, days of the week & month 899 0.949 0.721
a) R2 = 0.822 (adjusted R2 = -.010)
Discussion
Throughout the 5 year study period, no significant association was found in the incidence rate of hospital in-patient falls occurring during the time period of the full or new moon, neither was periodicity demonstrated for days of the week, months or seasons of the year. Despite significant fluctuations of the hospital's patient occupancy per day of the week and month, the patient fall rates remained relatively stable during the entire study period.
Our results contrast with the one other study that addressed the relationship between patient falls and lunar cycles [15]. Sutton et al reported significant findings in view of increased accident rates during the seven days prior to a full moon and the seven days prior to the new moon. In contrast, we examined whether there were associations between fall rates per day during the lunar cycle, throughout 62 lunar cycles.
In general, our findings are concordant with all other studies that, as with our study, did not show an association between lunar days and patient related events such as hospital admissions, emergency department visits, accessing psychiatric services, and violent behavior [28,30-33].
We assume that the belief of some health care professionals that frequency of in-hospital fall accidents increases with the time of the full moon rely on non-specific, non systematic observations within the realm of everyday practice. Such beliefs are probably influenced by lay press reports that highlight bizarre unusual activities when the moon is full [42]. Empirical evidence shows that the etiology of falls during hospitalization is multifactorial. Clinically identifiable risk factors such as impaired mobility, impaired mental status, special toileting needs, psychotropic medications, and a past history of falling have been consistently found to be relevant for predicting future falls [8,10,43]. Of note is that it has recently been shown that hospital system related factors such as nurse staffing and nurse skill mix also influence the frequency of patient falls [44-46].
The challenge for healthcare professionals will be to support patient safety and quality of care by early identification of patients at risk for falling, and implement interventions to prevent falls and related injuries.
Conclusion
The in-patient fall rates were neither associated with days of the week, months, or seasons, nor with lunar cycles such as the full moon or new moon. Preventive strategies should be focused on assessment of patients' modifiable fall risk factors, and the provision of organizational conditions which support a safe hospital environment.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RS contributed to the conception, design, data collection, analysis, interpretation of data, and drafted the manuscript. FJ contributed to the data collection and analysis. SDG contributed to the design, interpretation of data, and critical revision of the manuscript. KM contributed to the analysis, interpretation of data, and manuscript preparation. All authors gave final approval for this version of the manuscript to be published.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thankfully recognize the work of the staff nurses from the clinical departments in filling in the incident fall reports. We also thank the executive management of the Stadtspital Waid in Zurich, namely Hugo Bühler, MD, and Lukas Furler, RN, for their support conducting this study, and we are grateful to Richard Klaghofer, PhD, for his statistical advice.
==== Refs
Goodwin MB Westbrook JI An analysis of patient accidents in hospital Aust Clin Rev 1993 13 141 149 8250776
Halfon P Eggli Y Van_Melle G Vagnair A Risk of falls for hospitalized patients: a predictive model based on routinely available data J Clin Epidemiol 2001 54 1258 1266 11750195 10.1016/S0895-4356(01)00406-1
Hitcho EB Krauss MJ Birge S Claiborne Dunagan W Fischer I Johnson S Nast PA Costantinou E Fraser VJ Characteristics and circumstances of falls in a hospital setting: a prospective study J Gen Intern Med 2004 19 732 739 15209586 10.1111/j.1525-1497.2004.30387.x
Schwendimann R Frequency and circumstances of falls in acute care hospitals: a pilot study Pflege 1998 11 335 341 10427278
Tutuarima JA van der Meulen JH de Haan RJ van Straten A Limburg M Risk factors for falls of hospitalized stroke patients Stroke 1997 28 297 301 9040678
Vassallo M Azeem T Pirwani MF Sharma JC Allen SC An epidemiological study of falls on integrated general medical wards Int J Clin Pract 2000 54 654 657 11221278
von_Renteln_Kruse WKT Sturzereignisse stationärer geriatrischer Patienten Z Gerontol Geriatr 2004 37 9 14 14991290 10.1007/s00391-004-0204-7
Evans D Hodgkinson B Lambert L Wood J Falls risk factors in the hospital setting: a systematic review Int J Nurs Pract 2001 7 38 45 11811346 10.1046/j.1440-172x.2001.00269.x
Moreland J Richardson J Chan DH O_Neill J Bellissimo A Grum RM Shanks L Evidence-based guidelines for the secondary prevention of falls in older adults Gerontology 2003 49 93 116 12574670 10.1159/000067948
Oliver D Daly F Martin FC McMurdo ME Risk factors and risk assessment tools for falls in hospital in-patients: a systematic review Age Ageing 2004 33 122 130 14960426 10.1093/ageing/afh017
Grenier_Sennelier C Lombard I Jeny_Loeper C Maillet_Gouret MC Minvielle E Designing adverse event prevention programs using quality management methods: the case of falls in hospital Int J Qual Health 2002 14 419 426 10.1093/intqhc/14.5.419
Kerzman H Chetrit A Brin L Toren O Characteristics of falls in hospitalized patients J Adv Nurs 2004 47 223 229 15196196 10.1111/j.1365-2648.2004.03080.x
Vassallo M Sharma JC Briggs RS Allen SC Characteristics of early fallers on elderly patient rehabilitation wards Age Ageing 2003 32 338 342 12720623 10.1093/ageing/32.3.338
Vance DE Belief in lunar effects on human behavior Psychol Rep 1995 76 32 34 7770587
Sutton J Standen P Wallace A Incidence and documentation of patient accidents in hospital Nurs Times 1994 90 29 35 8084784
Barr W Lunacy revisited. The influence of the moon on mental health and quality of life J Psychosoc Nurs Ment Health Serv 2000 38 28 35 10820695
Ghiandoni G Secli R Rocchi MB Ugolini G Does lunar position influence the time of delivery? A statistical analysis Eur J Obstet Gynecol Reprod Biol 1998 77 47 50 9550200 10.1016/S0301-2115(97)00226-1
Guillon P Guillon D Pierre F Soutoul JH Seasonal, weekly and lunar cycles of birth. Statistical study of 12,035,680 births Rev Fr Gynecol Obstet 1988 83 703 708 3206096
Hicks-Caskey WE Potter DR Effect of the full moon on a sample of developmentally delayed, institutionalized women Percept Mot Skills 1991 72 1375 1380 1817481
Kollerstrom N Steffert B Sex difference in response to stress by lunar month: a pilot study of four years' crisis-call frequency BMC Psychiatry 2003 3 20 14664724 10.1186/1471-244X-3-20
Mikulecky M Rovensky J Gout attacks and lunar cycle Med Hypotheses 2000 55 24 25 11021320 10.1054/mehy.1999.0973
Roman EM Soriano G Fuentes M Galvez ML Fernandez C The influence of the full moon on the number of admissions related to gastrointestinal bleeding Int J Nurs Pract 2004 10 292 296 15544585 10.1111/j.1440-172x.2004.00492.x
Zettinig G Crevenna R Pirich C Dudczak R Waldhoer T Appointments at a thyroid outpatient clinic and the lunar cycle Wien Klin Wochenschr 2003 115 298 301 12793030
Culver RR J Kelly IW Geophysical variables and behavior: XLIX. Moon mechanisms and myths: A critical appraisal of explanations of purported lunar effects on human behavior Psychol Rep 1988 62 638 710 3406280
Gutierrez-Garcia JM Tusell F Suicides and the lunar cycle Psychol Rep 1997 80 243 250 9122334
Martin SJ Kelly IW Saklofske DH Suicide and lunar cycles: a critical review over 28 years Psychol Rep 1992 71 787 795 1454925
Nunez S Perez Mendez L Aguirre-Jaime A Moon cycles and violent behaviours: myth or fact? Eur J Emerg Med 2002 9 127 130 12131634
Owen C Tarantello C Jones M Tennant C Lunar cycles and violent behaviour Aust N Z J Psychiatry 1998 32 496 499 9711362 10.1046/j.1440-1614.1998.00442.x
Cohen-Mansfield J Marx MS Werner P Full moon: does it influence agitated nursing home residents? J Clin Psychol 1989 45 611 614 2768501
Amaddeo F Bisoffi G Micciolo R Piccinelli M Tansella M Frequency of contact with community-based psychiatric services and the lunar cycle: a 10-year case-register study Soc Psychiatry Psychiatr Epidemiol 1997 32 323 326 9299925 10.1007/BF00805436
Gorvin JJ Roberts MS Lunar phases and psychiatric hospital admissions Psychol Rep 1994 75 1435 1440 7886164
Wolbank S Prause G Smolle-Juettner F Smolle J Heidinger D Quehenberger F Spernbauer P The influence of lunar phenomena on the incidence of emergency cases Resuscitation 2003 58 97 102 12867315 10.1016/S0300-9572(03)00107-2
Thompson DA Adams SL The full moon and ED patient volumes: unearthing a myth Am J Emerg Med 1996 14 161 164 8924138 10.1016/S0735-6757(96)90124-2
Alves DW Allegra JR Cochrane DG Cable G Effect of lunar cycle on temporal variation in cardiopulmonary arrest in seven emergency departments during 11 years Eur J Emerg Med 2003 10 225 228 12972900 10.1097/00063110-200309000-00013
Eisenburger P Schreiber W Vergeiner G Sterz F Holzer M Herkner H Havel C Laggner AN Lunar phases are not related to the occurrence of acute myocardial infarction and sudden cardiac death Resuscitation 2003 56 187 189 12589993 10.1016/S0300-9572(02)00298-8
Sok M Mikulecky M Erzen J Onset of spontaneous pneumothorax and the synodic lunar cycle Med Hypotheses 2001 57 638 641 11735326 10.1054/mehy.2001.1432
Peters-Engl C Frank W Kerschbaum F Denison U Medl M Sevelda P Lunar phases and survival of breast cancer patients--a statistical analysis of 3,757 cases Breast Cancer Res Treat 2001 70 131 135 11768603 10.1023/A:1012907308443
Holzheimer RG Nitz C Gresser U Lunar phase does not influence surgical quality Eur J Med Res 2003 8 414 418 14555297
Eberhart LH Jakobi G Winterhalter M Georgieff M Impact of environmental factors on the incidence of posteropative nausea and vomiting. Influence of the weather and cycle of the moon Anasthesiol Intensivmed Notfallmed Schmerzther 2000 35 635 640 11116495 10.1055/s-2000-7364
Joshi R Bharadwaj A Gallousis S Matthews R Labor ward workload waxes and wanes with the lunar cycle, myth or reality? Prim Care Update Ob Gyns 1998 5 184 10838345 10.1016/S1068-607X(98)00100-0
Waldhoer T Haidinger G Vutuc C The lunar cycle and the number of deliveries in Austria between 1970 and 1999 Gynecol Obstet Invest 2002 53 88 89 11961380 10.1159/000052999
Kelly IR J Culver R Nickell JK, B Genoni J The moon was full and nothing happened: a review of studies of the moon and human behavior and human belief The Outer Edge 1996 New York: CSIOP 17 34
Chang JT Morton SC Rubenstein LZ Mojica WA Maglione M Suttorp MJ Roth EA Shekelle PG Interventions for the prevention of falls in older adults: systematic review and meta-analysis of randomised clinical trials BMJ 2004 328 680 15031239 10.1136/bmj.328.7441.680
Cho SH Ketefian S Barkauskas VH Smith DG The effects of nurse staffing on adverse events, morbidity, mortality, and medical costs Nursing Research 2003 52 71 79 12657982 10.1097/00006199-200303000-00003
Mark BA Burleson DL Measurement of patient outcomes: data availability and consistency across hospitals J Nurs Adm 1995 25 52 59 7714631
Sovie MD Jawad AF Hospital restructuring and its impact on outcomes: nursing staff regulations are premature J Nurs Adm 2001 31 588 600 11771464
|
16225704
|
PMC1274330
|
CC BY
|
2021-01-04 16:30:31
|
no
|
BMC Nurs. 2005 Oct 17; 4:5
|
utf-8
|
BMC Nurs
| 2,005 |
10.1186/1472-6955-4-5
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-1091622566510.1186/1471-2458-5-109Research ArticleKhat and alcohol use and risky sex behaviour among in-school and out-of-school youth in Ethiopia Kebede Derege [email protected] Atalay [email protected] Getnet [email protected] Fikre [email protected] Frehiwot [email protected] Yigeremu [email protected] Reta [email protected] Wuleta [email protected] Tamrat [email protected] Tewodros [email protected] Department of Community Health, Addis Ababa University, Addis Ababa, Ethiopia2 Department of Psychiatry, Addis Ababa University, P.O.Box 9086 Addis Ababa, Ethiopia3 Ethiopian Public Health Association, Addis Ababa, Ethiopia4 Ministry of Defence, Addis Ababa, Ethiopia5 Family Health International, Addis Ababa, Ethiopia6 Private consultant, Addis Ababa, Ethiopia2005 14 10 2005 5 109 109 23 6 2005 14 10 2005 Copyright © 2005 Kebede et al; licensee BioMed Central Ltd.2005Kebede et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Khat (an evergreen plant with amphetamine-like properties) and alcohol are widely consumed among the youth of Ethiopia. However, their relationship to risky sexual behaviour is not well described. This study was conducted to describe the magnitude of risky sexual behaviour (unprotected sex and early initiation of sexual activity) and its association with Khat and alcohol consumption in Ethiopian youths.
Methods
A probabilistic national sample of 20,434 in-school and out-of-school youths aged between 15 and 24 years of age was selected and interviewed regarding their sexual behavior and substance use.
Results
Over 20% of out-of-school youth had unprotected sex during the 12-month period prior to interview compared to 1.4% of in-school youth. Daily Khat intake was also associated with unprotected sex: adjusted OR (95% CI) = 2.26 (1.92, 2.67). There was a significant and linear association between alcohol intake and unprotected sex, with those using alcohol daily having a three fold increased odds compared to those not using it: adj. OR (95% CI) = 3.05 (2.38, 3.91). Use of substances other than Khat was not associated with unprotected sex, but was associated with initiation of sexual activity: adj. OR (95% CI) = 2.54 (1.84, 3.51).
Conclusion
A substantial proportion of out-of-school youth engage in risky sex. The use of Khat and alcohol and other substances is significantly and independently associated with risky sexual behaviour among Ethiopian youths.
==== Body
Background
Over 6% of Ethiopia's adult population is believed to be HIV positive. It is estimated that a large proportion of new infections occur in people aged 25 years or younger. As is the case elsewhere in Africa, transmission is almost exclusively through heterosexual contact [1], although parenteral transmission through injections account for some infections. Postponement of sex, adoption of safe sex practices, including protected sex, and prevention of parenteral transmission are the cornerstones of national and international HIV/AIDS control strategies[2]. Surveys have indicated a generally high level of awareness of HIV/AIDS[3], and increased condom use in the Ethiopian population[4]. Substance abuse is generally believed to be one of the associated factors for sexual risk behaviour in HIV transmission. Hard drugs like heroin and cocaine are very rarely available in Ethiopia. However, khat, a locally produced psycho-stimulant is commonly and widely used in the country.
Khat (Catha edulis) is an evergreen plant that grows mainly in Ethiopia, Yemen and other African countries along the coast of the Indian Ocean. It has been used for centuries as a mild stimulant. The fresh leaves are chewed or consumed as tea. For most youths chewing Khat is a method of increasing energy and elevating mood in order to improve work performance[5]. The psycho-stimulant effect of Khat is due to the alkaloid ingredient cathinone, which has a similar chemical structure to amphetamine[6]. Several case reports and population studies have shown that there is a clear association between heavy consumption of khat and psychosis [7-13]. Khat is widely consumed among the youth of Ethiopia as shown by several prevalence studies[14]. Although the deleterious physical effects of Khat has been shown by some studies[15], its role in altering sexual behaviour has not been studied or reported. While some attribute their sexual impotence to Khat use, others report increased libido[16]. A study assessing the magnitude of Chlamydia trachomatis and Neisseria gonorrhoeae infections together with self-reports of sexual risk behavior among youths (15–24 years old) in Addis Ababa, Ethiopia, reported that increased sexual activity was significantly associated with being male, aged 20 years or over, out-of-school status, and reported alcohol/khat consumption[17].
The importance of alcohol and substance misuse as causes of significant mortality and morbidity (particularly from injuries) and social harm (such as social disruptions from crime, underemployment and marital disharmony) have been well recognized[18]. Although the effect of alcohol and other psychoactive substances in interfering with condom use has also been studied to some extent in developed societies[19], this vital area of research has not been explored in Ethiopia. Physical injury and high-risk sexual behaviour under the influence of alcohol are common in teenagers. Alcohol-related physical injury appears closely related to patterns of alcohol consumption whereas alcohol-related sexual risk-taking is most closely associated with symptoms of depression and anxiety [20].
Young people are particularly vulnerable and are the key to the future course of the HIV/AIDS epidemic. As a group, they are an essential focus for prevention and control programs. Since most new infections are in young people, modest changes in behaviour will have a significant impact on the epidemic. Thus, they have been given higher priority in the prevention and control of HIV/AIDS in Ethiopia[21]. As both Khat and alcohol are widely consumed in these groups, description of the relationship between these substances and risky sexual behaviour would usefully guide national policy and decision making on HIV/AIDS.
The present study was undertaken to describe the magnitude of risky sexual behaviour and its association with Khat consumption and alcohol use among a national sample of in-school and out-of-school youth in Ethiopia.
Methods
Data collection for the study was conducted between December 2001 and May 2002.
Sampling procedures
The following criteria were used in selection of the study subjects: (1) In-school youth: aged 15–19 years, unmarried, daytime high school students attending grades 9–12 or vocational training schools. (2) Out-of-school youth: aged 15–24 years, not attending day or night school, unmarried, unemployed or employed informally. The sampling frames for selection of study subjects were prepared in consultation with the Ministry of Education, regional education bureaus and respective schools (to obtain details of classes and number of students in each grade). Probability proportional to size sampling (PPS) was used to select classes in the first stage and then systematic sampling was applied to select students in the second stage. A list of classes from each selected school with their corresponding measures of size was prepared. They were listed using the numbering system of the school so that they can be identified easily. Starting at the top of the list, we calculated the cumulative measure of size (per sex) and entered these figures in a column next to the measure of size for each class. Using an average cluster size, and equal sample size for males and females, we calculated the sampling interval (SI) by dividing the total cumulative measure of size by the number of classes to be selected.
We then selected a random starting number (RS) between 1 and SI and compared this number with the cumulated measure of size column. The unit within whose cumulated measure of size for the RS falls is the first sampling unit and subsequent classes were selected by adding the SI to the RS.
Once inside the classroom, starting from the front right hand seat of the class, we picked a random number and systematically selected the required number of students. Then the selected students were requested to meet the trained interviewers at the school compound where privacy is maintained; when this was not convenient for the respondents the interviewer made an appointment to meet the student, at any suitable time and place.
Out-of-school youth (OSY) were selected from urban centres in each of the ten regions of Ethiopia and Addis Ababa and the Dire Dawa cities. The sampling frames for selection of OSY were prepared using the 1994 census report[22]. Segmentation method was used for larger regions and cities that had 20 or more Kebeles (sub-districts), and it is the preferred sampling frame for household surveys. This approach had several advantages; notably, the household had already been mapped and numbered within enumeration areas and moreover, the areas had population sizes associated with them that could be used as measures of size during sample selection, making control of the fieldwork easier. Using records from each Kebele office that contain number and size of households, we created segments for each Kebele. For instance if a Kebele has 1000 households and a sample of 100 households are needed for the survey, dividing 1000 by 100 = 10 segments were created for this particular Kebele. All segments in each Kebele had the same number of households. In randomly selected segments targeted number of households was visited, until the required number of youth was identified for interview in each segment. Each interviewer had a short checklist to determine eligibility of respondents. A face-to-face household interview was conducted to obtain the needed information. If the identified respondent was not available on the day of visit to a household appointments were made to return for the interview. A further detailed description of the sampling method is given elsewhere[23]
Data collection and processing
Data collection was done using a standardized pre-coded and pre-tested questionnaire. Male and female interviewers were selected from each of the participating regions. Interviewers had completed high school and had some previous experience of collecting survey data. They were given a one week intensive training about the interview processes and on how to administer the questionnaire. Pilot-testing was carried out in Addis Ababa where they took the training before the interviewers were sent back to their respective sites.
Full-time editors scrutinised all completed interview forms for completeness, accuracy and consistency in the field. SPSS was used for both univariate and bivariate analysis.
History of substance use (alcohol, Khat, and others) in the last four weeks preceding the interview and history of unprotected sexual practice and initiation of sex were obtained using the interview instrument. Unprotected sex was defined as sex without the use of condom during the 12-month period preceding the interview. Irregular use of condom was also categorized as unprotected sex. Initiation of sexual activity was analysed for those younger than 20 years and was defined as a report of any sex encounter in the past. Logistic regression was employed to adjust for confounding. Thus, either unprotected sex or sex initiation was included in the logistic model as a dependent variable. As independent variables, the following were included in the model: sex, age, school status (in or out of school), educational attainment, khat, alcohol and other substance use.
Ethical clearance for the study was obtained from a national ethics committee and from Human Subjects Committee of Family Health International – USA. Participation of respondents was strictly on voluntary basis. Informed consent was solicited orally. Measures were taken to ensure the respect, dignity and freedom of each individual participating in the study. Measures were also taken to assure confidentiality.
Results
A total of 20,434 youth aged between 15 and 24 years were included in the study. Of these, 14,224 were out-of-school youth. The response rate was 91.2% and 94.1% for out of school and in-school youth, respectively. Approximately equal number of males and females were included in the study. (Table 1). Over 80% had attained higher education, and 62% were orthodox Christians. Among in-school youth, over 90% did not use Khat or alcohol, while among the out-of-school youth close to 73% did not use Khat or alcohol. Over 23% of out-of-school youth used Khat every day or once weekly while only 7.5% of in-school youth did so. Only 0.7% of the in-school youth reported use of substances other than Khat, compared to 5.1% for out-of-school youth.
Table 1 Socio-demographic characteristics of the study population of in-school and out-of-school youth, Ethiopia 2003
Characteristics Number (percent)*
In-school Out-of-school Total
Sex
Male 3,089 (49.7) 7,109 (50.0) 10,198 (49.9)
Female 3,121 (50.3) 7,115 (50.0) 10.236 (50.1)
Age
15–19 years 6,206 (100.0) 7,267 (51.1) 13,473 (66.6)
20–24 years - 6,955 (48.9) 6,955 (34.0)
Educational level
Not literate - - -
Elementary - 3,149 (29.1) 3,149 (18.5)
Secondary or above 6,210 (100.0) 7,687 (70.9) 13,897 (81.5)
Religion
Orthodox Christian 4,025 (65.1) 8,648 (61.1) 12,673 (62.3)
Catholic 61 (1.0) 375 (2.6) 436 (2.1)
Protestant 351 (5.7) 1,310 (9.2) 1,661 (8.2)
Muslim 1,691 (27.3) 3,709 (26.2) 5,400 (26.5)
Other 57 (0.9) 122 (0.9) 179 (0.9)
Alcohol intake
None or occasional 5,650 (91.1) 9,992 (73.0) 15,642 (78.7)
On a weekly basis 526 (8.5) 3,307 (24.2) 3,833 (19.3)
On a daily basis 26 (0.4) 384 (2.8) 410 (2.1)
Khat intake
None 5,611 (91.5) 10,500 (73.9) 16,111 (79.2)
Less than one per week 67 (1.1) 364 (2.6) 431 (2.1)
Once a week 353 (5.8) 1,885 (13.3) 2,238 (11.0)
Every day 103 (1.7) 1,459 (10.3) 1,562 (7.7)
Substance use other than Khat**
No 6,164 (99.3) 13,496 (94.9) 19,660 (96.2)
Yes 46 (0.7) 728 (5.1) 774 (3.8)
Total 6,210 (100.0) 14,224 (100.0) 20,434 (100.0)
* Missing values not shown.
** Substance use other than Khat include: Shisha, Benzene, Hashish, Mandrax, Cocaine, or Crack.
A total of 16,606 (82%) youth reported on their sexual behaviour and were included in the analysis (Table 2). Over 20% of out-of-school youth had unprotected sex during the 12-month period prior to the interview compared to 1.4% in-school youth. The odds of unprotected sex were slightly higher among males compared to females. Larger effect sizes were, however associated with younger age: In those aged 15–19 years compared to those age 20–24 years, adjusted OR (95% CI) = 2.54 (2.29, 2.81). Being out-of-school was strongly associated with self-report of unprotected sex: adjusted adj. OR (95% CI) = 8.48 (6.66, 10.8). Daily Khat intake was also associated with unprotected sex: adj. OR (95% CI) = 2.26 (1.92, 2.67). There was a significant and linear association between alcohol intake and unprotected sex (P-value for trend <0.01) with those using alcohol daily having a three fold increased odds compared to those not using it: adj. OR (95% CI) = 3.05 (2.38, 3.91). Use of substances other than Khat was not associated with unprotected sex.
Table 2 Socio-demographic and behavioral correlates of unprotected sex (during the 12 months prior to the interview) among the youth (aged 15–24 yrs), Ethiopia 2003
Characteristics Total population Reported unprotected sex (%) Crude odds ratio (95% confidence interval) Adjusted odds ratio* (95% confidence interval) P-value
Sex
Female 8,234 1,038 (12.6) 1.00 (Reference) 1.00 (Reference)
Male 8,372 1,338 (16.0) 1.32 (1.21, 1.44) 1.13 (1.02, 1.26) 0.02
Age
15–19 yrs 11,738 857 (7.3) 1.00 (Reference) 1.00 (Reference)
20–24 yrs 4,862 1,519 (31.2) 5.77 (5.26, 6.33) 2.54 (2.29, 2.81) <0.001
School status
In-school 5,502 75 (1.4) 1.00 (Reference) 1.00 (Reference)
Out-of-school 11,104 2,301 (20.7) 18.9 (15.0, 23.9) 8.48 (6.66, 10.8) <0.001
Alcohol intake
None 13,004 1,322 (10.2) 1.00 (Reference) 1.00 (Reference)
On a weekly basis 2,838 852 (30.0) 3.79 (3,44, 4.18) 2.02 (1.81, 2.25) <0.001
On a daily basis 318 148 (46.5) 7.69 (6.13, 9.66) 3.05 (2.38, 3.91) <0.001
Khat intake
None 13,438 1,362 (10.1) 1.00 (Reference) 1.00 (Reference)
Occasional 342 114 (33.3) 4.43 (3.52, 5.59) 2.42 (1.86, 3.13) <0.001
On a weekly basis 1,633 481 (29.5) 3.70 (3.28, 4.18) 2.06 (1.79, 2.36) <0.001
On a daily basis 717 414 (36.6) 5.12 (4.48, 5.85) 2.26 (1.92, 2.67) <0.001
Substance use other than Khat
No 16,034 2,147 (13.4) 1.00 (Reference) 1.00 (Reference)
Yes 572 229 (40.0) 4.32 (3.63, 5.14) 1.19 (0.97, 1.47)
Total 16,606 2,376 (14.3)
* Terms included in the logistic model were: sex, age, school status, alcohol intake (3 levels), Khat intake (4 levels), and substance use other than Khat.
** Substance use other than Khat include: Shisha, Benzene, Hashish, Mandrax, Cocaine, or Crack. (Shisha is a mixture that may include tobacco, honey, hashish and spices and is smoked from an oriental tobacco pipe)
Initiation of sexual activity was studied on a subset of the study population who were younger than 20 years of age. Of the total of 13,473 youths in this younger age group, 11,737 (84%) reported initiation of sexual activity and were included in the analysis. Being male or female was not associated with initiation of sexual activity when school status and substance use were adjusted for in the logistic regression model (Table 3). The odds of initiation of sexual activity were 3.6 fold higher among out-of-school youth compared to in-school youth: adj. OR (95% CI) = 3.57 (3.09, 4.12). Khat use was strongly associated with initiation of sexual activity with four-fold increased odds in both daily and weekly users. The odds ratios for daily use were: adj. OR (95% CI) = 4.13 (3.26, 5.23); and for weekly Khat use: 4.18 (3.50, 4.99). Alcohol use was strongly and linearly associated with initiation of sexual activity, those using alcohol having a four-fold increase, and daily users having six-fold increased odds of sex initiation (P for trend < 0.001). Use of substances other than Khat was also strongly associated with sex initiation: adj. OR (95% CI) = 2.54 (1.84, 3.51).
Table 3 Socio-demographic and behavioral correlates of initiating sex among in-school and out-of-school youth aged 15–19 yrs, Ethiopia 2003
Characteristics Total population Reported sexual initiation (%) Crude odds ratio (95% confidence interval) Adjusted odds ratio* (95% confidence interval) P-value
Sex
Female 5,932 620 (10.5) 1.00 (Reference) 1.00 (Reference)
Male 5,806 960 (16.5) 1.70 (1.52, 1.89) 1.12 (0.98, 1.27)
School status
In-school 5,498 286 (5.2) 1.00 (Reference) 1.00 (Reference)
Out-of-school 6,240 1,294 (20.7) 4.76 (4.17,5.44) 3.57 (3.09, 4.12) <0.001
Alcohol intake
None or occasional 9,944 939 (9.4) 1.00 (Reference) 1.00 (Reference)
On a weekly basis 1,407 542 (38.5) 6.01 (5.29, 6.82) 3.81 (3.31, 4.39) <0.001
On a daily basis 115 71 (61.7) 15.5 (10.6, 22.6) 5.75 (3.70, 8.93) <0.001
Khat intake
None 10,209 931 (9.1) 1.00 (Reference) 1.00 (Reference)
Occasional 185 92 (49.7) 9.86 (7.33, 13.3) 4.99 (3.56, 7.00) <0.001
On a weekly basis 840 335 (39.9) 6.61 (5.67, 7.71) 4.18 (3.50, 4.99) <0.001
On a daily basis 445 210 (47.2) 8.91 (7.31, 10.9) 4.13 (3.26, 5.23) <0.001
Substance use other than Khat**
No 11,498 1,423 (12.4) 1.00 (Reference) 1.00 (Reference)
Yes 240 157 (65.4) 13.4 (10.2, 17.6) 2.54 (1.84, 3.51) <0.001
Total 11,738 1,580 (13.5)
* Terms included in the logistic model were: sex, age, school status, alcohol intake (3 levels), Khat intake (4 levels), and substance use other than Khat.
** Substance use other than Khat include: Shisha, Benzene, Hashish, Mandrax, Cocaine, or Crack.
Discussion
Our findings show that Khat and alcohol use, sex, age, and school-status were independently associated with both self-report of unprotected sexual intercourse and initiation of sexual activity. The results are unlikely to be biased. Selection of the study population used a probabilistic sampling method and was not based on any of the factors under study. Thus, selection bias is unlikely. Although information bias due to a differential and systematic under-reporting of sexual behaviour among the various variables is a possibility, it is more likely that this under-reporting is randomly distributed than otherwise. If this is the case, the resulting random misclassification will tend to bias the odds ratio towards the null value [24]. The implication for our study is that the odds ratios reported are actually underestimates of the true effect sizes in the population. We have adjusted for potential confounding by using a multivariate logistic model.
Use of alcohol, Khat and other substances were measured over the four weeks prior to the interview, while that of risky sexual behaviour over the 12 months prior to the interview. It is assumed that the four-week window assessment period is representative of long-term individual pattern of use of these substances. The correctness of this assumption can not be verified from the present study.
In this study, the association between male sex and unprotected sex, although significant, was small when other factors such as age, school status, and substance use behavior were adjusted for in a logistic model. Although hormonal factors might be expected to increase impulsiveness and risk taking behavior in males, the association with male sex may also be mediated through other intermediate behaviors such as alcohol use. In male attendees at sexually transmitted disease clinics in southern Vietnam, being aged under 20, not married, not having a current girlfriend, using alcohol before sex and substance use were all factors independently associated with visiting a female sex worker (FSW) [25]. Similarly in Colombia, a longitudinal survey revealed that adolescents with increased drug use were more likely to engage in unprotected sex as well as multiple partnerships[26].
It is more difficult to speculate on the reasons for the fairly strong two-fold association between increasing age and unprotected sex. This finding goes against previous studies and our expectation. Although we have adjusted for the possible effects of alcohol and khat use, we can not exclude the effect of other potential confounding factors.
We have also shown that being out of school was strongly associated with unprotected sex. A similar finding was reported in a previously conducted study among Addis Ababa youths, where being out of school showed a strong association with sexual risk behaviour[17]. In New South Wales, 16-year old out-of-school adolescents had consistently higher rate of reported substance use compared to their age matches in school [27]. Although underlying behavioural problems or mental disorders could be linked to the reasons for youths being out of school, and although we have previously shown that mental and behavioural disorders are prevalent in adolescents in Ethiopia[28,29], we can not confirm the association from the present study. In Bonomo et al.'s [20] study of Australian 16–17-year-olds, alcohol-related sexual risk-taking, psychiatric morbidity and high frequency of alcohol consumption had strong independent associations. In a ten-year prospective study, others reported that adolescents who have experienced drinking alcohol even once or twice during the past 12 months were more likely to exhibit more substance use, face academic problems and become involved in delinquent behaviors during the latter stages of school (middle and high school) compared to non-drinkers [30]. Early drinking was also found to be associated with early sexual initiation [31].[32]
Khat is primarily used for its stimulant effect. Users report that Khat intake results in increased energy levels and alertness, improves self-esteem, creates a sensation of elation, enhances imaginative ability and the capacity to associate ideas, and improves the ability to communicate. It has not yet been associated with alteration of rational decision-making and has not been shown to increase risk-taking behaviour[5]. Although some users also take alcohol after Khat to counteract the stimulant properties and facilitate sleep, this should not confound the observed association as alcohol use was adjusted for in the logistic model.
Our finding of a linear and strong association between unprotected sex and alcohol is to be expected because of the nature of alcohol in decreasing inhibitions, altering rational decision making, and increasing risk-taking behaviour. A qualitative study among military conscripts in Northern Thailand drew the following conclusions regarding alcohol consumption and inconsistent condom use: Alcohol is (1) consciously used by men to reduce inhibitions that constrain their interpersonal interaction with women and with each other; (2) reduces inhibitions of individuals to sexual risk taking; (3) provides a socially acceptable excuse for non-use of condoms; (4) is associated by conscripts with brothel attendance; and (5) is seen to enhance male sexual pleasure, in contrast to condoms, which are said to reduce pleasure[33]. In a meta-analysis of the association between alcohol intake and condom use, drinking at first intercourse was associated with decreased condom use, but alcohol was not related to condom use in recent sexual encounters and in recent encounters with new partners. On the other hand drinking was related to non condom use among adolescents [19].
Initiation of sexual activity was not associated with being male when other factors were adjusted in the logistic model. Others have reported that males become sexually active at a younger age than females [34]. The reason why we did not detect any difference between males and females in initiation of sex is probably due to the influence of factors other than individual behavior. For example, although in most cultures females are expected not to initiate sex before they are married, they may be forced to have sex as has been shown by recent studies of sexual abuse in Ethiopia [35, 36]. Sex initiation was also associated with being out of school. As described above for unprotected sex, the reason for the association could be underlying behavioural problems that predispose youth to leave school and also engage in risky sexual behaviour. Thus, being out of school is likely to be just a marker of such underlying behavioural and mental problems.
Khat use was also strongly associated with sex initiation. However, Khat use was assessed over the four weeks prior to the interview, while initiation of sex was assessed for a longer period spanning from puberty to 19 years of age. It is thus possible that sex was initiated well before the initiation of Khat use or vice versa. As we do not have the pattern of use of Khat beyond the four-week period prior to the survey, it is difficult to speculate on the significance of this strong association. As is the case with Khat use, the strong and linear association of alcohol use with initiation of sex, and the association of use of substances other than Khat with sex initiation, are difficult to interpret because of the uncertain temporal relationships between the various variables.
Conclusion
This study has shown that a substantial proportion of out-of-school youths engage in risky sexual behaviours and that the use of Khat, alcohol and other substances is significantly and independently associated with risky sexual behaviour among these young people. HIV/AIDS prevention and control programmes targeting youths, should take into account this newly demonstrated association of risky sexual behaviour with use of Khat, given its widespread use among in-school and out-of-school youths, and formulate appropriate interventions to limit its use.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TA, TG and RY were supervisors of data collection (field data quality control) and then significantly involved in the analysis and write up.
YA, involved in proposal writing, design, implementation of project and national dissemination of the findings.
WL, contributed in the designing of the methodology and all stages of the project.
FB, project manager and involved in all stages of the project implementation and write up.
GM, lead investigator and involved in mapping, designing, implementation, analysis of data, write up and dissemination.
DK, project initiator and lead investigator involved in project proposal, design of questionnaires, recruitment and training of supervisors and data collectors. He did most of the analysis and write up of the paper.
AA, investigator, involved in designing, analysis and write up.
FE, designing of the methodology and sampling procedures, coordinated mapping (listing) of study sites and target groups and managed data quality and write up of the methodology.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The study was funded by USAID, with technical assistance from Family Health International. Additional support was also obtained from UNICEF and Save the Children – USA. Additional material support was given by the Department of Community Health of Addis Ababa University, HIV/AIDS Prevention and Control Office, Ministries of Health and Defence, regional HIV/AIDS Prevention and Control Secretaries and regional health bureaus. The following individuals have rendered technical advice during the planning and execution of the study: Drs. Damen Hailemariam, Dagnatchew Hailemariam, Endalamaw Abera, Yetnayet Asfaw, Asegid Woldu, Gail Davey, Ahmed Ali, and Ato Mirgissa Kaba, and Mr. Mohamed Ali Bhuiyan. We also thank the following for their administrative support: Ms. Francesca Stuer, Jeanette Bloem, and Ato Ali Beyene. Charlotte Hanlon for reviewing the earlier manuscript. All study participants are gratefully acknowledged for their time and for sharing their experiences with the study team.
==== Refs
MOH AIDS in Ethiopia. Fourth Edition. 2003 Addis Ababa , Ministry of Health
MOH Strategic Framework for the National Response to HIV/AIDS in Ethiopia for 2000-2004. 1999 Addis Ababa , Ministry of Health
CSA. Ethiopia Demographic and Health Survey-2000. Preliminary report. 2000 Addis Ababa , Central Statistical Authority, Addis Ababa and Marco International Inc., USA.
Kebede D Aklilu M Sanders E The HIV epidemic and the state of its surveillance in Ethiopia. Ethiop Med J 2000 38 283 302 11125503
Kalix P Khat: scientific knowledge and policy issues. Br J Addict 1987 82 47 53 2881570
Kalix P The pharmacology of Khat. Gen Pharmacol 1984 15 179 187 6376274
Alem A Shibre T Khat induced psychosis and its medico-legal implication: a case report. Ethiop Med J 1997 35 137 141 9577014
Yousef G Huq Z Lambert T Khat chewing as a cause of psychosis. Brit J Hosp Med 1995 54 322 326 8556211
Odenwald M Neuner F Schauer M Elbert RT Catani C Lingenfelder B Hinkel H Hafner H Rockstroh B Khat use as a risk factor for psychotic disorder: a cross-sectional and case-control study in Somalia BMC Medicine 2005 3 15707502
Pantelis C Hindler CG Taylor JC Use and abuse of khat (Catha edulis): a review of the distribution, pharmacology, side effects and a description of psychosis attributed to khat chewing Pharmacol Med 1989 19 657 668
Gough SP Cookson IB Khat induced schizophreniform psychosis in UK Lancet 1984 25 455 10.1016/S0140-6736(84)91789-6
Giannini AJ Castelani FS A manic like psychosis due to khat (Catha edulis Forsk.) J Toxicol-Clin Toxicol 1982 19 455 459 7175990
Dhadphale M Mengech HNK Chege SW Miraa (Catha edulis) as a cause of psychosis. East Afr Med J 1981 58 130 135 6113948
Alem A Kebede D Kullgren G The prevalence and socio-demographic correlates of khat chewing in Butajira, Ethiopia. Acta Psychiatrica Scand (Suppl) 1999 100 84 91
Belew M Kebede D Kassaye M Khat use and its associated health and socio-economic effects in a general population. Ethiop Med J 2000 38 11 26 11144876
Al Motareb A Baker K Broadley KJ Khat: Pharmacological and medical aspects and its social use in Yemen, A review article. Phytotherapy Research 2002 16 403 413 12203257 10.1002/ptr.1106
Taffa N Bjune G Sundby J Gaustad P Alestrøm A Prevalence of Gonococcal and Chlamydial Infections and Sexual Risk Behavior Among Youth in Addis Ababa, Ethiopia. Sexually transmitted diseases 2002 29 828 833 12466727 10.1097/00007435-200212000-00015
Ritson B Chick J CJ Epidemiology and primary prevention of alcohol misuse. Seminars in alcohol and drug misuse 1994 Glasgow , Bell and Bain Ltd. 75 93
Leigh BC Alcohol and condom use: A meta analysis of event-level studies Sexually transmitted diseases 2002 29 476 482 12172533 10.1097/00007435-200208000-00008
Bonomo Y Coffey C Wolfe R Lynskey M Bowes G Patton G Adverse outcomes of alcohol use in adolescents. Addiction 2001 96 1485 1496 11571067 10.1046/j.1360-0443.2001.9610148512.x
NAPCC National HIV/AIDS Control Program (in Amharic). 2000 Addis Ababa , National AIDS Prevention and Control Council
OPHCC The 1994 Population and Housing Census of Ethiopia. Abridged Statistical Report. 1996 Addis Ababa , Office of Population and Housing Census Commission
HAPCO/MOH. HIV/AIDS Behavioral Surveillance Survey (BSS), Round 1, 2002 Addis Ababa, Ethiopia
Rothman KJ Modern Epidemiology 1986 Boston , Little, Brown and Company
Thuy NT Lindan CP Phong TH Van dat Nhung VT Barclay J Kiem HB Predictors of visits to commercial sex workers by male attendees at sexually transmitted disease clinics in southern Vietnam. AIDS 1999 13 719 725 10397567 10.1097/00002030-199904160-00013
Brook D. W Brook J. S Pahl T I. M The Longitudinal Relationship between drug use and risky behaviours among Colombian adolescents. Arch Pediatr Adolesc Med 2002 156 1101 11107 12413337
Tresidder J Macaskill P Bennett D Nutbeam D Aust N Z J Public Health 1997 21 168 174 9161073
Ashenafi Y Kebede D Desta M Alem A Prevalence of mental and behavioural disorders in Ethiopian children. East Afr Med J 2001 78 308 311 12002109
Kebede M Kebede D Desta M Evaluation of the Amharic version of the Diagnostic Interview for Children and Adolescents (DICA) in Addis Ababa. Ethiop J Health Dev 2000 14 13 21
Ellickson P L. Tucker J S. Klein DJ Ten -Year Prospective Study of Public Health Problems Associated With Early Drinking. J Pediatrics 2003 111 949 955 10.1542/peds.111.5.949
MacQueen KM Nopkesorn T Sweat MD Sawaengdee Y Mastro TD Weniger BG Alcohol consumption, brothel attendance, and condom use: normative expectations among Thai military conscripts. Med Anthropol Q 1996 10 402 423 8873026 10.1525/maq.1996.10.3.02a00070
Burack R Teenage sexual behaviour: attitudes towards and declared sexual activity. Br J Fam Plann 1999 24 145 148 10023100
Mulugeta E Kassaye M Berhane Y Prevalence and outcomes of sexual violence among high school students. Ethiop Med J 1998 36 167 174 10214457
Gossaye Y Deyessa N Berhane Y Butajira Rural Health Program: Women's Health and Life Events Study in Rural Ethiopia. Ethiop J Health Dev 2003 17 1 47
|
16225665
|
PMC1274331
|
CC BY
|
2021-01-04 16:28:55
|
no
|
BMC Public Health. 2005 Oct 14; 5:109
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-109
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-1111622974010.1186/1471-2458-5-111Research ArticleRisk factors for adverse perinatal outcomes in imprisoned pregnant women: a systematic review Knight Marian [email protected] Emma [email protected] National Perinatal Epidemiology Unit, University of Oxford, Oxford, UK2 Department of Public Health, University of Oxford, Oxford, UK2005 17 10 2005 5 111 111 22 3 2005 17 10 2005 Copyright © 2005 Knight and Plugge; licensee BioMed Central Ltd.2005Knight and Plugge; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Imprisoned pregnant women constitute an important obstetric group about whom relatively little is known. This systematic review was conducted to identify the risk factors associated with adverse pregnancy outcome present in this group of women.
Methods
The review was conducted according to a prespecified protocol. Studies of any design were included if they described information on any of the pre-specified risk factors. We calculated the results as summary percentages or odds ratios where data was available on both cases and population controls.
Results
The search strategy identified 27 relevant papers of which 13 met the inclusion criteria, involving 1504 imprisoned pregnant women and 4571 population control women. Imprisoned women are more likely to be single, from an ethnic minority, and not to have completed high school. They are more likely to have a medical problem which could affect the pregnancy outcome and yet less likely to receive adequate antenatal care. They are also more likely to smoke, drink alcohol to excess and take illegal drugs.
Conclusion
Imprisoned women are clearly a high risk obstetric group. These findings have important implications for the provision of care to this important group of women.
==== Body
Background
Although women make up only a small proportion of the 9 million people imprisoned worldwide [1], their numbers are increasing rapidly and consistently across a number of countries [2,3]. For example, the number of women imprisoned in England and Wales has risen almost threefold over the past decade [3]. Most of these women will be of childbearing age and an estimated 6% of imprisoned women are pregnant [3,4]. This implies that in England and Wales alone there are about 240 pregnant women in prison at any one time, and in the United States of America over 6,000.
These women constitute an important obstetric group about whom relatively little is known. Available evidence suggests that they are more likely to come from socially deprived backgrounds and to smoke, drink alcohol to excess and abuse illegal drugs than the general population [5-10]. However, estimates of the prevalence of these risk behaviors in this population vary. These factors may affect both the health of the women themselves and also their offspring and are therefore of considerable public health significance. It is important to recognize these factors in order to allow appropriate planning of future services for this increasing number of women. The objective of this study therefore was to identify the risk factors associated with adverse pregnancy outcome present in imprisoned women through a systematic review of the literature.
Methods
The review was conducted according to a pre-specified protocol. We identified possible risk factors for poor perinatal outcomes from guidelines produced by the American College of Obstetricians and Gynecologists (ACOG) [11] and the National Institute for Clinical Excellence (NICE) [12], and literature review. We used multiple strategies to identify relevant articles, searching for any studies published from 1980 up to the end of May 2004. We searched Medline, Embase, CINAHL, Psycinfo, and the Cochrane library database. We identified search terms from database thesauri (indicated by italics) and terms were also included as free text. We used a combination of terms relating to pregnancy (e.g. pregnan*, pregnancy, pregnancy-outcome, pregnant-women) and to imprisonment (e.g. prison*, gaol*, jail*, incarcerat*, prisons, prisoners) combining them using Boolean operators. We also carried out hand searches of the references of selected papers and relevant policy documents. We identified grey literature and unpublished research by searching the National Research Register, the NHS Centre for Reviews and Dissemination database and the internet (accessed May 2004 using Google search engine).
We included studies if they described any of the pre-specified characteristics of pregnant women (Table 1) who were imprisoned at any stage during their pregnancy in any category of prison. No restrictions were placed on the study design that could be included nor on the basis of the language of publication or the geographical location of the study. Non-English articles were translated. We excluded studies which did not include information on the pre-specified risk factors. An assessment of methodological quality was made according to the principles recommended for assessing non-experimental studies in the Cochrane Reviewers Handbook [13]. Potential for selection, performance, attrition and detection bias was assessed. Studies were graded for quality as A, B or C indicating a low, moderate or high risk of bias respectively.
Table 1 Risk factors for poor perinatal outcomes pre-specified in the study protocol
Risk factors
Age over 40
Age less than 18
Primiparity
Parity >4
Previous preterm delivery or midtrimester loss
Previous stillbirth or neonatal death
Previous infant with a congenital anomaly
Smoking
Alcohol use in pregnancy
Illicit drug use
Maternal medical problem known to be associated with poor perinatal outcome: epilepsy
diabetes
autoimmune disorder
HIV
hypertension
cardiac disease
renal disease
Previous low birth weight infant
Non-white race
Educational level (Not completed high school/ A levels)
Inadequate antenatal care (first antenatal visit in second trimester or beyond, or fewer than six visits in total)
Single marital status
Body Mass Index (BMI) >35
BMI <18
Previous Cesarean section
Two investigators independently extracted the data according to a fixed protocol. Differences were resolved by discussion. As well as the pre-specified risk factors, we also collected data on study design, case selection, control selection, nature of the control group and location of the study. We expressed the frequencies of risk factors for poor perinatal outcomes among imprisoned women as percentages. Where data was available on both cases and population controls who were not matched for a particular risk factor, odds ratios were calculated using a fixed effects model (Mantel-Haenszel). The X2 test for heterogeneity was used to assess the extent to which the results of the studies were in agreement.
Results
We identified 27 relevant papers. Of these, 13 met the inclusion criteria (Table 2), the majority of which were conducted in the USA. All were written in English with the exception of one German paper which was translated. The excluded papers were either discussion documents or did not contain any data on the pre-specified risk factors. The included studies comprise a total of 1504 imprisoned pregnant women and 4571 population control women.
Table 2 Included studies
Reference Study Design Setting Participants (imprisoned pregnant women and population control women) Quality
Stauber 1984 [14] Cohort Berlin, Germany 43 pregnant women imprisoned between 1973 and 1982. 172 women from the same hospital matched with cases by age, parity, marital status and year of birth. B
Elton 1985 [15] Cohort Manchester, UK 298 pregnant women admitted to one prison 1975–1982. 298 non-imprisoned women selected from the same hospital antenatal clinic matched with cases by age, marital status, previous stillbirths and height. A
Shelton 1989 [16] Case series Missouri, Maryland, USA 26 imprisoned women who delivered in 1982. B
Cordero 1991 [9] Case Series Ohio, USA 53 pregnant women imprisoned for between 1 week and 90 days. 53 matched pregnant women imprisoned for greater than 120 days. 1986–1990. A
Cordero 1992 [8] Case Series Ohio, USA 233 pregnant women imprisoned in the state medium-security prison 1986–1990. B
Egley 1992 [7] Cohort North Carolina, USA 69 imprisoned pregnant women cared for at one hospital during 1988. 69 non-imprisoned pregnant women from the same hospital matched with cases by age, race, parity and date of entry into prenatal care. A
Fogel 1993 [6] Case Series North Carolina, USA 89 pregnant women imprisoned between 1986 and 1989. A
Terk 1993 [10] Cohort Texas, USA 76 imprisoned pregnant women 1987–1990. 117 unmatched randomly-chosen non-imprisoned pregnant women from the same hospital during the same time period. A
Martin 1997 [17] Cohort North Carolina, USA 168 imprisoned pregnant women who gave birth to one infant between 1988 and 1991 identified from state records. 3910 unmatched randomly selected women resident in and delivering in North Carolina over the same time period. B
Kyei-Aboagye 2000 [5] Cohort Massachusetts, USA 31 imprisoned pregnant women delivering at one hospital between 1993 and 1996. 71 unmatched randomly chosen non-imprisoned women delivering at the same hospital. B
Mertens 2001 [18] Cohort Illinois, USA 71 pregnant women imprisoned in a county jail in one calendar year. 51 pregnant women identified from state records and matched with cases by age, race, gravidity and zip code of residence. C
Siefert 2001 [19] Case Series Michigan, USA 120 pregnant women imprisoned before commencement of a residential program (1987–1991), 44 unmatched pregnant women imprisoned after the residential program (1991–1995). A
Barkauskas 2002 [20] Case Series Michigan, USA 90 imprisoned pregnant women in a residential care program and 40 unmatched imprisoned pregnant women not in the residential programme.1996–1998. C
The risk factor profile of imprisoned pregnant women is summarized in Table 3. We did not identify any studies which described the proportion of imprisoned pregnant women who were aged less than 18, who had a parity greater than four, who had had a previous infant with a congenital anomaly, who had a body mass index (BMI) greater than 35 or less than 18, or who had had a previous caesarean section. Five of the risk factors were identified to be present in more than 50% of imprisoned pregnant women: single marital status (83.7%), non-white race (66.7%), smoking (66%), low educational attainment (54.4%) and illicit drug use (53.7%).
Table 3 Risk factor profile of imprisoned pregnant women
Risk factor Number of women studied Number of studies Percentage of imprisoned pregnant women with identified risk factor
Single marital status 997 7 83.7
Non-white race 1042 10 66.7
Smoking 838 8 66.0
Educational level (not completed high school/ A levels) 529 6 54.4
Drug use 646 7 53.7
Inadequate prenatal care 704 6 30.5
Primiparity 944 8 26.7
Previous low birth weight infant 88 1 25.0
Alcohol use 363 4 19.8
Previous preterm delivery or mid-trimester loss 428 3 15.9
Maternal medical problem 100 2 11.0
Age over 40 233 1 2.6
Previous stillbirth or neonatal death 298 1 1.3
Eight outcomes were reported in studies which included an appropriate population control group of non-imprisoned pregnant women, enabling summary odds ratios to be calculated. These are shown in Figure 1.
Figure 1 Summary odds ratios for the predefined risk factors for poor perinatal outcomes of imprisoned women compared to controls. Figures greater than 1 indicate risk factors occurring more frequently in imprisoned pregnant women than population control pregnant women.
Five studies reported smoking status in imprisoned women and controls [5,7,10,14,17] (386 imprisoned, 4339 control women). Imprisoned women were significantly more likely to smoke during pregnancy than control women (OR 6.05 (95% CI 4.74–7.73). Imprisoned women were also significantly more likely to use alcohol during pregnancy (3 studies [5,10,17], 274 imprisoned, 4098 control women, OR 4.82 (3.23–7.19)) and to use illicit drugs (4 studies [5,7,10,14], 218 imprisoned, 429 control women, OR 25.86 (14.06–47.57)). It should be noted, however, that there was significant heterogeneity between the studies reporting alcohol use (X2 = 11.56, p = 0.003). This heterogeneity was entirely due to one small study [5] (weight in meta-analysis 2.78%, OR 54.63 (11.24–265.4)). It was not clear from this paper how alcohol use in control women was assessed, which may have resulted in performance bias in this study. The remaining included studies reported smaller, but still significant, increased use of alcohol in imprisoned pregnant women compared with the non-imprisoned group.
Only two studies reported maternal medical problems in imprisoned women and controls [5,7] (100 imprisoned, 140 control women). Imprisoned pregnant women were significantly more likely to have medical problems likely to impact on pregnancy outcome than control women (OR 5.64 (1.66–19.11). We identified only two studies which reported race in unmatched imprisoned women and controls [5,10] (244 imprisoned, 4027 control women); these studies showed that imprisoned women were significantly more likely to be of non-white race (OR 3.17 (2.39–4.19) than control women.
The remaining three outcomes were reported in imprisoned and control women in only one study [17] (168 imprisoned, 3910 control women). This study showed that imprisoned pregnant women were more likely not to have completed high school than control women (OR 3.30 (2.42–4.51)), were more likely to be of single marital status (OR 12.32 (8.21–18.50)), and were more likely to have received inadequate prenatal care (OR 5.15 (3.60–7.38).
Discussion
This is the first review to describe in detail the risk factor profile of imprisoned pregnant women. The results show that these women are clearly a group at high risk of poor perinatal outcomes. They are more likely to be single, from an ethnic minority (predominantly African-American and Hispanic), and not to have completed high school. They are more likely to have a medical problem which could affect the pregnancy outcome and yet less likely to receive adequate antenatal care. They are also more likely to smoke, drink alcohol to excess and take illegal drugs. These findings have implications for the provision of care to this important group of women.
Overall, more than 30% of imprisoned pregnant women received inadequate prenatal care, classified by the date of their first prenatal visit or by the total number of visits. Clearly, this inadequacy of care may relate to the period of pregnancy before these women were imprisoned, and it is therefore important to undertake work to ensure that prenatal care is available and accessible to the socially disadvantaged populations from which these women come. However, it also remains a priority to ensure that provision for prenatal care is adequate in all prison settings so that these women and their unborn infants are not further compromised by poor care during imprisonment.
The results of this review also indicate areas of prenatal care for imprisoned pregnant women on which there needs to be a specific focus. 66% of imprisoned pregnant women smoke, nearly 20% use alcohol to excess and over 50% use illegal drugs. This shows a clear need to provide programs particularly for pregnant drug users to assist not only with cessation of use during pregnancy as appropriate, but also with long-term drug-free maintenance once women have completed their pregnancy and period of imprisonment.
This review also provided some limited evidence that imprisoned pregnant women are more likely to suffer from medical problems known to be associated with poor perinatal outcomes, such as diabetes, epilepsy, hypertension, cardiac or renal disease. Prison prenatal services need therefore to be sufficiently flexible to allow for more specialist care for women with particular problems.
We were not able to extract information on a number of factors known to increase the risk of poor perinatal outcomes, for example over- and underweight, grand-multiparity and previous caesarean section. Although several studies included some information on weight, there were no consistent definitions used and no studies reported body mass index (BMI). Obesity is known to be associated with a range of adverse perinatal outcomes [21], and population prevalence is known to be increasing [22,23]. This is therefore an important risk factor to identify in imprisoned pregnant women. Previous cesarean section is also known to be associated with adverse perinatal outcomes [24], and identification of this factor in imprisoned pregnant women would give a more comprehensive picture of the pattern of risk in this group.
The results highlight the importance of pregnant prisoners as a high risk population. This has important policy implications. Clearly the level of provision of services for pregnant women in prison should not be based on that provided for the general population of pregnant women – need is much greater in the prisoners. Furthermore, there is some evidence to suggest that imprisonment might actually have a beneficial effect on particular pregnancy outcomes; limited research on the pregnancy outcomes of imprisoned women has shown that prison actually improves fetal outcomes and the longer spent in prison, the better the outcome [15,17]. The authors give a number of possible reasons for this; prison provides food, shelter and protection from abusive partners; it ensures access to antenatal care and moderates the use of alcohol and drugs. This suggests that appropriate health service provision for pregnant women in prison can be effective and should be prioritised in the health planning process. Clearly more research is needed in this area.
Undoubtedly there are problems with the generalisability of these findings. Most of the studies come from the USA although a few come from the UK and one from Germany. The demographics of female prison populations vary considerably across the world but tend to reflect the composition of the poorest, most disadvantaged sections of society from which they come. Thus whilst the prevalence of risk factors for adverse perinatal outcomes may show some variation from country to country, it is unlikely that in any country the prevalence is lower in the prison population. It is likely that pregnant prisoners have great health needs wherever they are from. Health planners and providers in all countries should not ignore this important population. Furthermore although some studies identified were conducted more than twenty years ago, we would argue that rates of smoking etc. have declined more slowly in disadvantaged groups and therefore are unlikely to have changed substantially from the figures reported.
Although this review encompasses information on more than 1500 pregnant women and 4500 ontrols, the number of studies identified were relatively few and we did not exclude any on the basis of quality. Of the thirteen studies identified, only 6 were assessed to be of high quality (low risk of bias), 5 were of moderate quality (moderate risk of bias), and 2 of low quality (high risk of bias). Because of the small number of studies reporting similar risk factors, we were unable to perform a sensitivity analysis to investigate the effect of study quality on the reported results. Thus the potential for bias needs to be taken into account when considering the overall results of the review.
The small study numbers have also impacted on the precision of some of the results. The confidence intervals for the odds ratios associated with drug use, a maternal medical problem and single marital status are wide. However, all are significantly greater than 1 indicating that these risk factors are present to a greater extent in imprisoned pregnant women and must be taken into account when planning services.
Conclusion
More research could be undertaken to describe more comprehensively the risk factors for poor perinatal outcomes in this group of women. However, it is clear from this review that the factors already identified are likely to impact significantly on both the health of imprisoned pregnant women and their infants. Imprisoned pregnant women are a socially disadvantaged group and prison presents an opportunity to engage with them effectively and meet their substantial needs. It is necessary to ensure the provision of adequate, tailored prenatal services for these women in order to prevent future maternal and perinatal morbidity and mortality.
List of abbreviations
ACOG American College of Obstetricians and Gynecologists
BMI Body Mass Index
NICE National Institute for Clinical Excellence
OR Odds Ratio
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MK and EP drafted the report and were involved in study design, data extraction, statistical analysis and interpretation of the results. Both authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
MK is employed by the Oxford Deanery Public Health Training Programme. EP is employed by the University of Oxford.
==== Refs
Walmsley R Prison Population List (fifth edition) 2003 London: Home office
Office of Justice Programs Prisoners in 2002 2003 Washington: US Department of Justice
Home Office Statistics on women in the criminal justice system 2001 London: Home Office
Office of Justice Programs Women in prison Bureau of Justice Statistics Special Report 1994 Washington: US Department of Justice
Kyei Aboagye K Vragovic O Chong D Birth outcome in incarcerated, high-risk pregnant women J Reprod Med 2000 45 190 4 10756495
Fogel CI Pregnant inmates: risk factors and pregnancy outcomes J Obstet Gynecol Neonatal Nurs 1993 22 33 9 8429412
Egley CC Miller DE Granados JL Ingram Fogel C Outcome of pregnancy during imprisonment J Reprod Med 1992 37 131 4 1538355
Cordero L Hines S Shibley KA Landon MB Perinatal outcome for women in prison J Perinatol 1992 12 205 9 1432273
Cordero L Hines S Shibley KA Landon MB Duration of incarceration and perinatal outcome Obstet Gynecol 1991 78 641 5 1923168
Terk JV Martens MG Williamson MA Pregnancy outcomes of incarcerated women J Matern Fetal Med 1993 2 246 50
American Academy of Pediatrics AcoOaG Guidelines for Perinatal Care 2002 Chicago: American Academy of Pediatrics
National Collaborating Centre for Women's and Children's Heath Antenatal care:routine care for the healthy pregnant woman 2003 London: Royal College of Obstetricians and Gynaecologists
Alderson P Green S Higgins JPT editors Cochrane Reviewers Handbook 4.2.2 [updated March 2004] 2003 Accessed May 2004
Stauber M Weingart B Koubenec J Schwangerschaft, Geburt und Wochenbett bei inhaftierten Frauen. [Pregnancy, labor and the puerperium in women prisoners] Geburtshilfe-Frauenheilkd 1984 44 731 7 6569008
Elton PJ Outcome of pregnancy among prisoners J Obstet Gynaecol 1985 5 241 4
Shelton BJ Gill DG Childbearing in prison: a behavioral analysis J Obstet Gynecol Neonatal Nurs 1989 18 301 8 2746379
Martin SL Kim H Kupper LL Meyer RE Hays M Is incarceration during pregnancy associated with infant birthweight? Am J Public Health 1997 87 1526 31 9314809
Mertens DJ Pregnancy outcomes of inmates in a large county jail setting Public-Health-Nurs 2001 18 45 53 11251873 10.1046/j.1525-1446.2001.00045.x
Siefert K Pimlott S Improving pregnancy outcome during imprisonment: a model residential care program Soc Work 2001 46 125 34 11329642
Barkauskas VH Low LK Pimlott S Health outcomes of incarcerated pregnant women and their infants in a community-based program J Midwifery Womens Health 2002 47 371 9 12361349 10.1016/S1526-9523(02)00279-9
Galtier-Dereure F Boegner C Bringer J Obesity and pregnancy: complications and cost Am J Clin Nutr 2000 71 1242S 1248S 10799397
Ogden CL Flegal KM Carroll MD Johnson CL Prevalence and trends in overweight among US children and adolescents, 1999–2000 JAMA 2002 288 1728 32 12365956 10.1001/jama.288.14.1728
Flegal KM Carroll MD Ogden CL Johnson CL Prevalence and trends in obesity among US adults, 1999–2000 JAMA 2002 288 1723 7 12365955 10.1001/jama.288.14.1723
Smith GC Pell JP Dobbie R Caesarean section and risk of unexplained stillbirth in subsequent pregnancy Lancet 2003 362 1779 84 14654315 10.1016/S0140-6736(03)14896-9
|
16229740
|
PMC1274332
|
CC BY
|
2021-01-04 16:28:55
|
no
|
BMC Public Health. 2005 Oct 17; 5:111
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-111
|
oa_comm
|
==== Front
BMC PharmacolBMC Pharmacology1471-2210BioMed Central London 1471-2210-5-141621613210.1186/1471-2210-5-14Research ArticleUrodynamic effects of oxybutynin and tolterodine in conscious and anesthetized rats under different cystometrographic conditions Angelico Patrizia [email protected] Cristina [email protected] Luciano [email protected] Giorgio [email protected] Amedeo [email protected] Rodolfo [email protected] Pharmaceutical R & D Division – Recordati S.p.A. – Via Civitali 1, 20148 Milano, Italy2005 11 10 2005 5 14 14 15 12 2004 11 10 2005 Copyright © 2005 Angelico et al; licensee BioMed Central Ltd.2005Angelico et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Antimuscarinic agents are the most popular treatment for overactive bladder and their efficacy in man is well documented, producing decreased urinary frequency and an increase in bladder capacity. During cystometry in rats, however, the main effect reported after acute treatment with antimuscarinics is a decrease in peak micturition pressure together with little or no effect on bladder capacity. In the present experiments we studied the effects, in rats, of the two most widely used antimuscarinic drugs, namely oxybutynin and tolterodine, utilising several different cystometrographic conditions. The aim was to determine the experimental conditions required to reproduce the clinical pharmacological effects of antimuscarinic agents, as seen in humans, in particular their ability to increase bladder capacity.
Results
Intravenous or oral administration of tolterodine or oxybutynin in conscious rats utilized 1 day after catheter implantation and with saline infusion at constant rate of 0.1 ml/min, gave a dose-dependent decrease of micturition pressure (MP) with no significant change in bladder volume capacity (BVC). When the saline infusion rate into the bladder was decreased to 0.025 ml/min, the effect of oral oxybutynin was similar to that obtained with the higher infusion rate. Also, experiments were performed in rats in which bladders were infused with suramin (3 and 10 μM) in order to block the non-adrenergic, non-cholinergic component of bladder contraction. Under these conditions, oral administration of oxybutynin significantly reduced MP (as observed previously), but again BVC was not significantly changed.
In conscious rats with bladders infused with diluted acetic acid, both tolterodine and oxybutynin administered at the same doses as in animals infused with saline, reduced MP, although the reduction appeared less marked, with no effect on BVC.
In conscious rats utilized 5 days after catheter implantation, a situation where inflammation due to surgery is reduced, the effect of tolterodine (i.v.) and oxybutynin (p.o.) on MP was smaller and similar, respectively, to that observed in rats utilized 1 day after catheter implantation, but the increase of BVC was not statistically significant.
In anesthetized rats, i.v. administration of oxybutynin again induced a significant decrease in MP, although it was of questionable relevance. Both BVC and threshold pressure were not significantly reduced. The number and amplitude of high frequency oscillations in MP were unmodified by treatment.
Finally, in conscious obstructed rats, intravenous oxybutynin did not modify frequency and amplitude of non-voiding contractions or bladder capacity and micturition volume.
Conclusion
Despite the different experimental conditions used, the only effect on cystometrographic parameters of oxybutynin and tolterodine in anesthetized and conscious rats was a decrease in MP, whereas BVC was hardly and non-significantly affected. Therefore, it is difficult to reproduce in rats the cystometrographic increase in BVC as observed in humans after chronic administration of antimuscarinic agents, whereas the acute effects seem more similar.
==== Body
Background
Overactive bladder is a chronic clinical syndrome characterized by urgency and frequency with or without urinary incontinence affecting millions of people worldwide [1,2]. Overactive bladder arises from uncontrolled contraction of the detrusor muscle during bladder filling [3]. Although what element(s) trigger(s) unstable contraction is not resolved; myogenic [4] or neural [5] theories have been suggested in the attempt to clarify the etiology of this bladder dysfunction.
Nevertheless, since contraction of the detrusor muscle and bladder emptying are primarily mediated by stimulation of muscarinic receptors by acetylcholine, anticholinergic agents are currently recommended as a first-line therapy for overactive bladder. Of the available antimuscarinic agents, oxybutynin and tolterodine are the most widely used to treat this condition [6,7].
As demonstrated by several investigations in patients, oxybutynin decreases urinary frequency, urgency and episodes of urge incontinence, in addition to increasing bladder volume at first desire to void, enhancing maximum bladder capacity and reducing maximum detrusor pressure during filling [6,8]. Similar results have been reported after administration of tolterodine in patients affected by overactive bladder [[6], and references therein].
Despite the favourable results observed in the clinical studies reported above with regard to the cystometrographic modifications induced by treatment with oxybutynin and tolterodine, the only effect generally observed in rats is a decrease in the maximum detrusor pressure at micturition [9-18].
Although purinergic mechanisms appear not to be involved to any extent in the normal function of human bladder [19], they are involved in human pathological conditions [19], and they are well documented for contraction of rat urinary bladder [20].
The aim of the present experiments was, therefore, to study the effect of oxybutynin and tolterodine in rats, using different cystometrographic conditions, attempting to find suitable experimental conditions able to reproduce the effects observed in humans, in particular the increase of bladder capacity.
A preliminary account of this work was presented in an abstract [21].
Results
Tolterodine and oxybutynin were evaluated as the most widely utilized antimuscarinics.
In general, 2–3 scaled doses of each compound were administered. In some pilot experiments each dose of compound was tested with a matched control group treated with vehicle. In other experiments, control rats and animals treated with different doses of test compound were evaluated simultaneously.
Effect of intravenous and oral administration in conscious rats utilized 1 day after catheter implantation
When i.v. injected in conscious rats, oxybutynin (0.3 mg/kg) induced a prompt and marked reduction of MP showing a peak 15 min after injection. BVC, in contrast, was not modified in comparison with the values observed in control animals. Similar results were obtained after administration of the same dose of tolterodine (Fig. 1). After different doses of both antimuscarinics administration, a dose-dependent reduction of MP was observed, with no significant changes of BVC, tolterodine being substantially more potent than oxybutynin (Fig. 2). After oral administration of oxybutynin (3 mg/kg) and tolterodine (10 mg/kg), the time-course of BVC and MP values in control and treated animals was similar to that observed after i.v. administration (Fig. 3). The compounds induced a marked and statistically significant fall of MP reaching a peak 1 hr after administration with no significant effects on BVC. Dose-response curves obtained after administration of the antimuscarinics are shown in Fig. 4. Oxybutynin and tolterodine dose-dependently decreased MP.
Figure 1 Time-course of the effect of i.v. administration of oxybutynin and tolterodine on BVC and MP in conscious rats. Data represent the mean (± S.E.) of BVC (ml) and MP (mmHg) values recorded before treatment (0 min) and at different times after administration of: upper panel – vehicle (0.1 ml/kg; circles; n = 6) or oxybutynin (0.3 mg/kg; squares; n = 8); lower panel – vehicle (0.1 ml/kg; circles; n = 7) or tolterodine (0.3 mg/kg; squares n = 8). Rats were utilized 1 day after catheter implantation. Statistical analysis indicates significativity vs vehicle-treated group at considered times.
Figure 2 Effect of i.v. administration of tolterodine and oxybutynin on BVC and MP in conscious rats. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section. Tolterodine was administered at 0.03 (n = 8), 0.1 (n = 9) and 0.3 (n = 8) mg/kg; oxybutynin at 0.1 (n = 5) and 0.3 (n = 8) mg/kg. Corresponding open bars represent changes recorded in the vehicle groups (n = 7). Rats were utilized 1 day after catheter implantation. Statistical significance was evaluated by ANOVA (and Dunnett's test).
Figure 3 Time-course of the effect of oral administration of oxybutynin and tolterodine on BVC and MP in conscious rats. Data represent the mean (± S.E.) of BVC (ml) and MP (mmHg) values recorded before treatment (0 min) and at different times after administration of: upper panel – vehicle (2 ml/kg; circles; n = 7) or oxybutynin (3 mg/kg; squares; n = 7); lower panel – vehicle (2 ml/kg; circles; n = 7) or tolterodine (10 mg/kg; squares n = 8). Rats were utilized 1 day after catheter implantation. Statistical analysis indicates significativity vs vehicle-treated group at considered times.
Figure 4 Effect of oral administration of tolterodine and oxybutynin on BVC and MP in conscious rats. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section. Tolterodine was administered at 1 (n = 6), 3 (n = 7) and 10 (n = 8) mg/kg; oxybutynin at 1 (n = 7) and 3 (n = 7) mg/kg. Corresponding open bars represent changes recorded in the vehicle groups (n = 7). Rats were utilized 1 day after catheter implantation. Statistical significance was evaluated by ANOVA (and Dunnett's test).
The effects of oral administration of oxybutynin (3 mg/kg) were also evaluated in freely-moving rats using discontinuous cystometry, increasing observation period up to 10 hr after treatment. Discontinuous cystometry allows parameters to be recorded for extended periods of time, avoiding the detrimental effects on the levels of basal physiological urodynamic parameters seen in restrained rats following continuous bladder infusion for a long time (for example decrease in peak micturition pressure). Oxybutynin induced a decrease of MP similar to that observed previously during the first 2–4 hr after treatment (see Fig. 5 and the corresponding Fig. 3). The difference between treated and control group resulted statistically significant in spite of the figure course because it should be considered that it was evaluated on Δ values, as reported in the Method section. BVC values resulted significantly (p < 0.05) different in comparison with vehicle treated animals at 4 hr after administration. This result, in contrast to that previously obtained during classical cystometry, however, can be related to the behaviour of the control group, whose BVC values were extremely constant during the period.
Figure 5 Time-course of the effect of oral administration of oxybutynin on BVC and MP in conscious freely-moving rats under discontinuous cystometry. Data represent the mean (± S.E.) of BVC (ml) and MP (mmHg) values recorded before treatment (0 min) and at different times after administration of vehicle (2 ml/kg; circles; n = 11) or oxybutynin (3 mg/kg; squares; n = 11).
To further study the effect of oxybutynin on BVC, the less inhibitory dose of oxybutynin on MP (oral dose of 1 mg/kg) was chosen and the saline infusion rate into the bladder was decreased from 0.1 ml/min to 0.05 and 0.025 ml/min. However, the effect of oxybutynin on BVC and MP was found to be similar to that obtained with the higher flow rate (Fig 6).
Figure 6 Effect of oral administration of oxybutynin on BVC and MP in conscious rats under cystometry at different infusion rates. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section. Oxybutynin was administered at 1 mg/kg (n = 9). Corresponding open bars represent changes recorded in the vehicle groups (n = 8). Infusion rates utilized were 0.025 and 0.05 ml/min. Rats were utilized 1 day after catheter implantation. Statistical significance was evaluated by Student's t test.
In order to block the non-adrenergic, non-cholinergic component of the contraction, experiments were performed in rats with bladders infused with suramin (3 and 10 μM). Suramin (10 μM) in vehicle-treated rats induced a significant increase (in comparison with basal values) in BVC (p < 0.01). Under these conditions, oral administration of 3 mg/kg of oxybutynin reduced significantly MP (as previously observed), but BVC was again not significantly increased (Fig 7). In normal unanesthetized rats, Igawa et al. [20] found that co-administration of atropine in rats with purinergic receptors desensitised by administration of α,β-methylene ATP induced urinary retention. In our experimental conditions, we did not observed overflow incontinence in rats with bladder infused with suramin and treated with oxybutynin, probably owing to the fact that suramin is not a specific antagonist [19] and, therefore, not all the P2X receptors were inhibited.
Figure 7 Effect of oral administration of oxybutynin on BVC and MP in conscious rats with bladder infused with different suramin concentrations. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section. Oxybutynin was administered at 3 mg/kg (n = 8). Corresponding open bars represent changes recorded in the vehicle groups (n = 8). Suramin concentrations were 3 × 10-6 and 1 × 10-5 M. Rats were utilized 1 day after catheter implantation. Statistical significance was evaluated by Student's t test.
Effect of intravenous and oral administration in conscious rats treated 5 days after catheter implantation
Catheterization of the bladder induces inflammation that reaches a peak between 1–3 days after surgery. At 5 days post catheterization, inflammation is considerably lower than at 1 day after catheter implantation, as confirmed by the higher values of BVC recorded (about 1.3 – 1.7 ml, data not shown) than those at 1 day after surgery (see basal values in Fig 1 and 3). Under these conditions, i.v. administration of tolterodine dose-dependently decreased MP without affecting BVC (Fig. 8). The effect of tolterodine on MP was less than that observed in rats 1 day after surgery. The effect of oral administration of oxybutynin (1 mg/kg) on BVC and MP was not substantially different from that observed in rats utilized 1 day after catheter implantation.
Figure 8 Effect of tolterodine and oxybutynin on BVC and MP in conscious rats utilized 5 days after catheter implantation. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section. Tolterodine was i.v. administered at 0.1 (n = 7) and 0.3 (n = 7) mg/kg (upper panel). Oxybutynin was administered at 1 mg/kg (n = 7; lower panel). Corresponding open bars represent changes recorded in the vehicle groups (n = 7). Statistical significance was evaluated by ANOVA (and Dunnett's test).
Effect of intravenous administration in anesthetized rats
Cystometrographic and electromyographic recordings performed during micturition in rats under anesthesia show that the striated muscle of external urethral sphincter (EUS) exhibits high-frequency bursting that induces corresponding high-frequency oscillations (HFO) in the pressure recorded intravescically. HFO are correlated with the bursting pattern in the EUS, and represent alternate contractions and relaxations of the urethral outlet, functioning like a pump to enhance urine flow. Consequently, pharmacological manipulations that may result in impaired EUS function might result in micturition disturbances, e.g. bladder-urethra dyssynergia.
In anesthetized rats, the i.v. administration of 0.3 mg/kg of oxybutynin (a dose inducing in conscious rats a significant decrease of MP) induced again a significant, although not so relevant decrease of MP, probably owing to the lower basal value of this parameter due to the anesthesia. Neither BVC nor threshold pressure showed a significant reduction. The number and amplitude of HFO were not modified by oxybutynin.
Effect of intravenous administration in conscious rats with bladder infused with diluted acetic acid
Bladders of conscious rats catheterized from 1 day were infused for a control period of 60 min with saline. Then the infusion was switched to 0.2% acetic acid. At the end of the first hour of bladder infusion with the irritant, BVC was markedly and significantly reduced (about 40 – 60%; p < 0.01) in all groups of animals, indicating bladder hyperactivity due to pain and an increase in afferent nerve firing. Micturition pressure was less affected and generally increased (Fig. 9).
Figure 9 Time-course of the effect of i.v. administration of oxybutynin and tolterodine on BVC and MP in conscious rats with bladder infused with diluted acetic acid. Data represent the mean (± S.E.) of BVC (ml) and MP (mmHg) values recorded before acetic acid infusion (-60 min), at the end of the first hr of acid infusion (0 min) and at different times after administration of: upper panel – vehicle (0.1 ml/kg; circles; n = 8) or oxybutynin (1 mg/kg; squares; n = 8); lower panel – vehicle (0.1 ml/kg; circles; n = 7) or tolterodine (0.3 mg/kg; squares n = 8). Rats were utilized 1 day after catheter implantation and during continuous infusion of acid. Statistical analysis indicates significativity vs vehicle-treated group at considered times.
In the groups of animals treated with vehicle, the continuous infusion of the bladder with acetic acid during the second hour of experiment induced a further decrease of bladder capacity (generally 10–30%), although this further reduction was not statistically significant. MP did not change during the period. Oxybutynin (1 mg/kg i.v.) and tolterodine (0.3 mg/kg i.v.) had no effect on BVC, but caused a rapid and marked decrease of MP similar to that observed in rats during saline infusion of the bladder (Fig. 9).
After the administration of the same i.v. doses as utilized in normal, saline-infused rats, both tolterodine and oxybutynin reduced MP, although the reduction appeared less marked (Fig. 10). The antimuscarinics generally showed only a trend to reverse the decrease in BVC observed during the second hr of acid infusion. However, the intermediate dose of tolterodine produced a significant difference (p < 0.05) between control and treated rats with regard to BVC.
Figure 10 Effect of i.v. administration of tolterodine and oxybutynin on BVC and MP in conscious rats with bladder infused with diluted acetic acid. Data represent the mean (± S.E.) AUC change of BVC and MP calculated as reported in the Method Section (basal values were those recorded at the end of the first hr of acid infusion). Tolterodine was administered at 0.03 (n = 7), 0.1 (n = 6) and 0.3 (n = 8) mg/kg; oxybutynin at 0.1 (n = 8) and 0.3 (n = 8) mg/kg. Corresponding open bars represent changes recorded in the vehicle groups (n = 7). Rats were utilized 1 day after catheter implantation. Statistical significance was evaluated by ANOVA (and Dunnett's test) or Student's t test.
Effect of intravenous administration in conscious obstructed rats
In rats, bladder hypertrophy secondary to bladder outlet obstruction induces bladder instability characterized by the presence of non-voiding contractions (NVC) during filling. Repeated cystometry in control obstructed rats injected with saline gave reproducible results in terms of frequency and amplitude of spontaneous contractile activity (NVC). BVC and MV in these animals were also not significantly changed. Oxybutynin (0.3 mg/kg i.v.) gave results similar to control animals (Table 2).
Table 2 Effect of intravenous administration of oxybutynin on cystometrographic parameters in conscious obstructed rats
BVC (ml) MV (ml) NVC: frequency (N°/2 min) NVC: amplitude (mmHg)
Controls
Before 1.27 ± 0.16 1.36 ± 0.24 5.3 ± 0.9 5.2 ± 0.9
After 1.14 ± 0.25 1.32 ± 0.30 4.9 ± 0.9 4.9 ± 0.9
Oxybutynin 0.3 mg/kg
Before 1.98 ± 0.12 2.02 ± 0.46 5.9 ± 0.4 11.7 ± 1.8
After 2.58 ± 0.70 2.77 ± 0.50 5.5 ± 0.6 9.7 ± 2.1
Data are expressed as mean ± s.e. mean. BVC: bladder volume capacity; MV: micturition volume; NVC: non-voiding contractions. * = p < 0.05; ** = p < 0.01 before vs after administration (Student's t test).
Discussion
A summary of the different models and conditions utilized to evaluate the effects of tolterodine and/or oxybutynin is shown in Table 3.
Table 3 Summary of results obtained after treatment with tolterodine and/or oxybutynin in the different models/conditions utilized.
Models/conditions Effect on BVC Effect on MP
Conscious rats – 1 day after surgery – i.v. and p.o. administration = ↓↓
Conscious (freely-moving/discontinuos cystometry) rats – 1 day after surgery – p.o. administration = ↓
Conscious (cystometry at different rate of filling) rats – 1 day after surgery – p.o. administration = ↓
Conscious (bladder infused with suramin) rats – 1 day after surgery – p.o. administration = ↓
Conscious (bladder infused with acetic acid) rats – 1 day after surgery – p.o. administration = (↑) ↓↓
Conscious (obstructed) rats – 2 days after surgery – i.v. administration = n.e.
Conscious rats – 5 days after surgery – i.v. and p.o. administration = ↓
Anesthetized rats – at the day of surgery – i.v. administration = ↓
= not significant change; ↓: decrease; ↓↓: dose-dependent decrease; n.e. = not evaluated (since the ligature around the urethra was not removed before cystometry, peak micturition pressure was not considered in these experiments); = (↑): some non dose-dependent effect was seen after tolterodine but not after oxybutynin.
Cystometrographic evaluation performed in conscious normal rats utilized one day after catheter implantation showed that neither oxybutynin nor tolterodine increased BVC after oral or i.v. administration. In agreement with previously reported data [9,13,15,16], however, both drugs induced a strong dose-dependent decrease of MP. Furthermore, this behaviour was maintained when a discontinuous cystometry was performed or when different filling rates were utilized.
Available evidence indicates that arachidonic acid metabolites produced along the cyclooxygenase pathway are involved in the physiological regulation of micturition during reflex activation of the urinary bladder. Furthermore, endogenous prostaglandins are produced locally following distension of the bladder wall and modulate the afferent branch of reflex micturition by lowering the threshold for eliciting voiding contractions, serving as a link between detrusor muscle stretch produced by bladder filling and activation of capsaicin-sensitive afferents [22-25]. Different authors [26,27] have shown that cystometrograms in conscious rats with bladders infused with saline and recorded during the first 1–3 days after catheter implantation showed bladder overactivity with relatively low urine volume and a high frequency of micturition. Cystitis is also present at this time and is characterized by edema in the submucosa and an increased tissue content of prostaglandins that stimulate capsaicin-sensitive sensory fibers in the afferent branch of the micturition reflex [27]. In the present study, cystometrographic recordings performed in conscious rats one day after catheter implantation and during saline infusion of bladder, confirmed that bladder capacity is significantly reduced in comparison with BVC values observed 5 days after catheter implantation, as previously reported [27]. It is therefore conceivable that in our experiments, performed under conditions of bladder inflammation, that a strong influence of prostaglandin levels on afferent firing was present. On the other hand, when the activity of oxybutynin or tolterodine was evaluated in rats 5 days after catheter implantation, the same effect as seen after 1 day, still occurred; namely, a decrease in MP with no change in BVC. These results indicate that the involvement of the inflammatory mediators is not the reason for the lack of antimuscarinic activity on BVC.
The involvement of ATP in non adrenergic, non cholinergic (atropine-resistant) contraction of urinary bladder is well documented [20]. Recently, it has also been demonstrated that ATP is released from the urothelium of isolated urinary bladder following increased intraluminal pressure [28]. Furthermore, it has been reported that intravesical ATP stimulates the micturition reflex in awake, freely moving rats [29] and that during cystometry the number of impulses generated in the afferent neurons was halved by treatment with suramin [19]. Consequently, suramin infusion into rat bladder has been reported to increase BVC [30]. Outflow obstruction may be associated with changes in the cholinergic function of the bladder associated with an increase of atropine-resistance. Again, the bladder instability seen in obstructed rats seems to be particularly related to the atropine-resistant contraction component, where ATP and prostaglandins play an important role [14]. However, we found that oxybutynin did not increase BVC either in rats under infusion of the bladder with suramin, or in obstructed rats. In addition, in this last experimental condition, oxybutynin did not modify the frequency and amplitude of the non-voiding contractions. These findings are in agreement with the lack of activity shown by tolterodine in the same model [31], whereas Kwak and Lee [14] reported a significant decrease of frequency and amplitude of the non voiding contractions in anesthetized rats after intraarterial administration of 1 mg/kg oxybutynin.
Although the mechanism of action of antimuscarinic agents used for the treatment of overactive bladder (such as oxybutynin and tolterodine) is thought to be due mainly to suppression of detrusor contraction through blockade of M3 muscarinic receptors on detrusor smooth muscle, an effect on central muscarinic receptors, located in the brain cannot be ruled out as both compounds pass into the central nervous system [32].
Effects on the lower urinary tract of drugs acting on central nervous system muscarinic receptors have been reported by several investigators [17,33-35] Following i.c.v. or i.t. administration in conscious normal rats of oxotremorine methiodide, a muscarinic agonist, a dose-dependent increase of BVC was observed. The muscarinic antagonist atropine did not change BVC after i.t. administration, and increased BVC after i.c.v. administration only in very high doses [36]. Both oxybutynin and tolterodine i.c.v. administered showed no (oxybutynin) or little (tolterodine) augmenting effect on BVC [17]. On the other hand, several Authors reported that oral oxybutynin administered in conscious rats with lesions at the basal forebrain [11,18], increases BVC. These findings seem to indicate that the inhibitory muscarinic mechanisms that can be activated by exogenously administered agonists seem to be inactive under normal conditions [36], but are working in lesioned animals. These considerations, therefore, can explain the lack of central activity on BVC of antimuscarinics tested in conscious normal rats.
Muscarinic receptors are also found on bladder urothelial cells, where their density may be even higher than in detrusor muscle. Evidence has been reported for a release of acetylcholine from urothelium and/or nerves during bladder filling, and acetylcholine may act directly on afferent nerves to initiate the micturition reflex or to enhance the myogenic contractile activity of the detrusor. If this is correct, blockade of muscarinic receptors should be expected to reduce bladder tone during storage and to increase bladder capacity. This effect is observed after treatment with antimuscarinics in normal individuals as well as in patients with detrusor overactivity [37]. In conscious rats, however, infravescical administration of different concentrations of oxybutynin did not increased BVC, but significantly decreased MP [12]. Furthermore, Kim et al. [38] recently showed that bladder infusion of different antimuscarinic agents (including oxybutynin) at concentrations equivalent to urine concentration in humans with oral application of these drugs, did not modified BVC when tested in normal rats but only inhibited bladder overactivity induced by intravesical instillation of carbachol.
Conclusion
Despite the different experimental conditions utilized, the main effect of the antimuscarinics tested on cystometrographic parameters in anesthetized and conscious rats is a decrease of MP, whereas BVC is hardly and, generally, non-significantly affected, suggesting that the block of bladder muscarinic receptors is the only mechanism that can be affected and evaluated after treatment with these compounds in this animal species.
Although urodynamic assessment showed significant comparable increases in bladder capacity following repeated treatment with oxybutynin and tolterodine in humans [6,7], the published papers reporting the acute effect of these antimuscarinics in man are confusing, since oxybutynin seems devoid of effect [39], whereas tolterodine increased the volumes at which subjects experienced the first sensation of bladder filling and normal desire to void [40]. It seems therefore difficult to reproduce in rats the cystometrographic effects observed in humans after chronic administration of these compounds. The effects observed in rats after acute administration (although obtained at doses about 10–100-fold higher than those used in ref. 39 and 40) seem more similar to those recorded in patients treated with a single dose, with the exception of the modification in symptoms that can not be recorded in animals.
Methods
The effects of the tested compounds on rats urodynamic parameters were evaluated by cystometrographic models in conscious and anesthetized animals. In anesthetized or obstructed animals (i.v. administration only) cystometry was performed on female rats (250–350 g b.w.). Conscious male rats (300–400 g b.w.) were used to evaluate the effect of tested compounds both after i.v. and oral administration.
Animals were housed with free access to food and water and maintained on a forced 12 hr light-dark cycle at 20–24°C for at least one week before the experiments were carried out. The animals were handled according to internationally accepted principles for the care and welfare of laboratory animals (E.E.C. Council Directive 86/609, O. J. no L358, 18/12/86).
Surgical procedures
To obtain bladder outlet obstruction, female rats were anesthetized with intraperitoneal administration of 3 ml/kg of equithensin solution (pentobarbital g 1.215, chloral hydrate g 5.312, magnesium sulphate g 2.657, ethanol ml 12.5, propylene glycol ml 49.5, distilled water to 125 ml of final volume) and then the bladder and proximal urethra were exposed via a lower abdominal midline incision. A silk ligature was placed around the urethra and tied in the presence of an intraluminally placed indwelling polyethylene cannula with an outside diameter of 1.22 mm. After removing the polyethylene cannula, the abdominal wall was sutured and then antibiotic medication (penicillin G 200,000 I.U./kg and streptomycin 300,000 I.U./kg i.m.) was administered. Obstructed rats were utilized for cystometry 3 weeks after urethral ligature.
To insert the catheter into the bladder, female rats were anesthetized with subcutaneous injection of urethane 1.25 g/kg (5 ml/kg). Animals were then placed in a supine position and an approximately 10 mm midline incision was made in the shaved and cleaned abdominal wall. The urinary bladder was gently freed from adhering tissues, emptied and then cannulated, via an incision at the dome, with a polyethylene cannula (ID 0.58 mm, OD 0.96 mm), which was permanently sutured with silk thread. For i.v. injection, another polyethylene cannula with the same characteristics and filled with heparine (40 UI/ml) in physiological saline was inserted into the jugular vein.
Obstructed female or male rats utilized 1 or 5 days after surgery were anesthetized with i.p. injection of equithensin (3 ml/kg) and catheters implantation was performed as above.
In female rats utilized upon anesthesia, the cannulae were exteriorized through a subcutaneous tunnel in the breast-bone area. In male and female rats awake utilized, the cannulae were exteriorized in the retroscapular area, where they were connected with a plastic adapter, in order to avoid the risk of removal by the animal. In male rats submitted to discontinuous cystometry, the free end of bladder cannula was connected to a swivel at the top of the cage, thus allowing free movements to animals in a 20 × 25 cm size cage. A rigid fluid-filled tubing connected to the swivel provided undistorted transmission of the bladder pressure to the transducer, which was placed outside the cage, at a height corresponding to the position of rat bladder.
Cystometry procedures
The free tip of the bladder cannula was connected by a T-shape tube to a pressure transducer and to a peristaltic pump for a constant rate continuous infusion of saline solution (at room temperature) into the urinary bladder.
In anesthetized female normal rats, the urodynamic parameters were recorded continuously using a MacLab/8SP interface with Chart Software v. 4.1.2. (AD Instrument). In conscious rats, the urodynamic parameters were obtained from the cystometrogram recorded on a chart poligraph (Rectigraph SAN-EI 8K).
In anesthetized female rats the following parameters were evaluated: bladder volume capacity (BVC), defined as the volume (in ml) of saline infused into the bladder and necessary to induce detrusor contraction followed by micturition; micturition pressure (MP, in mm Hg) defined as the maximal intravesical pressure induced by contraction of the detrusor during micturition; threshold pressure (TP, in mmHg) i.e. the difference between intraluminar basal pressure and pressure value recorded just before micturition; the number (n.HFO) and the amplitude (a.HFO, in mmHg) of high-frequency oscillations recorded in 1 sec at the middle of the expulsion time phase.
In conscious female obstructed rats, the number and the mean amplitude of the spontaneous bladder contractions, present during bladder filling but without urine emission and termed "non-voiding contractions" (NVC), were evaluated for the 2 min time prior to micturition. In addition, BVC, MP and micturition volume (MV) were evaluated.
In conscious normal male rats, only BVC and MP values (defined as above) were obtained from the cystometrograms.
Cystometric investigation in conscious normal rats
Rats were generally utilized one day after catheter implantation. Some experiments were carried out in rats utilized 5 days after surgery.
On the day of the experiment, conscious rats were placed in Bollman's cages and the recording of cystometrographic parameters started after a stabilization period of 20 min. Saline infusion into the bladder was generally at a constant rate of about 0.1 ml/min. Some experiments were performed decreasing the rate of saline infusion at 0.05 and 0.025 ml/min. Basal BVC and MP values were evaluated as the mean of two complete and reproducible cystometrograms during the pretreatment period (basal). Then the animals were treated intravenously or orally with the test compound (or vehicle) under continuous infusion of the bladder with saline, and changes in BVC and MP were evaluated for 60 or 300 min, respectively.
Discontinuous cystometry was performed for 12 hours (2 hours for basal recording and 10 hours after administration of drugs), through a discontinuous infusion of warm saline solution into the urinary bladder (1 h cycle) as reported in Fig. 11.
Figure 11 Scheme of discontinous cystometry in conscious rats. Bidirectional arrows indicate hourly periods of bladder infusion, vertical arrow indicates drug administration.
The effect of the following treatments was evaluated:
- i.v. administration of tolterodine (0.03 – 0.1 and 0.3 mg/kg) and oxybutynin (0.1 and 0.3 mg/kg) in conscious rats utilized 1 day after catheter implantation;
- p.o. administration of tolterodine (1 – 3 and 10 mg/kg) and oxybutynin (1 and 3 mg/kg) in conscious rats utilized 1 day after catheter implantation;
- p.o. administration of oxybutynin (3 mg/kg) in conscious freely-moving rats upon discontinuous cystometry;
- p.o. administration of oxybutynin (1 mg/kg) in conscious rats utilized 1 day after catheter implantation and with saline infusion rate of 0.025 and 0.05 ml/min;
- p.o. administration of oxybutynin (3 mg/kg) in conscious rats utilized 1 day after catheter implantation and with suramine infusion into the bladder at 3 × 10-6 and 1 × 10-5 M concentration.
- i.v. administration of tolterodine (0.1 and 0.3 mg/kg) in conscious rats utilized 5 days after catheter implantation;
- p.o. administration of oxybutynin (1 mg/kg) in conscious rats utilized 5 days after catheter implantation;
Cystometric investigation in anesthetized normal rats
Rats were utilized on the day of catheter implantation. Saline infusion into the bladder was at a constant rate of about 0.1 ml/min. After stabilization, basal values of the considered parameters were evaluated as mean value from the second and third cystometrogram recorded after i.v. injection of vehicle. Then, the animals were treated with the compound tested (or again with vehicle for the control group) and changes induced by treatment were evaluated by considering the value of the cited parameters as mean of the second and third cystometrogram after treatment.
The effect of i.v. administration of oxybutynin 0.3 mg/kg was evaluated.
Cystometric investigation in rats with bladder infused with diluted acetic acid
In conscious rats utilized 1 day after catheter implantation, saline solution was infused into bladder until stabilization of cystometrograms was achieved. At this point, bladder infusion was switched from saline to 0.2% acetic acid solution. Infusion rate was always 0.1 ml/min. One hr after, the rats were injected intravenously with the test compound (or vehicle) and changes in BVC and MP were evaluated for the following 60 min under continuous infusion of acetic acid. The effect of i.v. administration of tolterodine (0.03 – 0.1 and 0.3 mg/kg) and oxybutynin (0.1 and 0.3 mg/kg) was evaluated.
Cystometric investigation in conscious obstructed rats
Rats were utilized two days after catheter implantation. Saline infusion into the bladder was at a constant rate of about 0.17 ml/min. From the cystometrograms of the obstructed rats, the number and the mean amplitude of the spontaneous bladder contractions, present during bladder filling without urine emission and termed "non-voiding contractions" (NVC) were evaluated for the 2 min time prior to micturition. In addition, BVC and micturition volume (MV) were evaluated. Values of the parameters reported above were expressed as mean values obtained from two similar cystometrograms recorded just before (basal values) and as mean of the second and third cystometrogram after treatment. Since the ligature around the urethra was not removed before cystometry, peak micturition pressure was not considered in these experiments.
The effect of i.v. administration of oxybutynin 0.3 mg/kg was evaluated.
Data analysis
Data were always expressed as mean ± S.E. of the mean.
Statistical significance of the change of the different parameters recorded in anesthetized normal rats and conscious obstructed rats (before vs after treatment) was evaluated by Student's t test for paired data.
Statistical analysis of data involving time-course of BVC and MP values was performed by S.A.S./STAT software, version 6.12. The difference between vehicle and active treatments effect at different times was evaluated on values (for each rat the value at considered time minus basal value) of BVC and MP, using the general linear model procedure, repeated measures ANOVA, a univariate test of hypotheses for within subjects effects and ANOVA of contrast variables.
BVC and MP values of each rat were also transformed in AUC data. Statistical significance was evaluated by ANOVA (and Dunnett's test) or by Student's t test.
Authors' contributions
PA, CV and GS carried out the experiments on the different animal models. LG partecipated in the design of the study and performed the statistical analysis. RT conceived of the study, partecipated in its design coordination and drafted the manuscript. AL conceived and performed the supervision of the study.
All authors read and approved the final manuscript.
Table 1 Effect of intravenous administration of oxybutynin on cystometrographic parameters in anesthetized rats
BVC (ml) TP (mmHg) MP (mmHg) HFO (number) HFO (amplitude)
Controls
Before 0.47 ± 0.04 4.74 ± 0.55 25.0 ± 3.6 4.2 ± 0.2 1.82 ± 0.19
After 0.47 ± 0.04 4.88 ± 0.58 20.9 ± 2.0 4.6 ± 0.2 1.63 ± 0.11
Oxybutynin 0.3 mg/kg
Before 0.44 ± 0.11 3.44 ± 0.84 20.3 ± 2.3 4.6 ± 0.2 1.68 ± 0.27
After 0.27 ± 0.05 2.34 ± 0.51 15.5 ± 1.7 ** 4.2 ± 0.7 1.63 ± 0.24
Data are expressed as mean ± s.e. mean. BVC: bladder volume capacity; TP: threshold pressure; MP: micturition pressure; HFO: high frequency oscillations (number and amplitude). * = p < 0.05; ** = p < 0.01 before vs after administration (Student's t test).
Acknowledgements
Iris Simonazzi and Rita Cova provided excellent technical assistance.
Valuable criticism of the manuscript by D.E. Clarke, is also greatly appreciated.
==== Refs
Abrams P Cardozo L Fall M Griffiths D Rosier P Ulmsten U van Kerrebroeck P Victor A Wein A The standardisation of terminology of lower urinary tract function: report from the standardisation sub-committee of the International Continence Society Neurourol Urodyn 2002 21 167 178 11857671 10.1002/nau.10052
Milsom I Abrams P Cardozo L Roberts RG Thuroff J Wein AJ How widespread are the symptoms of an overactive bladder and how are they managed ? A population-based prevalence study BJU Int 2001 87 760 766 11412210 10.1046/j.1464-410x.2001.02228.x
Wein AJ Rovner ES The overactive bladder: an overview for primary care health providers Int J Fertil Womens Med 1999 44 56 66 10338262
Brading AF A myogenic basis for the overacting bladder Urology 1997 50 57 67 9426752 10.1016/S0090-4295(97)00591-8
De Groat WC A neurologic basis for the overactive bladder Urology 1997 50 36 52 9426749 10.1016/S0090-4295(97)00587-6
Yoshimura N Chancellor MB Current and future pharmacological treatment for overactive bladder J Urol 2002 168 1897 1913 12394674 10.1097/00005392-200211000-00004
Chapple RC Muscarinic receptor antagonists in the treatment of overactive bladder Urology 2000 55 33 46 10767450 10.1016/S0090-4295(99)00492-6
Yarker YE Goa KL Fitton A Oxybutynin. A review of its pharmacodynamic and pharmacokinetic properties, and its therapeutic use in detrusor instability Drugs & Aging 1995 6 243 262 7620236
Guarneri L Cova R Angelico P Colli E Testa R Effects of different drugs on the cystometrogram in conscious rats Pharmacol Res 1991 24 175 187 1749747 10.1016/1043-6618(91)90032-S
Guarneri L Ibba M Angelico P Testa R Effects of oxybutynin, terodiline and nifedipine on the cystometrogram in conscious rats with infravesical outflow obstruction Pharmacol Res 1991 24 263 272 1956871 10.1016/1043-6618(91)90090-K
Yamamoto T Koibuchi Y Miura S Sawada T Ozaki R Esumi K Ohtsuka M Effects of vamicamide on urinary bladder functions in conscious dog and rat models of urinary frequency J Urol 1995 154 2174 2178 7500484 10.1097/00005392-199512000-00056
Yokoyama O Ishiura Y Nakamura Y Ohkawa M Urodynamic effects of intravesical oxybutynin chloride in conscious rats J Urol 1996 155 768 771 8558722 10.1097/00005392-199602000-00110
Williamson IJR Newgreen DT Naylor AM The effects of darifenacin and oxybutynin on bladder function and salivation in the conscious rat Br J Pharmacol 1997 120 205P
Kwak TI Lee JG Inhibitory effects of propiverine, atropine and oxybutynin on bladder instability in rats with infravesical outlet obstruction BJU 1998 82 272 277 9722766 10.1046/j.1464-410X.1998.00724.x
Velasco C Guarneri L Leonardi A Testa R Influence of pump compliance (peristaltic vs. infusion) on urodynamic measurement during cystometry in conscious rats J Pharmacol Toxicol Methods 2001 45 215 21 11755385 10.1016/S1056-8719(01)00151-4
Modiri AR Alberts P Gillberg PG Effect of muscarinic antagonists on micturition pressure measured by cystometry in normal, conscious rats Urology 2002 59 963 968 12031395 10.1016/S0090-4295(02)01535-2
Ishizuka O Gu BJ Yang ZX Nishizawa O Andersson K-E Functional role of central muscarinic receptors for micturition in normal conscious rats J Urol 2002 168 2258 2262 12394771 10.1097/00005392-200211000-00110
Shimizu I Kawashima K Ishii D Oka M Urodynamics in a rat neurogenic bladder model with a unilateral electrolytic lesion of the basal forebrain BJU Int 2003 91 861 867 12780849 10.1046/j.1464-410X.2003.04234.x
Morrison J Steers WD Brading A Blok B Fry C De Groat WC Kakizaki H Levin R Thor K Abrams P, Cardozo L, Khoury S, Wein A Neurophysiology and neuropharmacology Incontinence 2002 2 Plymouth, United Kingdom: Plymbridge Distributors Ltd 85 163
Igawa Y Mattianson A Andersson K-E Functional importance of cholinergic and purinergic neuro-transmission for micturition contraction in the normal, unanesthetized rat Br J Pharmacol 1993 109 473 479 8395289
Testa R Angelico P Velasco C Guarneri L Leonardi A Urodynamic effects of oral oxybutynin in conscious rats under different cystometrographic conditions 33rd International Continence Society Annual Meeting, Florence 2003 Abs 316
Maggi CA Evangelista S Grimaldi G Santicioli P Giolitti A Meli A Evidence for the involvement of arachidonic acid metabolites in spontaneous and drug-induced contractions of rat urinary bladder J Pharmacol Exp Ther 1984 230 500 513 6431081
Maggi CA Santicioli P Furio M Meli A The effect of cyclooxygenase inhibitors on the low filling rate cystometrogram in urethane anesthetized rats J Urol 1985 134 800 803 3928908
Maggi CA Giuliani S Patacchini R Conte B Furio M Santicioli P Meli P Gragnani L Meli A The effect of SC-19220, a prostaglandin antagonist, on the micturition reflex in rats Eur J Pharmacol 1988 152 273 279 2906002 10.1016/0014-2999(88)90722-4
Maggi CA Prostanoids as local modulators of reflex micturition Pharmacol Res 1992 25 13 20 1738753 10.1016/S1043-6618(05)80059-3
Yaksh TL Durant PAC Brent CR Micturition in rats: a chronic model for study of bladder function on effect of anesthetics Am J Physiol 1986 251 R1177 R1185 3789199
Morikawa K Hichihashi M Kakiuchi M Yamauchi T Kato H Ito Y Gomi Y Effects of various drugs on bladder function in conscious rats Jpn J Pharmacol 1989 50 369 376 2779004
Ferguson DR Kennedy I Burton TJ ATP is released from rabbit urinary bladder epithelial cells by hydrostatic pressure changes – a possible sensory mechanism? J Physiol 1997 505 503 511 9423189 10.1111/j.1469-7793.1997.503bb.x
Pandita RK Andersson K-E Intravesical adenosine triphosphate stimulates the micturition reflex in awake, freely moving rats J Urol 2002 168 1230 1234 12187273 10.1097/00005392-200209000-00094
Cova R Guarneri L Angelico P Velasco C Leonardi A Testa R Effects of infravescical suramin and PPADS on cystometrographic parameters in conscious rats Urodinamica 1999 9 162 168
Fabiyi AC Gopalakrishnan M Lynch JJ Brioni JD Coghlan MJ Brune ME In vivo evaluation of the potency and bladder-vascular selectivity of the ATP-sensitive potassium channel opener (-)-cromakalim, ZD6169 and WAY-133537 in rats BJU Int 2003 91 284 290 12581020 10.1046/j.1464-410X.2003.03069.x
Andersson K-E Potential benefits of muscarinic M3 receptor selectivity Eur Urol 2002 23 28
Masuda H Tsujii T Azuma H Oshima H Role of a central muscarinic cholinergic pathway for relaxation of the proximal urethra during the voiding phase in rats J Urol 2001 165 999 1003 11176529 10.1097/00005392-200103000-00076
Yokoyama O Ootsuka N Komatsu K Kodama K Yotsuyanagi S Niikura S Nagasaka Y Nakada Y Kanie S Namiki M Forebrain muscarinic control of micturition reflex in rats Neuropharmacol 2001 41 629 638 10.1016/S0028-3908(01)00102-2
Ogasawara M Shimoda N Sato K Kato T Sugaya K The effects of the antagonists of muscarinic acethylcholine receptor subtypes in rat brain on urinary bladder contraction Nippon Hinyokika Gakki Zasshi 2002 93 427 434
Ishiura Y Yoshiyama M Yokoyama O Namiki M De Groat WC Central muscarinic mechanism regulating voiding in rats J Pharmacol Exp Ther 2001 297 933 939 11356913
Andersson K-E Yoshida M Antimuscarinics and the overactive detrusor-which is the main mechanism of action ? Eur Urol 2003 43 1 5 12507537 10.1016/S0302-2838(02)00540-7
Kim Y Yoshimura N Masuda H de Miguel F Chancellor MB Antimuscarinic agents exhibit local inhibitory effects on muscarinic receptors in bladder-afferent pathways Urology 2005 65 238 242 15708029 10.1016/j.urology.2004.11.021
Kawabe K Abe S Kanda T Tei K Clinical reevaluation of the effect of oxybutynin chloride on unhinibited neurogenic and reflex neurogenic bladder Urol Int 1986 41 16 20 3962017
Stahl MMS Ekstrom B Sparf B Mattiasson A Andersson K-E Urodynamic and other effects of tolterodine: a novel antimuscarinic drug for the treatment of detrusor overactivity Neurourol Urodyn 1995 14 647 655 8750383
|
16216132
|
PMC1274333
|
CC BY
|
2021-01-04 16:32:59
|
no
|
BMC Pharmacol. 2005 Oct 11; 5:14
|
utf-8
|
BMC Pharmacol
| 2,005 |
10.1186/1471-2210-5-14
|
oa_comm
|
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-221621266010.1186/1471-2229-5-22Research ArticleCallose (β-1,3 glucan) is essential for Arabidopsis pollen wall patterning, but not tube growth Nishikawa Shuh-ichi [email protected] Gregory M [email protected] Robert J [email protected] Daisuke [email protected] Daphne [email protected] Howard Hughes Medical Institute, Department. of Molecular Genetics and Cell Biology, University of Chicago, Chicago IL 60637, USA2 Department of Chemistry, Graduate School of Science, Nagoya University, Chikusaku, Nagoya 464-8602, Japan2005 7 10 2005 5 22 22 13 6 2005 7 10 2005 Copyright © 2005 Nishikawa et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Callose (β-1,3 glucan) separates developing pollen grains, preventing their underlying walls (exine) from fusing. The pollen tubes that transport sperm to female gametes also contain callose, both in their walls as well as in the plugs that segment growing tubes. Mutations in CalS5, one of several Arabidopsis β-1,3 glucan synthases, were previously shown to disrupt callose formation around developing microspores, causing aberrations in exine patterning, degeneration of developing microspores, and pollen sterility.
Results
Here, we describe three additional cals5 alleles that similarly alter exine patterns, but instead produce fertile pollen. Moreover, one of these alleles (cals5-3) resulted in the formation of pollen tubes that lacked callose walls and plugs. In self-pollinated plants, these tubes led to successful fertilization, but they were at a slight disadvantage when competing with wild type.
Conclusion
Contrary to a previous report, these results demonstrate that a structured exine layer is not required for pollen development, viability or fertility. In addition, despite the presence of callose-enriched walls and callose plugs in pollen tubes, the results presented here indicate that callose is not required for pollen tube functions.
==== Body
Background
Pollination begins when pollen grains adhere to stigma cells on the female pistil surface. In species with dry stigmas, including Arabidopsis thaliana, this interaction is highly selective, allowing binding of only a limited pollen grain set. Pollen adhesion occurs within seconds, is extremely strong, and is mediated by adhesives that reside in the exine, the outer pollen cell wall [1]. Exine fragments washed with organic solvents and salts retain their binding capacity, suggesting that adhesion does not require protein-protein interactions [1]. To discern the genes that mediate Arabidopsis pollen capture, we employed simple binding assays, identifying Arabidopsis mutants with less adherent pollen. Here, we cloned LAP1 [2], showing it is identical to CalS5 [3,4] a male-specific β-1,3 glucan synthase that plays a role in exine development and pollen tube composition.
Exine patterns are highly variable across plant families, making them an important feature of plant taxonomic classifications and forensic identifications [5]. These lattices are composed primarily of a chemically resistant polymer, sporopollenin, deposited in multiple layers. The ridges and spaces that comprise the exine lattice are established soon after male meiosis when an exine precursor, primexine, is deposited along the plasma membrane of the microspores, just under a layer of β-1,3 glucan (callose) that temporarily separates the developing grains [6,7] Although the molecular mechanisms that yield the vast diversity of exine patterns are poorly understood, species-specific primexine patterning requires coordination between the extracellular callose layer, the plasma membrane, and the underlying cytoskeleton, vesicles and endoplasmic reticulum [5-7].
In addition to CALS5 (discussed below), several mutations that affect exine development have been identified, providing opportunities for defining the requirements for sporopollenin synthesis and exine patterning. In the male-sterile Arabidopsis nef1 mutant, the pollen surface completely lacks sporopollenin; NEF1 encodes a predicted membrane protein that affects lipid accumulation in plastids, potentially in the tapetal cells that surround developing pollen grains [8]. While sporopollenin decorates the surface of Arabidopsis male-sterile dex1 pollen grains, it is deposited in large disorganized aggregates [7]. DEX1 encodes a predicted membrane-associated protein that is required for temporal regulation of primexine deposition [7]. Arabidopsis qrt mutants produce tetrads of fused pollen grains, each with a normally patterned cell surface [9,10]. QRT3 encodes a endopolygalacturonase required for degradation of the residual mother cell wall surrounding developing microspores [11]. Mutations that alter exine organization have also been identified in other species; for example, a pollen mutant of Haplopappus gracilis (yellow daisy) lacks densely packed exine spines, yet is fertile [12]. Genes involved in exine patterning may be particularly diverse, varying in coding sequence or site and timing of expression, and consequently contributing to the evolution of exine patterns [reviewed in [13]].
Callose plays multiple roles in pollen development, not only blanketing microspores as they form independent exine layers, but also as a major constituent of pollen tubes, the polarized extensions that deliver sperm to female gametes. Like most plant cells, pollen tubes contain pectin and cellulose, yet they have an additional layer of callose, a feature that is common to the hundreds of species examined to date [14,15]. Pollen tubes can extend hundreds to thousands of times the pollen grain's diameter, transporting all of the cytoplamsic contents, including the vegetative nucleus and two sperm, to the elongating tip. In order to maintain a manageable cytosolic volume, callose plugs are deposited periodically, separating the growing tip from the evacuated portions [14,15].
Recently plants carrying T-DNA insertions in the CALS5 gene, which encodes a male-specific β-1,3 glucan synthase were characterized [3,4]. These studies indicated that CALS5 is required to produce the temporary callose walls that separate developing microspores, and that this callose layer is critical for exine formation and fertility [4]. Here, we examined plants carrying three additional cals5 alleles; in all cases, we saw a similarly aberrant exine layer, yet the plants retained their fertility. One mutant (cals5-3) also lacked detectable callose in its pollen tubes, suggesting it is not essential for their growth or guidance through female tissues.
Results and discussion
CalS5 mutants produce adhesion-deficient pollen
We recovered cals5 mutations in a screen for genes that mediate Arabidopsis pollen capture. Using an assay that measures the number of Arabidopsis pollen grains bound to the stigma following a vigorous wash [1], we screened for plants with weak pollen adhesion, surveying a collection of 374 T-DNA insertion lines [2] and ~6700 M2 ethyl nitrosourea (ENU) mutagenized plants. This screen yielded 9 adhesion-defective plants, two of which (cals5-3 and cals5-4, Ws and Landsberg ecotypes, respectively) had severe pollen surface defects apparent with light, scanning, or transmission electron microscopy (Fig. 1). Consistent with a previously identified role for exine in pollen adhesion [1], exine patterning was altered in the adhesion mutants; the regular network of projections typical of wild type was abolished, resulting in structurally weakened pollen grains that easily collapsed (Fig. 1b,c). Transmission electron micrographs indicated that sporopollenin, the chemically resistant polymer that comprises exine, assembled into irregular aggregates in the cals5 mutants (Fig. 1j,k); these structures stained readily with the exine-specific dye, auramine O (Fig. 1f,g). Previous analysis of cals5-1 and cals5-2 mutants, showed that developing microspores degenerated, with anthers producing only 1–15 pollen grains [4]; in contrast, the mutants we identified generated fertile pollen grains in abundance (see below).
We mapped cals5-3 to chromosome 2; DNA sequencing confirmed mutations in At2g13680 (Fig. 2), the callose synthase that encodes CalS5 [3]. The SIGnAL T-DNA insertion collection [16] contained an additional allele, cals5-5 (Columbia ecotype); this insertion also caused an exine defect (Fig. 1d,h). We amplified a cDNA corresponding to CALS5, confirming intron-exon boundaries [GenBank:AY337762 and BK001470] and found that the CALS5 gene has two additional exons in the 5' UTR region; this annotated structure differs from that previously deposited in GenBank, with 41 exons, instead of the 39 described before [4]. The cals5-3, 5-4, and 5-5 lesions were localized to exon 41, exon 18, and intron 5, respectively (Fig. 2). In contrast to the null defects identified previously [4], RT-PCR analysis showed that each of mutant lines we characterized produced cals5 mRNA, although levels were reduced in cals5-5, perhaps reflecting inefficient splicing of intron 5, due to the large T-DNA insertion (not shown). All three cals5 exine defects were fully recessive (e.g. Fig. 1l), and pair-wise crosses between each mutant failed to complement.
T-DNA constructs containing a full-length genomic copy of CALS5 were lethal in Agrobacterium tumefaciens, the organism commonly used for Arabidopsis transformation; thus, we were unable to test a transgene for complementation. Nonetheless, because three independent non-complementing mutations were identified, each of which caused a pollen surface phenotype, and two additional exine-deficient alleles have been described [4], it is highly likely that we identified the locus that caused the adhesion defect. The predicted CALS5 catalytic domain shares 59 – 69% amino acid identity to 11 other putative Arabidopsis callose synthase genes [3,17] and 82% amino acid identity to a callose synthase expressed in Nicotiana alata pollen tubes (NaGsl1, [18], [GenBank:AAK49452]).
CALS5 expression is male-specific
We detected CALS5 expression throughout pollen development, both in pollen mother cells and in developing and mature pollen grains (Fig. 3). By Northern blot analysis, CALS5 messages were sufficiently abundant for detection in flower buds, but not vegetative tissues or whole flowers (Fig. 3a), consistent with publicly available microarray datasets showing low, but detectable CALS5 expression, beginning in immature flowers and persisting through pollen development [19]. With in situ hybridization we showed CALS5 mRNA was present at the beginning of pollen development, within pre-meiotic pollen mother cells (Fig. 3b,c). Transgenic plants carrying CALS5 fused to reporter genes yielded somewhat different results. Previously, a fusion of 2.2 kb upstream and 20 amino acids of the CaslS5 coding sequence to β-glucuronidase (GUS) revealed expression in root tips and in the vegetative vasculature of stems and leaves, and when fused to GFP, showed expression in pollen tetrads, but not pollen mother cells [4]. Here, we fused a 1.8 kb fragment upstream of the CALS5 start codon to GUS, and found abundant expression in pollen grains starting one day before maturity, but not at earlier stages, and no expression in female reproductive tissues or in other vegetative parts of the plant (not shown); these results were replicated in multiple lines, all of which expressed high levels of GUS in mature pollen. Because direct detection of mRNA showed a different pattern, it may be that an essential promoter element is absent from the fragments used to construct the GUS and GFP fusions. Taken together, the phenotypic and expression data point to a role for CALS5 in pollen mother cells and pollen grains, and, although no vegetative phenotypes were observed, expression in vegetative tissues is also possible, due to a putative enhancer more than 1.8 kb upstream of the coding sequence.
The role of callose in exine formation
Tobacco plants overexpressing callase produce inviable pollen with aberrant exine [20,21]. To further explore the relationship between callose formation and exine patterning, we used aniline blue staining to examine the callose that surrounds and separates newly formed cals5 microspores (Fig. 4). cals5-3 showed a strong defect; unlike wild-type, cals5-3 tetrads completely lacked a peripheral callose layer and had little callose between the microspores (Fig. 4a,b). This peripheral callose layer was also missing from cals5-4 and 5-5, but significant callose was present between the microspores (Fig. 4c,d). Interestingly, qrt2 mutants have an opposite callose phenotype – they produce fused pollen tetrads with normally patterned exine, and have an intact peripheral callose layer, but discontinuous callose walls between microspores [9]. The cals5-1 and 5-2 alleles appear to be stronger than those reported here; neither peripheral nor interstitial callose walls were detected around microspores, and upon release, developing pollen grains were deformed and often burst [4]. Thus, the cals5 allelic series makes it possible to discern varied roles for callose in developing microspores. A complete absence (cals5-1 and 5-2) disrupts both exine formation and pollen viability, while an absence of pheripheral, but not interstitial walls (cals5-3, 5-4, and 5-5), results in viable pollen with aberrantly patterned exine.
Exine formation requires intracellular functions that nucleate sporopollenin deposition at the pollen plasma membrane; the Arabidopsis DEX1 gene product may play such a role [6,7]. Our data indicate that an extracellular callose layer is also required. Conceivably, callose could trap primexine subunits, increasing their local concentration and preventing them from diffusing into the anther locule. These subunits may self-assemble onto a scaffold, nucleated by intracellular components, at the plasma membrane. Subsequently, the removal of callose walls may allow rod-shaped baculae to form on the primexine template. This model is consistent with the cals5 phenotype, where disorganized aggregates of sporopollenin collect on the pollen surface (Fig. 1e–k), suggestive of fewer sites of primexine nucleation. It is also possible that callose provides a physical support for primexine assembly, or interacts directly with primexine nucleation sites on the microspore plasma membrane. Distinguishing these models would be facilitated by an in vitro system capable of nucleating the assembly of sporopollenin polymers in a manner that recapitulates species-specific patterning.
The role of callose in pollen tube growth and fertility
Mature pollen contained CalS5-GUS, suggesting a role for callose synthase during pollination. To explore this possibility, we used aniline blue staining to assess callose content in pollinated pistils. Wild type pollen tube walls were readily visualized on the stigma surface and callose plugs appeared as bright punctate dots in the pistil interior (Fig. 4e). These features were also apparent in pollen tubes germinated in vitro (Fig. 4i,j[22]); 81.5% of wild-type tubes (n = 351) contained callose plugs, and every tube showed intense callose staining in the outer wall (Fig. 4m,n). In contrast, cals5-3 pollen tubes had a severe callose defect; in pollinated pistils, callose was absent from tube walls and bright staining plugs were not detected (Fig. 4f). Consistent with their weaker effect on callose synthesis in pollen mother cells, cals5-4 and cals5-5 mutants produced pollen tubes with apparently normal callose levels (Fig. 4g,h). We confirmed the cals5-3 phenotype in vitro; pollen tubes grew normally, yet they lacked aniline staining above background levels and no callose plugs were observed, even with differential interference contrast microscopy (n = 150, Fig. 4k,l,o,p). We also showed that callose-specific staining with an anti-β-1,3 glucan antibody was virtually abolished in cals5-3 pollen tubes (Fig. 4q), while cellulose and pectin (Fig. 4r,s) were unaffected. Lastly, to verify that the same genetic defect disrupts callose synthesis in cals5-3 pollen mother cells and pollen tubes, we examined 63 F2 plants derived from a cross of cals5-3 to wild-type; 15 homozygous mutant lines (verified by PCR) lacked detectable callose in pollen tubes and had aberrant exine, while the remaining cals5-3 /+ and +/+ lines had normal exine. Together these results indicate that cals5-3 is a stronger allele, with severe disruption of callose synthesis both in pollen mother cells and pollen tubes; cals5-4 and cals5-5 have weaker phenotypes, suggesting a reduction, but not complete loss of function.
The absence of callose in pollen tubes did not eliminate pollen germination, polarized tube growth, pollen tube guidance to the ovules, or the delivery of sperm cells. Despite their deficiency in callose content, cals5-3 pollen tubes germinated with normal efficiencies in vitro (cals5-3, 577/876, 65.9%; wild-type, 466/702, 66.4%), and grew at normal rates (after 6 hrs at 22°C, cals5-3, 99 ± 37 μm, n = 110; wild-type, 93 ± 43 μm, n = 140). In vivo, cals5-3 pollen tubes extended down the length of the pistil at apparently the same rate as wild type, and exhibited wild-type targeting behaviour (n = 300 ovules; Fig. 5a,b). Moreover, self-pollinated cals5-3 plants were fertile, producing normal seed sets. This was somewhat unexpected, as callose plugs have been assumed to be essential for limiting the cytoplasmic volume of the pollen tube. Mature Arabidopsis pistils measure ~2.5 mm and pollen tubes have 6 μm diameters; thus, we estimate that tubes would undergo at least a 20-fold increase in cytosolic volume if they lacked callose plugs, a factor that would be increased by meandering tube growth. Our results suggest that Arabidopsis pollen tubes can tolerate a large expansion in cell volume and that efficient organelle transport can occur within tubes over relatively long distances.
Callose and pollen tube fitness
CALS5 gene expression patterns and cals5 mutant phenotypes suggest it functions in both pollen mother cells and pollen tubes (sporophytic and gametophytic tissues [23,24]). The cals5 exine patterning defect was fully recessive and therefore showed sporophytic inheritance (Fig. 1l), while cals5-3 /+ plants segregated 1:1 for pollen tube callose content (310:278 callose+:callose-), indicating gametophytic segregation. This gametophytic role allowed us to test for subtle defects in cals5-3 pollen fitness by placing mutant pollen in competition with wild type. First, we noted that self-pollinated cals5-3 /+ plants yielded exine-defective progeny at a distorted ratio (1904:362 CALS5+: CALS5-; ~5:1). Because flowers from cals5 plants consistently produce pollen with exine defects, it is unlikely that this distorted segregation pattern reflects incomplete penetrance; rather, cals5 haploid cells are less functional than wild type. This effect was restricted to male gametophytes: crossing cals5-3 /+ pistils to wild type pollen yielded normal segregation, while the reciprocal cross showed 31% of the progeny inherited cals5-3, differing significantly from the expected 50% ratio (P <0.01, Fig. 5c). The ratio of cals5-3 /+ : +/+ seeds was not significantly different (P > 0.4, χ2 test) in the upper and lower halves of these developing siliques (upper, 37:60; lower, 15:31, 5 pistils sampled), suggesting that cals5-3 pollen tube deficiencies were not exacerbated with additional tube growth. Because these crosses did not produce full seed sets, competition was not complete, and it remains possible that the inability of cals5-3 pollen to sire as many progeny as wild type is due to slower pollen tube growth in vivo. It is also possible that the pollen tube volume increase predicted to result from the absence of callose plugs could account for a loss in pollen competence at these final stages, perhaps diluting key components in an expanded cytosol. Lastly, the heightened demand on systems that transport and secrete materials to the pollen tube tip could reduce the ability of cals5-3 pollen tubes to deliver sperm. In either case, the selective pressure of a natural environment could place pollen grains lacking callose at a disadvantage.
Conclusion
Here we characterized three alleles of the Arabidopsis CALS5 gene, dissecting the roles of this callose synthase in pollination. In early pollen development, CALS5 has a sporophytic function, forming callose that surrounds pollen mother cells and separates developing microspores. All cals5 mutants showed defects at this stage and mutant pollen grains had a pronounced exine defect, providing compelling evidence that callose deposition determines exine patterning. While strong CALS5 alleles (cals5-1 and cals5-2) caused pollen degeneration and sterility [4], we identified CALS5 mutants (cals5-3, cals5-4, cals5-5) that are fertile. This work points to the value of an allelic series: other than an exine patterning defect, cals5-4 and cals5-5 pollen was normal, while cals5-3 pollen produced tubes that lacked callose walls and plugs. These differences did not correlate with genetic background; cals5-1, 5-2, and 5-5 were from the Columbia ecotype, cals5-3 was Ws, and cals5-4 was Landsberg. Thus, we can define three distinct roles for CALS5 in pollen development: i) patterning the exine layer, ii) forming callose in pollen tubes, and iii) preventing pollen degeneration early in development.
The cals5 mutants also provide important insight into how variation in callose synthesis could contribute to taxonomic diversity in exine patterns. For example, changes in the site and timing of CalS5 activity could have profound changes on exine structure. In this respect, the callose layer at the pollen mother cell periphery is key; cals5-4 and cals5-5 retained callose between developing microspores yet produced uniformly aberrant exine. The gametophytic role of CALS5 in pollen tubes may provide an added evolutionary constraint on CALS5 variation. Direct competition between cals5-3 and wild-type pollen showed that callose does provide a fitness advantage at late stages of pollen tube growth. Thus, changes in patterns of callose synthesis at the pollen mother cell must be balanced with the selective pressures that lead to a nearly universal presence of callose plugs; such changes would be possible by differentially modulating sporophytic and gametophytic CALS5 expression profiles.
Methods
Arabidopsis strains and growth conditions
Although it is caused by a point mutation, cals5-3 (Ws-2, CS 2330) was derived from a pool of T-DNA mutagenized seed; cals5-4 was isolated from cer6-2 (Landsberg erecta, CS 6242) mutagenized with 1.25 mM ethyl nitrosourea overnight at 22°C; cals5-5 (Columbia, SALK 072226) was obtained from the SIGnAL project [Genbank:BH850992]. Plants were grown on soil with 24 h fluorescent lighting at 22°C, or in a greenhouse. Pollen-stigma adhesion was assayed in self-pollinated M2 pistils as described [1].
Genetic analysis
cals5-3 was not linked to a T-DNA insertion; consequently, we used PCR-based genetic markers to map CALS5 in a population generated by crossing cals5-3 to wild-type Columbia. Analysis of 210 F2 plants localized the mutation between nga1145 (10 cM) and mi398 (29 cM) on chromosome 2. The mutation was narrowed to a 122 kb region on BAC F13J11, between T10F5g14B (F-GCTTGTCGGCGATTAAGGTTGGTC; R-GCCCGCATGCAGAAAGAACACC; Hpy188I) and F13J11.8 (F-ACCAATCGGGCCTGACCTTATC; R-CTAGTGGCGTTACACCATTATTTTCAGTTTAG; MspI); restriction digestion cuts only the Columbia products. cals5-3 has a 6-bp deletion in At2g13680, which removes an EcoRV restriction site; PCR amplification of the region flanking this site (primers: F-CCGAATGGGAAAACAGCACAG, R-CTTACCACCGGCGAGAATACG) followed by EcoRV digestion was used to score cals5-3 alleles in segregating populations.
RNA analysis
Total RNA was prepared by the TRIzol method [25] or with an RNeasy kit (Qiagen). CALS5 cDNA was isolated by RT-PCR from flower bud mRNA and extended by 5'RACE (Invitrogen). PCR products were cloned into pCR2.1 (Invitrogen), and the nucleotide sequence of the CALS5 cDNA was determined. Semi-quantitative RT-PCR for 22, 25, 28, and 30 cycles was used to assess the levels of expression of each cals5 allele (not shown), using primers that amplified four different portions of the gene (regions adjacent to each cals5 mutation, spanning exons 3–5, 6–8, 19–22, and 40–41). For in situ hybridization analysis, a 2.8-kb CALS5 cDNA SacI fragment was cloned in both orientations into pGEM4z (Promega) and antisense and sense digoxigenin-labeled probes were prepared with in vitro transcription using SP6 RNA polymerase. This SacI fragment corresponds to the N-terminal portion of CALS5 and lacks significant similarity to other Arabidopsis callose synthases; the probe for Northern blotting included this region, as well as the 5' UTR.
CALS5 promoter-GUS fusion
A 0.26-kb HindIII/EcoRV fragment of p35S-2 containing the 35S terminator region was cloned into the HindIII and StuI sites of pGreenII 0229 [26] to give pDM1. The CALS5 regulatory region 1815 bp upstream from the translation initiation codon was amplified by PCR using primers 5'-GCGGTCGACACCATATTTTGTCCATGTAAGAC-3' and 5'-GCGAAGCTTCATATGCTCTCCCTGTTACAAAACATTG-3'. The amplified fragment was cloned upstream of the Escherichia coli GUS gene. The resultant plasmid, pDM10, was introduced into Agrobacterium tumefaciens strain GV3101 with pSOUP [26] by electroporation. Arabidopsis was transformed using the floral dip method [27], and transformants were selected on soil spraying the seedlings with 300 μM glufosinate ammonium. Flowers from the T2 transgenic plants were stained for GUS activity as described [9]. Samples were cleared in 70% ethanol and photographed.
Microscopy
Pollen grains were coated with 10 nm of gold (Denton DESK II sputter coater) and viewed with a JEOL JSM-840A scanning electron microscope at 10 kV. LR White resin sections of microspores, transmission electron microscopy, and aniline blue staining were performed as described [20,28]. Pollen germination and tube growth assays were performed as described [29]. Pollen tubes germinated on stigmas or in vitro were stained with 0.1% Congo red [22], 1% alcian blue 8GX in 3% acetic acid, or 0.01% calcofluor white (fluorescent brightener 28) in 0.1 M K2HPO4. For auramine O staining, pollen grains were suspended in 0.1% auramine O and 50 mM Tris-HCl, pH 7.5 and observed under a Zeiss LSM-510 laser-scanning confocal microscope using the filter set suitable for FITC. Indirect immunofluorescence detection of callose was performed on in vitro germinated pollen tubes, fixed and treated with cellulase and macerozyme as described [29], labelled with monoclonal mouse anti-β-1,3 glucan antibodies (1:100, Biosupplies, Parkville, Australia), and visualized with Alexa488-labeled goat anti- mouse antibodies (1:1000, Molecular Probes). In situ hybridization was carried out as described [30].
Authors' contributions
SN and GMZ carried out the mutant screens, participated in the mapping and cloning CALS5 and helped draft the manuscript. RJS carried out adhesion assays and assembled constructs for complementation. DM performed in situ hybridization experiments. DP, SN, and GMZ conceived of this study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We dedicate this paper to the memory of H. Swift, a generous and valued colleague who provided advice for this study. We also thank R. Palanivelu, J. Greenberg, and J. Malamy for helpful discussions; A. Edlund, N. Yao, and E. Williamson for assistance with microscopy; B. Alexander, J. Hill and C. Shroedl for mapping and cloning; S. Swanski, J. Zdenek and J. Coswell for greenhouse support; and H. Tanaka for assistance with in situ hybridization. This work was supported by a Yamada Science fellowship (SN), USDA awards (GMZ and RJS), the Howard Hughes Medical Institute and University of Chicago Materials Research Science and Engineering Center (DP).
Figures and Tables
Figure 1 Arabidopsis mutants defective in exine pattern formation. Scanning electron micrographs of pollen grains (a-d), confocal images of pollen stained with 0.1% auramine O (e-h,l), and transmission electron micrographs of mature pollen sections (i-k, arrows indicate sporopollenin). Wild type Landsberg (a,e and i), cals5-3 (b,f, and j), cals5-4 (c, g, and k), cals5-5 (d and h), and cals5-3 /+ (l). Bars, 10 μm (a-h, l); 1 μm (i-k).
Figure 2 CALS5 mutations. The CALS5 5769 bp open reading frame encodes a predicted protein measuring 1923 amino acids. Boxes, exons; filled boxes, similar to domains in glycogen synthase; gray boxes, 5' untranslated region. cals5-3 is a deletion of amino acids 1848–1849; cals5-4, a G651E substitution; cals5-1, cals5-2 and cals5-5, T-DNA insertions.
Figure 3 CALS5 is expressed in mature pollen and pollen mother cells. Northern blot of mRNA isolated from various tissues and probed with CALS5 and EF1α (a). In situ hybridization of pollen mother cells (pm) within sectioned anthers (b-c); CALS5 anti- sense (b) and sense probe (c). Bar, 100 μm.
Figure 4 Callose content of cals5 pollen tubes. Aniline blue staining (a-h,j,l,n,p) and differential interference contrast optics (i,k,m,o) of sections of tetrad stage anthers (a-d), self-pollinated pistils (e-h), pollen tubes germinated on an excised stigma and elongating in vitro (i-l), in vitro germinated pollen tubes; arrowheads, callose plugs (m-p). Wild-type (a,e,i,j,m,n), cals5-3 (b,f,k,l,o,p), cals5-4 (c,g), and cals5-5 (d,h). Wild type (upper) and cals5-3 (lower) pollen tubes stained with anti-β-1,3 glucan primary and FITC secondary antibodies (q), cellulose staining with calcofluor white (r), and pectin staining with Alcian blue 8GX (s). Bars, 10 μm (a-d, m-s); 100 μm (e-l).
Figure 5 cals5-3 pollen tube function. Pollen tubes from self-pollinated wild-type (a) and cals5-3 (b) pistils, stained with 0.1% Congo red and viewed with a confocal microscope. Bar, 50 mm; arrowheads, pollen tubes entering the micropyle; ov, ovule. Offspring ratios determined by PCR from reciprocal crosses of wild-type and heterozygous cals5-3 plants (c); P, probability determined by χ2.
==== Refs
Zinkl GM Zwiebel BI Grier DG Preuss D Pollen-stigma adhesion in Arabidopsis : a species-specific interaction mediated by lipophilic molecules in the pollen exine Development 1999 126 5431 5440 10556067
Zinkl GM Preuss D Dissecting Arabidopsis pollen-stigma interactions reveals novel mechanisms that confer mating specificity Annals Bot Suppl 2000 85 15 21 10.1006/anbo.1999.1066
Verma DPS Hong Z Plant callose synthase complexes Plant Mol Biol 2001 47 693 701 11785931 10.1023/A:1013679111111
Dong X Hong Z Sivaramakrishnan M M M Verma DPS Callose synthase (CalS5) is required for exine formation during microgametogenesis and for pollen viability Arabidopsis Plant J 2005 42 315 328 15842618 10.1111/j.1365-313X.2005.02379.x
Scott RJ RJ Scott, MA Stead Pollen exine – the sporopollenin enigma and the physics of pattern Molecular and Cellular Aspects of Plant Reproduction Society for Experimental Biology Seminar Series 55 1994 Cambridge, UK: Cambridge University Press 49 81
Paxon-Sowders DM Owen HA Makaroff CA A comparative ultrastructural analysis of exine pattern development in wild-type Arabidopsis and a mutant defective in pattern formation Protoplasma 1997 198 53 65 10.1007/BF01282131
Paxon-Sowders DM Dodrill CH Owen HA Makaroff CA DEX1, a novel plant protein, is required for exine pattern formation during pollen development in Arabidopsis Plant Physiol 2001 127 1739 1749 11743117 10.1104/pp.127.4.1739
Ariizumi T Hatakeyama K Hinata K Inatsugi R Nishida I Sato S Kato T Tabata S Toriyama K Disruption of the novel plant protein NEF1 affects lipid accumulation in the plastids of the tapetum and exine formation of pollen, resulting in male sterility in Arabidopsis thaliana Plant J 2004 39 170 181 15225283 10.1111/j.1365-313X.2004.02118.x
Preuss D Rhee SY Davis RW Tetrad analysis possible in Arabidopsis with mutation of the QUARTET (QRT) genes Science 1994 264 1458 1460 8197459
Rhee SY Somerville CR Tetrad pollen formation in quartet mutants of Arabidopsis thaliana is associated with persistence of pectic polysaccharides of the pollen mother cell wall Plant J 1998 1 79 88 10.1046/j.1365-313X.1998.00183.x
Rhee SY Osborne E Poindexter PD Somerville CR Microspore separation in the quartet3 mutants of Arabidopsis is impaired by a defect in a developmentally regulated polygalacturonase required for pollen mother cell wall degradation Plant Physiol 2003 133 1170 1180 14551328 10.1104/pp.103.028266
Jackson RC Skvarla JJ Chissoe WF A unique pollen wall mutation in the family Compositae: Ultrastructure and genetics Am J Bot 2000 87 1571 1577 11080106
Edlund AF Swanson R Preuss D Pollen and stigma structure and function: The role of diversity in pollination Plant Cell Suppl 2004 16 84 97 10.1105/tpc.015800
Kauss H M Smallwood, JP Knox, DJ Bowles Oxford Callose synthesis Membranes: Specialized functions in plants 1996 UK: BIOS Scientific Publishers Ltd 77 92
Taylor LP Hepler PK Pollen germination and tube growth Annu Rev Plant Physiol Plant Mol Biol 1997 48 461 491 15012271 10.1146/annurev.arplant.48.1.461
Alonso JM Stepanova AN Leisse TJ Kim CJ Chen H Shinn P Stevenson DK Zimmerman J Barajas P Cheuk R Gadrinab C Heller C Jeske A Koesema E Meyers CC Parker H Prednis L Ansari Y Choy N Deen H Geralt M Hazari N Hom E Karnes M Mulholland C Ndubaku R Schmidt I Guzman P Aguilar-Henonin L Schmid M Weigel D Carter DE Marchand T Risseeuw E Brogden D Zeko A Crosby WL Berry CC Ecker JR Genome-wide insertional mutagenesis of Arabidopsis thaliana Science 2003 301 653 657 12893945 10.1126/science.1086391
Østergaad L Petersen M Mattson O Mundy J An Arabidopsis callose synthase Plant Mol Biol 2002 49 559 566 12081364 10.1023/A:1015558231400
Doblin MS De Melis L Newbigin E Bacic A Read SM Pollen tubes of Nicotiana alata express two genes from different β-glucan synthase families Plant Physiol 2001 125 2040 2052 11299383 10.1104/pp.125.4.2040
Honys D Twell D Comparative analysis of the Arabidopsis pollen transcriptome Plant Physiol 2003 132 640 652 12805594 10.1104/pp.103.020925
Worrall D Hird DL Hodge T Paul W Draper J Premature dissolution of the micosporocyte callose wall causes male sterility in transgenic tobacco Plant Cell 1992 4 759 771 1392594 10.1105/tpc.4.7.759
Tsuchiya T Toriyama K Yoshikawa M Ejiri S Hinata K Tapetum-specific expression of the gene for an endo- β-1,3-glucanase causes male sterility in transgenic tobacco Plant Cell Physiol 1995 36 487 494 7757338
Palanivelu R Brass L Edlund AF Preuss D Pollen tube growth and guidance is regulated by POP2, an Arabidopsis gene that controls GABA levels Cell 2003 114 47 59 12859897 10.1016/S0092-8674(03)00479-3
Mascarenhas JP Gene activity during pollen devlopment Annu Rev Plant Physiol Plant Mol Biol 1990 41 317 338 10.1146/annurev.pp.41.060190.001533
Johnson MA Preuss D Plotting a course: multiple signals guide pollen tubes to their targets Dev Cell 2002 2 273 281 11879633 10.1016/S1534-5807(02)00130-2
Chomczynski P Sacchi N Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction Anal Biochem 1987 162 156 159 2440339 10.1016/0003-2697(87)90021-2
Hellens RP Edwards EA Leyland NR Bean S Mullineaux PM pGreen: a versatile and flexible binary Ti vector for Agrobacterium-mediated plant transformation Plant Mol Biol 2000 42 819 832 10890530 10.1023/A:1006496308160
Clough SJ Bent AF Floral dip: a simplified method for Agrobacterium-mediated transformation of Arabidopsis thaliana Plant J 1998 16 735 743 10069079 10.1046/j.1365-313x.1998.00343.x
Preuss D Lemieux B Yen G Davis RW A conditional sterile mutation eliminates surface components from Arabidopsis pollen and disrupts cell signaling during fertilization Genes Dev 1993 7 974 985 8504936
Lin Y Wang Y Zhu JK Yang Z Localization of a Rho GTPase implies a role in tip growth and movement of the generative cell in pollen tubes Plant Cell 1996 8 293 303 12239385 10.1105/tpc.8.2.293
Long JA Barton MK The development of apical embryonic pattern in Arabidopsis Development 1998 125 3027 3035 9671577
|
16212660
|
PMC1274334
|
CC BY
|
2021-01-04 16:03:52
|
no
|
BMC Plant Biol. 2005 Oct 7; 5:22
|
utf-8
|
BMC Plant Biol
| 2,005 |
10.1186/1471-2229-5-22
|
oa_comm
|
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-351622567510.1186/1471-244X-5-35Research ArticleZDHHC8 as a candidate gene for schizophrenia: Analysis of a putative functional intronic marker in case-control and family-based association studies Faul Thomas [email protected] Micha [email protected] Martin [email protected] Sven [email protected] Bruno [email protected] Burkhard [email protected] Michael [email protected]öber Gerald [email protected] Department of Psychiatry and Psychotherapy, University of Würzburg, Füchsleinstraße 15, 97080 Würzburg, Germany2 Department of Forensic Medicine, University of Würzburg, Lindleinstraße 15, 97080 Würzburg, Germany3 Institute of Medical Biometry, Informatics and Epidemiology, University of Bonn, Sigmund-Freud-Str. 25, 53105 Bonn, Germany2005 14 10 2005 5 35 35 15 7 2005 14 10 2005 Copyright © 2005 Faul et al; licensee BioMed Central Ltd.2005Faul et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The chromosome 22q11 region is proposed as a major candidate locus for susceptibility genes to schizophrenia. Recently, the gene ZDHHC8 encoding a putative palmitoyltransferase at 22q11 was proposed to increase liability to schizophrenia based on both animal models and human association studies by significant over-transmission of allele rs175174A in female, but not male subjects with schizophrenia.
Methods
Given the genetic complexity of schizophrenia and the potential genetic heterogeneity in different populations, we examined rs175174 in 204 German proband-parent triads and in an independent case-control study (schizophrenic cases: n = 433; controls: n = 186).
Results
In the triads heterozygous parents transmitted allele G preferentially to females, and allele A to males (heterogeneity χ2 = 4.43; p = 0.035). The case-control sample provided no further evidence for overall or gender-specific effects regarding allele and genotype frequency distributions.
Conclusion
The findings on rs175174 at ZDHHC8 are still far from being conclusive, but evidence for sexual dimorphism is moderate, and our data do not support a significant genetic contribution of rs175174 to the aetiopathogenesis of schizophrenia.
==== Body
Background
Systematic approaches designed to identify schizophrenia susceptibility genes have implicated the chromosome 22q11 microdeletion region as a promising hot-spot [1,2]. Recently, Mukai and colleagues [3] proposed a gender-specific effect of the ZDHHC8 gene at 22q11 to liability to schizophrenia based on animal models followed by human association studies. ZDHHC8 is a brain expressed putative palmitoyltransferase of 765 amino acids and may, thus, be involved in synaptic transmission and posttranslational modification of yet unidentified substrates. The ZDHHC8 gene consists of 11 exons spanning ~16 kb on genomic DNA (; OMIM: *608784). A single nucleotide polymorphism (SNP), rs175174 at intron 4, provided the highest significance from the entire 22q11 locus [4], and cell experiments gave evidence that rs175174 (A/G) may modify ZDHHC8 expression by causing imperfect splicing, intron retention and reduced enzyme activity. Zdhhc8 knockout mice had sexually dimorphic deficits in prepulse inhibition and other features proposed to mimic schizophrenia in mice. Subsequently, fitting into the animal model analysis of 389 proband-parent trios showed significant over-transmission of allele A in female, but not male subjects with schizophrenia [3]. Studies aiming replication of the genetic evidence challenged the initial findings in populations of Asian or European origin either by a lack of association or by reporting the allele G over-transmitted to schizophrenia in a non-gender specific manner [5-7]. Given the genetic complexity of schizophrenia and the potential genetic heterogeneity in different populations as highlighted by the recent findings, we decided to conduct an association study on rs175174 in ZDHHC8 using both population- and family-based samples of an ethnically homogenous German population.
Methods
The proband-parent trio sample encompassed 204 cases, who fulfilled diagnostic criteria of DSM IV for schizophrenia [8], and their biological parents. The index cases (136 males; 67%) had a mean age at onset of 23.8 years (SD 7.5) and an age at assessment of 32.4 years (SD 8.8). In 60 triads (29%) we found a parent (33 females) previously treated for psychosis. There was no evidence for bilineal transmission in the sample. The sample for the independent case-control association study involved 433 probands with schizophrenia (321 males; 74%). The probands had a mean age of first hospitalization at 28.9 years (SD 10.6), and an age at assessment of 39.2 (SD 13.5) years. In 15% (n = 64) a further first-degree relative had been recorded as hospitalized for schizophrenia (familial cases). Subjects fulfilled diagnosis of cycloid psychoses (n = 149), unsystematic (n = 155) and systematic schizophrenia (n = 129) according to differentiated psychopathology [9]. The probands were recruited at the Department of Psychiatry and Psychotherapy at the University of Würzburg. The 186 volunteer control subjects (105 males; 56%) were recruited from the blood donor centre at the University of Würzburg at a mean age of 30.0 years (SD 10.3). All subjects were unrelated and of German Caucasian descent. The Ethics Committee of the University of Würzburg had approved the study, and informed consent was obtained from all subjects.
PCR for allelic discrimination at SNP rs175174 (A/G) was performed in a final reaction volume of 10 μl containing 20 ng genomic DNA and 5 μl of 20× TaqMan®Universal PCR Master Mix (Applied Biosystems) and 0.5 μl of 20× TaqMan™ validated SNP genotyping assay including fluorescent tags specific for the wild type allele and the variant allele. Marker amplification was performed in microtiter plates on Biometra machines (Whatman). PCR amplification conditions were according to the facturer's recommendation [10 min at 95°C followed by 15 sec at 92°C and 60 sec at 60°C for 40 cycles]. Allelic discrimination with endpoint detection of fluorescence was performed at 60°C on an ABI prism 7000 sequence detection system followed by analysis with an appropriate software package (Applied Biosystems).
Fisher's exact and Armitage's trend test and were used to compare allelic and genotypic distributions between cases and controls. One-way ANOVA as implemented in the SAS was applied to compare the age-at-onset between genotypes. The data of the proband-parent triads were analyzed by the transmission-disequilibrium-test (TDT) [10]. Power calculations are based on the approximation described by Jackson et al. [11]; the exact test proposed by Weir [12] was applied for Hardy-Weinberg equilibrium.
Results
In the sample of 204 triads, no transmission distortion was apparent in the whole sample (TDT = 0.13, p = 0.72; Table 1) and in the sub-samples of families with maternal affection (TDT = 0.13, p = 0.72) or paternal affection (TDT = 0.17, p = 0.68). Even after stratifying for gender of the proband, there was no significant over-transmission of an allele to males (TDT = 2.31, p = 0.13) and to females (TDT = 2.25, p = 0.14; Table 1). However, allele A is preferentially transmitted to males, whereas allele G is more often transmitted to females and this gender-related heterogeneity of allele transmission was significant (heterogeneity χ2 = 4.43; p = 0.035).
Table 1 Transmission of marker rs 175174 A/G in 204 traids with schizophrenic psychoses
total sample (N = 204) females (N = 68) males (N = 136)
MT AA AG GG AA AG GG AA AG GG
AAxAA 32 - - 10 - - 22 - -
AAxAG 37 31 - 9 13 - 28 18 -
AAxGG - 32 - - 13 - - 19 -
AGxAG 14 23 14 4 7 8 10 16 6
AGxGG - 9 10 - 2 2 - 7 8
GGxGG - - 2 - - 0 - - 2
n 83 95 26 23 35 10 60 60 16
% 40.7 46.6 12.7 33.8 51.5 14.7 44.1 44.1 11.8
HWE p-value 0.37 0.78 0.37
TDT 97 T/92 NT 26 T/38 NT 71 T/54 NT
p-value 0.72 0.13 0.13
Transmission from heterozygous parents: MT = Mating Type; HWE = Hardy-Weinberg-Equilibrium; TDT = transmission-disquilibrium test; T = transmitted, NT = Non-transmitted.
In our case-control panel of 619 individuals, allele and genotype frequencies were not significantly different between cases and controls, and there was no evidence for gender-specific effects (Table 2). The familial/sporadic distinction produced negative results as did a sub-classification according to differentiated psychopathology (data not shown). Finally, genotypes were not associated with different mean age-at-onset in both panels (ANOVA p = 0.066). Genotypes were in appropriate Hardy-Weinberg equilibrium (HWE), and the allele frequency of rs175174G was equal in the volunteer control sample (0.63) and the non-transmitted alleles in triads (0.63).
Table 2 Allele and genotype frequency at marker rs 175174 in schizophrenic psychoses
Allele (%) Genotypes (%)
Samples A G p-value AA AG GG p-value
total
SCZ (N = 433) 527 (60.9) 339 (39.1) 0.48 154 (35.6) 219 (50.6) 60 (13.9) 0.43
CON (N = 186) 235 (63.2) 137 (36.8) 71 (38.2) 93 (50.0) 22 (11.8)
females
SCZ (N = 112) 125 (55.8) 99 (44.2) 0.25 30 (26.8) 65 (58.0) 17 (15.2) 0.20
CON (N = 81) 100 (61.7) 62 (38.3) 27 (33.3) 46 (56.8) 8 (9.9)
males
SCZ (N = 321) 402 (62.6) 240 (37.4) 0.68 124 (38.6) 154 (48.0) 43 (13.4) 0.66
CON (N = 105) 135 (64.3) 75 (35.7) 44 (41.9) 47 (44.8) 14 (13.3)
family history
sporadic SCZ (N = 369) 441 (59.8) 297 (40.2) 0.12 124 (33.6) 193 (52.3) 52 (14.1) 0.10
familial SCZ (N = 64) 86 (67.2) 42 (32.8) 30 (46.9) 26 (40.6) 8 (12.5)
SCZ: probands with schizophrenia; CON: controls; p-values for alleles (Fisher's exact test) and genotypes (Armitage's trend test).
Discussion
In independent case-control and proband-parent samples we attempted replication of a gender-specific transmission of allele rs175174 A to females at the ZDHHC8 gene locus. In contrast to the original report of Mukai and colleagues [3] we found moderate evidence at p-level 0.035 of a sex-related heterogeneity of allele transmission with preferential transmission of allele A to males and allele G to females. This observation remained the only signal for a genetic involvement of ZDHHC8 in schizophrenia from our samples. The preferential transmission of allele G corresponded partially to the report of Chen and colleagues [5] that allele G instead of A is preferentially transmitted to schizophrenic subjects, in opposite to the animal model and the initial genetic data from U.S. and Afrikaner pedigrees [3]. However, in the present study gender-specific effects were restricted to proband-parent triads, whereas one further study found non-gender related associations in both family-based and case-control samples [5] and two studies from different ethnic background were clearly negative in case-control [6] and both case-control and family-based designs [7]. Within European populations frequencies of allele G are consistent at 0.61 [7] and of 0.63 [present study], whereas in Asian populations allele G was the minor allele with 0.37 and 0.40 [5,6]. However, these differences hardly explain the contradictory genetic results, and assuming polygenic inheritance in schizophrenia, the negative findings could be only partially explained by sample stratification.
To replicate weak gender-specific effects at rs175174 the study sample was relatively small, but the case-control sample possesses a power of 80% to detect (at α = 0.05) an association with a susceptibility allele, under the assumption that the susceptibility allele has a population frequency of 0.63, and the effect of this allele is recessive with a relative risk of 2.5. Assuming a multiplicative model, the effect size of the female replication sample of Mukai et al. [3] is 1.6 [7]. The power against this alternative was 57% for our case-control sample of females.
The inconclusive results obtained in several genetic studies raise the question whether SNP rs175174 is itself the susceptibility allele or is in linkage disequilibrium (LD) with a further functional variant that increases risk for schizophrenia. Whereas Saito et al. [6] proposed that rs175174 is representative of ZDHHC8, differences in LD structure between the different populations may account for that the same disease causative variants yet to be identified are associated with different haplotypes. The functional variants could locate within ZDHHC8 or other genes around this locus at 22q11 and may independently or synergistically exert increased risk for schizophrenia. Therefore, the preferential transmission of individual alleles in at least a proportion of samples suggests the existence of additional functional variants in LD with SNP rs175174 or an interaction effect on the risk haplotype. In addition, the effect of the non-spliced product bearing a premature termination codon on cell homeostasis and function is still unknown [3]. On the basis of these inconsistent genetic findings on rs175174, the potential genetic diversity of the ZDHHC8 locus and the exact functional effect of related SNPs require further examination.
In conclusion, the findings at rs175174 at ZDHHC8 are still far from being conclusive, but evidence for a sexual dimorphism -if any- is weak. Given the large sample sizes studied together with the failure to replicate the initial findings in key respects, we agree with others [2] that the balance of evidence for ZDHHCH8 still favours the null hypothesis.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TF carried out the molecular genetic studies and drafting of the manuscript, MG, SJ performed laboratory assays, MB participated in the design of the study and its coordination, BP, BJ participated in the recruitment of the study sample, MK performed the data-analysis, interpretation of the data, and drafting of the manuscript, GS participated in the design and coordination of the study, interpretation of the data, and drafting of the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Karayiorgou M Gogos JA The molecular genetics of the 22q11-associated schizophrenia Brain Res Mol Brain Res 2004 132 95 104 15582150 10.1016/j.molbrainres.2004.09.029
Craddock N O'Donovan MC Owen MJ The genetics of schizophrenia and bipolar disorder: dissecting psychosis J Med Genet 2005 42 193 204 15744031 10.1136/jmg.2005.030718
Mukai J Liu H Burt R Swor DE Lai W-S Karayiorgou M Gogos JA Evidence that the gene encoding ZDHHC8 contributes to the risk of schizophrenia Nat Genet 2004 36 725 731 15184899 10.1038/ng1375
Liu H Abecasis GR Heath SC Knowles A Demars S Chen YJ Roos JL Rapoport JL Gogos JA Karayiorgou M Genetic variation in the 22q11 locus and susceptibility to schizophrenia Proc Natl Acad Sci USA 2002 99 16859 16864 12477929 10.1073/pnas.232186099
Chen WY Shi YY Zheng YL Zhao XZ Zhang GJ Chen SQ Yang PD He L Case-control study and transmission disequilibrium test provide consistent evidence for association between schizophrenia and genetic variation in the 22q11 gene ZDHHC8 Hum Mol Genet 2004 13 2991 2995 15489219 10.1093/hmg/ddh322
Saito S Ikeda M Iwata N Suzuki T Kitajima T Yamanouchi Y Kinoshita Y Takahashi N Inada T Ozaki N No association was found between a functional SNP in ZDHHC8 and schizophrenia in a Japanese case-control population Neurosci Lett 2004 374 21 24 15631889 10.1016/j.neulet.2004.10.015
Glaser B Schumacher J Williams HJ Jamra RA Ianakiev N Milev R Ohlraun S Schulze TG Czerski PM Hauser J Jonsson EG Sedvall GC Klopp N Illig T Becker T Propping P Williams NM Cichon S Kirov G Rietschel M Murphy KC O'Donovan MC Nothen MM Owen MJ No association between the putative functional ZDHHC8 single nucleotide polymorphism rs175174 and schizophrenia in large European samples Biol Psychiatry 2005 58 78 80 15992527 10.1016/j.biopsych.2005.03.017
American Psychiatric Association Diagnostic and statistical manual of mental disorders 4th ed text revision (DSM-IV-TR) 2000 Washington, DC: American Psychiatric Association
Leonhard K Classification of endogenous psychoses and their differentiated etiology 2nd rev and enlarged ed 1999 Wien, New York: Springer
Schaid DJ Sommer SS Genotype relative risks: methods for design and analysis of candidate-gene association studies Am J Hum Genet 1993 53 1114 26 8213835
Jackson MR Genin E Knapp M Escary JL Accurate power approximations for χ2-tests in case-control association studies of complex disease genes Ann Hum Genet 2002 66 307 321 12418971 10.1046/j.1469-1809.2002.00120.x
Weir BS Genetic data analysis II 1996 Sunderland, Massachusetts: Sinauer Associates
|
16225675
|
PMC1274335
|
CC BY
|
2021-01-04 16:33:02
|
no
|
BMC Psychiatry. 2005 Oct 14; 5:35
|
utf-8
|
BMC Psychiatry
| 2,005 |
10.1186/1471-244X-5-35
|
oa_comm
|
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-361622567710.1186/1471-244X-5-36Research ArticleSystematic mutation analysis of KIAA0767 and KIAA1646 in chromosome 22q-linked periodic catatonia Stöber Gerald [email protected] Bernd [email protected] Markus [email protected] Claudia [email protected] Micha [email protected]öller-Ehrlich Kerstin [email protected] Thomas [email protected] Thomas [email protected] Department of Psychiatry and Psychotherapy, University of Würzburg, Füchsleinstraße 15, 97080 Würzburg, Germany2 Department of Child and Youth Psychiatry and Psychotherapy, University of Würzburg, Füchsleinstraße 15, 97080 Würzburg, Germany3 Department of General, Vascular and Paediatric Surgery, University of the Saarland, Homburg/Saar 66421, Germany4 Institute of Human Genetics, Technical University of Munich & GSF Ingolstädter Landstr. 1, 85764 Neuherberg, Germany5 Max-Planck Institute of Psychiatry, Kraepelinstr. 2–10, 80804 Munich, Germany2005 14 10 2005 5 36 36 9 6 2005 14 10 2005 Copyright © 2005 Stöber et al; licensee BioMed Central Ltd.2005Stöber et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Periodic catatonia is a familial subtype of schizophrenia characterized by hyperkinetic and akinetic episodes, followed by a catatonic residual syndrome. The phenotype has been evaluated in two independent genome-wide linkage scans with evidence for a major locus on chromosome 15q15, and a second independent locus on chromosome 22qtel.
Methods
In the positional and brain-expressed candidate genes KIAA0767 and KIAA1646, we searched for variants in the complete exons and adjacent splice-junctions as well as in parts of the 5'- and 3'-untranslated regions by means of a systematic mutation screening in individuals from chromosome 22q-linked pedigrees.
Results
The mutation scan revealed 24 single nucleotide polymorphisms, among them two rare codon variants (KIAA0767: S159I; KIAA1646: V338G). However, both were neither found segregating with the disease in the respective pedigree nor found at a significant frequency in a case-control association sample.
Conclusion
Starting from linkage signals at chromosome22qtel in periodic catatonia, we screened two positional brain-expressed candidate genes for genetic variation. Our study excludes genetic variations in the coding and putative promoter regions of KIAA0767 and KIAA1646 as causative factors for periodic catatonia.
==== Body
Background
The phenotype of periodic catatonia is characterized by hyperkinetic and akinetic episodes with parakinetic movements in a bipolar course, accompanied by delusional or hallucinatory symptoms, and followed by residual states with psychomotor features [1,2]. The estimated lifetime prevalence of periodic catatonia is ~0.001 in the general population. Evidence for significant linkage to chromosome 15q15 was obtained and replicated in two independent genome-wide linkage scans [3,4]. Mainly supported by a single four-generation pedigree, a second locus was identified on chromosome 22q with a maximum multipoint LOD score of 2.59 (θ = 0.0) under an autosomal dominant model at marker D22S1169, and with a heterogeneity Zmax of 1.57 with an estimated 38% of families linked [3]. Preliminary findings had suggested a link between MLC1 and catatonia via a putative dominantly acting missense mutation cosegregating in a large pedigree, but further analyses have excluded sequence variants of MLC1 as causing chromosome 22q-linked catatonia. MLC1 is the disease causing gene for autosomal recessively inherited megalencephalic leukoencephalopathy with subcortical cysts [5-7]. A systematic mutation screening in 140 index cases with periodic catatonia and five cases with MLC detected a high degree of sequence diversity of MLC1 with evidence for further allelic heterogeneity of MLC1 mutations in MLC, but unfortunately the study failed to validate an association of schizophrenia to genetic variants of MLC1. Among periodic catatonia index cases, the mutation scan revealed 15 different single nucleotide polymorphisms, among them three coding variants: two of them were observed in controls at a significant frequency, and the L309M variant, that was previously supposed to be the causative factor for chromosome 22qtel linked periodic catatonia, was found non-segregating in a further multiplex pedigree [6]. In addition, MLC1 is a 377 amino acid protein with preferential expression in the brain and peripheral white blood cells. In the brain, MLC1 is specifically expressed in distal astroglial processes in perivascular, subependymal, and subpial regions. Although MLC1 shares low homology with human voltage-gated potassium channels, addressing the membrane topology and cellular localization of MLC1 supports the possible transport function of MLC1 for a specific, yet unknown substrate [8,9]. Thus, mutations in MLC1 are causative for MLC, but can be excluded as a susceptibility factor in schizophrenia. Reports of other groups also failed to support an aetiological relevance of this gene in schizophrenia [10-12] and more importantly, these studies [6,12] rule out that spongiform leukodystrophies and subtypes of schizophrenia are allelic disorders.
Under the assumption of disease heterogeneity and allelic heterogeneity [13], we screened in a systematic approach the positional candidate genes KIAA1646 and KIAA0767, located between 45.34–45.45 Mb [14], for genetic variation. Human KIAA0767 is a mitochondrial protein of 578 amino acids (aa) with high expression in adult brain and strong pro-apoptotic effect [15,16]. KIAA1646 is a cytoplasmatic and membrane-associated ceramide kinase (CERK) of 537 amino acids [17]. CERK is acting in the signal transduction cascade and is suggested to be involved in the process of synaptic vehicle fusion [18]. KIAA0767 and KIAA1646 consist of 18 and 13 exons, respectively, spanning each ~50 kb on genomic DNA.
Methods
We selected two affected individuals (933, 1045) from different branches of F20, and two cases (727, 857) of smaller pedigrees F15 and F17, which were compatible with linkage to chromosome 22q [19], as well as DNA of a healthy individual as control. In the association study we included 115 unrelated cases with periodic catatonia (66 males; mean age first hospitalisation: 26.2 years, SD 10.5; age at assessment: 44.6 years, SD 17.1) and 110 blood donors as controls (60 males; mean age at assessment: 29.5 years, SD 9.4). All subjects actively participated in the study after giving informed consent. The Ethics Committee of the University of Würzburg had approved the study.
Primers covered in overlapping fragments parts of the 5'-UTR containing putative promoter regions [20] and were allocated in intronic regions to encompass the complete exon and adjacent splice-junctions as well as parts of the 3'-UTR up to ~1.0 kb. PCR (30 sec at 94°C, 30 sec at 57°/60°C, and 30 sec at 72°C for 32 cycles) was carried out in 25 μl reaction volumes containing ~80 ng genomic DNA, 20 pmol of each primer, 200 μM of each dNTP, 0.5 U Taq DNA polymerase (Fermentas), and buffer as supplied by the manufacturer in a MJ Research 96-well block Tetrad thermocycler (Waltham, MA). Exon 1 of CERK was not sequenced because of technical difficulties. PCR products were purified by solid phase extraction and bidirectionally sequenced with ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) followed by computer-assisted analyses. For restriction fragment analyses (RFLP) PCR products were amplified on a Thermocycler (Biometra, Göttingen), and subsequently digested with the appropriate enzyme (CERK V338G: DraIII; KIAA0767 S159I: MboI). Fragments were resolved on a 10% PAA gel containing 1.0 × TBE at 15 V/cm2 followed by silverstaining.
Results
In the mutation screening sample we identified seven single nucleotide polymorphisms (SNPs) at the KIAA0767/DIP locus (Table 1). Nt 45379466 G>T changes a serine to isoleucine (S159I). We identified case 857 as a heterozygote carrier; the variant, however, did not co-segregate with the disease in the multiplex pedigree. The SNP was introduced through an unaffected individual, who married into the family. In addition, we did not observe 159I in a sample of 450 chromosomes (115 index cases with periodic catatonia and 110 controls, respectively). These findings coincidently indicated that 159I is a rare variant, but not associated with disease susceptibility.
Table 1 sequence variable at the KIAA0767/DIP locus on chromosome 22q13. 3
Nucleotide position * DNA level Nucleotide change Codon SNP-database Genotypes of the individuals from each family
933 1045 727 857 Ctrl Allele frequency
45377716 Intron 3 C>T - rs 2076708 CC CC CC CC CT C: 0.9 T: 0.1
45379466 Exon 5 G>T S1591 - GG GG GG GT GG G: 0.9 T: 0.1
45385184 Intron 11 C>T - - CT CC CC CC CT C: 0.8 T: 0.2
45385356 Intron 12 A>C - rs 2076711 AA AA AA AA AC A: 0.9 C: 0.1
45391987 Intron 16 C>T - - CC CC CC CC CT C: 0.9 T: 0.1
45392051 Intron 16 T>C - - TT TT TC TC TC T: 0.7 C: 0.3
45392996 Intron 16 G>A - rs 2236028 GA GG GG GG GA G: 0.8 A: 0.2
* nt position according the UCSC Genome Browser May 2004 assembly [14; NM\_022766].
Nucleotide changes found by automated sequencing of amplicons of four individuals from pedigrees segregating periodic catatonia evaluated in a genome-wide linkage scan [3, 19]: 933 and 1045 (F20), 727 and 857 (F15,F17), and a control subject.
The systematic scan of KIAA1646/CERK resulted in a total of 17 SNPs (Table 2). We observed the coding variant 338G by sequencing DNA-amplicons of a control subject. In a subsequent case-control association study, we found a frequency of the heterozygous genotype of 0.9% (2 out of 220 alleles) compared to a frequency of the heterozygous genotype of 2.2% in periodic catatonia (5 out of 230 alleles; ns). In addition, the variants C50 and D377 were both non-segregating within the multiplex pedigree F20, as did the synonymous N500 in the respective pedigrees.
Table 2 Polymorphisms at the KIAA1646/CERK locus in periodic catatonia
Nucleotide position* DNA level Nucleotide change Codon SNP-database Genotypes of the individuals from each family
933 1045 727 857 Ctrl Allele frequency
45437424 Exon 2 G>A C50 rs 12166204 GG GA GG GG GG G: 0.9; A: 0.1
45418024 Intron 7 C>T - rs 16995595 GT CC CC CC CC C: 0.9; T: 0.1
45415928 Intron 7 C>G - - CG GG CG CC CG C: 0.5; G: 0.5
45415691 Intron 8 C>G - rs 5767329 CC CC CG CG CC C: 0.8; G: 0.2
45411750 Intron 8 G>A - rs 9616098 GG GG GG GG GA G: 0.9; A: 0.1
45411662 Exon 9 A>C V338G - AA AA AA AA AC A: 0.9; C: 0.1
45409793 Intron 10 G>A - - GG GG GA GA GG G: 0.8; A: 0.2
45408189 Exon 11 G>A D377 - GA GG GG GG GA G: 0.8; A: 0.2
45406443 Exon 12 G>A D502 - GG GG GA GA GG G: 0.8; A: 0.2
45403531 3'- UTR T>A - - TA TT TA TA AA T: 0.5; A: 0.5
45403529 3'-UTR T>C - rs 2542014 TC TT TC TC CC T: 0.5; C: 0.5
45403096 3'-UTR G>T - rs 8143065 GG GG GT GT GG G: 0.8; T: 0.2
45402864 3'-UTR C>T - rs 801719 CC CC CT CT CT C: 0.7; T: 0.3
45402678 3'-UTR C>A - rs 801720 CC CC CA CA AA C: 0.6; A: 0.4
45402542 3'-UTR G>A - rs 3747258 GG GG GA GA GG G: 0.8; A: 0.2
45402502 3'-UTR C>T - - CC CC CT CT CC C: 0.8; T: 0.2
45402374 3'-UTR A>G - rs 2748348 AA AA AG AG AA A: 0.8; G: 0.2
* nt position according the UCSC Genome Browser May 2004 assembly [14; NM\_022766].
Eleven of the 24 SNPs had not yet been deposited in current databases. At the CERK locus rs 2542014 was linked to a nearby (+2 bp) variant at nt 45403531. The intronic SNPs at nt 45415691C>G (KIAA1646) and nt 45392050 T>C (KIAA0767) were found at allelic frequencies useful for further LD-mapping studies.
Discussion
Sustained interest in schizophrenia susceptibility on chromosome 22q is substantiated by several genome-wide linkage studies on schizophrenic psychoses [21-23]. Because of reports of schizophrenia susceptibility genes within the 22q11 region, this locus has obtained a high marker density compared to the 22q13 region [24]. Although only weak signals for linkage to chromosome 22q11 were received by recent multicenter and meta-analytic studies [25,26], subsets of pedigrees in the study by Mowry et al. [25] gave positive scores for chromosome 22q13, particularly if accounting for intersample heterogeneity. The pedigrees analyzed here were not part of these multicenter samples. We focused on the phenotype periodic catatonia [[27]; #605419] with a major disease locus at chromosome 15q15, and at least one further independent locus at chromosome 22q13, mainly supported by a large four-generational pedigree [3]. As it has been already demonstrated in several chromosome 15q15-linked pedigrees, haplotype analyses in pedigree F20 showed autosomal dominant transmission [3,6]. The lack of common haplotypes in unrelated families, however, added further evidence to genetic and allelic heterogeneity in periodic catatonia.
The chromosome 22qtel candidate locus of ~4 Mb comprises more than 45 genes [28], and harbours several genes involved in severe neuropsychiatric disorders, such as metachromatic leukodystrophy and megalencephalic leukoencephalopathy [[27]: #250100; #604004]. In a systematic mutation scan of positional candidates at the chromosome 22qtel candidate region we screened ~3400 nt of coding sequence including splice-donor sites and additionally parts of the 5'-UTR and 3'-UTR of KIAA0767 and KIAA1646, respectively. The mutation scan revealed 24 sequence variants, among them two rare codon variants (KIAA0767: S159I; KIAA1646: V338G). However, both neither were found to segregate with the disease in the respective pedigrees nor found at a significant frequency in periodic catatonia.
Human KIAA0767/DIP [GenBank: NM_015124; LocusLink: 23151] consists of 18 exons, spanning ~50 kb on genomic DNA [NT_011523.2]. KIAA0767 is a protein of 578 amino acids (aa) putatively localized in the mitochondrion [16]. KIAA0767 expression resulted in a significant loss of cell viability independent of the p53 status, and thus, termed death-inducing-protein (DIP) because of its strong pro-apoptotic effect [16]. DIP shows at least two transmembrane domains and multiple putative phosphorylation and glycolisation sites. The variant S159I, however, seems not to affect any of the various functional DIP/KIAA0767 protein domains.
Human KIAA1646/CERK [GenBank: NM_022766; LocusLink: 64781] consists of 13 exons, spanning ~50 kb on genomic DNA. CERK is a cytoplasmatic and membrane-associated protein of 537 amino acids with a calculated molecular weight of 60 kDa. It is a member of a new class of lipid kinases and acts in the signal transduction cascade of the control of apoptosis catalyzing specifically the phosphorylation of ceramide [17]. The component ceramide is thought to regulate apoptotic responses to stress, particularly those initiated by the mitochondria. CERK is highly expressed in brain and leukocytes, and is suggested to be involved in synaptic neurotransmitter release [18] or in phagocytosis and vehicle fusion in neutrophils and mast cells [17,29,30]. In CERK, the pleckstrin homology domain and the diacylglycerol kinase (DGK) catalytic domain were found non-polymorphic, but V338G was located in the central homologous region, near the casein kinase II phosphorylation site at S340, and V338 is conserved between human and mouse CERK [17,31]. Although V338G may be of functional relevance, it was not found associated with periodic catatonia. Recent reports that mutations in the related gene CERKL on chromosome 2q31 cause autosomal recessive retinitis pigmentosa (RP26) indicate a link of the ceramide kinase gene family to retinal neurodegeneration [32].
Starting from linkage signals at chromosome22qtel in periodic catatonia, we screened two positional brain-expressed candidate genes for genetic variation (KIAA0767/DIP; KIAA1646/CERK). Our study excludes variants at coding and putative promoter regions as causative factors in periodic catatonia, but do not exclude the involvement of other regulatory elements in intronic or extended promoter regions. Although negative, the present study narrowed down the putative susceptibility region, and provides a systematic SNP generation for forthcoming LD studies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BK, MS, CR performed laboratory assays, MG, KME performed the data-analysis and drafted the manuscript, TM, TB participated in the design of the study and its coordination, GS participated in the design of the study, interpretation of the data, and drafting of the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by the Deutsche Forschungsgemeinschaft.
==== Refs
Leonhard K Classification of endogenous psychoses and their differentiated etiology 1999 2nd rev. and enlarged Wien, New York: Springer 104 112 301–308
Stöber G Caroff SN, Mann SC, Francis A, Fricchione GL Genetics Catatonia From psychopathoplogy to neurobiology 2004 Washington DC, London: American Psychiatric Press 173 187
Stöber G Saar K Rüschendorf F Meyer J Nürnberg G Jatzke S Franzek E Reis A Lesch KP Wienker TF Beckmann H Splitting schizophrenia: periodic catatonia-susceptibility locus on chromosome 15q15 Am J Hum Genet 2000 67 1201 1207 11001582
Stöber G Seelow D Rüschendorf F Ekici A Beckmann H Reis A Periodic catatonia: confirmation of linkage to chromosome 15 and further evidence for genetic heterogeneity Hum Genet 2002 111 323 330 12384773 10.1007/s00439-002-0805-4
Leegwater PAJ Yuan BQ van der Stehen J Mulders J Könst AAM Boor PKI Mejaski-Bosnjak V van der Maarel SM Frants RR Oudejans CBM Schutgens RBH Pronk JC van der Knaap MS Mutations of MLC1 (KIAA0027), encoding a putative membrane protein, cause megalencephalic leucoencephalopathy with subcortical cysts Am J Hum Genet 2001 68 831 838 11254442 10.1086/319519
Rubie C Lichtner P Gärtner J Siekiera M Uzig G Kohlmann B Kohlschütter A Strom T Meitinger T Stöber G Bettecken T Sequence diversity of KIAA0027/MLC1: are megalencephalic leukoencephalopathy and schizophrenia allelic disorders? Hum Mutat 2003 21 45 52 12497630 10.1002/humu.10145
Meyer J Huberth A Ortega G Syagailo YV Jatzke S Mössner R Strom T Ulzheimer-Teuber I Stöber G Schmitt A Lesch KP A missense mutation in a novel gene encoding a putative cation channel is associated with catatonic schizophrenia in a large pedigree Mol Psychiatry 2001 6 304 308 10.1038/sj.mp.4000869
Kaganovich M Peretz A Ritsner M Bening Abu-Shach U Attali B Navon R Is the WKL1 gene associated with schizophrenia? Am J Med Genet Neuropsychiatr Genet 2004 125 31 37 10.1002/ajmg.b.20115
Boor PK de Groot K Waisfisz Q Kamphorst W Oudejans CB Powers JM Pronk JC Scheper GC van der Knaap MS MLC1: a novel protein in distal astroglial processes J Neuropathol Exp Neurol 2005 64 412 419 15892299
Devaney JM Donarum EA Brown KM Meyer J Stöber G Lesch KP Nestadt G Stephan DA Pulver AE No missense mutation of WKL1 in a subgroup of probands with schizophrenia Mol Psychiatry 2002 7 419 423 11986987 10.1038/sj.mp.4001022
McQuillin A Kalsi G Moorey H Lamb G Mayet S Quested D Baker P Curtis D Gurling HM A novel polymorphism in exon 11 of the WKL1 gene, shows no association with schizophrenia Eur J Hum Genet 2002 10 491 494 12111645 10.1038/sj.ejhg.5200837
Leegwater PA Boor PK Pronk JC van der Knaap MS Association of WKL1/MLC1 with catatonic schizophrenia Mol Psychiatry 2002 7 1037 12476316 10.1038/sj.mp.4001142
Pritchard JK Cox NJ The allelic architecture of human disease genes: common disease-common variant...or not? Hum Mol Genet 2002 11 2417 2423 12351577 10.1093/hmg/11.20.2417
Kikuno R Nagase T Nakayama M Koga H Okazaki N Nakajima D Ohara O HUGE: a database for human KIAA proteins, a 2004 update integrating HUGEppi and ROUGE Nucleic Acids Res 2004 32 502 504 14742665 10.1093/nar/gkh035
Stanelle J Tu-Rapp H Pützer BM A novel mitochondrial protein DIP mediates E2F1-induced apoptosis independently of p53 Cell Death Differ 2005 12 347 357 15565177 10.1038/sj.cdd.4401532
Sugiura M Kono K Liu H Shimizugawa T Minekura H Spiegel S Kohama T Ceramide kinase, a novel lipid kinase. Molecular cloning and functional characterization J Biol Chem 2002 277 23294 23300 11956206 10.1074/jbc.M201535200
Bajjalieh SM Martin TFJ Floor E A synaptic vesicle ceramide kinase. A calcium-stimulated lipid kinase that co-purifies with brain synaptic vesicles J Biol Chem 1989 264 14354 14360 2547795
Stöber G Pfuhlmann B Nürnberg G Schmidtke A Reis A Franzek E Wienker TF Towards the genetic basis of periodic catatonia: pedigree sample for genome scan I and II Eur Arch Psychiatry Clin Neurosci 2001 251 25 30
Coon HS Jensen M Holik J Hoff M Myles-Worsley M Reimherr F Wender P Leppert M Byerley W Genomic scan for genes predisposing to schizophrenia Am J Med Genet 1994 54 59 71 7909992 10.1002/ajmg.1320540111
Blouin JL Dombroski BA Nath SK Lasseter VK Wolyniec PS Nestadt G Thornquist M Ullrich G McGrath J Kasch L Lamacz M Thomas MG Gehrig C Radhakrishna U Snyder SE Balk KG Neufeld K Swartz KL DeMarchi N Papadimitriou GN Dikeos DG Stefanis CN Chakravarti A Childs B Housman DE Kazazian HH Antonarakis S Pulver AE Schizophrenia susceptibility loci on chromosomes 13q32 and 8p21 Nat Genet 1998 20 70 73 9731535 10.1038/1734
DeLisi LE Shaw SH Crow TJ Shields G Smith AB Larach VW Wellman N Loftus J Nanthakumar B Razi K Stewart J Comazzi M Vita A Heffner T Sherrington R A genome-wide scan for linkage to chromosomal regions in 382 sibling pairs with schizophrenia or schizoaffective disorder Am J Psychiatry 2002 159 803 812 11986135 10.1176/appi.ajp.159.5.803
Harrison PJ Owen MJ Genes for schizophrenia? Recent findings and their pathophysiological implications Lancet 2003 361 417 419 12573388 10.1016/S0140-6736(03)12379-3
Mowry BJ Holmans PA Pulver AE Gejman PV Riley B Williams NM Laurent C Schwab SG Wildenauer DB Bauche S Owen MJ Wormley B Sanders AR Nestadt G Liang KY Duan J Ribble R Norton N Soubigou S Maier W Ewen-White KR DeMarchi N Carpenter B Walsh D Williams H Jay M Albus M Nertney DA Papadimitriou G O'Neill A O'Donovan MC Deleuze JF Lerer FB Dikeos D Kendler KS Mallet J Silverman JM Crowe RR Levinson DF Multicenter linkage study of schizophrenia loci on chromosome 22q Mol Psychiatry 2004 9 784 795 15007391 10.1038/sj.mp.4001481
Lewis CM Levinson DF Wise LH DeLisi LE Straub RE Hovatta I Williams NM Schwab SG Pulver AE Faraone SV Brzustowicz LM Kaufmann CA Garver DL Gurling HM Lindholm E Coon H Moises HW Byerley W Shaw SH Mesen A Sherrington R O'Neill FA Walsh D Kendler KS Ekelund J Paunio T Lonnqvist J Peltonen L O'Donovan MC Owen MJ Wildenauer DB Maier W Nestadt G Blouin JL Antonarakis SE Mowry BJ Silverman JM Crowe RR Cloninger CR Tsuang MT Malaspina D Harkavy-Friedman JM Svrakic DM Bassett AS Holcomb J Kalsi G McQuillin A Brynjolfson J Sigmundsson T Petursson H Jazin E Zoega T Helgason T Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: Schizophrenia Am J Hum Genet 2003 73 34 48 12802786 10.1086/376549
Online Mendelian Inheritance in Man (OMIM)
Hinkovska-Galcheva V Kjeldsen L Mansfield PJ Boxer LA Shayman JA Suchard SJ Activation of a plasma membrane-associated neutral sphingomyelinase and concomitant ceramide accumulation during IgG-dependent phagocytosis in human polymorphonuclear leukocytes Blood 1998 91 4761 4769 9616175
Mitsutake S Kim TJ Inagaki Y Kato M Yamashita T Igarashi Y Ceramide kinase is a mediator of calcium-dependent degranulation in mast cells J Biol Chem 2004 279 17570 17577 14769792 10.1074/jbc.M312885200
Bornancin F Mechtcheriakova D Stora S Graf C Wlachos A Devay P Urtz N Baumruker T Billich A Characterization of a ceramide kinase-like protein Biochim Biophys Acta 2005 1687 31 43 15708351
Tuson M Marfany G Gonzalez-Duarte R Mutation of CERKL, a novel human ceramide kinase gene, causes autosomal recessive retinitis pigmentosa (RP26) Am J Hum Genet 2004 74 128 138 14681825 10.1086/381055
|
16225677
|
PMC1274336
|
CC BY
|
2021-01-04 16:33:02
|
no
|
BMC Psychiatry. 2005 Oct 14; 5:36
|
utf-8
|
BMC Psychiatry
| 2,005 |
10.1186/1471-244X-5-36
|
oa_comm
|
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-571620737010.1186/1475-925X-4-57EditorialIn Memorium: Swamy Laxminarayan [1939–2005] Foster Kenneth R [email protected] Luis G [email protected] Department of Bioengineering, University of Pennsylvania, 220 South 33rd Street, Philadelphia, PA 19104, USA2 IRM College, National Defense University, Fort McNair, Washington, DC 20319, USA2005 5 10 2005 4 57 57 4 10 2005 5 10 2005 Copyright © 2005 Foster and Kun; licensee BioMed Central Ltd.2005Foster and Kun; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Swamy Narasimha Laxminarayan, known to his many friends and colleagues as Swamy, passed away on September 29, 2005. He was one of the most prominent biomedical engineers on the international scene, and contributed immensely to the globalization of this new field.
==== Body
Swamy grew up in Bangalore, India. Records list his birthdate as May 24, 1939, although he has reportedly confessed to colleagues that he misrepresented his age to gain early admission to university; his real birthdate may be as late as 1943. He received his undergraduate degree in physics and mathematics from the University of Mysore, India (1957) and Masters (1967) and Ph.D. (1972) from the University of Southampton in the UK. His Ph.D. thesis concerned on-line signals processing and physiological systems, interests that he maintained through his entire professional career.
Beginning in 1970, Swamy held research faculty positions at two universities in the Netherlands (1970–78). A large part of his career was spent at the University of Medicine and Dentistry with an adjunct appointment at the New Jersey Institute of Technology, both in Newark, New Jersey, where he led programs in research computing and health care informatics. He joined Idaho State University in 2002, where he was a Research Associate Professor at the university and chief of Biomedical Information Engineering at Telehealth Idaho. He had numerous other affiliations along the way, including several adjunct clinical appointments, and visiting or honorary appointments at the Technical University of Brno (Czech Republic) and Tsinghua University (China), as well as several affiliations with private industry.
Swamy was a prolific researcher. His resume lists more than 300 technical papers in fields as diverse as biomedical engineering, medical image processing, telehealth, bioterrorism and homeland security. These include numerous original papers, together with a great many conference proceedings and special issues that he edited. Most recently, he was awarded a $3.8 M grant for developing training materials to combat bioterrorism.
Swamy's contributions to engineering are varied and prominent. Many were connected with his enthusiastic leadership in engineering societies and editorial work with professional journals. He was the Founding Editor-in-Chief of the IEEE Transactions on Information Technology in Biomedicine, and served on the editorial boards or as associate editor of numerous other biomedical engineering journals. He had many leadership roles in the Institute of Electronics and Electrical Engineers (IEEE). Most of his work over the years was with the IEEE Engineering in Medicine and Biology Society, but he also held leadership roles in the IEEE Computer and Communications Societies, and in the IEEE at large. He was also active in a host of other societies in medical informatics, telemedicine, medical instrumentation, and related fields, where he worked tirelessly to organize projects related to biomedicine, and to promote individuals' careers and professional recognition.
Swamy's immense contribution came from his ability to move people and projects along, propelled by his incredible enthusiasm and assisted by a vast network of friends around the world. For many years, he was a prominent figure in organizing international meetings in biomedical engineering, and he was exceedingly effective in bringing scientists and engineers from around the world together at these conferences. He was a longtime leader in the International Federation for Medical and Biological Engineering. Biomedical engineering, which began largely as an American enterprise, now has a global scope, thanks in no small part to Swamy's work over the years.
Swamy received numerous awards, including the Purkinje award (Czech Academy of Medical Societies, 1994) for 'pioneering contributions in the field of advanced computer applications to cardiovascular, neuro and pulmonary physiology and for international leadership in information technology in medicine and healthcare', the IEEE 3rd Millennium Medal and career achievement awards, and fellowship in the American Institute of Medical and Biological Engineering. This past spring, Swamy was honored by Idaho State University as a Distinguished Researcher.
A dedicated family man, Swamy is survived by his wife Marijke, daughter Malini, and son, Vinod. He will also be very much missed by his many friends and colleagues throughout the world.
Figure 1 Swamy Laxminarayan [1939–2005]
|
0
|
PMC1274337
|
CC BY
|
2021-01-04 16:37:37
|
no
|
Biomed Eng Online. 2005 Oct 5; 4:57
|
utf-8
|
Biomed Eng Online
| 2,005 |
10.1186/1475-925X-4-57
|
oa_comm
|
==== Front
Clin Pract Epidemiol Ment HealthClinical Practice and Epidemiology in Mental Health : CP & EMH1745-0179BioMed Central 1745-0179-1-191620216310.1186/1745-0179-1-19ReviewHeart failure and health related quality of life Coelho Rui [email protected] Sónia [email protected] Joana [email protected] Paulo [email protected] António [email protected] Mário [email protected] Department of Psychiatry, Hospital de S. João, Porto Medical School, Porto, Portugal2 Department of Psychiatry, Hospital de Magalhães Lemos, Porto, Portugal3 Department of Psychiatry, Centro Hospitalar de Vila Nova de Gaia, Vila Nova de Gaia, Portugal4 Department of Internal Medicine, Hospital de S. João, Porto, Portugal5 Unit of Cardiovascular Research and Development. Hospital de S. João, Porto Medical School, Porto, Portugal2005 4 10 2005 1 19 19 16 9 2005 4 10 2005 Copyright ©2005 Coelho et al; licensee BioMed Central Ltd.2005Coelho et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Quality of life is a major goal in the context of preventive and therapeutic cardiology. It is important, both as an outcome measure in clinical trials of congestive heart failure (CHF) and as a consideration in individual physicians' therapeutic decisions. In this article, quality of life concepts are reviewed, methods of measurement are explored and clinically significant changes on prognosis are discussed. There is a need for more research which is based on carefully selected measures of quality of life chosen as being of particular importance to patients and to the hypotheses being tested.
Congestive Heart failureQuality of lifePrognosis
==== Body
1. Introduction
Congestive heart failure (CHF) is a major health problem, with an increasing incidence and a gloomy prognosis, that is often accompanied by restricted physical activity and severe complaints in several areas of quality of life [1]. Despite its high prevalence, there have been few studies of the impact of CHF on quality of life, especially in the community and in the elderly patient [2], suggesting the need of epidemiology of quality of life as a part of epidemiology of "positive mental health" strictly closed with mental health. According to Stott, to the population of ageing, elderly heart failure patients who have multiple co-morbidities, quality of life, disability and cognitive impairment are the elements which matter to measure [3].
Since 1948, when the World Health Organization defined health as being not only the absence of disease and infirmity but also the presence of physical, mental and social well-being, quality of life issues have become steadily more important in health care practice and research [4].
2. Definition of quality of life
The concept of quality of life is not yet defined in a uniform way, lacks clarity and even creates confusion. It seems that in medicine, the term has become a bandwagon concept for all those human needs which are often neglected in a health care field increasingly dominated by technology. It is justifiable to say that it is a term describing a field of interest rather than a single variable [5].
Health status, functional status and quality of life are three concepts often used interchangeably to refer to the same domain of "health". Guyatt used the term "health related quality of life" (HRQL) because many widely valued aspects of life are not generally considered as "health-related" include income, freedom and quality of the environment [6].
As a rule, "quality of life" is used in medicine for characterising an individual patient's quality of life from his or her own subjective perspective [5]. HRQL, to Schipper and associates, can be defined as "... the functional effect of an illness and its consequent therapy upon a patient, as perceived by the patient" [6]. It has been found logical to distinguish sharply between diseases, which explain illness behaviour, and other states of health, which do not have an explanatory element but might be seen as a consequence of having one or more illnesses [7]. So, HRQL measures the illness experience as opposed to the disease [6], it defines the patient reality, his or her point of view as opposed to the reality defined by professional medical knowledge [8].
3. Measuring the quality of life
Measures of the quality of life in chronically ill patients provide an important source of medical information in addition to laboratory or diagnostic tests [8] and are becoming increasingly relevant to controlled clinical trials [6]. One goal of the measurement of quality of life is to have objective evaluations of how and how much the disease influences patients'life and how patients cope with it. These evaluations may be useful as a baseline and outcome measures and should provide framework to determine the impact of any change on patients'quality of life [9].
In the first phase of quality of life research in the 1970s and early 1980s, already available psychological well-being scales were used or new ones were specifically developed for this purpose. Examples are the "Affect Balance Scale" (ABS) by Bradburn (1969); the "Quality of Well-Being Scale" (QWBS) by Kaplan et al (1976) and the "Psychological General Well-Being Index" (PGWB) by Dupuy (1984). This particular development has connections to the "happiness research" tradition within psychology, where well-being is discussed not only in terms of absence of negative factors (such as depressed mood), but as a positive concept.
From the 1980s onwards, in addition to the assessment of well-being and satisfaction, instruments for assessing functioning in daily life were developed. This development is subsumed under the term "health status research". Well- known examples of "health status research" instruments are the Sickness Impact Profile (SIP) by Bergner et al (1981); the Nottingham Health Profile (NHP) by Hunt and McEwen (1980) and the Rand SF-36 Health Status Profile by Ware and Sherbourne (1992). Although these instruments do not use the term "quality of life", studies employing them are today generally regarded as belonging to health- related quality of life research. Later – in contrast to these "generic instruments" – disease specific quality of life instruments were developed [5].
General (or generic) measures attempt to provide a summary of quality of life and they can be standardized and applied widely to those with different types of illness to enable comparisons [10], but they lack the range, sensitivity and flexibility to deal with the special problems of particular illness.
Specific measures focus on problems associated with individual disease states, patients groups or areas of function [6]. Specific instruments tend to be more responsive to changes and more sensitive in discriminating the range of impairment because of their focus on the most relevant aspects of quality of life for the problems assessed. Nevertheless, using a disease specific quality of life instrument alone can miss important aspects of the impact of a disease in quality of life [10-12]. There is a need for more research, which is based on carefully selected specific measures of quality of life chosen as being of particular importance to patients and to the hypotheses being tested [13]. Some examples of specific instruments to measure quality of life in heart failure patients are: the Chronic Heart Failure Questionnaire; the Minnesota Living with Heart Failure Questionnaire; the Yale Scale; the Quality of Life Questionnaire in Severe Heart Failure and the Kansas City Cardiomyopathy Questionnaire, etc.
As there are benefits with each type of instrument, it is recommended often that both (generic and specific instruments) should be used [10,14]. However, a critical analysis of the properties of the growing range of generic and disease specific measures is necessary in order to guide and direct researchers and clinicians towards the most appropriate measures in terms of reliability, validity and sensitivity to change [10].
Instruments used in measuring quality of life must be valid (if it is really measuring what is supposed to measure); reliable (if it gives the same measurement after repeated administration in stable patients); sensitive (if it is able to reflect clinically meaningful differences in quality of life across the broad spectrum of the clinical conditions) and responsive (if it detects changes when the patients' conditions change) [11].
4. Domains of quality of life
Domains of quality of life refer to areas of behaviour that are measured.
The subjective domains of quality of life are: physical functioning (the capacity to perform physical tasks); occupational functioning (quality of life should focus on the ability to perform multiple essential roles and not just on return to work); perceptions about health status (health perceptions are personal beliefs and evaluations of general health status, they are the result of integration of information and feelings about health and health limitations from the self, the medical system, the family and the society); psychological functioning and social functioning. Siegrist and Junge (see Swenson and Clinch) [11] defined social health as the dimension of an individual's well-being that concerns how the individual gets along with others, how other people react to him or her, and how the person interacts with social institutions and norms.
In the domain of objectivity we can include: health status (measured by laboratory or diagnostic tests); psycophatology (CID-10/DSM-IV-TR); socio-economic status and social support (number and quality of the contacts).
Thompson et al (see Majani et al, 1999) stated that "objective measures of quality of life often bear little relationship to life satisfaction, whereas subjective indicators are often found to correlate highly with a global sense of well-being, as well as being more meaningful and sensitive barometers of quality of life". Accordingly, patients'subjective satisfaction should always be included in routine assessment and clinical interventions; they are a useful source of information on patient distress and psychological resources [15].
5. Assessing quality of life among people with CHF
Advanced CHF is the final outcome of many cardiovascular disorders. Despite recent improvement in survival related to newer therapies, CHF remains a condition with such a generally poor prognosis that the critical therapeutic advantages are more likely to be those that maintain and stabilize the patient's limited functional abilities and, by ameliorating symptoms, improve the comfort of the patient for the remaining duration of life. Prolongation of survival without these benefits may be viewed as a less important objective [16,17]. According to Coats et al optimally treated patients are often left with unrelieved symptoms and have a correspondingly low quality of life [18]. The development in the recent years of standardised measures of quality of life in CHF reflects the growing perception of the importance of these outcomes in patients [19]. The achievement of quality of life measures in CHF can be summarised as follows: general profiles of populations with CHF in terms of quality of life have been developed; the validity of the measuring instruments has been tested and correlated with clinical outcome; the need to include quality of life studies as outcome measures in clinical trials such as SOLVD has been recognised [9].
5.1 Quality of life in patients with CHF: comparison with other chronic diseases
Juenger et al compared the quality of life of patients with CHF with a previously characterised general population and with those with other chronic diseases. Patients'self assessment of quality of life was measured by the SF-36. The authors concluded that the total CHF sample was characterised by significantly reduced scores in all aspects of quality of life compared with a healthy reference group. Patients with CHF showed the same pattern of reduced quality of life as patients on chronic haemodialysis, thus it could be argued that all chronic disease conditions have a similar impact on quality of life, however patients with chronic hepatitis C had higher scores in physical functioning, role functioning, physical and general health than the CHF population. In comparison with patients with major depression from the Medical Outcome Study, the total CHF sample was characterised by significantly worse physical health (as expected) and better mental health. However, patients with more advanced CHF (NYHA III) had similar scores to patients with major depression on the mental health scales in addition to their already dramatically reduced physical health. These data are in accord with the findings of some recent studies showing that a large proportion of patients with CHF suffer from depression. [20]
5.2 Quality of life, medical treatment and clinical outcome
As has occurred with other chronic medical problems for which several therapies have produced comparable survival benefit, additional attributes or consequences of these therapies are then considered to warrant evaluation, among them the effects on quality of life, as a means to select an optimal drug regimen. The emergence of quality of life as an outcome measure in clinical trials of cardiovascular, as well as other therapies appears associated with additional changes in contemporary medical care [16].
Wenger suggested that quality of life measurements are particularly useful with respect to investigating treatment of cardiovascular disease in three instances: 1) when there is little likelihood of one treatment showing a major improvement in survival over another in a clinical trial. Quality of life measurement in such a trial might point toward the choice of the therapy showing the greatest benefit for improving it; 2) when a treatment is effective in reducing mortality, but has toxic or unacceptable side effects. Quality of life measurement in this case may help physicians and their patients weight the benefits and risks of such a treatment; 3) when patients are asymptomatic or have mild symptoms, the morbidity and mortality rates are low, and the therapy is long term. Quality of life measurement would ensure that quality of life is not diminished unacceptably and there is reasonable chance of compliance with therapy [11]. This is one of the areas where most of medical research is currently centred on and it is strongly supported by the pharmaceutical industry [21].
According to Konstam et al (1996) attention has recently been directed to quality of life in the treatment of CHF patients; however, few systematic studies actually depict the relation between quality of life and clinical outcome. These authors found that the baseline assessment of quality of life independently predicted mortality and CHF related hospitalisations in symptomatic and asymptomatic patients randomised to enalapril or placebo treatment (Studies of Left Ventricular Dysfunction trial- SOLVD – patients with an ejection fraction < 0.35 were followed for a mean of 36.5 months). The domains of activities of daily living (ADL) and general health predicted mortality and CHF related hospitalisations in both univariate and multivariate analysis. Quality of life indexes provided additional predictive values with respect to both mortality and CHF related hospitalisations, above and beyond the predictive power of variables such as ejection fraction, age, treatment (enalapril vs placebo) and the NYHA classification. These findings support the argument that a patient functional and psychological status represent independent risk factors for morbidity and mortality and warrant the need for further investigations regarding the mechanisms by which self-rated activity level predicts mortality and CHF related hospitalisations (this might include a closer look on psychological mechanisms and the pathophysiology underlying activity level). This study also supports the need to include quality of life in the evaluation and course of treatment of CHF [22].
Jenkinson et al (see O'Keefe et al, 1998) noted that the SF-36 and the Darmouth CCOP, two generic measures of quality of life were not responsive to self-reported improvements in global health in a study of elderly CHF patients starting treatment with ACE inhibitor. O'Keefe et al used the Chronic Heart Failure Questionnaire (CHQ), a CHF specific quality of life measure, in a similar population and found different results. It maybe possible that ACE inhibitors, despite their beneficial effect on mortality, do not lead to major improvement in quality of life; however, the better responsiveness of the CHQ may simply reflect that it is an instrument specifically designed for use in CHF [19].
Many studies have been performed in patients with CHF of various aetiologies and the beneficial effect of long-term beta blockade, has been confirmed [23]. According to these authors, numerous studies have shown that beta blockade when given for more than two months, elicit significant improvement in functional class, exercise capacity, cardiac function, quality of life and/or morbidity. Several large studies have also reported benefits on mortality and morbidity. Randomised placebo controlled trials with β1 blockade in patients with CHF reporting quality of life or NYHA functional class have yielded inconsistent results. While studies with β1 selective agents have shown improvement in quality of life or NYHA class, others with non-selective agents have shown a less clear picture and still others have failed to show significant improvement in quality of life or NYHA class [19].
5.3 Quality of life and somatic variables
According to Juenger et al few studies have investigated the relation between quality of life and clinical variables (reflecting the severity of disease) in CHF and most of these studies achieved inconclusive results. These authors studied the relation between quality of life (assessed by a generic quality of life measure:SF-36) and somatic indices (that included the assessment of New York Heart Association – NYHA, functional class, left ventricular ejection fraction, peak oxygen uptake and the distance covered during a standardised six minute walk test). They concluded that quality of life decreases as NYHA functional class worsens. The peak oxygen uptake and the six minute walk test also showed some relation to the quality of life (patients with a more severe impairment of functional capacity had, in general, significantly lower SF-36 values). In contrast, left ventricular ejection fraction showed no clear association with quality of life. These findings may also explain why beta blockade treatment in CHF has no consistent effect on the quality of life, despite a pronounced improvement in left ventricular ejection fraction, whereas the increase in peak oxygen uptake achieved by exercise training is associated with an improvement in quality of life [20].
5.4 Social support as a measure of clinical outcome
Numerous studies of cardiac patients (coronary heart disease patients) have reported that lack of social support and social isolation are associated with increased risk of mortality, but the study undertaken by Murberg et al (2000) was the first detecting an effect of social isolation (a patient's perception that he or she is no longer able to maintain the same degree of social contacts and activities with family, other relatives and friends as previously) on mortality among CHF patients. Another finding concerns the relationship between lack of intimate network support and mortality: for the CHF patients, most of them elderly, lack of social support from a spouse seems to be a more critical factor of fatal outcome than lack of social support from primary and secondary network. The authors pointed two possible explanations for that: poor network support might be associated with poor compliance to physical and medical regimens and that poor compliance may lead, in turn, to a dismal outcome; it could be also that the association between social isolation and mortality is a reflection of some underlying factors such as subjective health or hopelessness, which have been reported in several studies to be strong predictors of mortality independently of depression [24].
5.5 Socio-economic status and hospital readmission
Philbin et al noted that the socio-economic status was an independent risk factor for CHF. The principal findings of their study suggest that differences in age, sex, race, insurance, coexistent illness and location of care do not fully explain the higher frequency of readmission among low-income patients, but rather imply that other causes may exist. The authors discuss possible causes to these findings: low-income patients may have diminished access to such care (rural hospitals may be less likely to offer comprehensive programs for management of CHF); financial constraints and educational limitations, more common among low-income persons, could compromise compliance with treatment recommendations and lead to higher rate of hospitalisations; substance abuse and cigarette smoking, more common among minorities with heart failure could also play an important role [25].
5.6 Quality of life and depression
Investigation of the links between emotion and the development and prognosis of CHF have been the focus of much psychosomatic research of patients with cardiac disease over the last years. Depression is the most explored psychosocial factor in patients with CHF. Anxiety, however, was far less investigated. From the available data it seems that anxiety is not affecting heart failure patients to a greater extent. However, it has been shown that emotional distress prior to hospitalization was twice as common in patients with CHF when compared to other patients [26].
With respect to depression, recent studies have shown that the presence of depressive symptoms below the severity threshold for a depressive disorder is associated with elevated cardiac mortality. However, the nature of the relationship between depressive symptoms and elevated risk of cardiac mortality is unclear.
Longitudinal studies of quality of life in people at risk for heart disease may help clarify the nature of interactions between affective states, physical and social function, health perceptions and cardiac events [11].
Rumsfeld et al (2003) conducted a multicenter prospective cohort study of 460 outpatients with CHF with the purpose of assess whether depressive symptoms are independently associated with changes in CHF specific health status. The patients completed a baseline Medical Outcomes Study Depression Questionnaire and both a baseline and follow-up (6 ± 2 weeks) Kansas City Cardiomyopathy Questionnaire (KCCQ) were analyzed. Approximately 30% of the patients had significant depressive symptoms at baseline. Depressed patients had markedly lower KCCQ summary scores. After adjustment for potencial confounders, depressed patients were at risk for significant worsening of their CHF symptoms, physical and social function, and quality of life. The authors concluded that depressive symptoms were a strongest predictor of short-term worsening of CHF-specific health status [27]. Also Gottlieb et al (2004) studied the prevalence of depression in an outpatient heart failure population and its relationship to quality of life. A total of 155 patients were evaluated with the Medical Outcomes Study-Depression questionnaire, the Minnesota Living with Heart Failure questionnaire and the Beck Depression Inventory (BDI). A total of 48% of the patients scored as depressed and these scored significantly worse than non-depressed patients on all components of both the questionnaires measuring quality of life. In this study depression was observed more commonly among younger than older patients. This study is consistent with the notion that depressed CHF patients may perceive their quality of life to be lower and to underestimate their functional status. The higher incidence of depression in the young suggests that depression is due to a larger disparity between the perception of functional status and the expectation. Patients'perceptions of their health status are more important than their absolute physiological impairment in determining both degree of depression and quality of life. This may lead physicians caring for depressed CHF patients to classify them as more severely compromised and rate their NYHA functional class higher [28].
Havraneck EP et al (2004) tried to identify the sociodemographic and clinical factors associated with the onset of depressive symptoms in outpatients with CHF. The patients were evaluated at baseline and one year later with a Medical Outcomes Study-Depression questionnaire, a Kansas City Cardiomyopathy Questionnaire (KCCQ) and a full clinical evaluation including patients social and economic status. Of 245 patients without depressive symptoms at baseline, 52 (21,2%) developed depressive symptoms one year later. In multivariable analysis, living alone, alcohol abuse, perception of medical care as being a substancial economic burden, and health status as measured by the KCCQ were independent predictors of developing depressive symptoms. The results of this study support the use of measures of quality of life in evaluating patients with CHF. Health status measures have been shown to predict mortality and cardiac events in cardiovascular populations, including CHF. The KCCQ scores have previously been associated with subsequent mortality and hospitalization in patients with CHF. In this study, KCCQ score was one of four independent risk factors for the development of depression among outpatients with CHF, so, it may aid in identification of patients at risk for a wide range of adverse outcomes. Future studies are warranted to evaluate whether health status-guided management of patients with CHF can improve outcomes [29].
The finding of an association between psychosocial factors and morbidity and mortality in patients with CHF underscore the importance of beginning to target these factors and not just tradicional medical factors for intervention. The failure of medical therapy to produce marked improvement in quality of life is sobering and highlights the need to examine other methods of improving psychosocial outcomes in patients with CHF. Although few investigators have examined the effect of interventions on psychosocial variables, there is evidence that nonpharmacologic interventions may be quite effective in improving psychosocial outcomes. For example, exercise, CHF disease management programs, stress management and cognitive therapy, biofeedback relaxation [30], well-being therapy [31] have all been shown to improve quality of life or depression. Interventions aimed at improving social support have been successful in other populations and deserve attention in patients with CHF. In general, nonpharmacological intervention appears to be a fruitful area for future research and practice.
5.7 Quality of life and reabilitations programs
Quality of life is a major goal in the context of preventive and therapeutic cardiology. Randomized controlled trials have demonstrated that comprehensive cardiac rehabilitation can enhance quality of life by decreasing specific symptoms, augmenting functional capacity and enhancing mood state [32,33].
Fonarow et al (1997) pointed out that a comprehensive heart failure management (which included adjustments in medical therapy and intensive patient education) led to improved functional status and an 85% decrease in the hospital admission rate for transplant candidates discharged after evaluation. The potential to reduce both symptoms and costs suggests that referral to a heart failure program may be appropriate not only for potential heart transplantation, but also for medical management of persistent functional class III and IV heart failure [34].
There is little information concerning the impact of therapeutic exercise programs on quality of life of patients with CHF, but benefit is likely from an earlier return to work (a gain in physical function) and an increase in aerobic power (physical function, physical role, fatigue and vitality) [35].
Wielenga et al (1997) studied the effect of exercise training on quality of life and exercise capacity in 35 patients with mild to moderate chronic heart failure. Three measures were used to evaluate quality of life: the Heart Patients Psychological Questionnaire; the Sickness Impact Profile and the Self- Assessment of General Well-Being. With this study the authors confirmed that exercise training can be performed safely by patients with mild to moderate CHF and that after 12 weeks of physical training an increase in exercise performance was observed. These results support the concept that exercise training is an important modality to increase quality of life in CHF [1] and may improve the psychological outlook and self confidence of patients [9]. Furthermore, the observation that the increase in training level is not reflected as an increase in peak VO2, suggests that the increase in exercise test duration is related to a psychological improvement, rather than to a physical improvement. This is in accordance with the well-known belief that exercise duration is a motivation-dependent test parameter [1], often explains why patients are able to accomplish many more of the activities of daily living than exercise testing suggests is possible. However, there are no standard ways to assess motivation, which is influenced by many factors such as emotional state, personality, financial gain or loss, and interpersonal relationships [11].
6 Conclusion
In the first documented medical journal when the term "quality of life" was used (Annals of Internal Medicine of 1966), the author (J.R.Elkinton) quotes Francis Bacon's view that "the office of medicine is but to tune this curious harp of man's body and reduce it to harmony". This is a most remarkable definition of quality of life because it stresses not only "well-being" and "satisfaction" ("the harmony within a man"), but also the relationship of a person to the environment ("harmony between a man and his world") [5]. Such an approach has obvious relevance in the assessment of quality of life in patients with chronic medical illnesses (such as congestive heart failure). Although quality of life research has its roots in the social sciences, it will be fully accepted by health care practitioners only when it answers questions directly related to clinical programs and therapeutic choices. To answer these and similar questions, future research should be used to demonstrate the links among medical interventions, clinical and physiologic changes, and quality of life [4,36].
Those involved in psychosomatic research can benefit by knowledge and familiarity with quality of life measures; their use and adaptation to psychosomatic research may help foster greater awareness and comprehension of outcomes of psychosomatic investigations by colleagues from other disciplines. Some of the generic health profiles and cardiac-disease-specific quality of life measures discussed in this review would likely be most useful alongside instruments used in psychosomatic investigations of personality, hostility, depression and social isolation, which have already been shown to be relevant to patients with heart disease [11].
The future will show whether quality of life research was a fashionable and transient movement at the end of the twentieth century or a serious endeavour with profound implications for the daily practice of medicine, for outcome assessment in clinical trials and health services research, for health needs assessment of populations, and for resource allocation [5].
==== Refs
Wielenga RP Erdman RAM Huisveld IA Bol E Dunselman PH Baselier MR Mosterd WL Effect of exercise training on quality of life in patients with chronic heart failure J Psychosom Res 1997 45 459 464 10.1016/S0022-3999(97)00309-7
Davis RC Hobbs RFD Kenkre JE Roalfe AK McLeod S Hare R Davies MK Quality of life in heart failure and others conditions JACC 1999 33 211A
Stott DJ Chronic heart failure and cognitive impairment: co-existence of conditions or true association? Eur J Heart Failure 2002 4 7 9 10.1016/S1388-9842(01)00182-9
Testa MA Simonson DC Assessment of quality of life outcomes The New England Journal of Medicine 1996 18 835 840 10.1056/NEJM199603283341306
Katschnig H Katschnig H, Freeman H, Sartorius N How useful is the concept of quality of life in Psychiatry? Quality of Life in Mental Disorders 1997 Wiley, Chichester 3 16
Guyatt GH Measurement of health-related quality of life in heart failure JACC 1993 22 4 185A 91A 8376691
Beck P Measurement of psychological distress and well-being Psychother Psychosom 1990 54 77 89 2098783
Siegrist J Junge A Measuring the social dimension of subjective health in chronic illness Psychother Psychosom 1990 54 90 98 2098784
Soler-Soler J Permanyer-Miralda G How do changes in lifestyle complement medical treatment in heart failure? Br Heart J 1994 72 suppl 87 91
Dempster M Donnelly M Measuring the health-related quality of life of people with ischaemic heart disease Heart 2000 83 641 644 10.1136/heart.83.6.641 10814620
Swenson RJ Clinch JJ Assessment of quality of life in patients with cardiac disease: the role of psychosomatic medicine J Psychosom Res 2000 48 405 415 10.1016/S0022-3999(99)00092-6 10880662
Spertus JA Winder JA Dewhurst TA Deyo RA Stephan DF Monitoring the quality of life in patients with coronary artery disease Am J Cardiol 1994 74 15 1240 44 10.1016/0002-9149(94)90555-X 7977097
Mayou R Quality of life in cardiovascular disease Psychother Psychosom 1990 54 99 109 2098785
Akhras KS Burch SP Dedhiya SD Sood VC Health-related quality of life in treatment of angina Heart Dis 2000 2 3 201 216 11728262
Majani G Pierobon A Giardini A Callegari S Opasich C Cobelli F Tavazzi L Relationship between psychological profile and cardiological variables in chronic heart failure- the role of patient subjectivity Eur Heart J 1999 20 1579 1586 10.1053/euhj.1999.1712 10529326
Wenger NK Quality of life. Can it should be assessed in patients with heart failure? Cardiology 1989 76 391 398 2688878
Desnos M Qualité de vie et insuffisance cardiaque: la quête du Graal? Archives des maladies du cœur et des vaisseaux 1997 90 12 1575 1576
Coats AJS Clarck AL Piepoli M Volterrani M Poole-Wilson PA Symptoms and quality of life in heart failure: the muscle hypothesis Br Heart J 1994 72 suppl S36 39
O'Keeffe ST Lye M Donnellan C Carmichael DN Reproducibility and responsiveness of quality of life assessment and six minute walk test in elderly heart failure patients Heart 1998 80 377 382 9875117
Juenger J Schellberg D Kraemer S Haunstetter A Zugck C Herzog W Haass M Health-related quality of life in patients with congestive heart failure: comparison with other chronic diseases and relation to functional variables Heart 2002 87 235 241 10.1136/heart.87.3.235 11847161
Fava GA Methodological and conceptual issues in research on quality of life Psychother Psychosom 1990 54 70 76 2098782
Konstam V Salem D Pouleur H Kostis J Gorkin L Shumaker S Mottard I Woods P Konstam MA Yusuf S Baseline quality of life as a predictor of mortality and hospitalisation in 5.025 patients with congestive heart failure Am J Cardiol 1996 78 890 895 10.1016/S0002-9149(96)00463-8 8888661
Hjalmarson A Kneider M Waagstein F The role of beta-blockers in left ventricular dysfunction and heart failure Drugs 1997 54 4 501 510 9339958
Murberg TA Bru E Social relationships and mortality in patients with congestive heart failure J Psychosom Res 2001 51 521 527 10.1016/S0022-3999(01)00226-4 11602222
Philbin EF William G Jenkins PL DiSalvo TG Socioeconomic status as an independent risk factor for hospital readmission for heart failure Am J Cardiol 2001 87 1367 1371 10.1016/S0002-9149(01)01554-5 11397355
Lainscak M Keber I Patient's view of heart failure: from the understanding to the quality of life Eur J Cardiov Nur 2003 2 275 281 10.1016/S1474-5151(03)00064-1
Rumsfeld JS Havranek E Masoudi FA Peterson ED Jones P Tooley JF Krumholz HM Spertus JA Depressive symptoms are the strongest predictors of short-term declines in health status in patients with heart failure JACC 2003 42 10 1811 1817 14642693
Gottlieb SS Khatta M Friedmann E Einbinder L Katzen S Baker B Marshall J Minshall S Robinson S Fisher ML Potenza M Sigler B Baldwin C The influence of age, gender, and race on the prevalence of depression in heart failure patients JACC 2004 43 9 1542 1549 15120809
Havranek EP Spertus JA Masoudi FA Jones PG Rumsfeld JS Predictors of the onset of depressive symptoms in patients with heart failure JACC 2004 44 12 2333 2338 15607395
Moser DK Worster PL Effect of psychological factors on physiologic outcomes in patients with heart failure J Cardiovasc Nurs 2000 14 4 106 115 10902107
Fava GA Well-being therapy: conceptual and technical issues Psychother Psychosom 1999 68 171 179 10.1159/000012329 10396007
Ades PA Coello CE Effects of exercise and cardiac rehabilitation on cardiovascular outcomes Med Clin North Am 2000 84 251 265 10.1016/S0025-7125(05)70217-0 10685138
Dugmore LD Tipson RJ Philips MH Changes in cardiorespiratory fitness, psychological well-being, quality of life and vocational status following a 12-month cardiac exercise rehabilitation programme Heart 1999 81 259 366
Fonarow GC Stevenson WL Walden JA Livingston NA Steimle AE Hamilton MA Moriguchi J Tillisch JH Woo MA Impact of a comprehensive heart failure management program on hospital readmission and functional status of patients with advanced heart failure JACC 1997 30 3 725 732 9283532
Sullivan M Levy WC Russo JE Spertus JA Depression and heath status in patients with advanced heart failure: a prospective study in tertiary care J Card Fail 2004 10 5 390 396 10.1016/j.cardfail.2004.01.011 15470649
Lehtinen V Sohlman B Kovess-Masfety V Level of positive mental health in the European Union: Results from the Eurobarometer 2002 survey Clin Pract Epid Ment Health 2005 1 9 10.1186/1745-0179-1-9
|
16202163
|
PMC1274338
|
CC BY
|
2021-01-04 18:01:01
|
no
|
Clin Pract Epidemiol Ment Health. 2005 Oct 4; 1:19
|
utf-8
|
Clin Pract Epidemiol Ment Health
| 2,005 |
10.1186/1745-0179-1-19
|
oa_comm
|
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-621621910910.1186/1477-7525-3-62ResearchDevelopment and validation of a quality of life questionnaire for patients with colostomy or ileostomy Prieto Luis [email protected] Hanne [email protected] Kristian [email protected] Health Outcomes Consultant. C/ Rioja, 7. 28750 San Agustin de Guadalix, Spain2 Institute of Public Health, Department of General Practice, University of Copenhagen, Denmark3 Ostomy Division, Clinical Documentation Department, Coloplast A/S, Holtedam 1, DK-3050 Humlebæk, Denmark2005 12 10 2005 3 62 62 9 5 2005 12 10 2005 Copyright © 2005 Prieto et al; licensee BioMed Central Ltd.2005Prieto et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Quality of life of stoma patients is increasingly being addressed in clinical trials. However, the instruments used in the majority of these studies have not been validated specifically for stoma patients. The aim of this paper is to describe the development and validation of a quality-of-life instrument, "Stoma-QOL", specifically for patients with colostomy or ileostomy.
Methods
Potential items were formulated in English on the basis of the results of a series of semi-structured interviews with 169 adult stoma patients. The process resulted in a preliminary 37-item version, which was translated into French, German, Spanish and Danish, and administered repeatedly to 182 patients with colostomy or ileostomy. A psychometric selection of items was performed through Rasch Analysis. The measurement properties of the final questionnaire version were subsequently tested.
Results
The 20 items in the final questionnaire covered four domains – sleep, sexual activity, relations to family and close friends, and social relations to other than family and close friends. These items were found to define a unidimensional variable according to Rasch specifications (Infit MNSQ < 1.3). Internal consistency reliability calculated as Cronbach's alpha was 0.92, i.e., highly reliable. Spearman's correlation coefficients of scores across times of administration was >0.88 (p < 0.01), indicating a high test-retest reliability. Item calibrations by country calculated as ICC were 0.81 (0.67–0.91 95% CI), confirming cross-cultural comparability across the European countries included in the study.
Conclusion
Given the adequacy of the metric properties of the Stoma-QOL suggested by the psychometric analyses, this study confirms the suitability of the instrument in clinical practice and in clinical research.
Stomaquality of lifedevelopmentvalidationreliabilityRasch analysis
==== Body
Background
Stoma patients have a surgically created opening on the abdomen involving parts of either the gastrointestinal or urinary tract. Colostomy involves discharging feces from the large intestine, ileostomy from the small intestine, while urostomy means discharging urine through the surgical opening. Due to this major change in physical appearance and bodily function, patients with stoma are challenged with a number of quality of life (QOL) issues.
In recent years QOL of stoma patients has been addressed in a number of studies [1-6], some covering a broad range of different stomas, others focusing more narrowly on just one or two of the conditions – colostomy, ileostomy or urostomy.
With few exceptions, the abovementioned instruments were not reported to have been validated specifically for stoma patients. Since the development and validation of Olbrisch's "Ostomy Adjustment Scale" in the early 1980's [1], to our knowledge the only contemporary instrument constructed and psychometrically tested specifically for colo-, ileo- and urostomy has been the "Stoma Care Quality of Life Index" (1998) [6]. This 34-item questionnaire was validated in the UK and France, showing a satisfactory reliability.
However, the psychometric properties of both the "Ostomy Adjustment Scale" [1] and the "Stoma Care Quality of Life Index" [6] were assessed solely within a classical theoretical approach, the so-called Classical Test Theory [7], which is a valid method, but in our opinion may not be the optimal solution. The main problem with the Classical Test Theory is that it presupposes that one can directly infer, e.g., a stoma patient's quality of life by summing responses and calculating a total score, assuming that each item contributes equally to this total score. However, treating items equally implies that all items are of identical importance, which in our experience might not be the case. A stoma patient's strong agreement with an item like "I worry that my family feel awkward around me" indicates a greater problem than does a strong agreement with an item like "I become anxious when the pouch is full". Thus, when items represent different levels of importance to the stoma patients' quality of life, should the data not be analysed so that the total score reflects this value of "importance" of the item's contribution to the total scale value?
To address this question, which to our knowledge has never been addressed before in association with measurements of quality of life in stoma patients, our aim was to develop a simple, cross-cultural and reliable measurement of quality of life in stoma patients, "Stoma-QOL", and to validate this instrument according to both the Classical Test Theory and the modern Item Response Theory [8], thereby taking into account the "importance" weight of each item in the test (see Methods section for details).
Methods
The psychometric models used in the development and validation of Stoma-QOL
The content of the new stoma-specific QOL instrument was developed on the basis of Hunt and McKenna's needs-based model of QOL [9]. This model draws on the work of theorists in the field of human motivation who postulate that individuals are motivated or driven by their needs, e.g., as defined in Maslow's well-known hierarchy of needs pyramid [10]. For this study, this approach implied that rather than relying on literature or experts to determine needs important to patients, the original content of the questionnaire was derived from qualitative interviews with stoma patients.
For the reasons briefly mentioned in the Introduction, the Item Response Theory [8] was our primary model for analysis of the questionnaire resulting from the interviews with stoma patients. This method is built around the idea that the probability of a patient's answer when confronted with a certain item ideally can be described as a simple function of the patient's position on a latent trait (e.g., quality of life) plus one or more parameters characterising the particular item (e.g., its "severity" or "importance" or "weight").
The Item Response Theory [8]measures the quality of a given test as a measurement instrument, and helps to predict its performance in future applications. However, the Item Response Theory can also assist in improving the quality of the test, e.g. by indicating which items are inappropriate and should be changed, deleted, or replaced. After this process, the test can be used as a standard instrument to measure similar patients. Skipping mathematical explanations, the Item Response Theory can do more things than the Classical Test Theory [7,8] when it comes to modelling existing tests, constructing new ones, and, above all, interpreting the results of measurement.
We chose the Rasch model [11,12] as this is a simple but at the same time very powerful Item Response Theory model for measurement. The Rasch model uses the traditional total score (i.e., the sum of the item ratings) as a starting point for estimating response probabilities. The model is based on the simple idea that some items are more important to patients than other items. Thus, the Rasch model constructs a line of measurement with the items placed hierarchically on this line according to their importance to patients. The validity of a given test can be assessed through examination of this item ordering, i.e. by assessing whether all items work together to measure a single variable.
See the section "Analysis and item reduction of the 37-item questionnaire" for details on the practical implementation of these theoretical models.
Development of the Stoma-QOL questionnaire
Figure 1 details the process of the cross-cultural development of the Stoma-QOL questionnaire.
Figure 1 Development of the Stoma-QOL questionnaire.
The development was initiated by the formulation of potential stoma-related items in English on the basis of the results of a series of semi-structured interviews conducted by stoma care nurses with 169 stoma patients in France, Denmark, Spain and the United Kingdom.
The interviews were structured to cover the following five broad domains, which are included in Maslow's hierarchy of needs pyramid [10] and at the same time were supported by the experience of stoma care nurses in their daily routine with stoma patients:
1. What, if any, concerns do you have about what you can eat?
2. What, if any, concerns do you have about sleeping?
3. What, if any, concerns do you have about intimate relations?
4. What, if any, concerns do you have regarding your relationship with family and close friends?
5. What, if any, concerns do you have regarding your relationship with people other than family and close friends?
Stoma care nurses put these questions to the patients in their respective national languages, and the answers were collected on a special form. Answers given in non-English-speaking countries were translated into English. A common listing of the answers was generated from these interviews. Redundancies and answers like "no concerns" were removed.
The next step consisted in the selection, at a meeting between the national investigators, of items that could be translated from English into the four non-English languages involved in the project (German, Spanish, French and Danish). It was ensured that all the items that were chosen were consistent with the need-based theory of quality of life [9]. Furthermore, it was decided that the items should be formulated so that they could be meaningfully answered with the following four response categories: "Always", "Sometimes", "Rarely" and "Not at all".
Following an international accepted protocol [13], the translation of the questionnaire from English into the four non-English languages proceeded in three steps. First, the items were translated by a panel of bilingual translators. In the second step, this intermediate translation was assessed by a panel of lay persons for linguistic clarity, understandability and easiness to complete. Thirdly, field tests were conducted as individual interviews with 12 stoma patients in each country, after which, where necessary, items were again semantically adjusted without distorting the content of the items. The process resulted in a 37-item translated version for each country.
Finally, a specific validation study was initiated in each country, aiming to test the psychometric properties of the preliminary 37-item questionnaire. This study also aimed at reducing the number of items through psychometric analysis. We wanted to preserve as much as possible of the structure of the preliminary questionnaire and, in addition, to allow for calculation of one global score. The measurement properties of the final 20-item questionnaire were subsequently tested as described in the following section.
Analysis and item reduction of the 37-item questionnaire
The original 37-item questionnaire was analysed with the Rasch Rating Scale model in a special version allowing more than two answer categories to be modelled for each item for the overall sample. Rasch analyses were performed with Version 3.0.1 of the WINSTEPS computer program [14]. An item calibration was obtained for each item.
In order to determine how well each item contributed to common global health measurement, chi-square fit statistics, known as Infit Mean Square (Infit MNSQ), were also calculated [12]. In this analysis values greater than 1.3 imply a potential misfit to the Rasch model [15], and items with values above this threshold were consequently removed from the test. Successive Rasch analyses were performed until all the remaining items showed acceptable goodness-of-fit properties.
Subsequently, the performance of the final 20-item questionnaire was determined as the index of person separation (PSEP) [7,8,12]. The index of person separation describes how reliably the patients are separated by the scale and has to exceed 2 (or 3) in order to confirm an optimal level of reliability of 0.80 (or 0.90).
Stratified Rasch analyses were also performed for each country in the study. The concordance of item calibrations by country was assessed through an Intraclass Correlation Coefficient (ICC) [15]. This is a statistical procedure used to determine the reproducibility of a measurement of a variable. The ICC is based on variance components analysis and measures the homogeneity within groups relative to the total variation. The ICC is large when there is little variation within the groups compared to variation among group means, where groups consist of replicate measurements. A small ICC occurs when within-group variation is large compared with between-group variability, indicating that some unknown variable has introduced nonrandom effects in the different groups. The maximum value of the ICC is 1, and the minimum value is theoretically 0. The ICC is routinely used in epidemiological studies to address the test-retest reliability, validity of questionnaires and interlaboratory concordance.
As a secondary model for analysis, the final 20 items were also subjected to a traditional item analysis. We used the following gold standard statistical procedures based on Classical Test Theory [8]: a) classical index of discrimination was calculated to measure the spread of scores between the patients; b) difficulty indices were determined by calculating the mean response choice for each item; c) Cronbach's alpha coefficient was calculated to estimate internal-consistency; d) Exploratory Factor Analysis (EFA) was performed in order to test the unidimensionality of the reduced version; e) test-retest reliability estimates were obtained for the reduced scale by calculating Spearman's coefficients of correlation across the different times of administration of the questionnaire (T1, T2 and T3); and f) distribution patterns of scores were described for each reduced questionnaire, overall and by country.
The Statistical Package for Social Sciences (SPSS), version 10, was used to perform all the above analyses.
Results
Patients
One hundred and eighty-two patients from four different European countries with various backgrounds for the creation of a stoma (Crohn's, cancer, diverticulitis, etc.) were included in the validation study Mean age was 53 years ranging from 18 to 84 years, with slightly more males than females (Table 1). 52% had colostomy and 48% had ileostomy. No urostomy patients were included. All patients were in a stable period or cured, with a duration since stoma creation ranging from 0 to 43 years, when they participated in the study. Incomplete or missing data for up to 12 patients resulted in a sample less than 182 on some of the abovementioned demographic variables.
Table 1 Demographics of patients in the validation study
DK Germany Spain France Total
N*) 49 43 58 32 182
Age (years), mean (SD) 58.2 (12.5) 61.9 (11.9) 40.7 (11.6) 58.5 (14.5) 53.4 (15.3)
Sex (Male/Female), N (%) 14/32 (30%/70%) 33/8 (80%/20%) 28/30 (48%/52%) 18/14 (56%/44%) 93/84 (52.5%/47.5%)
Type of stoma
Colostomy, N (%) 23 (50.0%) 34 (100.0%) 0 (0.0%) 32 (100.0%) 89 (52.4%)
Ileostomy, N (%) 23 (50.0%) 0 (0.0%) 58 (100.0%) 0 (0.0%) 81 (47.6%)
Duration from stoma creation (years), mean (SD) 15.4 (11.0) 4.8 (3.4) 4.6 (4.8) 2.7 (5.4) 7.1 (8.4)
*) Missing demographic data for up to N = 12 means that sample N will be less than 182 on some variables. Means, SD and percentages refer to valid cases only.
Responses to the 37-item questionnaire came from two different study sources. The patients from France, Germany and Spain were included as a part of controlled clinical trials (randomised cross-over designs) conducted to test a new stoma pouch. Where required these trials were approved by ethics committees and informed consent was obtained. In accordance with the protocol of the clinical trials, the patients responded to the questionnaire on three different occasions. The patients from Denmark were recruited directly through a stoma patient database approved by the Danish Data Protection Agency. For logistic reasons, these patients did not test any stoma products and only responded twice to the instrument.
Results from psychometric analysis
The overall Rasch analysis of the 37 items of the original questionnaire (Table 2) showed 6 misfitting items. Infit MNSQ statistics ranged from 0.69 to 1.40 (SD = 0.19). Misfitting items in this and subsequent analyses were removed until no further improvement in fit requirements was found. Seventeen items (items 1, 2, 6, 7, 9, 10, 11, 12, 14, 15, 20, 28, 30, 31, 32, 36 and 37) were discarded in this process (performed in seven different steps), reducing the initial questionnaire to 20 items, the final Stoma-QOL. During the process all questions belonging to the domain related to food were omitted because these items did not contribute to constructing, with the remaining items, a common and single health-related quality of life variable. With four response choices per question (1. Always; 2. Sometimes; 3. Rarely; 4. Not at all), the highest possible raw score for the reduced questionnaire is 80 (best QOL) and the lowest possible score is 20 (worst QOL).
Table 2 Content of the Original 37-Item Questionnaire and of the Final, Reduced Version, Stoma-QOL (4 response choices: 1-Always, 2-Sometimes, 3-Rarely, 4-Not at all)
Original 37-Item Questionnaire Stoma-QOL + : item included - : item excluded
1. I worry about skin problems where the pouch attaches -
2. Because of my stoma I prefer eating at home -
3. I feel the need to know where the nearest toilet is +
4. I become anxious when the pouch is full +
5. I feel tired during the day +
6. I worry that my family will reject me -
7. I avoid sexual intimacy because of my stoma -
8. I am afraid of meeting new people +
9. I am preoccupied by what I can eat and drink -
10. I worry that friends will reject me -
11. My sleep is interrupted because of my stoma -
12. I avoid sleeping in certain positions -
13. It is difficult to hide the fact that I wear a pouch +
14. I have to avoid drinks that I like -
15. I have problems falling asleep -
16. My stoma makes it difficult for me to be with other people +
17. I sleep badly during the night +
18. I feel lonely even when I am with other people +
19. I need to rest during the day +
20. I worry about the pouch leaking -
21. I worry that my condition is a burden to people close to me +
22. I avoid close physical contact with my friends +
23. I worry that my family feel awkward around me +
24. I feel embarrassed about my body because of my stoma +
25. It would be difficult for me to stay away from home overnight +
26. I worry that the pouch rustles +
27. I worry that the pouch may smell +
28. I am afraid of being rejected sexually because of my stoma -
29. My stoma makes me feel sexually unattractive +
30. I worry that my friends feel awkward around me -
31. I have to think about my pouch when planning my day -
32. I avoid close physical contact with my family -
33. I worry about noises from the stoma +
34. I worry that the pouch will loosen +
35. My stoma pouch limits the choice of clothes that I can wear +
36. I have to avoid situations where I over-perspire (for example, brisk walking or sports) -
37. I avoid getting changed in front of other people -
There were 178 individuals (out of 182) susceptible to measurement in the Rasch analysis. A total of four cases were not considered in the analysis, since they reported a maximum extreme score (n = 2), or lacked responses for the whole questionnaire (n = 2). Valid responses accounted for 98.3% of the sample. The item calibrations, or the item "weights", varied from -1.60 to 1.33 logits, all but one below the threshold of 1.3 for a potential misfit. The 20 items fit to define a unidimensional variable according to initial Rasch specifications (Infit MNSQ < 1.3) (Table 3).
Table 3 Rasch Analysis of the Items of the Final, Reduced Stoma-QOL: Item Statistics by Measure (or Item Calibration) Order.
Item no. Item text Calibration SE Infit MNSQ
i4 I become anxious when the pouch is full 1.33 0.10 0.78
i34 I worry that the pouch will loosen 1.16 0.10 1.13
i3 I feel the need to know where the nearest toilet is 1.02 0.10 1.16
i27 I worry that the pouch may smell 0.92 0.10 1.05
i33 I worry about noises from the stoma 0.72 0.09 0.88
i19 I need to rest during the day 0.42 0.09 0.93
i35 My stoma pouch limits the choice of clothes that I can wear 0.35 0.10 1.07
i5 I feel tired during the day 0.27 0.09 0.80
i29 My stoma makes me feel sexually unattractive 0.23 0.11 1.22
i17 I sleep badly during the night 0.08 0.10 1.16
i26 I worry that the pouch rustles -0.03 0.10 1.19
i24 I feel embarrassed about my body because of my stoma -0.10 0.10 0.93
i25 It would be difficult for me to stay away from home overnight -0.13 0.10 1.14
i13 It is difficult to hide the fact that I wear a pouch -0.22 0.10 0.92
i21 I worry that my condition is a burden to people close to me -0.41 0.11 1.18
i22 I avoid close physical contact with my friends -0.56 0.11 0.98
i16 My stoma makes it difficult for me to be with other people -0.97 0.12 0.76
i8 I am afraid of meeting new people -1.12 0.12 0.93
i18 I feel lonely even when I am with other people -1.35 0.13 0.73
i23 I worry that my family feel awkward around me -1.60 0.16 1.29
The index of person separation, PSEP, for the Stoma-QOL was 2.92, corresponding to a reliability of 0.90.
In Figure 2, the calibrations, or "weights", of the items are located within the spread of the Stoma-QOL patient scores. The mean of the item calibrations was adopted by default as the 0 point. Item 26 was calculated as having that exact middle "weight", so it is located at the 0 point on the item-person map. Patients above a given item are likely not to indicate any concerns with it. Thus, the higher a patient is positioned on the map relative to the items of the test, the better in terms of quality of life. As an example, it can be seen that most of the patients indicated being "anxious when the pouch is full", so the majority of the patients are below the level of item 4; on the other hand, almost no-one had any concerns that their "family feel awkward around" them (item 23), so most patients were located above the calibration of item 23.
Figure 2 Stoma-QOL Item Calibration of the items located within the spread of the person measures. (logit scores, the measurement unit common to both patients and items, are displayed down the middle of the map; patient, represented by a single dot, are arranged within the scale from better (top) to worse (down) quality of life; items are identified by the item number)
Rasch analysis was also performed on the categories used as response choices, indicating the 'distance' that separates the four response choices. These weights ranged from -0.94 to 1.66. The distance that separated response category "2. Sometimes" and "3. Rarely" (0.75) was found to be in the same range as the distance that separated "1. Always" and "2. Sometimes" (0.89) and "3. Rarely" and "4. Not at all" (0.96).
Item parameters by country also fitted to the Rasch model (Infit MNSQ<1.3) and had very similar item calibrations: ICC of the item calibrations by country was 0.81 (0.67–0.91 95% CI).
The secondary analysis according to classical test theory of the final reduced questionnaire gave the following results: The classical item discrimination index for the 20 items of the questionnaire ranged from 0.51 to 0.67, exceeding the minimum recommended value of 0.3. The mean response choice for each item (difficulty index) ranged from 2.11 to 3.60, which suggests that all items are moderately spread around the centre of the four response choices (1. Always; 2. Sometimes; 3. Rarely; 4. Not at all). Internal consistency reliability estimated as Cronbach's alpha for the Stoma-QOL was 0.92, exceeding the minimum recommended standard of 0.70.
In order to test whether a calculation of a global score from the Stoma-QOL is a valid measure, we inspected the scree plot in an Exploratory Factor Analysis. We found that a single component was an optimal solution with factor loadings in a single factor solution ranging from 0.53 to 0.72, accounting for 38% of the total variance.
Table 4 shows Spearman's correlation coefficients of the Stoma-QOL scores across times of administration of the questionnaire (T1 vs. T2 vs. T3) with all scores >0.88 (p < 0.01).
Table 4 Spearman's correlation coefficients of the Stoma-QOL scores across times of administration.
Pair wise comparisons of times of administrations Spearman
1st vs. 2nd r = 0.913
1st vs. 3rd r = 0.881
2nd vs. 3rd r = 0.946
Finally, to take into account each particular item calibration in the final Stoma-QOL scores, raw scores were transformed through Rasch modelling into a new score scale set to a minimum of 0 (Worst Quality of Life) and a maximum of 100 (Best Quality of Life) points. Overall mean on this scale was 58.5, with the lowest mean value in France, 53.8, significantly lower than Denmark with the highest mean value, 62.6 (p = 0.007).
Table 5 shows the scoring correspondence between the simple raw score of the four response choices to the 20 items of the Stoma-QOL and the final 0–100 score.
Table 5 Scoring Correspondence Between the Simple Raw Sum of the responses to the 20 items of the Stoma-QOL (responded in a scale: 1-Always; 2-Sometimes; 3-Rarely; 4-Not at all) and the Final 0–100 Score
Raw Score (Simple Sum of 20 Items each scored from 1 to 4) Final Score Raw Score (Simple Sum of 20 Items each scored from 1 to 4) Final Score
20 0.00 51 53.47
21 11.54 52 54.13
22 18.48 53 54.88
23 22.70 54 55.53
24 25.80 55 56.19
25 28.24 56 56.85
26 30.30 57 57.50
27 32.08 58 58.16
28 33.58 59 58.91
29 34.99 60 59.57
30 36.30 61 60.32
31 37.52 62 60.98
32 38.65 63 61.73
33 39.68 64 62.48
34 40.62 65 63.32
35 41.56 66 64.17
36 42.50 67 65.01
37 43.34 68 65.85
38 44.18 69 66.79
39 45.03 70 67.82
40 45.78 71 68.95
41 46.53 72 70.08
42 47.28 73 71.39
43 48.03 74 72.89
44 48.78 75 74.58
45 49.44 76 76.55
46 50.19 77 79.17
47 50.84 78 82.83
48 51.50 79 89.02
49 52.16 80 100.00
50 52.81
Discussion
Strengths and weaknesses of this study
Rooted in the lower three sections of Maslow's hierarchy of needs pyramid [10], Stoma-QOL items were generated by the asking the patients about their concerns over food, sleep, sexual activity, relations to family and close friends, and social relations to other than family and close friends. The subsequent Rasch analysis led to omission of the food domain, as items within this domain failed to contribute to construction of the global score. However, the final set of items within the remaining four domains of the Stoma-QOL successfully defined a meaningful measurement instrument with excellent psychometric properties with regard to validity and reliability.
It could be argued that a possible framing effect was present in the item generation phase. While we cannot rule out that some kind of biased selection of items may have been present during one or more steps of the development of the questionnaire, we find it important to underline that in our view selection of items is always a qualitative process, and thus somehow subjective. However, the resulting instrument has been subject to a quantitative analysis based on both classical test theory approaches and Rasch analysis, with fairly acceptable results.
Similarly, it could also be argued that the final four specific domains do not necessarily cover every relevant QOL issue specific to stoma patients, e.g. needs belonging in the top of Maslow's hierarchy of needs pyramid, such as "Ego needs" or "Self Actualisation" needs [10]. In fact, the relatively low level of variance explained by the Exploratory Factor Analysis (38%) indicates that there are other domains that impact on QOL, and this is an avenue for further research. However, the potential contribution of interventions, such as, e.g., stoma devices, to patients' QOL will by their nature primarily be correlated with improvements in the patients' basic functional status, and thus we consider these four domains to be adequate for the purpose. Furthermore, there are impracticalities associated with assessing an extensive number of domains in one instrument, e.g. increased length, complexity and time to completion.
In our opinion, the concept of responsiveness can be rejected as a separate measurement property of an evaluative instrument, a point of view supported by several authors [17-19], and for this reason responsiveness of the questionnaire was not tested. In opposition to a recent paper by Puhan et al. [20], we find that internal consistency coefficient adequately reflects an instrument's potential sensitivity to changes over time.
The statistically significant difference found between Denmark and France with regard to total Stoma-QOL scores, 8.8 points higher in Denmark than France, is notable, but it should be emphasised that the study was not designed to address this question – merely to show similar item calibrations between the countries.
For study-logistic reasons, only colostomy and ileostomy patients, but no urostomates, were included in the validation study. While we would predict that more similarities than differences exist between the different groups of stoma patients with regard to QOL issues, the use of Stoma-QOL in a population including urostomy patients would require testing for validity and reliability in the specific patient segment of urostomates.
Also, demographic data collection was limited to the main categories, colostomy and ileostomy, but for future studies it would be valuable to further subcategorise patients with regard to sigmoidostomies, transverse colostomies, loop stomas, etc., and perhaps also with regard to whether stoma surgery was performed on an elective or an acute basis. In the practical everyday surgical situation, the choice between performing the different types of stoma operation often has to be made more on the basis of experience than on the basis of evidence. If extended as suggested above, a tool such as Stoma-QOL would give the possibility of estimating potential differences between the different surgical techniques with regard to patients' expected post-operative quality of life. Used for this purpose, a tool such as Stoma-QOL could also contribute to providing more evidence with regard to variations in different age groups, in order to explore the importance of aiming to avoid stomas in those age groups particularly at risk of being affected by stoma-related reduced quality of life.
The patients included in this study were in general in a stable period or cured with a mean duration since stoma creation of more than seven years. A potential weakness of our study could therefore be a lack of sensitivity to QOL-related issues in patients whose stoma surgery took place recently. In the validation of the Ostomy Adjustment Scale, a small but significant relationship was found between global scores and the number of months elapsed since surgery [1]. Thus, for future studies of Stoma-QOL, we would recommend including a sufficient number of patients who have only recently undergone surgery to address these patients' specific concerns during their initial "adjustment process".
The full use of Stoma-QOL, for instance in a clinical trial, requires in principle that all 20 questions be answered. The reason is that each of the 20 questions will weigh with 1–4 points per question (1. Always; 2. Sometimes; 3. Rarely; 4. Not at all), and the summary patient score in the range of 20–80 will be converted to a global "0–100 score". In a clinical trial of an intervention, for instance a novel stoma pouch, some questions, more specifically i4, i34, i27, i35, i26 and i13 (Table 3), will be directly related to the intervention, while the remaining questions will be related, directly or indirectly, to the underlying condition (the stoma). In this situation it may be relevant to select only the questions related to the intervention and to omit the others. However, as the response burden of this 20-item questionnaire is already low (time to complete the questionnaire rarely exceeds 10 minutes), and in order to be able to calculate the overall QOL score, we generally recommend using the instrument in its full length.
Comparison with other studies
The most important difference between Stoma-QOL and other instruments intended for stoma-patients is that the items in Stoma-QOL were generated by the primary source, the stoma patients themselves. The initial item generation was designed to get the patients to describe all their daily life concerns within the five pre-selected domains. This method, in our opinion, is preferable compared to item generation based on literature, experts and other second-hand sources. In contrast, the "Stoma Care Quality of Life Index" [6] was constructed as a modification of an existing tool, "QLI" for cancer patients, without involving the stoma patients in the item generation.
Another important difference between the Stoma-QOL and other more traditionally validated tools such as the "Stoma Care Quality of Life Index" [6] or the "Ostomy Adjustment Scale" [1] is that, as a result of the Rasch analysis, items are ordered from top to bottom according to the importance of the health problems (Table 3). The importance of each item is also reflected in the calculation of the total Stoma-QOL score, and thus the resulting measure, in our view, will be more meaningful to the clinician. For future studies, we plan to compare scores obtained with Stoma-QOL with scores obtained with one or more of the abovementioned instruments in order to investigate their degree of correlation, notwithstanding the differences in development and validation methods.
A specific instrument such as Stoma-QOL has its self-evident strengths as compared with generic instruments by virtue of its increased sensitivity to the unique problems related to a particular disease. As an example of obtaining a different outcome when using a specific as against a generic instrument in colorectal QOL research, a lower health-related QOL was found in a Danish study of Crohn's disease patients using a specific instrument [21], while, contrary to that result, another study using a generic instrument found health-related QOL to be at a level equalling that of the general population [22]. However, specific instruments also have limitations. In contrast to generic instruments, Stoma-QOL is not comprehensive outside its final four domains and cannot be used to compare with other conditions than stoma [23].
National versions of Stoma-QOL in English, Danish, German, French or Spanish and a User Manual in English can be downloaded upon registration at the website:
Conclusion
In conclusion, the metric properties of the Stoma-QOL questionnaire were assessed by means of Rasch analysis, taking into account the importance to the patients of the items, as well as the standard Classical Test Theory approaches, which assume that each item contributes equally to the total score. The importance weights of the 20 items of the Stoma-QOL spread out in a way that shows a coherent and meaningful direction, defining a variable of useful generality. Results also showed that the Stoma-QOL conforms to the Rasch model expectation of item weight invariance between national versions in different European countries. Given the adequacy of the metric properties of the Stoma-QOL suggested by the Rasch results as well as by the classical analysis, this study has shown the suitability of the instrument both for clinical practice and for clinical research.
Authors' contributions
LP participated in the design of the study and performed the statistical analysis. HT planned and conducted the international collection of qualitative statements and supervised the translation including the field tests. KJ drafted the manuscript. All the authors read and approved the final manuscript.
Acknowledgements
The authors would like to express their thanks to Dr. Rune Sjödahl, Dr. Olof Hallböök and Dr. Peter Andersson, Linköping University Hospital, for reviewing and commenting on this article. This study was supported by a grant from Coloplast A/S.
==== Refs
Olbrisch ME Development and validation of the ostomy adjustment scale Rehabil Psychol 1983 28 3 12
Nugent KP Daniels P Stewart B Patankar R Johnson CD Quality of life in stoma patients Dis Colon Rectum 1999 42 1569 1574 10613475
Karadag A Mentes BB Uner A Irkorucu O Ayaz S Ozkan S Impact of stomatherapy on quality of life in patients with permanent colostomies or ileostomies Int J Colorectal Dis 2003 18 234 238 12673489
Silva MA Ratnayake G Deen KI Quality of life of stoma patients: temporary ileostomy versus colostomy World J Surg 2003 27 421 424 12658485 10.1007/s00268-002-6699-4
Scarpa M Barollo M Polese L Keighley MR Quality of life in patients with an ileostomy Minerva Chir 2004 59 23 29 15111829
Marquis P Marrel A Jambon B Quality of life in patients with stomas: the Montreux Study Ostomy Wound Manage 2003 49 48 55 12598701
Nunnally JC Bernstein IH Psychometric theory 1994 3 New York, McGraw-Hill
Crocker L Algina J Introduction to classical and modern test theory 1986 Fort Worth, Harcourt Brace Jovanovich
Hunt SM McKenna SP : The QLDS: a scale for the measurement of quality of life in depression Health Policy 1992 22 307 319 10122730 10.1016/0168-8510(92)90004-U
Maslow A Motivation and Personality 1970 2 New York Harper & Row
Prieto L Alonso J Lamarca R Wright BD Rasch measurement for reducing the items of the Nottingham Health Profile J Outcome Meas 1998 2 285 301 9803716
Bond TG Fox CM Applying the Rasch Model: Fundamental Measurement in the Human Sciences 2001 Lawrence Erlbaum Associates Publishers. Mahwah, New Jersey 7 34
Swaine-Verdier A Doward LC Hagell P Thorsen H McKenna SP Adapting quality of life instruments Value Health 2004 7 S27 30 15367241 10.1111/j.1524-4733.2004.7s107.x
Wright BD Linacre JM User's Guide to WINSTEPS: Rasch-Model Computer Program 1999 Chicago, MESA Press
Smith RM Schumacker RE Bush MJ Using item mean squares to evaluate fit to the Rasch model J Outcome Measurement 1998 2 66 78
Deyo RA Diehr P Patrick DL Reproducibility and responsiveness of health status measures. Statistics and strategies for evaluation Control Clin Trials 1991 12 142S 158S 1663851
Lindeboom R Sprangers MA Zwinderman AH Responsiveness: a reinvention of the wheel? Health Qual Life Outcomes 2005 3 8 15691385 10.1186/1477-7525-3-8
Hays RD Hadorn D Responsiveness to change: an aspect of validity, not a separate dimension Qual Life Res 1992 1 73 75 1301117 10.1007/BF00435438
Terwee CB Dekker FW Wiersinga WM Prummel MF Bossuyt PM On assessing responsiveness of health-related quality of life instruments: guidelines for instrument evaluation Qual Life Res 2003 12 349 62 12797708 10.1023/A:1023499322593
Puhan MA Bryant D Guyatt GH Heels-Ansdell D Schunemann HJ Internal consistency reliability is a poor predictor of responsiveness Health Qual Life Outcomes 2005 3 33 15877824 10.1186/1477-7525-3-33
Guassora AD Kruuse C Thomsen OO Binder V Quality of life study in a regional group of patients with Crohn disease. A structured interview study Scand J Gastroenterol 2000 35 1068 1074 11099060 10.1080/003655200451199
Casellas F Lopez-Vivancos J Badia X Vilaseca J Malagelada JR Impact of surgery for Crohn's disease on health-related quality of life Am J Gastroenterol 2000 95 177 182 10638579 10.1111/j.1572-0241.2000.01681.x
Guyatt G Feeny D Patrick D Issues in quality-of-life measurement in clinical trials Control Clin Trials 1991 12 81S 90S 1663862
|
16219109
|
PMC1274339
|
CC BY
|
2021-01-04 16:38:14
|
no
|
Health Qual Life Outcomes. 2005 Oct 12; 3:62
|
utf-8
|
Health Qual Life Outcomes
| 2,005 |
10.1186/1477-7525-3-62
|
oa_comm
|
==== Front
Int J Health GeogrInternational Journal of Health Geographics1476-072XBioMed Central London 1476-072X-4-231620971710.1186/1476-072X-4-23MethodologyThe Population Health Approach: health GIS as a bridge from theory to practice Barnard Deborah Kelly [email protected] Weimin [email protected] Population Health Surveillance Unit, Vancouver Island Health Authority, Victoria, British Columbia, Canada2005 6 10 2005 4 23 23 16 6 2005 6 10 2005 Copyright © 2005 Barnard and Hu; licensee BioMed Central Ltd.2005Barnard and Hu; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The Population Health Approach, proposed by Health Canada, is the articulation of a long advocated model of human health. This approach strives to ensure that the health system is appropriately oriented to improve health status by applying evidence based practices across the continuum from health determinants to service interventions. Although conceptually appealing, it has been difficult to implement widely in the existing program-based health care system. The Population Health Surveillance Unit (PHSU) of the Vancouver Island Health Authority (VIHA) has developed a health geographical information system (HGIS), where GIS is used as both platform for information integration and as an analytical tool supporting comprehensive data analysis. With the assistance of the HGIS, the theory of the population health approach can be transformed into a practical, stepwise process supporting health services and program planning.
Results
Three important components of a health service planning and evaluation framework grounded in population health theory are described in this article. In particular, a stepwise methodological process to enable the incorporation of the principles of a population health into practical applications is presented; the technical functionality to integrate multiple sources of information, with different levels and scales is discussed; and sources of information about the health of the population at the appropriate level to populate this frame are proposed. An application of the methodology in the planning of health services for a high needs neighbourhood is presented as an illustrative example.
Conclusion
The population health approach incorporates the consideration of health determinants and the context within which the health conditions arise in communities. The complexity of these relationships requires the application of innovative methodologies such as Health GIS to frame the issues practically. A population health based foundation for the planning and evaluation of health services can now move from theory to practice.
Community Health PlanningNeeds AssessmentPublic Health InformaticsMedical GeographyGeographic Information Systems
==== Body
Background
The Population Health Approach, proposed by Health Canada [1], is the articulation of a long advocated model of human health [2] [Figure 1]. This approach strives to ensure that the health system is appropriately oriented to improve health status by applying evidence based practices across the continuum from health determinants to service interventions. It is a proactive stance taking responsibility for the health outcomes of a defined group of people based on a longitudinal view of the influences on health.
Although conceptually appealing, it has been difficult to implement this approach widely in the existing program-based health care system, the bridge between theory and practice tenuous [3]. In Canada, our health system has been built through the development of "sectors", a mix of publicly and privately funded services. Hospital care, physician services, laboratories and diagnostic imaging services, public health, mental health care, home and continuing care, pharmaceutical programs, workers compensation services, ambulance services, were developed as independently planned and delivered services, often with central program and policy control. It has been noted that the alignment of resources is based on factors other than overall health benefit, resulting in gaps as well as overlap and duplication. In an attempt to streamline the delivery system, making it less fragmented and more responsive to local needs, most provinces have partially devolved health delivery responsibility (for some of the publicly funded services) to regional bodies. Stated additional goals were to increase community-based services, improve public participation in health care and to encourage policies and programs to promote health [4]. Notable exclusions from the regionalized models are physician services and pharmaceutical insurance. Many of these arrangements have been in place for over a decade, yet the tension remains, with uncertainty as to the appropriate distribution of services among preventive, community- based and acute care services.
In British Columbia, despite this investment in regionalized structures, the delivery of health care information continues to be entirely program utilization based. Information development has not moved with the conceptual and structural changes in the system. This mismatch between the format and availability of information, and the tasks faced by the health authorities in the delivery of their services, is an additional impediment to the realization of integrated and efficient health services. Providing, with consistent and readily accessible information, a conceptually coherent picture of the population in terms of influences on health, health status, and the distribution and outcome of services would enable the health system to plan, implement and evaluate its services in this context.
The Population Health Surveillance Unit (PHSU) of the Vancouver Island Health Authority (VIHA) has developed a health geographical information system (HGIS), where GIS is used as both platform for information integration and as an analytical tool supporting detailed data analysis. Description of HGIS is beyond the scope of this paper and can be found in recent comprehensive review [5].
Turning data assets (large program based data sets) into person and population-based perspectives to better serve our program needs is a significant challenge. Traditional epidemiological constructs (disease rates and distributions) and utilization-based reporting (numbers and costs of services) alone cannot support the integrated view that is required for a population health approach.
With the assistance of the HGIS and the use of a wide range of data (including underutilized administrative data), the theory of the population health approach can be transformed into a practical, stepwise process supporting health services and program planning. Compelling representations can be developed of the health needs of people in the context of health influences, services and outcomes. This article describes methodology, data sources and tools, and discusses the some specific initial results obtained through the application of this process in support of the planning of services for an urban neighbourhood.
Data and Methods
The methodology proposed for the application of population health theory involves five basic steps:
1. Population Identification
2. Population health assessment including the ecological profile
3. Description of existing service utilization and distribution
4. Analysis of the interface between needs and existing services (gaps and redundancies)
5. Outcome evaluation
These steps are iterative and cyclical, with outcomes evaluated in the context of population need. As population characteristics and service profiles shift, and more information becomes available, the framework is retained as an ongoing resource for decision-making.
Identification of the population to be served
The Health Geographical Information System (HGIS), developed by the Population health Surveillance Unit (PHSU), uses ArcGIS spatial technology to integrate most commonly available spatial information, such as Statistics Canada's census geographies, postal codes, the boundaries of BC Ministry of Health local health authorities, and commercial spatial data including street network files. These geographies are appropriately spatially associated with each other using ArcGIS. Non-spatial attribute data, which are specific to each layer of these spatial features, are also integrated by ArcGIS and linked to specific geography. Each spatial feature and its attribute information are linked in the ArcGIS, and spatially referred to other geographical features. This HGIS plays a critical role in preliminary stages of this population health approach process.
To identify the population to be served, the HGIS is used to define the geographical boundary of the catchment area, based on pre-determined street names proposed by the project team. By overlaying other layers spatial information, such as 6-digit postal codes and census tract polygons, on the tope of the catchment area polygon, the postal codes falling within the catchment area and the census tracts are extracted and spatially referred.
BC Medical Services Plan (MSP) maintains a client registry dataset, which contains every client and related demographic and geographical information in British Columbia, who register to the MSP. The postal code of each client residential address is matched to those extracted postal codes as described previously. The clients with the postal codes of their home address matched to previously queried postal codes are then aggregated to form the population to be served. It is important to note that although the data used is at an individual level, no identifiable information is available. Additionally, the information products resulting from this work are at aggregated levels to mask any possibly identifying characteristics.
Once the base catchment population has been defined geographically, further stratification based on characteristics such as health risk factors, high co-morbidity or specific health conditions or service utilization, can be undertaken to refine health service planning.
Population profile
The demographic characteristics of identified population are described based on the information contained in a reference client registry file. Also included in administrative data sets are the encounter records describing the interactions of individuals with physicians, hospitals and other providers. It is a significant challenge to utilize the diagnostic information contained in these administrative systems, to meaningfully describe health status. The Adjusted Clinical Group (ACG) system [6] developed by the Johns Hopkins University is used in the British Columbia Ministry of Health Services to provide a conceptually simple, statistically valid, and clinically relevant measure of population health status [7]. Measures of burden of illness and prevalence of co-morbidity of the population are also derived using the ACG system.
Ecological setting
The ecological setting of a population describes the social and physical environmental features of the population. Routinely collected census data [8] can be used to inform these broader perspectives. The advantage of the GIS frame is that once established, other ecological features pertinent to our clearly defined cohort can be examined. The "layering" of information in this way enables a wider context than previously feasible.
As illustrative examples, socio-economic characteristics in the ecological setting are important health determinants and thus determinants of demand for health services. The national census process collects a wide array of information that is made available for analysis at aggregated levels (in this case census tracts). As described, those postal codes falling within the catchment area are spatially referenced to the census tracts within the same catchment area. The census variables associated with these census tracts, including education, income, family structure, and population living alone, are derived to describe the socio-economic characteristics of the census tracts associated with these clients residing in these areas. The choice and configuration of variables for consideration can be customized depending on their importance in the population of interest.
Service distribution and outcomes
The existing use of services to the population of concern, are also specifically described. In British Columbia there are administrative databases that collect information about the wide array of publicly funded health services. The use of these services across the continuum of care, by the specific population of interest, provides an important view of the resources allocated. This information can now be framed with health determinants, risks and health status as well as place. Programs such as clinical services map their services to the community and understand the distribution relative both to some measures of need and other services. As evaluation is undertaken it is now appropriately anchored in an understanding of the population and outcomes can be examined through the development of more detailed models developed in this context.
Results
The following are some sample results from work that was done to support health and social service planning for the residents of the downtown core of the downtown City A, British Columbia, Canada. Routinely available health information is reported at a "Local Health Area" level that would indicate that this area is part of a region characterized by a relatively affluent, older population with good health status. The reported experience of the care providers in this neighbourhood differs greatly from the "average" presented in these routinely available health statistics. Their anecdotal accounts describe people of all ages with complex problems and poor social circumstances and health status. Our team was asked to help delineate and quantify the needs of the population of this neighbourhood to enable specific service planning and evaluation. The specific information presented to the care providers was used as a component of a needs assessment that continues to inform planning for a continuum of health and social services in this downtown neighbourhood. This information base has supported decisions such as the location of services, composition and numbers of case management teams and spectrum of services required.
Population identification
Figure 5 displays the geography of City A. Two layers of spatial information, average income in year 2000 for population aged 15+ by the census tracts and major streets, are overlaid. This thematic map shows a clear pattern of the distribution of annual income by the census tract and indicates the core area of downtown of the City A with the lowest average income for individuals aged 15 years over. This area fits with the general catchment described by the care providers and is thus chosen as the catchment area of this project.
Figure 5 Average Income for Population Aged 15+ by 2001 Census Tracts, City A, British Columbia, Canada. The average income for population aged 15+ in year 2000 is displayed in different color by census tract in City A, British Columbia, Canada. The center of the City A, labeled by Community V, presenting the lowest income level, is defined as the catchment area of this study.
Figure 6 displays the distribution of clients by their home address postal codes within the core area of Downtown City A (Census tract CT010.00), as described previously. There are 6,479 clients identified through this method.
Figure 6 Distribution of Clients by Postal Codes within the Core of Downtown City A (Total Clients = 6,479). The postal codes located within the Census Tract (CT 010.00), where the center of City A is located, are spatially queried and extracted. The queried postal codes are then matched to the residential postal codes of clients who reside in the same Census Tract. As a result, a total 6,479 clients are identified within the core of Downtown City A, and will be the population of this study. Demographic information and spatial distribution of these individual clients can be further analyzed.
Population profile
The demographic characteristics of this population are examined in comparison to the larger region (the entire region where City A is located). As shown in Figure 2, there is a clear difference in age distribution. Compared to the region as a whole, the service population has significant higher population at age 25–34 years (20.6% vs. 13.1%), and higher population at age groups of 25–34 and 35–44. Further analysis shows different distribution of population by gender. Males dominate in age groups 25–64 years, while there are more females in age groups 01–24 and 65–75+. For the senior population (aged 75+), females nearly triple males.
Figure 2 Downtown Core Area of City A versus the Regional Total: Comparison of Clients by Age and Gender, Year 2002. The HGIS, as a platform, integrates information on Census income, postal codes, and clients in Downtown core area of City A. The spatial analytical functions of the HGIS are used to identify and extract actual clients resided in the area (as displayed in Figure 6). These clients are then compared to total population of the region, to identify distinguishing demographic characteristics of the study population.
The demographic features of the population, described above, signal that this population will have an entirely different set of health needs, and thus different requirements for service delivery, compared with the larger region described by the routine administrative boundary.
With the cohort defined, the postal codes can be used to link to administrative health databases to develop a profile of health characteristics. As well as examining the prevalence of individual health problems, it is important to identify the number of people who have multiple problems and the need for more complex care. This population has been stratified based on the number of health conditions associated with the individuals. This is estimated based on the number of Adjusted Diagnosis Groups and further categorized into three groups. These groups are then described by the associated resource use (Figure 3).
Figure 3 Percent of Clients by Number of ADGs, Downtown Core Area, fiscal year 2001/2002. The study population, identified and extracted through HGIS, is further analyzed through the Johns Hopkins ACG system. The number of ADGs (Adjusted Diagnostic Group) per capita is a method of describing the co-morbidity of a population. This figure indicates that 223 clients out of the study population (6,479 in total) have 10 or more reported health conditions, while majority of them (4,013) possesses 4–9 co-morbidities.
Service utilization
For this small neighbourhood there are over 200 people (Figure 3) with more than 10 diagnoses and significant hospital and physician service use (Figure 4). Interventions that are specifically disease focused will be unlikely to meet the health care needs of this group. The profile of services will also allow quantification of the resource allocation for the spectrum of health services providing a valuable support for planning and evaluation. The services supplied by the existing downtown clinic were also mapped to the population to examine issues of access.
Figure 4 Services Utilization by Number of ADGs for Clients of Downtown Core Area, City A, fiscal year 2001/2002. The medical service utilization of the study population is analyzed to identify the pattern of the utilization in relation to their co-morbidities.
The original plan to provide additional services based on individual program delivery in the form of a stand-alone mental health center, for example, was reevaluated based on this finding of high co-morbidity. Expanded primary health care service models [9,10] are being explored to provide more patient-based care, including case management, as a direct result of the information provided by this analytical approach.
This information now serves as a foundation for service planning based on a broad base of information about the population to be served and allows quantification of need. The service team with the responsibility for providing care for this population can continue to examine issues of pertinence and to iteratively explore information about the population, the health services and their interface. With the geographic frame it is now possible to undertake neighbourhood specific, detailed analysis of data from a variety of sources. Thus the unique functionalities of GIS enable the ongoing development and maintenance of a spatially enabled information base, which allows consideration of a broad array of data sources and pertinent information products beyond those traditionally used in health service planning. As programs are developed, the data gathered and the further analyses generated, will feed into this information repository and allow evaluations of impact and reach that are strengthened through rich context.
Discussion and Conclusion
Three important components of a health service planning and evaluation framework grounded in population health theory have been described:
Firstly, a stepwise methodological process is delineated to enable the incorporation of the principles of a population health into practical applications.
Secondly, the technical functionality to integrate multiple sources of information, with different levels and scales is described. Geographical information systems provide this capability, building extensible profiles of populations and service delivery in context. GIS also provide unique functionality in the analysis of patterns and trends. The ability to understand groups of people through the development of information profiles unconstrained by the use of administrative boundaries and categories, transforms decision support at multiple levels of the system. GIS also can provide access tools for information products that can be useful to the full range of interested parties, from researchers and analysts to the general public.
Finally, sources of information about the health of the population at the appropriate level are identified to populate this frame. Epidemiological information from surveys and registries is not available to support population health diagnosis at the level at which service must be planned, often different from the levels delineated by existing boundaries. Through the use of the Johns Hopkins ACG system, administrative data can be sorted into meaningful information to contribute to the diagnosis of the health needs of a defined small population.
The optimal use of GIS in health applications requires the ongoing exploration of non-traditional data sources. As more administrative data becomes available, through initiatives such as the electronic health record, the utility and promise of tools such as geographical information systems will be realized. Policy and governance issues, including consideration of data access and the protection of privacy, require specific attention if these evidence-based decision support methods are to be appropriately implemented.
The population health approach incorporates the consideration of health determinants and the context within which the health conditions arise in communities. The complexity of these relationships requires the application of innovative methodologies such as Health GIS to frame the issues practically. A population health based foundation for the planning and evaluation of health services can now move from theory to practice.
Authors' contributions
These authors contributed equally to this work.
Figure 1 The Population Health Approach. The diagram depicting the Population Health Approach, proposed by the Population and Population Health Branch, Health Canada, July 2001.
==== Refs
Health Canada The Population Health Approach 2001
Lalonde M A new perspective on the health of Canadians (a working document.) Government of Canada, Ottawa
Frankish CJ Paluck E Williamson D Milligan CD Linking Health Research and Decision Making: The Use of Population Health & Health Promotion Research by Regional Health Authorities A Report to the National Health Research Development Program Health Canada 1999
The Canadian Centre for the Analysis of Regionalization and Health: What is Regionalization?
Boulos MN Towards evidence-based, GIS-driven national spatial health information infrastructure and surveillance services in the United Kingdom Int J Health Geogr 2004 3 1 14748927 10.1186/1476-072X-3-1
The Johns Hopkins University ACG System
Reid Robert J MD,PhDMacWilliam Leonard MSc,MNRNVerhulst Lorne MD,MPARoos Noralou PhDAtkinson Michael BSc Performance of the ACG Case-Mix System in Two Canadian Provinces Medical Care 2001 39 86 99 11176546
Donohoe MT Profile of All Levels of Geography in Canada 2001 Census 2005 Ottawa, ON: Statistics Canada (Catalogue No. 95F0296XCB)
Comparing generalist and specialty care: discrepancies, deficiencies, and excesses Arch Intern Med 1998 158 1596 1608 9701093 10.1001/archinte.158.15.1596
Bodenheimer T Wagner EH Grumbach K Improving primary care for patients with chronic illness JAMA 2002 288 1775 1779 12365965 10.1001/jama.288.14.1775
|
16209717
|
PMC1274340
|
CC BY
|
2021-01-04 16:39:06
|
no
|
Int J Health Geogr. 2005 Oct 6; 4:23
|
utf-8
|
Int J Health Geogr
| 2,005 |
10.1186/1476-072X-4-23
|
oa_comm
|
==== Front
J NeuroinflammationJournal of Neuroinflammation1742-2094BioMed Central London 1742-2094-2-221621266410.1186/1742-2094-2-22ResearchFocal glial activation coincides with increased BACE1 activation and precedes amyloid plaque deposition in APP[V717I] transgenic mice Heneka Michael T [email protected] Magdalena [email protected] Lucia [email protected] Ilse [email protected] Jochen [email protected] Thomas [email protected] Leuven Fred [email protected] Department of Neurology, University of Münster, 48149 Münster, Germany2 Department of Neurology, University of Bonn, 53127 Bonn, Germany3 Experimental Genetics Group, Dept Human Genetics, K.U.Leuven, B-3000 Leuven, Belgium2005 7 10 2005 2 22 22 2 5 2005 7 10 2005 Copyright © 2005 Heneka et al; licensee BioMed Central Ltd.2005Heneka et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Inflammation is suspected to contribute to the progression and severity of neurodegeneration in Alzheimer's disease (AD). Transgenic mice overexpressing the london mutant of amyloid precursor protein, APP [V717I], robustly recapitulate the amyloid pathology of AD.
Methods
Early and late, temporal and spatial characteristics of inflammation were studied in APP [V717I] mice at 3 and 16 month of age. Glial activation and expression of inflammatory markers were determined by immunohistochemistry and RT-PCR. Amyloid deposition was assessed by immunohistochemistry, thioflavine S staining and western blot experiments. BACE1 activity was detected in brain lysates and in situ using the BACE1 activity kit from R&D Systems, Wiesbaden, Germany.
Results
Foci of activated micro- and astroglia were already detected at age 3 months, before any amyloid deposition. Inflammation parameters comprised increased mRNA levels coding for interleukin-1β, interleukin-6, major histocompatibility complex II and macrophage-colony-stimulating-factor-receptor. Foci of CD11b-positive microglia expressed these cytokines and were neighbored by activated astrocytes. Remarkably, β-secretase (BACE1) mRNA, neuronal BACE1 protein at sites of focal inflammation and total BACE1 enzyme activity were increased in 3 month old APP transgenic mice, relative to age-matched non-transgenic mice. In aged APP transgenic mice, the mRNA of all inflammatory markers analysed was increased, accompanied by astroglial iNOS expression and NO-dependent peroxynitrite release, and with glial activation near almost all diffuse and senile Aβ deposits.
Conclusion
The early and focal glial activation, in conjunction with upregulated BACE1 mRNA, protein and activity in the presence of its substrate APP, is proposed to represent the earliest sites of amyloid deposition, likely evolving into amyloid plaques.
==== Body
Background
Alzheimer's disease (AD) is a neurodegenerative disorder that is characterized by progressive memory loss and decline of cognitive functions. Histopathological hallmarks include extracellular amyloid peptide (Aβ) deposition in neuritic plaques, and intracellular deposits of hyperphosphorylated Tau, causing formation of neurofibrillary tangles and finally neuronal death. Aβ peptides are generated from amyloid precursor protein (APP) by sequential actions of two proteolytic enzymes, i.e. the β-site APP cleavage enzyme (BACE1) and the γ-secretase [1,2]. Their formation and eventual deposition represents a key feature and possibly the triggering mechanism of AD. The importance of Aβ formation was instigated by dominantly inherited familial forms of AD that are linked to APP mutations in or close to the β- and γ-secretase cleavage sites [3]. This made it possible to generate transgenic mouse models of cerebral amyloidosis and AD-like histopathology, i.e. amyloid plaques and cerebral amyloid angiopathy (CAA) [4-6](3–8) [7,8].
The eventual deposition of Aβ and the neurofibrillary tangle formation may not account for all, and particularly not for the most early clinical symptoms in AD. Inflammatory changes are observed in AD brain overall, and particularly at the amyloid depots, invariably comprising activated microglia [9,10]. Once stimulated by beginning neuronal degeneration, microglia releases, a wide variety of pro-inflammatory mediators including cytokines, complement components, various free radicals and nitric oxide (NO), which all contribute to further neuronal dysfunction and eventually death. These create and feed a vicious cycle that could be essential in the pathological progression of AD [11]. Apart from any direct effects of microglial inflammation, the recruitment of astrocytes that assemble around and in amyloid plaques are likely to prolong the ongoing inflammation.
In addition to histopathological and biochemical data, several proinflammatory genes have been linked to an increased risk for AD, including interleukin1 (Il-1) [12], interleukin 6 (Il-6) [13] and tumor necrosis factor alpha (TNFα) [14]. The hypothesis that inflammatory changes actively contribute to AD pathogenesis is further supported by epidemiological data, i.e. long term medication with non-steroidal anti-inflammatory drugs (NSAIDs) appears to decrease the risk, delay the onset and slow the cognitive decline of AD patients [15-17].
The finding that cytokines are able to transcriptionally upregulate BACE1 mRNA, protein and enzyme activity levels and thereby increase total and fibrillogenic Aβ peptides in cell-biological models [18] prompted us to test the hypothesis that BACE1 is related to age-dependent parameters of inflammation in vivo, i.e. in the brain of APP [V717I] transgenic mice. The data presented are an important extension of the phenotypic characterization of APP [V717I] mice which recapitulate not only the amyloid [6] and cerebrovascular angiopathy [7] but various aspects of neuroinflammation. Moreover, they indicate that early and focal inflammation may feedback stimulate local APP processing via BACE1 and these sites therefore possibly represent the birthplaces of plaques.
Methods
Animals
Transgenic mice expressing APP [V717I] under the mouse thy1 gene promoter in the FVB/N genetic background [6] aged 3 and 16 months were used in this study with non-transgenic mice of the same genetic background, gender and age as controls. At the time of sacrifice, animals were anesthetised and transcardially perfused with heparinized sodium chloride (0.9%), brains were removed and several regions including frontal cortex and cerebellum dissected from one hemisphere using the mouse brain atlas coordinates [19]. Dissected sections were snap frozen in liquid nitrogen and stored at -80°C until analysis. The remaining hemisphere was fixed either in 4% paraformaldehyde followed by paraffin embedding or underwent cryofixation under tissue protection with tissue frezzing medium (Leica Instruments, Nussloch, Germany) according to standard protocols, before sectioning for immunohistochemistry. Animal care and handling was performed according to the declaration of Helsinki and approved by local ethical committees (approval #50.203.2BN 33,34/00).
Immunohistochemistry
Serial sagittal sections were cut (7 μm) from parrafin embedded tissue (Leica microtome RM2155) and mounted (Histobond adhesion slides, Marienfeld, Germany). Retrieval of antigen sites, blocking of endogenous peroxidase activity and blocking of non-specific binding sites was performed according to standard protocols. For immunostaining of paraffin-embedded tissue, sections were incubated overnight at 4°C with the following primary antibodies: 1) mouse mAb against GFAP, #MAB360 (1:800, Chemicon, Hofheim, Germany). 2) rabbit pAb against iNOS, 32030 (1:150, Transduction Laboratories, Heidelberg, Germany). 3) rabbit pAb against Aβ1–42, #44–344 (1:40, Biosource International, USA.). Immunohistochemical localization was performed using the avidin-biotin peroxidase complex method (ABC-Kit, Vector Laboratories, Burlingame, USA) with 3,3'-diaminobenzidine tetrahydrochloride as chromogen. For costaining in paraffin tissue of GFAP and Aβ1–42, slides were washed twice in PBS and blocked in 20% normal goat serum. After incubation with the primary antibody for 20 h slides were washed and incubated with biotinylated goat anti rabbit IgG. Immunohistochemical localisation was detected as described above using Vector-blue as substrate (Vector-blue substrate kit, Vector Laboratories, Burlingame, USA).
All other single or double immunostaining was performed on cryofixed sections cut (6 μm) and mounted as described above. Sections were dried at RT for 1 h and then fixed in 4% PFA or methanol for 15 min at RT. After washing with PBS the double staining was performed by adding simultaneously both first antibodies and followed by overnight incubation at 4°C. In addition to the above decribed antibodies the following antibodies were used: 4) rat mAb #MCA 711 against murine CD11b (CD11b, 1:250, Serotec Düsseldorf, Germany). 5) rat mAb against Il-1β, MAB401 (1:50, R&D Systems, Wiesbaden-Nordenstadt, Germany). 6) goat pAb against IL 6, M12 sc1265 (1:200, Santa Cruz, Biotechnology Inc., Heidelberg, Germany). 7) 7520 rabbit pAb against the C-terminal domain of BACE1 (gift from Dr. Christian Haass, Adolf-Butenandt-Institute, University of Munich). 8) mouse mAb anti nitrotyrosine # 05–233 (1:40, Upstate Inc., Biomol, Hamburg) 9) rabbit pAb GFAP, Z334 against glial fibrillary acidic protein (1:800, DAKO, Hamburg, Germany). 10) mouse mAb # MAB 377 against neuronal nuclei (neuN, 1:500, Chemicon, Hofheim, Germany). The goat secondary antibodies (Fluorescein DTAF conjugated anti rabbit 1:150, Texas Red conjugated anti mouse 1:80, Texas Red conjugated anti rat 1:80, Jackson Immuno Research Laboratories, West Grove, USA) were applied sequentially after washing in PBS. Negative controls included non-specific IgG instead of primary antibodies; pre-absorption with respective cognate peptides (150–200 μg of peptide/ml of antibody working solution), omission of the secondary antibody and absence of immunoreactivity in non-transgenic controls of the respective age.
Confocal laser scanning microscopy
Double-labeled specimens were analyzed with a confocal laser scanning microscope (Multiprobe 2001; Molecular Probes, Inc., Eugene, OR) equipped with an Ar/Kr laser with balanced emission at 488, 568, and 647 nm. Images were aquired at a 40 × magnification to ensure a high quality resolution of microglial cells. To achieve an optimal signal-to-noise ratio for each fluorophore, sequential scanning with 568 and 488 nm was used. The digitalized images were then processed with ImageSpace 3.10 software (Molecular Probes, Inc.) on a Silicon Graphics (Mountain View, CA) power series 310GTX work station. Original section series were subjected to Gaussian filtration to reduce noise and enhance weakly but specifically labeled parts. Original and filtered sections were projected on one plane using a maximum-intensity algorithm and in some cases using depth-coding and surface-rendering algorithms.
Thioflavine-S staining
Thioflavine-S staining consisted of reacting section in 0.015% aqueous thioflavine-S for 10 min, followed by differentiation in 50% ethanol, rinsing in water, air draining and clarification into xylene. Thereafter slides were covered and evaluated under fluorescent lighting using UV filtration and a standard microscope (Nikon, Eclipse E-800).
Quantification of immunohistochemistry
For quantitative image analysis of hippocampal and cortical immunostaining, serial sagittal sections taken from lateral (+0.5–+2.25) were examined. iNOS, GFAP and CD11b staining cells as well as Aβ1–42-positive neuritic plaques were counted on sections of 6 animals per group. Antigens were detected in 10 parallel sections with defined distance of 70 μm showing both the hippocampus and cortex. In each section, 20 randomly choosen fields were evaluated. Cell number was determined using a counting grid at 20 × magnification and given as calculations of square millimeters. Images were aquired using a standard light and immunofluorescence microscope (Nikon, Eclipse E-800) connected to a digital camera (SONY, model DXC-9100P, Köln, Germany) and to a PC system with LUCIA imaging software (LUCIA 32G, version 4.11; Laboratory Imaging, Düsseldorf, Germany). Data were analysed by ANOVA with Tukey's post test using SYSTAT (Systat, Evanston, U.S.A.).
RNA preparation and RT-PCR
Brain sections from frontal cortex and cerebellum were dissected and RNA extracted from using Trizol reagent as recommended by the manufacturer (Sigma, St. Louis, MO), followed by RT-PCR. The primers were: iNOS forward 5'-TGGGAGCCACAGCAATATAG-3' and iNOS reverse 5'-ACAGTTTGGTGTGGTGTAGG-3'; GFAP forward 5'-TCCGCGGCACGAACGAGTC-3' and GFAP reverse 5'-CACCATCCCGCATCTCCACAGTCT-3'; MCSF-R forward 5'-GACCTGCTCCACTTCTCCAG-3' and MCSF-R reverse 5'-GGGTTC AGACCAAGCGAGAAG-3'; MHCII forward 5'-CTGATGGCTGCTCATCCTGTGC-3' and MHCII reverse 5'-TTCTGTTTTCTGTATGCTGTCC-3'; IL-1β forward 5'-CCTGTGTAATGAAAGACGGC-3' and IL-1β reverse 5'-AAGGGA GCTCCTTCACA TGC-3'; GAPDH forward 5'-TCACCAGGGCTGCCATTTGC-3' and GAPDH reverse 5'-GACTCCACGACATACTCAGC-3'; IL-6 forward 5'- CAGAAA CCGCTATGAAGT TCC-3' and IL-6 reverse 5'-TGTACTCCAGGTAGCTATGG-3'. TGF-β1 forward 5'-CAAGTGTGGAGCAACATGTG-3' and TGFβ-1 reverse 5'-CACAGCAGTTCTTCTCT GTG-3', BACE1 forward 5'-CCGGCG GGAGTGG TATTATGAAGT-3' and BACE1 reverse 5'GATGGTGATGCGGAAGGACTGATT-3'. PCRs were carried out on RNA from n = 6 animals in each group, and representative gels of 2 animals per group are shown. PCR conditions were 35 cycles (iNOS, GFAP, IL-1β, TGF-β1, IL-6, MHCII, MCSF-R, BACE1) and 24 cyles (GAPDH) of denaturation at 95°C for 30s; annealing at 63°C for 45s, and extension at 72°C for 45s using a PX2 (ThermoHybaid, Ulm, Germany). PCR products were separated by electrophoresis through 2% agarose containing 0.5 μg/ml ethidium bromide and imaged using an AlphaInotech imaging system (Temeculah, USA).
Determination of Aβ
Frontal cortex from transgenic mice were homogenized in RIPA buffer (1% Triton, 1% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 50 mM Tris-HCl, pH 7.2) using an Ultraturrax T25 (Janke&Kunkel, IKA-Labortechnik). Aβ was immunoprecipitated from 100 μg protein using antibody 2964 and protein A beads (Amersham Pharmacia, Freiburg, Germany), separated on 10–20% Tris tricine gels (Anamed, Darmstadt, Germany) and transferred onto nitrocellulose membranes. Aβ was detected by immunoblotting with antibody 6E10 (Signet labs Inc, Dedham, MA).
Determination of BACE activity
The enzymatic activity of BACE1 was measured in membrane extracts from frontal cortex by fluorimetric reaction as suggested by the supplier (BACE activity kit FP002, R&D Systems, Wiesbaden, Germany). In addition, BACE1 activity was determined in situ using serial cryosections. Sections were stored at -70°C and immediately before analysis kept at -20°C for 15 min and 4°C for 10 min. Thereafter sections were incubated at 4°C in PBS plus 0.4% TritonX for 30 min. After addition of 5 μl of fluorgenic BACE1 substrate and 100 μl of 1x substrate buffer, sections were incubated at 37°C for 1 hr. Then, sections were rinsed in PBS and mounted with Mowiol4-88 (Calbiochem, San Diego, CA, USA). BACE1 activity was visualized using a DAPI filter set (Ex. 340–380, Emis:435-485) and a standard light and immunofluorescence microscope (Nikon, Eclipse E-800) connected to a digital camera (SONY, model DXC-9100P, Köln, Germany) and to a PC system with LUCIA imaging software (LUCIA 32G, version 4.11; Laboratory Imaging, Düsseldorf, Germany). Addition of a BACE1 inhibitor served as control as previously described [20]. Parallel sections were used to detect GFAP immunostaining as described above. Computational overlay analysis was employed to estimate the colocalisation of BACE1 activity/GFAP expression.
Quantification of RT-PCR and immunoblot results
RT-PCR was quantified by densitometry of at least 6 animals per age. Band intensities were determined using Image-J software (NIH). Data were analyzed by ANOVA with Tukey's post test (Systat, Evanston, U.S.A.).
Results
Brain amyloid plaque load was determined in APP [V717I] mice and completely in line with previous studies [6,7]. Amyloid plaques were undetectable by Thioflavin-S or Aβ immunostaining in brains of APP [V717I] mice at 3 months of age but were abundantly present in 16 month old transgenic mice (Fig. 1A). By western blotting, amyloid peptides were evidently detected in brains of APP [717I] mice at both ages (Fig. 1B). In parallel, age-dependent inflammatory changes were assessed in the frontal cortex and hippocampus by immunohistochemistry for CD11b and GFAP, as markers for microglial and astrocytic activation, respectively. In 3 month old wild type controls, clustered CD11b was undetectable but labelled uniformly distributed ramified microglia (Figure 1A). In contrast, brains of APP [V717I] mice showed a focally activated CD11b immunostaining already at 3 months. Microglial morphology identified different activation states, but only round or oval appearing cells were quantified and counted as being "activated" in the hippocampus and frontal cortex (Figure 1A, see insert). In APP [V717I] of 16 months, an even more pronounced excess of activated microglia was obvious in both brain areas (Figure 1B). In keeping with microglial activation, cortical GFAP-immunostaining was practically absent in non-transgenic control mice at 3 months (data not shown), whereas APP [V717I] transgenic mice had randomly distributed foci of astrocytes strongly expressing GFAP within the cortex and hippocampus at that age. While the majority of GFAP-positive foci appeared to be randomly distributed within the cortex and hippocampus, some of these GFAP postive foci were found to surround brain vessels. Quantification of GFAP-positive cells (Figure 1B) demonstrated an even greater increase in the number of activated astrocytes at 16 months compared to age-matched non transgenic mice.
Figure 1 Comparison of Aβ deposition, micro- and astroglial activation. (A) Representative detection of Aβ1–42 immunostaining, Thioflavin-S histochemistry, microglial (CD11b) and astroglial activation (GFAP) in APP [V717I] mice and non-transgenic controls of the identical genetic background at 3 (3 m) and 16 (16 m) months (Bar graph = 50 μm (Aβ1–42, Thioflavin-S), = 25 μm (CD11b, GFAP)) Focal microglial activation is indicated by black arrows. Focal astroglial activation within the parenchyma by black arrows and at the side of of a brain vessel by a white arrow (B). Quantification of hippocampal (HC, open bar) and cortical (FC, filled bar) Aβ1–42-positive plaques of APP [V717I] mice at 3 and 16 months (tg 3 m, tg 16 m) (n = 12, ANOVA followed by a TUKEY test, **p < 0.01, ***p < 0.001.) and total Aβ detection by immunoprecipitation/western blot and subsequent quantification by densitometry (n = 3, Students t-test, *p < 0.05). (C) Quantification of CD11b positive, activated microglia (see insert, arrows) and GFAP positive astrocytes in the hippocampus (HC, open bar) and frontal cortex (FC, filled bars) (n = 12, ANOVA followed by a TUKEY test, *p < 0.05, **p < 0.01,***p < 0.001).
Confocal analysis of immunostaining for CD11b in combination with Il-1β (Figure 2A) or Il-6 at (Figure 2A) at 3 month demonstrated that microglia already produced both cytokines in young APP [V717I] transgenics. Similar results were obtained by double staining for CD11b and MHCII or MCSF-R (not shown). In brains of 16 month old APP [V717I] transgenic mice, Cd11b positive and activated microglia cells were predominantly associated with amyloid plaques as revealed by co-staining with Aβ1–42 (Figure 2B). Further analysis demonstrated that these microglial cells also expressed Il-1β, Il-6 (Figure 2B), MCSF-R and MHC II (not shown). The mRNA coding for Il-1β, Il-6, MHC II and MCSF-R were already detectable in frontal cortex brain lysates of 3 month old APP [V717I] transgenic mice, while absent in non-transgenics (data not shown) and most significantly increased in the brain of old APP [V717I] transgenic mice at 16 months (Figure 2C). Several other cytokines, i.e. tumor necrosis factor alpha, interferon gamma, interleukin-10 and interleukin-4 were undetectable at either age (results not shown). In contrast, TGFβ-1 mRNA levels showed an inversed pattern with signifcantly decreased levels in the brain of 16 month old, relative to young APP [V717I] transgenic mice (Figure 2C).
Figure 2 Characterisation of microglial inflammation. (A) Representative confocal immunohistochemistry of APP [V717I] mice revealed that Il-1β and Il-6 colocalized with activated CD11b-positive microglial cells at 3 month. (B) At 16 month CD11b positive cells were almost exculsively detected in close proximity to Aβ1–42 positive plaques. At this time point, CD11b-positive and plaque associated microglia were also found to be colocalized with Il-1β and Il-6. (C) RT-PCR analysis was performed with frontal cortex brain lysates and is being displayed from two single animals at each age (3 and 16 month) for Il-1β, Il-6, MCSF-R, MHCII and TGFβ-1 and showed increased gene transcription at 16 months. (D) Densitometry of PCR products of APP [V717I] mice at 3 (open bars) and 16 months (filled bars) for the indicated inflammatory molecule. RT-PCR for GAPDH served as control. (n = 6, ANOVA followed by a TUKEY test, ***p < 0.001. Bar graphs in A-B are = 25 μm).
Analysis of astroglial activation by double staining for Aβ1–42 and GFAP demonstrated that GFAP-positive cells were mostly located around amyloid plaques in the brains of aged transgenic mice (Figure 3A). At this age a subset of plaque-associated astrocytes was immunopositive for iNOS in both the hippocampus and the frontal cortex (Figure 3B, C). Confocal staining for GFAP and iNOS confirmed that iNOS positive cells were astrocytes (not shown), and demonstrated their close spatial relation to amyloid plaques (Figure 3A). Additionally, co-staining for nitrotyrosine and Aβ revealed an increased NO-dependent peroxynitrite generation in close proximity to the amyloid plaques (Figure 3A). This result was paralleled by increased iNOS and GFAP mRNA levels in brain of 16 month old APP [V717I] mice (Figure 3B, D). In brains of non-transgenic mice, the iNOS mRNA was not detectable (data not shown). Remarkebly, activated microglial and astrocytic cells were colocalized as demonstrated by double staining for CD11b and GFAP, already in the brain of young APP [V717I] mice, suggesting the formation of inflammatory foci in both brain regions evaluated (not shown).
Figure 3 Astrocytic iNOS expression and plaque associated nitrotyrosine. (A) Costaining of Aβ1–42 and GFAP at 16 months detected activated astrocytes nearby Aβ plaques. Astrocytic iNOS and confocal staining of iNOS (red) and Aβ1–42 (green) or nitrotyrosine (red) and Aβ1–42 (green). (B) RT-PCR for GFAP and iNOS in APP [V717I] mice at 3 (3 m) and 16 months (16 m) of age. (C) Quantification of iNOS-positive astrocytes in the hippocampus (HC, black bar) and frontal cortex (FC, hatched bars) of APP transgenic mice (tg) and wild type controls (wt) at 3 and 16 months. (D) Densitometry of GFAP, iNOS and GAPDH mRNA from APPV [7171I] mice at 3 (black bars) and 16 months (hatched bars). (E) Confocal staining of CD11b positive microglia and GFAP labelled astrocytes showed that both cells were located in close neighbouring in APP [V717I] mice at 3 month of age. (n = 6, ANOVA followed by a TUKEY test, n.s. = non significant, ***p < 0.001). Bar graph = 50 μm.
Since we demonstrated that cytokine stimulated neuronal cells increased production of Aβ by transcriptional BACE1 up-regulation in vitro [18], and the latter cytokines were detectable at sites of early inflammation in young APP [V717I] mice, we next analysed whether early inflammatory foci would be accompanied by BACE1 expression. Co-staining for BACE1 and neuN demonstrated that neurons expressed BACE1 in the 3 month old APP [V717I] mice throughout the cortex and hippocampus, confirming a previous observation in another transgenic mouse model [21] (data not shown). Despite the fact that BACE1 was expressed widely, a clear and focal upregulation of neuronal BACE1 immunostaining was observed in brain of APP [V717I] transgenic mice at 3 months of age. Costaining for CD11b and BACE1 or for GFAP and BACE1 showed that the upregulation was predominantly confined to neurons which were located in close proximity to CD11b positive microglia (Fig. 4A). The neuronal nature of BACE expressing cells was further confirmed by confocal immunostaining for the neuronal marker neuN and BACE 1 (Figure 5). Subsequent quantification of BACE1 expressing neurons confirmed that the highest number of BACE1 positive cells were in close distance to both CD11b and GFAP activated micro- and astroglial cells (Figure 4B).
Figure 4 Sites of focal and early inflammation show BACE1 upregulation in neurons. (A) Representative confocal immunostaining of CD11b positive microglia and BACE1 and GFAP and BACE1 in 3 month old APP transgenics showed that BACE positive neurons were found close to focally activated microglia cells in 3 month old APP [V717I] mice. (B) Quantitation of the number of BACE1 positive cells in relation to the distance to CD11b or GFAP positive cells. (C) Representative image of focal GFAP expression, BACE1 activity and overlay in APP [V717I] mice at 3 m of age. (D) Measurement of BACE1-activity was calculated as percentage of 3 month old controls (wt 3 m) and showed that enzyme activity was already elevated in APP [V717I] mice at 3 month (tg 3 m) (n = 5, ANOVA followed by a TUKEY test, *p < 0.05). (E) RT-PCR detection of BACE1 mRNA levels of cortical (frontal cortex, FC) and cerebellar (Cb) lysates from wild type controls (wt), APPV [7171I] (tg) mice at 3 (3 m) and 16 months (16 m). (F) Densitometrical analysis and quantitation of BACE1 mRNA levels of frontal cortex lysates of APP [V717I] transgenic and controls at the respective age (n = 6, ANOVA followed by a TUKEY test, *p < 0.05). Bar graphs are = 50 μm for CD11b/neuN and GFAP/BACE and = 25 μm for CD11b/BACE1).
Figure 5 Expression of BACE1 in neurons in APP [V717I] transgenic mice at 3 month of age. Representative confocal immunostaining of BACE1 positive cells and neuN positive neurons in the cortex of 3 month old APP [V717I] mice. Bar graphs are = 50 μm for BACE1, neuN and BACE1/neuN.
In situ fluorescence detection of BACE1 activity revealed that sites of increased BACE1 activation colocalised to GFAP positive and activated astrocytes (Figure 4C). Addition of a previously described BACE1 inhibitor served as control [20] and abrogated the signal (data not shown). Quantitative determination of BACE1 activity from cortical lysates showed that BACE1 enzyme activity was significantly increased in brains of 3 month old APP [V717I] mice when compared to controls and did not further increase at 16 month (Figure 4D). This phenomenon was paralleled by increased BACE1 mRNA levels in the frontal cortex, whereas at the same time cerebellar BACE1 mRNA levels did not reveal any significant regulation (Figure 4E, F). Combined, these data indicate the inflammation-associated increase in BACE1 levels in brain of young, 3 month old APP [V717I] mice compared to age-matched non-transgenic mice.
Discussion
In AD, the deposition of amyloid peptides and neurofibrillary tangles are invariably associated with an inflammatory component, mainly characterized by activated microglial cells and astrocytes. Aβ peptides and secreted APPs are potent activators of glia cells [22]. Once activated, micro- and astroglia release a variety of cytokines, chemokines and free radical oxygen species, which can contribute to neuronal dysfunction and death. In addition, some specified glia-derived cytokines may also increase Aβ generation [23]. The finding that several cytokines increase total and fibrillogenic Aβ by transcriptional upregulation of BACE1 mRNA, protein and activity levels [18] suggests a morbid feedback mechanism by which neurodegenerative and neuroinflammatory mechanisms interact. Activated microglia may, however, play a dual role in AD, since clearance of Aβ through phagocytosis [24] may be advantageous. To define the active contribution of inflammation in AD, experimental animal models are needed that recapitulate both the neurodegenerative and the inflammatory components of the disease.
Whereas transgenic mouse models are widely used to study APP processing, only a limited number of studies has addressed neuroinflammation in these animals, yielding in part controversial results. Thus, APP695 transgenic mice aged 2 to14 month failed to reveal mRNA for several cytokines including Il-1α/β, Il-6, Il-10, Il-12 and IFNγ by ribonuclease protection assay [25]. In the same study, however, Il-1β-positive astrocytes were detected in close proximity to amyloid deposits in older mice, whereas immunohistochemistry for TNFα, Il-1α, Il-6, and MCP-1 was negative. In contrast, TNFα mRNA was evident as early as 6 month [26] and IFNγ and Il-12 mRNA and protein was detected by in situ hybridization and immunohistochemistry in 9 month old APP695 transgenic mice [27]. Moreover, Il-1β, TNFα and Il-10 was detected by immunohistochemistry in animals at 12 and 13 month of age [28,29]. The differences reported in the same strain of APP transgenic mice are likely to be caused by different techniques employed and demonstrate the difficulties encountered in assessing inflammatory changes in the brain of this mouse model.
In contrast to these studies, the present work revealed a significant increase in focally activated microglia cells expressing cytokines such as Il-1β and Il-6 already at 3 months, which was paralleled by mRNA levels for Il-1β, Il-6, MHC II and MCSF-R. At this age, these APP [V717I] mice do not yet deposit amyloidogenic Aβ peptides as verified by the complete absence of immunopositive and Thioflavin-S-positive plaques, confirming previous results [6,7]. Microglial foci seemed to be randomly distributed in the cortex and hippocampus of 3 month old APP transgenic mice. However, since total levels of Aβ were already detectable at this age and soluble fragments also act as potent stimulators of microglial cytokine secretion [30], soluble Aβ along with secreted APP [22] may cause this early microglial activation long before amyloidogenic fragments deposit.
It is most interesting to note that the APP [V717I] transgenic mice develop cognitive impairment, decreased long-term potentiation (LTP) and neophobia already at 3 month of age [6]. Importantly, this phenomenon was not correlated with the actual APP isoform expressed nor with the levels of a single APP metabolite [6]. Because cytokines including Il-1β and Il-6 directly impair neuronal function and suppress hippocampal LTP [31,32] the current data allow us to propose that early and focal inflammatory events contribute to neuronal dysfunction at this age. The foci contain moreover all the ingredients needed to generate amyloid peptides and are tentatively identified as "birth-places" of amyloid plaques, resulting from a viscious circle instilled by amyloid peptides and immuno-modulatory factors.
Focal microglia activation was surrounded by GFAP-positive astrocytes in young mice, but GFAP mRNA levels were almost undetectable at 3 month [33]. However, GFAP and iNOS mRNA levels became detectable in transgenic mice at 16 month indicating strong astrocytic activation. Increased GFAP mRNA levels were paralleled by increased numbers of GFAP-positive and iNOS expressing astrocytes. Importantly, iNOS mRNA and protein levels were undetectable in non-transgenic controls and in young APP [V717I] mice. In addition, expression of iNOS in Aβ plaque-associated astrocytes was paralleled by an increase of nitrotyrosine staining indicating enhanced generation of NO dependent peroxynitrite. Because iNOS expression and increased nitrotyrosine staining has been attributed to AD before [34,35], APP [V717I] mice also resemble this aspect of neurodegeneration-induced glial inflammation. In contrast to other cytokines, TGFβ-1 mRNA levels decreased in ageing APP [V717I] mouse brain. Because TGFβ-1 acts mostly as an anti-inflammatory cytokine, age-related loss may facilitate the observed neuroinflammation.
Since we showed most recently that several cytokines, alone and potently in concert, increased Aβ40 and Aβ42 levels by transcriptional upregulation of BACE1 [18], we tested and demonstrated that microglia-derived cytokine generation in early inflammatory foci was accompanied by BACE1 upregulation in brain of young APP [V717I] transgenic mice. At 3 month of age, BACE1 expression was exclusively restricted to neurons confirming studies by in sity hybridisation in Tg2576 and PDAPP mice [36,37]. However, in both major brain regions, i.e. hippocampus and cortex, the increased neuronal BACE1 expression appeared to be clustered. Costaining with CD11b or GFAP and subsequent quantification demonstrated that neuronal BACE1 expression was upregulated in close proximity to activated microglia and astrocytes. Irrespective whether inflammatory mediators or β-site APP-cleavage derived products occur first, the early and focal presence of immunoactive microglia, cytokines and BACE-expressing neurons strongly points to an interaction between neurodegenerative and neuroinflammatory events. In keeping with this finding, hippocampal BACE1 mRNA levels were significantly increased in 3 month old APP [V717I] transgenics compared to non-transgenic mice and this phenomenon was paralleled by strongly increased BACE1 enzymatic activity as determined from brain lysates. In old mice BACE1 expression was also detected in activated astrocytes as observed in Tg2576 mice, but not different from non-transgenic mice [21]. However, the fact that the observed changes of BACE1 RNA levels are higher than those observed for activity; parallels our previous in vitro results [18] and may just indicate that BACE1 activity is regulated not only by gene transcription but at multiple steps thereafter. Interestingly, BACE1 mRNA levels did not significantly change between 3 and 16 month of age in APP [V717I] transgenics. While the current study did not identify the underlying reason of this phenomenon, It can be hypothesized, that irregardeless of higher levels of inflammatory mediators present at 16 month, it is possible that (i) the total spectrum of pro-inflammatory and antiinflammatory mediators is more or equally permissive for BACE 1 upregulation at a very early age, or (ii) the increase of pro-inflammatory cytokines at later ages are accompanied by counteracting anti-inflammatory molecules, resulting in a similar netto induction of BACE1. In addition, several other mechanisms may account for the almost equal levels of BACE1 mRNA at 3 and 16 month of age including a desensitized transcriptional activation, a rebalance between production and degradation of the BACE1 transcript a later age or a lower contribution from disease affected neurons in the close proximity to amyloid plaques.
Conclusion
APP [V717I] transgenic mice do not only model the late amyloid pathology in parenchym and vasculature as in AD patients, but exhibit also many inflammatory parameters ascribed to the AD pathology. The early and focal neuro-inflammatory changes are demonstrated here to be parallelled closely by upregulated neuronal BACE1 mRNA and protein expression and by increased BACE1 enzyme activity, already in young APP transgenic mice, before any amyloid deposition is evident. The vicious cycle of APP proteolytic cleavage giving rise to soluble and amyloidogenic immunostimulators, causing microglial activation, cytokine generation, is closed by the upregulation of BACE1, ultimately enhancing further APP processing. This cycle appears to operate locally, in focal nidi of disease that could represent the birthplaces of amyloid plaques, already present early in the disease process in brain of young APP transgenic mice.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Michael Heneka: conception and design, immunostaining, data aquisition, interpretation, article writing
Magdalena Sastre: conception, BACE1 measurements
Lucia Dumitrescu-Ozimek: Immunostaining, data aquisition
Ilse Dewachter: amyloid determination
Jochen Walter: BACE1 measurements in situ,
Thomas Klockgether: conception and design,
Fred van Leuven: conception and design, data analysis and interpretation
Acknowledgements
This investigation was supported by the Deutsche Forschungsgemeinschaft Collaborative Research Grant (SFB 400, A8) and the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen (FWO-Vlaanderen), the KULeuven GOA-Research Fund and KULeuvenR&D. We thank Christian Haass for generous gift of BACE-1 antibody.
==== Refs
Dingwall C Spotlight on BACE: the secretases as targets for treatment in Alzheimer disease J Clin Invest 2001 108 1243 1246 11696563 10.1172/JCI200114402
Esler WP Wolfe MS A portrait of Alzheimer secretases--new features and familiar faces Science 2001 293 1449 1454 11520976 10.1126/science.1064638
Hardy J Allsop D Amyloid deposition as the central event in the aetiology of Alzheimer's disease Trends Pharmacol Sci 1991 12 383 388 1763432 10.1016/0165-6147(91)90609-V
Hsiao K Chapman P Nilsen S Eckman C Harigaya Y Younkin S Yang F Cole G Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice Science 1996 274 99 102 8810256 10.1126/science.274.5284.99
Sturchler-Pierrat C Abramowski D Duke M Wiederhold KH Mistl C Rothacher S Ledermann B Burki K Frey P Paganetti PA Waridel C Calhoun ME Jucker M Probst A Staufenbiel M Sommer B Two amyloid precursor protein transgenic mouse models with Alzheimer disease-like pathology Proc Natl Acad Sci U S A 1997 94 13287 13292 9371838 10.1073/pnas.94.24.13287
Moechars D Dewachter I Lorent K Reverse D Baekelandt V Naidu A Tesseur I Spittaels K Haute CV Checler F Godaux E Cordell B Van Leuven F Early phenotypic changes in transgenic mice that overexpress different mutants of amyloid precursor protein in brain J Biol Chem 1999 274 6483 6492 10037741 10.1074/jbc.274.10.6483
Van Dorpe J Smeijers L Dewachter I Nuyens D Spittaels K Van Den HC Mercken M Moechars D Laenen I Kuiperi C Bruynseels K Tesseur I Loos R Vanderstichele H Checler F Sciot R Van Leuven F Prominent cerebral amyloid angiopathy in transgenic mice overexpressing the london mutant of human APP in neurons Am J Pathol 2000 157 1283 1298 11021833
Lamb BT Bardel KA Kulnane LS Anderson JJ Holtz G Wagner SL Sisodia SS Hoeger EJ Amyloid production and deposition in mutant amyloid precursor protein and presenilin-1 yeast artificial chromosome transgenic mice Nat Neurosci 1999 2 695 697 10412057 10.1038/11154
McGeer PL Itagaki S Tago H McGeer EG Reactive microglia in patients with senile dementia of the Alzheimer type are positive for the histocompatibility glycoprotein HLA-DR Neurosci Lett 1987 79 195 200 3670729 10.1016/0304-3940(87)90696-3
Eikelenboom P van Gool WA Neuroinflammatory perspectives on the two faces of Alzheimer's disease J Neural Transm 2004 111 281 294 14991455 10.1007/s00702-003-0055-1
Akiyama H Barger S Barnum S Bradt B Bauer J Cole GM Cooper NR Eikelenboom P Emmerling M Fiebich BL Finch CE Frautschy S Griffin WS Hampel H Hull M Landreth G Lue L Mrak R Mackenzie IR McGeer PL O'Banion MK Pachter J Pasinetti G Plata-Salaman C Rogers J Rydel R Shen Y Streit W Strohmeyer R Tooyoma I Van Muiswinkel FL Veerhuis R Walker D Webster S Wegrzyniak B Wenk G Wyss-Coray T Inflammation and Alzheimer's disease Neurobiol Aging 2000 21 383 421 10858586 10.1016/S0197-4580(00)00124-X
Nicoll JA Mrak RE Graham DI Stewart J Wilcock G MacGowan S Esiri MM Murray LS Dewar D Love S Moss T Griffin WS Association of interleukin-1 gene polymorphisms with Alzheimer's disease Ann Neurol 2000 47 365 368 10716257 10.1002/1531-8249(200003)47:3<365::AID-ANA13>3.0.CO;2-G
Papassotiropoulos A Bagli M Jessen F Bayer TA Maier W Rao ML Heun R A genetic variation of the inflammatory cytokine interleukin-6 delays the initial onset and reduces the risk for sporadic Alzheimer's disease Ann Neurol 1999 45 666 668 10319892 10.1002/1531-8249(199905)45:5<666::AID-ANA18>3.0.CO;2-3
McCusker SM Curran MD Dynan KB McCullagh CD Urquhart DD Middleton D Patterson CC McIlroy SP Passmore AP Association between polymorphism in regulatory region of gene encoding tumour necrosis factor alpha and risk of Alzheimer's disease and vascular dementia: a case-control study Lancet 2001 357 436 439 11273064 10.1016/S0140-6736(00)04008-3
Rogers J Kirby LC Hempelman SR Berry DL McGeer PL Kaszniak AW Zalinski J Cofield M Mansukhani L Willson P . Clinical trial of indomethacin in Alzheimer's disease Neurology 1993 43 1609 1611 8351023
McGeer PL Schulzer M McGeer EG Arthritis and anti-inflammatory agents as possible protective factors for Alzheimer's disease: a review of 17 epidemiologic studies Neurology 1996 47 425 432 8757015
in t'Veld BA Ruitenberg A Hofman A Launer LJ van Duijn CM Stijnen T Breteler MM Stricker BH Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease N Engl J Med 2001 345 1515 1521 11794217 10.1056/NEJMoa010178
Sastre M Dewachter I Landreth GE Willson TM Klockgether T Van Leuven F Heneka MT Nonsteroidal anti-inflammatory drugs and peroxisome proliferator-activated receptor-gamma agonists modulate immunostimulated processing of amyloid precursor protein through regulation of beta-secretase J Neurosci 2003 23 9796 9804 14586007
Lim GP Yang F Chu T Chen P Beech W Teter B Tran T Ubeda O Ashe KH Frautschy SA Cole GM Ibuprofen suppresses plaque pathology and inflammation in a mouse model for Alzheimer's disease J Neurosci 2000 20 5709 5714 10908610
Fluhrer R Multhaup G Schlicksupp A Okochi M Takeda M Lammich S Willem M Westmeyer G Bode W Walter J Haass C Identification of a beta-secretase activity, which truncates amyloid beta-peptide after its presenilin-dependent generation Journal of Biological Chemistry 2003 278 5531 5538 12471021 10.1074/jbc.M211485200
Rossner S Apelt J Schliebs R Perez-Polo JR Bigl V Neuronal and glial beta-secretase (BACE) protein expression in transgenic Tg2576 mice with amyloid plaque pathology J Neurosci Res 2001 64 437 446 11391698 10.1002/jnr.1095
Barger SW Harmon AD Microglial activation by Alzheimer amyloid precursor protein and modulation by apolipoprotein E Nature 1997 388 878 881 9278049 10.1038/42257
Blasko I Marx F Steiner E Hartmann T Grubeck-Loebenstein B TNFalpha plus IFNgamma induce the production of Alzheimer beta-amyloid peptides and decrease the secretion of APPs FASEB J 1999 13 63 68 9872930
Schenk D Barbour R Dunn W Gordon G Grajeda H Guido T Hu K Huang J Johnson-Wood K Khan K Kholodenko D Lee M Liao Z Lieberburg I Motter R Mutter L Soriano F Shopp G Vasquez N Vandevert C Walker S Wogulis M Yednock T Games D Seubert P Immunization with amyloid-beta attenuates Alzheimer-disease-like pathology in the PDAPP mouse Nature 1999 400 173 177 10408445 10.1038/22124
Mehlhorn G Hollborn M Schliebs R Induction of cytokines in glial cells surrounding cortical beta-amyloid plaques in transgenic Tg2576 mice with Alzheimer pathology Int J Dev Neurosci 2000 18 423 431 10817926 10.1016/S0736-5748(00)00012-5
Sly LM Krzesicki RF Brashler JR Buhl AE McKinley DD Carter DB Chin JE Endogenous brain cytokine mRNA and inflammatory responses to lipopolysaccharide are elevated in the Tg2576 transgenic mouse model of Alzheimer's disease Brain Res Bull 2001 56 581 588 11786245 10.1016/S0361-9230(01)00730-4
Abbas N Bednar I Mix E Marie S Paterson D Ljungberg A Morris C Winblad B Nordberg A Zhu J Up-regulation of the inflammatory cytokines IFN-gamma and IL-12 and down-regulation of IL-4 in cerebral cortex regions of APP(SWE) transgenic mice J Neuroimmunol 2002 126 50 57 12020956 10.1016/S0165-5728(02)00050-4
Benzing WC Wujek JR Ward EK Shaffer D Ashe KH Younkin SG Brunden KR Evidence for glial-mediated inflammation in aged APP(SW) transgenic mice Neurobiol Aging 1999 20 581 589 10674423 10.1016/S0197-4580(99)00065-2
Apelt J Schliebs R Beta-amyloid-induced glial expression of both pro- and anti-inflammatory cytokines in cerebral cortex of aged transgenic Tg2576 mice with Alzheimer plaque pathology Brain Res 2001 894 21 30 11245811 10.1016/S0006-8993(00)03176-0
Szczepanik AM Rampe D Ringheim GE Amyloid-beta peptide fragments p3 and p4 induce pro-inflammatory cytokine and chemokine production in vitro and in vivo J Neurochem 2001 77 304 317 11279286
Murray CA Lynch MA Evidence that increased hippocampal expression of the cytokine interleukin-1 beta is a common trigger for age- and stress-induced impairments in long-term potentiation J Neurosci 1998 18 2974 2981 9526014
Tancredi V D'Antuono M Cafe C Giovedi S Bue MC D'Arcangelo G Onofri F Benfenati F The inhibitory effects of interleukin-6 on synaptic plasticity in the rat hippocampus are associated with an inhibition of mitogen-activated protein kinase ERK J Neurochem 2000 75 634 643 10899938 10.1046/j.1471-4159.2000.0750634.x
Graber HU Zurbriggen A Vandevelde M Identification of canine glial cells by nonradioactive in situ hybridization Zentralbl Veterinarmed A 1993 40 665 671 8135072
Vodovotz Y Lucia MS Flanders KC Chesler L Xie QW Smith TW Weidner J Mumford R Webber R Nathan C Roberts AB Lippa CF Sporn MB Inducible nitric oxide synthase in tangle-bearing neurons of patients with Alzheimer's disease J Exp Med 1996 184 1425 1433 8879214 10.1084/jem.184.4.1425
Smith MA Richey Harris PL Sayre LM Beckman JS Perry G Widespread peroxynitrite-mediated damage in Alzheimer's disease J Neurosci 1997 17 2653 2657 9092586
Bigl M Apelt J Luschekina EA Lange-Dohna C Rossner S Schliebs R Expression of beta-secretase mRNA in transgenic Tg2576 mouse brain with Alzheimer plaque pathology Neurosci Lett 2000 292 107 110 10998560 10.1016/S0304-3940(00)01452-X
Irizarry MC Locascio JJ Hyman BT beta-site APP cleaving enzyme mRNA expression in APP transgenic mice: anatomical overlap with transgene expression and static levels with aging Am J Pathol 2001 158 173 177 11141490
|
16212664
|
PMC1274341
|
CC BY
|
2021-01-04 16:38:22
|
no
|
J Neuroinflammation. 2005 Oct 7; 2:22
|
utf-8
|
J Neuroinflammation
| 2,005 |
10.1186/1742-2094-2-22
|
oa_comm
|
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-4-211619754210.1186/1476-511X-4-21ResearchPostprandial lipemia in men with metabolic syndrome, hypertensives and healthy subjects Kolovou Genovefa D [email protected] Katherine K [email protected] Antonis N [email protected] Klelia D [email protected] Stella A [email protected] Konstantinos [email protected] Dimitris S [email protected] Athanasios [email protected] Dennis V [email protected] Cardiology Department, Onassis Cardiac Surgery Centre, Athens, Greece2 Molecular Biology Department, Onassis Cardiac Surgery Centre, Athens, Greece3 Medical Department, Tzanio State Hospital, Piraeus, Greece4 Boston University, School of Medicine, Boston, USA2005 30 9 2005 4 21 21 12 9 2005 30 9 2005 Copyright © 2005 Kolovou et al; licensee BioMed Central Ltd.2005Kolovou et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The metabolic syndrome (MetS), as well as postprandial hypertriglyceridemia, is associated with coronary heart disease. This study aimed to evaluate the postprandial lipemia after oral fat tolerance test (OFTT) in subjects with MetS and compare them to hypertensive (HTN) and healthy subjects.
Results
OFTT was given to 33 men with MetS (defined by the Adult Treatment Panel III), 17 HTN and 14 healthy men. The MetS group was further divided according to fasting triglycerides (TG) into TG ≥ 150 [MetS+TG, (n = 22)] or <150 mg/dl [MetS-TG (n = 11)], and into those with or without hypertension [MetS+HTN (n = 24), MetS-HTN (n = 9), respectively]. TG concentrations were measured before and at 4, 6 and 8 h after OFTT and the postprandial response was quantified using the area under the curve (AUC) for TG.
The postprandial response was significantly higher in MetS compared to HTN and healthy men [AUC (SD) in mg/dl/h; 2534 ± 1016 vs. 1620 ± 494 and 1019 ± 280, respectively, p ≤ 0.001]. The TG levels were increased significantly in MetS+TG compared to MetS-TG subjects at 4 (p = 0.022), 6 (p < 0.001) and 8 hours (p < 0.001). The TG were increased significantly in MetS-TG compared to healthy subjects at 4 (p = 0.011), 6 (p = 0.001) and 8 hours (p = 0.015). In linear regression analysis only fasting TG levels were a significant predictor of the AUC (Coefficient B = 8.462, p < 0.001).
Conclusion
Fasting TG concentration is the main determinant of postprandial lipemia. However, an exaggeration of TG postprandialy was found in normotriglyceridemic MetS and HTN compared to healthy subjects. This suggests that intervention to lower fasting TG levels should be recommended in MetS subjects.
==== Body
Background
In 1988, Reaven et al. proposed insulin resistance as the underlying metabolic aberration linking essential hypertension (HTN), dyslipidemia, type 2 diabetes and other abnormalities associated with an increased risk of atherosclerotic cardiovascular disease in adults [1]. This constellation is now designated as the metabolic syndrome (Mets). Since then, many epidemiological studies have shown that MetS is associated with an increased incidence of coronary heart disease (CHD) [2,3]. MetS is clinically characterized by the presence of abnormal fasting triglyceride (TG) levels, low high-density lipoprotein (HDL) cholesterol levels, elevated plasma glucose, HTN and abdominal obesity. However, according to the National Cholesterol Education Program – Adults Treatment Panel (ATP III) guidelines, the diagnosis of MetS is present when at least three of the above five clinical criteria co-exist [4]. Although estimates of the prevalence are critically dependent on the exact definition used, the MetS has reached epidemic proportions. The age-adjusted prevalence of the MetS in the United States is estimated at 24% and increases to 44% in adults who are over 60 years old [5]. Similar results were found in Greek population (19.8%) [6].
Postprandial hypertriglyceridemia is also associated with cardiovascular disease [7]. During the postprandial state, TG rich lipoproteins such as chylomicrons, very low-density lipoproteins and their remnants, may promote the development of atherogenic small dense low-density lipoprotein (LDL) particles [8]. Abnormal postprandial lipemia has been observed in normolipidemic men with or without CHD [9], in heterozygous familial hypercholesterolemia (hFH), in women with CHD, in hypertensives and others, according to previous studies including ours [10-14]. Only few data exist regarding the response of TG to fatty meal in subjects with MetS [15]. The aim of this study was to evaluate the postprandial lipemia after an oral fat tolerance test (OFTT) in patients with MetS (defined by ATP III).
Results
All participants ingested their fatty meal and tolerated it well. The amount of fatty meal ingested by the MetS group was 362(28) g, by the HTN was 333(23) g, and by the Healthy was 347(16) g.
Clinical Characteristics
Clinical characteristics of main groups (MetS, HTN, and Healthy) are shown in Table 1. Clinical characteristics of subgroups (MetS+HTN, MetS-HTN, MetS+TG and MetS-TG) are shown in Table 2.
Table 1 Clinical characteristics of the 3 main study groups.
Characteristics MetS n = 33 HTN n = 17 Healthy n = 14 P values
Age (years) 50(11) 52(10) 49(9) 0.632
BMI (kg/m2) 30(4) 26(2) 26(2) <0.001
Waist (cm) 107(11) 93(5) 95(6) <0.001
CHD -/+ 25/8 17/0 14/0
Hypertension -/+ 9/24 0/17 14/0
Systolic BP (mmHg) 141(19) 149(18) 121(7) <0.001
Diastolic BP (mmHg) 83(12) 92(11) 74(9) <0.001
Smokers -/+ 12/21 9/8 14/0
Diabetes mellitus -/+ 26/7 16/1 14/0
TC (mg/dl) 241(39) 214(29) 188(28) <0.001
HDL (mg/dl) 35(7) 48(15) 51(15) <0.001
LDL (mg/dl) 165(36) 145(31) 113(33) <0.001
Apo A (mg/dl) 133(19) 148(30) 159(48) 0.038
Apo B (mg/dl) 137(28) 124(23) 125(29) 0.232
Lp (a) (mg/dl) 16(13) 21(16) 16(11) 0.57
Glucose (mg/dl) 109(29) 93(10) 91(10) 0.013
Insulin (μU/ml) 14(9) 8(9) 7(5) 0.036
HOMA-IR 3.9(2.7) 2.1(2.2) 1.7(1.1) 0.015
TG0 (mg/dl) 204(86) 105(30) 89(30) <0.001
TG4 (mg/dl) 364(153) 242(84) 145(44) <0.001
TG6 (mg/dl) 366(155) 269(94) 142(53) <0.001
TG8 (mg/dl) 297(124) 200(87) 131(56) <0.001
AUC (mg/dl/h) 2534(1016) 1620(494) 1019(280) <0.001
All values are presented as means (standard deviation) or percentages for categorical variables. P values arose by ANOVA analysis and denote difference among the three groups. MetS: men with metabolic syndrome, HTN: Hypertensive men, BMI: body mass index, CHD: coronary heart disease, TC: total cholesterol, HDL: high-density lipoprotein cholesterol, LDL: low-density lipoprotein cholesterol, Apo: apolipoprotein, Lp: lipoprotein, TG0: fasting plasma triglyceride concentration, TG4, TG6, TG8: plasma triglyceride concentration 4, 6, 8 h after the fat load, respectively, HOMA-IR: index of homeostasis model of insulin resistance, AUC: area under the curve for triglycerides. To convert from mg/dl to mmol/L for TC, HDL, LDL, divide by 38.7. To convert from mg/dl to mmol/L for glucose, divide by 18. To convert from mg/dl to mmol/L for TG divide by 88.6.
Table 2 Clinical characteristics of different subgroups.
Characteristics MetS+HTN MetS-HTN MetS-TG MetS+TG
Number of patients 24 9 11 22
Age (years) 53(10) 40(6) 52(10) 48(11)
BMI (kg/m2) 30(4) 31(3) 30(4) 30(4)
Waist (cm) 105(10) 115(10) 104(9) 108(11)
CHD -/+ 16/8 9/0 6/5 19/3
Hypertension -/+ 0/24 9/0 1/10 8/14
Systolic BP (mmHg) 148(17) 123(11) 143(12) 140(22)
Diastolic BP (mm Hg) 85(13) 78(6) 82(15) 83(11)
Smokers -/+ 10/14 2/7 4/7 8/14
Diabetes mellitus -/+ 17/7 9/0 7/4 19/3
TC (mg/dl) 234(37) 259(40) 228(34) 247(40)
HDL (mg/dl) 36(7) 30(3) 34(6) 35(7)
LDL (mg/dl) 159(34) 179(38) 166(33) 164(38)
Apo A (mg/dl) 137(19) 124(17) 130(10) 135(23)
Apo B (mg/dl) 135(29) 143(25) 134(36) 139(24)
Lp (a) (mg/dl) 16(11) 17(19) 16(10) 17(15)
Glucose (mg/dl) 113(33) 101(11) 114(35) 107(26)
Insulin (mg/dl) 12(6) 22(14) 14(9) 15(10)
HOMA 3.4(2.1) 5.8(3.9) 3.7(2.4) 4.1(2.9)
TG0 (mg/dl) 187(86) 248(72) 125(22) 243(78)
TG4 (mg/dl) 327(116) 454(199) 232(96) 411(143)
TG6 (mg/dl) 344(150) 426(161) 256(95) 421(151)
TG8 (mg/dl) 287(126) 321 (121) 200(73) 343(117)
AUC (mg/dl/h) 2349(945) 2984(1099) 1639(549) 2960(907)
All values are presented as means (standard deviation) or percentages for categorical variables. MetS: men with metabolic syndrome, HTN: Hypertensive men, BMI: body mass index, CHD: coronary heart disease, TC: total cholesterol, HDL: high-density lipoprotein cholesterol, LDL: low-density lipoprotein cholesterol, Apo: apolipoprotein, Lp: lipoprotein, TG0: fasting plasma triglyceride concentration, TG4, TG6, TG8: plasma triglyceride concentration 4, 6, 8 h after the fat load, respectively, HOMA-IR: index of homeostasis model of insulin resistance, AUC: area under the curve for triglycerides. To convert from mg/dl to mmol/L for TC, HDL, LDL, divide by 38.7. To convert from mg/dl to mmol/L for glucose, divide by 18. To convert from mg/dl to mmol/L for TG divide by 88.6.
Postprandial Changes
Plasma total and HDL cholesterol, apolipoproteins A and B, and lipoprotein (a) were measured postprandialy in a 60% of the study population and no significant difference was found between groups.
The following changes were noticed pre and post OFTT:
A) Comparison of the TG concentrations at 0, 4, 6, 8 h in MetS, HTN and Healthy subject groups are shown in Figure 1. Schematic representation of postprandial response (AUC) of the above three groups is shown in Figure 2.
Figure 1 Comparison of triglyceride (TG) concentrations at 0, 4, 6, 8 h between pairs of the three main groups [men with metabolic syndrome (MetS), Hypertensive men (HTN) and Healthy].
Figure 2 Schematic representation of postprandial response (area under the curve) of the three main groups [men with metabolic syndrome (MetS), Hypertensive men (HTN) and Healthy]. p = 0.001 for differences between MetS and HTN groups. p < 0.001 for differences between MetS and Healthy groups. p < 0.001 for differences between HTN and Healthy groups.
B) In MetS+HTN and HTN groups: The TG levels were significantly increased in MetS+HTN vs. HTN subjects at 4 hours (p = 0.022), 6 (p < 0.001) and 8 hours (p < 0.001). [Mean (SD) values are presented in Table 1, 2].
C) In MetS+TG and MetS-TG: The TG levels were increased significantly in MetS+TG vs. MetS-TG subjects at 4 (p = 0.022), 6 (p < 0.001) and 8 hours (p < 0.001). [Mean (SD) values are presented in Table 2].
D) In MetS-TG and Healthy: The TG levels were increased significantly in MetS-TG vs. Healthy subjects at 4 (p = 0.011), 6 (p = 0.001) and 8 hours (p = 0.015). [Mean (SD) values are presented in Table 1, 2].
E) No significant differences were found in MetS+HTN vs. MetS-HTN and in MetS-TG vs. HTN subjects.
F) In linear regression analysis, where the AUC was the dependent variable and age, body mass index (BMI), HDL cholesterol, fasting TG, fasting glucose and HOMA-IR were the independent variables, only fasting TG levels were independently associated with high AUC values (Coefficient B = 8.462, p < 0.001).
Discussion
In the present study we investigated the effect of a fatty meal on plasma TG concentration in untreated men with MetS, HTN and Healthy. We showed that MetS men have a different response to a fatty meal when compared to HTN and Healthy men. Linear regression analysis showed that for every rise of 1 mg/dl of the fasting TG values, the AUC increased by 8.462 mg/dl/h.
The fasting TG values were 129% higher in MetS compared to Healthy men. The observed difference was smaller than that reported by Chan et al [16]. Several facts could explain this difference: First, the fasting TG levels in their control group were much lower compared to our group. Fasting TG levels of their MetS subjects were also lower even though the BMI was higher. Indeed, when the clinical characteristics were similar to those of our subjects, the fasting TG values did not differ substantially [17]. Furthermore, the regulation of plasma TG in MetS men may differ according to concomitant characteristics. An enhanced TG rise postprandialy has been reported not only in patients with CHD [18], but also in patients with diabetes mellitus [19] and obesity [20], all of which are common findings in MetS subjects.
Obesity is associated with a range of metabolic abnormalities including fasting and postprandial dyslipidemia. This point is considered in greater detail; see our review [21]. In summary, visceral adiposity with increased intra-abdominal fat has been shown to precede the development of insulin resistance. The insulin resistance is associated with dyslipidemia [22] and may affect chylomicron remnant metabolism by down regulating LDL-receptor expression and increasing hepatic cholesterol synthesis and very low-density lipoprotein secretion [23]. These effects increase competition between chylomicron and very low density lipoprotein remnants for hepatic receptors, thereby impairing the uptake of chylomicron remnants by this pathway [24]. These changes result in elevated postprandial TG levels. Obesity has similar potential effects on chylomicron remnant metabolism. However, a weight reduction of 10 kg in insulin-resistant obese men was insufficient to reduce the elevated chylomicron remnant levels [25]. In contrast, type 2 diabetic patients treated with insulin, showed a 28% decrease in fasting plasma TG levels, a 17% increase in total HDL cholesterol, while the magnitude of postprandial lipemia after the ingestion of fatty meal decreased by 38%. Additionally, insulin treatment was accompanied by a 21% increase in lipoprotein lipase activity and a 20% increase in cholesteryl ester transport protein activity [26].
Essential hypertension constitutes one of the features included in the MetS and is further associated with insulin resistance. We have previously reported [13] that hypertension is associated with an abnormal response to a fatty meal (increased TG levels and delayed TG clearance), which seems to aggravate the prognosis of hypertensive patients. In our current subjects with MetS and hypertension (24 out of 33 MetS subjects) the exaggerated postprandial TG response to a fatty meal was even more pronounced than in 17 subjects with hypertension as a single abnormality [at 4 hours (p = 0.022), 6 (p < 0.001) and 8 hours (p < 0.001)]. It is difficult to distinguish which mechanism is responsible for this. The variability of BMI or fasting TG levels between subjects with MetS and hypertension compared to subjects with hypertension only [30(4) vs. 26(2), or 187(86) vs. 105(30), respectively] can probably account for these differences. However, other mechanisms may also be responsible.
In our current study HDL cholesterol levels were found to be lower in the MetS subjects compared to Healthy men as expected. However, other studies, including ours [27], have suggested that the levels of HDL cholesterol do not affect postprandial TG magnitude. Specifically, in a subgroup of subjects with isolated low fasting HDL cholesterol, fasting hypertriglyceridemia was a prerequisite in order to have an exaggerated postprandial TG response [28]. Additionally, Cohen and co-workers [29] measured the response of plasma TG and retinyl palmitate to different fatty meals in endurance-trained men with a wide range of plasma HDL cholesterol concentrations (36–105 mg/dl). Their data indicated that the magnitude of postprandial lipemia is not primarily affected by HDL cholesterol concentration.
The remnants of TG-rich lipoproteins accumulated in the postprandial state are involved in atherogenesis. They act as carriers of cholesteryl ester to the vessel wall and also induce endothelial dysfunction with their toxic influence on endothelial cells [30]. Furthermore, the hypertriglyceridemic state is accompanied by small dense LDL particles that are more susceptible to oxidation and low concentrations of HDL cholesterol [31]. Hypertriglyceridemia may also be involved in events leading to thrombosis [32] and probably provoke activation of nuclear factor kappa B, a proposed key mediator of atherosclerosis [33]. In subjects with MetS, the hypertriglyceridemia is one of the main lipid abnormalities, which potentially is responsible for exaggerated TG postprandialy. However, in subjects with MetS and normal fasting TG [125(22) mg/dl], defined by ATP III guidelines (< 150 mg/dl), the TG levels postprandialy were exaggerated and delayed clearance was also observed. Besides, the postprandial response was similar in MetS subjects with hypertension compared to MetS subjects with normal blood pressure, even though the baseline TG levels were lower in the former [187(86) vs. 248(72) mg/dl, p = 0.041, respectively]. This suggests that after certain fasting TG levels, no further increase of postprandial TG occurs; alternatively other mechanisms could be involved such as genetic influence.
Conclusion
Our data indicate that fasting TG concentration is the main determinant of postprandial lipemia. However, an exaggeration of TG postprandialy is observed in MetS and HTN subjects with normal fasting TG levels. This suggests that intervention to lower fasting TG levels should be recommended in MetS subjects.
Methods
Study population
The study population consisted of 64 Greek men. Heavy drinking, liver and renal disease, hypothyroidism, professional sport activity and the use of hypolipidemic drugs were exclusion criteria. The study population was divided into 3 groups:
1. The MetS group consisted of 33 men, mean age 50 ± 11 years. The diagnosis of MetS was based on the co-existence of at least three of the following five clinical criteria: a) abdominal obesity (waist >102 cm), b) serum TG levels ≥ 150 mg/dl, c) serum HDL cholesterol levels <40 mg/dl, d) blood pressure ≥ 130/85 mmHg, e) fasting glucose >110 mg/dl.
2. The Hypertensive (HTN) group consisted of 17 men, mean age 52 ± 10 years with a negative CHD history. Additionally, their serum TG levels were < 150 mg/dl, and none of them fulfilled the criteria for MetS.
3. The Healthy group consisted of 14 healthy men, age-matched, mean age 49 ± 9 years with no family history of premature atherosclerosis, diabetes mellitus, HTN and dyslipidemia. Their fasting TG levels were < 150 mg/dl, their fasting glucose < 110 mg/dl, their total cholesterol < 240 mg/dl, their LDL <160 mg/dl, and their blood pressure < 130/85 mmHg. All healthy subjects were never smokers.
The MetS group was further divided into those with baseline TG levels < 150 mg/dl (MetS-TG, n = 11) and into those with TG levels ≥ 150 mg/dl (MetS+TG, n = 22), and into MetS with and without HTN [MetS+HTN (n = 24) and MetS-HTN (n = 9), respectively].
All participants gave their informed consent and the Ethics Committee of the Onassis Cardiac Surgery Centre, Athens, Greece, approved the study protocol.
Study Protocol
All patients were studied in the outpatient clinic between 8.00–9.00 am after 12 h overnight fast. The fatty meal was consumed within 20 min and plasma TG concentrations were measured before and 4, 6 and 8 h after the fat load. During this 8 h period, the participants did not eat and did not smoke. They were only allowed to drink water. Blood samples were drawn at 8:00 (before the meal), at 12:30 am (4 h after the meal), at 2:30 pm (6 h after the meal), and at 4:30 pm (8 h after the meal). In all four samples total cholesterol, TG, HDL cholesterol, apolipoproteins A and B, and lipoprotein (a) were measured. The content of the fatty meal has been described in a previous study [13]. Briefly, the fatty meal was a slight modification of that introduced by Patsch et al. [9], consisting of 83.5% fat, 14.0% carbohydrates and 2.5% proteins and was given in a dose based on the patient's body surface area (350 g for 2 m2).
Determination of blood lipids and glucose
Plasma total cholesterol, TG and HDL cholesterol were measured using enzymatic colorimetric methods on a Roche Integra Biochemical analyzer with commercially available kits (Roche Diagnostics Gmbh, Hannheim, Germany). The serum LDL cholesterol levels were calculated using the Friedewald formula [34] only in patients with TG levels <400 mg/dl. Apolipoproteins A, B and lipoprotein (a) were measured by nephelometery (Nephelometer: BN-100, Behring, Germany). Blood glucose was measured by the hexokinase method with a Dade Behring reagent on a Dimension (Dade Behring) instrument. Blood insulin was measured with the IMX ABBOTT Diagnostics instrument. All samples were analyzed within 24 h.
The body mass index was calculated as weight divided by height expressed in kg/m2. We assessed the whole-body insulin resistance with the following formula: HOMA-IR = (fast glucose × fast insulin)/22.5 [13].
Statistical Analysis
Categorical variables are presented as percentages. Values of numerical characteristics were tested for normality and are presented as mean value (± SD), if normally distributed. An ANOVA or Kruskal-Wallis H test with a Bonferroni correction (whichever appropriate) was performed for three group comparisons. The t-test for independent samples or the Mann Whitney U test was used for the comparison of numerical values between two groups. Areas under the curve (AUC) for serial measurements of TG levels at baseline and after the fatty meal were calculated using the trapezoid rule. Linear regression analysis, where the AUC was the dependent variable and age, BMI, HDL cholesterol, fasting TG, fasting glucose and HOMA-IR were the independent variables was performed in order to uncover the significant predictors of postprandial lipemia (i.e. high AUC values). The level of significance was set at p < 0.05.
Authors' contributions
GDK conceived the study and participated in the development of the hypothesis, the study design and drafting of the manuscript. KKA is a research associate who participated in the development of the hypothesis, the study design and drafting of the manuscript. ANP is a physician who participated in the study design, recruitment of subjects and clinical evaluation. KDS is a research associate who participated in data analysis and interpretation of the findings. SAI, KT and DSD are physicians who participated in the study design, recruitment of subjects and clinical evaluation. AM is a physician who participated in the study design and drafting of the manuscript. DVC is a physician who participated in the design of the study and its coordination.
==== Refs
Reaven GM Banting Lecture 1988: role of insulin resistance in human disease Diabetes 1988 37 1597 1607
Meigs JB Invited commentary: insulin resistance syndrome? Syndrome X? Multiple metabolic syndrome? A syndrome at all? Factor analysis reveals patterns in the fabric of correlated metabolic risk factors Am J Epidemiol 2000 152 908 911 11092432 10.1093/aje/152.10.908
Lakka HM Laaksonen DE Lakka TA Niskanen LK Kumpusalo E Tuomilehto J Salonen JT The metabolic syndrome and total and cardiovascular disease mortality in middle-aged men JAMA 2002 288 2709 2716 12460094 10.1001/jama.288.21.2709
Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (AdultTreatment Panel III) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults JAMA 2001 285 2486 2497 11368702 10.1001/jama.285.19.2486
Ford ES Giles WH Dietz WH Prevalence of the metabolic syndrome among US adults: findings from the third National Health and Nutrition Examination Survey JAMA 2002 287 356 359 11790215 10.1001/jama.287.3.356
Panagiotakos DB Pitsavos C Chrysohoou C Skoumas J Tousoulis D Toutouza M Toutouzas P Stefanadis C Impact of lifestyle habits on the prevalence of the metabolic syndrome among Greek adults from the ATTICA study Am Heart J 2004 147 106 112 14691427 10.1016/S0002-8703(03)00442-3
Uiterwaal CS Grobbee DE Witteman JC van Stiphout WA Krauss XH Havekes LM de Bruijn AM van Tol A Hoffman A Postprandial triglyceride response in young adult men and familial risk for coronary atherosclerosis Ann Intern Med 1994 121 576 583 8085689
Tan KC Cooper MB Ling KL Griffin BA Freeman DJ Packard CJ Shepherd J Hales CN Betteridge DJ Fasting and postprandial determinants for the occurrence of small dense LDL species in non-insulin-dependent diabetic patients with and without hypertriglyceridaemia: the involvement of insulin, insulin precursor species and insulin resistance Atherosclerosis 1995 113 273 287 7605366 10.1016/0021-9150(94)05454-Q
Patsch JR Miesenbock G Hopferwieser T Muhlberger V Knapp E Dunn JK Gotto AM JrPatsch W Relation of triglyceride metabolism and coronary artery disease. Studies in the postprandial state Arterioscler Thromb 1992 12 1336 1345 1420093
Kolovou G Anagnostopoulou K Pilatis N Iraklianou S Hoursalas I Liberi S Pavlidis A Dritsas A Mikhailidis D Cokkinos DV Heterozygote men with Familial Hypercholesterolaemia (FH) may have an Abnormal Triglyceride Response Postprandially. Evidence for another predictor of vascular risk in FH Int J Clin Pract 2005 59 311 317 15857328 10.1111/j.1742-1241.2004.00223.x
Kolovou G Anagnostopoulou K Pilatis N Giannopoulou M Hoursalas I Pavlidis A Adamopoulou E Valaora A Mikhailidis D Cokkinos DV The Influence of Natural Menopause on Postprandial Lipaemia in Heterozygotes for Familial Hypercholesterolaemia J Womens Health (Larchmt) 2004 13 1119 26 15650345
Meyer E Westerveld HT de Ruyter-Meijstek FC van Greevenbroek MM Rienks R van Rijn HJ Erkelens DW de Bruin TW Abnormal postprandial apolipoprotein B-48 and triglyceride responses in normolipidemic women with greater than 70% stenotic coronary artery disease: a case-control study Atherosclerosis 1996 124 221 235 8830935 10.1016/0021-9150(96)05832-7
Kolovou GD Daskalova CD Iraklianou SA Adamopoulou EN Pilatis ND Hatzigeorgiou GC Cokkinos DV Postprandial lipemia in hypertension J Am Coll Nutr 2003 22 80 87 12569118
Kolovou G Daskalova D Anagnostopoulou K Hoursalas I Voudris V Mikhailidis DP Cokkinos DV Postprandial hypertriglyceridaemia in patients with Tangier disease J Clin Pathol 2003 56 937 941 14645354 10.1136/jcp.56.12.937
Schrezenmeir J Fenselau S Keppler I Abel J Orth B Laue C Sturmer W Fauth U Halmagyi M Marz W Postprandial triglyceride high response and the metabolic syndrome Ann N Y Acad Sci 1997 827 353 368 9329767
Chan DC Watts GF Barrett PH O'Neill FH Redgrave TG Thompson GR Relationships between cholesterol homoeostasis and triacylglycerol-rich lipoprotein remnant metabolism in the metabolic syndrome Clin Sci (Lond) 2003 104 383 388 12653682
Watts GF Barrett PH Ji J Serone AP Chan DC Croft KD Loehrer F Johnson AG Differential regulation of lipoprotein kinetics by atorvastatin and fenofibrate in subjects with the metabolic syndrome Diabetes 2003 52 803 811 12606523
Pirro M Mauriege P Tchernof A Cantin B Dagenais GR Despres JP Lamarche B Plasma free fatty acid levels and the risk of ischemic heart disease in men: prospective results from the Quebec Cardiovascular Study Atherosclerosis 2002 160 377 384 11849661 10.1016/S0021-9150(01)00588-3
Ginsberg HN Illingworth DR Postprandial dyslipidemia: an atherogenic disorder common in patients with diabetes mellitus Am J Cardiol 2001 88 9H 15H 11576520 10.1016/S0002-9149(01)01831-8
Mamo JC Watts GF Barrett PH Smith D James AP Pal S Postprandial dyslipidemia in men with visceral obesity: an effect of reduced LDL receptor expression? Am J Physiol Endocrinol Metab 2001 281 E626 632 11500319
Kolovou GD Anagnostopoulou KK Daskalopoulou SS Mikhailidis DP Cokkinos DV Clinical Relevance of Postprandial Lipaemia Curr Med Chem 2005 12 1931 45 16101498 10.2174/0929867054546609
Kolovou G Anagnostopoulou K Cokkinos DV Pathophysiology of Dyslipidemia in the Metabolic Syndrome Post Med J 2005 81 358 66 10.1136/pgmj.2004.025601
Cummings MH Watts GF Umpleby AM Hennessy TR Kelly JM Jackson NC Sonksen PH Acute hyperinsulinemia decreases the hepatic secretion of very-low-density lipoprotein apolipoprotein B-100 in NIDDM Diabetes 1995 44 1059 1065 7657029
Cooper AD Hepatic uptake of chylomicron remnants J Lipid Res 1997 38 2173 2192 9392416
James AP Watts GF Barrett PH Smith D Pal S Chan DC Mamo JC Effect of weight loss on postprandial lipemia and low-density lipoprotein receptor binding in overweight men Metabolism 2003 52 136 141 12601621 10.1053/meta.2003.50032
Geltner C Lechleitner M Foger B Ritsch A Drexel H Patsch JR Insulin improves fasting and postprandial lipemia in type 2 diabetes Eur J Intern Med 2002 13 256 263 12067822 10.1016/S0953-6205(02)00038-9
Kolovou GD Anagnostopoulou KK Pilatis N Kafaltis N Sorodila K Psarros E Cokkinos DV Low fasting low high-density lipoprotein and postprandial lipemia Lipids Health Dis 2004 3 18 15271218 10.1186/1476-511X-3-18
Couillard C Bergeron N Bergeron J Pascot A Mauriege P Tremblay A Prud'homme D Bouchard C Despres JP Metabolic heterogeneity underlying postprandial lipemia among men with low fasting high density lipoprotein cholesterol concentrations J Clin Endocrinol Metab 2000 85 4575 4582 11134111 10.1210/jc.85.12.4575
Cohen JC Stray-Gundersen J Grundy SM Dissociation between postprandial lipemia and high density lipoprotein cholesterol concentrations in endurance-trained men Arterioscler Thromb 1991 11 838 843 2065037
Zilversmit DB Atherogenesis: a postprandial phenomenon Circulation 1979 60 473 485 222498
Malloy MJ Kane JP A risk factor for atherosclerosis: triglyceride-rich lipoproteins Adv Intern Med 2001 47 111 136 11795072
Sanders TA Oakley FR Cooper JA Miller GJ Influence of a stearic acid-rich structured triacylglycerol on postprandial lipemia, factor VII concentrations, and fibrinolytic activity in healthy subjects Am J Clin Nutr 2001 73 715 721 11273845
Collins T Cybulsky MI NF-kappaB: pivotal mediator or innocent bystander in atherogenesis? J Clin Invest 2001 107 255 264 11160146
Friedewald WT Levy RI Fredrickson DS Estimation of the concentration of the low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge Clin Chem 1972 18 499 502 4337382
|
16197542
|
PMC1274342
|
CC BY
|
2021-01-04 16:39:19
|
no
|
Lipids Health Dis. 2005 Sep 30; 4:21
|
utf-8
|
Lipids Health Dis
| 2,005 |
10.1186/1476-511X-4-21
|
oa_comm
|
==== Front
Mol PainMolecular Pain1744-8069BioMed Central London 1744-8069-1-301624203310.1186/1744-8069-1-30ReviewRole of spinal cord glutamate transporter during normal sensory transmission and pathological pain states Tao Yuan-Xiang [email protected] Jianguo [email protected] Robert L [email protected] Department of Anesthesiology and Critical Care Medicine, Johns Hopkins University School of Medicine, 355 Ross, 720 Rutland Ave., Baltimore, Maryland 21205, USA2 Department of Oral and Maxillofacial Surgery, Mcknight Brain Institute and College of Dentistry, University of Florida, Gainesville, Florida, 32610, USA3 Department of Physiology and Cell Biology, The Ohio State University College of Medicine, Columbus, Ohio 43210, USA2005 21 10 2005 1 30 30 22 8 2005 21 10 2005 Copyright © 2005 Tao et al; licensee BioMed Central Ltd.2005Tao et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Glutamate is a neurotransmitter critical for spinal excitatory synaptic transmission and for generation and maintenance of spinal states of pain hypersensitivity via activation of glutamate receptors. Understanding the regulation of synaptically and non-synaptically released glutamate associated with pathological pain is important in exploring novel molecular mechanisms and developing therapeutic strategies of pathological pain. The glutamate transporter system is the primary mechanism for the inactivation of synaptically released glutamate and the maintenance of glutamate homeostasis. Recent studies demonstrated that spinal glutamate transporter inhibition relieved pathological pain, suggesting that the spinal glutamate transporter might serve as a therapeutic target for treatment of pathological pain. However, the exact function of glutamate transporter in pathological pain is not completely understood. This report will review the evidence for the role of the spinal glutamate transporter during normal sensory transmission and pathological pain conditions and discuss potential mechanisms by which spinal glutamate transporter is involved in pathological pain.
==== Body
In addition to its essential metabolic role, glutamate is a major mediator of excitatory signals in the central nervous system and is involved in many physiologic and pathologic processes, such as excitatory synaptic transmission, synaptic plasticity, cell death, stroke, and chronic pain [1,2]. Glutamate exerts its signaling role by acting on glutamate receptors, including N-methyl-D-aspartate (NMDA), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate, and metabotropic glutamate receptors. These receptors are located on the pre- and post-synaptic membranes, as well as, at extra-synaptic sites. Glutamate concentration in the synaptic cleft determines the extents of receptor stimulation and excitatory synaptic transmission. It is of critical importance that the extracellular glutamate concentration be kept at physiological levels, as excessive activation of glutamate receptors can lead to excitotoxicity and neuronal death [3]. The clearance of glutamate from the synaptic cleft is principally dependent on Na+-dependent, high-affinity, neuronal glutamate transporters present presynaptically, postsynaptically, and perisynaptically, and on glial glutamate transporters (Fig. 1). Currently, five isoforms of glutamate transporters have been identified [3]: namely, GLAST (glutamate/aspartate transporter), GLT-1 (glutamate transporter-1), EAAC (excitatory amino acid carrier) 1, EAAT (excitatory amino-acid transporter) 4, and EAAT5. The human homologues of the three more ubiquitous subtypes (GLAST, GLT-1, and EAAC1) are named EAAT1, EAAT2, and EAAT3, respectively. The five isoforms belong to the same gene-family and share 50–60% amino acid sequence identity [3]. However, they have discrete cellular and regional localizations. GLAST is present in glial cells throughout the central nervous system, with strong labeling in cerebellar Bergmann glia and more diffuse labeling in the forebrain [3]. It is also transiently expressed in a small number of neurons [4]. GLT-1 is almost exclusively expressed on glia and is widespread and abundant throughout the forebrain, cerebellum, and spinal cord [4]. In contrast, EAAC1 is found predominantly in neurons of the spinal cord and brain [4,5]. EAAT4 has properties of a ligand-gated Cl-channel and is localized mainly in cerebellar Purkinje cells [6]. EAAT5 is retina-specific [7].
Figure 1 Glutamate (Glu) uptake and Glu/glutamine (Gln) cycle. Glu released from the nerve terminal by exocytosis is taken up by neuronal Glu transporter present presynaptically (1) and postsynaptically (2) and by glial Glu transporter (3). Glu/Gln cycle is one type of Glu recycling, but the significance is still unclear in vivo (see references 37 and 38). Astroglia detoxifies Glu by converting it to Gln. Glu is subsequently released from the glial cells by glial Gln transporter (4) and taken up by neuronal Gln transporter (5). Neurons convert Gln back to Glu, which is loaded into synaptic vesicles by vesicular Glu transporter (6). 7: postsynaptic Glu receptors.
Given the well-documented evidence that glutamate acts as a major excitatory neurotransmitter in primary afferent terminals [2], it is expected that glutamate transporter might be involved in excitatory sensory transmission and pathological pain. Indeed, recent studies have revealed that inhibition of spinal glutamate transporter produced pro-nociceptive effects under normal conditions [8] and have unexpected antinociceptive effects under pathological pain conditions [9-11]. It is not completely understood why the effects of spinal glutamate transporter inhibition under pathological pain conditions are opposite to its effects under normal conditions. In this review, we will illustrate the expression and distribution of the glutamate transporter in two major pain-related regions: spinal cord and dorsal root ganglion (DRG). We will also review the evidence for the role of the glutamate transporter during normal sensory transmission and pathological pain conditions and discuss potential mechanisms by which glutamate transporter is involved in pathological pain.
Expression and distribution of glutamate transporter in the spinal cord and dorsal root ganglion
In the spinal cord, three isoforms of glutamate transporter (GLAST, GLT-1, and EAAC1) have been reported [4,12]. They are expressed in highest density within the superficial dorsal horn of the spinal cords of rats and mice (Fig. 2). GLT-1 and GLAST are exclusively distributed in glial cells at perisynaptic sites in the superficial dorsal horn [13]. EAAC1, in addition to its expression in the spinal cord neurons, is detected in the DRG and distributed predominantly in the small DRG neurons (but not in DRG glial cells) [12] (Fig. 3). Some of these EAAC1-positive DRG neurons are positive for calcitonin gene-related peptide (CGRP) or are labeled by IB4 [12,13]. Unilateral dorsal root rhizotomy shows less intense EAAC1 immunoreactivity in the superficial dorsal horn on the ipsilateral side, compared to the contralateral side [12]. Moreover, confocal microscopy demonstrates that some EAAC1-positive, small dot- or patch-like structures in the superficial laminae are labeled by IB4 or are positive for CGRP [12]. Under electron microscope, EAAC1 is associated with the axon terminal and dendritic membranes at synaptic and non-synaptic sites and is present with CGRP in the axons and the terminals in the superficial dorsal horn [13]. The expression level and distribution pattern of neuronal and glial glutamate transporters in the superficial dorsal horn suggest an important role for spinal glutamate transporter in spinal nociceptive transmission. In addition, the unique expression of EAAC1 in the small DRG neurons and nociceptive primary afferent terminals suggests that EAAC1 might have a distinct role in pain processing, compared to GLT-1 and GLAST.
Figure 2 Expression and distribution of the glutamate transporter in the dorsal root ganglion (A) and the spinal cord (B-D). EAAC1 is expressed mainly in small dorsal root ganglion cells (A) and distributed predominantly in the superficial dorsal horn of the spinal cord (B). GLAST (C) and GLT-1 (D) are expressed highly in the superficial dorsal horn and the region around the central canal. Scale bars: 10 μm in A and 125 μm in B, C, and D.
Figure 3 Double-immunofluorescence histochemistry for EAAC1 (A) and NeuN (B, a marker for neuronal nuclei), and their overlapping (C) in the dorsal root ganglion. Scale bar: 10 μm.
Role of the spinal cord glutamate transporter in normal sensory transmission
Recently evidence suggests that spinal glutamate transporter might play an important role in normal sensory transmission. Liaw et al. [8] reported that intrathecal injection of glutamate transporter blockers DL-threo-β-benzyloxyaspartate (TBOA) and dihydrokainate (DHK) produced significant and dose-dependent spontaneous nociceptive behaviors, such as licking, shaking, and caudally directed biting, phenomena similar to the behaviors caused by intrathecal glutamate receptor agonists, such as glutamate, NMDA, or AMPA, when given intrathecally [14-16]. Intrathecal TBOA also led to remarkable hypersensitivity in response to thermal and mechanical stimuli [8]. These findings are consistent with a previous report that showed an increase in spontaneous activity and responses of wide dynamic range neurons to both innocuous mechanical (brush, pressure) and noxious mechanical (pinch) stimuli after topical application of L-trans-pyrrolidine-2,4-dicarboxylic acid (PDC), a glutamate transporter blocker [17,18]. TBOA-induced behavioral responses could be significantly blocked by intrathecal injection of the NMDA receptor antagonists MK-801 and AP-5, the non-NMDA receptor antagonist CNQX or the nitric oxide synthase inhibitor L-NAME [8]. The effects of DHK and PDC were thought to be partially due to their non-specific interactions with glutamate receptors. However, unlike DHK and PDC, TBOA does not act as an agonist or antagonist at glutamate receptors [9,19,20]. Thus, spontaneous pain-related behaviors and sensory hypersensitivity evoked by TBOA directly support the involvement of glutamate transporter in normal excitatory synaptic transmission in the spinal cord. In vivo microdialysis analysis showed that intrathecal injection of TBOA produced short-term elevation of extracellular glutamate concentration in the spinal cord [8]. Topical application of TBOA on the dorsal surface of the spinal cord also resulted in a significant elevation of extracellular glutamate concentrations demonstrated by in vivo glutamate voltametry [8]. These findings indicate that a decrease of spinal glutamate uptake can lead to excessive glutamate accumulation in the spinal cord, which might, in turn, result in over-activation of glutamate receptors, and production of spontaneous nociceptive behaviors and sensory hypersensitivity. Thus, glutamate uptake through spinal glutamate transporters is critical for maintaining normal sensory transmission under physiological conditions.
Expression and function of the spinal cord glutamate transporter in pathological pain states
Glutamate uptake and expression of glutamate transporters in the spinal cord have been found to be changed under pathological conditions associated with chronic pain status. Chronic constriction nerve injury upregulated glutamate transporter expression at day 1 and 4 postoperatively, but it downregulated glutamate transporter expression at days 7 and 14 postoperatively [21]. Moreover, chronic constriction nerve injury significantly reduced spinal glutamate uptake activity at day 5 postoperatively [21]. Recently, another study showed that spinal nerve ligation also markedly reduced glutamate uptake activity, as demonstrated in spinal deep dorsal and ventral horn 4–6 weeks after the nerve ligation [22]. Although the underlying mechanism by which neuropathic inputs cause the decrease in spinal glutamate uptake is unclear, it is thought that this decrease might contribute to the central mechanisms of the development and maintenance of pathological pain[21,22].
As shown above, inhibition of glutamate uptake produces pronociceptive effects in normal animals [8]. Unexpectedly, in pathological pain states, inhibition of glutamate transporter activity produced antinociceptive effects. For example, glutamate transporter inhibitors attenuated the induction of allodynia induced by PGE2, PGF2α, and NMDA [9]. Inhibition or transient knockdown of spinal GLT-1 led to a significant reduction of nociceptive behavior in the formalin model [10]. Consistent with these findings, the preliminary work from Yuan-Xiang Tao's laboratory showed that three different glutamate transporter inhibitors (TBOA, DHK, threo-3-hydroxyaspartate) reduced formalin-induced nociceptive responses and Complete Freund's adjuvant (CFA)-evoked thermal hyperalgesia [11]. On the other hand, the glutamate transporter activator MS-153, which is reported to accelerate glutamate uptake in in vivo and in vitro studies [23-26], had no effect in formalin tests when MS-153 was applied via intrathecal injection, even at the highest dose (1,000 μg/10 μl) [11]. Interestingly, Sung et al. reported that riluzole, a glutamate transporter regulator, significantly attenuated thermal hyperalgesia and mechanical allodynia after chronic constriction nerve injury [21], but this drug was ineffective against peripheral neuropathic pain in a clinical setting [27]. The reason for the discrepancy between the two studies is unclear, but it is worth noting that, in addition to increasing glutamate uptake, riluzole has multiple actions on many systems [neuroprotective, anticonvulsant, anxiolytic, and anesthetic qualities by its blockade of sodium channel α-subunits, glutamate receptors, and γ-aminobutyric acid (GABA) reuptake and its stabilization of voltage-gated ion channels] [28-31]. Thus, more selective drugs that promote spinal glutamate transporter function are needed to demonstrate whether glutamate transporter activators have possible efficacy in the treatment of chronic pain.
The intriguing question remains as to why glutamate transporter inhibitors have antinociceptive effects under pathologic pain conditions that are opposite to their pro-nociceptive effects under normal conditions. Several mechanisms potentially contribute to the role of the glutamate transporter inhibitors under pathological pain states (Fig. 4). First, the blockade of spinal glutamate transporter uptake inhibits clearance of glutamate, leading to the chronic elevation of spinal extracellular glutamate, possibly subsequently causing excitotoxicity, compromising or destroying subsceptible dorsal horn neurons, and interfering with the transmission of pain signaling. However, preliminary data from Dr. Tao's laboratory showed that transient glutamate transporter inhibition did not produce significant spinal neuronal damage in rat formalin or CFA model [11]. Second, GABA, an inhibitory transmitter, is synthesized from glutamate by glutamic acid decarboxylase. Do the increased glutamate levels caused by glutamate transporter inhibition lead to increased GABA in the spinal cord? Recent work showed that inhibition of glutamate transporter activity depleted both glutamate and GABA neurotransmitter pools and reduced inhibitory postsynaptic current (IPSC) and miniature IPSC amplitudes [32,33]. Thus, if this property extends to the spinal cord, one would expect that blockade of spinal glutamate transporter would decrease the amount of GABA in GABAergic terminals and reduce IPSP or IPSC. This expectation would not explain the mechanisms of antinociception by glutamate transporter blocker in pathological pain. Third, inhibition of reuptake through presynaptic EAAC1 and/or the glutamate/glutamine cycle in the spinal cord might result in a depletion of glutamate in synaptic vesicles and a decrease in presynaptically released glutamate, leading to a reduction in glutamate receptor-mediated nociceptive transmission. It is documented that abundant glutamate is distributed in intracellular space, particularly inside nerve terminals [3,34,35]. As a precursor for transmitter glutamate, glutamine is also rich in the intracellular and extracellular fluid [3]. Moreover, although it has been demonstrated in vitro that glutamine is a precursor of transmitter glutamate, the in vivo evidence regarding the glutamate/glutamine cycle is less convincing [36]. The ability of glutamatergic neurons to sustain release of glutamate independently of glutamine might be related to a newly found capacity for pyruvate carboxylation [37,38]. Pyruvate carboxylation replenishes the loss of α-ketoglutarare from the tricarboxylic acid cycle that is inherent in release of glutamate. Thus, inhibition of spinal glutamate transporter might not cause the depletion of glutamate in presynaptic vesicles in pathological pain. Fourth, chronic elevation of extracellular glutamate caused by glutamate uptake inhibition might activate the inhibitory presynaptic metabotropic glutamate receptors (mGluRs) [39-41] and promote a postsynaptic desensitization of glutamate receptors [40]. It is possible that, during pathological pain conditions, glutamate transporter inhibitor-produced antinociception might be due to the decreased release of pre-synaptic glutamate via activation of inhibitory mGluRs in primary afferent terminals and/or reduced postsynaptic efficacy of glutamate via desensitization of glutamate receptors in the spinal dorsal horn neurons. Finally, glutamate transporter inhibitors might produce antinociception in pathological pain by the mechanism of blocking inverse operation of the glutamate transporter. It is well documented that the glutamate transporter imports one glutamate ion and co-transports three Na+ ions into the cell [42] and that transporter function is dependent upon both the membrane potential and the transmembrane ion gradients established by Na+-K+ATPase as driving forces [43,44]. Under physiological conditions, these forces are sufficient to maintain the concentration gradient of micromolar extracellular glutamate against millimolar intracellular glutamate through glutamate transporter uptake [3,42]. However, under pathological conditions, metabolic insults that deplete intracellular energy and alter ionic gradients can lead to reversed action of the glutamate transporter [3]. For example, during brain ischemia, ATP is depleted and impairment of Na+-K+ATPase results in the increases in intracellular Na+ ions and extracellular K+ ions, which causes inverse operation of the glutamate transporter and release of glutamate into the extracellular space [3]. Indeed, the glutamate transporter inhibitors (e.g., TBOA) reduce glutamate release and have neuroprotective actions in brain ischemia [20]. Does pathological persistent pain cause cellular energy insufficiency that inverses the glutamate transporter operation to release glutamate in the spinal cord? Metabolic activity and energy demand significantly increase in the spinal cord under pathological pain conditions [45-48]. Such hyperactive states of spinal neuronal and glial cell activities might not only consume large amounts of cell energy, but also disturb energy metabolism, decrease ATP, and result in energy insufficiency that might reverse spinal glutamate transporter operation to release glutamate. It is possible that blocking the reversed glutamate transporter-mediated glutamate release is an underlying mechanism of antinociception produced by glutamate transporter inhibition under chronic pain conditions.
Figure 4 Two distinct models for the role of glutamate (Glu) transporter inhibitors in pathological pain. Pathological pain might cause Glu uptake decrease and energy insufficiency in the spinal cord. The latter may, in turn, results in Glu uptake decrease as well as inverse operation of spinal Glu transporter to release Glu. In one model, Glu transporter inhibitors further block Glu uptake and enhance the increase in extracellular Glu levels, perhaps leading to dorsal horn neuronal death, increase of spinal GABA contents, activation of inhibitory presynaptic mGluRs, and desensitization of postsynaptic Glu receptors. Glu transporter inhibitors also block neuronal Glu transporter reuptake and/or the Glu/Gln cycle via inhibition of glial Glu transporter, resulting in Glu depletion of synaptic vesicles in primary afferent terminals. In contrast, Glu transporter inhibitors block inverse operation of Glu transporter to release Glu and decrease extracellular Glu levels in another model.
Taken together, it is evident that at least five potential mechanisms are involved in the action of glutamate transporter inhibitors during pathological pain (Fig. 4). In the first four mechanisms, glutamate transporter inhibitors lead to an increase in spinal extracellular glutamate levels, whereas, in the last one, glutamate transporter inhibitors block the reversed glutamate transporter-mediated glutamate release, and reduce extracellular glutamate levels (Fig. 4). Therefore, two distinct models explain the role of spinal glutamate transporter in pathological pain (Fig. 4). Determining extracellular glutamate levels in the spinal cord following glutamate transporter inhibition during pathological pain might be a key to determine the mechanisms of glutamate transporter inhibitor-produced antinociception in the state of pathological pain.
Conclusion
Pathological pain, particularly as a result of nerve injury, is poorly managed by current drugs, such as opioids and non-steroidal anti-inflammatory drugs. Glutamate receptor antagonists are effective in reducing pain hypersensitivity in animal models and clinical settings, but with unacceptable side effects. Glutamate transporter inhibitors have recently been shown to produce antinociceptive effects in several preclinical pathological pain models. Further studies to delineate the role of the spinal glutamate transporters during chronic pain states might lead to better strategies for the prevention and therapy of chronic pain.
Acknowledgements
This work was supported by the Johns Hopkins University Blaustein Pain Research Fund and in part by NIH grant NS44219. The corresponding author would like to thank Drs. John A. Ulatowski and Roger A. Johns for their support. The authors thank Tzipora Sofare, MA, for her editorial assistance.
==== Refs
Mayer BL Westbrook GL The physiology of excitatory amino acids in the vertebrate central nervous system Prog Neurobiol 1987 28 197 276 2883706 10.1016/0301-0082(87)90011-6
Basbaum AI Woolf CJ Pain Curr Biol 1999 9 R429 431 10375535 10.1016/S0960-9822(99)80273-5
Danbolt NC Glutamate uptake Prog Neurobiol 2001 65 1 105 11369436 10.1016/S0301-0082(00)00067-8
Rothstein JD Martin L Levey AI Dykes-Hoberg M Jin L Wu D Nash N Kuncl RW Localization of neuronal and glial glutamate transporter Neurons 1994 13 713 725 10.1016/0896-6273(94)90038-8
Kugler P Schmitt A Glutamate transporter EAAC1 is expressed in neurons and glial cells in the rat nervous system Glia 1999 27 129 142 10417812 10.1002/(SICI)1098-1136(199908)27:2<129::AID-GLIA3>3.0.CO;2-Y
Dehnes Y Chaudhry FA Ullensvang K Lehre KP Storm-Mathisen J Nanbolt NC The glutamate transporter EAAT4 in rat cerebellar Purkinje cells: a glutamate-gated chloride channel concentrated near the synapses in parts of the dendritic membrane facing astroglia J Neurosci 1998 18 3606 3619 9570792
Arriza JL Eliasof S Kavanaugh MP Amara SG Excitatory amino acid 5, a retinal glutamate transporter coupled to a chloride conductance Proc Natl Sci USA 1997 94 4155 4160 10.1073/pnas.94.8.4155
Liaw WJ Stephens RL JrBinns BC Chu Y Sepkuty JP Johns RA Rothstein JD Tao Y-X Spinal glutamate uptake is critical for maintaining normal sensory transmission in rat spinal cord Pain 2005 115 60 70 15836970 10.1016/j.pain.2005.02.006
Minami T Matsumura S Okuda-Ashitaka E Shimamoto K Sakimura K Mishina M Mori H Ito S Characterization of the glutamatergic system for induction and maintenance of allodynia Brain Res 2001 895 178 85 11259776 10.1016/S0006-8993(01)02069-8
Niederberger E Schmidtko A Rothstein JD Geisslinger G Tegeder I Modulation of spinal nociceptive processing through the glutamate transporter GLT-1 Neurosci 2003 116 81 87 10.1016/S0306-4522(02)00547-X
Tao Y-X Tao F Liaw WJ Zhang B Yaster M Rothstein JD Johns RA Evidence for the involvement of spinal cord glutamate transporters in the development of chronic inflammatory pain ASA 2003 Annual Meeting San Francisco, CA, USA Oct. 12–16, 2003
Tao F Liaw WJ Zhang B Yaster M Rothstein JD Johns RA Tao Y-X Evidence of neuronal excitatory amino acid carrier 1 expression in rat dorsal root ganglion neurons and their central terminals Neuroscience 2004 123 1045 51 14751295 10.1016/j.neuroscience.2003.11.026
Tao Y-X Petralia RS Liaw W-J Zhang B Johns RA Rothstein JD Expression and distribution of glutamate transporters in the spinal cord and dorsal root ganglion Soc Neurosci Abstract 2004 Program No. 484.5
Aanonsen LM Wilcox GL Nociceptive action of excitatory amino acids in the mouse: effects of spinally administered opioids, phencyclidine and sigma agonists J Pharmcol Exp Ther 1987 243 9 19
Brambilla A Prudentino A Grippa N Borsini F Pharmacological characterization of AMPA-induced biting behaviour in mice Europ J Pharmacol 1996 305 115 7 10.1016/0014-2999(96)00145-8
Kontinen VK Meert TF Vocalization responses after intrathecal administration of ionotropic glutamate receptor agonists in rats Anesth Analg 2002 95 997 1001 12351283 10.1097/00000539-200210000-00038
Weng H Cata J Dougherty P Glutamate transporters play a key role in sensory transmission of the spinal dorsal horn J Pain 2004 5 10.1016/j.jpain.2004.02.557
Weng H Chen J Cata J Dougherty P Inhibition of glutamate transporters increases hind paw withdrawal frequencies and dorsal horn neurons responses to peripheral stimuli in rats Soc Neurosci Abstr 2004 Program No 291.6
Shimamoto K Lebrun B Yasuda-Kamatani Y Sakaitani M Shigeri Y Yumoto N Nakajima T DL-TBOA, a potent blocker of excitatory amino acid transporters Mol Pharmacol 1998 53 195 201 9463476
Phillis JW Ren J O'Regan MH Transporter reversal as a mechanism of glutamate release from the ischemic rat cerebral cortex: studies with DL-TBOA Brain Res 2000 868 105 12 10841893 10.1016/S0006-8993(00)02303-9
Sung B Lim G Mao J Altered expression and uptake activity of spinal glutamate transporters after nerve injury contribute to the pathogenesis of neuropathic pain in rats J Neurosci 2003 23 2899 910 12684477
Binns BC Huang Y Goettl VM Hackshaw KV Stephens RL Jr Glutamate uptake is attenuated in spinal deep dorsal and ventral horn in the rat spinal nerve ligation model Brain Res 2005 1041 38 47 15804498 10.1016/j.brainres.2005.01.088
Umemura K Gemba T Mizuno A Nakashima M Inhibitory effect of MS-153 on elevated brain glutamate level induced by rat middle cerebral artery occlusion Stroke 1996 27 1624 8 8784139
Shimada F Shiga Y Morikawa M Kawazura H Morikawa O Matsuoka T Nishizaki T Saito N The neuroprotective agent MS-153 stimulates glutamate uptake Eur J Pharmacol 1999 386 263 70 10618478 10.1016/S0014-2999(99)00735-9
Abekawa T Honda M Ito K Inoue T Koyama T Effect of MS-153 on the development of behavioral sensitization to locomotion- and ataxia-inducing effects of phencyclidine Psychopharmacology (Berl) 2002 160 122 31 11875629 10.1007/s00213-001-0958-1
Abekawa T Honda M Ito K Inoue T Koyama T Effect of MS-153 on the development of behavioral sensitization to stereotypy-inducing effect of phencyclidine Brain Res 2002 926 176 80 11814421 10.1016/S0006-8993(01)03164-X
Galer BS Twilling LL Harle J Cluff RS Friedman E Rowbotham MC Lack of efficacy of riluzole in the treatment of peripheral neuropathic pain conditions Neurology 2000 55 971 5 11061253
Kretschmer BD Kratzer U Schmidt WJ Riluzole, a glutamate release inhibitor, and motor behavior Naunyn Schmiedebergs Arch Pharmacol 1998 358 181 90 9750003
Hebert T Drapeau P Pradier L Dunn RJ Block of the rat brain IIA sodium channel alpha subunit by the neuroprotective drug riluzole Mol Pharmacol 1994 45 1055 60 8190096
Mantz J Riluzole CNS Drug Rev 1996 2 40 51
Azbill RD Mu X Springer JE Riluzole increases high-affinity glutamate uptake in rat spinal cord synaptosomes Brain Res 2000 871 175 80 10899284 10.1016/S0006-8993(00)02430-6
Mathews GC Diamond JS Neuronal glutamate uptake contributes to GABA synthesis and inhibitory synaptic strength J Neurosci 2003 23 040 8
Rae C Hare N Bubb WA McEwan SR Broer A McQuillan JA Balcar VJ Conigrave AD Broer S Inhibition of glutamine transport depletes glutamate and GABA neurotransmitter pools: further evidence for metabolic compartmentation J Neurochem 2003 85 503 14 12675927
Ottersen OP Laake JH Reichelt W Haug FM Torp R Ischemic disruption of glutamate homeostasis in brain: quantitative immunocytochemical analysis J Chem Neuroanat 1996 12 1 14 9001944 10.1016/S0891-0618(96)00178-0
Storm-Mathisen J Zhang N Ottersen OP Electron microscopic localization of glutamate, glutamine and GABA at putative glutamatergic and GABAergic synapses Mol Neuropharmacol 1992 2 7 13
Broman J Hassel B Rinvik E Ottersen OP Ottensen OP, Storm-Mathisen J Biochemistry and anatomy of transmitter glutamate Glutamate 2000 Elsevier, Amsterdam 1 44
Hassel B Brathe A Cerebral metabolism of lactate in vivo: evidence for neuronal pyruvate carboxylation J Cereb Blood Flow Metab 2000 20 327 336 10698070 10.1097/00004647-200002000-00014
Hassel B Brathe A Neuronal pyruvate carboxylation supports formation of transmitter glutamate J Neurosci 2000 20 1342 1347 10662824
Scanziani M Salin PA Vogt KE Malenka RC Nicoll RA Use-dependent increases in glutamate concentration activate presynaptic metabotropic glutamate receptors Nature 1997 385 630 4 9024660 10.1038/385630a0
Maki R Robinson MB Dichter MA The glutamate uptake inhibitor L-trans-pyrrolidine-2,4-dicarboxylate depresses excitatory synaptic transmission via a presynaptic mechanism in cultured hippocampal neurons J Neurosci 1994 14 6754 6762 7965076
Dube GR Marshall KC Activity-dependent activation of presynaptic metabotropic glutamate receptors in locus coeruleus J Neurophysiol 2000 83 1141 1149 10712444
Nicholls D Attwell D The release and uptake of excitatory amino acids Trends Pharmacol Sci 1990 11 462 8 1980041 10.1016/0165-6147(90)90129-V
Erecinska M Stimulation of the Na+/K+ pump activity during electrogenic uptake of acidic amino acid transmitters by rat brain synaptosomes J Neurochem 1989 52 135 9 2562804
Sarantis M Attwell D Glutamate uptake in mammalian retinal glia is voltage- and potassium-dependent Brain Res 1990 516 322 5 1973066 10.1016/0006-8993(90)90935-5
Vikman KS Duggan AW Siddall PJ Increased ability to induce long-term potentiation of spinal dorsal horn neurones in monoarthritic rats Brain Res 2003 990 51 7 14568329 10.1016/S0006-8993(03)03385-7
Wieseler-Frank J Maier SF Watkins LR Glial activation and pathological pain Neurochem Int 2004 45 389 95 15145553 10.1016/j.neuint.2003.09.009
Schadrack J Neto FL Ableitner A Castro-Lopes JM Willoch F Bartenstein P Zieglgansberger W Tolle TR Metabolic activity changes in the rat spinal cord during adjuvant monoarthritis Neuroscience 1999 94 595 605 10579220 10.1016/S0306-4522(99)00186-4
Benani A Vol C Heurtaux T Asensio C Dauca M Lapicque F Netter P Minn A Up-regulation of fatty acid metabolizing-enzymes mRNA in rat spinal cord during persistent peripheral local inflammation Eur J Neurosci 2003 18 1904 14 14622223 10.1046/j.1460-9568.2003.02930.x
|
16242033
|
PMC1274343
|
CC BY
|
2021-01-04 16:40:07
|
no
|
Mol Pain. 2005 Oct 21; 1:30
|
utf-8
|
Mol Pain
| 2,005 |
10.1186/1744-8069-1-30
|
oa_comm
|
==== Front
Part Fibre ToxicolParticle and Fibre Toxicology1743-8977BioMed Central London 1743-8977-2-91620970910.1186/1743-8977-2-9ResearchROS-mediated genotoxicity of asbestos-cement in mammalian lung cells in vitro Dopp Elke [email protected] Santosh [email protected] Furquan Ahmad [email protected] Kunal [email protected] Recklinghausen Ursula [email protected] Ursula [email protected]ödelsperger Klaus [email protected] Behnaz [email protected] Stefan [email protected] Qamar [email protected] Institute of Hygiene and Occupational Medicine, University Hospital Essen, Germany2 Fibre Toxicology Division, Industrial Toxicology Research Centre, Lucknow, India3 Institute of Physiological Chemistry, University Hospital Essen, Germany4 Institute of Occupational Medicine, University Hospital Giessen, Germany2005 6 10 2005 2 9 9 23 3 2005 6 10 2005 Copyright © 2005 Dopp et al; licensee BioMed Central Ltd.2005Dopp et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Asbestos is a known carcinogen and co-carcinogen. It is a persisting risk in our daily life due to its use in building material as asbestos-cement powder. The present study done on V79-cells (Chinese hamster lung cells) demonstrates the cytotoxic and genotoxic potential of asbestos-cement powder (ACP) in comparison with chrysotile asbestos. A co-exposure of chrysotile and ACP was tested using the cell viability test and the micronucleus assay. The kinetochore analysis had been used to analyse the pathway causing such genotoxic effects. Thiobarbituric acid-reactive substances were determined as evidence for the production of reactive oxygen species. Both, asbestos cement as well as chrysotile formed micronuclei and induced loss of cell viability in a concentration- and time- dependent way. Results of TBARS analysis and iron chelator experiments showed induction of free radicals in ACP- and chrysotile exposed cultures. CaSO4 appeared to be a negligible entity in enhancing the toxic potential of ACP. The co-exposure of both, ACP and chrysotile, showed an additive effect in enhancing the toxicity. The overall study suggests that asbestos-cement is cytotoxic as well as genotoxic in vitro. In comparison to chrysotile the magnitude of the toxicity was less, but co-exposure increased the toxicity of both.
asbestos cementchrysotilecytotoxicitymicronucleikinetochorefree radicals
==== Body
Background
Asbestos has been well documented to be a carcinogen and co-carcinogen associated with the induction of mesothelioma, lung cancers and other benign lung diseases [1,2]. 'Asbestos' is a generic term for a group of six naturally occurring fibrous silicate minerals. It is grouped into two major classes: Serpentine, which contains a magnesium silicate called chrysotile and Amphibole, which includes crocidolite, amosite, anthophyllite, actinolite and tremolite [3]. Asbestos has been used in more than 3,000 products because of its high tensile strength, relative resistance to acid and temperature, varying textures and degrees of flexibility. It does not evaporate, dissolve, burn, or undergo significant reactions with other chemicals, which make asbestos non-biodegradable and environmentally cumulative. Over 95% of the total commercial asbestos use all over the world is chrysotile asbestos [4]. Chrysotile has the morphology of being curly and pliable [5]. Size, geometry, chemical composition and surface charge of various asbestos types play an important role in interactions with cells that lead to cell injury and disease [6,7]. Respiratory impairment, bronchial asthma, chronic bronchitis was noticed in asbestos cement factory workers [8]. However, in the case of chrysotile asbestos, its positive surface charge is more important than its morphology in rendering a toxic and lytic potential [9]. The iron content in chrysotile, primarily present as a surface contaminant [7] is low (~1–6%), but has to be considered in its toxicity.
Asbestos fibres in the environment can result from mining, milling and weathering of asbestos-bearing rocks, and from the manufacture, wear, and disposal of asbestos-containing products [10]. Because of the widespread use of asbestos, its fibres are ubiquitous in the environment. Indoor air can become contaminated with fibres released from building materials, especially if they are damaged or crumbling. Common sources of asbestos in homes are ceilings, pipe insulation, boiler coverings, wallboard, floor, ceiling tiles, sheets, pipes and jointings, etc. Asbestos-cement products, e.g. roof tiles, contain as much as 11–12% of chrysotile asbestos. As a result of continuing exposure to the weather and to acid rain, the surface of asbestos-cement products becomes corroded and weathered. Cement particles, asbestos fibres and agglomerates of particles and fibres are therefore released from the surface and may be dispersed in air and water in large amounts [11].
The toxicity of asbestos is characterized by a number of processes, among which the production of reactive oxygen and reactive nitrogen species (ROS and RNS) are thought to be the most important ones. Highly reactive oxygen species such as the hydroxyl radical can be produced through Fenton-type reaction catalysed by iron impurities present on the surface. ROS/RNS are also produced in the lungs by the chronic inflammatory reaction produced by the prolonged phagocytic activity of macrophages against the bio-persistent fibres [12]. ROS/RNS can cause various types of DNA damages. The most extensively studied are lesion of 8-oxodeoxyguanosine (8-oxodGuo) or the corresponding base (8-oxoGua). These altered nucleotides can be detected in the DNA of cell lines of human or animal origin after treatment with asbestos fibres [13,14].
Smailyte et al. [15] analyzed the cancer risk in Lithuanian cement producing workers and found that exposure to cement dust may increase lung and bladder cancer. He further reported a dose related risk for stomach cancer. Fatima et al. [16] have reported chromosomal abnormalities in asbestos cement factory workers. Rahman et al. [17] found chromosomal aberrations, sister chromatid exchanges and micronuclei formation in the blood lymphocytes of asbestos cement factory workers in comparison to their controls. Dušinská et al. [12] investigated chromosomal and DNA damage in former asbestos cement plant workers. As discussed above chrysotile is the most commercially exploited variety of asbestos and mostly used as asbestos-cement for building material. There are not many studies that assess the cyto- and genotoxicity of asbestos cement in vitro using cell lines. In the present study, we have investigated if asbestos-cement causes similar effects in cellular systems regarding cytotoxicity and genotoxicity than chrysotile asbestos. The micronucleus assay was applied to test the genotoxic effects of asbestos cement in V79-cells (Chinese hamster lung cells), an established cell culture model. Application of kinetochore analysis, radical measurements and iron chelator experiments gave more informations about the mechanistic background, which seems to be based on the formation of free radicals.
Results
Light microscopy showed the average percentage of fibre sizes in asbestos-cement samples to be 50.3% (< 5 – 10 μm), 31.2% (11–20 μm) and 18.5% (21 – 30 μm) and in chrysotile 49.7% (< 5 – 10 μm), 30.7% (11 – 20 μm) and 19.5% (21 – 30 μm) (Table 1). The cytotoxic potential of asbestos cement, chrysotile asbestos and CaSO4 (negative control) was determined after an exposure time of 24, 48 and 72 hrs. The results show a decrease in cell viability of ACP- and chrysotile-exposed V79-cells with increasing fibre/dust concentrations and exposure times. The results showed chrysotile to be more cytotoxic than the ACP after 24, 48 and 72 hrs exposure (Figure 2). CaSO4 was seen to be negligibly cytotoxic up to the highest concentration (20 μg/cm2) and also did not have any effect at longer exposure times.
Table 1 Analysis of asbestos-cement and chrysotile samples using light microscopy (magnification: 2000×) Data represent the mean of 33 counting.
Sample WHO-fibres F/ml × 106 % of fibres/sample Distribution of fibres according to the length (%)
< 5–10 μm 10–12 μm 20–30 μm
Asbestos cement 144 12.8 50.3 31.2 18.5
Chrysotile 697 100 49.7 30.7 19.5
Table 2 Fibre counting by electron microscopy
Sample Suspension WHO-fibre counts TEM
mg/ml Filter deposit F/mg F/ml Magnification, Number of fields
Original Diluted μg/cm2 n ×106 ×106
Asbestos cement 11 0.05 73.2 14 38.3 425 ×10000, 10 fields
34 12.9 144 ×2000, 2 fields
UICC-chrysotile 5 0.01 13.2 32 486 2432 ×10000, 10 fields
33 139 697 ×2000, 1 field
Figure 1 Transmission electron microscopy pictures of asbestos-cement (A) and chrysotile (B) samples. (Magnification: 2000×)
Figure 2 Cytotoxicity of asbestos-cement in V79-cells. Cells were treated with various doses (1 μg/cm2, 5 μg/cm2, 10 μg/cm2, 20 μg/cm2) of asbestos-cement and chrysotile for 24 hrs (A), 48 hrs (B) and 72 hrs (C). The percentage of decreased cell viability is shown in relation to the untreated control. The cytotoxicity was determined by trypan blue-staining. The experiments were repeated twice. SD < 1%
Figure 3 shows the level of induced micronuclei (MN) by ACP in V79-cells at 24, 48 and 72 hrs consecutively to a concentration of 1, 3, 5 and 10 μg/cm2 of ACP. ACP induced a significant number of micronucleated cells at all applied concentrations after 24 hrs and 48 hrs of exposure with the highest induction at 5 μg/cm2 after 24 hrs. The reduced number of MN after 72 hrs exposure can be explained by increased cytotoxic effects at the applied ACP-concentrations. On comparing ACP to chrysotile the latter induced at a concentration of 1 μg/cm2 almost equal numbers of MN as ACP at a concentration of 3 μg/cm2 (p < 0.01). The results of a co-exposure of ACP (3 μg/cm2) and chrysotile (1 μg/cm2) are shown in Figure 4. Additive effects can be seen through an increased formation of MN as compared to induction by ACP (3 μg/cm2) or chrysotile (1 μg/cm2) alone. The difference between ACP or chrysotile alone and co-exposed V79-cells is statistically not significant.
Figure 3 Micronucleus induction after exposure of V79-cells to different doses of asbestos-cement for 24, 48 and 72 hrs. The experiment was repeated twice. Significance: * p < 0.05, ** p < 0.01, *** p < 0.001.
Figure 4 Micronucleus induction in V79-cells after co-exposure of asbestos and asbestos- cement (chrysotile: 1 μg/cm2, ACP: 3 μg/cm2; exposure time: 48 h). The experiment was repeated twice. Significance: *** p < 0.001.
The kinetochore analysis revealed a slight increase in kinetochore-negative micronuclei in cells exposed to ACP (3 μg/cm2), chrysotile (3 μg/cm2) and co-exposure of ACP (3 μg/cm2) and chrysotile (1 μg/cm2) indicating clastogenic events (p < 0.05). However, the differences compared to the untreated control are statistically not significant. CaSO4 induced a negligible amount of kinetochore-negative micronuclei, which was almost equal to controls. Co-exposure of ACP and chrysotile induced kinetochore-negative micronuclei almost equal to that induced by chrysotile (1 μg/cm2) alone (Table 3).
Table 3 Kinetochore analysis after exposure of V79-cells to asbestos cement, chrysotile and gypsum (negative control) for 48 h. The experiment was repeated twice.
Samples No of scored MN Mean K- (± SD)
Control 200 68.5 ± 0.70
CaSO4 (3 μg/cm2) 200 67 ± 2.12
Asbestos cement (3 μg/cm2) 200 71.5 ± 1.4
Chrysotile (1 μg/cm2) 200 73.5 ± 0.70
Co-exposure (3 μg/cm2 ACP and 1 μg/cm2 chrysotile) 200 75 ± 2.8
Addition of the iron chelators 2,2'-DPD and desferal reduced the number of induced micronuclei (see Fig. 4) to the control level or even lower (Figure 5). The iron chelators 2, 2' DPD and desferal are able to prevent radical formation in cellular systems by complexation of free metal and iron ions. A stronger reducing effect in MN-formation can be observed in chrysotile-exposed V79 cells compared to asbestos cement-exposed cells (Fig. 5A). However, application of desferal induced stronger reducing effects in ACP-exposed V79-cells (Fig. 5B). This reduction is statistically significant (p < 0.05).
Figure 5 Reduction of micronucleus formation in exposed V79-cells after addition of the iron chelators 2,2'-dipyridyl (final concentration: 100 μM) (A) and desferal (final concentration: 100 μM) (B). The treatment of cells with 2,2'-DPD and desferal, respectively, was done simultaneously with the fiber and particle treatment (concentrations: 3 μg/cm2 CaSO4; 3 μg/cm2 asbestos cement; 1 μg/cm2 chrysotile; exposure time: 48 h).
The formation of TBARS was detected in ACP, chrysotile and co-exposed (ACP and chrysotile) V79-cells (Figure 6). Fe/8HQ (1.6 μl/ml) was used as positive control and induced TBARS formation up to 0.25 nmol/mg protein. After 24 h exposure to ACP or chrysotile, V79-cells started to release low levels of TBARS, which enhanced in quantity longer exposure times of 36 h (0.063 nmol/mg, ACP) and 48 h (0.089 nmol/mg, chrysotile), respectively. We observed a delayed formation of TBARS after co-exposure of V79-cells to ACP and chrysotile after 72 h exposure time (0.089 nmol/mg) (Figure 6).
Figure 6 Thiobarbituric acid-reactive substances (TBARS) released by V79-cells exposed to asbestos-cement (3 μg/cm2), chrysotile (1 μg/cm2) or asbestos cement and chrysotile in co-exposure. The experiment was repeated twice. Significance: * p < 0.05
Discussion
The present study demonstrates a time- and concentration- dependent loss of cell viability in chrysotile-exposed V79-cells. These results are in agreement to those found by Hong and Choi [18] in V79-cells. The genotoxicity analysis using micronuclei (MN) as biomarker proved that chrysotile gave a maximum damage to the cells at relatively low concentrations. Similar observations were also found in our earlier studies done with human mesothelial cells (HMC) [19-21] and human peripheral blood lymphocytes [22]. The studies suggest that clastogenic factors are responsible for the genotoxic effect shown as kinetochore-negative MN. In the past Dopp et al. [23], Dopp and Schiffmann [24], Rahman et al. [3] and Poser et al. [21] have shown that clastogenic events caused by chrysotile are responsible for the formation of micronuclei in different cell types. The results of the TBARS analysis in the present study further strengthened chrysotile-induced clastogenic events by suggesting the release of free radicals. The present study showed AC-induced release of TBARS (evidence for lipid peroxidation) after 24 – 48 hrs exposure. The highest amount of MN induction in V79-cells was also found during this period demonstrating an interrelation between these two features. In the case of chrysotile asbestos, a delayed release of TBARS (> 24 hrs exposure) can also be observed. These findings are in concordance with findings of Burmeister et al. [25]. These authors did not observe an increase in Fpg-sensetive sites indicative of oxidative DNA-base modification in asbestos-treated human mesothelial cells up to an exposure time of 24 hrs. Abidi et al. [26] and Afaq et al. [27] reported about the production of high amounts of TBARS and alteration of the GSH redox system by chrysotile fibres. Kopnin et al. [28] showed that fibroblasts as well as mesothelial cells are able to generate reactive oxygen species (ROS) in response to asbestos exposure whereas fibroblasts have a lower ability to produce ROS compared to mesothelial cells.
ACP induced pronounced cytotoxicity and genotoxicity in V79-cells, even though its toxic effects were lower than that of chrysotile both in dosage and induction levels. Tilkes and Beck [29] reported similar findings on macrophages in which asbestos cement caused lower cytotoxicity than UICC-chrysotile. Exposure to the different concentrations of ACP showed increased formation of micronuclei in V79-cells.
The co-exposure of V79-cells to ACP and chrysotile resulted in a weak additive effect. However, the amount of induced kinetochore-negative MN did not vary much from that induced by chrysotile. In summary, it can be stated that both ACP and chrysotile have cytotoxic and genotoxic properties. However, the toxicity of chrysotile is more pronounced than that of ACP. The co-exposure (ACP and chrysotile) of V79-cells showed weak additive genotoxic effects. The release of TBARS in ACP- and chrysotile exposed V79-cells suggests the involvement of free radicals in fibre/dust-induced toxicity.
Methods
Fibres and dust samples
Asbestos cement powder (ACP) was prepared from asbestos cement sheet by grinding with mortar and pestle (Industrial Toxicology Research Centre, Lucknow, India). The main type of asbestos fibre in the ACP-sample is chrysotile asbestos (12.8% fibres/sample, Tab. 1). The sample was sieved through a 30 μm brass sieve. Sterilization was carried out at 120°C for 2 hours (hrs) and the samples were subsequently suspended in sterile PBS buffer. The suspensions were analysed by transmission electron microscopy (Hitachi H600) according to fibrous and non-fibrous material and the number of fibres/ml were calculated according to the counting rules of VDI 3492 (VDI Richtlinie, 1994) (Tab. 2). Further fibre and particle analyses were carried out using light microscopy [30] and transmission electron microscopy (TEM, Phillips). TEM pictures of asbestos-cement and chrysotile samples are shown in Fig. 1 (magnification: 2000×). The large non-fibrous conglomerate material in Fig. 1A represents agglomerates of cement particles.
A final concentration of asbestos cement powder (ACP) in PBS (phosphate buffered saline) of 1.1 mg/ml was used. Chrysotile (UICC) and commercially available CaSO4 (Sigma, Taufkirchen, Germany) were applied as positive and negative controls, respectively.
Cell culture
V79-cells (Chinese hamster lung fibroblast cells) were obtained from ECC (European Collection of Cell Cultures, ECC No.: 86041102). Cells were cultivated in RPMI-1640 (Gibco) with Fetal Calf Serum (10%) (Gibco), L-glutamine (1%) (Gibco) and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin) (Gibco) at 37°C and 5% CO2.
Cytotoxicity test
V79-cells (state of confluence: max. 70%) were treated with ACP, CaSO4 and chrysotile, respectively, at different doses (1 μg/cm2 – 10 μg/cm2) for 24 hrs, 48 hrs and 72 hrs. Cell viability was evaluated immediately after exposure. Treated and untreated cells were harvested by trypsin treatment (Sigma). Cell counting was performed following trypan blue staining. The cell suspension was mixed with an equivalent volume of 0.4% trypan blue solution (Sigma) and subsequently evaluated under the light microscope. The membrane of dead cells is permeable to trypan blue (blue stained cells), whereas living cells remain unstained. Cell viability is expressed as percentage of surviving cells compared to the total number of cells. A substance is considered to be cytotoxic if the decrease in cell viability is ≤ 50%.
Micronucleus assay and kinetochore analysis
For micronucleus (MN) analysis, 2 × 105 V79-cells were seeded in each well of Quadriperm-dishes (Viva-Science, Sartorius, Göttingen, Germany) and cultured overnight. Then the fibre and dust samples were applied for 24 hrs, 48 hrs or 72 hrs at different concentrations. At the end of the exposure times cells were fixed and stored in cold methanol (-20°C) for at least 30 minutes before staining. For micronucleus assay the cells were washed with PBS/CMF (calcium- and magnesium-free phosphate buffered saline) and the nuclei were stained with bisbenzimide (Hoechst 33258, concentration: 5 μg/ml, 4 minutes). The slides were then mounted for fluorescence microscopy and examined for the presence of micronuclei. Each data point represents the mean of 3 treated cultures from different experiments with 2000 nuclei evaluated in each case. The significance was tested by using the Chi2-test.
For further analysis of the induced micronuclei after treatment of cells with ACP or chrysotile, kinetochores were stained by incubating the fixed cell preparations with CREST antibodies (Chemicon, Temecula, CA, U.S.A.) for 1 hr in a humidified chamber at 37°C. After rinsing with PBS with 0.5% Tween 20 (Sigma, Germany), the cells were incubated with fluorescein isothiocyanate (FITC)-conjugated anti-human IgG (Antibodies Incorporated, Davis, USA) for 30 min before applying bisbenzimide. At least 200 micronuclei were examined for the presence of kinetochores in each case. The significance was tested by using the Chi2-test.
Application of iron chelators
The iron chelators 2,2'-dipyridyl (DPD) (Fluka, Germany) and desferal (Novartis, Germany) were used to investigate the reduction of the particle induced genomic effects by binding to metal/iron ions. Herewith, the formation of free radicals can be reduced. DPD (final concentration: 100 μM) and desferal (final concentration: 10 mM) were dissolved in ddH2O and added to the culture medium. The treatment of cells with DPD or desferal was done simultaneously with the fibre/dust treatment (concentrations: ACP 3 μg/cm2, chrysotile 1 μg/cm2, CaSO4 as negative control 3 μg/cm2, exposure time: 48 hrs).
Radical measurement
Thiobarbituric acid-reactive substances (TBARS) were determined as indication for the formation of reactive oxygen species. TBARS were determined in the supernatant after various incubation times. V79-cells were cultivated in Ham's F12 culture medium for 24 hrs. The cells were then exposed to asbestos cement (3 μg/cm2), chrysotile (1 μg/cm2) or asbestos cement and chrysotile in co-exposure for 6 h, 12 h, 24 h, 36 h, 48 h, and 72 h. Cells cultured in Ham's F12 medium were used as negative control. After the different exposure times, 1 ml of the probe was mixed with 200 μl iced trichloracetic acid (30%) to precipitate the protein and thereafter centrifuged at 10,000 rpm for 5 min. Further, 1 ml of the supernatant was incubated with 500 μl thiobarbituric acid (1%) in a water bath at 95°C for 10 min. After centrifuging at 3000 rpm for 5 min, the absorbance of the supernatant was measured in a spectral photometer at 532 nm. The amount of TBARS formed was expressed as malondialdehyde (MDA) equivalents in the supernatant. The concentration of the TBARS was calculated by a calibration curve (standard substance: 1, 1, 3, 3-tetramethoxypropane, Sigma, Germany). The experiments were performed in duplicates.
Statistical analysis
The chi2-test was used for comparison of micronucleus and kinetochore results with the untreated control in each set of experiments.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
QR had the initial idea of performing the investigations together with ED, who had coordinated the experiment and had edited the final version of the manuscript. SY and SG have carried out the genotoxicity studies. KB had drafted the manuscript and BS had participated with him in the genotoxicity studies with different iron chelators. UvR has done the cytotoxicity studies and the TBARS assay with help of UR. FAA had done the light microscopic study of the samples. Finally, KR had done the electron microscopic study of the samples. All authors read and approved the final manuscript.
Acknowledgements
We thank Gabriele Zimmer, Ute Zimmermann and Mohd. Ashquin for their excellent technical assistance. We are also thankful to the Director of the Industrial Toxicology Research Centre, Lucknow, and Human Resource Development of Council of Scientific and Industrial Research, India, for providing facilities. The authors thank Prof. Dr. Gerrit M. Alink (Wageningen University, The Netherlands) for critical reading of the manuscript.
==== Refs
Mossman BT Bignon J Corn M Seaton A Gee JB Asbestos: scientific developments and implications for public policy Science 1990 247 294 301 2153315
Paris C Beniuchou J Saunier F Matayer J Brochard P Thiberville L Nouvet G Smopking status, occupational asbestos exposure and bronchial location of lung cancer Lung Cancer 2003 40 17 24 12660003 10.1016/S0169-5002(02)00538-X
Rahman Q Dopp E Lohani M Schiffmann D Occupational and environmental factors enhancing the genotoxicity of asbestos Inhal Toxicol 2000 12 157 165
Landrigan PJ Asbestos–still a carcinogen N Engl J Med 1998 338 1618 9 9603801 10.1056/NEJM199805283382209
Shukla A Ramos-Nino M Mossman B Cell signaling and transcription factor activation by asbestos in lung injury and disease Int J Biochem Cell Biol 2003 35 198 209 10.1016/S1357-2725(02)00315-1
Guthrie G Mossman B Eds Health effects of mineral dusts 1993 28 Minerological Society of America, Washington D.C 1 584
Kamp DW Weitzman SA The molecular basis of asbestos induced lung injury Thorax 1999 54 638 652 10377212
Lim HH Rampal KG Joginder S Bakar CM Chan KH Vivek TN Respiratory conditions of Malaysian asbestos cement workers Med J Malaysia 2002 340 347 12440274
Light W Wei E Surface charge and asbestos toxicity Nature 1977 265 537 539 834305 10.1038/265537a0
Rahman Q Dhar BB, Thakur DN Asbestos: An occupational and environmental carcinogen Mining and Environment 1995 Oxford & IBH Publ. Co. Pvt. Ltd., New Delhi 549 564
Spurny KR Asbestos fibre release by corroded and weathered asbestos-cement products IARC Sci Publ 1989 90 367 71 2744837
Dušinská M Collins A Kazimirova A Barancokova M Harrington V Volkovova K Staruchova M Horska A Wsolova L Kocan A Petrik J Machata M Ratcliffe B Kyrtopoulos S Genotoxic effects of asbestos in humans Mutat Res 2004 553 91 102 15288536
Kim HN Morimoto Y Tsuda T Ootsuyama Y Hirohashi M Hirano T Tanaka I Lim Y Yun IG Kasai H Changes in DNA 8-hydroxyguanosine formation levels, 8-hydroxyguanosine repair activity, and hOGG1 and hMTH1 mRNA expression in human lung alveolar epithelial cells induced by crocidolite asbestos Carcinogenesis 2001 22 265 269 11181447 10.1093/carcin/22.2.265
Fung H Kow YW Van-Houten B Mossman BT Patterns of 8-hydroxydeoxyguanosine formation in DNA and indications of oxidative stress in rat and human pleural mesothelial cells after exposure to crocidolite asbestos Carcinogenesis 1997 18 825 832 9111221 10.1093/carcin/18.4.825
Smailyte G Kurtinaitis J Andersen A Mortality and cancer incidence among Lithuanian cement producing workers Occup Environ Med 2004 61 529 534 15150393 10.1136/oem.2003.009936
Fatima N Jain AK Rahman Q Frequency of sister chromatid exchange and chromosomal aberrations in asbestos cement workers Br J Ind Med 1991 48 103 105 1998603
Rahman Q Prasad R Das M Pandey US Ashquin M Lohani M Follow up study in an asbestos cement factory with special emphasis on preventive and diagnostic measures Project: Industrial Toxicology research Centre, India, Annual Report 1996
Hong YC Choi SS Cytotoxicity and multinucleate giant cell formation in Chinese hamster lung fibroblast caused by crocidolite and chrysotile J Korean Med Sci 1997 12 99 104 9170013
Dopp E Poser I Papp T Interphase FISH detection of cell cycle genes in asbestos-treated human mesothelial cells, MeT-5A- and mesothelioma cells Cell Molec Biol 2002 48 OL271 OL277 12643444
Lohani M Yadav S Schiffmann D Rahman Q Diallylsulfide attenuates asbestos-induced genotoxicity Toxicol Lett 2003 143 45 50 12697379 10.1016/S0378-4274(03)00112-7
Poser I Rahman Q Lohani M Yadav S Becker HH Weiss DG Schiffmann D Dopp E Modulation of genotoxic effects in asbestos-exposed primary human mesothelial cells by radical scavengers, metal chelators and a glutathione precursor Mutat Res 2004 11:559 19 27 15066570
Bhattacharya K Yadava S Papp T Schiffmann D Rahman Q Reduction of chrysotile asbestos-induced genotoxicity in human peripheral blood lymphocytes by garlic extract Toxicol Lett 2004 153 327 32 15454308 10.1016/j.toxlet.2004.06.005
Dopp E Schuler M Schiffmann D Eastmond DA Induction of micronuclei, hyperdiploidy and chromosomal breakage affecting the centric/pericentric regions of chromosomes 1 and 9 in human amniotic fluid cells after treatment with asbestos and ceramic fibers Mutat Res 1997 377 77 87 9219582
Dopp E Schiffmann D Analysis of chromosomal alterations induced by asbestos and ceramic fibers Toxicol Lett 1998 96–97 155 62 10.1016/S0378-4274(98)00063-0
Burmeister B Schwerdtle T Poser I Hoffmann E Hartwig A Muller WU Rettenmeier AW Seemayer NH Dopp E Effects of asbestos on initiation of DNA damage, induction of DNA-strand breaks, P53-expression and apoptosis in primary, SV40-transformed and malignant human mesothelial cells Mutat Res 2004 558 81 92 15036122
Abidi P Afaq F Arif JM Lohani M Rahman Q Chrysotile-mediated imbalance in the glutathione redox system in the development of pulmonary injury Toxicol Lett 1999 106 31 9 10378448 10.1016/S0378-4274(99)00013-2
Afaq F Abidi P Rahman Q N-acetyl L-cysteine attenuates oxidant-mediated toxicity induced by chrysotile fibers Toxicol Lett 2000 117 53 60 11033233 10.1016/S0378-4274(00)00236-8
Kopnin PB Kravchenko IV Furalyov VA Pylev LN Kopnin BP Cell type-specific effects of asbestos on intracellular ROS levels, DNA oxidation and G1 cell cycle checkpoint Oncogene 2004 23 8834 8840 15480427 10.1038/sj.onc.1208108
Tilkes F Beck EG Cytotoxicity and carcinogenicity of chrysotile fibres from asbestos-cement products IARC Sci Publ 1989 90 190 196 2545608
Trivedi AK Ahmad I Musthapa MS Ansari FA Rahman Q Environmental contamination of chrysotile asbestos and its toxic effects on growth and physiological and biochemical parameters of Lemna gibba Arch Environ Contam Toxicol 2004 47 281 9 15386121 10.1007/s00244-004-3161-7
|
16209709
|
PMC1274344
|
CC BY
|
2021-01-04 16:38:30
|
no
|
Part Fibre Toxicol. 2005 Oct 6; 2:9
|
utf-8
|
Part Fibre Toxicol
| 2,005 |
10.1186/1743-8977-2-9
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-531619755010.1186/1477-7827-3-53ResearchPotential targets of transforming growth factor-beta1 during inhibition of oocyte maturation in zebrafish Kohli Gurneet [email protected] Eric [email protected] Chun [email protected] Department Of Biology, York University, Toronto, Ontario, M3J 1P3, Canada2005 30 9 2005 3 53 53 11 8 2005 30 9 2005 Copyright © 2005 Kohli et al; licensee BioMed Central Ltd.2005Kohli et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
TGF-beta is a multifunctional growth factor involved in regulating a variety of cellular activities. Unlike mammals, the function of TGF-beta in the reproduction of lower vertebrates, such as fish, is not clear. Recently, we showed that TGF-beta1 inhibits gonadotropin- and 17alpha, 20beta-dihydroxyprogesterone (DHP)-induced maturation in zebrafish. The aim of the present study was to investigate the mechanisms underlying this action.
Method
To determine if the effect of TGF-beta1 on oocyte maturation involves transcription and/or translation, ovarian follicles were pre-treated with actinomycin D, a blocker of transcription, and cyclohexamide, an inhibitor of translation, and incubated with hCG or DHP, either alone or in combination with TGF-beta1 and oocyte maturation scored. To determine the effect of TGF-beta1 on mRNA levels of several key effectors of oocyte maturation, three sets of experiments were performed. First, follicles were treated with control medium or TGF-beta1 for 2, 6, 12, and 24 h. Second, follicles were treated with different concentrations of TGF-beta1 (0 to 10 ng/ml) for 18 h. Third, follicles were incubated with hCG in the absence or presence of TGF-beta1 for 18 h. At the end of each experiment, total RNA was extracted and reverse transcribed. PCR using primers specific for 20beta-hydroxysteroid dehydrogenase (20beta-HSD) which is involved in DHP production, follicle stimulating hormone receptor (FSHR), luteinizing hormone receptor (LHR), the two forms of membrane progestin receptor: mPR-alpha and mPR-beta, as well as GAPDH (control), were performed.
Results
Treatment with actinomycin D, a blocker of transcription, reduced the inhibitory effect of TGF-beta1 on DHP-induced oocyte maturation, indicating that the inhibitory action of TGF-beta1 is in part due to regulation of gene transcription. Treatment with TGF-beta1 caused a dose and time-dependent decrease in mRNA levels of 20beta-HSD, LHR and mPR-beta in follicles. On the other hand, TGF-beta1 had no effect on mPR-alpha mRNA expression and increased FSHR mRNA levels. Furthermore, hCG upregulated 20beta-HSD, LHR and mPR-beta mRNA levels, but this stimulatory effect was blocked by TGF-beta1.
Conclusion
These findings suggest that TGF-beta1 acts at multiple sites, including LHR, 20beta-HSD and mPR-beta, to inhibit zebrafish oocyte maturation.
==== Body
Background
Transforming Growth Factor-β1 (TGF-β1) is the prototypical member of the TGF-β family [1,2]. Members of this family are implicated in diverse physiological processes, including reproduction. Three isoforms of TGF-β (TGF-β1, -β2, and -β3) are expressed in the mammalian ovary [2-4]. They have been shown to regulate follicle development, steroidogenesis, oocyte maturation, ovulation and follicular atresia [2-4]. There is molecular evidence for the presence of TGF-β1–3 in fish [5-7]. However, the role of TGF-β in fish reproduction is not well understood. Studies in zebrafish have suggested that TGF-β inhibits oocyte maturation [8]. In the goldfish, TGF-β has been reported to inhibit ovarian steroid production [9].
Ovarian development in fish is broadly divided into two major phases: growth and maturation. During oocyte growth, follicle stimulating hormone (FSH) stimulates production of estradiol-17β from the ovary. Estradiol-17β stimulates the production of vitellogenin by the liver. Vitellogenin is taken up by the developing oocyte and cleaved to yolk protein, which serves as a nutritional reserve for the developing embryo [8,10,11]. Oocyte maturation in teleosts is triggered by the release of leutinizing hormone (LH) from the pituitary. LH stimulates a number of signaling cascades culminating in the production of 17α-hydroxyprogesterone (17α-HP). In the granulosa cells, under the action of 20β-hydroxysteroid dehydrogenase (20β-HSD), 17α-HP is converted to 17α, 20β-dihydroxyprogesterone (DHP), the maturation inducing hormone (MIH) in cyprinids, such as zebrafish and goldfish. MIH activates the cytoplasmic maturation promoting factor (MPF), which is made up of two subunits: cyclin B (a regulatory subunit) and cdc2 (a catalytic subunit). MIH stimulates the de novo synthesis of cyclin B. Cyclin B protein binds to cdc2 to form MPF. The newly formed MPF is activated by phosphorylation of cdc2 on threonine 161. The active MPF, then, stimulates all the changes associated with oocyte maturation, such as germinal vesicle break down (GVBD), spindle formation, chromosome condensation and allows the transition from G2/M phase of meiosis [12-15].
Two isoforms of the MIH receptor, designated as membrane progestin receptor-α (mPR-α) and mPR-β, have recently been cloned in zebrafish [16]. Microinjection of zebrafish oocytes with antisense oligonucleotides to either mPR-α or mPR-β or both receptors has been shown to block MIH-induced maturation, indicating that both play a role in zebrafish oocyte maturation [17]. Originally discovered in sea-trout oocytes, several isoforms of mPR have also been discovered in humans and other vertebrates [16-20].
The zebrafish model has been used extensively for studies on early embryonic development. This model is also very useful for the investigation of ovarian follicle development and maturation because the zebrafish ovary contains ovarian follicles at different stages of development. We and others have been using zebrafish to examine the role of TGF-β superfamily in oocyte maturation [8,21-23]. Our previous work has shown that TGF-β1 inhibits human chorionic gonadotropin (hCG) and MIH-induced maturation in zebrafish [8]. We have also observed a decrease in TGF-β1 mRNA expression in maturing follicles. These findings suggest that TGF-β1 may play a role in preventing premature oocyte maturation in zebrafish. The present study is an attempt to elucidate the mechanisms underlying the inhibitory effect of TGF-β1 by identifying the potential target genes of TGF-β1. We report here our recent findings on the effect of TGF-β1 on the important effectors of oocyte maturation in zebrafish, including: 20β-HSD, LHR, mPR-α, and mPR-β.
Materials and methods
Animals
Adult zebrafish, Danio rerio, were purchased from a local pet supplier (Fish & Bird Emporium, Brampton, ON) and maintained in 10 L tanks of an AHAB System(Aquatic Habitats, FL) at 26°C, under a 14-h light, 10-h dark photoperiod. The fish were fed twice daily with tropical fish food and supplemented with freshly hatched brine shrimp two or three times a week. Experiments were performed according to the Guide to the Care and Use of Experimental Animals published by the Canadian Council on Animal Care.
In vitro culture of zebrafish follicles
Female zebrafish that have a full-grown ovary were anesthetized using 3-aminobenzoicacid ethyl ester (Sigma-Aldrich Canada Inc., Oakville, ON) and decapitated. The ovaries were removed and follicles greater than 0.52 mm in diameter were collected, since previous studies in the zebrafish have shown that only follicles of this size can undergo maturation in vitro in response to hormones [21]. Approximately 20 follicles were placed into each well of a 24-well culture plate and pre-treated at 26°C in either 1 ml of modified Cortland's medium (MCM) or MCM + chemicals such as Actinomycin D or Cyclohexamide (Sigma-Aldrich, Mississauga, ON) for 2 hours. Pre-treated follicles were then incubated with the control medium, human recombinant TGF-β1 (R&D Systems, Minneapolis, MN), hCG (kindly provided by Dr. A. F. Parlow, National Hormone and Peptide Program, Torrance, CA), MIH (17α, 20β-DHP;Sigma-Aldrich Canada), either alone or in combination, as previously described [8]. Maturation was scored after 18 hours of incubation. Follicles that underwent germinal vesicle breakdown (GVBD) could be identified by their acquired translucency.
Total RNA extraction
Approximately 80 follicles per treatment group were used for RNA extraction. Total RNA was extracted using an RNeasy Mini kit (Qiagen Inc., Mississauga, ON) according to the manufacturer's suggested protocol. The RNase-free deoxyribonuclease Set (Qiagen) was also used during RNA isolation to remove any potential genomic DNA contamination.
Reverse Transcription (RT) and polymerase chain reaction (PCR)
Five micrograms of total RNA were reverse transcribed to cDNA at 37°C for 1.5 h in a total volume of 50 μl as previously described [21]. PCR was carried out in the presence of 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 2.0 mM MgCl2, 50 μM deoxynucleotide triphosphate, 1U Hotstar Taq (Qiagen) and 5 pmol primers. The primers for 20β-HSD, FSHR, LHR, mPR-α and mPR-β were designed according to the sequences from GenBank (Table 1). The semi-quantitative RT-PCR was validated by peforming PCR reactions for different cycles to determine the cycle number that generated half maximal PCR product for each gene studied. PCR was carried out on an Eppendorf Master Cycler (Eppendorf AG, Hamburg, Germany) with the cycling profile: 20 s at 94°C, 30 s at 60–65°C (depending on the primer sets used), 30 s at 72°C followed by a 7-minute final extension at 72°C. The PCR products were electrophoresed on 1.5% agarose gels and visualized by ethidium bromide staining. The spot density of each PCR product was determined using the Fluorchem v2.0 Software (Alpha Innotech Corporation, San Leandro, CA). GAPDH was used as an internal control to normalize the variation in mRNA concentration in the RT reaction. The mRNA level of each gene was first normalized with the GAPDH level and then expressed as percent of the control.
Table 1 List of primers and their sequences
Gene Sequence (5'-3') Accession #
20β-HSD Forward: TGC ACG AGT GGT CAA TGT GTC
Reverse: ACT AGC TGT CCA TGC GGC TCT AF298898
FSHR Forward: GGA TTC TTC ACC GTC TTC TCC
Reverse: TGT AGC TGC TCA ACT CAA ACA AY278107
LHR Forward: GGC GAA GGC TAG ATG GCA CAT
Reverse: TCG CCA TCT GGT TCA TCA ATA AY424302
mPR-α Forward: CAG CGC CTA CTT CTT CTC GT
Reverse: CAC TGC ATC ATG AGC CAA AT AY149121
mPR-β Forward: ACA ACG AGC TGC TGA ATG TG
Reverse: ATG GGC CAG TTC AGA GTG AG AY14920
Statistical Analysis
All values are expressed as mean ± SEM of 3–4 replicates in one representative experiment. All experiments were performed three times to confirm the results using different batches of animals. To determine the statistical difference among different groups at the same time point, multiple group comparisons were performed by one-way ANOVA, followed by a Student-Newman-Keuls multiple group comparisons test, using the GraphPad InStat Software (GraphPad Inc., San Diego, CA). P < 0.05 was considered significant.
Results
Effects of transcriptional and translational inhibitors on final oocyte maturation
To determine if the effect of TGF-β on oocyte maturation requires transcription and/or translation, follicles were pretreated with cyclohexamide, an inhibitor of translation, or actinomycin D, an inhibitor of transcription. Pretreatment with cyclohexamide completely blocked hCG- and MIH-induced oocyte maturation. Pretreatment with actinomycin D blocked hCG-, but not MIH-, induced oocyte maturation (Fig. 1A). Interestingly, the inhibitory effect of TGF-β1 on MIH-induced maturation was partially reversed in the presence of actinomycin D (Fig. 1B), suggesting that the inhibitory effect of TGF-β1 on oocyte maturation involves in part transcriptional regulation.
Figure 1 Effects of transcriptional and translational inhibitors on oocyte maturation. Ovarian follicles were pre-treated for 2 hours with 1 mg/ml Actinomycin D, Cyclohexamide or medium (untreated) and then incubated with: (A) Medium only (control), TGF-β1 (10 ng/ml), hCG (100 IU/ml), or a combination of hCG and TGF-β1 or (B) Medium only (control), TGF-β1 (10 ng/ ml), MIH (100 ng/ml) or a combination of MIH and TGF-β1 for 18 hours. The rate of maturation was scored as percentage of follicles that underwent GVBD. Data represent the mean ± SEM of one representative experiment with four replicates. Different letters denote statistical significance (P < 0.05).
Semi-quantitative RT-PCR Validation
Semi-quantitative RT-PCR assays were developed to examine the effect of TGF-β1 on the mRNA expression of 20β-HSD, FSHR, LHR, mPR-α and mPR-β. PCR assays were performed using ovarian cDNA as template for varying cycle numbers and spot densities of the resulting products were measured. A cycle number within the exponential phase of the amplification curve was chosen for quantifying the expression of each gene in subsequent experiments. Accordingly, 20, 23, 30, 31, 33 and 32 cycles of PCR for GAPDH, 20β-HSD, FSHR, LHR, mPR-α and mPR-β, respectively, were selected as the optimal cycle numbers for measuring the levels of mRNA expression (Fig. 2).
Figure 2 Validation of semi-quantitative RT-PCR for GAPDH (A), 20β-HSD (B), FSHR (C), LHR (D), mPR-α (E) and mPR-β (F). PCRs were performed using zebrafish ovarian cDNA as the template, amplified for varying cycle numbers and the density of the PCR products was quantified. Each value represents the mean ± SEM of three replicates in one representative RT-PCR. Representative ethidium bromide stained gel pictures were included. C = negative control; number on each lane represents the number of PCR cycles performed.
Effect of TGF-β1 on the mRNA expression of 20β-Hydroxysteroid dehydrogenase
To test whether TGF-β1 affects the mRNA expression of 20β-HSD, a time course study was performed where follicles were treated with 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours. A set of untreated follicles was included for each time point as controls. Total RNA was extracted and RT-PCRs were performed. A time-dependant decrease in 20β-HSD mRNA expression relative to control levels was found in TGF-β1 treated follicles. A significant inhibitory effect was observed at 24 hours of treatment (Fig. 3A). A dose-response study was performed where follicles were treated with increasing doses of TGF-β1 (0, 0.1, 1 and 10 ng/ml) for 18 hours. Total RNA was extracted from each set of treated follicles and subjected to RT-PCR analyses. A dose-dependant decrease in 20β-HSD expression in response to TGF-β1 treatment was found (Fig. 3B). The strongest inhibition was seen upon treatment with 10 ng/ml of TGF-β1. Finally, the combinational effect of hCG and TGF-β1 on 20β-HSD mRNA expression was tested. Follicles were incubated with control medium, hCG (100 IU/ml), TGF-β1 (10 ng/ml), or hCG plus TGF-β1, for 18 hours. Treatment with hCG resulted in an increase in 20β-HSD mRNA levels. However, treatment with TGF-β1 decreased basal and hCG-induced 20β-HSD mRNA levels (Fig. 3C).
Figure 3 TGF-β1 inhibits mRNA expression of 20β-HSD. (A) Follicles were treated with control medium or 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours. (B) Follicles were treated with different concentrations (0, 0.1,1 and 10 ng/ml) of TGF-β1 for 18 hours. (C) Follicles were treated with control medium, hCG (100 IU/ml), TGF-β1 (10 ng/ml), or hCG+ TGF-β1 for 18 hours. Total RNA was extracted and subjected to RT-PCR using primers for 20β-HSD and GAPDH. Each value represents the mean ± SEM of three replicates in one representative RT-PCR reaction. 20β-HSD mRNA levels are expressed as percent of control after normalized with the GAPHD levels. Different letters above the bars denote statistical significance (P < 0.05). *, P < 0.05 vs. control. The insets show representative ethidium bromide stained gels. GAPDH gels are the same for Figs. 4-7.
Effect of TGF-β1 on the mRNA expression of follicle stimulating hormone receptor
Treatment with TGF-β1 resulted in a significant increase in FSHR mRNA levels from 6 to 18 hours after treatment, with the maximal stimulation at 12 hours post treatment (Fig. 4A, B). The stimulatory effect of TGF-β1 on FSHR mRNA expression was observed for all doses tested (Fig. 4B). Treatment with hCG had no significant effect on basal FSHR mRNA expression when compared to the control (Fig. 4C). Similarly, hCG did not affect TGF-β1-induced FSHR mRNA levels (Fig. 4C).
Figure 4 TGF-β1 stimulates FSHR mRNA expression. (A) Follicles were incubated with control medium or 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours. (B) Follicles were treated with different concentrations (0, 0.1,1 and 10 ng/ml) of TGF-β1 for 18 hours. (C) Follicles were treated with medium (control), hCG (100 IU/ml), TGF-β1 (10 ng/ml), or a combination of hCG and TGF-β1 for 18 hours. At the end of each incubation, total RNA was extracted and reverse transcribed. PCR was carried out using primers for FSHR and GAPDH. Each value represents the mean ± SEM of three replicates in one representative RT-PCR. Statistical significance (P < 0.05) is indicated by either an * or a different letter. The insets show the representative ethidium bromide stained gels.
Effect of TGF-β1 on the mRNA expression of the luteinizing hormone receptor
TGF-β1 significantly decreased LHR mRNA expression at 18 and 24 hours after treatment, but had no effect at 2 to 12 hours post treatment (Fig. 5A, B). At 18 hours after treatment, TGF-β1 inhibited LHR mRNA expression in a dose-dependant manner (Fig. 5B). Treatment with hCG increased LHR mRNA levels. However, when TGF-β1 was added together with hCG, the stimulatory effect of hCG on LHR mRNA expression was blocked (Fig. 5C).
Figure 5 TGF-β1 suppresses LHR mRNA expression. Follicles were treated with (A) 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours; (B) 0.1, 1, or 10 ng/ml of TGF-β1 for 18 hours; and (C) hCG (100 IU/ml), TGF-β1 (10 ng/ml), either alone or in combination, for 18 hours. Each value represents the mean ± SEM of three replicates in one representative RT-PCR. Different letters denote statistical significance (P < 0.05). The insets show the original ethidium bromide stained gels.
Effect of TGF-β1 on the mRNA expression of membrane progestin receptors
No significant effect of TGF-β1 on mPR-α mRNA expression was observed. The time course study showed no significant difference in the mRNA expression of mPR-α at any of the time points examined (Fig. 6A). Similarly, treatment with different concentrations of TGF-β1 for 18 hours did not result in a significant change in mPR-α mRNA levels (Fig. 6B). Neither hCG nor TGF-β1 had a significant effect on mPR-α mRNA expression (Fig. 6C). In contrast to mPR-β, similar treatment with TGF-β1 resulted in a dose- and time-dependent inhibition of mPR-β mRNA levels. The inhibitory effect was observed at 18 and 24 hours after treatment and all doses of TGF-β1 tested caused a significant decrease in mPR-β mRNA levels (Fig. 7A, B). Incubation of follicles with hCG resulted in a strong stimulation of mPR-β mRNA expression. The stimulatory effect of hCG on mPR-β mRNA expression of was partially blocked by TGF-β1 (Fig. 7C).
Figure 6 TGF-β1 has no effect on mPR-α mRNA expression. Follicles were incubated with (A) 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours; (B) different concentrations of TGF-β1 for 18 hours; and (C) medium only (control), hCG (100 IU/ml), TGF-β1 (10 ng/ml) or hCG + TGF-β1 for 18 hours. Each value represents the mean ± SEM of three replicates in one representative RT-PCR. The insets show representative ethidium bromide stained gels. Neither hCG nor TGF-β1 had an effect on mPR-α mRNA expression.
Figure 7 TGF-β1 downregulates mPR-β mRNA expression. (A) Time course study of the effect of TGF-β1 on mPR-β mRNA expression. Follicles were treated with 10 ng/ml of TGF-β1 for 2, 6, 12 and 24 hours. (B) Follicles were treated with different concentrations (0, 0.1,1 and 10 ng/ml) of TGF-β1 for 18 hours. (C) Follicles were incubated with medium (control), hCG (100 IU/ml), TGF-β1 (10 ng/ml) or hCG + TGF-β1 for 18 hours. Each value represents the mean ± SEM of three replicates in one representative RT-PCR. Different letters or "*" denote statistical significance (P < 0.05). The insets show the representative ethidium bromide stained gels.
Discussion
Currently, little is known about the role of TGF-β in regulating ovarian functions in lower vertebrates, such as fish. Recent studies have suggested that TGF-β inhibits steroid production in the goldfish ovary [9] and oocyte maturation in zebrafish [8]. In the present study, we further examined the cellular mechanisms underlying the inhibitory effect of TGF-β in oocyte maturation. We have shown that TGF-β1 inhibits mRNA expression of 20β-HSD, the key enzyme involved in MIH production, as well as LHR and mPR-β. These novel findings suggest that TGF-β1 inhibits multiple targets in the oocyte maturation pathway, both upstream and downstream of MIH.
One of the potential targets of TGF-β1 identified in this study is 20β-HSD. We found that TGF-β1 inhibited basal and hCG-induced 20β-HSD mRNA levels. TGF-β1 alone induced a dose- and time-dependent inhibitory effect on 20β-HSD mRNA expression. Treatment with hCG increased 20β-HSD mRNA levels; however, in the presence of TGF-β1, the effect of hCG was blocked. 20β-HSD activity in the ovary and its stimulation by gonadotropins and cAMP-enhancing drugs has been demonstrated in many species [10,11,24-28]. Our finding that hCG stimulates 20β-HSD is consistent with studies in mammals [24,28] and a recent study in Nile Tilapia, which reported an increase in the mRNA expression of 20β-HSD in response to hCG treatment [26]. The observation of decreased 20β-HSD mRNA levels after TGF-β1 treatment suggests that TGF-β1 may inhibit 20β-HSD activity, leading to a decrease in MIH production. This notion is supported by a recent report that TGF-β1 inhibits the conversion of 17α-HP to DHP in the goldfish [9]. It is possible that one of the actions of TGF-β1 is to decreases MIH production, and thus inhibits oocyte maturation.
In this study, we observed that hCG increased LHR but had no effect on FSHR mRNA levels. Several studies conducted in fish on the effect of gonadotropin on LHR and FSHR expression have yielded inconsistent results. In African and Channel catfish, it has been reported that hCG treatment caused an activation of the LHR and cAMP mediated pathways, as well as a slight increase in FSHR mRNA levels [29-31]. However, a two-fold increase in FSHR expression, but no change in LHR expression in pre-maturational follicles of rainbow trout in response to treatment with partially purified gonadotropins has also been reported [32]. The reason for such discrepancy is unclear, however, it could be due to species specificity or variation in treatment conditions.
An interesting finding from this study is that TGF-β1 differentially regulates FSHR and LHR mRNA levels. We observed that basal FSH receptor mRNA levels were increased, while basal- and hCG-induced LH receptor mRNA levels were decreased, upon incubation of follicles with TGF-β1. TGF-β1 has been reported to regulate LH and FSH binding sites in mammals, however, whether TGF-β1 has a stimulatory or inhibitory effect appears to be species-dependant. TGF-β1 decreased basal and FSH-induced LH binding sites in porcine granulosa cells but enhanced FSH-induced LH binding in rat granulosa cells [33]. FSH induced FSH binding in porcine granulosa cells and this effect was attenuated by TGF-β1. On the other hand, FSH decreased its own binding levels in rat granulosa cells and this effect was also blocked by TGF-β1 [33]. Our finding that TGF-β1 increased FSHR mRNA levels is consistent with previous studies in rat granulosa cells [34,35]. However, our observation that TGF-β1 decreased LHR mRNA levels is opposite to studies in rat [36] and chicken [37] granulosa cells. In these studies, it was reported that TGF-β1 induced LHR mRNA expression. Whether this is due to species variation as in the case of rat and pig or to the difference in follicle development between fish and higher vertebrates awaits more studies in fish species. The finding that TGF-β1 inhibited LHR mRNA expression suggests that the inhibitory effect of TGF-β1 on hCG-induced oocyte maturation may be due, in part, to the downregulation of LH receptor by TGF-β1. Since FSH is known to be a major regulator of oocyte growth [10,11], by stimulating FSH receptor expression TGF-β1 may play a role in promoting follicle growth. These findings, together with our previous observation that TGF-β1 mRNA levels are higher in growing follicles than in maturing follicles, suggest that TGF-β1 may stimulate follicle development and inhibit precocious oocyte maturation in the zebrafish ovary. This possibility will be investigated further in the future.
A recent study has shown that microinjection of zebrafish oocytes with antisense oligonucleotides to either mPR-α or mPR-β or both causes similar marked decreases in the rates of oocyte maturation, suggesting that both subtypes are obligatory for oocyte maturation in zebrafish [17]. In this study, we observed a strong induction of mPR-β mRNA levels by hCG and an inhibitory effect of TGF-β1 on both basal and hCG-induced mPR-β mRNA expression. These findings support the role of mPR-β in oocyte maturation and suggest that one of the mechanisms by which TGF-β1 inhibits hCG- and MIH-induced oocyte maturation is by the downregulation of MIH receptors, specifically mPR-β. However, we found that neither hCG nor TGF-β1 had an effect of mPR-α mRNA levels, suggesting that the two membrane progestin receptors are under differential regulation. It has been reported that hCG caused an increase in the mPR protein expression in sea-trout oocytes [18]. The sea-trout mPR has a higher homology to mPR-α (80%) than to mPR-β (46%). It remains to be determined if hCG regulates mPR-α and mPR-β protein levels in the zebrafish. Recently, Kazeta et al. (2005) reported that hCG did not change mRNA levels of mPR-α and mPR-β at 5 and 10 h after hCG treatment [38]. The difference between this and our studies may be due to the duration of hCG treatment as 18 h was used in our study.
Conclusion
Based on our findings that TGF-β1 downregulates basal and hCG-induced LHR, 20β-HSD, and mPR-β mRNA levels, we propose that TGF-β1 acts upon multiple targets to exert its inhibitory effect on oocyte maturation. TGF-β1 may downregulate LHR, leading to decreased signal transduction and decreased production of 17α-hydroxyprogesterone. TGF-β1 may inhibit basal and gonadotropin-induced MIH production by inhibiting 20β-HSD. Finally, TGF-β1 may inhibit the expression of the MIH receptor, such as mPR-β, on the oocyte surfaces, leading to a reduction of MPF activation and subsequent oocyte maturation. This model can be tested in the future by examining the protein levels on these molecules once antibodies become available. Similarly, the physiological significance of TGF-β on oocyte maturation will be confirmed using loss-of-function approaches.
Acknowledgements
This study was supported by a grant from Natural Science and Engineering Research Council (NSERC) of Canada and an Ontario Premier's Research Excellence Award to CP. CP is a recipient of a Mid-Career Award from Ontario Women's Health Council and CIHR. We thank Dr. A.F. Parlow and NHPP for providing human chorionic gonadotropin.
==== Refs
Massague J TGF-beta signal transduction Annu Rev Biochem 1998 67 753 791 9759503 10.1146/annurev.biochem.67.1.753
Peng C The TGF-beta superfamily and its roles in the human ovary and placenta J Obstet Gynaecol Can 2003 25 834 844 14532952
Knight PG Glister C Local roles of TGF-beta superfamily members in the control of ovarian follicle development Anim Reprod Sci 2003 78 165 183 12818643 10.1016/S0378-4320(03)00089-7
Ingman WV Robertson SA Defining the actions of transforming growth factor beta in reproduction Bioessays 2002 24 904 914 12325123 10.1002/bies.10155
Hardie LJ Laing KJ Daniels GD Grabowski PS Cunningham C Secombes CJ Isolation of the first piscine transforming growth factor beta gene: analysis reveals tissue specific expression and a potential regulatory sequence in rainbow trout (Oncorhynchus mykiss) Cytokine 1998 10 555 563 9722928 10.1006/cyto.1997.0334
Yin Z Kwang J Molecular isolation and characterization of carp transforming growth factor ß1 from activated leucocytes Fish Shellfish Immunol 2000 10 309 318 10938741 10.1006/fsim.1999.0241
Harms CA Kennedy-Stoskopf S Horne WA Fuller FJ Tompkins WA Cloning and sequencing hybrid striped bass (Morone saxatilis x M. chrysops) transforming growth factor-beta (TGF-beta), and development of a reverse transcription quantitative competitive polymerase chain reaction (RT-qcPCR) assay to measure TGF-beta mRNA of teleost fish Fish Shellfish Immunol 2000 10 61 85 10938723 10.1006/fsim.1999.0230
Kohli G Hu S Clelland E Di Muccio T Rothenstein J Peng C Cloning of Transforming Growth Factor-ß1 (TGF-ß1) and Its Type II receptor from Zebrafish Ovary and Role of TGF-ß1 in Oocyte Maturation Endocrinology 2003 144 12697700
Calp MK Matsumoto JA Van Der Kraak G Activin and transforming growth factor-beta as local regulators of ovarian steroidogenesis in the goldfish Gen Comp Endocrinol 2003 132 142 150 12765654 10.1016/S0016-6480(03)00060-1
Nagahama Y Yoshikuni M Yamashita M Tokumoto T Katsu Y Regulation of oocyte growth and maturation in fish Curr Top Dev Biol 1995 30 103 145 7555044
Nagahama Y 17 alpha,20 beta-dihydroxy-4-pregnen-3-one, a maturation-inducing hormone in fish oocytes: mechanisms of synthesis and action Steroids 1997 62 190 196 9029736 10.1016/S0039-128X(96)00180-8
Kajiura-Kobayashi H Yoshida N Sagata N Yamashita M Nagahama Y The Mos/MAPK pathway is involved in metaphase II arrest as a cytostatic factor but is neither necessary nor sufficient for initiating oocyte maturation in goldfish Dev Genes Evol 2000 210 416 425 11180847 10.1007/s004270000083
Katsu Y Yamashita M Nagahama Y Translational regulation of cyclin B mRNA by 17alpha,20beta-dihydroxy-4-pregnen-3-one (maturation-inducing hormone) during oocyte maturation in a teleost fish, the goldfish (Carassius auratus) Mol Cell Endocrinol 1999 158 79 85 10630408 10.1016/S0303-7207(99)00177-X
Kondo T Yanagawa T Yoshida N Yamashita M Introduction of cyclin B induces activation of the maturation-promoting factor and breakdown of germinal vesicle in growing zebrafish oocytes unresponsive to the maturation-inducing hormone Dev Biol 1997 190 142 152 9331337 10.1006/dbio.1997.8673
Kondo T Kotani T Yamashita M Dispersion of cyclin B mRNA aggregation is coupled with translational activation of the mRNA during zebrafish oocyte maturation Dev Biol 2001 229 421 431 11150242 10.1006/dbio.2000.9990
Zhu Y Bond J Thomas P Identification, classification, and partial characterization of genes in humans and other vertebrates homologous to a fish membrane progestin receptor Proc Natl Acad Sci U S A 2003 100 2237 2242 12601167 10.1073/pnas.0436133100
Thomas P Pang Y Zhu Y Detweiler C Doughty K Multiple rapid progestin actions and progestin membrane receptor subtypes in fish Steroids 2004 69 567 573 15288771 10.1016/j.steroids.2004.05.004
Zhu Y Rice CD Pang Y Pace M Thomas P Cloning, expression, and characterization of a membrane progestin receptor and evidence it is an intermediary in meiotic maturation of fish oocytes Proc Natl Acad Sci U S A 2003 100 2231 2236 12574519 10.1073/pnas.0336132100
Hammes SR Niimura S Kawakami SY Senthilkumaran B Sudhakumari CC Chang XT Kobayashi T Oba Y Guan G Yoshiura Y Yoshikuni M Nagahama Y Kazeto Y Adachi S Yamauchi K Steroids and oocyte maturation--a new look at an old story Mol Endocrinol 2004 18 769 775 14630999 10.1210/me.2003-0317
Kazeto Y Goto-Kazeto R Thomas P Trant JM Molecular characterization of three forms of putative membrane-bound progestin receptors and their tissue-distribution in channel catfish, Ictalurus punctatus J Mol Endocrinol 2005 34 781 791 15956347 10.1677/jme.1.01721
Wu T Patel H Mukai S Melino C Garg R Ni X Chang J Peng C Activin, inhibin, and follistatin in zebrafish ovary: expression and role in oocyte maturation Biol Reprod 2000 62 1585 1592 10819759
Pang Y Ge W Gonadotropin and activin enhance maturational competence of oocytes in the zebrafish (Danio rerio) Biol Reprod 2002 66 259 265 11804937
Pang Y Ge W Epidermal Growth factor and TGF-a promote zebrafish oocyte maturation in vitro: potential role of the ovarian activin regulatory system Endocrinology 2002 143 47 54 11751590 10.1210/en.143.1.47
Niimura S Kawakami SY Changes in the activities of hydroxysteroid dehydrogenases in mouse oocytes during meiotic maturation J Reprod Dev 2003 49 451 456 14967895 10.1262/jrd.49.451
Kazeto Y Adachi S Yamauchi K 20beta-Hydroxysteroid dehydrogenase of the Japanese eel ovary: its cellular localization and changes in the enzymatic activity during sexual maturation Gen Comp Endocrinol 2001 122 109 115 11352559 10.1006/gcen.2001.7624
Senthilkumaran B Sudhakumari CC Chang XT Kobayashi T Oba Y Guan G Yoshiura Y Yoshikuni M Nagahama Y Kazeto Y Adachi S Yamauchi K Ovarian carbonyl reductase-like 20beta-hydroxysteroid dehydrogenase shows distinct surge in messenger RNA expression during natural and gonadotropin-induced meiotic maturation in nile tilapia Biol Reprod 2002 67 1080 1086 12297521
Nagahama Y KHYG Stimulation of 17 ,20ß-dihydroxy-4-pregnen-3-one production in the granulosa cells of amago salmon, Oncorhynchus rhodurus, by cyclic nucleotides. J Exp Zool 1985 236 371 375 10.1002/jez.1402360316
Stocco CO Chedrese J Deis RP Luteal expression of cytochrome P450 side-chain cleavage, steroidogenic acute regulatory protein, 3beta-hydroxysteroid dehydrogenase, and 20alpha-hydroxysteroid dehydrogenase genes in late pregnant rats: effect of luteinizing hormone and RU486 Biol Reprod 2001 65 1114 1119 11566732
Vischer HF Bogerd J Cloning and functional characterization of a gonadal luteinizing hormone receptor complementary DNA from the African catfish (Clarias gariepinus) Biol Reprod 2003 68 262 271 12493722 10.1095/biolreprod.102.004515
Kumar RS Ijiri S Trant JM Molecular biology of the channel catfish gonadotropin receptors: 2. Complementary DNA cloning, functional expression, and seasonal gene expression of the follicle-stimulating hormone receptor Biol Reprod 2001 65 710 717 11514332
Kumar RS Ijiri S Trant JM Molecular biology of channel catfish gonadotropin receptors: 1. Cloning of a functional luteinizing hormone receptor and preovulatory induction of gene expression Biol Reprod 2001 64 1010 1018 11207219
Bobe J Maugars G Nguyen T Rime H Jalabert B Rainbow trout follicular maturational competence acquisition is associated with an increased expression of follicle stimulating hormone receptor and insulin-like growth factor 2 messenger RNAs Mol Reprod Dev 2003 66 46 53 12874798 10.1002/mrd.10334
Gitay-Goren H Kim IC Miggans ST Schomberg DW Transforming growth factor beta modulates gonadotropin receptor expression in porcine and rat granulosa cells differently Biol Reprod 1993 48 1284 1289 8318582
Dunkel L Tilly JL Shikone T Nishimori K Hsueh AT Follicle-stimulating hormone receptor expression in the rat ovary: increases during prepubertal development and the regulation by the opposing actions of transforming growth factors and alpha. Biol Reprod 1994 50 940 948 8199274
Inoue K Nakamura K Abe K Hirakawa T Tsuchiya M Oomori Y Matsuda H Miyamoto K Minegishi T Mechanisms of action of transforming growth factor beta on the expression of follicle-stimulating hormone receptor messenger ribonucleic acid levels in rat granulosa cells Biol Reprod 2003 69 1238 1244 12801997 10.1095/biolreprod.102.014753
Inoue K Nakamura K Abe K Hirakawa T Tsuchiya M Matsuda H Miyamoto K Minegishi T Effect of transforming growth factor beta on the expression of luteinizing hormone receptor in cultured rat granulosa cells Biol Reprod 2002 67 610 615 12135904
Johnson AL Bridgham JT Woods DC Cellular mechanisms and modulation of activin A- and transforming growth factor beta-mediated differentiation in cultured hen granulosa cells Biol Reprod 2004 71 1844 1851 15269104 10.1095/biolreprod.104.032573
Kazeto Y Goto-Kazeto R Trant JM Membrane-bound progestin receptors in channel catfish and zebrafish ovary: changes in gene expression associated with the reproductive cycles and hormonal reagents Gen Comp Endocrinol 2005 142 204 211 15862564 10.1016/j.ygcen.2005.01.017
|
16197550
|
PMC1274345
|
CC BY
|
2021-01-04 16:37:13
|
no
|
Reprod Biol Endocrinol. 2005 Sep 30; 3:53
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-53
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-1121621265910.1186/1465-9921-6-112ResearchThe role of pro- and anti-inflammatory responses in silica-induced lung fibrosis Barbarin Virginie [email protected] Aurélie [email protected] Pierre [email protected] Mohammed [email protected] Monique [email protected] Isabelle [email protected] Dominique [email protected] Francois [email protected] Industrial Toxicology and Occupational Medicine Unit, Faculty of Medicine, Université catholique de Louvain, Clos Chapelle-aux-champs 30.54, 1200 Brussels, Belgium2 Laboratory of Pathology, University Hospital of Mont Godinne, Université catholique de Louvain, Avenue Dr. G. Thérasse 1, 5530 Yvoir, Belgium3 Unit of Gastro-enterology, Faculty of Medicine, Université catholique de Louvain, 53–79, Avenue E. Mounier 53,1200 Brussels, Belgium2005 7 10 2005 6 1 112 112 7 4 2005 7 10 2005 Copyright © 2005 Barbarin et al; licensee BioMed Central Ltd.2005Barbarin et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
It has been generally well accepted that chronic inflammation is a necessary component of lung fibrosis but this concept has recently been challenged.
Methods
Using biochemical, histological, immunohistochemistry, and cellular analyses, we compared the lung responses (inflammation and fibrosis) to fibrogenic silica particles (2.5 and 25 mg/g lung) in Sprague-Dawley rats and NMRI mice.
Results
Rats treated with silica particles developed chronic and progressive inflammation accompanied by an overproduction of TNF-α as well as an intense lung fibrosis. Dexamethasone or pioglitazone limited the amplitude of the lung fibrotic reaction to silica in rats, supporting the paradigm that inflammation drives lung fibrosis.
In striking contrast, in mice, silica induced only a limited and transient inflammation without TNF-α overproduction. However, mice developed lung fibrosis of a similar intensity than rats. The fibrotic response in mice was accompanied by a high expression of the anti-inflammatory and fibrotic cytokine IL-10 by silica-activated lung macrophages. In mice, IL-10 was induced only by fibrotic particles and significantly expressed in the lung of silica-sensitive but not silica-resistant strains of mice. Anti-inflammatory treatments did not control lung fibrosis in mice.
Conclusion
These results indicate that, beside chronic lung inflammation, a pronounced anti-inflammatory reaction may also contribute to the extension of silica-induced lung fibrosis and represents an alternative pathway leading to lung fibrosis.
==== Body
Background
Lung fibrosis is often associated with an inflammatory process which precedes or coexists with fibroblast proliferation and deposition of extracellular matrix proteins [1]. Abundant human and experimental data have highlighted inflammation as a major effector in the development of lung fibrosis [2-4]. Persistent inflammation characterized by an accumulation of macrophages, neutrophils and lymphocytes in the lung causes the release of degradative enzymes and oxidants capable of inducing lung injury and DNA damage [5,6]. Lung inflammatory cells are also a source of growth factors [7], cytokines [8] and chemokines [9] that amplify and maintain alveolitis and activate fibroblasts. It has been demonstrated, for instance, that macrophages obtained from animal models of silicosis [10] or from patients with lung fibrosis [11] overproduce pro-inflammatory cytokines and growth factors such as TNF-α, IL-1, PDGF and TGF-β. All these mediators clearly possess strong stimulating activities on fibroblasts [4,12].
Beside this strong evidence of a major role of inflammation, other studies did not find a clear relationship between lung inflammation and fibrosis and, thus, have challenged this paradigm.
First, several studies, mainly conducted in mice, reported that lung inflammation is not always followed by a fibrotic disease. Adamson and colleagues showed that an increased pulmonary inflammation induced by the leukocyte chemoattractant FMLP (N-formyl-L-methionyl-leucyl-phenylalanine), significantly reduced the fibrotic response induced by silica in mice [13]. In addition, IL-10 deficient mice treated with silica showed an intense alveolitis but a reduced fibrotic lung response compared to their wild-type counterparts [14]. Also, αvβ6 integrin knockout mice developed marked lung inflammation in response to bleomycin but failed to develop fibrosis [15].
Second, the control of inflammation is not always associated with a reduction of fibrosis. In mice, treatment with anti-MIF (macrophage migration inhibitory factor) antibodies significantly reduced the accumulation of inflammatory cells in the alveolar space as well as TNF-α production after treatment with bleomycin but did not affect the lung fibrotic response [16]. Also, IL-12p40-/- mice treated with bleomycin exhibited reduced pulmonary inflammation but increased fibrosis compared to the wild type mice [17].
Finally, in humans, anti-inflammatory therapy has never been definitely shown to significantly alter the course of pulmonary fibrosis [18,19].
Collectively, these observations suggest that inflammation is not necessarily related to the fibrotic response and that additional pathogenic routes can be responsible for the development of a pulmonary fibrotic response. Alternative paradigms have therefore been proposed [20,21].
It has been postulated that pulmonary fibrosis may result from sequential epithelial cell injury and abnormal wound repair, independently from an inflammatory reaction [22]. Angiogenesis may also constitute a pivotal process in the development of lung fibrosis, independent of lung inflammation [23]. We have recently proposed a third pathogenic pathway which is based on the profibrotic activity of anti-inflammatory cytokines such as IL-4, IL-13, TGF-β but also IL-10 [24]. Indeed, these cytokines stimulate, directly or indirectly, the fibroblasts to proliferate and/or to produce extracellular matrix proteins. Their protracted overproduction in the lung, e.g. in response to a sustained insult, may in turn drive a fibrotic process.
In this study, we compared the lung responses to silica in Sprague-Dawley rats and NMRI mice, two animal models largely used to study the pathogenesis of lung fibrosis [25,26]. By comparing these species, we show that the extension (rats) but also the control of the inflammatory reaction (mice) induced by silica particles may both lead to the emergence of similar fibrotic lesions.
Methods
Animals
Female NMRI mice and Sprague-Dawley rats weighing respectively 20 to 25 g and 200 to 300 g, were purchased from Charles River (Brussels, Belgium). Female BALB/c, C57BL/6J, and DBA2 mice were obtained from our local breeding facility (Ludwig Institute, Brussels). The animals were housed in positive pressure air-conditioned units (25°C, 50% relative humidity) on a 12-hr light/dark cycle. The experimental protocol was approved by the local committee for animal use at the Université catholique de Louvain.
Instillation method
To allow sterilization and inactivation of any trace of endotoxin, particles were heated at 200°C for 2 h immediately before suspension and administration.
A suspension of crystalline silica particles (DQ12; d50 = 2.2 μm, a gift from Dr L. Armbruster Essen, Germany) in sterile 0.9% saline was injected directly into the lungs of mice and rats by intratracheal instillation. To allow a comparison between both species, the doses of silica were adjusted to administer 2.5 and 25 mg of silica per g of lung (lung weight; mouse about 200 mg and rat about 1.2 g; ratio = 6; instillation of 0.5 or 5 mg silica in 60 μl of saline in mice and 3 or 30 mg silica in 360 μl saline in rats). These doses of silica are comparable to those usually used in the literature to induce intense lung fibrosis [27,28]. All instillations were performed on anesthetized animals after surgical opening of the neck. Two mg of sodium pentobarbital (Certa, Braine-l'Alleud, Belgium) or a mix of 10 mg of Ketalar (N.V. Warner-Lambert, Zaventem, Belgium) and 2 mg of Rompun (Bayer A6, Leverkussen, Germany) were used to anesthetize rats or mice, respectively.
For the particle comparative model, 2.5 mg of silica, tungsten carbide (WC, d50 = 1 μm) or manganese dioxide (MnO2, d50 = 3.7 μm) particles in sterile 0.9% saline (60 μl) were injected directly into the lungs of mice by intratracheal instillation.
Bronchoalveolar lavage and whole lung homogenates
At selected time intervals after silica treatment (3, 30 and 60 days), mice and rats were sacrificed with sodium pentobarbital (20 mg/mice and 120 mg/rat, i.p) and a bronchoalveolar lavage (BAL) was performed by cannulating the trachea and infusing the lungs twice with sterile 0.9 % saline. The volume of saline used for BAL was determined on the basis of the lung weight (ratio between rat and mouse weight = 6). Since mice are usually lavaged with 1 ml, we used 6 ml for lavaging the rats. BAL fluid fractions were centrifuged (1500 rpm, 10 min, 4°C) and the cell-free supernatant of the first fraction was used for biochemical measurements. The cell pellets of the two BALF fractions were pooled and resuspended with 2 ml of sterile saline for mice and 12 ml for rats. Aliquots of the cell suspensions were used to determine cell numbers (200 cells counted). Cell differentials were performed on cytocentrifuge preparations fixed in methanol and stained with Diff-Quik (Baxter, Lessines, Belgium).
Separately, at day 3, 30 or 60 after treatment, non-lavaged whole lungs were perfused and excised. The right lobes were placed into a Falcon tube chilled on ice and 3 ml (mice) or 18 ml (rats) of cold 0.9% NaCl were added. The content of each tube was then homogenized with a Ultra-Turrax T25 homogenizer (Janke and Kunkel, Brussels, Belgium) during 30 second. The homogenates were kept frozen at -80°C until use.
The time points analyzed in this study were selected to correspond to the peaks of inflammation (3 days) and fibrosis (2 months) in the mouse and rat models [25,29,28].
Biochemical analyses
Lactate dehydrogenase (LDH) activity in BALF was assayed spectrophotometrically by monitoring the reduction of nicotinamide adenine dinucleotide (NAD+) at 340 nm in the presence of lactate. Total proteins in BALF were determined by the pyrogallol red staining method (Technicon RA system; Bayer Diagnostics, Domont, France).
Silica measurement
The amount of silica particles remaining in the lungs of rats and mice was measured after 3, 30 and 60 days following administration. The concentration of silica was determined colorimetrically with the molybdenum blue method after digestion in sodium hypochlorite [30].
Anti-inflammatory therapy
Dexamethasone (2.5 μg/ml) was administered in the drinking water starting 3 days before silica (25 mg/ g lung) or saline instillation both in rats (0.25 mg/kg/d) and mice (0.375 mg/kg/d). Two times per week throughout the experimental protocol, 1.25 mg of dexamethasone phosphate (Sigma) was diluted to 500 ml of drinking water [31]. Pioglitazone (Takeda, Japan, commercialized by Eli Lilly, Belgium) was added to powdered standard rodent chow (0.01% wt/wt, ad libitum) [32]. This treatment started 3 days before silica or saline pulmonary administration (10 and 15 mg/kg/d in rats and mice, respectively). Control animals were given powdered standard lab chow ad libitum and tap water. The selection of dose of anti-inflammatory molecules were based on those reported in the literature to significantly attenuate inflammation in rats and mice [31,33,34,32]. Two months after silica or saline treatment, animals were sacrificed and BALF inflammatory parameters (see above) as well as lung collagen deposition (see below) were quantified.
Collagen assay
Collagen deposition was estimated by measuring the lung hydroxyproline content. Lung homogenates were hydrolyzed in 6N HCl overnight at 110°C. Hydroxyproline was assessed by high- performance liquid chromatography analysis [35] and data are expressed as micrograms of hydroxyproline per ml of lung homogenate.
Enzyme-linked immunosorbent assays (ELISA)
Type I collagen contents were measured in lung homogenate supernatants (5000 rpm, 4°C, for 10 min) using standardized ELISA as previously described [36].
Mouse and rat IL-10 (Biosource International, Camarillo, CA, USA), TNF-α (Pharmingen, BD Biosciences, San Diego, USA) concentrations were measured in lung and BAL supernatants using ELISA kits following the manufacturer's protocols. The detection limits of these ELISA are respectively 0.9, 5, 5 and 5 (pg/ml).
Histopathology and immunohistochemical staining
The left lung of silica-treated or control mice was excised and fixed in Bouin solution (Merck-Belgolabo, Belgium). Paraffin-embedded sections were stained with hematoxylin and eosin or Masson's trichrome for light microscopic examination. For immunohistochemistry stainings, dewaxed and rehydrated tissue sections were subjected to endogenous peroxidase inactivation (0.5% H2O2 for 20 min) followed by three washes of 5 minutes in PGT buffer (phosphate-buffered saline [PBS], 0.05% Tween 20, and 0.02% gelatine). An incubation was then performed for 1 h in a humidified room with a rat monoclonal anti-mouse IL-10 antibody (SXC1) diluted 250 times in PBS. After 3 washes with PGT buffer (5 min each), tissue sections were exposed for 1 h to the second antibody (polyclonal rabbit against rat IgG coupled with peroxidase as second antibody, Dako, Copenhagen, Denmark) diluted 40-fold in PBS supplemented with 1% mouse serum. Tissue sections were then rinsed and washed three times in PGT buffer. The peroxidase activity was revealed by 3-3'-diaminobenzidine tetrahydrochloride (Aldrich, Beerse, Belgium)-H2O2 substrate. The staining was enhanced by incubation in a solution of 0.5% CuSO4 in saline for 15 min. Sections were counterstained with Harris hematoxylin, rinsed, dehydrated, and mounted in DPX (BDH, Poole, UK).
Statistics
Treatment-related differences were evaluated using t tests and one-way analysis of variance, followed by pairwise comparisons using the Student-Newman-Keuls test, as appropriate. Statistical significance was considered at P < 0.05.
Results
Pulmonary inflammation induced by silica particles was persistent in rats but limited in mice
LDH activity, protein levels and neutrophil numbers measured in BALF were used to estimate the amplitude of pulmonary inflammation induced by silica particles both in rats and mice (2.5 and 25 mg/g of lung) (Figure 1). At all time points and in a dose-dependent manner, silica induced a significant increase in BALF LDH and protein levels in rats (Figure 1 A &1 B). Similarly, an accumulation of lung neutrophils (Figure 1 C), macrophages and lymphocytes (data not shown) was observed in silica-treated rats in a dose-related manner. These effects were progressive and the most pronounced 30 and 60 days after particle treatment. Thus, these observations showed the establishment of a chronic alveolitis in silica-treated rats.
Figure 1 Lactate dehydrogenase (LDH) activity, total protein content and neutrophil numbers in bronchoalveolar lavage of Sprague-Dawley (SD) rats (A-C) and NMRI mice (D-F) after intratracheal instillation of saline or silica particles (2.5 or 25 mg/ g lung). Bars represent means +/- SEM of 5 to 7 animals. Significant differences between treated animals and controls: *P < 0.05, **P < 0.01, ***P < 0.001 (Student-Newman-Keuls multiple comparison test). Please note the different scales between mice and rats for LDH and neutrophils.
In mice, BALF LDH and protein levels were found increased only at day 3 after silica treatment and both parameters had returned to control values after 60 days. A dose-dependent recruitment of pulmonary neutrophils was noted at day 3 in the BALF of silica-treated mice but this neutrophil accumulation did not last after 30 days. The modifications of BALF numbers of macrophages and lymphocytes were similar to that observed with neutrophils (data not shown). On the basis of these results, we concluded that mice controlled silica-induced alveolitis and did not develop chronic inflammation.
Rats and mice developed a similar lung fibrotic reaction to silica
To estimate silica-induced lung fibrosis, hydroxyproline and type-1 collagen levels were measured in lung homogenates in both species. A clear accumulation of extracellular matrix components was noted both in rats and mice after 60 days (Figure 2). The amplitude of the lung fibrotic reaction was relatively similar in both species since, at the highest dose tested, silica induced a 2.9- and 2.1-fold increase in OH-proline, in rats and mice respectively (Figure 2 A &2 C). Type-1 collagen contents were 2.2- and 3-fold increased in silica-treated rats and mice, respectively (Figure 2 B &2 D). As shown in Figure 3, both species developed clear silicotic lesions characterized by the formation of well defined and organized silicotic nodules. No significant lung fibrosis was noted at days 3 and 30 in the two silica-treated species (data not shown).
Figure 2 Hydroxyproline and type 1 collagen contents in lung homogenates of SD rats (A-B) and NMRI mice (C-D) 60 days after intratracheal instillation of saline or silica particles (2.5 or 25 mg/ g lung). Bars represent means +/- SEM of 5 to 7 animals. Significant differences between treated animals and controls: *P < 0.05, **P < 0.01, ***P < 0.001 (Student-Newman-Keuls multiple comparison test).
Figure 3 Representative silicotic nodules from (A) SD rats and (B) NMRI mice 60 days after silica instillation (25 mg/g lung). Masson trichrome staining. Magnification 200X.
These results indicated therefore that, while rats and mice treated with equivalent doses of silica developed contrasting inflammatory responses, both species showed in turn similar lung fibrotic reactions.
The lung persistence of silica particles was similar in rats and mice
The amplitude of silica-induced lung inflammation and fibrosis directly depends on the amount of particles retained in the lung. We therefore assessed the amount of silica particles that remained in the lungs in both species. Similar amounts of silica particles were retrieved 30 and 60 days after treatment in both rats and mice (day 30, respectively 96.9 ± 9.8 and 80.4 ± 24.9 percent of the mean silica content measured 3 days, values represent means ± SEM; day 60, respectively 113.1 ± 22.6 and 110.3 ± 28.5 %; n = 5 to 7). On the basis of these data, we excluded a difference in the clearance of silica particles to explain the varying lung responses between rats and mice.
The lung response to silica was characterized by the production of pro-inflammatory mediators (TNF-α) in rats and of anti-inflammatory mediators (IL-10) in mice
It is well demonstrated that pro-inflammatory cytokines such as TNF-α are involved in the pathogenesis of silica-induced lung disease [12]. TNF-α levels were therefore measured in BALF and lung homogenates of rats and mice treated with silica (Figure 4 A–C). A dose-dependent and progressive increase of TNF-α in BALF was observed after silica treatment in rats (Figure 4 A). Similar data were obtained by measuring TNF-α in lung homogenates by ELISA or by assessing the amounts of TNF-α transcripts (semi-quantitative RT-PCR) in BALF cells and whole lungs (data not shown). In striking contrast, no such induction was found in BALF (Figure 4 C) or lung homogenates (data not shown) of silica-treated mice. Since it is well demonstrated that the anti-inflammatory cytokine IL-10 may downregulate the expression of TNF-α, we assessed IL-10 levels in the BALF and lung homogenates of silica-treated rats and mice (Figure 4 B–D). A significant and dose-dependent increase of IL-10 production was observed in lung homogenates of silica-treated mice at day 60 (Figure 4 D). No similar IL-10 induction was noted in treated rats (Figure 4 B). Moreover, it was noteworthy that the basal IL-10 content (saline) was 20-fold higher in the lung of mice than in rats (at day 3, saline rats: 175.5 ± 1.2 vs saline mice: 3618.1 ± 422.8 pg/ml). IL-10 was not detected in BALF neither in rats nor in mice. The differences in IL-10 expression were confirmed by semi-quantitative RT-PCR in BALF cells and lung homogenates (data not shown). No similar effect on the levels of IL-4, IL-13 or IFN-γ was noted in this model. Altogether, these results indicated that the mediators associated to the lung response to silica were opposite in both species. The rat lung response was characterized by the expression of a pro-inflammatory cytokine such as TNF-α while the mouse lung response involved an anti-inflammatory cytokine such as IL-10. These observations also suggested that the limited lung inflammation observed in mice could be related, at least in part, to their increased expression of IL-10.
Figure 4 Levels of TNF-α in BALF and IL-10 in lung homogenates of SD rats (A-B) and NMRI mice (C-D) after intratracheal instillation of saline or silica particles (2.5 or 25 mg/ g lung). Bars represent means +/- SEM of 5 to 7 animals. Significant differences between treated animals and controls: **P < 0.01, ***P < 0.001 (Student-Newman-Keuls multiple comparison test). At day 3, absolute levels of TNF-α in saline rats and mice were respectively 6.3 ± 0.7 and 11.7 ± 1.2 pg/ml. For IL10, 175.5 ± 1.2 and 3618.1 ± 422.8 pg/ml were respectively detected in saline rats and mice (day 3).
IL-10 expression was intimately related to silica-induced lung fibrosis in mice
In order to further explore the role of IL-10 in the establishment of lung fibrosis in mice, we used a mouse model that allows a comparison among three different types of particles (tungsten carbide, WC; manganese dioxide, MnO2; and crystalline silica, SiO2) and the identification of specific events leading to the extension of lung fibrosis [37]. After intratracheal instillation of these mineral dusts, the pulmonary responses in NMRI mice were characterized respectively by no inflammation (NI), resolutive alveolitis (RA) or fibrosing alveolitis (FA). As already observed, a persisting increase of IL-10 production was observed in the FA model (silica) which paralleled the establishment of lung fibrosis (Figure 5 A). No significant change in lung IL-10 content was noted in the saline, NI (poorly soluble particles of low toxicity, WC) or RA (inflammatory but not fibrogenic particles) groups, indicating that IL-10 induction in mice seems specific to the fibrotic process. No similar effect on the levels of IL-4, IL-13 or IFN-γ was noted in this comparative model.
Figure 5 A: Time-dependent IL-10 production in the lung of NMRI mice in the fibrosing alveolitis (FA), resolving alveolitis (RA) and non-inflammatory models (NI). Bars represent means +/- SEM of 5 to 6 animals. Significant differences from controls: *P < 0.05 (Student-Newman-Keuls multiple comparison test). B: Cellular immuno-localization of IL-10 production in the FA model at day 120. Mainly macrophages having phagocytozed silica particles expressed IL-10. C: IL-10 contents in lung homogenates of BALBc, C57BL/6 and DBA/2 mice 60 days after intratracheal instillation of silica (2.5 mg). DBA2 mice developed the most severe fibrotic lesions and expressed the highest IL-10 levels. The results for individual mice are shown. The bars denote the mean values for each group (n = 8 to 9). Significantly different from controls: *P < 0.05 (Student-Newman-Keuls multiple comparison test).
To determine the localization and cellular sources of IL-10 in the fibrotic model (FA), we evaluated lung tissue sections obtained at the late stage of the disease (120 days). This analysis showed that, in silica-treated mice, alveolar macrophages appeared as the major cells expressing IL-10. The corresponding sections showed no staining with control non-immune IgG (data not shown). Using polarized light and high magnification, we found that mainly macrophages that had phagocytozed silica particles expressed IL-10 (Figure 5 B).
To complete the data obtained in mice, we also determined the IL-10 levels in the lung of several strains of mice presenting different sensitivities to silica [38]. BALB/c, C57BL/6J and DBA2 mice received silica particles (2.5 mg/mouse) or saline. Two months after treatment, collagen deposition and histology were assessed to monitor the lung fibrotic response. As previously described [38], we observed that DBA2 mice developed the most severe fibrotic lesions. While C57BL/6 mice developed intermediate lung fibrosis, BALB/c mice responded weakly to silica. This gradient of susceptibility was illustrated by measuring pulmonary collagen contents and by histological analysis (data not shown). Lung IL-10 levels were related to the amplitude of pulmonary fibrosis. Indeed, IL-10 contents were significantly higher after silica treatment in lung homogenates of DBA2 mice in comparison to their respective controls (Figure 5 C). While no difference between saline- and silica-treated groups was noted in C57BL/6, in BALB/c mice silica treatment induced a reduction of IL-10 contents in comparison to saline (Figure 5 C). Levels of IL-10 in lung tissue of saline mice were as follows: BALB/c = 605 ± 36; C57BL/6 = 683 ± 62 and DBA/2 = 1049 ± 123 pg/lung. No similar effect on the levels of IL-4, IL-13 or IFN-γ was noted in this comparative model.
Altogether, we concluded that expression of IL-10 was intimately correlated with the amplitude of silica-induced lung fibrosis.
Reduction of inflammation prevented silica-induced lung fibrosis in rats but not in mice
Our observations suggested two opposite lung responses in association with the development of silica-induced lung fibrosis (inflammatory and anti-inflammatory, respectively in rats and mice). In order to delineate the role of lung inflammation in both models of lung fibrosis, we treated silica-administered rats and mice with anti-inflammatory molecules. Dexamethasone (corticosteroids) and pioglitazone (a peroxisome proliferator-activated receptor-gamma agonist) were used in this study because they have been shown to control lung inflammation [33,34]. In rats, silica-induced lung fibrosis was significantly reduced both after dexamethasone or pioglitazone treatment as estimated by OH-proline or type I collagen lung levels (Figure 6 A and 6 B). This reduction of lung fibrosis was accompanied by a limited accumulation of leukocytes in the lung but not by an amelioration of biochemical parameters (i.e., LDH and protein BALF levels; at day 60) (data not shown). In striking contrast, the anti-inflammatory treatments had no similar effect on the amplitude of lung fibrosis in silica-treated mice. Moreover, pioglitazone administration increased the pulmonary levels of OH-proline and type I collagen after silica (Figure 6 C and 6 D), denoting an exacerbated lung fibrotic process in this group. No significant effect of the anti-inflammatory treatments was observed on lung inflammatory parameters as well as cytokine production (IL-10, IL-4, IL-13 and IFN-γ) in mice. Altogether, these data indicated that the inflammatory process drives the pathogenesis of lung fibrosis in rats but not in mice.
Figure 6 Hydroxyproline and type 1 collagen contents in lung homogenates of SD rats (A-B) and NMRI mice (C-D) 60 days after intratracheal instillation of saline or silica particles (2.5 or 25 mg/ g lung) and after dexamethasone or pioglitazone treatment. Bars represent means +/- SEM of 5 to 7 animals. Significant differences between treated animals and controls: *P < 0.05, **P < 0.01, ***P < 0.001 (Student-Newman-Keuls multiple comparison test).
Discussion
Lung fibrosis, in humans as well as in experimental models, is often associated with pulmonary inflammation characterized by the accumulation of macrophages, lymphocytes and granulocytes [1]. These inflammatory cells release toxic oxygen derivatives and proteolytic enzymes which cause cellular damage and disruption of the extracellular matrix which, in turn, leads to destroying lung architecture [39]. Inflammatory cells are also considered to be a major source of mediators such as cytokines and growth factors, which possess the ability to stimulate fibroblast functions critical to fibrogenesis [1].
This scenario was largely described in models of lung fibrosis induced by silica or asbestos mainly in rats [40]. Indeed, compelling evidence demonstrates that pro-inflammatory cytokines such as IL-1 and TNF-α not only regulate chronic lung inflammation but also fibrosis. For instance, activated alveolar macrophages purified from rats instilled with silica particles release pro-inflammatory mediators such as IL-1 and TNF-α as well as MIP-2, responsible for the persistence of inflammation and the development of fibrosis [10,41]. The fact that lung inflammation and production of pro-inflammatory cytokines leads to the development of fibrosis was also reported in models using bleomycin or ionizing radiation to induce lung fibrosis [12,42]. In the present study, we also found that Sprague-Dawley rats injected with silica particles developed progressive inflammation, with a dramatic accumulation of neutrophils at the latter stage of the disease. As already demonstrated [10,29], the lung response to silica in rats was accompanied by an overproduction of TNF-α. The key role of the inflammatory reaction in the extension of lung fibrosis induced by silica was demonstrated by the efficacy of anti-inflammatory therapy in this study. Indeed, in rats, silica-induced fibrosis was strongly attenuated by dexamethasone or pioglitazone administration (Figure 6).
By contrast, in the mouse model used in this study, the lung response to silica was not associated with chronic inflammation or a significant up-regulation of TNF-α expression. This control of inflammation in mice was accompanied by a pronounced expression of IL-10, both at the basal level as well as in response to silica. The fact that this marked overproduction of IL-10 in the lung may contribute to limit inflammation induced by silica is largely supported by existing experimental studies. In mice genetically deficient in IL-10, we have previously reported that the administration of silica particles induced an enhanced inflammatory reaction compared to wild type animals [14]. Furthermore, IL-10 has been shown to suppress tissue inflammation in other mouse models of lung insult induced by Pneumocystis Carinii infection [43], endotoxin [44], bleomycin [45] or immune complexes [46]. The absence of response to anti-inflammatory drugs in mice further supports the fact that fibrosis was not driven by inflammation in this species.
Our data indicate that despite its anti-inflammatory properties, IL-10 participates to the extension of the fibrotic reaction. An obvious association between the extent of IL-10 overexpression and the amplitude of fibrosis was shown in this study. First, alveolar macrophages, strongly implicated in the pathogenesis of lung fibrosis induced by silica particles [47], were identified as the main cellular source of IL-10 in mice (Figure 5 B). Moreover, using several mineral particles inducing different lung responses, we showed that IL-10 production in the lung of mice was up-regulated during the development of fibrosing alveolitis (FA, silica) but not in the resolutive alveolitis (RA, MnO2) or the non-inflammatory models (NI, WC) (Figure 5 A). In addition, IL-10 was up-regulated in silica-susceptible mice (DBA2) but reduced in a resistant strain (BALB/c) (Figure 5 C). It is noteworthy that these last characteristics recapitulate those reported for the pro-fibrotic and pro-inflammatory cytokine TNF-α in the rats. Indeed, TNF-α is overproduced mainly by activated lung macrophages, induced in the rat lung only by fibrotic particles [10], and overexpressed in silica-treated sensitive animals [48,27]. Together with our previous observations demonstrating a pro-fibrotic activity of IL-10 in the lung of mice treated with silica particles [14,24,49], we can conclude that the overexpression of IL-10 documented in the lung of mice in the present study contributed to the establishment of the fibrotic response.
The mechanism by which IL-10 exerts its pro-fibrotic effect is still unclear. The possibility that IL-10 may directly stimulate fibroblasts has already been tested with conflicting results. Thus, Liu and colleagues showed that IL-10 had no significant effect on human fetal lung fibroblasts [50]. In contrast, it has been reported that IL-10 might downregulate type I procollagen mRNA expression in skin fibroblasts [51], as well as constitutive and TGF-β-stimulated type I collagen mRNA expression in a human lung fibroblast cell line [45]. By using primary cultures of mouse lung fibroblasts, we previously showed that unlike other Th2 cytokines such as IL-4 and IL-13 [52,53], IL-10 did not directly modulate fibrosis-associated functions, such as proliferation and collagen or α-SMA expression [24]. IL-10 could, however, exert its pro-fibrotic action by up-modulating the expression of pro-fibrotic mediators such as TGF-β. Indeed, in IL-10 transgenic mice, the expression of TGF-β was found increased in the lung [54], which is consistent with our previous study showing a significant reduction of TGF-β lung levels in IL-10 deficient mice exposed to silica [24]. Moreover, IL-10 enhances the expression of the type II TGF-β receptor and restores TGF-β responsiveness on activated T cells [55] but similar effects have not been explored in fibroblasts. Recent work from this laboratory showed that, in mice treated with silica, lung overexpression of IL-10 upregulated the production of other Th2 cytokines such as IL-4 and IL-13. Since there is evidence that pulmonary fibrosis is a Th2-mediated process, we speculate that elevated lung levels of IL-10 may also contribute to the progression of fibrosis via its capacity to stimulate Th2 polarized responses. Collectively, these observations strongly indicate that IL-10, an anti-inflammatory/Th2 cytokine, exacerbates the severity and pathology of lung fibrosis, at least in the mouse.
Conclusion
On the basis of the comparative models studied here, we suggest that at least two different types of lung response to silica can lead to the development of lung fibrosis. First, as observed in the Sprague-Dawley rats, lung fibrosis is the consequence of a chronic and exaggerated inflammatory response associated with an overproduction of pro-inflammatory mediators also possessing pro-fibrotic activities such as TNF-α. In mice, marked expression of anti-inflammatory cytokines such as IL-10 have beneficial effects by limiting and controlling inflammation. However, in this species, because of its pro-fibrotic properties, IL-10 participates to the extension of fibrosis. Thus, in mice, the strong anti-inflammatory response established to control inflammation could contribute to the fibrotic reaction induced by silica. These data clearly suggest that several pathogenic routes are responsible for the development of a pulmonary fibrotic response.
It remains however, to learn how to extrapolate these observations to human diseases. Anti-inflammatory therapies used in this study had no effect on lung fibrosis in mice while they were very efficient in the control of the disease in rats. This may indicate that the treatment of pulmonary fibrosis may need to be modulated according to the type of pathogenic mechanism involved.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
VB: planned the experimental design and drafted the manuscript.
AN: participated in the study design and performed biochemical and cellular studies.
PM: participated in the study design and performed biochemical and cellular studies.
MA: participated in the study design and performed animal studies.
MD: participated in the study design and performed histological studies.
IL: participated in the study design and performed anti-inflammatory therapy.
DL: participated in the study design, helped to draft the manuscript and coordinated the research group.
FH: participated in the study design, helped to draft the manuscript and coordinated the research group.
Acknowledgements
We thank Johan Casters, Yousof Yakoub and Francine Uwambayinema for their excellent technical assistance. This work was supported in part by the Fonds de la Recherche Scientifique Médicale and Actions de Recherche Concertées, Communauté française de Belgique, Direction de la Recherche Scientifique. F.H. is a Scientific Research Worker with the Fonds de la Recherche Scientifique (FNRS), Belgium.
==== Refs
Gross TJ Hunninghake GW Idiopathic pulmonary fibrosis N Engl J Med 2001 345 517 525 11519507 10.1056/NEJMra003200
Kolb M Margetts PJ Anthony DC Pitossi F Gauldie J Transient expression of IL-1beta induces acute lung injury and chronic repair leading to pulmonary fibrosis J Clin Invest 2001 107 1529 1536 11413160
Sime PJ Marr RA Gauldie D Xing Z Hewlett BR Graham FL Gauldie J Transfer of tumor necrosis factor-alpha to rat lung induces severe pulmonary inflammation and patchy interstitial fibrogenesis with induction of transforming growth factor-beta1 and myofibroblasts Am J Pathol 1998 153 825 832 9736031
Zhang Y Lee TC Guillemin B Yu MC Rom WN Enhanced IL-1 beta and tumor necrosis factor-alpha release and messenger RNA expression in macrophages from idiopathic pulmonary fibrosis or after asbestos exposure J Immunol 1993 150 4188 4196 8473757
Atzori L Chua F Dunsmore SE Willis D Barbarisi M McAnulty RJ Laurent GJ Attenuation of bleomycin induced pulmonary fibrosis in mice using the heme oxygenase inhibitor Zn-deuteroporphyrin IX-2,4-bisethylene glycol Thorax 2004 59 217 223 14985557 10.1136/thx.2003.008979
Pardo A Ruiz V Arreola JL Ramirez R Cisneros-Lira J Gaxiola M Barrios R Kala SV Lieberman MW Selman M Bleomycin-induced pulmonary fibrosis is attenuated in gamma-glutamyl transpeptidase-deficient mice Am J Respir Crit Care Med 2003 167 925 932 12468440 10.1164/rccm.200209-1007OC
Vanhee D Gosset P Wallaert B Voisin C Tonnel AB Mechanisms of fibrosis in coal workers' pneumoconiosis. Increased production of platelet-derived growth factor, insulin- like growth factor type I, and transforming growth factor beta and relationship to disease severity Am J Respir Crit Care Med 1994 150 1049 1055 7921435
Rom WN Relationship of inflammatory cell cytokines to disease severity in individuals with occupational inorganic dust exposure Am J Ind Med 1991 19 15 27 1846507
Driscoll KE Hassenbein DG Carter J Poynter J Asquith TN Grant RA Whitten J Purdon MP Takigiku R Macrophage inflammatory proteins 1 and 2: expression by rat alveolar macrophages, fibroblasts, and epithelial cells and in rat lung after mineral dust exposure Am J Respir Cell Mol Biol 1993 8 311 318 8383510
Driscoll KE Lindenschmidt RC Maurer JK Higgins JM Ridder G Pulmonary response to silica or titanium dioxide: inflammatory cells, alveolar macrophage-derived cytokines, and histopathology Am J Respir Cell Mol Biol 1990 2 381 390 2157474
Vanhee D Gosset P Marquette CH Wallaert B Lafitte JJ Gosselin B Voisin C Tonnel AB Secretion and mRNA expression of TNF alpha and IL-6 in the lungs of pneumoconiosis patients Am J Respir Crit Care Med 1995 152 298 306 7599837
Piguet PF Collart MA Grau GE Sappino AP Vassalli P Requirement of tumour necrosis factor for development of silica- induced pulmonary fibrosis Nature 1990 344 245 247 2156165 10.1038/344245a0
Adamson IYR Prieditis H Bowden DH Instillation of Chemotactic Factor to Silica-Injected Lungs Lowers Interstitial Particle Content and Reduces Pulmonary Fibrosis Am J Pathol 1992 141 319 326 1323217
Huaux F Louahed J Hudspith B Meredith C Delos M Renauld JC Lison D Role of interleukin-10 (IL-10) in the lung response to silica in mice Am J Respir Cell Mol Biol 1998 18 51 59 9448045
Munger JS Huang X Kawakatsu H Griffiths MJ Dalton SL Wu J Pittet JF Kaminski N Garat C Matthay MA Rifkin DB Sheppard D The integrin alpha v beta 6 binds and activates latent TGF beta 1: a mechanism for regulating pulmonary inflammation and fibrosis Cell 1999 96 319 328 10025398 10.1016/S0092-8674(00)80545-0
Tanino Y Makita H Miyamoto K Betsuyaku T Ohtsuka Y Nishihira J Nishimura M Role of macrophage migration inhibitory factor in bleomycin-induced lung injury and fibrosis in mice Am J Physiol Lung Cell Mol Physiol 2002 283 L156 L162 12060572
Sakamoto H Zhao LH Jain F Kradin R IL-12p40(-/-) mice treated with intratracheal bleomycin exhibit decreased pulmonary inflammation and increased fibrosis Exp Mol Pathol 2002 72 1 9 11784117 10.1006/exmp.2001.2409
Lynch JPIII White E Flaherty K Corticosteroids in idiopathic pulmonary fibrosis Curr Opin Pulm Med 2001 7 298 308 11584180 10.1097/00063198-200109000-00009
Mason RJ Schwarz MI Hunninghake GW Musson RA NHLBI Workshop Summary. Pharmacological therapy for idiopathic pulmonary fibrosis. Past, present, and future Am J Respir Crit Care Med 1999 160 1771 1777 10556155
Selman M King TE Pardo A Idiopathic pulmonary fibrosis: Prevailing and evolving hypotheses about its pathogenesis and implications for therapy Annals of Internal Medicine 2001 134 136 151 11177318
Noble PW Homer RJ Idiopathic pulmonary fibrosis: new insights into pathogenesis Clin Chest Med 2004 25 749 758 15564020 10.1016/j.ccm.2004.04.003
Selman MS Thannickal VJ Pardo A Zisman DA Martinez FJ Lynch JP Idiopathic pulmonary fibrosis - Pathogenesis and therapeutic approaches Drugs 2004 64 405 430 14969575
Burdick MD Murray LA Keane MP Xue YY Zisman DA Belperio JA Strieter RM CXCL11 attenuates bleomycin-induced pulmonary fibrosis via inhibition of vascular remodeling Am J Respir Crit Care Med 2005 171 261 268 15502109 10.1164/rccm.200409-1164OC
Barbarin V Arras M Misson P Delos M McGarry B Phan SH Lison D Huaux F Characterization of the effect of interleukin-10 on silica-induced lung fibrosis in mice American Journal of Respiratory Cell and Molecular Biology 2004 31 78 85 14975940 10.1165/rcmb.2003-0299OC
Lardot C Huaux FA Broeckaert FR Declerck PJ Delos M Fubini B Lison DF Role of urokinase in the fibrogenic response of the lung to mineral particles Am J Respir Crit Care Med 1998 157 617 628 9476881
Mariani TJ Roby JD Mecham RP Parks WC Crouch E Pierce RA Localization of type I procollagen gene expression in silica- induced granulomatous lung disease and implication of transforming growth factor-beta as a mediator of fibrosis Am J Pathol 1996 148 151 164 8546202
Carter JM Driscoll KE The role of inflammation, oxidative stress, and proliferation in silica- induced lung disease: a species comparison J Environ Pathol Toxicol Oncol 2001 20 Suppl 1 33 43 11570672
Blake TL DiMatteo M Antonini JM McCloud CM Reasor MJ Subchronic pulmonary inflammation and fibrosis induced by silica in rats are attenuated by amiodarone Exp Lung Res 1996 22 113 131 8838139
Huaux F Lasfargues G Lauwerys R Lison D Lung toxicity of hard metal particles and production of interleukin-1, tumor necrosis factor-alpha, fibronectin, and cystatin-c by lung phagocytes Toxicol Appl Pharmacol 1995 132 53 62 7747285 10.1006/taap.1995.1086
Sweet DV F.R. W Crable JV Spectrophotometric determination of free silica Am Ind Hyg Assoc J 1973 34 500 506 4358769
Weiner FR Czaja MJ Giambrone MA Takahashi S Biempica L Zern MA Transcriptional and Posttranscriptional Effects of Dexamethasone on Albumin and Procollagen Messenger-Rnas in Murine Schistosomiasis Biochemistry 1987 26 1557 1562 3109466 10.1021/bi00380a010
Kawaguchi K Sakaida I Tsuchiya M Omori K Takami T Okita K Pioglitazone prevents hepatic steatosis, fibrosis, and enzyme-altered lesions in rat liver cirrhosis induced by a choline-deficient L-amino acid-defined diet Biochemical and Biophysical Research Communications 2004 315 187 195 15013444 10.1016/j.bbrc.2004.01.038
Kim J McKinley L Siddiqui J Bolgos GL Remick DG Prevention and reversal of pulmonary inflammation and airway hyperresponsiveness by dexamethasone treatment in a murine model of asthma induced by house dust American Journal of Physiology-Lung Cellular and Molecular Physiology 2004 287 L503 L509 15132954 10.1152/ajplung.00433.2003
Birrell MA Patel HJ McCluskie K Wong S Leonard T Yacoub MH Belvisi MG PPAR-gamma agonists as therapy for diseases involving airway neutrophilia European respiratory journal 2004 24 18 23 15293600 10.1183/09031936.04.00098303
Biondi PA Chiesa LM Storelli MR Renon P A new procedure for the specific high-performance liquid chromatographic determination of hydroxyproline J Chromatogr Sci 1997 35 509 512 9358626
Huaux F Arras M Tomasi D Barbarin V Delos M Coutelier JP Vink A Phan SH Renauld JC Lison D A profibrotic function of IL-12p40 in experimental pulmonary fibrosis J Immunol 2002 169 2653 2661 12193738
Huaux F Lardot C Arras M Delos M Many MC Coutelier JP Buchet JP Renauld JC Lison D Lung fibrosis induced by silica particles in NMRI mice is associated with an upregulation of the p40 subunit of interleukin-12 and Th-2 manifestations American Journal of Respiratory and Critical Care Medicine 1999 20 561 572
Callis AH Sohnle PG Mandel GS Wiessner J Mandel NS Kinetics of inflammatory and fibrotic pulmonary changes in a murine model of silicosis J Lab Clin Med 1985 105 547 553 2985721
Weiss SJ Tissue destruction by neutrophils [see comments] N Engl J Med 1989 320 365 376 2536474
Bissonnette E Rola-Pleszczynski M Pulmonary inflammation and fibrosis in a murine model of asbestosis and silicosis. Possible role of tumor necrosis factor Inflammation 1989 13 329 339 2753523 10.1007/BF00914399
Driscoll KE Maurer JK Higgins J Poynter J Alveolar Macrophage Cytokine and Growth-Factor Production in A Rat Model of Crocidolite-Induced Pulmonary Inflammation and Fibrosis Journal of Toxicology and Environmental Health 1995 46 155 169 7563215
Johnston CJ Piedboeuf B Rubin P Williams JP Baggs R Finkelstein JN Early and persistent alterations in the expression of interleukin-1 alpha, interleukin-1 beta and tumor necrosis factor alpha mRNA levels in fibrosis-resistant and sensitive mice after thoracic irradiation Radiat Res 1996 145 762 767 8643837
Ruan S Tate C Lee JJ Ritter T Kolls JK Shellito JE Local delivery of the viral interleukin-10 gene suppresses tissue inflammation in murine Pneumocystis carinii infection Infect Immun 2002 70 6107 6113 12379687 10.1128/IAI.70.11.6107-6113.2002
Dokka S Malanga CJ Shi X Chen F Castranova V Rojanasakul Y Inhibition of endotoxin-induced lung inflammation by interleukin-10 gene transfer in mice Am J Physiol Lung Cell Mol Physiol 2000 279 L872 L877 11053022
Arai T Abe K Matsuoka H Yoshida M Mori M Goya S Kida H Nishino K Osaki T Tachibana I Kaneda Y Hayashi S Introduction of the interleukin-10 gene into mice inhibited bleomycin-induced lung injury in vivo Am J Physiol Lung Cell Mol Physiol 2000 278 L914 22 10781421
Mulligan MS Jones ML Vaporciyan AA Howard MC Ward PA Protective effects of IL-4 and IL-10 against immune complex- induced lung injury J Immunol 1993 151 5666 5674 7901280
Olbruck H Seemayer NH Voss B Wilhelm M Supernatants from quartz dust treated human macrophages stimulate cell proliferation of different human lung cells as well as collagen-synthesis of human diploid lung fibroblasts in vitro Toxicol Lett 1998 96-97 85 95 9820652 10.1016/S0378-4274(98)00054-X
Seiler F Rehn B Rehn S Bruchs J Significant differences in the cellular and molecular reactions of rat and hamster lung after quartz exposure Toxicol Lett 2001 119 11 19 11275417 10.1016/S0378-4274(00)00289-7
Barbarin V Xing Z Delos M Lison D Huaux F Pulmonary overexpression of IL-10 augments lung fibrosis and TH2 responses induced by silica particles. Am J Physiol Lung Cell Mol Physiol 2005 Epub ahead of print
Liu X Kohyama T Wang H Zhu YK Wen FQ Kim HJ Romberger DJ Rennard SI Th2 cytokine regulation of type I collagen gel contraction mediated by human lung mesenchymal cells Am J Physiol Lung Cell Mol Physiol 2002 282 L1049 L1056 11943670
Wangoo A Laban C Cook HT Glenville B Shaw RJ Interleukin-10- and corticosteroid-induced reduction in type I procollagen in a human ex vivo scar culture Int J Exp Pathol 1997 78 33 41 9166103 10.1046/j.1365-2613.1997.d01-241.x
Sempowski GD Derdak S Phipps RP Interleukin-4 and interferon-gamma discordantly regulate collagen biosynthesis by functionally distinct lung fibroblast subsets J Cell Physiol 1996 167 290 296 8613470 10.1002/(SICI)1097-4652(199605)167:2<290::AID-JCP13>3.0.CO;2-C
Ingram JL Rice A Geisenhoffer K Madtes DK Bonner JC Interleukin-13 stimulates the proliferation of lung myofibroblasts via a signal transducer and activator of transcription-6-dependent mechanism - A possible mechanism for the development of airway fibrosis in asthma Chest 2003 123 422S 424S 12629010 10.1378/chest.123.3_suppl.422S
Lee CG Homer RJ Cohn L Link H Jung S Craft JE Graham BS Johnson TR Elias JA Transgenic overexpression of IL-10 in the lung causes mucus metaplasia, tissue inflammation and airways remodeling Via IL-13-dependent and - independent pathways Journal of Biological Chemistry 2002
Cottrez F Groux H Regulation of TGF-beta response during T cell activation is modulated by IL-10 J Immunol 2001 167 773 778 11441082
|
16212659
|
PMC1274346
|
CC BY
|
2021-01-04 16:36:26
|
no
|
Respir Res. 2005 Oct 7; 6(1):112
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-112
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1624867610.1371/journal.pbio.0030370Research ArticleEcologyEvolutionMicrobiologyEubacteriaExploitative and Hierarchical Antagonism in a Cooperative Bacterium Antagonism in a Social BacteriumFiegna Francesca
1
Velicer Gregory J [email protected]
1
1Max-Planck Institute for Developmental Biology, Tüebingen, GermanyLevin Simon Academic EditorPrinceton UniversityUnited States of America11 2005 1 11 2005 1 11 2005 3 11 e37024 3 2005 31 8 2005 Copyright: © 2005 Fiegna and Velicer.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Antisocial Behavior in Cooperative Bacteria (or, Why Can't Bacteria Just Get Along?)
Social organisms that cooperate with some members of their own species, such as close relatives, may fail to cooperate with other genotypes of the same species. Such noncooperation may take the form of outright antagonism or social exploitation. Myxococcus xanthus is a highly social prokaryote that cooperatively develops into spore-bearing, multicellular fruiting bodies in response to starvation. Here we have characterized the nature of social interactions among nine developmentally proficient strains of M. xanthus isolated from spatially distant locations. Strains were competed against one another in all possible pairwise combinations during starvation-induced development. In most pairings, at least one competitor exhibited strong antagonism toward its partner and a majority of mixes showed bidirectional antagonism that decreased total spore production, even to the point of driving whole populations to extinction. Differential response to mixing was the primary determinant of competitive superiority rather than the sporulation efficiencies of unmixed populations. In some competitive pairings, the dominant partner sporulated more efficiently in mixed populations than in clonal isolation. This finding represents a novel form of exploitation in bacteria carried out by socially competent genotypes and is the first documentation of social exploitation among natural bacterial isolates. Patterns of antagonistic superiority among these strains form a highly linear dominance hierarchy. At least some competition pairs construct chimeric, rather than segregated, fruiting bodies. The cooperative prokaryote M. xanthus has diverged into a large number of distinct social types that cooperate with clone-mates but exhibit intense antagonism toward distinct social types of the same species. Most lengthy migration events in nature may thus result in strong antagonism between migratory and resident populations, and this antagonism may have large effects on local population sizes and dynamics. Intense mutual antagonism appears to be more prevalent in this prokaryotic social species than has been observed in the eukaryotic social slime mold Dictyostelium discoideum, which also exhibits multicellular development. The finding of several cases of facultative social exploitation among these natural isolates suggests that such exploitation may occur frequently in nature in many prokaryotes with cooperative traits.
Experimentally competing different natural strains of the social bacterium Myxococcus xanthus reveals that some strains exploit others, with implications for the evolution of intraspecific cooperation and the generation of bacterial diversity.
==== Body
Introduction
The microbial world is replete with cooperative behaviors that appear to produce density-dependent fitness benefits [1], including biofilm formation [2], quorum sensing [3], siderophore production [4,5], and fruiting body construction [6–8]. Also present, however, are strong negative social interactions that have evolved repeatedly among distinct lineages of relatively asocial species of bacteria such as Escherichia coli [9]. In highly social bacteria, such as those that form multicellular fruiting structures, the degree of social compatibility or antagonism among divergent strains classified within a single species remains unexplored. In the eukaryotic social slime mold Dictyostelium discoideum, which produces spores within multicellular fruiting bodies upon starvation, total social productivity (i.e., spore production) does not appear to suffer when distinct genotypes are mixed [10,11]. It is unclear whether intraspecific mixing of multicellular prokaryotes is similarly benign with respect to the benefits of social development or rather has more severe effects at the population level. It is also unknown whether natural, socially competent genotypes of a cooperative bacterial species are capable of exploiting other genotypes of the same species in a social context.
The soil-dwelling myxobacteria are unique among prokaryotes in the variety and sophistication of their social behaviors. They swarm cooperatively throughout a vast range of global soil ecosystems [12] while feeding on other microorganisms and detritus [13]. Upon starvation, local groups aggregate and develop into multicellular fruiting bodies [6]. In the model species Myxococcus xanthus, stress-resistant spores are formed by a minority of the fruiting body population, whereas the remainder appear to either undergo suicidal autolysis or remain undifferentiated [14].
The motility of M. xanthus and external migration vectors such as animals, insects, water, and wind should frequently cause distinct genotypes to encounter one another over both small and large spatial scales. Such encounters may result in a wide variety of interactions, including neutral social compatibility, antagonism, synergism, and social exploitation of one genotype by another. Smith and Dworkin [15] previously showed that two distinct Myxococcus species (M. xanthus andM. virescens) are incompatible during cooperative development. Clones of these two species separate into distinct, unmixed fruiting bodies in initially mixed cultures, and M. virescens strongly dominates over M. xanthus in spore production. The dominance of M. virescens is largely due to the production of compounds toxic to M. xanthus. The nature of social interactions among isolates classified within a single species of myxobacteria, however, remains unclear.
In this study, we quantify the effect of pairwise mixing among nine developmentally proficient global isolates of M. xanthus on the developmental performance of these strains and address the following questions. (i) Do most strains respond positively, negatively, or neutrally to mixing with other genotypes? (ii) Do competitors segregate into distinct fruiting bodies or mix within individual fruiting bodies? (iii) In any given pair, do both strains respond similarly to mixing (e.g., negatively), or are there cases where one competitor exploits another by simultaneously showing enhanced performance in mixture while inhibiting the performance of the other? (iv) Which better predicts the outcome of mixed developmental competitions: the relative performance of two competitors in clonal isolation or the relative effect of mixing on two competitors' developmental performance? (v) What effects do observed antagonisms have on the total social productivity (i.e., spore production) of competitive mixtures relative to clonal controls? (vi) Are most fitness relationships among three or more strains transitive (i.e., linear) or nontransitive (i.e., circular)?
Paired competitors were mixed at equal proportions in all possible combinations at the onset of starvation-induced development. Previous theory has predicted that antagonism among bacteriocin-producing competitors should be highest when the degree of relatedness within a mixed group is approximately 0.5 [16], suggesting that equal proportions of pairwise competitors may be the optimal condition for detecting any antagonism that exists between them. Mixing effects on fruiting body formation were observed, and the sporulation efficiency of each strain in mixture was contrasted to its spore production in isolation. The majority of strain interactions were strongly negative, and many pairings resulted in bidirectional inhibition of sporulation. In some pairings, one competitor performed better in mixed competition than in clonal isolation, revealing the capacity for social exploitation among natural prokaryotes and demonstrating that socially competent bacteria are capable of exploiting other genotypes. All except one set of three-way fitness relationships were linear rather than circular, supporting the view that the social incompatibilities among these genotypes are the result of local adaptation by isolated lineages.
Results
Mixing Effects on Population Morphology
Strong strain-by-strain interactions were clearly evident in the effects of mixing on the shape, size, and distribution of fruiting bodies in mixed cultures relative to their corresponding pure cultures (Figure 1). In almost all competition pairs, the developmental morphology of the mixed population was clearly distinct from those of both pure-culture counterparts. Mixed populations usually showed fewer fruiting bodies than each competitor in isolation, thus illustrating that cooperative development is normally less efficient when performed by chimeric, rather than clonal, populations. Because sporulation efficiency decreases at lower cell densities, we tested whether the effects of mutual antagonism are even more severe when mixed populations begin development at reduced density (109 cells/ml). Under these conditions, mixing of strains E and F completely eliminated developmental aggregation (Figure 2).
Figure 1 Effects of Pairwise Mixing on Fruiting Body Size and Distribution
Pairings AE, DH, AF, and BE are shown. Pure-culture fruiting bodies are shown in the top and middle photographs of each column (first and second listed strains, respectively), and fruiting bodies of the corresponding mixed culture are shown at the bottom.
Figure 2 Extinction Caused by Mutual Antagonism between Strains E and F
(A) At an initial developmental density of 109 cells/ml, no spores survived starvation in mixed populations, but both competitors performed well in clonal isolation. Error bars indicate 95% confidence intervals.
(B) Developmental phenotypes of E and F in pure culture (left and right, respectively) and in mixed competition (middle).
Sporulation Mixing Effects
The use of antibiotic resistance markers allowed us to measure the effects of mixing on both the total spore production of mixed populations and each competitor individually. Mixing decreased the overall social productivity (i.e., total spore production) of mixed populations (Table 1). Across the 36 competition pairs, the average total spore production by mixed populations was lower (57.2%) than that expected from spore production in the pure-culture controls (p = 0.0002, two-tailed one-sample t-test of nine strain averages, df = 8). Four pairs sporulated at less than 10% of expected levels (1.6%-6.8%), 14 pairs produced between 10% and 50% of the expected spore number, and 16 pairs produced between 53% and 87%. One pairing (EF) was repeated at a lower initial density of 109 cells/ml. No spores of either competitor survived development, whereas both competitors produced many spores in pure culture at the same density (Figure 2). At the standard higher density of 5 × 109 cells/ml, two pairings appeared to have enhanced spore production, but neither increase was statistically significant.
Table 1 Average One-Way (Ci [j]) and Bidirectional (Bij) Mixing Effects by Strain
Of 72 one-way mixing effects (Ci[j]), 56 were negative (strains sporulated less efficiently in mixture than in isolation), whereas 16 were positive (strains sporulated more efficiently in mixture than in isolation) (Figure 3). Thus, 20 pairs exhibited bidirectional antagonism, and 16 pairs had one positive response and the other negative (strain pairs AB, AF, AG, AI, BG, BI, CF, CG, CI, DB, DF, DG, DH, DI, EG, IG [positive response strain listed first in each pair]). The mean one-way mixing effect of each strain was negative for all nine strains (Table 1), and the grand mean of one-way mixing effects was strongly negative (−1.85, p = 0.0010, two-tailed one-sample t-test of nine strain averages, df = 8) and corresponded to an approximately 70-fold average decrease in sporulation efficiency in mixed competitions relative to pure-culture development.
Figure 3 Distribution of Mixing Effects on the Sporulation Efficiencies of Individual Competitors (Ci[j])
The 16 positive values corresponded to competition mixes in which one strain appeared to sporulate more efficiently in mixture with a competitor than in clonal isolation and simultaneously inhibited its competitor's performance. These outcomes provide evidence of facultative social exploitation in bacteria. By “facultative,” we mean that the ability to exploit other genotypes is not derived from a genetic social defect that prevents normal cooperation under clonal conditions. Thus, the exploiters do not obligately defect from cooperation in all contexts but do so only in mixes with some genetically distinct social partners.
The distributions of strain representations among the 16 exploitative responses and the 16 corresponding negative responses by the inferior (exploited) partners are nonrandom, with some strains being disproportionately represented (p < 0.0001 in both cases, chi-square test). Among the 16 positive responses, 12 are made by strains A, C, and D (four, three, and five cases, respectively). These are the three most dominant competitors among the nine strains examined here (see explanation of Table 2 below). Inversely, among the partners exploited by dominant strains, strain G accounts disproportionately for six of the 16 apparent cases of exploitation.
Table 2 Relative Sporulation Efficiency (Wij) Dominance Hierarchy
Exploitative responses to mixing were demonstrated more robustly in two particular strain pairs that involved relatively strong positive responses in the larger matrix of competition mixes. Two competition pairings (DI and IG) with one strain showing a positive response to mixing were repeated with the dominant strain (D and I, respectively) at a lower initial frequency of 0.32 (Figure 4). Both D (vs. I) and I (vs. G) showed significantly increased sporulation efficiency in mixed competition relative to their efficiencies in pure culture (1.6- and 2.1-fold increases in efficiency, respectively; p = 0.04 and 0.005, respectively, one-tailed t-tests, df = 2).
Figure 4 Facultative, Antagonistic Exploitation by Two Genotypes (D against I and I against G) during Mixed Development
The log-scale effect of mixing strains i and j on the sporulation efficiency of strain i is given as Ci(j). Open bars show the effect of mixing on sporulation efficiency for the dominant, exploitative competitor in each pair (D and I, respectively) in response to its inferior competitor. Shaded bars indicate the effect of mixing on the inferior strain (I and G, respectively). Error bars indicate 95% confidence intervals.
Among the 36 competition pairs initiated at a 50:50 ratio, very different mixing effect patterns were observed among all competitors against a given strain. Three examples are shown in Figure 5. Strain A showed dramatic inhibition of competitor sporulation in most pairings (greater than 1,000-fold mean inhibitory effect) but its own sporulation efficiency was relatively unaffected by these strains (Figure 5A). Strain B, in contrast, hindered other strains to approximately the same degree, on average, as its sporulation was hindered by the other strains (Figure 5B). Its interaction patterns, however, varied greatly among partners. Three competitors (A, C, and D) strongly inhibited the sporulation of B but were not harmed by the presence of B in return, while three other competitors (F, G, and H) showed the opposite pattern. The developmental performance of strain G was strongly inhibited by every competitor, with none but strain H being hindered by G in return (Figure 5C).
Figure 5 Bidirectional Mixing Effects for Strains A, B, and G against All Competitors
The log-scale effect of mixing strains i and j on the sporulation efficiency of strain i is given as Ci(j). Open bars show the effect of mixing on sporulation efficiency for strains A (A), B (B), and G (C) in response to each competitor shown along the horizontal axis. Shaded bars indicate the effect of mixing on the sporulation of the variable competitors. Error bars indicate 95% confidence intervals.
Developmental Performance Predictors
We asked whether the relative performance of two competitors in clonal isolation or rather the relative effect of mixing better predicts the outcome of mixed developmental competitions. To compare the relative performance of the two strains within each competing pair, we calculated W*ij (predicted relative sporulation), Wij (actual relative sporulation), and Cij (relative mixing effect) for each of the 36 unique competition pairings. From equations 1–6 (see Materials and Methods), Cij = Wij − W*ij. The relative effect of mixing on competing strains' sporulation efficiencies (Cij) is highly predictive of actual competitive performance (Wij) (r
2 = 0.967, p < 0.001, linear regression; Figure 6). In contrast, predicted relative performance based on pure-culture sporulation efficiencies (W*ij) showed no relationship with actual competitive performance (Wij) (r
2 = 0.010, p = 0.556, linear regression; Figure 6). These results show antagonistic social interaction between competing genotypes to be the primary determinant of actual relative sporulation efficiency.
Figure 6 Relationships between Relative Mixing Effect (Cij) or Predicted Relative Performance (W*ij) (y-Axis) and Actual Relative Performance (Wij) (x-Axis)
Dominance Hierarchies
We examined whether the structure of competitive dominance relationships across all pairwise comparisons is strongly hierarchical (i.e., linear). For measures of relative mixing effect (Cij) and relative sporulation efficiency in mixture (Wij), each strain was ranked by its number of positive Cij or Wij values among competitions with the other eight strains. The two resulting matrices are identical in structure and reveal a clear dominance hierarchy among the strains. Table 2 shows the Wij values of the competitors listed by column (i) relative to the competitors listed by row (j). Positive Wij values indicate that the column strain is superior to the row strain, whereas the only negative value indicates that the column strain (A) is inferior to the row strain (C). Values not shown in the opposite half of the matrix are simply the negative of the value shown in the reciprocal cell. For example, WDB = 4.43 (shown), whereas WBD = −4.43 (not shown). Both values mean that strain D produces approximately 27,000-fold more spores than strain B in mixed development starting at a 50:50 ratio (log[27,000] = 4.43).
Strains A, D, and C shared the highest rank dominance (superiority over seven competitors each) and are ordered by decreasing average Wij. The remaining strains, B, I, F, E, H, and G, exhibit five, four, three, two, one, and zero cases of superiority, respectively. The Wij values among the three pairings of A, D, and C reflect the sole nontransitive triad (A beats D, D beats C, and C beats A) of performance relationships in each respective hierarchy. The top four competitors were all isolated from North America (as opposed to Eurasia), but such a ranking distribution has a 9.2% likelihood of occurring by chance (binomial test).
Kendall's technique [17] was used to test for significant linearity in the dominance hierarchy. This test is most commonly used to test for patterns of behavioral dominance in animals but has also been used to examine fitness among multiple isolates of D. discoideum [11]. The null hypothesis assumes that the relationship between any two particular strains does not predict anything about the relationship that each strain has with other strains. Statistical significance is derived from a chi-square test with K representing the degree of linearity on a scale from zero (total absence of linearity) to one (perfect linearity). With our group size of nine individuals, dominance hierarchies with 16 or fewer nontransitive triads exhibit significant linearity. Our hierarchy had only one such triad, representing extremely strong and significant linearity (K = 0.965, p < 0.001), meaning that this set of strains shows many distinct fitness ranks (seven of nine possible) rather than few.
Competitor Distributions across Fruiting Bodies
Distinct antibiotic-resistance states of mixed competitors also allowed us to examine whether competitors ever mix within fruiting bodies or rather segregate during development after forced mixing in liquid. For two pairings (AC and BI) in which both competitors sporulated at relatively high levels in the mixed populations, we tested whether only one competitor could be found within individual fruiting bodies or rather both. For each pair, fruiting bodies were transferred to two nutrient agar plate types (ten to each type for AC and 40 to each type for BI), with each plate type containing distinct antibiotic corresponding to the competitors' resistance states. Fruiting bodies were then scored for subsequent growth after 5 d.
If competitors always segregate into separate fruiting bodies, only half of the total number of transferred fruiting bodies should germinate, regardless of the fruiting-body type ratio, because each fruiting body will germinate on only one of the two selective agar types. In the AC competition, all ten fruiting bodies germinated on whichever selective media they were placed, indicating that A and C form chimeric populations within fruiting bodies and likely do so in all fruiting bodies. The probability of all 20 fruiting bodies germinating under the total segregation hypothesis is extremely low (p < 0.0001, binomial test).
In the BI competition, however, the 40 fruiting bodies under selection for strain B (plates with rifampin) all germinated, whereas only three of the 40 fruiting bodies under selection for strain I (novobiocin plates) germinated. This result might reflect either total segregation of the competitors into distinct fruiting bodies (with I fruiting bodies in the minority) or segregation of B in most fruiting bodies with a small percentage of chimeras in which both B and I are present. Under the total segregation model, our novobiocin result gives the estimate that approximately 7.5% of fruiting bodies are of type I, which leads to the prediction that only 92.5% (37) of fruiting bodies will grow on rifampin agar. The total segregation hypothesis is unlikely to be correct, however, because the probability of all 40 fruiting bodies growing on rifampin when only 37 are expected is only 4.4% (binomial test). In the quantitative spore production assays, strain B sporulated approximately 28-fold more efficiently than strain I, averaged over all fruiting bodies and replicates. However, strain B was found in only 12.3-fold more fruiting bodies than strain I, suggesting that even if the total segregation hypothesis were correct, strain I should sporulate less efficiently within its respective fruiting bodies than strain B.
Discussion
Intraspecific antagonism among relatively asocial bacteria is well known [9,18–20], but the nature of intraspecific interactions among divergent genotypes of a highly social bacterium has not been previously characterized. Here we have shown that intense antagonism occurs between distinct genotypes of the cooperative prokaryote M. xanthus under environmental conditions in which social cooperation is crucial to fitness and survival. This antagonism negatively affected social productivity in almost all cases, with several pairings of antagonistic competitors reducing total spore production by greater than 90%. Most individual responses to mixing were strongly negative, but some strains exploited others by performing better in mixture than in isolation. Overall fitness ranks among the isolates were strongly hierarchical, with the three best strains forming a dominant triad (with nontransitive fitness ranks within the triad) and the remaining six exhibiting a perfect hierarchy.
These patterns reveal that M. xanthus as a species is not composed of a small number of cooperative units. Rather, this species has diverged into a large number of distinct social types that cooperate with clone-mates and perhaps very close relatives but exhibit intense antagonism toward distinct social types of the same species. This leads to the prediction that most long-distance transports of soil particles will result in strong antagonism between migratory and resident populations of M. xanthus. Our results show that such antagonism often causes large reductions in the total spore production of chimeric populations and can even cause whole populations to become extinct (see Figure 2). In migratory encounters involving intense mutual antagonism, resulting reductions in total population size may jeopardize the prosperity or even survival of mixed populations when density-dependent cooperation is important for fitness, such as during starvation. Interestingly, chimeric mixing of distinct D. discoideum genotypes does not appear to reduce total spore production during its eukaryotic version of cooperative development, although group mobility is impaired by chimeric mixing [11]. This difference between D. discoideum and M. xanthus in the effects of mixing on spore production may reflect fundamental differences in the mechanisms of intraspecific competition between eukaryotic and prokaryotic social microorganisms more generally.
A previous Myxococcus study found that two genotypes classified as distinct species (M. virescens and M. xanthus) segregated into separate fruiting bodies after forced mixing and that the M. virescens isolate inhibited sporulation of the M. xanthus isolate [15]. In contrast, we have shown that at least some isolate pairs classified as M. xanthus are capable of forming chimeric fruiting bodies, suggesting that such chimerism might also occur under natural conditions, even if rarely. Such within-species chimerism has also been observed in D. discoideum fruiting bodies [7]. It has been suggested that spontaneous mixing might be advantageous for all involved genotypes under conditions in which fruiting is important for survival but in which distinct co-existing genotypes must combine in order to generate a group size larger than the minimum threshold required to undergo development [21]. Under such a scenario, even competitors that perform relatively poorly in mixed groups will be better off mixing than segregating, as they would produce at least some spores rather than none. When distinct genotype populations are all sufficiently large to sporulate efficiently as clonal groups, however, only dominant competitors would benefit from chimerism. The degree to which distinct Myxococcus genotypes may actually mix in the wild is under further investigation.
Until recently, myxobacterial classification was based exclusively on characterization of morphological traits, including the size, shape, and color of fruiting bodies and growing swarms. A recent comparison of a 16S rRNA-based phylogeny of the myxobacteria with traditional classification showed that the two approaches yielded largely consistent results across the monophyletic group [22]. At the genus level, however, the 16S rRNA approach did not always resolve strains bearing the same traditional species classification into a single group distinct from other strains with a different species classification. More comprehensive molecular-based approaches to classification are now feasible, including rapid whole-genome sequencing [23], and will allow a much finer resolution of species and strain distinctions within Myxococcus and other genera. Questions of future interest emerge from comparison of the M. virescens- vs.-M. xanthus study with our results, including whether genotypes of distinct species classifications ever form fruiting chimeras and whether qualitative antagonism patterns during interspecific competitions are consistent or highly variable across multiple isolates from each species.
Relatively asocial bacterial species such as E. coli have evolved intense forms of intraspecific antagonism [24]. Our results demonstrate that such antagonisms evolve readily in highly cooperative bacteria as well. In E. coli, the best known form of intraspecific antagonism is mediated by the production of colicin proteins from plasmid-borne genes that can be readily exchanged among genotypes via conjugative plasmid transfer [25]. In contrast, there is no evidence that M. xanthus hosts self-replicating plasmids [26], and M. xanthus has no known mechanism of conjugative DNA transfer between cells. Thus, both the evolutionary and biochemical mechanisms of antagonism are likely to be fundamentally different between these species.
At the biochemical level, M. xanthus carries genes for the production of a variety of secondary metabolites not present in most bacteria that may serve as anticompetitor agents [27]. Initial results indicate that the strains used in this study vary significantly in the types of secondary metabolites they produce (R. Müller and G. Velicer, unpublished data). This variation may be at least partially responsible for the asymmetric antagonisms observed here. Alternatively, inhibition of another competitor's sporulation may result from interference with the competitor's developmental signaling ability or motility rather than production of compounds that are directly toxic to competitor cells.
Regardless of its underlying biochemical mechanisms, the hierarchical antagonism observed during M. xanthus development among spatially isolated genotypes suggests that local diversity generated by migration of previously isolated genotypes will, on average, be short lived. Previous studies have shown that nontransitive fitness relationships among three or more competitors in structured habitats can serve to maintain biological diversity [18,20,28]. In our set of nine strains, however, only one triad of pairwise fitness relationships (among strains A, C, and D) of a total of 84 total triads showed nontransitivity. Competition experiments similar to those described here are under way with spatially proximate isolates of M. xanthus, and it will be of interest to determine whether nontransitivity is detected more frequently among close neighbors in the soil than was observed here among isolates from distant origins.
Although the average effect of competitive encounters here is strongly negative, 44% of competitive pairs had one partner that showed enhanced sporulation efficiency in response to mixing (see Figures 3 and 4). A variety of mechanisms might mediate such exploitation, including the ability to induce disproportionate cell death in a competitor and consume its remains as a growth substrate prior to spore differentiation. The exploitation observed here is facultative [29], whereas previous instances of bacterial cheating [4,8,30] have involved obligate defection from cooperation. Such obligate cheaters inherently (i.e., genetically) defect from cooperation regardless of the genotypic composition of their social neighbors (e.g., pure defectors or mixed defectors and cooperators). These obligate defectors perform poorly as clonal groups under selective conditions favoring sociality but perform well as a minority among cooperators. The isolates examined here, however, are highly proficient at sporulation in clonal isolation, but some are also able to exploit other genotypes in mixture.
Previous reports of bacterial social exploitation have involved genotypes modified or evolved in laboratory environments, whereas this study demonstrates intraspecific social exploitation among distinct natural bacterial isolates. The superiority of exploitative clones over their competitors in mixture observed here is not due to major differences in pure-culture performance. Rather, it is accomplished by the formidable competitive ability of some dominant genotypes to both inhibit (and in some cases, abolish) a competitor's sporulation ability and simultaneously convert a higher percentage of cells into spores specifically due to the presence of the inferior competitor.
Materials and Methods
Strains
Nine clonal isolates previously classified as M. xanthus based on morphological criteria (e.g., fruiting body color, shape, and size) were used in this work. DK801 (here, strain D, Tracy, California, United States), DK816 (B, Ontario, Canada), DK836 (A, Albany, New York, United States), and DK897 (E, Maryhill, Washington, United States) were obtained as clonal stocks from D. Kaiser; clones of Mxx15 (F, Olympia, Greece), Mxx23 (C, Minneapolis, Minnesota, United States), Mxx41 (G, Madras, India), Mxx104 (H, Bodrum, Turkey), and Mxx144 (I, Mt. Ar-Li, Taiwan) were purified from culture isolates provided by H. Reichenbach.
Three criteria were used in strain selection: (i) robust fruiting body formation and high sporulation efficiency, (ii) ability to obtain strain variants resistant to novobiocin, rifampin, or kanamycin, and (iii) zero or small effect of antibiotic resistance on sporulation efficiency. Strains were selected for use prior to collection of any data about their competitive ability against other strains. Strains A, B, D, E, and F were marked with spontaneous mutations conferring resistance to rifampin, while strains C, D, G, and I were likewise marked with spontaneous mutations to novobiocin resistance. Kanamycin resistance was conferred to strains C and E by chromosomal integration (via electroporation) of the plasmid pREG1727 [31] into the Mx8 phage attachment site and to strains D, F, and H by random chromosomal integration of the mariner transposon carried by the suicide plasmid pMycoMar [32]. Positions 46–445 of the 16S rRNA gene were sequenced in all nine strains and are identical to the standard M. xanthus lab strain DK1622.
In all reported results, estimates of marked strain performance in mixtures were compared to performance of the same marked strains in pure culture, not to pure cultures of unmarked parental strains. Nonetheless, it is worth noting that any marker effects on sporulation efficiency were small relative to total spore production (unpublished data). Factoring in marker effects to predict the outcome of hypothetical competitions between unmarked strains did not alter the structure of the Wij competition matrix (Table 2).
Developmental competition assays
All agar plates were incubated at 32 °C and 90% rH, and all liquid cultures were grown at 32 °C, 300 rpm. Strains were grown on CTT agar medium [33] from ultralow freezer stocks, inoculated into CTT liquid, and kept in exponential-phase growth for 2 d. Cultures were then centrifuged at 20 °C for 15 min at 4,500 × g, and the pellets resuspended with TPM liquid [34] to a density of approximately 5 × 109 cells/ml unless specified otherwise. For pure-culture assays, 100 μl of pure-culture resuspension was dispensed in the center of TPM plates (1.5% agar) for starvation. For paired competition assays, pure resuspended cultures from paired competitors were mixed at the appropriate ratio (50:50 in the complete matrix of pairwise mixes or 32:68 in a later experiment), and the total mixed volume (100 μl) was dispensed on the center of TPM plates. Thus, the total initial cell number and density were the same in competitive mixtures and clonal treatments. Developmental plates were incubated for 5 d, after which the cell populations were harvested with a scalpel blade, transferred into 1 ml of double-distilled H2O, and heated at 50 °C for 2 h to select for viable spores. Samples were then sonicated by microtip and diluted into CTT soft agar (0.5% agar) with the appropriate antibiotic treatment (40 μg/ml kanamycin, 5 μg/ml rifampin, 10 μg/ml novobiocin, or no antibiotic). In cases of mixed competitions where no colonies grew at our lowest dilution factor (101), a value of ten spores/assay was entered for data analysis, providing conservatively low estimates of fitness inferiority in these cases.
In each competition mixture, the paired strains had distinct antibiotic-resistance marker states, allowing quantification of each competitor's spore production. In most cases, both competitors bore resistance to distinct antibiotics, allowing for direct mixing of one resistant representative from each strain. For two pairings (AB and GI), however, only variants with the same resistance marker were obtained. In these two cases, the resistant variants of both strains were mixed with the original antibiotic-sensitive clone of the respective competitor and their sporulation efficiencies in mixture were measured independently.
In the complete competition matrix, each of all possible pairwise competitions were performed in at least three temporally independent blocks, with two replicates of each particular competition performed within each block. Assays of pure-culture sporulation efficiency for all competitors were performed simultaneously with the respective competition assays, also with 2-fold replication within each block. Because variation across blocks was larger than within blocks, we conservatively used within-block means of relevant values as data points for statistical analysis.
The untransformed sporulation efficiency (D) of strain i in pure culture is the frequency of cells surviving development as viable spores:
where Ni represents the viable population size of i before starvation (t
0) and after 5 d of starvation (t
5). The efficiency of i in mixed competition with strain j is similarly given as
The effect of mixing strains i and j on the sporulation efficiency of strain i is given as
Thus, a positive value of Ci(j) indicates that strain i sporulates more efficiently in the presence of strain j than in clonal isolation, whereas a negative value indicates that mixing with j negatively affects the efficiency of i. The relative mixing effect of strains i and j is given as
The predicted relative sporulation efficiency (log-transformed) of strains i andj during development assumes that there are no effects of mixing on sporulation efficiency and is defined as
The actual relative sporulation efficiency (log-transformed) of two strains is defined as the log difference of the strains' actual sporulation efficiencies in mixed competition with each other:
The bidirectional effect of mixing on total spore production (Bij) was calculated for all 36 matrix pairs as the frequency of spores produced in each mix ij relative to that expected from pure-culture assays of i and j, with
To estimate the significance of the negative overall mean of Bij, the average value of Bij for all eight pairs containing strain i was calculated for each of the nine strains. These nine means were used as statistical units in a one-sample t-test of the difference between the grand mean and the null expectation of zero, giving 8 df. Similarly, the mean Ci(j) value of each strain (across each strain's eight pairings) was used as the statistical unit to estimate the significance of the overall one-way effect of mixing.
Supporting Information
Accession Number
The NCBI Entrez (http://www.ncbi.nlm.nih.gov/gquery/gquery.fcgi) accession number for DK1622 is M34114.
We thank D. Gilliland, M. Grbić, K. Hillesland, C. Page, M. Travisano, M. Vos, S. West, and two anonymous reviewers for helpful discussions or comments; D. Kaiser and H. Reichenbach for strains; and S. Deiss for technical assistance. This work was partially funded by a grant from the Deutsche Forschungsgemeinschaft (VE 362).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. GJV conceived the experiments. FF and GJV designed and performed the experiments, and analyzed the data. GJV contributed reagents/materials/analysis tools. FF and GJV wrote the paper.
Citation: Fiegna F, Velicer GJ (2005) Exploitative and hierarchical antagonism in a cooperative bacterium. PLoS Biol 3(11): e370.
==== Refs
References
Crespi BJ The evolution of social behavior in microorganisms Trends Ecol Evol 2001 16 178 183 11245940
Webb JS Givskov M Kjelleberg S Bacterial biofilms: Prokaryotic adventures in multicellularity Curr Opin Microbiol 2003 6 578 585 14662353
Bassler BL Small talk. Cell-to-cell communication in bacteria Cell 2002 109 421 424 12086599
Griffin AS West SA Buckling A Cooperation and competition in pathogenic bacteria Nature 2004 430 1024 1027 15329720
West SA Buckling A Cooperation, virulence and siderophore production in bacterial parasites Proc R Soc Lond B Biol Sci 2003 270 37 44
Shimkets LJ Intercellular signaling during fruiting-body development of Myxococcus xanthus
Annu Rev Microbiol 1999 53 525 549 10547700
Strassmann JE Zhu Y Queller DC Altruism and social cheating in the social amoeba Dictyostelium discoideum
Nature 2000 408 965 967 11140681
Velicer GJ Kroos L Lenski RE Developmental cheating in the social bacterium Myxococcus xanthus
Nature 2000 404 598 601 10766241
Riley MA Goldstone CM Wertz JE Gordon D A phylogenetic approach to assessing the targets of microbial warfare J Evol Biol 2003 16 690 697 14632232
Foster KR Fortunato A Strassmann JE Queller DC The costs and benefits of being a chimera Proc R Soc Lond B Biol Sci 2002 269 2357 2362
Fortunato A Queller DC Strassmann JE A linear dominance hierarchy among clones in chimeras of the social amoeba Dictyostelium discoideum
J Evol Biol 2003 16 438 445 14635843
Dawid W Biology and global distribution of myxobacteria in soils FEMS Microbiol Rev 2000 24 403 427 10978544
Rosenberg E Varon M Rosenberg E Antibiotics and lytic enzymes Myxobacteria: Development and cell interactions 1984 New York Springer-Verlag 109 125
Wireman JW Dworkin M Developmentally induced autolysis during fruiting body formation by Myxococcus xanthus
J Bacteriol 1977 129 798 802 402359
Smith DR Dworkin M Territorial interactions between two Myxococcus species J Bacteriol 1994 176 1201 1205 8106334
Gardner A West SA Buckling A Bacteriocins, spite and virulence Proc Biol Sci 2004 271 1529 1535 15306326
Appleby M The probability of linearity in hierarchies Anim Behav 1983 31 600 608
Czaran TL Hoekstra RF Pagie L Chemical warfare between microbes promotes biodiversity Proc Natl Acad Sci U S A 2002 99 786 790 11792831
Lenski RE Riley MA Chemical warfare from an ecological perspective Proc Natl Acad Sci U S A 2002 99 556 558 11805313
Kirkup BC Riley MA Antibiotic-mediated antagonism leads to a bacterial game of rock-paper-scissors in vivo Nature 2004 428 412 414 15042087
Kaplan HB Plamann L A Myxococcus xanthus cell density-sensing system required for multicellular development FEMS Microbiol Lett 1996 139 89 95 8674994
Sproer C Reichenbach H Stackebrandt E The correlation between morphological and phylogenetic classification of myxobacteria Int J Syst Bacteriol 1999 49 1255 1262 10425789
Margulies M Egholm M Altman W Attiya S Bader J Genome sequencing in microfabricated high-density picolitre reactors Nature 2005 E-pub ahead of print
Riley MA Gordon DM The ecological role of bacteriocins in bacterial competition Trends Microbiol 1999 7 129 133 10203843
Riley MA Wertz JE Bacteriocin diversity: Ecological and evolutionary perspectives Biochimie 2002 84 357 364 12423779
Hartzell P Streips UN Zasbin RE Genetic tools for dissecting motility and development of Myxococcus xanthus
Modern microbial genetics 2002 New York Wiley-Liss 289 322
Gerth K Pradella S Perlova O Beyer S Muller R Myxobacteria: Proficient producers of novel natural products with various biological activities—Past and future biotechnological aspects with the focus on the genus Sorangium
J Biotechnol 2003 106 233 253 14651865
Kerr B Riley MA Feldman MW Bohannan BJ Local dispersal promotes biodiversity in a real-life game of rock-paper-scissors Nature 2002 418 171 174 12110887
Velicer GJ Social strife in the microbial world Trends Microbiol 2003 11 330 337 12875817
Vulic M Kolter R Evolutionary cheating in Escherichia coli stationary phase cultures Genetics 2001 158 519 526 11404318
Fisseha M Gloudemans M Gill RE Kroos L Characterization of the regulatory region of a cell interaction-dependent gene in Myxococcus xanthus
J Bacteriol 1996 178 2539 2550 8626320
Rubin EJ Akerley BJ Novik VN Lampe DJ Husson RN In vivo transposition of mariner-based elements in enteric bacteria and mycobacteria Proc Natl Acad Sci U S A 1999 96 1645 1650 9990078
Hodgkin J Kaiser D Cell-to-cell stimulation of movement in nonmotile mutants of Myxococcus
Proc Natl Acad Sci U S A 1977 74 2938 2942 16592422
Kroos L Kuspa A Kaiser D A global analysis of developmentally regulated genes in Myxococcus xanthus
Dev Biol 1986 117 252 266 3017794
|
16248676
|
PMC1275521
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e370
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030370
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1624867710.1371/journal.pbio.0030378Research ArticleBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyInfectious DiseasesHIV/AIDSHomo (Human)The Case for Selection at CCR5-Δ32
The Case for Selection at CCR5-Δ32Sabeti Pardis C [email protected]
1
2
Walsh Emily
1
Schaffner Steve F
1
Varilly Patrick
1
Fry Ben
1
Hutcheson Holli B
3
Cullen Mike
3
Mikkelsen Tarjei S
1
Roy Jessica
1
Patterson Nick
1
Cooper Richard
4
Reich David
1
5
Altshuler David
1
5
6
O'Brien Stephen
3
Lander Eric S
1
7
8
9
1 Broad Institute of MIT and Harvard, Cambridge, Massachusetts, United States of America,2 Harvard Medical School, Boston, Massachusetts, United States of America,3 Laboratory of Genomic Diversity, National Cancer Institute, Frederick, Maryland, United States of America,4 Department of Preventive Medicine and Epidemiology, Loyola University Medical School, Maywood, Illinois, United States of America,5 Departments of Genetics and Medicine, Harvard Medical School, Boston, Massachusetts, United States of America,6 Department of Molecular Biology and Center for Human Genetic Research, Diabetes Unit, Department of Medicine, Massachusetts General Hospital, Boston, Massachusetts, United States of America,7 Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America,8 Whitehead Institute for Biomedical Research, Cambridge, Massachusetts, United States of America,9 Department of Systems Biology, Harvard Medical School, Boston, Massachusetts, United States of AmericaClark Andy Academic EditorCornell UniversityUnited States of America11 2005 1 11 2005 1 11 2005 3 11 e37813 4 2005 8 9 2005 Copyright: © 2005 Sabeti et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
A Well-Studied Disease-Resistance Gene Shows No Signs of Selection
The C-C chemokine receptor 5, 32 base-pair deletion (CCR5-Δ32) allele confers strong resistance to infection by the AIDS virus HIV. Previous studies have suggested that CCR5-Δ32 arose within the past 1,000 y and rose to its present high frequency (5%–14%) in Europe as a result of strong positive selection, perhaps by such selective agents as the bubonic plague or smallpox during the Middle Ages. This hypothesis was based on several lines of evidence, including the absence of the allele outside of Europe and long-range linkage disequilibrium at the locus. We reevaluated this evidence with the benefit of much denser genetic maps and extensive control data. We find that the pattern of genetic variation at CCR5-Δ32 does not stand out as exceptional relative to other loci across the genome. Moreover using newer genetic maps, we estimated that the CCR5-Δ32 allele is likely to have arisen more than 5,000 y ago. While such results can not rule out the possibility that some selection may have occurred at C-C chemokine receptor 5 (CCR5), they imply that the pattern of genetic variation seen atCCR5-Δ32 is consistent with neutral evolution. More broadly, the results have general implications for the design of future studies to detect the signs of positive selection in the human genome.
Sabeti and colleagues use dense genetic maps to show that the HIV-resistance CCR5-Δ32 allele is more than 5,000 years old and is likely to have been under mainly neutral selection.
==== Body
Introduction
The impact of evolutionary selection on the human population is of central interest and, with increasing information about genetic variation, has become a subject of intense examination [1–6]. Knowledge of selective events and selected loci provide insight into the genetic etiology of human disease, past and present, and into the events that have shaped our species. As infectious diseases pose a major selective force, selected variants may give insight into immunological defense mechanisms—highlighting important pathways in pathogen resistance.
Evolutionary pressure generates a number of potentially detectable signals at a locus under selection as compared to the neutrally evolving genome. Because different populations are subject to distinct selective environments, selection may produce population-specific alleles and greater population differentiation at an affected gene, which can be measured with theFST statistic [7]. Positive selection may also cause a rapid rise in an allele's frequency, creating a disparity in the age of an allele estimated from its high frequency in the population (characteristic of an old allele) and its long-range linkage disequilibrium (LD, characteristic of a young allele). LD-based methods such as the Long-Range Haplotype test have been developed to detect this signal [3,8–10].
C-C chemokine receptor 5 (CCR5) is one of the most prominent reported cases of recent natural selection in the human genome. First identified as encoding a principal entry receptor for HIV-1 infection of CD4-bearing T lymphocytes, CCR5 has been the subject of intense focus by geneticists [8,11–14]. A well-established association exists between a 32 base-pair deletion variant in CCR5 (CCR5-Δ32) and protection from HIV infection, demonstrating that CCR5 plays an important biological role in HIV entry into cells.
The first suggestion that CCR5 may have been subject to positive selection was a high proportion of nonsynonymous mutations at CCR5, suggesting selective pressure for amino acid divergence [12]. More compelling evidence for selection on CCR5-Δ32 came from work by Stephens et al. [8]. This study found that Δ32 occurs at high frequency in European Caucasians (5%–14%, with north-south and east-west clines) but is absent among African, Native American, and East Asian populations, suggesting that the Δ32 mutation occurred after the separation of the ancestral founders of these populations. Moreover, Stephens et al. [8] reported strong LD between CCR5-Δ32 and two microsatellite markers, suggesting an estimated age for the allele of only ∼700 y (range 275–1,875 y). The apparent rapid rise in frequency implied strong positive selection, and the specific age raised intriguing possibilities for the selective agent, such as the bubonic plague in Medieval Europe.
With the recent availability of comprehensive information about patterns of allelic diversity in the human genome, we can now reexamine the case for selection at CCR5 by comparison with extensive empirical data and more sophisticated predicted distributions. We carried out high-density single-nucleotide polymorphism (SNP) genotyping around CCR5 in multiple populations, and analyzed the data with the benefit of large genomic comparison datasets and revised physical and genetic maps. Our results show that CCR5-Δ32 does not clearly stand out in terms of genetic diversity or long-range haplotypes relative to other variants at the locus or throughout the human genome.
Results/Discussion
We genotyped CCR5-Δ32, two microsatellites, and 70 SNPs (dbSNP data release 120, www.ncbi.nlm.nih.gov/SNP) extending 837 kbp centromere-distal and 430 kbp centromere-proximal to the CCR5 locus (Table S1). We studied 340 chromosomes from three populations: European-Americans, Chinese, and Yoruba from Nigeria. Eight of the European-American chromosomes bore the Δ32 mutation. In addition, we genotyped a subset of the SNPs in 12 Δ32/Δ32 individuals from the original study. This provided a total of 32 chromosomes bearing the Δ32 allele. We carried out all analyses on both datasets (Table S2).
We first examined the allele frequencies at SNPs around CCR5 in the European-American, Yoruba, and Chinese population samples for evidence of selection. As a genome-wide empirical comparison, we used two datasets. The first is 2,359 SNPs genotyped in the same 340 samples in the three populations. These SNPs are distributed in 168 immunologic genes from 64 loci across the genome; they were chosen according to the same methodology and have a similar physical distribution as for CCR5 [15] (see Materials and Methods). The second is data for 63,149 SNPs on Chromosome 3 from the International Haplotype Map Project (HapMap, data release 16) genotyped in the same three populations.
CCR5 is not a significant outlier relative to the 168 genes or HapMap Chromosome 3 with respect to heterozygosity and FST (Table 1; Figure S1). The heterozygosity statistic assesses the genetic diversity in a population; a selective sweep can reduce genetic diversity and balancing selection can increase genetic diversity. The FST statistic [7] compares the frequency of an allele between populations; a population-specific selective pressure may produce greater population differentiation at an affected gene. We also looked at the derived allele frequency (DAF) distribution, which can detect the genetic hitchhiking of variation linked to an allele under positive selection, and found no evidence for selection [16] (Table 1; Figure S2). All of these tests have limited power, with genotyping data ascertained to favor common shared SNPs and using the chimpanzee sequence for comparison. Therefore, while the results provide no evidence for selection, it can not be ruled out; this could be further explored with sequencing of a large number of chromosomes.
Table 1 Genetic Diversity at CCR5 in Comparison with Genetic Diversity for Regions from Two Large Empirical Datasets
We also assessed the significance of the observation that Δ32 is at moderately high frequency (8%) in the European-Americans but absent in the Chinese and Yoruba populations sampled. The observation is not exceptional in our available polymorphic data: of SNPs present at similar frequency (7%–9%) in European-Americans, ~7% are not found in the Chinese and Yoruba populations for the 168 genes, and 6% are not found for the same populations for the HapMap data. These estimates are likely to be conservative considering that the ascertainment of these studies favors shared polymorphisms. As more data become available, this analysis should be extended by larger sample sizes, more populations, and more closely matched data (including insertion/deletion polymorphisms and functional polymorphisms).
We next tested for signatures of selection by examining the extent of LD around CCR5-Δ32. For this purpose, we used the Long-Range Haplotype test for selection [3] (see Materials and Methods). Specifically, we calculated the relative extended haplotype homozygosity (REHH), which is sensitive to recent directional positive selection, and extended haplotype homozygosity (EHH), which is more sensitive to multiple selective sweeps at a locus. To estimate the recombination rate, we used two measures: the genetic distance from a family-based linkage study [17] and the number of observed historical recombination events [3] (Material and Methods).
We initially examined the centromere-distal side of CCR5 using the approach of Stephens et al. [8] (Figure 1A). Specifically, we sorted the chromosomes into two groups: Δ32-bearing and non-Δ32-bearing chromosomes. Consistent with the previous study [8], we found that the Δ32-bearing chromosomes have much longer LD than non-Δ32-bearing chromosomes: the EHH is 5.96 times greater than the average EHH of other variants at this locus (REHH = 5.96 at a distance of 500 kbp or 0.25 centimorgans [cM]) (Figure 1B).
Figure 1 EHH Breakdown of EHH over Distance between the CCR5-Δ32 Mutation and 63 SNPs at Increasing Distances from the Mutation
(A) Map of SNPs typed.
(B) Comparison between Δ32 and a single non-Δ32 class of haplotypes. It should be noted that the Δ32-bearing chromosomes appear (red) to have greatly extended LD compared to the non-Δ32-bearing chromosomes (black).
(C) Breakdown using the eight-marker haplotype containing the Δ32 mutation. There are five haplotypes in European-Americans (frequencies: 42%, 31%, 10%, 8%, and 8%, respectively). Full haplotype sequences and frequencies in other populations are given in Table S3. Notice that two of the non-Δ32-bearing chromosomes (black) appear to have the similar extended LD when compared to the Δ32-bearing chromosomes centromere-distal to CCR5 (red). Centromere-proximal Δ32-bearing chromosomes still have the most extended LD, indicated with an arrow.
We reasoned, however, that the apparent long-range LD might be a result of sorting the chromosome into only two classes based on their genotype at CCR5-Δ32, rather than dividing them according to the full variation seen at CCR5.
Figure 2 shows how an apparent signal of long-range LD can readily arise in this fashion. Briefly, one class (for example, the non-Δ32) may contain multiple distinct haplotypes whose individual signals of long-range LD may be obscured when grouped together, with the result that the other class (for example, the Δ32) appears to have much longer relative LD.
Figure 2 Model of Haplotype-Based Selection Approach
The image compares this approach, where the variants at the gene being studied are fully elaborated, to a model where the variants are not fully elaborated. At the top, multiple SNPs are genotyped to fully define the variants that exist in the gene. The resultant observed haplotype structure is shown in both bifurcation diagram and EHH plot formats (see Materials and Methods). At the bottom, only one SNP is genotyped, collapsing all other variants into a seemingly diverse super-haplotype and creating an impression of extension for the remaining haplotype.
In fact, this is precisely the case for CCR5. We fully delineated the variation at CCR5 by genotyping seven additional SNPs within the gene and defined haplotypes as previously described [18] (Figure S3). There are five distinct haplotypes, including the Δ32-bearing haplotype with frequency 8% (Table S3). The relative LD of the Δ32-bearing haplotype is significantly lower than for two other haplotypes (REHH = 1.92 versus 6.77 and 3.29 at distance 500 kbp or 0.25 cM; see Figure 1C), indicating that there is no significant evidence of long-range LD on the centromere-distal side of CCR5.
We next analyzed LD on the centromere-proximal side of CCR5. We first employed the approach used in the original study and again found the Δ32-bearing chromosomes had much longer LD than non-Δ32-bearing chromosomes (REHH = 20.22 at a distance of 250 kbp or 0.25 cM; see Figure 1B). We then reanalyzed the data by disaggregating the chromosomes into the five haplotypes described above. The relative long-range LD for Δ32-bearing chromosomes is much lower (REHH = 7.26), although it is still the highest among the five haplotypes.
We sought to assess whether the extent of LD in the centromere-proximal direction on Δ32-bearing chromosomes is unusual relative to that seen across the human genome. We first compared the results to the genome-wide distribution of REHH scores for the HapMap (Release 16, www.hapmap.org, and found that Δ32-bearing chromosomes do not clearly stand out from other haplotypes of similar frequency (6%–10%) (Figures 3A and S4). Because the 120 European-American chromosomes genotyped in the HapMap project have limited power for studying low-frequency haplotypes (P. V., B. F., E. S. L., and P. C. S., unpublished data), we augmented the analysis by comparing all 32 Δ32-bearing chromosomes to simulations with larger sample size [19] (see Materials and Methods). We simulated 1,000 1-mbp regions in 400 European-American chromosomes under a neutral model, generating 5,915 haplotypes matched with a frequency similar to the Δ32-bearing haplotype (6%–10%). The level of EHH for the Δ32-bearing haplotype was not unusually high on either the centromere-distal (p = 0.49) or centromere-proximal (p = 0.15) side of CCR5 when compared to the level seen at an equivalent recombination distance for the simulated regions. The REHH (we used the EHH of the two common haplotypes for a relative value) was also not unusually high (Figure 3B).
Figure 3 Comparisons with Empirical and Simulation Data
(A) and (B) Plots of relative EHH versus frequency for CCR5 in comparison to HapMap data (release 16) for Chromosome 3 in European-Americans (A) and 1,000 simulations of 400 chromosomes in European-Americans (B). Green dots represent the comparison haplotypes and the lines represent, from bottom to top, the 50th, 75th, 95th, and 99th percentiles. The red dots represent the results for eight CCR5-Δ32-bearing chromosomes in (A) and 32 CCR5-Δ32-bearing chromosomes in (B) for the centromere-proximal side, and the blue dots represent results for the centromere-distal side.
(C) EHL of the haplotypes of frequency 6%–10% from the HapMap (solid green line) and from simulations (dotted green line) in comparison to CCR5-Δ32 (red line).
We further examined the extent of the Δ32-bearing haplotype in comparison to other haplotypes of similar frequency. For this purpose, we defined the extended haplotype length (EHL) on each side of a haplotype to be the genetic distance at which the EHH score falls to 0.5. The EHL for the Δ32-bearing haplotype is 0.212 cM on the centromere-distal side and 0.258 cM on the centromere-proximal side, corresponding to a total of 0.470 cM (Figures 3 and S5). We then determined the EHL for haplotypes of comparable frequency (6%–10%) for both the HapMap data (average EHL is 0.354; CCR5-Δ32 is the 88th percentile) and for the simulated data (average EHL is 0.453; CCR5-Δ32 is the 64th percentile). The distribution is presented in Figure 3. Long-range LD around rare alleles is a prevalent feature in the genome, and the EHL for CCR5-Δ32 therefore does not stand out in comparison to either the HapMap or simulation dataset (Table S4). The EHL for CCR5-Δ32 would only be significant if the recombination rate in this region were several-fold higher than that measured by the current recombinational maps or by counting of historical recombination events (Protocol S1).
Given that long-range LD is a common feature of rare alleles in European-Americans, we wanted to test if our method would have the power to detect selection of an 8% allele over the time scale previously proposed [8]. We simulated 500 regions of 1 mbp length in 400 and 120 European-American chromosomes that had undergone a partial selective sweep beginning either 700 or 2,000 y ago for both groups of chromosomes, carrying the selected allele to a frequency of 8%. We were able to detect recent selection in the 400 chromosomes; 69% of selected alleles originating 700 y ago and 39% of selected alleles originating 2,000 y ago have EHL values above the 95th percentile when compared to the neutral distribution. There is far less power in the 120 chromosomes (30% and 10% of selected alleles originating 700 or 2,000 y ago, respectively), suggesting that the HapMap dataset will be insufficient to scan for rare selected alleles in European-Americans.
Finally, we revisited the estimated date of origin for the CCR5-Δ32 mutation. The original estimate [8] was based on the analysis of two microsatellites that were in strong LD despite apparently being at a considerable genetic distance away (0.91-cM interval and both centromere-distal, according to the genetic maps that were current at the time). With improvements in the genetic map over the past 7 y [17], the microsatellites were shown to be on opposite sides of CCR5 and at a much shorter genetic distance (0.18 cM, Figure S6). Using the methodology and data employed by Stephens et al. [20] (Table S5), but with the revised genetic map, the estimated age rises from 688 y (275–1,875 y, 95% confidence interval) to 7,000 y (2,900–15,750 y, 95% confidence interval ). When we expanded the analysis to include 32 genetic markers that have been genotyped in the Δ32-bearing chromosomes, the estimated age also rises, to a similar value of 5,075 y (3,150–7,800 y, 95% confidence interval). The SNP-based estimate of the age differs and has tighter error bars because the denser map holds more information about historical recombination events than the two microsatellites, whose genetic diversity is roughly equivalent to two SNPs (Figure S7). The older age estimate is consistent with unpublished work on DNA extracted from 3,000-y-old burial sites in central Germany showing that the CCR5-Δ32 was present at an appreciable frequency several millennia ago, at least in central Germany [21].
The revised age estimate suggests the high frequency of the CCR5-Δ32 allele cannot be attributed solely to a strong selective event within the past millennium. If selection did play a role in the high frequency of the allele, the initial selection pressure must have occurred before the period calculated in the previous estimate [8]. It should be noted that the data do not rule out some additional selection occurring within the past millennia, but none that would be detected by the methodology used in Stephens et al. or in the current paper.
Our reanalysis of CCR5 shows that CCR5-Δ32 does not clearly stand out from the rest of the genome in terms of allele frequency distribution, population differentiation, or long-range LD (Figure S8). The high population differentiation and long-range LD found for CCR5-Δ32 are, in fact, far more common in the genome than previously believed, and therefore do not provide support for the hypothesis of strong selection for CCR5-Δ32. Using methods described both in the previous study [8] and in the current study, and by examining currently available data, there is no detectable evidence for recent selection for CCR5-Δ32. Of course, the lack of support does not exclude the possibility of selection for the allele or the locus. Given the biology of the gene, it is certainly possible that it has been subject to some selection despite the lack of clear evidence. We note that small-scale studies of the distribution of mutations [12–14,22] have provided suggestive evidence for selection, but these results may be less convincing in comparison to recently available genome-wide distributions [23].
Beyond the specific results for CCR5, our results have important implications for studies of selection in the human genome. First, accurate assessment of LD benefits from fully delineating the core haplotypes at a locus; it may not be sufficient to compare a haplotype of interest to the set of all other haplotypes. Second, long-range LD around specific alleles is a prevalent feature in the genome; the significance of LD results should therefore be assessed relative to empirical distributions observed in genome-wide studies with larger numbers of samples. Third, accurate estimates of an allele's age require accurate genetic maps.
With the growing availability of genome-wide datasets, it should soon be possible to search the genome for signs of strong selective events [3] by studying the pattern of variation at every gene relative to a comprehensive genome-wide distribution. The results should shed light on important factors that have shaped our species and may provide valuable information about natural mechanisms of disease resistance.
Materials and Methods
Samples
DNA samples for 93 individuals from 12 multigenerational pedigrees of European-American ancestry were obtained from Coriell Repositories (http://locus.umdnj.edu/ccr). DNA samples from 93 healthy individuals (31 mother–father–child clusters) from the Yoruba in Nigeria were obtained as part of the International Collaborative Study of Hypertension in Blacks. DNA samples from 30 Han Chinese trios from Guanchi were included. DNA samples from a chimpanzee, gorilla, and orangutan were obtained from Coriell Repositories.
Genotype data
We genotyped 71 SNPs in and around and the CCR5-Δ32 using the mass spectrometry-based MassArray platform provided by Sequenom (San Diego, California, United States), implemented as previously described [18]. The names, locations, alleles, and flanks for all SNPs used are given in Table S1.
Microsatellite genotyping was conducted at the McGill University and Genome Quebec Innovation Center (Quebec, Canada), by use of MultiProbe and MiniTrak Liquid Handling Systems (Perkin-Elmer, Wellesley, California, United States) and 3730 DNA sequencers (Advanced Biosystems, Foster City, California, United States). PCR was performed with fluorescently labeled markers in standard conditions (annealing temperature of 54 °C).
We also used genotypes of 2,359 SNPs, distributed in 168 immunologic genes from 64 loci throughout the genome in the same three populations [15]. SNPs were selected from public databases in multiple batches over a 1.5-y period from July 2002 to December 2003. Preference was given to “double hit” SNPs which have been shown to be more likely to be validated [24]. These criteria may bias our ascertainment of haplotype structure and may reduce the representation of rare and population-specific variation; we comment in the paper where this bias might affect our observations.
We used publicly available data from the International Haplotype Map Project as a comparative distribution of variance in the genome with which to compare our results (http://www.hapmap.org).
Phasing
We prepared these files using Genehunter (http://www.broad.mit.edu/ftp/distribution/software/genehunter/) to uncover unambiguous phasing using family data [25]. The child chromosomes were then discarded, and we kept only the independent parent chromosomes. We then used PHASE (http://www.stat.washington.edu/stephens/software.html [26,27]) to obtain complete phased data.
Simulations
We used a computer program that simulates gene history with recombination based on a neutral model of evolution described elsewhere [19,28]. The program was modified to generate data comparable with that collected from the three populations—Chinese, European-American, and Yoruba. The simulations were calibrated to provide data consistent with the HapMap with respect to various genetic measures (including FST, heterozygosity, and minor-allele frequency distribution) and used model parameters (including demography and recombination rate) consistent with current estimates [19]. We simulated a long region (1 mbp in length) of DNA and then mimicked the SNP selection strategy used by the SNP Consortium [29], which was the source of most of the SNPs in our study. We modified the program to generate simulations of a partial selective sweep in 400 European-American chromosomes, where 32 chromosomes had a common ancestor 700 y ago as per Stephens et al. [8]. We also tested where the 32 chromosomes had a common ancestor 2,000 y ago.
FST
Mean pairwise distance fixation index, FST, was used to calculate genetic differentiation between the three populations [30,31]. FST partitions the total variance into within- and between-population components, quantifying the inbreeding effect of population substructure.
Heterozygosity
Nei's measure of heterozygosity [32], the probability that any two randomly chosen samples from a population are the same, was used to calculate SNP diversity:
where n is the number of chromosomes in the sample, k is the number of alleles at a locus, and pi is the frequency of the ith allele.
DAF distribution
We calculated the DAF distribution for all SNPs where it was likely that the ancestral allele could be determined by genotyping a representative chimpanzee, gorilla, and orangutan. If there was a consensus primate allele across all successfully genotyped primates, it was identified as the ancestral allele. Otherwise, no ancestral allele was defined.
EHH
EHH assesses the age of each haplotype at a gene by measuring the decay of the extended ancestral haplotype (i.e., SNPs far away from the gene), which occurs over time with recombination. For a population of individuals sharing core haplotype X, EHH is the probability that any two randomly chosen samples of core haplotype X have the same extended haplotype [3]. It is a measure of the decay of LD across a region of the genome that has two advantages: first, it can be used with multi-allelic markers so a core haplotype model can be studied if desired, and second, it measures LD across a region with many loci and not just between a pair of loci. The EHH is calculated as:
where t is the core haplotype tested, c is the number of samples of a particular core haplotype, e is the number of samples for a particular extended haplotype, and s is the number of unique extended haplotypes.
To correct for local variation in recombination rates, we can compare the EHH of a tested core haplotype to that of other core haplotypes present at the locus, using the relative EHH measure (i.e., REHH). REHH is the factor by which EHH decays on the tested core haplotype compared to the decay of EHH on all other core haplotypes combined. One must first calculate the “
,” the decay of EHH on all other core haplotypes combined. For this, we use the following equation where n is the number of different core haplotypes:
The relative EHH (i.e., REHH) is simply
. EHH and REHH were calculated for all haplotypes in all haplotype blocks for CCR5, HapMap Release 16 Chromosome 3, and the 1,000 simulated regions (120-chromosome and 500-chromosome sample sets). Haplotypes were placed into 20 bins based on their frequency. p-Values were obtained by log-transforming the EHH and REHH in the bins to achieve normality, and calculating the mean and standard deviation. All analysis was carried out using the Sweep software program (P. V., B. F., E. S. L., and P. C. S., unpublished data).
Observed historical recombination (marker breakdown, all EHH)
When comparing EHH/REHH values across regions, it is important to make sure that the value is being calculated at a similar genetic distance. This will soon be replaced with better cM values, but, where they are not known, this can be matched by the “marker breakdown,” that is the degree to which each added marker at a further distance causes the extended haplotypes to decay for all core haplotypes [3]. This gives an evaluation of how much historical recombination (observed recombinants) has occurred over a distance from the core, and therefore what genetic distance is being looked at. This can be calculated as “all EHH.”
where n is the number of different core haplotypes, c is the number of samples of a particular core haplotype, e is the number of samples of a particular extended haplotype, and s is the number of unique extended haplotypes.
Bifurcation diagram
To visualize the breakdown of LD on core haplotypes, we used bifurcation diagrams [3]. The root of each diagram is a core haplotype, identified by a dark-blue circle. The diagram is bidirectional, portraying both proximal and distal LD. Moving in one direction, each marker is an opportunity for a node; the diagram either divides or not, depending on whether both or only one allele is present. Thus, the breakdown of LD on the core haplotype background is portrayed at progressively longer distances. The thickness of the lines corresponds to the number of samples with the indicated long-distance haplotype.
Supporting Information
Figure S1
FST and Heterozygosity for SNPs within 100 kbp of CCR5 Compared to 100-kbp Sliding Windows for HapMap Release 16 for European-Americans
(54 KB DOC).
Click here for additional data file.
Figure S2 The DAF Distribution of CCR5 Compared to 100-kbp Sliding Windows for HapMap Release 16 for European-Americans
(62 KB DOC).
Click here for additional data file.
Figure S3 Haplotype Bifurcation Diagrams in European-Americans
(231 KB DOC).
Click here for additional data file.
Figure S4 The REHH versus Frequency Distribution at Matched Genetic Distance [17]
(99 KB DOC).
Click here for additional data file.
Figure S5 Estimating the Rate of Degradation of EHH
(40 KB DOC).
Click here for additional data file.
Figure S6 Remapping of Microsatellite Markers from First Study Given the Improved Genomic Maps
(45 KB DOC).
Click here for additional data file.
Figure S7 Microsatellite Genotyping
(71 KB DOC).
Click here for additional data file.
Figure S8 Comparison of Overall Genetic Diversity and Specific Haplotype EHH in Different Populations
(38 KB DOC).
Click here for additional data file.
Protocol S1 Recombination-Rate Estimates for CCR5 from Family-Based Linkage Studies (deCODE and Marshfield Maps), from Preliminary Sperm-Typing, and from Population Genetics (LDhat)
(32 KB DOC).
Click here for additional data file.
Table S1 Information for Δ32 (rs333), 70 SNPs, and Two Microsatellites Used in the Study
(30 KB XLS).
Click here for additional data file.
Table S2
CCR5-Δ32 EHH Values for Eight European-American Chromosomes versus the 32 Total Genotyped Chromosomes
(23 KB DOC).
Click here for additional data file.
Table S3 Haplotype Frequencies for Six Variants in Strong LD at CCR5, Genotyped in the Three Population Samples
(23 KB DOC).
Click here for additional data file.
Table S4 Extended Haplotype Length for Haplotypes of Different Frequency on HapMap Chromosome 3 in European-Americans
(23 KB DOC).
Click here for additional data file.
Table S5 Details of CCR5-Δ32 Date Estimates
(26 KB DOC).
Click here for additional data file.
Accession Number
The LocusLink (http://www.ncbi.nlm.nih.gov/LocusLink) accession number for the C-C chemokine receptor 5 is 1234.
PCS is funded by the Damon Runyon Cancer Research Foundation and by a L'Oreal award for Women in Science. EW was funded by the Cancer Research Institute. We thank Andrei Verner and his colleagues at McGill University and Genome Quebec Innovation Center for their work on microsatellite genotyping. We thank Mary Carrington, Dan Richter, Parisa Sabeti, and three anonymous reviewers for their suggestions and reviews of our manuscript.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PCS, EW, MC, DA, SO, and ESL conceived and designed the experiments. PCS, EW, MC, and JR performed the experiments. PCS, SFS, PV, BF, TSM, NP, and DR analyzed the data. SFS, PV, BF, RC, HH, and ESL contributed reagents/materials/analysis tools. PCS, EW, DA, and ESL wrote the paper.
Citation: Sabeti PC, Walsh E, Schaffner SF, Varilly P, Fry B, et al. (2005) The case for selection at CCR5-Δ32. PLoS Biol 3(11): e378.
Abbreviations
cMcentimorgan
DAFderived allele frequency
EHHextended haplotype homozygosity
EHLextended haplotype length
LDlinkage disequilibrium
REHHrelative extended haplotype homozygosity
SNPsingle-nucleotide polymorphism
==== Refs
References
Olson S Population genetics. Seeking the signs of selection Science 2002 298 1324 1325 12434033
Bamshad M Wooding SP Signatures of natural selection in the human genome Nat Rev Genet 2003 4 99 111 12560807
Sabeti PC Reich DE Higgins JM Levine HZ Richter DJ Detecting recent positive selection in the human genome from haplotype structure Nature 2002 419 832 837 12397357
Akey JM Eberle MA Rieder MJ Carlson CS Shriver MD Population history and natural selection shape patterns of genetic variation in 132 genes PLoS Biol 2004 2 e286 10.1371/journal.pbio.0020286 15361935
Walsh EC Mather KA Schaffner SF Farwell L Daly MJ An integrated haplotype map of the human major histocompatibility complex Am J Hum Genet 2003 73 580 590 12920676
Rockman MV Hahn MW Soranzo N Loisel DA Goldstein DB Positive selection on MMP3 regulation has shaped heart disease risk Curr Biol 2004 14 1531 1539 15341739
Taylor MF Shen Y Kreitman ME A population genetic test of selection at the molecular level Science 1995 270 1497 1499 7491497
Stephens JC Reich DE Goldstein DB Shin HD Smith MW Dating the origin of the CCR5 Δ32 AIDS-resistance allele by the coalescence of haplotypes Am J Hum Genet 1998 62 1507 1515 9585595
Tishkoff SA Varkonyi R Cahinhinan N Abbes S Argyropoulos G Haplotype diversity and linkage disequilibrium at human G6PD: Recent origin of alleles that confer malarial resistance Science 2001 293 455 462 11423617
Toomajian C Ajioka RS Jorde LB Kushner JP Kreitman M A method for detecting recent selection in the human genome from allele age estimates Genetics 2003 165 287 297 14504236
Dean M Carrington M Winkler C Huttley GA Smith MW Genetic restriction of HIV-1 infection and progression to AIDS by a deletion allele of the CKR5 structural gene. Hemophilia Growth and Development Study, Multicenter AIDS Cohort Study, Multicenter Hemophilia Cohort Study, San Francisco City Cohort, ALIVE Study Science 1996 273 1856 1862 8791590
Carrington M Kissner T Gerrard B Ivanov S O'Brien SJ Novel alleles of the chemokine-receptor gene CCR5
Am J Hum Genet 1997 61 1261 1267 9399903
Carrington M Dean M Martin MP O'Brien SJ Genetics of HIV-1 infection: Chemokine receptor CCR5 polymorphism and its consequences Hum Mol Genet 1999 8 1939 1945 10469847
Bamshad MJ Mummidi S Gonzalez E Ahuja SS Dunn DM A strong signature of balancing selection in the 5′ cis-regulatory region of CCR5
Proc Natl Acad Sci U S A 2002 99 10539 10544 12149450
Walsh EC Sabeti P Hutcheson HB Fry B Schaffner SE Searching for signals of evolutionary selection in 168 genes related to immune function Hum Genet 2005 In press
Fay JC Wu CI Hitchhiking under positive Darwinian selection Genetics 2000 155 1405 1413 10880498
Kong A Gudbjartsson DF Sainz J Jonsdottir GM Gudjonsson SA A high-resolution recombination map of the human genome Nat Genet 2002 31 241 247 12053178
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J The structure of haplotype blocks in the human genome Science 2002 23 2225 2229
Schaffner S Foo C Gabriel SB Reich D Daly MJ Calibrating a coalescent simulation of human genome sequence variation Genome Res 2005 In press
Reich DE Goldstein DB Goldstein DB Schlotter C Estimating the age of mutations using variation at linked markers Microsatellites: Evolution and applications 1999 Oxford Oxford University Press 128 138
Duncan SR Scott S Duncan CJ Reappraisal of the historical selective pressures for the CCR5 Δ32 mutation J Med Genet 2005 42 205 208 15744032
Wooding S Stone AC Dunn DM Mummidi S Jorde LB Contrasting effects of natural selection on human and chimpanzee C-C chemokine receptor 5 Am J Hum Genet 2005 76 291 301 15625621
Nielsen R Bustamante C Clark AG Glanowski S Sackton TB A scan for positively selected genes in the genomes of humans and chimpanzees PLoS Biol 2005 3 e170 10.1371/journal.pbio.0030170 15869325
Reich DE Gabriel SB Altshuler D Quality and completeness of SNP databases Nat Genet 2003 33 457 458 12652301
Kruglyak L Daly MJ Reeve-Daly MP Lander ES Parametric and nonparametric linkage analysis: A unified multipoint approach Am J Hum Genet 1996 58 1347 1363 8651312
Stephens M Donnelly P A comparison of bayesian methods for haplotype reconstruction from population genotype data Am J Hum Genet 2003 73 1162 1169 14574645
Stephens M Smith NJ Donnelly P A new statistical method for haplotype reconstruction from population data Am J Hum Genet 2001 68 978 989 11254454
Hudson RR Gene genealogies and the coalescent process Oxf Surv Evol Biol 1990 7 1 44
Sachidanandam R Weissman D Schmidt SC Kakol JM Stein LD A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms Nature 2001 409 928 933 11237013
Nei M Chesser RK Estimation of fixation indices and gene diversities Ann Hum Genet 1983 47 253 259 6614868
Akey JM Zhang G Zhang K Jin L Shriver MD Interrogating a high-density SNP map for signatures of natural selection Genome Res 2002 12 1805 1814 12466284
Nei M Molecular evolutionary genetics 1987 New York Columbia University Press 512
|
16248677
|
PMC1275522
|
CC BY
|
2021-01-05 08:21:45
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e378
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030378
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1624867810.1371/journal.pbio.0030381Research ArticleEvolutionMicrobiologyVirologyVirusesEvolution of Mutational Robustness in an RNA Virus Virus Co-Infection and Evolved RobustnessMontville Rebecca
1
Froissart Remy
1
¤aRemold Susanna K
1
¤bTenaillon Olivier
2
Turner Paul E [email protected]
1
1 Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut, United States of America,2 INSERM, Equipe, Ecologie et Evolution de Micro-Organismes, Henri Huchard, Paris, FrancePenny David Academic EditorMassey UniversityNew Zealand11 2005 1 11 2005 1 11 2005 3 11 e3816 6 2005 8 9 2005 Copyright: © 2005 Montville et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
A Role for Selection in the Evolution of Genetic Robustness
Mutational (genetic) robustness is phenotypic constancy in the face of mutational changes to the genome. Robustness is critical to the understanding of evolution because phenotypically expressed genetic variation is the fuel of natural selection. Nonetheless, the evidence for adaptive evolution of mutational robustness in biological populations is controversial. Robustness should be selectively favored when mutation rates are high, a common feature of RNA viruses. However, selection for robustness may be relaxed under virus co-infection because complementation between virus genotypes can buffer mutational effects. We therefore hypothesized that selection for genetic robustness in viruses will be weakened with increasing frequency of co-infection. To test this idea, we used populations of RNA phage φ6 that were experimentally evolved at low and high levels of co-infection and subjected lineages of these viruses to mutation accumulation through population bottlenecking. The data demonstrate that viruses evolved under high co-infection show relatively greater mean magnitude and variance in the fitness changes generated by addition of random mutations, confirming our hypothesis that they experience weakened selection for robustness. Our study further suggests that co-infection of host cells may be advantageous to RNA viruses only in the short term. In addition, we observed higher mutation frequencies in the more robust viruses, indicating that evolution of robustness might foster less-accurate genome replication in RNA viruses.
RNA phage viruses evolved under high co-infection of host cells are less robust to mutations than those propagated under low co-infection, suggesting that co-infection may be advantageous to RNA viruses only in the short term.
==== Body
Introduction
Mutational (genetic) robustness can be defined as constancy of phenotype in the face of mutational perturbation [1]. Genetic variance and differences in phenotypic performance among genotypes underlie all of Darwinian evolution. Thus, robustness is crucial to the understanding of evolution because it dictates phenotypic expression of genetic variation [2]. But it remains unclear whether robustness is merely accidental or a consequence of natural selection. The most straightforward explanation for the evolution of robustness is adaptationist. For a well-adapted population, almost all mutations lead to deviations from optimal performance in the selective environment. Populations at equilibrium should therefore experience selection for mutational robustness. However, evolution of genetic robustness is hard to observe in most laboratory systems because equilibrium states are difficult to achieve (or definitively prove) and the benefit of mutational robustness is not experienced until offspring carrying mutations arise [2]. For these reasons, the evidence for adaptive evolution of mutational robustness in biological populations remains controversial [3–5]. Therefore, the vast majority of studies demonstrating the phenomenon have relied on theory [6] or artificial life systems [7].
High mutation rate is perhaps the most important prerequisite for adaptive genetic robustness [2], so mutational robustness should be strongly selected in biological systems experiencing elevated mutation rates [6]. Thus, strong candidates for observing adaptive robustness would be RNA viruses, which typically feature mutation rates that are orders of magnitude higher than in DNA systems [8]. In general, the theoretical predictions for adaptive robustness under elevated mutation rates assume that phenotype expression results solely from the underlying genotype. However, many viruses feature complementation, a mechanism whereby low-fitness genotypes can phenotypically profit from intracellular proteins made by co-infecting strains of high fitness [9–11]. Co-infection coupled with complementation can therefore act as a mechanism that provides phenotypic buffering in the event of genomic mutations, similar to other buffering mechanisms such as gene duplication and diploidy that might have evolved to facilitate canalization in higher organisms [12,13]. We recently demonstrated that complementation can buffer the harmful fitness effects of deleterious alleles in co-infecting populations of the segmented RNA phage φ6 [10]. Here we further examine the evolutionary consequences of virus co-infection, by predicting that co-infecting phages are likely to experience weakened selection for mutational robustness.
We previously conducted laboratory experimental evolution of φ6 populations at low and high levels of co-infection [14]. In that study, a single clone of the wild-type virus was used to found six replicate populations, which underwent adaptation to Pseudomonas syringae pathovar (pv) phaseolicola bacteria for hundreds of virus generations. Three of the populations were evolved at low multiplicity of infection (MOI; ratio of infecting viruses to bacterial cells), and three at high MOI. Co-infection level was controlled by mixing viruses and bacteria in liquid medium at a given MOI, allowing sufficient time for virus attachment to cells, and then plating a dilution of the mixture onto agar with a superabundance of cells. During overnight incubation, viruses formed distinct plaques in the lawn, which resulted from lysis of infected cells and the release of viral progeny that infected neighboring cells. The passage cycle was repeated by harvesting plaques, removing the bacteria by filtration, and mixing viruses and naive bacteria at the controlled MOI. A total of 60 passage cycles were conducted, which equaled roughly 300 viral generations (i.e., five generations occur during overnight plaque formation [14]). Every fifth generation, populations in the low co-infection treatment experienced MOI = 0.002 whereas those in the high co-infection treatment experienced MOI = 5; otherwise, all aspects of the treatment environments were equal. Assuming Poisson sampling [15], the proportion of cells infected with n phages is P(n) = e−MOI × MOI. Therefore, at MOI = 0.002, only approximately 0.1% of all infected (n > 0) cells are co-infected and clonal infections predominate. In contrast, approximately 97% of infected cells should be multiply infected at MOI = 5, and co-infection predominates, generally with two to three viruses (the limit to co-infection in φ6 [16]) infecting each host. (The percentage difference in multiply-infected cells across the MOI treatments is probably less than 1,000-fold, because we observe greater than Poisson-expected entry of φ6 particles into the same cell when viruses are grown at low MOI [16].) Thus, the treatment environments were the same, except that populations at high MOI more often experienced co-infection, an environment demonstrated to allow intracellular virus interactions such as complementation [10].
Here we examined whether evolution of mutational robustness occurs differently for viruses evolved at low and high levels of co-infection. We tested the hypothesis by randomly isolating clones from each of the six previously evolved populations, and using these clones to found lineages that were subjected to a mutation accumulation experiment [17–19]. Mutation accumulation was achieved by serially propagating the lineages in a new environment where they experienced severe bottlenecking. The sampling of mutations in these experiments is nearly unbiased because genetic drift overwhelms natural selection during the extreme bottlenecks. By removing selection, all non-lethal mutations can increase to fixation with roughly the same probability, regardless of whether they are deleterious, advantageous, or neutral [17–19]. However, because most mutations are deleterious, mutation accumulation experiments tend to cause reduced fitness [17–19]. We compared the fitness consequences of mutation accumulation for lineages drawn from the different co-infection treatments, by measuring the mean magnitude and variance in fitness change that occurred as a result of bottlenecking. Our data confirmed the hypothesis that viruses historically evolved under high co-infection are relatively less robust than those evolved under low co-infection, demonstrated by their greater mean magnitude and variance in fitness changes generated by addition of random non-lethal mutations.
Results
We isolated ten clones at random from each replicate population in the low co-infection and high co-infection level treatments at generation 300, and used these to found 60 independent virus lineages (ten clones × three populations × two co-infection treatments). We then conducted a mutation accumulation experiment [17,19], in which the lineages were subjected to a new habitat containing extreme population bottlenecks consisting of single-virus passages on P. phaseolicola for 20 d (see Materials and Methods). Plaque formation (i.e., five generations of ordinary virus growth) occurred in between bottlenecks, resulting in approximately 100 generations of virus evolution via mutation accumulation. Under these conditions, lineages may become fixed for any non-lethal mutation through the process of genetic drift. The genomic mutation rate in φ6 is gauged to be 0.067 deleterious mutations per generation [19]. We therefore estimated that one mutation on average had been fixed in each lineage (i.e., 0.067 × 20 bottleneck events ≍ 1.3), in which it is assumed that the majority of spontaneous mutations are deleterious. At the end of the mutation accumulation experiment, we conducted replicated (n = 3) fitness assays against a common virus competitor for the focal genotypes (pre-bottleneck founding clone and post-bottleneck final clone) of each lineage. In this way, we were able to measure the mean change in log10 fitness (Δlog10
W) for each lineage as a result of mutation accumulation; Δlog10
W = log10
W
post-bottleneck − log10
W
pre-bottleneck. This design resulted in 360 total fitness estimates (60 lineages × two focal genotypes per lineage × three replicate estimates), in which focal genotypes of a lineage were always assayed within the same temporal block.
The Δlog10
W values estimate the fitness effect of adding roughly one random non-lethal mutation to the founding genotype of a lineage. Mutational robustness is defined as decreased phenotypic variability in the face of mutational change. Thus, one set of genotypes can be considered more robust than another collection of genotypes if the first group has a significantly lower variance in the fitness change brought on by addition of mutation(s) to the genome [2,20]. In our study, greater robustness would be indicated by relatively lower among-lineage variance in the fitness changes brought on by the bottlenecking experiment. We therefore analyzed the Δlog10
W values using mixed linear models that can fit different variances for random factors, depending on the level of the fixed factor with which they are associated [21]. Our approach is uniquely powerful for testing effects of co-infection treatment on the evolution of robustness, because we are able to estimate and test differences between the among-lineage variance for lineages drawn from low versus high co-infection level populations.
We detected significantly lower variance in Δlog10
W among the lineages evolved under low co-infection (Table 1); these data supported the hypothesis that selection for mutational robustness is stronger when viruses rarely experience co-infection. The relatively greater robustness of the low co-infection–evolved viruses is evident in Figure 1, which shows the tighter clumping of mean Δlog10
W values around the grand mean for virus lineages evolved under low co-infection, and the lower deviation in Δlog10
W associated with these estimates. However, close inspection of the data revealed the possibility that a low value (mean Δlog10
W = −0.698 ± 0.215 standard error of the mean) in the high co-infection group of lineages might be driving this result. To test this possibility, we reanalyzed our data by excluding the value. Results showed that presence or absence of this value did not affect the outcome of our analysis or its conclusions; rather, we still observed a statistically greater variance in fitness change for the high co-infection lineages relative to their low co-infection counterparts (Table S1).
Figure 1 Viruses Evolved under Low Co-Infection Are More Robust than Those Evolved under High Co-Infection
Each point is the mean (n = 3, ± standard error of the mean) change in log10 fitness (Δlog10
W) resulting from mutation accumulation, for an independent lineage founded by a virus clone evolved under low level of co-infection (circles) or high level of co-infection (squares). Horizontal lines are grand means for Δlog10
W among lineages within a treatment, and the dashed lines indicate one standard deviation away from the grand mean.
Table 1 Mixed Linear Models Testing Differences in Mean and Variance of Change in Log10 Fitness for Bottlenecked Virus Lineages Differing in Co-Infection History
Our analysis comparing variance in Δlog10
W between the two groups of evolved lineages is generally regarded as a rigorous method to test for differences in robustness among groups of lineages with differing ecological histories [2,20]. However, some theory also suggests that increased mutational robustness should coincide with reduced effects of mutations on mean fitness [22]. It is widely assumed that the majority of spontaneous mutations are deleterious; thus, mutation accumulation experiments that foster the action of genetic drift over natural selection should on average lead to decreased fitness in a bottlenecked lineage. Our results were consistent with this logic; the grand mean of Δlog10
W was below zero for both sets of treatment lineages following mutation accumulation (Figure 1). Within each group, at least one-half of the lineages showed mean values below zero, but these were rarely statistically significant, most likely due to the high measurement error expected in RNA systems. Many of the lineages founded by viruses that evolved under low co-infection showed a positive change in fitness, but none of these values was significantly different from zero; thus, we cannot conclude from the data whether the fixed mutation (if a mutation fixed at all) was neutral, mildly deleterious, or mildly beneficial. However, by definition this result is consistent with the relatively greater robustness of low co-infection–evolved viruses, as we would expect phenotypic constancy to cause their Δlog10
W values to hover close to zero. Also consistent was the greater number of high co-infection lineages experiencing a statistically significant drop in fitness (Table S2), but this difference among groups was minimal (i.e., four versus three significant lineages). More importantly, however, the significantly smaller magnitude of mean Δlog10
W for the low co-infection lineages (Table 1) was a further indicator that these viruses were more robust than the lineages historically evolved under high co-infection. Once again, we found this conclusion was unaffected when we reanalyzed the data by removing the outlying lineage in the high co-infection group (Table S1).
Claims of mutational robustness are easily confounded by mutational differences among genotypes [5]. For example, it could be that viruses evolved at low and high levels of co-infection are equally robust, but less frequent mutation of low co-infection lineages is responsible for the lower observed variance in Δlog10
W. To test this alternative explanation, we measured mutant frequencies (see Materials and Methods) occurring on two novel hosts, P. syringae pv.atrofaciens and pv. tomato, using one pre-bottleneck clone randomly chosen from each population in the two treatments. Host-range mutations on the two hosts occur at different viral loci, meaning that the vast majority of φ6 mutants arising on P. atrofaciens cannot infect P. tomato, confirmed by genome sequencing of host-range mutants (S. Duffy, C. Burch, and P.E. Turner, unpublished data). The data (Figure 2) did not support the alternative hypothesis because mutant frequencies on each host were not significantly higher for the strains historically evolved under high co-infection. Rather, mutants on P. atrofaciens occurred significantly less often for strains evolved under high co-infection (analysis of variance [ANOVA] with F
[1,4] = 30.49, p = 0.0053), and mutant frequencies on P. tomato did not differ according to past ecological history (ANOVA with F
[1,4] = 6.48, p = 0.0636). It is unknown whether the frequencies of host-range mutants are representative of mutations occurring elsewhere in the genome, but the robustness differences we observed cannot be alternatively explained by these data.
Figure 2 Differences in Virus Robustness Are Not Confounded by Increased Mutation Frequencies in Viruses Evolved under High Co-Infection
Mean mutation frequency (± 95% confidence interval) was assayed for wild-type φ6 (ANC), and for one pre-bottleneck clone isolated from each population evolved under low co-infection (L1–L3) and high co-infection (H1–H3). Assays on P. tomato (circles) and P. atrofaciens (squares) were replicated 5-fold. Assays for P. atrofaciens mutants challenged with growth on P. tomato (diamonds) were replicated 6-fold, except for H3 (n = 2).
Discussion
Mutation and natural selection are cornerstones of evolutionary biology, but elucidating their interplay has proved challenging. To fill this intellectual gap, we conducted experiments to better understand how selection can drive the evolution of genetic robustness. The inherently high error rates in genome replication occurring in RNA viruses make them an obvious choice for examining the adaptive evolution of mutational robustness. We predicted that selection for robustness should be weakened when the RNA phage φ6 experiences high levels of co-infection, owing to virus complementation that buffers mutational effects [10]. Consistent with this hypothesis, our data showed that φ6 genomes evolved under high co-infection were less robust than those propagated under low co-infection, demonstrated by their greater mean magnitude and variance in fitness changes brought on by accumulation of random non-lethal mutations. Our mutation frequency data suggested that the observed difference in robustness was not alternatively explained by elevated mutation frequencies in the viruses evolved under high co-infection.
One possibility is that the rare occurrence of complementation at low co-infection caused these viruses to adapt by evolving greater robustness than the wild-type φ6 ancestor; but this is unknown. If so, our data suggest it is unlikely that this greater robustness was accompanied by more accurate RNA replication; ANOVA shows mutation frequencies of the low co-infection genotypes do not differ from the ancestor on either host (P. atrofaciens: F
[1,2] = 0.00, p = 0.9940; P. tomato: F
[1,2] = 0.17, p = 0.7175). This result is intriguing because the robust viruses seem to feature higher mutation frequencies on average. In turn, the data suggest that evolution of mutational robustness (whatever the underlying molecular mechanism) allows RNA viruses to tolerate less-accurate genome replication, perhaps explaining why these viruses remain highly mutable. Unraveling the exact molecular mechanism(s) for these results would be of great interest. Furthermore, our study implies a need for population genetic models that consider the impact of adaptive robustness on the evolution of replication fidelity, theory that has not been previously explored.
For the wild-type φ6 ancestor and evolved genotypes in our study, mutation frequencies were lower than expected given the ease with which host-range–marked mutants of φ6 were obtained in previous studies. That is, host-range mutants occurred at frequencies ranging between 10−8 and 10−10 on P. atrofaciens and P. tomato, whereas we have observed frequencies between 10−4 and 10−6 for φ6 on the host P. pseudoalcaligenes [10,23]. To explain this disparity, we hypothesized that the host-range phenotypes in the current study resulted from a two-step mutation requiring a precursor allele change. We tested this idea by measuring the frequency with which P. atrofaciens mutants were able to further mutate to infect P. tomato hosts (see Materials and Methods). The data (Figure 2) supported the assumption that the host-range mutations occur at different loci (i.e., further mutation was needed for growth on the second host); the two-step requirement was also evident because these assays yielded the expected mutant frequencies between 10−4 and 10−6. Also, these results further suggested that the decreased fitness effects in the low co-infection viruses did not result from more accurate RNA replication, as values for the ancestor and these strains did not differ in the two-step experiment (P. atrofaciens mutants challenged with growth on P. tomato: F
[1,4.26] = 0.07, p = 0.8126). Most important, we found that mutation frequency in these assays was significantly lower in the high co-infection viruses (ANOVA: with F
[1,4.26] = 9.30, p = 0.0351), which again argues that our robustness conclusions are not confounded by elevated mutation rates in these viruses. Overall, our data suggest that mutation rate governed by the accuracy of the viral replicase is a trait under selection, especially in the viruses evolved at high co-infection. It is unclear why viruses that frequently experience complementation would evolve higher fidelity of replication, and this may be a pleiotropic effect of selection occurring elsewhere in the genome. The possible adaptive significance of this result is unclear and merits future exploration.
We considered the potential relevance of two other possible confounding factors in our study. First, we examined whether the two groups of viruses differed in fitness prior to mutation accumulation; the alternative explanation is that the low co-infection viruses were already of low fitness (in comparison to the high co-infection strains), and that fixation of one or more additional mutations via bottlenecking did not lead to a further reduction in their fitness. That is, lower fitness genotypes of φ6 are shown to be less affected by addition of further deleterious mutations, a result demonstrating that diminishing-returns epistasis can operate in phage φ6 [19]. This phenomenon would provide an alternative explanation for our data if the average fitness of pre-bottleneck clones drawn from the low co-infection populations was lower than the fitness of those sampled from the high co-infection treatment. We tested this idea by determining whether the fitness of the 60 pre-bottleneck clones in our study differed according to co-infection treatment. Results showed no statistical difference (ANOVA with F
[1,3.94] = 0.05, p = 0.84). Thus, we rejected the possibility that diminishing-returns epistasis, or any argument that hinged on differing fitness among the starting clones, provided an alternative explanation for our results.
We also considered that erroneous conclusions may be drawn from mutation accumulation experiments if conditions impose a change in selective environment for only a subset of test lineages. A similarity between the mutation accumulation experiment and the low MOI treatment is that viruses must infect cells alone, whereas in every fifth generation of the high MOI treatment viruses are forced to undergo co-infection. Thus, it might be argued that the high co-infection lineages (but not low co-infection lines) experienced a change in growth conditions, and that the magnitude and variability in their performance in the bottlenecking experiment results from adaptation to the new conditions. However, we believe that this could not have been a confounding factor in our study for several reasons.
First, the pre-bottleneck clones drawn from both MOI treatments performed equally well prior to mutation accumulation (see above), as measured using fitness assays conducted at low MOI. These data may seem surprising, given that at generation 200 the treatment populations showed significantly different fitness under low-MOI conditions [14]. However, this difference was at the level of populations, whereas the current study used genotypes drawn from these populations; mean fitness of a microbial population can differ from mean fitness of clones drawn from the population [24,25]. One possibility is that the parent population contains variants of very high fitness in the assay environment. Because fitness assays in microbial systems typically span multiple generations (i.e., five generations in our case), a genetic variant within the mixed population will be overrepresented by the end of the assay, which elevates the overall performance of the population revealing its “evolutionary potential” in the assay environment. In contrast, a clone's performance in two habitats reveals only its phenotypic plasticity and not the evolutionary potential of its parent population. This could easily account for the very large fitness differences at low MOI observed for populations evolved under low and high co-infection, contrasted with the equal fitness of pre-bottleneck clones drawn from these populations assayed in the same environment.
Second, the mutation accumulation experimental conditions can be considered novel for all test lineages. In particular, the experimental evolution necessitated that viruses attached to host cells in liquid medium prior to overnight growth on agar plates, whereas the mutation accumulation habitat did not. Even if phage φ6 attachment to cells in liquid versus on agar surfaces is fundamentally the same (this is not well described), it is plausible that these two habitats differ substantially in the dispersal of phages between infected cells (T. Berngruber and L. Chao, unpublished data).
Third, it is highly unlikely that the test lineages could have adapted to the mutation accumulation conditions at all. Plaque formation on agar necessarily involves expansion of the bottlenecked population to large size, which allows for some positive selection. That is, a plaque forms from a single virus experiencing five generations of growth on the plate, where the average number of progeny made by an infecting virus is 100 particles, yielding ~108 virus particles within a plaque. However, the mutation accumulation design prevents virus adaptation because the action of genetic drift overwhelms that of natural selection; i.e., lineages evolve at an effective population size of N ≍ 2, the harmonic mean of the serial passage fluctuating between 1 and 108 viruses.
Our study suggests several intriguing possibilities for future research in φ6 and other RNA viruses. Identifying whether the low co-infection populations, the high co-infection populations, or both have changed in robustness relative to their common ancestor could help shed light on the genetic mechanism(s) underlying this difference in the φ6-derived viruses. (Our goal was to examine the impact of co-infection history on evolution of robustness, which was achieved by comparing effects of mutation accumulation among lineages founded by random clones drawn from each treatment, so our study omitted lineages founded by the ancestor because these would lack standing genetic variation, preventing any direct comparisons to results from the evolved groups.) The 300-generation experiment [14] that preceded mutation accumulation was probably too short for evolution of de novo adaptive epistatic interactions [26]. Therefore, it is perhaps more likely that the high co-infection populations became less robust (relative to the ancestral state) due to weakened selection for robustness owing to genetic buffering provided by complementation. Future viral-genomic research could determine whether this outcome occurred due to fixation of non-adaptive alleles, or due to antagonistic pleiotropy (i.e., selection for alleles that enhance virus fitness under co-infection, but diminish performance in the unselected trait robustness). Recent studies have addressed the evolutionary consequences of co-infection, especially in RNA viruses [9–11]. During co-infection, viruses can be considered analogous to polyploid organisms because several copies of the virus genome are present inside the same host cell. Co-infection may be advantageous in the short term because it allows for complementation, which can partially or fully mask the cost of deleterious mutations. However, co-infection may be detrimental in the long term, because complementation can slow the rate at which deleterious alleles are eliminated from the virus population [10] and can promote selection for defective-interfering particles or other cheater genotypes that reduce mean population fitness [16]. Our experiments suggest that an additional cost of co-infection is the weakened selection for robustness that may cause evolution of relatively brittle genomes. We obtained this result by contrasting the strength of selection for adaptive robustness in low versus high co-infection populations. In this way, we purposefully avoided the more difficult task of examining de novo evolution of robustness that current theory suggests is most likely to occur after populations have reached equilibrium; other researchers might benefit from such an approach in their study systems.
Materials and Methods
Strains and culture conditions
Phages and bacteria were cultured at 25 °C in LC medium, Luria-Bertani broth (pH 7.5) [14]. Overnight cultures of bacteria were grown from a single colony placed in 10 ml LC medium, with shaking incubation (120 rpm). P. phaseolicola was purchased from American Type Culture Collection (
ATCC #21781; Manassas, Virginia, United States). L. Mindich (Public Health Research Institute, Newark, New Jersey, United States) kindly provided host strain P. pseudoalcaligenes ERA (East River isolate A); G. Martin (Cornell University, Ithaca, New York, United States) kindly providedP. syringae pv.tomato and P. syringae pv.atrofaciens. Bacterial stocks were stored in 4:6 glycerol/LC (v/v) at −80 °C. Viruses were grown on lawns made from overnight bacterial cultures. Agar concentrations in plates were 1.5% and 0.7% for bottom and top LC agar, respectively. Plates contained 3 ml of top agar and a 200 μl bacterial lawn. Phage lysates were prepared by growing viruses on a P. phaseolicola lawn for 24 h; plaques were then collected and filtered (0.22 μm filter, Durapore; Millipore, Billerica, Massachusetts, United States) to remove bacteria. Phage lysates were stored at −20 °C in 4:6 glycerol/LC (v/v).
Virus evolution
Virus populations were experimentally evolved at MOI of 0.002 or 5 for 250 generations [14], followed by an additional 50 generations in our laboratory. MOI was imposed every fifth generation by allowing viruses to attach to P. phaseolicola cells in a test tube containing liquid LC medium, incubated for 40 min at 25 °C. These infected cells were then diluted and plated on a P. phaseolicola lawn on which viruses form distinct plaques. Population size of N ≍ 500 was held constant across treatment populations by controlling the number of harvested plaques [14]. On average, each plaque in the low co-infection treatment was produced by one phage, whereas that in the high co-infection treatment was produced by five co-infecting parents; thus, N ≍ 500 was controlled by harvesting 500 plaques from a low MOI population and 100 plaques from a high MOI population. After the experimental evolution, ten clones were chosen at random from each population. Each of the 60 clones was then used to found a single lineage subjected to 20 d of mutation accumulation via bottlenecking [17,19]. Bottlenecking was achieved by cutting out a single 24-h-old plaque from a P. phaseolicola lawn, placing it in sterile liquid medium, and vortexing gently to disperse the progeny viruses contained within the plaque. A sterile platinum loop was then used to streak the viruses onto agar containing a naive host lawn. After 24-h incubation, the viruses formed distinct plaques, and the process was repeated for 20 consecutive days. Because each plaque grows from a single virus, daily passage caused a test lineage to be forced through an extreme bottleneck (population size of one individual), in which the intense drift allows fixation of non-lethal mutations, the majority of which are presumed to be deleterious. Because the phage expand to approximately 108 phage in five generations within a plaque, this creates the opportunity for selection to operate within a population propagated by such one-plaque transfers. Thus, positive selection on the plate should bias against fixation of highly deleterious mutations, and it is possible to observe the fixation of a rare beneficial mutation in a bottlenecked lineage [17]. However, the intensity of selection is not sufficient to overcome the intense genetic drift generated by the bottlenecks, causing fitness to generally decline due to accumulation of deleterious mutations [17–19]. Also, we note that in our experiments any confounding effects of co-infection on the plate could be ignored because the virus dilution (or streak) grew within a superabundant lawn (i.e., no more than 400 viral plaques were allowed to form on a P. phaseolicola lawn containing 8 × 108 cells; MOI on the plate of 5 × 10−7).
Fitness assays
Fitness assays consisted of paired-growth experiments [17], which compared 24-h growth on P. phaseolicola of a focal genotype relative to a common competitor of φ6 bearing a genetic marker (i.e., ability to infect P. pseudoalcaligenes bacteria [27]). The competitors were mixed at a 1:1 volumetric ratio, and then a dilution of this mixture containing approximately 400 viruses was plated on a P. phaseolicola lawn. Because no pre-attachment occurred before plating, every virus in the lawn infected a cell alone. After 24-h incubation, the approximately 400 plaques were harvested and filtered to obtain a cell-free lysate. The ratios of competing genotypes in the starting mixture (R
0) and in the harvested lysate (R
1) were obtained by plating on mixed lawns of P. phaseolicola and P. pseudoalcaligenes (200:1 mixture) on which ordinary and host-range genotypes form turbid and clear plaques, respectively. Fitness (W) was defined as the relative change in ratio of ordinary to host-range virus, or W =R
1
/R
0
.
Mutant frequency estimates
We measured the appearance of host-range mutants formed on selective plates: bacterial lawns of P. tomato or P. atrofaciens. A high-titer lysate (typically ~1010 viruses per ml) of a virus genotype was grown and titered on P. phaseolicola, and a sample of the lysate was then screened on a selective plate. Mutant frequency was calculated as the number of plaque-forming mutants per viruses in the inoculum; when no mutants were found, we used the limit of detection as a conservative estimate. Acquisition of a second host-range mutation was examined by rearing phage mutants on P. atrofaciens, followed by selective plating on P. tomato.
Statistical analyses
A mixed general linear model controlling for experimental assay day was run [21], testing the effects of the fixed-factor co-infection treatment (low MOI or high MOI) and the random-factors population nested within treatment, and the lineage nested within population and treatment on the difference in the (log-transformed) Wrightian relative fitnesses of the founding and final clone from each bottleneck lineage. For these two random factors, a term testing both for variability in means among levels and for differences in variance between the treatments was included. For example, population (treatment) means is a term testing for variability in means among populations, whereas population (treatment) variances is a term testing for whether low-MOI populations and high-MOI populations differ in variance. The fixed effect was tested using an approximate F-test with the denominator degrees of freedom estimated using the Satterthwaite approximation. These denominator degrees-of-freedom estimates depend both on sample sizes and variance structure. Random factors were tested using likelihood-ratio tests, which compare the restricted likelihood of models with and without the parameter of interest [21]. This method is asymptotically based and must be adjusted when the null hypothesis is on the boundary of the parameter space. Although approximate, this test has the advantage that it can be used for all the hypotheses considered here, including those that test for variance heterogeneity [28].
Supporting Information
Table S1 Summary of Changes in Log10 Fitness for Bottlenecked Virus Lineages
(91 KB DOC).
Click here for additional data file.
Table S2 Reanalysis of Mixed Linear Models Shown in Table 1, Excluding an Unusually Low Value (Population H2, Lineage 5)
(26 KB DOC).
Click here for additional data file.
We thank C. Burch, J. Hermisson, G. Wagner, and members of our lab group for useful comments and enlightening discussion. This work was supported by grants from the French government to RF and OT, a Yale Institute for Biospheric Studies fellowship to SKR, and grants to PET from the United States National Science Foundation (DEB-01–29089 and DEB-02–01860) and the Woodrow Wilson Foundation.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RF, SKR, OT, and PET conceived and designed the experiments. RM performed the experiments. RM, SKR, and PET analyzed the data. RM, RF, OT, and PET wrote the paper.
¤a Current address: UMR 385 Biologie and Genetique des Interactions Plante-Parasite INRA-CIRAD-ENSAM, UMR 2724 Génétique et Évolution des Maladies Infectieuses CNRS-IRD, Montpellier, France
¤b Current address: Department of Biology, University of Louisville, Louisville, Kentucky, United States of America
Citation: Montville R, Froissart R, Remold SK, Tenaillon O, Turner PE (2005) Evolution of mutational robustness in an RNA virus. PLoS Biol 3(11): e381.
Abbreviations
ANOVAanalysis of variance
MOImultiplicity of infection
pvpathovar
==== Refs
References
Waddington CH The strategy of the genes; A discussion of some aspects of theoretical biology 1957 London Allen and Unwin 262
de Visser J Hermisson J Wagner GP Meyers LA Bagheri-Chaichian H Perspective: Evolution and detection of genetic robustness Evolution 2003 57 1959 1972 14575319
Scharloo W Canalization: Genetic and developmental aspects Ann Rev Ecol Syst 1991 22 65 93
Stearns SC Kawecki TJ Fitness sensitivity and the canalization of life-history traits Evolution 1994 48 1438 1450
Hermisson J Wagner GP The population genetic theory of hidden variation and genetic robustness Genetics 2004 168 2271 2284 15611191
Wagner GP Booth G Bagheri-Chaichian H A population genetic theory of canalization Evolution 1997 51 329 347
Wilke CO Evolution of digital organisms at high mutation rates leads to survival of the flattest Nature 2001 412 331 333 11460163
Drake JW Holland JJ Mutation rates among RNA viruses Proc Natl Acad Sci U S A 1999 96 13910 13913 10570172
Bretscher MT Althaus CL Muller V Bonhoeffer S Recombination in HIV and the evolution of drug resistance: For better or for worse? Bioessays 2004 26 180 188 14745836
Froissart R Wilke CO Montville R Remold SK Chao L Co-infection weakens selection against epistatic mutations in RNA viruses Genetics 2004 168 9 19 15454523
Novella IS Reissig DD Wilke CO Density-dependent selection in vesicular stomatitis virus J Virol 2004 78 5799 5804 15140977
Krakauer DC Plotkin JB Redundancy, antiredundancy, and the robustness of genomes Proc Natl Acad Sci U S A 2002 99 1405 1409 11818563
Gu ZL Steinmetz LM Gu X Scharfe C Davis RW Role of duplicate genes in genetic robustness against null mutations Nature 2003 421 63 66 12511954
Turner PE Chao L Sex and the evolution of intrahost competition in RNA virus phi6 Genetics 1998 150 523 532 9755186
Sokal RR Rohlf FJ Biometry. 3rd edition 1995 San Francisco W. H. Freeman 887
Turner PE Chao L Prisoner's dilemma in an RNA virus Nature 1999 398 441 443 10201376
Chao L Fitness of RNA virus decreased by Muller's ratchet Nature 1990 348 454 455 2247152
Burch CL Chao L Evolution by small steps and rugged landscapes in the RNA virus phi 6 Genetics 1999 151 921 927 10049911
Burch CL Chao L Epistasis and its relationship to canalization in the RNA virus phi6 Genetics 2004 167 559 567 15238511
Gibson G Wagner G Canalization in evolutionary genetics: A stabilizing theory? Bioessays 2000 22 372 380 10723034
Littell RC Milliken GA Stroup WW Wolfinger RD SAS system for mixed models 1996 Cary (North Carolina) SAS Institute
Hartl DL Taubes CH Compensatory nearly neutral mutations: Selection without adaptation J Theoret Biol 1996 182 303 309 8944162
Chao L Rang CU Wong LE Distribution of spontaneous mutants and inferences about the replication mode of the RNA bacteriophage phi6 J Virol 2002 76 3276 3281 11884552
Lenski RE Rose MR Simpson SC Tadler SC Long-term experimental evolution in Escherichia coli . I. Adaptation and divergence during 2,000 generations Am Nat 1991 138 1315 1341
Travisano M Adaptation and divergence in experimental populations of the bacterium Escherichia coli The roles of environment, phylogeny and chance [dissertation] 1993 East Lansing (Michigan) Michigan State University 180 p. Available from ProQuest, Ann Arbor, Michigan; 9418079
Wagner A Redundant gene functions and natural selection J Evol Biol 1999 12 1 16
Mindich L Cohen J Weisburd M Isolation of nonsense suppressor mutants in Pseudomonas
J Bacteriol 1976 126 177 182 816771
Remold SK Lenski RE Pervasive joint influence of epistasis and plasticity on mutational effects in Escherichia coli
Nature Genet 2004 36 423 426 15072075
|
16248678
|
PMC1275523
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e381
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030381
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1624867910.1371/journal.pbio.0030384Research ArticleBioinformatics/Computational BiologyGenetics/Genomics/Gene TherapyMolecular Biology/Structural BiologySaccharomycesVariant Histone H2A.Z Is Globally Localized to the Promoters of Inactive Yeast Genes and Regulates Nucleosome Positioning H2A.Z Genome-Wide Binding MapGuillemette Benoît
1
Bataille Alain R
2
Gévry Nicolas
1
Adam Maryse
1
Blanchette Mathieu
3
Robert François [email protected]
2
Gaudreau Luc [email protected]
1
1 Centre de Recherche sur les Mécanismes du Fonctionnement Cellulaire, Département de Biologie, Faculté des Sciences, Université de Sherbrooke, Sherbrooke, Québec, Canada,2 Laboratoire de Chromatine et Expression du Génome, Institut de Recherches Cliniques de Montréal, Montréal, Québec, Canada,3 McGill Center for Bioinformatics, Lyman Duff Medical Building, Montréal, Québec, CanadaGroudine Mark Academic EditorFred Hutchinson Cancer Research CenterUnited States of America12 2005 1 11 2005 1 11 2005 3 12 e3842 8 2005 12 9 2005 Copyright: © 2005 Guillemette et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Preparing for Transcription: The Role of Histone H2A.Z
H2A.Z is an evolutionary conserved histone variant involved in transcriptional regulation, antisilencing, silencing, and genome stability. The mechanism(s) by which H2A.Z regulates these various biological functions remains poorly defined, in part due to the lack of knowledge regarding its physical location along chromosomes and the bearing it has in regulating chromatin structure. Here we mapped H2A.Z across the yeast genome at an approximately 300-bp resolution, using chromatin immunoprecipitation combined with tiling microarrays. We have identified 4,862 small regions—typically one or two nucleosomes wide—decorated with H2A.Z. Those “Z loci” are predominantly found within specific nucleosomes in the promoter of inactive genes all across the genome. Furthermore, we have shown that H2A.Z can regulate nucleosome positioning at the GAL1 promoter. Within HZAD domains, the regions where H2A.Z shows an antisilencing function, H2A.Z is localized in a wider pattern, suggesting that the variant histone regulates a silencing and transcriptional activation via different mechanisms. Our data suggest that the incorporation of H2A.Z into specific promoter-bound nucleosomes configures chromatin structure to poise genes for transcriptional activation. The relevance of these findings to higher eukaryotes is discussed.
A high-resolution ChIP-on-chip mapping of H2A.Z across the yeast genome suggests a relationship to transcriptional activity of the genes with which it is associated.
==== Body
Introduction
Eukaryotic chromosomes are packaged into a complex nucleoprotein structure termed chromatin, from which the fundamental unit is the nucleosome. Nucleosomes are composed of octamers of histone proteins designated as H2A, H2B, H3, and H4 that wrap approximately 146 bp of DNA [1,2]. Chromatin is generally highly repressive to cellular processes that involve DNA transactions such as gene transcription [3]. Cells have thus devised at least three schemes to overcome the nucleosomal barrier: the first involves adenosine triphosphate hydrolysis to mechanically displace nucleosomes [4,5], and the second consists of chemically modifying the tails of histones, such as by acetylation, phosphorylation, methylation, and ubiquitination [2]. This last scheme can take the form of a “code” where modifications act as signals for subsequent chromatin modifications [6,7]. A third scheme that has been devised by cells to overcome the nucleosomal barrier is to alter the composition of nucleosomes through the incorporation of histone variants that can—directly or indirectly—alter the permissiveness of chromatin to gene expression [8,9].
Nonallelic variants have been described for many classes of histones, and one of the best-studied examples is the Z variant of H2A. H2A.Z, formerly called H2A.F/Z, is evolutionarily conserved from yeast to mammals [10,11]. H2A.Z is essential for the viability of Tetrahymena, Drosophila, and mice [12–14] and constitutes approximately 10% of all cellular H2A molecules in S-phase [15,16]. Experiments carried out in many organisms have shown that H2A.Z has a function distinct from the canonical H2A [13,17,18]. H2A variant histones differ mainly in their N- and C-terminal portions as compared to S-phase H2A, whereas the core region is highly conserved [16]. In fact, experiments carried out in Drosophila and in yeast have revealed that the unique feature of H2A.Z important for its specific function resides in the carboxyl-terminal region of the variant histone and not in its histone fold [18,19].
In Tetrahymena thermophila, H2A.Z has been found to be associated with the transcriptionally active macronucleus, thus implying a positive role for H2A.Z in gene transcription [20]. Subsequently, experiments carried out in yeast have shown that H2A.Z could regulate transcription [19,21] and that its function is partially redundant with nucleosome remodeling complexes such as SAGA and Swi/Snf [21]. H2A.Z has been shown to be quickly remodeled soon after gene induction [19,21–23], a result that implies that the role of H2A.Z in positive gene transcription occurs at an early step. In line with this idea, we have shown that the transcriptional machinery is not efficiently recruited to the GAL1 promoter in htz1Δ cells [19]. Another mechanism by which H2A.Z could contribute to gene regulation has been suggested by Farris et al. [23], where they have shown at the c-myc gene that transcription-induced chromatin transactions involve the exchange of H2A.Z for canonical H2A.
In addition to its well-established role in transcriptional activation, a recent important piece of work by the Madhani group showed that H2A.Z positively regulates clusters of genes that are proximal to telomeres [24]. The gene expression defect observed at those gene clusters (called HZAD domains) in H2A.Z mutants can be suppressed by the deletion of silencing proteins such as Sir2. It has therefore been suggested that H2A.Z functions as an antisilencing factor by antagonizing the spread of heterochromatin to euchromatic regions [24]. The expression of other H2A.Z-affected genes, however, is not sensitive to Sir proteins suggesting H2A.Z positively regulates gene expression via both antisilencing and other mechanisms.
Evidence also suggests that H2A.Z can negatively regulate gene expression, notably through gene silencing. In Drosophila and in yeast, H2A.Z was reported to be involved in heterochromatic silencing [25,26]. In mammalian cells, H2A.Z colocalizes with, and directly interacts with, heterochromatin protein 1 or HP1-α (LocusLink:23468) and with INCENP (LocusLink:3619)—a protein involved in chromosome segregation—in pericentric heterochromatin [27,28]. In addition, HP1-α and H2A.Z may function together to establish a specialized conformation at constitutive heterochromatin [27]. Thus, from these findings it is clear that H2A.Z is important both for the positive and negative regulation of gene expression.
To add to all of these recent findings, there has been the codiscovery of an adenosine triphosphate-dependent chromatin remodeling complex that specifically loads H2A.Z into chromatin and exchanges it with H2A [29–31]. This complex, from which the catalytic subunit is Swr1, also shares essential subunits with the NuA4 histone acetyltransferase complex [29]. In addition to their importance in gene regulation, the Swr1 complex, H2A.Z, and NuA4 are all involved in the regulation of chromosome stability in yeast [32]. This is interesting and important since it was also demonstrated that in mammalian cells, knockdown of H2A.Z causes the genome to become unstable, a process that is attributed to defects in chromosome segregation [33]. Of equal importance is the fact that H2A.Z is required for early mouse and Xenopus development [13,14].
Despite all the work reviewed above, the mechanisms by which H2A.Z regulates transcription, silencing, genome stability, and developement remains largely unknown. A key piece of data that is missing is a global view of the distribution of H2A.Z across the genome. Polytene chromosome stainings have shown that Drosophila H2A.Z is widely distributed in the genome, but with nonuniform distribution that is characterized by a complex banded pattern [34]. From these data, it is thus conceivable that certain regions of the genome are preferentially occupied by variant histones and interspersed with regular histones. The resolution provided by this assay, however, is not high enough to allow one to correlate H2A.Z localization with its effect on gene expression. Experiments aimed at localizing proteins on defined chromatin segments in vivo—chromatin immunoprecipitation (ChIP)—have been performed at specific loci, but these experiments do not allow determining whether H2A.Z assembles preferentially within large chromosomal domains, whether it is randomly spread out across the genome, or whether it is associated with specific loci along chromosomes. Answering this question is a crucial step toward understanding the various functions of this important protein.
Here we make use of a genome-wide location assay (ChIP–chip) to map H2A.Z across the yeast genome with a 300-bp resolution. Our data show that H2A.Z preferentially and specifically associates to small regions within the promoters of inactive yeast genes. We also show that replacement of H2A.Z for regular H2A within the GAL1 promoter region perturbs nucleosome positioning, a situation that might account for the transcription defect at the latter gene in the absence of H2A.Z. Interestingly, the very strict promoter-specific association of H2A.Z that is otherwise observed elsewhere is not observed for genes contained within the HZAD domains. At those genes, the presence of H2A.Z can be detected within wider domains encompassing both the promoter region and part of the open reading frame (ORF).
Results
Genome-Wide Location Analysis of H2A.Z
In order to investigate the genomic distribution of H2A.Z in yeast, we performed genome-wide location experiments (also known as ChIP–chip) on whole-genome tiling microarrays. The assay consists of first performing ChIP assays to immunoprecipitate (IP) our target DNA molecules of interest. Thus, yeast cells are first cross-linked with formaldehyde prior to being lysed. Chromatin fragments (approximately 400 bp) are next obtained by sonication and specific antibodies are then used to IP our protein or proteins of interest along with their target DNA. Purified DNA is next amplified by ligation-mediated PCR (LM-PCR), labeled, and hybridizing to an array that contains a total of 44,290 60-mer probes (including controls) for an average of one probe every 266 bp (see Materials and Methods for more details). Thus, ChIPs were performed using yeast strains bearing Myc epitopes on H2A.Z and on the canonical H2A. In every case, a ChIP for histone H2B (performed using an HA-tagged H2B strain) was hybridized together with the H2A.Z (or H2A) ChIP to control for nucleosome density. In addition, other control experiments where two Myc-H2A.Z ChIPs are hybridized against each other were also performed. All experiments were performed in duplicate and combined using a weighted average method. A Gaussian filter was then applied to each dataset in order to create a smooth interpolation occupancy curve. The method allowed for the identification of 4,862 loci enriched in H2A.Z, referred to below as “Z loci.” The enrichment ratio for every probe on the microarray is available in Table S1, whereas the smoothed data can be found in Tables S2–S4. The coordinates of all Z loci are listed in Table S5. All datasets were deposited into ArrayExpress under accession number E-MEXP-396.
H2A.Z Is Associated with Small Genomic Regions Present All across the Yeast Genome
The 4,862 loci identified here are widely distributed across all chromosomes, and no regions show any bias (see Figure 1 for a map of Chromosome III and Table S6 to see the whole genome). Typical Z loci are less than 400 bp long, suggesting that the variant histone occupies regions that are generally no larger than two nucleosomes. This estimation is based on four observations. First, the probes enriched in the H2A.Z experiments were not depleted in the H2A experiments (data not shown). Because the average size of the DNA fragments used in our ChIP assays is 400 bp, this suggests that most 400-bp fragments within the genome that contains H2A.Z will also contain H2A. Since, in principle, a nucleosome cannot contain both H2A and H2A.Z simultaneously [35], the data suggest that H2A.Z occupies regions that are smaller than 400 bp. Second, the size of the Z loci as they appear from the microarray data (see Figure 1, for example) are consistent with H2A.Z occupying small regions—especially considering the fact that ChIP experiments necessarily overestimate the size of the region bound by proteins due to the size of the chromatin fragments generated. Third, high-resolution quantitative PCR (Q-PCR) analysis of ChIP experiments using small overlapping PCR probes allowed us to map the Z loci within the promoter of the SRB8 genes to an approximately 260-bp region using TFIIB as a reference [36] (Figure 1D; see Protocol S1 and Figure S1 for more details). Last, alignment of our data with global nucleosome positioning information from Yuan et al. [37] revealed that H2A.Z maps to one or two nucleosomes in promoters (see below). Taken together, these analyses show that H2A.Z occupies very small regions scattered across the genome, rather than concentrating within large chromosomal domains.
Figure 1 Genome-Wide Location Analysis of H2A.Z and H2A; a Zoom on Chromosome III
(A) The location of the Z loci along Chromosome III is depicted by gray bars.
(B) A zoom in a 120 KB region between position 196611–316611 shows the raw H2A.Z/H2B log2 ratios for each probe in that region (green bars) and the smoothed data resulted from the Gaussian plot analysis of the raw data for H2A.Z/H2B log2 ratios (red line). The position of the Z loci is shown by gray lines. The genes present in that regions are shown in blue.
(C) Same as in (B) but for a zoom in region 237700–277700 (40 KB).
(D) The size of the Z loci within the promoter of the SRB8 gene as determined by Q-PCR analysis of ChIP experiments. The binding ratios for H2A.Z (red) and TFIIB (blue) are shown relative to the center of the probe that generates the maximum enrichment. The data were smoothed by a sliding median over three probes (see Figure S1). The size of the observed H2A.Z domain is about 250 bp larger than that of TFIIB. Since TFIIB covers about 10 bp of DNA, we can infer that H2A.Z covers about 260 bp of DNA at that locus (shaded area). More details can be found in Protocol S1 and Figure S1.
H2A.Z-Containing Nucleosomes Are Essentially Located within Promoter Regions
We next examined the actual location of the Z loci relative to genes. Manual inspection of the data suggests that Z loci are predominantly located within promoter regions. This feature is illustrated in Figure 1, that shows (i) only one ORF within the 40-KB window shown in Figure 1C contains enrichment for H2A.Z; (ii) none of the seven intergenic regions that do not contain any promoter in that region (for example, between SOL2 and ERS1) contain a Z locus; (iii) there is a Z locus upstream of every ORF shown in Figure 1C; and (iv) interestingly, the intergenic region between the divergently transcribed YCR090C and KIN82 ORFs contains two separate Z loci, supporting the idea that each promoter is decorated with its own H2A.Z nucleosome(s). In order to systematically evaluate the H2A.Z preference for promoter regions, we analyzed the relation between the H2A.Z/H2B ratio and the distance to the closest 5′ gene boundary (for example, the ATG for protein-coding genes) for all probes on the microarray. As shown in Figure 2A, H2A.Z binding (green line) is higher around the 5′ boundary with a small bias toward the regions located just upstream compared to downstream (see inlet for a zoom into the 5′ region). This relationship is not observed on random data (Figure 2A, black line) or when the proximity to the 3′ end of genes is examined (not shown). Because 5′ UTRs are very small in yeast, these data are consistent with H2A.Z being localized over promoters, although we cannot rule out the possibility that H2A.Z is rather present within the 5′ UTR since only a few transcriptional start sites have been annotated in yeast. For the sake of simplicity the word “promoter” will be used here to refer to the regions localized between 500 bp upstream and 100 bp downstream of the 5′ boundary of transcribed elements (this includes ORFs, but also other transcribed elements such as Ty elements, tRNAs, etc.). This interval was chosen based on a recent genome-wide analysis of yeast promoters [38]. As shown in Figure 2B, 74% (3,639 out of 4,862) of the Z loci identified in this study are located within the promoter of annotated genes. Considering that promoters account for about 31% of the yeast genome, this translates into a 2.4-fold enrichment over a random uniform distribution (p < 0.001). In addition, we find that 63% (4,742 out of 7,488) of yeast promoters are decorated with H2A.Z (Figure 2C). Conversely, only 2% of the H2A.Z loci are located within intergenic regions that contain no promoters (where two genes are transcribed convergently).
Figure 2 H2A.Z Is Predominantly Localized within Promoter Regions
(A) A sliding median of the H2A.Z/H2B ratio for all probes on the microarray is plotted against the distance from the 5′ boundary of their closest transcribed element (including ORF, tRNA, transposon, etc.) (green). Randomized data (where the H2A.Z/H2B ratios were scrambled prior to calculating the sliding median) are plotted in black.
(B) Most Z loci are found within promoters. The fraction of the 4,862 Z loci present within promoters (gold) and nonpromoter region (aqua) is shown. Promoter regions are defined as the −500/+100 interval relative to the 5′ boundary of transcribed elements as annotated in the SGD database.
(C) The fraction of the 7,571 promoter regions (defined as in [B]) containing (green) or not containing (purple) a Z loci is shown.
(D) H2A.Z is preferentially localized downstream of the NFR in yeast promoters. A sliding median of the ratio for H2A.Z/H2B (green) and H2A/H2B (gold) was plotted against the distance from the NFR (defined as a linker of more than 100 bp located less than 500 bp upstream of a 5′ gene boundary) for regions where the NFR can be unambiguously assigned to a single gene (i.e., NFR that map to two diverging genes were filtered out). A similar sliding window was applied to the nucleosome positioning data from Yuan et al. [37] (blue). A cartoon representation of the nucleosomal organization of a typical gene is shown at the bottom.
In a recent elegant study, the Rando group mapped nucleosomes over 482 KB of yeast DNA, including most of the entire Chromosome III. They showed that the scored yeast promoters contain a nucleosome-free region (NFR) flanked with well-positioned nucleosomes [37]. A median sliding window of their data is reproduced in Figure 2D (blue) along with our genome-wide location data with H2A.Z/H2B (green) and H2A/H2B (gold) data for a composite of 1,850 genes [37]. Despite both assays having different resolution, the figure clearly shows that H2A.Z is preferentially localized downstream of the NFR in yeast promoters. More important, data from the figure suggest that H2A.Z may occupy the first, and perhaps also the second, well-positioned nucleosome downstream of the NFR (see cartoon in Figure 2D). Taken together, these data suggest that H2A.Z targets specific nucleosomes within promoters across the genome and therefore contributes to create a promoter-specific chromatin structure.
In addition to class II promoter regions, we found H2A.Z loci at other transcribed units such as tRNAs, small nucleolar RNAs, rDNAs, and Ty elements, suggesting that the role of H2A.Z in transcription is not limited to protein-coding genes (data not shown). Interestingly, we also found H2A.Z loci centered around centromeres and replication origins (data not shown). The association of H2A.Z to centromeres was also reported previously [32], and the role of H2A.Z in DNA replication and chromosomal segregation will be studied elsewhere (Bataille et al., unpublished data).
H2A.Z Is Not Enriched in the Promoter Region of Highly Active Genes
Data from Figures 1 and 2 have clearly shown that H2A.Z has a preference for promoter regions. However, Figure 2B shows that up to 37% of promoters are not associated with a Z locus. In order to investigate what determines the association of H2A.Z with some promoters but not others, we investigated the correlation between H2A.Z promoter occupancy and the transcription rate for all genes. As shown in Figure 3, there is an inverse correlation between the transcription rate and H2A.Z occupancy. Accordingly, when the genome-wide location of H2A.Z was determined for cells grown in galactose, we noticed that the occupancy of H2A.Z was reduced, compared to its occupancy in glucose, at the promoters of 147 out of 173 genes whose expression levels have been demonstrated to be induced the most under those conditions according to the study of Ren et al. [39]. These data are consistent with H2A.Z being dynamically associated with inactive genes. These genomic observations are also in line with our ChIP/Q-PCR data (see Protocol S2 and Figure S2) and with previously published evidence at specific promoters [19,21,22]. We would like to point out that data represented in the experiment of Figure 3 were collected from a population of cells that are heterogeneous in the sense that each cell is in a different physiological state and phase of the cell cycle. For that reason, and also because promoters have inherently different strengths, a given gene is never completely off or on within the population. The observed transcription rate across all genes is therefore a continuum ranging from near zero to about 300 mRNA per hour for genes that are the most transcriptionally active. The transcription rate being a continuum, it is expected that the H2A.Z occupancy will also be a continuum. In that regard, RNA pol II or general transcription factors like TFIIB also show continuum-like binding patterns in those assays [36,40,41]. We also propose that the low enrichment of H2A.Z from the promoter of highly active genes most likely reflects the low abundance of nucleosomes in those regions since recent evidence shows that the promoters of highly active genes are depleted of nucleosomes [42–46]. Collectively, these data show that H2A.Z specifically associates with one or a few nucleosomes within the promoter of some inactive genes but is generally absent from the promoter of highly active genes.
Figure 3 H2A.Z/H2B Ratios Are Inversely Correlated with Transcription Rate
The binding trend was calculated by computing the moving median of the H2A.Z/H2B flat ratios over a sliding window of 50 genes across all genes ordered by transcription rate as described previously [36,40,41,60–62]. Each probe was assigned to its closest ATG, and the probe with the highest H2A.Z/H2B value for each gene was used. The transcription rate for yeast genes was determined previously [63].
H2A.Z Regulates Nucleosome Positioning
Having established that H2A.Z is present at most yeast promoters, we sought to investigate the functional significance of the promoter-specific localization of H2A.Z. First, we investigated nucleosome positioning of genes that have Z loci (Figure 4A) versus genes that are devoid of those loci (Figure 4B) as previously defined in Figure 2, using data presented in Yuan et al. [37]. The figure clearly shows that nucleosomes within the genes that bear Z loci are well positioned, whereas nucleosomes positioning at genes that do not have Z loci are significantly less well organized. This observation is consistent with the fact that H2A.Z-free promoters are less covered by nucleosomes than the more frequent H2A.Z-containing promoters (not shown). The figure further suggests that H2A.Z may play an important role in organizing chromatin structure at promoters. Alternatively, however, H2A.Z may associate with promoters as a consequence of their well-organized chromatin structure. In order to test directly whether H2A.Z can regulate chromatin structure at promoters, we studied nucleosome positioning at a model gene in wild-type (WT) and H2A.Z-null (htz1Δ) cells. The GAL1 gene was chosen for these experiments, since H2A.Z is required for its full expression and since the variant histone preferentially binds the repressed promoter of that gene [19,21]. Furthermore, chromatin structure at the GAL1 promoter has been thoroughly investigated [47–50], so any changes in chromatin structure should be experimentally tractable. First, we performed indirect end-labeling experiments on chromatin samples that were partially digested by MNase. Figure 4C, left panel, shows that the gross nucleosome positioning pattern is not significantly altered in htz1Δ cells when the latter are grown in media containing glucose. However, we do note a slight shift of nucleosome C (nucC) in a downstream position as well as a few other nucleosomes. The right panel of that figure shows the cleavage pattern of MNase-digested naked GAL1 promoter DNA. The UASG region can be easily identified as it is quite resistant to MNase cleavage [49]. Since indirect end-labeling experiments only have a resolution of approximately 20–30 bp, we decided to carry out high-resolution LM-PCR assays of MNase-digested mononucleosomes at two positioned nucleosomes within the GAL1 promoter region (nucB and nucC; see Figure 4D). Thus, LM-PCR linkers were ligated to mononucleosome-purified DNA, and PCR reactions were carried out with external oligonucleotides corresponding to the linkers in combination with oligonucleotides corresponding approximately to the center of the nucleosome-protected DNA as previously determined by Fisher-Adams and Grunstein [49]. These last oligonucleotides are designed to either amplify the right-end portion of nucleosome-embedded DNA (R) of nucB and nucC, or the left-end portion (L) (Figure 4D, above). The results show that nucB does not shift position in the htz1Δ mutant as compared to WT cells when PCR reactions were performed at either the right or left boundaries of the nucB. However, in the case of nucC, there is an approximately 20-bp shift to a downstream position in htz1Δ cells as compared to the WT case (Figure 4D, right panel). More important, this shift was observed with both the right and left boundaries of nucC. These results confirm our observations in indirect end-labeling experiments and suggest that the absence of H2A.Z induces a specific shift of nucC to a downstream position but has no effect on the positioning of nucB. Our results suggest that H2A.Z may be enriched at the promoters of inactive genes in order to modulate the position of key regulatory promoter-associated nucleosomes. Because we found that most Z loci are located within well-positioned nucleosomes and because H2A.Z-free promoters have a less organized chromatin structure than H2A.Z-containing genes, it is tempting to speculate that the variant histone also plays a role in nucleosome positioning at other genes.
Figure 4 H2A.Z Regulates Nucleosome Positioning
(A) Nucleosome positioning map of genes associated with a Z locus. The data from Yuan et al. [37] were used to compute the nucleosome occupancy curve for all genes containing a Z locus aligned on their ATG. Peaks represent nucleosomes, and valleys represent linker regions. An NFR is detected approximately 200 bp upstream of ATGs as described by Yuan et al. [37]. The vertical thickness of the curves contains 1-SD error bars for the mean log2 ratio.
(B) Same as (A) but for genes containing no Z locus.
(C) Indirect end-labeling of MNase-digested chromatin from WT and htz1Δ cells grown in the presence of glucose.
(D) High-resolution LM-PCR analysis of MNase digested nucB–C mononucleosomes. Upper part: structure of the GAL1 promoter and PCR probes used; left: nucB analysis probing the right (R), and left (L) boundaries in WT and htz1Δ (Δ) cells; right: same as left part of the figure, but the analysis is with nucC.
H2A.Z Is Unusually Localized at HZAD Genes
Because genes contained within the so-called HZAD domains—unlike other H2A.Z regulated genes—are proposed to be regulated by H2A.Z through an antisilencing mechanism [24], we investigated more closely the distribution of H2A.Z at those genes. First, manual inspection suggested that H2A.Z occupied larger regions at those genes. As shown in Figure 5A, the regions occupied by H2A.Z around the GIT1 gene (an HZAD gene) extend both upstream as well as downstream within the coding region as compared to a non-HZAD gene such as SRB8 (see Figure 5B). Similar examples can be found in Figure S3. In order to investigate this in an unbiased manner, we have calculated the size of the regions localized between the coordinates where the H2A.Z occupancy curve approaches baseline (log2 ratio = 0.1). This size provides an estimation of the width of the Z loci. Figure 5C shows a plot of that size relative to the distance to chromosome ends. The figure shows that H2A.Z covers larger uninterrupted regions within the first 20 KB from telomeres compared to telomere distal regions. This is consistent with H2A.Z occupying wider regions around HZAD genes than at other genes in the genome. This conclusion was also reached using a different analysis described in Figure S4 and Protocol S3 and is not dependent on a difference in gene density at telomeres (not shown). This phenomenon is not restricted to HZAD genes since many telomere-proximal non-HZAD genes also show the same pattern (not shown). It is therefore possible that the antisilencing function observed at HZAD genes could actually be extended to more (perhaps most) telomere-proximal genes. Three possibilities could account for this wider distribution. First, H2A.Z could occupy more nucleosomes at those genes. Second, the number of nucleosomes occupied by H2A.Z may be the same, but these nucleosomes may not always be the same in different cells. Finally, it is also possible that chromatin at those genes is more fluid such that the H2A.Z nucleosomes can be found anywhere along a long stretch of DNA.
Figure 5 H2A.Z Is Unusually Localized in HZAD Genes
(A) The raw (green) and smoothed (red) H2A.Z/H2B log2 ratios, as determined by our microarray experiment, are shown across a 5-KB region around the GIT1 genes. The genes present in that region are shown in blue.
(B) Same as (A) but for a 5-KB region around the SRB8 gene.
(C) H2A.Z covers wider regions near telomeres. The size of the regions covered by H2A.Z (as determined by distance between the coordinates where the smoothed H2A.Z/H2B log2 ratio reaches “0.1”) is plotted against the distance to telomeres. A moving average (window = 10 KB) was applied to the data.
Discussion
We have investigated the genome-wide distribution of histone variant H2A.Z as compared to the regular S-phase histone H2A. Our results show that H2A.Z can preferentially associate to narrow regions of yeast promoters. Surprisingly, the variant histone is present within most nucleosome-containing promoters. Conversely, the promoters of highly transcribed genes are not predominantly enriched in H2A.Z. Furthermore, and much to our surprise, H2A.Z does not appear to be as sharply localized within HZAD genes. The wider distribution of H2A.Z at HZAD genes may be important for preventing Sir protein spreading at those regions. Since H2A.Z shows a strong preference for inactive gene promoter regions, we also investigated whether the variant histone had an important role in modulating chromatin structure at the GAL1 model gene. More important, we find that the replacement of H2A.Z for regular H2A in cells induces a shift in the positioning of a nucleosome located over the transcription start site. In accordance with this finding is the fact that nucleosomes at promoter regions bearing Z loci are well positioned as compared to genes not enriched in H2A.Z. Taken together, our results support a model where H2A.Z associates to specific nucleosomes in the promoters of inactive genes in order to poise, and perhaps organize, chromatin structure in a fashion that is permissive to transcription initiation. However, it remains to be determined whether the simple incorporation of H2A.Z into specific nucleosomes is by itself necessary and sufficient to facilitate gene activity or whether additional factors are required for that action. The first possibility is supported by the fact that in vitro-assembled chromatin arrays containing H2A.Z are inherently more resistant to condensation as compared to H2A-containing nucleosomal arrays [51,52]. On the other hand, work from our laboratory has shown that H2A.Z can interact with regulatory proteins and components of the transcriptional machinery [19], arguing that H2A.Z may regulate gene expression through interactions with such downstream effectors.
Although H2A.Z has been previously shown to be enriched at the GAL1, PHO5, and PUR5 genes [21,22], the fact that H2A.Z globally associates to defined promoter regions of inactive genes is surprising since previous studies had suggested that even though H2A.Z was generally enriched in euchromatic regions, there did not appear to be any correlation between deposition of the variant histone and promoters [24,30]. However, we note that in those cases where H2A.Z was found more enriched in the ORF as compared to the promoter, the genes were highly transcribed housekeeping genes (e.g.,ACT1 and ADH1) [24,30]. Recent studies have clearly established that the promoters of such highly active genes are rather depleted of nucleosomes [42–46], and thus the apparent low enrichment of H2A.Z at those promoters could actually reflect low nucleosome occupancy. In support of this conclusion, we find that H2B levels are significantly reduced at the ACT1 promoter, while the H2A.Z/H2B ratio shows a slight enrichment in the promoter region (see Figure S2C). Because housekeeping genes are by definition transcribed in most, if not all, conditions, and since we found that H2A.Z is present at promoters of genes when they are inactive, we favor a model where H2A.Z would be more important for the regulation of nonhousekeeping genes that are generally repressed when grown in rich medium but are strongly induced under specific growth conditions (e.g.,GAL genes upon galactose induction, etc.). It is interesting to note that this last possibility is in line with published evidence regarding the role of H2A.Z in higher organisms. For example, H2A.Z is not significantly expressed in the nondifferentiated inner cell mass tissue of the mouse embryo, whereas it is highly expressed in the differentiating tissues [28]. Furthermore, H2A.Z is required for the proper development of Xenopus [13]. These observations may be taken to suggest that H2A.Z is only required when a particular gene expression program is engaged in an organism and not so much for gene expression that pertains to basic cellular metabolisms. Since H2A.Z has been shown to be abundantly associated with heterochromatin in higher eukaryotes [28]—which is much more preponderant in metazoans as compared to yeast [53,54]—it remains to be determined whether or not the variant histone plays additional functions linked to this localization in metazoans. It is worth emphasizing, however, that this colocalization of H2A.Z with metazoan heterochromatin does not exclude the possibility that it may also be associated with the promoter regions of inactive genes as we have shown with yeast. This possibility is currently under investigation.
Our genome-wide location assay has revealed that—as opposed to what is observed at other genes—H2A.Z occupies wider regions that spread out well beyond the promoter region and into the ORF, at HZAD genes. As previously suggested by Meneghini et al. [24], it is conceivable that H2A.Z has a dual role in gene expression, the first being that of poising genes for transcription initiation in euchromatin, and the second being that of protecting euchromatin from silencing. We propose that at most inactive genes where H2A.Z has a sharp promoter-binding pattern, H2A.Z will simply be incorporated into specific repressive nucleosomes that need to be remodeled quickly prior to gene induction, whereas at HZAD genes, H2A.Z would create nucleated domains that are refractory to gene silencing by heterochromatin. Alternatively to this scenario, it is possible that the broad distribution of H2A.Z at HZAD genes reflects unstable positioning of variant histone-bearing nucleosomes, perhaps due to a chromatin structure that is less well organized at those genes. This last possibility, however, is less likely since examination of the nucleosome positioning data from Yuan et al. [37] reveals that nucleosomes are well positioned over the Chromosome III-contained HZAD genes. Nevertheless, it is likely that the specific chromatin structure exemplified by this unusual localization of H2A.Z renders the HZAD genes more sensitive to Sir2 silencing.
Our observations that H2A.Z is also important in regulating nucleosome positioning, in particular at the GAL1 promoter, may account for a mechanism by which H2A.Z regulates transcription. We had previously observed that absence of H2A.Z prevents RNA polymerase II and TBP from being efficiently recruited to the GAL1 promoter, whereas the Gal4 activator was efficiently recruited to the gene [19]. In addition to these findings, recent experiments from our laboratory have shown that Mediator, SAGA, and Swi/Snf also cannot be efficiently recruited to GAL1 in the absence of the variant histone (K. Lemieux and L. Gaudreau, unpublished data). We thus propose that improper positioning of at least one nucleosome over the GAL1 transcription start site may be sufficient to severely impair the recruitment of the entire transcriptional machinery to the gene. Our results also emphasize the fact that chromatin architecture is exquisitely regulated, and even subtle changes in its structure may trigger transcription defects. A similar situation was recently reported with bromodomain factor 1 (Bdf1) and histone H4-tail acetylation at the PHO5 gene. It was reported that a shift of 2–3 bp of a positioned nucleosome at a PHO5-lacZ reporter construct resulted in a defect in PHO5 induction in the absence of Bdf1 and H4-acetylation [55]. This is of particular significance since Bdf1 is a component of Swr1.com that loads H2A.Z into chromatin [29,30]. It is conceivable that deletion of BDF1 prevents loading of H2A.Z into chromatin, and therefore the Bdf1-related defect would be owed to improper loading of H2A.Z at the PHO5 promoter. Alternatively, it is conceivable that Bdf1 could be a downstream target of H2A.Z and important in mediating its function in nucleosome positioning. These possibilities are currently under investigation. Finally, we point out that our data do not exclude the possibility that H2A.Z could also be important in transcription elongation since the latter histone variant has been shown to interact genetically with the TFIIS elongation factor in yeast [30] and since it appears to be replaced by H2A downstream of the c-myc promoter region during elongation [23].
Materials and Methods
Yeast strains and genetic methods
Saccharomyces cerevisiae W303a strains carrying 3xMyc-H2A.Z, 3xMyc-H2A or 3xHA-H2B were generated using PCR epitope tagging as described previously [56]. All URA3 markers were “popped-out” and strains were verified by PCR and immunoblotting analyses. MAY424 (htz1Δ) was described previously [19].
Antibodies
Antibodies in this study were from the following source: anti-Myc, (9E10; Santa Cruz Biotechnology, Santa Cruz, California, United States); anti-HA (12CA5; Roche, Basel, Switzerland); and anti-TFIIB, a kind gift from Richard A. Young.
ChIPs
ChIP experiments were performed essentially as described previously [19,22] with some modifications. Briefly, 50 ml of cells were grown in yeast extract-peptone supplemented with 2% dextrose, or 2% raffinose/5% galactose for GAL-induced conditions, to an OD600 of 0.7 and fixed with 1% formaldehyde as described [19]. Next, 500 μl of whole-cell extract was incubated with the appropriate antibody coupled to magnetic beads (Dynal Biotech, Brown Deer, Wisconsin, United States). Immunoprecipitated DNA was either used for genome-wide location analysis (see below) or Q-PCR analysis. Briefly, immunoprecipitated DNA and input DNA were mixed with the appropriate oligonucleotides and Brilliant SYBR Green Q-PCR Master Mix (Stratagene) in 20 μl final volume. Q-PCR was performed on an Mx3000P Real-Time PCR System (Stratagene, La Jolla, California, United States). The sequences of the oligonucleotides used for all of the Q-PCR experiments are available in Table S7.
DNA microarrays
The microarrays used for location analysis were purchased from Agilent Technologies (Palo Alto, California, United States). The microarrays contain a total of 44,290 60-mer probes (including 2,306 controls), covering the entire genome (except for repetitive regions) for an average density of one probe every 270 bp (±100 bp) within the probed regions. The greatest distance between the end of one probe and the beginning of the next is 540 bp, and 97% of the probes are within 350 bp of another one. The probes were mapped to the genome (UCSC) using BLAST. More than 99.9% of the probes mapped to a single locus, whereas 370 probes (<0.1%) had multiple BLAST outputs and were mapped accordingly. These microarrays were previously used to map histone modifications [57].
DNA labeling and hybridization
Labeling was done as described before [58]. Briefly, the immunoprecipitated DNA fragments were blunted with T4 DNA polymerase and ligated to unidirectional linkers. The DNA was then amplified by LM-PCR in the presence of aminoallyl-modified dUTP. The labeling was carried out post-PCR using monoreactive Cy-dye NHS esters that react specifically with the aminoallyl-modified dUTP. Hybridization was performed as described in Pokholok et al. [57], using salmon sperm in place of herring sperm DNA. Detailed hybridization and labeling protocols can be found at http://www.ircm.qc.ca/microsites/francoisrobert/en.
Data analysis
The data were normalized and replicates were combinined using a weighted average method as described previously [39]. To interpolate between probes and identify Z loci, a standard Gaussian filter (SD = 200 bp) was applied twice to the data. Z loci were defined as the maxima of this smoothed occupancy curve (excluding maxima smaller than zero).
Chromatin analysis
S. cerevisiae nuclei were prepared as previously described [59]. Nuclei were washed once in digestion buffer (15 mM Tris-HCl [pH 8.0], 50 mM NaCl, and 1.4 mM CaCl2) and resuspended in 1.2 ml digestion buffer per gram of dry weight.
For indirect end-labeling experiments, resuspended nuclei were digested with either 0, 1, 3, 5, 25, or 50 U/ml of MNase (USB, Cleveland, United States). DNA was prepared as described [59]. Southern blot analysis was conducted using an EcoRI-BanII fragment from the GAL1–10 region as a probe.
For LM-PCR nucleosome boundary mapping, 200-μl aliquots of resuspended nuclei were digested with a final concentration of 100 U/ml of MNase (USB) for 10 min at 37 °C and stopped with 4 mM EDTA. DNA was then prepared as described [59]. MNase-digested DNA was ligated overnight to unidirectional linkers, and mononucleosome-length fragments were then purified by agarose gel electrophoresis. PCR reactions were conducted with linker-specific and nucleosome-specific oligonucleotides, using JumpStart Taq DNA Polymerase (Sigma, St. Louis, Missouri, United States), using the manufacturer's recommendations. Oligonucleotide sequences are available in Table S8.
Supporting Information
Figure S1 High-Resolution Mapping of H2A.Z at the SRB8 Locus by Q-PCR
(A) Theory for domain size calculation using a control protein. A protein for which previous literature provides the actual size of the domain occupied by the protein on DNA (e.g., TFIIB) can be used to estimate the actual domain of an unknown protein (e.g., H2A.Z). Data from the control protein are used to estimate the amount of DNA that is added due to the limit of the resolution of the ChIP assay. This “resolution addition” is removed from the “observed” domain of the tested protein to generate its “actual” domain.
(B) Probes used in our Q-PCR assay. The regions amplified by PCR to probe our ChIP assay are depicted in red.
(962 KB TIF).
Click here for additional data file.
Figure S2 Q-PCR Analysis of H2A.Z ChIP Experiments
(A) The locus including the PRS2, UBC6, AST2, and RPS8B genes is drawn to scale, and the positions of the PCR probes used to monitor H2A.Z occupancy are shown as black bars. The H2A.Z/H2B ratios are shown by gray bars.
(B) Same as in (A) but for the GIT1 locus, a gene contained within an HZAD domain.
(C) Same as in (A) and (B) but for the ACT1 locus. The H2B/input ratio is also shown as white bars. In each panel, the probe with the lowest ratio was set to “1.”
(763 KB TIF).
Click here for additional data file.
Figure S3 H2A.Z Occupancy at HZAD10
(A) The raw (green) and smoothed (red) H2A.Z/H2B ratios are shown across a 15-KB region from Chromosome IX containing HZAD10 (gray shaded box). The genes in that region are shown in blue.
(B) Same as in (A) but for a non-HZAD region from Chromosome III.
(749 KB TIF).
Click here for additional data file.
Figure S4 Z Loci Are Wider around Telomeres
(A) Graphical representation of the width index calculation.
In order to look at the spread of the H2A.Z domains across chromosomes we have calculated an index estimating the relative width of these domains. The width index (i) is defined as x/y, where x is the width observed at ½ y, y being the height at the maximum of the H2A.Z smooth occupancy curve. The width is divided by the height in order to normalize for the fact that higher peaks will necessarily take more space to reach baseline and will therefore look like wider domains than would small peaks.
In order to avoid “contamination” by overlapping peaks, maxima localized within less than 1.5 KB from one another were not considered in the analysis.
(B) The width trend was determined by computing a sliding median of the width index (see A) across all yeast genes ordered by their distance from the chromosomal end and plotted against that distance.
(265 KB TIF).
Click here for additional data file.
Protocol S1 Determination of the Size of the Z Locus within the SRB8 Promoter
(27 KB DOC).
Click here for additional data file.
Protocol S2 Validation of ChIP–Chip Data by Q-PCR
(26 KB DOC).
Click here for additional data file.
Protocol S3 Z loci Are Wider around Telomeres
(24 KB DOC).
Click here for additional data file.
Table S1 Raw Data for the H2A.Z/H2B, H2A/H2B, and H2A.Z/H2A.Z ChIP–Chip Experiments
For each experiment, the normalized enrichment ratio is shown for every probe on the microarray.
(4.3 MB ZIP).
Click here for additional data file.
Table S2 Smoothed Data for the H2A.Z/H2B ChIP–Chip Experiment
For every coordinate within the genome, the H2A.Z/H2B ratio derived from the raw data by the Gaussian plot analysis is shown. This file is a text file ready for display within the UCSC Genome Browser.
(10.8 MB ZIP).
Click here for additional data file.
Table S3 Smoothed Data for the H2A/H2B ChIP–Chip Experiment
For every coordinate within the genome, the H2A/H2B ratio derived from the raw data by the Gaussian plot analysis is shown. This file is a text file ready for display within the UCSC Genome Browser.
(10.5 MB ZIP).
Click here for additional data file.
Table S4 Smoothed Data for the H2A.Z/H2A.Z ChIP–Chip Experiment
For every coordinate within the genome, the H2A.Z/H2A.Z ratio derived from the raw data by the Gaussian plot analysis is shown. This file is a text file ready for display within the UCSC Genome Browser.
(10.6 MB ZIP).
Click here for additional data file.
Table S5 List of Z Loci
A table listing the position of each Z loci identified in this study.
(340 KB XLS).
Click here for additional data file.
Table S6 Genome Browser-Ready File
A BED file ready for displaying the H2A.Z/H2B raw (green) and smoothed (red) data for the entire yeast genome as in Figure 1.
(11.2 MB ZIP).
Click here for additional data file.
Table S7 Sequences of the Oligonucleotides Used to Quantify ChIPs by PCR in This Study
(28 KB XLS).
Click here for additional data file.
Table S8 Sequences of the Oligonucleotides Used in the LM-PCR Assay to Map Nucleosomes within the GAL1 Promoter
(18 KB XLS).
Click here for additional data file.
Accession Numbers
The Saccharomyces Genome Database (http://www.yeastgenome.org/) accession numbers for the genes and proteins discussed in this paper are ACT1 (SGDID S000001855), ADH1 (SGDID S000005446), Bdf1 (SGDID S000004391), ERS1 (SGDID S000000671), GAL1 (SGDID S000000224), Gal4 (SGDID S000006169), GIT1 (SGDID S000000695), KIN82 (SGDID S000000687), PHO5 (SGDID S000000297), PUR5 (SGDID S000001259), Sir2 (SGDID S000002200), SOL2 (SGDID S000000718), SRB8 (SGDID S000000677), Swr1 (SGDID S000002742), TBP (SGDID S000000950), TFIIB (SGDID S000006290), and YCR090C (SGDID S000000686).
We thank Alain Lavigueur and Benoît Leblanc for critical comments on the manuscript and Svetlana Sadekova for sharing equipment. This work was supported by grants from the Canadian Institutes of Health Research (CIHR) awarded to LG and FR and from Genome Québec to MB. LG holds a Canada research chair on mechanisms of gene transcription, and FR holds a new investigator award from the CIHR. BG is a recipient of a doctoral studentship from Natural Sciences and Engineering Research Council (NSERC), ARB is a recipient of a doctoral studentship from the Institut de recherches cliniques de Montréal/CIHR cancer research program, and NG is a recipient of a postdoctoral fellowship from NSERC.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. FR and LG conceived and designed the experiments. BG, ARB, NG, and MA performed the experiments. ARB and MB analyzed the data. FR and LG wrote the paper.
Citation: Guillemette B, Bataille AR, Gévry N, Adam M, Blanchette M, et al. (2005) Variant histone H2A.Z is globally localized to the promoters of inactive yeast genes and regulates nucleosome positioning. PLoS Biol 3(12): e384.
Abbreviations
Bdf1bromodomain factor 1
ChIPchromatin immunoprecipitation
ChIP–chipchromatin immunoprecipitation coupled with tiling microarray analysis
IPimmunoprecipitate
LM-PCRligation-mediated PCR
nucCnucleosome C
NFRnucleosome-free region
ORFopen reading frame
Q-PCRquantitative PCR
==== Refs
References
Khorasanizadeh S The nucleosome: From genomic organization to genomic regulation Cell 2004 116 259 272 14744436
Vaquero A Loyola A Reinberg D The constantly changing face of chromatin Sci Aging Knowledge Environ 2003 2003 RE4 12844523
Felsenfeld G Groudine M Controlling the double helix Nature 2003 421 448 453 12540921
Flaus A Owen-Hughes T Mechanisms for ATP-dependent chromatin remodelling: Farewell to the tuna-can octamer? Curr Opin Genet Dev 2004 14 165 173 15196463
Lusser A Kadonaga JT Chromatin remodeling by ATP-dependent molecular machines Bioessays 2003 25 1192 1200 14635254
Jenuwein T Allis CD Translating the histone code Science 2001 293 1074 1080 11498575
Turner BM Cellular memory and the histone code Cell 2002 111 285 291 12419240
Kamakaka RT Biggins S Histone variants: Deviants? Genes Dev 2005 19 295 316 15687254
Henikoff S Ahmad K Assembly of variant histones into chromatin Annu Rev Cell Dev Biol 2005 In press
Thatcher TH Gorovsky MA Phylogenetic analysis of the core histones H2A, H2B, H3, and H4 Nucleic Acids Res 1994 22 174 179 8121801
Wu RS Nishioka D Bonner WM Differential conservation of histone 2A variants between mammals and sea urchins J Cell Biol 1982 93 426 431 7096447
Liu X Li B Gorovsky MA Essential and nonessential histone H2A variants in Tetrahymena thermophila
Mol Cell Biol 1996 16 4305 4311 8754831
Ridgway P Brown KD Rangasamy D Svensson U Tremethick DJ Unique residues on the H2A.Z containing nucleosome surface are important for Xenopus laevis development J Biol Chem 2004 279 43815 43820 15299007
Faast R Thonglairoam V Schulz TC Beall J Wells JR Histone variant H2A.Z is required for early mammalian development Curr Biol 2001 11 1183 1187 11516949
Wu RS Bonner WM Separation of basal histone synthesis from S-phase histone synthesis in dividing cells Cell 1981 27 321 330 7199388
Redon C Pilch D Rogakou E Sedelnikova O Newrock K Histone H2A variants H2AX and H2AZ Curr Opin Genet Dev 2002 12 162 169 11893489
Jackson JD Gorovsky MA Histone H2A.Z has a conserved function that is distinct from that of the major H2A sequence variants Nucleic Acids Res 2000 28 3811 3816 11000274
Clarkson MJ Wells JR Gibson F Saint R Tremethick DJ Regions of variant histone His2AvD required for Drosophila development Nature 1999 399 694 697 10385122
Adam M Robert F Larochelle M Gaudreau L H2A.Z is required for global chromatin integrity and for recruitment of RNA polymerase II under specific conditions Mol Cell Biol 2001 21 6270 6279 11509669
Allis CD Glover CV Bowen JK Gorovsky MA Histone variants specific to the transcriptionally active, amitotically dividing macronucleus of the unicellular eucaryote, Tetrahymena thermophila
Cell 1980 20 609 617 7418000
Santisteban MS Kalashnikova T Smith MM Histone H2A.Z regulates transcription and is partially redundant with nucleosome remodeling complexes Cell 2000 103 411 422 11081628
Larochelle M Gaudreau L H2A.Z has a function reminiscent of an activator required for preferential binding to intergenic DNA EMBO J 2003 22 4512 4522 12941702
Farris SD Rubio ED Moon JJ Gombert WM Nelson BH Transcription-induced chromatin remodeling at the c-myc gene involves the local exchange of histone H2A.Z J Biol Chem 2005 280 25298 25303 15878876
Meneghini MD Wu M Madhani HD Conserved histone variant H2A.Z protects euchromatin from the ectopic spread of silent heterochromatin Cell 2003 112 725 736 12628191
Swaminathan J Baxter EM Corces VG The role of histone H2Av variant replacement and histone H4 acetylation in the establishment of Drosophila heterochromatin Genes Dev 2005 19 65 76 15630020
Dhillon N Kamakaka RT A histone variant, Htz1p, and a Sir1p-like protein, Esc2p, mediate silencing at HMR Mol Cell 2000 6 769 780 11090616
Fan JY Rangasamy D Luger K Tremethick DJ H2A.Z alters the nucleosome surface to promote HP1alpha-mediated chromatin fiber folding Mol Cell 2004 16 655 661 15546624
Rangasamy D Berven L Ridgway P Tremethick DJ Pericentric heterochromatin becomes enriched with H2A.Z during early mammalian development EMBO J 2003 22 1599 1607 12660166
Kobor MS Venkatasubrahmanyam S Meneghini MD Gin JW Jennings JL A protein complex containing the conserved Swi2/Snf2-related ATPase Swr1p deposits histone variant H2A.Z into euchromatin PLoS Biol 2004 2 e131 10.1371/journal.pbio.0020131 15045029
Krogan NJ Keogh MC Datta N Sawa C Ryan OW A Snf2 family ATPase complex required for recruitment of the histone H2A variant Htz1 Mol Cell 2003 12 1565 1576 14690608
Mizuguchi G Shen X Landry J Wu WH Sen S ATP-driven exchange of histone H2AZ variant catalyzed by SWR1 chromatin remodeling complex Science 2004 303 343 348 14645854
Krogan NJ Baetz K Keogh MC Datta N Sawa C Regulation of chromosome stability by the histone H2A variant Htz1, the Swr1 chromatin remodeling complex, and the histone acetyltransferase NuA4 Proc Natl Acad Sci U S A 2004 101 13513 13518 15353583
Rangasamy D Greaves I Tremethick DJ RNA interference demonstrates a novel role for H2A.Z in chromosome segregation Nat Struct Mol Biol 2004 11 650 655 15195148
Leach TJ Mazzeo M Chotkowski HL Madigan JP Wotring MG Histone H2A.Z is widely but nonrandomly distributed in chromosomes of Drosophila melanogaster
J Biol Chem 2000 275 23267 23272 10801889
Suto RK Clarkson MJ Tremethick DJ Luger K Crystal structure of a nucleosome core particle containing the variant histone H2A.Z Nat Struct Biol 2000 7 1121 1124 11101893
Robert F Pokholok DK Hannett NM Rinaldi NJ Chandy M Global position and recruitment of HATs and HDACs in the yeast genome Mol Cell 2004 16 199 209 15494307
Yuan GC Liu YJ Dion MF Slack MD Wu LF Genome-scale identification of nucleosome positions in S. cerevisiae
Science 2005 309 626 630 15961632
Harbison CT Gordon DB Lee TI Rinaldi NJ Macisaac KD Transcriptional regulatory code of a eukaryotic genome Nature 2004 431 99 104 15343339
Ren B Robert F Wyrick JJ Aparicio O Jennings EG Genome-wide location and function of DNA binding proteins Science 2000 290 2306 2309 11125145
Ng HH Robert F Young RA Struhl K Genome-wide location and regulated recruitment of the RSC nucleosome-remodeling complex Genes Dev 2002 16 806 819 11937489
Ng HH Robert F Young RA Struhl K Targeted recruitment of Set1 histone methylase by elongating Pol II provides a localized mark and memory of recent transcriptional activity Mol Cell 2003 11 709 719 12667453
Reinke H Horz W Histones are first hyperacetylated and then lose contact with the activated PHO5 promoter Mol Cell 2003 11 1599 1607 12820972
Lee CK Shibata Y Rao B Strahl BD Lieb JD Evidence for nucleosome depletion at active regulatory regions genome-wide Nat Genet 2004 36 900 905 15247917
Boeger H Griesenbeck J Strattan JS Kornberg RD Nucleosomes unfold completely at a transcriptionally active promoter Mol Cell 2003 11 1587 1598 12820971
Boeger H Griesenbeck J Strattan JS Kornberg RD Removal of promoter nucleosomes by disassembly rather than sliding in vivo Mol Cell 2004 14 667 673 15175161
Schwabish MA Struhl K Evidence for eviction and rapid deposition of histones upon transcriptional elongation by RNA polymerase II Mol Cell Biol 2004 24 10111 10117 15542822
Fedor MJ Lue NF Kornberg RD Statistical positioning of nucleosomes by specific protein-binding to an upstream activating sequence in yeast J Mol Biol 1988 204 109 127 3063825
Fedor MJ Kornberg RD Upstream activation sequence-dependent alteration of chromatin structure and transcription activation of the yeast GAL1-GAL10 genes Mol Cell Biol 1989 9 1721 1732 2657404
Fisher-Adams G Grunstein M Yeast histone H4 and H3 N-termini have different effects on the chromatin structure of the GAL1 promoter EMBO J 1995 14 1468 1477 7729422
Axelrod JD Reagan MS Majors J GAL4 disrupts a repressing nucleosome during activation of GAL1 transcription in vivo Genes Dev 1993 7 857 869 8491382
Abbott DW Ivanova VS Wang X Bonner WM Ausio J Characterization of the stability and folding of H2A.Z chromatin particles: Implications for transcriptional activation J Biol Chem 2001 276 41945 41949 11551971
Fan JY Gordon F Luger K Hansen JC Tremethick DJ The essential histone variant H2A.Z regulates the equilibrium between different chromatin conformational states Nat Struct Biol 2002 9 172 176 11850638
Rusche LN Kirchmaier AL Rine J The establishment, inheritance, and function of silenced chromatin inSaccharomyces cerevisiae
Annu Rev Biochem 2003 72 481 516 12676793
Grewal SI Moazed D Heterochromatin and epigenetic control of gene expression Science 2003 301 798 802 12907790
Martinez-Campa C Politis P Moreau JL Kent N Goodall J Precise nucleosome positioning and the
TATA box dictate requirements for the histone H4 tail and the bromodomain factor Bdf1
Mol Cell 2004 15 69 81 15225549
Schneider BL Seufert W Steiner B Yang QH Futcher AB Use of polymerase chain reaction epitope tagging for protein tagging in Saccharomyces cerevisiae
Yeast 1995 11 1265 1274 8553697
Pokholok DK Harbison CT Levine S Cole M Hannett NM Genome-wide map of nucleosome acetylation and methylation in yeast Cell 2005 122 517 527 16122420
Drouin S Robert F Genome-wide location analysis of chromatin-associated proteins by ChIP on CHIP: Controls matter Methods 2005 In press
Svaren J Venter U Hörz W In vivo analysis of nucleosome structure and transcription factor binding in Saccharomyces cerevisiae
Methods in molecular genetics: Molecular microbiology techniques, Part B 1995 New York Academic Press 153 167
Lieb JD Liu X Botstein D Brown PO Promoter-specific binding of Rap1 revealed by genome-wide maps of protein-DNA association Nat Genet 2001 28 327 334 11455386
Kurdistani SK Tavazoie S Grunstein M Mapping global histone acetylation patterns to gene expression Cell 2004 117 721 733 15186774
Wang B Matsuoka S Carpenter PB Elledge SJ 53BP1, a mediator of the DNA damage checkpoint Science 2002 298 1435 1438 12364621
Holstege FC Jennings EG Wyrick JJ Lee TI Hengartner CJ Dissecting the regulatory circuitry of a eukaryotic genome Cell 1998 95 717 728 9845373
|
16248679
|
PMC1275524
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Dec 1; 3(12):e384
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030384
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1624868010.1371/journal.pbio.0030386Research ArticleAnimal BehaviorNeurosciencePhysiologyMus (Mouse)Ultrasonic Songs of Male Mice Songs of Male MiceHoly Timothy E [email protected]
1
Guo Zhongsheng
1
1Department of Anatomy and Neurobiology, Washington University School of Medicine, St. Louis, Missouri, United States of AmericaKauer John Academic EditorTufts University School of MedicineUnited States of America12 2005 1 11 2005 1 11 2005 3 12 e3867 1 2005 14 9 2005 Copyright: © 2005 Holy and Guo.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Music to Her Ears? Male Mice Sing an Ultrasonic Tune
Previously it was shown that male mice, when they encounter female mice or their pheromones, emit ultrasonic vocalizations with frequencies ranging over 30–110 kHz. Here, we show that these vocalizations have the characteristics of song, consisting of several different syllable types, whose temporal sequencing includes the utterance of repeated phrases. Individual males produce songs with characteristic syllabic and temporal structure. This study provides a quantitative initial description of male mouse songs, and opens the possibility of studying song production and perception in an established genetic model organism.
Vocalizations emitted by male mice when encoutering female pheromones have the characteristics of song, including temporal structure and repeated syllables.
==== Body
Introduction
Many animals communicate using sound. Often, brief sounds are produced to warn of danger or mediate aggressive encounters. Some species, however, produce long sequences of vocalizations often called “songs.” Most commonly, these long sequences are generated as a part of courtship. For example, many insects and amphibians [1] advertise their presence and identity with a single type of utterance—which, depending on the species, might be described as a chirp, click, or whine—repeated several times to form a “phrase,” with silent gaps between phrases. The utterance, its repetition rate, and the number of repetitions in a phrase are characteristic of the species [1]. More complex vocalizations are observed in many birds [2], as well as in a few mammals such as whales [3] and bats [4]. These species generate multiple types of sounds organized in more intricate phrases.
Rodents produce a variety of social vocalizations, including vocalizations audible to humans, like postpartum sounds and distress calls, as well as ultrasonic vocalizations [5,6]. In mice, ultrasonic vocalizations utilize frequencies higher than 30 kHz [7], and therefore cannot be detected directly by human ears. A number of studies have shown that mice produce ultrasonic vocalizations in at least two situations: pups produce “isolation calls” when cold or when removed from the nest [8], and males emit “ultrasonic vocalizations” in the presence of females or when they detect their urinary pheromones [6,9–11]. Most commonly, these sounds have been recorded using a detector with narrow frequency tuning [9,10], which suffices to estimate the amount of vocalization. However, because of its narrow frequency tuning, such a detector does not record the acoustical details of these vocalizations.
While numerous studies have focused on the circumstances leading to ultrasound production, few have examined the sounds themselves. Sales [7] observed that these vocalizations consisted of a series of discrete utterances, with species-specific differences in vocalizations. Some diversity was also noted among the utterances within a species [6,7], but it was not determined whether this latter variability was continuous—as in the case, for example, of the “random” variability observed when a single word is spoken many times—or whether the utterances fall into distinct categories. In a recent quantitative study of mouse vocalizations, Liu et al. [12] studied changes in pup vocalizations during the first 2 wk after birth, and compared these to adult vocalizations. However, this study focused only on the aggregate properties of vocalizations, measuring parameters such as median pitch and call rate, which, if applied to humans, would be more analogous to “voice” than to speech. To date, no study that we know of has examined whether the discrete utterances consist of distinct syllable types, or whether these vocalizations have significant temporal structure.
Here, we provide a quantitative description of the ultrasonic vocalizations of the adult male mouse, and show that they display unexpected richness, including several syllable types organized into phrases and motifs. Thus, these vocalizations display the characteristics of song [1,3,13]. Different males, even though genetically identical, show small but significant differences in syllable usage and the temporal structure of their songs. These results indicate that communication among mice may be more complex than previously appreciated. Because of the ubiquity of the mouse for physiological and genetic investigations, these observations may lead to new opportunities in studies of the biological basis of song production and perception.
Terminology
As the terminology used to describe animal vocalizations is varied, we adopt the following definitions. A “syllable” is a unit of sound separated by silence from other sound units [14]; it may consist of one or more “notes,” continuous markings on a sonogram. A “syllable type” is a category of syllable, observed regularly in the animal's vocalization, distinct from other syllable types. A “phrase” is a sequence of syllables uttered in close succession. A “phrase type” or “motif” is a sequence of several syllables, falling into one or more syllable types, where the entire sequence is observed repeatedly in the animal's vocalization.
The term “song” has been used with a variety of connotations, so that Broughton [13] offers three different definitions of song: a sensu latissimo, a “sound of animal origin which is not both accidental and meaningless,” which includes relatively simple vocalizations often described as “calls”; a sensu stricto, “a series of notes [or syllables], generally of more than one type, uttered in succession and so related as to form a recognizable sequence or pattern in time”; and a sensu strictissimo, “a complete succession of periods or phrases,” in which a song consists of several distinct motifs, often delivered in a characteristic sequence.
Results
Listening to Ultrasonic Vocalizations
To induce ultrasonic vocalizations, male mice of the B6D2F1 strain were presented with sex-specific odors applied on cotton swabs (Figure 1). We tested dilute urine of either sex (BALB/c strain) and mixtures of urine from both sexes. (The correspondence between stimulus identity and vocal response will be reported elsewhere.) We recorded all sounds in the chamber with a microphone with flat frequency response from 20 Hz to 100 kHz. While these vocalizations are well beyond the range of human hearing, we make them audible through two techniques. Most straightforward is to play them back slowly. When slowed 16×, these vocalizations sound like a series of breathy whistles (Audio S1). However, slow playback makes it difficult for human listeners to develop an appreciation of the temporal sequence of the vocalizations. Using a phase vocoder algorithm [15], the pitch of these vocalizations can be dropped several octaves without lengthening the duration of the playback. These pitch-shifted vocalizations are reminiscent of birdsong (Audio S2). Readers are urged to listen to these recordings.
Figure 1 Male Mice Vocalize in the Ultrasound after Olfactory Exploration of Urinary Cues
A cotton swab containing female mouse urine (top) was introduced at approximately 30 s into a 210-s trial. Arrow indicates the time of first contact with the cotton swab. Recorded acoustical power is represented as a function of time and frequency, with shading increasing with power. Power below 25 kHz was truncated. Bottom, an expansion of a 2-s period showing vocalizations in greater detail. Individual syllables, as identified by an automated algorithm, are spanned by magenta lines below.
Elementary Features of Vocalizations
Male mouse ultrasonic vocalizations consisted of a rapid series of “chirp-like” syllables in the 30–110 kHz band (Figure 1). Syllables were of varying duration (approximately 30–200 ms), uttered at rates of about ten per second. Most syllables involved rapid sweeps in frequency, with rates of approximately 1 kHz/ms typical. Over tens of seconds, periods of closely spaced syllables alternated with periods of silence. These features of adult male vocalizations, and their analogs for the isolation calls of mouse pups, have been previously described [7,12].
The microphone recorded a variety of sounds in the test chamber, including noises from movement, gnawing, contact with the cage wall, audible squeaks, and ultrasonic vocalizations. For the purposes of this study, we excluded sounds other than ultrasonic vocalizations. The majority of extraneous sounds fell below 30 kHz, and were excluded by selecting the appropriate frequency band. However, some sounds, particularly brief “snaps,” penetrated into the frequency band of the ultrasonic vocalizations. We developed an automated algorithm to recognize ultrasonic vocalizations in terms of their generic features. Subjectively, the algorithm appears no worse than a well-trained human in identifying these vocalizations (see Materials and Methods; Figure 1).
Features of Syllables: Pitch Changes
As reported previously [7], inspection (Figure 1) suggests that some syllables involve relatively sudden, large changes (“jumps”) in frequency. To determine whether these frequency jumps are stereotyped or random, we analyzed a collection of 750 syllables uttered by one mouse in a single 210-s trial. We simplified our description of each syllable by extracting the dominant frequency (the “pitch”) as a function of time (Figure 2A). For each syllable, we compared the pitch at one moment with the pitch in the next time bin, approximately 1 ms later. These pitch pairs were pooled for all 750 syllables, resulting in a total of 31,303 consecutive pitch pairs. This analysis (Figure 2B) revealed four distinct clusters of pitch changes. The long cluster along the diagonal corresponds to the gradual shift in pitch occurring at most time points in all syllables. Two distinct off-diagonal clusters reveal large, stereotyped jumps to or from comparatively low frequencies (35–50 kHz). These downward (“d”) and upward (“u”) jumps are often paired in a syllable (see below and insets for Figure 2B), and will be collectively described as “low jumps.” The cluster just below the diagonal, containing transitions from 70–90 kHz down to 55–70 kHz, results from a third type of jump (“high jump,” or “h”). These jumps were often, but not exclusively, associated with a brief “grace note” at the beginning of a syllable (see jump labeled “h” in lower inset, Figure 2B).
Figure 2 Characterization of Pitch Changes during Syllables
(A) Two examples of syllables, represented in terms of their sonogram (top member of each pair of panels) and the extracted pitch versus time (bottom member of pairs).
(B) Plot of pitch at one time point versus the next time point (Δt = 1.02 ms). All such pitch pairs in all syllables from a single trial with 750 syllables are shown, representing a total of 31,303 pitch changes. Particular pitch jumps are placed within the context of their individual syllables at right (top syllable, 98 ms in duration; bottom syllable, 33 ms in duration).
(C) Pitch pairs analyzed for single 210-s trials from 45 different mice, containing in aggregate 15,543 syllables and over 600,000 pitch pairs. The distribution of pitch pairs is represented as a two-dimensional histogram; the correspondence between grayscale and number of observations is indicated in the color bar at right. Polygons define the clusters corresponding to the three jump types “u,” “h,” and “d.”
(D) Numbers of each type of pitch jump per trial (45 mice, one trial each).
These pitch jumps were identified in Figure 2B from a single 210-s recording from one mouse. To determine whether these jumps are stereotypic features of the ultrasonic vocalizations of all male mice, we performed the same analysis for a 210-s trial from each of 45 different males. The pitch changes in adjacent time bins are pooled across mice in Figure 2C. Both the number of clusters and their positions and sizes are essentially unchanged, and examples of all three types of jumps were broadly distributed across mice (Figure 2D). Thus, at least for similarly aged males of the B6D2F1 strain, these pitch jumps are a universal feature of ultrasonic vocalizations.
Pitch Jumps and Mechanisms of Sound Production
Many syllables with low jumps display both a fundamental frequency and a faint first harmonic during the low-frequency period (Figure 3A; see also Figures 1 and 2A). The frequency of the harmonic is almost precisely twice that of the fundamental, suggesting the involvement of a resonator in the production of these sounds. A priori, this resonator might be the vocal folds of the larynx. However, based on the effect of partial replacement of air with helium, Roberts [16] argued that these sounds are not produced by the vibration of vocal cords. Instead, he proposed that ultrasound arises from an aerodynamic whistle, and showed that mechanical whistles can produce sounds similar to the examples described by Sales [7], including pitch jumps.
Figure 3 Features of Vocalizations Relating to Mechanisms of Sound Production
(A) Syllable with both a fundamental and first harmonic.
(B) Abundance of frequency (vertical axis is frequency, continued from [A]) in syllables with (LJ+) and without (LJ−) low jumps.
(C) Average pitch (top) and mean ± standard deviation log10(power) (bottom) as a function of time, surrounding a downward low jump (for syllables with low jumps) or surrounding the upward crossing of 75 kHz (for syllables without low jumps). Power units are arbitrary but consistent between syllable types. Color scheme is as in (B).
(D) Syllable showing extensive temporal overlap and independent frequency modulation among the different notes in the syllable. Syllables are from the same trial analyzed in Figure 2B.
While our recordings appear largely consistent with Roberts's results, several features of these vocalizations indicate that their production is more sophisticated than that of a whistle from a rigid, static pipe. The rigid whistles investigated by Roberts had a characteristic relationship between frequency and fluid velocity [16]. Frequency was fairly stable over a range of velocities, and would suddenly jump to a new frequency at yet higher or lower velocities. In contrast, the pitch of mouse vocalizations is modulated considerably, in both a continuous and discrete (jump) fashion. Despite their stereotyped form, jumps were not obligatory upon reaching a particular frequency. While down-type jumps began from frequencies of 65–80 kHz (see Figure 2B), these frequencies were well-sampled even in syllables that lack these jumps (Figure 3B). Furthermore, if jumps were produced by changes in air velocity, one might expect to see differences in vocal power between cases where jumps do and do not occur. In contrast with this expectation, the power distributions of syllables both with and without “d” jumps overlap considerably (Figure 3C), although variability in the mouse's head position and orientation relative to the microphone could obscure a true relationship.
Finally, the fine-scale temporal structure of pitch jumps appears to be inconsistent with the nonlinear properties of purely static whistles. During a downward low jump, the pitch of the preceding phase overlaps in time with the pitch in the succeeding phase (Figure 3A), often by 5–10 ms. This behavior is apparently not observed in pitch jumps arising from mode-locking nonlinearities [17], where changes in pitch are nearly instantaneous. In a few cases, both tones were present simultaneously for longer periods, with one frequency modulated and the other nearly fixed (Figure 3D). In birdsong, similar observations were used by Greenewalt [18] to posit two sites of sound production—specifically, that birds could independently control the left and right sides of their syrinx. This assertion was later confirmed directly [19]. Examples such as Figure 3D may indicate that mice have at least two sites of ultrasound production. However, the strength of this conclusion is tempered by our incomplete knowledge of the nonlinear properties of aerodynamic whistles [20].
Classifying Syllables into Distinct Types
Because pitch jumps exist in three distinct categories, their presence or absence serves as a basis for classifying individual syllables into types. However, it is possible that other features of these vocalizations might also be a basis for classification. We therefore analyzed these syllables using multidimensional scaling, a technique that has been used previously to classify syllables in birdsong [21]. Multidimensional scaling provides a method to represent high-dimensional data in lower dimensions; it takes as an input the pairwise “distance” between any two pitch waveforms, and attempts to faithfully represent the set of all pairwise distances by proper placement of points, each representing a single syllable, in two dimensions (Figure 4A). Because inspection suggested that a given syllable type can be uttered quickly or slowly, we first aligned the pairs by warping their time axes to place the pitch waveforms in maximal correspondence with each other (Figure 4A; [22]). We also used a variant of multidimensional scaling, called isomap [23], which assembles global information from local features.
Figure 4 Multidimensional Scaling Analysis of Syllable Types
(A) Pairs of pitch waveforms are temporally aligned using dynamic time warping, and pairwise distances (root mean squared difference) are computed. Using multidimensional scaling (MDS)/isomap, projections are found that approximately preserve the distances between pairs.
(B) Isomap analysis of all pitch waveforms in the trial analyzed in Figure 2B. Points are colored according to the presence or absence of low jumps as in Figure 3B.
(C) A different isomap projection, focusing only on syllables containing low jumps. Sonograms of representative syllables in both clusters are shown in the insets. Pitch waveforms are from the same trial analyzed in Figure 2B.
The isomap analysis revealed the presence of several clusters, indicating distinct syllable types. The most prominent distinction is illustrated in Figure 4B, with an almost perfect correspondence between cluster membership and the presence or absence of low-jump transitions. Closer examination of the cluster representing syllables containing low jumps reveals further clusters within this overall category. An example is shown in Figure 4C, in which syllables again group into types that can be described in terms of their pitch jumps: one distinct cluster contains almost entirely syllables with an “h” jump followed by a “d” jump. Further projections (not shown) confirm the presence of additional clusters, which also correspond to particular sequences of pitch jumps.
Therefore, general classification techniques confirm that syllables are naturally grouped by their pitch jumps. In fact, from the isomap analysis we have not found evidence for any other means to categorize syllables; in all cases we have examined, clear isomap clusters correspond to types defined by their sequence of pitch jumps. However, it remains possible that further subtypes exist, but that the isomap analysis fails to reveal these clusters. We therefore focused on the simplest syllable type, with no pitch jumps at all. These syllables take a variety of forms, some of which are illustrated in Figure 5A. We noted that many had an oscillatory appearance. We therefore fit each pitch waveform to a sine wave, scaling and shifting both the time and frequency axes for maximal alignment. (We did not permit local time warping, as used in Figure 4.) The quality of the fit could be assessed by scaling and shifting each pitch waveform to a common axis, revealing that the vast majority of these waveforms lie on top of each other, as well as the underlying sine wave (Figure 5B). Based on this result, we call syllables lacking any pitch jumps “sinusoidal sweeps” (SSs).
Figure 5 Pitch Waveforms of Syllables Lacking Jumps
(A) Sonograms of representative syllables, showing a range of oscillatory behavior.
(B) Overlay of pitch waveforms for all 361 syllables lacking pitch jumps from the trial analyzed in Figure 2B. Time and frequency axes have been shifted and globally stretched independently for each syllable to bring waveforms into maximal overlap with a sine wave. The root mean squared error in fit to the sine wave is indicated by dashed lines.
(C and D) Histogram of starting (C) and ending (D) phases.
(E) Relationship between the oscillation rate (measured in periods/millisecond) and amplitude of the best-fit sine wave. Only syllables with at least 0.3 periods (160/361) are shown; syllables spanning a smaller fraction of a period do not permit an accurate measurement of oscillation rate or amplitude.
This analysis suggests that the pitch waveforms of SS syllables can be accurately described in terms of five variables (see Materials and Methods): the starting and ending phases, the rate of oscillation, the center frequency, and the pitch sine amplitude. Analysis of these parameters reveals that most SSs begin during (or just before) the rising phase of the sine wave (Figure 5C), and that a large subset terminate at the peak of the sine wave (Figure 5D). There is also a strong inverse relationship between the oscillation rate and the oscillation amplitude (Figure 5E; see example in bottom two waveforms in Figure 5A). An analogous inverse relationship has been found in birdsong, between the trill rate and the amplitude of pitch variation [24]. In birdsong, this relationship has been interpreted as evidence of a performance limit in the rate at which frequency can be modulated by changes in beak conformation. An analogous limit may constrain a mouse's ability to modulate the frequency of its whistle.
While syllables are naturally grouped by their pitch jumps, and indeed we have not found any clear means of classifying them in a different way, it remains possible that other groupings exist. In particular, short stretches of a recording sometimes seem to provide evidence for further subtypes; an example is shown in the next section (Figure 6A). Table 1 shows a breakdown by prevalence of the most common syllable types in mouse ultrasonic vocalizations.
Figure 6 Examples of Temporal Regularities in Mouse Song
(A) Sequences of syllables in a phrase. Here, “hdu” syllables have been classified as “A” or “B” depending on whether the lower frequency band fell or rose, respectively. SS and “h” (with a brief grace note) are labeled “C”.
(B) Example of a phrase repeated three times without interruption in the original song. The three repeats are shown one above the other, aligned on the start time for the phrase. See Audio S4 for entire sound recording.
(C) Long time scale changes in syllable type. Syllable type is categorized by whether low jumps are present (LJ+) or absent (LJ−). Shown is the number of syllables without low jumps, out of the most recent 20 syllables. Insets contain sonograms from the indicated portions of the sequence.
(A) and (C) are from the same trial analyzed in Figure 2B; (B) is from a different mouse.
Table 1 The Most Common Syllable Types in Mouse Ultrasonic Vocalizations, Labeled by Pitch Jump
Temporal Sequencing of Syllables
In sonograms of mouse vocalizations, complex syllable sequences can be identified: Figure 6A shows an example of a phrase in which three “hdu” syllables with descending low-frequency bands (labeled “A”) are followed by six “hdu” syllables with ascending low-frequency bands (labeled “B”); the phrase is finished off by an “h” syllable (almost a SS, but for the brief grace note), an A-type “hdu,” and an SS (Audio S3).
An example of a motif can be seen in Figure 6B, in which a phrase beginning with 2–3 SSs followed by 6–8 “du” syllables is repeated three times. The consistency of this repeated sequence, in the context of the whole, is easily noted in pitch-shifted playbacks (Audio S4).
Finally, there are regularities in the syllable types over longer time scales. Figure 6C shows an example of a trial that begins with a series of SSs, has a middle period with many syllables containing low jumps, and ends with repeated blocks of “h” syllables.
To determine whether such examples are statistically significant, we investigated the temporal structure of these vocalizations quantitatively in terms of two models of syllable selection. To simplify the analysis, we grouped syllables into only two categories, depending on whether they did (“1”) or did not (“0”) contain one or more low jumps. We considered whether individual syllables might be selected randomly. In the first model, we tested whether the probability of selecting a syllable was based purely on the prevalence of each type, so that each syllable is selected independently of all others. In the second model, the selection probability depended on the identity of the previous syllable (Figure 7A): from the data, we calculated the conditional probability pi
→j to choose a syllable of type j after a syllable of type i (i, j = 0, 1). We also used a third state (a “gap”) to represent a silent period lasting more than 0.5 s, to ensure that the analyzed state transitions occurred within a phrase. Omitting the gap state from the model did not qualitatively change the results.
Figure 7 Quantitative Modeling of Syllable Temporal Sequences
(A) A three-state Markov model, where the states correspond to syllables with (“1”) or without (“0”) low jumps, and to a gap of greater than 0.5 s in the sequence. Arrows indicate possible choices for the next state; transition probabilities are calculated from the observed sequence of syllables and gaps.
(B) Observed numbers of the eight distinct three-syllable combinations, and the number expected from two models: “syllable prevalence” picks the next syllable randomly based on the proportion of each type, whereas “transition probability” employs the Markov model diagrammed in (A).
(C) Comparison of transition probabilities to type 1 syllables with the prevalence of type 1 syllables. “Prevalence of 1” is n
1/(n
0 + n
1), where ni is the number of syllables of type i; prevalence of transition g→1 is calculated as ng
→1/(ng
→0 + ng
→1), where ni
→j is the number of observed transitions from state i to state j (g = gap); and prevalence of 1→1 is n
1→1/(n
0→1 + n
1→1). Each point represents the results from a single trial, of 81 qualifying trials (see text).
We then examined the prevalence of all possible three-syllable combinations (see Materials and Methods) in terms of these two models. As shown in Figure 7B, the first model, based purely on prevalence, does a poor job of predicting the distribution of three-syllable combinations (p << 10−10). The transition-probability model provides a much more accurate description of the temporal structure. However, it, too, is insufficient (p ≍ 10−6) to capture all of the higher-order structure of these three-syllable sequences. Similar conclusions apply to four- and five-syllable sequences.
Therefore, we find that syllables are not chosen independently in random order. From examples of raw sonograms (see Figure 1), it appears that type 1 syllables (those with low jumps) tend to be grouped in blocks. To examine this aspect of sequencing, we compared the prevalence of type 1 syllables against the likelihood that the next syllable after a type 1 would also be a type 1. For the example in Figure 7B, 258/750 (34%) of syllables were of type 1, but the likelihood of a successive type 1 was much higher (58%). On the basis of counting statistics (binomial distribution), this difference is highly significant (p << 10−10).
To determine whether this tendency to repeat low-jump syllables is a universal feature of these vocalizations, we recorded the vocalizations of 45 socially experienced males over a period of 3 wk. Over the 3 wk, each animal participated in nine trials, each 210 s in duration, during which the male was presented with either a blank (non-odorized) cotton swab or one with 20 μl of dilute mouse urine (see Materials and Methods). Of the more than 400 trials, 81 (from a total of 32 different males) contained ten or more examples each of type 0 and 1 syllables, and were tagged as “qualifying trials.” These qualifying trials contained sufficient numbers of each syllable type to allow measurement of the syllable prevalence and transition probabilities. We consistently found that type 1 syllables were more likely following another type 1 (Figure 7C): in 78/81 qualifying trials, type 1 syllables were more likely following another type 1 than would have been predicted from their overall prevalence. This demonstrates a strong tendency for male mice to utter low-jump syllables in blocks.
Similarly, we found that type 1 syllables were very unlikely to be used at the beginning of a phrase: after a gap, the likelihood of a type 1 syllable was lower (in 78/81 trials) than would have been predicted from chance selection of syllable types (Figure 7C). A related phenomenon is seen in zebra finch song, in which phrases often begin with an “introductory note” [25].
We conclude that these vocalizations display strong temporal regularities. Therefore, mouse ultrasonic vocalizations contain the two elements most commonly used to define song [1,3,13]: the vocalizations contain multiple syllable types, and these syllables are uttered in regular, repeated temporal sequences. We therefore label these vocalizations as songs.
The Songs of Individual Males
In many songbirds, individual males produce a characteristic song, which in the case of oscine songbirds is learned from a tutor. To determine whether individual male mice produce stereotyped songs, we recorded the songs of 45 males over a period of 3 wk. Seven of the 45 mice had four or more “qualifying trials” (see above) with enough syllabic diversity to permit analysis.
As shown in Figure 8A, individual mice displayed tendencies to use particular syllable types. For example, mouse 2 tended to utter an abundance of “du” syllables, whereas mouse 3 used a larger number of “h” syllables. To determine whether these tendencies were sufficiently reliable to characterize the song of individual males, we again used isomap to generate a graphical representation of the syllable selection probability across mice and trials (see Figure 8B and Materials and Methods). Importantly, the isomap analysis was blind to the singer's identity, so that any differences between mice were a feature discovered by, rather than imposed upon, this analysis. As shown in Figure 8B, the choice of syllable types was fairly consistent over the 3-wk period among trials from an individual. With a single exception (the mouse labeled by cyan stars), trials from a given mouse tended to occupy one of the three arms, or the center, of this distribution. This tendency to cluster is corroborated by the fact that the mean “distance” (see Materials and Methods) between trials from a particular mouse (1.7 ± 0.1%, mean ± standard error of the mean) was significantly smaller than the mean distance between trials from different mice (2.09 ± 0.04%, p < 0.001, one-sided t-test). In a two-alternative forced-choice experiment, individuals could typically be recognized by their song on the basis of a single trial.
Figure 8 Regularities in the Songs of Individual Mice
All seven mice with four or more qualifying trials (see text) are analyzed.
(A) Syllable usage for three of the more common syllable types for three different mice. Error bars represent the standard error of the mean across trials.
(B) Patterns of syllable usage on individual trials across mice. Each point corresponds to a single trial, where the trials from a particular mouse are marked with a consistent color and marker. For mice 1–3, colors are consistent between (A) and (B). Placement of points in the plane reflects the pairwise “distance” between trials, where that distance measures the overall differences in percentage of each syllable type (see Materials and Methods). Projection into two dimensions was performed by isomap.
(C) Temporal regularities in song structure. Transition probabilities for all qualifying trials grouped by mouse.
To determine whether these individual differences extend to the temporal domain, we calculated the syllable transition probabilities p
0→1 and p
1→0 from the observed sequence of syllables. For these seven males, the transition probabilities for all qualifying trials are plotted in Figure 8. Inspection suggests that the spread in values for a single male is smaller than the spread for the population as a whole. To determine whether this effect is significant, we analyzed the spread of transition-probability values across trials using a bootstrap analysis, comparing the spread in the true dataset to simulated datasets in which the singer's identity was scrambled across trials (see Materials and Methods). The average spread using the correct identities was significantly (p = 0.02, 10,000 bootstraps) less than the spread for scrambled datasets. Therefore, we conclude that individual males also have characteristic temporal structure to their songs: across males, the differences in temporal structure are larger than the variability across trials by a single male.
Discussion
This study reveals that the ultrasonic vocalizations of the mouse have the characteristics of song. Qualitatively, this is apparent directly from playback of pitch-shifted audio recordings; we have also provided quantitative evidence for the usage of distinct syllable types arranged in nonrandom, repeated temporal sequences. These songs satisfy Broughton's sensu stricto definition of song [13], as well as many aspects of his sensu strictissimo (see Figure 6). While courtship songs are common among birds, insects, and frogs, song has only rarely been documented in mammals, and to our knowledge only in humans, whales, and bats [3,4]. However, some rodent species display a variety of calls [26] and at least one other, the rat Dactylomys dactylilnus, utters long sequences of vocalizations that contain some syllabic diversity [27]. More generally, a number of Central and South American rodent species display complex vocalization (L. H. Emmons, personal communication), but none has been characterized in detail. However, it seems likely that song is more widely distributed than we currently appreciate. While the neural and motor mechanisms used to produce song and other communication sounds vary across species, recent work has indicated some commonality at the molecular level: the Foxp2 transcription factor, expressed in the brain of zebra finches during vocal learning [28], seems to be required both for mouse ultrasonic vocalization [29] and normal human speech [30].
Subjectively, mouse song has a diversity and complexity that exceeds that of most insect and amphibian advertisement songs, which often contain only a single syllable type [1], perhaps modulated in amplitude and cadence [31]. At the syllable level, diversity in mouse song comes in two forms, discrete and continuous. Discrete categories of syllables exist, as evidenced by the appearance of distinct clusters, by two criteria: in terms of the sequence of stereotyped frequency jumps (see Figure 2), and by a comparison of the pitch waveforms of individual syllables (see Figure 4). Within syllable types, there also exists considerable continuous variability (see Figures 5 and 6A). Because of our adoption of a strict quantitative classification of types, we have not used this continuous variability to define subtypes. This does not, however, argue that additional types are not present, merely that our analysis does not yet support further subdivision of types. Our quantitative classification scheme may be stricter than that employed in some analyses. A comparison of both subjective and quantitative classification has been carried out for the song of swamp sparrows: subjective methods [32] were used to classify notes into either 96 subtypes (which they termed the “splitter's classification”), or into six major categories (termed the “lumper's classification”). A later quantitative analysis carried out by the same laboratory, using techniques related to those employed here, found that notes clustered in general agreement with the major categories identified in the “lumper's classification,” with no evidence for further subdivision [21].
The richness and complexity of mouse song appear to approach that of many songbirds. For example, in the zebra finch, a widely used model organism for studying song production, individuals have a number (3–7) of syllable types [25,33] similar to the number of common types we find in mice (Table 1). There are other species, for example, canaries, whose vocal repertoire would appear to exceed that of mice [34]. Both mice (see Figure 6) and birds [25,33] exhibit regular temporal structure in their songs, including the production of repeated themes with sharp transitions between syllable types. However, mice also exhibit more gradual changes in syllable structure (see Figure 1). Overall, the tendency to repeat a syllable, with sharp transitions between types, appears to be stronger in some birds [34] and whales [3] than in mice. However, in birds these sharp transitions are a feature of the adult “crystallized” song; juvenile or isolation-reared birds are more experimental and less predictable in terms of the temporal structure of their song [33,35]. Indeed, our pitch-shifted recordings of mouse song sound similar to the early “plastic” song of species such as swamp sparrows (Audio S5). For this reason, any comparison between birds and mice should consider the development of mouse song over the lifetime of the animal. Such a study has been undertaken for properties like mean pitch and cadence over the first 2 wk of life [12], but is lacking for the more complex features that compose song.
Because mouse songs are ultrasonic and therefore inaudible to human ears, it is worth noting that laboratory domestication has probably not acted to preserve the full richness of mouse song through generations inbreeding. One study documented considerable variability in the amount of vocalization by different laboratory strains [36]. In contrast, domesticated bird populations have been subject to song selection, and indeed sub-strains such as the Waterschlager canary have been bred for particular vocal qualities. It therefore seems possible that wild mice might exhibit considerably greater diversity and/or more complex structure in their songs. Future comparisons between the songs of mice and birds may benefit from using wild mice.
A final question is whether mice, like birds, learn their songs through experience. The fact that different males have characteristic syllable usage and temporal structure to their songs (see Figure 8) is evidence for individual variability in song. Directly testing the role of experience will require that the auditory environment during development be explicitly controlled.
In sum, we have demonstrated that the ultrasonic vocalizations of mice are songs, containing different syllable types sequenced in regular temporal patterns. Different individuals sing recognizably different songs. These results open new possibilities for molecular and physiological studies of the production and perception of song in a well-studied laboratory organism.
Materials and Methods
Signal acquisition and testing environment
Sounds were recorded with a microphone and amplifier (1/4” microphone, model 4939, Brüel and Kjær, Nærum, Denmark) with flat frequency response out to 100 kHz and diminishing sensitivity at higher frequencies. Sounds were digitized at 250 kHz, 16 bits (National Instruments, Austin, Texas, United States) and captured to disk within a custom MATLAB-based program. To attenuate environmental noise, trials were conducted in a wooden enclosure with a transparent Plexiglas front. A slow stream of fresh air flowed through each chamber.
Experimental design
Four-week-old males of the B6D2F1 strain (an F1 cross between C57Bl/6 and DBA2/J) were purchased from Jackson Laboratory (Bar Harbor, Maine, United States). Mice were kept on a 12 h/12 h light/dark cycle and were individually housed starting at 8 wk. Trials were conducted in April and May when the animals were at least 100 d old. Males were given two 3-min social experiences per day for 4 d, one to a BALB/c female and one to a BALB/c castrated male painted with 50 μl of 0.316× intact BALB/c male mouse urine [11]. The order of female/male social experiences was balanced across days.
Animals were acclimatized to the testing environment with three 12–15 min episodes in the testing chamber. A given male was then tested 1 d/wk over a period of 3 wk. A day's test consisted of 15 min of acclimatization followed by three 210-s trials: presentation of an odorized (20-μl drop of mouse urine, see below) swab, presentation of a non-odorized (blank) swab, and presentation of a second odorized swab. For a given mouse, the gap between trials was typically approximately 20 min. Swabs were introduced through a hole in the lid of the enclosure at 30 s into the trial; they were removed immediately after the end of a trial.
Urine stimuli used to trigger vocalizations were either 0.316× male mouse urine, 0.316× female mouse urine, or any of nine different mixtures of male and female mouse urine, where the concentration of each component was one of 0.316×, 0.1×, or 0.0316×. The correlation between stimulus identity and vocal response will be reported separately.
Data analysis
Stored acoustical waveforms were processed in MATLAB to compute the sonogram (512 samples/block, half-overlap, resulting in a time resolution of 1.02 ms and a frequency resolution of 0.98 kHz). Sonograms were thresholded to eliminate the white noise component of the signal, frequencies outside 25–110 kHz were truncated, and the resulting sonograms were stored to disk as sparse complex matrices. This procedure greatly reduced the storage and processing requirements for later analysis, and also eliminated hiss when playing back reconstructed acoustical waveforms.
Syllables were identified by computing three time-varying parameters from the sparse sonograms: the mean frequency, the “spectral purity” (fraction of total power concentrated into a single frequency bin), and the “spectral discontinuity,” which is computed in the following manner: if pi(fj)
is the normalized power as a function of frequency fi in the ith time bin, then the spectral discontinuity δi between two neighboring time bins is
Essentially, δ measures the change in the allocation of power across frequencies between two adjacent time bins; to accommodate the fact that syllables involve rapid sweeps in frequency, we permit a slight shift in frequency (Δj up to three bins in either direction, almost 3 kHz) to maximize the alignment between adjacent time bins. Note that because p^
is normalized, 0 ≤ δi ≤ 2.
These three parameters were median-filtered over 10 ms, and syllables were identified as periods longer than 5 ms in which mean frequency exceeded 35 kHz, spectral purity exceeded 0.25, and δ was less than one. Because faint syllables were occasionally interrupted by brief periods of dropout when the power approached the noise level, candidate syllables separated by less than 30 ms were merged.
The performance of this algorithm was compared with human inspection on 50 randomly selected 210-s trials. The algorithm successfully identified the vast majority (>95%) of human-identified syllables, with systematic omission occurring only on the faintest and briefest syllables. False positives were encountered so rarely (two clear examples in 10,500 s of recording) that it was difficult to estimate their frequency, but they were clearly rarer than true syllables by several orders of magnitude. The algorithm also identified numerous vocalizations that were initially missed by a human observer, but which proved upon closer inspection to be correctly identified (verified graphically and by audible playback). The algorithm also identified the timing of the beginning and end of each syllable with high accuracy; occasional discrepancies with a human observer arose from interfering sounds or when the beginning or end of the syllable was unusually faint.
Pitch was defined as the dominant frequency as a function of time, discarding periods of dropout. Note that pitch was occasionally corrupted by other noises, contributing particularly to background “hash” in Figure 2. The criteria used to define the three pitch jump types “d,” “u,” and “h” are illustrated in Figure 2C.
Alignment of pairs of pitch waveforms (see Figure 4) was performed by dynamic time warping [22]. The distance between any two pitch waveforms was defined as the root mean squared distance between the aligned waveforms. The isomap analysis of pitch waveforms used a neighborhood distance criterion of 3 kHz; in Figure 4 and other such figures, only the largest connected component is shown.
In fitting SSs to a sine wave (see Figure 5), the sine wave was described in terms of the following parameters:
, where ϕ0 is the starting phase, f is the rate of oscillation, y¯
is the center frequency, and A is the pitch sine amplitude. The total duration of the syllable, T, defines the ending phase ϕend by ϕend = 2πft + ϕ0.
In analyzing the temporal structure of mouse songs (see Figure 7), the prevalence of a syllable type was defined as follows: if n
0 and n
1 are the numbers of type 0 and type 1 syllables, respectively, then the prevalence of type 0 (within that trial) is defined as p
0 = n
0/(n
0 + n
1)
. The prevalences of other types are defined similarly. The prevalence of a particular transition, for example, p
1→1, is defined analogously in terms of the numbers of each transition type n
1→0 and n
1→1 observed during the trial. In Figure 7B, sequences interrupted by a gap were discarded. The expected number of a given three-syllable combination “abc” was calculated as Npapbpc for the “syllable prevalence” model, and as Npapa
→b
pb
→c for the transition-probability model, where N is the total number of three-syllable combinations.
The analysis of syllable usage across mice (see Figure 8B) defined the distance between trials in terms of the differences in percentage utilization of each syllable type. More precisely, if pi
1 is the fraction of syllables of type i used in trial 1, and pi
2 is the fraction of the same syllable type in trial 2, then d
12 = 〈|pi
1 − pi
2|〉i. The pairwise distances were used to project into two dimensions using isomap, much as schematized for pitch waveforms in Figure 4A (rightmost panel). The isomap analysis used a local neighborhood definition consisting of the five closest points; this criterion incorporated all trials into a single connected component.
The bootstrap analysis of the spread in transition probabilities across individuals (see Figure 8C) was performed as follows: starting from the median value (calculated separately for each condition, 0→1 or 1→0, and for each mouse), we calculated the absolute deviation for each trial. We then calculated the mean absolute deviation across all mice, conditions, and trials. We compared this mean deviation to the same quantity calculated from synthesized datasets in which the singers' identities were scrambled across trials.
Supporting Information
Audio S1 Slowed (16×) Playback of the 2-s Section Expanded in the Lower Panel of Figure 1
(977 KB WAV).
Click here for additional data file.
Audio S2 Pitch-Shifted (16×) Playback of the 2-s Section Expanded in the Lower Panel of Figure 1
(61 KB WAV).
Click here for additional data file.
Audio S3 Pitch-Shifted (16×) Playback of the Phrase in Figure 6A
(48 KB WAV).
Click here for additional data file.
Audio S4 Pitch-Shifted (16×) Playback of a Longer Segment of Song
The triply repeated phrase shown in Figure 6B begins at 40 s into the recording.
(1.6 MB WAV).
Click here for additional data file.
Audio S5 Recording of Juvenile Swamp Sparrow Song
For comparison between bird and mouse songs. Courtesy of Peter Marler.
(2.4 MB WAV).
Click here for additional data file.
We are grateful to Markus Meister for encouraging these investigations, to Eric Dorner for his help in initiating these experiments, to Bence Ölveczky for drawing our attention to the fine-scale temporal structure of pitch jumps (see Figure 3A), to Peter Marler for the swamp sparrow sound recording and helpful discussions, and to Louise Emmons and Robert Pless for helpful discussions. We thank Greg DeAngelis, Rebecca Hendrickson, Markus Meister, Bence Ölveczky, and the anonymous referees for comments on the manuscript. This work was supported by startup funds from Washington University School of Medicine, a grant from the National Institutes of Health (TEH, 5R01DC005964–02), and the Pew Scholars Program (TEH).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. TEH conceived the experiments. TEH and ZG designed the experiments, built the custom hardware, and wrote the software. ZG collected the data. TEH analyzed the data and wrote the paper.
Citation: Holy TE, Guo Z (2005) Ultrasonic songs of male mice. PLoS Biol 3(12): e386.
Abbreviations
“d,”downward low jump
“h,”high jump
SSsinusoidal sweep
“u,”upward low jump
==== Refs
References
Simmons AM Popper AN Fay RR Acoustic communication. Volume 16, Springer handbook of auditory research 2003 New York Springer 404
Marler P Slabbekoorn H Nature's music: The science of birdsong 2004 Boston Elsevier 513
Payne RS McVay S Songs of humpback whales Science 1971 173 585 597 17833100
Behr O von Helversen O Bat serenades—Complex courtship songs of the sac-winged bat Saccopteryx bilineata
Behav Ecol Sociobiol 2004 56 106 115
Whitney G Nyby J Willott JR Sound communication among adults The auditory psychobiology of the mouse 1983 Springfield (Illinois) C. C. Thomas 98 129
Gourbal BEF Barthelemy M Petit G Gabrion C Spectrographic analysis of the ultrasonic vocalisations of adult male and female BALB/c mice Naturwissenschaften 2004 91 381 385 15278217
Sales GD Ultrasound and mating behavior in rodents with some observations on other behavioural situations J Zool Lond 1972 168 149 164
Haack B Markl H Ehret G Willott JF Sound communication between parents and offspring The auditory psychobiology of the mouse 1983 Springfield (Illinois) C. C. Thomas 57 97
Wysocki C Nyby J Whitney G Beauchamp G Katz Y The vomeronasal organ: Primary role in mouse chemosensory gender recognition Physiol Behav 1982 29 315 327 7146137
Sipos M Kerchner M Nyby J An ephemeral sex pheromone in the urine of female house mice (Mus domesticus)
Behav Neural Biol 1992 58 138 143 1456933
Stowers L Holy T Meister M Dulac C Koentges G Loss of sex discrimination and male-male aggression in mice deficient for TRP2 Science 2002 295 1493 1500 11823606
Liu RC Miller KD Merzenich MM Schreiner CE Acoustic variability and distinguishability among mouse ultrasound vocalizations J Acoust Soc Am 2003 114 3412 3422 14714820
Broughton WP Burnel RG [Glossary] Accoustic behavior of animals 1963 Boston Elsevier 883
Doupe AJ Kuhl PK Birdsong and human speech: Common themes and mechanisms Annu Rev Neurosci 1999 22 567 631 10202549
Moulines E Laroche J Non-parametric techniques for pitch-scale and time-scale modification of speech Speech Commun 1995 16 175 205
Roberts L The rodent ultrasound production mechanism Ultrasonics 1975 13 83 88 1167711
Fee MS Shraiman B Pesaran B Mitra PP The role of nonlinear dynamics of the syrinx in the vocalizations of a songbird Nature 1998 395 67 71 12071206
Greenewalt C Bird song: Acoustics and physiology 1968 Washington (D. C.) Smithsonian Institution Press 194
Suthers RA Contributions to birdsong from the left and right sides of the intact syrinx Nature 1990 347 473 477
Chanaud RC Aerodynamic whistles Sci Am 1970 222 40 46
Clark CW Marler P Beeman K Quantitative analysis of animal vocal phonology: An application to swamp sparrow song Ethology 1987 76 101 115
Rabiner L Rosenberg A Levinson S Considerations in dynamic time warping algorithms for discrete word recognition IEEE Trans Acoust 1978 26 575 582
Tenenbaum JB de Silva V Langford JC A global geometric framework for nonlinear dimensionality reduction Science 2000 290 2319 2323 11125149
Podos J A performance constraint on the evolution of trilled vocalizations in a songbird family (Passeriformes: Emberizidae) Evolution 1997 51 537 551
Price PH Developmental determinants of structure in zebra finch song J Comp Physiol Psychol 1979 93 260 277
Hafner MS Hafner DJ Vocalizations of grasshopper mice (genus Onychomys ) J Mammol 1979 60 85 94
Emmons LH Morphological, ecological, and behavioral adaptations for arboreal browsing in Dactylomys dactylinus (Rodentia, Echimyidae) J Mammal 1981 62 183 189
Haesler S Wada K Nshdejan A Morrisey EE Lints T FoxP2 expression in avian vocal learners and non-learners J Neurosci 2004 24 3164 3175 15056696
Shu W Cho JY Jiang Y Zhang M Weisz D Altered ultrasonic vocalization in mice with a disruption in the Foxp2 gene Proc Natl Acad Sci U S A 2005 102 9643 9648 15983371
Lai CS Fisher SE Hurst JA Vargha-Khadem F Monaco AP A forkhead-domain gene is mutated in a severe speech and language disorder Nature 2001 413 519 523 11586359
Kelley DB Tobias ML Horng S Ryan M Producing and perceiving frog songs; dissecting the neural bases for vocal behaviors in Xenopus laevis
Anuran communication 2001 Washington (D. C.) Smithsonian Institution Press 156 166
Marler P Pickert R Species-universal microstructure in the learned song of the swamp sparrow, Melospiza geogiana
Anim Behav 1984 32 673 689
Immelmann K Hinde RA Song development in the zebra finch and other estrildid finches Bird vocalizations: Their relations to current problems in biology and psychology 1969 Cambridge Cambridge University Press 61 74
Nottebohm F Nottebohm ME Crane L Developmental and seasonal changes in canary song and their relation to changes in the anatomy of song-control nuclei Behav Neural Biol 1986 46 445 471 3814048
Ölveczky BP Andalman AS Fee MS Vocal experimentation in the juvenile songbird requires a basal ganglia circuit PLoS Biol 2005 3 e153 10.1371/journal.pbio.0030153 15826219
Maggio J Whitney G Ultrasonic vocalizing by adult female mice (Mus musculus)
J Comp Psychol 1985 99 420 436 4075780
|
16248680
|
PMC1275525
|
CC BY
|
2021-01-05 08:21:46
|
no
|
PLoS Biol. 2005 Dec 1; 3(12):e386
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030386
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030398SynopsisDevelopmentEcologyEvolutionMicrobiologyEubacteriaAntisocial Behavior in Cooperative Bacteria (or, Why Can't Bacteria Just Get Along?) Synopsis11 2005 1 11 2005 1 11 2005 3 11 e398Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Exploitative and Hierarchical Antagonism in a Cooperative Bacterium
==== Body
Bacteria are defined as unicellular organisms, but they don't typically function as single cells in nature. Social behavior among bacteria is well established, and makes a lot of sense when you consider that billions of bacteria—representing as many as 1,000,000 species—can be found in just one gram of fertile soil. Cooperative bacteria coordinate a range of complex behaviors through a density-dependent mechanism called quorum sensing: when bacterial numbers reach a critical mass, individual cells secrete signaling molecules that control the behavior of the colony. Through quorum sensing, individual cells amass into biofilms (bacterial colonies that exude slime and other molecules that help them stick to everything from ship hulls to teeth), and some species are able to form structures called fruiting bodies to weather nutrient-poor conditions.
One group of bacterial species, known as the myxobacteria, exhibit several sophisticated social behaviors. They socially swarm and hunt other microbes in a manner analogous to wolf pack hunting. Even more dramatically, when cells of the species Myxococcus xanthus fall upon hard times due to lack of food, some 100,000 individuals band together and form fruiting structures. This process is marked by distinct gene expression programs, differentiation, and morphological changes. Inside the fruiting body, rod-shaped cells differentiate into spherical, stress-resistant spores designed to wait out a famine. But only a portion of the population turns into spores; the vast majority either commit cell suicide, making the ultimate sacrifice, or remain undifferentiated.
But how far does this cooperative behavior go? One species of bacteria can comprise many divergent strains, with different genotypes. It's been shown that when two distinct Myxococcus species are mixed together, the species segregate and form separate fruiting bodies, with one species dominating the other in spore production. Would mixing divergent strains of the same species produce similar results? In a new study, Francesca Fiegna and Gregory Velicer investigated this question using nine strains of the “highly social” ubiquitous soil bacterium M. xanthus isolated from different regions of the world.
To see how divergent strains behave in mixed company, Fiegna and Velicer placed the divergent strains in nutrient-poor cultures, pitting every possible combination of one strain against another. After starving the mixed cultures for five days, the authors observed each pair's fruiting body formation, as well as the spore production of each strain in the mixtures and in isolation.
Some 100,000 Myxococcus xanthus cells amassed into a fruiting body with spores, above. Experimental competitions showed that some strains of this social bacterium exploited others. (Image: Michiel Vos)
The shape, size, and distribution of fruiting bodies were different for nearly every mixed pair relative to their clonal cultures, with most pairs producing fewer fruiting bodies than each strain in isolation. Mixing also decreased the overall social productivity (indicated by total spore production) of the pairs, with some antagonistic pairs reducing total spore production as much as 90%. Even though most strains responded poorly to mixing, some performed better in competition than in isolation—revealing that naturally occurring social bacteria are capable of exploiting their neighbors.
Fiegna and Velicer went on to rank the dominant strains (that is, determined which strain produced the most spores), based on the possible pairing interactions, and showed that their fitness ratings were largely hierarchical, with only one case of a rock-paper-scissors (circular) fitness relationship among any three strains out of 82 such comparisons. This hierarchy suggests that diversity would be quickly lost if all nine strains resided together in one mixed population, with only one strain (or a small number of strains) dominating and eliminating the others over time. Thus, these strains do not tend to act as cooperative subunits when mixed, and M. xanthus as a species has diverged into multiple, distinct social types that cooperate with clone-mates (and perhaps close relatives) but have no qualms about exploiting distant relatives of the same species.
Since M. xanthus can travel great distances carried by water, wind, and an array of animals and insects, the authors conclude, it's possible that resulting antagonisms between introduced foreign strains and resident bacterial populations might decimate some native populations. The degree to which this type of mixing occurs in nature is an active area of research. With the help of whole-genome sequencing and molecular techniques, scientists can refine their traditional morphological classifications of this social soil bacterium to better understand its distribution and likely encounters in soil communities—whether the fitness hierarchies seen here are more typical of mixed distant rather than local strains, for example—and to begin unraveling the molecular agents of subjugation. —Liza Gross
|
0
|
PMC1275526
|
CC BY
|
2021-01-05 08:21:45
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e398
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030398
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030400SynopsisBioinformatics/Computational BiologyEvolutionGenetics/Genomics/Gene TherapyInfectious DiseasesHIV/AIDSHomo (Human)A Well-Studied Disease-Resistance Gene Shows No Signs of Selection Synopsis11 2005 1 11 2005 1 11 2005 3 11 e400Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
The Case for Selection at CCR5-Δ32
==== Body
When our ancestors switched from hunting and gathering to farming about 10,000 years ago, they unwittingly unleashed new selective pressures associated with different diets, increased population density, and novel infectious diseases. This period of rapid change, it is widely thought, also precipitated many genetic adaptations, some conferring resistance to disease. One gene presumed to have undergone positive selection has generated significant interest because of its role in HIV infection. The gene encodes a chemokine receptor called CCR5 on the surface of white blood cells that, along with CD4, mediates HIV entry into the cells.
In 1996, independent research groups discovered a 32–base pair deletion, called ▵32, in the gene's coding region that confers HIV resistance to individuals with two copies of the gene variant, or allele, and delayed AIDS progression to those with one copy. The mutation was relatively common among northern Europeans but virtually absent in non-Caucasians. In a 1998 study, researchers estimated the mutation's age roughly 700 years by analyzing the genetic variation patterns of 192 Caucasian chromosomes. Because it is unlikely for a young mutation to reach a high population frequency by chance alone, it was thought that some selective agent—alternately thought to be bubonic plague and smallpox—had accelerated its spread.
But now Pardis Sabeti, Eric Lander, and their colleagues report that the mutation may be much older than previously thought—and find no evidence of positive selection. The 1998 study based CCR5 ▵32's age, in part, on evidence that the allele was inherited along with two genomic markers, called microsatellites, positioned farther away than would be expected under neutral evolution. Higher than expected linkage disequilibrium (LD)—the distance between linked sequences— suggests the linked sequences are under positive selection. LD shrinks over time because the recombination that occurs during sperm and egg cell development reshuffles sequences around the mutation. Using recently available genome-wide sequence variation data, Sabeti et al. show that the LD and pattern of genetic variation at the allele's locus isn't unusual when compared to the rest of the genome.
For their study, Sabeti et al. analyzed the CCR5 ▵32 polymorphism, two microsatellites, and 70 single nucleotide polymorphisms (SNPs) on Chromosome 3, where the mutation resides, in 340 chromosomes from European, Chinese, and African (Yoruba, Nigeria) populations. They compared the genetic diversity at CCR5 to genomic regions from two large empirical datasets: 2,359 SNPs in 168 immunological genes, located across the genome studied in the same 340 chromosomes, and 63,149 SNPs on Chromosome 3 studied for the same populations from the International Haplotype Map Consortium project. The frequency with which CCR5 variants appeared within and between populations was not unusual compared to the 168 genes or to other regions of Chromosome 3. For the gene to be under selection, it should have shown either decreased or increased diversity within a population or greater variation in distribution among populations. As for CCR5 ▵32's higher frequency in European populations, that isn't unusual either, it turns out: many other polymorphisms found in similar frequencies (7%–9%) in Europeans do not occur in the other populations.
To estimate the LD around CCR5 ▵32, Sabeti et al. first applied the technique used in the 1998 study, and similarly found that chromosomes with the mutation have much longer LD than chromosomes without the mutation. But when the authors analyzed the full range of variation at the CCR5 locus, rather than simply examining ▵32 versus non-▵32 variation, they found two variants with longer LD than ▵32. They further compared the LD around CCR5 ▵32 to LD around other similarly prevalent mutations on Chromosome 3, and found that it was not unusual. Finally, Sabeti et al. used the new sequence variation data to remap the microsatellite markers, and showed that they're much closer to the mutation than previously thought, pushing the mutation's age back to roughly 5,000 years ago.
These results show that genetic variation around CCR5 ▵32 is not so different from the rest of the genome, and find no sign of recent selection for the allele. The absence of evidence is not evidence of absence, but the study raises the important issue that evidence for selection should now be examined and re-examined in a genome-wide context. As more genome-wide datasets become available, scientists will be able to compare the pattern of variation at every gene with overall genomic variation. And by finding the signs of selection in our genes, these tools can point to the evolutionary events that shaped our history and shed light on the genetic roots of disease resistance. —Liza Gross
|
0
|
PMC1275527
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e400
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030400
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030407SynopsisEvolutionMicrobiologyVirusesA Role for Selection in the Evolution of Genetic Robustness Synopsis11 2005 1 11 2005 1 11 2005 3 11 e407Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Evolution of Mutational Robustness in an RNA Virus
==== Body
When Darwin proposed the interplay between variation and natural selection as the driving force of evolution, he had no idea what material produced that variation. Ninety-one years later, Hershey and Chase's famous blender experiment identified the source of variation as DNA. Today, biologists are still struggling to elucidate the details of that interplay.
Natural selection works on genetic variations that produce physical changes in an organism. Think of an organism as a collection of genes (its genotype), and its physical characteristics (its phenotype) as its genotype interfacing with the environment, a natural laboratory that saves or discards a genotype based on the performance of its phenotype. Assuming that a population of genotypes is well adapted to its environment, most mutations are likely to reduce performance in that environment. Thus, populations at equilibrium should experience selection for mechanisms that protect the phenotype against mounting deleterious mutations—a phenomenon called mutational (or genetic) robustness. Most of the evidence that robustness arises from selection rather than chance comes from theoretical studies and from studies of “digital organisms”—computer programs that self-replicate, mutate, and evolve—and has proved difficult to establish in the lab.
RNA phage viruses that co-infect host cells with more fit viruses can withstand high mutation loads with the help of their fitter counterparts—but only for so long
In a new study, Rebecca Montville, Paul Turner, and their colleagues provide experimental evidence for adaptive genetic robustness by working with a mutation-prone virus that infects bacteria, called RNA phage ψ6. (Viruses that infect bacteria are called phages.) Though theoretical predictions for mutational robustness assume that phenotype expression results solely from the underlying genotype, many viruses can overcome their own mutational deficiencies by co-opting the proteins produced by more fit viruses co-infecting the same host, a feature called complementation.
In a previous study, the authors demonstrated for RNA phage ψ6 that complementation buffers less fit viruses against the harmful effects of mutations. They then created six replicate populations of phages and allowed them to adapt to their bacterial host over hundreds of virus generations; three populations evolved at a low ratio of infecting viruses to bacteria (called low multiplicity of infection [MOI]) and three at a high MOI. Populations at high MOI experienced higher rates of co-infection. In this study, the authors investigated the evolutionary consequences of this phenomenon with the hypothesis that selection for mutational robustness should be relaxed for co-infecting phages, since phenotype constancy is bolstered by co-infection with their fitter viral companions.
Using clones from their six replicate populations, Montville et al. generated 60 new lineages and subjected them to a mutation accumulation experiment under conditions that allowed mutations to accumulate at roughly the same probability in high and low MOI lineages. The authors then evaluated the fitness consequences of mutation accumulation on the lineages by comparing their growth rate in the bacteria before and after mutation accumulation. The authors found greater variance in fitness change for the high co-infection lineages compared to the low-infection lineages, supporting the hypothesis that selection for mutational robustness is stronger in the absence of co-infection.
The authors go on to rule out the notion that different fitness effects were produced by different mutation rates in the lineages. Interestingly, the less robust viral genomes were copied more accurately than their more robust counterparts; why less accurate genome replication might accompany the evolution of robustness is a question for future study.
While complementation appears to buffer the damage of mutational onslaughts in the short-term, this benefit of co-infection eventually disappears because the buffer slows the rate that harmful mutations are culled from the virus population. The authors highlight an additional cost of co-infection: by weakening selection for robustness, co-infection may favor the evolution of genomes that are more vulnerable to the harmful effects of mutation. Future work is needed to examine this seeming tug-of-war between short-term and long-term consequences of co-infection, and its impact on the evolution of virus traits. —Liza Gross
|
0
|
PMC1275528
|
CC BY
|
2021-01-05 08:28:16
|
no
|
PLoS Biol. 2005 Nov 1; 3(11):e407
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030407
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030413SynopsisBioinformatics/Computational BiologyGenetics/Genomics/Gene TherapyMolecular Biology/Structural BiologySaccharomycesPreparing for Transcription: The Role of Histone H2A.Z Synopsis12 2005 1 11 2005 1 11 2005 3 12 e413Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Variant Histone H2A.Z Is Globally Localized to the Promoters of Inactive Yeast Genes and Regulates Nucleosome Positioning
==== Body
Every cell in our body contains the instructions for life encoded in around two meters of DNA. In eukaryotic cells (cells with nuclei), all this DNA is squeezed into the cell's nucleus, a region about one-hundredth of a millimeter across. Cells accomplish this improbable task with the help of histones. These proteins combine with DNA to form chromatin, which is made up of structural units called nucleosomes. Nucleosomes, in turn, consist of about 146 base pairs of DNA wrapped around an eight-unit structure containing two molecules each of four core histones—H2A, H2B, H3, and H4. Each nucleosome is separated from its neighbors by a short “linker” DNA. This “beads-on-a-string” arrangement folds into a smooth fiber, which folds into thicker fibers so that all the DNA packs neatly into the nucleus.
Unfortunately, this tidy solution renders chromatin-packaged DNA mostly inaccessible to the transcription machinery. Consequently, cells have devised several ways to adjust the position and/or characteristics of nucleosomes to allow gene expression, including the incorporation of variant histones into nucleosomes. The evolutionarily conserved histone variant H2A.Z (also called Htz1 in yeast) is implicated in transcriptional regulation and gene silencing (inactive, or silent, genes are packaged into dense chromatin called heterochromatin), but little is known about how H2A.Z, which replaces H2A in some nucleosomes, regulates these biological functions. Knowing exactly where in the genome H2A.Z takes the place of H2A should provide insights into how this particular variant histone regulates genome structure and function—which is why Benoît Guillemette and colleagues set out to map H2A.Z binding sites throughout the yeast genome.
The researchers used a technique called chromatin immunoprecipitation to isolate DNA sequences bound to specific histones in living yeast cells, then amplified them, and determined their position in the yeast genome using microarray analysis. They report that H2A.Z binds to 4,862 small regions (which they call Z loci) scattered across the yeast genome. 74% of these regions lie over promoters, regulatory DNA sequences at the start of transcribed genes; 63% of yeast promoters are decorated with H2A.Z. The authors show that H2A.Z specifically associates with one or two nucleosomes within the promoter of some inactive genes but is generally absent from promoters of highly active genes. In addition, they provide evidence that the chromatin structure is more organized in terms of exact positioning of the nucleosomes and that H2A.Z may play a role in that promoter-specific chromatin organization.
The authors also describe the physical pattern of binding of H2A.Z to a second type of region in the genome—Htz1-activated domains (HZADs). These domains are found in euchromatin (loosely packed, actively transcribed chromatin) lying next to heterochromatin, and it is thought that H2A.Z binding stops the spread of silencing into euchromatin, a function called antisilencing. The researchers report that H2A.Z occupies a wider region around HZAD genes than it does at Z loci, indicating that H2A.Z may affect gene transcription and antisilencing through different mechanisms.
Overall, the authors propose that H2A.Z has two roles in yeast cells: to poise genes for transcription initiation in euchromatin and to protect euchromatin from silencing. Whether H2A.Z incorporation into nucleosomes is necessary and sufficient for these activities, and whether it has additional transcriptional effects in yeast and other organisms, is not yet known, but this map of H2A.Z binding sites in the yeast genome will be invaluable in future investigations into the mechanisms of gene regulation. —Jane Bradbury
|
0
|
PMC1275529
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Dec 1; 3(12):e413
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030413
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030420SynopsisAnimal BehaviorNeurosciencePhysiologyMus (Mouse)Music to Her Ears? Male Mice Sing an Ultrasonic Tune Synopsis12 2005 1 11 2005 1 11 2005 3 12 e420Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Ultrasonic Songs of Male Mice
==== Body
Readers of a certain age may remember Mighty Mouse singing, “Here I come to save the day!” as he raced in to save the innocent from evildoers. Fanciful as his cartoon antics may have seemed at the time (it first aired in the 1950s), the TV show's creators weren't so off the mark. There's still no evidence of superhero behavior in real mice, but in a new study, Timothy E. Holy and Zhongsheng Guo show that mice have a gift for song.
Much of what we know about the biology of song and song learning comes from research on songbirds, but birds are difficult subjects for genetic studies. Song commonly figures in courtship rituals among birds, insects, and frogs, but such behavior in mammals had been restricted to whales, bats, and humans. Evidence of similar behavior in the mouse—a long-established genetic model, often referred to as the pocket human—could open whole new avenues of research into the genetic contributions to song and song learning.
To a lay person, the vocal stylings of rodents appear restricted to squeaks and chatter. But mouse social encounters prompt many vocalizations; some are audible to humans, such as the distress calls of pups, and some aren't, such as the ultrasonic calls of males presented with females or urine pheromones. Previous studies characterized the situations that prompt these vocalizations, but did not detail their acoustics. In this study, Holy and Guo went beyond the conditions that prompt rodent discourse to focus on the sounds themselves. Far from random patter, male ultrasonic calls contain complex passages with long sequences composed of diverse syllable types.
Male mice serenade potential partners with ultrasonic song
The authors used cotton swabs coated with either female mouse urine, male mouse urine, or a combination of the two to elicit the male mouse's ultrasonic sounds, and then recorded their vocal responses. The authors manipulated the recordings to hear the ultrasonics. One approach used a slow playback (at one-sixteenth of the recorded speed), but this distorted the temporal structure of the sounds, and the calls sounded like low, intermittent whistles. The other dropped the pitch to an audible level without interfering with the time sequence—the pitch-shifted recording sounds remarkably like birdsong. (To listen, go to DOI: 10.1371/journal.pbio.0030386.sa004.) To bolster this subjective conclusion, the authors then undertook a quantitative analysis of the sounds.
The males produced rapid “chirp-like” syllables of varying duration, spaced at about ten syllables per second, with a burst of closely spaced syllables followed by periods of silence. Some of the syllables showed sudden, significant changes in frequency (or pitch)—all in keeping with previous reports. To determine whether these frequency jumps, or pitch changes, followed a stereotyped pattern or occurred randomly, the authors first analyzed a set of 750 syllables produced by one mouse in a single 210-second trial. They identified discrete clusters of pitch changes, including two clusters with stereotyped jumps to or from a low frequency, which they called low jumps. Repeating the trial and analysis with 45 different mice produced similar results, indicating that the pitch changes are a universal feature of mouse ultrasonic vocalizations.
These pitch jumps formed three distinct categories, which together with two other techniques—borrowed from previous research in birdsong, speech, and pattern recognition—confirmed that syllables are naturally grouped by their pitch changes. The syllables, in turn, occurred in complex sequences that, in some cases, constituted regularly repeated motifs. Syllables with low jumps were repeated in blocks, and phrases tended to be introduced with syllables containing no pitch jumps at all. Since the mice produced multiple syllable types arranged in regular, repeated time signatures, their vocalizations meet the definition of song. Finally, the authors showed that individual males produced songs distinct from those of other males. “The richness and diversity of mouse song appear to approach that of many songbirds,” Holy and Guo write. And just like songbirds, the mice appear to be singing their own tune. Future studies can begin to unravel the physiological basis and mechanics of ultrasonic mouse song—and perhaps decipher the messages encoded in the notes and melody. —Liza Gross
|
0
|
PMC1275530
|
CC BY
|
2021-01-05 08:21:45
|
no
|
PLoS Biol. 2005 Dec 1; 3(12):e420
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030420
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2491622131010.1186/1471-2105-6-249DatabaseSPdb – a signal peptide database Choo Khar Heng [email protected] Tin Wee [email protected] Shoba [email protected] Department of Biochemistry, National University of Singapore, Singapore2 Department of Chemistry and Biomolecular Sciences & Biotechnology Research Institute, Macquarie University, Sydney, Australia2005 13 10 2005 6 249 249 15 4 2005 13 10 2005 Copyright © 2005 Choo et al; licensee BioMed Central Ltd.2005Choo et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The signal peptide plays an important role in protein targeting and protein translocation in both prokaryotic and eukaryotic cells. This transient, short peptide sequence functions like a postal address on an envelope by targeting proteins for secretion or for transfer to specific organelles for further processing. Understanding how signal peptides function is crucial in predicting where proteins are translocated. To support this understanding, we present SPdb signal peptide database , a repository of experimentally determined and computationally predicted signal peptides.
Results
SPdb integrates information from two sources (a) Swiss-Prot protein sequence database which is now part of UniProt and (b) EMBL nucleotide sequence database. The database update is semi-automated with human checking and verification of the data to ensure the correctness of the data stored. The latest release SPdb release 3.2 contains 18,146 entries of which 2,584 entries are experimentally verified signal sequences; the remaining 15,562 entries are either signal sequences that fail to meet our filtering criteria or entries that contain unverified signal sequences.
Conclusion
SPdb is a manually curated database constructed to support the understanding and analysis of signal peptides. SPdb tracks the major updates of the two underlying primary databases thereby ensuring that its information remains up-to-date.
==== Body
Background
Günter Blobel discovered that "proteins have intrinsic signals that govern their transport and localization in the cell" [1]. Proteins synthesised at the ribosome (cytoplasm or rough endoplasmic reticulum), mitochondria or chloroplast are transported to their site of function. This process is known as protein targeting and it depends on targeting signals to direct the proteins to their specific locations.
There are many different classes of targeting signals. One of the commonly occurring signals is formed by short, transient peptides known as signal peptides or leader sequences, which are usually found at the amino terminus of secreted proteins. Signal peptides are present in both prokaryotic and eukaryotic cells, indicating its ancient universal origins. They function like a postal address label on an envelope by targeting the proteins for secretion or to specific organelle for further processing. The signal peptides are cleaved off and degraded upon reaching their targeted locations. Interestingly, not all proteins possess signal peptides [2,3], suggesting that other mechanisms for protein targeting exist.
Over the years, several prediction tools [4-10] have been developed to predict the cleavage sites of signal peptides. These prediction tools require training and testing datasets. As a preparatory step for prediction work, researchers often devote considerable time sifting through primary databases such as Swiss-Prot [11], EMBL [12] and other databases to collate and construct their own datasets. This repetitive process can and should be eliminated by creating a centralised repository of signal peptide sequences.
Searching through popular search engines and reviewing the Nucleic Acids Research database list [13] reveal several databases that provide information on protein subcellular localisation [14,15], nuclear proteins [16] and secreted proteins [17]. These databases do not provide signal peptide specific information except for SPD [17]. SPD or secreted protein database [17] is a collection of proteins from the human, mouse and rat proteomes originating from databases such as TrEMBL [18], Ensembl [19] and Refseq [20]. It also includes datasets from the Secreted Protein Discovery Initiative (SPDI) [21], a large-scale effort to identify novel human secreted and transmembrane proteins; the Riken mouse secretome and seven other related datasets [22]. SPD aims to be a comprehensive repository for secreted proteins, but it suffers from providing datasets that may still contain many erroneous annotations from its underlying data sources for example TrEMBL. TrEMBL is generated from an automated pipeline and has yet to undergo manual curation. In addition, the entries in SPD were not checked manually against the publications. Then, there is also the issue that the datasets are not being updated.
Besides the SPD which offer downloadable datasets, there are sites that offer downloadable datasets namely the SignalP datasets (1997) [10,23] and the datasets (2000) used by Meene et al. [24,25] in their evaluation of signal peptide prediction methods. More recently, there is the dataset consisting of 270 secreted recombinant human proteins with experimentally determined cleavage sites from Zhang and Henzel [26,27]. These datasets are often either limited in size or otherwise lacking in tools for querying the datasets. Moreover, these datasets although valuable but they are often outdated [28] especially when GenBank/EMBL, Swiss-Prot and other publicly accessible primary databases continue to churn out new entries or sequences.
Many researchers are confronted by similar obstacles in accessing up-to-date data, which are withheld from public access by method developers [29,30] and hence, we strongly believe that there is a urgent need to provide a publicly-accessible, manually curated and regularly updated database specialised for signal peptides. These datasets will not only be important for prediction work but they will also serve as the common datasets needed when researchers are performing benchmarking against each other methods or programs, without which we think it is difficult to perform proper or fair benchmarking of the multitudes of prediction methods.
Construction and content
Construction and implementation of database
Recognizing the need for a curated, specialised and up-to-date database, we have developed a composite signal peptide database SPdb [31] that offers researchers a singular point for depositing their signal peptide annotations. SPdb integrates information from Swiss-Prot (part of UniProt) [18] and EMBL. It is updated when there is a major release of Swiss-Prot. First released in May 2004, SPdb was upgraded to release 3.2 recently to add on new features and to synchronise with the release of its underlying data sources.
SPdb is a relational database built using MySQL database management system [32] and using PERL/CGI [33] for processing web forms. An easy-to-navigate web interface was built to allow user to search through the database. Some of the web features were added in response to the requests by some of the users that have used our database since its inception. Through the web interface, users are able to download the returned results as FASTA formatted files or view the results as HTML web page. We have also provided a link from the search page to the Swiss-Prot ID tracker to verify whether an entry has been renamed e.g. ANL3_HUMAN from the Zhang and Henzel dataset [27], which is now ANGL3_HUMAN.
We deployed the bioinformatics pipeline shown in (Figure 1) to construct SPdb. The pipeline to construct the database was semi-automated with specific checkpoints for manual checking of the results to minimise errors in the database.
Construction method
Signal sequences and coding sequences obtained from Swiss-Prot (TrEMBL entries were not taken into consideration) were filtered initially using the data extraction and redundancy reduction methodology proposed by Nielsen et al. [34] to segregate the dataset into two sets (a) the preliminary filtered set and (b) the unverified sequences set. The Nielsen et al. [34] methodology has been employed to generate the training and testing data used in SignalP [10,35]. We adapted and omitted some of the criteria proposed by the method since our goal is to build a repository of signal peptides with as many relevant and accurate entries as possible. We observed that the proposed methodology still renders many undesirable entries upon the filtering process. Thus, we have constructed SPdb by building on the strength of the proposed methodology [34] and improved it with our own criteria and filtering rules.
Any entries with the SIGNAL keyword indicated in the feature table FT field [36] of Swiss-Prot entries were presumed to contain information on signal sequence. This simple selection process yielded 18,146 entries out of the total 170,140 Swiss-Prot entries (Release 46.1). Entries that connoted uncertainty namely those with annotations like PROBABLE, POTENTIAL, BY SIMILARITY, HYPOTHETICAL and entries with ambiguous cleavage or signal peptide positions were tagged as unverified sequences. Then, entries with signal sequences length less than 11 were relegated to the unverified sequences set. Signal sequences are generally considered to be of length 15 to 40. This initial step filtered off 13,701 entries from the preliminary filtered set leaving behind 4,445 entries. These entries include type I signal peptides, type II signal peptides (lipoproteins) and TAT-containing signal peptides. Using the SIGNAL keyword, mitochondria and chloroplast transit peptides were excluded from the preliminary set since transit peptides are identified by the TRANSIT keyword in Swiss-Prot.
We proceeded to integrate information from the EMBL database. By integrating complementary information, besides providing extra information not found in Swiss-Prot, we could use the information from EMBL to cross check against Swiss-Prot, allowing us to discover erroneous annotations. This practice of using complementary information from other data sources has been found useful in data evaluation [37].
The first cross-reference entry to EMBL database was used for the respective Swiss-Prot entry. Based on the data categorisation of EMBL found in its release note [38], only sequences from the data groups fungi, human, invertebrate, mouse, organelle, bacteriophage, plant, prokaryote, rodent, viral, mammals and vertebrate were taken into consideration. Entries belonging to the data groups expressed sequence tags, genome survey sequences, high-throughput genome sequences, unfinished DNA sequences generated by high-throughput sequencing, patent sequences, synthetic sequences, contig sequences and unclassified were omitted. We extracted out relevant annotations from EMBL whenever available including coding sequence, signal sequence and its length, subcellular location, authors' notes and so on.
The annotations, specifically the sig_region and misc fields from the EMBL entry were utilised in the subsequent step to cross-check against the preliminary filtered entries. This step again filtered out many inconsistent entries where the positions are quoted wrongly by either source e.g. [Swiss-Prot:CD166_CHICK] where Swiss-Prot quoted cleavage position of 33 while EMBL provided 32. As a result, another 866 entries were eliminated to retain 3,579 entries in this newly filtered Swiss-Prot/EMBL set. It must be noted that there were some Swiss-Prot entries in the Swiss-Prot/EMBL set without any EMBL reference e.g. [Swiss-Prot:APOE_CAVPO]; or with insufficient annotations in the EMBL entries e.g. [Swiss-Prot:17KD_RICAU]; or their EMBL cross-references were indicated with annotation such as NOT_ANNOTATED_CDS e.g. [Swiss-Prot:2B31_HUMAN], ALT_TERM e.g. [Swiss-Prot:CD1E_HUMAN], ALT_INIT e.g. [Swiss-Prot:1A03_PANTR] and ALT_SEQ e.g. [Swiss-Prot:17KD_RICPR]. In these cases, all these entries were earmarked for manual curation. These terms "NOT_ANNOTATED_CDS", "ALT_TERM" and so on are known as status identifiers and they are found at the DR field in Swiss-Prot entries. The reader is referred to the detailed explanation found in the Swiss-Prot manual [39].
Following this step, all the entries in this Swiss-Prot/EMBL set are manually checked against the referred publications. We located numerous entries with discrepancies on signal peptide cleavage site between the Swiss-Prot annotations and the accompanying papers e.g. [Swiss-Prot:CECC_DROME, Swiss-Prot:AMCY_PARVE]. Entries that we do not have access to the accompanying papers e.g. [Swiss-Prot:ZEAL_MAIZE, Swiss-Prot:ZEA6_MAIZE] or those entries that we could not locate their cleavage site information in the papers e.g. [Swiss-Prot:GUX1_TRIRE]; these entries in addition to those entries which are inadequately labelled or entries with inconsistent positional information were all relegated to the unverified sequences set. In this manual curation step, we eliminated 995 entries from the Swiss-Prot/EMBL set of 3,579 entries. These 995 entries were entries that (a) both Swiss-Prot and the quoted papers provided the same putative position (b) we found differing positions quoted by Swiss-Prot as compared to the quoted papers (c) we did not have access to the quoted subscription-only papers or the papers referred to were old and in some cases there were no paper or no relevant paper quoted (d) we could not find or locate the cleavage site information (Table 1).
Table 1 Distribution of the signal sequences filtered out in the manual curation step
Description No. of Entries/Sequences
Swiss-prot and the accompanying papers quoted same putative position 311
Swiss-Prot and the accompanying papers quoted different position;
The position quoted maybe confirmed or putative 100
No references or relevant references were provided;
No access to some subscription-only papers;
No access to some very old papers 194
Unable to locate or obtain the position information from the papers 390
TOTAL 995
Content of database
The result from filtering and manually curating the entries culminated in the latest release of SPdb release 3.2, with a total of 18,146 signal sequence entries, out of which 2,584 are filtered sequences (Table 2). These filtered sequences, known as the filtered sequences set include the mature endogenous proteins that were sequenced on their N-terminal and have been checked against the accompanying reference paper to be considered as experimentally verified positions. The remaining 15,562 unverified sequences contain putative or experimentally unverified cleavage site signal sequences. This unverified set also contains entries with erroneous database annotations. It is worth noting that this unverified set might contain some experimentally verified signal sequences since we may not have access to the accompanying papers.
Table 2 Distribution of signal sequences in SPdb according to archaea (AR), bacteria (BA), viruses (VR) and eukaryotes (EU).
AR BA EU VR SUB-TOTAL
Verified sequences 7 540 1,945 92 2,584
Unverified sequences 101 3,528 11,239 694 15,562
TOTAL 108 4,068 13,184 786 18,146
With the two primary databases integrated, SPdb contains four data groups namely archaea, bacteria, eukaryotes and viruses and (Table 2). SPdb provides key extracted information (Figure 2) such as organism source, organelle, subcellular location and other accompanying important notes. For full annotation, cross-referenced links to the originating database are provided. The signal peptide cleavage site is explicitly marked if such information is available. The signal peptide sequences and 30 residues [40] after the cleavage site are colour-coded using the convention as specified by RasMol amino acids colour scheme [41] which is based on the traditional amino acid properties. In the process of manually curating the 3,579 entries, we have added our own annotation for the 995 entries that were removed from this dataset later on.
Utility and discussion
SPdb provides users with an easy-to-use web interface with flexibility to select for an entry or a collective set of entries matching users' criteria such as name of organism, data group, length of signal sequences, keyword searches and more importantly the option to choose between including or excluding certain entries. We have taken the approach to allow users to omit or filter any sequences since every user may have different requirements on the returned results. Each of the entry will be indicated whether it is verified or unverified (Figure 2).
In the process of creating SPdb, we realise that although Swiss-Prot provides better quality annotation, it still contains erroneous or conflicting annotations as evident when we compare the positions or length of the signal sequences reported by Swiss-Prot with EMBL e.g. [Swiss-Prot:A2AP_HUMAN, Swiss-Prot:BTD_HUMAN]. We notice that the inconsistencies usually arise when there is more than one reference. The referred papers may quote different positions thus this may have caused the confusion. To help to resolve this issue, we have combined the annotations from -EMBL, and we managed to identify and filter off many such entries. The annotations on signal peptide found in EMBL were mostly accurate though there were cases when the information was reported incorrectly as well e.g. [EMBL:M19077] in [Swiss-Prot:CHR1_BOMMO]. In this respect, we have included a link in each entry for users to report to us if they encounter any errors or discrepancies in SPdb.
Apart from the errors and inconsistencies just described, there was also the issue on experimental support from the journal publication. Many of the entries with annotation on signal sequences positions or length were predicted or deemed putative or potential (Table 1) by the researchers when they reported the positions in their papers. Nonetheless, the entries were not labelled with words like POTENTIAL, BY SIMILARITY and PROBABLE in the Swiss-Prot entry as previously assumed. We learnt that many of the referred papers were using prediction or sequence alignment software to identify or suggest the cleavage site of signal sequences. Therefore, we think it would be more appropriate and useful if the references to the papers were also indicated at the relevant fields so that any users of the entries can easily check and read up on the papers that mentioned about signal peptides or any other features.
All these issues and problems have made automation of the construction of SPdb immensely difficult if not remotely impossible. Prior to manually curating the entries, we have considered using text-mining approach but we forwent the method eventually when we discovered that many of abstracts did not contain the cleavage site information rather the information was found in the body of the paper, usually located under the results or discussion section. Moreover, the words or phrases used to express the positional information were also varying and difficult to express as extraction rules e.g. in the paper [42] quoted in entry [Swiss-Prot:PRRP_BOVIN], we encountered this sentence "... its N-terminal portion before Ser-23 showed the typical profile of a secretory signal peptide ...". Then there is also the problem where many of the papers require subscriptions, rendering the extraction program useless unless we can obtain the papers. Unless each of the paper submitted in future provides a short note on the features of the proteins described coupled with the improvement in text-mining accuracy, we will have to resort to manual curation.
In SPdb, datasets are classified into filtered sequences and unverified sequences. By classifying the entries into these two classifications, researchers can use them in the work of machine learning approaches, where datasets are sought after as training and testing sets in signal peptide cleavage site prediction.
Apart from facilitating test datasets, SPdb provides other information such as amino acid composition of the protein which have been suggested to correlate with the subcellular localisation of the protein [43]; amino acid residues properties (aromatic, non-polar, polar, charged and so on) are shown in graphical format to indicate which residues possess the properties visibly; also accompanying each entry are the hydropathy plots based on Kyte and Doolittle [44], Sweet and Eisenberg [45], Eisenberg et al. [46] of the signal sequences and the sequences downstream of the signal sequence cleavage. The plots are rendered using pepinfo within the computational analysis package of EMBOSS [47], an open source software suite for sequence analysis. Each signal peptide exhibits three distinct regions at the sequence level: the n-region (a positive charged region), the h-region (hydrophobic region) and the c-region (polar and neutral region) [9]. The hydropathy plots help to visualize and identify these regions.
De Gier et al. [48] showed that signal peptide processing by the signal recognition particle (SRP) requires certain contextual cues in the sequence downstream. SRP binds to N-terminus signal or signal-anchor sequence when the nascent polypeptide chain is synthesised by the ribosome up to ~60 amino acid residues. At this length, this segment is conveniently exposed and translation will resume upon dissociation of SRP from the nascent chain. In the effort to capture this information for the co-translational translocation mechanism, SPdb includes both the signal peptide sequences and 30 residues after the cleavage site.
For the future releases, we hope to include other information which maybe useful such as functional classification of signal peptides according to target destination, the profiles of signal peptides from various organisms and so on. As differentially targeted organelles or locations have variations on the general theme of signal peptide target proteins, we would like to include these different targeting signals for comparison and studies. Concerning secreted proteins that lack cleavable signal peptide [49] e.g. ovalbumin, a secreted glycoprotein and the major protein of egg-white which does not have a cleavable signal peptide [50], we would like to include this information and analyse how they differ from those proteins with cleavable signal peptide.
Conclusion
Signal peptide plays an important role in the transport of secretory proteins. Understanding of signal peptide recognition and mechanisms of targeting, transport and translocation will unleash many applications in the area of drug design and medicine. We have provided a freely accessible, manually curated signal peptide database that is regularly updated and synchronised with the release of the two major primary databases, Swiss-Prot and EMBL. By integrating information from both databases, SPdb is able to eliminate some of the discrepancies and minimise the errors found in the sequence entries, thereby providing a better quality of the downloadable datasets that can be used by the research community for prediction work and other research.
Availability and requirements
SPdb is freely accessible through the website . We have made available a dedicated page to allow user to download the dataset in full based on certain criteria available to user.
Authors' contributions
KHC built the database pipeline and the web interface. SR developed the signal peptide project and provided comments and suggestions on the features of the database while TWT provided assistance for the database design and the manuscript.
Figure 1 Schematic diagram of the construction pipeline of SPdb.
Figure 2 SPdb entry information includes a short description of the protein, the hydropathy plots and amino acids properties and more. (A) Each entry is marked as verified or unverified, with (B) a "report-error" link for users to inform us on any error or updated information pertaining to an entry for us to rectify/update. (C) users can deposit their signal sequences with us and add on their own annotation.
Acknowledgements
We would like to acknowledge Vivek Gopalan (while at the Dept. of Biochemistry, NUS) and the anonymous reviewers for their comments and suggestions. We also thank our users who have mailed us to provide their support, encouragement and comments.
==== Refs
Nobel Prize in Physiology or Medicine 1999
Bowden GA Baneyx F Georgiou G Abnormal fractionation of beta-lactamase in Escherichia coli: evidence for an interaction with the inner membrane in the absence of a leader peptide J Bacteriol 1992 174 3407 3410 1577708
Flower AM Doebele RC Silhavy TJ PrlA and PrlG suppressors reduce the requirement for signal sequence recognition J Bacteriol 1994 176 5607 5614 8083155
SIG-Pred: Signal Peptide Prediction
SIGFIND - Signal Peptide Prediction Server (Eukaryotes)
Hiller K Grote A Scheer M Munch R Jahn D PrediSi: prediction of signal peptides and their cleavage positions Nucleic Acids Res 2004 32 W375 9 15215414
Juncker AS Willenbrock H Von Heijne G Brunak S Nielsen H Krogh A Prediction of lipoprotein signal peptides in Gram-negative bacteria Protein Sci 2003 12 1652 1662 12876315 10.1110/ps.0303703
Kall L Krogh A Sonnhammer EL A combined transmembrane topology and signal peptide prediction method J Mol Biol 2004 338 1027 1036 15111065 10.1016/j.jmb.2004.03.016
Nielsen H Engelbrecht J Brunak S von Heijne G Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites Protein Eng 1997 10 1 6 9051728 10.1093/protein/10.1.1
Bendtsen JD Nielsen H von Heijne G Brunak S Improved prediction of signal peptides: SignalP 3.0 J Mol Biol 2004 340 783 795 15223320 10.1016/j.jmb.2004.05.028
Bairoch A Boeckmann B Ferro S Gasteiger E Swiss-Prot: juggling between evolution and stability Brief Bioinform 2004 5 39 55 15153305 10.1186/1471-2105-5-39
Kanz C Aldebert P Althorpe N Baker W Baldwin A Bates K Browne P van den Broek A Castro M Cochrane G Duggan K Eberhardt R Faruque N Gamble J Diez FG Harte N Kulikova T Lin Q Lombard V Lopez R Mancuso R McHale M Nardone F Silventoinen V Sobhany S Stoehr P Tuli MA Tzouvara K Vaughan R Wu D Zhu W Apweiler R The EMBL Nucleotide Sequence Database Nucleic Acids Res 2005 33 Database Issue D29 33 15608199
NAR Database Category List - Protein localization and targeting
PSORTdb
DBSubLoc
Nuclear Protein Database (NPD)
Secreted Protein Database (SPD)
Bairoch A Apweiler R Wu CH Barker WC Boeckmann B Ferro S Gasteiger E Huang H Lopez R Magrane M Martin MJ Natale DA O'Donovan C Redaschi N Yeh LS The Universal Protein Resource (UniProt) Nucleic Acids Res 2005 33 D154 9 15608167 10.1093/nar/gki070
Birney E Andrews TD Bevan P Caccamo M Chen Y Clarke L Coates G Cuff J Curwen V Cutts T Down T Eyras E Fernandez-Suarez XM Gane P Gibbins B Gilbert J Hammond M Hotz HR Iyer V Jekosch K Kahari A Kasprzyk A Keefe D Keenan S Lehvaslaiho H McVicker G Melsopp C Meidl P Mongin E Pettett R Potter S Proctor G Rae M Searle S Slater G Smedley D Smith J Spooner W Stabenau A Stalker J Storey R Ureta-Vidal A Woodwark KC Cameron G Durbin R Cox A Hubbard T Clamp M An overview of Ensembl Genome Res 2004 14 925 928 15078858 10.1101/gr.1860604
Pruitt KD Tatusova T Maglott DR NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins Nucleic Acids Res 2005 33 D501 4 15608248 10.1093/nar/gki025
Clark HF Gurney AL Abaya E Baker K Baldwin D Brush J Chen J Chow B Chui C Crowley C Currell B Deuel B Dowd P Eaton D Foster J Grimaldi C Gu Q Hass PE Heldens S Huang A Kim HS Klimowski L Jin Y Johnson S Lee J Lewis L Liao D Mark M Robbie E Sanchez C Schoenfeld J Seshagiri S Simmons L Singh J Smith V Stinson J Vagts A Vandlen R Watanabe C Wieand D Woods K Xie MH Yansura D Yi S Yu G Yuan J Zhang M Zhang Z Goddard A Wood WI Godowski P Gray A The secreted protein discovery initiative (SPDI), a large-scale effort to identify novel human secreted and transmembrane proteins: a bioinformatics assessment Genome Res 2003 13 2265 2270 12975309 10.1101/gr.1293003
SPD and its collated related datasets
SignalP Training Datasets (1997)
Test sets used in evaluating the different signal prediction methods
Menne KM Hermjakob H Apweiler R A comparison of signal sequence prediction methods using a test set of signal peptides Bioinformatics 2000 16 741 742 11099261 10.1093/bioinformatics/16.8.741
Signal peptide prediction based on analysis of experimentally verified
cleavage sites - supplementary data
Zhang Z Henzel WJ Signal peptide prediction based on analysis of experimentally verified cleavage sites Protein Sci 2004 13 2819 2824 15340161 10.1110/ps.04682504
Datasets used in SignalP and provided for download
Pennisi E Keeping genome databases clean and up to date Science 1999 286 447 450 10577208 10.1126/science.286.5439.447
Wiley HS Michaels GS Should software hold data hostage? Nat Biotechnol 2004 22 1037 1038 15286656 10.1038/nbt0804-1037
Signal Peptide Database (SPdb)
MySQL database management system
Perl Programming Language
Nielsen H Engelbrecht J von Heijne G Brunak S Defining a similarity threshold for a functional protein sequence pattern: the signal peptide cleavage site Proteins 1996 24 165 177 8820484 10.1002/(SICI)1097-0134(199602)24:2<165::AID-PROT4>3.0.CO;2-I
SignalP - A Popular Signal Peptide Prediction Server
Swiss-Prot Manual (FT Field)
Bork P Powers and pitfalls in sequence analysis: the 70% hurdle Genome Res 2000 10 398 400 10779480 10.1101/gr.10.4.398
EMBL Database Release Note
Swiss-Prot Manual (DR Field)
Andersson H von Heijne G A 30-residue-long "export initiation domain" adjacent to the signal sequence is critical for protein translocation across the inner membrane of Escherichia coli Proc Natl Acad Sci U S A 1991 88 9751 9754 1946398
RasMol amino colour scheme
Hinuma S Habata Y Fujii R Kawamata Y Hosoya M Fukusumi S Kitada C Masuo Y Asano T Matsumoto H Sekiguchi M Kurokawa T Nishimura O Onda H Fujino M A prolactin-releasing peptide in the brain Nature 1998 393 272 276 9607765 10.1038/30515
Nakashima H Nishikawa K Discrimination of intracellular and extracellular proteins using amino acid composition and residue-pair frequencies J Mol Biol 1994 238 54 61 8145256 10.1006/jmbi.1994.1267
Kyte J Doolittle RF A simple method for displaying the hydropathic character of a protein J Mol Biol 1982 157 105 132 7108955 10.1016/0022-2836(82)90515-0
Sweet RM Eisenberg D Correlation of sequence hydrophobicities measures similarity in three-dimensional protein structure J Mol Biol 1983 171 479 488 6663622 10.1016/0022-2836(83)90041-4
Eisenberg D Weiss RM Terwilliger TC The helical hydrophobic moment: a measure of the amphiphilicity of a helix Nature 1982 299 371 374 7110359 10.1038/299371a0
EMBOSS : Sequence analysis application suite
de Gier JW Scotti PA Saaf A Valent QA Kuhn A Luirink J von Heijne G Differential use of the signal recognition particle translocase targeting pathway for inner membrane protein assembly in Escherichia coli Proc Natl Acad Sci U S A 1998 95 14646 14651 9843943 10.1073/pnas.95.25.14646
Ye RD Wun TC Sadler JE Mammalian protein secretion without signal peptide removal. Biosynthesis of plasminogen activator inhibitor-2 in U-937 cells J Biol Chem 1988 263 4869 4875 3127394
Lingappa VR Lingappa JR Blobel G Chicken ovalbumin contains an internal signal sequence Nature 1979 281 117 121 471058 10.1038/281117a0
|
16221310
|
PMC1276010
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Oct 13; 6:249
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-249
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-841622569110.1186/1471-2334-5-84Research ArticleUse of "biokit HSV-2 Rapid Assay" to improve the positive predictive value of Focus HerpeSelect HSV-2 ELISA Morrow Rhoda Ashley [email protected] David [email protected] Amalia [email protected] Lawrence [email protected] Department of Laboratory Medicine, University of Washington, Seattle, Washington, USA2 Epidemiology, University of Washington, Seattle, Washington, USA3 Medicine, University of Washington, Seattle, Washington, USA4 Childrens Hospital and Regional Medical Center; Seattle, Washington, USA5 Fred Hutchinson Cancer Research Center; Seattle, Washington, USA2005 14 10 2005 5 84 84 26 5 2005 14 10 2005 Copyright © 2005 Morrow et al; licensee BioMed Central Ltd.2005Morrow et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Commercially available assays to detect antibodies to the herpes simplex virus type 2 (HSV-2)-specific glycoprotein gG-2 have markedly improved serologic diagnosis of HSV-2 infection. However, even tests with high specificity can have low positive predictive values in low prevalence populations. HSV-2 is a chronic, life-long viral infection that requires both medical attention and potential alterations in health care strategy. As such, the concern for false positive diagnoses is high confirmatory testing is routine for other viral serologies such as HIV and hepatitis C. We evaluated such a strategy for HSV-2 serology by using an easily performed commercial test, biokitHSV-2 rapid test ("Biokit"; Biokit USA, Lexington MA) as a confirmatory test for the widely used gG-2 specific serology ("Focus;" HerpeSelect HSV-2 ELISA; Focus Diagnostics, Cypress CA).
Methods
We tested 782 sera by Focus HSV-2 ELISA, Biokit, and the current gold standard test, Western blot (WB).
Results
The positive predictive value of the Focus HSV-2 ELISA increased from 80.5% to 95.6% when Biokit testing was performed on sera that were initially positive by Focus HSV-2 ELISA. Confirmatory testing increased the specificity markedly among sera with Focus EIA values between 1.1 and 3.5: only 35% of low positive (index values 1.1–3.5) Focus HSV-2 ELISA results confirmed as positive by Biokit and WB compared with 92% of those with index values >3.5. Mathematical modeling of the data resulted in expected positive predictive values over 98% for populations with antibody prevalences typical of clinical practices in the US and Europe.
Conclusion
Confirmatory Biokit testing of positive Focus HSV-2 ELISA results is fast, easy, and effective in reducing falsely positive HSV-2 antibody results. Patients, clinicians, and laboratories could benefit from the enhanced specificity of this simple HSV-2 serologic test combination.
==== Body
Background
Several studies over the last decade have shown the importance of subclinical HSV-2 reactivation in the epidemiology of HSV-2 infection. Over 95% of persons who are HSV-2 seropositive will reactivate and shed HSV-2 from genital sites and 70% of sexual and maternal-fetal transmission occurs from such subclinical shedding. As such, serologic detection of past HSV-2 increasingly is being recommended for a variety of immunocompetent and immunosuppressed populations. Several enzyme linked immunoassays for HSV-1 and HSV-2 antibodies to the type-specific glycoproteins, gG-1 and gG-2, respectively, are approved by the U.S. Food and Drug Administration. These methods are cost effective, widely available, and are the only commercial methods that accurately differentiate HSV-1 from HSV-2 antibodies. The HerpeSelect HSV-2 gG2 ELISA test (Focus Diagnostics) demonstrated a sensitivity of 96% in a group of pregnant women and 95% in an STD population of men and women [1]. Specificity of Focus HSV-2 ELISA also was high in these groups: 97 % in pregnant women and 96% in the STD population [1]. Both groups had relatively high HSV-2 seroprevalence by Western blot (WB); 25% of the pregnant women and 22% of the STD group had antibodies to HSV-2.
However, in select patient groups from several African countries, the Focus HSV-2 ELISA may give falsely positive results when compared with other gG-based tests such as the gG-2 monoclonal antibody inhibition assay [2] or WB [3,4]. A recent study of a low prevalence population suggests that falsely positive tests may not be limited to African populations [5]. All of these studies have found that Focus HSV-2 ELISA false positive results are far more likely with sera that have index values in the low positive range (1.1–3.5) than those that have index values above 3.5 [4,6]. As such, a confirmatory test to improve test specificity is desirable.
In 2000, a gG-2-based point of care membrane test, POCkit-HSV-2, was cleared by the US Food and Drug Administration for use with capillary blood and sera. This test showed high sensitivity and specificity in premarket trials against WB [7,8]. This test is now available as "biokitHSV-2 Rapid Test" from Biokit USA or as "SureVue-HSV-2" Rapid Test from Fisher HealthCare, Houston, TX. The biokitHSV-2 Rapid Test ("Biokit") is a readily accessible alternative to WB for confirmatory testing and can be performed easily on sera within a few minutes. Performing Biokit tests on sera that are initially positive by Focus HSV-2 ELISA could provide a useful strategy to increase the specificity of this HSV-2 serology.
To assess the value of biokit-HSV-2 as a confirmatory assay after an initial screening by Focus HSV-2 ELISA, we selected two sets of sera to study: 1) one from men at high risk for genital herpes and 2) one from an all-comer group of sera received by the University of Washington laboratory for HSV antibody testing. Biokit results were the same as WB results in 93.7% of these sera. Concordance of WB and Focus HSV-2 ELISA was 88.9%; concordance of Biokit and Focus HSV-2 ELISA was 86.7%. Using the Biokit result for sera positive by Focus HSV-2 ELISA increased the specificity from 93.2% to 98.7%. Positive predictive values increased from 80.5% for Focus HSV-2 ELISA to 95.6% when Biokit results were applied to sera that were positive by Focus HSV-2 ELISA.
Methods
Serology
Focus HerpeSelect HSV-2 ELISA ("Focus HSV-2 ELISA"; Focus Diagnostics, Cypress CA) was performed on each serum according to kit instructions. Sera with index values <0.9 were considered negative, those >3.5 as positive, values .9–1.1 (inclusive) were considered equivocal. Index values >1.1 to 3.5 were considered low positive.
The biokitHSV-2 Rapid Assay ("Biokit") Biokit USA, Lexington, MA) was performed according to kit instructions. Positive results were those in which the test spot was clearly colored red or pink. Negative results were those that had very faint or no color on the test spot. In a few cases, a colored ring appeared around an uncolored spot. These were scored negative.
The Western blot assay ("WB") for HSV-1 and HSV-2 was performed as described previously [9].
Study subjects
Two groups were studied: 1) High-risk population: Through Dec 1, 2004, 1125 adult "men who have sex with men" (MSM) were tested by Focus HSV-2 ELISA to screen for enrollment into a National Institutes of Health trial evaluating acyclovir therapy to reduce acquisition of HIV-1 infection among HSV-2 seropositive persons (HIV Prevention Trial Network Protocol 039). As such, this serum subset was biased toward men who felt they had previously acquired HSV-2. Of the 388 sera found to be positive by Focus HSV-2 ELISA, 301 (77.6%) had sufficient volume to be used in the comparison study. Of the 718 that were seronegative by Focus HSV-2 ELISA, every 10th sample with sufficient volume (N = 67) was selected to test by Biokit and WB. Sera that were indeterminate by Focus HSV-2 ELISA (N = 19) or WB (N = 14) were not tested by Biokit.
2) All-comer sera: We also tested 624 consecutive sera submitted for HSV WB testing to the University of Washington Virology Laboratory during a 4 week period in 2004. Of all subjects whose sera were received in 2004, 10% were under 20 years of age; 86% were 21–60 years old and 3% were over 60 years of age. Thus, some of our sample was likely to be pediatric. Of 148 sera that were positive by Focus HSV-2 ELISA, 141 were used in the comparison study. We selected 273 of the 469 samples that were negative by Focus HSV-2 ELISA for continued testing. Sera that were equivocal by Focus HSV-2 ELISA (N = 6) or by WB (N = 22) or by both tests (N = 1) were not analyzed further.
Samples from both groups were stripped of identifiers before performing the tests for this study. After de-identifying, each serum was stored at -20C and was thawed once for this study.
Statistical measures
All Focus positive and low positive samples with sufficient volume were tested by WB and Biokit. Only a subset of Focus negative sera were run by WB or Biokit. Therefore, raw estimates of sensitivity, specificity and other predictive values were expected to be biased [10]. To adjust for this bias, we calculated expected confirmation rates among unconfirmed Focus negative results and incorporated these hypothetical results. Estimation was performed conservatively using binomial theory, first using positive rates for WB and Biokit in confirmed data to compute endpoints of a 95% confidence interval for the hypothetical number confirmed positive, then applying the extremes of this range to form two potential confidence intervals for each accuracy measure. A maximally-wide interval was built by combining these.
Results
HSV-2 prevalence of the study populations was between 23.7 and 34.5% by Focus HSV-2 ELISA (Table 1). Overall, one-third of the HSV-2 seropositive sera had low positive results.
Table 1 HSV-2 status of study populations by Focus HSV-2 ELISA
Number of Sera (% of Group Total)
MSM Group All-Comer Group All Subjects
HSV-2 Status (Index value) Population Sera Tested Population Sera Tested Population Sera Tested
Positive
(All >1.1) 388 (34.5) 301 148 (23.7) 141 536 442
(Only >3.5) 245 (21.8) 194 110 (17.6) 106 355 300
(Only 1.1 – 3.5) 143 (12.7) 107 38 (6.1) 35 181 142
Equivocal (.9 – 1.1) 19 (1.7) 0 7 (1.1) 0 26 0
Negative (<0.9) 718 (63.8) 67 469 (75.2) 273 1187 340
Total 1125 368 624 414 1749 782
Because we were interested in evaluating the ability of the Biokit assay to serve as a confirmatory test for positive Focus HSV-2 ELISA samples, all seropositive sera with sufficient volume were used for subsequent WB and Biokit testing. In particular, 142 (78.5%) of the 181 low positive sera and 300 (84.5%) of the 355 high positive samples were tested by WB and Biokit. A subset of 340 (28.6%) of the 1187 seronegative sera by Focus HSV-2 ELISA was tested further (Table 1).
Effect of index value on confirmation
The proportion of sera that was positive by Focus HSV-2 ELISA and either WB or Biokit rose from <12–15% of sera with initial index values of 1.1–1.5 to >90% for index values >3.5 (Figure 1). The proportion of sera that confirmed at each index value level was very similar between Biokit and WB (Figure 1).
Figure 1 Western blot and Biokit results sorted by Focus HSV-2 ELISA index value. The proportion of sera that were positive by western blot (hatched bars) or by biokitHSV2 Rapid Test (solid bars) is shown for each range of index values obtained by Focus HerpeSelect HSV-2 ELISA. Numbers of sera (N) contributing to each subset are given below the designated index value ranges.
Concordance among the 3 tests
The 3 tests had the same result in 662 (85%) of the 782 study sera (Figure 2) with 337 sera having concordantly negative results and 325 sera having concordantly positive results. Focus HSV-2 ELISA and WB were concordant in 339 negative sera and in 356 positive sera; overall 695 (88.9%). Focus HSV-2 ELISA and Biokit were concordant in 338 negative and 340 positive results; overall 678 (86.7%). WB and Biokit were concordant in 408 negative and 325 positive results for an overall concordance of 93.7%.
Figure 2 HSV-2 serology results by HerpeSelect HSV-2 ELISA ("Focus"), biokitHSV2 Rapid Test ("Biokit") and western blot (WB) in 782 sera.
Of 300 sera that were positive by Focus HSV-2 ELISA with index values over 3.5, 275 (92%) confirmed by both Biokit and WB. Fifteen results confirmed as positive by WB, only (Figure 2). Low positive (index values 1.1–3.5) index values occurred in 142 sera. Only 50 (35.2%) confirmed as positive by both Biokit and WB while 62 (43.7%) confirmed by neither test (Figure 2). The low positive Focus HSV-2 ELISA index value group yielded the majority (N = 30) of the study's 49 discordant sera between Biokit and WB. Sera that were negative for HSV-2 antibody by Focus HSV-2 ELISA were nearly always negative by WB (339; 99.7%) or Biokit (338; 99.4%).
Effect of Biokit confirmatory testing on test accuracy
To compute test accuracy, we applied data from the 340 sera that were seronegative by Focus HSV-2 ELISA to those Focus negative sera that were not run by WB or Biokit. We determined that approximately 2 (confidence interval [CI] 0–7) of the additional 847 seronegative sera would be positive by WB and about 5 (CI 0–12) would be positive by Biokit. Specificity of the Focus test was, in this way, estimated to be 93.2 (CI 91.8–94.6) with WB results as the gold standard. Specificity of the Biokit test, alone, against WB, was 98.4 (CI 97.5–99.3) (Table 2). Use of Biokit testing on all Focus HSV-2 ELISA positive sera improved estimated specificity from 93.2 to 98.7% without sacrificing sensitivity (99.1%). Positive predictive value improved from 80.5% to 95.6% and the negative predictive value remained at 99.7% (Table 2).
Table 2 Estimated sensitivity and specificity of HSV-2 tests
Focus Biokit Focus Plus Biokit
Sensitivity 99.2 (96.3,100.0) 90.5 (86.1,94.0) 99.1 (96,100.0)
Specificity 93.2 (91.8,94.6) 98.4 (97.5,99.3) 98.7 (98.1,99.4)
Positive Predictive Value 80.5 (76.9,84.2) 94.5 (90.5,97.3) 95.6 (93.4,97.8)
Negative Predictive Value 99.7 (98.9,100.0) 97.5 (96.6,98.4) 99.7 (98.9,100.0)
Ranges indicated show the confidence intervals for accuracy; these also incorporate the uncertainty from having confirmed only a portion of Focus negatives. Biokit was used to confirm all sera that were initially positive for HSV-2 antibodies by Focus HSV-2 ELISA; the combined test results are shown in the "Focus Plus Biokit" column.
Effect of HSV-1 antibody on discordance for HSV-2 between tests
Seventeen sera were false positive by Biokit as compared with WB. Two were also negative by Focus HSV-2 ELISA; 14 were low positive by Focus HSV-2 ELISA and 1 was positive by Focus HSV-2 ELISA with an index value >3.5. All 17 were positive for HSV-1 antibody by WB.
Of 86 sera with positive Focus test results but negative WB results, 68 (79.1%) were positive for HSV-1. The overall HSV-1 prevalence in the study was 64% (502 of 782) by WB. Thus, Biokit, and to a lesser extent, Focus HSV-2 ELISA specificity values appeared to be affected by the presence of HSV-1 antibodies. Conversely, WB could be falsely negative for HSV-2 in the presence of HSV-1 antibody.
Predictive value of the test algorithm by population prevalence
We used the data for estimated sensitivity and specificity to evaluate the positive and negative predictive values that would be expected with populations that differ in prevalence of HSV-2 infection. As shown in Table 3, the use of the Biokit test to confirm initial Focus HSV-2 ELISA positive results (i.e. considering only results that are positive by both Focus HSV-2 ELISA and Biokit to be true positives) enhanced the positive predictive value (PPV) in all populations. PPV increased with increasing prevalence as predicted by Bayes theorem (PPV = Θp/ [Θp + (1-Φ)(1-p)], where p = prevalence, Θ = test sensitivity, and Φ = test specificity). For example, in groups with very low HSV-2 prevalence (10%) the PPV increased from 61.9 to 89.8%. For prevalences typical of antenatal practices (30%) and STD clinics (40–50%), the PPV of the test combination was over 97% and 98%, respectively.
Table 3 Positive and negative predictive values of HSV-2 test approaches by population prevalence
Positive Predictive Value
True Prevalence Focus HSV-2 ELISA Biokit Focus HSV-2 ELISA+Biokit
10 61.9% 86.5% 89.8%
20 78.5% 93.5% 95.2%
30 86.3% 96.1% 97.1%
40 90.7% 97.5% 98.1%
50 93.6% 98.3% 98.8%
70 97.2% 99.3% 99.5%
90 99.2% 99.8% 99.9%
Negative Predictive Value
10 99.9% 98.9% 99.9%
20 99.8% 97.7% 99.8%
30 99.6% 96.0% 99.6%
40 99.4% 94.0% 99.4%
50 99.1% 91.2% 99.1%
70 98.0% 81.7% 97.9%
90 92.5% 53.6% 92.3%
Ranges indicated show the confidence intervals for accuracy; these also incorporate the uncertainty from having confirmed only a portion of Focus negatives. Biokit was used to confirm all sera that were initially positive for HSV-2 antibodies by Focus HSV-2 ELISA; the combined test results are shown in the "Focus Plus Biokit" column.
Discussion
Commercially available type specific serologic testing for HSV-2 has markedly improved the ability to diagnose this common, widespread infection with significant clinical and therapeutic implications. While these assays have been extremely useful for research and epidemiological studies, there has been concern about their specificity for case management of HSV-2, especially in populations in which definitive data on seroprevalence are lacking [5]. Sera that have index values between 1.1 and 3.5 by Focus HSV-2 ELISA have the highest probability of being falsely positive. In our study, over one third of positive sera fell into this category of "low positive" index values and the majority of these sera did not confirm as positive by the Western blot assay (WB) [6].
One approach to this problem is to perform confirmatory assays on sera initially positive in the screening test. For example, the HerpeSelect immunblot assay is easy to run, relatively inexpensive at about $25 per test, and is FDA approved. However, the test is based on the same gG-2 antigen as the Focus HSV-2 ELISA and performs almost identically with that test [1]. As such, the immunoblot test has not been highly effective in discriminating falsely positive from truly positive Focus HSV-2 ELISA results [5]. Other confirmatory testing options require shipping sera to a reference laboratory for western blot (WB) or inhibition ELISAs that, while effective [6], can be time-consuming and expensive. We selected the biokit HSV-2 Rapid HSV-2 Test ("Biokit") as a confirmatory test because it is relatively inexpensive (about $20 per test), requires less than 10 minutes and no special equipment to perform, is FDA approved, and can be purchased and easily performed by any laboratory. The gG-2 antigen is a lectin-purified native protein as compared with the recombinant gG-2 used by Focus. Biokit's test also differs in presenting the antigen within a membrane while, in the Focus test, the gG-2 is sterically bound to a plastic microwell. These differences may result in different subpopulations of antibodies being detected so that a positive result in both tests is a more rigorous result than one provided by a single test.
Our study demonstrated that sera in the "low positive" index value range (1.1–3.5) by Focus HSV-2 ELISA had the highest proportion of discordant results among 3 tests: Focus, Biokit, and WB. Only 35.2% of low positive sera were positive by all 3 tests. In contrast, 99% of the Focus HSV-2 ELISA-negative sera were concordantly negative by all 3 tests and 92% of 300 sera with index values >3.5 were concordantly positive by all 3 tests. Low positive results are not rare. In the high risk MSM group in our study, 12.7% of sera, overall, and 36.9% of positive sera had index values of 1.1–3.5. In the all-comer group, 6.1% of sera, overall, and 25.7% of positive sera had low positive index values.
When sera with positive Focus HSV-2 ELISA results were run by Biokit, the combination provided a confirmed positive result in 340 (76.9%) of the 442 sera. By comparison, WB confirmed 356 (80.5%) of positive results. Moreover, 71 (16%) of the 442 positive sera were negative by both Biokit and WB; thus additional testing successfully identified false positive results in a significant proportion of the study population (4.1%). Overall, 89.6% of positive sera were given a higher quality answer as defined by confirming a positive result or by identifying a false positive. Either of these testing outcomes would serve to enhance the confidence of the laboratory providing results or the clinician preparing to counsel a patient.
The few false positive results from our testing strategy were all seen among those who were HSV-1 seropositive, a finding that has been previously described in a different patient population [5]. Several explanations are possible: First, the WB, which was the comparator assay, could be falsely negative; HSV-2 antibodies are more difficult to detect against a background of HSV-1 antibodies. Second, cross-reactive epitopes on glycoprotein G may affect the tests. While glycoprotein G molecules from HSV-1 and HSV-2 are predominantly distinct immunologically, there are regions of sequence homology and this effect cannot be definitively ruled out [11,12].
The two-step testing algorithm resulted in an increase in indeterminate results as defined by a positive Focus HSV-2 ELISA and a negative Biokit test. When Focus HSV-2 ELISA, alone, is used, 1% (26 of 1749) of study sera had index values in the indeterminate or "equivocal" range (0.9–1.1). When Biokit confirmatory testing was done, the additional unconfirmed results led to a total of 7.3% (128 of 1749) of samples with indeterminate outcomes. We feel this is a reasonable tradeoff for higher accuracy in determining HSV-2 infection status. Most of these additional indeterminate sera (N = 78) were low positive and most (N = 71) did not confirm by WB.
Several options are reasonable to determine HSV-2 serostatus in patients with Focus HSV-2 ELISA positive results that fail to confirm by Biokit. First, laboratories might consider retesting by Focus HSV-2 ELISA to rule out laboratory error. Reconstruction experiments have shown that low positive results can be artificially introduced by splashing or by dipping the pipette tip into a positive ELISA well, then into a negative well [4]. Seventy-four of the 86 sera that were positive by Focus HSV-2 ELISA but negative by WB were repeated and 11 sera (14.8%) retested as negative; all 11 were negative by Biokit. Thus, a repeat test can resolve the serostatus quickly, without redrawing blood from the patient, and, ideally, before the serology result leaves the laboratory. A second option is to re-test the patient in 6–12 weeks to rule out early seroconversion [13,14]. This option also applies to patients whose first sample was repeatedly low positive. A third option to establish HSV-2 serostatus is to send the sample to a reference laboratory for WB or to Focus Diagnostics for a gG-2 inhibition assay [6].
Conclusion
Our study shows that Biokit confirmation of positive Focus results (index values >1.1) can substantially improve the positive predictive value of serologic screening for HSV-2 antibodies by Focus HSV-2 ELISA. Clinicians who counsel and manage patients with suspected herpes infections or asymptomatic, low-risk patients who wish to be screened for herpes may consider confirmatory testing to be both justified and cost-effective to increase accuracy. Laboratories now have available a quick and inexpensive means of providing a highly specific test combination for HSV-2.
Competing interests
RAM has received speaking honoraria or consulting fees from Focus Technologies within the last 5 years. LC, AM, and DF report no competing interest. This study was supported by NIH grant AI 30731 and The Bill and Melinda Gates Foundation's Partners in Prevention Project Grant #26469; these entities do not have financial interest in the outcome of the study.
Authors' contributions
RAM and LC designed the study and wrote the manuscript; DF established and maintained study data files and performed the basic data analyses. AM contributed to the study design and performed statistical analyses. All authors reviewed and approved the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank the University of Washington Diagnostic Virology Laboratory for serum banking and retrieval and for performing the Focus and WB testing and Ms. Stacy Selke for conducting the selection process for sera included in the study. This study was supported by NIH grant AI 30731 and The Bill and Melinda Gates Foundation's Partners in Prevention Project Grant #26469.
==== Refs
Ashley RL Performance and use of HSV type specific serology test kits Herpes 2002 9 38 45 12106510
Van Dyck E Buve A Weiss HA Glynn JR Brown DWG De Deken B Parry J Hayes RJ Performance of commercially available enzyme immunoassays for detection of antibodies against herpes simplex virus type 2 in African populations J Clin Microbiol 2004 42 2961 2965 15243045 10.1128/JCM.42.7.2961-2965.2004
Laeyendecker O Henson C Gray RH Nygun H-N Horne BJ Wawer MJ Serwadda D Kiwnuka N Morrow RA Hogrefe W Quinn TC Performance of a commercial, type-specific enzyme-linked immunosorbent assay for detection of herpes simplex virus type 2-specific antibodies in Ugandans J Clin Microbiol 2004 42 1794 6 15071053 10.1128/JCM.42.4.1794-1796.2004
Morrow RA Nollkamper J Robinson NJ Bishop N Smith J Performance of Focus ELISA tests for herpes simplex virus type 1 (HSV-1) and HSV-2 antibodies among women in ten diverse geographic locations Clin Microbiol Infect 2004 10 530 6 15191381 10.1111/j.1469-0691.2004.00836.x
Golden MR Morrow-Ashley R Swenson P Hogrefe WR Handsfield HH Wald A HSV-2 Western blot confirmatory testing among men testing positive for HSV-2 using the Focus ELISA in an STD clinic Sex Trans Dis
Hogrefe W Su X Song J Ashley R Kong L Detection of herpes simplex virus-2 immunoglobulin G antibodies in African sera by using recombinant gG2, Western blotting, and gG2 inhibition J Clin Microbiol 2002 40 3635 3640 12354858 10.1128/JCM.40.10.3635-3640.2002
Ashley RL Eagleton M Pfeiffer N Ability of a rapid serology test to detect seroconversion to herpes simplex virus type 2 glycoprotein G soon after infection J Clin Microbiol 1999 37 1632 1633 10203544
Ashley RL Wald A Eagleton M Premarket evaluation of the POCkit HSV-2 type specific serologic test in culture-documented cases of genital herpes simplex virus type 2 Sex Trans Dis 2000 27 266 269 10.1097/00007435-200005000-00005
Ashley RL Militoni J Lee F Nahmias A Corey L Comparison of Western blot (immunoblot) and glycoprotein G-specific immunodot enzyme assay for detecting antibodies to herpes simplex virus types 1 & 2 in human sera J Clin Microbiol 1988 26 662 667 2835389
Begg CB Greenes RA Assessment of diagnostic tests when disease verification is subject to selection bias Biometrics 1983 39 207 15 6871349
Liljeqvist J-A Trybala E Svennerholm B Jeansson S Sjogren-Jansson E Bergstrom Localization of type-specific epitopes of herpes simplex virus type 2 glycoprotein G recognized by human and mouse antibodies J Gen Virol 1998 79 1215 1224 9603337
Tunback P Liljeqvist J-A Lowhagen G-B Bergstrom T Glycoprotein G of herpes simplex virus type 1: identification of type-specific epitopes by human antibodies J Gen Virol 2000 81 1033 1040 10725430
Ashley-Morrow R Krantz E Wald A Time course of seroconversion by HerpeSelect ELISA after acquisition of genital herpes simplex virus type 1 (HSV-1) or HSV-2 Sex Trans Dis 2003 30 310 314 10.1097/00007435-200304000-00007
Wald A Ashley-Morrow R Serological testing for herpes simplex virus (HSV)-1 and HSV-2 infection Clin Infect Dis 2002 35 S173 182 12353203 10.1086/342104
|
16225691
|
PMC1276011
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Oct 14; 5:84
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-84
|
oa_comm
|
==== Front
Ann Clin Microbiol AntimicrobAnnals of Clinical Microbiology and Antimicrobials1476-0711BioMed Central London 1476-0711-4-161620213910.1186/1476-0711-4-16ResearchA multicenter point-prevalence study: antimicrobial prescription frequencies in hospitalized patients in turkey Usluer Gaye [email protected] Ilhan [email protected] Hakan [email protected] Turkish Antibiotic Utilization Study Group [email protected] Osmangazi University, Faculty of Medicine, Department of Infectious Diseases, Eskisehir-Turkey2 Ondokuz Mayis University, Faculty of Medicine, Department of Infectious Diseases, Samsun-Turkey2005 3 10 2005 4 16 16 20 6 2005 3 10 2005 Copyright © 2005 Usluer et al; licensee BioMed Central Ltd.2005Usluer et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Accurate information about prescribing patterns in hospitals is valuable in improving the quality of antimicrobial prescriptions.
Methods
Data on the use of antimicrobial agents in eighteen tertiary care hospitals were collected on March 20th 2002.
Results
One or more antimicrobials were ordered in 2900 (30.6 %)of 9471 hospitalized patients. The reasons of hospitalization of the patients receiving antimicrobials were medical treatment (42.5 %), elective surgery (39.6 %), treatment of infectious disease (17.1 %) and emergent surgical procedures (10.4 %). The highest consumption frequencies were found in surgical (81.6 %) and medical (55.2 %) intensive care units. The 48.8 % of antimicrobials were given for treatment and 44.2 % for prophylactic use. The most common reasons for treatment were found as lower respiratory tract, urinary tract, surgical wound infections and febrile neutropenia. Antimicrobials were ordered empirically in 78.4 % of patients. The proven infection ratio was found as 30.7 %. The 56.4 % and 13.4 % of orders were evaluated as clinically and microbiologically appropriate respectively.
Conclusion
These results suggest that antimicrobial prescription and empirical treatment ratios were high and inappropriate at inpatient groups.
antimicrobial useappropriate antimicrobial usecost
==== Body
Introduction
Since antimicrobial chemotherapy was introduced in medical practice, there have been calls for its rational use. Appropriate antimicrobial treatment greatly improves the prognosis of infectious diseases. There has been a very significant reduction in morbidity and mortality associated with the use of antimicrobials since they were first introduced [1]. However, the overuse of antimicrobials may increase the risks of drug resistant pathogens, side effects and costs of medical care. The right agent at the right dose and dosing interval and right duration can achieve both a favorable clinical outcome and prevent the selection of resistance. It was reported that 20–50% of antimicrobial use in humans was questionable or inappropriate [1,2]. Accurate information about prescribing patterns in hospitals is valuable in improving the quality of antimicrobial prescriptions.
Only very limited data on the usage of antimicrobials in Turkey [1,3]. The over use of antimicrobials increases the risk of drug resistant pathogens, side affects and cost of medical care [4]. This multicenter study was planned as a point-prevalence study to evaluate antimicrobial prescription frequency and patterns in tertiary care hospitals in Turkey.
Materials and methods
This prospective study was conducted in eighteen tertiary care hospitals from 14 different cities located in seven geographical regions of Turkey. These hospitals were representing approximately 30% of all tertiary care hospitals in Turkey. Data on the use of antimicrobial agents in these hospitals were collected by infectious diseases consultants on March 20th 2002. The same methodology was used for all hospitals. All patients who have received antimicrobials for any reason were included to this study. The data was included to the study within the first week of the study day, if there was a delay in the recording of data for any reason such as unavailable culture results which the specimen for this culture was collected before or on March 20th 2002.
In the study, total bed capacities of hospitals, number of hospitalized patients, the type and number of antimicrobial prescriptions, the main diagnosis which the prescription was made, clinical and microbiological evidences for treatment were recorded. The presence of an infectious disease was diagnosed according to signs and symptoms, non-microbiological and microbiological laboratory findings and defined as proven infection.
For patients receiving antimicrobials, demographic data, reason for admission and hospitalization, results of microbiological samples, name and dosage of antimicrobials and indication for antimicrobials were recorded on special forms. The antimicrobial regimes were evaluated according to choice, combination, duration and dose of the antimicrobials.
Hospital and patients details were recorded in two different forms. The first one was for the hospital details such as the name of the hospital, total bed capacity, departments and their bed capacities, total number of hospitalized patients of the hospital and of each department on the study day. The second form was used for recording data of patients which were receiving an antimicrobial agent on the study day. All of the records were collected and evaluated by the principal investigators in the study center. All of the data was transferred to the computer using a file designed by Dr. Ozgunes with Microsoft Access.
The antimicrobial prescription ratio, hospitalization reason of antimicrobial receiving patients and combination therapies were evaluated. Antimicrobial prescriptions were globally considered inappropriate if any of the assessed criteria appeared unacceptable, according to indication or antimicrobial choice, dosage errors, and duration of treatment. Appropriateness of antimicrobial prescriptions was evaluated according to the clinical and laboratory findings on the beginning of the therapy.
Statistical analyses were made by chi square test.
Results
Eighteen tertiary care hospitals from 14 different cities of Turkey included to the study. 9471 hospitalized patients were evaluated. One or more antimicrobials were ordered in 2900(30.6%) of 9471 patients. In the antimicrobial receiving group 1232 (42.5 %) patients were hospitalized for medical treatment, 1147 (39.6%) for elective surgery, 497(17.1%) for infectious diseases, 303 (10.4 %) for emergent surgery and 73(2.5%) for other reasons. There were more than one hospitalization reasons for some patients (Table 1).
Table 1 Hospitalized, antimicrobial receiving patients and hospitalization reasons of antimicrobial prescribed patients in 18 centers.
n %
Hospitalized patients 9471 100
Antimicrobial receiving patients 2900 30.6
Hospitalization reason
Medical treatment 1231 42.5
Elective surgery 1147 39.6
Treatment of an infectious diseases 497 17.1
Emergent surgery 303 10.4
Other reasons 73 2.5
The highest antimicrobial consumption ratios were found in intensive care units (ICU) (Surgical ICU 81%, medical ICU 52.5%). Antimicrobial consumption frequencies according to departments (surgical/medical) were shown in table 2.
Table 2 The distribution of antimicrobial prescribed patients to hospitalized patients.
Antimicrobial prescribed Total hospitalized Percentage of antimicrobial prescribed
Surgical Clinics 1414 4172 33.9%*
Medical Clinics 1138 4529 25.1%*
Surgical ICU 107 132 81%*
Medical ICU 83 158 52.5%*
Total 2900** 9471 30.6%*
ICU: Intensive care unit.
*The ratios were found statistically different (x2 = 119 SD = 2, P < 0.001).
**The wards of 158 patients were not reported.
The indications of antimicrobial therapy were also evaluated. The 48.8 % of antimicrobials were given for treatment of an infectious disease and 44.2 % for surgical antimicrobial prophylaxis. It wasn't found any reason for antimicrobial prescriptions in 204 (7 %) patients' records. More than one reason was reported for some patients.
Antimicrobial prescriptions were made empirically in 2275 (78.4 %) of patients and according to microbiological data in 334 (11.5%).
The proven infection ratio was found as 30.7 % in 2900 patients and 57.15% (807 of 1412) in treatment group. The antimicrobial prescriptions were evaluated by the investigator if or not they were appropriate to clinical and microbiological data. The 56.4 % and 13.4 % of orders were evaluated as clinically and microbiologically appropriate respectively in 2900 patients. In patients receiving prophylactic antimicrobials 671(52.46%) of 1279 prescription were evaluated as appropriate (Table 3). The 61.54% (869 of 1412) prescriptions were evaluated as clinically appropriate in patients receiving antimicrobials for treatment (Table 3). There was not any microbiological data in 986 (69.83%) patients in this group. The microbiologically appropriate prescription ratio was found 84.04% in 326 patients with microbiological data. The appropriate and inappropriate prescription in treatment group was given in table 3.
Table 3 The appropriate prescription in patients receiving prophylactic antimicrobials and the proven infection, clinically and microbiologically appropriate treatment ratios in patients that were treated for an infection.
n %
Prophylactic antimicrobial use Appropriate 671 52.46*
Inappropriate 423 33.07*
Not reported 185 14.46
TOTAL 1279
Clinically Proven infection 807 57.15
Appropriate 869 61.54**
Inappropriate 364 25.77**
Not reported 179 12.67
TOTAL 1412 100
Microbiologically Appropriate 274 84.04***
Inappropriate 52 15.95***
Not reported 88 6.23
TOTAL 414**** 100
No microbiological data 986 69.83
*t = 9.91, SD = 1092, p < 0.00
**t = 19.16, SD = 1231, p < 0.001
*** t = 15.79, SD = 324, p < 0.001
****The microbiological data was not available for all patients.
The combination therapy ratio was found as 33%. 50 patients including tuberculosis cases were receiving more than three antimicrobials. 25 patients were receiving combination therapy because of tuberculosis. 453 (15.6 %) of patients were receiving three antimicrobials and 428 of them (14.7%) were non-tuberculosis patients.
The most common prescribed antibiotics were cefazolin, ampicillin-sulbactam, ceftriaxone, ciprofloxacin, amikacin, gentamicin, ornidazole, cefuroxime, meropenem and vancomycin. The prescription ratios of antibiotic groups were given in table 4.
Table 4 The most common prescribed antibiotic groups in hospitalized patients and the most common used antibiotics in combinations.
Antibiotic group Prescription % Combination %
Penicillines 23.6 18.8
1.Generation Cephalosporins 14.6 7.1
2.Generation Cephalosporins 5.3 0.0
3.Generation Cephalosporins 23.7 21.1
4.Generation Cephalosporin 4.2 4.5
Aminoglycosides (Excluding streptomycin) 17.2 30.8
Carbapenems 6.5 10.9
Glycopolypeptides 4.8 13.1
Ornidazole-Metronidazole-Clindamycin 9.9 18.2
Quinolones 14.4 11.9
Macrolides 3.0 4.7
Tetracyclines 0.7 1.2
Antifungal agents 3.4 4.3
The most common used antibiotics in combinations were aminoglycosides (30.8%), 3rd generation cephalosporins (21.1%), penicillins (18,8%), ornidazole-metronidazol-clindamycin (18.2%), glicopolypeptides (13.1%), quinolones (11.9%) and carbapenems (10.9%) (Table 4). The 88.5% of combined aminoglycosides were used in combination with beta-lactams and glycopolypeptides. There were 15 combinations of sulbactam-ampicillin with clindamycin, ornidazole or metronidazol. We determined that 67.44 % of the patients were in official health insurance systems and 19.7 % of them were in official social assistance system.
Discussion
Although the principles of rational antimicrobial usage have been well defined for many years, inappropriate use of antimicrobials remains wide spread. The cost, adverse effects and development of resistance are main problems in wide spread usage of antimicrobials. The emergence and spread of drug resistant pathogens have already become a very serious problem internationally. It was reported that 14% and 43% of all courses of antimicrobial chemotherapy were deemed unnecessary because there was no evidence of infection [2,5,6].
In this study, antimicrobial prescription frequency was found as 30.6% in hospitalized patients. The antibiotic prescription frequency was reported as 77.8% from a university hospital in China, and as 65% from a pediatric teaching in Costa Rica [6,7]. Empirical antimicrobial prescription and combination antimicrobial treatment ratios were high (78.4%, 33%) in the study group also. The problem is more serious in ICU and surgical departments than medical departments. The antimicrobial prescription ratios were higher in ICU's (81% of surgical ICU, 52.5% of medical ICU) than other departments of hospitals (P < 0.001). It was reported that the 58.0% of surgical ICU patients in a university hospital from Germany were receiving antibiotics [8]. The antibiotic prescription frequency was reported as 6.55 and 14.4% from two different pediatric ICUs from Israel [9]. The proven infection ratio was found as 30.7% in the appropriate antimicrobial treatment given group and 57.9% in the inappropriate antimicrobial treatment group. The results of the study showed that inappropriate antimicrobial prescription was an obvious problem in the study hospitals of Turkey. More than 40 % of antimicrobial prescriptions were made without a proven infection. Inappropriate antimicrobial usage is a worldwide problem. 40% of antibiotic prescriptions were reported that had no record of justification and 55% of prescriptions had no indication of planned duration of therapy [7].
The 44% of antimicrobial prescriptions were made for surgical prophylaxis and 52.4% of them were appropriate. This group was seemed to be increasing the inappropriate prescription ratios because of the long duration usage and wrong selection of antimicrobials. Hu et al reported that 30% of hospitalized patients were receiving perioperative antibiotics and 20% of them received antibiotics before or during operation and 80% of them after operation. The duration of perioperative antibiotic prophylaxis was less than or equal to seven days in 42.7% of patients, 8–13 days in 31%, and 14 days or more in 26.3% [6]. In another study reported by Bailly et al, the rate of compliant prescription for surgical prophylaxis was 41.7% [10].
Also the combination therapy ratios were found as high as 33% of total antimicrobial prescribed patients. It can be thought that there is a relation between high empirical antimicrobial treatment and high combination therapy ratios. The limited microbiological evidence for the diagnosis of infection can be thought as another reason for high ratios of empirical and combination therapies because of the microbiologically appropriate and inappropriate usage ratios were found as 84.04% and 15.95% respectively in the treatment group. These results suggest that a multidisciplinary antimicrobial management system is required in hospitals because of the high proportion of empirically treatment and inappropriate use of antimicrobials. The system must have legal support and the antimicrobial control teams must be include the departments of infectious diseases, microbiology, pharmacy, and infection control [1]. Also there is need good microbiological support for clinicians to increase the appropriate prescription rate. Local and practical antimicrobial treatment guidelines for clinicians and continuous education programs may decrease the inappropriate, empirical and combination therapy ratios.
The cost of antimicrobials is another serious problem for insurance systems in Turkey. The anti-infective drugs are the most used drugs (22% of all drugs) in our country. The annually antimicrobial and total drug cost for per person was calculated as $8,4 and $38 in Turkey [11].
In conclusion, this point-prevalence study revealed that more than 50% of patients received inappropriate antimicrobial prescriptions. We thought that only restricted prescription procedures are not enough for the reduction of inappropriate antimicrobial rates. A general antimicrobial treatment program must include education, guidelines, restricted usage, control of the hospital pharmacies and automatic discontinuation by the hospital pharmacies.
Acknowledgements
Turkish Antibiotic Utilization Study Group: Halis Akalın, Celal Ayaz, Rahmet Caylan, Yesim Cetinkaya Sardan, Nese Demirturk, Ilknur Erdem, Funda Ergin, Serpil Erol, Saban Esen, Sibel Gündes, Iftihar Koksal, Oral Oncul, Kazim Ozdamar, Recep Ozturk, Fatma Sirmatel, Irfan Sencan, Yesim Tasova, Gunay Tuncer, Sercan Ulusoy, Serhat Unal, Haluk Vahaboglu, Tansu Yamazhan (In alphabetical order)
==== Refs
Guven GS Uzun O Principles of good use of antibiotics in hospitals J Hosp Infect 2003 53 91 96 12586566 10.1053/jhin.2002.1353
Hecker MT Aron DC Patel NP Lehmann MK Donskey CJ Unnecessary use of antimicrobials in hospitalized patients Arch Intern Med 2003 163 972 978 12719208 10.1001/archinte.163.8.972
Usluer G Ozgunes I Kılıc Z Enfez N Evaluation of antimicrobic use in a university hospital. [Abstract] Spanish Journal of Chemotherapy 2000 13 s50
John FJ JrFishmen NO Programmatic role of the infectious diseases physician in controlling antimicrobial costs in the hospital Clin Infect Dis 1997 24 471 485 9114203
Knox K Lawson W Dean B Holmes A Multidisciplinary antimicrobial management and the role of the infectious diseases pharmacist-a UK perspective J Hosp Infect 2003 53 85 90 12586565 10.1053/jhin.2002.1350
Hu S Liu X Peng Y Assessment of antibiotic prescription in hospitalized patients at a Chinese university hospital J Hosp Infect 2003 46 161 3
Mora Y Avila-Agüero ML Umana MA Jimenez AL Paris MM Faingezich I Epidemiologic observations of the judicious use of antibiotics in pediatric teaching hospital Int J Infect Dis 2002 6 74 7 12044307 10.1016/S1201-9712(02)90141-4
Hartmann B Junger A Brammen D Röhring R Klasen J Quinzio L Benson M Hempelmann G Review of antibiotic drug use in a surgical ICU: Management with a patient data management system for additional outcome analysis in patients staying more than 24 hours Clin Ther 2004 26 915 24 15262462 10.1016/S0149-2918(04)90135-X
Gavrilov V Berkovitch M Ling G Brenner-Zadda G Lifshitz M Gorodischer R Unapproved prescriptions in two pediatric intensive care units in Israel Curr Ther Res Clin Exp 2003 64 734 42 10.1016/j.curtheres.2003.09.016
Bailly P Lallemand S Thouverez M Talon D Multicentre study on the appropriateness of surgical prophylaxis J Hosp Infect 2001 49 135 8 11567560 10.1053/jhin.2001.1064
Kurt H Infeksiyon hastaliklarıinda tedavi ve maliyet Klimik Derg 2003 16 261 263
|
16202139
|
PMC1276781
|
CC BY
|
2021-01-04 16:38:21
|
no
|
Ann Clin Microbiol Antimicrob. 2005 Oct 3; 4:16
|
utf-8
|
Ann Clin Microbiol Antimicrob
| 2,005 |
10.1186/1476-0711-4-16
|
oa_comm
|
==== Front
Aust New Zealand Health PolicyAustralia and New Zealand Health Policy1743-8462BioMed Central London 1743-8462-2-221620971810.1186/1743-8462-2-22ResearchA network approach for researching partnerships in health Lewis Jenny M [email protected] Department of Political Science, University of Melbourne, Parkville, 3010, Australia2005 7 10 2005 2 22 22 8 7 2005 7 10 2005 Copyright © 2005 Lewis; licensee BioMed Central Ltd.2005Lewis; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The last decade has witnessed a significant move towards new modes of governing that are based on coordination and collaboration. In particular, local level partnerships have been widely introduced around the world. There are few comprehensive approaches for researching the effects of these partnerships. The aim of this paper is to outline a network approach that combines structure and agency based explanations to research partnerships in health. Network research based on two Primary Care Partnerships (PCPs) in Victoria is used to demonstrate the utility of this approach. The paper examines multiple types of ties between people (structure), and the use and value of relationships to partners (agency), using interviews with the people involved in two PCPs – one in metropolitan Melbourne and one in a rural area.
Results
Network maps of ties based on work, strategic information and policy advice, show that there are many strong connections in both PCPs. Not surprisingly, PCP staff are central and highly connected. Of more interest are the ties that are dependent on these dedicated partnership staff, as they reveal which actors become weakly linked or disconnected without them. Network measures indicate that work ties are the most dispersed and strategic information ties are the most concentrated around fewer people. Divisions of general practice are weakly linked, while local government officials and Department of Human Services (DHS) regional staff appear to play important bridging roles. Finally, the relationships between partners have changed and improved, and most of those interviewed value their new or improved links with partners.
Conclusion
Improving service coordination and health promotion planning requires engaging people and building strong relationships. Mapping ties is a useful means for assessing the strengths and weaknesses of partnerships, and network analysis indicates concentration and dispersion, the importance of particular individuals, and the points at which they will fragment. A narrative approach adds an assessment of whether the partnerships are being used and valued. The approach outlined here, which examines structure and agency as separate but related explanations, has much to offer in examining partnerships.
==== Body
Introduction
Much discussion of policy making and governing at the beginning of the 21st Century, indicates a significant shift in the model of governance across many sectors, away from an emphasis on competition between agencies (markets), to a model of inter-agency coordination and collaboration (networks). In Australia, this general trend is spelt out in the recent 'Connecting Government' paper [1]. This reflects the increasingly complicated arrangements for organizing and delivering services, which cross horizontal (levels of government) and vertical (government, private and third sector agencies) boundaries, and are in part a legacy of the earlier managerial and market based modes of governing [2]. It is helpful to conceive of many organizational contexts as network-like in order to understand them, without claiming that they are necessarily "good".
Partly driven by this shift, but also indicating a desire to find better approaches, there has been an astonishing growth in public policies which embrace the concepts of partnerships, alliances, collaborations and networks. Health policy has been a part of this broader trend, and there is no shortage of discussions of a range of collaborative forms of governing. In particular, partnerships of many varieties have become a key means for governing a range of policy initiatives at the local level [3].
This paper does not intend to make a full explication of what the characteristics of these partnerships are, or how they are being used. Neither does it address the question of whether there is in fact a clear commitment to them as a new form of governance, or whether they are better than previous approaches. Instead it explores new theories and methods for examining partnerships in health. As the current emphasis on partnerships shows no signs of abating, health policy research is in need of new concepts, methodologies and techniques for establishing their positive and negative effects.
While partnerships range from the bottom-up, locally self generated and voluntary, to the top-down, centrally steered and government mandated arrangements, those of central interest here are the latter, and are considered to be 'managed networks' [4]. Partnerships in the public sector often reflect efforts to institutionalise the positive effects of networking (such as increasing diversity by involving a greater range of actors) by requiring organisations and programs to have more formal connections to each other [5]. Those in focus here are created by government and are centrally steered with specific deliverables and targets defined by the centre rather than the individual partnerships. They cover defined geographical areas and have dedicated network coordinators. Leadership is undertaken by formal agencies rather than by mobilised communities. They have some ability to shape their own local priorities, but within limits set by a central authority [4].
Evaluations of a number of these types of partnerships in the UK have demonstrated that there are benefits, but that partnerships also face substantial difficulties. It is a slow process which clashes with the demands of government for results in the short term, and the need for local partnerships to reflect national priorities [4]. The most ambitious of these was Health Action Zones, and a recent evaluation of this program proclaimed the need for a new body of theory about what these kinds of programs can reasonably be expected to deliver in the face of bewildering complexities [6]. This paper argues that it is not only theory that is needed but also more appropriate methods for exploring their pluses and minuses.
A number of partnership tools have been created. One of the most relevance here is VicHealth's Partnerships Analysis Tool [7], which encourages partners to examine the reason for the partnership, map their relationships, and complete a checklist on a number of features of the partnership. While the map of the partnership has a similar focus to what is of concern here – understanding relationships – it is based on people's views of different types of engagement at an organizational level. This approach is straightforward to apply but more detailed views of relationships between people are required to understand what is happening beyond structures, and in accounting for the agency of individuals.
This paper attempts to walk the line between structure and agency, by combining network mapping and analysis with narratives, both of which are based on the observations of individual participants in partnerships. Pure description explains nothing, yet reflects the complexity of reality, while abstract theorising and modelling explains much but only by ignoring the complexity of reality [8]. A seemingly fruitful way of examining both structure and agency stems from Gidden's structuration theory [9]. He argues that structures constrain and facilitate actions, and also bind actions so that patterns are generated and reproduced. In other words, people work from within a set of structural constraints and opportunities, but also create and sustain these structures through their actions.
While Gidden's approach is to examine structure and action in isolation, Jessop's strategic-relation approach goes beyond this to examine structure in relation to action and action in relation to structure [10]. He argues for combining structural and discursive approaches. This is what this paper attempts, by using an approach that examines social networks as a set of connections, as well as a narrative about those network connections.
A little explored set of concepts and analytical techniques useful for evaluating these partnerships is available from social network analysis, which focuses on analysing relational data. It encompasses tools for network visualisation and network analysis using graph theory, statistical and algebraic models [11], and a range of concepts aimed at examining global network structure, network sub-structures, and the position of individuals within these networks (see these and other books dealing with these methods: [11-13]). The second means for examining whether and how partnerships improve relationships, build trust and foster better collaboration and cooperation between agencies is to examine the use and value of relationships through narratives.
The aim of this paper is to outline a network approach for use in researching partnerships in health. In doing so, research based on Primary Care Partnerships in Victoria is used to illustrate how combining network concepts and methods with narratives can be used to answer important questions about partnerships. The approach used examines connections (ties) between people, through the use of network mapping and analysis, particularly looking at multiple ties between people. This provides information on whether there are connections between people, in relation to various purposes (structure), but uncovers little about how they use and value them (agency), which requires an exploration of the quality of relationships within partnerships. More information on this approach is contained in the methods section.
Primary Care Partnerships
Primary Care Partnerships (PCPs) were introduced in Victoria in 2001. The stated aim of PCPs is to improve the health and well being of a catchment's population by better coordination of planning and service delivery. A second aim is to improve the experience of and outcomes for recipients and reduce the preventable use of hospital, medical and residential services [14].
The core agencies in PCPs are community health services, local governments, district nursing services, divisions of general practice, and aged care assessment services. In each locality, other agencies are also partners, based on local priorities. The initial PCP policy document emphasised consumer, carer and community involvement in the partnership [14]. In establishing PCPs, the Department essentially provided funding for each to employ a network coordinator, and project workers who took on roles that reflect the main priorities of service coordination and health promotion. In general, all PCPs began with a steering committee, and committees to deal with service coordination and health promotion. Each PCP has a chair, often drawn from one of the partner organisations.
The state health authority, the Department of Human Services (DHS) centrally steers these partnerships. The DHS central office role is one of policy direction and advice, while DHS regional offices are responsible for monitoring and accountability. Local governments are an important partner and PCPs usually cover two or three local government areas. Some 32 partnerships were established across the state initially, and each of them received an establishment grant from DHS on signing a partnership agreement.
Methods
Combining considerations of structure and agency into one approach is no easy task. However, this is where network theories and methods provide great promise. To demonstrate the utility of social network analysis for examining partnerships, research on two of the original 32 (now 31) PCPs, is used to examine whether and how relationships between individuals and organizations changed and developed since the inception of these partnerships.
One of the PCPs is within the Melbourne metropolitan area, and one is in rural Victoria. The information presented is based on a survey of 19 people from the metropolitan PCP, and 18 from the rural PCP. Interviews in the rural PCP were conducted with the three people in the PCP office, nine of the 14 people from the steering committee, three DHS regional office personnel, and three members of the health promotion steering committee. Three of the people from the consumers and carers group were interviewed together but network information was not collected from them. Interviews in the metropolitan PCP were conducted with three people from the PCP office, 11 of the 15 people who were on the steering committee at the time, two DHS regional office personnel, and three members of the health promotion steering committee. This information is summarised in Table 1.
Table 1 Information on the two PCP surveys
Number of interviewees Composition of interviewees Date of interviews
Metropolitan PCP 19 3 PCP
11/15 Steering Committee
2 DHS Regional Office
3 Health Promotion Late 2002
Rural PCP 18 3 PCP
9/14 Steering Committee
3 DHS Regional Office
3 Health Promotion Late 2002 to early 2003
Members of the partnerships were interviewed using face to face or telephone interviews, and all were recorded and then fully transcribed. Name generators (that is, asking people who they would contact in relation to something) are commonly used to collect network information based on a range of relationships (see [15,16] for examples). In network terms, people have multiple types of ties with each other. To capture these multiple relations, interviewees were asked the following:
1. Looking back over the last 6 months, who are the people you had the most contact with in order to do your work?
2. Over the last 6 months, who did you go to most when you wanted to get strategic information about something in the PCP?
3. Over the last 6 months, who did you go to most when you wanted to talk about policy in relation to this PCP or PCPs in general?
No set number of nominations was required, and a set of prompts was used if people were having trouble with recall (the prompts were: in your agency; in your PCP; in DHS regional and central offices; and elsewhere). The names were written into a form by the interviewer during the interview and the tape recording was used to check names later if they had been missed during the interview. While the second and third questions gave similar lists of people (with a number of respondents saying that list was the same), not all people nominated the same set of others for both. So many interviewees made a clear distinction between strategic information and policy ties.
The interviews provide information on both network structure, in terms of three different types of ties, and on agency, as described by people within the partnerships. The information on network ties forms the basis of the network maps and analysis. More people were mentioned than appear in these diagrams (as interviewees were free to nominate whoever they chose), but only interviewees are included here. That is, a larger number of people were mentioned in both PCPs, but those named but not surveyed did not have the chance to nominate people in return. Network analysis relies on people being able to both be nominated and to nominate others in return, so only the 19 and 18 who were interviewed are included in the analysis. This does not mean that the others are non-respondents in the traditional sense, it simply reflects that networks in effect have no boundaries.
During the interviews, open-ended questions were used to gather narrative descriptions of relationships. Questions were centred around:
• involvement with people in other agencies before the PCP was established
• level of contact since its establishment
• whether and how relationships had changed because of the PCP.
The narratives from the transcriptions were simply grouped under these three headings.
Results
Network mapping
The visualisation of relationships generated by mapping network connections between participants provides a useful pictorial means for describing links between people. The figures that follow show the three types of ties (work, strategic information, policy advice) combined, with the thickness of the lines between nodes (people) reflecting the number of different types of ties. For the purposes of this mapping exercise, more different types of ties between people is taken to indicate a stronger relationship. The maps presented here are generated by Netdraw, which is part of the UCInet package [17]. Those people with the most ties are placed at the centre of the map by this software.
The different colours of the dots (people) reflect which organization a person is from, based on Table 2:
Table 2 Red PCP office
Black Local goverment
Blue Hospital
Pink Community health service
Green DHS Regional office
White Division of General Practice
Yellow Other
Figure 1 shows the ties for each of the 19 people interviewed in the metropolitan PCP, including the three PCP project staff. It shows a network with many strong connections, particularly to the PCP staff, and (unsurprisingly) centred around them. Community health service staff are also central, and this reflects that the chair of this PCP was located in a community health service. The ties between people are all based on different aspects of PCP engagement, so a point of interest here is how many of these ties are neither to or through PCP staff.
Figure 1 Network based on strength of ties for metropolitan PCP.
Figure 2 shows the same network map with the three PCP staff excluded. The Chair of the PCP (3) becomes the most connected person once the PCP staff are removed, indicating the importance of this person and/or this organization in this locality. Without PCP staff, the remaining people are still connected into a single graph, but a number of them are now only weakly linked to others. Of interest is the divisions of general practice (11 and 12) which are especially weakly linked without the PCP in place.
Figure 2 Network based on strength of ties for metropolitan PCP, without PCP staff.
Figure 3 shows the ties for the 18 people surveyed in the rural PCP. Again, this network has many strong connections, particularly to the PCP staff, and is centred around them. Figure 4 is the same network minus the three PCP project staff. In contrast to the metropolitan PCP, the most connected person in this network, once the PCP staff have been removed, is located in the DHS regional office (208). Without the PCP staff, the rural PCP becomes disconnected, with two individuals isolated from the rest of the graph. These two people are from the division of general practice (219) and a hospital (215).
Figure 3 Network based on strength of ties for rural PCP.
Figure 4 Network based on strength of ties for rural PCP, without PCP staff.
These figures illustrate the usefulness of examining ties between individuals across organisational boundaries in partnerships. It is not surprising that PCP staff are the most central in providing a connecting role to others, since this is after all what PCP staff are supposed to do. Removing the ties that directly involve PCP staff, indicates what a policy change to abolish PCPs would do, at least in the short term. Divisions of general practice provide an interesting case in point, as the maps show that they are likely to be the first to become disconnected or weakly connected without the impetus of the PCP. They are funded by the Commonwealth and have few financial incentives to get involved in state government policy initiatives, including through PCPs, so it is not surprising that they are weakly linked. This result concurs with an evaluation of PCPs across the state, that identifies these and other reasons why Divisions have low engagement in PCPs [18].
Network measures
A large range of global network measures and network measures for individuals are available (see the growing literature on social network analysis for a more comprehensive guide). The following tables are based on the three different types of network ties discussed above – work, strategic information and policy advice. Table 3 contains overall network centralisation information. In-degree centrality is (in this case) a measure of the extent to which people choose others in relation to doing their work, gathering strategic information, or seeking advice about policy. Network centralization provides a measure of how concentrated all the ties are within a particular network, with a higher percentage of in-degree ties meaning that fewer people in a network have more of the ties directed to them, and a lower percentage indicating that the ties are more dispersed. The second measure – betweenness centrality – is an indication of the strategic importance of people within a network. A higher percentage means that fewer people provide bridging roles across the network, while a lower percentage means that more actors are playing this role.
Table 3 Centrality measures for networks (percentages)
PCP Work Strategic information Policy advice
Metropolitan
In-degree centrality 53.1 77.2 56.1
Betweenness centrality 28.4 6.0 27.8
Rural
In-degree centrality 41.5 69.9 60.9
Betweenness centrality 21.2 21.3 15.6
Table 3 indicates that strategic information is the most centralised of these three types of networks, for both PCPs. This means that fewer people are sought out for this purpose than for policy advice, and work ties are the most dispersed of the three types. The pattern differs for betweenness centrality, with the measures being approximately equal across the three types in the rural PCP, but a much smaller percentage (and therefore many more people) playing a linking role for strategic information in the metropolitan PCP. This reinforces the earlier examination of the maps, which shows the fragmentation of the rural PCP without the PCP staff.
Table 4 includes individual measures of in-degree centrality and betweenness centrality for the three highest ranked people in each case. An individual's in-degree centrality score is the number of ties received, and so, a proxy measure for how important a person is seen to be by others in terms of informal resources possessed. Betweenness centrality is the number of non-redundant ties received. That is, it is the number of single ties that connect someone to others in the network. A high betweenness centrality (a high number of these ties) means that a person is in a position to act as a gatekeeper or bridge for information to flow throughout a network.
Table 4 Centrality measures for individuals within networks – highest ranked actors
PCP Work Strategic information Policy advice
Metropolitan
In-degree centrality 2 PCP staff
PCP Chair PCP CEO
PCP Chair
1 DHS Region 2 PCP staff
PCP Chair
1 DHS Region
Betweenness centrality 2 PCP staff
PCP Chair 2 PCP staff
PCP Chair 2 PCP staff
Local Govt
Rural
In-degree centrality 3 PCP staff 3 PCP staff
1 DHS Region 2 PCP staff
1 DHS
Region
Betweenness centrality 2 PCP staff
1 Local Govt 2 PCP staff
1 DHS Region 2 PCP staff
1 DHS Region
A glance at Table 4 confirms the picture from the network maps that PCP staff are highly central in all three types of networks, and in both PCPs. In the metropolitan PCP, the Chair and one DHS regional staff member are also very central for strategic information and policy advice. In the rural PCP, one DHS regional office staff member is important for strategic information and policy advice. The betweenness centrality statistics show a similar picture, with PCP staff providing the bridges. Again the Chair in the metropolitan PCP is important, and a DHS regional officer is important in the rural case. An interesting addition is that local government staff appear to play bridging roles in both PCPs, although they are not central in terms of in-degree in either case.
Table 5 shows where the cut-points are for these three network types for both PCPs. A cut-point is a point which, if removed, causes the network to fragment. That is, it is a point at which a single, connected graph will become disconnected into two or more components. So cut-points indicate those people that are holding a network together. PCP staff, and particularly PCP CEOs feature in all three network types for the metropolitan PCP and two of the rural ones. The exception is policy advice in the rural PCP, where a DHS regional officer is the cut-point. Two local government officials in the metropolitan PCP are holding the policy advice network together, and a community health actor is important in the rural PCP in regard to work.
Table 5 Cut-points for individuals within networks
PCP Work Strategic information Policy advice
Metropolitan
PCP CEO PCP CEO
PCP PCP CEO
PCP Chair
2 Local Govt
Rural
PCP CEO
Comty Health PCP CEO
PCP DHS Region
Network narratives
Since the approach taken in this paper assumes that those involved in PCPs are not simply passive points in a network structure, but are actively creating and sustaining relationships with others in the PCP, this paper now considers how those involved in these PCPs use and value them.
Many interviewees in both PCPs argued that relationships between agencies, prior to the establishment of PCPs, had been adversely effected by an atmosphere of competition in their locality, reflecting the policy emphasis on purchaser-provider separation, compulsory competitive tendering and also local government amalgamations, under the previous state government. The following comment from the Chair in the metropolitan PCP (3), who is very central (see Figures 1 and 2), indicates how relationships were shaped by this:
"Relationships with other agencies were often strained. It was often the case that you were trying to work out what other agencies might be doing in terms of the tendering process."
A hospital person in the rural PCP (215), situated on the top edge of Figures 3 and 4 remarked on the previous lack of working together:
"In the pre-PCP environment we all conceptualised it, we all sat around and dreamed. That's as far as it got."
Everybody interviewed in these two PCPs thought they had greater engagement with other agencies than prior to the PCP, and almost all saw this as positive. The metropolitan PCP CEO (1), who is very central in Figures 1 and 2, commented:
"We've got people around the table who haven't been around the table before ... through the regular meetings and forums ... you can eye-ball someone and know who it is"
A less positive comment came from one (fairly peripheral) local government actor:
"I don't think that people have got time to waste ... I think it's just quite cumbersome to go to meeting after meeting and run into the same people" (14)
In the rural PCP, one community health actor (206) said:
"So networking is what it's all about and it really has opened a lot of doors as to who is out there, what they do and I guess getting them to recognise that they do have a health promotion role."
A telling comment on connections to the divisions of general practice, made by a DHS regional officer (208) was:
"I've probably dealt more with the Division of GPs in the last six months than I have in the previous five years and, even though they have a strong relationship with a few hospitals, there was never a need or a perceived need to talk to them directly but now that I've made the contacts I talk to them about lots of different things."
Finally, comments about the quality of relationships revealed that most people valued the partnerships and were using them for a variety of purposes. Several spoke of how trust had been built through the PCPs, leading to opportunities to do more things together. One of the weakly linked peripheral people (10 on figures 1 and 2) claimed:
"I think we're a lot better off, we're a lot better networked."
Another peripheral person (18) said:
"I think what happens now is there's more of a commitment and an understanding for working together rather than just knowing each other."
The division of general practice person in the rural PCP (219) argued:
"The PCP basically formalised that into a process with MOUs or contracts to in fact strengthen the ties or get people to have a better understanding of what was going on."
One centrally located local government person (205 in figures 3 and 4) said:
"I feel that the hospitals aren't as committed as the other agencies. The hospital management, I think they still see their primary role as, obviously, acute care and everything else is a tack on...."
The rural PCP Chair (217) said:
"I think it's about establishing networks. Because you know more people involved in different services, if something crosses your desk and it might be a funding application or whatever, if it rings bells you think, 'Oh yes I could talk to so and so about that, that links in to this program and we could do so and so together.' "
These comments add agency to the network structures discussed earlier, highlighting how relationships have changed and developed through the PCPs. Interestingly, a number of people were using networks as a concept in their comments and some spoke in quite specific network or partnership terms (overcoming boundaries and cut offs, more and better networking, developing relationships, working together, strengthening ties). Comments about the divisions of general practice and hospitals also resonate with their positions on the network maps (tending to be around the edges, as could be expected) and the network measures, which show they are not as central as local government, community health and DHS regional officials. They are the most likely to be disconnected or weakly linked without the intervention of PCP staff.
Discussion and conclusion
Clear indications of how these partnerships were progressing in their first year can be drawn from this analysis. The aim of PCPs is to improve health and well being through better service coordination and health promotion planning. A first step towards this is engaging more people and building stronger relationships. Mapping reported ties provides a useful means for assessing structure and where the strengths and weaknesses of partnerships lie. Using network analysis techniques such as measuring centrality of different kinds and the points at which networks fragment, clearly highlights the overall concentration and dispersion of different types of networks, and the importance of particular individuals. And narrative descriptions of partnerships provide insights on agency – in this case, whether those involved are using and valuing the partnerships.
The focus in this paper is on evaluating new governance modes which could ultimately lead to better outcomes through a coordinated service where agencies understand each others roles and have ongoing relationships. This is intuitively better than one based on fragmentation, lack of information and few relationships, but the benefits need to be settled by empirical examination. The approach outlined here, based on analysing network structures and narratives, is useful for examining these changes. It is particularly valuable as a means for analysing linkages that should indicate increasing capacity in the early stages of such policies when clear improvements in outcomes might be some time off. But this approach cannot reveal whether outcomes have improved, and the intention of this paper is not to suggest that it can.
This paper has combined structural and agency-based explanations by collecting information on both and examining them as separate but related parts of the same puzzle. This introduction to some of the techniques available for collecting, visualising, and analysing network data, highlights their potential for researching partnerships. The toolbox of social network analysis is much bigger than this overview indicates, and the many and varied measurement techniques can be found in the growing literature on methods [11-13], in specialist journals (most notably Social Networks), and the websites of relevant associations, such as the International Network for Social Network Analysis [19]. Enthusiasm for these techniques needs to be tempered with considerations of agency. Conversely, a focus on narratives holds much appeal but tends to ignore structure. Both are required, as a reliance on either goes only half way to a strong approach to researching partnerships.
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
The author gratefully acknowledges the assistance of a research fellowship, funded by VicHealth and the Victorian Department of Human Services.
==== Refs
Management Advisory Committee Connecting government Whole of government responses to Australia's priority challenges 2004 Canberra: Commonwealth of Australia
Lewis JM Health policy and politics: networks, ideas and power 2005 Melbourne: IP Communications
Geddes M Benington J Local partnerships and social exclusion in the European Union 2001 London: Routledge
Lewis JM eds Partnerships, primary health care and health inequalities Australian Journal of Primary Health 2004 10 38 45
O'Toole L Treating networks seriously: practical and research-based agendas in public administration Public Administration Review 1997 57 45 51
Bauld L Judge K Barnes M Benzeval M Mackenzie M Sullivan H Promoting social change: the experience of Health Action Zones in England Journal of Social Policy 2005 34 427 445 10.1017/S0047279405008858
VicHealth The Partnerships Analysis Tool accessed on 22nd September 2005
Hay C Political analysis: a critical introduction 2002 London: Palgrave
Giddens A The constitution of society: outline of the theory of structuration 1984 Cambridge: Polity
Jessop B Institutional (re)turns and the strategic-relational approach Environment and Planning A 2001 33 1213 1235 10.1068/a32183
Wasserman S Faust K Social network analysis: methods and applications 1994 New York: Cambridge University Press
Degenne A Forse M Introducing social networks 1999 London: Sage
Scott J Social network analysis: a handbook 2000 London: Sage
Department of Human Services Primary Care Partnerships: going forward 2000 Melbourne: State of Victoria
Burt RS Network items and the General Social Survey Social Networks 1984 6 293 339 10.1016/0378-8733(84)90007-8
Straits BC Ego's important discussants or significant people: An experiment in varying the wording of personal network name generators Social Networks 2000 22 123 140 10.1016/S0378-8733(00)00018-6
Borgatti S Everett M Freeman L UCINET 60 2002 Natick MA: Analytic Technologies
Australian Institute for Primary Care Evaluation of the Primary Care Partnership Strategy, Baseline Report, July 2002 2002 Melbourne: La Trobe University
The International Network for Social Network Analysis website
|
16209718
|
PMC1276782
|
CC BY
|
2021-01-04 16:38:28
|
no
|
Aust New Zealand Health Policy. 2005 Oct 7; 2:22
|
utf-8
|
Aust New Zealand Health Policy
| 2,005 |
10.1186/1743-8462-2-22
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2441620737710.1186/1471-2105-6-244SoftwareASPIC: a novel method to predict the exon-intron structure of a gene that is optimally compatible to a set of transcript sequences Bonizzoni Paola [email protected] Raffaella [email protected] Graziano [email protected] DISCo, University of Milan Bicocca, via Bicocca degli Arcimboldi, 8, Milan, 20135, Italy.2 Dipartimento di Scienze Biomolecolari e Biotecnologie, University of Milan, via Celoria, 26, Milan, 20133, Italy.2005 5 10 2005 6 244 244 26 5 2005 5 10 2005 Copyright © 2005 Bonizzoni et al; licensee BioMed Central Ltd.2005Bonizzoni et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background:
Currently available methods to predict splice sites are mainly based on the independent and progressive alignment of transcript data (mostly ESTs) to the genomic sequence. Apart from often being computationally expensive, this approach is vulnerable to several problems – hence the need to develop novel strategies.
Results:
We propose a method, based on a novel multiple genome-EST alignment algorithm, for the detection of splice sites. To avoid limitations of splice sites prediction (mainly, over-predictions) due to independent single EST alignments to the genomic sequence our approach performs a multiple alignment of transcript data to the genomic sequence based on the combined analysis of all available data. We recast the problem of predicting constitutive and alternative splicing as an optimization problem, where the optimal multiple transcript alignment minimizes the number of exons and hence of splice site observations.
We have implemented a splice site predictor based on this algorithm in the software tool ASPIC (Alternative Splicing PredICtion). It is distinguished from other methods based on BLAST-like tools by the incorporation of entirely new ad hoc procedures for accurate and computationally efficient transcript alignment and adopts dynamic programming for the refinement of intron boundaries. ASPIC also provides the minimal set of non-mergeable transcript isoforms compatible with the detected splicing events. The ASPIC web resource is dynamically interconnected with the Ensembl and Unigene databases and also implements an upload facility.
Conclusion:
Extensive bench marking shows that ASPIC outperforms other existing methods in the detection of novel splicing isoforms and in the minimization of over-predictions. ASPIC also requires a lower computation time for processing a single gene and an EST cluster. The ASPIC web resource is available at .
==== Body
Background
The completion of several genome projects has, rather surprisingly, revealed that despite a remarkable heterogeneity in organism complexity and genome size, the variation in total gene number is much less pronounced, with a less than a 10-fold increase in gene number between prokaryotes (e.g. E. coli) and vertebrates (e.g. human) [1].
However, the level of protein complexity in humans and other vertebrates is much higher than expected from the estimated gene number. Alternative splicing, leading to the generation of multiple transcripts from single genes, is believed to be the major mechanism expanding protein diversity in higher organisms [2]. These transcripts can differ both in the untranslated (UTR) and in coding regions. Thus, using a different combination of donor and acceptor splice sites, transcripts encoding different proteins can be produced with alternative UTRs regulating their fate in the cell. Indeed, recent large scale genomic studies have shown that alternative splicing occurs in 40–60% of human genes [3] and that it is a likely determinant of species-specificity since an unexpectedly low level of alternative splicing pattern conservation has been observed in pairs of orthologous genes [4]. Recent studies have also shown that alternative splicing is important for determining developmental- and tissue-specific- gene expression [5,6]. Aberrant splicing forms are also associated with human diseases [7]. For these reasons, there is a growing interest in the high-throughput identification of alternative splicing forms in human and other organisms [8].
Recently, there has been a growing interest in the design of computational methods to predict alternative splicing. Published methods may be classified in three groups: methods based on the comparison of expressed sequences to each other (i.e. [9], [10], [11]), methods based on the alignment of ESTs to the genomic sequence [12-14] and more recently methods that combine the previous two approaches, i.e. EST comparison and genome comparison, as proposed in [15] and [16]: we call such methods multiple EST alignment methods. A wide ranging discussion of the limitations of the first two methods has been presented and it has been shown that combining the two approaches leads to clear improvements in alternative splicing identification [16]. Computational methods may be also classified according to the computational approach used to produce EST alignments. Indeed, it must be pointed out that the majority of tools uses BLAST, sim4 or most recently BLAT to map ESTs to the genome (see Table 1 in [11]). These tools are often error prone when aligning ESTs because they have not been designed to consider either the relationship between ESTs and their corresponding genomic sequences or sequencing errors in ESTs – for example the presence of large gaps, short exons or specific constraints on the alignment near intron boundaries.
Table 1 Benchmark comparison of ASPIC with other similar tools
ASPIC ASAP ASD ACEVIEW
GENE #introns (#novel) #TS #EST/splice #introns (#ASPIC) #TS #introns (#ASPIC) #TS #introns (#ASPIC) #TS
ABCB10 12(0) 2 12.42 12(12) 1 Not Found 13(12) 3
ACADM 21(1) 15 31.52 15(14) 6 Not Found 22(20) 14
ACTN2 23(0) 28 19.09 20(20) 1 22(22) 4 23(23) 8
ADAM15 41(4) 67 40.07 13(13) 4 29(29) 11 56(37) 25
ADAMTS4 8(0) 3 8.63 7(7) 1 8(8) 4 8(8) 3
ADORA1 13(1) 10 4.69 8(8) 4 12(12) 9 10(9) 7
ADORA3 15(13) 5 6.13 2(2) 2 Not Found 3(2) 3
AGL 40(1) 12 14.48 38(38) 1 Not Found 39(39) 10
AGRN 41(4) 21 16.98 35(32) 1 Not Found 45(37) 11
AGT 7(2) 17 52.86 4(3) 1 Not Found 9(5) 8
AHCYL1 26(4) 35 48.50 19(19) 5 19(19) 4 26(22) 13
AKR7A2 9(1) 6 73.33 6(6) 1 7(7) 2 31(8) 15
ALDH9A1 17(4) 10 39.29 11(11) 2 Not Found 16(13) 7
ALPL 15(0) 8 19.47 14(13) 3 13(13) 3 17(15) 9
AMPD1 14(0) 4 7.64 13(13) 1 Not Found 45(14) 12
ANGPTL1 6(0) 3 10.50 5(5) 1 Not Found 6(6) 4
ANGPTL3 6(0) 5 19.17 6(6) 2 Not Found 8(6) 7
ANXA9 15(1) 3 14.40 13(13) 1 14(13) 2 16(14) 6
AP4B1 18(0) 22 14.61 12(12) 1 17(16) 12 16(16) 14
APCS 2(1) 5 62.50 1(1) 1 Not Found 1(1) 1
ARHGEF2 32(1) 37 15.19 22(22) 3 26(25) 6 35(31) 17
ARHGEF11 47(1) 9 7.70 42(42) 2 41(40) 6 46(45) 17
ARHGEF16 14(2) 12 18.64 10(10) 1 Not Found 15(12) 5
ARNT 26(1) 14 11.73 20(18) 1 22(21) 3 38(26) 14
ARPC5 4(1) 2 120.75 3(3) 1 4(2) 2 6(3) 4
ARTN 8(0) 10 5.25 7(7) 4 6(6) 3 7(7) 10
ATAD3A 22(2) 10 36.41 16(16) 2 Not Found 58(21) 27
ATP1B1 11(2) 12 59.27 7(7) 1 10(9) 4 11(8) 10
ATP2B4 29(3) 11 8.07 22(22) 5 23(23) 3 26(26) 14
Clorf10 2(0) 1 7.00 2(2) 1 Not Found 2(2) 2
Clorf26 22(0) 7 7.91 17(17) 1 Not Found 23(22) 6
C1QB 3(1) 3 25.67 5(2) 3 5(2) 3 6(2) 5
CAPZA1 15(4) 15 68.40 11(11) 2 10(9) 3 12(11) 7
CTRC 9(1) 4 36.22 8(8) 1 Not Found 8(7) 3
DMRTA2 2(0) 2 1.00 1(1) 1 Not Found 2(2) 1
DPH2L2 12(1) 11 31.25 10(10) 7 12(11) 12 12(11) 14
EPHA2 20(1) 8 13.45 16(16) 1 17(17) 7 20(19) 8
EYA3 20(0) 9 11.40 15(15) 1 Not Found 21(20) 10
FBXO2 9(0) 6 13.67 9(8) 2 6(5) 3 9(8) 5
FCGR3B 7(0) 5 22.57 4(4) 1 Not Found 7(7) 6
FUCA1 11(3) 8 18.00 8(8) 2 Not Found 7(7) 2
GBP2 17(3) 8 27.82 12(12) 2 Not Found 26(14) 10
GMEB1 12(1) 5 16.67 9(9) 2 11(11) 3 11(11) 6
HNRPR 20(2) 38 45.70 16(15) 7 12(12) 7 21(18) 17
LGALS8 22(2) 25 16.00 12(11) 4 13(13) 4 27(19) 21
LRRN5 3(0) 3 3.00 5(3) 2 Not Found 5(3) 4
LYPLA2 15(0) 14 96.07 14(14) 6 Not Found 15(14) 14
MASP2 11(0) 5 7.00 11(11) 2 11(11) 5 11(11) 6
MOV10 35(4) 42 25.29 29(29) 7 24(23) 8 33(31) 21
NPPB 2(0) 1 30.50 2(2) 1 Not Found 3(2) 3
PAFAH2 18(3) 11 15.06 12(11) 2 13(13) 5 16(14) 8
PALMD 8(0) 9 35.63 7(7) 1 Not Found 9(8) 8
PEX10 12(1) 13 21.58 6(6) 1 10(10) 7 12(10) 11
PINK1 10(2) 15 40.20 7(7) 2 Not Found 9(8) 10
PTPRU 38(3) 15 12.89 20(20) 1 Not Found 35(35) 9
RHOC 17(3) 5 8.35 13(2) 7 15(1) 9 39(14) 31
SDC3 6(2) 5 5.33 4(3) 2 8(4) 5 9(5) 6
SDHB 11(0) 12 97.27 9(9) 3 Not Found 13(11) 11
SERPINC1 12(2) 7 18.75 8(8) 2 8(8) 3 16(10) 11
SFPQ 12(3) 11 74.75 9(9) 1 9(9) 3 17(9) 25
TARDBP 21(3) 20 29.38 15(13) 4 9(9) 4 18(16) 15
TCN2 13(1) 10 26.15 9(9) 2 12(12) 4 13(10) 11
TOR3A 15(2) 9 19.20 9(9) 4 11(11) 8 15(13) 12
VAMP3 5(0) 3 80.60 6(5) 2 6(5) 3 7(5) 10
Total 1009(94) 11.9 28.3 753(721) 2.3 495(461) 5.1 1194(905) 9.7
ASPIC results from a random sample of 64 human genes from Chromosome 1 compared to those from the ASAP, ASD and AceView resources. The first column reports the HUGO name of the examined gene. The ASPIC data include the total number of predicted introns (novel introns in brackets), the minimum number of compatible transcripts and the average number of ESTs supporting gene splices. Introns with 2 non canonical splices are accepted by ASPIC only if confirmed by at least two ESTs. For the other resources the number of predicted introns (in brackets those also predicted by ASPIC) and the minimum number of compatible transcripts are reported. Other resources: AceView (July 2003 and August 2004 releases), ASD (July 2004) and ASAP (July 2004).
In this paper we propose a method that is not based on traditional BLAST-like (or BLAT-like as in [17]) alignment tools for spliced alignment, but which relies on a new heuristic for multiple EST alignments that allows – as in [12] – the use of a high number of insertions/deletions and specific scoring criteria for the spliced alignment in order to generate more accurate splice site predictions (see [18]). Indeed, even recent tools such as BLAT [19] produce erroneous alignments when used for EST-genome comparison as observed in [17] and require further corrections to the alignments produced. For example BLAT tends to create many small gaps in the alignment in cases of low sequence quality.
Through a combined analysis of all EST data and their genomic alignments our heuristic method aims to reduce over predictions of splice sites due to EST sequence errors or erroneous single EST alignments. This goal is achieved by minimizing the set of splice sites that is compatible with a multiple alignments of all transcript data. This approach overcomes the limitations of methods that (incorrectly) assume independency of single transcript-genome alignments. Indeed, tools based on independent single EST alignments (for example, Spidey [14] and Squall [20]) may produce false splice forms that would not be supported by a combined multiple alignment of all ESTs against the genomic sequence.
Implementation
Methods
Our method is based on the formalization of the problem of detecting splice sites as an optimization problem (Multiple EST Factorization Compatibility, MEFC) as proposed in [15]: it implements an heuristic that extends – and greatly improves – a basic algorithmic approach proposed by the same authors in [15]. An evident shortcoming of computational methods to predict splice sites is represented by the large number of false positive predictions produced by these methods. To overcome this limitation, we propose that an optimization criterion may be required to construct a multiple transcript alignment: the objective function of such a criterion is to minimize the number of exon predictions and hence of alignment-inferred splice sites. There is theoretical evidence for this assumption which is also supported by several real cases encountered while analyzing EST alignments. Indeed, such an optimization criterion is required when there are multiple possible adequate alignments of an EST region (or candidate exon) to the genomic sequence, even when restrictive rules are used (i.e. GT – AG splice sites) to restrict the alignment to biologically plausible solutions. The use of the optimization criterion, the combined EST analysis and the fact that our method is entirely based on a novel alignment procedure all differentiate our approach from those previously presented. The method we propose here is also different from the ones suggested in [21] and [11] where a combined analysis of EST alignments is done after all EST alignments have been generated. The method we propose also aims to reduce the computational time as in [20], while retaining a high accuracy of predictions. It is specifically designed to process a whole gene and large number of ESTs – the databases currently contain about 6 millions human ESTs and the number is growing rapidly. As shown in [20], computational times for a single EST alignment may range from a fraction of a second to the several seconds required by programs such as sim4 [22].
The software tool ASPIC (Alternative Splicing PredICtion) has been designed and implemented in a user-friendly web-server accepting as input a gene sequence and transcript data, typically a Unigene cluster related to the gene. Major features of ASPIC include its applicability to the analysis of splice variants in several organisms, and the fact that it collects together several sources of information on splice sites in a single web-based tool.
ASPIC also provides a minimal set of transcript isoforms explaining all alternative splice events occurring among the set of transcripts considered. Furthermore, it includes a module for detecting and scoring splice junctions (canonical and non-canonical) by using quality measures based on [18] and [23]. An extensive benchmark comparison of ASPIC with respect to other similar tools [24,25] shows that our method calculates the location of splice sites with high sensitivity and accuracy but still retaining an high computational efficiency such that in [20]. Remarkably, ASPIC differently from [20] combines EST alignment to splice site prediction.
Algorithm overview
In the following, we will use the term EST to denote a transcript and genomic sequence to refer to a gene related to a set of transcripts. We will use G to denote a genomic sequence, that is, a sequence over alphabet Σ = {A, C, G, T} ∪ {N}, with N denoting any nucleotide. Genomic sequences containing sequence repeats or short exons may be alignable to the same EST sequence in a number of equally probable ways. This fact further complicates the problem of identifying the correct exon-intron structure. However, it is reasonable to assume that a correct exon-intron structure can be obtained by aligning all EST sequences so that regions that are common to different ESTs are aligned to the same region of the gene. This assumption leads to the framing of the problem of predicting gene structure from a set of ESTs as an optimization problem as introduced in [15] with the MEFC problem (Minimum EST Factorization Compatible with a genomic sequence). In this context, the gene structure prediction problem has an instance consisting of a set of EST sequences and a genomic sequence: the question is to compute the constitutive exons of the genomic sequence and the factorization of each EST into such genomic exons with the objective of minimizing the number of predicted exons.
In fact, as illustrated in the examples below, a minimum length exon-factorization of a genomic sequence would forbid multiple unsupported EST alignments. However, with real data, situations frequently occur where multiple EST alignments are generated and additional criteria to find an exon-factorization are required, thus justifying (as discussed in the following sections) the use of the optimization criterion in our method.
1. Terminal EST factors may be short (10–30 bp in length) and may have multiple plausible alignments to the genomic sequence, particularly when the EST sequence contains errors.
2. Part of a factor may be repeated along the genomic sequence. A theoretical example of this situation, and how optimization may be used to find correct predictions, is reported in Fig 1. Additional file 1 illustrates a specific example of this situation, occurring in the Unigene cluster related to the human AMY2A gene.
Figure 1 The figure illustrates two gene-factorizations into 7 and 4 pseudo-exons of the genomic sequence G. Let S1, S2 and S3 be EST sequences in S agreeing to the genomic sequence G, where sequence S1 = ABDEF, S2 = ABCDE and S3 = BDEFG, each letter in {A, B, C, D, E, F, G} denotes a sequence (A). In (B) and (C) two alternative EST-genome alignments of sequences S1, S2 and S3 are represented: each EST factorization of Si associated with the EST-genome alignment is shadowed. Pseudo-exons in the gene-factorization are colored white, while introns are in grey. Segments labelled by letters represent regions of the genomic sequence that align to a substring of the input sequence of the corresponding letter. Note that an approach that aligns independently each sequence S1, S2 and S3 to G, one after the other, may produce the gene-factorization <A, B, C, D, F, E, G> consisting of 7 pseudo-exons (B), while the one minimizing the number of pseudo-exons provides only 4 pseudo-exons (C). Indeed, there are EST factorizations of each Si that are compatible or variant compatible with the gene-factorization GE = <AB, C, DE, FG>. More precisely, <AB, DE, F> is an EST-factorization of S1 that is compatible to GE. Then <AB, C, DE> is an EST-factorization of S2 compatible to GE. Finally, <B, DE, FG> is an EST-factorization of S3 compatible with GE (C).
3. Short repeats may occur in the genomic sequence and EST sequences may contain errors near splice junctions.
The MEFC problem: definition
In the following we introduce some basic notions that allow us to define the MEFC problem and describe the method we propose to face it.
We recall that there are four main patterns of alternative splicing that potentially may occur in nature [2]:
1) exon-skipping; 2) mutually exclusive exons; 3) competing 5'/3' ends; and 4) intron retention. While the first two splicing modes simply determine whether an exon is used or not during splicing, in the third mode the transcript splicing variants derive from competing partially overlapping exons. Finally, intron retention occurs when an exon is present in a transcript, while in another it appears with a missing internal region.
Then, a gene factorization GE of G is a sequence <f1, ..., fn> of n substrings fi of G, we define pseudo-exons, such that G is given by the concatenation of the pseudo-exons fi interspersed by other substrings called introns. In particular, a pseudo-exon defines a contiguous genome region corresponding to and/or containing one or more exon splice variants.
An EST factorization of an EST sequence S is an ordered sequence <s1, s2, ..., sk> such that S = s1s2 ... sk, where each substring si is called a factor of the EST S. The edit distance ed(x, y) between two sequences x and y measures the number of mismatches in the alignment of x and y.
We define an EST factorization <s1, s2, ..., sk> compatible with a gene-factorization GE of a genomic sequence G if there exists a sequence of genomic pseudo-exons fi1,fi2,fi3,…,fik
MathType@MTEF@5@5@+=feaafeart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabigdaXaqabaaaleqaaOGaeiilaWIaemOzay2aaSbaaSqaaiabdMgaPnaaBaaameaacqaIYaGmaeqaaaWcbeaakiabcYcaSiabdAgaMnaaBaaaleaacqWGPbqAdaWgaaadbaGaeG4mamdabeaaaSqabaGccqGGSaalcqWIMaYscqGGSaalcqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdUgaRbqabaaaleqaaaaa@41F0@ of G such that for each factor sj, with 2 ≤ j ≤ k - 1, ed(sj, fij
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdQgaQbqabaaaleqaaaaa@311D@
) is bounded by a given parameter bound, factors s1 and sk differ from a suffix of pseudo-exon fi1
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabigdaXaqabaaaleqaaaaa@30B0@
and a prefix of fik
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdUgaRbqabaaaleqaaaaa@311F@
, respectively, by a number of alignment mismatches bounded by bound.
Because of alternative splicing, we further provide the notion of EST factorization variant compatible with a gene-factorization GE. This is simply obtained by requiring in the previous notion that ed(sj, factor(fij
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdQgaQbqabaaaleqaaaaa@311D@
)) is bounded by a given parameter bound, where factor (fij
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdQgaQbqabaaaleqaaaaa@311D@
) is a prefix, suffix or even a proper factor of the pseudo-exon fij
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdQgaQbqabaaaleqaaaaa@311D@.
An EST factor sj, corresponding to a gene exon factor(fij
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaacqWGMbGzdaWgaaWcbaGaemyAaK2aaSbaaWqaaiabdQgaQbqabaaaleqaaaaa@311D@) is defined as internal or external depending on whether both donor and acceptor splices are or are not present, respectively at its genome boundaries after alignment. Thus, factors s1, sk of the EST factorization <s1, s2, ..., sk> are called external factors while s2, ..., sk-1 are called internal factors.
In other words, an EST factorization is induced by an alignment of the EST to exons of the genomic sequence. Each EST factor must correspond or align to an exon. The external EST factors can correspond to a fragment (a prefix or a suffix) of the relative exons.
By using the above stated notions, the MEFC problem is defined as follows. The instance of the problem consists of a genomic sequence G and a set of EST sequences (transcripts), while a solution consists of one gene-factorization GE of G and EST factorizations that are compatible or variant compatible with GE. Thus an optimal solution in the MEFC problem (that is an optimal gene-factorization and optimal compatible EST factorizations) is the one that minimizes the number of distinct pseudo-exons in the gene-factorization of the genomic sequence.
Generation of nearly optimal compatible genome-EST alignments
The ASPIC software implements an heuristic method for the MEFC problem stated before.
The general structure of the method consists of:
(a) an initial pre-processing of the genomic sequence,
(b) two main procedural phases applying criteria to minimize splice sites.
In the following we provide a detailed description of the method by first describing the initial pre-processing phase and then the main algorithmic steps of the two phases.
Pre-processing of the genomic sequence
The alignment of a single EST factor to the genomic sequence is based on the notion of a component: a component is a substring of the genomic sequence that perfectly matches a portion of an EST factor. The length of a component is a critical parameter used to accelerate the alignment of EST factors as well as for finding error-free matching regions between ESTs and the genomic sequence. Indeed, components of a given length (for example 15 bp) may have very few occurrences on a genomic sequence, thus making the process of locating EST factors very fast. For this reason, the length of a component is computed automatically as a function of the gene sequence length, but it can be also modified by the user as an input parameter. The algorithm starts with an initial pre-processing of the genomic sequence G that consists in building a hash-table containing all occurrences of each component in G. Thus a key list of components (i.e. substrings of the genome) provides the entry of a Hash Table used to speed up the alignment process of an EST factor to the genomic sequence. Since the algorithm locates the intron regions by validating the splice sites using first the GT-AG rule, a second hash-table for all GT and AG occurrences on the genomic sequence, is initially computed and stored.
Phase 1: iterative computation of all EST internal factors
The first phase is an iterative processing of each EST in the set S = {S1, ..., Sm} such that the general i iteration produces an alignment of each EST in the set {S1, ..., Si} compatible with a partial gene-factorization of G – the generation of an EST alignment against the genomic sequence implying an EST factorization. The generic step of the iteration in our algorithm consists of finding the next factor sj of a partial EST factorization <s2, ..., sj-1> and the corresponding exon along the genomic sequence. In this phase the EST-factorization is produced using a criterion, called concatenating exons, to minimize the number of exons. This criterion consists of concatenating two or more consecutive EST factors into a unique exon whenever a true exon may have been over factorized because of repeated regions in the genomic sequence (see as an example Figure 1).
More precisely, given the alignment of the internal factors <s2, ..., si> of an EST, then the genomic alignment of a new EST factor si+1 is computed in four main steps.
In step (1) the EST suffix to be aligned after factor si is divided into consecutive strings x1, x2, ..., xn of the predefined length of a component. Indeed, the first possible genomic location of EST factor si+1 is determined by finding the leftmost string xj of the EST suffix that is a component and allows the optimal alignment of the entire EST factor si+i (see Fig. 2(a), (b)). In step (2), for each occurrence of a component xj along the genomic sequence, a genomic region of maximal length containing xj is optimally aligned in linear time and space (using the edit-distance within a Kband [26]) to the new EST factor si+1, until a compatible alignment is found (i.e. few errors are allowed and possibly canonical splice sites are located). Note that step two may fail to compute the new EST factor si+1, whenever the previous EST internal factors <s2, ..., si> do not allow the generation of an EST-factorization compatible with the partially computed gene-factorization. Indeed, some EST factors may have been incorrectly computed because of a wrong alignment of the EST sequence. Backtracking allows the relocation of exons. This consists of trying alternative occurrences in the genomic sequence of components of previous factors starting from si up to s2.
Figure 2 Location of a new EST internal factor si+1 given previous computed factors s2, ..., si. (a) Consecutive sequence components c1 ... cj are tested to find the first one that allows the identification of a genomic region that optimally aligns factor si+1 (i.e. alignment extension on one or both sides of the component): such a region is determined in (b) by the component cj. Figure (b) shows that some intervening positions (sequence x) may occur between factor si and si+1. Indeed, in this case the placement of si+1 gives the correct right end of previous factor si, since the larger factor inducing canonical splice sites on the genomic sequence can be optimally aligned before si+1 thus leading to an optimal location of both si and si+1.
Once the location of factor si+1 is determined, the concatenating exon criterion is applied in step (3) which consists of testing whether one or more consecutive EST factors preceding factor si+1 can be concatenated to si+1 to obtain a unique factor s such that it optimally aligns to the genomic sequence. In this case, s replaces a list of consecutive EST factors, thus minimizing the number of exonic regions in the gene-factorization (see for example exons AB and DE in Figure 1(C) produced by the application of concatenating exon criterion to A and B first, and then to D and E). Clearly, after the minimization, the new EST factor si+1 as well as previous factor si are redefined so that the EST alignments define a smaller number of exons.
Finally, in step (4), a dynamic programming (DP) algorithm is used to refine the intron boundaries between the defined EST factors si and si+1. This crucial step of the algorithm is detailed in the next section Refining intron boundaries.
Observe that the location of a new EST factor si+1 is based on the use of a single component (that is a perfect matching region) and that such a component is located on the factor by testing consecutive positions in the EST suffix after factor si. This approach may imply that several positions after the right end of EST factor si are skipped before placing the left end of the new factor si+1. Indeed, in such cases the placement of factor si+1 may imply an extension (or a reduction) of the right end of previous factor si thus optimizing exon definition (see Fig. 2(c)). This strategy makes the alignment process more flexible and faster with reference to other approaches (such as BLAT [19]) that apply strict matching criteria.
Indeed a feature of ASPIC alignment algorithm is that it allows a fast exact location of the alignment regions of EST factors without necessarily comparing all EST sequences against large portions of the genomic sequence. Consequently, ASPIC also allows EST alignment in the presence of a relatively high number of errors that are located in specific regions. Moreover, even though the alignment process relies on dynamic programming (DP) it turns out to be very fast in most of the cases, as indeed DP is only applied to short portions of the EST and genome sequence.
Phase 2: refining internal factors and placing external factors
This phase of the algorithm completes the computation of all EST factorizations (i.e. EST alignments) by first correcting all internal EST factors pre-computed in the first phase in order to make all factorizations compatible with the same gene-factorization GE of G minimizing the number of splice sites. More precisely, the minimization relies on the use of a criterion called merging splice sites. Merging splice sites consists of comparing computed exons x and y supported by EST factors to reduce the intron boundary of x to the one of y or vice versa, whenever they differ at only a few positions, likely because of sequencing errors in the EST factors (see an example in Fig. 3). Clearly, this step may avoid over prediction of splice sites due to the erroneous location of intron boundaries because of sequencing errors. This criterion is also implemented to allow the detection of possibly true splice variants determined by competing 3' or 5' junctions induced by few bases (two bases or more).
Figure 3 Example of intron detection in the human ATP1B1 (UG:Hs.291196) gene without (A) or with (B) the refinement of exon-intron boundaries. The first row shows the genomic sequence aligned to the EST sequences (below). In (A) four different introns are detected (A, B, C, D) that can be merged to only two (A, D) in B. Absolute coordinate (NCBI 35 assembly) are shown for each intron and acceptor/donor splice sites are in black-background.
Finally, after the localization of EST internal factors, all EST external factors are computed. The concatenating exons and merging splice sites criteria are used again since errors in EST sequences are more prevalent in terminal regions, which may be as short as few bases – thus permitting several alternative alignments. The procedure that finds external EST factors tries to align the EST leftmost (or rightmost) factor as a suffix (or a prefix) of some previously computed exon. If that is not possible, the factor is placed in a new location in correspondence with a GT (or AG) pattern and then the DP algorithm is used again to refine intron boundaries.
Refining exon-intron boundaries
Because of sequence repeats and sequencing errors in ESTs, the exact location of splice junctions is a critical issue [27]. Our method combines different strategies to evaluate and hence improve the quality of splice data produced. These are listed below:
1. Finding intron boundaries via dynamic programming. A first criterion used to find the exact location of intron boundaries is the evaluation of alignment quality. We have designed an algorithm, based on dynamic programming (DP), to produce optimal alignments of regions close to splice sites. It computes the genomic alignment of a suffix w and a prefix y of two consecutive EST factors, si and si+1, in order to locate in the genomic sequence the optimal position for a single large gap corresponding to the intron region. This gap may not be delimited by canonical splice sites following the GT – AG rule, which is recognized as a basic one for the validation of splice sites, as more than 98.7% annotated splice sites in GenBank are canonical in this respect [18]. Indeed, there may be different optimal alignments leaving a gap with the same error rate. Thus a second important algorithmic step is applied by ASPIC to locate splice sites.
2. Canonical patterns and weight matrices. Whenever the optimal alignment computed via DP does not lead to canonical splice junctions, then the algorithm looks for alternative alignments with the same error rate with preference for the couple of splice boundaries more frequently represented in the weight matrix provided in [18] (see Table 2 in [18]). If different alignments of the same quality (i.e. number of errors) are possible near intron boundaries, the choice of the alignment is done by using the weight matrix. For example, the base-pairs GC-AG are selected before the pair AT-AG if compatible with an alignment of splice sites leaving the same number of errors, as GC-AG is more frequent than AT-AG in the weight matrix. Clearly, an high quality alignment may also lead to the acceptance of splice sites with null frequency in [18] matrices.
Table 2 Splice sites in known and novel ASPIC-predicted introns
Splice Site Known introns Novel introns
N % N %
GT-AG 897 98.14 57 60.64
GC-AG 8 0.77 15 15.96
GT-other 3 0.33 5 5.10
other-AG 4 0.44 13 13.27
other-other 3 0.33 4 4.08
Total 915 94
Number (N) and percentage (%) of splice site types in known and novel ASPIC-predicted introns.
Actually, the presence of sequencing errors may often complicate the location of the correct splice sites junctions. For these reasons, the use of agreement criteria among EST alignments turns out to be crucial in many practical cases to detect highly confirmed splice junctions and thus to correct ambiguous alignments.
Moreover, in order to evaluate the quality of splice sites we annotate each detected splice site, either donor or acceptor, with a consensus sequence and a score: the score derives from the formula and tabular nucleotide frequencies reported in [23]. Indeed, conserved splice sequences provide further evidence for splice junctions.
3. Congruence of ESTs on the location of splice sites. Since the merging splice site criterion discussed in the previous section is based on a combined analysis of all EST factorizations, it is crucial also for validating intron boundaries. Indeed, by comparing EST factors it is possible to discover sequencing errors in ESTs that show that some intron boundaries must be considered as coincident if few errors are tolerated (typically at most one error for each splice site) or even by shifting the location of canonical splice sites. For example, in many cases the GT-AG rule may be applied to locate an EST factor boundary in two very close locations of the genomic sequence, thus making the choice of the alignment near intron boundaries for a single EST difficult. In these cases, an independent EST alignment does not allow the determination of the EST splice sites, while the presence of other EST factorizations having a better quality alignment to the genomic sequence may solve the aforementioned dilemma because of the common compatibility to the exon-intron structure. This situation is detailed in the example shown in Fig. 3.
4. Filtering artifacts and locating gene strand. Our implementation has automatic procedures to locate the strand from which each EST originates (independently from the cluster annotation) and a filtering of possible artifacts and polyA ends. Moreover, EST alignments of poor quality are filtered out based on several criteria, including a percentage of sequence identity below the fixed cutoff.
As an example, Figure 3 reports the optimal alignments of ESTs close to intron boundaries illustrating the need for specific criteria to locate all plausible intron boundaries. The basic criterion is the congruence of ESTs near splice sites, combined with the use of known frequencies of splice patterns (see [18]). ATP1B1 introns B and C (Fig. 3A) can disappear by merging them to intron A (confirmed by a large number of ESTs) after the introduction of a A-insertion or of a C-deletion in the relative alignments. On the other hand, intron D is likely to represent a genuine variant. In all these cases it is likely that the relevant EST sequences are not correct due to a typical base miscalling in single-read automatic sequencing, i.e. AAA instead of AA for BG705986 and C instead of CC for BG699442.
Clustering ESTs by common splice sites
For each splice site predicted, ASPIC provides the list of ESTs supporting such splice sites, thus allowing the evaluation of the quality of the prediction in terms of number of ESTs confirming it. Moreover, this step allows the grouping of ESTs that strongly support a common transcript (by sharing the same sequence of splice sites).
Minimal set of full-length transcript isoforms
Since a feature of ASPIC is to report splice sites and corresponding factorization into genomic exons for each EST (EST-exon-factorization in our terminology), we have designed and implemented in the module Transview of ASPIC an efficient algorithm that combines EST-exon-factorization data into a set of minimal full-length transcripts that are supported by the evidence, i.e. by the set of available ESTs. Our algorithm is based on the use of directed acyclic graphs (DAG): nodes of the graph are EST-exon-factorizations, while edges connect nodes (sequences) that are related by a binary relation among EST-exon-factorization (extension). Paths in the graph represent possible full-length transcripts. Various methods based on graphs have been reported to predict transcripts from ESTs such as in [28,10] and [17]: our method is different from those approaches in the construction of the graph as well as in the way the graph is visited to report full-length transcripts. In contrast to graph based approaches proposed in [17] or [11] where nodes are exons or nucleotide sequences, our approach uses a reduced graph and an efficient visiting process that allows the reporting of all plausible paths, without requiring a trimming phase as in [17] to remove redundant models. Indeed, our algorithm aims to reduce over predictions or false positives as well as to reduce the execution time required by the construction of a potentially exponential number of paths (putative full-length transcripts) in the graph. Moreover, the construction of the graph in our model is guided by input parameters that allows the user to specify the quality of predicted full-length transcript with respect to the set of transcripts supporting them.
Transview provides a visualization of full-length isoforms and for each predicted full-transcript their composition in terms of the ESTs that support the full-transcript. Details on the algorithm will be discussed elsewhere.
Results
The capability of ASPIC method to computationally produce high quality gene predictions has been tested by performing two types of experiments. A first experiment consisted in comparing ASPIC data with data available from other database sources that collect intron-exon data obtained through computational as well as experimental methods. This first experiment shows the ability of ASPIC in predicting novel splice variants as well as in detecting good quality splice sites confirmed by other sources. In order to assess the quality and reliability of novel predictions, a second experiment has been carried out: this one consisted in comparing ASPIC data with those produced within the ENCODE project [29] aimed at providing a reliable annotation of 1% of the human genome. In particular, we investigated the occurrence of false positives in ASPIC-predicted introns as determined by RT-PCR analysis for 22 genes located in 13 Encode regions.
Comparing ASPIC with other similar tools
The ASPIC method has been tested on a sample of 64 genes randomly chosen from the human Chromosome 1. Results are summarized in Table 1 where they are also compared with those obtained by other publicly available resources. A total of 1009 introns were predicted by ASPIC as compared to 753 by ASAP, 495 by ASD and 1194 by AceView. ASPIC predicted 95.7%, 93.1% and 75.8% of introns predicted by ASAP, ASD and AceView, respectively. In general, predicted introns were well supported by genome-transcript alignments with 28.3 ESTs supporting each splice site on average. Missing introns may derive from additional ESTs not present in the UNIGENE cluster used by ASPIC or by the stringent parameter thresholds adopted in ASPIC to consider an intron prediction reliable. The large number of additional introns detected by AceView, but not by other resources, are partly due to the wrong selection – in some cases – of the genomic region to be considered for the analysis. For example, AceView predicts 45 introns in the gene AMPD1 w.r.t. the 14 introns predicted by ASPIC (13 in ASAP). In this case the genome region selected by AceView encompasses 113 kb covering AMPD1 and two additional genes. A similar problem can be observed with several other genes where the number of AceView introns is remarkably higher than that detected from other resources (e.g. ADAM15, AKR7A2, ARNT, ARPC5, ATAD3A, etc.). Also, AceView intron over-prediction is likely due to the use of less stringent parameters in genome-transcript alignments, as in the example shown in Fig. 4.
Figure 4 Example of intron boundaries detected for the human AHCYL1 gene by AceView and ASPIC. The hypothetical novel intron predicted by AceView (July 2003 release) with non-canonical splices can be reduced to a known intron by a single A-insertion. Intron coordinates are referred to Ensembl release 26.35.1.
However, ASPIC detected a total of 94 novel introns, each confirmed by 2.18 ESTs on average. It is interesting to note that our data show a higher occurrence of non-canonical splice sites with respect to previous estimates [30]. Table 2 shows splice sites for known and novel ASPIC predicted introns. These data are not unexpected as previous estimates did not consider most of the splicing variants of annotated genes. While some of the predicted introns may simply be artifactual it is likely that rarer splicing isoforms involve a higher proportion of non-canonical splice sites. Another striking observation from our analysis is that 62/64 genes (97%) show alternative splicing with an average of 11.9 transcripts/gene, a value similar to that from AceView data (see Table 1) but significantly higher than 2.3 and 5.1 estimated by ASAP and ASD repectively. It is worth mentioning that data reported by ASAP are not updated w.r.t. the latest Unigene/genome data and several genes (28/64) were not annotated in ASD. It should be considered that Unigene clusters are enlarging at a great rate and genomic sequences are also continuously updated. To address this problem ASPIC data are stored in a dynamic database. The relevant data for each gene query are stored in the ASPIC database so that if another user does a similar query the results are immediately available without carrying out a new analysis. However, the user can choose to overwrite stored data with updated genome and transcript data directly extracted from Ensembl and Unigene databases. The new data remain stored in the ASPIC database until a new overwrite request for the same gene query is made.
False positive incidence of ASPIC introns
In order to compare the false positive rate of introns predicted by ASPIC and other methods we analyzed the GENCODE experimental verification of computationally predicted introns for a set of 22 genes in 13 Encode regions (see the GENCODE annotations in the Additional file 2). Of the total 44 introns not supported by RT-PCR experiments (labeled RT_negative) ASPIC supported only 12/44 whereas AceView supported 41/44 (Table 3). Interestingly, 7/12 ASPIC introns were supported by more than 2 ESTs, also showing high-scoring slice patterns (see Additional file 3). This finding suggests possible leakages in experimental validations carried out within the Encode project.
Table 3 RT-negative introns supported by ASPIC
Encode Region Gene Intron position Prediction Method
Chr Start End
ENm004 SLC5A1 22 30779886 30787475 ASPIC (3), ECgene, acembly
ENm004 PISD 22 30350425 30351061 ASPIC (1), ensEstGene
ENm004 PISD 22 30337622 30338657 acembly
ENm004 PISD 22 30346557 30351299 acembly
ENm004 PISD 22 30365972 30366216 acembly
ENm004 RFPL3 22 31075439 31078694 acembly
ENm004 SYN3 22 31727364 31734939 acembly
ENm004 TIMP3 22 31521971 31522263 acembly
ENm004 TIMP3 22 31521971 31522271 acembly
ENr223 MTO1 6 74249065 74253041 ASPIC (6), ECgene, acembly
ENr223 MTO1 6 74253206 74258677 ASPIC (5), ECgene, acembly
ENr231 PSMD4 1 148044771 148047709 ASPIC (2), ECgene, acembly
ENr231 PIP5K1A 1 148035078 148039516 acembly
ENr231 PIP5K1A 1 148035192 148035350 acembly
ENr231 PSMB4 1 148187431 148194586 acembly
ENr231 PSMD4 1 148040228 148047741 acembly
ENr231 PSMD4 1 148040358 148044611 acembly
ENr231 PSMD4 1 148044684 148047709 acembly
ENr231 PSMD4 1 148046796 148047709 acembly
ENr231 SNX27 1 148423496 148424527 acembly
ENr231 TUFT1 1 148350164 148356015 acembly
ENr231 TUFT1 1 148356492 148372163 acembly
ENr232 CRAT 9 128949911 128950731 ASPIC (1), acembly, softberryGene
ENr232 PPP2R4 9 128953625 128962345 ASPIC (1), ECgene
ENr232 PPP2R4 9 128952305 128953168 acembly
ENr232 PPP2R4 9 128952336 128953268 acembly
ENr232 PPP2R4 9 128952981 128953304 acembly
ENr232 PPP2R4 9 128953105 128953150 acembly
ENr232 SH3GLB2 9 128835746 128849868 acembly
ENr232 SH3GLB2 9 128860722 128862923 acembly
ENr323 LACE1 6 108794230 108829892 ASPIC (5), sgpGene
ENr323 LACE1 6 108747689 108751721 acembly
ENr323 SNX3 6 108688727 108690771 acembly
ENr333 RNPC2 20 33764744 33765167 ASPIC (1), ECgene, acembly
ENr333 RNPC2 20 33786418 33787848 ASPIC (1), ECgene
ENr333 CEP2 20 33527835 33529106 acembly
ENr333 CEP2 20 33554537 33568378 acembly
ENr333 CEP2 20 33561298 33568224 acembly
ENr333 ITGB4BP 20 33335958 33343927 acembly
ENr333 RNPC2 20 33776436 33777255 acembly
ENr333 RNPC2 20 33780699 33780701 acembly
ENr333 SDBCAG84 20 33585738 33593682 acembly
ENr334 TFEB 6 41766952 41811861 ASPIC (5), ECgene, acembly
ENr334 TFEB 6 41766952 41799176 ASPIC (3), ECgene, acembly
List of computationally predicted introns of 22 genes contained in 13 Encode regions (see the GENCODE annotations in the Additional file 2) but not validated by RT-PCR analysis. For each intron are shown the Encode region, the gene ID, location (NCBI 35 assembly) and prediction methods (Acembly/AceView, ; ECgene [17]; ensEstGene [28]; softberryGene [35]. For ASPIC predictions the number of supporting in shown in the brackets.
The ASPIC Web Resource
The ASPIC program can be accessed online at: . ASPIC standard input data consist of a genomic sequence and a set of transcripts. Such data are acquired either automatically or by uploading files specified by the user. In the first case, a basic form permits the input of an official HUGO gene name for the genomic sequence (e.g. ABCB10, HUGO names are permitted only for human genes) and/or a Unigene cluster identifier (e.g. Hs.1710). EST clusters are automatically retrieved from Unigene, while genome sequences are retrieved by using the API provided from Ensembl. All results presented here are based on one of the latest releases (September 2004 Ensembl API release .25 and 2004 Unigene database release).
The automatic acquisition of clusters is allowed for human and every other organism whose data may be acquired from the Ensembl database. A specific upload function allows the user to query ASPIC processing of arbitrary genomic sequences and transcript data in FASTA format.
An advanced search form allows the user to run the ASPIC program by specifying basic parameters used to produce compatible EST alignments.
We have tested our method using standard parameters suggested by experimental analysis of real data. For example, we choose a minimum exon length of 15 nt. The component length for building hash tables is computed by using a formula that relates the minimum exon length to the component length in such a way that the existence of an error-free substring in an EST factor is guaranteed.
ASPIC outputs a complete description of each EST exon-factorization, with a view of the alignment to the genomic sequence, as well as a tabulated view of splice sites. The program provides an output file that contains detailed information about all EST exon-factorizations. This file is also processed by Perl scripts in order to produce and make available to the user from the ASPIC web site: i) a table view listing all detected introns; ii) a graphical view showing the general exon-intron arrangement of the queried gene; and iii) a transcript view showing all non-mergeable transcript models compatible with detected introns. In particular, the table reports the relative and absolute coordinates of each detected intron derived from the genomic sequence and genome build considered, respectively, as well as the number of confirming ESTs. Absolute coordinates, not provided by other resources, are particularly useful for the comparison of intron coordinates for a gene to those annotated in genome browsers. The main graphical view is a visualization of the intron structure of the genomic sequence derived from the tabulated data. Such a graphical view also provides links to a visualization of the alignment of the 15 base pairs of EST sequences closest to intron boundaries. Figure 5 shows an example of the table, the graphical and the transcript view.
Figure 5 Snapshot of the ASPIC output for the gene HNRPR (human chromosome 1). The Table View (A) lists all detected introns, their coordinates and the number of supporting ESTs. The Alignment View (B) shows the alignment between genomic and EST sequences around splice sites. The Graphical View (C) provides a general scheme of the splicing pattern. The Transcript View (D) shows the minumum set of different transcripts compatible with the detected splicing patterns.
ASPIC Execution time
The performance of ASPIC has been evaluated on a Pentium IV class PC, with 256 MB of main memory running the Linux operating system.
The processing time for a single EST varied from 0.007 sec cpu time to a maximum of 2.5 sec cpu time, where the gene length varied from 5014 bp to 287011 bp, requiring on average around 71 seconds cpu time per gene. Thus ASPIC can process about 5000 ESTs in about half an hour of cpu time (against the four hours required in [16]).
Experimental results: WEB-sources
The comparison of ASPIC data with other sources of splice sites has been carried out by accessing available databases from the web at the following sites: ASD [31], ASAP [32], Acembly [33].
Conclusion
The ASPIC algorithm implements a novel methodology that optimizes the overall compatibility between genomic and transcript sequences to detect splice sites – thus minimizing mispredictions due to repetitive sequences or sequence errors in the ESTs. It does not impose constraints on the splice boundaries (i.e. strict observance of the GT-AG rule) but in case of equally likely alternative alignments adjusts splice boundaries to those observed to occur more frequently in known genes [18]. Hence, it is able to detect non-canonical splice boundaries such as those of U12-dependent introns [34] in the presence of suitable supporting transcripts (see Additional file 3). Finally, ASPIC allows the user to carry out splicing predictions on a wide range of species as well as on user-submitted genome and transcript sequences.
Availability and requirements
The ASPIC web tool is available to scientists wishing to use it at . To submit a query to ASPIC the user needs to fill a form specifying the organism, the gene ID (Ensembl or HUGO), the Unigene cluster ID (optional) and providing an email address. The request is processed by the ASPIC software and when the results are available an email is automatically sent back to the address specified by the user, providing a link to processed data.
ASPIC collects all the results of submitted queries in a dynamic database.
Project name: ASPic Alternative Splicing Prediction
Project home page:
Programming language: C
Operating system: Debian GNU/Linux 3.1, kernel 2.6.8
Other requirements: Apache 1.3, Perl 5.8.4, Php 4.3.10, MySQL 4.1, gcc 3.3.5
Authors' contributions
GP conceived the study. PB and RR designed the algorithms and the general ASPIC method. RR implemented the method, realized the web resources and performed the experimental analysis. All authors participated in the design of the ASPIC tool and the experimental study. All authors have contributed in drafting the article.
Supplementary Material
Additional File 1
Splicing site prediction with and without the optimization strategy.
Click here for file
Additional File 2
Gencode annotation of 13 Encode regions.
Click here for file
Additional File 3
RT-negative introns detected by ASPIC.
Click here for file
Additional File 4
U12 dependent introns detected by ASPIC.
Click here for file
Acknowledgements
This work was supported by FIRB projects "Bioinformatica per la Genomica e la Proteomica" and "Laboratorio Italiano di Bioinformatica – L.I.BI." (Ministero dell'Istruzione e Ricerca Scientifica, Italy), Associazione Italiana Ricerca sul Cancro and Telethon. We thank Gianluca Delia Vedova for his helpful suggestions on the preliminary design of ASPIC software, David Horner and Giulio Pavesi for helpful comments on the manuscript and Gabriele Ravanelli for providing a Perl library to visualize ASPIC data.
==== Refs
International Human Genome Sequencing Consortium IHGSC Initial sequencing and analysis of the human genome Nature 2001 409 860 921 11237011 10.1038/35057062
Graveley B Alternative splicing: increasing diversity in the proteomic world Trends Genet 2001 17 100 107 11173120 10.1016/S0168-9525(00)02176-4
Modrek B Lee C A genomic view of alternative splicing Nat Genet 2002 30 13 19 11753382 10.1038/ng0102-13
Nurtdinov RN Artamonova II Mironov AA Gelfand MS Low conservation of alternative splicing patterns in the human and mouse genomes Hum Mol Genet 2003 12 1313 1320 12761046 10.1093/hmg/ddg137
Xu Q Modrek B Lee C Genome-wide detection of tissue-specific alternative splicing in the human transcriptome Nucleic Acids Res 2002 30 3754 3766 12202761 10.1093/nar/gkf492
Xie H Zhu WY Wasserman A Grebinskiy V Olson A Mintz L Computational analysis of alternative splicing using EST tissue information Genomics 2002 80 326 330 12213203 10.1006/geno.2002.6841
Caceres JF Kornblihtt AR Alternative splicing: multiple control mechanisms and involvement in human disease Trends Genet 2002 18 186 193 11932019 10.1016/S0168-9525(01)02626-9
Boue S Vingron M Kriventseva E Koch I Theoretical analysis of alternative splice forms using computational methods Bioinformatics 2002 18 S65 S73 12385985
Brett D Hanke J Lehmann G Haase S Delbruck S Krueger S Reich J Bork P EST comparison indicates 38% of human mRNAs contain possible alternative splice forms FEBS Letters 2000 474 83 86 10828456 10.1016/S0014-5793(00)01581-7
Heber S Alekseyev M Sze S Tang H Pevzner P Splicing graphs and EST assembly problem Bioinformatics 2002 18 S181 S188 12169546
Leipzig J Pevzner P Heber S The Alternative Splicing Gallery (ASG): bridging the gap between genome and transcriptome Nucleic Acids Res 2004 32 3977 3983 15292448 10.1093/nar/gkh731
Brendel V Xing L Zhu W Gene structure prediction from consensus spliced alignment of multiple ESTs matching the same genomic locus Bioinformatics 2004 20 1157 1169 14764557 10.1093/bioinformatics/bth058
Kan Z Rouchka EC Gish WR States DJ Gene structure prediction and alternative splicing analysis using genomically aligned ESTs Genome Res 2001 11 889 900 11337482 10.1101/gr.155001
Wheelan SJ Church DM Ostell JM Spidey: a tool for mRNA-to-genomic alignments Genome Res 2001 11 1952 1957 11691860
Bonizzoni P Pesole G Rizzi R A method to detect gene structure and alternative splice sites by agreeing ESTs to a genomic sequence Proc WABI Lectures Notes in Bioinformatics 2003 2812 63 77
Grasso C Modrek B Xing Y Lee C Genome-wide detection of alternative splicing in expressed sequences using partial order multiple sequence alignment graphs Pac Symp Biocomput 2004 29 41 14992490
Kim N Shin S Lee S ECgene: Genome-based EST clustering and gene modeling for alternative splicing Genome Res 2005 15 566 576 15805497 10.1101/gr.3030405
Burset M Seledtsov IA Solovyev VV Analysis of canonical and non-canonical splice sites in mammalian genomes Nucleic Acids Res 2000 28 4364 4375 11058137 10.1093/nar/28.21.4364
Kent WJ BLAT-The BLAST-Like Alignment Tool Genome Res 2002 12 656 664 11932250 10.1101/gr.229202. Article published online before March 2002
Ogasawara J Morishita S A Fast and sensitive algorithm for aligning ESTs to the human genome Journal of Bioinformatics and Computational Biology 2003 1 363 386 15303288 10.1142/S0219720003000058
Wheeler RL A method of consolidating and combining EST and mRNA alignments to a genome to enumerate supported splice variants Proc WABI 2002, Lectures Notes in Computer Science 2002 2452 201 209
Florea L Hartzell G Zhang Z Rubin GM Miller W A computer program for aligning a cDNA sequence with a genomic DNA sequence Genome Res 1998 8 967 974 9750195
Shapiro MB Senapathy P RNA splice junctions of different classes of eukaryotes: sequence statistics and functional implications in gene expression Nucleic Acids Res 1987 15 7155 7174 3658675
Thanaraj TA Stamm S Clark F Riethoven JJ Le Texier V Muilu J ASD: the alternative splicing database Nucleic Acids Res 2004 32 D64 D69 14681360 10.1093/nar/gkh030
Lee C Atanelov L Modrek B Xing Y ASAP: the alternative splicing annotation project Nucleic Acids Res 2003 31 101 105 12519958 10.1093/nar/gkg029
Setubal J Meidanis J Introduction to computational molecular biology 1997 PWS Publishing Company
Gupta S Zink D Korn B Vingron M Haas S Genome wide identification and classification of alternative splicing based on EST data Bioinformatics 2004 20 2579 2585 15117759 10.1093/bioinformatics/bth288
Eyras E Caccamo M Curwen V Clamp M ESTGenes: alternative splicing from ESTs in Ensembl Genome Res 2004 14 976 987 15123595 10.1101/gr.1862204
ENCODE PC The ENCODE (ENCyclopedia Of DNA Elements) Project Science 2004 306 636 640 15499007 10.1126/science.1105136
Burset M Seledtsov IA Solovyev VV SpliceDB: database of canonical and non-canonical mammalian splice sites Nucleic Acids Res 2001 29 255 259 11125105 10.1093/nar/29.1.255
ASD – Alternative Splicing Database
ASAP – Alternative Splicing Annotation Project
Acembly
Levine A Durbin R A computational scan for U12-dependent introns in the human genome sequence Nucleic Acids Res 2001 29 4006 4013 11574683
SoftberryGene
|
16207377
|
PMC1276783
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Oct 5; 6:244
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-244
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2501622344210.1186/1471-2105-6-250Methodology ArticleMASQOT: a method for cDNA microarray spot quality control Bylesjö Max [email protected] Daniel [email protected]ödin Andreas [email protected]öström Michael [email protected] Stefan [email protected] Henrik [email protected] Johan [email protected] Research group for Chemometrics, Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden2 Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Sweden3 Umeå Plant Science Centre, Department of Plant Physiology, Umeå University, SE-901 87 Umeå, Sweden2005 13 10 2005 6 250 250 30 6 2005 13 10 2005 Copyright © 2005 Bylesjö et al; licensee BioMed Central Ltd.2005Bylesjö et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
cDNA microarray technology has emerged as a major player in the parallel detection of biomolecules, but still suffers from fundamental technical problems. Identifying and removing unreliable data is crucial to prevent the risk of receiving illusive analysis results. Visual assessment of spot quality is still a common procedure, despite the time-consuming work of manually inspecting spots in the range of hundreds of thousands or more.
Results
A novel methodology for cDNA microarray spot quality control is outlined. Multivariate discriminant analysis was used to assess spot quality based on existing and novel descriptors. The presented methodology displays high reproducibility and was found superior in identifying unreliable data compared to other evaluated methodologies.
Conclusion
The proposed methodology for cDNA microarray spot quality control generates non-discrete values of spot quality which can be utilized as weights in subsequent analysis procedures as well as to discard spots of undesired quality using the suggested threshold values. The MASQOT approach provides a consistent assessment of spot quality and can be considered an alternative to the labor-intensive manual quality assessment process.
==== Body
Background
At present, the DNA microarray technology allows simultaneous monitoring of the expression levels of thousands of genes. The technique produces large and complex datasets that are relatively easy to generate but non-trivial to analyze and extract information from. So far, much of the data mining efforts have been focused on the statistical analysis (see, for instance [1-4]) and less on acquiring high quality data from the image analysis. Image analysis is the process of extracting information from the scanned microarray images, which is an important step due to the sequential nature of the microarray analysis [5]. Consequently, problems in the initial steps have large impact on the interpretation of the final results from the experiment.
A number of technical issues during microarray preparation potentially affect the spot quality.
• Low signal intensity is perhaps the most generally acknowledged property that affects spot quality due to the subsequent problems in distinguishing signal from noise for spots with weak signals. Weak signal intensities should result from physiologically low expression levels but might also be related to surface properties of the slide, signal bleaching, scanner problems or incomplete or irregular hybridization.
• Intensity distribution issues appear as regions of pixels containing signals that clearly deviate from the average signal, typically as distinct sub-areas within the foreground region. As the signal of any given spot on a microarray slide is expected to be uniform over the entire spot foreground area, intensity distribution issues are usually a consequence of non-specific binding or irregular distribution of the printed DNA on the slide.
• Morphological issues refer to unexpected shape-related variations of the spot foreground region. This includes very small or very large spot sizes, low spot circularities or spot mixing. Size aberrations might be a consequence of precipitates or impurities in the printing solution or needle clogging during printing. Spots are expected to be roughly circular in shape, but manufacturing issues might result in deviation from the circularity norm. Furthermore, imperfections on the slide or washing problems might cause the dye from several spots to mix, referred to as bleeding, making the separation of these signals difficult or even impossible.
• Background issues appear as intensity fluctuations in the local background region immediately surrounding the foreground region. An increase in local background intensity or variance compared to the global slide background typically result from dye contaminants due to non-specific binding or incomplete washing.
Microarray spot quality control is essentially the identification and removal of spots with properties that cause the subsequent interpretation of the signal from these spots to be unreliable or misleading. Analogously, it is the recognition of characteristics that enable dependable interpretations and conclusions from the generated data. The field of microarray spot quality control has been largely neglected in the past but has recently become an area of interest. Existing documented semi-automatic methodologies include Bayesian networks [6] as well as linear combinations of quality parameters allegedly related to the quality of the spot [7-10]. Interestingly, manual evaluation is still a common quality assessment procedure, despite the labor-intensive nature of visually inspecting spots in the order of hundreds of thousands. Increasing availability and usage of microarrays in the field of transcriptomics has caused experiments to involve an escalating number of slides, which in turn has highlighted the primary bottleneck of manual quality assessment.
Partial least squares (PLS) is a generalized regression method which aims to maximize the covariance between the X (descriptor) and Y (response) matrices. PLS can handle large data sets of multi-collinear and noisy data with moderate amounts of missing data in both X and Y. PLS-DA can be seen as a special case of PLS where the response matrix Y is categorical (numerically represented as 0 or 1) and determines class belonging of observations. PLS-DA has been widely applied in microarray analysis (see, for instance [11,12]) as well as other areas of life science (see, for instance [13,14]). For a more detailed description of the properties of PLS, please consult [15-17] and references therein.
Here, we propose the microarray spot quality control (MASQOT) methodology for assessment of cDNA microarray spot quality, outlined in figure 1. A set of existing and novel spot descriptors were identified that aimed to characterize spot quality in terms of physical attributes of the spot. Prior to the extraction of descriptors, manual assessment of the spot quality was performed independently by three experienced microarray users on roughly eighty thousand spots in order to provide a sufficiently large data set of known quality.
Figure 1 Flowchart of the classification procedure. The classification process involves an 8-bit image, optimized for segmentation, as well as a 32-bit image, used for information extraction. During the training phase, visual classification results are required while this is not necessary for external data.
Spot descriptors were subsequently subjected to multivariate discriminant analysis by means of PLS-DA with the aim to categorize spots of low quality and spots of high quality by treating these spots as separate classes. The utilized descriptors aimed to describe foreground and background irregularity measures, spot morphology and foreground density attributes that were potentially useful for discriminating between the reliable (not bad) and unreliable (bad) spots. For instance, the circularity measure is a descriptor ranging from 0% to 100%. If the circularity descriptor approaches its minimum value, the predicted class belonging should typically be higher (closer to 1) for the bad class compared to the not bad class. The coefficient of variation for the foreground and background regions is an example of an employed descriptor that illustrates reverse characteristics; higher values should typically provide greater class conformity with the bad class compared to the not bad class. However, all employed descriptors together contribute to the regression model and consequently to the final class determination at varying degrees depending on the properties of the spot.
The MASQOT approach aims to provide a consistent assessment of spot quality, applicable to various types of microarray data, thus avoiding the labor-intensive manual quality assessment process. The methodology generates continuous values of spot quality which can be utilized to discard spots of undesired quality or used as weights in subsequent analysis procedures.
Results
Five cDNA microarray slides using the Populus second generation microarray slide layout (POP2) where the samples originate from a previous investigation of leaf development (Sjödin et al, in preparation) were used for classification training. Five additional POP2 slides, not included in the training set, were employed for external validation. Segmentation of raw images was performed using an implementation of the Seeded Region Growing (SRG) algorithm [18]. The properties of each spot were subsequently characterized using a large set of descriptors allegedly linked to spot quality. These properties include foreground and background variability measures, spot morphology and foreground intensity distribution measures. Please consult Additional file 1 for a complete list of all the utilized descriptors.
Following segmentation, spots were inspected by three experienced microarray users and independently assigned to the two quality categories {bad, not bad}. Spots in the bad category consisted of all the spots that were classified as bad by at least one of the experienced users while the remaining spots were categorized as not bad. For classification and evaluation purposes, the spots in the bad category were subsequently partitioned into different sub-classes based on visual properties as described in table 1. This can be seen as characterizing each spot as exhibiting
Table 1 The different sub-classes of bad spots.
Class Description
not bad No issue. Contains all spots with no apparent problems according to the classification by the three experienced users.
HIFI High-Intensity Foreground Issue. Typically intensity distribution issues, such a dye debris in the foreground region or donut-shaped spots, with very distinct characteristics.
LIFI Low-Intensity Foreground Issue. Weak intensity distribution issues in the foreground region or morphological issues.
HIBI High-Intensity Background Issue. Typically intensity distribution issues, such a dye debris in the background region, with very distinct characteristics.
LIBI Low-Intensity Background Issue. Weak intensity distribution issues or faint increases in noise level in the background region.
HIFI/HIBI A combination of HIFI and HIBI.
HIFI/LIBI A combination of HIFI and LIBI.
LIFI/HIBI A combination of LIFI and HIBI.
LIFI/LIBI A combination of LIFI and LIBI.
HIFI/LIFI A combination of HIFI and LIFI.
HIFI/LIFI/HIBI A combination of HIFI and LIFI and HIBI.
1. No issues (not bad); or
2. Foreground issues (FI); or
3. Background issues (BI); or
4. Any combination of 2 and 3
To avoid confounding of properties, only spots displaying pure issues (entries 1–3) were used in the classification training, although all spots were used in the model evaluation. The three-class problem was subjected to multivariate analysis by means of PLS [16] regression coupled with discriminant analysis (PLS-DA) with the aim to discriminate not bad spots from FI spots and BI spots. The result from the PLS-DA regression model is a predicted class conformity (CC) value for each of the classes: not bad (CCnb), foreground issues (CCFI) and background issues (CCBI) with the added restriction that CCnb = 1 - (CCFI + CCBI). Due to this restriction, only the conformity value of the not bad spots (CCnb) will be interpreted in the upcoming sections. A CCnb value approaching 1 denotes high compliance with the not bad class, which can be interpreted as a quality measure of the spot. Spots visually categorized as bad should thus exhibit a value of CCnb close to 0 whereas spots categorized as not bad should exhibit a value of CCnb close to 1.
Receiver Operating Characteristics (ROC) plot of the classifications of the POP2 training set and POP2 test set, respectively, is available in figure 2. A density plot of the CCnb value for the bad and not bad spots in the POP2 training set is shown in figure 3, showing a partial overlap between the discrimination of the two classes. Due to this overlap, discrimination accuracy was dependent on a threshold value t denoting the separation point between the two classes.
Figure 2 Receiver Operating Characteristics (ROC) plot. The relation between true positives (bad spots classified as bad) and false positives (not bad spots classified as bad) for the training and test data. The solid line denotes training data whereas the dashed line denotes test data.
Figure 3 Density plot of the predicted class conformity of the not bad class. A class conformity value of 1 signifies perfect class conformity while a value of 0 signifies no class conformity. The dashed line illustrates the density for the prediction of the bad spots in the POP2 training set whereas the solid line illustrates the density of the prediction of the not bad spots in the POP2 training set.
Spots with a predicted CCnb value below t were classified as bad whereas the remaining spots were classified as not bad. The threshold value can be set more or less stringently depending on the quality filtering requirements. This is illustrated in figures 4a–b, depicting different views of interpreting the classification accuracy of the POP2 training set and the POP2 test set. The threshold values were set either to maximize the overall classification accuracy or to maximize the class-wise classification accuracy. The overall classification accuracies for the POP2 training set (38 627 spots) and POP2 test set (39 421 spots) were calculated using equation 1 for all t in the interval (0,1) using CCnb values for the {bad, not bad} spots. The classification accuracy peak at a level of approximately 98% where t = 0.4 (see figure 4a). Predicted class-wise accuracies were calculated using equation 1 employing CCnb values for the bad spots and CCnb values for the not bad spots separately. The predicted class-wise accuracies intersect at a level of 95% (see figure 4b) for t = 0.5. Exact classification accuracies per sub-class, based on the intersection threshold value t = 0.5 as illustrated in figure 4b, are available in table 2.
Figure 4 Relationship between classification accuracy and threshold value for the POP2 data. The threshold value t defines the boundary between bad and not bad spots for the POP2 training set (38 627 spots) and the POP2 test set (39 421 spots). Spots with a predicted class conformity value for the not bad class (CCnb) below the threshold value t are classified as bad while the remaining spots are classified as not bad. a) Overall classification accuracy vs. threshold value calculated as the fraction of correctly classified spots in the data set for a given threshold value. The solid line represents the POP2 training set whereas the dashed line represents the POP2 test set. The dotted vertical line at threshold value t = 0.4 illustrates an approximate maximum. b) Classification accuracy of the bad and not bad spots vs. threshold value. For the POP2 training set, the solid line represents the classification accuracy of the not bad spots and the dashed line represents the classification accuracy of the bad spots. For the POP2 test set, the dot-dashed line represents the classification accuracy of the not bad spots and the long-dashed line represents the classification accuracy of the bad spots. The dotted vertical line at threshold value t = 0.5 denotes the intersection point.
Table 2 Classification accuracy of the POP2 training data. The classification accuracy for each sub-class as calculated using threshold value t = 0.5.
Class Number of spots Classification accuracy (%)
not bad 35983 94.7
HIFI 942 98.7
LIFI 76 86.8
HIBI 987 96.5
LIBI 284 85.9
HIFI/HIBI 81 97.5
HIFI/LIBI 69 98.6
LIFI/HIBI 66 98.5
LIFI/LIBI 44 77.3
HIFI/LIFI 29 89.7
HIFI/LIFI/HIBI 62 100.0
The presented MASQOT approach was compared to three existing quality control methods: the composite quality score qcom proposed by Wang et al [8], the mean-median correlation factor mmcorr evaluated by Tran et al [9] and the coefficient of variation (CV) parameter CVspot evaluated by Sauer et al [10]. Threshold values for all quality control parameters were set to achieve maximum overall classification accuracy. The result, shown in table 3, demonstrates that the MASQOT approach provides a greater level of class discrimination for the POP2 test set compared to the remaining evaluated quality control methods.
Table 3 Comparison to other quality control methods. The presented quality control parameter CCnb was compared to the composite quality score qcom, the mean-median correlation factor mmcorr and the CVspot value. Threshold values for all quality control parameters were set to maximize overall classification accuracy. The classification accuracy was determined from classification of the POP2 test set.
Quality control parameter Threshold Classification accuracy (%)
CCnb 0.40 98.1%
qcom 0.32 94.5%
mmcorr 0.65 94.3%
CVspot 1.05 95.0%
Discussion
cDNA microarray spot quality control is, in many aspects, a complex problem. Naturally, the automatic assessment of quality of each spot is highly reliant on the characterization of the spot. Incorrect approximations of the spatial location will affect the properties of the segmented foreground region, which in turn will influence the values of the quality control descriptors. In such a sequential process, where each step is dependent on the preceding steps, errors will propagate down-stream at a high rate. However, the most striking intricacy is perhaps the visual assessment, which is the foundation of this computer-based classification, where even experienced microarray users tend to disagree. As shown from the results presented here, it is the spots with unanimous visual quality assessment that are the most complicated to reproduce accurately. This disagreement stems from the more fundamental issue of defining 'quality' and in understanding the basal aspects that affect this quality.
The approach described here aims to assess the technical precision of each spot, which is believed to be linked to the biological accuracy (see [19] for a discussion regarding precision and accuracy in the microarray field). It should thus be stated, in this context, that lack of precision in a microarray spot measurement does not necessarily infer lack of accuracy. However, it is arguably reasonable to handle spots of questionable precision with specific care during the analysis procedure to aid the concluding biological interpretations.
The spot quality control assessment is commonly treated as a discrete problem (essentially, separating 'bad' spots from 'good' spots) but the spot quality varies on a continuous scale, ranging from very bad to very good. Instead of discarding spots, one might weight the spots according to the quality assessment. The concept of relative spot weights has previously been acknowledged in microarray normalization techniques (see, for instance [2,20]) but might also prove to be valuable in additional analysis steps. However, evaluation of the usage and validity of spot weights based on the quality assessments provided here remains the scope of a future paper.
The rate of accuracy in prediction of the true positives (bad spots predicted as bad) and the true negatives (not bad spots predicted as not bad) are illustrated separately since these accuracies are not consistently of equal importance. For instance, depending on the user and the question at hand, it might be more important to avoid the risk of removing spots of decent quality than to eliminate all of the bad spots from the data set. Simply using the overall classification accuracy could be rather illusive, merely since the number of bad spots is much lower than the number of not bad spots in a typical data set.
The methodology presented here is based on the scaled sum of the intensities from both channels but can, with minor adjustments, also be based on single channel intensities. By using the scaled channel intensity levels, one avoids the risk of drowning information, in particular when there is a great difference in intensity level between the channels. In addition, the presented approach greatly resembles the visual illustrations of the spots which, by design, will provide an advantage in finding correlation between the visual quality assessment and the spot descriptors. Furthermore, it is more feasible for the average user to achieve a per-spot quality measure than a per-channel quality measure since this avoids raising questions with regard to what to do when only one channel is of low or moderate quality.
The recent advances in spot quality control have clearly shown that a good explanation of training data is possible using several different methodologies of varying complexity. However, very few efforts have been made to evaluate further aspects of the quality (for instance, more refined descriptors) and, most importantly, the reproducibility of the classification on external data. External reproducibility has been the major aim here, partly overshadowing the aim of internal reproducibility on the training data, which is shown by the clear agreement in accuracy between the independent POP2 training set and the POP2 test set.
Conclusion
The presented MASQOT technique provides a robust methodology for semi-automated cDNA microarray spot quality control with high accuracy of training data as well as external data compared to other evaluated methods. The MASQOT methodology generates non-discrete values of spot quality which can be utilized as weights in subsequent analysis procedures as well as to discard spots of undesired quality using the proposed threshold values.
Methods
Microarray preparation
Samples for the microarray slides used in this paper originate from an experiment of Populus tremula leaves, investigating regulation of leaf development (Sjödin et al, in preparation). The utilized microarray layout, referred to as POP2, consist of 25 278 single spotted cDNA clones from a recent assembly of more than 100 000 expressed sequence tags (ESTs) from the Populus genus [21]. All sequence information is available in the online sequence resource PopulusDB [22] and a full array layout is available for download from the online microarray resource UPSC-BASE [23].
Ten out of a total of 28 POP2 slides were randomly chosen for classification and were subsequently grouped into two equally large sets of five slides each; the POP2 training set and POP2 external test set. See additional file 2 for a complete list of the POP2 microarray slides used here. All POP2 slides were printed using a QArray arrayer (Genetix, Hampshire, U.K.). The preparation, labeling and hybridization of cDNA clones and mRNA samples were carried out according to the protocol described by Smith et al [24]. The arrays were scanned on a ScanArray 4000 (Perkin-Elmer Wellesley, MA) at 5 μm resolution to obtain raw image files for the red-fluorescent dye Cy5 and the green-fluorescent dye Cy3. All POP2 raw image files are available online for download at the UPSC-BASE microarray database [23] from experiment number 0013.
Image analysis
The workflow from scanned cDNA images to computer-based classification was separated into seven sub-procedures, outlined below and illustrated in figure 1.
1. Image merging generated a combined image from the intensity measurements of both the red-fluorescent dye Cy5 and the green-fluorescent dye Cy3, which was used in the subsequent gridding and segmentation steps.
2. Gridding attempted to identify the precise spatial center of the spots on the scanned microarray images.
3. Segmentation classified the pixels as either representing the cDNA expression level (foreground pixels) or an estimation of the local noise level (background pixels). In addition, a thin strip of pixels in the boundary region between the two segments (border pixels) was identified.
4. Information extraction refers to the characterization of the foreground and background regions from the segmentation process. In general terms, information extraction should provide a description of each region that is relevant in some sense (for instance, the spatial location of the foreground region or the foreground intensity level.) The focus here was on features that captured the overall quality of the spot.
5. Manual classification provided a measure of the spot quality by means of visual inspection carried out by three experienced microarray users.
6. Computer-based classification of spot quality (the training phase) generated a model for the differences between the spot quality classes using discriminant analysis based on the PLS regression method (PLS-DA).
7. Verification of the computer-based classification (the test phase) validated the predictive ability of the model using processed data not included in the training phase.
Image merging
Both segmentation and gridding were based on a combined eight-bit image constructed from the intensity measurements of the red-fluorescent dye Cy5 and the green-fluorescent dye Cy3. The merged eight-bit image lacks some details of the original images but is computationally efficient, in particular concerning memory requirements. Details of the utilized damping and scaling procedures are described by Yang et al [5] and briefly outlined below.
• The intensity levels in both images were square-root transformed. The square-root transform utilizes damping, which ensures that the relative impact of high-intensity pixels is decreased during gridding and segmentation.
• Median intensity values were computed from the transformed images.
• A joint intensity value was calculated using the sum of the square-root transformed intensities from both channels scaled according to the median values, respectively.
• Intensity values greater than 255 were truncated.
Gridding
Approximate spatial centers of each spot, referred to as the grid points, were manually located using an in-house developed Java application. This procedure is the only step in the classification process that requires user intervention. A more precise midpoint of the foreground region was found using a square pixel mask with the expected spot diameter (100 μm for the POP2 data) surrounding the initial grid point. The pixel mask was spatially reallocated in all directions, deviating at most 30 μm from the initial grid point, and the center position of the square pixel mask containing the highest total sum of intensities was selected as seed point.
Segmentation
The employed segmentation method was an implementation of the seeded region growing (SRG) algorithm, initially proposed by Adams and Bischof [18]. The SRG method has earlier been utilized in microarray spot segmentation by Yang et al [5]. Implementation details are available in additional file 3. The result from the segmentation process was a pixel mask categorizing each pixel into one of the four groups {foreground, border, background, un-assigned}. Each spot thus consisted of a distinct foreground region with the following characteristics:
• All pixels within the foreground region were spatially connected.
• No pixels overlapped with the foreground region of another spot.
• Minor fluctuations in intensity level within the region were accepted.
• The maximum Euclidean distance between any two pixels in the foreground region was restricted.
• Spot circularity was not assumed.
Information extraction
The data utilized in the information extraction originates from the sum of the raw intensities of both the Cy5 and the Cy3 channels scaled according to the respective median intensity value. The scaling was applied to decrease the impact of the channel demonstrating the highest median intensity value. A large set of different features were extracted which were believed to be linked to spot quality and these were subsequently used in the upcoming computer-based classification. For purposes of repeatability and applicability to various types of microarray slides, all descriptors were corrected by an approximation of the slide background mean based on the mean intensity level of the local background regions from all spots on the slide.
Furthermore, spots where the saturation contents in at least one of the channels exceeded 10% of the total number of pixels, as suggested by Wang et al [8], were not included in the classification.
A complete table of all extracted descriptors including a description or definition is available in additional file 1. The descriptors aimed to capture foreground and background variability properties, spot morphology and foreground intensity and density properties.
Manual classification
The spots from ten POP2 slides were independently inspected by three experienced microarray users and assigned to the two quality categories {bad, not bad}. The spots in the bad category consisted of all the spots that were classified as bad by at least one of the experienced users while the remaining spots were categorized as not bad.
During the visual classification, the experienced users worked according to four basic rules of thumb related to the technical precision of each spot.
• The signal within the foreground region should have low variability.
• The foreground region should be circular.
• The foreground region should be spatially located at the expected position.
• The background region should have low variability and low intensity level compared to the global slide background.
The relation between these, that is, how much each factor was allowed to deviate, alone and in combination with other factors, was the task for the multivariate classification model. The utilized data sets, subsequent to segmentation, are available at additional file 4 (training set) and additional file 5 (test set). A summary of the manual classifications as performed by the experienced users is available in additional file 6.
The POP2 slides were randomly partitioned into two equally large sets of five slides each; the POP2 training set and the POP2 test set. For classification and evaluation purposes, the bad spots of the training set were subsequently divided into different sub-classes based on visual properties as described in table 1. The HIFI and LIFI sub-classes were used as representatives of the pure foreground issues (FI) during classification training. Analogously, the HIBI and LIBI sub-classes were used as representatives of the pure background issues (BI) during classification training. Typical examples of the described sub-classes can be found in additional file 7.
Computer-based classification
The computer-based classification was performed using PLS-DA as implemented in SIMCA-P+ 10.0 (Umetrics AB, Umeå, Sweden). Cross-validation [25] with seven groups was used to determine the number of latent variables. Prior to analysis, all descriptors were column-wise mean-centered and scaled to unit variance (UV) by dividing each descriptor with the standard deviation of the descriptor. The UV scaling procedure in combination with mean-centering translates the distribution of each descriptor to unit variance. Results and model statistics from the PLS-DA training phase are described in additional file 8.
Classification training was based on discriminant analysis of subsets of the foreground issues (FI) class, the background issues (BI) class and the not bad class. See table 1 for a more detailed description of the available sub-classes. Prior to discriminant analysis, a representative subset of each class consisting of 355 spots each was selected using D-optimal design [26] in order to eliminate the large differences in data set size between the three classes. See additional file 9 for the designed data set and additional file 10 for details regarding the D-optimal design.
Classification accuracies were calculated using equation 1, where n is the number of observations; t is a threshold value and x the predicted class conformity values for a given set of spots. Equation 1 utilizes the corrpred(i, y, t) function that returns 1 if i ∈ {bad} and y < t or if i ∈ {not bad} and y ≥ t; or 0 otherwise.
predacc(x,t)=100n∑i=1ncorrpred(i,xi,t) (1)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGWbaCcqWGYbGCcqWGLbqzcqWGKbazdaWgaaWcbaGaemyyaeMaem4yamMaem4yamgabeaakiabcIcaOiabdIha4jabcYcaSiabdsha0jabcMcaPiabg2da9maalaaabaGaeGymaeJaeGimaaJaeGimaadabaGaemOBa4gaamaaqahabaGaem4yamMaem4Ba8MaemOCaiNaemOCai3aaSbaaSqaaiabdchaWjabdkhaYjabdwgaLjabdsgaKbqabaGccqGGOaakcqWGPbqAcqGGSaalcqWG4baEdaWgaaWcbaGaemyAaKgabeaakiabcYcaSiabdsha0jabcMcaPaWcbaGaemyAaKMaeyypa0JaeGymaedabaGaemOBa4ganiabggHiLdGccaWLjaGaaCzcamaabmaabaGaeGymaedacaGLOaGaayzkaaaaaa@6051@
List of abbreviations used
PLS Partial Least Squares
PLS-DA Partial Least Squares Discriminant Analysis
SRG Seeded Region Growing
ROC Receiver Operating Characteristics
EST Expressed Sequence Tag
Authors' contributions
MB implemented the segmentation and gridding tools, performed visual classification, implemented the D-optimal design, conceived and generated the classification model and drafted the manuscript. DE generated the microarray slides, performed visual classification and helped to draft the manuscript. AS conceived the study, collected leaf samples, performed visual classification and helped to draft the manuscript. MS, SJ, HA and JT supervised the project. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Definition of the employed spot descriptors. Provides a definition of the employed spot descriptors used to assess the quality of each spot.
Click here for file
Additional File 2
A list of the employed POP2 slides. Provides a list of the employed POP2 slides.
Click here for file
Additional File 3
Implementation details of the segmentation process. Provides in-depth information regarding the implementation of the seeded region growing (SRG) algorithm.
Click here for file
Additional File 4
The processed POP2 training data. Provides the processed POP2 training data set, which contains per-spot values of all descriptors employed here.
Click here for file
Additional File 5
The processed POP2 test data. Provides the processed POP2 test data set, which contains per-spot values of all descriptors employed here.
Click here for file
Additional File 6
Manual quality assessments. Provides a summary of the quality assessments as classified by the three experienced microarray users.
Click here for file
Additional File 7
Visual representations of the sub-classes of bad spots. Provides images of typical examples of the 4 main sub-classes of bad spots.
Click here for file
Additional File 8
Details of the PLS-DA model. Provides details and statistics from the utilized PLS-DA model.
Click here for file
Additional File 9
The designed subset of the POP2 training data. Provides the processed and filtered POP2 training data set, containing only the spots from the not bad, FI and BI classes which were selected according to the D-optimal design.
Click here for file
Additional File 10
Description of the utilized D-optimal design. Provides information regarding generation of the D-optimal design used in to select subsets of the three classes.
Click here for file
Acknowledgements
This work was supported by grants from
• The Swedish Foundation for Strategic Research (MB, HA)
• The Knut and Alice Wallenberg Foundation (JT)
• The European Commission through the Directorate General Research within the Fifth Framework for Research – Quality of Life and Management of the Living Resources Programme, contract No.QLK5-CT-2002-00953 coordinated by the University of Southampton (AS, SJ).
• The Swedish Research Council (AS, SJ, DE, MS)
• EU-strategic funding (DE)
• The Functional Genomics Initiative at Swedish University of Agricultural Sciences (DE)
==== Refs
Kerr MK Martin M Churchill GA Analysis of variance for gene expression microarray data J Comput Biol 2000 7 819 837 11382364 10.1089/10665270050514954
Smyth GK Speed T Normalization of cDNA microarray data Methods 2003 31 265 273 14597310 10.1016/S1046-2023(03)00155-5
Wolfinger RD Gibson G Wolfinger ED Bennett L Hamadeh H Bushel P Afshari C Paules RS Assessing gene significance from cDNA microarray expression data via mixed models J Comput Biol 2001 8 625 637 11747616 10.1089/106652701753307520
Efron B Tibshirani R Storey JD Tusher V Empirical Bayes analysis of a microarray experiment J Am Stat Assoc 2001 96 1151 1160 10.1198/016214501753382129
Yang YH Buckley MJ Dudoit S Speed TP Comparison of methods for image analysis on cDNA microarray data J Comput Graph Stat 2002 11 108 136 10.1198/106186002317375640
Hautaniemi S Edgren H Vesanen P Wolf M Jarvinen AK Yli-Harja O Astola J Kallioniemi O Monni O A novel strategy for microarray quality control using Bayesian networks Bioinformatics 2003 19 2031 2038 14594707 10.1093/bioinformatics/btg275
Chen YD Kamat V Dougherty ER Bittner ML Meltzer PS Trent JM Ratio statistics of gene expression levels and applications to microarray data analysis Bioinformatics 2002 18 1207 1215 12217912 10.1093/bioinformatics/18.9.1207
Wang X Ghosh S Guo SW Quantitative quality control in microarray image processing and data acquisition Nucleic Acids Res 2001 29 e75 11470890 10.1093/nar/29.15.e75
Tran PH Peiffer DA Shin Y Meek LM Brody JP Cho KW Microarray optimizations: increasing spot accuracy and automated identification of true microarray signals Nucleic Acids Res 2002 30 e54 12060692 10.1093/nar/gnf053
Sauer U Preininger C Hany-Schmatzberger R Quick and simple: quality control of microarray data Bioinformatics 2005 21 1572 1578 15615693 10.1093/bioinformatics/bti238
Man MZ Dyson G Johnson K Liao B Evaluating methods for classifying expression data J Biopharm Stat 2004 14 1065 1084 15587980 10.1081/BIP-200035491
Perez-Enciso M Tenenhaus M Prediction of clinical outcome with microarray data: a partial least squares discriminant analysis (PLS-DA) approach Hum Genet 2003 112 581 592 12607117
Eriksson L Antti H Gottfries J Holmes E Johansson E Lindgren F Long I Lundstedt T Trygg J Wold S Using chemometrics for navigating in the large data sets of genomics, proteomics, and metabonomics (gpm) Anal Bioanal Chem 2004 380 419 429 15448969 10.1007/s00216-004-2783-y
Jonsson P Bruce SJ Moritz T Trygg J Sjöström M Plumb R Granger J Maibaum E Nicholson JK Holmes E Antti H Extraction, interpretation and validation of information for comparing samples in metabolic LC/MS data sets Analyst 2005 130 701 707 15852140 10.1039/b501890k
Martens H Naes T Multivariate Calibration 1992 Chichester , John Wiley & Sons
Burnham AJ MacGregor JF Viveros R Latent variable multivariate regression modeling Chemom Intell Lab Syst 1999 48 167 180 10.1016/S0169-7439(99)00018-0
Eriksson L Johansson E Kettaneh-Wold N Wold S Multi- and Megavariate Data Analysis: Principles and Applications 2001 Umeå , Umetrics Academy
Adams R Bischof L Seeded Region Growing IEEE T Pattern Anal 1994 16 641 647 10.1109/34.295913
van Bakel H Holstege FC In control: systematic assessment of microarray performance EMBO Rep 2004 5 964 969 15459748 10.1038/sj.embor.7400253
Smyth GK Linear models and empirical Bayes methods for assessing differential expression in microarray experiments Stat Appl Genet Mol Biol 2004 3 Article 3 16646809
Sterky F Bhalerao RR Unneberg P Segerman B Nilsson P Brunner AM Charbonnel-Campaa L Lindvall JJ Tandre K Strauss SH Sundberg B Gustafsson P Uhlen M Bhalerao RP Nilsson O Sandberg G Karlsson J Lundeberg J Jansson S A Populus EST resource for plant functional genomics Proc Natl Acad Sci USA 2004 101 13951 13956 15353603 10.1073/pnas.0401641101
Sterky F Bhalerao RR Unneberg P Segerman B Nilsson P Brunner AM Charbonnel-Campaa L Lindvall JJ Tandre K Strauss SH Sundberg B Gustafsson P Uhlen M Bhalerao RP Nilsson O Sandberg G Karlsson J Lundeberg J Jansson S PopulusDB: A Populus EST resource for plant functional genomics
Sjödin A Bylesjö M UPSC-BASE: Populus transcriptomics online
Smith CM Rodriguez-Buey M Karlsson J Campbell MM The response of the poplar transcriptome to wounding and subsequent infection by a viral pathogen New Phytol 2004 164 123 136 10.1111/j.1469-8137.2004.01151.x
Wold S Cross Validatory Estimation of the Number of Components in Factor and Principal Components Models. Technometrics 1978 20 397 406
deAguiar PF Bourguignon B Khots MS Massart DL PhanThanLuu R D-optimal designs Chemom Intell Lab Syst 1995 30 199 210 10.1016/0169-7439(94)00076-X
|
16223442
|
PMC1276784
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Oct 13; 6:250
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-250
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2561622569010.1186/1471-2105-6-256Research ArticleEvaluating eukaryotic secreted protein prediction Klee Eric W [email protected] Lynda BM [email protected] Department of Laboratory Medicine and Pathology, University of Minnesota, Mayo Mail Code 609, 420 SE Delaware Street, Minneapolis, MN 55455, USA2005 14 10 2005 6 256 256 15 3 2005 14 10 2005 Copyright © 2005 Klee and Ellis; licensee BioMed Central Ltd.2005Klee and Ellis; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Improvements in protein sequence annotation and an increase in the number of annotated protein databases has fueled development of an increasing number of software tools to predict secreted proteins. Six software programs capable of high throughput and employing a wide range of prediction methods, SignalP 3.0, SignalP 2.0, TargetP 1.01, PrediSi, Phobius, and ProtComp 6.0, are evaluated.
Results
Prediction accuracies were evaluated using 372 unbiased, eukaryotic, SwissProt protein sequences. TargetP, SignalP 3.0 maximum S-score and SignalP 3.0 D-score were the most accurate single scores (90–91% accurate). The combination of a positive TargetP prediction, SignalP 2.0 maximum Y-score, and SignalP 3.0 maximum S-score increased accuracy by six percent.
Conclusion
Single predictive scores could be highly accurate, but almost all accuracies were slightly less than those reported by program authors. Predictive accuracy could be substantially improved by combining scores from multiple methods into a single composite prediction.
==== Body
Background
Predicting secreted proteins from primary sequence is a major component of automated protein annotation and is critical to a wide range of studies. Embryology, tumor maker detection, and agricultural animal performance are investigated using eukaryotic secreted proteins and their role in cell-to-cell communication, cellular differentiation, morphological development, and cellular response to disease. Many software tools have been developed for ab initio cellular localization prediction, using machine learning techniques such as neural networks, hidden Markov models and support vector machines. Identifying the program best suited for a researcher's needs requires familiarity with several different programs. Prediction accuracy depends on the methods employed by a program and the integrity of the data used to develop the program. Additionally, unbiased comparison using an independent protein sequence set is needed to compare programs, as system characteristics reported by program authors are often inflated [1].
The ambiguity of terminology used to describe and label secreted proteins often results in confusion on just what type of protein is being predicted or discussed. To eliminate this confusion, biologically concrete labels will be used in lieu of the term "secreted protein" or "secretory protein", here. Proteins possessing an N-terminal signal sequence and entering the classical secretory pathway via the endoplasmic reticulum, will be called CoTranslationally Translocated (CTT) proteins. Proteins transported out of the cell (regardless of mechanism) will be called extracellular proteins, proteins exported through the CTT pathway will be called classical extracellular proteins, and proteins exported by other mechanisms will be called non-classical extracellular proteins.
Most prediction programs predict CTT (not extracellular) proteins by identifying an N-terminal signal sequence, a signal sequence cleavage site, or a combination of both features, in a target sequence. New programs try to improve this approach through refinements in the sequence data used for program development and the application of new decision making algorithms to the problem. Novel methods, including predictions based on base composition across the entire protein sequence, identifying localization specific protein domains, homology to annotated protein databases, and mining partial protein annotations for key words, are also being used. Though programs that predict signal sequences often also predict signal sequence cleavage sites, we here focus on the former, since the latter has been recently reviewed [2]. We also focus on prediction of eukaryotic signals; the prokaryotic signal pathway has been recently reviewed [3].
Here, six programs, selected for their applicability for high throughput analysis, are described, and their ability to predict CTT proteins in eukaryotic proteins, are evaluated: SignalP 3.0 [4], SignalP 2.0 [5], TargetP 1.01 [6], PrediSi [7], Phobius [8], and ProtComp 6.0 [9]. SignalP 3.0, Phobius, PrediSi, and ProtComp 6.0 are recently released and have not been extensively reviewed nor independently compared. TargetP 1.01 and SignalP 2.0 are older programs, previously demonstrated to have high accuracy; they provide a basis for comparing our results with other studies [1,10-12].
Prediction programs
SignalP versions 2.0 and 3.0 both use Neural Networks (NN) and Hidden Markov Models (HMM) to predict CTT proteins, through the analysis of protein sequence N-termini. These programs are among the most accurate methods for CTT protein prediction [1,10,11] and the programs' HMMs have an uncommon ability to discriminate N-terminal signal peptides from N-terminal signal anchors. SignalP 2.0 neural networks were trained using N-terminal subsequences containing CTT signal peptides and subsequent 30 residue of the mature peptides of 1137 eukaryotic CTT proteins and 70 residue N-terminal subsequences of 1451 eukaryotic non-CTT proteins, abstracted from SwissProt 35.0. SignalP 2.0 outputs four predictors computed by independent neural networks and two predictors computed by the Hidden Markov Models. NN outputs include the position and probability of the residue most likely to belong to a signal peptide (S-score max), the average probability all residue analyzed belong to a signal peptide (S-score mean), the position and probability of the residue most likely to be the first N-terminal residue of the mature peptide (C-score max), and a geometric average of the C-score and smoothed derivative of the S-score (Y-score). For each predictor a Boolean flag denoting CTT or non-CTT protein is returned, along with a composite neural network prediction which identifies CTT proteins in sequences which possess a high average S-score from the first N-terminal residue to the residue with the maximum Y-score, followed by a predicted cleavage site.
SignalP uses two HMM's, one that models the CTT signal peptide and a second that models a signal anchor. An N-terminal signal peptide is a short polypeptide (average length 20–25 residues), has no strongly conserved sequence motifs, but has three distinct sequential regions, the n (N-terminal)-region, the h (hydrophobic)-region, and the c (C-terminal)-region [13,14]. The signal peptide model contains submodels that describe each of these three regions. The signal anchor model contains two submodels that represent its n-region and h-region. In the signal peptide model, the h-region is limited to between six and twenty residues, the n-region must have at least one residue (and start with a methionine), and the c-region must have at least three residues. The n-region and c-region contain self-cyclic states with exponentially decaying transitions. This type of transition state allows the model to fit signal peptides possessing n-regions and c-regions with variable lengths, while still constraining the system, preventing unusually long region lengths, and thereby encapsulating the known properties of these regions. In the signal anchor module, the architecture of the n-region is the same, but the h-region also possesses a self-cyclic, exponentially decaying, transition state. The HMM outputs the position and score of the residue with the maximum C-score and a mean S-score for the entire sequence analyzed. In addition, Boolean flags for both predictors and a composite predictor characterizing the analyzed sequence as CTT, signal anchor, or other, are also output [5,15].
SignalP 3.0 possesses updated neural network architecture, new selection criteria for training sequences, and a composite score for signal peptide prediction. The neural networks were modified to include input nodes for sequence composition characteristics. Also, a symmetric sliding window of size 27 for signal peptide prediction and an asymmetric window of size 24 for cleavage site prediction were implemented after an exhaustive analysis of 27,000 neural networks determined these non-uniform window sizes provided the best performance. The networks were retrained using protein sequences from SwissProt 40.0, which were filtered to remove sequences likely to be mis-annotated. The new filtering process limited eukaryotic training data to sequences containing an alanine, cysteine, glycine, leucine, proline, glutamine, serine, or threonine at the first position upstream of the annotated cleavage site [4]. Additionally, ProP [16] predictions were used to identify and remove ten sequences likely to contain mis-annotated signal peptide cleavage sites. Finally, the D-score, computed from the mean S-score and the maximum Y-score, was added, thereby incorporating data from cleavage site predictions into the signal peptide predictions and improving their accuracy [4].
SignalP 2.0 and SignalP 3.0 were evaluated and compared by the program authors using five-fold cross-validation. Overall, version 3.0 outperformed version 2.0 in cleavage site predictions and signal peptide presence predictions. Internal testing showed SignalP 3.0 NN and HMM differentiated CTT proteins from non-CTT proteins with 98% and 94% accuracy, respectively, and SignalP 2.0 NN and HMM differentiated these proteins with 97% and 94% accuracy, respectively. Accuracy was assessed for the analysis of the first 70 N-terminal residues of target proteins and accuracy may decrease if more residues are analyzed.
TargetP differs from the SignalP software by predicting CTT (SP) and mitochondrial (mTP) or chloroplastic (cTP) proteins through the analysis of N-terminal sequence data. The program has a two-layer architecture; the first layer uses independently-trained networks to predict SP, mTP, or cTP localization, and the second layer integrates first layer outputs into a final prediction. Non-redundant, equal size, sequence sets from SwissProt release 36 (for plants) and release 37 (for non-plants) were used to train the networks. cTP cleavage site predictions are performed using the methods implemented in ChloroP [17], SP cleavage site predictions are performed using the methods implemented in SignalP [15], and mTP cleavage site predictions are made with a motif identifying matrix. Both the overall prediction and individual numeric scores from each network are output. TargetP also assigns a reliability class (RC) to each prediction based on the difference between the highest scoring prediction and the second highest scoring prediction. TargetP was tested using cross-validation and shown to correctly predict CTT localization with 92% accuracy, 92% specificity and 95% sensitivity, in non-plant sequences. When compared to PSORT [18,19], MitoProt [20,21], ChloroP and SignalP, TargetP CTT predictions in non-plants had a higher specificity than PSORT and a higher sensitivity than SignalP [6].
PrediSi predicts CTT proteins through the analysis of N-terminal sequence data by positional weighted matrices. Matrices were developed for the n-region, h-region, and c-region of the signal peptide using 2,783 eukaryotic, 557 Gram-negative, and 236 Gram-positive CTT proteins and 5,547 eukaryotic, 2,013 Gram-negative, and 1,077 Gram-positive control sequences (cytoplasmic and nuclear), obtained from SwissProt 42.9. The resulting amino acid frequency values were corrected to account for baseline proteome levels. PrediSi outputs a single numeric score, predicted cleavage site and Boolean flag denoting CTT signal peptide presence or absence. Self-consistency testing correctly identified 72.66% of eukaryotic CTT proteins and correctly excluded 98.31% of control proteins. PrediSi was outperformed by SignalP-NN and SignalP-HMM in eukaryotic and Gram-negative predictions, but displayed improved performance in Gram-positive predictions. PrediSi is designed for extremely fast analysis and is well suited for high throughput processing. These valuable characteristics were achieved at the cost of slightly reduced accuracy [7].
Phobius predicts CTT proteins using Hidden Markov Models to analyze full-length protein sequences. The program also predicts transmembrane domains and is designed to differentiate N-terminal transmembrane domains from CTT signal peptides. The HMMs were trained using 146 sequences from the TMHMM dataset [22], 140 sequences from TMPDB [23], 2 sequences from the Moller dataset [24], and 4 TM sequences from SWISS-PROT. These sequences were divided into TM-only and TM-and-SP sequence sets. Additionally, SP-only and not-TM-not-SP sequence sets were created using SWISS-PROT 41.0 proteins. Phobius outputs a Boolean flag denoting the presence or absence of a CTT signal peptide, the number of transmembrane domains predicted and a position labeled protein orientation schematic. In ten-fold cross-validation testing, Phobius correctly predicted 91.1% of TM-and-SP sequences, 63.6% of TM-only sequences, 96.1% of SP-only sequences and 98.2% of not-TM-not-SP sequences. In comparisons to other programs, Phobius out performed TMHMM, HMMTOP, TMHMM – SignalP combination, and HMMTOP-SignalP combination predicting TM-and-SP sequences, while being outperformed by HMMTOP and TMHMM predicting TM-only sequences. None of the software's options which allow users to constrain predictions based using known information about the presence of CTT signal peptides and TM domains or use a homology modeling component to perform BLAST comparisons against NCBI's nr database, were used for the testing described here [8].
ProtComp 6.0, from Softberry, Inc., predicts protein localization, including extracellular proteins, using a combination of neural networks and sequence homology. Sequences are assigned localization through homology to experimentally and theoretically annotated databases, neural network predictions and pentamer distribution comparisons to the homology databases. Softberry reports 86% correct prediction of extracellular proteins as tested with approximately 200 extracellular proteins. In this study, only ProtComp neural network predictions were evaluated [9].
Results
Individual predictions
CTT signal peptide predictive accuracies for individual predictive scores are shown in Table 1. Based on Matthew's Correlation Coefficient (MCC) [25,26], SignalP 3.0 D-score was the most accurate predictor, closely followed by the SignalP 3.0 maximum S-score and the TargetP prediction. The most sensitive predictors were the SignalP 2.0 neural network Mean S-score and Hidden Markov Model S-probability score. The maximum prediction specificity was obtained using the SignalP 2.0 HMM maximum C-score predictor.
Table 1 System performance measures. Performance was measured based on the program's ability to correctly discriminate CTT proteins from non-CTT proteins. MCC = Mathews' Correlation Coefficient [1].
Program TP FP TN FN Sensitivity Specificity MCC
TargetP 55 8 307 2 96% 87% 90%
SignalP3 NN – Cmax 52 48 267 5 91% 52% 62%
SignalP3 NN – Ymax 55 9 306 2 96% 86% 89%
SignalP3 NN – Smax 56 9 306 1 98% 86% 90%
SignalP3 NN – Smean 55 17 298 2 96% 76% 83%
SignalP3 NN – D 55 7 308 2 96% 89% 91%
SignalP3 HMM Cmax 46 9 306 11 81% 84% 79%
SignalP3 HMM Sprob 56 17 298 1 98% 77% 84%
SignalP2 NN – Ymax 56 14 301 1 98% 80% 86%
SignalP2 NN – Cmax 54 28 287 3 95% 66% 75%
SignalP2 NN – Smean 57 13 302 0 100% 81% 88%
SignalP2 NN – Smax 56 21 294 1 98% 73% 81%
SignalP2 HMM Sprob 57 21 294 0 100% 73% 83%
SignalP2 HMM Cmax 36 3 312 21 63% 92% 73%
Phobius 55 13 302 2 96% 81% 86%
PrediSi 52 12 303 5 91% 81% 83%
ProtComp NN 46 32 283 11 81% 59% 62%
Combined predictions
The combinatorial analysis examined 14,892 unique predictor combinations; for each combination, all program performance measures were calculated. A maximum MCC value of 97% was obtained in 58 different score combinations. The t-score for the highest MCC value associated with the combinatorial prediction method (0.97) is ts = 76.8, corresponding to a significance level well below 0.05%. The Fisher's Z-transformation testing the significance between the combinatorial correlation (0.97) and the best correlation arising from a single prediction score (0.91) returned a p ≤ 1.72 e-14. These results support the significance of the reported finding.
The minimum number of predictors needed to obtain the 0.97 MCC value was four and occurred in five different combinations. In each of these combinations, the TargetP prediction, the SignalP 2.0 maximum Y-score and the SignalP 3.0 maximum S-score were included. The fourth predictor was the SignalP 2.0 mean S-score, SignalP 2.0 Hidden Markov Model S-probability, the SignalP 3.0 mean S-score, the SignalP 3.0 D-score, or the SignalP 3.0 Hidden Markov Model S-probability. The most accurate pairs of predictors had MCC values of 95%: all included TargetP combined with either the SignalP 2.0 maximum Y-score, the SignalP 2.0 mean S-score, the SignalP 3.0 maximum S-score or the SignalP 3.0 D-score.
A prediction specificity of 98% was reached by 43 score combinations. The minimum number of scores required to reach this level of specificity was four and occurred in five different combinations. These five combinations all included the SignalP 2.0 maximum Y-score and SignalP 3.0 maximum C-score. The highest sensitivity obtained during the combinatorial analysis corresponded to the individual predictive scores with the highest sensitivity (combination set size 1), SignalP 2.0 NN mean S-score and SignalP 2.0 Hidden Markov Model S-probability.
Sequences can be analyzed using TargetP, SignalP 2.0 and SignalP 3.0 on the Vertebrate Secretome and CTT-ome Database [27]. Two sets of criteria can use used: positive TargetP prediction, SignalP 2.0 maximum Y-score and SignalP 3.0 maximum S-score (sensitivity 0.96, specificity 0.96, MCC 0.96), or positive prediction using TargetP or SignalP 3.0 D-Score (sensitivity 0.96, specificity 0.87, MCC 0.90).
Discussion
The single most accurate predictors for discriminating CTT proteins from other proteins were the TargetP prediction, the SignalP 3.0 maximum S-score and the SignalP 3.0 D-score. The high accuracy of the SignalP 3.0 D-score is not a surprise, as it was designed to increase the overall prediction accuracy of CTT signal peptides, is itself a composite score of the neural network mean S-score and maximum Y-score, and incorporates cleavage site prediction information. Likewise, we expect the SignalP 3.0 maximum S-score to perform well, as this is obtained from networks trained on very current sequence data and the score is designed to specifically quantify CTT signal peptide presence. It is surprising the TargetP predictor performed so strongly, as it was expected predictors from older software, trained on older and generally smaller numbers of proteins, would be outperformed by the more recent programs. The TargetP accuracy can be attributed to its high specificity and ability to minimize false positive predictions; this is likely a result of TargetP's capacity to differentiate mitochondrial proteins from CTT proteins.
The CTT protein prediction sensitivity almost unilaterally decreased for common predictive scores shared by SignalP 2.0 and SignalP 3.0; the exceptions being the neural network maximum S-score and the Hidden Markov Model maximum C-score. Changes in specificity and accuracy were more variable, with the values of some predictors increasing and others decreasing for both performance measures. Exactly why the predictive sensitivity dropped between SignalP versions is not known, but it could be a by-product of the new screening protocols used to select positive CTT proteins for the version 3.0 training set. This protocol was particularly sensitive to the inclusion of protein sequences containing rare amino acids in the -1 and -3 residue locations relative to the CTT signal peptide cleavage site. While excluding these should improve the accuracy of cleavage site location prediction, it may have also caused the drop in CTT protein prediction sensitivity.
The PrediSi and Phobius predictions were 5% to 8% less accurate than the best predictors from TargetP and SignalP. While these programs fall short in predicting CTT proteins, they both possess characteristics that address niche analyses. The program developers state that the value of PrediSi is its computational speed. In our analysis this claim was validated; PrediSi was clearly the fastest program evaluated and did not restrict the size of sequence set analyzed. Therefore, if users are working with extremely large datasets, PrediSi can be used for rapid initial screens. However, for more accurate results, PrediSi's analyses should be combined with more rigorous, computationally expensive, methods.
The results obtained for Phobius are not surprising, as this program was not developed to specifically differentiate CTT proteins from non-CTT proteins, but to differentiate CTT signal peptides from N-terminal transmembrane domains. Since proteins with N-terminal transmembrane domains were intentionally not included in the test set, we could not assess this function. Phobius, like PrediSi, is not as accurate as TargetP and SignalP for CTT protein prediction. However, Phobius could add value to an analysis pipeline processing protein sequences containing N-terminal transmembrane domains.
ProtComp 6.0 returned the most disappointing results, despite the possibility of inflated results due to duplication between the test set and neural network training set. This program is not designed to strictly identify CTT proteins, but predicts localization to multiple cellular compartments. As such, quantifying CTT proteins required combining multiple prediction categories output by the program, which may have added to the poor performance. ProtComp is the only program tested which differentiates extracellular proteins from other CTT proteins. However, it is unclear if this program predicts non-classical extracellular proteins. ProtComp is more suitable for general localization screening than for specific locale predictions, as discussed here.
Combinatorial analysis is a systematic method for identifying the complementary CTT protein predictors best suited for incorporation into an analysis pipeline. A maximum combination of six scores was chosen to limit the exponential explosion of combinations evaluated, while still allowing for a single predictive score from each program to be included in the optimal combination. Fifty-eight different combinations provided a CTT signal peptide prediction accuracy of 97%, which exceeded the highest single score accuracy by 6%. A minimum of four predictive scores was necessary to obtain this accuracy and occurred in five different combinations. Interestingly, the TargetP prediction, the SignalP 2.0 maximum Y-score and the SignalP 3.0 maximum S-score, but not the SignalP 3.0 D-score (one of the highest individual scores), were included in all five. The fourth predictor in these five combinations varied. Accuracy was only slightly reduced (decreased by less than 1%) when the fourth predictor was eliminated.
It is possible the combinatorial predictors identified in this study may prove to be non-optimal in large applications, since the optimal 4-predictor combinations may have been over-fit to the data. The combination of the TargetP prediction, the SignalP 2.0 maximum Y-score and the SignalP 3.0 maximum S-score, the three predictors common to all five of the most accurate predictor combinations, may be best used in an analysis pipeline, to avoid the over-fitting while still generating high accuracy predictions. These are the three used for multiple predictor sequence analysis on the Vertebrate Secretome and CTT-ome Database [27].
It has been suggested that program authors overstate the predictive accuracy of their programs [1]. Almost all of the predictive accuracies reported here were lower than those reported in the original publications. TargetP was reported to correctly predict localization with 92% MCC accuracy, slightly higher than the 90% we calculated. For the SignalP 2.0 Hidden Markov Model the highest predictive score had an accuracy of 83% MCC, 11% lower than the published accuracy. The program's neural network accuracies, however, were comparable, with a published accuracy of 87% and an accuracy of 86% found in our testing. The published accuracy of the SignalP 3.0 neural network predictions was 7% higher and that of the Hidden Markov Model predictions was 10% higher than the accuracies obtained in our testing. Phobius predictions were reported to be 96.1%, 10% higher then the accuracy found in this study. PrediSi is one of the only programs to report a lower accuracy that what was found in this testing; reporting 73% accuracy compared to the 83% in this evaluation. ProtComp 6.0 website reports correctly predicting extracellular sequences 86% of the time, 14% higher than the accuracy found in our testing. It is important to independently assess predictive performance.
Conclusion
This study of eukaryotic CTT protein prediction software evaluates six programs. Each offers a different analysis method, which in many cases is designed for a particular type of analysis. Understanding the differences between prediction programs is critical. The independent assessment of the predictive accuracy described here can provide a good basis for selecting software. TargetP, SignalP maximum S-score and the SignalP 3.0 D-score were shown to be the most accurate individual scores for CTT prediction. Prediction accuracy is significantly improved through use of multiple analysis methods and combining multiple predictive scores into a single composite prediction. Older predictive programs retain value; both SignalP 2.0 and TargetP contained predictive scores that were among the top predictive scores in both single and composite prediction analysis.
Methods
Protein test set
Full-length test proteins were abstracted from the SwissProt database. To eliminate bias caused by redundancy between the training and testing sequence sets only sequences modified or entered during 2004 (version 43.0+) were used for testing. Homo sapiens, Mus musculus, Sus scrofa, Rattus norvegicus, and Bos taurus proteins were identified using the SwissProt Sequence Retrieval System. For each organism the FASTA sequence and full SwissProt record were downloaded for proteins annotated as "Date=20040101:20041220 AND Comment Type=Subcellular Location". Sequences were categorized by the SwissProt record Subcellular Localization Comment entry and sequences annotated as "Putative", "Possible", "by Similarity", and, in one case, "by Similatity" were eliminated. Secreted, Mitochondrial, Nuclear, and Cytoplasmic proteins were selected from the remaining sequences and manually reviewed for ambiguous annotation. The final test set consisted of 372 full-length proteins; breakdown by cellular localization and organism is given in Table 2, the sequences themselves are in Additional file 1.
Table 2 Summary of the protein test set. 372 protein sequences from five vertebrate organisms and four localizations taken from the SwissProt database. The sequences themselves are in Additional file 1.
Organism Secreted Mitochondrial Nuclear Cytoplasmic Total
Human 22 19 106 57 204
Mouse 19 10 37 35 101
Rat 7 2 14 21 44
Pig 4 0 3 4 11
Cow 5 3 1 3 12
Total 57 34 161 120 372
CTT protein prediction
All CTT protein predictions were carried out using each program's web-servers. TargetP, SignalP 2.0 and SignalP 3.0 analyzed 125-residue, N-terminal subsequences of the 372 test proteins. TargetP was configured with non-plant settings, prediction of cleavage site enabled and default thresholds for assigning classifications. SignalP 3.0 was configured with settings for "Eukaryotic" sequence data, "Both" analysis methods, "No graphics" output, "Short" output and sequence truncation set to 70 residues. SignalP 2.0 was configured to analyze eukaryotic sequence data, output "no graphics", and otherwise use default parameters. PrediSi, Phobius and ProtComp analyzed the full-length test proteins. PrediSi was run using the eukaryotic organism group and a text based output. Phobius was run in normal prediction mode with short output. ProtComp predictions were executed on the animal and fungi ProtComp 6.0 server and positive CTT predictions assigned when the neural network predictions were extracellular, Golgi, endoplasmic reticulum or lysosome. Output files from all analyses were parsed, and program performance measures calculated, using custom Perl scripts.
Program performance measures
To assess predictive accuracy, program sensitivity, program specificity (also known as positive predictive value), and program accuracy using MCC, were calculated. MCC value significance was validated using a standard t-test and Fisher's z-transformation [28]. For programs reporting multiple predictive scores, system characteristics were calculated for each score independently. Performance characteristics were also calculated for multiple combinations of predictors. An exhaustive analysis of two to six score combinations was performed. A protein was predicted to be secreted in a combinatorial analysis only when all predictive scores included in the combination independently predicted the protein to be secreted. The combinatorial analysis did not attempt to integrate or algorithmically combine the numeric scores values associated with each individual predictive score.
Authors' contributions
EWK designed the evaluation, chose the six servers to be evaluated, designed the evaluation, analyzed the results, and drafted the paper; LBE supervised the overall work, and critically revised the paper. Both authors read and approved the final manuscript.
Supplementary Material
Additional file 1
Protein test set 372 eukaryotic Swiss-Prot protein sequences used to evaluate the six servers, in fasta format.
Click here for file
Acknowledgements
We thank the developers and maintainers of all software used in this study. They have given the scientific community excellent resources. We thank John Attewell for helpful discussions. This work was supported in part by NIH R01-GM63904 and a predoctoral traineeship to EK (NLM TG-0704l).
==== Refs
Menne K Hermjakob H Apweiler R A comparison of signal sequence prediction methods using a test set of signal peptides Bioinformatics 2000 16 741 742 11099261 10.1093/bioinformatics/16.8.741
Zhang Z Henzel WJ Signal peptide prediction based on analysis of experimentally verified cleavage sites Protein Sci 2004 13 2819 24 15340161 10.1110/ps.04682504
Luirink J Sinning I SRP-mediated protein targeting: structure and function revisited Biochim Biophys Acta 2004 1694 17 35 15546655
Bendtsen JD Nielsen H von Heijne G Brunak S Improved prediction of signal peptides: SignalP 3.0 J Mol Biol 2004 340 783 795 15223320 10.1016/j.jmb.2004.05.028
Nielsen H Krogh A Prediction of signal peptides and signal anchors by a hidden Markov model Proc Int Conf Intell Syst Mol Biol 1998 6 122 130 9783217
Emanuelsson O Nielsen H Brunak S von Heijne G Predicting subcellular localization of proteins based on their N-terminal amino acid sequence J Mol Biol 2000 300 1005 1016 10891285 10.1006/jmbi.2000.3903
Hiller K Grote A Scheer M Munch R Jahn D PrediSi: prediction of signal peptides and their cleavage positions Nucleic Acids Res 2004 32 W375 9 15215414
Kall L Krogh A Sonnhammer EL A combined transmembrane topology and signal peptide prediction method J Mol Biol 2004 338 1027 36 15111065 10.1016/j.jmb.2004.03.016
Softberry ProtComp 6.0
Nielsen H Brunak S von Heijne G Machine learning approaches for the prediction of signal peptides and other protein sorting signals Protein Eng 1999 12 3 9 10065704 10.1093/protein/12.1.3
Emanuelsson O von Heijne G Prediction of organellar targeting signals Biochim Biophys Acta 2001 1541 114 9 11750667 10.1016/S0167-4889(01)00145-8
Zhang Z Henzel WJ Signal peptide prediction based on analysis of experimentally verified cleavage sites Protein Sci 2004 13 2819 24 15340161 10.1110/ps.04682504
von Heijne G A new method for predicting signal sequence cleavage sites Nucleic Acids Res 1986 14 4683 90 3714490
von Heijne G Signal sequences. The limits of variation J Mol Biol 1985 184 99 105 4032478 10.1016/0022-2836(85)90046-4
Nielsen H Engelbrecht J Brunak S von Heijne G Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites Protein Eng 1997 10 1 6 9051728 10.1093/protein/10.1.1
Duckert P Brunak S Blom N Prediction of proprotein convertase cleavage sites Protein Eng Des Sel 2004 17 107 112 14985543 10.1093/protein/gzh013
Emanuelsson O Nielsen H von Heijne G ChloroP, a neural network-based method for predicting chloroplast transit peptides and their cleavage sites Protein Sci 1999 8 978 984 10338008
Nakai K Kanehisa M A knowledge base for predicting protein localization sites in eukaryotic cells Genomics 1992 14 897 911 1478671 10.1016/S0888-7543(05)80111-9
Horton P Nakai K Better prediction of protein cellular localization sites with the k nearest neighbors classifier Proc Int Conf Intell Syst Mol Biol 1997 5 147 152 9322029
Claros MG MitoProt: a Macintosh application for studying mitochondrial proteins Comput Appl Biosci 1995 11 441 447 8521054
Claros MG Vincens P Computational method to predict mitochondrially imported proteins and their targeting sequences Eur J Biochem 1996 241 779 786 8944766 10.1111/j.1432-1033.1996.00779.x
Krogh A Larsson B von Heijne G Sonnhammer EL Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes J Mol Biol 2001 305 567 80 11152613 10.1006/jmbi.2000.4315
Ikeda M Arai M Okuno T Shimizu T TMPDB: a database of experimentally-characterized transmembrane topologies Nucl Acids Res 2003 31 406 409 12520035 10.1093/nar/gkg020
Moller S Kriventseva EV Apweiler R A collection of well characterised integral membrane proteins Bioinformatics 2000 16 1159 1160 11159338 10.1093/bioinformatics/16.12.1159
Matthews BW Comparison of predicted and observed secondary structure of T4 phage lysozyme Biochim Biophys Acta 1975 405 442 451 1180967
Baldi P Brunak S Chauvin Y Andersen CAF Nielsen H Assessing the accuracy of prediction algorithms for classification: an overview Bioinformatics 2000 16 412 424 10871264 10.1093/bioinformatics/16.5.412
Vertebrate Secretome and CTT-ome Database
David FN The Moments of the z and F Distributions Biometrika 1949 36 394 403 15402074
|
16225690
|
PMC1276785
|
CC BY
|
2021-01-04 16:27:46
|
no
|
BMC Bioinformatics. 2005 Oct 14; 6:256
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-256
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2571622569610.1186/1471-2105-6-257Research ArticleStatistical distributions of optimal global alignment scores of random protein sequences Pang Hongxia [email protected] Jiaowei [email protected] Su-Shing [email protected] Shiheng [email protected] School of Life Science, Northwest A&F University, Yangling, Shaanxi, China2 Institute of Bioinformatics, Northwest A&F University, Yangling, Shaanxi, China2005 15 10 2005 6 257 257 11 7 2005 15 10 2005 Copyright © 2005 Pang et al; licensee BioMed Central Ltd.2005Pang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The inference of homology from statistically significant sequence similarity is a central issue in sequence alignments. So far the statistical distribution function underlying the optimal global alignments has not been completely determined.
Results
In this study, random and real but unrelated sequences prepared in six different ways were selected as reference datasets to obtain their respective statistical distributions of global alignment scores. All alignments were carried out with the Needleman-Wunsch algorithm and optimal scores were fitted to the Gumbel, normal and gamma distributions respectively. The three-parameter gamma distribution performs the best as the theoretical distribution function of global alignment scores, as it agrees perfectly well with the distribution of alignment scores. The normal distribution also agrees well with the score distribution frequencies when the shape parameter of the gamma distribution is sufficiently large, for this is the scenario when the normal distribution can be viewed as an approximation of the gamma distribution.
Conclusion
We have shown that the optimal global alignment scores of random protein sequences fit the three-parameter gamma distribution function. This would be useful for the inference of homology between sequences whose relationship is unknown, through the evaluation of gamma distribution significance between sequences.
==== Body
Background
Sequence alignment is a central problem in computational molecular biology. Every branch of molecular biology- from database search, phylogeny reconstruction to protein structure prediction- takes sequence alignments as its foundation. The functional and/or structural properties of a new sequence could be predicted from its degree of similarity with some clearly defined and known sequences. If the similarity between two sequences is statistically significant and does not simply stem from sequence repeats of low complexity, then the two sequences are likely to be homologous and thus may have similar functions and/or structures.
To understand whether an observed sequence similarity implies indeed a functional or evolutionary link, or is just a chance event is the central problem for the evaluation of the statistical significance of sequence alignment scores. The basic question to be answered is: what is the probability that a similarity score as great as that actually observed in a comparison between real sequences could have arisen by chance, when sampling from suitably-defined populations of random unrelated sequences? This question is generally addressed by analyzing the distribution of optimal alignment scores from random or real but unrelated sequences [1], which is the approach applied in this research.
Accurate statistical estimate for similarity searches for local alignments has been studied comprehensively [1-4], and the Gumbel distribution is used to estimate the statistical significance for local alignments [5]. However, we still lack a complete theoretical solution to the optimal global alignment between sequences, due to the fact that global alignment scores grow proportionally to the length of the sequences and small changes in the scoring system can produce a different alignment [6].
Abagyan and Batalov suggested that global alignment scores between unrelated protein sequences followed the Gumbel distribution [7]. However, since the scoring system that they used favoured local alignments, these alignments they produced may not be global but local. Dayhoff et al (1978) and Dayhoff et al(1983) evaluated global alignment scores for randomized sequences (maintaining the amino acid composition and sequence length of the real sequences) using their log-odds scoring matrix at PAM250 and a constant gap penalty. The distribution of the resulting scores matched a normal distribution [8,9]. Webber and Barton used sequences belonging to different folds of the SCOP database whose percent identity to each other is less than 40 to analyze the distribution of global alignment z-scores between sequences [10]. They found that the gamma distribution describes the distribution of z-scores better than either the normal or Gumbel distribution.
The determination of homolog is also affected by the reference datasets used for statistical estimation. Lipman et al analyzed the distribution of scores among 100 vertebrate nucleic acid sequences and compared these scores with randomized sequences prepared in different ways [11]. When the randomized sequences were prepared by shuffling the sequence to conserve base composition, the standard deviation was approximately one-third less than the distribution of scores of the natural sequences, thus leading to an overestimate of the significance if such randomized sequences were used for a significant test. When sequences were locally shuffled for randomization, the standard deviation was similar to that of the natural sequences.
In this research, we chose 6 different ways to generate random and real but unrelated protein sequences as reference datasets for significance estimation. Four scoring matrices were applied for global alignments. The PAM120 and PAM250 matrices were selected because they imply different evolutionary time [8], and the BLOSUM50 and BLOSUM62 matrices were selected for their sound performance in database search [12,13]. Most alignments were carried out with the affine gap penalty (see Methods) as it penalizes less for additional gaps, which is more reasonable. The resulting alignment scores were then fitted with three distributions to obtain the statistical characteristic of the global alignment scores.
Results
Derivation of distributions with different sequence sets
We have generated random and real but unrelated sequences in six different ways as reference datasets. The datasets are abbreviated as LAC, LCA, GS, LS, SMP and RUS sequences respectively. The definition of the abbreviations can be found in the List of Abbreviations Used below.
In general the three-parameter gamma distribution performs the best in the goodness of fit test with the distribution of global alignment scores. When the shape parameter of the gamma distribution is sufficiently large, the gamma distribution closely approximates a normal distribution. Thus the normal distribution agrees with the data as well. The Gumbel distribution deviates from the alignment score distribution over the majority of the score frequencies, however its performance gets a little better for the LS sequence set than for the other sequence sets.
The distributions of global alignment scores of the LAC, LCA and GS sequence sets are similar (Figure 1). The three-parameter gamma distribution defines the empirical distributions extremely well. The fitting quality of the normal distribution is better than that of the Gumbel distribution, but both of them diverged from the majority of the global alignment frequencies.
Figure 1 Distribution of scores from global alignments of the GS sequence sets of 300aa long. The alignments were carried out with the BLOSUM62 matrix and an affine gap penalty of -7 and -1. The distribution of the scores was fitted with three distribution curves indicated by the solid line. (A) The scores were fitted with the gamma distribution. The fitted parameters are: γ = 44.7296, λ = 0.381642, μ = 158.666, χ2 value is 14.7643 with d.f. = 18; (B) The scores were fitted with the normal distribution. The fitted parameters are: σ = 17.5243, μ = -41.4629; d.f. = 19, χ2 = 729.532; (C) The scores were fitted with the Gumbel distribution. The fitted parameters are: β = 13.66367, μ = -49.3489; d.f. = 19, χ2 = 10563.8.
For the LS sequence sets the majority of the empirical score frequencies disagree with the normal or Gumbel distributions, whereas the three-parameter gamma distribution describes the alignment data extremely well (Figure 2). It also can be seen that the performance of the Gumbel distribution is better with this sequence set than with the others.
Figure 2 Distribution of scores from global alignments of the LS sequence set whose sequence length is 200aa. The alignments were carried out with the PAM250 matrix and an affine gap penalty of -10 and -2. The sequences were permutated within windows of 10 residues. The distribution of the scores was fitted with three distribution curves indicated by the solid line. (A) The scores were fitted with the gamma distribution. The fitted parameters are: γ = 11.5053, λ = 0.296252, μ = -37.4408, χ2 value is 15.6515 with d.f. = 11; (B) The scores were fitted with the normal distribution. The fitted parameters are: σ = 11.4495, μ = 1.39533; d.f. = 12, χ2 = 815.04; (C) The scores were fitted with the Gumbel distribution. The fitted parameters are: β = 8.927137, μ = -3.75695; d.f. = 12, χ2 = 1001.89.
The distribution of alignment scores of the SMP sequence set is quite different. The empirical score distribution of sequences generated with the PAM120 mutation probability matrix and aligned with the PAM120 log-odds scoring matrix is different from those generated with the PAM250 matrix. Both the gamma and normal distributions fit the score frequencies of the former well (Figure 3), whereas for the latter, the normal distribution disagrees with the majority of the score curve.
Figure 3 Distribution of scores from global alignments of the SMP sequence set of 300aa long. The alignments were carried out with the PAM120 matrix and an affine gap penalty of -16 and -4. The distribution of the scores was fitted with three distribution curves indicated by the solid line. (A) The scores were fitted with the gamma distribution. The fitted parameters are: γ = 3294940, λ = 31.7263, μ = -103798, χ2 value is 10.8053 with d.f. = 12; (B) The scores were fitted with the normal distribution. The fitted parameters are: σ = 57.2142, μ = 56.5328; d.f. = 13, χ2 = 11.0862; (C) The scores were fitted with the Gumbel distribution. The fitted parameters are: β = 44.6098, μ = 30.7864; d.f. = 13, χ2 = 21244.1.
Most of the score distributions of the RUS sequence set agree well with the gamma distribution, no matter what the sequence length is. Although there are occasions when no distribution agrees perfectly well with the score distribution, the three-parameter gamma distribution is still the closest to the empirical distribution (Figure 4).
Figure 4 Distribution of scores from global alignments of the RUS sequence set of 100aa long. The alignments were carried out with the BLOSUM50 matrix and an affine gap penalty of -10 and -2. This sequence set include 300 sequences whose E-value to each other is greater than 10. The distribution of the scores was fitted with three distribution curves indicated by the solid line. (A) The scores were fitted with the gamma distribution. The fitted parameters are: γ = 53.0052, λ = 0.406777, μ = -176.022, χ2 value is 13.7903 with d.f. = 10; (B) The scores were fitted with the normal distribution. The fitted parameters are: σ = 17.898, μ = -45.717; d.f. = 11, χ2 = 308.925; (C) The scores were fitted with the Gumbel distribution. The fitted parameters are: β = 13.95498, μ = -53.771; d.f. = 11, χ2 = 4659.8.
In database search, it is always the sequences, with higher scores (i.e., tail of the distribution), are of interest. So the score frequencies were also plotted against the natural logarithm of scores at the tail of the distribution (Figure 5). It shows clearly that the gamma distribution outperforms both the normal and Gumbel distributions even at the tails of those distributions.
Figure 5 Plots of the tail of the global alignment optimal score distribution. The score frequencies were plotted against the natural logarithm of scores at the tail of the distribution of Figure 1. The three theoretical distributions were indicated in solid lines. The score distribution was fitted with (A) the three-parameter gamma distribution; (B) the normal distribution; and (C) the Gumbel distribution.
The effect of sequence length on the theoretical distribution
We chose the LAC sequence set to analyze the impact of the sequence length on the gamma distribution because the amino acid composition of the LAC sequence set is the same. The result shows that the shape parameter of the fitted gamma distribution increases and scale parameter decreases gradually as the sequence length increases, at the same time the performance of the normal distribution for curve fitting improves slowly (Table 2 and Table 3).
Table 2 Fitting of the global alignment scores aligned with affine gap penalty. All sequences were generated with the LAC approach with different sequence lengths and the alignments were carried out with the BLOSUM62 matrix and an affine gap penalty of -7/-1. Global alignment scores were fitted to the gamma and normal distribution respectively.
sequence fitted gamma distribution fitted normal distribution score
length γ λ μ d.f. χ2 p-value d.f. χ2 mean variance
50aa 41.00 0.63 -84.60 10 11.99 0.29 11 89.22 -19.19 104.36
100aa 49.16 0.55 -115.44 12 6.37 0.9 13 63.32 -25.40 164.90
200aa 52.24 0.44 -153.25 12 8.26 0.77 13 86.91 -33.58 274.14
300aa 58.60 0.41 -183.39 11 8.58 0.64 12 81.78 -39.55 353.04
400aa 56.17 0.36 -200.13 11 4.20 0.95 12 29.30 -45.36 426.44
500aa 66.17 0.36 -235.17 11 9.27 0.58 12 61.39 -50.31 516.48
600aa 68.51 0.34 -253.30 9 7.52 0.58 10 27.55 -54.30 578.04
700aa 73.11 0.34 -273.95 10 12.28 0.27 11 50.20 -58.67 633.89
800aa 74.73 0.32 -295.61 11 12.00 0.36 12 55.39 -62.46 727.41
1200aa 96.17 0.32 -376.88 12 7.52 0.82 13 26.49 -76.42 938.75
Table 3 Fitting of the global alignment scores aligned with constant gap penalty. All sequences were generated with the LAC approach with different sequence lengths and the alignments were carried out with the PAM250 matrix and a constant gap penalty of -10. Global alignment scores were fitted to the gamma and normal distribution respectively.
sequence fitted gamma distribution fitted normal distribution score
length γ λ μ d.f. χ2 p-value d.f. χ2 mean variance
50aa 156.20 0.90 -201.46 10 9.19 0.51 11 14.56 -25.97 197.16
100aa 113.08 0.58 -237.29 10 9.38 0.49 11 29.91 -41.55 338.82
200aa 341.64 0.78 -504.83 11 10.78 0.45 12 20.40 -68.55 557.15
300aa 110.45 0.39 -376.02 10 9.86 0.45 11 18.42 -91.93 730.71
400aa 110.74 0.35 -431.71 11 9.05 0.59 12 19.41 -114.46 908.81
500aa 158.16 0.38 -551.37 12 14.69 0.26 13 15.67 -136.15 1090.13
600aa 118.10 0.31 -537.57 11 11.67 0.39 12 30.80 -157.00 1226.38
700aa 118.40 0.29 -582.49 10 9.85 0.45 11 12.88 -178.59 1377.88
800aa 129.03 0.30 -635.51 11 7.47 0.74 12 33.63 -198.74 1478.53
1200aa 131.35 0.25 -798.05 9 11.07 0.26 10 12.83 -278.65 2053.78
The difference of window size for local shuffling
The impact of the window size of local shuffling approach on the gamma distribution was also studied (Table 4). The result shows that when the window size increases, the shape parameter of the fitted gamma distribution decreases and the fitting performance of the normal distribution gets worse.
Table 4 Fitting of the global alignment scores of the LS sequence set generated with different window sizes. The sequence length is 200aa and all alignments were carried out with the BLOSUM62 matrix with an affine gap penalty of -7/-1. Global alignment scores were fitted to the gamma and normal distribution respectively.
window fitted gamma distribution fitted normal distribution score
size γ λ μ d.f. χ2 p-value d.f. χ2 mean variance
5aa 13.90 0.39 -40.14 9 9.82 0.36 10 239.33 -4.73 90.14
10aa 11.19 0.37 -29.01 10 12.13 0.28 11 236.91 0.99 80.40
15aa 10.60 0.35 -20.57 8 9.31 0.32 9 277.47 9.60 85.90
20aa 10.28 0.36 -16.85 13 17.04 0.19 14 334.87 11.93 80.56
The impact of scoring scheme
The effects of the four scoring matrices are minor on the distribution of global alignment scores. The only exception is the empirical distribution of the SMP sequence set aligned with the PAM120 log-odds scoring matrix, in which the normal distribution performs as well as the gamma distribution.
Discussion
Dayhoff et al (1978) and Dayhoff et al (1983) evaluated global alignment scores for randomized sequences generated as our LCA sequence set using the PAM250 log-odds scoring matrix and a constant gap penalty [8,9]. The distribution of the resulting scores matched a normal distribution. We tried both constant and affine gap penalty for the LCA sequence set and found that the distribution of optimal scores of the LCA sequence set agrees better with the gamma distribution than with the normal distribution.
Webber and Barton took sequences of different folds with less than 40 percent identity from the SCOP database for global alignments and fitted the z-scores to peak distributions [10]. They found that the gamma distribution describes the alignment scores between different folds better than either the normal or Gumbel distribution. The problem is that the derivation of z-scores requires additional alignments to be calculated, and the z-score carries with it an implicit and possible incorrect assignment of significance by the normal distribution.
The score distribution of the SMP sequence set simulated from the evolution of the ancestor sequence at PAM120 is an exception. The fitted shape parameter of the gamma distribution is very large, and the normal distribution fits with the data equally well. This is because sequences generated with the PAM120 mutation probability matrix are around 40 percent similar with each other, so they can hardly be viewed as random sequences for distributional statistical analysis.
This study specifies three-parameter gamma distribution as the theoretical distribution of global alignment scores of random protein sequences. It could be used for the inference of homology between sequences whose relationship is unknown through significance evaluation. One issue worth exploring further is to formulate a function taking sequence length, scoring scheme and amino acid composition into account so that rapid statistics conclusion is reached.
Conclusion
The global alignment optimal scores have been regarded as normal [7] or Gumbel distributed [8]. We have shown here that the normal distribution holds only for sequences simulated with the PAM120 matrix, while the Gumbel distribution disagrees with the optimal alignment score frequencies for all sequence sets in this research. The study shows that the three-parameter gamma distribution describes the random sequence alignment scores better than the normal or Gumbel distribution. The normal distribution performs as well as three-parameter gamma distribution when the shape parameter of the gamma distribution is sufficiently large.
Methods
Dataset
The SCOP (Structural Classification of Proteins) database [14] provides a detailed and comprehensive description of the structural and evolutionary relationships between all proteins whose structure is known [15]. The classification is on hierarchical levels that embody the evolutionary and structural relationships. The hierarchy of the database from top to bottom is fold, superfamily and family. Proteins that share clear evolutionarily relationship are clustered in families, those that have low sequence identities but whose structural and functional features suggest that a common evolutionary origin is probable are placed together in superfamilies. Proteins are defined as having a common fold if they have the same major secondary structures in the same arrangement and with the same topological connections. The SCOP 1.65 (released on December 2003) was used in this study. It contains 40,452 domains organized in 2,327 families, 1,294 superfamilies and 800 folds. These domains correspond to 20,619 entries in the Protein Data Bank (PDB) [16].
The amino acid compositions of all the sequences in the SCOP 1.65 were calculated as shown in Table 1. The amino acid composition in Table 1 was taken as the average amino acid composition of proteins for random sequence generation.
Table 1 The amino acid composition of proteins in SCOP 1.65
amino acid frequency amino acid frequency amino acid frequency amino acid frequency
Aln 0.0819 Gln 0.0372 Leu 0.0871 Ser 0.0607
Arg 0.0489 Glu 0.0634 Lys 0.0593 Thr 0.0582
Gly 0.0775 Met 0.0216 Asn 0.0444 Asp 0.0577
His 0.0233 Trp 0.0150 Tyr 0.0358 Cys 0.0151
Phe 0.0397 Pro 0.0466 Val 0.0709 Ile 0.0557
10 pairs of sequences with different sequence lengths- 50aa, 100aa, 200aa, 300aa, 400aa, 500aa, 600aa, 700aa, 800aa and 1200aa-were randomly chosen from the SCOP database to be managed with different approaches described below. The SCOP identifies of the 10 pairs of sequences are: d1aqt_1 a.2.10.1, d1gjja1 a.140.1.1, d1foka3 a.4.5.12, d1mk7d1 a.11.2.1, d1hx9a1 a.102.4.1, d1h6pb_ a.146.1.1, d1qgjb_ a.93.1.1, d1lj8a1 a.100.1.9, d1fppb_ a.102.4.3, d1jdpb_ c.93.1.1, d1eswa_ c.1.8.1, d1jv1a_ c.68.1.5, d1dhx_1 b.121.2.2, d1jqkf_ e.26.1.2, d2fcpa_ f.4.3.3, d1br2a2 c.37.1.9, d1qgra_ a.118.1.1, d1jqna_ c.1.12.3, d1i50b_ e.29.1.2, d1muka_ e.8.1.4.
Sequence randomization approaches
1) Maintaining the sequence length and average amino acid composition (LAC)
Sequences were generated as strings of independent characters where the amino acid in any position is proportional to its composition in proteins (Table 1) with a given sequence length.
2) Maintaining the sequence length and the amino acid composition of the authentic sequences (LCA)
The amino acid compositions of the two authentic sequences were calculated and random sequences were generated retaining both the distributions of the amino acid composition and the sequence length of the original sequences.
3) Global shuffling or permutation (GS)
Each residue in the authentic sequences is randomly repositioned anywhere in the sequence, so that both the amino acid composition and sequence length were maintained.
4) Local shuffling or permutation (LS)
The sequence is broken into n/w windows (n denotes the length of the sequence and w is the length of the window, typically 5–20 residues) and the residues in each window are randomly shuffled. Both the sequence length and the local amino acid composition are retained with this approach. We selected four window sizes-5aa, 10aa, 15aa and 20aa- to compare their effects on the statistical distributions.
5) Simulation of the mutational process (SMP)
To get sequences of a known evolutionary distance, we simulated the mutational process of the ancestor sequence. First, the PAM1 mutation probability matrix is multiplied by itself 120 or 250 times to get the PAM120 and PAM250 mutation probability matrices. The matrices provide information about each amino acid stayed unchanged or substituted by any other one at the given evolutionary distance. Then the fate of each residue in the new sequence was determined according to the PAM120 or PAM250 mutation probability matrix through computer simulation [8].
6) Real but unrelated sequences (RUS)
The SCOP database also provides filtered sub datasets selected with different criteria, such as sequence identity percentage, E-value, or different hierarchy representatives [17]. We chose three subsets, one in which the sequences are less than 10 percent identity to each other, another with sequences whose E-value with one another is greater than 10, and the third with 800 sequences each represents one fold in the SCOP 1.65.
The sequence lengths in the three subsets vary from 23aa to 1504aa. As global alignment between sequences of highly different sequence lengths is not appropriate, we extracted sequences of similar length in each filtered sequence set further. 300 sequences of 50aa, 100aa, 200aa, 300aa and 400aa were extracted respectively from each of the filtered sequence set, and the tails of longer ones were cut off.
Alignment algorithm
We wrote a C program for all the global alignments in this study. The program implements the Needleman-Wunsch dynamic programming algorithm [18] and penalized on end gaps.
For the LAC and SMP sequence sets, two sequences were generated at a time and global alignments were carried out between them, this process was repeated 10,000 times. For the LCA, GS and LS sequence sets the first sequence was aligned with 5,000 randomizations of the second, then vice versa. Global alignments were carried out between every pair of the 300 sequences for the RUS sequence set.
Scoring scheme
As the evolutionary distance of the SMP sequence set is known, the scoring matrix matching the giving evolutionary time was chosen for the alignments. For other sequence sets two matrices, the PAM120 and PAM250 from the PAM series and another two, the BLOSUM50 and BLOSUM62 from the BLOSUM series were employed. The respective gap open/extension penalty combination for the PAM120 is -16/-4, PAM250 -11/-1, BLOSUM50 -10/-2, BLOSUM62 -7/-1, as recommended by Vingron and Waterman, Mount, and Pearson [6,19,20]. Gap extension penalty is added for the first gap.
We also used constant gap penalty of -10 for the PAM250 matrix for the LAC and LCA sequence sets.
Curve fitting
Optimal global alignment scores of the different sequence sets aligned with different scoring scheme were fitted with the Gumbel, normal and three-parameter gamma distributions respectively. Methods of moments were used for the estimation of the parameters of the optional distributions.
The probability density function of the gamma distribution is given as
f(x)=λγ(x−μ)γ−1e−λ(x−μ)Γ(γ)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGMbGzcqGGOaakcqWG4baEcqGGPaqkcqGH9aqpdaWcaaqaaiabeU7aSnaaCaaaleqabaGaeq4SdCgaaOGaeiikaGIaemiEaGNaeyOeI0IaeqiVd0MaeiykaKYaaWbaaSqabeaacqaHZoWzcqGHsislcqaIXaqmaaGccqWGLbqzdaahaaWcbeqaaiabgkHiTiabeU7aSjabcIcaOiabdIha4jabgkHiTiabeY7aTjabcMcaPaaaaOqaaiabfo5ahjabcIcaOiabeo7aNjabcMcaPaaaaaa@4E19@
where γ (γ > 0) is the shape parameter, λ (λ > 0) is the scale parameter, and μ (x - μ ≥ 0) is the location parameter.
The normal distribution is given as
f(x)=12πσe−(x−μ)22σ2
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGMbGzcqGGOaakcqWG4baEcqGGPaqkcqGH9aqpdaWcaaqaaiabigdaXaqaamaakaaabaGaeGOmaiJaeqiWdahaleqaaOGaeq4WdmhaaiabdwgaLnaaCaaaleqabaGaeyOeI0YaaSaaaeaacqGGOaakcqWG4baEcqGHsislcqaH8oqBcqGGPaqkdaahaaadbeqaaiabikdaYaaaaSqaaiabikdaYiabeo8aZnaaCaaameqabaGaeGOmaidaaaaaaaaaaa@4516@
where μ is the location parameter and σ (σ > 0) is the scale parameter.
The Gumbel distribution, given as
f(x)=1βexp{−exp[−(x−μ)β]−x−μβ}
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGMbGzcqGGOaakcqWG4baEcqGGPaqkcqGH9aqpdaWcaaqaaiabigdaXaqaaiabek7aIbaacyGGLbqzcqGG4baEcqGGWbaCcqGG7bWEcqGHsislcyGGLbqzcqGG4baEcqGGWbaCcqGGBbWwdaWcaaqaaiabgkHiTiabcIcaOiabdIha4jabgkHiTiabeY7aTjabcMcaPaqaaiabek7aIbaacqGGDbqxcqGHsisldaWcaaqaaiabdIha4jabgkHiTiabeY7aTbqaaiabek7aIbaacqGG9bqFaaa@52D0@
where μ is the location parameter and β (β > 0) is the scale parameter.
The χ2 goodness of fit test was used to evaluate the fitting result. The degree of freedom for the fitting with gamma distribution is the number of intervals subtracts 4, while for normal and Gumbel distribution is the number of intervals subtracts 3.
List of abbreviations used
LAC: Maintaining the sequence length and average amino acid composition
LCA: Maintaining the sequence length and the amino acid composition of the authentic sequences
GS: Global shuffling or permutation
LS: Local shuffling or permutation
SMP: Simulation of the mutational process
RUS: Real but unrelated sequences
Authors' contributions
HP developed the code, tested program performance, analyzed the results, and drafted the manuscript. JT proposed many additional suggestions for improving algorithm performance. SC supported the research and improved the writing. ST conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.
Acknowledgements
We would like to thank Fei Lu and Jingjing Li for stimulating discussions and advice on program improvement. We are also grateful to the anonymous reviewers for helpful and constructive comments.
==== Refs
Altschul SF Bundschuh R Olsen R Hwa T The estimation of statistical parameters for local alignment score distributions Nucleic Acid Res 2001 26 351 361 11139604 10.1093/nar/29.2.351
Mott R Accurate formula for p-value of gapped local sequence and profile alignment J Mol Bio 2000 300 649 659 10884359 10.1006/jmbi.2000.3875
Brenner SE Chothia C Hubbard T Assessing sequence comparison methods with relative structurally identified distant evolutionary relationships Proc Natl Acad Sci 1998 95 6073 6078 9600919 10.1073/pnas.95.11.6073
Waterman MS Vingron M Rapid and accurate estimates of statistical significance for sequence database searches Proc Nat Acad Sci 1994 91 4625 4628 8197109
Karlin S Altschul SF Methods for assessing the statistical significance of molecular sequence features by using general scoring schemes Proc Natl Acad Sci 1990 87 2264 2268 2315319
Mount DW Bioinformatics: sequence and genome analysis 2001 Cold Spring Harbor: Cold spring harbor laboratory press
Abagyan RA Batalov S Do aligned sequences share the same fold? J Mol Biol 1997 273 355 368 9367768 10.1006/jmbi.1997.1287
Dayhoff MO Schwartz RM Orcutt BC Dayhoff MO A model of evolutionary change in proteins. Matrices for detecting distant relationships Atlas of protein sequence and structure 1978 5 Washington DC: National biomedical research foundation 345 358
Dayhoff MO Barker WC Hunt LT Establishing homologies in protein sequences Methods Enzymol 1983 91 524 545 6855599
Webber C Barton GJ Estimation of P-values for global alignments of protein sequences Bioinformatics 2001 17 1158 1167 11751224 10.1093/bioinformatics/17.12.1158
Lipman DJ Wilbur WJ Smith TF Waterman MS On the statistical significance of nucleic acid similarities Nucleic Acid Res 1984 12 215 226 6694902
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks Proc Natl Acad Sci 1992 89 10915 10919 1438297
Henikoff S Henikoff JG Performance evaluation of amino acid substitution matrices Proteins: Struct Funct Genet 1993 17 49 61 8234244 10.1002/prot.340170108
Structural Classification of Proteins
Murzin AG Brenner SE Hubbard T Chothia C SCOP: a structural classification of proteins database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
The RCSB Protein Data Bank
ASTRAL SCOP Sequence
Needleman SB Wunsch CD A general method applicable to the search for similarities in the amino acid sequences of two proteins J Mol Biol 1970 48 444 453 10.1016/0022-2836(70)90057-4
Vingron M Waterman MS Sequence alignment and penalty choice: Review of concepts, case studies and implications J Mol Biol 1994 235 1 12 8289235
Pearson WR Empirical statistical estimates for sequence similarity searches J Mol Biol 1998 276 71 84 9514730 10.1006/jmbi.1997.1525
|
16225696
|
PMC1276786
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Oct 15; 6:257
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-257
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2601623617010.1186/1471-2105-6-260SoftwareSpectralNET – an application for spectral graph analysis and visualization Forman Joshua J [email protected] Paul A [email protected] Stuart L [email protected] Stephen J [email protected] The Broad Institute of MIT & Harvard University, 320 Bent Street, Cambridge, MA 02141, USA2 Howard Hughes Medical Institute, Department of Chemistry & Chemical Biology, Harvard University, Cambridge, MA 02138, USA2005 19 10 2005 6 260 260 1 6 2005 19 10 2005 Copyright © 2005 Forman et al; licensee BioMed Central Ltd.2005Forman et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Graph theory provides a computational framework for modeling a variety of datasets including those emerging from genomics, proteomics, and chemical genetics. Networks of genes, proteins, small molecules, or other objects of study can be represented as graphs of nodes (vertices) and interactions (edges) that can carry different weights. SpectralNET is a flexible application for analyzing and visualizing these biological and chemical networks.
Results
Available both as a standalone .NET executable and as an ASP.NET web application, SpectralNET was designed specifically with the analysis of graph-theoretic metrics in mind, a computational task not easily accessible using currently available applications. Users can choose either to upload a network for analysis using a variety of input formats, or to have SpectralNET generate an idealized random network for comparison to a real-world dataset. Whichever graph-generation method is used, SpectralNET displays detailed information about each connected component of the graph, including graphs of degree distribution, clustering coefficient by degree, and average distance by degree. In addition, extensive information about the selected vertex is shown, including degree, clustering coefficient, various distance metrics, and the corresponding components of the adjacency, Laplacian, and normalized Laplacian eigenvectors. SpectralNET also displays several graph visualizations, including a linear dimensionality reduction for uploaded datasets (Principal Components Analysis) and a non-linear dimensionality reduction that provides an elegant view of global graph structure (Laplacian eigenvectors).
Conclusion
SpectralNET provides an easily accessible means of analyzing graph-theoretic metrics for data modeling and dimensionality reduction. SpectralNET is publicly available as both a .NET application and an ASP.NET web application from . Source code is available upon request.
==== Body
Background
The field of graph theory concerns itself with the formal study of graphs – structures containing vertices and edges linking these vertices. Scientifically, graphs can be used to represent networks embodying many different relationships among data, including those emerging from genomics, proteomics, and chemical genetics. Networks of genes, proteins, small molecules, or other objects of study can be represented as nodes (vertices) and interactions (edges) that can carry different weights.
Graph-theoretic metrics, including eigenspectra, have been used to analyze diverse sets of data in the fields of computational chemistry and bioinformatics. Protein-protein interaction networks in Saccharomyces cerevisiae, for example, have been shown to exhibit scale-free properties [1], and databases of mRNAs can be mined using spectral properties of graphs created by their secondary structure [2]. Graph theory has also been used in conjunction with combinations of small-molecule probes to derive signatures of biological states using chemical-genomic profiling [3].
Despite the widespread use of graph theory in these fields, however, there are few user-friendly tools for analyzing network properties. SpectralNET is a graphical application that calculates a wide variety of graph-theoretic metrics, including eigenvalues and eigenvectors of the adjacency matrix (a simple matrix representation of the nodes and edges of a graph) [4], Laplacian matrix [5], and normalized Laplacian matrix, for networks that are either randomly generated or uploaded by the user. SpectralNET is available both as an ASP.NET web application and as a standalone .NET executable. While SpectralNET was originally written to analyze chemical genetic assay data, it should be of use to any researcher interested in graph-theoretic metrics and eigenspectra.
Implementation
SpectralNET was originally written as an ASP.NET application in C#, and has subsequently been ported to a standalone .NET executable version (also written in C#). ASP.NET was originally chosen because it offered a fast, easy way to offer a thin client to users, obviating the need for large amounts of computational power on the client machine, as is often needed to perform large matrix calculations. A standalone version was created for three primary reasons: it avoids the problem of time-outs inherent when using a web interface (a potential issue when performing long-duration calculations), it is more easily distributable, and porting from ASP.NET to a .NET executable is a relatively simple matter.
Many computations are performed directly in C#, such as graph instantiation and metric calculation. Matrix computations (including eigendecomposition) are performed using the NMath Suite (CenterSpace Software, Corvallis, Oregon). Because the NMath Suite is a commercially licensed library, those receiving source code from the authors must supply their own means of performing matrix eigendecomposition in order to modify and redeploy the application. The implementation of the Fermi-Dirac integral, used in the calculation of spectral density, is ported from Michele Goano's implementation in FORTRAN (Goano, 1995). Because SpectralNET uses a third-party library for matrix calculations that is partially implemented using Managed Extensions for C++, SpectralNET will not be portable to Linux until the Mono implementation of this C++ language feature is complete.
Results and discussion
Graph creation
Idealized random networks can be automatically generated by the application, or networks can be uploaded by the user for analysis. SpectralNET can automatically generate random Erdos-Renyi graphs [7], Barabasi-Albert (scale-free) graphs [8], re-wiring Barabasi-Albert graphs [9], Watts-Strogatz (small-world) graphs [10], or hierarchical graphs [11]. Each automatically generated graph type is customizable with algorithmic parameters. SpectralNET was designed with extensibility in mind, so that users may request additional random graph types provided they submit a succinct algorithm to the author or create their own.
Networks can be uploaded by the user in the form of a Pajek file [12] or a tab-delimited text file with one edge per line (see additional file 1: HumanPPI_nodenodeweight.txt for an example network definition file defining a network of human protein-protein interactions). Raw data files can also be uploaded to the application, where each line of data is represented as a labeled vertex. Vertices can be connected with edge weights equal to the square of the correlation of their associated input data, or according to their Euclidean distance as defined by the Eigenmap algorithm [13]. If raw data is uploaded by the user, principal component analysis (PCA) [14,15] can optionally be performed on the data before calculating edge weights.
Graph analysis
After processing the input network, SpectralNET displays for the user a wide variety of graph-analytic metrics. For example, the degree and clustering coefficient is displayed for each vertex. The degree of a vertex is the number of edges incident upon that vertex; for weighted graphs, SpectralNET calculates this as the sum of these edges' weights. The clustering coefficient of a node represents the proportion of its neighbors that are connected to each other, and is calculated for a node i as:
Ci(ki)=2niki(ki−1), (1)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGdbWqdaWgaaWcbaGaemyAaKgabeaakmaabmaabaGaem4AaS2aaSbaaSqaaiabdMgaPbqabaaakiaawIcacaGLPaaacqGH9aqpdaWcaaqaaiabikdaYiabd6gaUnaaBaaaleaacqWGPbqAaeqaaaGcbaGaem4AaS2aaSbaaSqaaiabdMgaPbqabaGcdaqadaqaaiabdUgaRnaaBaaaleaacqWGPbqAaeqaaOGaeyOeI0IaeGymaedacaGLOaGaayzkaaaaaiabcYcaSiaaxMaacaWLjaWaaeWaaeaacqaIXaqmaiaawIcacaGLPaaaaaa@46A8@
where ni denotes the number of edges connecting neighbors of node i to each other, and ki denotes the number of neighbors of node i [16]. In addition, the minimum, average, and maximum distances of each vertex are displayed, which are defined as the shortest, average, and maximum distances, respectively, from the node to any other node in the graph. The components of the adjacency, Laplacian, and normalized Laplacian eigenvectors corresponding to the vertex are also shown, where the adjacency matrix is defined as the matrix A with the following elements:
Aij(G)={1if i≠j and ∃ edge (i,j)0otherwise; (2)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akYBe9v8qqaqFD0xXdHaVhbbf9v8qqaqFr0xc9pk0xbba9q8WqFfeaY=biLkVcLq=JHqVepeea0=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@8470@
the Laplacian matrix is defined as the matrix L with the following elements:
Lij(G)={diifi=j−w(e)ifi≠jand∃edgee(i,j)0otherwise, (3)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=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@7915@
where w(e) denotes the weight of edge e; and the normalized Laplacian matrix is defined as the matrix L
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBamXvP5wqSXMqHnxAJn0BKvguHDwzZbqegm0B1jxALjhiov2DaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaWaaeGaeaaakeaaimaacaWFmbaaaa@3967@ with the following elements:
Lij(G)={1if i=j−1didjif i and j are adjacent,0otherwise (4)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBamXvP5wqSXMqHnxAJn0BKvguHDwzZbqegm0B1jxALjhiov2DaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=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@8D64@
where di denotes the degree of node i [5]. It should be noted that Chung defines the Laplacian matrix as the normalized form above, but we use the more commonly found definition (for an example, see Mohar [17]).
Many large networks derived from biological data are composed of multiple subgraphs that are not always connected together. SpectralNET computes many properties based on the selected or "active" connected component. For the active connected component, its size and average diameter are displayed in addition to graphs of degree distribution [18], clustering coefficient by degree, and average distance by degree [19]. Graphs of eigenvalues, eigenvectors, inverse participation ratios, and spectral densities of the three matrix types are also displayed. The inverse participation ratio is defined for each eigenvector as:
Ij=∑k=1N[(ej)k]4 (5)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGjbqsdaWgaaWcbaGaemOAaOgabeaakiabg2da9maaqahabaWaamWaaeaadaqadaqaaiabdwgaLnaaBaaaleaacqWGQbGAaeqaaaGccaGLOaGaayzkaaWaaSbaaSqaaiabdUgaRbqabaaakiaawUfacaGLDbaadaahaaWcbeqaaiabisda0aaaaeaacqWGRbWAcqGH9aqpcqaIXaqmaeaacqWGobGta0GaeyyeIuoakiaaxMaacaWLjaWaaeWaaeaacqaI1aqnaiaawIcacaGLPaaaaaa@43FB@
where ej represents the eigenvector. Spectral density, or the density of the eigenvalues, is plotted for each eigenvalue as λ/Np(1−p)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqaH7oaBcqGGVaWldaGcaaqaaiabd6eaojabdchaWnaabmaabaGaeGymaeJaeyOeI0IaemiCaahacaGLOaGaayzkaaaaleqaaaaa@36C0@ on the horizontal axis and pNp(1−p)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGWbaCdaGcaaqaaiabd6eaojabdchaWjabcIcaOiabigdaXiabgkHiTiabdchaWjabcMcaPaWcbeaaaaa@35B8@ on the vertical axis, with the function p defined on any eigenvalue as:
p(λ)=1N∑j=1Nδ(λ−λj) (6)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGWbaCdaqadaqaaiabeU7aSbGaayjkaiaawMcaaiabg2da9maalaaabaGaeGymaedabaGaemOta4eaamaaqahabaGaeqiTdq2aaeWaaeaacqaH7oaBcqGHsislcqaH7oaBdaWgaaWcbaGaemOAaOgabeaaaOGaayjkaiaawMcaaaWcbaGaemOAaOMaeyypa0JaeGymaedabaGaemOta4eaniabggHiLdGccaWLjaGaaCzcamaabmaabaGaeGOnaydacaGLOaGaayzkaaaaaa@4820@
where λ is the eigenvalue and δ represents the delta function, implemented as described above [20]. Most graphs can be mouse-clicked to select the vertex corresponding to a desired data point, and eigenvalue graphs can be sorted by value or by vertex degree. All calculated graph metrics can be exported as a tab-delimited text file for further analysis.
Visualization and dimensionality reduction
The main graph display window of SpectralNET offers two interactive graphical networks displays that support zooming and allow vertex selection by mouse-click. The default display view is the resulting graph processed by the Fruchterman-Reingold algorithm [21], which positions vertices by force-directed placement. The other available display is the network's Laplacian embedding, which locates vertices in two-dimensional Euclidean space using the corresponding second and third Laplacian eigenvector components (the first eigenvector component of the Laplacian matrix is degenerate). Exportation of the other Laplacian eigenvector components allows for visualization in higher dimensions.
In conjunction with uploaded raw data, Laplacian embedding allows the user to see a reduced-dimensionality view of high-dimensionality input, once this input is converted into a network. If the user chooses to process input data using the Eigenmap algorithm, Laplacian embedding shows the reduced-dimensionality result [13]. Dimensionality reduction has proven to be a useful tool in computational chemistry and bioinformatics; for example, Agrafiotis [22] used multidimensional scaling (MDS) to reduce the dimensionality of combinatorial library descriptors, and Lin [15] used PCA to analyze single nucleotide polymorphisms from genomic data. We chose to implement Laplacian embedding rather than MDS or other algorithms in SpectralNET because of promising results in the field of machine learning [23]. Although dimensionality reduction is especially useful for analyzing high-dimensional data, Laplacian embedding is an elegant display choice for any input network (see the next section for an example using a scale-free biological network). For a simpler (linear) dimensionality-reduced view of the input data, SpectralNET also has the option of viewing the results of PCA (though this view is not available when a network definition file, such as a Pajek file, is used). Both Laplacian embedding and PCA can be viewed in three dimensions with a Virtual Reality Modeling Language (VRML) viewer.
Example analysis of a randomly-generated small-world network and a biological scale-free network
SpectralNET provides an easy-to-use interface for creating a randomly generated small-world network. All that is required is to supply the desired number of nodes, the desired number of neighbors to which to connect each node, and the desired random probability that an edge is re-wired. For this example we create a network with 300 nodes in which each node is connected to four neighbors, and edges are rewired with 4% probability.
The default view of the graph is its Fruchterman-Reingold display, which, as noted above, uses force-directed placement to draw graph nodes (Figure 1). While the Fruchterman-Reingold display offers a quickly generated view of large networks, relatively little information about the global organization of the network is observable in the display of this small-world network (one cannot tell, for example, that the graph is a small-world network by its Fruchterman-Reingold display alone). In order to see the graph as drawn by the Laplacian eigenvector components of each node, the "Laplacian Embedding" radio button underneath the graph display is selected. In contrast to the Fruchterman-Reingold display, the Laplacian embedding of this small-world network (Figure 2) conveys significantly more information about its topology. In this display, it is clear that the small-world network was generated by placing neighboring nodes next to each other in a ring-like fashion – the theoretical ring-structure is represented literally in the Laplacian embedding.
Figure 1 Fruchterman-Reingold display of a small-world network. Fruchterman-Reingold display of a randomly generated small-world graph. The node selection panel and node information panel are visible to the left of the display.
Figure 2 Laplacian embedding of a small-world network. Laplacian embedding of the randomly generated small-world network depicted in Figure 2, as drawn by SpectralNET.
Real-world biological networks are also amenable to topological analysis using Laplacian embeddings. In order to generate a suitable biological network to analyze, the MIPS Mammalian Protein-Protein Interaction Database [26] was downloaded and parsed into a node-node-weight file for import into SpectralNET (see: additional file 1: HumanPPI_nodenodeweight.txt). The Laplacian embedding of the largest connected component of the resulting graph (Figure 3) shows a central hub of highly connected proteins connected to four connected branches. Spectral analysis similar to that performed below shows that the network is scale-free in nature, as is further evidenced by the fact that there are many more low-degree proteins than high-degree proteins, with the relationship between number of proteins and protein degree following a power-law distribution (data not shown). The scale-free nature of this network suggests that highly-connected proteins in the central hub may perform a coordinating role for the proteins in this interaction network. Examining the most highly connected protein in the central hub of the network (indicated in Figure 3) shows that, indeed, it is the transcriptional co-activator SRC-1, which receives and augments signals from multiple pathways [27]. Readers with further interest in topological analysis of biological networks are encouraged to read Farkas et al. [28] for a global analysis of the transcriptional regulatory network of S. cerevisiae or Jeong et al. [29] for an analysis of the protein interaction network of the yeast.
Figure 3 Virtual Reality Modeling Language (VRML) diagram of a human protein interaction network. Laplacian embedding of a scale-free biological network generated from a curated online database of protein interactions in humans (MIPS Mammalian Protein-Protein Interaction Database). For data see additional file 1 HumanPPI_nodenodeweight.txt.
In addition to the graphical display of networks, SpectralNET enables analysis of spectral properties of input networks, which can shed light on graph topology. One way this can be achieved is to compare a small-world network similar to, but not identical to, the randomly generated small-world network described above. This graph is a small-world network created by attaching complete subgraphs, varying in size from three to six nodes, to nodes arrayed in a ring (see additional file 2: Small-world_nodenodeweight for the network definition file, originally described by Comellas [24]) (Figure 4). The spectral properties of this graph can be used to help identify the topology of the original graph, in this case by comparing their adjacency and Laplacian spectral densities (Figure 5) [5,20]. Spectral density measures the density of surrounding eigenvalues at each eigenvalue and serves as an especially useful metric of global graph topology. The plot of these values for the example network is most similar to the corresponding plots for a Watts-Strogatz network (in this network, there are 500 nodes connected to 6 neighbors, with a re-wiring probability of 1%), despite the fact that there are only 33 nodes in the example network. Thus, even when an example network has relatively few nodes, comparison of spectral properties of the graph to idealized graphs can yield clues about network's topology.
Figure 4 Laplacian embedding of an uploaded small-world network. Laplacian embedding of a small-world network (n = 33) created by attaching complete subgraphs to nodes arrayed in a ring. The subgraphs each appear as a single point because their constituent nodes have identical connectivity profiles, yielding identical Laplacian eigenvector components.
Figure 5 Comparison of spectral properties of two small-world networks. Plots of spectral density of the adjacency and Laplacian eigenvalues for a randomly-generated Erdos-Rényi graph, a randomly-generated Watts-Strogatz graph, a randomly-generated Barabási-Albert graph, and the small-world network depicted in Figure 4 consisting of complete subgraphs attached to nodes arrayed in a ring. The input small-world network is most similar to the randomly-generated Watts-Strogatz network, since they have the most similar topologies.
Dimensionality reduction of a real-world chemical dataset to analyze QSAR
In addition to performing spectral analysis of networks, SpectralNET can also perform dimensionality reduction on chemical datasets to analyze quantitative structure activity relationships (QSAR). In this example, we upload a set of chemical descriptor data into SpectralNET and analyze it using the Laplacian Eigenmap algorithm originally developed by Belkin and Niyogi [13]. This dataset contains one small molecule, each created by the same diversity-oriented synthesis pathway [25], per row of the input file. Each column of the data represents a different molecular descriptor – metrics used to capture an aspect of the compound, such as volume, surface area, number of rings, etc.
The Laplacian Eigenmap algorithm in SpectralNET connects these small molecules to their K-nearest neighbors (measured by Euclidean distance), where K is an algorithmic parameter supplied by the user. In this example, we choose K = 7 to yield a reasonable number of edges in the resulting graph. Weights are assigned to each edge in one of two ways – every edge can have a weight of one, or weights can be assigned to edges by the following formula:
Wij=e−‖x1−x2‖2t (7)
MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqabeGadaaakeaacqWGxbWvdaWgaaWcbaGaemyAaKMaemOAaOgabeaakiabg2da9iabdwgaLnaaCaaaleqabaGaeyOeI0YaaSaaaeaadaqbdaqaaiabdIha4naaBaaameaacqaIXaqmaeqaaSGaeyOeI0IaemiEaG3aaSbaaWqaaiabikdaYaqabaaaliaawMa7caGLkWoadaahaaadbeqaaiabikdaYaaaaSqaaiabdsha0baaaaGccaWLjaGaaCzcamaabmaabaGaeG4naCdacaGLOaGaayzkaaaaaa@441D@
where Wij represents the weight of an edge connecting edges i and j and t is an algorithmic parameter [13]. For the molecular descriptor dataset, edge weights of one were chosen (it should be noted that when applying the second method to this dataset, increasing values of t eventually resulted in convergence to the same result as this method around t = 20,000). SpectralNET also offers the choice of performing PCA on input data before performing the Laplacian Eigenmap algorithm, which is performed by default and remains enabled for this example.
The resultant Laplacian embedding of the graph, which can be viewed by selecting the "Laplacian Embedding" radio button underneath the graph view pane, is the reduced dimensionality result of the Laplacian Eigenmap algorithm (Figure 6). Like PCA, the Laplacian Eigenmap algorithm performs dimensionality reduction on an input dataset such that relationships among the data are captured by fewer dimensions. Unlike PCA, however, it is not a linear transformation of the data, and the resulting non-linear dimensionality reduction can offer a more powerful view of the data than does PCA.
Figure 6 Laplacian Eigenmap result for a molecular descriptor dataset. A network of small molecules encoded as molecular descriptors, connected by similarity and displayed using the Laplacian Eigenmap algorithm, which plots each small molecule according to its corresponding Laplacian eigenvector components. Small molecules are colored according to the value of their minimized energy, one of the molecular descriptors of the original dataset.
Because Laplacian Eigenmaps is a local, rather than global, algorithm, it seeks to preserve local topological features of the data in its reduced-dimensionality space [13]. Thus, it is difficult to compare its performance relative to a linear, global algorithm like PCA without labeled features on which to classify the data and a rigorous comparison across multiple datasets and datatypes. However, by visual inspection of points clustered together in the Laplacian Eigenmap result (from the highlighted areas in Figure 6), one can see that they are structurally similar relative to a set of random compounds selected from the space as a whole (Figure 7), and the two outlier groups visible in the original image are also chemically similar (data not shown). The same dataset plotted on its first two principal components (via PCA) yields no significant clustering comparable to that of Laplacian Eigenmaps with instead one large and a second smaller diffuse cluster visible (Figure 8). Additional support for nonlinear QSAR methods comes from Douali et al. [30], which found that a nonlinear QSAR approach using neural networks predicted activities very well, outperforming other methods found in the literature. A more rigorous comparison of these algorithms in the context of molecular descriptor data is ongoing.
Figure 7 Comparison of chemical structures from Laplacian Eigenmap clusters. Comparison of chemical structures from the example real-world dataset of molecular descriptors depicted in Figure 5, taken either (A) from the group labeled "A", (B) from the group labeled "B", or (C) at random from the entire set.
Figure 8 Principal Components Analysis result for a molecular descriptor dataset. The network of small molecules depicted in Figure 5, displayed using the first two principal components of the data as derived from PCA. Small molecules are colored according to the values of their minimized energies.
Conclusion
SpectralNET provides an easily accessible means of analyzing graph-theoretic metrics for data modeling and dimensionality reduction. The software allows users to analyze idealized random networks or uploaded real-world datasets, and exposes metrics like the clustering coefficient, average distance, and degree distribution in an easy-to-use graphical manner. In addition, SpectralNET calculates and plots eigenspectra for three important matrices related to the network and provides several powerful graph visualizations.
SpectralNET is available as both a standalone .NET executable and an ASP.NET web application. Source code is available by request from the author.
Availability and requirements
Project name: SpectralNET
Project home page:
Operating system(s): Windows
Programming language: C#
Other requirements: The .NET framework v1.1 or higher
License: The SpectralNET software is provided "as is" with no guarantee or warranty of any kind. SpectralNET is freely redistributable in binary format for all non-commercial use. Source code is available to non-commercial users by request of the primary author. Any other use of the software requires special permission from the primary author.
Any restriction to use by non-academics: Contact authors
Authors' contributions
JF developed and tested the software, wrote the initial version of the manuscript, and co-designed the software; PC provided feedback and data for molecular descriptor analysis, assisted with design of the software, and edited the manuscript; SS provided project guidance and edited the manuscript; SH initially conceived of and co-designed the software and edited the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Human PPI network definition file. Network definition file representing a network of human protein-protein interactions. Data for this network was parsed from the MIPS Mammalian Protein-Protein Interaction Database. The numbers contained in this file correspond to the "shortLabel" annotation of proteins in the XML representation of the MIPS database.
Click here for file
Additional File 2
Small-world network definition file. Network definition file for a 33-node small-world network with attached complete subgraphs.
Click here for file
Acknowledgements
We gratefully acknowledge the Broad Institute of Harvard University and MIT, the National Cancer Institute (Initiative for Chemical Genetics), and the National Institute of General Medical Sciences (Center of Excellence for Chemical Methodology and Library Development) for support of this research. S.L.S. is an Investigator at the Howard Hughes Medical Institute.
==== Refs
Eisenberg E Levanon E Preferential attachment in the protein network evolution Phys Rev Lett 2003 91 138701 14525344 10.1103/PhysRevLett.91.138701
Fera D Kim N Shiddelfrim N Zorn J Laserson U Gan HH Schlick T RAG: RNA-As-Graphs web resource BMC Bioinformatics 2004 5 88 15238163 10.1186/1471-2105-5-88
Haggarty S Clemons P Schreiber S Chemical genomic profiling of biological networks using graph theory and combinations of small molecule perturbations J Am Chem Soc 2003 125 10543 10545 12940736 10.1021/ja035413p
Chartrand G Introductory Graph Theory 1985 New York: Dover
Chung F Spectral Graph Theory 1997 Providence: American Mathematical Society
Goano M Algorithm 745: Computation of the complete and incomplete Fermi-Dirac integral ACM Trans Math Software 1995 21 221 232 10.1145/210089.210090
Erdõs P Rényi A On random graphs I Publ Math Debrecen 1959 6 290 297
Barabási AL Albert R Emergence of scaling in random networks Science 1999 286 509512
Albert R Barabási AL Topology of evolving networks: Local events and universality Phys Rev Lett 2000 85 5234 5237 11102229 10.1103/PhysRevLett.85.5234
Watts D Strogatz S Collective dynamics of 'small-world' networks Nature 1998 393 440 442 9623998 10.1038/30918
Barabási AL Dezso Z Ravasz E Yook SH Oltvai Z Scale-free and hierarchical structures in complex networks Modeling of Complex Systems 2003 661 1 16
Batagelj V Mrvar A PAJEK – program for large network analysis Connections 1998 21 47 57
Belkin M Niyogi P Laplacian eigenmaps for dimensionality reduction and data representation Neural Computation 2003 15 1373 1396 10.1162/089976603321780317
Hotelling H Analysis of a complex of statistical variables into principal components J Educ Psychol 1931 24 417441
Lin Z Altman R Finding haplotype tagging SNPs by use of principal components analysis Am J hum Genet 2004 75 850 861 15389393 10.1086/425587
Nacher JC Ueda N Yamada T Kanehisa M Akutsu T Clustering under the line graph transformation: application to reaction network BMC Bioinformatics 2004 5 207 15617578 10.1186/1471-2105-5-207
Mohar B Hahn G, Sabidussi G Some applications of Laplace eigenvalues of graphs Graph Symmetry: Algebraic Methods and Applications 1997 Dordrecht: Kluwer Academic Publishers 225 275
Barabási AL Ravasz E Vicsek T Deterministic scale-free networks Physica A 2001 299 559564 10.1016/S0378-4371(01)00369-7
Chung F The average distance and the independence number J Graph Theory 1988 12 229 235
Farkas I Derenyi I Barabási AL Vicsek T Spectra of "real-world" graphs: Beyond the semicircle law Phys Rev E 2001 64 026704 10.1103/PhysRevE.64.026704
Fruchterman T Reingold E Graph drawing by force-directed placement Software: Practice and Experience 1991 21 1129 1164
Agrafiotis D Lobanov V Multidimensional scaling of combinatorial libraries without explicit enumeration J Comput Chem 2001 22 1712 1722 10.1002/jcc.1126
Belkin M Niyogi P Semi-supervised learning on Riemannian manifolds Machine Learning 2004 56 209 239 10.1023/B:MACH.0000033120.25363.1e
Comellas F Sampels M Deterministic small-world networks Physica A 2002 309 231 235 10.1016/S0378-4371(02)00741-0
Stavenger R Schreiber S Asymmetric Catalysis in Diversity-Oriented Organic Synthesis: Enantioselective Synthesis of 4320 Encoded and Spatially Segregated Dihydropyrancarboxamides Angew Chem Intl Ed 2001 40 3417 3421 10.1002/1521-3773(20010917)40:18<3417::AID-ANIE3417>3.0.CO;2-E
Pagel P Kovac S Oesterheld M Brauner B Dunger-Kaltenback I Frishman G Montrone C Mark P Stümpflen V Mewes H Ruepp A Frishman D The MIPS mammalian protein-protein interaction database Bioinformatics 2005 21 832 834 15531608 10.1093/bioinformatics/bti115
Kalkhoven E Valentine J Heery D Parker M Isoforms of steroid receptor co-activator 1 differ in their ability to potentiate transcription by the oestrogen receptor The EMBO Journal 1998 17 232 243 9427757 10.1093/emboj/17.1.232
Farkas I Jeong H Vicsek T Barabási AL Oltvai Z The topology of the transcription regulatory network in the yeast, S. cerevisiae Physica A 2003 318 601 612 10.1016/S0378-4371(02)01731-4
Jeong H Mason S Barabási AL Oltvai Z Lethality and centrality in protein networks Nature 2001 411 41 42 11333967 10.1038/35075138
Douali L Villemin D Cherqauoi C Neural networks: Accurate nonlinear QSAR model for HEPT derivatives Chem Inf Comput Sci 2003 43 1200 1207 10.1021/ci034047q
|
16236170
|
PMC1276787
|
CC BY
|
2021-01-04 16:27:46
|
no
|
BMC Bioinformatics. 2005 Oct 19; 6:260
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-260
|
oa_comm
|
==== Front
BMC BiolBMC Biology1741-7007BioMed Central London 1741-7007-3-231624626010.1186/1741-7007-3-23Research ArticleWnt5 signaling in vertebrate pancreas development Kim Hyon J [email protected] Jack R [email protected] Jose [email protected] Saulius [email protected] Anna [email protected] Shuo [email protected] Stephen C [email protected] Department of Genetics, Cell Biology, and Development, University of Minnesota, Minneapolis, MN 55455 USA2 Department of Molecular, Cellular, and Developmental Biology, University of California, Los Angeles, Los Angeles, CA 90095 USA3 Department of Pediatrics, University of Minnesota, Minneapolis, MN 55455 USA4 Department of Laboratory Medicine and Pathology, University of Minnesota, Minneapolis, MN 55455 USA2005 24 10 2005 3 23 23 10 6 2005 24 10 2005 Copyright © 2005 Kim et al; licensee BioMed Central Ltd.2005Kim et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Signaling by the Wnt family of secreted glycoproteins through their receptors, the frizzled (Fz) family of seven-pass transmembrane proteins, is critical for numerous cell fate and tissue polarity decisions during development.
Results
We report a novel role of Wnt signaling in organogenesis using the formation of the islet during pancreatic development as a model tissue. We used the advantages of the zebrafish to visualize and document this process in living embryos and demonstrated that insulin-positive cells actively migrate to form an islet. We used morpholinos (MOs), sequence-specific translational inhibitors, and time-lapse imaging analysis to show that the Wnt-5 ligand and the Fz-2 receptor are required for proper insulin-cell migration in zebrafish. Histological analyses of islets in Wnt5a-/- mouse embryos showed that Wnt5a signaling is also critical for murine pancreatic insulin-cell migration.
Conclusion
Our results implicate a conserved role of a Wnt5/Fz2 signaling pathway in islet formation during pancreatic development. This study opens the door for further investigation into a role of Wnt signaling in vertebrate organ development and disease.
==== Body
Background
Wnt signaling pathways play important roles in both normal development and in the pathogenesis of a variety of diseases, including cancer [1]. Activation of a Wnt signaling pathway requires interaction between a secreted glycoprotein, Wnt, and a seven-pass transmembrane receptor protein, Frizzled (Fz). Different combinations of Wnt and Fz ligand-receptor pairs can transduce at least three distinct kinds of intracellular signaling pathways. The "canonical" Wnt signaling pathway (Wnt/β-catenin pathway) results in changes of intracellular β-catenin levels and is thought to be involved in cell fate specification and proliferation. Wnt pathway activation can also lead to changes in either intracellular Ca2+ concentration (Wnt/Ca2+ pathway) or actin cytoskeleton reorganization (Wnt/tissue polarity pathway) [2]. The role(s) of these 'non-canonical' Wnt signaling pathway(s) in organ formation are largely unknown.
A function for Wnt signaling has been suggested by the expression patterns of Wnt and Fz genes during development of the mouse embryonic pancreas [3]. At embryonic day E11, Wnt5a and Fz-2 are expressed in the mesenchyme and epithelium of the pancreas. At E17.5, both Wnt5a and Fz-2 are co-localized with insulin- and glucagon-expressing cells. In contrast, only Wnt5a is expressed in the surrounding mesenchyme. In situ hybridization and RT-PCR gene expression analysis showed that both Wnt5a and Fz-2 are expressed in the embryonic pancreas from E11 until the end of gestation as well as postnatally. The highest level of expression is at E12 for Wnt5a, and at E12-E13 for Fz-2. Overexpression of Wnt5a in the pancreas results in the formation of multiple small and scattered islets, but the mechanism for such abnormality has not been characterized [3].
We explored a role of Wnt signaling in insulin-positive cell migration to form a pancreatic islet. In the mouse embryo at about E9.5, the primitive pancreatic epithelial cells express a transcription factor, pdx-1. Glucagon-positive cells are first detected around E10.5, and insulin-positive cells around E11.5 within the pancreatic ductal epithelium [4]. At E15.5, clusters of intermingled insulin-positive and glucagon-positive cells are found in the pancreatic interstitium, largely associated with the ducts [4,5]. During the last 4 days of gestation and postnatally, endocrine cells detach from the ducts, increase in number, and, at E17.5-E18.5, reorganize to form mature islets with the core of insulin-expressing cells surrounded by glucagon-expressing cells [6]. Formation of mature islets is thought to require migration of endocrine cells out of the pancreatic ductal epithelium to the pancreatic mesenchyme. These processes are partially controlled by matrix metalloproteinases (MMPs), a family of enzymes that degrade extracellular matrix proteins [7]. TGF-β signaling is necessary for the activation of MMP-2, which affects islet morphogenesis in vitro [8]. Recently, it has been reported that EGF signaling also regulates activation of MMP-2 and affects insulin-positive cell migration [6]. In mice lacking EGF-receptors, the majority of insulin-positive cells remain associated with pancreatic ducts in the newborn period.
The zebrafish pancreas functions similarly to that of other vertebrates by secreting hormones and exocrine enzymes to regulate blood glucose level and participate in digestion, respectively [9,10]. As in other vertebrates, synthesis and secretion of endocrine hormones in zebrafish occur in an islet called the Brockmann body, but unlike other vertebrates, only a single islet initially forms [11]. Thus, the zebrafish islet provides a model for the simplest endocrine pancreas with the core biological complexity of other vertebrates. Insulin-positive cells are specified as bilateral patches around the 14-somite stage and subsequently form a single islet in the midline [12]. Immunohistochemical studies using antibodies against insulin and glucagon revealed that the zebrafish islet consists of a core of insulin-expressing cells surrounded by glucagon-expressing cells, a structural organization similar to that observed in the mouse and the human [11,12]. However, the pattern of cell migration and the molecular mechanisms of zebrafish islet morphogenesis have not been previously investigated.
We characterized the process of insulin-positive cell migration in zebrafish and examined a role of Wnt signaling in pancreatic islet formation in zebrafish and the mouse. We show that wnt-5/fz-2 signaling is required for proper islet formation in zebrafish, and we demonstrate that Wnt5a signaling is required for the separation of islets from the ducts in the mouse. These phenotypes in zebrafish and mouse are consistent with defects in insulin-positive cell migration demonstrating a new and conserved role of Wnt signaling in vertebrate endocrine pancreas formation.
Results
Loss of direction results in scattered insulin-positive cells in embryos injected with fz-2 morpholinos
To study the function of fz-2, we used morpholino-modified oligonucleotides (MOs) as sequence-specific translational inhibitors in zebrafish [13]. Injection of fz-2 MOs [14] into transgenic zebrafish embryos ('morphants') carrying the GFP reporter gene under the control of the insulin promotor (insulin:GFP transgenic fish: [15]) resulted in scattered GFP-positive cells compared to the single islet observed in wild-type (WT) embryos (Fig. 1A,B). To determine the cause of abnormal insulin expression in FZ-2 morphants, we conducted a time-lapse imaging analysis. In wild-type embryos, bilateral insulin-positive cells first appeared at approximately the 14-somite stage. These cells divided and actively migrated to the posterior. As insulin-positive cells migrate, these cells converge in the midline. By the 24 hours-post-fertilization (hpf) stage, all insulin-positive cells associate to form a single islet at the midline (Fig. 1C–H [also see additional file 1]). We have also performed a time-lapse analysis on insulin:GFP transgenic embryos injected with rhodamine-conjugated dextran (Molecular Probe, D-1841). This created embryos with GFP expression in the insulin-positive cells and rhodamine localized to random cells in a mosaic pattern. We found that the GFP-negative cells labeled with rhodamine did not change their relative positions and did not show any morphological changes associated with migrating cells, while GFP-positive cells moved posteriorly and medially, arguing that the GFP positive cells move relative to their neighboring cells [see additional file 2]. Plotting the trajectory of GFP-positive cells demonstrated that the pattern of cell migration is mostly along a straight line from the initial to the final position (Fig. 1O). The average migration rate of a GFP positive cell was 0.3 ± 0.12 μm/minute and average A-P progression was 0.3 ± 0.04 μm/minute (n = 10).
Figure 1 Time-lapse imaging of insulin:GFP transgenic embryos shows cell migration defects in Fz-2 morphants. (A, C-H) Uninjected insulin:GFP transgenic embryo, (B, I-N) fz-2 MO-injected insulin:GFP transgenic embryo. All panels are dorsal views and anterior is to the left. Scale bar represents 100 μm. (A) Uninjected transgenic embryo, 24 hpf. (B) Fz-2 MO-injected transgenic embryo, 24 hpf. (C) At the 14-somite stage, bilateral patches of GFP-positive cells are visible in uninjected embryo. (D) At the 15–16 somite stage, GFP-positive cells have started proliferating. (E-G) At the 17 somite to 24 hpf stages, GFP-positive cells are aligned in bilateral rows of cells and undergo a medial and posterior migration. (H) At 24 hpf, all GFP-positive cells have merged to form one islet. (I) At the 14-somite stage, bilateral patches of GFP expression are apparent in fz-2 MO-injected embryos similar to uninjected embryos. (J-M) GFP-positive cells migrate in random directions in fz-2 morphant embryos. (N) At 24 hpf, GFP-positive cells have still not merged. (O) Trajectory of GFP-positive cells in uninjected insulin:GFP embryo. Notice that cells are uniformly moving posteriorly. (P) Trajectory of GFP-positive cells in fz-2 MO-injected insulin:GFP embryo. Notice cells are moving in random directions. A: anterior, P: posterior, T: time, L: left, R: right, O: origin.
In transgenic insulin:GFP zebrafish embryos injected with fz-2 MO, bilateral insulin-positive cells also appeared at the 14-somite stage, as observed in uninjected transgenic embryos. However, these cells failed to migrate to the posterior. At 24 hpf, insulin-positive cells remained scattered in fz-2 morphants (Fig. 1I–N [also see additional file 3]). The average migration rate of a GFP-positive cell was 0.3 ± 0.08 μm/minute, similar to uninjected transgenic embryos, indicating that the GFP-positive cells are capable of moving at the normal speed. However, the trajectory of insulin-positive cells in fz-2 MO-injected embryos showed that the cells were moving randomly (Fig. 1P). The average A-P progression of an insulin-expressing cell in fz-2 morphant embryos was -0.02 ± 0.07 μm/minute (n = 10). These data argue that GFP-positive cells still can move normally but have lost directional information in fz-2 MO-injected transgenic embryos.
Fz-2 and insulin are expressed in neighboring cells
As we reported earlier, fz-2 is restricted to somitic mesoderm and posterior paraxial mesoderm starting at the one somite stage. This expression pattern remains similar during somitogenesis, although the expression level of fz-2 in somitic mesoderm gradually decreases after 12–14 somite stage. (Fig. 2A, [14]). Double in situ hybridization of insulin and fz-2 and subsequent sectioning of the embryo showed that fz-2 expression is concentrated on the surface of somatic mesoderm and the entire endoderm adjacent to insulin-expressing cells (Fig. 2A,B). To determine if fz-2 is expressed in the insulin-positive cells, we sorted GFP-positive and -negative cells from the 20 somite stage insulin:GFP trangenic zebrafish embryos using Fluorescent Automated Cell Sorting (FACS), isolated total RNA from each sample, synthesized cDNA, and performed RT-PCR analysis for the presence of fz-2 transcript in the GFP-positive cells. As expected, EF-1α, a positive control, but not insulin was detected in the GFP-negative cells (Fig. 2C). In contrast, both EF1-α and insulin transcripts were detected in GFP-positive sample (Fig. 2C) indicating that the cell sorting procedure was effective. Two different sets of primers designed to amplify fz-2 transcripts produced bands from the GFP-negative cells but not from the GFP-positive cells (Fig. 2C). Our analysis shows that even though fz-2 and insulin are expressed in neighboring cells as detected by in situ hybridization, they are not co-expressed in the same cells.
Figure 2 Migration defects in Fz-2 morphant embryos can be rescued by synthetic fz-2 mRNA. (A) Double in situ hybridization with fz-2 and insulin at 20 somite stage of development. Arrow, insulin; arrowhead, fz-2 expression in the endoderm; dotted line, approximate position of the section in (B). (B) A section of double in situ hybridization with fz-2 and insulin. Fz-2 is expressed more strongly on the surface of mesoderm and entire endoderm. Arrow, insulin; arrowhead, fz-2 expression in the endoderm; a, arteries; asterisk, neural tube; d, pronephric duct. (C) RT-PCR using cDNA made from sorted cells of transgenic insulin:GFP zebrafish embryos. L: ladder; lanes 1–5: GFP-negative cells; lanes 6–10: GFP-positive cells; lanes 1, 6: EF1α lanes2, 7: insulin; lanes 3, 8: fz-2 primer set #1; lanes 4, 9: fz-2 primer set #2; lanes 5, 10: wnt-5. (D) High-dose injection of either fz-2 MO1 or MO2 resulted in scattered insulin expression, whereas low dose injection of either MO caused such defects in less than 10% of embryos. Co-injection of low dose fz-2 MO1 and MO2 resulted in synergistic increase of percentage of embryos with scattered insulin expression. (E) 80% of fz-2 MO-injected embryos displayed scattered insulin expression. Co-injection of fz-2 MO and fz-2 RNA reduced the percentage of embryos with abnormal insulin expression down to 45%. (F-I) In situ hybridization with insulin at 24 hpf stage, anterior is to the left, (F) fz-2 MO1-injected embryo, (G) fz-2 mismatch MO-injected embryo, (H) fz-2 RNA-injected embryo, (I) fz-2 MO- and fz-2 RNA-co-injected embryo. Notice the compact islet in this embryo that displays an undulated notochord.
Fz-2 plays a specific role in pancreatic islet formation
To confirm the identity of the scattered GFP-positive cells in fz-2 MO-injected insulin:GFP embryos, fz-2 morphant embryos were fixed at 24 hpf and analyzed for insulin expression by in situ hybridization. At 24 hpf, wild-type zebrafish embryos have a single islet consisting of 15–20 cells at the midline. In contrast, insulin-expressing cells in fz-2 morphant embryos were scattered along the A-P axis similar to the pattern of GFP-expressing cells in fz-2 MO-injected transgenic embryos (Fig. 2F). The number of insulin-positive cells in fz-2 morphants was not significantly different from control embryos despite the abnormal islet morphology, indicating that cell proliferation is not significantly affected (average: 12 ± 2 cells in wild-type and 10 ± 2 cells in MO-injected, n = 25 for each group).
Injection of two different previously published MOs targeting non-overlapping 5' regions of fz-2 mRNA generated similar effects on insulin-expressing cell migration. Furthermore, co-injection of two fz-2 MOs resulted in a synergistic increase in a percentage of embryos with scattered insulin expression, suggesting a specific role of fz-2 in proper islet formation (Fig. 2D). To assess whether scattered insulin expression observed in fz-2 MO-injected embryos was specific to the loss of fz-2 function, we also injected a five-base mismatch fz-2 MO, which elicited no effect on insulin-expressing cell migration (Fig 2E,G). To test the specificity of the observed phenotype in fz-2 MO-injected embryos further, we examined the ability of synthetic fz-2 mRNA to reverse the observed defects in insulin-positive cell migration. The injected synthetic fz-2 mRNA contains a β-globin 5' leader sequence and does not contain the fz-2 MO target-site sequence. Injection of fz-2 MO resulted in 84% of the embryos (n = 77) with scattered insulin expression (Fig. 2E,F). None of the fz-2 mRNA-injected embryos (n = 81) displayed scattered insulin expression (Fig. 2E,H). In contrast, co-injection of fz-2 MO and fz-2 mRNA resulted in a significantly lower frequency of embryos (n = 112, p value = 0.001) with scattered insulin expression compared with embryos injected with fz-2 MO only (Fig. 2E,I). These results demonstrate that the scattered insulin expression observed in fz-2 morphant embryos is specific to the loss of fz-2 function.
Wnt-5 has a similar and specific role in islet formation
Because the Fz protein family can function as receptors of Wnt proteins, we wanted to test if any known Wnt protein is similarly required for insulin-positive cell migration. Loss-of-function of three Wnts – wnt-5, -8, and -11 – are each reported to cause developmental abnormalities grossly similar to the ones observed in fz-2 morphants [16,17]. Among these, wnt-5 exhibits the most similar expression pattern to fz-2 during somitogenesis. Zebrafish wnt-5 is weakly expressed in the posterior half of the somitic and lateral mesoderm and strongly in the tail bud during somitogenesis [18]. Interestingly, double in situ hybridization against wnt-5 and pdx-1 revealed that wnt-5 is expressed as a gradient of RNA from the posterior limit of pdx-1 expression to the tail-bud at the 10-somite stage when pdx-1 is first detected (Fig. 3A). However, we did not detect wnt-5 in the GFP-positive cells isolated from insulin:GFP transgenic zebrafish embryos indicating that wnt-5 is not co-expressed with insulin (Fig. 2C, lane 10: compare to wnt-5 band from GFP-negative cells in lane 5).
Figure 3 Wnt-5 has a specific role in islet formation. (A) Double in situ hybridization with pdx-1 and wnt-5, 10 som stage, dorsal view, the anterior is to the left. Arrow, pdx-1 expression, bracket, wnt-5 expression. (B-H) In situ hybridization with insulin at 24 hpf. (B) wild-type, (C) WNT-8 morphant embryos, (D) WNT-11 morphant embryos, (E) WNT-5 morphant embryos, (F)wnt-5 mismatch MO-injected embryos, (G) wnt-5 RNA injected embryo, (H) wnt-5 MO and wnt-5 RNA co-injected embryo. Notice the compact islet in this embryo that displays an undulated notochord. (I) Percentage of embryos with scattered insulin expression resulting from injection of wnt-5 MO reduced significantly from 60% to 10% when wnt-5 RNA was co-injected with wnt-5 MO. (J-L) Morphology at 24 hpf, (J) wild-type, (K) wnt-5 insertional mutant, (L) wnt-5 translation-blocking MO-injected embryos. Notice that wnt-5 MO injected embryos have more severe morphological phenotype than wnt-5 insertional mutant embryos. (M) RT-PCR analysis of wnt-5 transcript in wnt-5 exon-intron MO injected embryos. Injection of wnt-5 exon-intron MO results in severely shortened wnt-5 transcript. L:ladder, 1:EF-1α control, 2:wnt-5.
To identify a candidate Wnt ligand, we utilized injections of MOs against wnt-5, -8, and -11 that effectively generate mutant phenocopies [17,19]. We generated Wnt-5, -8, and -11 morphants as described and analyzed insulin expression by in situ hybridization at 24 hpf [17,19]. Islet formation was normal in Wnt-8 and -11 morphant embryos as analyzed for the expression of insulin, pdx-1, glucagon and somatostatin (Fig. 3B–D, data not shown). In contrast, analysis of insulin expression at 24 hpf of Wnt-5 morphant embryos revealed scattered insulin expression along the A-P axis, similar to the islet phenotype noted in fz-2 morphant embryos (Fig. 3E).
To address the specificity of wnt-5 function in pancreatic islet formation, we designed a second MO targeting the 3' splice-site of exon 3. Injection of this MO resulted in effective skipping of exons 2 and 3, leading to a transcript that is only about 200 bp in length (Fig. 3M). When we injected this MO, embryos showed scattered insulin expression similar to embryos injected with wnt-5 translation blocking MO (Fig. 3I), arguing that the zygotic function of wnt-5 is involved in insulin-positive cell migration. In addition, injection of a five-base-mismatch wnt-5 MO resulted in embryos with normal insulin expression (Fig. 3F,I). Furthermore, the observed scattered insulin expression pattern in Wnt-5 morphant embryos was ameliorated by the addition of wnt-5* mRNA that encodes an altered wnt5 open reading frame engineered with degenerate nucleotides in the region around the translation start site to avoid targeting by the wnt-5 MO (see Methods). When translation blocking wnt-5 MO was injected alone, 65% of embryos (n = 75) showed scattered insulin expression (Fig. 3E,I). Embryos injected with wnt-5* mRNA alone resulted in 94% of embryos (n = 55) with normal insulin expression (Fig. 3G,I). The percentage of embryos with abnormal insulin expression decreased to 12% (n = 53, p value = 0.003) when embryos were co-injected with wnt-5* mRNA (Fig. 3H,I).
We also analyzed insulin expression in two different published wnt-5 mutant alleles (alleles of the mutant locus pipetail; kind gifts of Dr. M. Hammerschmidt (pptti265) [20] and Dr. N. Hopkins(ppthi1780b) [21]). In situ analysis using the insulin marker did not detect any significant effect on pancreas development in embryos homozygous for either of the tested pipetail alleles (data not shown). Gross morphological examination of embryos from either pipetail allele demonstrated a significant inter-embryo variation in manifestation of the pipetail embryonic phenotype, with the insertional allele (ppthi1780b) exhibiting an overall more severe phenotype. The ppthi1780b allele showed phenotypes that were less extreme than the effects noted for the wnt-5 morphants (Fig. 3J–L). To investigate further whether the scattered insulin expression observed in WNT-5 morphant embryos is specific to the loss of WNT-5 function, we tested to see if carriers of the ppthi1780b allele are more sensitive to wnt-5 MO injection (Table 1). Injecting 2 ng of either translation-blocking or exon-intron junction targeting wnt-5 morpholinos in wild-type embryos resulted in less than 5% of embryos with scattered insulin expression. We injected the same low dose wnt-5 MOs into embryos obtained from a cross between wnt-5 ppthi1780b allele carrier and wild-type adult fish. In this batch of embryos, approximately 50% are expected to be carriers of the ppthi1780b allele. This study resulted in approximately 50% of embryos exhibiting scattered insulin expression (Table 1). Furthermore, genotyping of individual embryo showed that 90% of embryos with abnormal insulin expression were ppthi1780b carriers (n = 20) whereas 100% with normal insulin expression were wild-type (n = 20), arguing that carriers of the ppthi1780b allele are more sensitive to wnt-5 MO injection [see additional file 4]. These data argue that the observed wnt-5 MO phenotype is specific to the targeting of the wnt-5 gene and that zygotic wnt-5 function is required for normal pancreatic islet formation.
Table 1 Wnt-5 heterozygous embryos are more sensitive to wnt-5 MO injection in pancreatic islet formation.
WT WT X wnt-5/WT
trans-block wnt-5 MO exon-intron wnt-5 MO trans-block wnt-5 MO exon-intron wnt-5 MO
% of embryos with abnormal insulin expression 5 ± 2% 0% 46 ± 5% 52 ± 3%
total # of embryos N = 124 N = 113 N = 53 N = 57
2 ng of either wnt-5 translation-blocking or exon-intron junction targeting morpholinos were injected into either wild-type embryos or embryos generated from an outcross of a wnt-5 mutant carrier adult. In wild-type embryos, injection of either wnt-5 MO resulted in less than 5% of embryos with scattered insulin expression. However, injection of either wnt-5 MO resulted in about 50% of embryos with scattered insulin expression in embryos obtained from the wnt-5 mutant carrier outcross.
WNT-5 and FZ-2 morphant embryos display a similar pancreatic developmental defect
To determine which step of pancreatic development is perturbed, we analyzed the expression patterns of different markers in fz-2 and wnt-5 morphant embryos. We examined early endodermal markers, mixer at 50% epiboly stage and sox-17 at 90% epiboly stage in fz-2 and wnt-5 morphant embryos, and found that the expression of mixer and sox-17 were normal in those embryos (Fig. 4A–C, 4D–F). We also examined the anterior endoderm marker, fox-A2, and the posterior endoderm marker, gata-6, at 24 hpf in wild-type, fz-2 and wnt-5 morphant embryos. Expression of fox-A2 and gata-6 was normal in both fz-2 and wnt-5 morphant embryos (Fig. 4G–I, 4J–L). These results indicate that the abnormal insulin expression observed in both fz-2 and wnt-5 morphant embryos is not due to secondary defects because of abnormal endoderm specification or migration.
Figure 4 Early endoderm markers are not affected in Wnt-5 and Fz-2 morphant embryos. All pictures are dorsal views. (A, D, G, J, M) wild-type, (B, E, H, K, N) Fz-2 morphants, (C, F, I, L, O) Wnt-5 morphants. (A-C) mixer, 50% epiboly, (D-F) sox-17, 90% epiboly, (G-I) fox-A3, 24 hpf, (J-L) anterior endoderm expression of fox-A3, arrow, pancreatic endoderm, 24 hpf, (M-O) gata-6, 24 hpf. Scale bar = 300 μm.
At 24 hpf, somatostatin, a marker for mature pancreatic δ-cells, was scattered similarly to insulin in both fz-2 and wnt-5 morphant embryos (Fig. 5A–C). Glucagon, which is expressed in pancreatic endocrine α-cells, was scattered along the A-P axis but still expressed at the edge of the islet in both fz-2 and wnt-5 morphants (Fig. 5D–F). Other endocrine pancreas markers, islet-1 and fspondin-2b, were also scattered along the A-P axis (Fig. 5G–I, 5J–L).
Figure 5 Wnt-5 and Fz-2 morphant embryos exhibit similar pancreatic islet defects at 24 hpf. In all panels, anterior is to the left and 24 hpf. .A-I, dorsal view; J-L, lateral view. (A, D, G, J) Wild-type embryos. (B, E, H, K) Fz-2 morphants. (C, F, I, L) Wnt-5 morphants. In situ hybridization analysis of (A, B, C) somatostatin, (D, E, F) glucagon, notice a hollow spot in the middle of each patch, (G, H, I) islet-1, (J, K, L) fspondin-2b. Note scattered pancreatic cells in Fz-2 and Wnt-5 morphants.
At the 3dpf stage, insulin-positive cells remain scattered in both wnt-5 and fz-2 morphants (Fig. 6A–C). Expression of an exocrine pancreas marker, carboxypeptidase-A, and a liver marker, ceruloplasmin, was reduced in both fz-2 and wnt-5 morphant embryos (Fig. 5D–I respectively). The hollow spot indicating the position of the islet within the exocrine pancreas was not observed in either wnt-5 or fz-2 morphants (Fig. 5D–F). This indicates that pancreatic islets in wnt-5 and fz-2 morphants are not completely embedded within exocrine tissue as found in wild-type embryos. However, scattered insulin expression is not likely to be affected by the reduction of the exocrine pancreas, because the exocrine pancreas in zebrafish does not develop until after the completion of insulin-positive cell migration.
Our results suggest that fz-2 and wnt-5 have similar function in pancreatic islet development. Interestingly, examination of the transcription factor pdx-1 expression profile revealed an unexpected phenotype in these embryos. Pdx-1 is one of the earliest known markers for the entire pancreas and is important for both early pancreatic development and adult β-cell function [22]. The area of pdx-1 expression was expanded in both fz-2 and wnt-5 morphants as analyzed by in situ hybridization at 24 hpf (Fig. 6J–L). At 3dpf, pdx-1 expression is restricted to the pancreatic islet, duct, and part of the intestine (Fig. 6M). Pdx-1-expressing cells in fz-2 and wnt-5 morphant embryos remain scattered and do not coalesce into a single islet until 3dpf (Fig. 6K,L) similar to insulin-expressing cells. In addition, the area of pdx-1 expression in the intestine becomes reduced in fz-2 and wnt-5 morphant embryos compared to wild-type embryos at 3dpf (Fig. 6M–O). The expanded pdx-1 expression at 24 hpf and misexpression at 3dpf in fz-2 and wnt-5 morphants requires further study (see discussion).
Figure 6 Wnt-5 and Fz-2 morphant embryos have other similar defects. In all panels, view is dorsal, anterior is to the left. (A-I, M-O) 3dpf, (J-L) 24 hpf stage. (A, D, G, J, M) Wild-type embryos. (B, E, H, K, N) Fz-2 morphants. (C, F, I, L, O) Wnt-5 morphants. In situ hybridization analysis of (A-C) insulin, (D-F) carboxypeptidase A, notice the hollow spot indicating the position of the islet, (G-I) ceruloplasmin, (J-O) pdx-1, (M) arrow, pdx-1-staining in islet.
Wnt-5 and fz-2 can genetically interact
We next determined if wnt-5 and fz-2 could interact genetically. First, we co-injected low doses of wnt-5 and fz-2 MOs. If wnt-5 and fz-2 were in two independent signaling pathways, co-injection of MOs against these genes would result in an additive increase of embryos with abnormal insulin expression. In contrast, a synergistic increase of embryos with abnormal insulin expression would indicate that wnt-5 and fz-2 are either in the same signaling pathway or in two pathways that can interact in the normal insulin-positive cell migration process. Injection of a low dose of wnt-5 MO resulted in 10% of embryos (n = 60) with abnormal insulin expression. Injection of a low dose of fz-2 MOs resulted in 12% of embryos (n = 48) with abnormal insulin expression. When we co-injected wnt-5 MO and fz-2 MOs, 55% of embryos (n = 59) displayed abnormal insulin expression, a synergistic increase in the percentage of embryos with abnormal insulin expression (Fig. 7A). This result suggests that wnt-5 and fz-2 can interact genetically in the same signaling pathway during islet formation.
Figure 7 Wnt-5 and fz-2 are in the same signaling pathway. (A) Injection of either wnt-5 MO or fz-2 MO mix results in less than 10% of embryos with scattered insulin expression. Co-injection of wnt-5 and fz-2 MOs results in 50% of embryos with defects. (B) Injection of either wnt-5 mRNA or fz-2 mRNA did not cause secondary axis in Xenopus embryos, whereas co-injection with both mRNAs resulted in 40% of embryos with secondary axis. Control injections of GFP mRNA alone or together with fz-2 mRNA resulted in no embryos with secondary axis. (C-F) Xenopus embryos, tailbud stage, (C) wild-type, (D) wnt-5 and fz-2 mRNA co-injected, black arrows-point to the primary and secondary hatching glands, (E) wnt-5 mRNA injected, (F) fz-2 mRNA injected.
To test a possible direct interaction between zebrafish wnt-5 and fz-2 further, we used a Xenopus secondary axis induction assay [23]. In this system, a Fz-dependent induction of the secondary axis in Xenopus has been shown to be the result of direct receptor (Fz) activation by the Wnt ligand. This effect can be mediated by Wnt ligands and Fz receptors that normally function via either canonical or non-canonical signaling processes. In this experiment, we injected either wnt-5 mRNA or fz-2 mRNA into two ventral blastomeres of 4-cell stage Xenopus embryos. Injecting zebrafish wnt-5 RNA or fz-2 RNA elicited a shortened and mildly bent body at the tailbud stage without the induction of any secondary axes (Fig. 7B,E,F). When we co-injected both RNAs in Xenopus embryos, more than 50% of embryos showed a secondary axis at the tailbud stage (Fig. 7B,D). This suggests that zebrafish Wnt-5 and Fz-2 proteins can interact functionally and signal when placed in an appropriate cellular environment.
Fz-2 is epistatic to wnt-5
To determine if wnt-5 acts genetically upstream of fz-2, we removed Wnt-5 protein by injecting wnt-5 MO and assayed whether fz-2 mRNA can rescue the resulting migration defects. When wnt-5 MO was injected alone, 56% of injected embryos (n = 42) showed scattered insulin expression (Fig. 8A,B). Injection of fz-2 RNA alone did not cause significant abnormalities in insulin expression (3% of embryos, Fig. 8A,C). The percentage of embryos with scattered insulin expression significantly decreased to 20% (n = 50, p value = 0.019) when fz-2 RNA was co-injected with wnt-5 MO (Fig. 8A,D). In a control experiment, co-injection of wnt-5 MO and GFP RNA did not reduce the percentage of embryos with abnormal insulin expression (Fig. 8A).
Figure 8 Wnt-5 acts genetically upstream of fz-2. (A) Injection of wnt-5 MO alone results in 50% embryos with insulin cell defect. Injection of fz-2 mRNA results in 5% embryos with insulin cell defects. Co-injection of wnt-5 MO and fz-2 mRNA results in 20% of embryos with abnormal insulin expression. In a control experiment, co-injection of wnt-5 MO and GFP mRNA results in 45% of embryos with defects. (B-D) Insulin expression as analyzed by in situ hybridization at 24 hpf, (B) wnt-5 MO injected, (C) fz-2 RNA injected, (D) wnt-5 MO- and fz-2 RNA-injected embryos. Note that fz-2 mRNA rescues insulin cell migration defect in wnt-5 morphants. (E) Same dose of wnt-5 mRNA that can rescue the insulin cell migration defects in wnt-5 morphants cannot rescue the defects in fz-2 morphants.
In contrast, injection of wnt-5 RNA cannot ameliorate defects in fz-2 morphant embryos (Fig. 8E). Injecting fz-2 MO resulted in 80% of embryos (n = 106) with scattered insulin expression. Injecting wnt-5 RNA caused 5% of embryos (n = 42) with scattered insulin expression. Co-injection of fz-2 MO and wnt-5 RNA resulted in 88% of embryos (n = 84) with scattered insulin expression. These results demonstrate that fz-2 RNA can rescue abnormal islet morphology caused by the wnt-5 MO, whereas wnt-5 RNA cannot rescue abnormal islet morphology caused by fz-2 MO, placing fz-2 genetically downstream of wnt-5 in this process.
Wnt-5a knockout mice exhibit similar islet precursor cell migration defects and abnormal islet morphology
To test if the role of Wnt-5 signaling in pancreatic development is conserved between different vertebrates, we obtained mice heterozygous for a previously described Wnt5a null allele [24]. Wild type and Wnt5a-/- embryos were harvested at E16.5, E17.5 and E18.5 and examined for defects in pancreatic development. Although Wnt5a-/- embryos have defects in many structures, the size and macroscopic morphology of the pancreas were normal with a characteristic loose lobular appearance and normal anatomical positioning: with the head of the pancreas located at the curve of the duodenum, the body associated with the greater omentum at the ventro-posterior surface of the stomach, and the narrower tail region pointed towards the hilum of the spleen (data not shown). We next performed immunohistochemical analysis of consecutive pancreatic sections with antibodies against insulin and glucagon between E16.5, E 17.5 and E18.5. In both wild type and mutant pancreata at E16.5, insulin-positive cells were intermingled with glucagon-positive cells as described previously [4] and did not have a typical mature distribution with insulin cells located centrally and glucagon cells located peripherally (Fig. 9A–D). At E17.5, insulin-staining cells were becoming more compact and glucagon-staining cells began to assume a peripheral distribution in both wild type and mutant mice (Fig. 9E–H). Since this distinct cell architecture is not observed in wild type mice before E17.5 [4,25], the appearance of the peripheral distribution of glucagon-positive cells in the mutant islets at E17.5 argues against maturational delay.
Similar pattern was observed in the islets at E18.5 with insulin-producing β-cells in the center of the islet (Fig. 9I,K) and the glucagon-producing α-cells on the periphery of the islet (Fig. 9J,L). However, when compared to compact and round islets in wild type embryos, the mutant islets were streaked along the pancreatic ducts (Fig. 10C), similar to mice lacking EGF receptor [6]. This phenotype is also very similar to that of zebrafish wnt-5 and fz-2 morphants, in which glucagon positive cells are scattered along the A-P axis while positioned at the periphery of insulin-positive cells (Fig. 5E,F).
Figure 9 Pancreatic islet development in Wnt5a-/- mouse embryos is not delayed. (A-D) E16.5, (E-H) E17.5, (I-L) E18.5, (A, C, E, G, I, K) insulin antibody staining, (B, D, F, H, J, L) glucagon antibody staining, (A, B, E, F, I, J) pancreas tissue from wild-type siblings, (C, D, G, H, K, L) pancreas tissue from Wnt5a-/- mouse embryos. Notice that glucagon staining is round and spherical at E16.5, but positioned at the periphery of insulin cells at E17.5 and E8.5.
Figure 10 Wnt5a-/- islets remain in ductal proximity and have a streaked appearance at E18.5. (A) In Wnt5a-/- embryos, most islets are associated with ducts. Both small and large β-cell aggregates are more frequently associated with pancreatic ducts in Wnt5a-/- embryos at E18.5 than in wild-type embryos. (B, C) Insulin antibody staining. (B) Round and compact islets in wild-type embryos. A normal pancreas consists of islets that are associated and separated from ducts. Arrows: pancreatic duct, asterisk: an islet separated from duct. (C) Streak-like, fragmented islets in Wnt5a-/- mutant embryos.
To examine whether defects in endocrine cell migration during islet formation can cause streaked islets of Wnt5a-/- embryos we used a previously described morphometry method [6]. Since insulin-positive cells have to detach and migrate away from the pancreatic ducts to form mature islets, one of the measures of islet cell migration is the degree to which insulin-positive cells remain in direct contact with the ductal epithelium. To assess migration of the endocrine cells in Wnt5a null embryos, we concentrated our analysis at E18.5, the latest time point at which we can assess separation of insulin positive cells from the ducts. Lethality at birth of Wnt5a knockout mice prohibits later analysis of pancreas development in these mice. When compared to wild-type embryos, the number of islets separated from the ducts was greatly reduced in the pancreata of Wnt5a-/-embryos (Fig. 10A). Only 15% of small islets and 6% of larger islets were separated from the ducts in the mutants, compared to 56% (P < 0.001) and 40% (P < 0.05) respectively in the wild type mice (Fig. 10B,C). Our results implicate that Wnt5a signaling is also important in the insulin-positive cell migration process during morphogenesis of the pancreas in higher vertebrates.
Discussion
A hypothetical role of Wnt signaling in early pancreatic development
The function of a wnt-5/fz-2 pathway in insulin-positive cell migration is most likely to provide the right environment and/or signals for insulin-positive cell migration during islet formation. Our time-lapse analysis of fz-2 MO-injected transgenic embryos shows that the insulin-positive cells still migrate, but their direction of migration is random indicating these cells have lost key positional information in the absence of the wnt5/fz2 signal (Fig. 1). In addition, our cell sorting analysis shows that wnt-5 RNA is expressed in a posterior-anterior gradient, potentially serving as a key step in establishing the positional information required by these migrating cells. Analysis of the receptor, fz-2, however, reveals a further complexity. Fz-2 RNA is detected in the cells immediately adjacent to the mis-migrating insulin-positive cells, suggesting that the function of wnt-5/fz-2 pathway activation in islet cell formation is a cell-non-autonomous and/or cell-cell instructive function in the transmission of this positional information (Fig. 2B). Interestingly, non-autonomous fz-mediated signaling has been observed during Drosophila development [51], but this process of information exchange does not require any cell migration movements or, apparently, any wnt ligand [52]. Further study is required to determine how the wnt-5/fz-2 signaling produces the right environment that allows normal insulin-positive cell migration during islet formation.
The activation of Wnt-5 signaling can change intracellular Ca2+ concentration [26]. One hypothesized function of the Wnt/Ca2+ pathway is to antagonize a second and simultaneously active Wnt/β-catenin pathway [27,28]. Wnt/β-catenin signaling is known to specify cell fate directly [2]. For example, reduction of maternal and zygotic wnt-5 activity using genetic approaches results in hyperdorsalized embryos [27]. Further investigation is required to determine whether wnt-5/fz-2 signaling provides the right environment and/or signals for normal insulin-positive cell migration by antagonizing the Wnt/β-catenin pathway or by another novel mechanism.
We observed an altered expression pattern of pdx-1 transcript in both wnt-5 and fz-2 morphant embryos. In our time-lapse imaging analysis, however, we observed that the initial specification of insulin-expressing cells was normal, and we observed no ectopic insulin-positive cells in spite of expanded pdx-1 expression in fz-2 or wnt-5 morphant embryos. Furthermore, the number of insulin-expressing cells is not increased in either fz-2 or wnt-5 morphant embryos even though these embryos displayed two or more insulin expression domains, sometimes with single cell patches, as observed in our time-lapse imaging analyses. These data suggest that the scattered insulin-expression is not due to ectopic islet formation. Future work will be required to detail the role (if any) for this expanded pdx-1 expression in abnormal insulin-positive cell migration.
A role for Wnt-5 in pancreatic islet formation is conserved between mammals and teleosts
Vertebrate pancreatic development is described here in one of its simplest forms in zebrafish (represented by a single islet at 24 hours of development) and by a more complex vertebrate system in mouse. Certain processes of pancreatic islet formation in zebrafish and mouse appear to be incongruent. For example, sonic hedgehog (shh), which is expressed in the entire endoderm except for the pancreatic endoderm, was shown to be a negative regulator of the adoption of pancreatic endoderm fate in mice [29,30]. In zebrafish, however, shh and other hh proteins are not expressed in the endoderm [31]. In addition, ectopic expression of shh induces a pancreatic cell fate rather than represses it, a result opposite to what was observed in mammalian models [31,32]. Despite these differences, the initial specification of insulin-expressing cells occurs before an islet is formed in both animals, and these insulin-positive cells migrate and form an islet at a distance from where they are initially specified. Even though a comparable structure to a duct does not exist in zebrafish at the stage when islet formation occurs, previous studies in mouse and our time-lapse imaging analysis show that this migration process of islet formation is similarly required to form normal islets in both mouse and zebrafish. The final disorganized islet phenotypes were observed as a result of abnormal migration of insulin-positive cells in both zebrafish and mouse embryos with disrupted Wnt5 signaling. These results argue that Wnt5 signaling is required in pancreatic islet formation of both mouse and zebrafish, a role most likely shared throughout the vertebrate lineage.
Fz-2 is a putative receptor for Wnt-5 in pancreatic islet formation
Our results suggest that wnt-5 is upstream of fz-2 in the same genetic pathway and show that Fz-2 functions as a receptor for Wnt-5 when placed in the right environment, as assayed using the Xenopus secondary axis assay system. In mouse, fz-2 and wnt-5 transcripts are expressed at the highest level among different wnts and fzs in the developing mouse pancreas, providing evidence that wnt-5 and fz-2 are expressed at the right place and time to be involved in islet formation as a functional ligand and receptor pair [3]. In zebrafish, early fz-2 expression is observed during somitogenesis. This expression pattern coincides with the timing and location of insulin cell migration, indicating that fz-2 is expressed at the right time and place to be involved in normal insulin-positive cell migration in zebrafish. Although wnt-5, wnt-11 and wnt-8 are expressed at the 14-somite stage, only wnt-5 and wnt-11 transcripts are detected at the location close to the presumptive pancreatic endoderm marked by pdx-1. wnt-5 is not expressed in the endoderm, but it is expressed in adjacent mesoderm close to the posterior boundary of early pdx-1 expression. wnt-5 proteins may be secreted from the mesoderm and signal to adjacent endoderm. Furthermore, our data show that wnt-11 morphant embryos have a normal pancreatic islet formation, whereas wnt-5 morphant embryos have similar pancreatic islet formation defects as fz-2 morphant embryos placing wnt-5 as the best candidate ligand for fz-2 in pancreatic islet formation.
Wnt-5 signaling has a specific role in pancreatic islet formation that is separable from other Wnt signaling events in embryogenesis
In embryos with defects in body axis elongation, some significant secondary effects on other organizing centers may occur. For example, zebrafish wnt-5 [20], wnt-8 [19], wnt-11 [33] and knypek [34] morphant or mutant embryos display undulating notochord, reduced body length and a number of other less apparent abnormalities. These defects could conceivably result in abnormal pancreatic islet formation. However, wnt-8 and wnt-11 morphant embryos display normal islet formation despite body elongation defects. This supports a specific role of Wnt-5 signaling in pancreatic islet formation that is not secondary to body elongation defects. In addition, injection of wnt-5 and fz-2 mRNA ameliorates the pancreatic islet formation defects in wnt-5 and fz-2 morphant embryos, respectively, but not the body elongation defects under the same conditions. These results strongly suggest that pancreatic islet formation is at least partially independent of the body axis elongation process. In addition, the initial specification of insulin-positive cells occurs at the right place and time in zebrafish embryos. The congregation of insulin-positive cells occurs normally, and glucagon cells surround the insulin-positive cells in mouse Wnt5a-/-knockout embryos and wnt-5 and fz-2 morphant zebrafish embryos. Finally, the development of a grossly normal pancreas in Wnt5a-/- mouse knockout embryos, and the subsequent defect in islet formation, emphasizes the possible function of Wnt5 signaling during later stages of pancreatic islet morphogenesis.
Conclusion
We examined a role of Wnt-5 signaling in pancreatic islet morphogenesis in zebrafish and mouse embryos. Time-lapse imaging of insulin-positive cells using transgenic zebrafish embryos revealed that gradual formation of a single islet in zebrafish is a result of an active cell migration process. Here we demonstrate that wnt-5 and fz-2 morphant embryos display defects in insulin-positive cell migration providing evidence for a role of Wnt5 signaling in this process. Comparative analyses of pancreatic tissue from Wnt5a knockout mice show that endocrine cells are specified normally within the primitive pancreatic epithelium. However, these insulin-positive cells fail to migrate from the ducts and form fragmented islets around ducts with glucagon-expressing cells positioned at the periphery. This argues that Wnt5a is required for insulin-positive cell migration, but not local organization within a forming islet in both mouse and zebrafish. Taken together, these results provide evidence that wnt-5/fz-2 signaling is required for the migration of insulin-positive cells during pancreatic development, and that this function is conserved. This study will help us to understand better the role of Wnt signaling during vertebrate pancreatic development and broaden our understanding of a role of Wnt signaling in pancreatic diseases. Wnt signaling pathways have been implicated in the pathogenesis of cancer [35], regulation of adipogenesis and insulin secretion [36,37]. Wnt5b, for example, has recently been identified as a candidate gene for conferring susceptibility to type 2 diabetes [38]. Wnt5a may inhibit the canonical Wnt pathway by promoting β-catenin degradation [28]. Interestingly, dysregulation of β-catenin has been identified in various forms of malignancy, including pancreatic tumors [39]. Our study reports a new and conserved role of Wnt-5 signaling in pancreatic islet formation and provides an example of how Wnt signaling functions in organogenesis.
Methods
Synthesis and microinjection of mRNAs
Wnt-5 ORF cDNA was amplified from a gastrula stage library and subcloned into pT3TS [40]. The modified wnt-5* construct was made using a degenerate 5' primer designed to have a mismatch with the published wnt-5 MO (5'-ATG GAC GTA CGT ATG AAT CAG GGT CAC CTA CTT CTG GCA G-3'). The degenerate PCR fragment was subsequently subcloned into pT3TS vector at SpeI site. To synthesize sense wnt-5 mRNA, the wnt-5-T3TS construct was digested with XbaI and transcribed using T3 mMessage mMachine kit (Ambion). Modified wnt-5 mRNA (4 pg) was injected for the wnt-5 rescue experiment.
Fz-2 ORF cDNA was amplified from the gastrula library and subcloned into pT3TS. Fz-2 construct was cut with XbaI and transcribed with T3 mMessage mMachine kit (Ambion). For Xenopus assay, 500 pg of fz-2 mRNA was injected into two ventral blastomeres of 4-cell stage Xenopus embryos. For the fz-2 MO specificity experiment and wnt-5 and fz-2 epistasis experiment, 25 pg of fz-2 mRNA was injected.
Microinjection of MOs
We injected 2.5 ng of the previously described wnt-5 MO for the wnt-5 and fz-2 genetic interaction experiment and 8 ng of the same wnt-5 MO for all other experiments [17]. We also used 5 ng of wnt-5 MO (5'-TGTTTATTTCCTCACCATTCTTCCG-3') targeting the 3' end of the exon-intron junction of exon 3. As a control, 8 ng of a five-base mismatch wnt-5 MO (5'-GTCGTTGCTTCTTTCACACTTCCAT-3') was injected.
For the analysis of different pancreatic markers, we injected 5 ng of fz-2 MO mix (2.5 ng of fz MO-1 and 2.5 ng of MO-2) as described [14]. As a control, 5 ng of a five-base mismatch fz-2 morpholino (5'-CCTCCATAGTCACGATAAGTTCGGC-3') was injected. For the wnt-5/fz-2 genetic interaction experiment, we injected 2 ng of fz-2 MO mix (1 ng of fz-2 MO-1 and 1 ng of fz-2 MO-2). For the FZ-2 morphant rescue experiment, we used 2 ng fz-2 MO-1.
We injected 2–4 ng of the previously described wnt-11 MO to generate WNT-11 morphant embryos [17]. Two specific MOs targeting WNT-8 have been previously described [16,19]. Embryos were injected with 1–8 ng of wnt-8 MO1, 1–5 ng of wnt-8 MO2 or 2 ng wnt-8 MO mix (1 ng of wnt-8 MO1 and 1 ng of wnt-8 MO2).
All three groups of embryos were allowed to develop for 24 hours and subsequently analyzed for pancreatic markers. Microinjections were performed at the one-cell stage as described [40].
In situ hybridization
Single color in situ hybridization was performed as described [41]. Double color in situ hybridization was performed as described [42]. Fz-2 antisense riboprobe was synthesized as described [14]. To synthesize wnt-5 probe, wnt-5 ORF cDNA was sub-cloned into 4-TOPO vector (Invitrogen) and subsequently cut with NsiI and transcribed with T3 RNA polymerase. The following probes were used: pdx-1, insulin [43], glucagon, somatostatin [11], islet-1 [44], fspondin-2b [45], sox-17 [46], fox-A2 [47], gata-6 [48], mixer [35], ceruloplasmin [49] and carboxypeptidase A [32]. The mixer probe construct was synthesized by sub-cloning a 1 kb PCR fragment into the 4-TOPO vector. To synthesize the probe, DNA was cut with NotI and transcribed with T3. The fox-A2 probe construct was synthesized by sub-cloning a 1 kb PCR fragment of fox-A2 into the 4-TOPO vector. To synthesize the probe, DNA was digested with NotI and transcribed with T3. The carboxypeptidase A probe construct was synthesized by subcloning a 500 bp PCR fragment into the TOPO vector. To synthesize the probe, DNA was digested with NotI and transcribed with T3. The gata-6 probe was synthesized by digesting an EST clone obtained from ZFIN (cb 603) with Sal1 and transcribing with SP6.
RNA microinjection into Xenopus embryos
Wild-type zebrafish wnt-5 mRNA (250 pg) and 500 ng of zebrafish fz-2 RNA was injected alone or in conjunction into the two ventral blastomeres of the 4-cell stage Xenopus embryos (total of 500 pg wnt-5 RNA and 1 ng of fz-2 RNA). As a control, 500 ng of GFP RNA was injected alone or together with fz-2 RNA. The injected embryos were scored for double axes at the tailbud stage. Injections were performed as described [40].
Time-lapse imaging
The 14 somite stage embryos were dechorionated and embedded in 0.8% low melting point agarose with 0.05% 3-aminobenzoic acid ethyl ester in a home made holding chamber. To prevent drying, the embedded embryo was covered with mineral oil and then overlaid with a cover slip. The images were captured every minute for 4–6 hours using a digital camera (Hamamastu, C4742-95) attached to a compound microscope with automated base (Zeiss) and OpenLab program.
Cell dissociation and FACS analysis
Insulin:GFP transgenic zebrafish embryos at the 20 som stage were dechorionated and homogenized in cold 1X Danieau/ 15% fetal bovine serum (FBS) solution using a tissue grinder. Homogenized embryos were washed twice with cold 1X Dan/15% FBS and incubated in 1X Danieau/ 15% FBS containing 10 mg/ml collagenase/dispase (Roche, #10269638001), 10 mg/ml trypsin and 10 unit/ml DNase 1 at 30°C for one hour with gentle agitation. Dissociated cells were subsequently washed twice, filtered through a cell strainer and finally suspended in 1X Dan/15% FBS supplemented with 1 μg/ml DAPI (Roche, #10236276001: to label dead cells and other debris). Cells were sorted using a cell sorter (BD science), collected in 100% FBS, and processed for subsequent total RNA isolation using an RNAqueous-4PCR kit (Ambion, #1914). For subsequent RT-PCR, cDNA was synthesized using Powerscript Reverse Transcriptase (BD science, #8460). The following primers were used. EF1-a: TCACCCTGGGAGTGAAACAGC, ACTTGCAGGCGATGTGAGCAG, insulin: CCATATCCACCATTCCTCG, CGGAGAGCATTAAGGCCTG, fz-2 set#1: ATGCAGGCGAGTGGAAGTG, CTTATAGCTGAGGTAGGCTG, fz-2 set#2: TGGCGTCTGGACAGCATC, TCCTGGTCTGTGAAGAACAT, wnt-5: ATGGATGTGAGAATGAACCAA, CTACTTGCACACAAACTGGTC.
Genotyping of zebrafish ppthi1780b allele
Zebrafish embryos injected with low dose exon-intron blocking wnt-5 MO were raised up to 3 days. The heads of these embryos were excised and frozen in -80°C for further genotyping and the bodies were further processed for insulin expression using in situ hybridization. After the insulin expression pattern analysis, genomic DNA of 20 embryos with normal and abnormal insulin expression pattern was isolated from the frozen head tissue and genotyped as described [21].
Genotyping of mice
Genomic DNA was isolated from tails or yolk sacs by standard methods [50] and amplified by PCR using the following primers: W1 (5'-GAC TTC CTG GTG AGG GTG CGT G-3'), W2 (5'-GGA GAA TGG GCA CAC AGA ATC AAC-3'), and W3 (5'-GGG AGC CGG TTG GCG CTA CCG GTG G-3'). The PCR settings were: 30 cycles, 94°C for 30 seconds, 65°C for 30 seconds, 72°C for 1 minute. A 360 bp band identified a wild-type allele, while a 200 bp band identified the mutant allele.
Histological analysis
Embryos used in this study were obtained by intercrossing Wnt5+/- mice [24]. Plug date was considered E0.5. The wild type (N = 5) and Wnt5a-/- (N = 8) embryos were harvested between E16.5-E18.5. The pancreatic tissue was dissected from the embryos, fixed in 4% paraformaldehyde, washed in PBS, dehydrated through alcohols, embedded in paraffin and cut into 7–9 μm thick sections.
Immunohistochemistry
Immunohistostaining to detect the islets of Langerhans was done using commercially available mouse monoclonal anti-insulin antibodies (SIGMA I-2018) for β-cells, and anti-glucagon (G-2654) antibodies for α-cells as previously described [51]. The sections were rehydrated, incubated in methanol containing 1% H2O2, and washed in PBS. The sections were then incubated with primary antibodies (concentration 1:500) at 4°C overnight. After washing the slides in PBS, the sections were incubated with biotinylated mouse IgG at 1:500 (Vector Laboratories) at room temperature for one hour. In order to detect the staining, the slides were incubated in Vectastain reagent for 30 minutes, followed by the diaminobenzidine detection method. The slides were then background stained by dipping them in hematoxylin 2 for 5 seconds, washed and cover slipped.
Morphometry
We defined small islets as insulin-positive aggregates 25–70 μm in diameter and large islets as aggregates greater than 70 μm in diameter. A total of 31 islets from 3 wild type embryos and 56 islets from 5 Wnt5a-/- embryos were counted and classified as being either associated with ductal tissue or not associated. Islets were defined as associated with ductal tissue if insulin-positive cells were in direct contact with duct epithelium [6]. 16 WT and 38 knockout islets 20–70 μm in size, and 15 WT and 18 knockout islets over 70 μm in size, were counted and classified.
Statistical analysis
The statistical significance between the groups was determined by the Student's T-test for zebrafish experiments and by a Chi-Square test for the mouse experiment.
Authors' contributions
SCE conceived and designed this study. HJK conducted zebrafish and Xenopus experiments, analyzed results and drafted the manuscript. SS made the initial observation of defects in pancreatic islets of fz-2 morphants. JRS and AP performed immunohistochemistry of pancreatic islets from wild-type and wnt-5a KO mice, and JJ analyzed the results. JRS conducted islet morphometric analyses. SCE, AP and SL supervised the study. SCE and AP edited the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
(QuickTime format) WT movie. Figure 1C–H panels were taken from this movie at the indicated time points. GFP-positive cells are first observed around the 14 somite stage as bilateral patches in insulin:GFP transgenic embryos. These cells divide and actively migrate to the posterior. The net movement towards the midline is an indirect result of convergent-extension since active migration towards the midline is not observed. By the 24 hpf stage, GFP-positive cells are coalesced into a single islet in the posterior and midline relative to the starting position.
Click here for file
Additional File 2
Two color time-lapse imaging shows that GFP-positive cells migrate relative to the neighboring cells. (A) GFP-positive cells are first visible posterior to rhodamine-labeled cells as bilateral lows of cells. (B-G) GFP-positive cells migrate posteriorly and medially, whereas rhodamine-labeled cells do not change their relative position. (H) At 24 hpf, all GFP-positive cells coalesced into a single islet, but the rhodamine-labeled cells remain separated at their original position. Arrows: rhodamine-labeled cells; t: time (minutes).
Click here for file
Additional File 3
(QuickTime format) FZ-2 morphant movie. Figure 1I–N panels were taken from this movie at the indicated time points. As also noted in non-injected insulin:GFP transgenic embryos, GFP-positive cells are first observed around the 14 somite stage as bilateral patches, and these cells divide normally in fz-2 MO-injected embryos. However, the uniform migration towards the posterior observed in the non-injected transgenic embryo is not observed in fz-2 MO-injected embryos. Instead, the polarity of cell migration is entirely lost, with no net movement along the anterior-posterior axis observed for these cells (see text).
Click here for file
Additional File 4
Genotyping of embryos with normal and abnormal insulin expression. (A) 90% of embryos with abnormal insulin expression are carriers of ppthi1780b. Lane 4 and lane 15 represent wild-type embryos with abnormal insulin expression. (B) Control PCR for exon 1 of wnt-5, N = water only control. (C) None of the embryos with normal insulin expression are carriers of ppthi1780b. (D) Control PCR for exon 1 of wnt-5, N = water only control.
Click here for file
Acknowledgements
We thank A. McMahon for providing us with Wnt5a knockout mouse, M. Jarcho for technical assistance, and R. Sorenson for helpful discussions. This work was supported by NIH Medical Scientist Training Grant to HK (5 T32 GM08244-16), NIH grants to SCE (GM55877 and GM63904), NIH grants to SL, and Student Research Fellowship Award from the Lawson Wilkins Pediatric Endocrine Society to JRS. JRS is a recipient of The Howard Hughes Fellowship for Medical Students.
==== Refs
Giles RH van Es JH Clevers H Caught up in a Wnt storm: Wnt signaling in cancer Biochim Biophys Acta 2003 1653 1 24 12781368
Miller JR The Wnts Genome Biol 2002 3 REVIEWS3001 11806834
Heller RS Dichmann DS Jensen J Miller C Wong G Madsen OD Serup P Expression patterns of Wnts, Frizzleds, sFRPs, and misexpression in transgenic mice suggesting a role for Wnts in pancreas and foregut pattern formation Dev Dyn 2002 225 260 270 12412008 10.1002/dvdy.10157
Herrera PL Huarte J Sanvito F Meda P Orci L Vassalli JD Embryogenesis of the murine endocrine pancreas; early expression of pancreatic polypeptide gene Development 1991 113 1257 1265 1811941
Slack JM Developmental biology of the pancreas Development 1995 121 1569 1580 7600975
Miettinen PJ Huotari M Koivisto T Ustinov J Palgi J Rasilainen S Lehtonen E Keski-Oja J Otonkoski T Impaired migration and delayed differentiation of pancreatic islet cells in mice lacking EGF-receptors Development 2000 127 2617 2627 10821760
Birkedal-Hansen H Moore WG Bodden MK Windsor LJ Birkedal-Hansen B DeCarlo A Engler JA Matrix metalloproteinases: a review Crit Rev Oral Biol Med 1993 4 197 250 8435466
Miralles F Battelino T Czernichow P Scharfmann R TGF-beta plays a key role in morphogenesis of the pancreatic islets of Langerhans by controlling the activity of the matrix metalloproteinase MMP-2 J Cell Biol 1998 143 827 836 9813100 10.1083/jcb.143.3.827
Agrawal VP Sastry KV Kaushab SK Digestive enzymes of three teleost fishes Acta Physiol Acad Sci Hung 1975 46 93 98 14477
Khanna SS Rekhari K Effect of glucose loading on the blood sugar and the histology of the pancreatic islets in a freshwater teleost, Heteropneustes fossilis Acta Anat (Basel) 1972 82 126 137 4592086
Argenton F Zecchin E Bortolussi M Early appearance of pancreatic hormone-expressing cells in the zebrafish embryo Mech Dev 1999 87 217 221 10495291 10.1016/S0925-4773(99)00151-3
Biemar F Argenton F Schmidtke R Epperlein S Peers B Driever W Pancreas development in zebrafish: early dispersed appearance of endocrine hormone expressing cells and their convergence to form the definitive islet Dev Biol 2001 230 189 203 11161572 10.1006/dbio.2000.0103
Nasevicius A Ekker SC Effective targeted gene 'knockdown' in zebrafish Nat Genet 2000 26 216 220 11017081 10.1038/79951
Sumanas S Kim HJ Hermanson S Ekker SC Zebrafish frizzled-2 morphant displays defects in body axis elongation Genesis 2001 30 114 118 11477686 10.1002/gene.1043
Huang H Vogel SS Liu N Melton DA Lin S Analysis of pancreatic development in living transgenic zebrafish embryos Mol Cell Endocrinol 2001 177 117 124 11377827 10.1016/S0303-7207(01)00408-7
Solnica-Krezel L Stemple DL Mountcastle-Shah E Rangini Z Neuhauss SC Malicki J Schier AF Stainier DY Zwartkruis F Abdelilah S Driever W Mutations affecting cell fates and cellular rearrangements during gastrulation in zebrafish Development 1996 123 67 80 9007230
Lele Z Bakkers J Hammerschmidt M Morpholino phenocopies of the swirl, snailhouse, somitabun, minifin, silberblick, and pipetail mutations Genesis 2001 30 190 194 11477706 10.1002/gene.1063
Blader P Strahle U Ingham P Three Wnt genes expressed in a wide variety of tissues during development of the zebrafish, Danio rerio: developmental and evolutionary perspectives Dev Genes Evol 1996 206 3 13 10.1007/s004270050025
Lekven AC Thorpe CJ Waxman JS Moon RT Zebrafish wnt8 encodes two wnt8 proteins on a bicistronic transcript and is required for mesoderm and neurectoderm patterning Dev Cell 2001 1 103 114 11703928 10.1016/S1534-5807(01)00007-7
Hammerschmidt M Pelegri F Mullins MC Kane DA Brand M van Eeden FJ Furutani-Seiki M Granato M Haffter P Heisenberg CP Jiang YJ Kelsh RN Odenthal J Warga RM Nusslein-Volhard C Mutations affecting morphogenesis during gastrulation and tail formation in the zebrafish, Danio rerio Development 1996 123 143 151 9007236
Golling G Amsterdam A Sun Z Antonelli M Maldonado E Chen W Burgess S Haldi M Artzt K Farrington S Lin SY Nissen RM Hopkins N Insertional mutagenesis in zebrafish rapidly identifies genes essential for early vertebrate development Nat Genet 2002 31 135 140 12006978 10.1038/ng896
Offield MF Jetton TL Labosky PA Ray M Stein RW Magnuson MA Hogan BL Wright CV PDX-1 is required for pancreatic outgrowth and differentiation of the rostral duodenum Development 1996 122 983 995 8631275
He X Saint-Jeannet JP Wang Y Nathans J Dawid I Varmus H A member of the Frizzled protein family mediating axis induction by Wnt-5A Science 1997 275 1652 1654 9054360 10.1126/science.275.5306.1652
Yamaguchi TP Bradley A McMahon AP Jones S A Wnt5a pathway underlies outgrowth of multiple structures in the vertebrate embryo Development 1999 126 1211 1223 10021340
Dahl U Sjodin A Semb H Cadherins regulate aggregation of pancreatic beta-cells in vivo Development 1996 122 2895 2902 8787762
Slusarski DC Yang-Snyder J Busa WB Moon RT Modulation of embryonic intracellular Ca2+ signaling by Wnt-5A Dev Biol 1997 182 114 120 9073455 10.1006/dbio.1996.8463
Westfall TA Brimeyer R Twedt J Gladon J Olberding A Furutani-Seiki M Slusarski DC Wnt-5/pipetail functions in vertebrate axis formation as a negative regulator of Wnt/{beta}-catenin activity J Cell Biol 2003 162 889 898 12952939 10.1083/jcb.200303107
Topol L Jiang X Choi H Garrett-Beal L Carolan PJ Yang Y Wnt-5a inhibits the canonical Wnt pathway by promoting GSK-3-independent beta-catenin degradation J Cell Biol 2003 162 899 908 12952940 10.1083/jcb.200303158
Hebrok M Kim SK St Jacques B McMahon AP Melton DA Regulation of pancreas development by hedgehog signaling Development 2000 127 4905 4913 11044404
Apelqvist A Ahlgren U Edlund H Sonic hedgehog directs specialised mesoderm differentiation in the intestine and pancreas Curr Biol 1997 7 801 804 9368764 10.1016/S0960-9822(06)00340-X
Roy S Qiao T Wolff C Ingham PW Hedgehog signaling pathway is essential for pancreas specification in the zebrafish embryo Curr Biol 2001 11 1358 1363 11553330 10.1016/S0960-9822(01)00402-X
diIorio PJ Moss JB Sbrogna JL Karlstrom RO Moss LG Sonic hedgehog is required early in pancreatic islet development Dev Biol 2002 244 75 84 11900460 10.1006/dbio.2002.0573
Heisenberg CP Brand M Jiang YJ Warga RM Beuchle D van Eeden FJ Furutani-Seiki M Granato M Haffter P Hammerschmidt M Kane DA Kelsh RN Mullins MC Odenthal J Nusslein-Volhard C Genes involved in forebrain development in the zebrafish, Danio rerio Development 1996 123 191 203 9007240
Topczewski J Sepich DS Myers DC Walker C Amores A Lele Z Hammerschmidt M Postlethwait J Solnica-Krezel L The zebrafish glypican knypek controls cell polarity during gastrulation movements of convergent extension Dev Cell 2001 1 251 264 11702784 10.1016/S1534-5807(01)00005-3
Kikuchi Y Trinh LA Reiter JF Alexander J Yelon D Stainier DY The zebrafish bonnie and clyde gene encodes a Mix family homeodomain protein that regulates the generation of endodermal precursors Genes Dev 2000 14 1279 1289 10817762
Ross SE Hemati N Longo KA Bennett CN Lucas PC Erickson RL MacDougald OA Inhibition of adipogenesis by Wnt signaling Science 2000 289 950 953 10937998 10.1126/science.289.5481.950
Fujino T Asaba H Kang MJ Ikeda Y Sone H Takada S Kim DH Ioka RX Ono M Tomoyori H Okubo M Murase T Kamataki A Yamamoto J Magoori K Takahashi S Miyamoto Y Oishi H Nose M Okazaki M Usui S Imaizumi K Yanagisawa M Sakai J Yamamoto TT Low-density lipoprotein receptor-related protein 5 (LRP5) is essential for normal cholesterol metabolism and glucose-induced insulin secretion Proc Natl Acad Sci U S A 2003 100 229 234 12509515 10.1073/pnas.0133792100
Kanazawa A Tsukada S Sekine A Tsunoda T Takahashi A Kashiwagi A Tanaka Y Babazono T Matsuda M Kaku K Iwamoto Y Kawamori R Kikkawa R Nakamura Y Maeda S Association of the Gene Encoding Wingless-Type Mammary Tumor Virus Integration-Site Family Member 5B (WNT5B) with Type 2 Diabetes Am J Hum Genet 2004 75 832 843 15386214 10.1086/425340
Tanaka Y Kato K Notohara K Hojo H Ijiri R Miyake T Nagahara N Sasaki F Kitagawa N Nakatani Y Kobayashi Y Frequent beta-catenin mutation and cytoplasmic/nuclear accumulation in pancreatic solid-pseudopapillary neoplasm Cancer Res 2001 61 8401 8404 11731417
Hyatt TM Ekker SC Vectors and techniques for ectopic gene expression in zebrafish Methods Cell Biol 1999 59 117 126 9891358
Jowett T Analysis of protein and gene expression Methods Cell Biol 1999 59 63 85 9891356
Hauptmann G Gerster T Multicolor whole-mount in situ hybridization Methods Mol Biol 2000 137 139 148 10948532
Milewski WM Duguay SJ Chan SJ Steiner DF Conservation of PDX-1 structure, function, and expression in zebrafish Endocrinology 1998 139 1440 1449 9492081 10.1210/en.139.3.1440
Korzh V Edlund T Thor S Zebrafish primary neurons initiate expression of the LIM homeodomain protein Isl-1 at the end of gastrulation Development 1993 118 417 425 8223269
Higashijima S Nose A Eguchi G Hotta Y Okamoto H Mindin/F-spondin family: novel ECM proteins expressed in the zebrafish embryonic axis Dev Biol 1997 192 211 227 9441663 10.1006/dbio.1997.8760
Alexander J Stainier DY A molecular pathway leading to endoderm formation in zebrafish Curr Biol 1999 9 1147 1157 10531029 10.1016/S0960-9822(99)80266-8
Odenthal J Nusslein-Volhard C fork head domain genes in zebrafish Dev Genes Evol 1998 208 245 258 9683740 10.1007/s004270050179
Yee NS Yusuff S Pack M Zebrafish pdx1 morphant displays defects in pancreas development and digestive organ chirality, and potentially identifies a multipotent pancreas progenitor cell Genesis 2001 30 137 140 11477692 10.1002/gene.1049
Korzh S Emelyanov A Korzh V Developmental analysis of ceruloplasmin gene and liver formation in zebrafish Mech Dev 2001 103 137 139 11335121 10.1016/S0925-4773(01)00330-6
Hogan BL Developmental signalling. Sorting out the signals Curr Biol 1994 4 1122 1124 7704577 10.1016/S0960-9822(00)00252-9
Freemark M Driscoll P Maaskant R Petryk A Kelly PA Ontogenesis of prolactin receptors in the human fetus in early gestation. Implications for tissue differentiation and development J Clin Invest 1997 99 1107 1117 9062370
|
16246260
|
PMC1276788
|
CC BY
|
2021-01-04 16:39:16
|
no
|
BMC Biol. 2005 Oct 24; 3:23
|
utf-8
|
BMC Biol
| 2,005 |
10.1186/1741-7007-3-23
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1341622974610.1186/1471-2407-5-134Research ArticleTP53 mutations in ovarian carcinomas from sporadic cases and carriers of two distinct BRCA1 founder mutations; relation to age at diagnosis and survival Kringen Pedro [email protected] Yun [email protected] Vanessa [email protected] Jahn M [email protected] Gunnar [email protected] Anne-Lise [email protected] Anne [email protected] Department of Genetics, The Norwegian Radium Hospital, Montebello, 0310 Oslo, Norway2 Department of Gynecologic Oncology, The Norwegian Radium Hospital, Montebello, 0310 Oslo, Norway3 University of Oslo, The Norwegian Radium Hospital, Montebello, 0310 Oslo, Norway4 Department of Pathology, The Norwegian Radium Hospital, Montebello, 0310 Oslo, Norway5 Institute of Community Medicine, University of Tromsø2005 17 10 2005 5 134 134 14 1 2005 17 10 2005 Copyright © 2005 Kringen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Ovarian carcinomas from 30 BRCA1 germ-line carriers of two distinct high penetrant founder mutations, 20 carrying the 1675delA and 10 the 1135insA, and 100 sporadic cases were characterized for somatic mutations in the TP53 gene. We analyzed differences in relation to BRCA1 germline status, TP53 status, survival and age at diagnosis, as previous studies have not been conclusive.
Methods
DNA was extracted from paraffin embedded formalin fixed tissues for the familial cases, and from fresh frozen specimen from the sporadic cases. All cases were treated at our hospital according to protocol. Mutation analyses of exon 2 – 11 were performed using TTGE, followed by sequencing.
Results
Survival rates for BRCA1-familial cases with TP53 mutations were not significantly lower than for familial cases without TP53 mutations (p = 0.25, RR = 1.64, 95% CI [0.71–3.78]). Median age at diagnosis for sporadic (59 years) and familial (49 years) cases differed significantly (p < 0.001) with or without TP53 mutations. Age at diagnosis between the two types of familial carriers were not significantly different, with median age of 47 for 1675delA and 52.5 for 1135insA carriers (p = 0.245). For cases ≥50 years at diagnosis, a trend toward longer survival for sporadic over familial cases was observed (p = 0.08). The opposite trend was observed for cases <50 years at diagnosis.
Conclusion
There do not seem to be a protective advantage for familial BRCA1 carriers without TP53 mutations over familial cases with TP53 mutations. However, there seem to be a trend towards initial advantage in survival for familial cases compared to sporadic cases diagnosed before the age of 50 both with and without TP53 mutations. However, this trend diminishes over time and for cases diagnosed ≥50 years the sporadic cases show a trend towards an advantage in survival over familial cases. Although this data set is small, if confirmed, this may be a link in the evidence that the differences in ovarian cancer survival reported, are not due to the type of BRCA1 mutation, but may be secondary to genetic factors shared. This may have clinical implications for follow-up such as prophylactic surgery within carriers of the two most frequent Norwegian BRCA1 founder mutations.
==== Body
Background
Ovarian cancer is one of the leading causes of cancer-related death in women today. It is the 4th most common cancer in women in Norway and accounts for 5 – 6% of all cancers [1,2]. Mean age at diagnosis for sporadic cases have been reported to 62.3 years [3], and in Norway to 65 years. Age-standardized incidence rates were 13.5 pr 100.000, and close to 40% of the patients is achieving 5-year survival according to The Norwegian Cancer Registry (OVANOR 1991 – 1996).
Almost 10% of epithelial ovarian cancer cases are associated with dominant genetic predisposition, in most cases (80 – 90%), linked to mutations in BRCA1 or BRCA2 [4-6]. Mean age at diagnosis for these inherited cases have been reported to be from 49 to 54.3 years [3,7]. The penetrance of the disease in mutation carriers varies, and has been reported to be from 27 – 80% [8-10]. It should be noted that both the incidence rate for hereditary cases and the penetrance of the disease may differ depending on geographic and ethnic origin [11]. The survival rate may also vary depending on type and localization of the mutation. Some studies have reported that ovarian cancer patients carrying germ-line BRCA1 mutations have an enhanced survival rate compared to sporadic cases [3,12-14]. Other studies demonstrated only an initial survival advantage that disappeared with time, and concluded that no enhanced survival rates follows BRCA1 dysfunction [15-17]. These studies predict a survival rate for BRCA1 familial ovarian cancer that is equal to or higher than non-familial cases.
Both the penetrance estimates and the survival rates are based on studies in populations with strong founder effects, and may therefore be biased. The type of mutation in the BRCA1 gene may affect the timing of the diagnosis of the disease, the response to environmental exposure causing DNA damage, the efficiency of DNA repair, and the frequency of somatic mutations developing in the tumor. These factors may in turn affect the survival rate.
Mutations in the TP53 tumor suppressor gene are the most common genetic alteration in human tumors and have been suggested as a molecular marker for prognosis. TP53 encodes a nuclear phosphoprotein located at chromosome region 17p13 involved in cell cycle arrest and DNA repair and somatic TP53 mutations are known to associate with familial ovarian cancer. In ovarian tumors from BRCA1 mutation carriers, somatic TP53 mutations are found in 60 – 80% of the cases [18-22]. Thirty to 50% of all ovarian cancers have been reported to harbor a TP53 mutation [18-20,23,24]. Further, in 30 – 85% of the sporadic ovarian carcinomas both a TP53 mutations and a somatic BRCA1/BRCA2 mutation have been found [18,20,25].
These findings implicate that TP53 and BRCA1 directly interacts and may play an important role in DNA repair processes and tumor suppression [26,27]. However, despite the high frequency of mutations in the tumor suppressor gene TP53, there are several reports concluding that TP53 is not a good predictor of prognosis in sporadic ovarian cancer patients [24,28,29].
We have previously reported two Norwegian BRCA1 founder mutations; 1135insA [30] and 1675delA [11]. Carriers of these mutations show almost the same penetrance for ovarian- and breast cancer and the penetrance is also high compared to most reported BRCA1 mutation carriers. By age 50, 48% of mutation carriers had experienced breast- and/or ovarian cancer. Mean age of ovarian cancer diagnosis was ~55 years [10]. Three per cent of all Norwegian ovarian cancers are caused by either of the two founder mutations [31]. As a result of a clinical follow-up program for early diagnosis in women from breast-ovarian cancer kindreds, these two mutations may account for more than half of those with a BRCA1 mutation in Norway. The histopathological characteristics of both breast and ovarian cancer indicated an unfavorable prognosis in these mutation carriers [32].
In the present study, we have screened epithelial ovarian tumors from 30 familial cases and 100 sporadic cases for somatic mutations in the TP53 gene. The cancer treatment was according to our hospital protocol. The familial cases consisted of one group with the BRCA1 1135insA mutation and the other had the BRCA1 1675delA mutation [10]. The TP53 mutation status was correlated to survival, age at diagnosis and histopathological features.
Materials
Formalin fixed and paraffin embedded ovarian cancer tissue from 30 BRCA1 germ line mutation carriers were collected and used for DNA extraction. Of the familial cases 20 patients carried the 1675delA mutation and another 10 patients the 1135insA mutation, which is a representative distribution between the two mutations in the Norwegian population. The BRCA1 carriers were from families with at least two first-degree relatives, or second-degree relatives through male, with ovarian cancer and/or breast cancer under age 60. All cases were sampled from pedigree regardless of survival status, as ovarian cancer treatment is centralized to our hospital. Analysis of fresh frozen specimen of tumor DNA from the 100 sporadic cases sampled from 1992–2003, included in this study has previously been reported [29]. Both groups were diagnosed and treated at the Norwegian Radium Hospital according to protocol. The patient characteristics are shown in Table 1. All tumors were reviewed at our department of pathology, the familial tumors by our team pathologist, and were classified and graded according to the World Health Organization (WHO) criteria. Follow-up time for each case was calculated from the date of diagnosis up to date of death or end of study (15th April, 2004).
Methods
DNA extraction and TP53 mutation analysis
DNA was manually extracted from paraffin-embedded tissue sections of tumor material using 5 sections of 10 μ. A modification of the procedure described by Miller [33] was used. The modification included using as much as possible of the top water layer of the 700 ml DNA/lysis buffer and 1 ml phenol/chloroform/water mix, and repeating the extraction step once. The protocol was optimized to give high yield of good quality DNA.
Mutation analyses of exons 2–11 of the TP53 gene in the 30 cases with BRCA1 germ line mutations were performed by TTGE followed by sequencing. Primers, PCR conditions and gel running conditions were as described elsewhere [34]. Samples with aberrantly migrating bands on TTGE were isolated, submitted to a new PCR and sequenced. Analysis of the fresh frozen specimen of tumor DNA from the 100 sporadic cases has previously been reported [29].
Statistical analyses
In univariate analyses, a log rank test have been used to investigate the effect of age at diagnosis, BRCA1 and TP53 mutations on the survival rate. In multivariate analyses, Cox proportional hazards regression analysis was used. Hazard ratios (HR's) are given with 95% confidence intervals (CI's). Statistical significance rates were set at 0.05. The software SAS® version 8.2 was used for statistical analyses.
Results
TP53 characterizations and novel mutations
Nineteen of the 30 ovarian carcinomas showed one or more aberrant migrating bands on TTGE in one or more exons and was sequenced (Table 2). A total of 21 sequence changes were detected. Two cases had two different TP53 sequence changes in their tumors, one being a silent mutation. Nine mutations were missense mutations, four nonsense, three were silent sequence changes (not previously reported as polymorphisms) and two were intronic sequence changes of unknown function.
The frequency of transitions vs. transversion in this hereditary cohort (85.7% and 14.3%) was also quite similar to that reported in the IARC database for sporadic cases (88% and 12%), but differed slightly from the sporadic cases in this study (76.4% and 23.6%). The frequency of mutations likely to cause protein alteration were 68.0% (68/100) for the sporadic cases and 53.3%(16/30) for the familial cases. The TP53 mutation frequency in the two different BRCA1 carriers differed slightly with 11/20 (55.0%) in the BRCA1 1675delA carriers and 5/10 (50.0%) in the BRCA1 1135insA carriers. The 1675delA carriers had 7.7% transversions and 92.3% transitions while the 1135insA carriers had 12.5% transversion and 82.5% transitions. Four of the TP53 mutations were novel and not previously reported in ovarian cancer in the IARC TP53 Database [35] or the SOUSSI database. These mutations affected codon 205 (tyr>ser), 260 (ser>ser), 267 (arg>gln) and 293 (gly>arg). All mutations detected resided in exons 5–8. When comparing the TP53 mutation spectrum in these familial cases with that of ovarian cancers cases reported in the IARC database and to the 100 sporadic ovarian cancer cases with a TP53 mutation, no obvious differences were seen either with respect to exon distribution or codon wise (data not shown), although a slightly lower frequency of mutations in exon 5 and a slightly higher in exon 8 were seen in the hereditary cases. The TP53 mutations in the 100 sporadic cases used in this study is reported elsewhere [29].
Age at diagnosis, survival, BRCA1 and TP53 status
Median age at diagnosis among sporadic cases and familial cases that carried 1675delA or 1135insA mutations is presented in Table 1. As expected, the familial cases are diagnosed earlier in life than sporadic cases (p < 0.001). The difference in median age of onset between the 1135insA and 1675delA mutation carriers was not significant (p = 0.245).
In the univariate analysis of the combined group, neither BRCA1 status nor age at diagnosis was significantly associated to survival (p = 0.87 and p = 0.50 for BRCA1 status and age at diagnosis (categorized into < 50 and ≥ 50 years), respectively). TP53 mutation did not significantly reduce the survival rates (p = 0.35). Notably, interaction between BRCA1 status and TP53 status was borderline significant (test for interaction: p = 0.06) while the one between BRCA1 status and age at diagnosis was statistically significant (test for interaction: p = 0.05). We further analyzed these factors adjusted for tumor grade, however, results did not substantially change (test interaction: p = 0.04 and p = 0.05 for BRCA1*TP53 and BRCA1*age at diagnosis, respectively).
No association between age at diagnosis and survival time was found among sporadic cases (p = 0.88). Familial cases with late age at diagnosis (≥50 years) had a slightly higher risk of dying than the cases with an early age at diagnosis, however the association did not reach significance, possibly due to a lack of statistical power (RR = 1.65, 95% CI [0.79–3.43], p = 0.14). Among cases diagnosed at age 50 years or more, familial cases had a trend towards a higher risk of dying than sporadic cases (RR = 1.75, 95% CI [0.93–3.30], p = 0.08). After adjustment for the effect of tumor grade and TP53 status (RR = 1.80, 95% CI [0.94–3.43], p = 0.08) (data not shown). Table 3 shows the risk ratios associated to TP53 mutations after stratification for BRCA1 status.
There was no significant difference in survival observed among TP53 mutations carriers compared to non-TP53 mutations carriers, neither for the familial nor the sporadic cases (Log-rank test for TP53 in familial cases: p = 0.25 and log-rank test for TP53 in sporadic cases: p = 0.88) (Table 3).
Discussion
Some studies have reported an enhanced survival for BRCA1 carriers with ovarian cancer compared to sporadic cases [12-14,36,37], but these studies have not taken TP53 status in the tumors in to consideration. Other studies in which TP53 status have been included concludes that there is no difference in survival [16].
Our results do not show an enhanced survival rate for familial cases compared to sporadic cases, even after adjustment for TP53 status when all age groups were included. Further, no significant difference in survival rates was observed between the familial cases with and without TP53 mutations (Table 3).
These results do not support earlier observations regarding the importance of the p53/BRCA1 interaction on cell proliferation and ovarian carcinogenesis. Most penetrance estimates and survival rates are based on studies in populations with strong founder effects, and may therefore be biased [15-17,38]. Two Ashkenazi founder mutations occur in BRCA1 185delAG and 5382insC (carrier frequencies of 0.9% and 0.13%), with mean age at diagnosis 54 years. How the type of mutation in the BRCA1 gene affects survival, age at diagnosis of the disease, the response to environmental exposure causing DNA damage, and the efficiency of DNA repair, is not clarified. Heterozygote advantage or an increase in biological fitness conferred on carriers of a disease causing mutation (like BRCA1?), often a resistance to certain infections that were common in times past, can cause an increase in allele frequency [39]. Genetic factors with impact on survival and age at onset of disease, to after childbearing age, would be preferential.
The trend for an increased survival in favour of the early age at onset in familial cases compared to late age at onset in familial cases may be attributed to younger patients having greater physical strength, less somatic mutations, and manage illness better than older patients. On the other hand, one might also expect this trend in sporadic cases, which was not the case. One limitation of our study is the small numbers of BRCA1 carriers. In our study, the statistical power to detect BRCA1 effect was 76%.
Consequently, our findings should be confirmed in larger studies. The conflicting literature on the impact of BRCA1 mutation status on ovarian cancer survival should promote additional studies from different ethnic populations, and thereby allow investigators to study whether or not there is a survival benefit due to BRCA1 mutation, or may be secondary to other common inherited genetic factors, which may be shared in ethnic or geographic isolated populations.
Alterations in the TP53 gene have been shown to affect breast cancer survival and in particular patients with mutations in the zinc-binding domains have poor survival [40]. In sporadic ovarian carcinoma several studies reports that no or little effect of TP53 mutations have been seen [17,24,28], which is similar to the results reported here. TP53 alterations are also suggested to alter ovarian cancer survival in BRCA1 germ line patients [13,14], while other groups concludes with a failure of BRCA1 dysfunction to alter ovarian cancer survival [16]. It should also be noted that a considerable fraction (60–80%) of all familial BRCA1 ovarian cancers harbor TP53 mutations [18,19,21,22]. Only a few studies have reported analysis of TP53 mutations in relation to BRCA1 associated ovarian cancer [20,41]. The present study is the first investigating somatic TP53 mutations in ovarian tumors from carries of two distinct high penetrant BRCA1 germ-line mutations, relating it to survival and age at diagnosis of disease and compares it to sporadic cases. We have previously studied the distribution in age at diagnosis in BRCA1 carriers and non-carriers as a part of a cohort study. Three percent of Norwegian ovarian cancers are caused by BRCA1 1675delA or 1135insA [31,42], with a distribution similar to that found in this study (Table 1). Further, Bjørge et al. [43] found that 87.0% of Norwegian sporadic ovarian cancers was papillary serous adenocarcinoma, an aggressive histo-prognostic factor.
Eigthy percent of both familial and sporadic ovarian cancer cases in this study were papillary serous adenocarcinoma. Questions need to be addressed concerning the clinical effects of mutations in the BRCA1 gene, why some mutation carriers develop breast cancer, others develop ovarian cancer, and some develop both. We do not know whether the cancers occurring in mutation carriers are significantly different from those occurring in non carriers. The frequency of TP53 mutations in the familial cases altering the protein was 53.3%, which is somewhat higher than other studies of familial BRCA1 ovarian cancer (31–50%) [18-22,24,25]. Although the number of familial cases in this study is limited, a slightly higher frequency of mutations was found in exon 8 and a lower frequency in exon 5 compared to sporadic cases in the IARC database. The same tendency has been reported by others [20]. However, a non-significant difference in TP53 mutation frequency was observed between the familial and sporadic cases in this study.
Of the novel mutations found in the familial cohort the codon 205 mutation has previously been reported in several other tumors like head and neck SCC as well as breast- and colorectal carcinoma. The amino acid change in codon 255 and 293 are only reported once, in oesophageal SCC and bladder cancer, respectively. The silent codon 260 mutation are reported in two different cancer tissues; lung (SCLC) and colorectal carcinoma. Environmental exposure, both external and internal, is known to influence the spectrum of mutations. Whether hormonal disturbance may affect the mutation rate and spectrum is not known, but if so, it may be expected that BRCA1 carriers are more sensitive to such exposure.
Conclusion
Interestingly, no difference in survival was observed between TP53 mutation carriers among the familial carries or among the sporadic cases (Table 3). Further, we did not find an overall difference in survival between familial BRCA1 carriers and sporadic epithelial ovarian cancer cases, even after adjustment for TP53 status. For cases diagnosed over the age 50 there was a trend toward higher survival for sporadic cases.
List of abbreviations
TTGE; temporal temperature gradient gel electrophoresis; PCR, polymerase chain reaction; FIGO, International Federation of Gynaecology and Obstetrics. RR; risk ratio.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PK participated in the design of the study, carried out the molecular genetic studies, sequence alignment and drafted the manuscript. YW participated in screening of the sporadic cases and sequence alignment. VD performed the statistical analyses. GK performed clinical updates of sporadic cases.
JMN evaluated pathology sections of sporadic and familial cases. ALBD conceived the study, participated in its design and helped draft the manuscript. AD conceived the study, performed clinical dates of the familial cases and helped draft the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Pedro Kringen is a research fellow of The Norwegian Cancer Society. Guro Elisabeth Lind is acknowledged for working out the optimal condition for DNA extraction from paraffin embedded tissues. We thank Sigrid Lystad and Phuong Vu for providing excellent technical assistance.
This work was supported by grants from The Norwegian Cancer Society and The Norwegian Research Council.
Figures and Tables
Table 1 Patient characteristics
Sporadic cases BRCA1 carriers
All familial cases 1135insA 1675delA
No of cases 100 30 10 20
Age at diagnosis: Median 59 49 52.5 47
range 39 – 80 39 – 80 41 – 80 39 – 65
FIGO stage
I – II 9 (9.0%) 8 (26.7%) 2 (20.0%) 6 (30.0%)
III 65 (65.0%) 14 (46.7%) 5 (50.0%) 9 (45.0%)
IV 26 (26.0%) 8 (26.7%) 3 (30.0%) 5 (25.0%)
Histology
Serous 82 (82.0%) 24 (80.0%) 7 (70.0%) 17 (85.0%)
Mixed 7 (7.0%) 2 (6.6%) 1 (10.0%) 1 (5.0%)
Endometroid 5 (5.0%) 3 (10.0%) 1 (10.0%) 2 (10.0%)
Unclassified 6 (6.0%) 1 (3.3%) 1 (10.0%) 0
Grade of differentiation
1 7 (7.0%) 0 0 0
2 26 (26.0%) 5 (16.6%) 1 (10.0%) 4 (20.0%)
3 67 (67.0%) 24 (80.0%) 9 (90.0%) 15 (75.0%)
Unknown 0 1 (3.3%) 0 1 (5.0%)
Survival >5 years 23 (23.0%) 10 (33.3%) 3 (30.0%) 7 (35.0%)
TP53 mutation status 72.0% 53.3% 50.0% 55.0%
All tumors are epithelial adenocarcinomas
Table 2 TP53 mutations, survival and histopathological features for each case.
Case aBRCA1 carrier type Stage bGrade Type TTGE Exon Codon Mutation aa change Age of diagnosis cSurvival months dVital status
3453 1 3 2 ser pos 5 559+1G>A splice 53 83 1
10 1 3 3 ser pos 6 213 CGA>TGA arg>stop 65 22 3
2857 1 3 3 ser pos 6 213 CGA>TGA arg>stop 53 83 3
4 1 4 3 ser pos 6 216 GTG>ATG val>met 59 11 3
27 1 2 3 endo pos 7 237 ATG>ATA met>ile 39 61 3
8e 1 3 3 ser pos 7 260 TCC>TCT ser>ser 49 84 3
8 306 CGA>TGA arg>stop
14 1 1 3 ser pos 8 267 CGG>CAG arg>gln 51 25 3
13 1 4 3 ser pos 8 273 CGT>AGT arg>ser 39 15 3
29 1 3 3 mix pos 8 273 CGT>TGT arg>cys 50 36 3
2842 1 1 4 3 ser pos 8 273 CGT>TGT arg>cys 39 35 3
21 1 3 3 ser pos 8 293 GGG>AGG gly>arg 39 21 3
26 1 3 2 ser pos intron G>A ivs5 53 93 1
9 1 4 2 ser neg 50 39 3
11 1 3 3 ser neg 44 8 3
17 1 3 3 ser neg 47 30 3
24 1 2 4 ser neg 59 20 3
34 1 1 3 ser neg 46 120 1
30 1 1 2 ser neg 43 199 2
32 1 2 3 ser neg 48 108 1
3351 1 4 3 endo neg 44 9 3
1 2 4 3 ser pos 5 144 CAG>TAG gln>stop 41 1 3
3 2 3 3 uncl pos 6 196 CGA>TGA arg>stop 48 46 3
20 2 3 3 ser pos 6 205 TAT>TCT tyr>ser 52 36 3
28 2 4 3 ser pos 7 261 AGG>AGA arg>arg 80 12 3
7e 2 3 3 ser pos 7 255 ATC>GTC ile>val 58 30 3
5 141 TGC>TGT cys>cys
12 2 1 3 ser pos 8 280 AGA>GGA arg>gly 49 134 1
15 2 3 3 ser pos intron C>T ivs7 50 19 3
5 2 4 3 ser neg 60 45 3
18 2 2 3 endo neg 44 117 3
22 2 3 2 mix neg 47 96 3
a: 1 = 1675delA and 2 = 1135insA. b: undifferentiated (4). c: all patients were followed until diseased or to 15th April, 2004. d: alive without cancer (1), alive with cancer (2), and dead by cancer (3). e: two different mutations detected in sample.
Table 3 Hazard risk of TP53 mutations on familial and sporadic ovarian cancer cases.
# cases HR CI (95%) p-value
Familial Cases 30
TP53 no mutation 14 1 (ref)
TP53 mutation 16 1.64 0.71 – 3.78 0.25
Sporadic cases 100
TP53 no mutation 26 1 (ref)
TP53 mutation 76 0.96 0.55 – 1.66 0.88
==== Refs
The Norwegian Cancer Registry Cancer in Norway 1998. The Cancer Registry of Norway
Engeland A Haldorsen T Tretli S Hakulinen T Horte LG Luostarinen T Magnus K Schou G Sigvaldason H Storm HH Prediction of cancer incidence in the Nordic countries up to the years 2000 and 2010. A collaborative study of the five Nordic Cancer Registries APMIS Suppl 1993 38 1 124 8297639
Zweemer RP Verheijen RH Menko FH Gille JJ van Diest PJ Coebergh JW Shaw PA Jacobs IJ Kenemans P Differences between hereditary and sporadic ovarian cancer Eur J Obstet Gynecol Reprod Biol 1999 82 151 153 10206407 10.1016/S0301-2115(98)00218-8
Miki Y Swensen J Shattuck-Eidens D Futreal PA Harshman K Tavtigian S Liu Q Cochran C Bennett LM Ding W A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1 Science 1994 266 66 71 7545954
Wooster R Bignell G Lancaster J Swift S Seal S Mangion J Collins N Gregory S Gumbs C Micklem G Identification of the breast cancer susceptibility gene BRCA2 Nature 1995 378 789 792 8524414 10.1038/378789a0
Boyd J Molecular genetics of hereditary ovarian cancer Oncology (Huntingt) 1998 12 399 406; discussion 409-10, 413
Lynch HT Watson P Lynch JF Conway TA Fili M Hereditary ovarian cancer. Heterogeneity in age at onset Cancer 1993 71 573 581 8420679
Ford D Easton DF Peto J Estimates of the gene frequency of BRCA1 and its contribution to breast and ovarian cancer incidence Am J Hum Genet 1995 57 1457 1462 8533776
Ford D Easton DF Stratton M Narod S Goldgar D Devilee P Bishop DT Weber B Lenoir G Chang-Claude J Sobol H Teare MD Struewing J Arason A Scherneck S Peto J Rebbeck TR Tonin P Neuhausen S Barkardottir R Eyfjord J Lynch H Ponder BA Gayther SA Zelada-Hedman M Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. The Breast Cancer Linkage Consortium Am J Hum Genet 1998 62 676 689 9497246 10.1086/301749
Dorum A Heimdal K Hovig E Inganas M Moller P Penetrances of BRCA1 1675delA and 1135insA with respect to breast cancer and ovarian cancer Am J Hum Genet 1999 65 671 679 10441573 10.1086/302530
Dorum A Moller P Kamsteeg EJ Scheffer H Burton M Heimdal KR Maehle LO Hovig E Trope CG van der Hout AH van der Meulen MA Buys CH te Meerman GJ A BRCA1 founder mutation, identified with haplotype analysis, allowing genotype/phenotype determination and predictive testing Eur J Cancer 1997 33 2390 2392 9616287 10.1016/S0959-8049(97)00328-6
Rubin SC Benjamin I Behbakht K Takahashi H Morgan MA LiVolsi VA Berchuck A Muto MG Garber JE Weber BL Lynch HT Boyd J Clinical and pathological features of ovarian cancer in women with germ- line mutations of BRCA1 N Engl J Med 1996 335 1413 1416 8875917 10.1056/NEJM199611073351901
Ben David Y Chetrit A Hirsh-Yechezkel G Friedman E Beck BD Beller U Ben-Baruch G Fishman A Levavi H Lubin F Menczer J Piura B Struewing JP Modan B Effect of BRCA mutations on the length of survival in epithelial ovarian tumors J Clin Oncol 2002 20 463 466 11786575 10.1200/JCO.20.2.463
Cass I Baldwin RL Varkey T Moslehi R Narod SA Karlan BY Improved survival in women with BRCA-associated ovarian carcinoma Cancer 2003 97 2187 2195 12712470 10.1002/cncr.11310
Johannsson OT Ranstam J Borg A Olsson H Survival of BRCA1 breast and ovarian cancer patients: a population-based study from southern Sweden J Clin Oncol 1998 16 397 404 9469321
Buller RE Shahin MS Geisler JP Zogg M De Young BR Davis CS Failure of BRCA1 dysfunction to alter ovarian cancer survival Clin Cancer Res 2002 8 1196 1202 12006538
Smith-Sorensen B Kaern J Holm R Dorum A Trope C Borresen-Dale AL Therapy effect of either paclitaxel or cyclophosphamide combination treatment in patients with epithelial ovarian cancer and relation to TP53 gene status Br J Cancer 1998 78 375 381 9703286
Ramus SJ Bobrow LG Pharoah PD Finnigan DS Fishman A Altaras M Harrington PA Gayther SA Ponder BA Friedman LS Increased frequency of TP53 mutations in BRCA1 and BRCA2 ovarian tumours Genes Chromosomes Cancer 1999 25 91 96 10337991 10.1002/(SICI)1098-2264(199906)25:2<91::AID-GCC3>3.0.CO;2-5
Tworek H Peng R Fetzer S Werness BA Piver MS Allen HJ DiCioccio RA Mutation analysis of BRCA1, TP53, and KRAS2 in ovarian and related pelvic tumors Cancer Genet Cytogenet 1999 112 105 118 10686936 10.1016/S0165-4608(98)00267-2
Buller RE Lallas TA Shahin MS Sood AK Hatterman-Zogg M Anderson B Sorosky JI Kirby PA The p53 mutational spectrum associated with BRCA1 mutant ovarian cancer Clin Cancer Res 2001 7 831 838 11309330
Rhei E Bogomolniy F Federici MG Maresco DL Offit K Robson ME Saigo PE Boyd J Molecular genetic characterization of BRCA1- and BRCA2-linked hereditary ovarian cancers Cancer Res 1998 58 3193 3196 9699640
Schorge JO Muto MG Lee SJ Huang LW Welch WR Bell DA Keung EZ Berkowitz RS Mok SC BRCA1-related papillary serous carcinoma of the peritoneum has a unique molecular pathogenesis Cancer Res 2000 60 1361 1364 10728699
Wertheim I Muto MG Welch WR Bell DA Berkowitz RS Mok SC p53 gene mutation in human borderline epithelial ovarian tumors J Natl Cancer Inst 1994 86 1549 1551 7932812
Fallows S Price J Atkinson RJ Johnston PG Hickey I Russell SE P53 mutation does not affect prognosis in ovarian epithelial malignancies J Pathol 2001 194 68 75 11329143 10.1002/path.857
Ozalp S Yalcin OT Minsin TH Expression of p53 in epithelial ovarian cancer Int J Gynaecol Obstet 2000 71 277 278 11102625 10.1016/S0020-7292(00)00309-X
Ludwig T Chapman DL Papaioannou VE Efstratiadis A Targeted mutations of breast cancer susceptibility gene homologs in mice: lethal phenotypes of Brca1, Brca2, Brca1/Brca2, Brca1/p53, and Brca2/p53 nullizygous embryos Genes Dev 1997 11 1226 1241 9171368
Sourvinos G Spandidos DA Decreased BRCA1 expression levels may arrest the cell cycle through activation of p53 checkpoint in human sporadic breast tumors Biochem Biophys Res Commun 1998 245 75 80 9535786 10.1006/bbrc.1998.8379
Wang Y Helland A Holm R Skomedal H Abeler VM Danielsen HE Trope CG Borresen-Dale AL Kristensen GB TP53 mutations in early-stage ovarian carcinoma, relation to long-term survival Br J Cancer 2004 90 678 685 14760384 10.1038/sj.bjc.6601537
Wang Y Kringen P Kristensen GB Holm R Baekelandt MM Olivier M Skomedal H Hainaut P Trope CG Abeler VM Nesland JM Borresen-Dale AL Helland A Effect of the codon 72 polymorphism (c.215G>C, p.Arg72Pro) in combination with somatic sequence variants in the TP53 gene on survival in patients with advanced ovarian carcinoma Hum Mutat 2004 24 21 34 15221786 10.1002/humu.20055
Andersen TI Borresen AL Moller P A common BRCA1 mutation in Norwegian breast and ovarian cancer families? Am J Hum Genet 1996 59 486 487 8755943
Dorum A Hovig E Trope C Inganas M Moller P Three per cent of Norwegian ovarian cancers are caused by BRCA1 1675delA or 1135insA Eur J Cancer 1999 35 779 781 10505039 10.1016/S0959-8049(99)00050-7
Borg A Dorum A Heimdal K Maehle L Hovig E Moller P BRCA1 1675delA and 1135insA account for one third of Norwegian familial breast-ovarian cancer and are associated with later disease onset than less frequent mutations Dis Markers 1999 15 79 84 10595257
Miller SA Dykes DD Polesky HF A simple salting out procedure for extracting DNA from human nucleated cells Nucleic Acids Res 1988 16 1215 3344216
Sorlie T Johnsen H Vu P Lind GE Lothe R Borresen-Dale AL Mutation Screening of the TP53 Gene by Temporal Temperature Gradient Gel Electrophoresis Methods Mol Biol 2004 291 207 216
Olivier M Eeles R Hollstein M Khan MA Harris CC Hainaut P The IARC TP53 database: new online mutation analysis and recommendations to users Hum Mutat 2002 19 607 614 12007217 10.1002/humu.10081
Zweemer RP Shaw PA Verheijen RM Ryan A Berchuck A Ponder BA Risch H McLaughlin JR Narod SA Menko FH Kenemans P Jacobs IJ Accumulation of p53 protein is frequent in ovarian cancers associated with BRCA1 and BRCA2 germline mutations J Clin Pathol 1999 52 372 375 10560359
Zweemer RP Verheijen RH Coebergh JW Jacobs IJ van Diest PJ Gille JJ Skates S Menko FH Ten Kate LP Kenemans P Survival analysis in familial ovarian cancer, a case control study Eur J Obstet Gynecol Reprod Biol 2001 98 219 223 11574135 10.1016/S0301-2115(01)00318-9
Goldgar DE Reilly PR A common BRCA1 mutation in the Ashkenazim Nat Genet 1995 11 113 114 7550331 10.1038/ng1095-113
Offman J Opelz G Doehler B Cummins D Halil O Banner NR Burke MM Sullivan D Macpherson P Karran P Defective DNA Mismatch Repair in Acute Myeloid Leukemia/Myelodysplastic Syndrome After Organ Transplantation Blood 2004 15090454
Borresen AL Andersen TI Eyfjord JE Cornelis RS Thorlacius S Borg A Johansson U Theillet C Scherneck S Hartman S TP53 mutations and breast cancer prognosis: particularly poor survival rates for cases with mutations in the zinc-binding domains Genes Chromosomes Cancer 1995 14 71 75 8527388
Buller RE Sood AK Lallas T Buekers T Skilling JS Association between nonrandom X-chromosome inactivation and BRCA1 mutation in germline DNA of patients with ovarian cancer [see comments] J Natl Cancer Inst 1999 91 339 346 10050867 10.1093/jnci/91.4.339
Bjorge T Engeland A Sundfor K CG. T Prognosis of 2,800 patients with epithelial ovarian cancer diagnosed during 1975-94 and treated at the Norwegian Radium Hospital. Acta Obstet Gynecol Scand 1998 77 777 781 9740528 10.1034/j.1600-0412.1998.770714.x
Bjorge T Engeland A Hansen S Trope CG Prognosis of patients with ovarian cancer and borderline tumours diagnosed in Norway between 1954 and 1993 Int J Cancer 1998 75 663 670 9495231 10.1002/(SICI)1097-0215(19980302)75:5<663::AID-IJC1>3.0.CO;2-X
|
16229746
|
PMC1276789
|
CC BY
|
2021-01-04 16:03:04
|
no
|
BMC Cancer. 2005 Oct 17; 5:134
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-134
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1351623617610.1186/1471-2407-5-135Research ArticleDifferential biologic effects of CPD and 6-4PP UV-induced DNA damage on the induction of apoptosis and cell-cycle arrest Lo Hsin-Lung [email protected] Satoshi [email protected] Lisa [email protected] Barbara [email protected] Akira [email protected] Douglas W [email protected] Laurie B [email protected] Division of Biomedical Sciences, University of California Riverside, Riverside, CA 92521 USA2 Department of Molecular Genetics, Institute of Development, Aging and Cancer, Tohoku University, Sendai, 980-8575 Japan3 BioLegend, Inc., 8395 Camino Santa Fe, Suite E, San Diego, CA 92121 USA2005 19 10 2005 5 135 135 17 5 2005 19 10 2005 Copyright © 2005 Lo et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
UV-induced damage can induce apoptosis or trigger DNA repair mechanisms. Minor DNA damage is thought to halt the cell cycle to allow effective repair, while more severe damage can induce an apoptotic program. Of the two major types of UV-induced DNA lesions, it has been reported that repair of CPD, but not 6-4PP, abrogates mutation. To address whether the two major forms of UV-induced DNA damage, can induce differential biological effects, NER-deficient cells containing either CPD photolyase or 6-4 PP photolyase were exposed to UV and examined for alterations in cell cycle and apoptosis. In addition, pTpT, a molecular mimic of CPD was tested in vitro and in vivo for the ability to induce cell death and cell cycle alterations.
Methods
NER-deficient XPA cells were stably transfected with CPD-photolyase or 6-4PP photolyase to specifically repair only CPD or only 6-4PP. After 300 J/m2 UVB exposure photoreactivation light (PR, UVA 60 kJ/m2) was provided for photolyase activation and DNA repair. Apoptosis was monitored 24 hours later by flow cytometric analysis of DNA content, using sub-G1 staining to indicate apoptotic cells. To confirm the effects observed with CPD lesions, the molecular mimic of CPD, pTpT, was also tested in vitro and in vivo for its effect on cell cycle and apoptosis.
Results
The specific repair of 6-4PP lesions after UVB exposure resulted in a dramatic reduction in apoptosis. These findings suggested that 6-4PP lesions may be the primary inducer of UVB-induced apoptosis. Repair of CPD lesions (despite their relative abundance in the UV-damaged cell) had little effect on the induction of apoptosis. Supporting these findings, the molecular mimic of CPD, (dinucleotide pTpT) could mimic the effects of UVB on cell cycle arrest, but were ineffective to induce apoptosis.
Conclusion
The primary response of the cell to UV-induced 6-4PP lesions is to trigger an apoptotic program whereas the response of the cell to CPD lesions appears to principally involve cell cycle arrest. These findings suggest that CPD and 6-4 PP may induce differential biological effects in the UV-damaged cell.
==== Body
Background
More than one million cases of skin cancer are diagnosed in the U.S. annually, resulting in 9,600 deaths, 7,400 of which are from metastatic melanoma [1]. Absorption of ultraviolet (UV) light produces two predominant types of DNA damage, cyclobutane pyrimidine dimers (CPD) [2] and pyrimidine (6-4) pyrimidone photoproducts (6-4PP) [3]. The result is a transition of C to T and CC to TT [4], which are the most frequent mutations of p53 in both human and mouse skin cancers [5,6]. UV damaged DNA causes torsional strain and is usually repaired by nucleotide excision repair (NER) or base excision repair (BER). After UV exposure, cells activate p53 and stall the cell cycle for repair [7,8]. If the damage is too severe, the cell will trigger apoptosis to get rid of a DNA damaged, potentially mutant cell [9]. But how does the cell determine when UV-damaged DNA can be repaired or when apoptosis should be initiated?
Previous studies have reported that CPD, but not 6-4PP, lesions are the major source of lasting UV-induced mutations [10], as CPD is usually at a 5 to 10 fold higher frequency than 6-4PP [11,12]. The repair of CPD lesions by exogenous CPD photolyase has been shown to reduce apoptosis both in vivo [13] and in vitro [14], but such studies did not examine the effect of 6-4PP lesions. Indeed, most previous studies have assumed that the amount of DNA damage is the deciding factor in a cell's fate with both lesions weighing in equally for biologic effect.
In this study, we have investigated whether UV-induced CPD and 6-4PP DNA damage have differential effects on the induction of apoptosis and cell-cycle arrest. To control DNA repair in UV-exposed human cells we have used photolyases, the DNA repair enzymes that are employed widely in the animal kingdom, but not in man [15]. Bacteria, fungi, plants, invertebrates, and many vertebrates, but not humans, use photolyases to repair UV-induced DNA damage [15,16]. This light-dependent (360–420 nm, visible~UVA) photolyase mediated repair of CPD or 6-4PP is referred to as photoreactivation (PR). By expressing photolyases in NER-defective cells we have been able to individually analyze the impact of residual CPD and 6-4PP lesions on apoptosis and cell cycle. Our findings demonstrate that the "type" of DNA lesion is as important as the total "amount" of damaged DNA in the balance of life and death, as it relates to apoptosis after UVB exposure.
Materials and methods
Chemicals and Cell Culture
All chemicals were purchased from Sigma (St. Louis, MO) unless otherwise noted. The SV40-transformed human cell line derived from an XPA patient, XP12ROSV, was stably transfected with expression vectors containing cDNAs for either 6-4PP photolyase, CPD photolyase, both types of photolyases, or empty vector controls as previously described [10,17]. Cells were maintained in Eagle's minimum essential medium (MEM) containing 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 1 mM sodium pyruvate, 100 units/ml penicillin, 100 μg/ml streptomycin, 1× MEM vitamins solution, and 1× MEM non-essential amino acids at 37°C in a humidified 5% CO2 incubator.
UV-irradiation and photoreactivation
Twenty-four hours prior to UV-irradiation, media was changed to phenol-red free culture medium to prevent the absorption of energy by phenol red. The UV source used was a Spectrolinker XL-1000 (Spectronics Corp., Westbury, NY) containing five 8-watt UV tubes. Two different UV wavelength tube sets were used, UVA 365 nm, and UVB 312 nm. For induction of DNA damage and apoptosis, cells were irradiated at 300 J/m2 UVB in 100 mm cell culture dishes without lids. After irradiation, cells were treated with UVA, which serves as photo-reactivation energy for photolyase activity (30 min, 60 kJ/m2). Following treatments, cells were incubated for 24 h prior to harvesting. Cells were harvested by washing with pre-warmed PBS and adding trypsin for 2 min. Cell pellets were collected after centrifugation at 1000 × g for 5 min.
Radio-immunoassay (RIA) of 6-4PP and CPD
Genomic DNA from UV exposed cells was isolated with a Wizard Genomic DNA purification kit (Promega, Madison, WI) 12 hours after 300 J/m2 UVB irradiation. UV-induced DNA damage was detected using specific antibodies and a radioimmunoassay described previously [18].
DNA damage detection by Immuno-Dot-Blot
The repair of CPD and 6-4PP lesions after photoreactivation was measured by an Immuno-Dot-Blot assay using the CPD-specific monoclonal antibody and the 6-4PP-specific monoclonal antibody (Kamiya Biomedical, WA). After UV treatment and PR, cellular DNA was isolated and then denatured in TE buffer (10 mM Tris-CL and 1 mM EDTA, pH 7.5) by boiling for 5 min. Thesamples were dot-blotted onto a Hybond N+ membrane (Amersham, NJ) using 200 ng of DNA for the CPD assay and 1 μg of DNA for the 6-4PP assay. The membrane was dried by baking on a 80°C plate for 1 hour then DNA was fixed to the membrane for 20 min by soaking in 0.4 N NaOH on a filter paper. The membranes were blocked overnight in phosphate-buffered saline, 0.2% Tween 20 (PBS-T) containing 5% (w/v) skim milk. After washing in PBS-T, the membranes were incubated for 2 h at room temperature with anti 6-4PP antibody and anti CPD monoclonal antibody. After washing, membranes were incubated for 1 h with peroxidase conjugated goat anti-mouse antibody (Jackson ImmunoResearch, PA). Signals were detected with a chemiluminescence kit (Amersham, NJ).
Western blot
Whole cell lysates were prepared from pelleted cells in lysis buffer [20 mM Tris-HCl (pH 8.0), 150 mM sodium chloride, 10% glycerol, 1% Triton-X 100, 2 mM sodium orthovanadate, 2 mM EDTA and 1× complete protease inhibitor cocktail (Roche, Germany)]. Total protein concentrations in lysate preparations were determined using a BCA assay and 15 μg total protein was loaded in each lane of a 10% SDS-PAGE gel. Proteins were resolved by electrophoresis and blotted onto nitrocellulose membranes (Hybond ECL, Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom). Blots were blocked with 5% milk in PBS for 1 h prior to incubation with primary antibodies. To confirm photolyase expression, blots were probed with rabbit antibodies specific for 6-4PP photolyase or CPD photolyase (Dr. Yasui, Tohoku University, Sendai, Japan). The blots were washed 3× with PBST (0.5% Tween 20) and incubated with secondary peroxidase-conjugated goat anti-rabbit antibodies for 1 h (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Blots were rinsed 3× with PBST and antibody binding was detected with ECL (Amersham Pharmacia Biotech).
Flow cytometric measurement of cell cycle and apoptosis
Following UV or pTpT treatment, cells were harvested and fixed in 70% ethanol at -20°C overnight. Cells were then pelleted and resuspended in 125 μl of PBS containing 0.2% Triton X-100 solution and 12 Kunitz units of ribonuclease A (Sigma) for 15 min at room temperature, followed by the addition of 130 μl propidium iodide (PI) at 0.1 mg/ml. Cells were maintained in the dark and analyzed by flow cytometry using a Becton Dickinson FACScan (San Jose, CA). After excluding cell debris and doublets, DNA content and cell cycle were analyzed using Cellquest-Pro software. Cells showing DNA content less than the peak for G1/G0 were considered sub-diploid and apoptotic.
Mimic of damaged DNA by pTpT
Thymidine dinucleotides, pTpT, (Midland Certified Reagent, Midland, TX) were used to mimic UV-induced DNA damage. In vitro, 100 μM pTpT was added to Jurkat cell culture medium, the same concentration of adenosine dinucleotides (pApA) was used for a negative control. Cells were harvested 1 h or 24 h after treatment to examine cell cycle. For in vivo studies, pTpT and pApA were dissolved in propylene glycol and 100 μl was applied to the shaved backs of C57BL/6 mice to cover an area at 200 μM per 2 cm2 (Jackson Labs, Bar Harbor, ME). All mouse studies were performed in accordance with NIH and Institutional Animal Care and Use Committee guidelines. The skin was shaved 24 hours before treatment. Mice were exposed to a single dose of 4 kJ/m2 UVB as positive controls. Mice were sacrificed 24 hours after treatment, with 3 mice for each treatment and 3 independent experiments performed. Formaldehyde-fixed skin biopsies were paraffin embedded. Seven-micron sections were made with a microtome and mounted onto charged glass slides. Apoptosis was assessed by staining for DNA fragmentation using the DeadEnd Fluorometric TUNEL System (Promega, Madison, WI). All TUNEL-positive cells in the epidermis of each section were counted using a fluorescence microscope. Each counted section was measured for length using calipers and using an apoptotic index generated to show the number of TUNEL-positive cells per unit of length.
Results and discussion
CPD is the dominant DNA lesion in NER-deficient cells after UVB irradiation
XPA cells mutated in the xpa gene cannot perform nucleotide excision repair (NER) and therefore accumulate large amounts of UV-induced DNA damage after UVB exposure. XPA cells were irradiated with 300 J/m2 UVB and assayed for UV-induced DNA damage using mutation-specific radioimmunoassays (RIA). The 300 J/m2 UVB dose was chosen as there was little acute, UV-induced oxidative cell death induced under these conditions, while there was sufficient levels of DNA damage to reproducibly measure decreases following photolyase repair (doses between 50~1500 J/m2 were screened initially). Because 6-4PP lesions occur at a lower frequency than CPD lesion, the 300 J/m2 dose was chosen as the lowest dose of UVB that produced high enough numbers of 6-4PP lesions such that repair could be accurately measured following photolyase treatment (>10 lesions per 106 DNA pairs). At this UVB dose, a substantial number of CPD lesions were observed (averaging 55 CPD per 106 DNA pairs) and fewer, but reproducible number of 6-4PP lesions (averaging at 12 6-4PP per 106 DNA pairs). The average number of lesions in unrepaired cells in two independent experiments is shown in Figure 1. Our results show that CPD are the predominant UV-induced DNA lesions and are approximately five-fold more prevalent than 6-4PP. These results are consistent with other studies that show 6-4PP lesions occur at 15–30% the frequency of CPD lesion following UV damage [12].
Repair of 6-4PP significantly reduces apoptosis
We examined the effects of repairing either CPD or 6-4PP lesions by stably expressing photolyase cDNAs in XPA cells. Several studies have shown that these photolyase enzymes can efficiently and specifically repair DNA lesions within one hour when provided with photo-reactivation light [10,14,19]. Stable expression and function of each photolyase in XPA cells was confirmed by western blot and immuno-dot blot (Fig 2). XPA cells expressing 6-4PP-PL only repair 6-4PP type lesions, vice versa, CPD-PL only repair CPD type lesions (Fig 2C and 2D). Using this system we were able to independently repair one kind of lesion and thereby examine the biological effects of the other in relative isolation. Background apoptosis of XPA cells, as measured by sub-G1 DNA, is 1–3% for non-irradiated cells. Twenty-four hours after exposure to UVB and PR (60 kJ/m2 UVA), 20.9% of empty vector transfected cells were apoptotic. UV-exposed XPA cells that expressed both photolyases and received PR, showed only 5.3% apoptosis, a reduction of approximately 75%. Cells expressing only one of the photolyases showed considerably different responses depending upon which enzyme was expressed. Repair of 6-4PP lesions reduced apoptosis to 6.6%, while repair of CPD lesions only reduced apoptosis to 12.9%. Although CPD is the major type of UV-induced DNA lesion (82%, about five times more than 6-4PP) repair of CPD only reduces apoptosis by 40% (Fig 3). In contrast, 6-4PP lesions comprise only 18% of UVB-induced DNA lesions, but account for 70% of the apoptosis (Fig 3). Indeed, repairing 6-4PP lesions alone suppressed apoptosis as much as repairing both types of DNA lesions. Similar findings were observed when Annexin V was used to measure apoptotic responses; suggesting that both sub-diploid DNA and phosphatidylserine exposure on the outer membrane leaflet are impacted by photolyase repair (data not shown). These findings indicate that 6-4PP lesions are more potent inducers of apoptosis than CPD lesions.
The CPD mimic, pTpT, stalls the cell cycle but does not significantly induce apoptosis
UV exposure induces mainly CPD lesions (Fig. 1) that are typically repaired by NER. Thymidine dinucleotides, pTpT a mimic of CPD, are structurally similar to the small DNA fragments released during NER and have been reported to mimic UV-damaged DNA by activating p53 and proliferating cell nuclear antigen (PCNA) to enhance DNA repair [20]. These dinucleotides have also been reported to mimic damaged DNA and to increase melanogenesis [21] in normal human melanocytes and in human melanoma cells. The monomeric form of thymidine, pT, [22] or pApA [21] does not increase melanogenesis indicating the effect of pTpT in mimicking CPD lesions.
UV can upregulate the death receptor apoptotic pathway to eradicate DNA damaged cells [23,24]. Lymphocytes are especially sensitive to DNA damage-induced apoptosis. Jurkat cells, a human T-cell line, provides a well-established model for death receptor (Fas-FasL) mediated apoptosis in response to either UV and/or chemotherapeutic drugs [25]. We examined if pTpT could induce Fas-FasL mediated apoptosis in vitro and in vivo. The addition of pTpT to Jurkat cells stalled the cell cycle, but did not induce FasL (data not shown) or apoptosis (as seen by sub-diploid peak analysis, Fig. 4A). Jurkat cells exposed to the same concentration of control oligonucleotide, pApA, did not show any cell cycle disruption or increased apoptosis. The relative percentage cells in each phase of the cell cycle for pTpT- and pApA-treated cells is shown in Fig. 4B. Note the increased proportion of cells in S-phase and decreased proportion of cells in G2/M phase for pTpT treated cells as compared to controls and pApA treated cells.
We also tested whether pTpT could alter FasL-mediated apoptosis in an established in vivo model. For these experiments, solvent containing 200 μM pTpT or 200 μM pApA was applied on the shaved dorsal skin of C57BL/6 mice. Mice exposed to a single dose of 4 kJ/m2 of UVB were used as positive controls. Apoptosis was assessed on paraffin embedded formaldehyde-fixed skin biopsies using the DeadEnd Fluorometric TUNEL System (Promega). Although UV-induced apoptosis was observed 12 hours after UV treatment and continued to increase at least until 24 hours following UV exposure (Fig 5A), pTpT did not induce apoptosis in vivo in concert with our findings in vitro (Fig. 5B). Taken together, our findings suggest that pTpT, the mimic of CPD, is not a potent inducer of apoptosis, but inhibits cell cycle progression. Similar results have been observed with human melanocytes in vitro, where pTpT has been shown to induces melanogenesis and stall the cell cycle in the absence of increased apoptosis [22].
Our finding that repair of 6-4 PP lesions was more effective in the blockade of apoptosis than the repair of CPD lesions would seem to be contradictory to published findings that transgenic mice expressing the CPD photolyase in the skin showed dramatically reduced levels of UV-induced apoptosis, erythema, and hyperplasia [13]. One underlying factor that may contribute to the effects of CPD on apoptosis induction is the p53 status of the cell. Because we [9] have shown that p53 can induce upregulation of death receptors such as Fas which plays a key role in UV-induced apoptosis in vivo [26], it is imperative to consider the p53 status of the cells/animals used for study. In the case of the XP cells used for the current studies, p53 is mutant (data not shown). When p53 is wild-type, CPD lesions may be more robust in the induction of apoptosis. Other studies have shown that transgenic animals expressing the CPD photolyase show superior resistance to sunlight-induced carcinogenesis [27]. These findings may relate to the fact that UV-induced apoptosis is decreased in the skin of chronically irradiated mice consistent with the induction p53 mutations [23]. Alternately stated, when p53 is wild-type CPD lesions may be a more powerful inducer of apoptosis, however, once p53 becomes mutated (as is commonly observed for non-melanoma skin cancers) CPD lesions may act to induce cell cycle arrest. It will be especially interesting to examine transcription coupled repair efficiency following photolyase treatments in wild-type and mutant p53 containing cells as reports have indicated that transcription coupled repair efficiency can determine cell cycle progression and apoptosis in UV-irradiated cells [28]. Future studies will address the impact of p53 status on the biological effects of CPD and 6-4 PP lesion repair as well as the effects of such repair on transcription coupled repair and global genome repair.
Conclusion
The UVB-induced DNA lesions, CPD and 6-4PP, show differential biological effects with respect to the induction of apoptosis and cell cycle arrest. CPD lesions are the most abundant following UVB exposure and photolyase-induced repair of these lesions decreased the overall apoptotic response by approximately 40% compared with cells receiving no PR repair. Repair of 6-4PP lesions which account for approximately 20% of the total UV-induced DNA damage, on the other hand, decreased the overall apoptotic response by approximately 75% (almost equivalent to that using both CPD and 6-4PP photolyases together). These findings suggest that the 6-4 PP lesion is more potent in the induction of UV-induced apoptosis. In contrast, the CPD lesion appears to be more potent in the induction of cell cycle arrest. Using the CPD mimic, pTpT, we documented that cell cycle arrest occurred in vitro after treatment in the absence of increased apoptosis. Similar findings were also observed in vivo using the pTpT molecular mimic of CPD lesions. Therefore CPD lesions will halt the cell cycle to effect repair that may or may not eliminate the damage, thus increasing the relative probability that these kinds of lesions will accumulate leading to mutation. In support of this premise, You et al. [10] have documented that CPD lesions account for the majority of UV-induced mutations in mammalian cells. The more potent apoptosis-inducing activity of 6-4PP lesions may induce apoptosis and "erase" the damaged cell, thus decreasing its carcinogenic potential.
Authors' contributions
HL and LOS initiated the research design, execution, coordination and writing. SN and AY carried out the XPA stable transfection cell line production. BW performed the FAScan assay. LM participated in the pTpT mimics CPD study. DE participated in study design and manuscript editing. All the authors have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are thankful to Dr. David Mitchell's (UT, MD Anderson Cancer Center) for help with the CCD and 6-4PP RIA assay. This research was supported by an American Cancer Society grant RPG-96-070-07-CIM to L.O.
Figures and Tables
Figure 1 Radio-immunoassay (RIA) of UV-induced 6-4PP and CPD. Twelve hours after UVB irradiation, genomic DNA was isolated from treated cells and assayed for UV-induced DNA damage using lesion-specific antibodies. An average of fifty-five CPD and twelve 6-4PP lesions per million bases were detected after exposure to 300 J/m2 UVB. Results are the average of duplicate assays from two independent experiments. Error bars show standard deviation.
Figure 2 Detection of photolyase expression and function. Photolyase expression was confirmed by western blots of whole cell lysates from stably transfected XPA cell lines containing empty vector or both photolyases. Blots were probed with antibodies specific for CPD-PL (A) and 6-4PP-PL (B). Photolyase repair of UV induced DNA damage in XPA cells was detected by immuno-dot-blot, using the same antibodies. Representative examples of 6-4PP (C) and CPD (D) lesion studies are shown.
Figure 3 Impact of photoreactivation on UVB-induced apoptosis. Apoptosis in XPA cells was monitored at a 24 hour time point following exposure to 300 J/m2 UV and photoreactivation-induced repair. Apoptosis was examined by PI staining and FACS, gating the sub-G1 population as apoptotic. Apoptotic values were normalized to empty vector containing cells (amount of apoptosis set as 100%). Bar graphs show the relative amount of apoptosis for control cells, cells expressing CPD-PL alone (40% rescue), cells expressing 6-4PP-PL alone (70% rescue), or cell expressing both-PLs (75% rescue).
Figure 4 Effects of pTpT on cell cycle and apoptosis in vitro. A. Jurkat cells were exposed to pTpT (100 μM) or pApA (100 μM) in vitro and cell cycle analysis performed after 1 and 24 hours by flow cytometry as described in Materials and Methods. Results show representative histograms from five independent experiments. Note that only pTpT stalls the cell cycle, not pApA, and no increase in apoptotic sub-G1 DNA content was observed with either pTpT or pApA. B. Bar graph showing percentage of cells in each phase of the cell cycle in cells treated with pTpT or pApA for 24 hrs.
Figure 5 CPD mimic, pTpT, does not induce apoptosis in vivo. TUNEL-stained sections from mouse dorsal skin coated with pTpT or pApA, or exposed to a single dose 4 kJ/m2 UVB as a positive control. (A) Green fluorescence indicates 3' DNA breaks, representative of apoptotic cells. (B) Numbers of TUNEL positive cells in skin sections from three separate experiments, consisting of three mice in each experimental group. Apoptotic values indicate the number of TUNEL-positive cells by tissue length. Pictures were taken on a Nikon ECLIPSE TE2000-U microscope under FITC illumination at 400×.
==== Refs
American Cancer Society [http://wwwcancerorg/docroot/PED/content/ped_7_1_What_You_Need_To_Know_About_Skin_Cancerasp?sitearea=PED]
Lippke JA Gordon LK Brash DE Haseltine WA Distribution of UV light-induced damage in a defined sequence of human DNA: detection of alkaline-sensitive lesions at pyrimidine nucleoside-cytidine sequences Proc Natl Acad Sci U S A 1981 78 3388 3392 6943547
Mitchell DL Nairn RS The biology of the (6-4) photoproduct Photochem Photobiol 1989 49 805 819 2672059
Ananthaswamy HN Loughlin SM Cox P Evans RL Ullrich SE Kripke ML Sunlight and skin cancer: inhibition of p53 mutations in UV-irradiated mouse skin by sunscreens Nat Med 1997 3 510 514 9142118
Soehnge H Ouhtit A Ananthaswamy ON Mechanisms of induction of skin cancer by UV radiation Front Biosci 1997 2 D538 D551 9343491
Ananthaswamy HN Bowden GTFSM Ultraviolet light as a carcinogen. Chemical Carcinogens and Anticarcinogens Comprehensive Toxicology 1997 12 Oxford, Pergamon pp. 255 279
el-Deiry WS Tokino T Velculescu VE Levy DB Parsons R Trent JM Lin D Mercer WE Kinzler KW Vogelstein B WAF1, a potential mediator of p53 tumor suppression Cell 1993 75 817 825 8242752
Hermeking H Lengauer C Polyak K He TC Zhang L Thiagalingam S Kinzler KW Vogelstein B 14-3-3 sigma is a p53-regulated inhibitor of G2/M progression Mol Cell 1997 1 3 11 9659898
Hill LL Ouhtit A Loughlin SM Kripke ML Ananthaswamy HN Owen-Schaub LB Fas ligand: a sensor for DNA damage critical in skin cancer etiology Science 1999 285 898 900 10436160
You YH Lee DH Yoon JH Nakajima S Yasui A Pfeifer GP Cyclobutane pyrimidine dimers are responsible for the vast majority of mutations induced by UVB irradiation in mammalian cells J Biol Chem 2001 276 44688 44694 11572873
Meador JA Walter RB Mitchell DL Induction, distribution and repair of UV photodamage in the platyfish, Xiphophorus signum Photochem Photobiol 2000 72 260 266 10946581
Bissonauth V Drouin R Mitchell DL Rhainds M Claveau J Rouabhia M The efficacy of a broad-spectrum sunscreen to protect engineered human skin from tissue and DNA damage induced by solar ultraviolet exposure Clin Cancer Res 2000 6 4128 4135 11051266
Schul W Jans J Rijksen YM Klemann KH Eker AP de Wit J Nikaido O Nakajima S Yasui A Hoeijmakers JH van der Horst GT Enhanced repair of cyclobutane pyrimidine dimers and improved UV resistance in photolyase transgenic mice Embo J 2002 21 4719 4729 12198174
Chigancas V Miyaji EN Muotri AR de Fatima Jacysyn J Amarante-Mendes GP Yasui A Menck CF Photorepair prevents ultraviolet-induced apoptosis in human cells expressing the marsupial photolyase gene Cancer Res 2000 60 2458 2463 10811124
Yasui A McCready SJ Alternative repair pathways for UV-induced DNA damage Bioessays 1998 20 291 297 9619100
Yasui AAPME Hoekstra JANMF DNA Photolyase DNA Damage and Repair 1998 2 Totowa, NJ, Humana Press Inc 9 32
Nakajima S Lan L Kanno S Takao M Yamamoto K Eker AP Yasui A UV light-induced DNA damage and tolerance for the survival of nucleotide excision repair-deficient human cells J Biol Chem 2004 279 46674 46677 15342631
Mitchell DL Byrom M Chiarello S Lowery MG Effects of chronic exposure to ultraviolet B radiation on DNA repair in the dermis and epidermis of the hairless mouse J Invest Dermatol 2001 116 209 215 11179995
Chigancas V Batista LF Brumatti G Amarante-Mendes GP Yasui A Menck CF Photorepair of RNA polymerase arrest and apoptosis after ultraviolet irradiation in normal and XPB deficient rodent cells Cell Death Differ 2002 9 1099 1107 12232798
Maeda T Eller MS Hedayati M Grossman L Gilchrest BA Enhanced repair of benzo(a)pyrene-induced DNA damage in human cells treated with thymidine dinucleotides Mutat Res 1999 433 137 145 10102040
Eller MS Gilchrest BA Tanning as part of the eukaryotic SOS response Pigment Cell Res 2000 13 Suppl 8 94 97 11041364
Pedeux R Al-Irani N Marteau C Pellicier F Branche R Ozturk M Franchi J Dore JF Thymidine dinucleotides induce S phase cell cycle arrest in addition to increased melanogenesis in human melanocytes J Invest Dermatol 1998 111 472 477 9740243
Ouhtit A Gorny A Muller HK Hill LL Owen-Schaub L Ananthaswamy HN Loss of Fas-ligand expression in mouse keratinocytes during UV carcinogenesis Am J Pathol 2000 157 1975 1981 11106570
Hill LL Shreedhar VK Kripke ML Owen-Schaub LB A critical role for Fas ligand in the active suppression of systemic immune responses by ultraviolet radiation J Exp Med 1999 189 1285 1294 10209045
Kasibhatla S Brunner T Genestier L Echeverri F Mahboubi A Green DR DNA damaging agents induce expression of Fas ligand and subsequent apoptosis in T lymphocytes via the activation of NF-kappa B and AP-1 Mol Cell 1998 1 543 551 9660938
|
16236176
|
PMC1276790
|
CC BY
|
2021-01-04 16:03:04
|
no
|
BMC Cancer. 2005 Oct 19; 5:135
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-135
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-551622974510.1186/1471-2148-5-55Research ArticleEvolutionary, structural and functional relationships revealed by comparative analysis of syntenic genes in Rhizobiales Guerrero Gabriela [email protected] Humberto [email protected] Alejandro [email protected]íaz Rafael [email protected] Miguel Angel [email protected] Arturo [email protected] Jaime [email protected] Program of Functional Genomics of Prokaryotes, Centro de Ciencias Genómicas, Universidad Nacional Autónoma de México. Ave. Universidad s/n (P.O. Box 565-A), Cuernavaca, Morelos, 62210, México2 Program of Computational Genomics, Centro de Ciencias Genómicas, Universidad Nacional Autónoma de México. Ave. Universidad s/n (P.O. Box 565-A), Cuernavaca, Morelos, 62210, México2005 17 10 2005 5 55 55 18 7 2005 17 10 2005 Copyright © 2005 Guerrero et al; licensee BioMed Central Ltd.2005Guerrero et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Comparative genomics has provided valuable insights into the nature of gene sequence variation and chromosomal organization of closely related bacterial species. However, questions about the biological significance of gene order conservation, or synteny, remain open. Moreover, few comprehensive studies have been reported for rhizobial genomes.
Results
We analyzed the genomic sequences of four fast growing Rhizobiales (Sinorhizobium meliloti, Agrobacterium tumefaciens, Mesorhizobium loti and Brucella melitensis). We made a comprehensive gene classification to define chromosomal orthologs, genes with homologs in other replicons such as plasmids, and those which were species-specific. About two thousand genes were predicted to be orthologs in each chromosome and about 80% of these were syntenic. A striking gene colinearity was found in pairs of organisms and a large fraction of the microsyntenic regions and operons were similar. Syntenic products showed higher identity levels than non-syntenic ones, suggesting a resistance to sequence variation due to functional constraints; also, an unusually high fraction of syntenic products contained membranal segments. Syntenic genes encode a high proportion of essential cell functions, presented a high level of functional relationships and a very low horizontal gene transfer rate. The sequence variability of the proteins can be considered the species signature in response to specific niche adaptation. Comparatively, an analysis with genomes of Enterobacteriales showed a different gene organization but gave similar results in the synteny conservation, essential role of syntenic genes and higher functional linkage among the genes of the microsyntenic regions.
Conclusion
Syntenic bacterial genes represent a commonly evolved group. They not only reveal the core chromosomal segments present in the last common ancestor and determine the metabolic characteristics shared by these microorganisms, but also show resistance to sequence variation and rearrangement, possibly due to their essential character. In Rhizobiales and Enterobacteriales, syntenic genes encode a high proportion of essential cell functions and presented a high level of functional relationships.
==== Body
Background
A huge amount of information has been obtained from sequencing projects. More than two hundred complete microbial genomes are available to date in public databases and sequencing of a similar number is in progress [1]. Many questions remain unsolved. For example, what is the biological meaning, if any, of gene arrangement in the bacterial chromosome?
Changes in gene sequence and chromosomal rearrangements constitute the main sources of genomic variability. Nonsynonymous substitutions in the first or second nucleotides of the codon change the encoded residue and are thus a driving force of natural selection. Genomic studies in bacteria regarding synonymous and nonsynonymous substitution rates have been published elsewhere [2,3]. Chromosomes show constraints on rearrangement and works dealing with that aspect were recently reviewed by Rocha [4]; he suggested that there is a balance between conservation and change in the organization of the chromosome.
The operon represents the first level of the gene organization. Neighboring genes, especially those in co-directional and in divergent orientation, represent a second organization level because they show a certain functional association revealed by genomic context analysis [5]. Regarding comparisons among closely related species, the gene order conservation, or synteny, represents a third level of organization. Synteny depends on shared ancestry and inter- and intrachromosomal exchanges, and represents a higher relationship between taxa. It was suggested that physiologically important gene clusters could be positively selected, and synteny perhaps reveals the functional constraints of these genes [6]. For the detection of synteny it is necessary first to determine the set of orthologous genes in pairs of organisms and recently an inverse method has proven useful for this [7].
Recombination/transposition events can easily disrupt synteny. Species of Buchnera and Corynebacterium have low levels of chromosomal rearrangements and lack recA and recBCD orthologs, respectively [8,9], thus suggesting that recombination is an important factor for loss of synteny. Synteny studies have focused in short gene clusters in eukaryotes [10,11] while whole chromosome comparative analysis has been done in bacteria and archaea [12-18]. For example, there is a striking conservation between the chromosomes of Escherichia coli and Salmonella typhimurium [19], and also between those of Brucella melitensis, B. suis and Mesorhizobium loti [20]. However, the synteny analyses for alpha proteobacteria such as those reported in the genome sequence determinations of Brucella melitensis, Sinorhizobium meliloti and Agrobacterium tumefaciens are highly schematic [20-22]. A thorough analysis of Rhizobiales genomes would determine if the key set of genes covering the most important metabolic functions in these organisms are syntenic.
Another factor affecting chromosomal rearrangement is horizontal gene transfer (HGT) which occurs in bacteria [23,24], but estimating its impact on genome organization has proved a daunting task for two reasons. First, the reliability of compositional methods to detect HGT events has been questioned [25,26]. Second, phylogenetic methods, albeit more reliable, are not always applicable and can easily be misleading without proper care. The results of a recently published analysis suggest that codon usage compatibility between alien genes and recipient genomes [27] is a prerequisite for successful HGT events. This premise has been supported by other evidence [28,29].
Among the most accepted methods to deduce functional relationships of proteins are phylogenetic profile [30], gene neighboring [31], and the Rossetta stone method [32]. These methods can give additive information about metabolic networks existing in organisms [33]. ProLinks is a program based on these methods with an extensive library of predicted functional interactions from 83 genomes [34]. Von Mering et al. [35], also applying these approaches with the STRING program, found global modularities in functional protein networks. A question remains about whether functional linkage differences exist in syntenic and non-syntenic gene clusters.
Rhizobiales is a prokaryotic order belonging to the alpha proteobacteria subdivision; some rhizobial species are intensively studied for their nitrogen-fixing ability when in symbiosis with leguminous plants. The order comprises both plant symbionts and plant and animal pathogens such as Rhizobium, Agrobacterium and Brucella, respectively. In rhizobia, genes responsible for the symbiotic interaction are commonly found on large plasmids or incorporated in a particular stretch of the chromosome called the symbiotic island [36-38]. The physiological potential of the rhizobial chromosome allows cell survival under different conditions. For example, an A. tumefaciens strain containing the symbiotic plasmid from Rhizobium etli induced nodules on legume plants [39,40], and conversely, an S. meliloti derivative strain with Ri, the rhizogenic induction plasmid, formed root mats on alfalfa plants [41]. Additionally, there are similarities in the parasitic/symbiotic strategies employed by species of the Rhizobiales [20]. Also, it is possible to find diverse life-styles among the members of Enterobacteriales order (gamma proteobacteria): for example, Buchnera is an obligate aphid symbiont; E. coli and S. typhimurium are common gut inhabitants in mammals; and Shigella flexneri, Yersinia pestis and Erwinia carotovora are pathogens, either for animals or plants [42].
The complete genome sequences of seven species of Rhizobiales were available by 2004, namely S. meliloti [21], Mesorhizobium loti [43], Bradyrhizobium japonicum [44], A. tumefaciens [22,45], B. melitensis [46], B. suis [20] and Rhodopseudomonas palustris [47]. Although their genomes show a certain degree of conservation, variability corresponding to their evolutionary divergence points, microbial life styles and ecological niches was also found. Comparative genomics has captured the attention of researchers as a way of achieving a better understanding of the molecular basis underlying phenomena such as symbiosis and pathogenesis.
We classified the genes of several rhizobial species in order to gain a comprehensive insight into chromosomal conservation and genome rearrangement. Conserved genes among these species can reveal phylogenetic relationships, but also show metabolic strategies useful in understanding the niche diversity in which these organisms usually grow. In particular, syntenic/non-syntenic genes among these species were analyzed in terms of their sequence identity/similarity, physical characteristics of the encoded products and functional relationships among them. Additionally, in order to find more general trends, we compared these results with an analysis performed on genomes belonging to the Enterobacteriales.
Results
Approach, strategy and outline
Our main objective was to enhance our understanding of the functional meaning of the gene arrangement on the bacterial chromosome, taking as examples some genomes from the Rhizobiales and Enterobacteriales. We consider that gene neighboring is not a random trait and gives an adaptive advantage to the cell because the proteins produced are likely to perform related functions. Our belief is that the coordinated expression of genes, organized on the chromosome either as operons or clusters, permits the correct integration of metabolic functions.
Our approaches were: i) to obtain a comprehensive gene classification, applicable to each of the species analyzed and suitable to make comparisons among them, and ii) to detect specific gene characteristics (if any) of each of the classes. In the first approach we identified orthologs among chromosomes, defined those that were syntenic, those in a different replicon, and those that were species-specific. For the second approach, we analyzed gene/protein sequences for identity, calculated the horizontal gene transfer rate for each class and the predicted molecular weight and isoelectric point of the peptides, and inferred the functional relationship in syntenic or non-conserved chromosomal regions. The results are presented in the following order: 1) Synteny in Rhizobiales (gene classification of Rhizobiales; gene organization and microsyntenic region formation; synteny and insertion sequences, horizontal gene transfer and codon usage), 2) Synteny in Enterobacteriales, 3) Sequence analysis of the chromosomal predicted orthologs, 4) Physical characteristics of the translated products of syntenic genes, and 5) Functional roles and linkage of chromosomal predicted orthologs. S. meliloti was taken as reference organism for the comparisons with each A. tumefaciens, M. loti, B. melitensis and E. coli. E. coli was used as base to compare with S. typhimurium, E. carotovora and S. meliloti.
1) Synteny in Rhizobiales
Gene classification of Rhizobiales
The selection criteria mentioned in Methods were applied to the chromosomes of S. meliloti (Sm), A. tumefaciens (At), M. loti (Ml), and B. melitensis (Bm). As compared to the chromosome of S. meliloti, we found that more than 60% of genes were chromosomal predicted orthologs in the At circular chromosome (At-C) and Bm chromosome I (Bm-I), one third in Ml chromosome and Bm chromosome II (Bm-II) and one quarter in the At linear chromosome (At-L); however, the number of chromosomal predicted orthologs in each organism was similar, about two thousand (Table 1). The Sm chromosome presents 3341 genes in 3.65 Mb.
Table 1 Gene classification of Rhizobiales.
In comparison with the Sinorhizobium meliloti chromosome
At-C At-L At Ml Bm-I Bm-II Bm
Genes in chromosome (length in Mb) 2721(2.84) 1833(2.07) 4554(4.92) 6750(7.04) 2059(2.12) 1139(1.18) 3198(3.30)
Chromosomal orthologs (% of chr. genes) 1737(63.8) 478(26.1) 2215(48.6) 2279(33.8) 1310(63.6) 415(36.4) 1725(53.9)
Syntenic genes (% of chr. orthologs) 1480(85.2) 357(74.7) 1837(82.9) 1624(71.3) 1039(79.3) 272(65.5) 1311(76.0)
Nonsyntenic genes (% of chr. orthologs) 257(14.8) 121(25.3) 378(17.1) 655(28.7) 271(20.7) 143(34.5) 414(24.0)
Microsyntenic regions 160 45 205 227 132 41 173
Syntenic genes in regions (% of synt. genes) 1394(94.2) 325(91.0) 1719(93.6) 1428(87.9) 965(92.9) 230(84.6) 1195(91.1)
Plasmidic homologs (% of chr. genes) 159(5.8) 490(26.7) 649(14.2) 924(13.7) 70(3.4) 134(11.8) 204(6.4)
Homologs in rest of Rhizobiales (% of chr. genes) 459(16.9) 523(28.5) 982(21.6) 1724(25.5) 349(16.9) 368(32.3) 717(22.4)
Species specific genes (% of chr. genes) 366(13.4) 342(18.7) 708(15.5) 1823(27.0) 330(16.0) 222(19.5) 552(17.3)
Operons (except homologs in rest of Rhizobiales) 527 336 863 1140 353 192 545
Syntenic operons (% of operons) 282(53.5) 68(20.2) 350(40.6) 259(22.7) 168(47.6) 43(22.4) 211(38.7)
Nonsyntenic operons (% of operons) 7(1.3) 0 7(0.8) 12(1.0) 8(2.3) 3(1.6) 11(2.0)
Plasmidic operons (% of operons) 15(2.8) 47(14.0) 62(7.2) 88(7.7) 1(0.3) 9(4.7) 10(1.8)
Species specific operons (% of operons) 26(4.9) 27(8.0) 53(6.1) 150(13.2) 20(5.7) 14(7.3) 34(6.2)
Mixed operons (% of operons) 197(37.4) 194(57.7) 391(45.3) 631(55.3) 156(44.2) 123(64.1) 279(51.2)
Chr., chromosomal. Synt., syntenic. At-C, A. tumefaciens circular chromosome. At-L, A. tumefaciens linear chromosome. At, A. tumefaciens chromosomes. Ml, M. loti chromosome. Bm-I, B. melitensis chromosome I. Bm-II, B. melitensis chromosome II. Bm, B. melitensis chromosomes.
We assessed the chromosomal genes with conserved order or synteny (see Methods). Syntenic genes represented about 70–80% of chromosomal predicted orthologs (see Table 1). That is, the conserved chromosomal order of these genes seems favored. The remarkable synteny level is highlighted by a group of 1038 common syntenic genes in all these species. Non-syntenic genes represented from 17% to 35% of the chromosomal predicted orthologs in these organisms. Only 98 non-syntenic genes were common in the four Rhizobiales.
The remaining categories obtained in this analysis were: homologs present in plasmids, homologs with the rest of rhizobial chromosomes, and those with no orthologs in the public databases (species-specific genes, also known as orphan). Homologs in plasmids were more abundant in the At-L,Ml and Bm-II chromosomes (Table 1). These replicons have special features as commented below and in Discussion. Aside from the two organisms being compared, some genes also matched with other rhizobial chromosomes (including B. japonicum USDA110) and comprised about a quarter of the chromosomal genes in these species (Table 1). Some of them only matched with unidirectional best hits (homologs). In this context, in an additional analysis between S. meliloti and A. tumefaciens, we found that 129 genes matched with unidirectional best hit (white fraction, Figure 1), but the rest were predicted orthologs with bidirectional best hits, either syntenic or non-syntenic, and are represented in the red and blue striped bars (denoted as rest, Figure 1), respectively. Species-specific genes were especially abundant in the M. loti chromosome, while in the other replicons this class covered at most 20%. One half of these species-specific genes were not present in the COG database [48] and the rest were denoted as hypothetical (data not shown).
Figure 1 Schematic representation of the S. meliloti chromosome (compared with A. tumefaciens) according to the classification of predicted orthologs and homologs. Striped bars in red, from the bottom to the top: genes syntenic with the A. tumefaciens circular chromosome (denoted with C); syntenic with the A. tumefaciens linear chromosome (L); syntenic with the rest of Rhizobiales (rest). Striped bars in blue, from the bottom to the top: genes non-syntenic with the A. tumefaciens circular chromosome (C); non-syntenic with A. tumefaciens linear chromosome (L); non-syntenic with the rest of Rhizobiales (rest). White bar, homologs in other Rhizobiales chromosomes matched with unidirectional best hits (other). Green bar, homologs in plasmids (plasmid). Gray bar, species-specific genes (sp). Numbers indicate genes in each of the categories.
In this way, all chromosomal genes were assigned to the categories mentioned, as shown in Figure 1 for the comparison of S. meliloti with A. tumefaciens. This approach gives a panoramic view about shared and unshared genes with other rhizobial species. Additional file 1 shows the categories for the other comparisons.
A schematic representation of rhizobial chromosomes was obtained relative to the classification of the genes. A high proportion of chromosomal predicted orthologs of the analyzed species were syntenic; however, in the A. tumefaciens linear, M. loti and B. melitensis I chromosomes, plasmidic homologs and species-specific genes were particularly abundant.
Gene organization in operons and microsyntenic regions in Rhizobiales
A relationship between predicted orthologs and operons was also explored. For S. meliloti (in comparison with A. tumefaciens), one half of the syntenic genes was found organized in 303 syntenic operons, a half of the total predicted operons (606); similar proportions were found for the A. tumefaciens circular and B. melitensis I chromosomes. In the A. tumefaciens linear, B. melitensis II and M. loti chromosomes, the proportion was about 21% (Table 1). The first two can be considered as accessory chromosomes and the last is the largest chromosome of the analyzed species. Non-syntenic genes were found organized in operons in a very small proportion in all analyzed Rhizobiales (Table 1).
A relevant level of operon conservation was found. In the comparison of S. meliloti with A. tumefaciens, 50% of the syntenic genes organized in operons were in identical operons, and taken together with those in similar operons (differing by one or two genes), 82% of total syntenic genes were in conserved operons. Similar proportions were found for the other comparisons (data not shown).
The operons formed with plasmidic homologs constituted a small fraction of the predicted operons (Table 1). Also, a reduced proportion was found for species-specific operons, except for Ml. Finally, mixed operons were present in a higher amount and ranged from 37 to 64% of the predicted operons (Table 1). This category revealed the highest rate of chromosomal rearrangements in these organisms. The mixed operons contained 17% of the syntenic genes, on average.
Syntenic genes were found in clusters and were assigned to microsyntenic regions (see Methods). When comparing S. meliloti with both A. tumefaciens chromosomes, 205 regions were common. In particular, At-C regions in common with Sm chromosome were found along all the chromosome, except in the third quarter (Figure 2). The third quarter had colinearity with At-L (Figure 2, see also Figure 3 panel a). In the other comparisons, a similar amount of common regions were observed (Table 1, see also Additional files 2 and 3). 146 regions were shared in Sm, At, and Ml and 94 regions were common to the four Rhizobiales. About 90% of syntenic genes were located in the microsyntenic regions (Table 1).
Figure 2 Synteny histogram of the S. meliloti chromosome in comparison to A. tumefaciens chromosomes. Red bars, syntenic genes. Framed with yellow boxes, microsyntenic regions with the A. tumefaciens (At) circular chromosome. Framed with light green boxes, microsyntenic regions with the At linear chromosome. Microsyntenic regions are denoted by letters (and numbers) in progressive order. Dark blue bars, non-syntenic genes with the At circular chromosome. Light blue bars, non-syntenic genes with the At linear chromosome. Green bars, homologs in plasmids. Gray bars, species-specific genes. White bars, homologs with other Rhizobiales chromosomes. Direction of transcription is denoted by upper (plus) or lower (minus) positions in respect to the central line. Predicted operons are denoted by red arrows. Scale in bp.
Figure 3 Schematic rearrangement of microsyntenic regions among S. meliloti, A. tumefaciens, M. loti and B. melitensis chromosomes. Panels: (a), S. meliloti chromosome (middle line) compared with A. tumefaciens circular (bottom line) and linear (upper line) chromosomes. (b), S. meliloti chromosome (lower line) compared with M. loti (upper line) chromosome. The M. loti chromosome was segmented in two fragments to maximize colinearity (see Methods). (c), S. meliloti chromosome (middle line) compared with B. melitensis chromosome I (bottom line), and II (upper line). oriC of Bm I was inverted to obtain maximal colinearity (see Methods). Red lines (orange for At-L and Bm-II chromosomes) represent the initial positions of the microsyntenic regions in each of the species analyzed.
When At-C syntenic genes were compared with Sm chromosome, a high level of colinearity was found (see Additional file 4 panel a). In the case of Ml, the synteny was disrupted possibly due to a conflicting annotation; for Bm-I, an inverse colinearity was obtained, possibly by oriC inversion relative to Sm (see Additional file 4 panels b and c, respectively). In regard to the rearrangement of microsyntenic regions, an extensive chromosomal colinearity was observed in long tracts. For example, in the comparison of Sm and At circular chromosomes, 93 microsyntenic regions were colinear, 24 almost colinear, 18 with drastic changes, and 13 inverted. The schematic representation is shown in Figure 3 panel a, lower part. In the comparison with the At linear and Bm-II chromosomes, highly rearranged structures were found (Figure 3 panel a, upper part and panel c, upper part). Rearrangement of microsyntenic regions on the chromosomes of M. loti (Figure 3 panel b) and B. melitensis I (Figure 3 panel c, lower part), showed a high level of colinearity when the conflicting annotation and the oriC inversion, respectively, were modified (see Methods).
Syntenic genes were organized mainly at two levels: operons and microsyntenic regions. Syntenic operons were as abundant as mixed operons. Extensive blocks of chromosomal colinearity were found, despite rearrangements.
Synteny and insertion sequences, horizontal gene transfer and codon usage
The mobile elements play an important role in chromosomal rearrangement. To determine how these elements were dispersed among microsyntenic regions, insertion sequence (IS) and transposase locations were analyzed. The S. meliloti chromosome contains 51 IS and 68 transposases belonging to diverse families [21]. Of the total transposases, 40 were found with homologs in plasmids, 20 were common with other rhizobial chromosomes, 6 were denoted as species-specific and 2 were common with the A. tumefaciens linear chromosome. Only 14 pairs of IS/transposases (27% of the total) were found inside microsyntenic regions.
We assessed the influence of horizontal gene transfer (HGT) on the genomic structure of rhizobial genomes (see Methods). Table 2 shows calculated HGT events for each of the classes syntenic, non-syntenic and plasmid homologs. Even though syntenic genes were in the largest class analyzed, they displayed the lowest number of HGT events (Table 2). Predicted HGT rates of non-syntenic genes were 2 to 8 times higher than those of syntenic genes. On the other hand, unlike any other gene class, species-specific genes had the strongest bias toward a low codon richness index in the four Rhizobiales (denoted with crosses, Figure 4 panels a to d). No significant differences were found for syntenic or non-syntenic genes in each organism.
Table 2 Horizontal gene transfer prediction for Rhizobiales.
A. Gene class HGT events* S. meliloti genes HGT ratio
Syntenic 5 1837 0.0027
Non-syntenic 8 378 0.0212
Plasmidic 3 129 0.0233
B. Gene class A. tumefaciens genes
Syntenic 1 1837 0.0005
Non-syntenic 1 378 0.0026
Plasmidic 0 649 0.0000
C. Gene class M. loti genes
Syntenic 7 1624 0.0043
Non-syntenic 6 655 0.0092
Plasmidic 11 924 0.0119
D. Gene class B. melitensis genes
Syntenic 6 1311 0.0046
Non-syntenic 3 414 0.0072
Plasmidic 5 204 0.0245
*Calculated by using the method described in Medrano-Soto et al. [27].
Figure 4 Codon richness index (CRI) for rhizobial genomes. All gene classifications were based on comparisons against S. meliloti, unless explicitly stated otherwise. Panels: (a), S. meliloti (compared with A. tumefaciens). (b), A. tumefaciens. (c), M. loti. (d), B. melitensis. CRIs were calculated according to the method described by Medrano-Soto et al. [27]. Symbols: red circles, syntenic genes. Blue plus signs, non-syntenic. Green squares, homologs in plasmids. Gray crosses, species-specific genes. Horizontal lines denote the species-specific thresholds for low and high CRI.
2) Synteny in Enterobacteriales (gamma proteobacteria)
To assess the adequacy of applying our synteny analysis to another bacterial clade, we chose two members from the Enterobacteriales (gamma proteobacteria), the closely related E. coli and S. typhimurium genomes and defined their orthologous genes. These organisms contained 3092 predicted orthologs and 95% of them (2943) were syntenic. An extensive chromosomal colinearity with few rearrangements was found (data not shown).
Also, we chose a more phylogenetically distant species, Erwinia carotovora subsp. atroseptica to compare with E. coli. Their genomes comprise 4254 and 4477 genes, respectively. They shared 2477 orthologous genes and these represented about half the total genes in each chromosome. In the detection of synteny between these organisms, 1993 genes (80.4% of the orhologs) fulfilled our requirement and the rest, 484, were classified as non-syntenic. When the genes were assigned to microsyntenic regions, 230 regions were found and contained 92.8% (1849) of total syntenic genes and the rest, 7.2% (144 genes), were detected in the non-conserved tracts (see Additional file 5). 172 non-orthologous genes also formed part of microsyntenic regions. The 230 non-conserved regions contained 2233 genes.
To define the conservation of orthology and synteny of Enterobacteriales genomes in comparison with the Rhizobiales, we compared the S. meliloti and E. coli chromosomes. We found 777 predicted orthologs between them, a proportion that represents only a third of the genes shared in the Rhizobiales. By visual examination no synteny was apparent between the Sm and E. coli chromosomes (see Additional file 4 panel d). With an algorithm, only 198 syntenic genes were assigned to 65 microsyntenic regions.
3) Sequence analysis of the chromosomal predicted orthologs
To determine the level of sequence identity among chromosomal predicted orthologs of the Rhizobiales, distribution curves of their translation products were obtained. Two different types of curves were found. By comparing S. meliloti with both A. tumefaciens chromosomes, the syntenic products presented a Gaussian distribution, with a tendency to high identity levels (asymmetry coefficient g1 = -0.48, significant at P < 0.001) (Figure 5 panel a) and a mean value of 71.3%. The non-syntenic products showed a non- asymmetric distribution (g1 = -0.09) with a lower mean value (61.9%) (Figure 5 panel a). Similar curves were found for the chromosomal predicted orthologs of the other comparisons (Additional file 6 panels a and b, for Sm-Ml and Sm-Bm comparisons, respectively). When At replicons were separately compared with Sm, syntenic products of both At-C or At-L showed a similar bias toward high identity levels; in contrast, non-syntenic products of At-L showed a strongly deviated distribution to low identity values (data not shown). The tendency of syntenic products to higher identity levels reflects not only restriction to change but also functional constraints, possibly due to an essential character. Conversely, the lower identity levels of non-syntenic products represent lower restrictions to change and higher functional versatility.
Figure 5 Sequence identity distribution of chromosomal predicted orthologs. Panels: (a), syntenic and non-syntenic products from the S. meliloti-A. tumefaciens (both chromosomes) comparison. Y-axis, relative proportions. (b), syntenic and non-syntenic products from the E. coli-S. typhimurium comparison. Y-axis, number of proteins in each range. (c), syntenic and non-syntenic products from the S. meliloti-E. coli comparison. Y-axis, relative proportions. Red bars, syntenic products. Blue bars, non-syntenic products.
A comparison of E. coli and S. typhimurium genomes (belonging to the Enterobacteriales) was performed. A very asymmetric distribution curve with a tendency to high identity levels for syntenic products was obtained (Figure 5 panel b); it remarkably resembled that obtained in the comparisons of syntenic products of the Rhizobiales. Interestingly, in the comparison of S. meliloti with E. coli both chromosomal non-syntenic and syntenic products had asymmetric curves with strong tendency to low identity levels (Figure 5, panel c).
To obtain a complete view of sequence variation of chromosomal predicted orthologs in the four Rhizobiales, either syntenic or non-syntenic genes were graphed in relation to their identity levels in comparison with S. meliloti. Figure 6 panel a shows the sequence identity of 1038 common syntenic genes in the four Rhizobiales, for Sm-At, Sm-Ml and Sm-Bm comparisons. We referred the comparison to the identity of Sm-Ml genes ordered progressively. We found that each comparison showed a particular clustering, probably related to the phylogenetic distances between these organisms. The Pearson correlation coefficients were r = 0.81 and r = 0.88 for Sm-Bm and Sm-At comparisons, respectively. In the clustering of 98 non-syntenic genes common to the four Rhizobiales (data not shown), lower correlation coefficients were obtained (r = 0.71 and r = 0.79, for Sm-Bm and Sm-At comparisons, respectively). In Figure 6 panel b the sequence identity of 140 common syntenic genes in the Rhizobiales and Enterobacteriales are shown. While the comparison for Rhizobiales follows a similar tendency to the previously observed, for E. coli and Salmonella there is a higher relatedness level. However, when comparing S. meliloti and E. coli a different tendency with very low identity level is observed. A high level of clustering among closely related species belonging to the same subdivision is visible and also, a tendency to reduced relatedness in members of different subdivisions; that is, synteny is a common trait for members of each subdivision.
Figure 6 Sequence identity analysis of common syntenic genes. Panels: (a), in the four Rhizobiales. Comparisons: Red dots, S. meliloti-A tumefaciens. Green dots, S. meliloti-B. melitensis. (b), in Rhizobiales and Enterobacteriales. Comparisons: Red dots, S. meliloti-A tumefaciens. Green dots, S. meliloti-B. melitensis. Magenta dots, E. coli-S. meliloti. Blue dots, E coli-S. typhimurium. Reference line (in black) is the identity percentage of S. meliloti-M. loti syntenic genes in progressive order.
To determine the meaning of sequence differences we analyzed the translated syntenic product ArgC, which participates in the arginine biosynthetic pathway. The alignment presented in Additional file 7 panel a shows 111 positions with identical residues and a range from 33 to 113 different residues, particular for each of the species compared. However, sequences from Brucella melitensis and Brucella suis showed only one difference between them (not shown). Synteny is almost absolute in these organisms (Additional file 4 panel e). Changing residues (possibly species signatures) were dispersed along the sequences and varied according to the identity level.
To determine more comprehensively sequence differences and similarities in Rhizobiales (alpha proteobacteria) and Enterobacteriales (gamma proteobacteria), we selected five organisms from each: S. meliloti, A. tumefaciens, M. loti, B. melitensis and R. palustris for the first and E. coli, S. typhimurium, S. flexneri, Buchnera sp. and E. carotovora for the last. We chose some syntenic products from the arginine biosynthetic pathway, namely ArgB, ArgC, ArgD, ArgF, ArgG, and ArgH. The alignment belonging to ArgC is shown in Additional file 7 panel b. As can be observed, there are sequence identities and differences among species from the same order ("species signatures"), but also similarities between species of different orders, albeit at smaller level. Additionally, we found an interesting pattern: proteins from Enterobacteriales showed an almost uniform level of sequence identity and similarity (about 50 and 70%, on average, respectively), however, sequences from Rhizobiales showed a clear increasing tendency, from 25 to 62% in identity, and from 44 to 81% in similarity (see Additional file 7 panel c). These different profiles possibly are related with the particular conditions of the niches occupied by these organisms.
4) Physical characteristics of the translated syntenic genes
Molecular weight (MW) and isoelectric point (pI) are the main traits for proteomic comparisons. To determine whether syntenic and non-syntenic genes could present differential protein characteristics, we graphed pairs of translated predicted orthologs for both MW and pI parameters. In all comparisons, MW graphs showed lower dispersion than pI ones. Figure 7 shows the pI graph for the S. meliloti-A. tumefaciens comparison (Sm-M. loti and Sm-B. melitensis comparisons are in Additional file 8 panels a and b, respectively). A large group of proteins (82% in each comparison) was located on the diagonal. The correlation coefficient for this group was r = 0.96 in all comparisons. However, the rest of the predicted proteins showed differential pI's and were assigned to sectors (Figure 7). Sector I had acidic proteins in S. meliloti and basic proteins in the organism compared with it. Sector II had basic proteins in S. meliloti and acidic in organism compared to it. In sector III were neutral proteins in S. meliloti and covered all pI range in each of the organisms compared. Proteins in sector IV were neutral in the comparing organism and covered all pI ranges in S. meliloti. In Additional file 9 there is a summary of pI variability of common syntenic products from comparisons with the chromosomes of S. meliloti, A. tumefaciens and M. loti. About 75% of products showed similar pI and 14% and 8% presented high and low variation, respectively. High level is defined as pI variation from acid in one or two organisms and basic in the other, and viceversa. Low level is defined as pI variation from neutral to acid or basic (Additional file 9). Proteins of the sectors mainly corresponded to the functional categories of energy generation, post-translational modification, and transport. In the case of non-syntenic products, a pattern similar to that described above was found (see Additional file 10). Proteins in the diagonal were the most conserved group with subtle pI changes possibly responding to species adaptation, whereas those with deviated pI's may represent a group with higher functional versatility.
Figure 7 Theoretical isoelectric points (pI) of the S. meliloti-A. tumefaciens syntenic products. Dots represent translated products. Red dots, products on the diagonal. Yellow, dots, sector I. Brown dots, sector II. Green dots, sector III. Blue dots, sector IV (see Results). Scales in pH units.
Since the functional categories mentioned above for sectors are known to often interact with the cell membrane, a membrane prediction for all syntenic products from the S. meliloti-A. tumefaciens circular chromosomes comparison was assessed. Strikingly, 790 syntenic products were predicted to contain membranal segments. This amount represents almost all membranal proteins coded in the S. meliloti chromosome, considering that bacterial genomes have 18–28% membranal proteins [49-51]. About 70% of predicted membranal proteins with assigned function belonged to transport, energy generation, post-translational modification, cofactor synthesis, amino acid metabolism and central intermediary metabolism categories (Additional file 11). There are reports about membrane-interacting proteins with functions such as amino acid and cofactor biosynthesis and central intermediary metabolism [52-56].
To determine whether the charged amino acid residues were clustered in proteins from the sectors described above, we selected several proteins from each. The residues determining radical changes in pI were observed scattered along the sequences (data not shown).
5) Functional roles and linkage of the chromosomal predicted orthologs
To define the metabolic participation of the chromosomal predicted orthologs, a functional classification was made with chromosomal syntenic and non-syntenic genes. As shown in Figure 8, syntenic genes of S. meliloti-A. tumefaciens chromosomes contained a high proportion of the most important house-keeping functions. The relative proportion of non-syntenic genes grew with decreasing functional essentiality, for example transport and binding proteins, cellular processes and regulatory functions (Figure 8). Furthermore, the syntenic products covered 85% of the main metabolic pathways as defined in MetaCyc for S. meliloti (see Additional file 12). Similar results in functional coverage were obtained with the other comparisons (see Additional file 13 panels a and b). Common syntenic genes in the four Rhizobiales also included a large fraction of the house-keeping functions (lower segment of the red bars, Figure 8). In the case of the comparison between E. coli and E. carotovora, syntenic genes also covered a high proportion of essential functions, however, the first two positions were occupied by cofactor and nucleotide synthesis (see Additional file 13 panel c). In this regard, it is important to note that E. coli and E. carotovora possess no large plasmids.
Figure 8 Coverage of fuctional classes with syntenic and non-syntenic genes in the S. meliloti-A. tumefaciens comparison. X-axis, functional classes: 1) Transcription, 2) Translation, 3) Fatty acid and phospholipid metabolism, 4) Cell envelope, 5) Biosynthesis of cofactors, prosthetic groups and carriers, 6) Purine, pyrimidine, nucleoside and nucleotide metabolism, 7) DNA metabolism, 8) Amino acid metabolism, 9) Cellular processes, 10) Energy metabolism, 11) Transport and ATP binding proteins, 12) Regulatory functions, 13) Central intermediary metabolism. Red bars, lower fraction: syntenic genes in the four Rhizobiales; upper fraction, syntenic genes in the S. meliloti-A. tumefaciens comparison. Blue bars, lower fraction: non-syntenic genes in the four Rhizobiales; upper fraction: non-syntenic genes in the Sm-At comparison. Y-axis, % of coverage.
When functional classes of genes in the microsyntenic regions were divided into Informational, Operational and Cellular processes superclasses and graphed for the S. meliloti chromosome with a 100 Kb window, an interesting pattern was observed (Figure 9), with the majority of the peaks belonging to a superclass matching with valleys of the other(s). This could represent functionally specialized blocks of chromosomal tracts, which were part of the ancestral rhizobial chromosome. For instance, the existence of genomic domains is accepted in eukaryotes [57].
Figure 9 Distribution of syntenic genes (in regions) by functional superclasses in the S. meliloti chromosome. Blue squares, informational processes. Red triangles, cellular processes. Green diamonds, operational functions. For distribution only microsyntenic regions (from the S. meliloti-A. tumefaciens comparison) with at least two genes belonging to a given class were considered.
The ProLinks program was used to determine how the chromosomal predicted orthologs are functionally related. Functional links were calculated for all genes in the S. meliloti and E. coli chromosomes and then correlated to their neighbors. For the S. meliloti-A.tumefaciens comparison, the microsyntenic regions presented, on average, 3.68 connections per node (a protein in a network), almost twice the value obtained in the non-conserved regions (2.06). From 205 microsyntenic regions, 104 had functional networks; in the case of non-conserved regions, only 35 presented networks. Networks with less than 6 connections were omitted. The networks of microsyntenic regions presented 1057 syntenic genes and from these 795 (75.2%) were organized in operons. 99 non-syntenic genes were in the networks of the non-conserved tracts, and only 56 (56.5%) were in operons. In the case of the synteny comparison between E. carotovora-E. coli (Enterobacteriales), 230 microsyntenic regions were obtained and from these, 161 presented functional networks, with a connectivity average of 5.33 connections/node. The non conserved regions with networks were 71 with a connectivity average of 2.04 connections/node. From 1497 syntenic genes in the networks, 1106 (73.8%) formed part of operons. Network connectivity obtained in S. meliloti and E. coli is shown in the graph of Additional file 14 (panels a and b, respectively). There is a striking difference in connectivity level in the networks from syntenic (gray bars) or non-conserved regions (black bars). The connectivity levels in syntenic vs non-conserved regions in both organisms were similar using the STRING program (data not shown).
Discussion
The comparative genomic analysis reported here was useful in finding interesting gene properties. Orthologs with conserved replicon and neighborhood were the principal component of the chromosomes. Compared with non-syntenic genes, syntenic ones had higher identity levels, lower horizontal gene transfer (HGT) rates, showed strongly organized structures as operons and microsyntenic regions and a relative absence of mobile elements. Thus, the syntenic genes can be considered as the chromosomal backbone of the order. Plasmidic homologs were scattered on the chromosomes, and higher HGT rate and linkage to transposases support their extrachromosomal origin. Species-specific genes had the lowest codon richness index, and possibly were acquired in the evolutionary history of each of the species.
In this way, a rhizobial chromosomal origin can be envisioned. The chromosomal orthologs were the gene set derived from the common cenancestor. From these, syntenic genes conserved a relative chromosomal order (and operonic organization) and encode the essential functions of the cell; non-syntenic genes lost the clustering and possibly some came from HGT events. The plasmidic homologs were obtained possibly by mobilization throughout replicons, a nonrare process in the rhizobial phylogenetic branch. The species-specific genes represent the particular gene set of the species and are the most intriguing group due to their unknown functional roles and origin. Work with members of the last group will help to define traits not shared with other species.
The Rhizobiales species analyzed showed a striking proportion of orthologous genes, mainly chromosomal syntenic; non-syntenic genes were found in lower proportion. A large fraction of the first class was common in the four organisms. Syntenic genes had a strong tendency to form operons and almost all were clustered in microsyntenic regions. Additionally, these operons were conserved in pairs of organisms. Therefore, a strong restriction for chromosomal rearrangement is visible. Given that these organisms cover a wide spectrum of environmental distributions, from plant rhizosphere to animal host, the conserved chromosomal tracts may be important to determine the metabolic properties common to the order. Similar results were observed in the Enterobacteriales comparison. In a recent report, using computational inference, Boussau et al. [58] proposed a common ancestral set of about 3000 genes for proteobacterial genomes.
Although rhizobial genomes shared common traits, important differences were also observed. For example, abundance of plasmidic homologs and species-specific genes in A. tumefaciens (linear), B. melitensis II and M. loti chromosomes confirmed their complex evolutionary histories [20,22,59,60]. The M. loti chromosome presents intensive incorporation of foreign genes by horizontal transfer, such as those belonging to the symbiotic island [59]. In regard to species-specific genes, a large number presented low codon richness index. This category will be reduced with incorporation of other rhizobial genomes into the databases. For example, R. leguminosarum biovar viciae 3841, R. tropici PRF81, R. sp. ANU265 and R. etli CFN42 genomes soon will be available [1]. Mobilization elements participate in chromosomal rearrangement and are abundant in Rhizobiales. These elements can decompose the microsyntenic regions where they are. In this way, all synteny approaches consist of snapshots in chromosomal evolution.
From the comparison of sequence identities among chromosomal predicted orthologs in the rhizobial species analyzed, an interesting characteristic was the differential distribution curves obtained. The asymmetric curves of syntenic products, deviated to high identity levels (with peaks at 70–75%, Figure 5 panel a and Additional file 6 panels a and b), possibly reveal their essentiality and can be compared with those from gamma proteobacteria with high identity (with peak at 95%, Figure 5 panel b). Conversely, non-syntenic genes had curves with lower sequence identity levels reflecting a higher functional versatility. In the case of S. meliloti-E. coli comparison, the identical curves for syntenic and non-syntenic genes with the majority at very low identity values (35%, Figure 5 panel c) reflect their greater phylogenetic distances.
The high identity of syntenic genes indirectly reveals their essential character; for the non-syntenic genes the low identity could represent adaptability to the ecological niche of the species. The identity relatedness in the syntenic genes among rhizobial species (Figure 6 panel a) revealed a cohesively evolved group; additionally the sequence differences were reflected in the theoretical pI plots of the proteins encoded by these genes, with a majority in the diagonal and the rest in sectors with strong pI deviation. Species signatures of the sequences (see Additional file 7) showed a differential level of changed residues and these could represent functional adaptation to a niche; this proposal is supported with the almost identical sequences of B. melitensis and B. suis. On the other hand, invariant peptides perhaps contribute to structural conformation [61]. Experiments in progress in our lab will determine the validity of our proposal. Recently, a conservative change which altered the function of the transcriptional regulator BosR [62], and pathogenic differences in enteric bacteria due to the expression of PmrD regulators with divergent sequences [63] were reported.
A typical trait of Enterobacteriales is its pathogenic character, and this means a very intimate, frequent contact with their hosts and almost constant, homogeneous conditions: the extreme case is that of Buchnera sp., an aphid-obligate symbiont. In contrast, Rhizobiales are commonly found in the soil in saprophytic living style, and occasionally they associate with hosts, and therefore face more heterogeneous, variable environments. These features were reflected in the sequence comparisons of proteins from the arginine biosynthetic pathway (see Additional file 7). The syntenic gene organization was different in the Rhizobiales compared with Enterobacteriales, however it is important to note the high orthology and synteny degrees between members of each clade, the essentiality of functions covered by the syntenic genes and the high functional linkage in the microsyntenic regions in each chromosome (see below). From the previous observations we can obtain a general trend: bacterial clades present a particular chromosomal gene arrangement and such plasticity possibly was selected in relation to the niches occupied by these organisms.
As have others, some time ago we found a proteomic bimodal distribution when MW and pI were plotted for rhizobial gene products. The main fraction was located at acid and basic pI's, in a shape resembling butterfly's wings and the body, at neutral pI, presenting a low number of proteins. Recently Knight et al. [64], in a vast genomic study, reported similar results and additionally, related proteome similarities to shared metabolic features. In this respect, we plotted pI of syntenic products from pair comparisons of studied species in order to detect proteomic similarities in these organisms. It was possible to analyze the pI variability in each of the common syntenic products; however, it is necessary to carry out experimentation to determine the biological role of pI variation. The overrepresented membranal fraction in the syntenic products must be further explored to determine whether they respond to different extracellular/intracellular signals. A rapid evolution for the membranal proteomic fraction was suggested in the same study [64].
The high proportion of syntenic operons in the microsyntenic regions and the duplicated connections per gene in the network-forming regions (in regard to non-conserved regions), in Rhizobiales but also in Enterobacteriales, supported the functional linkage and interaction of these genes in the conserved tracts (see Additional file 14), and this is a factor which could help to define the role of selective pressure in maintaining the gene order.
Functional characterization of the predicted orthologs deepened our understanding of their cellular roles. In both Rhizobiales and Enterobacteriales, a clear essentiality was observed for chromosomal syntenic genes, agreeing with their sequence restrictions for change. Non-syntenic genes, on the other hand, appeared abundant in functions granting metabolic versatility to the cell. By calculating nonsynonymous/synonymous substitution rates, other authors have shown that in bacteria most conserved genes cover the essential functions of the cell [65].
Conclusion
Our synteny analysis defined a multi-level gene organization in the bacterial chromosome. Restriction of sequence variation in these genes, with clear essential functional roles, appeared extended to the conservation of chromosomal arrangement. In this way, synteny possibly has an important biological significance in these organisms.
Methods
Identification of orthologs
Available genome sequences of the fast growing Rhizobiales, S. meliloti 1021 [accession number GenBank: NC_003047], A. tumefaciens C58 (Cereon) [GenBank: NC_003062 and NC_003063], M. loti MAFF303099 [GenBank: NC_002678] and B. melitensis 16 M [GenBank: NC_003317 and NC_003318] were obtained from the Genome division of the NCBI Entrez system [66]. Genomes of E. coli K12, S. typhimurium LT2 and Erwinia carotovora subsp. atroseptica SCRI1043 [accession numbers GenBank: NC_00913NC_003197, and NC_004547], were used for synteny and functional analysis. The genome of B. suis 1330 [accession numbers GenBank: NC_004310 and NC_004311 for chromosomes I and II, respectively] was used for synteny and sequence identity analysis. Genomes of Bradyrhizobium japonicum USDA110 [accession number GenBank: NC_004463] and Rhodopseudomonas palustris CGA009, also belonging to the alpha proteobacteria, were not considered in the main analysis because of their more distant phylogenetic relationship with the fast-growing rhizobia. To obtain a comprehensive view of shared genes among Rhizobiales, we differentiated genes by orthology and their presence in the same or different replicons in the analyzed species. Chromosomal orthologs were assigned by the best bidirectional hit between pairs of organisms, using the Fasta34 program [67]. Unidirectional best hits (homologs) were considered to cover complete chromosomal gene number (see Results) and for detection of chromosomal genes with plasmidic homologs. Parameters were: an identity of at least 50%, overlapping by at least in 150 nt and an expectance (E) score of <10-3. The base organism for comparisons was S. meliloti 1021. GenBank accession numbers of proteins used for alignments in the order ArgB, ArgC, ArgD, ArgF, ArgG and ArgH, were as follows. Rhizobiales. R. palustris: NP_945982.1, NP_947833.1, NP_950107.1, NP_950106.1, NP_945745.1, NP_950077.1; B. melitensis: NP_541250.1, NP_540088.1, NP_540538.1, NP_540537.1, NP_540787.1, NP_539004.1; M. loti: NP_105609.1, NP_108547.1, NP_106269.1, NP_106270.1, NP_105253.1, NP_104594.1; A. tumefaciens: NP_353412.1, NP_354256.1, NP_353456.1, NP_353457.1, NP_355604.1, NP_357013.1; S. meliloti: NP_384545.1, NP_385346.1, NP_384623.1, NP_384624.1, NP_387315.1, NP_386753.1. Enterobacteriales. Buchnera sp.: NP_239886.1, NP_239885.1, NP_240341.1, NP_240186.1, NP_239887.1, NP_239888.1; E. coli: NP_418394.1, NP_418393.1, NP_417818.1, NP_414807.1, NP_417640.1, NP_418395.1; E. carotovora: YP_048320.1, YP_048319.1, YP_052152.1, YP_048510.1, YP_048232.1, YP_048321.1S. typhimurium: NP_457937.1, NP_457938.1, NP_458434.1, NP_458882.1, NP_457671.1, NP_457936.1; S. flexneri: NP_838925.1, NP_838926.1, NP_839526.1, NP_839629.1, NP_838682.1, NP_838924.1. All data sets are available on request.
Procedures for detection of syntenic genes, microsyntenic region formation and operon similarity
To consider chromosomal orthologs as syntenic among pairs of organisms, at least two genes must remain contiguous in both chromosomes. Microsyntenic region formation and extension fulfilled the following criterion: a pair of predicted orthologs separated from at least one other by no more than three genes (from the rest of categories). The minimal region was formed by a stretch containing three syntenic genes. Operon prediction was performed as reported by Moreno-Hagelsieb and Collado-Vides [68]. Rearrangements were graphed using initial positions of the microsyntenic regions in each chromosome. The syntenic regions of M. loti and B. melitensis I chromosomes were graphed so as to increase the colinearity in these replicons. The M. loti chromosome was segmented into two halves at 3.5 Mb position and the fragment covering from 3.5 to 7.0 Mb was located in the first position and then both halves were aligned with microsyntenic regions of S. meliloti. In the case of Brucella chromosome I, the origin was inverted. Graphs were obtained using the GenVision program (DNAStar Inc., Madison, WI). For operon similarity calculation, a limit of three different genes in each operon was allowed. All data sets are available on request.
Detection of horizontal gene transfer
Prediction of horizontally transferred genes in the Rhizobiales genomes was performed using the method described by Medrano-Soto et al. [27]. Briefly, it on based in similar gene length, maximum global protein identity, conflicting phylogenies and codon usage of xenologous genes. In the case of A. tumefaciens, this rate was calculated using sequences and annotation obtained from the U. of Washington Sequencing Project [45]. Species-specific (or orphan) genes were not considered by two reasons: (1) the lack of orthologs in other genomes precluded phylogenetic analysis, and (2) impossibility of correlating these genes to a synteny.
Sequence comparison of chromosomal orthologs in Rhizobiales
The identity of peptidic sequences of chromosomal predicted orthologs were used to graph the distribution curves. The asymmetry of distribution curves, or skewness, was calculated by the asymmetry coefficient of Pearson (g1) as described elsewhere [69]. To correlate nucleotide sequence identities of the chromosomal predicted orthologs in the four rhizobial genomes, the gene identity of the predicted orthologs from the S. meliloti-M. loti comparison was graphed in progressive order. Then, corresponding predicted orthologs of the other comparisons were located at their corresponding identity percentages. Correlation coefficient values were calculated by the Pearson method. Plasmidic homologs and species-specific genes were not graphed because they have no counterparts in the pairs of analyzed genomes.
Theoretical proteome and transmembranal protein prediction
Theoretical proteomes were obtained by calculating molecular weight and isoelectric point for each translated chromosomal predicted ortholog. Both parameters were estimated with the pI/MW prediction tool of the Laquip Proteomic Team page [70]. To determine the set of orthologs coding proteins predicted to interact with the cell membrane, we used the TMAP program, version 46 [71], available in the EMBOSS package (European Molecular Biology Open Software Suite, [72]). Correlation coefficient values were obtained with the Pearson method. Alignments were performed with ClustalW [73].
Functional classification of chromosomal predicted orthologs
Chromosomal predicted orthologs were assigned to the functional classes used for Agrobacterium tumefaciens C58 in the U. of Washington genome report [45]. For the distribution of functions coded in the microsyntenic regions along the S. meliloti chromosome, classes were assigned into Operational (including amino acid, fatty acid, carbohydrate and nucleotide metabolism, energy generation, central intermediary metabolism, transport and cofactor synthesis), Informational (DNA metabolism, transcription, translation and regulatory functions), and Cellular processes (cell envelope, cell division, secretion and chemotaxis) superclasses. This grouping, except for Cellular processes, is similar to that of Rivera et al. [74]. For functional relationship inference, the ProLinks [34] and STRING databases [35] were used with permission. The confidence levels used were 0.6 and 0.9, respectively. Resulting networks with ProLinks with less than 6 links were omitted from the count. Networks were constructed with the Pajek Program (written by A. Vlado), version 1.02, available in the web [75]. Assignment of metabolic pathways was performed using the MetaCyc database [76], with permission.
Authors' contributions
GG and AA performed the computer predictions. HP made the functional analysis, interpreted the data and wrote the paper. RD and MAV participated in the work design. AM-S made the HGT and CRI analysis. JM conceived and directed the project. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Schematic representation of the Rhizobiales chromosomes in comparison with S. meliloti, according to the gene classification of predicted orthologs and homologs. Panels: (a), S. meliloti-A. tumefaciens comparison. (b), S. meliloti-M. loti comparison. (c), S. meliloti-B. melitensis comparison. Red striped bars, syntenic genes with the organism in comparison. Blue striped bars, non-syntenic genes with the organism in comparison. White bars, homologs with other Rhizobiales chromosomes (for S. meliloti, compare with Fig. 1, white fraction). Green bars, homologs in plasmids. Gray bars, species-specific genes. In panels a and c, the S. meliloti chromosome shows syntenic and non-syntenic genes with both replicons of the organisms under comparison. Red striped bars, syntenic genes, lower fraction: with (a) At-C and (c) Bm-I chromosomes; upper fraction: with (a) At-L and (c) Bm-II chromosomes. Blue striped bars, non-syntenic genes, lower fraction: with (a) At-C and (c) Bm-I chromosomes; upper fraction: with (a) At-L and (c) Bm-II chromosomes.
Click here for file
Additional File 2
Synteny histogram of S. meliloti in comparison with M. loti chromosome. Red bars, syntenic genes. Surrounded with yellow boxes, microsyntenic regions. Microsyntenic regions are denoted by letters (and numbers) in progressive order. Blue bars, non-syntenic genes. Green bars, homologs in plasmids. Gray bars, species-specific genes. White bars, homologs with other Rhizobiales chromosomes. Direction of transcription is denoted by upper (plus) or lower (minus) positions in respect to central line. Predicted operons are denoted by red arrows. Scale in bp.
Click here for file
Additional File 3
Synteny histogram of S. meliloti in comparison with B. melitensis chromosomes. Red bars, syntenic genes. Surrounded with yellow boxes, microsyntenic regions with B. melitensis chromosome I (Bm-I). Surrounded with light green boxes, microsyntenic regions with B. melitensis chromosome II (Bm-II). Microsyntenic regions are denoted by letters (and numbers) in progressive order. Dark blue bars, non-syntenic genes with Bm-I. Light blue bars, nonsyntenic genes with Bm-II. Green bars, homologs in plasmids. Gray bars, species-specific genes. White bars, homologs with other Rhizobiales chromosomes. Direction of transcription is denoted by upper (plus) or lower (minus) positions in respect to central line. Predicted operons are denoted by red arrows. Scale in bp.
Click here for file
Additional File 4
Synteny of Rhizobiales and Enterobacteriales. Panels: (a), S. meliloti-A. tumefaciens circular chromosomes comparison. (b), S. meliloti-M. loti comparison. (c), S. meliloti-B. melitensis chromosome I comparison. (d), S. meliloti-E. coli comparison. (e), B. suis-B. melitensis chromosomes I comparison. Red dots, syntenic genes. Blue dots, non-syntenic genes. Scales in bp.
Click here for file
Additional File 5
Synteny histogram of E. coli in comparison with E. carotovora chromosome. Red bars, syntenic genes. Surrounded with yellow boxes, microsyntenic regions with E. carotovora chromosome Microsyntenic regions are denoted by letters (and numbers) in progressive order. Blue bars, non-syntenic genes. Gray bars, non-orthologous genes. Scale in bp.
Click here for file
Additional File 6
Sequence identity distribution of chromosomal translated orthologs. Panels: (a), syntenic and non-syntenic products from the S. meliloti-M. loti comparison. (b), syntenic and non-syntenic products from the S. meliloti-B. melitensis (chromosomes I and II) comparison. Y-axis, relative proportions. Red bars, syntenic genes. Blue bars, non-syntenic genes. Y-axis, relative proportions.
Click here for file
Additional File 7
Sequence alignments and data from the alignments of proteins from the arginine biosynthetic pathway in Rhizobiales and Enterobacteriales. Panels: (a), ArgC in Rhizobiales. Identical residues for aech position are marked with yellow. Least abundant residues for a given position are denoted with an specific color for each of the species: dark blue, differences in R. palustris; green, differences in B. melitensis; red, differences in M. loti; gray, differences in A. tumefaciens; violet, S. meliloti. (b), ArgC in Rhizobiales and Enterobacteriales. Identical residues for aech position are marked with yellow. Least abundant residues for a given position are denoted with an specific color for each of the species: Rhizobiales, same code of panel (a). Enterobacteriales: brown, Buchnera; pink, E. carotovora; blue, S. typhimurium. E. coli and S. flexneri, none. (c), data of the identity (*) and similarity (:*) in residues and in percentage (bold) of the alignments of the proteins in Rhizobiales and Enterobacteriales.
Click here for file
Additional File 8
Theoretical isoelectric points (pI) of syntenic products. Panels: (a), S. meliloti-M. loti comparison. (b), S. meliloti-B. melitensis (chromosomes I and II) comparison. Dots represent translated products. Scales in pH units.
Click here for file
Additional File 9
Summary of proteins with differential pI's from comparisons with chromosomes of S. meliloti, A. tumefaciens (circular) and M. loti.
Click here for file
Additional File 10
Theoretical isoelectric points (pI) of nonsyntenic products. Panels: (a), S. meliloti-A. tumefaciens (both chromosomes) comparison. (b), S. meliloti-M. loti comparison. (c), S. meliloti-B. melitensis (both chromosomes) comparison. Dots represent translated products. Scales in pH units.
Click here for file
Additional File 11
Functional categories of syntenic products of the membranal prediction of S. meliloti- A. tumefaciens (circular chromosome) comparison.
Click here for file
Additional File 12
Metabolic pathways covered by syntenic products of the S. meliloti-A. tumefaciens comparison. Scheme belongs to the MetaCyc pathways for S. meliloti chromosome, used with permission. Highlighted with green, reactions covered with syntenic products.
Click here for file
Additional File 13
Coverage of functional classes with syntenic and non-syntenic genes. Panels: (a), S. meliloti-M. loti comparison. Classes: 1) Translation, 2) Transcription, 3) Purine, pyrimidine, nucleoside and nucleotide metabolism, 4) Cellular processes, 5) Energy metabolism, 6) Cell envelope, 7) Fatty acid and phospholipid metabolism, 8) Biosynthesis of cofactors, prosthetic groups and carriers, 9) Transport and ATP binding proteins, 10) Amino acid metabolism, 11) DNA metabolism, 12) Regulatory functions, 13) Central intermediary metabolism. (b), S. meliloti-B. melitensis comparison. Classes: 1) Transcription, 2) Translation, 3) Cellular processes, 4) Biosynthesis of cofactors, prosthetic groups and carriers, 5) Cell envelope, 6) Energy metabolism, 7) Fatty acid and phospholipid metabolism, 8) Purine, pyrimidine, nucleoside and nucleotide metabolism, 9) Amino acid metabolism, 10) Transport and ATP binding proteins, 11) DNA metabolism, 12) Regulatory functions, 13) Central intermediary metabolism. (c), E. coli-E. carotovora comparison. Classes: 1) Biosynthesis of cofactors, prosthetic groups and carriers, 2) Purine, pyrimidine, nucleoside and nucleotide metabolism, 3) Translation, 4) Fatty acid and phospholipid metabolism, 5) Transcription, 6) Cellular processes, 7) DNA metabolism, 8) Energy metabolism, 9) Amino acid metabolism. 10) Cell envelope, 11) Regulatory functions, 12) Transport and ATP binding proteins, 13) Central intermediary metabolism. Note that order of classes is different to that in Fig. 8. Red bars, syntenic genes. Blue bars, non-syntenic genes.
Click here for file
Additional File 14
Connectivity values from networks formed by microsyntenic and non-conserved regions in (a) S. meliloti (in comparison with A. tumefaciens) and (b) E. coli (in comparison with E. carotovora). Y-axis, connections per network. First syntenic networks, with 1060 (S. meliloti) and 810 (E. coli) connections, were omitted for clarity. Arranged in decrecent connectivity order. Gray bars, microsyntenic regions. Black bars, non-conserved regions. Successive networks, with connectivity values lower than 6, were omitted.
Click here for file
Acknowledgements
This work was partially supported by the grant 028 from the Consejo Nacional de Ciencia y Tecnología-México.
We wish thank to Gabriel Moreno-Hagelsieb for providing the operon prediction, César Bonavides for help with metabolic pathway assignment, Rafael Palacios, Michael Dunn and Yolanda Mora for manuscript review, and an anonymous referee for valuable suggestions.
==== Refs
Genomes Online Database
Itoh T Martin W Nei M Acceleration of genomic evolution caused by enhanced mutation rate in endocellular symbionts Proc Natl Acad Sci U S A 2002 99 12944 12948 12235368 10.1073/pnas.192449699
Jordan IK Rogozin IB Wolf YI Koonin EV Microevolutionary Genomics of Bacteria Theoretical Population Biology 2002 61 435 12167363 10.1006/tpbi.2002.1588
Rocha EP Order and disorder in bacterial genomes Curr Opin Microbiol 2004 7 519 527 15451508 10.1016/j.mib.2004.08.006
Korbel JO Jensen LJ von Mering C Bork P Analysis of genomic context: prediction of functional associations from conserved bidirectionally transcribed gene pairs Nat Biotechnol 2004 22 911 917 15229555 10.1038/nbt988
Tamames J Evolution of gene order conservation in prokaryotes Genome Biol 2001 2 R0020 10.1186/gb-2001-2-6-research0020
Zheng XH Lu F Wang ZY Zhong F Hoover J Mural R Using shared genomic synteny and shared protein functions to enhance the identification of orthologous gene pairs Bioinformatics 2005 21 703 710 15458983 10.1093/bioinformatics/bti045
Nakamura Y Nishio Y Ikeo K Gojobori T The genome stability in Corynebacterium species due to lack of the recombinational repair system Gene 2003 317 149 155 14604803 10.1016/S0378-1119(03)00653-X
Shigenobu S Watanabe H Hattori M Sakaki Y Ishikawa H Genome sequence of the endocellular bacterial symbiont of aphids Buchnera sp. APS Nature 2000 407 81 86 10993077 10.1038/35024074
Cannon SB McCombie WR Sato S Tabata S Denny R Palmer L Katari M Young ND Stacey G Evolution and microsynteny of the apyrase gene family in three legume genomes Mol Genet Genomics 2003 270 347 361 14598165 10.1007/s00438-003-0928-x
Gualtieri G Bisseling T Microsynteny between the Medicago truncatula SYM2-orthologous genomic region and another region located on the same chromosome arm Theor Appl Genet 2002 105 771 779 12582492 10.1007/s00122-002-0949-6
Bankier AT Spriggs HF Fartmann B Konfortov BA Madera M Vogel C Teichmann SA Ivens A Dear PH Integrated mapping, chromosomal sequencing and sequence analysis of Cryptosporidium parvum Genome Res 2003 13 1787 1799 12869580
Bannantine JP Zhang Q Li LL Kapur V Genomic homogeneity between Mycobacterium avium subsp. avium and Mycobacterium avium subsp. paratuberculosis belies their divergent growth rates BMC Microbiol 2003 3 10 12740027 10.1186/1471-2180-3-10
Buchrieser C Rusniok C Kunst F Cossart P Glaser P Comparison of the genome sequences of Listeria monocytogenes and Listeria innocua: clues for evolution and pathogenicity FEMS Immunol Med Microbiol 2003 35 207 213 12648839 10.1016/S0928-8244(02)00448-0
Eppinger M Baar C Raddatz G Huson DH Schuster SC Comparative analysis of four Campylobacterales Nat Rev Microbiol 2004 2 872 885 15494744 10.1038/nrmicro1024
Horimoto K Fukuchi S Mori K Comprehensive comparison between locations of orthologous genes on archaeal and bacterial genomes Bioinformatics 2001 17 791 802 11590096 10.1093/bioinformatics/17.9.791
Moreira LM de Souza RF Almeida NFJ Setubal JC Oliveira JC Furlan LR Ferro JA da Silva AC Comparative genomics analyses of citrus-associated bacteria Annu Rev Phytopathol 2004 42 163 184 15283664 10.1146/annurev.phyto.42.040803.140310
Read TD Brunham RC Shen C Gill SR Heidelberg JF White O Hickey EK Peterson J Utterback T Berry K Bass S Linher K Weidman J Khouri H Craven B Bowman C Dodson R Gwinn M Nelson W DeBoy R Kolonay J McClarty G Salzberg SL Eisen J Fraser CM Genome sequences of Chlamydia trachomatis MoPn and Chlamydia pneumoniae AR39 Nucleic Acids Res 2000 28 1397 1406 10684935 10.1093/nar/28.6.1397
Krawiec S Riley M Organization of the bacterial chromosome Microbiol Rev 1990 54 502 539 2087223
Paulsen IT Seshadri R Nelson KE Eisen JA Heidelberg JF Read TD Dodson RJ Umayam L Brinkac LM Beanan MJ Daugherty SC Deboy RT Durkin AS Kolonay JF Madupu R Nelson WC Ayodeji B Kraul M Shetty J Malek J Van Aken SE Riedmuller S Tettelin H Gill SR White O Salzberg SL Hoover DL Lindler LE Halling SM Boyle SM Fraser CM The Brucella suis genome reveals fundamental similarities between animal and plant pathogens and symbionts Proc Natl Acad Sci U S A 2002 99 13148 13153 12271122 10.1073/pnas.192319099
Galibert F Finan TM Long SR Puhler A Abola P Ampe F Barloy-Hubler F Barnett MJ Becker A Boistard P Bothe G Boutry M Bowser L Buhrmester J Cadieu E Capela D Chain P Cowie A Davis RW Dreano S Federspiel NA Fisher RF Gloux S Godrie T Goffeau A Golding B Gouzy J Gurjal M Hernandez-Lucas I Hong A Huizar L Hyman RW Jones T Kahn D Kahn ML Kalman S Keating DH Kiss E Komp C Lelaure V Masuy D Palm C Peck MC Pohl TM Portetelle D Purnelle B Ramsperger U Surzycki R Thebault P Vandenbol M Vorholter FJ Weidner S Wells DH Wong K Yeh KC Batut J The composite genome of the legume symbiont Sinorhizobium meliloti Science 2001 293 668 672 11474104
Goodner B Hinkle G Gattung S Miller N Blanchard M Qurollo B Goldman BS Cao Y Askenazi M Halling C Mullin L Houmiel K Gordon J Vaudin M Iartchouk O Epp A Liu F Wollam C Allinger M Doughty D Scott C Lappas C Markelz B Flanagan C Crowell C Gurson J Lomo C Sear C Strub G Cielo C Slater S Genome Sequence of the Plant Pathogen and Biotechnology Agent Agrobacterium tumefaciens C58 Science 2001 294 2323 2328 11743194 10.1126/science.1066803
Doolittle WF Lateral genomics Trends Cell Biol 1999 9 M5 8 10611671 10.1016/S0962-8924(99)01664-5
Ochman H Lawrence JG Groisman EA Lateral gene transfer and the nature of bacterial innovation Nature 2000 405 299 304 10830951 10.1038/35012500
Koski LB Morton RA Golding GB Codon bias and base composition are poor indicators of horizontally transferred genes Mol Biol Evol 2001 18 404 412 11230541
Wang B Limitations of compositional approach to identifying horizontally transferred genes J Mol Evol 2001 53 244 250 11523011 10.1007/s002390010214
Medrano-Soto A Moreno-Hagelsieb G Vinuesa P Christen JA Collado-Vides J Successful lateral transfer requires codon usage compatibility between foreign genes and recipient genomes Mol Biol Evol 2004 21 1884 1894 15240837 10.1093/molbev/msh202
Hao W Golding GB Patterns of bacterial gene movement Mol Biol Evol 2004 21 1294 1307 15115802 10.1093/molbev/msh129
Jain R Rivera MC Moore JE Lake JA Horizontal gene transfer accelerates genome innovation and evolution Mol Biol Evol 2003 20 1598 1602 12777514 10.1093/molbev/msg154
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO Assigning protein functions by comparative genome analysis: protein phylogenetic profiles Proc Natl Acad Sci U S A 1999 96 4285 4288 10200254 10.1073/pnas.96.8.4285
Overbeek R Fonstein M D'Souza M Pusch GD Maltsev N The use of gene clusters to infer functional coupling Proc Natl Acad Sci U S A 1999 96 2896 2901 10077608 10.1073/pnas.96.6.2896
Marcotte EM Pellegrini M Ho-Leung N Rice D Yeates T Elsenberg D Detecting protein function and protein-protein interactions from genome sequences Science 1999 285 751 753 10427000 10.1126/science.285.5428.751
Yanai I DeLisi C The society of genes: networks of functional links between genes from comparative genomics Genome Biol 2002 3 R0064 10.1186/gb-2002-3-11-research0064
Bowers PM Pellegrini M Thompson MJ Fierro J Yeates TO Eisenberg D Prolinks: a database of protein functional linkages derived from coevolution Genome Biol 2004 5 R35 15128449 10.1186/gb-2004-5-5-r35
von Mering C Zdobnov EM Tsoka S Ciccarelli FD Pereira-Leal JB Ouzounis CA Bork P Genome evolution reveals biochemical networks and functional modules Proc Natl Acad Sci U S A 2003 100 15428 15433 14673105 10.1073/pnas.2136809100
Freiberg C Fellay R Bairoch A Broughton WJ Rosenthal A Perret X Molecular basis of symbiosis between Rhizobium and legumes Nature 1997 387 394 401 9163424 10.1038/387394a0
Gonzalez V Bustos P Ramirez-Romero MA Medrano-Soto A Salgado H Hernandez-Gonzalez I Hernandez-Celis JC Quintero V Moreno-Hagelsieb G Girard L Rodriguez O Flores M Cevallos MA Collado-Vides J Romero D Davila G The mosaic structure of the symbiotic plasmid of Rhizobium etli CFN42 and its relation to other symbiotic genome compartments Genome Biol 2003 4 R36 12801410 10.1186/gb-2003-4-6-r36
Gottfert M Rothlisberger S Kundig C Beck C Marty R Hennecke H Potential symbiosis-specific genes uncovered by sequencing a 410-kilobase DNA region of the Bradyrhizobium japonicum chromosome J Bacteriol 2001 183 1405 1412 11157954 10.1128/JB.183.4.1405-1412.2001
Hirsch AM Drake D Jacobs TW Long SR Nodules are induced on alfalfa roots by Agrobacterium tumefaciens and Rhizobium trifolii containing small segments of the Rhizobium meliloti nodulation region J Bacteriol 1985 161 223 230 3968028
Martinez E Palacios R Sanchez F Nitrogen-fixing nodules induced by Agrobacterium tumefaciens harboring Rhizobium phaseoli plasmids J Bacteriol 1987 169 2828 2834 3584072
Weller SA Stead DE Young JP Acquisition of an Agrobacterium Ri plasmid and pathogenicity by other alpha-Proteobacteria in cucumber and tomato crops affected by root mat Appl Environ Microbiol 2004 70 2779 2785 15128532 10.1128/AEM.70.5.2779-2785.2004
Moran NA Wernegreen JJ Lifestyle evolution in symbiotic bacteria: insights from genomics Trends In Ecology And Evolution 2000 15 321 326 10884696 10.1016/S0169-5347(00)01902-9
Kaneko T Nakamura Y Sato S Asamizu E Kato T Sasamoto S Watanabe A Idesawa K Ishikawa A Kawashima K Kimura T Kishida Y Kiyokawa C Kohara M Matsumoto M Matsuno A Mochizuki Y Nakayama S Nakazaki N Shimpo S Sugimoto M Takeuchi C Yamada M Tabata S Complete genome structure of the nitrogen-fixing symbiotic bacterium Mesorhizobium loti DNA Res 2000 7 331 338 11214968 10.1093/dnares/7.6.331
Kaneko T Nakamura Y Sato S Minamisawa K Uchiumi T Sasamoto S Watanabe A Idesawa K Iriguchi M Kawashima K Kohara M Matsumoto M Shimpo S Tsuruoka H Wada T Yamada M Tabata S Complete genomic sequence of nitrogen-fixing symbiotic bacterium Bradyrhizobium japonicum USDA110 DNA Res 2002 9 189 197 12597275 10.1093/dnares/9.6.189
Wood DW Setubal JC Kaul R Monks DE Kitajima JP Okura VK Zhou Y Chen L Wood GE Almeida NFJ Woo L Chen Y Paulsen IT Eisen JA Karp PD Bovee DS Chapman P Clendenning J Deatherage G Gillet W Grant C Kutyavin T Levy R Li MJ McClelland E Palmieri A Raymond C Rouse G Saenphimmachak C Wu Z Romero P Gordon D Zhang S Yoo H Tao Y Biddle P Jung M Krespan W Perry M Gordon-Kamm B Liao L Kim S Hendrick C Zhao ZY Dolan M Chumley F Tingey SV Tomb JF Gordon MP Olson MV Nester EW The genome of the natural genetic engineer Agrobacterium tumefaciens C58 Science 2001 294 2317 2323 11743193 10.1126/science.1066804
DelVecchio VG Kapatral V Redkar RJ Patra G Mujer C Los T Ivanova N Anderson I Bhattacharyya A Lykidis A Reznik G Jablonski L Larsen N D'Souza M Bernal A Mazur M Goltsman E Selkov E Elzer PH Hagius S O'Callaghan D Letesson JJ Haselkorn R Kyrpides N Overbeek R The genome sequence of the facultative intracellular pathogen Brucella melitensis Proc Natl Acad Sci U S A 2002 99 443 448 11756688 10.1073/pnas.221575398
Larimer FW Chain P Hauser L Lamerdin J Malfatti S Do L Land ML Pelletier DA Beatty JT Lang AS Tabita FR Gibson JL Hanson TE Bobst C Torres JL Peres C Harrison FH Gibson J Harwood CS Complete genome sequence of the metabolically versatile photosynthetic bacterium Rhodopseudomonas palustris Nat Biotechnol 2004 22 55 61 14704707 10.1038/nbt923
COGs: Phylogenetic classification of protein encoded in complete genomes database
Gerstein M Patterns of protein-fold usage in eight microbial genomes: a comprehensive structural census Proteins 1998 33 518 534 9849936 10.1002/(SICI)1097-0134(19981201)33:4<518::AID-PROT5>3.0.CO;2-J
Gerstein M Levitt M A structural census of the current population of protein sequences Proc Natl Acad Sci U S A 1997 94 11911 11916 9342336 10.1073/pnas.94.22.11911
Krogh A Larsson B von Heijne G Sonnhammer EL Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes J Mol Biol 2001 305 567 580 11152613 10.1006/jmbi.2000.4315
Fernandez-Herrero LA Badet-Denisot MA Badet B Berenguer J glmS of Thermus thermophilus HB8: an essential gene for cell-wall synthesis identified immediately upstream of the S-layer gene Mol Microbiol 1995 17 1 12 7476196
Mac Siomoin RA Nakata N Murai T Yoshikawa M Tsuji H Sasakawa C Identification and characterization of ispA, a Shigella flexneri chromosomal gene essential for normal in vivo cell division and intracellular spreading Mol Microbiol 1996 19 599 609 8830250 10.1046/j.1365-2958.1996.405941.x
Maggio-Hall LA Claas KR Escalante-Semerena JC The last step in coenzyme B(12) synthesis is localized to the cell membrane in bacteria and archaea Microbiology 2004 150 1385 1395 15133100 10.1099/mic.0.26952-0
Wu G Williams HD Gibson F Poole RK Mutants of Escherichia coli affected in respiration: the cloning and nucleotide sequence of ubiA, encoding the membrane-bound p-hydroxybenzoate:octaprenyltransferase J Gen Microbiol 1993 139 1795 1805 8409922
Xia T Song J Zhao G Aldrich H Jensen RA The aroQ-encoded monofunctional chorismate mutase (CM-F) protein is a periplasmic enzyme in Erwinia herbicola J Bacteriol 1993 175 4729 4737 8335631
Kosak ST Groudine M Gene order and dynamic domains Science 2004 306 644 647 15499009 10.1126/science.1103864
Boussau B Karlberg EO Frank AC Legault BA Andersson SG Computational inference of scenarios for alpha-proteobacterial genome evolution Proc Natl Acad Sci U S A 2004 101 9722 9727 15210995 10.1073/pnas.0400975101
Sullivan JT Ronson CW Evolution of rhizobia by acquisition of a 500-kb symbiosis island that integrates into a phe-tRNA gene Proc Natl Acad Sci U S A 1998 95 5145 5149 9560243 10.1073/pnas.95.9.5145
Wong K Golding GB A phylogenetic analysis of the pSymB replicon from the Sinorhizobium meliloti genome reveals a complex evolutionary history Can J Microbiol 2003 49 269 280 12897836 10.1139/w03-037
Prakash T Ramakrishnan C Dash D Brahmachari SK Conformational analysis of invariant peptide sequences in bacterial genomes J Mol Biol 2005 345 937 955 15644196 10.1016/j.jmb.2004.11.008
Seshu J Boylan JA Hyde JA Swingle KL Gherardini FC Skare JT A conservative amino acid change alters the function of BosR, the redox regulator of Borrelia burgdorferi Mol Microbiol 2004 54 1352 1363 15554974 10.1111/j.1365-2958.2004.04352.x
Winfield MD Groisman EA Phenotypic differences between Salmonella and Escherichia coli resulting from the disparate regulation of homologous genes Proc Natl Acad Sci U S A 2004 101 17162 17167 15569938 10.1073/pnas.0406038101
Knight CG Kassen R Hebestreit H Rainey PB Global analysis of predicted proteomes: functional adaptation of physical properties Proc Natl Acad Sci U S A 2004 101 8390 8395 15150418 10.1073/pnas.0307270101
Jordan IK Rogozin IB Wolf YI Koonin EV Essential genes are more evolutionarily conserved than are nonessential genes in bacteria Genome Res 2002 12 962 968 12045149 10.1101/gr.87702. Article published online before print in May 2002
Completed Microbial Genomes
Persson B Argos P Prediction of transmembrane segments in proteins utilising multiple sequence alignments J Mol Biol 1994 237 182 192 8126732 10.1006/jmbi.1994.1220
Moreno-Hagelsieb G Collado-Vides J A powerful non-homology method for the prediction of operons in prokaryotes Bioinformatics 2002 18 Suppl 1 S329 36 12169563
Sokal RR Rohlf FJ 3rd Biometry: The pinciples and practice of statistics in biological research 2003 New York , WH Freeman and Company 109 117
Laquip Proteomic Team
Pearson WR Lipman DJ Improved tools for biological sequence comparison Proc Natl Acad Sci U S A 1988 85 2444 2448 3162770
Software EMBO EMBOSS
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Rivera MC Jain R Moore JE Lake JA Genomic evidence for two functionally distinct gene classes Proc Natl Acad Sci U S A 1998 95 6239 6244 9600949 10.1073/pnas.95.11.6239
Pajek
Krieger CJ Zhang P Mueller LA Wang A Paley S Arnaud M Pick J Rhee SY Karp PD MetaCyc: a multiorganism database of metabolic pathways and enzymes Nucleic Acids Res 2004 32 D438 42 14681452 10.1093/nar/gkh100
|
16229745
|
PMC1276791
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Oct 17; 5:55
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-55
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-431624204810.1186/1471-2296-6-43Research ArticleSatisfaction is not all – patients' perceptions of outcome of general practice consultations, a qualitative study Andén Annika [email protected] Sven-Olof [email protected] Carl-Edvard [email protected] Bergnäsets Vårdcentral, Box 80074, SE-97433 Luleå, Sweden2 Department of Public Health and Clinical Medicine, Family Medicine, University of Umeå, SE-90185, Umeå, Sweden3 Kalmar County Council, Vårdcentralen Esplanaden, SE-59330 Västervik, Sweden2005 24 10 2005 6 43 43 2 3 2005 24 10 2005 Copyright © 2005 Andén et al; licensee BioMed Central Ltd.2005Andén et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Evaluation of outcome in general practice can be seen from different viewpoints. In this study we focus on the concepts patients use to describe the outcome of a consultation with a GP.
Method
Patients were interviewed within a week after a consultation with a GP. The interviews were made with 20 patients in 5 focus groups and 8 individually. They were analysed with a phenomenographic research approach.
Results
From the patient's perspective, the outcome of a consultation is about cure or symptom relief, understanding, confirmation, reassurance, change in self-perception and satisfaction.
Conclusion
General practice consultations are often more important for patients than generally supposed. Understanding is the most basic concept.
==== Body
Background
Evaluation of outcome in general practice can be seen from different viewpoints.
Together, the patients' and the professionals' viewpoints make up the clinical perspective on outcome in medicine. Even when limited to those two, evaluation of general practice is a complex matter. Donabedian pointed out the difficulties in defining outcomes in general practice sufficiently to make them measurable [1]. To respond to the diversity and complexity of practice and its consequences, outcome research in general practice must take on many shapes.
On the whole, and according to the literature in the field, the professional perspective has dominated. Here, the measurable, bodily effects of treatment or prevention are studied and the methods include epidemiology and various forms of intervention studies, some of them fulfilling the criteria of the RCT. One example of the former is the UKDPS in which development of complications in relation to the care of the diabetic patient is studied [2], while RCTs like "Famciklovir for the treatment of acute herpes zoster" describe the effects of specific treatments [3]. If practising according to the evidence given, the GP has to rely on such research. In the individual case he does not know whether the patient is actually being treated or just taking the medicine. Outcome of treatment is here a forecast with reasonable accuracy. Intermediate outcomes, that are available in practice, such as the levels of blood pressure or HbA1c, may support the forecast, but should not be mistaken for factual outcomes.
In general practice patient-centred medicine has been increasingly important. Patient-centredness is the common denominator of a number of systematic efforts within general practice internationally, to introduce the patient as a person into the scope of the clinical method [4,5]. Patient-centredness has its roots in the Balint tradition [6] and the innovative research of Byrne and Long [7]. In the patient-centred clinical method described by Stewart et al five principal domains are included- exploring the illness experience, or expectations, the whole person, finding common ground, health promotion and enhancing the doctor-patient relationship [8].
According to the hypothesis, or hope, that it should improve health, and not just lead to good scores in psychological measures patient-centred medicine has been looked upon as an intervention. Since randomised interventions are difficult to achieve, the methods used are mainly indirect. Outcomes of patient-centred consultations have been compared to those less patient-centred. The viewpoint of the profession has decided also the evaluation of outcome of patient-centred medicine. Hard endpoints are looked for, and when psychological effects are brought within the scope, patients have to tick off predefined answering boxes to make quantification possible.
Accordingly, when studying the effects of patient-centredness, Stewart et al used the following outcome measures: patient's health measured for symptom discomfort and concern, self reported health and medical care utilization of diagnostic tests, referrals and visits to their GPs. They found that patients reported better recovery from discomfort and concern, better emotional health and fewer diagnostic tests and referrals if the patient himself perceived that the visit had been patient-centred. However this was not the case if an external judge scored audiotapes of the consultation as patient-centred [8]
Kinnersley et al inquired into whether the GPs' working style, especially patient-centredness, was related to outcome. Five generic outcomes were measured: doctor-patient agreement, patient satisfaction, resolution of symptoms, resolution of concern, and functional health. The only outcome that correlated with patient-centredness was satisfaction [9].
In the consultation, patient-centred medicine addresses the patient's view. In the extension from this should lie the interest in the patient's personal view on outcome, independent of the possible change of symptoms or disease. Studying patient satisfaction may look like doing patient-centred outcome research, but satisfaction is a multifarious concept. What are the implications of the findings of patients being satisfied? It could mean satisfaction with the doctor, the communication, the staff, the accessibility or the fulfilment of expectations. In a meta-analysis from 1988 it was noted that only 4 % of 221 studies related patient satisfaction to health outcome [10].
Jackson et al found, when evaluating satisfaction, that immediate post-visit satisfaction was heavily influenced by variables reflecting doctor-patient communication, while satisfaction within two weeks and three months was linked to symptom outcome of the consultation [11]. Clinging to patient-satisfaction, without specifying the term, or without asking patients directly, does not make the patient's experience and view become really expressed.
The limitations of patient-satisfaction as an outcome measure, has lead to the development of more nuanced protocols, such as in the PEI and PEQ instruments [12,13]. They have been developed to trace more crucial, personally-oriented outcomes than plain satisfaction, linked to patient-centredness. Enablement was greater if the doctor had been interested in the effect on the patients' life, health promotion and had had a positive approach [14].
The PEI and the PEQ spring out from the view of the profession on what should be the preferences in terms of personal outcomes. They are from the start integrated into the idea of patient-centredness. They are sensitive with regard to variations in the degree of patient-centredness in consultations. Still, there are also those outcomes that inevitably exist, beyond the doctor's aims and awareness. We find much less research in the data bases about them.
This study focuses on the patients' view of outcome as a phenomenon in its own right without looking at what actually happened in the consultation. Our aim was to draw up a systematic outline of the outcomes the patients may perceive after consultations with their GPs.
Method
We chose to make a focus group interview study with a phenomenographic approach.
Phenomenography
Phenomenography is a research approach originally developed when studying learning in pedagogic research [15-17]. As in other approaches of qualitative research, the aim is to describe the world as it is understood or experienced. People experience phenomena or situations in the world in qualitatively different ways, but in a limited number of ways, and the aim of phenomenography is to discern and describe such differences in a systematic way. The different ways of perceiving a phenomenon or a situation are called description categories. The description categories seen together, when they have been compared with regard to their differences and similarities, constitute what is called the outcome space.
Phenomenography has been used for health research, for instance attitudes towards physical activity among people with rheumatoid arthritis [18] or patients' long term relation to their asthma-allergy [19].
For our purpose phenomenography was a fruitful research approach. When dealing with patients, it is useful to understand how different patients can experience similar situations in different ways [20,21].
Patients
The patients were recruited from four health centres and one after-hours general practice clinic in Luleå and Piteå in northern Sweden. They were asked by members of the staff to come for a focus group interview within a week after their latest consultation. They received written and oral information about the study.
The first step of recruitment was broad, but finally, we asked selected patients to participate, as we wanted to roughly cover the range of patients in family practice regarding age and common health problems.
Our preference for a focus group was based on the expectation that the patients in a group would inspire each other, and thus more aspects would be obtained than in a one-to-one, doctor-patient interview [22]. On the other hand we did not want a selection of patients due to the method of recruitment. As some patients hesitated to join a group interview, we added individual ones to complete the selection. We managed to recruit 5 groups with 3–6 patients in each. Twenty patients were interviewed in this way, while 8 patients were interviewed individually.
The patients were from 2–74 years, median age being 47, nineteen were women and nine were men. In the case of the two-year-old child, the mother was interviewed. The diagnostic groups were common cold (4), problems with the back and joints (9), diseases of the circulatory system (9), internal medicine diseases (4), allergies (2) dermatological problems (1), psychiatric problems (1) and health check-ups (2). Three patients had more than one problem.
Nine patients lived in the countryside and 19 in the town. Five patients had another mother tongue than Swedish. Different social classes were represented.
Interviews
The interviews took place within a week after their latest consultation with a GP. We wanted to see them after a week, and not immediately, to give symptom change a chance to occur. We examined how they perceived the outcome of the latest consultation as this would render the freshest memories.
The group interviews were conducted by AA, assisted by a male colleague, and lasted about 1 1/2 hours. Since a pilot interview, not processed in this study, had shown that the women said very little when there were men in the group, the group interviews were carried out with men and women separately.
The individual interviews were conducted by AA and lasted 20–40 minutes.
The question introduced was: "What did you get out of your latest consultation?" An open discussion then followed which gradually was lead into a thorough discussion on the participants' experiences of their latest consultation.
In the individual interviews the interviewees became more like patients. However, in both situations the patients were outspoken about their dissatisfaction or other unpleasant facts.
Analyses
The interviews were tape-recorded and transcribed verbatim by AA. Statements from 28 patients with 32 problems were analysed. The analysis was made in accordance with the phenomenographic analysis as described by Sjostrom with the seven steps; familiarization, compilation, condensation, grouping and classification, comparison and revision, naming of categories and description and contrastive comparison [21]. We read the transcript several times to become familiar with them. Interviews describing the outcome in similar ways were grouped together. In the accounts of the latest consultation we picked out the statements describing the outcome. These were now detached from their contexts. We categorized the statements. The description categories were compared and their content further scrutinized. The outcome space gave the overall picture of the patients' conceptions of the outcome of their latest consultation.
The Regional Ethics Committee of Umeå University approved the study.
Results
The analysis brought forward six categories of outcome. From the patient's perspective, the outcome of a consultation is about
- cure or symptom relief
- understanding
- reassurance
- confirmation
- change in self-perception
- satisfaction
Each category contains what patients refer to as being important in their way of thinking and perception of the outcome of their latest consultation. The categories partly overlap, but are still distinct concepts.
Except for change in self-perception, the categories represent the evident needs and requests patients have when consulting. As far as the outcomes are "about" evident needs, the specific outcomes may be presented either positively or negatively; positively when the need in question is satisfied, and negatively when not.
In the following the categories will be exemplified by quotations.
Cure or symptom relief
In this category are statements where the outcome was cure or symptom relief; either as experienced or as expected but without being obtained. This is a desired outcome but often not possible as many patients have symptoms or diseases that cannot be cured. The patients presenting this outcome often had acute or semi acute symptoms and/or disease.
The patients who had been cured did not perceive a change in self-perception, probably because they had not been confronted enough with the illness experience.
Citation: Nike (woman, age 55):"I had a pain in the elbow for several months. I work with a physiotherapist and I had been asking over and over again if she could do something. -No she said. I went to a doctor and actually he gave me a diagnosis immediately. I got treatment and I was cured. So I was very satisfied."
Understanding
This category was relevant in all the statements about consultation outcomes. Understanding may be increased or may be a matter of frustration. All patients expressed that they wanted to "know what they had"; some wanted to know more even though they "knew what they had". The patients considered knowledge about their state to be a main outcome of the consultation. They requested knowledge of "what they had" based on their own condition, in their own circumstances and with their own understanding. An understanding might imply quite different things for different patients with a similar medical condition but may also vary over a period of time for the same patient. The name of the disease or a diagnosis was not always what they needed. Neither was the cause always a prerequisite for understanding, although the doctor may have found the explanation of the cause so obvious that there should be nothing left to wonder about.
Understanding is necessary to manage to live on with the health problems and the concern caused by them.
Citation Mari (woman age 46):"A diagnosis for me is completely unessential. What I want is that they realize why I have pain. So I can get rid of it."
Patients, who felt that they had not acquired an understanding of their condition, were dissatisfied with the outcome even if they had been cured.
Lejla (woman, age 39):"But I mean, just relieving the pain does not help, you also have to know why you have it. It doesn't help just to be relieved you must know in some way how to handle it to be able to prevent more pain."
Understanding must not be mixed up with explanation because an explanation in the abstract that does not respond to one's experienced needs is of no benefit.
Citation: Siri (woman 18):"They never explained why I got this skin infection, they just said it is some staphylo- or strepto-something."
In other situations, though, information about the condition and nothing more, may lead to a new understanding, which was then the outcome of the consultation.
Citation Erik (man age 63):"I have some stuff that goes from my kidneys into the blood, I don't know the name of it, I don't have to because I am not the doctor, but it could get worse if I stopped taking my antihypertensive, it could be dangerous for me and it would really be rotten to get kidney problems."
Confirmation
The patients had often had thoughts about their symptoms /disease before the consultation, and maybe also fantasies of how it could develop. The GP had observed and listened, added some tests in a few cases, but had then not taken any action beyond confirmation. An outcome for some patients was that their fears or fantasies were confirmed or unconfirmed.
Citation Anette (woman 33):"-So I asked that doctor -Do I have fibromyalgia? Because I had been thinking and wondering. – Yes she said that has been established. – Yes thank you, then I know, I said. I got to know this half a year ago when my ordinary doctor was not here, and I had had this pain for four and a half years but I had never got round to asking before." (This citation referred to a consultation half a year earlier)
A confirmation of a disease, even when serious, is at the same time a confirmation of an experience, and therefore not only negative. The prevailing uncertainty when the doctor does not know, or does not respond to the worrying experience, may however in itself be a torment. Outcomes of this nature, presented by patients who were thrown into uncertainty, were also placed in this category. They had been referred, and were waiting for further treatment or assessments. They did not know what they could expect and were left in a state of confusion. They did not express satisfaction, dissatisfaction or any other feelings with regard to the consultation. The uncertainty of the situation dominated.
Nils (man, age 73):"You will see then, you know, if it is the kidneys that are.... The function of the kidneys has been a little poor .... I get so tired all of a sudden and that's not good...I am not exactly ready to die yet...But you won't know from one day to another. I have to find out the reason for my being so tired. I mostly want to lie down, but you can't lie down all the time. You have to keep moving.
For others the confirmation dealt with the fact that they were doing the right thing. Their judgement was recognized by the doctor.
Citation Gudrun (woman 49):"She (the GP) examined me very well. I said I had been seeing a physiotherapist, and she told me to go on with the exercise that I had been taught. I told her I had been taking painkillers, aspirin and paracetamol, and she told me to go on with that, and to take it easy."
For some patients an assessment or information about their condition was the only outcome of the consultation. They had diseases that did not make them disabled but rather were to be considered as risk factors. Without having been worried they had got it confirmed that everything was well.
Citation Johnny (man age 62):"Now I have good tests on everything, everything was perfect. That was the good experience of my visit."
Some patients perceived lack of confirmation although they had expected it.
Citation Cecilia (woman, age 36):"I came to this doctor and told him about all the strange allergies I had had this last week, and showed him my wrist, that all of a sudden had become so swollen. So, he said you must have a sprain, and he gave me naproxene. I felt so misunderstood and I was so angry, that I went home and took the cortisone I had got the other day."
Reassurance
Some patients, who had been worried before the consultation, perceived reassurance as an outcome. Their fears were not confirmed. Once the cause of worry had been refuted, the worry itself was much diminished and almost forgotten.
A reassurance could be both explicit and implicit. An assessment saying that there is nothing dangerous going on is an explicit outcome.
Citation Nils (man 73):"I had felt extra beats from my heart and that made my pulse jump. I thought I would maybe need a pacemaker but my Doctor said I did not. It will probably disappear by itself. It was a good thing that I don't need a pacemaker."
A reassurance can also be implicit. Getting a diagnosis or an explanation of symptoms implies that it is not another, dangerous disease. The fear did not have to be mentioned. The fear of cancer was seldom openly expressed but often between the lines.
Citation Johnny (man 63):"Now I have good tests on everything... That was the good experience of my visit. I was not worried. But people around, they die. You are at the age for prostatic cancer.
Often reassurance was seen together with confirmation, especially when a worry had been confuted. But they could also be separate. Patients who had changed their image of themselves perceived confirmation but did not mention reassurance.
Change in self-perception – accepting the reality of the body
In this category we find statements from patients who had had the symptoms or the disease for a long time and now had reached the understanding that it would persist. The consultation had been the last in a row where earlier consultations had gradually prepared for a more definite change in self-perception. In this very consultation knowledge had turned into acceptance of the reality of the body and now they were ready to face their future searching for strategies to handle their situation and their lives. The illness/disease did not change but they were satisfied with the outcome.
This outcome was seen in some form in a fourth of the patients' descriptions.
Citation Hedvig (woman age 59):"It's something neurological...it's something in the brain you know...you can't know for sure what is actually the cause of it...How do you get on with your life after this? The consultation before last was about the whole of me and everything that concerns me... The last time I came here there were three ready suggestions for me; thus the sadness, but also relief. I'm not at all disappointed with that doctor or so, the sadness is about other things – about life."
Satisfaction
A manifestation of satisfaction or dissatisfaction was to be found in most of the patients' statements. It was rather expressed as positive or negative assessments of consultation outcome than as an explicit degree of satisfaction. Statements of satisfaction or dissatisfaction often functioned as a summing up or conclusion of the patient's evaluation of the outcome. Satisfaction was never the main outcome.
The patients were satisfied when they had acquired an understanding of the condition. Patients who "knew what they had" were satisfied even though they were not relieved or cured.
Citation Curt (man age 74):"Now the last time my blood pressure had gone down so it was just 170 over 70 and that was good in my case, it had gone down.
Patients who did not know what they had were dissatisfied even if they had been cured.
Citation Mia (woman age 36):"I didn't even get to know what I had. I was so angry with myself- why had I not asked? I had to call back to the nurse and ask and she said you have tonsillitis."
Maja (woman age 20):"I went to the doctor and he bent my knees back and forth and pressed them a little, and then he gave me a prescription for pain killers. But I wanted to know what it might be. He could not answer, because he did not know. I was very dissatisfied with going there. I had realized that myself, that I had pain and needed pain killers but I wanted to know what it was. If he couldn't help me, I think it was his duty to send me to a specialist."
All the patients who had acquired an understanding were satisfied. Some were satisfied if they had received confirmation but not an understanding.
All the patients who had not acquired an understanding or received confirmation were dissatisfied. They were dissatisfied even if they had been cured. They did not feel reassured.
Patients who were dissatisfied felt that they had not been seen or heard during the consultation.
They were mostly women whose mother tongue was different from the doctor's.
Discussion
The patients' perceptions of the outcome of their latest consultation with a GP can be described with the concepts; cure or symptom relief, understanding, reassurance, confirmation, change in self-perception and satisfaction.
The analysis illuminates a spectrum of categories- meanings contained within what is perceived as outcome by patients in general practice. Some are of course self evident and also well researched, while others probably are less well recognized, or less recognized as to their possible importance.
Satisfaction has been measured in many ways. Two satisfaction instruments are CSQ [23] and MISS [24]. They have the main focus on the consultation situations but few items on outcome.
Satisfaction is necessary as a part of an evaluation but not enough on its own.
Howie et al developed the Patient Enablement Instrument, PEI, with questions immediately after the consultation whether the patient could understand his illness, and cope with illness and life in a better way [12]. This covers the concept of understanding and is close to the concept "a change in the perception of oneself".
The PEQ questionnaire is broader and has focused on the patients' experiences of the consultation in terms of communication, emotions, outcome similar to the PEI questions, barriers and auxiliary staff [13].
There are several validated instruments to measure change of health or symptoms. The SF-36 [25], EuroQol [26]and MYMOP [27] are such instruments.
What emerges as an important finding, and which is not so much an item of other outcome studies, is that patients do not assess outcome predominately as a change of symptoms. Outcome is a change within a context, that embraces the person, the body as an aspect of the person, and the person's understanding of what is going on in his/her own, physical body.
They seldom mentioned prescriptions or sick-listing, and they did not regard such measures as outcomes.
The main position of understanding is in accordance with the goals for patient-centred care, but while understanding is there, above all an aspect of "finding common ground", our findings suggest that it should be regarded as an end in itself.
Our results indicate that outcomes of consultations, from the patients' point of view, to a great extent concern how to deal with life changes caused by ill health. In the first place this may be accomplished through increased knowledge and understanding, in the second through cure or relief and in the third through the acceptance of change and finding coping strategies. This is in accordance with Helen Ekströms recent thesis, "Keeping my ways of being", where she found that the patients who go through a bodily change finally reach a reappraisal of themselves [28].
One definition of health, which is relevant in this context, is a "home-likeness" in the world [29]. The patients who had experienced a change in their self-perception caused by disease had lost that feeling, but after this latest consultation they were on their way home again.
We believe that the overall, systematic picture of perceived outcome has relevance in itself. Being aware of the possible range of outcomes, in the consultation and in the longer term, the GP may trace the effects of his/her own actions in a more sensitive way. The signs should be there in the way patients talk, act and react. It is not necessary that all become matters of open discussion. Some of the outcomes, like confirmation or change in self-perception, may not even be a possible request for the patient. Still they should be parts of a valid doctor-patient relation. In recognizing the implicit, that which can not be verbalized, our findings may be a valuable addition where the spirit of explicitness, so eloquently argued for by the protagonists of patient-centredness, falls short.
Limitations of the study
The selection was purposeful in relation to the aim of the research [30]. The 28 patients represented both sexes, all ages, symptoms and diseases common in general practice and different social circumstances. Still we cannot maintain that the outcome space is complete. Other outcomes might have been described if more patients had been interviewed.
Adding individual interviews for those who preferred meant that it was not only patients who were positive to group participation that were included.
The focus group interviews were richer in content, but the statements were interwoven and had to be unravelled. In the individual interviews the patients took a more passive role. There was one essential difference between the content of the two types of interviews, since a change in the perception of oneself was only described in the focus groups. The other categories were represented in both sorts of interviews. This strengthens the presupposition that the patients were more free in group interviews.
The impact of the consultation works over a period of time. The interviews were made within a week, which might be too short a period to see some outcome for some patients, but with a longer interval other patients could have forgotten.
Conclusions of the study
- The categories of the perceptions of the patients' outcome that we have described have been investigated and measured to some extent in earlier studies, but here the picture is more complete.
- The results imply that general practice consultations often are more important for patients than generally supposed. The most radical outcome is a change in self- perception, which is a big thing in the individual's world.
- Understanding is the most basic outcome, being an aspect of all the others.
- Cure, or remedy, that many doctors regard as the most self-evident outcome, is quite often of limited importance.
- Satisfaction relates to all the other categories, but is first and foremost a function of the understanding.
- The predicament of the patient has a major impact on what turns out to be the outcome. Seemingly small contributions from the doctor, like the pure confirmation of the state of matters, may become great as to their effects.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AA made the interviews and transcribed them.
AA and SOA made the design of the study.
AA and SOA and CER made the analyses and wrote the article.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Donabedian A Evaluating the quality of medical care Milbank Mem Fund Q 1966 44 Suppl:166 206
Clarke PM Gray AM Briggs A Farmer AJ Fenn P Stevens RJ Matthews DR Stratton IM Holman RR A model to estimate the lifetime health outcomes of patients with type 2 diabetes: the United Kingdom Prospective Diabetes Study (UKPDS) Outcomes Model (UKPDS no. 68) Diabetologia 2004 47 1747 1759 15517152 10.1007/s00125-004-1527-z
Tyring S Barbarash RA Nahlik JE Cunningham A Marley J Heng M Jones T Rea T Boon R Saltzman R Famciclovir for the treatment of acute herpes zoster: effects on acute disease and postherpetic neuralgia. A randomized, double-blind, placebo-controlled trial. Collaborative Famciclovir Herpes Zoster Study Group Ann Intern Med 1995 123 89 96 7778840
McWhinney IR A textbook of family medicine 1997 second New York, Oxford University Press
Pendleton D STTPHP The Consultation an approach to learning and teaching 1984 6 Oxford, Oxford university press 117
Balint M Läkaren patienten och sjukdomen [The doctor, his patient and the illness] 1964 Lund, studentlitteratur 396
Byrne Long Doctors talking to patients 1976 London, Her Majestys Stationary Office
Stewart M Patient-centered medicine Patient-centered care 2003 Abdington, USA,
Kinnersley P Stott N Peters TJ Harvey I The patient-centredness of consultations and outcome in primary care Br J Gen Pract 1999 49 711 716 10756612
Hall JA Dornan MC Meta-analysis of satisfaction with medical care: description of research domain and analysis of overall satisfaction levels Soc Sci Med 1988 27 637 644 3067359 10.1016/0277-9536(88)90012-3
Jackson JL Chamberlin J Kroenke K Predictors of patient satisfaction Soc Sci Med 2001 52 609 620 11206657 10.1016/S0277-9536(00)00164-7
Howie JG Heaney DJ Maxwell M Walker JJ A comparison of a Patient Enablement Instrument (PEI) against two established satisfaction scales as an outcome measure of primary care consultations Fam Pract 1998 15 165 171 9613486 10.1093/fampra/15.2.165
Steine S Finset A Laerum E A new, brief questionnaire (PEQ) developed in primary health care for measuring patients' experience of interaction, emotion and consultation outcome Fam Pract 2001 18 410 418 11477049 10.1093/fampra/18.4.410
Little P Everitt H Williamson I Warner G Moore M Gould C Ferrier K Payne S Observational study of effect of patient centredness and positive approach on outcomes of general practice consultations Bmj 2001 323 908 911 11668137 10.1136/bmj.323.7318.908
Marton F Studying Conceptions of Reality- A Metatheoretical Note Scandinavian Journal of Educational Research 1981 25 159 169
Marton F Phenomenography- Describing Conceptions in the World around us Instructional Science 1981 10 177 200 10.1007/BF00132516
Pang MF Two faces of variation: on continuity in the phenomenographic movement Scandinavian Journal of Educational Research 2003 47 145 156 10.1080/00313830308612
Eurenius E BGSC Attitudes towards physical activity among people with reumatoid arthritis Physiotherapy Theory and Praxis 2003 19 53 62 10.1080/09593980307967
Hansson-Scherman M Refusing to be Ill; a Longitudinal Study of Patients' Relationships with their Asthma/Allergy Acta Universitatus Gothenburgensus 1994 Goteborg, University of Goteborg
Barnard A McCosker H Gerber R Phenomenography: a qualitative research approach for exploring understanding in health care Qual Health Res 1999 9 212 226 10558364 10.1177/104973299129121794
Sjostrom B Dahlgren LO Applying phenomenography in nursing research J Adv Nurs 2002 40 339 345 12383185 10.1046/j.1365-2648.2002.02375.x
Barry CA Bradley CP Britten N Stevenson FA Barber N Patients' unvoiced agendas in general practice consultations: qualitative study Bmj 2000 320 1246 1250 10797036 10.1136/bmj.320.7244.1246
Baker R Whitfield M Measuring patient satisfaction: a test of construct validity Qual Health Care 1992 1 104 109 10172105
Kinnersley P Stott N Peters T Harvey I Hackett P A comparison of methods for measuring patient satisfaction with consultations in primary care Fam Pract 1996 13 41 51 8671103
McHorney CA Ware JEJ Raczek AE The MOS 36-Item Short-Form Health Survey (SF-36): II. Psychometric and clinical tests of validity in measuring physical and mental health constructs Med Care 1993 31 247 263 8450681
Brooks R EuroQol: the current state of play Health Policy 1996 37 53 72 10158943 10.1016/0168-8510(96)00822-6
Paterson C Measuring outcomes in primary care: a patient generated measure, MYMOP, compared with the SF-36 health survey Bmj 1996 312 1016 1020 8616351
Ekstrom H Keeping my Ways of Being Department of medicine 2005 Lund, University of Lund
Svenaeus F The Hermeneutics of Medicine and the Phenomenology of Health Faculty of Arts and Science 1999 Linköping, Linköpings universitet 314
Malterud K Qualitative research: standards, challenges, and guidelines Lancet 2001 358 483 488 11513933 10.1016/S0140-6736(01)05627-6
|
16242048
|
PMC1276792
|
CC BY
|
2021-01-04 16:29:13
|
no
|
BMC Fam Pract. 2005 Oct 24; 6:43
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-43
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1451624202310.1186/1471-2164-6-145Softwarei-Tracker: For quantitative proteomics using iTRAQ™ Shadforth Ian P [email protected] Tom PJ [email protected] Kathryn S [email protected] Conrad [email protected] Department of Analytical Science and Informatics, Cranfield University at Silsoe, Silsoe, Bedfordshire, UK2 Cambridge Centre for Proteomics, Biochemistry Department, Cambridge University, Cambridgeshire, UK2005 20 10 2005 6 145 145 1 7 2005 20 10 2005 Copyright © 2005 Shadforth et al; licensee BioMed Central Ltd.2005Shadforth et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
iTRAQ™ technology for protein quantitation using mass spectrometry is a recent, powerful means of determining relative protein levels in up to four samples simultaneously. Although protein identification of samples generated using iTRAQ may be carried out using any current identification software, the quantitation calculations have been restricted to the ProQuant software supplied by Applied Biosciences. i-Tracker software has been developed to extract reporter ion peak ratios from non-centroided tandem MS peak lists in a format easily linked to the results of protein identification tools such as Mascot and Sequest. Such functionality is currently not provided by ProQuant, which is restricted to matching quantitative information to the peptide identifications from Applied Biosciences' Interrogator™ software.
Results
i-Tracker is shown to generate results that are consistent with those produced by ProQuant, thus validating both systems.
Conclusion
i-Tracker allows quantitative information gained using the iTRAQ protocol to be linked with peptide identifications from popular tandem MS identification tools and hence is both a timely and useful tool for the proteomics community.
==== Body
Background
In recent years several techniques for protein quantitation by mass spectrometry have emerged. These include isotope-coded affinity tags (ICAT), metabolic labelling and stable isotope labelling of amino acids in culture (SILAC) [1-3]. These techniques enable peptides derived from two samples to be distinguished by mass spectrometry (MS). This is achieved though protein labelling with isotopically distinct tags (ICAT) or through the incorporation of isotopically distinct amino acids (SILAC), or a stable isotope labelled compound which represents the sole source of an element, typically nitrogen or carbon (metabolic label-ling). Protein quantitation can then be achieved by comparing the MS intensity of the peptides derived from the two samples. iTRAQ™ is a recently developed protein quantitation technique that utilizes four isobaric amine specific tags [4]. In single MS mode the differentially labelled versions of a peptide are indistinguishable. However, in tandem MS mode (in which peptides are isolated and fragmented) each tag generates a unique reporter ion. Protein quantitation is then achieved by comparing the intensities of the four reporter ions in the MSMS spectra.
The principal advantage of iTRAQ over ICAT, SILAC and metabolic labelling is that four samples can be analyzed simultaneously, thereby reducing the amount of mass spectrometry time needed for analysis. In addition, the b- and y- ions derived from peptides labelled with the four iTRAQ tags are indistinguishable, resulting in higher MSMS intensity and therefore more confident peptide identifications in comparison to ICAT, SILAC and metabolic labelling, in which the MSMS spectra for the differentially labelled peptides are acquired independently.
The ProQuant software, from Applied Biosciences (ABI), enables the quantitation and identification of iTRAQ labelled peptides. Peptide identification is achieved using ABI's Interrogator™ search algorithm. However, for the purpose of reporting iTRAQ results it is desirable to verify the proteins identified using a second, more widely used, MSMS search engine such as Mascot [5] or Sequest [6]. For this reason, the i-Tracker software has been developed to calculate iTRAQ reporter ion ratios and report them in a format that can be easily integrated with Mascot and Sequest search results.
i-Tracker takes as its input non-centroided mass spectra, either in the .dta format, as created by a program such as wiff2dta [7], or the .mgf files generated by the mascot.dll script for ABI's Analyst™ software. The software returns the relative ratios of each reporter ion. Indicative errors are provided to highlight the large discrepancies that may arise in the reported ratios when very low ion counts are used; they do not provide a model of all errors in the system, such as the probability if successful ion detection or counts introduced by background noise.
Implementation
Overview
The iTRAQ™ protocol uses four reporter ions of 114.1, 115.1, 116.1 and 117.1 Da. These are singly-charged and so found in the region 114 – 117 m/z in the mass spectra. Relative quantitation is performed by comparing the peak areas of each of these reporter ions in the mass spectrum. The default setting for i-Tracker assumes that the bulk of the peak will occur in the region of the reporter ion mass ± 0.05 Da. Each ion peak within this region is captured. The area of each reporter ion is then calculated by summing the areas of the trapezoids formed between each captured peak. This can be user-adjusted to suit the characteristics of the mass analyser used.
The reagents supplied by ABI are not 100% pure, but come with a datasheet by batch indicating the percentages of each reporter ion reagent that differ by -2, -1, +1 and +2 Da from the quoted mass. This quality control measure is taken into account by the i-Tracker software by adjusting each peak area as appropriate. The simultaneous equations needed for making this adjustment are solved using Cramer's rule. If the determinant of the initial matrix of coefficients is zero, or if no purity information is supplied, the software will output a warning and proceed without purity correction.
Following any purity correction, the peak areas are normalized. These normalized areas are used to calculate the quantitative ratios between each reporter ion. If the maximum peak height of any reporter ion is below the user-defined threshold for consideration, the string "UT" for under-threshold is the output. If there is no peak in the spectrum associated with a reporter ion as comparator (i.e. the denominator of the ratio calculation) the string "NA" is output.
Very low peaks in the mass spectra may suffer from large errors due to the quantized nature of the ion current. In order to provide some idea as to the potential magnitude of this error, a set of ratio-errors is reported which represent the maximum percentage error due to quantization. It should be noted that this reported error does not account for errors in detection of the ion current nor systematic error, such as background noise, in the measurement, but merely serves as a warning against placing too high confidence in reported ratios when these have been based on peaks with low ion counts.
Detailed description
Items in this section are presented in the order in which data is processed by the software with one exception: The calculation for the determinant of coefficients, for purity correction, is performed very early in the processing sequence, in order to minimise repetitive calculation, whereas here it is presented as part of the purity correction section. Other than for this the following may be considered in parallel to the Perl code (i-Tracker.pl), which contains similar headings and flags for straightforward comparison.
Data input
Spectra must be non-centroided as the peak area calculations rely on the presence of all the peaks that would otherwise be combined in a centroided output.
i-Tracker can read spectra in .dta or .mgf formats. The two formats differ in the title information they contain and slightly in the format of the precursor ion information. However, the main difference in the way i-Tracker handles these files is that .dta files, which represent a single spectrum, are read into memory before processing whilst .mgf files are read in to memory spectrum by spectrum whilst keeping the input file open. Once a spectrum's information has been read, further processing is identical between input file formats.
Reporter ion peak collection
All peaks in the ranges:
114.1 ± 0.05
115.1 ± 0.05
116.1 ± 0.05
117.1 ± 0.05
are collected as a {mass}->intensity pair (hash). The default range of ± 0.05 was identified by considering the mass accuracy of the mass spectrometer and through manual inspection of a number of these peaks in the output files. This can be user-adjusted through an option presented at run-time.
Reporter ion area calculation
For each reporter ion peak range, the total area is calculated by summing the areas between ion peak pairs using the trapezoid approximation for calculating the area under a curve.
For example, a reporter ion peak may be comprised of four ions within the range considered. Here a, b, c and d are ion masses and a', b', c' and d' are their intensities. The total area (A) of this reporter ion peak is therefore:
A = (b-a) * 0.5 * (a'+b') + (c-b) * 0.5 * (b'+c') + (d-c) * 0.5 * (c'+d')
The maximum ion peak intensity is also identified at this point for comparison with the user-entered ion intensity threshold and for the calculation of quantisation errors.
Purity correction
Each batch of iTRAQ reagents supplied by ABI is labelled with sixteen purity values indicating the percentages of each reporter ion that have masses differing by -2, -1, +1 and +2 Da from the nominal reporter ion mass due to isotopic variants. This information can be used to correct the peak areas calculated for each reporter ion to account for the losses to, and gains from, other reporter ions. Losses to ion peaks not in the reporter ion range are also accounted for in this method.
The simultaneous equations needed to solve this problem are fairly complicated, but can be framed such that Cramer's rule may be applied. A detailed explanation of how to use Cramer's rule to solve simultaneous equations may be found in [8]. Briefly, if the determinant of the matrix of coefficients for the simultaneous equation is non-zero, the solution in each variable may be found. The four-way simultaneous equation for purity correction may be written as:
a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p are the sixteen purity correction values (as percentages) in the order:
114.1 – 2 Da, 115.1 – 2 Da, 116.1 – 2 Da, 117.1 – 2 Da, 114.1 – 1 Da, etc...
(NB This is a different logical order to that in which the user enters the values, they are rearranged within the program).
w,x,y,z represent the percentage of each peak expected to be present at the mass of the reporter ion associated with that peak. Here, w is for 114.1, x for 115.1 etc.:
w = (100 - (a + e + i + m))
x = (100 - (b + f + j + n))
y = (100 - (c + g + k + o))
z = (100 - (d + h + l + p))
The area (Ar) of each reporter ion peak (r), as calculated above, can now be written in terms of the true areas of peaks (Tr):
The task is now to calculate each Tr according to these equations.
The determinant of the matrix of coefficients can be found:
If |C| is zero, then there is either an infinity of solutions or there are no solutions to these equations and so the purity correction module is skipped. If |C| is non-zero, purity correction proceeds:
The Cramer determinants, Δr, are found to be:
The true areas, Tr, can now be found:
Tr = Δr/|C|
Peak normalisation
Providing that the sum of the total areas is non-zero, normalised areas (Nr) are calculated as:
Nr = Tr / (T114.1 + T115.1 + T116.1 + T117.1)
If the sum of all areas is zero, then each normalised area is also considered to be zero.
Under threshold checking
If the maximum ion peak intensity for any reporter ion peak area is equal to or less than the user-entered threshold, a flag of "UT" for "Under Threshold" is reported.
Ratio calculation
All ratios of true areas are calculated to three decimal places provided that the denominator is non-zero. If the denominator in any ratio calculation is zero, an "NA" flag is reported.
Quantisation error calculation
Very low ion counts may introduce a significant quantisation error. To some extent this is mitigated against by a sensible user-entered threshold of around 20 ion counts, but even so, comparing two reporter ion peaks that just pass such threshold could introduce an error of around 2.5% into the final ratios:
Eg. The user-entered ion count threshold is set to 19. The "correct" areas of peaks 1 and 2 should have been based on intensities of 20.5 and 19.5 respectively, but the reported ratio is 1:1 due to the quantum nature of ion counts. A quantisation error of 2.5% has been introduced in this case. For ion counts lower than this, the potential quantisation error will be much greater, but their ratios in this case would have been masked by the user-entered threshold.
These potential quantisation errors are reported alongside the peak ratios to highlight instances where results might be compromised by this effect. They are calculated as a percentage error between two ratios thus:
Err(1,2) = (100 * ((0.5 / Peak1Max) + (0.5 / Peak2Max))
these are output in the errors matrix for each ratio.
Similar potential quantisation errors in the normalised areas are calculated as:
Err(1) = (100 * (0.5 * Peak1Max))
these are output in the left-right diagonal of the quantisation errors matrix.
Output format
Results are output in a choice of two comma-separated-variable (CSV) formats readily imported into R, Excel and other packages. One of these is designed for human-readability whereas the other is more convenient for simple analysis in a spreadsheet. Both are simple to parse for more detailed analysis. Full descriptions of both formats are provided in the user-instructions. The title information for each spectrum's results is taken from that of the input spectrum and hence matching of quantitative data with peptide identifications is straightforward. Each set of ratios is reported by spectrum, in the order in which they appear in either the directory or the input file (depending on .dta or .mgf inputs). Linking iTRAQ ratios, as determined by i-Tracker, and peptide identifications is dependent on the user being able to accurately link the identified spectra to this output data. Either the filename, for .dta input, or the Mascot peptide "Title" information can be used for this purpose, as performed in the GAPP project prototype .
Results and discussion
Relative quantitation data using i-Tracker were compared to the output from ProQuant and found to be in good agreement, as shown in Figure 1. The data used were derived from Arabidopsis membrane protein samples labeled with iTRAQ reagents. Arabidopsis membrane protein samples were prepared as described in [9]. Membrane pellets were solubilised in 100 μl of labelling buffer (50 mM TEAB, 8 M Urea, 2 % Triton X-100 and 0.1 % SDS). 100 μg of protein were reduced with 5 mM TCEP for 1 h at 20°C and cysteines were blocked with 10 mM MMTS for 10 min at 20°C. Samples were then diluted with 50 mM TEAB, in order to reduce the urea concentration to 1 M. 5 μg trypsin were added to the samples, which were then incubated for 15 h at 37°C. The peptide samples were then labelled with the iTRAQ reagents as described in [4].
Figure 1 Comparison of ProQuant and i-Tracker results. Log ratios of reporter ions 115:114 are shown for ProQuant and i-Tracker. The user-entered intensity threshold for i-Tracker processing was set to 20 ion-counts.
The pooled labelled peptides were loaded onto a Dionex ProPac SCX-10 strong cation exchange column (250 mm × 2 mm i.d.) at 0.3 ml/min and separated using a gradient of 0 mM to 500 mM NaCl over 50 min during which time 17 fractions were collected and analyzed by LC-MSMS using an ABI QSTAR mass spectrometer.
The data reported in Figure 1 represent one of these 17 fractions, picked at random and processed using i-Tracker and ProQuant. i-Tracker was run using an ion count threshold of 20 and the appropriate purity correction factors as supplied by ABI. The ratios of the two reporter ions shown (114 and 115) range from 0.12 to 7.3. Other comparisons, not shown here, demonstrate similar performance. These results demonstrate that the output from i-Tracker is almost identical to that of ProQuant in terms of calculating the ratios of reporter ion peaks. As ProQuant was developed by ABI, this very high correlation is desirable, but it should be noted was not part of the original specification for i-Tracker. If the results had been markedly different, there would have been a question to answer as to the validity of the results from both systems. As i-Tracker was developed entirely independently from the developers of ProQuant, with no information as to ABI's algorithm being sought or provided, the convergence of the end results provides a positive validation of either a common sense of the design or of the implementation of both systems. The first of these assertions would apply if the algorithms are independently identical. In this case similar design decisions would have been made when presented with the same problem, leading to a corroborative validation of the design. On the other hand, if the algorithms are different, then the results provide a demonstration that both algorithms perform to the same specification. Which of these is the true position is unknown as ABI have as yet not released details of their ProQuant algorithm. This also prevents a complete analysis of the very few outliers, around 5 out of 1463 data points present in the comparison.
A current limitation of i-Tracker is that it only accepts two types of input file; the .dta and .mgf formats. Although a number of converters are publicly available, it would be beneficial for i-Tracker to be modified such that it can handle the more generic MS file types available, such as mzXML and mzData. The advantage of using these would be that they are set to become standard across the community. However, as both are XML-based and contain encoded peak lists, processing these is more complicated than for the file-types currently handled. This extension to i-Tracker will be addressed in future releases.
Conclusion
i-Tracker provides quantitative proteomic information for peptides when using the iTRAQ reagents supplied by ABI. The principal advantage to using i-Tracker is that the results are provided in a form that may be easily linked to peptide identifications made using software other than that provided by ABI, something which is currently time-consuming and difficult using ProQuant. Furthermore, both the algorithm and source code for i-Tracker are freely available and therefore may be reviewed and developed further by the proteomics community.
Authors' contributions
CB, KL, TD and IS together conceived and developed the initial specification for the tool. IS designed and built the i-Tracker application with extensive input in the design and user-requirements from TD. TD carried out the sample collection, mass spectrometry and data analysis. IS drafted the manuscript with input from TD, KL and CB. All authors read and approved the final manuscript.
Availability and requirements
Project name: i-Tracker
Project home page: The software may be freely downloaded from this website. As it is provided as a Perl script, the source code is naturally available.
Operating system(s): Platform independent. A Windows (XP) executable is provided as well as the Perl script.
Programming language: Perl
Other requirements: The freely available Perl modules Time::localtime and Math::MatrixReal are required unless using the Windows executable.
License: GNU GPL
Any restrictions for use by non-academics: None.
Acknowledgements
We are grateful to GlaxoSmithKline and the EPSRC for jointly funding Ian Shadforth through the Engineering Doctorate (EngD) programme, and to the BBSRC and Applied Biosystems for funding Tom Dunkley through a BBSRC CASE PhD studentship. We would also like to thank Applied Biosystems for their kind donation of iTRAQ™ reagents.
==== Refs
Gygi SP Rist B Gerber SA Turecek F Gelb MH Aebersold R Quantitative analysis of complex protein mixtures using isotope-coded affinity tags Nature Biotechnology 1999 17 994 999 10504701 10.1038/13690
Krijgsveld J Ketting RF Mahmoudi T Johansen J Artal-Sanz M Verrijzer CP Plasterk RHA Heck AJR Metabolic labeling of C-elegans and D-melanogaster for quantitative proteomics Nature Biotechnology 2003 21 927 931 12858183 10.1038/nbt848
Ong SE Blagoev B Kratchmarova I Kristensen DB Steen H Pandey A Mann M Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics Molecular & Cellular Proteomics 2002 1 376 386 12118079 10.1074/mcp.M200025-MCP200
Ross PL Huang YLN Marchese JN Williamson B Parker K Hattan S Khainovski N Pillai S Dey S Daniels S Purkayastha S Juhasz P Martin S Bartlet-Jones M He F Jacobson A Pappin DJ Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents Molecular & Cellular Proteomics 2004 3 1154 1169 15385600 10.1074/mcp.M400129-MCP200
Perkins DN Pappin DJC Creasy DM Cottrell JS Probability-based protein identification by searching sequence databases using mass spectrometry data Electrophoresis 1999 20 3551 3567 10612281 10.1002/(SICI)1522-2683(19991201)20:18<3551::AID-ELPS3551>3.0.CO;2-2
Eng JK McCormack AL Yates III An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database Journal of the American Society for Mass Spectrometry 1994 5 976 989 10.1016/1044-0305(94)80016-2
Boehm A Galvin R Sickmann A Extractor for ESI quadrupole TOF tandem MS data enabled for high throughput batch processing BMC Bioinformatics 2004 5 162 162 15507135 10.1186/1471-2105-5-162
Riley KF Hobson MP Bence SJ Mathematical Methods for Physics and Engineering 1998 Cambridge, UK, Cambridge University Press 208 212
Dunkley TP Watson R Griffin JL Lilley KS Localization of organelle proteins by isotope tagging (LOPIT). Molecular Cell Proteomics 2004 3 1128 1134 10.1074/mcp.T400009-MCP200
|
16242023
|
PMC1276793
|
CC BY
|
2021-01-04 16:32:46
|
no
|
BMC Genomics. 2005 Oct 20; 6:145
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-145
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-851622569410.1186/1471-2334-5-85Research ArticleDementia-specific risks of scabies: Retrospective epidemiologic analysis of an unveiled nosocomial outbreak in Japan from 1989–90 Tsutsumi Masae [email protected] Hiroshi [email protected] Toshio [email protected] School of Nursing, Yamaguchi Prefectural University, Miyanoshimo, Yamaguchi, Japan2 Graduate School of Health Sciences, Hiroshima University, Kasumi 1-2-3, Minamiku, Hiroshima, Japan3 Department of Medical Biometry, University of Tübingen, Westbahnhofstr. 55-D, 72070, Tübingen, Germany2005 14 10 2005 5 85 85 9 6 2005 14 10 2005 Copyright © 2005 Tsutsumi et al; licensee BioMed Central Ltd.2005Tsutsumi et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Although senile dementia patients in long-term care facilities are at leading risk of scabies, the epidemiologic characteristics of this disease have yet to be fully clarified. This study documents the findings of a ward-scale nosocomial outbreak in western Japan from 1989–90, for which permission to publish was only recently obtained.
Methods
A retrospective epidemiologic study was performed to identify specific risk factors of scabies among patients with dementia. Analyses were based on a review of medical and nursing records. All inpatients in the affected ward at the time of the outbreak were included in the study. Observational and analytical approaches were employed to assess the findings.
Results
Twenty of 65 inpatients in the ward met the case definition of scabies. The outbreak lasted for almost 10 months and as a result, the spatial distribution of infections showed no localized patterns in the latter phase of the outbreak. The duration of illness significantly decreased after initiation of control measures (P = 0.0067). Movement without assistance (Odds Ratio [OR] = 11.3; 95% Confidence Interval [CI]: 2.9, 44.8) and moving beyond the room (but within the ward) (OR = 4.1; 95% CI: 1.4, 12.5) were significantly associated with infection, while types of room (Western or Japanese) and sleeping arrangement (on beds or futons laid directly on the floor) appeared not to be risk factors.
Conclusion
Univariate analysis demonstrated the importance of patients' behaviours during daily activities in controlling scabies among senile dementia patients. The findings also support previous evidence that catching scabies from fomites is far less common. Moreover, since cognitive disorders make it difficult for individuals to communicate and understand the implications of risky contacts as well as treatment method, and given the non-specific nature of individual contacts that are often unpredictable, real-time observations might help improve control practices.
==== Body
Background
Scabies is a contagious skin irritation caused by the small translucent mite Sarcoptes scabiei(itch mite). Allergic responses to these mites and the waste products they produce lead to development of extensive areas of inflamed, reddened itchy skin [1]. The disease is transmitted from person to person by direct skin contact [2] and continues to be a major problem in nursing homes in industrialized countries, particularly among debilitated patients who require extensive hands-on care [3]. The clinical features of scabies in the elderly differ from those in younger individuals and such episodes are often the cause of nosocomial outbreaks because of delayed diagnosis due to the inspecificity of the lesions [4]. This is especially true among elderly individuals diagnosed with senile, psychogenic or degenerative diseases and unable to directly complain of their symptoms. A lack of attention to individual protection measures by healthcare workers (HCWs) has also been described as a cause of delayed diagnosis [4].
Even though several reports have documented local outbreaks and dermatological case descriptions, these remain insufficient in helping identify the epidemiologic characteristics of nosocomial outbreaks. Particularly, hospital-based epidemiologic investigations focusing on patterns of transmission not only among caregivers but also among elderly inpatients are necessary in establishing and activating an appropriate surveillance system. The specific trends of a scabies outbreak were previously observed in a geriatrics hospital in Japan from 1989–90 (Tsutsumi M, unpublished data). Although these observations were neither announced nor reported officially because of factors related to the reputation of the hospital, we recently obtained permission to study and report the epidemiologic details. This paper describes a ward-scale outbreak of scabies among elderly inpatients with senile dementia in an attempt to characterize the risk factors and patterns of spread of infection through a retrospective epidemiologic study based on unveiled outbreak records.
Methods
The outbreak
On 6 May 1989, an 85-year-old female patient with senile dementia presented with tiny red dots and surrounding skin redness on her abdomen and both femoral regions, and was consequently diagnosed with scabies. She was housed in a dementia ward in a 435-bed geriatrics hospital in western Japan. The hospital was equipped with specially authorized geriatric wards according to Japanese law. Diagnoses of scabies in the dementia ward continued until 7 December 1989. Preventive measures were not instituted until 4 months after diagnosis of the index case and no prophylactic treatment of uninfected inpatients or HCWs was performed throughout; staff awareness of and adherence to infection control practice seem to have been insufficient at this time.
Case definition and diagnosis
All inpatients in the dementia ward were diagnosed with senile dementia due to prior cerebrovascular or degenerative diseases. Suspected cases of scabies in this study were defined as persons 1) housed in the dementia ward and 2) who presented with clinical signs (generalized or localized pruritus of several days evolution or appearance of cutaneous lesions suggesting scabies regardless of their severity and extent) during the outbreak period (May 1989 to February 1990). Confirmed diagnoses were made by dermatologists through direct bedside microscopic examinations of Sarcoptes scabiei. Since there was no attending dermatologist in the hospital, dermatologists working part-time once a week under the support of an outpatient service conducted these consultations.
Study background (observational study)
The aim of this study was to identify specific features of scabies outbreaks in dementia wards. Although the hospital authority in question previously prohibited documentation of the outbreak, detailed clinical and epidemiologic information was obtained by the first author (MT) for academic purpose while working as a nurse in this institute. In addition to the data obtained through personal observations, clinical information was retrospectively obtained by reviewing medical and nursing records of inpatients housed in the ward during the outbreak period.
Exposure and statistical analysis
To assess the association between dementia ward-specific characteristics and scabies, we conducted a retrospective study of all inpatients in the ward during the outbreak period. Investigated dichotomous variables were as follows: 1) individual demographic characteristics (gender); 2) type of room (Western-type floor made of wood or traditional Japanese tatami mats made of straw); 3) sleeping arrangement (on a bed or futon laid directly on the floor); 4) ease of movement (need for assistance); and 5) range of movement (confined to the room or ward). Age was not included in the analyses since the ward in question only housed patients with senile dementia. Although multivariate analyses could not be conducted due to the lack of information on non-infected individuals, stored data based on hospital records enabled investigation of univariate associations.
Comparisons between groups were performed using Fisher's exact test to assess the univariate association between the investigated variables and infection. The level of statistical significance was set at P = 0.05. Comparisons of the duration of illness before and after the initiation of control measures were performed using the F-test followed by the Student t-test or Welch test. Duration was compared between those diagnosed before 30 Sep 1989 (n = 6) and those diagnosed thereafter (n = 15). All data entry was performed by two different persons using Microsoft Excel 2000 (Microsoft Corporation, Redmond, WA) and double-checked. The statistical data were analysed using the statistical software R (R Development Core Team, Vienna) [5] and JMP IN ver. 5.1 (SAS Institute Inc., Cary, NC).
Results
Personal characteristics
Twenty of the 65 dementia ward inpatients (30.8%; 18 females, 2 males) developed symptoms. The mean age at infection was 81.6 years (standard deviation (SD) = 7.9; median = 80.5). The chain of nosocomial transmission was observed only within the dementia ward investigated. Detailed descriptions of each case are provided in Additional file 1.
Temporal and spatial distribution
The outbreak continued for a total of 288 days. The epidemic curve peaked in early November (Figure 1) with 15 patients showing associated clinical signs in mid-November (Figure 2). None of the ward staff were reported as having scabies and no cases were reported among private individuals (i.e., relatives of patients).
Figure 1 Temporal distribution of the scabies cases (n = 20) according to the time of diagnosis. Biweekly numbers of cases diagnosed on the dates given are shown.
Figure 2 Duration of illness of the scabies cases. Duration is shown at two-week intervals. Note: illness onset was biased by the delay in diagnosis.
Figure 3 shows the spatial distribution of scabies cases in the ward with time. The ward is composed of 10 patient rooms arranged in a T-shape with a nurse station at the centre. No restriction of movement or occupational therapy (i.e., recreation activities) was performed to prevent worsening of disturbed cognitive function. For three months from the beginning of the outbreak, occurrence was localized to the tatami mat room where the index case was located. Thereafter, the disease spread to rooms at each end of the ward.
Figure 3 Spatial distribution of the scabies cases (n = 20) with time. Ward-scale spatial distribution of the scabies cases are shown as of a) 15 July, b) 5 October, c) 23 November and d) 7 December, 1989. Circles and squares denote inpatients sleeping on futons directly on tatami flooring and beds, respectively. Red denotes individuals able to move without assistance and beyond their room, purple denotes those able to move without assistance but only within their room, pink denotes those unable to move without assistance, and light blue denotes those who recovered from scabies. N, nurse station; R, recreation room (for occupational therapy); area enclosed by a dashed line, Western style room with wooden flooring. Slight relocation of individuals is not shown.
Disease characteristics
Five patients (25.0%) received treatment with corticosteroid prior to diagnosis. Only 60% (n = 12) complained of an itching sensation, and 4 patients (20.0%) showed difficulty sleeping at night. Major infected areas included the abdomen (n = 9; 45.0%), chest (n = 8; 40.0%), back (n = 6; 30.0%), femoral region (n = 5; 25.0%) and axillaries (n = 5; 25.0%). On the other hand, the neck (n = 1; 5.0%) and inguinal region (n = 1; 5.0%) were rarely affected.
Three patients (15.0%) received treatment with N-Ethyl-N-o-tolylcrotonamide ointment (Crotamiton; Eurax®) [6,7], and combination therapy with 1% γ-BHC (1,2,3,4,5,6-hexachlorocychlohexane; Lindane) [8] and Eurax was performed for the remainder (Note: presently, scabies is best treated with oral ivermectin) [9,10]. The mean duration of illness was 85.0 days (SD = 64.3, median = 63.0), and this gradually shortened among later cases (Figure 2). Figure 4 shows distributions of the duration of illness before and after implementation of control measures. Since there was a significant difference in variance between the groups investigated (F = 4.51, df (degree of freedom) = 5, df = 13, P = 0.0264), Welch analysis of variance testing was performed. The mean duration of the first 6 cases (163.3 days) was significantly longer than that of the later 14 infections (53.6 days) (F Ratio = 16.52, t Ratio = 4.06, df = 5.97, P = 0.0067).
Figure 4 Comparative distributions of the duration of illness after implementation of control measures.
Dementia-specific characteristics
Table 1 summarizes the scabies cases grouped according to investigated exposure factors, and compares non-infected (n = 45) and infected inpatients (n = 20) in each grouping. There was no significant sex-related difference in infection, and types of room and sleeping arrangement were also unassociated with this scabies outbreak. Those able to move without assistance had a significantly higher chance of infection than those who could not (P = 0.0001). Further, movement beyond the room was associated with risk of disease (P = 0.0136).
Table 1 Univariate analysis: senile dementia-specific risk factors related to scabies, and comparisons between infected and non-infected individuals
Scabies cases (n = 20) Non-infected individuals (n = 45) p-value$ Odds ratio† (95% CI)*
Gender (male) 2 6 NS§ 0.7 (0.1 – 4.0)
Type of room (Japanese style with tatami mats) 19 42 NS 1.4 (0.1 – 13.9)
Sleeping arrangement (on futons directly on the floor) 18 37 NS 1.9 (0.4 – 10.1)
Ease of movement (without assistance) 17 15 0.0001 11.3 (2.9 – 44.8)
Range of movement (outwith the room but within the ward) 12 12 0.0136 4.1 (1.4 – 12.5)
$Two-tailed; †Odds ratio of being infected through exposure; *95% Confidence interval; §NS: not significant.
Discussion
We conducted an epidemiologic investigation of a previous scabies outbreak based on individual records and personal observations. The outbreak involved 20 patients with senile dementia. Ward-based control measures were not facilitated until 4 months after diagnosis of the index case, contributing to the prolonged duration of the outbreak of almost 10 months. Although the duration of illness might be biased due to the delay in diagnosis, especially at the early stage of the outbreak, the duration significantly decreased in the latter phase. Using the findings, we attempted to identify dementia-specific risk factors of scabies transmission.
One important conclusion drawn from our study is that a high proportion of the scabies cases were able to move freely around the ward, with those able to move without assistance and outwith their rooms at significant risk of infection. Some cases with cognitive dysfunction were seen walking around the ward yet no specific restrictions were enforced by the HCWs. For example, the patients frequently visited room 106 for no particular reason other than perhaps because it was located at the end of the hallway. It should be noted that transmission of scabies within dementia wards is frequently observed not only among individuals housed in the same or neighbouring rooms but also among those located far from the index case. Since spatial spread enhanced by the behaviour of dementia patients is therefore suggested, our findings support a previous suggestion that transmission through objects (fomites) is far less common [11,12]. A practical dilemma in this ward, however, is that strict control of movement, such as restricted attendance of recreation therapy, is likely to worsen dementia-associated symptoms, and cases in which isolation has triggered mental disturbance have been observed [13]. Patients with dementia tend to have non-specific contacts that are sometimes unpredictable, and consequently, real-time observations of behavioural patterns and individual control measures during daily activities should be among the goals of control practice.
For a long time, senile dementia patients have been known to be at high risk of scabies [1,4,14] and Japan is not an exception [15]. Moreover, scabies among the elderly is known to accompany underlying diseases, causing clinical signs to sometimes last longer than in younger individuals [16]. Another practical dilemma of dealing with this population is that patients with cognitive disorders sometimes do not understand even the chemotherapy treatment they are undergoing [13]. In addition, as observed here, it is extremely difficult to identify probable contacts retrospectively among patients who have difficulty communicating. In addition to discussions on biological risks previously documented in Japan [17], this epidemiologic study suggests that behaviour- and cognitive function-specific characters also play a role in the spread of this disease.
In this outbreak, scabies-specific management among staff members was delayed in the early phase of the outbreak. Furthermore, awareness of infection control practices was insufficient, preventing even chemoprophylaxis. Consequently, no specific restrictions of movement of infectious individuals were made by the HCWs, and thus the dementia patients themselves, whose contacts were unpredictable, contributed to spatial spread. Nevertheless, as a result of careful precautions at a later stage, the observed outbreak eventually declined to extinction. Another difficulty in controlling outbreaks is that it is extremely rare for geriatric hospitals to have an attending dermatologist [16]. As with this outbreak, it seems common to keep the outbreak information confidential. Thus, sufficient institutional and administrative support should be provided to enhance hospital-based control practices.
Conclusion
This study documented a ward-scale scabies outbreak that lasted for almost 10 months. Initiation of control measures was delayed and, as a result, spatial distribution of cases showed no localized patterns in the later phase of the outbreak. Movement of individuals was identified as a dementia-specific risk factor, while type of room and sleeping arrangement were not significantly associated with infection. Dementia patients tend to have non-specific contacts that are sometimes unpredictable, indicating the necessity of real-time observations of patients' behaviours. Given the requirements of identifying specific risk factors of scabies in dementia wards, we believe our study partly emphasizes the need to enhance and establish hospital-based control practices in long-term care facilities [18].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MT carried out direct investigations of the outbreak, proposed and outlined the study, and was in charge of data handling. HN contributed to the study design and performed statistical analyses. MT and HN reviewed and drafted the manuscript. TK reviewed and commented on the early version of the manuscript. All authors have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional file 1
Case descriptions Details of individual clinical courses and discussions of management proposals are provided in pdf format.
Click here for file
Acknowledgements
We acknowledge the hospital for permitting us to document the outbreak. HN thanks Banyu Life Science Foundation International for supporting his study (Banyu Fellowship Programme) in Germany.
==== Refs
Juranek DD Currier RW Millikan LE Orkin M, Maiback HI Scabies control in institutions Cutaneous infestations and insect bites 1985 New York: Dekker
Burgess I Sarcoptes scabieiand scabies Adv Parasitol 1994 33 235 292 8122567
Centers for Disease Control (CDC) Scabies in health-care facilities – Iowa MMWR Morb Mortal Wkly Rep 1988 37 178 179 3126388
Larrosa A Cortes-Blanco M Martinez S Clerencia C Urdaniz LJ Urban J Garcia J Nosocomial outbreak of scabies in a hospital in Spain Euro Surveill 2003 8 199 203 14605375
Development R, Core team R; a language and environment for statistical computing 2004 Vienna: R Foundation for Statistical Computing
Tronstein AJ Ten per cent Crotonil-N-Ethyl-O-Toluidine ointment (Eurax), a new scabieticidal agent Ohio State Med J 1949 45 889 891
Couperus MJ The use of N-ethyl-o-croton-o-toluidide in the treatment of scabies and various pruritic dermatoses J Invest Dermatol 1949 13 35 43
Codington HB Coghlan SE Lindane for scabies in Bangladesh Lancet 1993 342 677 678 7689677 10.1016/0140-6736(93)91783-I
Meinking TL Taplin D Hermida JL Pardo R Kerdel FA The treatment of scabies with ivermectin N Engl J Med 1995 333 26 30 7776990 10.1056/NEJM199507063330105
Flinders DC De Schweinitz P Pediculosis and scabies Am Fam Physician 2004 69 341 348 14765774
Burkhart CG Scabies: an epidemiologic reassessment Ann Intern Med 1983 98 498 503 6340578
Andrews JR The origin and evolution of host associations of Sarcoptes scabiei and the subfamily Sarcoptinae Murray Acarologia 1983 24 85 94 6408884
Ohtaki M Current status of scabies in geriatric care facilities Practical Dermatology (Hifubyoh Shinryoh) 1997 19 468 472 (in Japanese)
Mellanby K Scabies 1943 London: Oxford University Press
Ohtaki M Outbreak of scabies Med Entomol Zool (Eisei Dobutsu) 1998 46 15 26 (in Japanese)
Tajiri A Inoue S Ogata K Kuroki Y Aoki Y Shimonuri Y Scabies among the elderly Practical Dermatology (Hifubyoh Shinryoh) 1987 9 274 278 (in Japanese)
Smith DR Atkinson R Guo YL Yamagata Z Systemic disease as a risk factor for dermatologic abnormality in elderly nursing home patients Yamanashi Med J 2003 17 99 103
Jiminez-Lucho VE Fallon F Caputo C Ramsey K Role of prolonged surveillance in the eradication of nosocomial scabies in an extended care. Veterans Affairs Medical Center Am J Infect Control 1995 23 44 49 7762874 10.1016/0196-6553(95)90008-X
|
16225694
|
PMC1276794
|
CC BY
|
2021-01-04 16:28:16
|
no
|
BMC Infect Dis. 2005 Oct 14; 5:85
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-85
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-871622974910.1186/1471-2334-5-87Research ArticleA simple and rapid approach for screening of SARS-coronavirus genotypes: an evaluation study Chung Grace TY [email protected] Rossa WK [email protected] Jo LK [email protected] Yongjie [email protected] Stephen SC [email protected] Paul KS [email protected] YM Dennis [email protected] Centre for Emerging Infectious Diseases, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong2 Department of Chemical Pathology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong3 Department of Microbiology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong2005 18 10 2005 5 87 87 8 6 2005 18 10 2005 Copyright © 2005 Chung et al; licensee BioMed Central Ltd.2005Chung et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The Severe Acute Respiratory Syndrome (SARS) was a newly emerged infectious disease which caused a global epidemic in 2002–2003. Sequence analysis of SARS-coronavirus isolates revealed that specific genotypes predominated at different periods of the epidemic. This information can be used as a footprint for tracing the epidemiology of infections and monitor viral evolution. However, direct sequencing analysis of a large number of clinical samples is cumbersome and time consuming. We present here a simple and rapid assay for the screening of SARS-coronavirus genotypes based on the use of fluorogenic oligonucleotide probes for allelic discrimination.
Methods
Thirty SARS patients were recruited. Allelic discrimination assays were developed based on the use of fluorogenic oligonucleotide probes (TaqMan). Genotyping of the SARS-coronavirus isolates obtained from these patients were carried out by the allelic discrimination assays and confirmed by direct sequencing.
Results
Genotyping based on the allelic discrimination assays were fully concordant with direct sequencing. All of the 30 SARS-coronavirus genotypes studied were characteristic of genotypes previously documented to be associated with the latter part of the epidemic. Seven of the isolates contained a previously reported major deletion but in patients not epidemiologically related to the previously studied cohort.
Conclusion
We have developed a simple and accurate method for the characterization and screening of SARS-coronavirus genotypes. It is a promising tool for the study of epidemiological relationships between documented cases during an outbreak.
==== Body
Background
The Severe Acute Respiratory Syndrome (SARS) is a recently emerged infectious disease which led to a global epidemic between 2002 and 2003. A novel coronavirus (SARS-CoV) was identified as the causative agent [1]. The genomic sequence of the SARS-CoV was promptly characterized [2,3]. Thereafter, studies had focused on the early detection of SARS-CoV and the development of diagnostic tools [4-6]. Systematic analysis of the SARS-CoV sequence information have demonstrated that characteristic viral genotypes predominated at certain periods during the course of the outbreak [7-10]. Furthermore, characterization of the viral sequences have been shown to be a useful tool for confirming epidemiological associations between infected individuals as suspected from conventional epidemiological investigations [9-11]. In-depth analysis of the available sequence data on SARS-CoV also revealed that the viral isolates could be readily subclassified into several major genotypes based on nucleotide variations at specific genomic positions [8,12]. In a large-scale phylogenetic analysis of SARS-CoV sequences [8], a 5-nucleotide motif at the GZ02 [GenBank :AY390556] reference nucleotide residues 17,564, 21,721, 22,222, 23,823, and 27,827 was identified to be most useful for distinguishing the major SARS-CoV genotypes. These major viral genotypes predominated at different periods of the epidemic [8]. Thus, it is evident that viral sequence and molecular epidemiological data provide valuable information and tools for our combat against infectious diseases. However, direct sequencing of viral isolates from a large number of clinical samples is cumbersome and time consuming. Therefore, a rapid system for the characterization and screening of viral genotypes, such as for SARS-CoV, would potentially be useful. In this study, we demonstrate the feasibility of the adoption of allelic discrimination assays based on the use of fluorogenic oligonucleotide probes for the genotyping of SARS-CoV isolates.
Methods
Study population
Viral culture isolates from 30 SARS patients who were admitted to the hospitals of the New Territories East Cluster of Hong Kong during the SARS epidemic were retrieved. The study was approved by the Institutional Review Board. SARS was confirmed in all cases either by positive reverse transcription-polymerase chain reaction (RT-PCR) detection of SARS-CoV RNA in clinical specimens or documented seroconversion.
Genotype analysis by Taqman allelic discrimination assay
We focused on the development of allelic discrimination assays for the five previously described characteristic single nucleotide variations (SNV) [8]. RNA was extracted from viral isolates cultured from SARS patients' clinical specimens using the QIAamp viral RNA mini kit (Qiagen, Valencia, CA, USA), according to manufacturer's instructions. Eleven microliters of the extracted viral RNA was reverse transcribed by Superscript III (Invitrogen, Carlsbad, CA, USA) with random hexamer according to manufacturer's instructions. Genotyping of the five SNVs was determined using TaqMan (Applied Biosystems, Foster City, CA, USA) allelic discrimination assays on an ABI Prism 7900HT sequence detection system (Applied Biosystems). Each assay consisted of two allele-specific minor groove binding probes associated with either, 6-carboxyfluorescein (FAM) or VIC™ as the fluorescent label, for the discrimination of the two respective alleles at each SNV locus. One assay was designed for each of the 5 SNVs. The primer and probe sequences, designed using the Primer Express 2.0 software (Applied Biosystems) are listed in Table 1. The probes were designed such that the discriminatory nucleotide is placed close to the middle portion of the oligonucleotide. The assays were set up according to the manufacturer's instructions (TaqMan Core PCR Kit; Applied Biosystems) in a reaction volume of 25 μL. Each reaction consists of 1X Buffer A, 4 mM MgCl2, 0.2 μM dATP, 0.2 μM dCTP, 0.2 μM dGTP and 0.4 μM dUTP, 900 nM forward and reverse primers, 200 nM of each fluorescent probe, 0.25U UNG, 0.625U Taq polymerase and 0.5 μl of cDNA as template. The thermal profile consists of an initial incubation at 50°C for 2 min, and then a denaturation period at 95°C for 10 min, followed by 40 cycles of denaturation at 92°C for 15 s, and 1 min of combined annealing and extension at 60°C. The genotypes were scored with the SDS2.1 software.
Table 1 Primers and probes
Sequence (all sequence starts from the 5' end)
SNV 17564*
Forward primer GACACTGTGAGTGCTTTAGTTTATGACA
Reverse primer CCTTTGTAGAACATTTTGAAGCATTG
Probes FAM-AGCTGACTTATCCTTGTGT
VIC-AGCTGACTTCTCCTTGTGT
Synthetic template for allele T GTTGACACTGTGAGTGCTTTAGTTTATGACAATAAGCTAAAAGCACACAAGGATAAGTCAGCTCAATGCTTCAAAATGTTCTACAAAGGTGT
Synthetic template for allele G GTTGACACTGTGAGTGCTTTAGTTTATGACAATAAGCTAAAAGCACACAAGGAGAAGTCAGCTCAATGCTTCAAAATGTTCTACAAAGGTGT
SNV 21721
Forward primer CCATTTTATTCTAATGTTACAGGGTTTCA
Reverse primer TTTCTCTGTGGCAGCAAAATAAATAC
Probes FAM-ATACGTTTGGCAACCCTGTC
VIC-ATACGTTTGACAACCCTGTC
Synthetic template for allele G CTTCCATTTTATTCTAATGTTACAGGGTTTCATACTATTAATCATACGTTTGGCAACCCTGTCATACCTTTTAAGGATGGTATTTATTTTGCTGCCACAGAGAAATCA
Synthetic template for allele A CTTCCATTTTATTCTAATGTTACAGGGTTTCATACTATTAATCATACGTTTGACAACCCTGTCATACCTTTTAAGGATGGTATTTATTTTGCTGCCACAGAGAAATCA
SNV 22222
Forward primer GAGCCATTCTTACAGCCTTTTTA
Reverse primer GCCAACAAAATAGGCTGCAG
Probes FAM-TGCTCAAGACACTTGGG-MGB
VIC-TGCTCAAGACATTTGGG-MGB
Synthetic template for allele C GCCATTCTTACAGCCTTTTTACCTGCTCAAGACACTTGGGGCACGTCAGCTGCAGCCTATTTTGTTGGCTATTTAAAGCCAACTACATTTATGCTCAAGTATGATG
Synthetic template for allele T GCCATTCTTACAGCCTTTTTACCTGCTCAAGACATTTGGGGCACGTCAGCTGCAGCCTATTTTGTTGGCTATTTAAAGCCAACTACATTTATGCTCAAGTATGATG
SNV 23823
Forward primer TCGCTCAAGTCAAACAAATGTACA
Reverse primer GAGGGTCAGGTAATATTTGTGAAAAATT
Probes FAM-CCAACTTTGAAATATTTTGG
VIC-CAACTTTGAAAGATTTTGG
Synthetic template for allele T TGTTCGCTCAAGTCAAACAAATGTACAAAACCCCAACTTTGAAATATTTTGGTGGTTTTAATTTTTCACAAATATTACCTGACCCTCTAA
Synthetic template for allele G TGTTCGCTCAAGTCAAACAAATGTACAAAACCCCAACTTTGAAAGATTTTGGTGGTTTTAATTTTTCACAAATATTACCTGACCCTCTAA
SNV 27827
Forward primer TCATTGTTTTGACTTGTATTTCTCTATGC
Reverse primer CTTCAAGCACATGAGGTTTATTAGATG
Probes FAM-TTGCATATGCACTGTAGT
VIC-TTGCATACGCACTGTAGT
Synthetic template for allele C TTCTCATTGTTTTGACTTGTATTTCTCTATGCAGTTGCATATGCACTGTAGTACAGCGCTGTGCATCTAATAAACCTCATGTGCTTGAAGATCC
Synthetic template for allele T TTCTCATTGTTTTGACTTGTATTTCTCTATGCAGTTGCATACGCACTGTAGTACAGCGCTGTGCATCTAATAAACCTCATGTGCTTGAAGATCC
* probes for SNV 17564 are anti-sense sequences
Sequence confirmation by direct sequencing
All viral sequences were confirmed by direct sequencing. RT-PCR was performed to specifically amplify genomic segments of SARS-CoV encompassing each of the 5 SNVs using primers and protocols previously described [7]. The DNA of each amplicon was sequenced by the dideoxy terminator method on an automated DNA sequencer (3100 Genetic Analyzer, Applied Biosystems) based on capillary electrophoresis.
Results and discussion
Taqman allelic discrimination assays for the 5 SNVs were first tested on synthetic templates (Sigma Genosys, Australia) (Table 1) and verified using 2 viral isolates, CUHK-W1 [GenBank :AY278554] and CUHK-Su10 [GenBank :AY282752]. CUHK-W1 is a SARS-CoV isolate with a G:A:C:T:C motif at the GZ02 reference nucleotide residues 17,564, 21,721, 22,222, 23,823, and 27,827, characteristic of SARS-CoV strains isolated before worldwide dissemination of SARS [8,13]. On the other hand, CUHK-Su10 demonstrates a T:G:T:T:T motif which is characteristic of SARS-CoV strains isolated after global spread was evident. As evident from figure 1, the newly developed allelic discrimination assays were able to differentiate the 2 viral isolates and genotype each SNV correctly (Table 1).
Figure 1 Allelic discrimination plot of CUHK-Su10 and CUHK-W1. Allelic discrimination at each of the 5 studied SNVs described in the text as demonstrated using the synthetic templates and cDNA from CUHK-Su10 and CUHK-W1 Vero cell culture isolates is presented in the successive plots. () synthetic template for the FAM-labeled allele, () synthetic template for the VIC-labeled allele, () CUHK-W1, () CUHK-Su10, () no template control.
Following initial development and optimization, the allelic discrimination assays were used to genotype SARS-CoV in clinical samples. We were able to successfully determine the SARS-CoV genotypes in all 30 samples. Genotypes of virus isolates at the 5 SNV positions are shown in Table 2. SARS-CoV from all but seven cases showed the T:G:T:T:T motif resembling that of the CUHK-Su 10 isolate. For the remaining seven cases, no allelic signal was detected at SNV position 27,827 (Figure 2), leading to a T:G:T:T:/ motif.
Table 2 Genotype of SARS-CoV culture isolates from 30 patients determined by Taqman Allelic Discrimination assays
Case number Nucleotide positions
17,564 21,721 22,222 23,823 27,827
TC01 T G T T T
TC02 T G T T T
TC07 T G T T T
TC10 T G T T T
TC11 T G T T T
TC14 T G T T T
TC19 T G T T T
TC20 T G T T T
TC22 T G T T T
TC26 T G T T T
TC27 T G T T T
TC28 T G T T T
TC29 T G T T ND
TC30 T G T T T
TC32 T G T T ND
TC34 T G T T T
TC37 T G T T T
TC38 T G T T T
TC39 T G T T T
TC41 T G T T T
TC42 T G T T T
TC43 T G T T T
TC44 T G T T T
TC45 T G T T ND
TC46 T G T T ND
TC48 T G T T T
TC51 T G T T ND
TC52 T G T T T
TC55 T G T T ND
TC59 T G T T ND
CUHK-W1 [AY278554] G A C T C
CUHK-Su10 [AY282752] T G T T T
ND: no allelic signal
Figure 2 Allelic discrimination plot of 30 SARS patients using the allelic discrimination assay on SNV 27827. () synthetic template for FAM-labeled allele, () synthetic template for VIC-labeled allele, () patients with SARS-CoV showing a T nucleotide at SARS-CoV SNV 27827, () patients with SARS-CoV showing no allelic signal at SNV 27827. This SARS-CoV variant was confirmed to contain a 386 nt deletion encompassing SNV 27827, () no template control.
The SARS-CoV genotypes isolated from the 30 patients were also confirmed by direct sequencing. The sequencing results were fully concordant with that based on the allelic discrimination assays at all the 5 SNVs. The seven samples which gave no allelic signal by the allelic discrimination assay at SNV 27,827 showed a shortened amplicon encompassing the region. Direct sequencing of this short amplicon revealed a deletion of 386 nt identical to a SARS-CoV deletion variant previously reported by our group [8,9]. This deletion variant was first isolated from a discrete cohort of 15 epidemiologically related SARS patients [9]. In the previous cohort of patients, the origin of the deletion variant was traceable to mid-April 2003 in two patients residing in an estate, T, in Hong Kong with subsequent spread predominantly at the North District Hospital, Hong Kong [9]. To further determine if the newly identified cases were epidemiologically related to the original patient cohort, the case histories were reviewed.
The seven patients had fever onset between April 4 to 15, 2003 which predated the disease onset dates of all cases in the original patient cohort. All but one patient presented to hospital A initially for other medical conditions without fever or evidence of chest infection and appeared to have acquired SARS nosocomially during admission. The remaining patient was a resident at Estate T and presented to hospital A with symptoms and signs of chest infection on April 8, 2003 and SARS was subsequently confirmed. Thus, it appeared that we have identified another cohort of patients harboring the SARS-CoV variant with the 386-nt deletion. It is interesting to note that this study provided additional anecdotal evidence pointing to Estate T as a propagation site for the deletion variant. In addition, we were able to trace the emergence of this deletion variant to early April 2003, weeks before the first appearance reported previously [9].
Our study has clearly demonstrated the feasibility of using allelic discrimination assays as a method for genetic characterization of SARS-CoV genotypes in patients. It is particularly useful when there is already extensive sequence information. Direct sequencing is still the gold standard for identifying new sequence variations when new agents of infectious disease continue to emerge and old ones reemerge. Once the variations have been identified, allelic discrimination assay is more efficient and suitable for large-scale population investigations. A recent study illustrated the use of mass spectrometry-based technology in characterizing SARS sequence variations [14]. However, this method requires post-PCR manipulations and the availability of specialized equipment. On the other hand, allelic discrimination assays have been widely used in the study of associations between single nucleotide polymorphisms and diseases such as cancers [15] and rheumatoid arthritis [16]. The validity of the approach for single nucleotide polymorphism genotyping has been previously demonstrated [17-19]. Thus, this study further extended the usefulness of allelic discrimination approach based on fluorogenic oligonucleotide probes. The approach provides a rapid and simple means to accurate genotype screening, making it ideal for epidemiological investigations.
Conclusion
We have evaluated a rapid approach for characterizing SARS-CoV genotypes. The assay is simple, easy to perform and reproducible. It can therefore be used as an efficient means to screen for virus genotypes and track the transmission of a particular viral strain in times of epidemics. Incidentally, we identified a previously reported deletion variant of the SARS-CoV in a new cohort of patients and traced the emergence of this variant to an earlier date than previously reported.
List of abbreviations
SARS: Severe acute respiratory syndrome
SARS-CoV: SARS-coronavirus
RT-PCR: reverse transcription-polymerase chain reaction
SNV: single nucleotide variation
Competing interests
YMDL, RWKC and SSCC have filed patent applications on aspects concerning the genomics and detection of the SARS-coronavirus.
Authors' contributions
GTYC, RWKC and YMDL have contributed in the preparation of the manuscript and the overall study design. GTYC, RWKC and YJ have contributed in the assay designs, data analysis and conducting the genotyping experiments. RWKC, SSCC, JLKC and PKSC have contributed in the collection and analysis of clinical data from the patients. JLKC and PKSC provided the viral samples.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The work is supported by the Research Fund for the Control of Infectious Disease (RFCID) from the Health, Welfare and Food Bureau of the Hong Kong Government SAR.
==== Refs
Fouchier RA Kuiken T Schutten M van Amerongen G van Doornum GJ van den Hoogen BG Peiris M Lim W Stohr K Osterhaus AD Aetiology: Koch's postulates fulfilled for SARS virus Nature 2003 423 240 12748632 10.1038/423240a
Marra MA Jones SJ Astell CR Holt RA Brooks-Wilson A Butterfield YS Khattra J Asano JK Barber SA Chan SY Cloutier A Coughlin SM Freeman D Girn N Griffith OL Leach SR Mayo M McDonald H Montgomery SB Pandoh PK Petrescu AS Robertson AG Schein JE Siddiqui A Smailus DE Stott JM Yang GS Plummer F Andonov A Artsob H Bastien N Bernard K Booth TF Bowness D Czub M Drebot M Fernando L Flick R Garbutt M Gray M Grolla A Jones S Feldmann H Meyers A Kabani A Li Y Normand S Stroher U Tipples GA Tyler S Vogrig R Ward D Watson B Brunham RC Krajden M Petric M Skowronski DM Upton C Roper RL The Genome sequence of the SARS-associated coronavirus Science 2003 300 1399 1404 12730501 10.1126/science.1085953
Rota PA Oberste MS Monroe SS Nix WA Campagnoli R Icenogle JP Penaranda S Bankamp B Maher K Chen MH Tong S Tamin A Lowe L Frace M DeRisi JL Chen Q Wang D Erdman DD Peret TC Burns C Ksiazek TG Rollin PE Sanchez A Liffick S Holloway B Limor J McCaustland K Olsen-Rasmussen M Fouchier R Gunther S Osterhaus AD Drosten C Pallansch MA Anderson LJ Bellini WJ Characterization of a novel coronavirus associated with severe acute respiratory syndrome Science 2003 300 1394 1399 12730500 10.1126/science.1085952
Ng EK Hui DS Chan KC Hung EC Chiu RW Lee N Wu A Chim SS Tong YK Sung JJ Tam JS Lo YM Quantitative analysis and prognostic implication of SARS coronavirus RNA in the plasma and serum of patients with severe acute respiratory syndrome Clin Chem 2003 49 1976 1980 14633867 10.1373/clinchem.2003.024125
Wang B Potter SJ Lin Y Cunningham AL Dwyer DE Su Y Ma X Hou Y Saksena NK Rapid and sensitive detection of severe acute respiratory syndrome coronavirus by rolling circle amplification J Clin Microbiol 2005 43 2339 2344 15872263 10.1128/JCM.43.5.2339-2344.2005
Poon LL Chan KH Wong OK Yam WC Yuen KY Guan Y Lo YM Peiris JS Early diagnosis of SARS coronavirus infection by real time RT-PCR J Clin Virol 2003 28 233 238 14522060 10.1016/j.jcv.2003.08.004
Chim SS Tong YK Hung EC Chiu RW Lo YM Genomic sequencing of a SARS coronavirus isolate that predated the Metropole Hotel case cluster in Hong Kong Clin Chem 2004 50 231 233 14709660 10.1373/clinchem.2003.025536
Chinese SMEC Molecular evolution of the SARS coronavirus during the course of the SARS epidemic in China Science 2004 303 1666 1669 14752165 10.1126/science.1092002
Chiu RW Chim SS Tong YK Fung KS Chan PK Zhao GP Lo YM Tracing SARS-coronavirus variant with large genomic deletion Emerg Infect Dis 2005 11 168 170 15714661
Guan Y Peiris JS Zheng B Poon LL Chan KH Zeng FY Chan CW Chan MN Chen JD Chow KY Hon CC Hui KH Li J Li VY Wang Y Leung SW Yuen KY Leung FC Molecular epidemiology of the novel coronavirus that causes severe acute respiratory syndrome Lancet 2004 363 99 104 14726162 10.1016/S0140-6736(03)15259-2
Chiu RW Chim SS Lo YM Molecular epidemiology of SARS--from Amoy Gardens to Taiwan N Engl J Med 2003 349 1875 1876 14602893 10.1056/NEJM200311063491923
Ruan YJ Wei CL Ee AL Vega VB Thoreau H Su ST Chia JM Ng P Chiu KP Lim L Zhang T Peng CK Lin EO Lee NM Yee SL Ng LF Chee RE Stanton LW Long PM Liu ET Comparative full-length genome sequence analysis of 14 SARS coronavirus isolates and common mutations associated with putative origins of infection Lancet 2003 361 1779 1785 12781537 10.1016/S0140-6736(03)13414-9
Tsui SK Chim SS Lo YM Coronavirus genomic-sequence variations and the epidemiology of the severe acute respiratory syndrome N Engl J Med 2003 349 187 188 12853594 10.1056/NEJM200307103490216
Liu J Lim SL Ruan Y Ling AE Ng LF Drosten C Liu ET Stanton LW Hibberd ML SARS Transmission Pattern in Singapore Reassessed by Viral Sequence Variation Analysis PLoS Med 2005 2 e43 15736999 10.1371/journal.pmed.0020043
Hughes DJ Ginolhac SM Coupier I Corbex M Bressac-de-Paillerets B Chompret A Bignon YJ Uhrhammer N Lasset C Giraud S Hardouin A Berthet P Peyrat JP Fournier J Nogues C Lidereau R Muller D Fricker JP Longy M Toulas C Guimbaud R Maugard C Olschwang S Yannoukakos D Durocher F Moisan AM Simard J Mazoyer S Lynch HT Szabo C Lenoir GM Goldgar DE Stoppa-Lyonnet D Sinilnikova OM Common BRCA2 variants and modification of breast and ovarian cancer risk in BRCA1 mutation carriers Cancer Epidemiol Biomarkers Prev 2005 14 265 267 15668505
Kilding R Akil M Till S Amos R Winfield J Iles MM Wilson AG A biologically important single nucleotide polymorphism within the toll-like receptor-4 gene is not associated with rheumatoid arthritis Clin Exp Rheumatol 2003 21 340 342 12846053
Letellier C Kerkhofs P Real-time PCR for simultaneous detection and genotyping of bovine viral diarrhea virus J Virol Methods 2003 114 21 27 14599675 10.1016/j.jviromet.2003.08.004
Campsall PA Au NH Prendiville JS Speert DP Tan R Thomas EE Detection and genotyping of varicella-zoster virus by TaqMan allelic discrimination real-time PCR J Clin Microbiol 2004 42 1409 1413 15070981 10.1128/JCM.42.4.1409-1413.2004
Namvar L Olofsson S Bergstrom T Lindh M Detection and typing of Herpes Simplex virus (HSV) in mucocutaneous samples by TaqMan PCR targeting a gB segment homologous for HSV types 1 and 2 J Clin Microbiol 2005 43 2058 2064 15872222 10.1128/JCM.43.5.2058-2064.2005
|
16229749
|
PMC1276795
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Oct 18; 5:87
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-87
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-891624201310.1186/1471-2334-5-89Research ArticleThe burden of diarrhoea, shigellosis, and cholera in North Jakarta, Indonesia: findings from 24 months surveillance Agtini Magdarina D [email protected] Rooswanti [email protected] Murad [email protected] Narain H [email protected] Cyrus [email protected] Ferry [email protected] Dazwir [email protected] Sri Pandam [email protected] Ainur [email protected] Hari [email protected] H [email protected] Agus [email protected] Pratiwi [email protected] Seidlein Lorenz [email protected] Jacqueline L [email protected] Mohammad [email protected] Hyejon [email protected] Deok Ryun [email protected] Oakpil [email protected] Jin Kyung [email protected] Agus [email protected] [email protected] Buhari A [email protected] James R [email protected] H James [email protected] Andrew L [email protected] John D [email protected] National Institute of Health Research and Development, Jakarta Indonesia, Ministry of Health, Jakarta, Indonesia2 United States Navy Medical Research Unit 2, Jakarta, Indonesia3 Infectious Disease Hospital Sulianti Saroso, Ministry of Health, Jakarta, Indonesia4 Communicable Disease Control and Environmental Health, Ministry of Health, Jakarta, Indonesia5 Department of Microbiology, University of Indonesia, Jakarta, Indonesia6 International Vaccine Institute, Seoul, Korea7 National Institute of Child Health and Human Development, Bethesda, Maryland, USA2005 20 10 2005 5 89 89 24 6 2005 20 10 2005 Copyright © 2005 Agtini et al; licensee BioMed Central Ltd.2005Agtini et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In preparation of vaccines trials to estimate protection against shigellosis and cholera we conducted a two-year community-based surveillance study in an impoverished area of North Jakarta which provided updated information on the disease burden in the area.
Methods
We conducted a two-year community-based surveillance study from August 2001 to July 2003 in an impoverished area of North Jakarta to assess the burden of diarrhoea, shigellosis, and cholera. At participating health care providers, a case report form was completed and stool sample collected from cases presenting with diarrhoea.
Results
Infants had the highest incidences of diarrhoea (759/1 000/year) and cholera (4/1 000/year). Diarrhea incidence was significantly higher in boys under 5 years (387/1 000/year) than girls under 5 years (309/1 000/year; p < 0.001). Children aged 1 to 2 years had the highest incidence of shigellosis (32/1 000/year). Shigella flexneri was the most common Shigella species isolated and 73% to 95% of these isolates were resistant to ampicillin, trimethoprim-sulfamethoxazole, chloramphenicol and tetracycline but remain susceptible to nalidixic acid, ciprofloxacin, and ceftriaxone. We found an overall incidence of cholera of 0.5/1 000/year. Cholera was most common in children, with the highest incidence at 4/1 000/year in those less than 1 year of age. Of the 154 V. cholerae O1 isolates, 89 (58%) were of the El Tor Ogawa serotype and 65 (42%) were El Tor Inaba. Thirty-four percent of patients with cholera were intravenously rehydrated and 22% required hospitalization. V. parahaemolyticus infections were detected sporadically but increased from July 2002 onwards.
Conclusion
Diarrhoea causes a heavy public health burden in Jakarta particularly in young children. The impact of shigellosis is exacerbated by the threat of antimicrobial resistance, whereas that of cholera is aggravated by its severe manifestations.
==== Body
Introduction
Diarrhoeal diseases remain a major cause of morbidity and mortality in all age groups in impoverished areas of South East Asia [1-5]. In Indonesia, diarrhoea is the third leading cause of overall morbidity and the leading cause of infant mortality [6,7]. Earlier studies extensively explored the causative organisms of diarrhoea in the slums of Indonesia's capital, Jakarta [8-11]. In a study conducted from 1997 to 1999, Shigella flexneri was found to be the most frequently isolated organism from diarrhoea patients in a community setting in Jakarta [9]. Similarly, S. flexneri was the most common organism isolated in eight hospitals throughout the Indonesian archipelago between 1999 and 2000 [10]. Cholera is another persistent problem in Indonesia. In studies of diarrhoeal outbreaks throughout Indonesia from 1993 to 1999, Vibrio cholerae O1 was isolated from 8% to 11% of stool samples collected [8,12,13].
To plan trials for the evaluation of vaccine candidates it is essential to understand the disease incidence in the study area. In preparation of vaccines trials to estimate protection against shigellosis and cholera we conducted a two-year community-based surveillance study in an impoverished area of North Jakarta. The experience gained during the surveillance will be useful for future vaccine trials as well as for the understanding of endemic disease rates in urban slums in the region.
Methods
Study area and population
North Jakarta, one of five municipal administrative areas of the city, is 154 km2, had a population of 1 435 207 in 2000, and an estimated population density of 9 314 individuals/square kilometre. The average annual income per person in this congested urban site was US$ 689. – in 2000 [14]. Water supply and sanitation are inadequate. Many homes are temporary structures without running water and more than a third of households have no access to tap water. Two adjacent districts (kecamatans) in North Jakarta, Tanjung Priok and Koja, were selected for the study based on expected high incidence of the target diseases, accessibility and previous research experience in the area [15,16]. Residents of the socio-economically better-off areas of the study site did not participate in the study due to their known preference for private health care. The total population enumerated by a study census in 2001 was 160 261 individuals of whom 15 741 (10%) were less than 60 months of age [17]. The local climate has 2 distinctive seasons: a rainy season (December–April) and a dry season (May–November)[18].
Health care system
In the Indonesian public sector, the first-level health care facility is the public health centre, locally known as the puskesmas. North Jakarta has 54 public health centres, 22 of which are located in Tanjung Priok and Koja. More severe conditions are referred to government hospitals. The Infectious Disease Hospital and Koja Hospital are the main government referral hospitals in North Jakarta. In 2001, there were 314 private practitioners, 36 polyclinics, 32 maternal clinics, and 29 mostly very small private hospitals in the study area.
Study surveillance procedures
The surveillance was conducted from August 2001 to July 2003. The study was designed to detect all patients residing in the two study districts presenting with diarrhoea at participating health care providers: public health centres in the study area, the Infectious Disease Hospital, and Koja Hospital. A 2001 survey in the study area asked 8 074 households that had diarrhoea cases in the preceding four weeks where care was sought [17], and 39% reported they would use a treatment provider taking part in the surveillance. Based on these findings, we estimate that 61% of diarrhoea episodes were not captured by our surveillance system.
In the present study, we invited consenting patients residing in the study area of all age groups presenting to a participating health care provider with loose bowel movement to join the study. Diarrhoea was defined as passage of three or more loose stools in 24 hours or one or more loose stool with visible blood. A new diarrhoea episode was defined if the diarrhoea definition was met after three of more days free of diarrhoea or dysentery. Shigellosis was defined as diarrhoea with isolation of Shigella species from a stool sample. Cholera was defined as diarrhoea with isolation of V. cholerae O1 or O139 from a stool sample. Fever was defined as axillary temperature of 37.5°C or higher.
Study personnel were trained in study procedures. For every patient presenting with diarrhoea agreeing to take part in the study, a case report form was completed describing address, demographics, medical history, physical examination results, and the treatment plan including the prescription of antibiotics. Patients who did not consent or had no diarrhea were excluded. Three rectal swabs were obtained from all study participants. Treatment was provided to patients in accordance with national guidelines and the treatment plan including hospital referral was recorded. The study did not include follow-up visits of patients.
For isolation of Shigella species, one rectal swab was placed in buffered glycerol saline (BGS) and another in potassium buffered saline. Both swabs were refrigerated and transported in a cool box to the central laboratory by motorcycle and plated on the same day they were obtained.
The specimens in BGS were plated on MacConkey agar and Salmonella-Shigella agar. Biochemical reactions of colonies were evaluated in Kligler's iron agar and motility indole ornithine medium. Colonies were serologically confirmed by slide agglutination with appropriate group-specific polyvalent antiserum, followed by type-specific monovalent antisera. Standardized, commercial antisera (Denka Seiken Co., Ltd. 3-4-2, Nihonbashi, Kayaba-Cho, Chuo-Ku, Tokyo 103-0025, Japan) were used for identification. In cases where no agglutination occurred with live bacteria, the test was repeated with boiled suspensions of bacteria.
For isolation of Vibrios, one rectal swab was placed in Cary-Blair transport medium, kept at room temperature, and transported to the central laboratory by motorcycle and plated on the same day that were obtained. From the Cary-Blair media, the specimens were plated directly onto thiosulfate citrate bile salt sucrose (TCBS) agar (Eiken Chemical Company, Tokyo, Japan) and also plated onto TCBS after enrichment in alkaline peptone water for 6 and 20 hours (pH 8.6, 37°C). After incubation overnight, suspected colonies on the TCBS plates were tested biochemically and agglutinated with polyvalent, Ogawa, and Inaba antisera (Difco Laboratories, Detroit, Michigan). Non-agglutinating strains were tested with antiserum to V. cholerae O139 strain. Identification of Vibrio parahaemolyticus was done according to standard methods [19].
Antimicrobial susceptibility testing of Shigella and V. cholerae isolates against ampicillin, trimethoprim-sulfamethoxazole, chloramphenicol, tetracycline, cefotaxime, ceftriaxone, ciprofloxacin, and nalidixic acid was performed using standard antibiotic discs (Becton, Dickinson and Co., Sparks, MD). Isolates were assessed as being resistant, intermediate or susceptible according to standard cut-off zone sizes [20]. A control strain, Escherichia coli ATCC 25922, was included in the test.
Data management and analysis
All case report forms were double entered into custom-made data entry programs using FoxPro software (Microsoft, Redmond, WA). The data management programs include error as well as consistency check programs. SAS software (SAS Institute Inc., Cary, NC) was used for statistical analyses. Because ascertainment of cases was passive, we did not know how many of the people included in the study stayed in the study area. We estimated person-years at risk assuming the study population stayed in the study area until the end of the study period. We used the age-specific number of disease episodes in residents of the study area as the numerator. Incidence was calculated based on the population residing in the catchment area in 2001 as the denominator. A test for trend (chi square) was applied to assess the statistical significance of increasing incidence rates of shigellosis with increasing age (after age 30 years). For nonparametric data, Wilcoxon rank sum test was used for the comparison of two groups. Chi square test was used for the analysis of binary data. Student's t test or chi square test was used, as appropriate, to compare clinical features of cases by etiologic organism.
Ethics
After the project's purpose was explained, patients, or in the case of minors, their parent or guardian, gave verbal consent prior to participation in the study. Participation consisted of providing a stool specimen and the information required to complete the case report forms. The study was approved by the Ethics Committee of the Ministry of Health, Indonesia; the Institutional Review Board, National Institute of Health Research and Development, Ministry of Health, Jakarta, Indonesia; the Institutional Review Board, United States Naval Medical Research Unit No 2; and from the Secretariat Committee on Research Involving Human Subjects, World Health Organization, Geneva, Switzerland.
Results
Diarrhoea
From August 2001 through July 2003, 16 872 patients presented to one of the participating treatment centres with loose bowel movements and agreed to take part in the study (Figure 1). 647 (4%) who did not fulfil the study criteria (consent and diarrhea) were excluded and 16 225 (96%) diarrhoea cases were analyzed. In all, 1 203 Shigella and 488 Vibrio organisms were isolated from the stool samples. More than one organism was isolated from 38 patients. We found an overall incidence of diarrhoea of 50 per 1 000 population per year. The highest diarrhoea incidence was in infants less than 1 year of age at 759 cases/1 000/year (Figure 2). Sixty-eight percent (10 998/16 225) of diarrhoea cases were in children less than 5 years of age (incidence of 349/1 000/year). Diarrhea incidence was significantly higher in boys under 5 years (387/1 000/year) than girls under 5 years (309/1 000/year; p < 0.001). In contrast the reported diarrhea incidence in females 5 years and older (21/1 000/year) was significantly higher than in males 5 years or older (16/1 000/year; p < 0.001). Diarrhoea cases had a clear seasonality. During the first three months of the year (January, February, March) 1578/16225 (38%) of diarrhea episodes were captured and during the remainig 9 months 10 026 (62%) of the diarrhea episodes were captured (p < 0.001; Figure 3).
Figure 1 Flow of diarrhoea patients.
Figure 2 Incidence of diarrhoea by sex and age group, North Jakarta, August 2001 to July 2003.
Figure 3 Seasonality of diarrhoea, shigellosis and Vibrio infections, North Jakarta a. Diarrhoea, b. Shigellosis, c. Vibrio infections.
Shigellosis
We found an overall incidence of shigellosis of 4/1 000/year. Shigellosis was most common among children and the elderly (Figure 4), with the highest incidence at 32/1 000/year in those 1 to 2 years of age. Of the 1 203 Shigella isolates, 866 (72%) were S. flexneri, 277 (23%) were S. sonnei, 21 (2%) were S. dysenteriae (none of which were S. dysenteriae type 1), and 39 (3%) were S. boydii (Figure 1). Incidence by type of specie was 2.6/1 000/year for S. flexneri, 0.8 for S. sonnei, 0.1 for S. dysenteriae, and 0.1 for S. boydii. The most frequently encountered S. flexneri serotypes were 2a and 3a, followed by 1b and 1c (Table 1). Two percent of the S. flexneri isolates could not be typed with commercially available antisera. We evaluated the antimicrobial resistance pattern of the Shigella isolates by species. In all, 73% to 95% of S. flexneri isolates were resistant to ampicillin, trimethoprim-sulfamethoxazole, chloramphenicol, and tetracycline. None showed resistance to ciprofloxacin and only a single isolate (of S. flexneri) was not susceptible to ceftriaxone and nalidixic acid. Significantly more shigellosis cases were detected between February and April (418/1772; 36%) during the rainy months compared to the remaining 9 months of the year (754/1172; 64%; p < 0.001; Figure 3).
Figure 4 Incidence of shigellosis by Shigella species and age group infected, North Jakarta, August 2001 through July 2003.
Table 1 Shigella flexneri serotypes detected in North Jakarta, August 2001 to July 2003
Serotypes Number (%)
1a 63 7
1b 105 12
1c 103 12
2a 297 34
2b 5 1
3a 142 16
3b 10 1
4 12 1
4a 53 6
4b 1 <1
4X 11 1
5a 2 <1
6 38 4
X 2 <1
Y 5 1
Non-typeable 17 2
Total 866 100
Cholera and other vibriosis
We found an overall incidence of cholera of 0.5/1 000/year. Cholera was most common in children (Figure 5), with the highest incidence at 4/1 000/year in those less than 1 year of age. Of the 154 V. cholerae O1 isolates, 89 (58%) were of the El Tor Ogawa serotype and 65 (42%) were El Tor Inaba. V. cholerae O139 was not isolated during the two-year study period. The majority of isolates (>90%) remain susceptible to the first-line antimicrobial agents, trimethoprim-sulfamethoxazole and tetracycline as well as nalidixic acid, ciprofloxacin and chloramphenicol. Cholera peaked in December. Significantly more cases were detected between December and March (162/443; 37%) during the cooler months compared to the remaining 9 months of the year (281/443; 37%; p < 0.001; Figure 3).
Figure 5 Incidence of Vibrio cholerae and Vibrio parahaemolyticus by age group, North Jakarta, August 2001 through July 2003.
During the surveillance, other Vibrio organisms were also detected. The incidence of diarrhoea due to V. cholerae non-O1 was 0.5/1 000/year and that due to V. parahaemolyticus was 0.4/1 000/year. V. cholerae non-O1 diarrhoea was most common among children and the elderly, whereas the incidence of V. parahaemolyticus diarrhoea gradually increased with age (Figure 5). Seasonality was not clear for diarrhoea associated with V. cholerae non-O1 and V. parahaemolyticus (Figure 3). Initially, V. parahaemolyticus infections were detected sporadically but increased from July 2002 onwards.
Clinical features and management
We assessed clinical features and management of the diarrhoea cases and separately, those from whom Shigella and Vibrio strains were isolated, excluding the 38 cases from whom more than one organism was isolated (Table 4). The mean and median ages of cases with diarrhoea due to V. parahaemolyticus were significantly higher than those of cases caused by other Vibrio or Shigella species (p < 0.001). Only 10 to 25% of shigellosis cases, depending on the specie, reported blood in the stool. Nearly all patients (range 90 to 100%) were prescribed antibiotics. Ten percent of patients with cholera were severely dehydrated, 34% were intravenously rehydrated, and 22% were referred for hospital admission. Compared to cholera cases, all other diarrhoea cases were less likely to have severe dehydration (P value < 0.001), rehydrated with intravenous fluids (P value < 0.001), or be hospitalized (P value = 0.001).
Discussion
This is the first community-based surveillance of diarrhoea in Jakarta and our findings underscore its public health significance especially amongst children less than 60 months of age. By utilizing passive surveillance, that is detecting patients coming to health centres for treatment, we found an annual incidence of diarrhoea requiring medical attention of 349 per 1 000 children and 759 per 1000 infants.
The study was designed in preparation of vaccine trials trying to find an appropriate population for the evaluation of vaccine candidates. The residents of two kecamatans, Tanjung Priok and Koja were selected as study population for the study based on their accessibility and the relative high incidence of the target diseases. We have found that in this population only 39% of diarrhea episodes overall would trigger a visit to a health care centre. Taking into account the lack of sensitivity of the classic culture methods used in the study it is reasonable to assume that the true disease incidence could be several fold higher than our estimates. Our study population is likely to be representative of the less affluent segment of the population of Jakarta which depends on the public health care system. However the study population was not chosen because it is representative for the population of Jakarta. More affluent households in North Jakarta may prefer to make use of the private sector and were not captured by the study. The incidence of the target diseases estimated by our study may differ in the more affluent population.
The incidence of shigellosis was highest in children between 1 and 2 years of age, consistent with previous data indicating that maternal antibodies and breast-feeding may protect newborns and infants [21-23]. We found increasing rates of shigellosis in those 70 years of age and older, which could be related to waning of immunity. S. flexneri was the most frequently detected Shigella species and four S. flexneri serotypes (2a, 3a, 1b, 1c) made up 74% of the S. flexneri isolates. This finding is consistent with those from a previous report that suggested that the most frequent S. flexneri serotypes are 2a, 1b, and 3b [24].
The highest number of diarrhea cases was detected each year during the rainy season between January and April when North Jakarta is particularly flood prone. The number of detected cholera cases peaked each year in January. Relatively few V. parahaemolyticus strains were detected during the first year of the study which became the dominant Vibrio species towards the end of the study. The increase in V. parahaemolyticus detection may well have been related to the pandemic spread of the V. parahaemolyticus serovar O3:K6 though the serovar of the strains has yet to be determined. The detection of all Shigella strains coincided with the rainy season (January through April) and had a distinct peak in April of the first year of surveillance.
The greatest burden of cholera was in children during the first year of life. Previously, cholera was believed to occur infrequently below 2 years of age [25] but more recent studies have found cholera to be a significant problem in young children [26,27]. We found significant numbers of diarrhoea cases due to other Vibrio species. These diarrhoea episodes were more likely to occur in adults than in children as previously reported [8,28]. V. parahaemolyticus infections are transmitted through unsafe foods, such as raw or undercooked seafood, which are more likely to be consumed by adults [8]. V. cholerae non-O1 and V. parahaemolyticus were associated with milder illness than V. cholerae O1.
The large majority of study patients received rehydration fluids (mostly orally) in accord with international guidelines. Only 4% of diarrhoea patients reported taking antibiotics prior to presentation which is surprising considering the ubiquitous presence of drug vendors in the study area. Resistance of Shigella to quinolones is not yet a major problem in the area but resistance to the more commonly used first-line antimicrobials is disturbing. Ampicillin, tetracycline, chloramphenicol, and trimethoprim-sulfamethoxazole are of little use for treatment of S. flexneri in the study area. Antimicrobial resistance of V. cholerae is not yet a major problem in Jakarta. However the surprisingly high percentage of diarrhea patients who were prescribed antibiotics is worrying as the drug pressure is likely to select resistant strains and is unlikely to benefit many the patients.
Conclusion
In conclusion, diarrhoea causes a heavy public health burden in Jakarta. The impact of shigellosis is exacerbated by the threat of antimicrobial resistance, whereas that of cholera is aggravated by its severe manifestations.
Competing interests
No author has political, personal, religious, academic, intellectual, commercial or any other interests that could lead to a conflict of interest. The views expressed in this paper are those of the authors and do not necessarily represent the views of the United States Department of the Navy or Department of Defence.
Authors' contributions
The principal investigators were CS during the first 12 months of the study and MA during the remainder of the study. NP and RS coordinated the local study activities during the first and second year respectively. ML, DN, BO, JRC, HJB, and ALC were responsible for the microbiological aspects of the study at United States Navy Medical Research Unit 2; ASj and PS coordinated the microbiological component at the University of Indonesia. SPP and HS coordinated the study activities in the hospitals. AR, AS, and I served as liaison with the Ministry of Health, the communities and the investigators. LvS coordinated the epidemiologic aspects of the overall project. DRK and MA developed and managed the database which was administered and supervised by FW, DRK, OH and JKP managed and analyzed the data. JLD wrote the first draft of the manuscript, JC conceived the project, attracted funding and oversaw all stages of the project. All authors contributed to the writing of the final version of this paper.
Table 2 Clinical features and management of diarrhoea cases, by organism isolated (excluding 38 cases with mixed infections)
Diarrhoea cases N = 16,187 Shigella Vibrio
flexneri N = 844 sonnei N = 269 dysenteriae N = 21 boydii N = 38 cholerae O1 N = 143 cholerae non-O1 N = 171 parahaemolyticus N = 129
Mean (median) age in years 11 (2) 19 (11) 7 (2) 22 (8) 23 (20) 15 (7) 14 (3) 33 (32)
Number (%) female 7,643 (47) 479 (57) 135 (50) 7 (33) 25 (66) 78 (55) 84 (49) 75 (58)
Number (%) with liquid stool 15,543 (96) 787 (93) 250 (93) 18 (86) 35 (92) 140 (98) 166 (97) 129 (100)
Number (%) with bloody stool 1,174 (7) 213 (25) 42 (16) 2 (10) 8 (21) 6 (4) 11 (6) 5 (4)
Number (%) with vomiting 7,053 (44) 231 (27) 71 (26) 7 (33) 4 (11) 75 (53) 72 (42) 77 (60)
Number (%) who had taken antibiotics 607 (4) 42 (6) 8 (3) 0 0 5 (4) 3 (2) 1 (1)
Number (%) with severe dehydration 168 (1) 4 (1) 0 0 0 14 (10) 5 (3) 0
Number (%) with fever 5,164 (33) 279 (34) 89 (33) 8 (38) 7 (19) 24 (17) 52 (31) 25 (19)
Number (%) orally rehydrated 11,597 (72) 547 (65) 194 (72) 14 (67) 26 (68) 108 (76) 126 (74) 91 (71)
Number (%) intravenously rehydrated 2,338 (14) 77 (9) 12 (4) 2 (10) 2 (5) 49 (34) 23 (14) 33 (26)
Number (%) prescribed antibiotics 14,800 (97) 803 (98) 253 (98) 20 (100) 29 (100) 127 (98) 153 (90) 120 (93)
Number (%) referred to hospital 2,044 (13) 57 (7) 13 (5) 0 2 (5) 31 (22) 20 (12) 20 (16)
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by the Diseases of the Most Impoverished Program, funded by the Bill and Melinda Gates Foundation, administered by the International Vaccine Institute, Korea.
==== Refs
Bern C Martines J de Zoysa I Glass RI The magnitude of the global problem of diarrhoeal disease: a ten-year update Bull World Health Organ 1992 70 705 714 1486666
Kosek M Bern C Guerrant RL The global burden of diarrhoeal disease, as estimated from studies published between 1992 and 2000 Bull World Health Organ 2003 81 197 204 12764516
Snyder JD Merson MH The magnitude of the global problem of acute diarrhoeal disease: a review of active surveillance data Bull World Health Organ 1982 60 605 613 6982783
Sunoto Diarrhoeal problems in Indonesia past, present and future J Singapore Paediatr Soc 1987 29 Suppl 1 141 152 3657084
Sunoto Diarrhoeal problems in South East Asia, 1986 Paediatr Indones 1987 27 114 130 3503972
Nazir M Pardede N Ismail R The incidence of diarrhoeal diseases and diarrhoeal diseases related mortality in rural swampy low-land area of south Sumatra, Indonesia J Trop Pediatr 1985 31 268 272 3877814
Simanjuntak CH Hardjining S Hasibuan MA Pujarwoto Koiman I Gastrointestinal infections in Southeast Asia (V): ; Japan. 1998
Lesmana M Subekti D Simanjuntak CH Tjaniadi P Campbell JR Oyofo BA Vibrio parahaemolyticus associated with cholera-like diarrhea among patients in North Jakarta, Indonesia Diagn Microbiol Infect Dis 2001 39 71 75 11248518 10.1016/S0732-8893(00)00232-7
Oyofo BA Subekti D Tjaniadi P Machpud N Komalarini S Setiawan B Simanjuntak C Punjabi N Corwin AL Wasfy M Campbell JR Lesmana M Enteropathogens associated with acute diarrhea in community and hospital patients in Jakarta, Indonesia FEMS Immunol Med Microbiol 2002 34 139 146 12381465 10.1016/S0928-8244(02)00351-6
Oyofo BA Lesmana M Subekti D Tjaniadi P Larasati W Putri M Simanjuntak CH Punjabi NH Santoso W Muzahar Sukarma Sriwati Sarumpaet S Abdi M Tjindi R Ma'ani H Sumardiati A Handayani H Campbell JR Alexander WK Beecham HJ Corwin AL Surveillance of bacterial pathogens of diarrhea disease in Indonesia Diagn Microbiol Infect Dis 2002 44 227 234 12493168 10.1016/S0732-8893(02)00454-6
Richie E Punjabi NH Corwin A Lesmana M Rogayah I Lebron C Echeverria P Simanjuntak CH Enterotoxigenic Escherichia coli diarrhea among young children in Jakarta, Indonesia Am J Trop Med Hyg 1997 57 85 90 9242325
Lesmana M Subekti DS Tjaniadi P Simanjuntak CH Punjabi NH Campbell JR Oyofo BA Spectrum of vibrio species associated with acute diarrhea in North Jakarta, Indonesia Diagn Microbiol Infect Dis 2002 43 91 97 12088614 10.1016/S0732-8893(02)00373-5
Simanjuntak CH Larasati W Arjoso S Putri M Lesmana M Oyofo BA Sukri N Nurdin D Kusumaningrum RP Punjabi NH Subekti D Djelantik S Sukarma Sriwati Muzahar Lubis A Siregar H Mas'ud B Abdi M Sumardiati A Wibisana S Hendarwanto Setiawan B Santoso W Putra E Sarumpaet S Ma'ani H Lebron C Soeparmanto SA Campbell JR Corwin AL Cholera in Indonesia in 1993-1999 AmJ TropMed Hyg 2001 65 788 797
Indonesian-Government National Statistics Bureaucratic(BPS) 2000
Richie EE Punjabi NH Sidharta YY Peetosutan KK Sukandar MM Wasserman SS Lesmana MM Wangsasaputra FF Pandam SS Levine MM O'Hanley PP Cryz SJ Simanjuntak CH Efficacy trial of single-dose live oral cholera vaccine CVD 103-HgR in North Jakarta, Indonesia, a cholera-endemic area Vaccine 2000 18 2399 2410 10738097 10.1016/S0264-410X(00)00006-2
Simanjuntak CH O'Hanley P Punjabi NH Noriega F Pazzaglia G Dykstra P Kay B Suharyono Budiarso A Rifai AR Safety, immunogenicity, and transmissibility of single-dose live oral cholera vaccine strain CVD 103-HgR in 24- to 59-month-old Indonesian children J Infect Dis 1993 168 1169 1176 8228350
Simanjuntak CH Punjabi NH Wangsasaputra F Nurdin D Pulungsih SP Rofiq A Santoso H Pujarwoto H Sjahrurachman A Sudarmono P von Seidlein L Acosta C Robertson SE Ali M Lee H Park J Deen JL Agtini MD Clemens JD Diarrhoea episodes and treatment-seeking behaviour in a slum area of North Jakarta, Indonesia J Health Popul Nutr 2004 22 119 129 15473515
Vollaard AM Ali S van Asten HA Widjaja S Visser LG Surjadi C van Dissel JT Risk factors for typhoid and paratyphoid fever in Jakarta, Indonesia Jama 2004 291 2607 2615 15173152 10.1001/jama.291.21.2607
McLaughlin JC Murray PR, Baron EJ, Pfaller MA, Tenover FC and Yolken RH Manual of Clinical Microbiology 1995 Sixth edition Washington DC, ASM Press 465 476 7671628
National-Committee-for-Clinical-Laboratory-Standards Performance standards for antimicrobial susceptibility testing: Tenth informational supplement M100-S10 2000 Wayne, PA, National Committee for Clinical Laboratory Standards
Perera BJ Ganesan S Jayarasa J Ranaweera S The impact of breastfeeding practices on respiratory and diarrhoeal disease in infancy: a study from Sri Lanka J Trop Pediatr 1999 45 115 118 10341510 10.1093/tropej/45.2.115
Escobar GJ Salazar E Chuy M Beliefs regarding the etiology and treatment of infantile diarrhea in Lima, Peru Soc Sci Med 1983 17 1257 1269 6635700 10.1016/0277-9536(83)90018-7
Nakao RM Kennedy KI Savina G Breastfeeding education and infant health in the rural Philippines Ecol Food Nutr 1992 27 115 126 12291168
Kotloff KL Winickoff JP Ivanoff B Clemens JD Swerdlow DL Sansonetti PJ Adak GK Levine MM Global burden of Shigella infections: implications for vaccine development and implementation of control strategies Bull World Health Organ 1999 77 651 666 10516787
World-Health-Organization. Management of the child with a serious infection or severe malnutrition: Guidelines for care at the first referral level in developing countries (WHO/FCH/CAH/00.1) 2000 Geneva, World Health Organization 47
Bhattacharya SK Datta D Bhattacharya MK Garg S Ramamurthy T Manna B Nair GB Nag A Moitra A Cholera in young children in an endemic area Lancet 1992 340 1549 1361637 10.1016/0140-6736(92)92804-O
Sack RB Siddique AK Longini IMJ Nizam A Yunus M Islam MS Morris JGJ Ali A Huq A Nair GB Qadri F Faruque SM Sack DA Colwell RR A 4-year study of the epidemiology of Vibrio cholerae in four rural areas of Bangladesh J Infect Dis 2003 187 96 101 12508151 10.1086/345865
Tuyet DT Thiem VD Von Seidlein L Chowdhury A Park E Canh do G Chien BT Van Tung T Naficy A Rao MR Ali M Lee H Sy TH Nichibuchi M Clemens J Trach DD Clinical, epidemiological, and socioeconomic analysis of an outbreak of Vibrio parahaemolyticus in Khanh Hoa Province, Vietnam J Infect Dis 2002 186 1615 1620 12447738 10.1086/345731
|
16242013
|
PMC1276796
|
CC BY
|
2021-01-04 16:28:16
|
no
|
BMC Infect Dis. 2005 Oct 20; 5:89
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-89
|
oa_comm
|
==== Front
BMC Mol BiolBMC Molecular Biology1471-2199BioMed Central London 1471-2199-6-201623617110.1186/1471-2199-6-20Research ArticlePseudouridine modification in Caenorhabditis elegans spliceosomal snRNAs: unique modifications are found in regions involved in snRNA-snRNA interactions Patton Jeffrey R [email protected] Richard W [email protected] Department of Pathology and Microbiology, University of South Carolina School of Medicine Columbia, SC 29208 USA2 Waksman Institute, Department of Molecular Biology and Biochemistry and Cancer Institute of New Jersey, Rutgers University, Piscataway, NJ 08854 USA2005 19 10 2005 6 20 20 18 7 2005 19 10 2005 Copyright © 2005 Patton and Padgett; licensee BioMed Central Ltd.2005Patton and Padgett; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Pseudouridine (Ψ) is an abundant modified nucleoside in RNA and a number of studies have shown that the presence of Ψ affects RNA structure and function. The positions of Ψ in spliceosomal small nuclear RNAs (snRNAs) have been determined for a number of species but not for the snRNAs from Caenorhabditis elegans (C. elegans), a popular experimental model system of development.
Results
As a prelude to determining the function of or requirement for this modification in snRNAs, we have mapped the positions of Ψ in U1, U2, U4, U5, and U6 snRNAs from worms using a specific primer extension method. As with other species, C. elegans U2 snRNA has the greatest number of Ψ residues, with nine, located in the 5' half of the U2 snRNA. U5 snRNA has three Ψs, in or near the loop of the large stem-loop that dominates the structure of this RNA. U6 and U1 snRNAs each have one Ψ, and two Ψ residues were found in U4 snRNA.
Conclusion
The total number of Ψs found in the snRNAs of C. elegans is significantly higher than the minimal amount found in yeasts but it is lower than that seen in sequenced vertebrate snRNAs. When the actual sites of modification on C. elegans snRNAs are compared with other sequenced snRNAs most of the positions correspond to modifications found in other species. However, two of the positions modified on C. elegans snRNAs are unique, one at position 28 on U2 snRNA and one at position 62 on U4 snRNA. Both of these modifications are in regions of these snRNAs that interact with U6 snRNA either in the spliceosome or in the U4/U6 small nuclear ribonucleoprotein particle (snRNP) and the presence of Ψ may be involved in strengthening the intermolecular association of the snRNAs.
==== Body
Background
Pseudouridine (Ψ) is an abundant modified nucleoside found in RNA, at one time considered the fifth nucleoside of RNA [1]. This modification has been found in many types of RNA and is particularly abundant in small stable RNAs such as transfer RNA (tRNA) and small nuclear RNA (snRNA) [2,3]. Because of its structure, with the carbon at the 5 position of the uracil ring attached to the sugar rather than the nitrogen at the 1 position, Ψ is potentially more versatile in its hydrogen bonding interactions [4]. The presence of Ψ appears to strengthen stems in RNA secondary structures and to stabilize base stacking in loops [5-7]. The splicing of pre-messenger RNAs (pre-mRNAs) in Xenopus oocytes is dependent on the presence of Ψ in U2 snRNA [8,9]. This requirement for Ψ in U2 snRNA was also seen in mammalian splicing extracts [10,11]. In addition, structural studies indicate that Ψ at a particular position in U2 snRNA is necessary to obtain the proper conformation of the pre-mRNA:U2 snRNA interaction during splicing [12-14]. The presence of Ψ at position 34 in human U2 snRNA enhances the formation of a splicing intermediate found when an RNA oligonucleotide containing the lariat branch point of the pre-mRNA is incubated with in vitro transcribed U2 and U6 snRNA segments [15].
The formation of Ψ is catalyzed by pseudouridine synthases after transcription and a number of site-specific synthases have been characterized from eukaryotes for the modification of tRNAs [16-22]. It has been shown that the pseudouridylation of several of the residues in metazoan snRNAs require small nucleolar RNA cofactors (snoRNAs) [23-25]. In yeast the formation of Ψ in U2 snRNA is accomplished by a combination of site-specific Ψ synthases [22,26,27] and a synthase that is RNA cofactor dependent [28].
The nucleotides that are modified to Ψ in the spliceosomal snRNAs (U1, U2, U4, U5 and U6 snRNA) have been completely mapped for two species of yeast [26,29] and rat [30]. In addition, the modifications in several of the spliceosomal snRNAs have been sequenced for other species such as fruit fly, frog, humans, and plants (for review see [3]). The positions that are pseudouridylated are, for the most part, conserved and nearly all of the Ψs appear in the 5' half of the snRNAs [3,29,30]. The systematic mapping of the sites of pseudouridylation on these essential splicing cofactors has not been attempted for one of the most popular model systems currently used in biology, Caenorhabditis elegans (C. elegans). This is especially surprising given the fact that cis-splicing in C. elegans exhibits a number of differences from other species. For example, there is a highly conserved 3' acceptor site and a lack of an identified lariat branch point consensus sequence within introns [31,32].
In this report we present the results of the mapping of the sites of Ψ modification on the spliceosomal snRNAs of C. elegans, the first step in a larger endeavor to determine the factors required for the formation of Ψ in worms and the function(s) of these modifications in an animal model of development. All of the C. elegans spliceosomal snRNAs contain Ψ, which distinguishes these results from those seen with the snRNAs from two yeast species [26,29]. C. elegans U2 snRNA is the most highly modified with a total of nine Ψs, all in the first 56 nucleotides of the snRNA. U5 snRNA has three Ψ residues in the highly conserved stem-loop known to be essential for pre-mRNA splicing. C. elegans U4 snRNA has two Ψs, with one in a region of the snRNA that participates in an RNA-RNA interaction with U6 snRNA and allows for the formation of the U4/U6 small nuclear ribonucleoprotein particle (snRNP). U1 and U6 snRNAs each have one Ψ identified. When the sites of modification are compared with snRNAs sequenced from other species, two of the positions modified on C. elegans snRNAs are unique, one at position 28 on U2 snRNA and one at position 62 on U4 snRNA. These modifications are in regions of each snRNA that are predicted to interact with U6 snRNA to form the U4/U6 snRNP or in the spliceosome during the removal of introns from pre-mRNA [33].
Results
A primer extension method [34,35] was used to identify the positions of Ψ in the snRNAs. The method involves chemically treating the total RNA from wild type worms with 1-cyclohexyl-3(2-morpholinoethyl) carbodi-imide metho-p-toluenesulphonate (CMCT) which covalently reacts with uridines, and other nucleotides, whether modified or not. Two concentrations of CMCT were used in the reactions and the reaction was allowed to proceed such that only a portion of the available uridines is covalently modified. This partial reaction is important since it is essential to be able to extend the primer past the first Ψ encountered in order to identify additional Ψs. The isolated total RNA is then treated with mild base, and as a consequence only the Ψ residues are left covalently modified. This covalent modification of Ψ creates a stop to reverse transcriptase (RT) progression during the specific primer extension phase of the assay. The products are electrophoresed on a 10% polyacrylamide/8.3 M urea gel along with sequencing lanes using the same end-labeled primer. When bands appear in the CMCT+ lanes but not in the CMCT- lanes and those bands are just prior to a uridine in the sequence, then there is a Ψ at that position.
Nearly all of the Ψ residues that have been mapped on any snRNA of any species are found in the 5' half of the RNA [3,30], the exceptions being one residue found near the 3' end of U6 snRNA (rat, mouse, bean) and two residues that are found at positions 131 and 132 on Drosophila U4 snRNA. Therefore this primer extension technique [34,35] can be used to determine the modification state of all but the very 3' 18 nucleotides of the snRNA. To map the positions of Ψ on C. elegans snRNAs, primers that hybridized to the extreme 3' end of each snRNA were used initially. Three additional primers were used to map sites of modification on U2 snRNA, and two additional primers were used to map U4, U5, and U6 snRNAs. Only one additional primer was used to map modifications on U1 snRNA.
U1 snRNA
With U1 snRNA, only one nucleotide, at the position shown in Figure 1A, was found to be converted into Ψ. This modification is at position 5 of the Thomas et al. [36] C. elegans numbering system for U1 snRNA. With the primer extension assay used in these experiments [34], the reverse transcriptase (RT) stops just before the nucleotide that is a Ψ, which is covalently modified by the CMCT. In the case of C. elegans U1 snRNA, there is a band before the U at position 5 and not position 6 (Fig. 1A) and therefore position 5 is modified and position 6 is not. The reason for the presence of the double band in the CMCT treated lanes is not known but it has been seen in other substrates by other investigators [37]. Other assays of U1 snRNA using this same primer with a separately isolated total RNA sample and electrophoresed on a different gel also showed this double band. We have noted a double band similar to this previously when Ψ was mapped on C. elegans tRNAAla (AGC) at a position where there was only one uridine [21]. Two primers were used in the primer extension reactions, with one hybridizing to the extreme 3' end of the U1 snRNA (U1RevA) and one closer to the 5' end (U1RevB). The modification shown in Fig. 1A was mapped with this latter primer and U1RevA yielded no additional Ψ residues (data not shown).
U4 snRNA
U4 snRNAs from other species that have been sequenced have from zero to five Ψs [3,29]. Two Ψs were identified on C. elegans U4 snRNA mapped with the U4RevB primer (Fig. 1B). They occur at positions 62 and 72, with stops to RT appearing in the CMCT treated lanes of the gel. Pseudouridines have been found at positions 4, 72, and 79 in all vertebrate U4 snRNAs sequenced, but no other U4 snRNA sequenced has a Ψ at position 62 [3]. The possible modifications of the uridine at positions 79 and 80 of C. elegans U4 snRNA, slightly obscured in a region of stops to RT progression even in the CMCT- lane, are marked on the figure. However, the use of additional primers in the assay has not confirmed Ψs at these positions (data not shown). With this specific primer extension method [34] there is a possibility of misidentifying Ψs [38], so since this is a region of considerable secondary structure, the conservative interpretation of the data was that there are no Ψs at these sites. No modification of uridines near the 5' end of C. elegans U4 snRNA, such as at position 4 (Fig. 1B), was observed with this or another primer (U4RevC), that is closer to the 5' end of the U4 snRNA (data not shown).
U2 snRNA
As with other U2 snRNAs that have been mapped for modifications [3], C. elegans U2 snRNA has the largest number of Ψs in the spliceosomal snRNAs. In a previous report of the characterization of Pseudouridine synthase 1 in C. elegans (CePus1p; [21]), we mapped at least five Ψ residues in C. elegans U2 snRNA at positions 39, 45, 46, 48, and 56 (C. elegans U2 snRNA numbering) using a primer (U2RevC) close to the middle of the snRNA [21]. When another primer (U2RevB) was used instead, two additional modifications, at positions 41 and 43 were seen (Fig. 2A), but the modification at position 48 was not detected. When the assay was repeated using the U2RevC primer used in the earlier report, the modification at position 48 was not detected (data not shown). In Fig. 2, and the other figures, the arrowheads on the left of the panels denote positions just before a U residue where bands occur in both CMCT+ lanes but are missing or greatly diminished in the CMCT- lane. Additional modifications were mapped to positions 10, 16, and 28 (Fig. 2B) of U2 snRNA when a primer closer to the 5' end of the RNA (U2RevD) was used. Three new primers were used to map the Ψ residues on U2 snRNAs in this report and the primer hybridizing to the extreme 3' end of the RNA (U2RevA) did not add any additional pseudouridylated positions to the ones identified with the U2RevB-D primers. This brings the total number of Ψ residues in C. elegans U2 snRNA to nine. Surprisingly, no Ψ was detected in C. elegans U2 snRNA at the position equivalent to 34 in vertebrates (Fig. 2A, stick arrow on right side of panel; position 36 in C. elegans U2 snRNA), even though a Ψ is found at this position in all other U2 snRNAs where the modifications have been determined [3]. This same result was seen with two different primers, there is no increase in band intensity in the CMCT treated lanes, which would indicate a stop to RT and therefore a Ψ (data not shown, see Discussion).
U5 snRNA
In many of the U5 snRNAs that have been sequenced there are several Ψ residues in the region of the terminal loop of the large stem-loop (see Fig. 4) [3]. Even yeast (S. pombe) has two modifications in this region [29], a portion of U5 snRNA that has been shown to be critical for the function of this splicing factor [39,40]. In the U5 snRNAs where Ψs have been found, there is a Ψ at position 43 and most have an additional one at position 46 (see Fig. 4, vertebrate U5 snRNA numbering). C. elegans has three Ψ residues in this loop region and the result with the U5RevB primer is shown in Fig. 3A. There are two Ψs at positions 45 and 48 in C. elegans U5 snRNA numbering, which correspond to positions 43 and 46 in vertebrates. The third Ψ, shown at position 36 in Fig. 3A, is in the stem at a position that it shares with several other species, including the pea, but not with vertebrates or with either S. pombe or S. cerevisiae (Fig. 4).
U6 snRNA
One Ψ was mapped on U6 snRNA from C. elegans at position 26 (Fig. 3B), which is equivalent to position 31 in the vertebrate U6 snRNA numbering system. This position is modified in mammals but not in either yeast U6 snRNA mapped [26,29]. A second primer (U6RevC, data not shown) was used to confirm the lack of modification at position 35, which corresponds to a Ψ in mammalian U6 snRNA (position 40). Based on the sequences of other U6 snRNAs [3] there is a possibility that there might be another Ψ residue close to the 3' end of the C. elegans U6 snRNA (at position 81). Since even the most 3' primer we used in these assays would hybridize to this position, we were not able to determine if worm U6 snRNA has this modification using this technique.
CePus1p knockout
In yeast, Pus1p modifies one position on U2 snRNA [26], but when total RNA from CePus1p knockout worms (VC110, see Experimental section) was used as template for the primer extensions with U1RevB, U5RevC, and U6RevC primers there was no difference in the Ψ modification pattern seen with wild type (N2) and VC110 RNA (data not shown). This result was also seen with U2 snRNA from VC110 worms, and was reported previously [21].
Discussion
In all the snRNAs that have been sequenced and the modifications determined, U2 snRNA has by far the most Ψs. C. elegans U2 snRNA has nine, less than the number seen with vertebrates and flowering plants but greater than the number seen with either S. pombe or S. cerevisiae [3]. The overall modification pattern seen with C. elegans U2 snRNA is unique, and there is one position that is modified to Ψ that is not found in the U2 snRNAs from other species (position 28). The other eight modifications were found at sites that have previously been identified as sites of pseudouridylation in U2 snRNA and two (positions 41 and 43 in rat U2 snRNA) have been found in all U2 snRNAs mapped [3]. These modifications have been implicated in the function of the pre-mRNA splicing cofactor [8-11]. It was shown by an in vitro reconstitution system for pre-mRNA splicing that the modified U2 snRNA reconstituted splicing but U2 snRNA without Ψs did not [10,11]. In addition, modifications at the 5' end of U2 snRNA were shown to be necessary for the cofactor's activity in the reconstitution of pre-mRNA splicing in Xenopus oocytes [8]. Using this same system, Zhao and Yu [9] have recently shown that the presence of Ψ in the region of U2 snRNA (nts. 34–46) that interacts with the pre-mRNA lariat branch point is required for function of the cofactor and for the formation of the active 17S U2 small nuclear ribonucleoprotein particle (snRNP).
The 'branch point interacting region' of U2 snRNA is highly pseudouridylated, with six modifications in vertebrates and V. faba, five in S. pombe and 3 in S. cerevisiae [3]. All of these species have U2 snRNAs that have Ψ at positions 34, 41, and 43. C. elegans also has five Ψs in this region, and although there are Ψs at positions equivalent to 41 and 43, we were unable to detect a modification at the position equivalent to 34 (C. elegans position 36). The sequence of this region is highly conserved [30], and so even though there is a uridine at position 36 in the U2 snRNA from C. elegans, it does not appear to be converted to Ψ. This is a surprising result given the importance attached to the presence of Ψ at this position. It is the nucleotide that is predicted to be directly across from the bulged adenosine in the helix formed by U2 snRNA and pre-mRNA in mammalian and yeast splicing mechanisms [33]. Newby and Greenbaum have shown that the presence of Ψ at this position stabilizes the bulged adenosine conformation in oligomers, which is predicted to be favorable for splicing [12-14]. In addition, Valadkhan and Manley have shown that in vitro, a Ψ at position 34 in U2 snRNA enhances the formation of a putative splicing intermediate between U2 and U6 snRNAs [15]. It has been shown recently that when the loss of Pseudouridine synthase 7 (Pus7p), responsible for Ψ35 in S. cerevisiae U2 snRNA (the Ψ34 equivalent), is coupled with the mutation of position 40 in U2 snRNA (a U to G substitution or the deletion of the U), a synthetic growth defect is seen. Neither the loss of Pus7p nor either mutation at position 40, showed a phenotype alone. This suggests there is a requirement for the Ψ at position 35 under certain conditions [41].
How is it that worms can do without Ψ at this position? It is possible our assays were incapable of detecting Ψ at this particular position in C. elegans U2 snRNA with the primers that we have used. But we think this is unlikely since we identify Ψs on either side of position 36. It is a formal possibility that CMCT does not react with this particular Ψ in worm U2 snRNA and therefore would not result in a stop to RT, but we believe this is also unlikely. Instead an answer might be found in the differences in consensus sequences seen with C. elegans introns. There is no identified lariat branch point consensus sequence in C. elegans introns and the consensus sequence at the 3' acceptor site is extended and highly conserved [31,32]. Since there is no branch point consensus in worms, there may be no requirement for a Ψ at position 36 in C. elegans U2 snRNA. In C. elegans U2 snRNP may be more closely associated with the 3' splice site since it has been found that the U2 associated factor (U2AF) is associated with this region of the intron rather than the lariat branch point [32].
Although C. elegans U2 snRNA does not appear to have a Ψ at position 36 it does have a Ψ at position 28, a modification that occurs in no other sequenced U2 snRNA [3]. This position is significant since it is in a region of C. elegans U2 snRNA that is predicted to interact with U6 snRNA in the active spliceosome (see Fig. 5A; [33,42]). The addition of an A-Ψ base pair would add stability to the inter-RNA stem and may favor the adoption of a particular conformation in the U6 or U2 snRNA that would promote catalytic function [5,6,33,43]. It is possible that this modification is critical due to the lack of an extensive lariat branch point interaction between U2 snRNA and the pre-mRNA in C. elegans [32].
Of the spliceosomal snRNAs, the modifications on U5 snRNA have been determined from the largest number of species [3]. One position (43, in human U5 snRNA numbering) is converted to Ψ in all of the U5 snRNAs where Ψ has been found (Fig. 4). In addition, many U5 snRNAs, C. elegans U5 snRNA included, also have a Ψ at position 46 (Fig. 4). The nucleotide sequence in the loop is highly conserved and is critical for the function of U5 snRNA in the splicing of pre-mRNA [30,39,40]. However, the position of Ψ in the stem is not highly conserved, except for a tendency to be located near the ends of the stem (Fig. 4). In tRNAs, Ψs located at the ends of stems are thought to stabilize base-pairing and therefore strengthen the stem [4]. This could be the reason the actual position is not conserved, only the presence of the Ψ in the stem. Interestingly, S. cerevisiae U5 snRNA does not have a Ψ in the stem, only at position 43 in the loop. This yeast has a U5 snRNA that can be folded into a different structure from the other U5 snRNAs [3]. It has an additional stem loop that could have a stabilizing effect on the stem-loop shown in Fig. 4, thereby abrogating the need for the presence of Ψ in the stem in U5 snRNA of S. cerevisiae.
C. elegans U1 snRNA has two uridines at positions 5 and 6, in this region of U1 snRNA that is essential for the function of the splicing factor, but only the uridine at position 5 is converted to Ψ. Many species, including vertebrates, Drosophila, and S. cerevisiae, also have two uridines at these positions but both are converted to Ψ. However, algae and S. pombe U1 snRNAs have two uridines, and Ψ appears only at the position nearest the 5' end of the RNA [3].
S. pombe, and S. cerevisiae U4 snRNAs have no detectable Ψs [26,29], whereas vertebrates, Drosophila, and V. faba (bean), all have several Ψs in their U4 snRNAs [3]. One of the three Ψs found in rat U4 snRNA (at position 4) is in a portion of U4 snRNA that is predicted to interact with U6 snRNA, suggesting the presence of Ψ in this region of U4 snRNA could strengthen the interaction. An equivalent Ψ is not present in C. elegans U4 snRNA but the Ψ at position 62, a Ψ that is found in no other sequenced U4 snRNA, is in an area that is predicted to participate in a second intermolecular hybridization with C. elegans U6 snRNA (see Fig. 5B). This A-Ψ base pair is in the middle of three G-U base pairs and so would lend greater stability to the stem formed by U4 and U6 snRNAs. This same snRNA-snRNA interaction is predicted to have only one G-U base pair with mammalian U4 and U6 snRNAs [44]. The A residue in C. elegans U6 snRNA that the Ψ at position 62 of U4 snRNA pairs with is the same residue that is predicted to base pair with the Ψ 28 of U2 snRNA (Fig. 5A), providing evidence of the need to strengthen each of the snRNA-snRNA interactions that involve this region of U6 snRNA.
Conclusion
The total number of Ψs found in the snRNAs of C. elegans is significantly higher than the minimal amount found in yeasts but it is lower than that seen in sequenced vertebrate snRNAs. When the actual sites of modification on C. elegans snRNAs are compared with other sequenced snRNAs most of the positions correspond to modifications found in other species. However, two of the positions modified on C. elegans snRNAs are unique, one at position 28 on U2 snRNA and one at position 62 on U4 snRNA. Both of these modifications are in regions of these snRNAs that interact with U6 snRNA either in the spliceosome or in the U4/U6 small nuclear ribonucleoprotein particle (snRNP). These Ψs and may be involved in strengthening the intermolecular association of the snRNAs.
The determination of the Ψ residues in the spliceosomal snRNAs of C. elegans will speed the identification of enzymes and cofactors that are involved in the modification of these uridine residues. The location of the modifications will also help in the identification of those cofactors by providing sequences that can be used to predict guide sequences in the snoRNAs. Once we can identify cofactors we will be able to use the powerful genetics of C. elegans to mutate the cofactors and affect the formation of Ψ at any position in the snRNAs, especially those that might be critical for the function of the splicing cofactors.
Methods
All enzymes and the fmole® Sequencing Kit were purchased from Promega Corporation (Madison, Wisconsin) except for Taq DNA polymerase, which was purchased from Fisher Scientific. The knockout of the C. elegans gene that codes for the homologue of Pseudouridine synthase 1 (Pus1p; strain VC110, W06H3.2 (gk38)) was provided by the C. elegans Reverse Genetics Core Facility at the University of British Columbia, which is funded by the Canadian Institute for Health Research, Genome Canada, and Genome BC. This strain was characterized in an earlier publication [21].
The snRNA genes from C. elegans were amplified using the polymerase chain reaction (PCR; [45,46]) from genomic DNA using the following primers: for U1 snRNA, 5'AAACTTACCTGGCTGGGGG3' (U1For) and 5'TTCAGGGCCGCGCGCACGC3' (U1RevA); for U2 snRNA, 5'ATCGCTTCTTCGGCTTATTAGC3' (U2For) and 5'ACTGGGCCGAGCCCGGCAGC3' (U2RevA); for U4 snRNA, 5'AGCTTTGCGCTGGGGCGATAACG3' (U4For) and 5'TTCTGCCTCCTAGAGGCGTTC3' (U4RevA); for U5 snRNA, 5'AACTCTGGTTCCTCTGCATT3' (U5For) and 5'TGACCCGCTTTCTCAAGGAG3' (U5RevA); and for U6 snRNA, 5'GTTCTTCCGAGAACATATAC3'(U6For) and 5'AAAAATTTGGAACGCTTCACG3' (U6RevA). The fragments were gel purified, inserted into pGEMT, and sequenced. The sequences of the inserts matched that of the published snRNA genes from C. elegans [36].
Isolation of total RNA from worms and assay for Ψ
Total RNA from N2 worms grown at 20°C on Nematode Growth Media plates was isolated as described [47]. Total RNA (10 μg) was treated with 0, 0.041, or 0.167 M 1-cyclohexyl-3(2-morpholinoethyl) carbodi-imide metho-p-toluenesulphonate (CMCT, Aldrich) for 15 minutes at 37°C, precipitated, and treated with sodium bicarbonate as described [34,35]. These treated RNAs (1 μg) were used in primer extension reactions using AMV reverse transcriptase and 32P-end-labeled primers as described [35] and the products separated on denaturing (8.3 M urea) 10% polyacrylamide gels (19:1; acrylamide:bis-acrylamide). Sequencing reactions using the same labeled primers were electrophoresed along with the primer extension reactions [35,48]. The reverse primers listed above for the cloning protocol were used first and the additional primers were: for U1 snRNA, 5'GGCCTAACCATGGGGATTCC3' (U1RevB); for U2 snRNA, 5' CATTCGAGTGTATACCGTAG 3' (U2RevB), 5'CCGAGTCTTCCCTAGGTTCC 3' (U2RevC) and 5'CCTTTATTACACTCATTCG3' (U2RevD); for U4 snRNA, 5'GCGTATGCCACCCATGTTTC3' (U4RevB) and 5'TAAACGCACCTCGGCAAAGC3' (U4RevC); for U5 snRNA, 5'GTAATACTCAGTAAATGCTCC3'(U5RevB) and 5'CTCAGTAAATGCTCCTTTGC3' (U5RevC); for U6 snRNA, 5'GCGTGTCATCCTTGCGCAGG3' (U6RevB) and 5'GCGCAGGGGCCATGCTAATC3' (U6RevC). The dried gels were exposed to Kodak XAR film for various lengths of time.
Authors' contributions
JRP participated in the conception of the study, carried out the mapping of the Ψs in the snRNAs, analyzed the data, wrote the initial draft of and edited the manuscript. RWP participated in the analysis of the data and in the conception of the study, edited the manuscript, provided experience with C. elegans, and provided materials for the study.
Acknowledgements
This work was supported by a Research and Productive Scholarship grant from the University of South Carolina (J.R.P.) and by a grant to R.W.P. from the National Institutes of Health (HD13465).
Figures and Tables
Figure 1 Location of Ψ in C. elegans U1 and U4 snRNAs. Total RNA was treated with either 0, 0.041, or 0.167 M CMCT and subsequently treated with mild base (see Materials and Methods). These samples were used as templates in primer extension reactions using 32P-end-labeled primers and electrophoresed on denaturing polyacrylamide gels. A portion of the autoradiograph is shown in panel (A) and is the result using primer U1RevB. The same primer was also used to generate the sequencing lanes, using C. elegans U1 pGEMT plasmid as the template, on the right of the panel. The lanes are labeled to correspond to the RNA sequence. In (B) the U4RevB primer was used to determine the positions of Ψ in U4 snRNA using treated RNA samples and a U4 pGEMT plasmid for the generation of the sequence. In this and the subsequent two figures, the arrow to the right of the sequence indicates the position of the Ψ and the arrowhead to the left of the panel indicates a stop to reverse transcriptase that is increased in the CMCT treated lanes. The reverse transcriptase stops just before the modified nucleotide.
Figure 2 Locations of Ψs in C. elegans U2 snRNA. The methods are the same as described in Fig. 1, but the primers used were specific for U2 snRNA and the template for the sequencing reaction was the C. elegans U2 pGEMT clone. In panel A the primer used was U2RevB and is closer to the 3' end of the U2 snRNA than the primer used in panel B, U2RevD. The stick arrow on the right of Panel A shows the position of U36, which is discussed in the text.
Figure 3 Locations of Ψs in C. elegans U5 and U6 snRNAs. The methods are the same as described in Fig. 1, but the primers used were specific for U5 or U6 snRNAs and the template for the sequencing reaction was the C. elegans U5 pGEMT or U6 pGEMT clones. Panel A shows the results with the U5RevB primer. Panel B shows the results of using a primer (U6RevB) that was specific for U6 snRNA and the template for the sequencing reaction was the C. elegans U6 pGEMT clone.
Figure 4 Phylogenetic comparison of the locations of Ψs in sequenced U5 snRNAs. Only the terminal portion of the large stem-loop is shown in the figure. The numbering system and primary sequence corresponds to human U5 snRNA and the positions of the modifications for the other species shown are for the homologous position in the human sequence [3]. At positions 29, 30, 32, 51, 52, and 53 the other nucleotides found at those positions in any of the species are indicated in a smaller font, the rest of the sequence is conserved.
Figure 5 The locations of Ψs in the predicted intermolecular interactions between C. elegans U6 snRNA and U2 and U4 snRNAs. Watson-Crick base pairs are denoted with a dash and G:U base pairs denoted with a dot. In A the U2-U6 Helix I of the spliceosome [33] is shown. In B the interaction between the same portion of U6 snRNA and U4 snRNA in the U4/U6 snRNP is shown.
==== Refs
Johnson L Soll D In vitro biosynthesis of pseudouridine at the polynucleotide level by an enzyme extract from Escherichia coli Proc Natl Acad Sci U S A 1970 67 943 950 4943184
Auffinger P Westhof E Grosjean H and Benne R Appendix 5: Location and distribution of Modified nucleotides in tRNA Modification and Editing of RNA 1998 Washinton, D.C., ASM Press 569 576
Massenet SA Mougin A Branlant C Grosjean H and Benne R Posttranscriptional modifications of the U small nuclear RNAs Modification and Editing of RNA 1998 Washington, D.C., ASM Press 201 227
Griffey RH Davis D Yamaizumi Z Nishimura S Bax A Hawkins B Poulter CD 15N-labeled Escherichia coli tRNAfMet, tRNAGlu, tRNATyr, and tRNAPhe. Double resonance and two-dimensional NMR of N1-labeled pseudouridine J Biol Chem 1985 260 9734 9741 3894360
Davis DR Stabilization of RNA stacking by pseudouridine Nucleic Acids Res 1995 23 5020 5026 8559660
Davis DR Poulter CD 1H-15N NMR studies of Escherichia coli tRNA(Phe) from hisT mutants: a structural role for pseudouridine Biochemistry 1991 30 4223 4231 2021615 10.1021/bi00231a017
Durant PC Davis DR Stabilization of the anticodon stem-loop of tRNALys,3 by an A+-C base- pair and by pseudouridine J Mol Biol 1999 285 115 131 9878393 10.1006/jmbi.1998.2297
Yu YT Shu MD Steitz JA Modifications of U2 snRNA are required for snRNP assembly and pre-mRNA splicing Embo J 1998 17 5783 5795 9755178 10.1093/emboj/17.19.5783
Zhao X Yu YT Pseudouridine in and near the branch site recognition region of U2 snRNA are required for snRNP biogenesis and pre-mRNA splicing in Xenopus oocytes RNA 2004 10 681 690 15037777 10.1261/rna.5159504
Segault V Will CL Sproat BS Luhrmann R In vitro reconstitution of mammalian U2 and U5 snRNPs active in splicing: Sm proteins are functionally interchangeable and are essential for the formation of functional U2 and U5 snRNPs Embo J 1995 14 4010 4021 7664740
Donmez G Hartmuth K Luhrmann R Modified nucleotides at the 5' end of human U2 snRNA are required for spliceosomal E-complex formation RNA 2004 10 1925 1933 15525712 10.1261/rna.7186504
Newby MI Greenbaum NL A conserved pseudouridine modification in eukaryotic U2 snRNA induces a change in branch-site architecture RNA 2001 7 833 845 11424937 10.1017/S1355838201002308
Newby MI Greenbaum NL Sculpting of the spliceosomal branch site recognition motif by a conserved pseudouridine Nat Struct Biol 2002 9 958 965 12426583 10.1038/nsb873
Xu D Greenbaum NL Fenley MO Recognition of the spliceosomal branch site RNA helix on the basis of surface and electrostatic features Nucleic Acids Res 2005 33 1154 1161 15728744 10.1093/nar/gki249
Valadkhan S Manley JL Characterization of the catalytic activity of U2 and U6 snRNAs RNA 2003 9 892 904 12810922 10.1261/rna.5440303
Motorin Y Keith G Simon C Foiret D Simos G Hurt E Grosjean H The yeast tRNA:pseudouridine synthase Pus1p displays a multisite substrate specificity RNA 1998 4 856 869 9671058 10.1017/S1355838298980396
Chen J Patton JR Cloning and characterization of a mammalian pseudouridine synthase RNA 1999 5 409 419 10094309 10.1017/S1355838299981591
Becker HF Motorin Y Planta RJ Grosjean H The yeast gene YNL292w encodes a pseudouridine synthase (Pus4) catalyzing the formation of psi55 in both mitochondrial and cytoplasmic tRNAs Nucleic Acids Res 1997 25 4493 4499 9358157 10.1093/nar/25.22.4493
Lecointe F Simos G Sauer A Hurt EC Motorin Y Grosjean H Characterization of yeast protein Deg1 as pseudouridine synthase (Pus3) catalyzing the formation of psi 38 and psi 39 in tRNA anticodon loop J Biol Chem 1998 273 1316 1323 9430663 10.1074/jbc.273.3.1316
Chen J Patton JR Pseudouridine Synthase 3 from Mouse Modifies the Anticodon Loop of tRNA Biochemistry 2000 39 12723 12730 11027153 10.1021/bi001109m
Patton JR Padgett RW Caenorhabditis elegans pseudouridine synthase 1 activity in vivo: tRNA is a substrate but not U2 small nuclear RNA Biochem J 2003 372 595 602 12597772 10.1042/BJ20021938
Behm-Ansmant I Urban A Ma X Yu YT Motorin Y Branlant C The Saccharomyces cerevisiae U2 snRNA:pseudouridine-synthase Pus7p is a novel multisite-multisubstrate RNA:Psi-synthase also acting on tRNAs RNA 2003 9 1371 1382 14561887 10.1261/rna.5520403
Ganot P Jady BE Bortolin ML Darzacq X Kiss T Nucleolar Factors Direct the 2'-O-Ribose Methylation and Pseudouridylation of U6 Spliceosomal RNA Mol Cell Biol 1999 19 6906 6917 10490628
Jady BE Kiss T A small nucleolar guide RNA functions both in 2'-O-ribose methylation and pseudouridylation of the U5 spliceosomal RNA Embo J 2001 20 541 551 11157760 10.1093/emboj/20.3.541
Zhao X Li ZH Terns RM Terns MP Yu YT An H/ACA guide RNA directs U2 pseudouridylation at two different sites in the branch point recognition region in Xenopus oocytes. RNA 2002 8 1515 1525 12515384
Massenet S Motorin Y Lafontaine DL Hurt EC Grosjean H Branlant C Pseudouridine mapping in the Saccharomyces cerevisiae spliceosomal U small nuclear RNAs (snRNAs) reveals that pseudouridine synthase pus1p exhibits a dual substrate specificity for U2 snRNA and tRNA Mol Cell Biol 1999 19 2142 2154 10022901
Ma X Zhao X Yu YT Pseudouridylation (Psi) of U2 snRNA in S. cerevisiae is catalyzed by an RNA-independent mechanism EMBO J 2003 22 1889 1897 12682021 10.1093/emboj/cdg191
Ma X Yang C Alexandrov A Grayhack EJ Behm-Ansmant I Yu YT Pseudouridylation of yeast U2 snRNA is catalyzed by either an RNA-guided or RNA-independent mechanism Embo J 2005 24 2403 2413 15962000 10.1038/sj.emboj.7600718
Gu J Patton JR Shimba S Reddy R Localization of modified nucleotides in Schizosaccharomyces pombe spliceosomal small nuclear RNAs: modified nucleotides are clustered in functionally important regions RNA 1996 2 909 918 8809017
Reddy R Busch H Birnstiel M Small nuclear RNAs: RNA sequences, structure, and modifications. Structure and function of major and minor snRNPs 1988 Berlin, Springer Verlag 1 37
Blumenthal T Steward K Riddle DL, Blumenthal T, Meyer BJ and Priess JR RNA Processing and Gene Structure C elegans II 1997 Cold Spring Harbor, NY, Cold Spring Harbor Laboratory Press 117 145
Hollins C Zorio DA MacMorris M Blumenthal T U2AF binding selects for the high conservation of the C. elegans 3' splice site Rna 2005 11 248 253 15661845 10.1261/rna.7221605
Madhani HD Guthrie C Dynamic RNA-RNA interactions in the spliceosome Annu Rev Genet 1994 28 1 26 7534458 10.1146/annurev.ge.28.120194.000245
Bakin AV Ofengand J Mapping of pseudouridine residues in RNA to nucleotide resolution Methods Mol Biol 1998 77 297 309 9770678
Baumstark T Ahlquist P The brome mosaic virus RNA3 intergenic replication enhancer folds to mimic a tRNA TpsiC-stem loop and is modified in vivo RNA 2001 7 1652 1670 11720293
Thomas J Lea K Zucker-Aprison E Blumenthal T The spliceosomal snRNAs of Caenorhabditis elegans Nucleic Acids Res 1990 18 2633 2642 2339054
Del Campo M Kaya Y Ofengand J Identification and site of action of the remaining four putative pseudouridine synthases in Escherichia coli RNA 2001 7 1603 1615 11720289
Del Campo M Recinos C Yanez G Pomerantz SC Guymon R Crain PF McCloskey JA Ofengand J Number, position, and significance of the pseudouridines in the large subunit ribosomal RNA of Haloarcula marismortui and Deinococcus radiodurans Rna 2005 11 210 219 15659360 10.1261/rna.7209905
Newman A Norman C Mutations in yeast U5 snRNA alter the specificity of 5' splice-site cleavage Cell 1991 65 115 123 2013092 10.1016/0092-8674(91)90413-S
Newman AJ Norman C U5 snRNA interacts with exon sequences at 5' and 3' splice sites Cell 1992 68 743 754 1739979 10.1016/0092-8674(92)90149-7
Yang C McPheeters DS Yu YT Psi35 in the branch site recognition region of U2 small nuclear RNA is important for pre-mRNA splicing in Saccharomyces cerevisiae J Biol Chem 2005 280 6655 6662 15611063 10.1074/jbc.M413288200
Madhani HD Guthrie C A novel base-pairing interaction between U2 and U6 snRNAs suggests a mechanism for the catalytic activation of the spliceosome Cell 1992 71 803 817 1423631 10.1016/0092-8674(92)90556-R
Davis DR Veltri CA Nielsen L An RNA model system for investigation of pseudouridine stabilization of the codon-anticodon interaction in tRNALys, tRNAHis and tRNATyr J Biomol Struct Dyn 1998 15 1121 1132 9669557
Wersig C Bindereif A Conserved domains of human U4 snRNA required for snRNP and spliceosome assembly Nucleic Acids Res 1990 18 6223 6229 2147057
Mullis K Faloona F Scharf S Saiki R Horn G Erlich H Specific enzymatic amplification of DNA in vitro: the polymerase chain reaction Cold Spring Harb Symp Quant Biol 1986 51 263 273 3472723
Mullis KB Faloona FA Specific synthesis of DNA in vitro via a polymerase-catalyzed chain reaction Methods Enzymol 1987 155 335 350 3431465
Johnstone IL Hope IA Molecular Biology C elegans: A practical approach The Practical Approach Series Hames BD 1999 Oxford, Oxford University Press 201 225
Sanger F Nicklen S Coulson AR DNA sequencing with chain-terminating inhibitors Proc Natl Acad Sci U S A 1977 74 5463 5467 271968
|
16236171
|
PMC1276797
|
CC BY
|
2021-01-04 16:22:25
|
no
|
BMC Mol Biol. 2005 Oct 19; 6:20
|
utf-8
|
BMC Mol Biol
| 2,005 |
10.1186/1471-2199-6-20
|
oa_comm
|
==== Front
BMC MedBMC Medicine1741-7015BioMed Central London 1741-7015-3-161624203410.1186/1741-7015-3-16Research ArticleSelf-prescribing among young Norwegian doctors: a nine-year follow-up study of a nationwide sample Hem Erlend [email protected] Guro [email protected] Reidar [email protected]ønvold Nina T [email protected] Per [email protected] Øivind [email protected] Department of Behavioural Sciences in Medicine, Institute of Basic Medical Sciences, University of Oslo, PO Box 1111 Blindern, NO-0317 Oslo, Norway2005 21 10 2005 3 16 16 22 7 2005 21 10 2005 Copyright © 2005 Hem et al; licensee BioMed Central Ltd.2005Hem et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Self-prescribing among doctors is common, but no longitudinal studies have documented this issue. We studied the self-prescribing behaviour among young Norwegian physicians and the predictors of self-prescribing.
Methods
We conducted a nationwide, prospective and longitudinal study following young Norwegian physicians from internship through the subsequent nine years using three postal questionnaires. Chi-square tests and logistic regression models were applied.
Results
About 54% of the physicians in their fourth and ninth postgraduate years had self-prescribed medication at least once during the previous year. Among those who had used prescription medication during the previous year, about 90% had self-prescribed. Self-prescribing behaviour did not differ significantly between men and women, or according to the type of work at any time. The most frequently self-prescribed medications were antibiotics (71%–81%), contraceptives (24%–25%), analgesics (18%–21%), and hypnotics (9%–12%). Those who had needed treatment for mental problems had self-prescribed hypnotics and sedatives to a greater extent than the others. Being male, having self-prescribed during internship, somatic complaints, mental distress, subjective health complaints, and not having sought help from a general practitioner, were significant adjusted predictors of self-prescribing in the ninth postgraduate year.
Conclusion
The level of self-prescribing among young Norwegian physicians is relatively high, and this behaviour is established early in their professional lives. Although self-prescribing is acceptable in some situations, physicians should seek professional help for illness. Efforts to inculcate more rational help-seeking behaviour should probably start in medical schools.
==== Body
Background
Doctors commonly self-prescribe medication, and estimates of the prevalence of self-prescribing vary widely from 39% to 99% [1-6]. Self-prescribing may be a matter of concern for several reasons. First, there is a lack of objectivity and professional distance, which normally exist in a physician-patient relation, and self-treatment can lead to delayed diagnosis and treatment [7] and worsening of the illness [8]. Second, many diseases need follow-up apart from medication, particularly for mental illness and chronic diseases. Third, self-prescribing may be an indicator that the doctor is neglecting his or her own health. Ethical rules for doctors in several countries explicitly state the importance of seeking help when ill [9,10]. Finally, the potential for addiction is the main concern when self-prescribing is discussed [11]. This topic is important because the health habits and attitudes of physicians influence the counselling and care they provide to patients [12].
The obstacles to seeking professional care are many and complex, and include both rational and irrational factors [13]. Many physicians find it difficult to enter the patient role (role reversal) for various individual and organizational reasons [8,11], such as time pressure, the stigmatizing nature of sickness, worries about bothering or letting down colleagues, fear of showing weakness or lack of medical knowledge, concerns about confidentiality, and fear of restrictions of medical licensing [14-16].
Several papers have been published on self-treatment among doctors in various countries [1-8,17-26]. Some findings are consistent across countries, such as a high level of self-treatment and a reluctance to seek professional care. However, the literature is ambiguous on several aspects of doctors' self-care.
Older doctors have a high prevalence of self-treatment [19,25], but young physicians also frequently self-treat [3,8,20]. Some studies have shown no gender differences in self-treatment [3,8,25], whereas others have found that female physicians use health services more than their male colleagues [23,24]. The influence of a medical specialty is also inconsistent: some studies have shown that general practitioners are more likely to self-treat than hospital physicians [20,25], others have found only small differences [7,19], and still others have shown that clinicians working outside a hospital are significantly less likely to undertake self-treatment than hospital physicians [8]. A substantial part of self-prescribing involves treating minor illnesses and acute infections. However, in Finland, one of the most common reasons for physician self-medication is mental disorder or insomnia [24], and more than 25% of American physicians have self-prescribed a psychoactive drug [4,27].
An important reason for the differences in results among studies is that most have been limited to small, selected or non-representative samples, such as general practitioners or graduates from one medical school. To our knowledge, all previous studies have been cross-sectional. We conducted a prospective longitudinal study of a nationwide sample of medical students from all four universities in Norway and assessed their self-prescribing behaviour in a nine-year follow-up study.
Self-prescribing is common among residents [4,28]. The aim of this study was to investigate the development of such behaviour during internship and the first nine postgraduate years, and to identify the predictors for self-prescribing. We have previously shown that a substantial proportion of doctors report mental health problems requiring treatment [13]. We were interested in whether these physicians are more likely to self-prescribe psychotropic medication than others, and thereby constitute a risk group for future misuse of drugs and addiction.
We believe that various symptoms and diseases might be important to self-prescribing behaviour, and included three symptom and disease measures in the multivariate analysis, including subjective health complaints with a particular focus on musculoskeletal problems, eight somatic disease categories, and mental distress tapping anxiety and depressive symptoms. Finally, we presumed that those who had sought help from a general practitioner would self-prescribe to a lesser extent than others. We asked the following specific research questions:
• What is the prevalence of self-prescribing among young doctors, and does it differ according to gender, type of medication, and type of work? Are there any changes in these patterns during the first postgraduate years?
• Is there an association between help-seeking behaviour for mental disorders and self-prescribing of medication in the postgraduate years?
• Can age, gender, self-prescribing during internship, help-seeking, somatic complaints, subjective health complaints, and mental distress predict self-medication?
Methods
Participants
A cohort, consisting of all medical students graduating from all four Norwegian medical schools in 1993–94 (N = 631), was sent a postal questionnaire three times: at the end of the internship year (one year after graduation) (T1), at the end of the fourth postgraduate year (T2), and in the 10 th year after graduation (T3).
At T1, 402 responded (64% of the original cohort); their mean age was 29 years (SD 2.8) and 56% were women. At T2, 422 responded (67%); their mean age was 31.4 years (2.4) and 56% were women. At T3, 390 responded (62%); their mean age was 37 years (2.7) and 58% were women. Across the entire study period, 252 (40%), 55% of whom were women, responded at all observational times.
Use of prescription medication
At T1, the residents were asked "Have you ever treated yourself with a prescription medication?" with the following response alternatives: "No, I have not had the need"; "No, I received it from another physician"; "Yes, on one or two occasions"; "Yes, sometimes"; and "Yes, often". Because medical students in Norway receive a temporary medical licence in their 10 th semester, the word "ever" indicates a lifetime prevalence of self-prescribing of 2.5 years. At T2 and T3, the prevalence of self-prescribing over the previous year was explored using the same response alternatives.
Prevalence of self-prescribing
At all three times, the participants were asked: "Do you prescribe at present, or have you previously prescribed, some of the following medications for yourself?" with the response alternatives: "Never"; "Yes, previously"; and "Yes, at present". The prevalence of self-medication for the different medications was determined by combining the two final responses. The following medications were asked about: antibiotics, contraceptive pills, migraine medications, prescription analgesics, antihypertensive medications, hypnotics, sedatives, and other psychotropic medications.
Type of work
At T2, the physicians were asked to report in which specialization they were working, according to a list of 53 medical specializations in Norway. At T3, 10 response alternatives of work situation were given, such as hospital physician with or without leader responsibilities, community physician, specialist in private practice, general practitioner, full-time researcher. Based on these categories, the physicians were divided into the following three main groups of work location: hospital physicians, general practitioners (including specialists in private practice), and others (includes all types of preclinical medicine, laboratory medicine, and full-time research).
Self-prescribing and mental health problems
At T2 and T3, the participants were asked: "If you had mental health problems during the last year, did you seek or receive help for them?" [13]. The response alternatives were: "Have had no mental health problems of importance"; "Have not sought help, although I have needed it"; "Yes, have consulted a general practitioner"; "Yes, have consulted a psychologist or psychiatrist"; and "Yes, have been admitted to a hospital psychiatric department". Participants who had needed treatment responded with one of the four latter answers, whereas those who had sought help answered with one of the three latter categories.
Help-seeking
At T3, the participants were asked: "Have you, during the last year, consulted some of the following health care workers with your own problems?" Among the nine response categories was "General practitioner" ("No" or "Yes").
Somatic complaints
Somatic complaints at T3 were measured by one question: "Do you suffer, or have you suffered, during the past year from ..." with the following eight items as answers: "menstrual pain", "impaired vision", "asthma, lung disease or respiratory disease", "allergy or skin problems", "cardiovascular disease or hypertension", "diabetes", "disease in kidneys or urinary tract", and "gynaecological problems or gynaecological disease". The items were scored on a five-point scale ranging from "not at all" (0) to "very much" (4). An index of the eight items was calculated by summing the responses.
Subjective health complaints
The subjective self-evaluation of health was assessed at T3 by a 10-item version of the Subjective Health Complaint (SHC) questionnaire [29,30]. This questionnaire consists of questions examining the occurrence, intensity, and duration of musculoskeletal pain, migraine or headache, and digestive problems for the previous 30 days [31]. The items are scored on a four-point scale ranging from "no complaints" (0) to "serious complaints" (3). Seven of the 10 items are related to musculoskeletal symptoms. The sum of the scores of all items was used to indicate the level of subjective health complaints.
Mental distress
Mental distress was measured at T3 using the SCL-5, a five-item version of the Symptom Check List-25, which captures anxiety and depressive symptoms [32-34]. The SCL-5 includes a question about how much one has been bothered by the following distress items during the last 14 days: "Feeling fearful"; "Nervousness or shaking inside"; "Feeling hopeless about the future"; "Feeling blue"; and "Worrying too much about things". Each item was measured on a five-point scale, from "not at all" (0) to "very much" (4). The sum of the scores of all items was used to indicate the level of mental distress.
Ethics
To ensure confidentiality, the questionnaire was answered so that the identity of respondents remained anonymous to the researchers, and the name and address codes were kept in Statistics Norway. The study was conducted in collaboration with the Norwegian Medical Association and with the approval of the Norwegian Data Inspectorate.
Statistical analyses
Chi-square (5% level of significance) was used to test for differences between categorical variables. Logistic regressions were used to test the associations between the predictor variables and the dichotomous outcome measure. Odds ratios (ORs) are presented with 95% confidence intervals in parentheses.
Results
As shown in Table 1, most of the physicians reported that they had self-prescribed medication. At the end of internship (T1), 69% reported lifetime self-prescribing, which comprised the preceding 2.5 years. The figures were somewhat lower at T2 and T3 (about 54%), because only the previous year's prevalence was asked at this time. At all three times, 10% or fewer had obtained all necessary medication from another physician. Most (74%–81%) physicians who had not self-prescribed (non-prescribers) had not needed drugs to treat themselves. Of the physicians using prescription medication, 90% (T1), 86% (T2), and 84% (T3) had self-prescribed the drugs.
Table 1 Prevalence of self-reported prescription behaviour among young physicians
Have not used prescription medication Have got all prescription medication from another physician Have self-prescribed on one or two occasions Have self-prescribed sometimes Have self-prescribed often
First postgraduate year (T1) # n = 397 90 (22.7%) 31 (7.8%) 215 (54.2%) 57 (14.4%) 4 (1.0%)
Fourth postgraduate year (T2) ## n = 422 158 (37.4%) 38 (9.0%) 177 (41.9%) 37 (8.8%) 12 (2.8%)
Ninth postgraduate year (T3) ## n = 387 136 (35.1%) 40 (10.3%) 162 (41.9%) 43 (11.1%) 6 (1.6%)
# At T1, lifetime prevalence implies the last 2.5 years, since the students are allowed to prescribe medicaments from the 10 th semester as a part of their temporary medical license
## last year prevalence
The prevalence of self-prescribing did not differ significantly between men and women, or by location of work at any of the three times. At T2 and T3 (previous year), 52% and 54% of the women, and 57% and 56% of the men, reported that they had self-prescribed. Among those who had taken prescription medication during the past year, the prevalence of self-prescribing did not differ between men and women. At T2 and T3, 85% and 83% of the women, and 85% and 88% of the men, had self-prescribed during the previous year.
Table 2 shows that the most frequently self-prescribed medications at the three times were antibiotics, contraceptives, analgesics, and hypnotics. Sedatives were self-prescribed by 3.1% (T3) of the physicians, and other psychotropic medications by 0.8%. Among women, 39% (T1), 43% (T2), and 40% (T3) had self-prescribed contraceptives.
Table 2 Prevalence of self-prescription at the first (T1), the fourth (T2) and the ninth (T3) postgraduate years
T1 lifetime T2 last year T3 last year
n (%) n (%) n (%)
Antibiotics 210 (53.0) 299 (71.0) 312 (80.6)
Contraceptive pills 88 (23.0) 102 (24.8) 90 (23.9)
Prescription analgesics 73 (18.5) 74 (17.7) 81 (20.9)
Hypnotics 26 (6.6) 39 (9.3) 48 (12.4)
Sedatives 4 (1.0) 11 (2.6) 12 (3.1)
Migraine medications 12 (3.1) 11 (2.6) 5 (1.3)
Other psychotropic medications 2 (0.5) 7 (1.7) 3 (0.8)
Antihypertensive medications 1 (0.3) 3 (0.7) 1 (0.3)
The prevalence of self-prescribing did not differ significantly between hospital physicians and general practitioners at T2 or T3. At T2, 54% of the hospital physicians and 56% of the general practitioners reported that they had self-prescribed medication during the previous year. At T3, 56% of both groups had self-prescribed.
Those who had needed treatment for mental problems had self-prescribed hypnotics and sedatives to a greater extent than the others, both at T2 (χ2 = 13.0, P < 0.001) and at T3 (χ2 = 11.8, P < 0.001). However, among those with mental problems needing treatment, those who had sought help did not differ from the others at any time point.
Table 3 shows the adjusted predictors for self-prescribing behaviour. At T3, being male, having self-prescribed during internship (T1), and having contemporary subjective health complaints, somatic complaints, mental distress, and not having consulted a general practitioner during the preceding year were significant predictors of self-prescribing sometimes or often. We performed multiple regression analyses separately for men and women. None of the variables were significant predictors in women, whereas all the variables in the model, except for mental distress and age, were significant predictors in men.
Table 3 Predictors of self-prescribing nine years after graduation (T3) among 252 Norwegian physicians
Crude Odds Ratio (95% CI) Adjusted Odds Ratio (95% CI)
Age 1.07 (0.97–1.18) 1.09 (0.94–1.25)
Gender (1 = male) 1.6 (0.9–3.0) 3.3 (1.3–8.5) *
Self-prescription at T1 3.1 (1.4–6.8) ** 2.7 (1.01–7.4) *
Somatic complaints at T3 1.4 (1.2–1.7) *** 1.5 (1.1–1.9) **
Mental distress at T3 1.2 (1.07–1.24) *** 1.13 (1.01–1.25) *
Subjective health complaints at T3 1.2 (1.1–1.3) *** 1.2 (1.08–1.4) **
Sought general practitioner (last year) at T3 0.9 (0.5–1.8) 0.2 (0.1–0.7) *
* P < 0.05.
** P < 0.01.
*** P < 0.001.
Discussion
There are several important findings from our study. Most of these young physicians had self-prescribed at least once during the previous year, and this behaviour started early in their careers and persisted at a relatively high level throughout the follow-up period. The most frequently self-prescribed medications were antibiotics, contraceptives, analgesics and hypnotics. Those who had mental health problems needing treatment were more likely to self-prescribe hypnotics than those without perceived mental health problems. Being male, having self-prescribed during internship, somatic complaints, mental distress, subjective health complaints, and lack of seeking help from a general practitioner were significant predictors of self-prescribing in the ninth postgraduate year.
The relatively high level of self-prescribing in our study concurs with the findings in other countries [4,28]. However, direct comparisons with other studies are difficult because of different questionnaires and time frames [1-6]. Previous studies have shown that physicians often self-prescribe medications, most practise self-treatment when they are ill, many have problems accepting their own illness, and many tend to avoid taking sick leave during an illness for which they would have sick-listed their patients [35]. However, this illness behaviour may be changing. For example, the sick-leave rate among physicians is rising [36,37]. This may indicate that their health is getting worse or that they have started to take better care of their own health.
We found no differences in self-prescribing behaviour between hospital physicians and general practitioners. One possible reason is that this cohort is young and that differences may be revealed later in their careers. Alternatively, this trend might relate to a similar pattern of self-prescribing behaviour among general practitioners and hospital physicians. We also found no difference in the self-prescribing behaviour between men and women, which agrees with previous data showing only small gender differences among physicians. However, being male was a significant predictor of self-prescribing in the ninth postgraduate year.
Physicians start self-prescribing early in their careers and at a relatively high level, which seems to be stable over the first postgraduate years. These findings are novel and have not been reported previously in longitudinal studies, although they are expected from the trends in previous cross-sectional studies [4,5,38,39]. Interestingly, in a recent study of medical students in London, most agreed that it is appropriate for doctors to self-investigate and self-refer, but fewer approved of doctors self-prescribing [39]. Presenting information about the dangers of self-prescribing in medical school and having medical students reflect on this topic might help reduce the prevalence of self-prescribing behaviour among physicians. Altering the climate of stoicism at an early stage in the medical career [3,40], and incorporating personal and professional development as part of the curriculum, might provide an opportunity to address issues of self-care [41].
Participants needing treatment for mental problems self-prescribed hypnotics and sedatives to a greater extent than the others. Among those with mental problems needing treatment, those who had sought help did not differ from the others at any time. However, the number of individuals was low. Physicians with mental health problems may be a target group for intervention. The most frequently self-prescribed medications in our sample were consistent to some extent with previous findings [4,5,24]. Few respondents reported that they self-prescribed sedatives or other psychotropic medications. However, those who do may be of concern.
One aim of our study was to determine whether subjective health complaints, mental distress, and somatic diseases predicted self-prescribing behaviour; the multivariate model showed that each of these was an independent predictor for self-prescribing. Interestingly, self-prescribing in internship was also an independent and significant predictor for self-prescribing in the ninth postgraduate year. This shows that the habit of self-medication starts early and supports the view that medical school is a relevant time to introduce students to the concepts of ethical thinking and reflection on this topic. Seeking help from a general practitioner was associated with less self-prescribing when we controlled for other predictors. This underpins the notion that more adequate help-seeking behaviour may reduce the level of self-prescribing.
Most papers view self-prescribing among physicians as a problem. Unfortunately, we do not have data to determine whether the self-prescribed medicaments were re-prescriptions of medicaments initiated by another physician, or if the self-prescription was totally self-initiated. It may be appropriate for physicians to self-prescribe under some circumstances [4,42], for example, to renew a prescription for long-term medication initiated by another doctor (e.g. contraceptives). It may also be appropriate for physicians to self-prescribe for minor illnesses, such as antibiotics for trivial infections. Such prescribing may also be practical and time saving in some parts of Norway because of the long distances between physicians. However, we did not identify any differences in self-prescribing behaviour between general practitioners and others.
Strengths and limitations
To our knowledge, this is the first longitudinal study of self-prescribing by physicians that followed the course and development of self-medication over time. One strength is the study's prospective design, because it allowed us to identify risk factors for self-prescribing behaviour. The validity of the data is reinforced by the study design, which used a nationwide, representative sample of an occupational group followed for nine years. Another strength is the inclusion of several previously validated instruments in the comprehensive questionnaire. The main findings may be generalized to other countries, because medical doctors belong to a relatively homogenous cross-national occupational group. A limitation of the study is that the response rates were not high in the longitudinal sample, and the results relied solely on self-report because no central prescription register was available. Some level of report bias may have occurred, because physicians may view self-prescribing as socially undesirable, at least as related to psychotropic medication.
Conclusion
The prevalence of self-prescribing was relatively high in this cohort of young physicians. Most self-prescribed medications are not addictive or prescribed for chronic or mental disorders. However, a substantial proportion of the respondents self-prescribed hypnotics. Moreover, those who self-prescribed early in their career were more prone to self-prescribe later on. These findings suggest that the issue of self-prescribing probably should be addressed in medical school.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EH was involved in designing the study, analysing the data and writing the paper. GS was involved in analysing the data and writing the paper. RT was involved in designing the study, analysing the data and writing the paper. NTG was involved in initiating and designing the study, was responsible for the collection of data, and entered the data into the computer. PV and ØE initiated and designed the study and supervised the collection of data, and were involved in writing the paper. EH is the guarantor.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Research Council of Norway, the Almus Foundation, and the Norwegian Medical Association are gratefully acknowledged for funding this research. We are also very grateful to all the young Norwegian doctors who conscientiously responded to the comprehensive questionnaires.
==== Refs
Wachtel TJ Wilcox VL Moulton AW Tammaro D Stein MD Physicians' utilization of health care J Gen Intern Med 1995 10 261 265 7616335
Waldron HA Sickness in the medical profession Ann Occup Hyg 1996 40 391 396 8806211 10.1016/0003-4878(96)00027-0
Baldwin PJ Dodd M Wrate RM Young doctors' health. II. Health and health behaviour Soc Sci Med 1997 45 41 44 9203269 10.1016/S0277-9536(96)00307-3
Christie JD Rosen IM Bellini LM Inglesby TV Lindsay J Alper A Asch DA Prescription drug use and self-prescription among resident physicians JAMA 1998 280 1253 1255 9786376 10.1001/jama.280.14.1253
Clarke J O'Sullivan Y Maguire N A study of self-care among Irish doctors Ir Med J 1998 91 175 176 9973754
Richards JG The health and health practices of doctors and their families N Z Med J 1999 112 96 99 10210296
Allibone A Oakes D Shannon HS The health and health care of doctors J R Coll Gen Pract 1981 31 728 734 7338866
Rosvold EO Bjertness E Illness behaviour among Norwegian physicians Scand J Public Health 2002 30 125 132 12028861 10.1080/14034940210133744
Norwegian Medical Association Ethical rules for doctors (in Norwegian), established 1961, latest update 2002 accessed September 30, 2005
British Medical Association Ethical responsibilities in treating doctors who are patients Guidance from the Ethics Department March revised March 2004 accessed September 30, 2005
Rosvold EO Tyssen R Should physicians' self-prescribing be restricted by law? Lancet 2005 365 1372 1374 15836873 10.1016/S0140-6736(05)66353-2
Gross CP Mead LA Ford DE Klag MJ Physician, heal Thyself? Regular source of care and use of preventive health services among physicians Arch Intern Med 2000 160 3209 3214 11088080 10.1001/archinte.160.21.3209
Tyssen R Røvik JO Vaglum P Grønvold NT Ekeberg Ø Help-seeking for mental health problems among young physicians: is it the most ill that seeks help? A longitudinal and nationwide study Soc Psychiatry Psychiatr Epidemiol 2004 39 989 993 15583907 10.1007/s00127-004-0831-8
Thompson WT Cupples ME Sibbett CH Skan DI Bradley T Challenge of culture, conscience, and contract to general practitioners' care of their own health: qualitative study BMJ 2001 323 728 731 11576981 10.1136/bmj.323.7315.728
Hendin H Maltsberger JT Haas AP A physician's suicide Am J Psychiatry 2003 160 2094 2097 14638576 10.1176/appi.ajp.160.12.2094
Center C Davis M Detre T Ford DE Hansbrough W Hendin H Laszlo J Litts DA Mann J Mansky PA Michels R Miles SH Proujansky R Reynolds CF 3rdSilverman MM Confronting depression and suicide in physicians: a consensus statement JAMA 2003 289 3161 3166 12813122 10.1001/jama.289.23.3161
Selley P Self-prescribing by doctors Health Trends 1988 20 128 129 10290668
Chambers R Belcher J Self-reported health care over the past 10 years: a survey of general practitioners Br J Gen Pract 1992 42 153 156 1586551
Forsythe M Calnan M Wall B Doctors as patients: postal survey examining consultants and general practitioners adherence to guidelines BMJ 1999 319 605 608 10473473
Pullen D Lonie CE Lyle DM Cam DE Doughty MV Medical care of doctors Med J Aust 1995 162 481, 484 7746206
Davidson SK Schattner PL Doctors' health-seeking behaviour: a questionnaire survey Med J Aust 2003 179 302 305 12964913
Bruguera M Guri J Arteman A Grau Valldosera J Carbonell J Doctors taking care of their own health. Results of a postal survey (in Spanish) Med Clin (Barc) 2001 117 492 494 11707204
Töyry S Räsänen K Kujala S Äärimaa M Juntunen J Kalimo R Luhtala R Mäkelä P Myllymääki K Seuri M Husman K Self-reported health, illness, and self-care among Finnish physicians: a national survey Arch Fam Med 2000 9 1079 1085 11115211 10.1001/archfami.9.10.1079
Töyry S Räsänen K Seuri M Äärimaa M Juntunen J Kujala S Husman K Increased personal medication use among Finnish physicians from 1986 to 1997 Br J Gen Pract 2004 54 44 46 14965406
Rosvold EO Vaglum P Moum T Use of minor tranquilizers among Norwegian physicians: a nation-wide comparative study Soc Sci Med 1998 46 581 590 9460837 10.1016/S0277-9536(97)00203-7
Rennert M Hagoel L Epstein L Shifroni G The care of family physicians and their families: a study of health and help-seeking behaviour Fam Pract 1990 7 96 99 2369989
McAuliffe WE Rohman M Santangelo S Feldman B Magnuson E Sobol A Weissman J Psychoactive drug use among practicing physicians and medical students N Engl J Med 1986 315 805 810 3748091
Campbell S Delva D Physician do not heal thyself. Survey of personal health practices among medical residents Can Fam Physician 2003 49 1121 1127 14526864
Ursin H Endresen IM Ursin G Psychological factors and self-reports of muscle pain Eur J Appl Physiol Occup Physiol 1988 57 282 290 3371336 10.1007/BF00635986
Eriksen HR Ihlebæk C Ursin H A scoring system for subjective health complaints (SHC) Scand J Public Health 1999 27 63 72 10847674 10.1080/14034949950153931
Aasland OG Olff M Falkum E Schweder T Ursin H Health complaints and job stress in Norwegian physicians: the use of an overlapping questionnaire design Soc Sci Med 1997 45 1615 1629 9428082 10.1016/S0277-9536(97)00093-2
Tambs K Moum T How well can a few questionnaire items indicate anxiety and depression? Acta Psychiatr Scand 1993 87 364 367 8517178
Tyssen R Vaglum P Aasland OG Grønvold NT Ekeberg Ø Use of alcohol to cope with tension, and its relation to gender, years in medical school and hazardous drinking: a study of two nation-wide Norwegian samples of medical students Addiction 1998 93 1341 1349 9926540 10.1046/j.1360-0443.1998.93913415.x
Strand BH Dalgard OS Tambs K Rognerud M Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36) Nord J Psychiatry 2003 57 113 118 12745773 10.1080/08039480310000932
Rosvold EO Physicians in illness and health. An epidemiological study on health- and illness behaviour among 19 th and 20 th century Norwegian physicians PhD thesis 2002 University of Oslo, Institute of General Practice and Community Medicine accessed September 30, 2005
Aasland OG Baker's children (in Norwegian) Tidsskr Nor Lægeforen 2001 121 3504 11808007
Runeson B Dahlin M Åsberg M Nygren A Stress and depressive symptoms in physicians and medical students in Sweden [abstract] Eur Psychiatry 2004 19 49s
Chew-Graham CA Rogers A Yassin N "I wouldn't want it on my CV or their records": medical students' experiences of help-seeking for mental health problems Med Educ 2003 37 873 880 12974841 10.1046/j.1365-2923.2003.01627.x
Hooper C Meakin R Jones M Where students go when they are ill: how medical students access health care Med Educ 2005 39 588 593 15910435 10.1111/j.1365-2929.2005.02175.x
McKevitt C Morgan M Illness doesn't belong to us J R Soc Med 1997 90 491 495 9370984
Spencer J Physician, heal thyself – but not on your own, please Med Educ 2005 39 548 549 15910427 10.1111/j.1365-2929.2005.02209.x
Fromme EK Hebert RS Carrese JA Self-doctoring: a qualitative study of physicians with cancer J Fam Pract 2004 53 299 306 15068775
|
16242034
|
PMC1276798
|
CC BY
|
2021-01-04 16:27:56
|
no
|
BMC Med. 2005 Oct 21; 3:16
|
utf-8
|
BMC Med
| 2,005 |
10.1186/1741-7015-3-16
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-621622568510.1186/1471-2180-5-62Research ArticleABC transporter FtsABCD of Streptococcus pyogenes mediates uptake of ferric ferrichrome Hanks Tracey S [email protected] Mengyao [email protected] Michael J [email protected] Benfang [email protected] Department of Veterinary Molecular Biology, Montana State University, Bozeman, Montana 59717, USA2005 14 10 2005 5 62 62 10 9 2005 14 10 2005 Copyright © 2005 Hanks et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The Streptococcus pyogenes or Group A Streptococcus (GAS) genome encodes three ABC transporters, namely, FtsABCD, MtsABC, and HtsABC, which share homology with iron transporters. MtsABC and HtsABC are believed to take up ferric (Fe3+) and manganese ions and heme, respectively, while the specificity of FtsABCD is unknown.
Results
Recombinant FtsB, the lipoprotein component of FtsABCD, was found to bind Fe3+ ferrichrome in a 1:1 stoichiometry. To investigate whether FtsABCD transports Fe3+ ferrichrome, GAS isogenic strains defective in lipoprotein gene ftsB and permease gene ftsC were generated, and the effects of the mutations on uptake of Fe3+ ferrichrome were examined using radioactive 55Fe3+ ferrichrome. FtsB was produced in the wild-type strain but not in the ftsB mutant, confirming the ftsB inactivation. While wild-type GAS took up 3.6 × 104 Fe3+ ferrichrome molecules per bacterium per min at room temperature, the ftsB and ftsC mutants did not have a detectable rate of Fe3+ ferrichrome uptake. The inactivation of ftsB or ftsC also decreased 55Fe3+ ferrichrome uptake by >90% under growth conditions in the case of limited uptake time. Complementation of the ftsB mutant with a plasmid carrying the ftsB gene restored FtsB production and 55Fe3+ ferrichrome association at higher levels compared with the parent strain. The inactivation of mtsA and htsA and Fe-restricted conditions enhanced the production of FtsB and Fe3+ ferrichrome uptake.
Conclusion
The FtsB protein bound Fe3+ ferrichrome, and inactivation of ftsB or ftsC, but not htsA or mtsA, diminished Fe3+ ferrichrome uptake, indicating that FtsABCD, but not HtsABC and MtsABC, is the transporter that takes up Fe3+ ferrichrome in GAS. Fe acquisition systems are virulence factors in many bacterial pathogens and are attractive vaccine candidates. The elucidation of the FtsABCD specificity advances the understanding of Fe acquisition processes in GAS and may help evaluating the GAS Fe acquisition systems as vaccine candidates.
==== Body
Background
Ferric iron (Fe3+), the stable iron form in an oxidative environment, has extremely low solubility in water under physiological conditions, and mammalian hosts thus do not have sufficient free Fe3+ to support bacterial growth [1]. The major sources of iron in vivo for bacteria are host heme-proteins and other iron complexes [1,2]. Many bacterial pathogens secrete low-molecular-weight iron chelators called siderophores to assimilate iron from host environments [3]. Ferrisiderophores formed are then transported across the cytoplasmic membrane by specific ATP-binding cassette (ABC) type transporters. ABC transporters consist of a solute-binding protein, a membrane protein (permease) encoded by one or two genes, and an ATPase [4]. The solute-binding proteins are located in the periplasmic space in Gram-negative bacteria and are lipoproteins in Gram-positive organisms. Siderophores can be divided into several types based on chemical structures [3]. Ferrichrome [5] belongs to the hydroxamate type.
Streptococcus pyogenes or Group A Streptococcus (GAS) is an important Gram-positive human pathogen causing both invasive and non-invasive infections [6]. Non-invasive infections, including pharyngitis, and post-infection sequelae, such as acute rheumatic fever, rheumatic heart disease, and glomerulonephritis, result in substantial morbidity and economic loss globally. Invasive GAS infections, such as necrotizing fasciitis and streptococcal toxic shock syndrome, are associated with high mortality rates. GAS can take up heme from hemoglobin and haptoglobin-hemoglobin complexes [7]. Exogenously-supplied heme and host heme proteins (hemoglobin, myoglobin, and catalase), but not iron-loaded transferrin and lactoferrin, support in vitro growth of GAS under iron-restricted conditions [8].
Iron acquisition processes in GAS are poorly understood, although progress has recently been made by us [9-11] and other groups [12-14]. GAS genomes [15-17] encode three ABC transporters, namely, HtsABC [9-11] or SiaABC [12], MtsABC [13,14], and one encoded by spy0383 to spy0386 [18] (designated FtsABCD), which are homologues of ABC transporters involved in iron acquisition. HtsABC and the cell-surface protein Shp [9-12] are believed to make up the machinery for heme acquisition. The lipoprotein component MtsA of MtsABC binds Fe3+, Zn2+, and Cu2+ [13], and MtsABC is important for acquisition of Mn2+ and Fe3+ [14]. The transcription of ftsABCD is up-regulated under iron-restricted conditions [18]. However, recombinant FtsB was heme-free and did not contain Fe, Mn, or Zn [10]. The specificity of FtsABCD is thus not known. It is not known whether GAS can use ferrisiderophores such as ferric ferrichrome (Fe3+ ferrichrome) as an iron source.
We found that recombinant FtsB bound Fe3+ ferrichrome, suggesting that FtsABCD is involved in the acquisition of Fe3+ ferrichrome. To test this hypothesis, GAS isogenic strains defective in lipoprotein gene ftsB and permease gene ftsC were generated, and the ftsB and ftsC inactivation dramatically decreased the uptake of Fe3+ ferrichrome, indicating that FtsABCD is the transporter that takes up Fe3+ ferrichrome in GAS.
Results
Binding of Fe3+ ferrichrome by FtsB
The specificity of an ABC transporter is believed to depend on the binding specificity of its solute binding protein. Purified recombinant FtsB did not contain free metal ions, suggesting that FtsABCD does not target free metal ions [10]. To test whether FtsABCD targets ferrisiderophores, the ability of FtsB to bind Fe3+ ferrichrome was examined by gel filtration. FtsB was incubated with excess Fe3+ ferrichrome and separated from free Fe3+ ferrichrome by chromatography on a Sephadex G-25 column. Fe3+ ferrichrome has an absorption peak at 425 nm and its elution profile was monitored by measuring A425 of each fraction. The elution profile displayed two A425 peaks (Fig. 1A). One co-migrated with the FtsB peak, which was localized by A280, and the other corresponded to free Fe3+ ferrichrome (Fig. 1A). When FtsB was replaced with MtsA in a control experiment, there was only one A425 peak corresponding to free Fe3+ ferrichrome, and no absorbance at 425 nm was associated with the protein peak (Fig. 1B). To check whether the species associated with FtsB was Fe3+ ferrichrome, the absorption spectra of the FtsB and MtsA peaks and free Fe3+ ferrichrome were compared. In addition to the protein absorption peak at A280, the FtsB sample had an absorption peak at 425 nm which was identical to that of free Fe3+ ferrichrome (Fig. 1C). As expected, the MtsA sample only had the protein peak (Fig. 1C). These results indicate that FtsB, but not MtsA, bound Fe3+ ferrichrome. The FtsB sample was found to have 1.08 Fe3+ ferrichrome per FtsB molecule on the basis of protein content and extinction coefficient of Fe3+ ferrichrome at 425 nm, indicating a 1:1 binding stoichiometry.
If FtsB binds Fe3+ ferrichrome, Fe3+ should co-migrate with the protein on the G-25 column. To test this idea, Fe3+ ferrichrome containing 1.1% 55Fe3+ of total Fe3+ was used to repeat the gel filtration experiment, and 55Fe3+ radioactivity was monitored. As expected, one of two 55Fe3+ peaks co-migrated with the FtsB peak, and the other peak corresponded to free Fe3+ ferrichrome (Fig. 1D). On the basis of 55Fe3+ percentage of total iron and specific activity and protein content, the FtsB peak fraction contained 37 μM FtsB and 36 μM Fe, consistent with the 1:1 binding stoichiometry determined above. These results confirmed that FtsB binds Fe3+ ferrichrome in a 1:1 molar ratio.
Effects of ftsB, ftsC, mtsA, and htsA inactivation on Fe3+ ferrichrome uptake
The binding results described above suggest that FtsABCD can acquire Fe3+ ferrichrome. To test this hypothesis, ftsB, ftsC, and mtsA (control) were first inactivated by insertional inactivation (Fig. 2). Inactivation was confirmed by PCR (Fig. 2C) and DNA sequencing. Western blotting analysis detected FtsB (Fig. 2D) and MtsA (Fig. 2E) in the wild-type strain but not in the corresponding mutant strain. These results indicate that ftsB and mtsA were indeed inactivated. The ftsB, ftsC, mtsA, and htsA mutants (construction of the htsA mutant will be described elsewhere) and parent strains were compared in the uptake of 55Fe3+ ferrichrome. Pilot experiments indicated that the wild-type strain could take up 55Fe3+ ferrichrome at both room temperature (24°C) and 37°C. The uptake was thus performed at room temperature for convenience. GAS cells harvested from exponential growth phase were incubated with 0.16 μM 55Fe3+ ferrichrome at room temperature for 1 h, and 55Fe3+ radioactivity associated with bacteria was determined. The ftsB and ftsC inactivation diminished the uptake of Fe3+ ferrichrome (Fig. 3), while the mtsA and htsA inactivation did not abolish but enhanced the uptake. To further examine the uptake of Fe3+ ferrichrome, 55Fe3+ radioactivity taken up by bacteria as a function of incubation time was determined. 55Fe3+ radioactivity with wild-type GAS increased linearly with time up to 30 min with a slope of 153 cpm/min (Fig. 4). The value of the slope could be translated into an uptake rate of 3.6 × 104 Fe3+ ferrichrome molecules per min per wild-type GAS bacterium. In contrast, the ftsB and ftsC mutants did not have a detectable rate of Fe3+ ferrichrome uptake (Fig. 4). The rate of uptake in the mtsA mutant was 2.6 times as that in the parent strain (Fig. 4). These results indicate that FtsABCD, not MtsABC and HtsABC, mediates the uptake of Fe3+ ferrichrome.
Complementation
To test whether the effect of the ftsB inactivation on the uptake of Fe3+ ferrichrome was due to the absence of FtsB, a complementing strain,ftsB mutant/pCMVftsB, was constructed to carry plasmid pCMVftsB with the ftsB gene. The intensity of FtsB band of the complemented strain in Western blotting analysis in Fig. 2D was 7/10 of that of the wild-type strain, and the amount of the complemented strain cells used was 1/125 of that of the wild-type cells, indicating that FtsB was produced in ftsB mutant/pCMVftsB at a level as 88 times as that in the wild-type strain. Wild-type, ftsB mutant, and ftsB mutant/pCMVftsB cells harvested in the exponential growth phase were incubated with 0.16 μM 55Fe3+ ferrichrome at 24°C for 1 h, and 55Fe3+ radioactivity associated with the bacteria was determined. 55Fe3+ radioactivity of ftsB mutant/pCMVftsB was 1.9 and 207 times higher than those of the wild-type and ftsB mutant cells, respectively (Fig. 5), suggesting that in trans expression of ftsB in the ftsB mutant restored the association of Fe3+ ferrichrome. The results support that FtsABCD targets Fe3+ ferrichrome.
Effect of Fe3+ ferrichrome on GAS growth
Since FtsABCD is involved in the uptake of Fe3+ ferrichrome, Fe3+ ferrichrome should be an iron source of GAS. This idea was tested by comparing GAS growth curves in THY treated with Chelex 100 to remove metal ions and supplemented with MgCl2 (DTHYMg) in the absence and presence of Fe3+ ferrichrome. The growth curve in the presence of 10 μM Fe3+ ferrichrome shifted to the left by about 40 min compared with that in the absence of Fe3+ ferrichrome under otherwise identical conditions. This small but repeatable stimulatory effect suggests that GAS can use Fe3+ ferrichrome as an iron source. Fe3+ ferrichrome had similar stimulatory effect on the growth of the mutants as that on the growth of the wild type strain, suggesting that either residual uptake of Fe3+ ferrichrome in the ftsB and ftsC mutants were enough to induce the stimulatory effect or an additional process was involved in the uptake of the iron extracted from Fe3+ ferrichrome under the growth conditions.
Effects of ftsB and ftsC inactivations on the uptake of Fe3+ ferrichrome under growth conditions
To examine why the stimulatory growth effect of Fe3+ ferrichrome was still observed in the ftsB and ftsC mutants, 55Fe3+ ferrichrome uptake was examined under growth conditions. 55Fe3+ ferrichrome was added into the cultures of wild-type, ftsB, and ftsC strains at mid-exponential growth phase, and, 30 min later, 55Fe3+ activity associated with bacteria was determined. 55Fe3+ radioactivity of ftsB and ftsC mutant cells were only 7.5% and 5.3% of that of wild-type cells, respectively (Fig. 6A), indicating that the ftsB and ftsC inactivation also dramatically diminished uptake of Fe3+ ferrichrome under growth conditions. However, 55Fe3+ radioactivity of ftsB and ftsC cells grown for 3 h after the addition of 55Fe3+ ferrichrome were about 50% of that of wild-type cells (Fig. 6B). These results suggest that there could be an additional acquisition process which could assimilate 55Fe3+ from its ferrichrome complex.
Effects of mtsA and htsA inactivation and Fe-restricted conditions on Fe3+ ferrichrome uptake and FtsB production
To further examine the factors to affect Fe3+ ferrichrome uptake, the effects of mtsA and htsA inactivation and Fe-restricted conditions on FtsB production and Fe3+ ferrichrome uptake were investigated. Inactivation of htsA and mtsA increased 55Fe3+ ferrichrome uptake by more than 100% (Fig. 7A). Wild-type, htsA, and htsA mutant cells harvested from THY containing 2,2'-dipyridyl took up 30%–50% more 55Fe3+ ferrichrome than those from THY without 2,2'-dipyridyl (Fig. 7A). Western blotting analysis indicated higher levels of FtsB in the htsA and mtsA mutants compared with the wild-type strain (Fig. 7B). Higher levels of FtsB were also detected in all strains grown in the presence of 2,2'-dipyridyl compared with those grown in the absence of 2,2'-dipyridyl (Fig. 7B). A control protein, phosphoglycerate kinase (PGK), had similar expression levels under all the conditions (Fig. 7B). The relative intensities of the FtsB bands in the Western blot correlated well with the relative uptake of 55Fe3+ ferrichrome under the same conditions. These results indicate that FtsB production was enhanced in the htsA and mtsB mutants and under Fe-restricted conditions, further supporting the role of FtsABCD in Fe acquisition.
Discussion
Evidences from this study indicate that FtsABCD is the transporter that takes up Fe3+ ferrichrome in GAS. The evidences include the binding of Fe3+ ferrichrome to FtsB, the effect of the insertional disruption of ftsB or ftsC on uptake of 55Fe3+ ferrichrome, and the non-involvement of MtsABC and HtsABC in Fe3+ ferrichrome uptake. Complementation data of the ftsB mutant with the ftsB gene expressed in trans indicated that the effect of ftsB inactivation was due to the lack of FtsB. MtsABC [13,14] and HtsABC [9-12] target free Fe3+ and heme, respectively. The specificity of the FtsABCD transporter elucidated in this study resolved the last piece of the puzzle regarding the roles of ABC transporters in Fe acquisition in GAS.
Uptake time was critical to the effects of ftsB and ftsC inactivation on 55Fe3+ activity associated with bacteria under growth conditions. The inactivation had a dramatic effect (>90% decrease in 55Fe3+ activity compared with wild-type strain) when uptake was performed for only 30 min. This decrease was reduced to about 50% when uptake was performed for 3 h. Although the reasons for this reduction are not known, the reduction is unlikely due to the existence of another transporter for Fe3+ ferrichrome. The other uptake results are not consistent with the existence of another Fe3+ ferrichrome transporter. Another possible reason is that 55Fe3+ ferrichrome exchanged its 55Fe3+ with another ferric complex in THY, resulting in a non-ferrichrome 55Fe3+ complex that could be taken up by another transporter. GAS has a putative secreted Fe binding protein (Spy1063). It is not known whether this protein can extract Fe3+ from Fe3+ ferrichrome.
The transcription of ftsABCD is up-regulated under Fe-reduced conditions [18]. Consistent with this observation, Fe-restricted conditions enhanced the production of FtsB and uptake of Fe3+ ferrichrome, further supporting the role of FtsABCD in Fe acquisition. Inactivation of htsA and mtsA also enhanced FtsB production and Fe3+ ferrichrome uptake. Apparently, all the three ABC transporters contributed to Fe3+ acquisition in GAS grown in THY, and inactivation of either mtsA or htsA might result in lower intracellular Fe levels and, in turn, enhanced the expression of ftsABCD. The results also suggest that the expression of ftsABCD, htsABC, and mtsABC is coordinately regulated.
Some Enterococcus faecium clinical strains do not produce siderophores but can acquire iron using exogenous siderophores produced by other bacteria living in the same habitats [19]. GAS is not known to produce siderophores, and GAS genomes [15-17] do not have genes encoding homologues of siderphore-production systems. We could not detect siderophore production in GAS under iron-restricted conditions. Therefore, GAS may not produce siderophors. However, GAS takes up Fe3+ ferrichrome, suggesting that GAS could acquire Fe3+ by using siderophores produced by other bacteria in the pharynx and skin, the noninvasive GAS infection sites.
Fe acquisition systems are virulence factors in many bacterial pathogens [20-24] and are attractive vaccine targets [25-30]. Elucidation of the specificities of the Fe transporters in GAS will facilitate determination of their relative importance in various infections and choose appropriate animal infection models to evaluate their efficacy as vaccine candidates. For examples, HtsABC could be more important in invasive GAS infection since heme should be the Fe source, and FtsABCD could be important in non-invasive infections because exogenous ferric siderophore complexes should be available.
In summary, we found that FtsB bound Fe3+ ferrichrome and that ftsB or ftsC inactivation dramatically decreased Fe3+ ferrichrome uptake. The results indicate that FtsABCD is the transporter that acquires Fe3+ ferrichrome in GAS.
Methods
Materials
Iron chelating agent 2,2'-dipyridyl was obtained from Aldrich. Sephadex G-25, iron-free ferrichrome A from Ustilago sphaerogena, Chelex 100, and other chemicals were purchased from Sigma (St. Louis, MO). 55Fe3+-labeled ferric chloride was purchased from RI Consultants LLC (Hudson, NH). Purified recombinant FtsB (Spy0385) and MtsA were prepared as described previously [10,31].
Mouse antisera
Five female outbred CD-1 Swiss mice (4- to 6-week-old) (Charles River Laboratories, Wilmington, MA) were immunized subcutaneously with 50 μg of recombinant FtsB, MtsA, or PGK suspended in 200 μL of saline emulsified in 44 μL of monophosphoryl lipid A-synthetic trehalose dicorynomycolate adjuvant (Corixa, Hamilton, MT). Mice were boosted at weeks 2 and 4. Immune sera were collected 5 days after the second boost.
Bacterial strains and growth
Serotype M1 GAS strain MGAS5005 has been described previously [32]. MGAS5005 and its isogenic mutants were grown routinely at 37°C in 5% CO2 in Todd-Hewitt broth (Difco Laboratories, Detroit, MI) supplemented with 0.2% yeast extract (THY). Spectinomycin (150 mg/L) was added into THY for mutant strains. Iron-restricted conditions were achieved by adding 0.3 mM 2,2'-dipyridyl into THY and by treating THY with the chelating resin Chelex 100 and supplementing it with 0.4 mM MgCl2 (DTHYMg). Tryptose agar with 5% sheep blood (Becton Dickinson, Cockeysville, MD) and THY agar were used as solid media.
Binding of Fe3+ ferrichrome to FtsB
Gel filtration was used to detect the binding of Fe3+ ferrichrome to FtsB. FtsB (0.3 ml of 0.18 mM) was incubated with 0.9 mM Fe3+ ferrichrome for 20 min at room temperature, loaded onto a Sephadex G-25 column (1.5 × 18 cm), and eluted with 20 mM Tris-HCl buffer, pH 8.0. Eluant was collected as fractions of 0.7 ml. The absorbance of each fraction attributable to protein and Fe3+ ferrichrome was measured. MtsA as a negative control was similarly tested for the binding of Fe3+ ferrichrome. The experiment was repeated with Fe3+ ferrichrome in which 1.1% of total Fe3+ was 55Fe, and A280 and 55Fe3+ radioactivity of each fraction were measured. 55Fe3+ radioactivity was measured using a window of 0–6 keV with a Packard 1500 Tri-Carb Liquid Scintillation Analyzer.
GAS isogenic mutants
MGAS5005 isogenic mutants defective in ftsC, ftsB, or mtsA were generated by insertional inactivation (Fig. 2). An internal ~400-bp fragment of each gene was PCR-amplified using MGAS5005 genomic DNA as template and the primers listed in Table 1. The PCR products were digested with NcoI and ligated to pFWaad at the NcoI site to yield suicide plasmids. The orientation of the fragments was determined by DNA sequencing, and the suicide plasmids in which the orientation of the fragments was opposite to that of the aad gene were chosen for generating the mutants. To obtain pFWaad, the spectinomycin-resistant gene aad [33] was amplified with primers 5'-AGTGTCGACTATAACTAATAACGTAACGTG-3' and 5'-ACCATGGGAATTCTATAATTTTTTTAATCTGTTATTTA-3' and cloned into pFW14 [34] at the SalI and NcoI restriction sites.
Each suicide plasmid was introduced into MGAS5005 by electroporation at 1.8 kV and 400 Ω. One ml of THY was added into the sample immediately after electroporation. The sample was incubated at 37°C for 2 h and plated on THY agar plates supplemented with 150 mg of spectinomycin per liter to select insertional mutants. The plates were incubated in 5% CO2 at 37°C for two days, and the colonies obtained were screened by PCR analysis using the primers listed in Table 1. Gene interruptions were then confirmed by sequencing the PCR products.
Construction of ftsB-complementing plasmid pCMVftsB
Plasmid pJRS525 [35] was modified to replace the spectinomycin-resistant gene with the chloramphenicol-resistance gene. A 2545-bp fragment without the spectinomycin-resistant gene was amplified from pJRS525 with primers 5'-TCGTGGATCC AAGCTTCACCATGG-3' and 5'-AAGAATTCCTTGCATAGACTTTCGTCAG-3'. The fragment containing the chloramphenicol-resistant gene was amplified from pFW14 [23] with primers 5'-GGAATTCCGGATGCATATGCATG-3' and 5'-GTCCGGATCCTCGAGCTCTAGATC-3'. The PCR products were digested with BamHI and EcoRI and ligated together to yield plasmid pCMV. A DNA fragment containing ftsB and its ribosome-binding site was amplified from MGAS5005 using primers 5'-CGGATCCAATAACTTTATTCTAGGAGAATTAG-3' and 5'-AGGGATCCTTAGTTTTCACTTGATAAGATTG-3'. The PCR product was digested with BamHI and ligated into pCMV at the BamHI site, yielding pCMVftsB containing ftsB. The cloned gene was sequenced to rule out spurious mutations and confirm the desired orientation. The resulting pCMVftsB was introduced into the ftsB mutant by electroporation, and the complement strain (designated ftsB/pCMVftsB) was selected by spectinomycin and chloramphenicol and confirmed for the existence of pCMVftsB by colony PCR using primers 5'-CAATTTCACACAGGAAACAGC-3' (pCMV-specific) and 5'-AGGGATCCTTAGTTTTCACTTGATAAGATTG-3' (ftsB-specific)
Uptake of Fe3+ ferrichrome by GAS
Fe3+ ferrichrome uptake by GAS was monitored using radioactive 55Fe3+ ferrichrome under non-growth and growth conditions. Under the non-growth conditions, wild-type MGAS5005 and its ftsB, ftsC, or mtsA mutant strains were harvested from the mid-exponential growth phase (OD600 of 0.4) by centrifugation. The bacterial pellets were washed with 10 ml of THY and resuspended in 1 ml of THY. To prepare 55Fe3+ ferrichrome working solution, 4.7 nmole 55FeCl3 was incubated with 14.1 nmole Fe-free ferrichrome A in 50 μL of Tris-HCl for 10 min, and the complex was diluted with 22 ml of THY. The 1-ml bacterial suspensions were mixed with 3 ml of the 55Fe3+ ferrichrome solution (final Fe3+ ferrichrome concentration 0.16 μM) to initiate the uptake process and rotated in a 10-ml tube from end to end at room temperature. A triplet of 0.2 ml samples were taken from each mixture at the indicated times, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. For uptake under growth conditions, 0.16 μM 55Fe3+ ferrichrome was added into the cultures of wild-type, ftsB, and ftsC strains at early exponential or mid-exponential growth phase. A triplet of 1-ml culture samples were taken at the indicated times after the 55Fe3+ ferrichrome addition and treated as described above. The pellets obtained were resuspended in 0.2 ml of THY and mixed with 3 ml of scintillation liquid. 55Fe3+ radioactivity associated with the bacteria was measured as described above.
FtsB and MtsA production
Production of FtsB or MtsA in wild-type and mutant strains was monitored by Western blotting analysis. To prepare samples for the analysis, bacteria were harvested from cultures at the indicated volumes in exponential phase. The bacterial pellets were washed twice with 1.3 ml of PBS, resuspended in 100 μl of PBS, and treated with 200 units of mutanolysin at 37°C for 2 h. The samples were briefly sonicated and mixed with equal volume of 2x SDS-PAGE loading buffer. Proteins in 10 μl of the samples were resolved by SDS-PAGE and transferred to nitrocellulose membranes. The proteins were detected by Western blotting using specific mouse antiserum as previously described [36].
Other procedures and measurements
Chromosomal DNA was isolated with the Puregene DNA Isolation kit (Gentra Systems, Minneapolis, MN). Sequence data were obtained with an ABI 310 automated DNA sequencer (Applied Biosystems, Inc., Foster City, CA). Absorbance and optical spectra were obtained with a SPECTRAmax 384 Plus spectrophotometer (Molecular Devices, Sunnyvale, CA). Protein concentrations were determined with the modified Lowry protein assay kit purchased from Pierce (Rockford, IL) with bovine serum albumin as a standard. Intensities of the bands in Western blotting analysis were determined using an AlphaImager 2000 Documentation & Analysis System (Alpha Innotech Corp.).
Authors' contributions
TSH carried out the ferric ferrichrome binding assay and the characterization of the GAS isogenic mutants and participated in writing the Methods section. ML generated the GAS isogenic mutants. MJM constructed the plasmid for complementing the ftsB mutant. BL designed the study and drafted the manuscript.
Acknowledgements
This work was supported in part by grants P20 RR-16455 and P20 RR-020185 from the National Center for Research Resources, K22AI057347 from National Institutes of Health, and the Montana State University Agricultural Experimental Station. We thank Dr. Kevin McIver for providing plasmid pJRS525.
Figures and Tables
Figure 1 Binding of Fe3+ ferrichrome to FtsB. (A) Co-migration of Fe3+ ferrichrome with FtsB. FtsB (0.3 ml of 0.18 mM) was incubated with 0.9 mM Fe3+ ferrichrome for 20 min at room temperature, loaded onto a Sephadex G-25 column (1.5 × 18 cm), and eluted with 20 mM Tris-HCl buffer, pH 8.0. Eluant was collected as fractions of 0.7 ml. A280 and A425 of each fraction were used to monitor FtsA and Fe3+ ferrichrome, respectively. (B) MtsA, as a control, did not bind Fe3+ ferrichrome. The experiment was performed under the conditions as those in panel A. (C) The absorption spectra of the protein peaks (fraction 9) in panel A (solid curve) and panel B (short dashed curve) and free Fe3+ ferrichrome (long dashed curve). The three samples contained 46 μM FtsB, 45 μM MtsA, and 100 μM Fe3+ ferrichrome, respectively. (D) Co-migration of partially 55Fe-labeled Fe3+ ferrichrome with FtsB. Fe3+ ferrichrome contained radioactive 55Fe3+ at 1.1% of total Fe3+. A280 and 55Fe3+ radioactivity of each fraction are presented.
Figure 2 Insertional inactivation of ftsB, ftsC, and mtsA. (A) Schematic showing the arrangement of ftsABCD genes and their neighbors in the MGAS5005 genome. The ftsA, ftsB, and ftsCD genes encode ATP-binding protein, lipoprotein, and permease, respectively. The numbers above the other arrows are the spy numbers assigned to the corresponding open reading frames in the M1 genome sequence (15). (B) Schematic for insertional gene inactivation. The paired solid or dotted arrows under the mutant genome indicate the locations of primers used to confirm the disruption of the gene using PCR and DNA sequencing. (C) PCR confirmation of insertional inactivation of ftsB, ftsC, and mtsA. The picture shows agarose gel analysis of PCR reactions using mutant (lanes labeled by 1) or wild type (lanes labeled by 2) genomic DNA as template and primers at the locations indicated by the solid arrows under the mutant genome in panel B. (D) Western blot showing the absence of FtsB in the ftsB mutant strain (lane 2) and the presence of FtsB in the wild-type (lane 1) and ftsB/pCMVftsB (lane 3) strains. Proteins from 5 × 108 wild-type cells, 5 × 108 ftsB mutant cells, and 4 × 106 ftsB/pCMVftsB were probed with FtsB-specific mouse antiserum. The amount of ftsB/pCMVftsB cells used was 1/125 of that of the wild-type cells. (E) Western blot showing the presence of MtsA in the wild-type strain (lane1) and the absence of MtsA in the mtsA mutant strain (lane 2). Proteins from 6 × 106 wild-type cells and 3 × 108 mtsA mutant cells were probed with MtsA-specific mouse antiserum.
Figure 3 Effects of ftsB, ftsC, mtsA, and htsA inactivation on 55Fe3+ ferrichrome uptake. Wild-type and mutant strains were grown to OD600 of 0.4 in 10 ml of THY and THY/spectinomycin, respectively. The bacterial pellets were washed with 1.0 ml of THY and incubated with 0.16 μM 55Fe3+ ferrichrome 1.0 ml of THY at room temperature for 1 h. A triplet of 0.2 ml samples were taken from each mixture, and bacteria were pelleted and washed twice with 0.4 ml of THY. 55Fe3+ radioactivity associated with the bacteria was measured. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria in a representative of three experiments.
Figure 4 The time course of 55Fe3+ ferrichrome uptake for mtsA (open circles), ftsB (open triangles), and ftsC (solid triangles) mutant and wild-type (solid circles) GAS strains. Each strain harvested at the mid-exponential growth phase from 40 ml culture, washed with 10 ml of THY, and incubated with 0.16 μM 55Fe3+ ferrichrome in 4 ml of THY at room temperature. A triplet of 0.2 ml samples were taken from each mixture at the indicated times, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria in a representative of three experiments.
Figure 5 Complementation of the ftsB mutant. Plasmid pCMVftsB carrying the ftsB gene was introduced into the ftsB mutant strain to obtain ftsB/pCMVftsB. The uptake experiment was performed exactly as in Fig. 3. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria.
Figure 6 Effects of ftsB and ftsC inactivation on 55Fe3+ ferrichrome uptake under growth conditions. (A) Uptake for 30 min. 55Fe3+ ferrichrome (0.16 μM) was added into cultures of the wild-type and mutant strains at 37°C in THY and THY supplemented with spectinomycin, respectively, when OD600 was 0.3. A triplet of 1 ml samples were taken from each culture 30 min later, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria. (B) Uptake for 3 h. 55Fe3+ ferrichrome (0.16 μM) was added into the wild-type and mutant cultures at 37°C when OD600 was 0.2, and, 3 h later, the samples were processed as in panel A for determination of 55Fe3+ radioactivity.
Figure 7 Effects of Fe-restricted conditions and mtsA and htsAinactivation on Fe3+ ferrichrome uptake and FtsB production. (A). 55Fe3+ radioactivity uptake by wild-type (wt), ftsB, ftsC, mtsA, and htsA mutant strains grown in THY in the absence (solid bars) or presence (slashed bars) of 0.3 mM 2,2'-dipyridyl. The experiment was performed as in Fig. 3. (B) Western blots showing the relative levels of FtsB and PGK as a control in the strains grown under the same conditions as in panel A. Proteins from 1 × 108 bacteria of each strain were probed with mouse anti-FtsB and anti-PGK antisera. The relative levels of FtsB and PGK were obtained by dividing the intensity of each band by the intensity of the FtsB or PGK band of the wild-type strain.
Table 1 Primers used in gene inactivation and confirmation
Primer Sequence (5'-3') Purpose
1 ACCATGGCTATGCCTCAAATTGCTGG ftsB inactivation
2 ACCATGGTTTTCCAATCTTTTAACCAC
3 ACCATGGAGCTTATGTGTTGCTATTTAC ftsC inactivation
4 ACCATGGATAGCTCCTGCTAAGACAAG
5 ACCATGGTCTTTTAGTAGCTTGTTCGTC mtsA inactivation
6 ACCATGGTTCATTTCGACAAGAGGCTG
7 AACTCTACTATTAACACTCTG confirmation of ftsB inactivation
8 GCAACGACGACATAATCACC
9 GTGTCTCGAGCGTTTTTGC confirmation of ftsC inactivation
10 CAATATAGCCGATCATTTG
11 ATGAGCCTCATTTTGGGTGC confirmation of mtsA inactivation
12 GCCACATAAGCTTTTAGGTTC
13 AGAATTTTGTTAGCAGTTCG aad-specific, paired with primer 7, 9, or 11 for inactivation confirmation
14 CTTTGAGTGAGCTGATACCG pFWaad-specific, paired with primer 8, 10, or 12 for inactivation confirmation
==== Refs
Weinberg ED Cellular iron metabolism in health and disease Drug Metab Health Dis 1990 22 532 579
Otto BR Verweij-van Vught AMJJ MacLaren DM Transferrins and heme-compounds as iron sources for pathogenic bacteria Crit Rev Microbiol 1992 18 217 233 1532495
Neilands JB Microbial iron compounds Ann Rev Biochem 1981 50 715 731 6455965 10.1146/annurev.bi.50.070181.003435
Higgins CF ABC transporters: From microorganisms to man Annu Rev Cell Biol 1992 8 67 113 1282354 10.1146/annurev.cb.08.110192.000435
Ecker DJ Passavant CW Emery T Role of two siderophores in Ustilago sphaerogena : Regulation of biosynthesis and uptake mechanisms Biochim Biophys Acta 1982 720 242 249 6213273 10.1016/0167-4889(82)90047-7
Cunningham MW Pathogenesis of group A streptococcal infections Clin Microbiol Rev 2000 13 470 511 10885988 10.1128/CMR.13.3.470-511.2000
Francis RT JRBooth JW Becker RR Uptake of iron from hemoglobin and the haptoglobin-hemoglobin complex by hemolytic bacteria Int J Biochem 1985 17 767 773 3902529 10.1016/0020-711X(85)90262-9
Eichenbaum Z Muller E Morse SA Scott JR Acquisition of iron from host proteins by group A Streptococcus Infect Immun 1996 64 5428 5429 8945602
Lei B Liu M Smoot LM Menning HM Voyich JM Kala SV DeLeo FR Musser JM Identification and characterization of a novel heme-associated cell-surface protein made by Streptococcus pyogenes Infect Immun 2002 70 4494 4500 12117961 10.1128/IAI.70.8.4494-4500.2002
Lei B Liu M Voyich JM Prater CI Kala SV DeLeo FR Musser JM Identification and Characterization of HtsA, A Second Heme-Binding Protein Made by Streptococcus pyogenes Infect Immun 2003 71 5962 5969 14500516 10.1128/IAI.71.10.5962-5969.2003
Liu M Lei B Heme transfer from streptococcal cell-surface protein Shp to HtsA of transporter HtsABC Infect Immun 2005 73 5086 5092 16041024 10.1128/IAI.73.8.5086-5092.2005
Bates CS Montanez GE Woods CR Vincent RM Eichenbaum Z Identification and characterization of a Streptococcus pyogenes operon involved in binding of hemoproteins and acquisition of iron Infect Immun 2003 71 1042 1055 12595414 10.1128/IAI.71.3.1042-1055.2003
Janulczyk R Pallon J Bjorck L Identification and characterization of a Streptococcus pyogenes ABC transporter with multiple specificity for metal cations Mol Microbiol 1999 34 596 606 10564500 10.1046/j.1365-2958.1999.01626.x
Janulczyk R Ricci S Bjorck L MtsABC is important for manganese and iron transport, oxidative stress resistance, and virulence of Streptococcus pyogenes Infect Immun 2003 71 2656 2664 12704140 10.1128/IAI.71.5.2656-2664.2003
Ferretti JJ McShan WM Ajdic D Savic DJ Savic G Lyon K Primeaux C Sezate S Suvorov AN Kenton S Lai HS Lin SP Qian Y Jia HG Najar FZ Ren Q Zhu H Song L White J Yuan X Clifton SW Roe BA McLaughlin R Complete genome sequence of an M1 strain of Streptococcus pyogenes Proc Natl Acad Sci USA 2001 98 4658 4663 11296296 10.1073/pnas.071559398
Smoot JC Barbian KD Van Gompel JJ Smoot LM Chaussee MS Sylva GL Sturdevant DE Mackie SM Porcella SF Parkins LD Beres SB Campell DS Smith TM Zhang Q Kapur V Daly JA Veasy LG Musser JM Genome sequence and comparative microarray analysis of serotype M18 group A Streptococcus strains associated with acute rheumatic fever outbreaks Proc Natl Acad Sci USA 2002 99 4668 4673 11917108 10.1073/pnas.062526099
Beres SB Sylva GL Barbian KD Lei B Hoff JS Mammarella ND Liu M Smoot JC Porcella SF Parkins LD McCormick JK Leung DYM Schlievert PM Musser JM Genome sequence of a serotype M3 strain of group A Streptococcus : phage-encoded toxins, the high-virulence phenotype, and clone emergence Proc Natl Acad Sci USA 2002 99 10078 10083 12122206 10.1073/pnas.152298499
Smoot LM Smoot JC Graham MR Somerville GA Sturdevant DE Migliaccio CA Sylva GL Musser JM Global differential gene expression in response to growth temperature alteration in group A Streptococcus Proc Natl Acad Sci USA 2001 98 10416 10421 11517341 10.1073/pnas.191267598
Szarapinska-Kwaszewska J Mikucki J Utilization of exogenous siderophores by enterococci Med Dosw Mikrobiol 2001 53 233 243 11785180
Viswanathan VK Edelstein PH Pope CD Cianciotto NP The Legionella pneumophila iraAB locus is required for iron assimilation, intracellular infection, and virulence Infect Immun 2000 68 1069 1079 10678909 10.1128/IAI.68.3.1069-1079.2000
Velayudhan J Hughes NJ McColm AA Bagshaw J Clayton CL Andrews SC Kelly DJ Iron acquisition and virulence in Helicobacter pylori : a major role for FeoB, a high affinity ferrous iron transporter Mol Microbiol 2000 37 274 286 10931324 10.1046/j.1365-2958.2000.01987.x
Takase H Nitanai H Hoshino K Otani T Impact of siderophore production on Pseudomonas aeruginosa infections in immunosuppressed mice Infect Immun 2000 68 1834 1839 10722571 10.1128/IAI.68.4.1834-1839.2000
Janakiraman A Slauch JM The putative iron transporter system SitABCD encoded on SPI1 is required for full virulence of Salmonella typhimurium Mol Microbiol 2000 35 1146 1155 10712695 10.1046/j.1365-2958.2000.01783.x
Brown JS Gilliland SM Holden DW A Streptococcus pneumoniae pathogenicity island encoding an ABC transporter involved in iron uptake and virulence Mol Microbiol 2001 40 572 585 11359564 10.1046/j.1365-2958.2001.02414.x
Rokbi B Mignon M Maitre-Wilmotte G Lissolo L Danve B Caugant DA Quentin-Millet M-J Evaluation of recombinant transferrin-binding protein B variants from Neisseria meningitidis for their ability to induce cross-reactive and bactericidal antibodies against a genetically diverse collection of serogroup B strains Infect Immun 1997 65 55 63 8975892
Myers LE Yang Y-P Du R-P Wang Q Harkness RE Schryvers AB Myers M Klein H Loosmore SM The transferrin binding protein B of Moraxella catarrhalis elicits bactericidal antibodies and is a potential vaccine antigen Infect Immun 1998 66 4183 4192 9712766
Potter AA Schryvers AB Ogunnariwo JA Hutchins WA Lo RYC Watts T Protective capacity of the Pasteurella haemolytica transferrin-binding proteins TbpA and TbpB in cattle Microb Pathog 1999 27 197 206 10502460 10.1006/mpat.1999.0297
Brown JS Ogunniyi AD Woodrow MC Holden DW Paton JC Immunization with components of two iron uptake ABC transporters protects mice against systemic Streptococcus pneumoniae infection Infect Immun 2001 69 6702 6706 11598041 10.1128/IAI.69.11.6702-6706.2001
Kuboniwa M Amano A Shizukuishi S Nakagawa I Hamada S Specific antibodies to Porphyromonas gingivalis Lys-Gingipain by DNA vaccination inhibit bacterial binding to hemoglobin and protect mice from infection Infect Immun 2001 69 2972 2979 11292714 10.1128/IAI.69.5.2972-2979.2001
West D Reddin K Matheson M Heath R Funnell S Hudson M Robinson A Gorringe A Recombinant Neisseria meningitidis transferrin binding protein A protects against experimental meningococcal infection Infect Immun 2001 69 1561 1567 11179327 10.1128/IAI.69.3.1561-1567.2001
Lei B Liu M Chesney G Musser JM Extracellular putative lipoproteins made by Streptococcus pyogenes : Identification of new potential vaccine candidate antigens J Infect Dis 2004 189 79 89 14702157
Hoe N Nakashima PK Grigsby D Pan X Dou SJ Naidich S Garcia M Kahn E Bergmire-Sweat D Musser JM Rapid molecular genetic subtyping of serotype M1 group A Streptococcus strains Emerg Infect Dis 1999 5 254 263 10221878
LeBlanc DJ Lee LN Inamine JM Cloning and nucleotide base sequence analysis of a spectinomycin adenyltransferase AAD(9) determinant from Enterococcus faecalis Antimicrob Agents Chemother 1991 35 1804 1810 1659306
Podbielski A Spellerberg B Woischnik M Pohl B Lutticken R Novel series of plasmid vectors for gene inactivation and expression analysis in group A streptococci (GAS) Gene 1996 177 137 147 8921859 10.1016/0378-1119(96)84178-3
McIver KS Scott JR Role of mga in growth phase regulation of virulence genes of the group A streptococcus J Bacteriol 1997 179 5178 5187 9260962
Lei B Mackie S Lukomski S Musser JM Identification and immunogenicity of Group A Streptococcus culture supernatant proteins Infect Immun 2000 68 6807 6818 11083799 10.1128/IAI.68.12.6807-6818.2000
|
16225685
|
PMC1276799
|
CC BY
|
2021-01-04 16:03:40
|
no
|
BMC Microbiol. 2005 Oct 14; 5:62
|
utf-8
|
BMC Microbiol
| 2,005 |
10.1186/1471-2180-5-62
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-621622568510.1186/1471-2180-5-62Research ArticleABC transporter FtsABCD of Streptococcus pyogenes mediates uptake of ferric ferrichrome Hanks Tracey S [email protected] Mengyao [email protected] Michael J [email protected] Benfang [email protected] Department of Veterinary Molecular Biology, Montana State University, Bozeman, Montana 59717, USA2005 14 10 2005 5 62 62 10 9 2005 14 10 2005 Copyright © 2005 Hanks et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The Streptococcus pyogenes or Group A Streptococcus (GAS) genome encodes three ABC transporters, namely, FtsABCD, MtsABC, and HtsABC, which share homology with iron transporters. MtsABC and HtsABC are believed to take up ferric (Fe3+) and manganese ions and heme, respectively, while the specificity of FtsABCD is unknown.
Results
Recombinant FtsB, the lipoprotein component of FtsABCD, was found to bind Fe3+ ferrichrome in a 1:1 stoichiometry. To investigate whether FtsABCD transports Fe3+ ferrichrome, GAS isogenic strains defective in lipoprotein gene ftsB and permease gene ftsC were generated, and the effects of the mutations on uptake of Fe3+ ferrichrome were examined using radioactive 55Fe3+ ferrichrome. FtsB was produced in the wild-type strain but not in the ftsB mutant, confirming the ftsB inactivation. While wild-type GAS took up 3.6 × 104 Fe3+ ferrichrome molecules per bacterium per min at room temperature, the ftsB and ftsC mutants did not have a detectable rate of Fe3+ ferrichrome uptake. The inactivation of ftsB or ftsC also decreased 55Fe3+ ferrichrome uptake by >90% under growth conditions in the case of limited uptake time. Complementation of the ftsB mutant with a plasmid carrying the ftsB gene restored FtsB production and 55Fe3+ ferrichrome association at higher levels compared with the parent strain. The inactivation of mtsA and htsA and Fe-restricted conditions enhanced the production of FtsB and Fe3+ ferrichrome uptake.
Conclusion
The FtsB protein bound Fe3+ ferrichrome, and inactivation of ftsB or ftsC, but not htsA or mtsA, diminished Fe3+ ferrichrome uptake, indicating that FtsABCD, but not HtsABC and MtsABC, is the transporter that takes up Fe3+ ferrichrome in GAS. Fe acquisition systems are virulence factors in many bacterial pathogens and are attractive vaccine candidates. The elucidation of the FtsABCD specificity advances the understanding of Fe acquisition processes in GAS and may help evaluating the GAS Fe acquisition systems as vaccine candidates.
==== Body
Background
Ferric iron (Fe3+), the stable iron form in an oxidative environment, has extremely low solubility in water under physiological conditions, and mammalian hosts thus do not have sufficient free Fe3+ to support bacterial growth [1]. The major sources of iron in vivo for bacteria are host heme-proteins and other iron complexes [1,2]. Many bacterial pathogens secrete low-molecular-weight iron chelators called siderophores to assimilate iron from host environments [3]. Ferrisiderophores formed are then transported across the cytoplasmic membrane by specific ATP-binding cassette (ABC) type transporters. ABC transporters consist of a solute-binding protein, a membrane protein (permease) encoded by one or two genes, and an ATPase [4]. The solute-binding proteins are located in the periplasmic space in Gram-negative bacteria and are lipoproteins in Gram-positive organisms. Siderophores can be divided into several types based on chemical structures [3]. Ferrichrome [5] belongs to the hydroxamate type.
Streptococcus pyogenes or Group A Streptococcus (GAS) is an important Gram-positive human pathogen causing both invasive and non-invasive infections [6]. Non-invasive infections, including pharyngitis, and post-infection sequelae, such as acute rheumatic fever, rheumatic heart disease, and glomerulonephritis, result in substantial morbidity and economic loss globally. Invasive GAS infections, such as necrotizing fasciitis and streptococcal toxic shock syndrome, are associated with high mortality rates. GAS can take up heme from hemoglobin and haptoglobin-hemoglobin complexes [7]. Exogenously-supplied heme and host heme proteins (hemoglobin, myoglobin, and catalase), but not iron-loaded transferrin and lactoferrin, support in vitro growth of GAS under iron-restricted conditions [8].
Iron acquisition processes in GAS are poorly understood, although progress has recently been made by us [9-11] and other groups [12-14]. GAS genomes [15-17] encode three ABC transporters, namely, HtsABC [9-11] or SiaABC [12], MtsABC [13,14], and one encoded by spy0383 to spy0386 [18] (designated FtsABCD), which are homologues of ABC transporters involved in iron acquisition. HtsABC and the cell-surface protein Shp [9-12] are believed to make up the machinery for heme acquisition. The lipoprotein component MtsA of MtsABC binds Fe3+, Zn2+, and Cu2+ [13], and MtsABC is important for acquisition of Mn2+ and Fe3+ [14]. The transcription of ftsABCD is up-regulated under iron-restricted conditions [18]. However, recombinant FtsB was heme-free and did not contain Fe, Mn, or Zn [10]. The specificity of FtsABCD is thus not known. It is not known whether GAS can use ferrisiderophores such as ferric ferrichrome (Fe3+ ferrichrome) as an iron source.
We found that recombinant FtsB bound Fe3+ ferrichrome, suggesting that FtsABCD is involved in the acquisition of Fe3+ ferrichrome. To test this hypothesis, GAS isogenic strains defective in lipoprotein gene ftsB and permease gene ftsC were generated, and the ftsB and ftsC inactivation dramatically decreased the uptake of Fe3+ ferrichrome, indicating that FtsABCD is the transporter that takes up Fe3+ ferrichrome in GAS.
Results
Binding of Fe3+ ferrichrome by FtsB
The specificity of an ABC transporter is believed to depend on the binding specificity of its solute binding protein. Purified recombinant FtsB did not contain free metal ions, suggesting that FtsABCD does not target free metal ions [10]. To test whether FtsABCD targets ferrisiderophores, the ability of FtsB to bind Fe3+ ferrichrome was examined by gel filtration. FtsB was incubated with excess Fe3+ ferrichrome and separated from free Fe3+ ferrichrome by chromatography on a Sephadex G-25 column. Fe3+ ferrichrome has an absorption peak at 425 nm and its elution profile was monitored by measuring A425 of each fraction. The elution profile displayed two A425 peaks (Fig. 1A). One co-migrated with the FtsB peak, which was localized by A280, and the other corresponded to free Fe3+ ferrichrome (Fig. 1A). When FtsB was replaced with MtsA in a control experiment, there was only one A425 peak corresponding to free Fe3+ ferrichrome, and no absorbance at 425 nm was associated with the protein peak (Fig. 1B). To check whether the species associated with FtsB was Fe3+ ferrichrome, the absorption spectra of the FtsB and MtsA peaks and free Fe3+ ferrichrome were compared. In addition to the protein absorption peak at A280, the FtsB sample had an absorption peak at 425 nm which was identical to that of free Fe3+ ferrichrome (Fig. 1C). As expected, the MtsA sample only had the protein peak (Fig. 1C). These results indicate that FtsB, but not MtsA, bound Fe3+ ferrichrome. The FtsB sample was found to have 1.08 Fe3+ ferrichrome per FtsB molecule on the basis of protein content and extinction coefficient of Fe3+ ferrichrome at 425 nm, indicating a 1:1 binding stoichiometry.
If FtsB binds Fe3+ ferrichrome, Fe3+ should co-migrate with the protein on the G-25 column. To test this idea, Fe3+ ferrichrome containing 1.1% 55Fe3+ of total Fe3+ was used to repeat the gel filtration experiment, and 55Fe3+ radioactivity was monitored. As expected, one of two 55Fe3+ peaks co-migrated with the FtsB peak, and the other peak corresponded to free Fe3+ ferrichrome (Fig. 1D). On the basis of 55Fe3+ percentage of total iron and specific activity and protein content, the FtsB peak fraction contained 37 μM FtsB and 36 μM Fe, consistent with the 1:1 binding stoichiometry determined above. These results confirmed that FtsB binds Fe3+ ferrichrome in a 1:1 molar ratio.
Effects of ftsB, ftsC, mtsA, and htsA inactivation on Fe3+ ferrichrome uptake
The binding results described above suggest that FtsABCD can acquire Fe3+ ferrichrome. To test this hypothesis, ftsB, ftsC, and mtsA (control) were first inactivated by insertional inactivation (Fig. 2). Inactivation was confirmed by PCR (Fig. 2C) and DNA sequencing. Western blotting analysis detected FtsB (Fig. 2D) and MtsA (Fig. 2E) in the wild-type strain but not in the corresponding mutant strain. These results indicate that ftsB and mtsA were indeed inactivated. The ftsB, ftsC, mtsA, and htsA mutants (construction of the htsA mutant will be described elsewhere) and parent strains were compared in the uptake of 55Fe3+ ferrichrome. Pilot experiments indicated that the wild-type strain could take up 55Fe3+ ferrichrome at both room temperature (24°C) and 37°C. The uptake was thus performed at room temperature for convenience. GAS cells harvested from exponential growth phase were incubated with 0.16 μM 55Fe3+ ferrichrome at room temperature for 1 h, and 55Fe3+ radioactivity associated with bacteria was determined. The ftsB and ftsC inactivation diminished the uptake of Fe3+ ferrichrome (Fig. 3), while the mtsA and htsA inactivation did not abolish but enhanced the uptake. To further examine the uptake of Fe3+ ferrichrome, 55Fe3+ radioactivity taken up by bacteria as a function of incubation time was determined. 55Fe3+ radioactivity with wild-type GAS increased linearly with time up to 30 min with a slope of 153 cpm/min (Fig. 4). The value of the slope could be translated into an uptake rate of 3.6 × 104 Fe3+ ferrichrome molecules per min per wild-type GAS bacterium. In contrast, the ftsB and ftsC mutants did not have a detectable rate of Fe3+ ferrichrome uptake (Fig. 4). The rate of uptake in the mtsA mutant was 2.6 times as that in the parent strain (Fig. 4). These results indicate that FtsABCD, not MtsABC and HtsABC, mediates the uptake of Fe3+ ferrichrome.
Complementation
To test whether the effect of the ftsB inactivation on the uptake of Fe3+ ferrichrome was due to the absence of FtsB, a complementing strain,ftsB mutant/pCMVftsB, was constructed to carry plasmid pCMVftsB with the ftsB gene. The intensity of FtsB band of the complemented strain in Western blotting analysis in Fig. 2D was 7/10 of that of the wild-type strain, and the amount of the complemented strain cells used was 1/125 of that of the wild-type cells, indicating that FtsB was produced in ftsB mutant/pCMVftsB at a level as 88 times as that in the wild-type strain. Wild-type, ftsB mutant, and ftsB mutant/pCMVftsB cells harvested in the exponential growth phase were incubated with 0.16 μM 55Fe3+ ferrichrome at 24°C for 1 h, and 55Fe3+ radioactivity associated with the bacteria was determined. 55Fe3+ radioactivity of ftsB mutant/pCMVftsB was 1.9 and 207 times higher than those of the wild-type and ftsB mutant cells, respectively (Fig. 5), suggesting that in trans expression of ftsB in the ftsB mutant restored the association of Fe3+ ferrichrome. The results support that FtsABCD targets Fe3+ ferrichrome.
Effect of Fe3+ ferrichrome on GAS growth
Since FtsABCD is involved in the uptake of Fe3+ ferrichrome, Fe3+ ferrichrome should be an iron source of GAS. This idea was tested by comparing GAS growth curves in THY treated with Chelex 100 to remove metal ions and supplemented with MgCl2 (DTHYMg) in the absence and presence of Fe3+ ferrichrome. The growth curve in the presence of 10 μM Fe3+ ferrichrome shifted to the left by about 40 min compared with that in the absence of Fe3+ ferrichrome under otherwise identical conditions. This small but repeatable stimulatory effect suggests that GAS can use Fe3+ ferrichrome as an iron source. Fe3+ ferrichrome had similar stimulatory effect on the growth of the mutants as that on the growth of the wild type strain, suggesting that either residual uptake of Fe3+ ferrichrome in the ftsB and ftsC mutants were enough to induce the stimulatory effect or an additional process was involved in the uptake of the iron extracted from Fe3+ ferrichrome under the growth conditions.
Effects of ftsB and ftsC inactivations on the uptake of Fe3+ ferrichrome under growth conditions
To examine why the stimulatory growth effect of Fe3+ ferrichrome was still observed in the ftsB and ftsC mutants, 55Fe3+ ferrichrome uptake was examined under growth conditions. 55Fe3+ ferrichrome was added into the cultures of wild-type, ftsB, and ftsC strains at mid-exponential growth phase, and, 30 min later, 55Fe3+ activity associated with bacteria was determined. 55Fe3+ radioactivity of ftsB and ftsC mutant cells were only 7.5% and 5.3% of that of wild-type cells, respectively (Fig. 6A), indicating that the ftsB and ftsC inactivation also dramatically diminished uptake of Fe3+ ferrichrome under growth conditions. However, 55Fe3+ radioactivity of ftsB and ftsC cells grown for 3 h after the addition of 55Fe3+ ferrichrome were about 50% of that of wild-type cells (Fig. 6B). These results suggest that there could be an additional acquisition process which could assimilate 55Fe3+ from its ferrichrome complex.
Effects of mtsA and htsA inactivation and Fe-restricted conditions on Fe3+ ferrichrome uptake and FtsB production
To further examine the factors to affect Fe3+ ferrichrome uptake, the effects of mtsA and htsA inactivation and Fe-restricted conditions on FtsB production and Fe3+ ferrichrome uptake were investigated. Inactivation of htsA and mtsA increased 55Fe3+ ferrichrome uptake by more than 100% (Fig. 7A). Wild-type, htsA, and htsA mutant cells harvested from THY containing 2,2'-dipyridyl took up 30%–50% more 55Fe3+ ferrichrome than those from THY without 2,2'-dipyridyl (Fig. 7A). Western blotting analysis indicated higher levels of FtsB in the htsA and mtsA mutants compared with the wild-type strain (Fig. 7B). Higher levels of FtsB were also detected in all strains grown in the presence of 2,2'-dipyridyl compared with those grown in the absence of 2,2'-dipyridyl (Fig. 7B). A control protein, phosphoglycerate kinase (PGK), had similar expression levels under all the conditions (Fig. 7B). The relative intensities of the FtsB bands in the Western blot correlated well with the relative uptake of 55Fe3+ ferrichrome under the same conditions. These results indicate that FtsB production was enhanced in the htsA and mtsB mutants and under Fe-restricted conditions, further supporting the role of FtsABCD in Fe acquisition.
Discussion
Evidences from this study indicate that FtsABCD is the transporter that takes up Fe3+ ferrichrome in GAS. The evidences include the binding of Fe3+ ferrichrome to FtsB, the effect of the insertional disruption of ftsB or ftsC on uptake of 55Fe3+ ferrichrome, and the non-involvement of MtsABC and HtsABC in Fe3+ ferrichrome uptake. Complementation data of the ftsB mutant with the ftsB gene expressed in trans indicated that the effect of ftsB inactivation was due to the lack of FtsB. MtsABC [13,14] and HtsABC [9-12] target free Fe3+ and heme, respectively. The specificity of the FtsABCD transporter elucidated in this study resolved the last piece of the puzzle regarding the roles of ABC transporters in Fe acquisition in GAS.
Uptake time was critical to the effects of ftsB and ftsC inactivation on 55Fe3+ activity associated with bacteria under growth conditions. The inactivation had a dramatic effect (>90% decrease in 55Fe3+ activity compared with wild-type strain) when uptake was performed for only 30 min. This decrease was reduced to about 50% when uptake was performed for 3 h. Although the reasons for this reduction are not known, the reduction is unlikely due to the existence of another transporter for Fe3+ ferrichrome. The other uptake results are not consistent with the existence of another Fe3+ ferrichrome transporter. Another possible reason is that 55Fe3+ ferrichrome exchanged its 55Fe3+ with another ferric complex in THY, resulting in a non-ferrichrome 55Fe3+ complex that could be taken up by another transporter. GAS has a putative secreted Fe binding protein (Spy1063). It is not known whether this protein can extract Fe3+ from Fe3+ ferrichrome.
The transcription of ftsABCD is up-regulated under Fe-reduced conditions [18]. Consistent with this observation, Fe-restricted conditions enhanced the production of FtsB and uptake of Fe3+ ferrichrome, further supporting the role of FtsABCD in Fe acquisition. Inactivation of htsA and mtsA also enhanced FtsB production and Fe3+ ferrichrome uptake. Apparently, all the three ABC transporters contributed to Fe3+ acquisition in GAS grown in THY, and inactivation of either mtsA or htsA might result in lower intracellular Fe levels and, in turn, enhanced the expression of ftsABCD. The results also suggest that the expression of ftsABCD, htsABC, and mtsABC is coordinately regulated.
Some Enterococcus faecium clinical strains do not produce siderophores but can acquire iron using exogenous siderophores produced by other bacteria living in the same habitats [19]. GAS is not known to produce siderophores, and GAS genomes [15-17] do not have genes encoding homologues of siderphore-production systems. We could not detect siderophore production in GAS under iron-restricted conditions. Therefore, GAS may not produce siderophors. However, GAS takes up Fe3+ ferrichrome, suggesting that GAS could acquire Fe3+ by using siderophores produced by other bacteria in the pharynx and skin, the noninvasive GAS infection sites.
Fe acquisition systems are virulence factors in many bacterial pathogens [20-24] and are attractive vaccine targets [25-30]. Elucidation of the specificities of the Fe transporters in GAS will facilitate determination of their relative importance in various infections and choose appropriate animal infection models to evaluate their efficacy as vaccine candidates. For examples, HtsABC could be more important in invasive GAS infection since heme should be the Fe source, and FtsABCD could be important in non-invasive infections because exogenous ferric siderophore complexes should be available.
In summary, we found that FtsB bound Fe3+ ferrichrome and that ftsB or ftsC inactivation dramatically decreased Fe3+ ferrichrome uptake. The results indicate that FtsABCD is the transporter that acquires Fe3+ ferrichrome in GAS.
Methods
Materials
Iron chelating agent 2,2'-dipyridyl was obtained from Aldrich. Sephadex G-25, iron-free ferrichrome A from Ustilago sphaerogena, Chelex 100, and other chemicals were purchased from Sigma (St. Louis, MO). 55Fe3+-labeled ferric chloride was purchased from RI Consultants LLC (Hudson, NH). Purified recombinant FtsB (Spy0385) and MtsA were prepared as described previously [10,31].
Mouse antisera
Five female outbred CD-1 Swiss mice (4- to 6-week-old) (Charles River Laboratories, Wilmington, MA) were immunized subcutaneously with 50 μg of recombinant FtsB, MtsA, or PGK suspended in 200 μL of saline emulsified in 44 μL of monophosphoryl lipid A-synthetic trehalose dicorynomycolate adjuvant (Corixa, Hamilton, MT). Mice were boosted at weeks 2 and 4. Immune sera were collected 5 days after the second boost.
Bacterial strains and growth
Serotype M1 GAS strain MGAS5005 has been described previously [32]. MGAS5005 and its isogenic mutants were grown routinely at 37°C in 5% CO2 in Todd-Hewitt broth (Difco Laboratories, Detroit, MI) supplemented with 0.2% yeast extract (THY). Spectinomycin (150 mg/L) was added into THY for mutant strains. Iron-restricted conditions were achieved by adding 0.3 mM 2,2'-dipyridyl into THY and by treating THY with the chelating resin Chelex 100 and supplementing it with 0.4 mM MgCl2 (DTHYMg). Tryptose agar with 5% sheep blood (Becton Dickinson, Cockeysville, MD) and THY agar were used as solid media.
Binding of Fe3+ ferrichrome to FtsB
Gel filtration was used to detect the binding of Fe3+ ferrichrome to FtsB. FtsB (0.3 ml of 0.18 mM) was incubated with 0.9 mM Fe3+ ferrichrome for 20 min at room temperature, loaded onto a Sephadex G-25 column (1.5 × 18 cm), and eluted with 20 mM Tris-HCl buffer, pH 8.0. Eluant was collected as fractions of 0.7 ml. The absorbance of each fraction attributable to protein and Fe3+ ferrichrome was measured. MtsA as a negative control was similarly tested for the binding of Fe3+ ferrichrome. The experiment was repeated with Fe3+ ferrichrome in which 1.1% of total Fe3+ was 55Fe, and A280 and 55Fe3+ radioactivity of each fraction were measured. 55Fe3+ radioactivity was measured using a window of 0–6 keV with a Packard 1500 Tri-Carb Liquid Scintillation Analyzer.
GAS isogenic mutants
MGAS5005 isogenic mutants defective in ftsC, ftsB, or mtsA were generated by insertional inactivation (Fig. 2). An internal ~400-bp fragment of each gene was PCR-amplified using MGAS5005 genomic DNA as template and the primers listed in Table 1. The PCR products were digested with NcoI and ligated to pFWaad at the NcoI site to yield suicide plasmids. The orientation of the fragments was determined by DNA sequencing, and the suicide plasmids in which the orientation of the fragments was opposite to that of the aad gene were chosen for generating the mutants. To obtain pFWaad, the spectinomycin-resistant gene aad [33] was amplified with primers 5'-AGTGTCGACTATAACTAATAACGTAACGTG-3' and 5'-ACCATGGGAATTCTATAATTTTTTTAATCTGTTATTTA-3' and cloned into pFW14 [34] at the SalI and NcoI restriction sites.
Each suicide plasmid was introduced into MGAS5005 by electroporation at 1.8 kV and 400 Ω. One ml of THY was added into the sample immediately after electroporation. The sample was incubated at 37°C for 2 h and plated on THY agar plates supplemented with 150 mg of spectinomycin per liter to select insertional mutants. The plates were incubated in 5% CO2 at 37°C for two days, and the colonies obtained were screened by PCR analysis using the primers listed in Table 1. Gene interruptions were then confirmed by sequencing the PCR products.
Construction of ftsB-complementing plasmid pCMVftsB
Plasmid pJRS525 [35] was modified to replace the spectinomycin-resistant gene with the chloramphenicol-resistance gene. A 2545-bp fragment without the spectinomycin-resistant gene was amplified from pJRS525 with primers 5'-TCGTGGATCC AAGCTTCACCATGG-3' and 5'-AAGAATTCCTTGCATAGACTTTCGTCAG-3'. The fragment containing the chloramphenicol-resistant gene was amplified from pFW14 [23] with primers 5'-GGAATTCCGGATGCATATGCATG-3' and 5'-GTCCGGATCCTCGAGCTCTAGATC-3'. The PCR products were digested with BamHI and EcoRI and ligated together to yield plasmid pCMV. A DNA fragment containing ftsB and its ribosome-binding site was amplified from MGAS5005 using primers 5'-CGGATCCAATAACTTTATTCTAGGAGAATTAG-3' and 5'-AGGGATCCTTAGTTTTCACTTGATAAGATTG-3'. The PCR product was digested with BamHI and ligated into pCMV at the BamHI site, yielding pCMVftsB containing ftsB. The cloned gene was sequenced to rule out spurious mutations and confirm the desired orientation. The resulting pCMVftsB was introduced into the ftsB mutant by electroporation, and the complement strain (designated ftsB/pCMVftsB) was selected by spectinomycin and chloramphenicol and confirmed for the existence of pCMVftsB by colony PCR using primers 5'-CAATTTCACACAGGAAACAGC-3' (pCMV-specific) and 5'-AGGGATCCTTAGTTTTCACTTGATAAGATTG-3' (ftsB-specific)
Uptake of Fe3+ ferrichrome by GAS
Fe3+ ferrichrome uptake by GAS was monitored using radioactive 55Fe3+ ferrichrome under non-growth and growth conditions. Under the non-growth conditions, wild-type MGAS5005 and its ftsB, ftsC, or mtsA mutant strains were harvested from the mid-exponential growth phase (OD600 of 0.4) by centrifugation. The bacterial pellets were washed with 10 ml of THY and resuspended in 1 ml of THY. To prepare 55Fe3+ ferrichrome working solution, 4.7 nmole 55FeCl3 was incubated with 14.1 nmole Fe-free ferrichrome A in 50 μL of Tris-HCl for 10 min, and the complex was diluted with 22 ml of THY. The 1-ml bacterial suspensions were mixed with 3 ml of the 55Fe3+ ferrichrome solution (final Fe3+ ferrichrome concentration 0.16 μM) to initiate the uptake process and rotated in a 10-ml tube from end to end at room temperature. A triplet of 0.2 ml samples were taken from each mixture at the indicated times, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. For uptake under growth conditions, 0.16 μM 55Fe3+ ferrichrome was added into the cultures of wild-type, ftsB, and ftsC strains at early exponential or mid-exponential growth phase. A triplet of 1-ml culture samples were taken at the indicated times after the 55Fe3+ ferrichrome addition and treated as described above. The pellets obtained were resuspended in 0.2 ml of THY and mixed with 3 ml of scintillation liquid. 55Fe3+ radioactivity associated with the bacteria was measured as described above.
FtsB and MtsA production
Production of FtsB or MtsA in wild-type and mutant strains was monitored by Western blotting analysis. To prepare samples for the analysis, bacteria were harvested from cultures at the indicated volumes in exponential phase. The bacterial pellets were washed twice with 1.3 ml of PBS, resuspended in 100 μl of PBS, and treated with 200 units of mutanolysin at 37°C for 2 h. The samples were briefly sonicated and mixed with equal volume of 2x SDS-PAGE loading buffer. Proteins in 10 μl of the samples were resolved by SDS-PAGE and transferred to nitrocellulose membranes. The proteins were detected by Western blotting using specific mouse antiserum as previously described [36].
Other procedures and measurements
Chromosomal DNA was isolated with the Puregene DNA Isolation kit (Gentra Systems, Minneapolis, MN). Sequence data were obtained with an ABI 310 automated DNA sequencer (Applied Biosystems, Inc., Foster City, CA). Absorbance and optical spectra were obtained with a SPECTRAmax 384 Plus spectrophotometer (Molecular Devices, Sunnyvale, CA). Protein concentrations were determined with the modified Lowry protein assay kit purchased from Pierce (Rockford, IL) with bovine serum albumin as a standard. Intensities of the bands in Western blotting analysis were determined using an AlphaImager 2000 Documentation & Analysis System (Alpha Innotech Corp.).
Authors' contributions
TSH carried out the ferric ferrichrome binding assay and the characterization of the GAS isogenic mutants and participated in writing the Methods section. ML generated the GAS isogenic mutants. MJM constructed the plasmid for complementing the ftsB mutant. BL designed the study and drafted the manuscript.
Acknowledgements
This work was supported in part by grants P20 RR-16455 and P20 RR-020185 from the National Center for Research Resources, K22AI057347 from National Institutes of Health, and the Montana State University Agricultural Experimental Station. We thank Dr. Kevin McIver for providing plasmid pJRS525.
Figures and Tables
Figure 1 Binding of Fe3+ ferrichrome to FtsB. (A) Co-migration of Fe3+ ferrichrome with FtsB. FtsB (0.3 ml of 0.18 mM) was incubated with 0.9 mM Fe3+ ferrichrome for 20 min at room temperature, loaded onto a Sephadex G-25 column (1.5 × 18 cm), and eluted with 20 mM Tris-HCl buffer, pH 8.0. Eluant was collected as fractions of 0.7 ml. A280 and A425 of each fraction were used to monitor FtsA and Fe3+ ferrichrome, respectively. (B) MtsA, as a control, did not bind Fe3+ ferrichrome. The experiment was performed under the conditions as those in panel A. (C) The absorption spectra of the protein peaks (fraction 9) in panel A (solid curve) and panel B (short dashed curve) and free Fe3+ ferrichrome (long dashed curve). The three samples contained 46 μM FtsB, 45 μM MtsA, and 100 μM Fe3+ ferrichrome, respectively. (D) Co-migration of partially 55Fe-labeled Fe3+ ferrichrome with FtsB. Fe3+ ferrichrome contained radioactive 55Fe3+ at 1.1% of total Fe3+. A280 and 55Fe3+ radioactivity of each fraction are presented.
Figure 2 Insertional inactivation of ftsB, ftsC, and mtsA. (A) Schematic showing the arrangement of ftsABCD genes and their neighbors in the MGAS5005 genome. The ftsA, ftsB, and ftsCD genes encode ATP-binding protein, lipoprotein, and permease, respectively. The numbers above the other arrows are the spy numbers assigned to the corresponding open reading frames in the M1 genome sequence (15). (B) Schematic for insertional gene inactivation. The paired solid or dotted arrows under the mutant genome indicate the locations of primers used to confirm the disruption of the gene using PCR and DNA sequencing. (C) PCR confirmation of insertional inactivation of ftsB, ftsC, and mtsA. The picture shows agarose gel analysis of PCR reactions using mutant (lanes labeled by 1) or wild type (lanes labeled by 2) genomic DNA as template and primers at the locations indicated by the solid arrows under the mutant genome in panel B. (D) Western blot showing the absence of FtsB in the ftsB mutant strain (lane 2) and the presence of FtsB in the wild-type (lane 1) and ftsB/pCMVftsB (lane 3) strains. Proteins from 5 × 108 wild-type cells, 5 × 108 ftsB mutant cells, and 4 × 106 ftsB/pCMVftsB were probed with FtsB-specific mouse antiserum. The amount of ftsB/pCMVftsB cells used was 1/125 of that of the wild-type cells. (E) Western blot showing the presence of MtsA in the wild-type strain (lane1) and the absence of MtsA in the mtsA mutant strain (lane 2). Proteins from 6 × 106 wild-type cells and 3 × 108 mtsA mutant cells were probed with MtsA-specific mouse antiserum.
Figure 3 Effects of ftsB, ftsC, mtsA, and htsA inactivation on 55Fe3+ ferrichrome uptake. Wild-type and mutant strains were grown to OD600 of 0.4 in 10 ml of THY and THY/spectinomycin, respectively. The bacterial pellets were washed with 1.0 ml of THY and incubated with 0.16 μM 55Fe3+ ferrichrome 1.0 ml of THY at room temperature for 1 h. A triplet of 0.2 ml samples were taken from each mixture, and bacteria were pelleted and washed twice with 0.4 ml of THY. 55Fe3+ radioactivity associated with the bacteria was measured. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria in a representative of three experiments.
Figure 4 The time course of 55Fe3+ ferrichrome uptake for mtsA (open circles), ftsB (open triangles), and ftsC (solid triangles) mutant and wild-type (solid circles) GAS strains. Each strain harvested at the mid-exponential growth phase from 40 ml culture, washed with 10 ml of THY, and incubated with 0.16 μM 55Fe3+ ferrichrome in 4 ml of THY at room temperature. A triplet of 0.2 ml samples were taken from each mixture at the indicated times, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria in a representative of three experiments.
Figure 5 Complementation of the ftsB mutant. Plasmid pCMVftsB carrying the ftsB gene was introduced into the ftsB mutant strain to obtain ftsB/pCMVftsB. The uptake experiment was performed exactly as in Fig. 3. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria.
Figure 6 Effects of ftsB and ftsC inactivation on 55Fe3+ ferrichrome uptake under growth conditions. (A) Uptake for 30 min. 55Fe3+ ferrichrome (0.16 μM) was added into cultures of the wild-type and mutant strains at 37°C in THY and THY supplemented with spectinomycin, respectively, when OD600 was 0.3. A triplet of 1 ml samples were taken from each culture 30 min later, and bacteria were immediately pelleted and washed twice with 0.4 ml of THY. Presented are the mean values ± SD of 55Fe3+ radioactivity associated with the bacteria. (B) Uptake for 3 h. 55Fe3+ ferrichrome (0.16 μM) was added into the wild-type and mutant cultures at 37°C when OD600 was 0.2, and, 3 h later, the samples were processed as in panel A for determination of 55Fe3+ radioactivity.
Figure 7 Effects of Fe-restricted conditions and mtsA and htsAinactivation on Fe3+ ferrichrome uptake and FtsB production. (A). 55Fe3+ radioactivity uptake by wild-type (wt), ftsB, ftsC, mtsA, and htsA mutant strains grown in THY in the absence (solid bars) or presence (slashed bars) of 0.3 mM 2,2'-dipyridyl. The experiment was performed as in Fig. 3. (B) Western blots showing the relative levels of FtsB and PGK as a control in the strains grown under the same conditions as in panel A. Proteins from 1 × 108 bacteria of each strain were probed with mouse anti-FtsB and anti-PGK antisera. The relative levels of FtsB and PGK were obtained by dividing the intensity of each band by the intensity of the FtsB or PGK band of the wild-type strain.
Table 1 Primers used in gene inactivation and confirmation
Primer Sequence (5'-3') Purpose
1 ACCATGGCTATGCCTCAAATTGCTGG ftsB inactivation
2 ACCATGGTTTTCCAATCTTTTAACCAC
3 ACCATGGAGCTTATGTGTTGCTATTTAC ftsC inactivation
4 ACCATGGATAGCTCCTGCTAAGACAAG
5 ACCATGGTCTTTTAGTAGCTTGTTCGTC mtsA inactivation
6 ACCATGGTTCATTTCGACAAGAGGCTG
7 AACTCTACTATTAACACTCTG confirmation of ftsB inactivation
8 GCAACGACGACATAATCACC
9 GTGTCTCGAGCGTTTTTGC confirmation of ftsC inactivation
10 CAATATAGCCGATCATTTG
11 ATGAGCCTCATTTTGGGTGC confirmation of mtsA inactivation
12 GCCACATAAGCTTTTAGGTTC
13 AGAATTTTGTTAGCAGTTCG aad-specific, paired with primer 7, 9, or 11 for inactivation confirmation
14 CTTTGAGTGAGCTGATACCG pFWaad-specific, paired with primer 8, 10, or 12 for inactivation confirmation
==== Refs
Weinberg ED Cellular iron metabolism in health and disease Drug Metab Health Dis 1990 22 532 579
Otto BR Verweij-van Vught AMJJ MacLaren DM Transferrins and heme-compounds as iron sources for pathogenic bacteria Crit Rev Microbiol 1992 18 217 233 1532495
Neilands JB Microbial iron compounds Ann Rev Biochem 1981 50 715 731 6455965 10.1146/annurev.bi.50.070181.003435
Higgins CF ABC transporters: From microorganisms to man Annu Rev Cell Biol 1992 8 67 113 1282354 10.1146/annurev.cb.08.110192.000435
Ecker DJ Passavant CW Emery T Role of two siderophores in Ustilago sphaerogena : Regulation of biosynthesis and uptake mechanisms Biochim Biophys Acta 1982 720 242 249 6213273 10.1016/0167-4889(82)90047-7
Cunningham MW Pathogenesis of group A streptococcal infections Clin Microbiol Rev 2000 13 470 511 10885988 10.1128/CMR.13.3.470-511.2000
Francis RT JRBooth JW Becker RR Uptake of iron from hemoglobin and the haptoglobin-hemoglobin complex by hemolytic bacteria Int J Biochem 1985 17 767 773 3902529 10.1016/0020-711X(85)90262-9
Eichenbaum Z Muller E Morse SA Scott JR Acquisition of iron from host proteins by group A Streptococcus Infect Immun 1996 64 5428 5429 8945602
Lei B Liu M Smoot LM Menning HM Voyich JM Kala SV DeLeo FR Musser JM Identification and characterization of a novel heme-associated cell-surface protein made by Streptococcus pyogenes Infect Immun 2002 70 4494 4500 12117961 10.1128/IAI.70.8.4494-4500.2002
Lei B Liu M Voyich JM Prater CI Kala SV DeLeo FR Musser JM Identification and Characterization of HtsA, A Second Heme-Binding Protein Made by Streptococcus pyogenes Infect Immun 2003 71 5962 5969 14500516 10.1128/IAI.71.10.5962-5969.2003
Liu M Lei B Heme transfer from streptococcal cell-surface protein Shp to HtsA of transporter HtsABC Infect Immun 2005 73 5086 5092 16041024 10.1128/IAI.73.8.5086-5092.2005
Bates CS Montanez GE Woods CR Vincent RM Eichenbaum Z Identification and characterization of a Streptococcus pyogenes operon involved in binding of hemoproteins and acquisition of iron Infect Immun 2003 71 1042 1055 12595414 10.1128/IAI.71.3.1042-1055.2003
Janulczyk R Pallon J Bjorck L Identification and characterization of a Streptococcus pyogenes ABC transporter with multiple specificity for metal cations Mol Microbiol 1999 34 596 606 10564500 10.1046/j.1365-2958.1999.01626.x
Janulczyk R Ricci S Bjorck L MtsABC is important for manganese and iron transport, oxidative stress resistance, and virulence of Streptococcus pyogenes Infect Immun 2003 71 2656 2664 12704140 10.1128/IAI.71.5.2656-2664.2003
Ferretti JJ McShan WM Ajdic D Savic DJ Savic G Lyon K Primeaux C Sezate S Suvorov AN Kenton S Lai HS Lin SP Qian Y Jia HG Najar FZ Ren Q Zhu H Song L White J Yuan X Clifton SW Roe BA McLaughlin R Complete genome sequence of an M1 strain of Streptococcus pyogenes Proc Natl Acad Sci USA 2001 98 4658 4663 11296296 10.1073/pnas.071559398
Smoot JC Barbian KD Van Gompel JJ Smoot LM Chaussee MS Sylva GL Sturdevant DE Mackie SM Porcella SF Parkins LD Beres SB Campell DS Smith TM Zhang Q Kapur V Daly JA Veasy LG Musser JM Genome sequence and comparative microarray analysis of serotype M18 group A Streptococcus strains associated with acute rheumatic fever outbreaks Proc Natl Acad Sci USA 2002 99 4668 4673 11917108 10.1073/pnas.062526099
Beres SB Sylva GL Barbian KD Lei B Hoff JS Mammarella ND Liu M Smoot JC Porcella SF Parkins LD McCormick JK Leung DYM Schlievert PM Musser JM Genome sequence of a serotype M3 strain of group A Streptococcus : phage-encoded toxins, the high-virulence phenotype, and clone emergence Proc Natl Acad Sci USA 2002 99 10078 10083 12122206 10.1073/pnas.152298499
Smoot LM Smoot JC Graham MR Somerville GA Sturdevant DE Migliaccio CA Sylva GL Musser JM Global differential gene expression in response to growth temperature alteration in group A Streptococcus Proc Natl Acad Sci USA 2001 98 10416 10421 11517341 10.1073/pnas.191267598
Szarapinska-Kwaszewska J Mikucki J Utilization of exogenous siderophores by enterococci Med Dosw Mikrobiol 2001 53 233 243 11785180
Viswanathan VK Edelstein PH Pope CD Cianciotto NP The Legionella pneumophila iraAB locus is required for iron assimilation, intracellular infection, and virulence Infect Immun 2000 68 1069 1079 10678909 10.1128/IAI.68.3.1069-1079.2000
Velayudhan J Hughes NJ McColm AA Bagshaw J Clayton CL Andrews SC Kelly DJ Iron acquisition and virulence in Helicobacter pylori : a major role for FeoB, a high affinity ferrous iron transporter Mol Microbiol 2000 37 274 286 10931324 10.1046/j.1365-2958.2000.01987.x
Takase H Nitanai H Hoshino K Otani T Impact of siderophore production on Pseudomonas aeruginosa infections in immunosuppressed mice Infect Immun 2000 68 1834 1839 10722571 10.1128/IAI.68.4.1834-1839.2000
Janakiraman A Slauch JM The putative iron transporter system SitABCD encoded on SPI1 is required for full virulence of Salmonella typhimurium Mol Microbiol 2000 35 1146 1155 10712695 10.1046/j.1365-2958.2000.01783.x
Brown JS Gilliland SM Holden DW A Streptococcus pneumoniae pathogenicity island encoding an ABC transporter involved in iron uptake and virulence Mol Microbiol 2001 40 572 585 11359564 10.1046/j.1365-2958.2001.02414.x
Rokbi B Mignon M Maitre-Wilmotte G Lissolo L Danve B Caugant DA Quentin-Millet M-J Evaluation of recombinant transferrin-binding protein B variants from Neisseria meningitidis for their ability to induce cross-reactive and bactericidal antibodies against a genetically diverse collection of serogroup B strains Infect Immun 1997 65 55 63 8975892
Myers LE Yang Y-P Du R-P Wang Q Harkness RE Schryvers AB Myers M Klein H Loosmore SM The transferrin binding protein B of Moraxella catarrhalis elicits bactericidal antibodies and is a potential vaccine antigen Infect Immun 1998 66 4183 4192 9712766
Potter AA Schryvers AB Ogunnariwo JA Hutchins WA Lo RYC Watts T Protective capacity of the Pasteurella haemolytica transferrin-binding proteins TbpA and TbpB in cattle Microb Pathog 1999 27 197 206 10502460 10.1006/mpat.1999.0297
Brown JS Ogunniyi AD Woodrow MC Holden DW Paton JC Immunization with components of two iron uptake ABC transporters protects mice against systemic Streptococcus pneumoniae infection Infect Immun 2001 69 6702 6706 11598041 10.1128/IAI.69.11.6702-6706.2001
Kuboniwa M Amano A Shizukuishi S Nakagawa I Hamada S Specific antibodies to Porphyromonas gingivalis Lys-Gingipain by DNA vaccination inhibit bacterial binding to hemoglobin and protect mice from infection Infect Immun 2001 69 2972 2979 11292714 10.1128/IAI.69.5.2972-2979.2001
West D Reddin K Matheson M Heath R Funnell S Hudson M Robinson A Gorringe A Recombinant Neisseria meningitidis transferrin binding protein A protects against experimental meningococcal infection Infect Immun 2001 69 1561 1567 11179327 10.1128/IAI.69.3.1561-1567.2001
Lei B Liu M Chesney G Musser JM Extracellular putative lipoproteins made by Streptococcus pyogenes : Identification of new potential vaccine candidate antigens J Infect Dis 2004 189 79 89 14702157
Hoe N Nakashima PK Grigsby D Pan X Dou SJ Naidich S Garcia M Kahn E Bergmire-Sweat D Musser JM Rapid molecular genetic subtyping of serotype M1 group A Streptococcus strains Emerg Infect Dis 1999 5 254 263 10221878
LeBlanc DJ Lee LN Inamine JM Cloning and nucleotide base sequence analysis of a spectinomycin adenyltransferase AAD(9) determinant from Enterococcus faecalis Antimicrob Agents Chemother 1991 35 1804 1810 1659306
Podbielski A Spellerberg B Woischnik M Pohl B Lutticken R Novel series of plasmid vectors for gene inactivation and expression analysis in group A streptococci (GAS) Gene 1996 177 137 147 8921859 10.1016/0378-1119(96)84178-3
McIver KS Scott JR Role of mga in growth phase regulation of virulence genes of the group A streptococcus J Bacteriol 1997 179 5178 5187 9260962
Lei B Mackie S Lukomski S Musser JM Identification and immunogenicity of Group A Streptococcus culture supernatant proteins Infect Immun 2000 68 6807 6818 11083799 10.1128/IAI.68.12.6807-6818.2000
|
16242028
|
PMC1276800
|
CC BY
|
2021-01-04 16:31:06
|
no
|
BMC Pediatr. 2005 Oct 21; 5:39
|
latin-1
|
BMC Pediatr
| 2,005 |
10.1186/1471-2431-5-39
|
oa_comm
|
==== Front
BMC Vet ResBMC Veterinary Research1746-6148BioMed Central London 1746-6148-1-51621612010.1186/1746-6148-1-5Research ArticleThe CD11a partner in Sus scrofa lymphocyte function-associated antigen-1 (LFA-1): mRNA cloning, structure analysis and comparison with mammalian homologues Vanden Bergh Philippe GAC [email protected] Thomas [email protected] Laurent LM [email protected] Anne VT [email protected] Daniel JM [email protected] Pathology Department, Faculty of Veterinary Medicine, University of Liege, Colonster Boulevard 20 B43, B-4000 Liege, Belgium2005 10 10 2005 1 5 5 6 7 2005 10 10 2005 Copyright © 2005 Vanden Bergh et al; licensee BioMed Central Ltd.2005Vanden Bergh et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Lymphocyte function-associated antigen-1 (LFA-1, CD11a/CD18, alphaLbeta2), the most abundant and widely expressed beta2-integrin, is required for many cellular adhesive interactions during the immune response. Many studies have shown that LFA-1 is centrally involved in the pathogenesis of several diseases caused by Repeats-in-toxin (RTX) -producing bacteria.
Results
The porcine-LFA-1 CD11a (alpha) subunit coding sequence was cloned, sequenced and compared with the available mammalian homologues in this study. Despite some focal differences, it shares all the main characteristics of these latter. Interestingly, as in sheep and humans, an allelic variant with a triplet insertion resulting in an additional Gln-744 was consistently identified, which suggests an allelic polymorphism that might be biologically relevant.
Conclusion
Together with the pig CD18-encoding cDNA, which has been available for a long time, the sequence data provided here will allow the successful expression of porcine CD11a, thus giving the first opportunity to express the Sus scrofa beta2-integrin LFA-1 in vitro as a tool to examine the specificities of inflammation in the porcine species.
==== Body
Background
Integrins constitute a large family of adhesion molecules with important roles in cell-extracellular matrix and cell-cell interactions which condition both the maintenance of tissue integrity and the promotion of cellular migration. They are heterodimeric membrane glycoproteins composed of non-covalently associated single-pass transmembrane α and β subunits, which are expressed on a wide range of cells [1]. The biggest part of each integrin subunit is extracellular while transmembrane region and cytoplasmic tail are typically reduced. The N-terminal domains of the α and β subunits associate to form the integrin headpiece, which contains the ligand binding site. The C-terminal segments traverse the plasma membrane and mediate interactions with the cytoskeleton and with signalling proteins [2,3].
Among the integrins, the leukocyte-specific β2-integrins (CD11/CD18) include four members: (i) CD11a/CD18 (LFA-1, αLβ2) on all leukocytes ; (ii) CD11b/CD18 (Mac-1, CR3, αMβ2) mainly on myeloid cells ; (iii) CD11c/CD18 (gp150/95, CR4, αXβ2, Leu-M5) and (iv) CD11d/CD18 (αDβ2) on monocytes and macrophages [4]. The individuals lacking functional β2 integrins due to mutations in the β2 (CD18) subunit develop the LAD (lymphocyte adhesion deficiency) I syndrome characterized by repeated infections. This disease demonstrated that β2 integrins are of relevant importance in (i) leukocyte development and maturation, (ii) naïve cells circulation in secondary lymphoid tissues and (iii) leukocytes transendothelial migration to injured tissue [5-7].
Lymphocyte function-associated antigen-1 (LFA-1, CD11a/CD18, αLβ2), the most abundant and widespread in expression β2-integrin, binds to the membrane proteins termed intercellular adhesion molecules ICAM-1 to ICAM-5 [4,8-12]. Several studies have shown that LFA-1 is centrally involved in the pathogenesis of diseases caused by Repeats-in-toxin (RTX) -producing bacteria. The virulence of both Actinobacillus actinomycetemcomitans (stomatitis in humans) and Mannheimia haemolytica (pneumonia in cattle) is clearly associated with the ligand-receptor interactions between their respective leukotoxin and CD11a/CD18, which triggers the synthesis and release of a wide array of cytokines and chemoattractants that exacerbate inflammation, and ultimately results in massive leukolysis [13,14]. As Actinobacillus pleuropneumoniae, the main causative agent of pneumonia in pigs, also produces toxins of the RTX family (Apx toxins) [15], it is tempting to hypothesize that the pathogenesis of the disease similarly rely on an interaction with the porcine LFA-1. On a more practical point of view, increasing our knowledge about this putative interaction could help the pig industry in controlling the economical losses and antibiotics abuses that are currently associated with A. pleuropneumoniae pneumonia [16]. The Sus scrofa CD18 (β2-) subunit has been well characterized [17], which is not the case of its partner in the LFA-1 heterodimer, CD11a. The purpose of this study is to report the cloning, sequencing and analysis of a cDNA encoding porcine CD11a, thus giving the first opportunity to produce recombinant LFA-1 for studies focused on interactions between Apx toxins and swine LFA-1.
Results and discussion
Characterization of PoCD11a-encoding cDNA and deduced amino-acid sequence
The PoCD11a cDNA sequence contains an ORF of 3519 bp [GenBank:DQ013285] or 3522 bp [GenBank:DQ013284] that codes for 1172 or 1173 aa followed by ~500 bp that constitute the 3'-UTR (Fig. 1). The 1173 aa mature PoCD11a contains a 23-residue putative leader peptide (residues 1–23), an extracellular domain of 1064 residues (24–1087), a single hydrophobic transmembrane region of 24 residues (1088–1111) and a short cytoplasmic tail of 62 residues (1112–1173) (Fig. 1). Six N-linked putative glycosylation sites (Asn-Xaa-Ser/Thr) are found in the extracellular domain (Fig. 1). The PoCD11a possesses 22 cysteine residues, among which one is located into the cytoplasmic tail (Fig. 1). A subset of integrin α chains (α1, α2, α10, α11, αD, αE, αL, αM and αX), including CD11a, contain an I-domain (for Inserted domain and also called αLI-domain or LA-domain) that is homologous to the family of von Willebrand Factor (vWF) A-type domains and to cartilage matrix protein [1,18]. The I-domain has been associated with ligand binding. Its three-dimensional structure consists of a five-stranded parallel β-sheet core surrounded on both faces by seven α-helices. A short antiparallel strand occurs on one edge of this sheet [19]. The I-domain (148–331) contains a metal ion-dependent adhesion site (MIDAS) (residues 160–164, 229, 262) [19] (Fig. 1). The I-domain crystallisation has demonstrated that a "closed" (low affinity) and an "open" (high affinity) forms exist, and that the major conformational changes during transition from the closed to open states include a rearrangement of the cation-coordinating residues in the MIDAS site, accompanied by a small inward movement of the α1 helix and a large downward shift of the mobile C-terminal α7 helix [20]. The extracellular domain of PoCD11a contains seven internal repeats that surround the I-domain [21]. The degree of identity is highest among the three COOH-terminal repeats (18–31%) and their central region (466–474, 528–536 and 588–596) is similar to the EF hand divalent cation-binding motifs of troponin C, parvalbumin and galactose binding protein [21] (Fig. 1). All the N-glycosylation sites and all but one cysteine residues are found outside the I-region and divalent cation binding motifs (Fig. 1), consistent with the hypothesis that these regions may undergo conformational changes important in ligand binding [21,22]. The cytoplasmic portion of PoCD11a contains three potential phosphorylation sites and also a conserved "GFFKR" basic sequence near the transmembrane region (Fig. 1). The integrins become constitutively active when this sequence is deleted, the GFFKR motif thus normally fixes the integrins in an inactive state [4,23].
Figure 1 The nucleotide and deduced amino acid sequences of Sus scrofa CD11a cDNA. The putative leader peptide and transmembrane region are respectively represented by green and dark blue boxes. The αI-domain is showed by a yellow box. Its five-stranded β-sheets and seven α-helices are underlined. Its metal-ion dependent adhesion site (MIDAS) is represented by red boxes. The sequences of the seven repeats that surround the αI-domain are framed. Light blue boxes represent the central divalent cation-binding motifs (DCBM) of the three COOH-terminal repeats. The important Glu-333 residue (E) and the supplementary Gln-744 (Q) are in black boxes. The conserved sequence "GFFKR" of the cytoplasmic tail, near the transmembrane region, is in a dark grey box. Cysteine residues (¤), potential N-glycosylation sites (#) and potential cytoplasmic-tail phosphorylation sites (+) are marked at the top of the sequences. Seven independent clones were sequenced in both directions. Sequence data have been deposited at GenBank under accession nos. DQ013285 and DQ013284 (shown here). Both sequences differ by a glutamine deletion in position 744 in sequence DQ013284.
Among the seven positive clones sequenced, two presented a supplementary "cag" codon (2230–2232) that codes for a glutamine (Gln, Q) in position 744 [GenBank:DQ013284]. This addition is located in the extracellular domain of PoCD11a, outside of the I-domain and divalent cation-binding motifs and, according to the GORIV bioinformatic program, increased the length of an α-helix. The Gln-744 addition was also observed in the human [GenBank:NM_002209, GenBank:AY892236] and ovine [24] CD11a cDNAs. The Gln addition could thus have a biological importance for the mature CD11a. Studies of genomic sequences will permit to know if this addition represents two alleles or if it is generated by an alternative splicing.
Comparison among species
Overall, the general organization of porcine, human [21], murine [25], bovine [26] and ovine [24] CD11a proteins is quite similar (Fig. 2). Comparison between mature PoCD11a sequence and its human, murine, bovine and ovine counterparts shows overall 77%, 69%, 78% and 77% identity, respectively, with the highest identity for the MIDAS, the cation binding motifs and the transmembrane region and the lowest identity for the cytoplasmic tail (Table 1). The high conservation of the MIDAS and the putative cation binding motifs is consistent with an involvement of these regions in the functional activity of LFA-1 α subunit, as suggested by the requirement of Mg2+ and Ca2+ for CD11a/CD18-dependent cellular interactions [22] or binding to purified ICAM-1 [27,28]. The transmembrane region shows also a high degree of conservation that could be explained by (i) physicochemical, and (ii) functional constraints. Indeed, (i) residues lying in the membrane have to possess a hydrophobic character to warrant liposolubility, which is confirmed by the presence of many leucine residues (fig. 2) and (ii) bi-directional integrin signalling (inside-out and outside-in) is accomplished by transmission of information across the plasma membrane [29]. By contrast, the low conservation of the COOH-terminal part of the cytoplasmic tail suggests that it is not required to guarantee adequate functioning of LFA-1. This is in agreement with the observation that truncation of the LFA-1 α subunit cytoplasmic domain has no effect on binding to ICAM-1, whereas binding is markedly diminished by β subunit cytoplasmic domain truncation [30]. However, the part near the transmembrane region of the cytoplasmic tail, containing the "GFFKR" sequence, is highly conserved (Table 1). This is consistent with the stabilizing role of this motif for the alpha/beta complex, possibly because of its direct involvement in heterodimer formation [23]. Residue Glu-333 that is located in the linker following the I domain and that is known to be critical for communication to the β2 I-like domain, rolling, integrin extension and activation by Mn2+ of firm adhesion [31] is strictly conserved, too (fig. 2).
Table 1 Between-species percent identities of CD11a constitutive blocks. Po, Hu, Mu, Bo and Ov : porcine, human, murine, bovine and ovine CD11a, respectively ; MIDAS: metal-ion dependent adhesion site ; vs : versus.
Block Po vs. Hu (%) Po vs. Mu (%) Po vs. Bo (%) Po vs.Ov (%)
Overall 77 69 78 77
Putative signal peptide 56 45 86 78
Extracellular domain 77 70 77 78
I-domain 79 72 82 82
MIDAS 85 85 85 85
Putative cation binding motif 1 77 77 66 66
Putative cation binding motif 2 77 77 77 77
Putative cation binding motif 3 88 88 100 100
Transmembrane region 91 75 83 87
Cytoplasmic tail 55 47 55 53
"GFFKR" motif 100 100 100 100
Figure 2 Comparison of the porcine (Po-), human (Hu-), murine (Mu-), bovine (Bo-) and ovine (Ov-) α subunits amino acids sequences. Black column with white letter, dark gray column with white letter and light gray column with black letter represent identity among 5, 4 and 3 species, respectively. Cysteine residues (¤), potential N-glycosylation sites (#) and potential cytoplasmic-tail phosphorylation sites (+) are marked at the bottom of the sequences. The important Glu-333 residue (E) and the Gln-744 residue (Q) are respectively identified by ($) and (=). The stripes above the sequences represent the deduced different constitutive parts of the protein: signal peptide (), extracellular domain (), transmembrane region (), cytoplasmic tail (),αI-domain () and its metal-ion dependent adhesion site (), and the central divalent cation-binding motifs of the three COOH-terminal repeats (). The highly conserved "GFFKR" motif of the cytoplasmic tail is framed for the different species.
Every cysteine residue in the mature porcine CD11a is present at the same location in human, murine, bovine and ovine CD11a, which is consistent with a role in maintaining the global structure of the protein. The mouse version distinguishes by an additional cysteine residue at position 199 within the extracellular portion. Of six potential Asn-glycosylation sites in porcine CD11a, the ones present at amino acids 186, 668, 724 and 860 are strictly conserved. Without predictable consequences on a functional point of view, one glycosylation site is only absent from porcine and murine CD11a (residue 897 of human sequence). The mouse sequence shows additional glycosylation sites at position 270 and 776. Furthermore, the porcine Asn-Xaa-Ser/Thr sites in position 87 and 728 are also found in human and murine homologues but not specially in the ruminant sequences, and the ovine sequence owns two supplementary sites at position 646 and 1000.
Conclusion
This study reports for the first time the isolation and sequencing of the porcine LFA-1 αL subunit (CD11a) cDNA, and demonstrates that, despite some focal differences, it shares all the main characteristics of its known mammalian homologues. Along with the porcine CD18-encoding cDNA which is available [17], the sequence data provided here allow the successful cloning of PoCD11a, thus giving the first opportunity to express porcine LFA-1 in vitro as a tool to examine the specificities of inflammation in the porcine species.
Methods
RNA isolation
Total RNA from spleen of a freshly slaughtered pig (Sus scrofa domestica) was extracted with TRIzol (Invitrogen, USA) as described by the manufacturer.
Amplification of cDNA ends
We used SMART RACE technology (Clontech Laboratories Inc., USA) to obtain porcine CD11a (PoCD11a) 5'- and 3'- ends and RT-PCR to amplify full-length PoCD11a CDS. For first strand cDNA synthesis, and according to the bovine CD11a sequence available [GenBank:AY267467], gene-specific primers were designed which were expected to give nonoverlapping ~1,5 kb rapid amplification of cDNA ends (RACE) products : a sense primer for the 3'-RACE PCR 5'-tgcaatgtragctctcccatcttc-3' (corresponding to nt 2575–2598) and an antisense primer for the 5'-RACE PCR 5'-aagatgtacacrgccccctgctcctcca-3' (nt 1628–1655). Reverse transcription and polymerase chain reactions (PCR) were carried out according to the instruction manual of the SMART RACE cDNA Amplification Kit. The 5'- and 3'-RACE products were gel-purified using the Qiaquick Gel Extraction Kit (Qiagen), TA-cloned into pCRII-TOPO (Invitrogen, USA) and seeded on kanamycin IPTG plates. Minipreps were obtained from colonies grown in 5 ml LB-Kan broth and the clones were sequenced on the ABI-3100 Genetic Analyzer using the Big Dye terminator chemistry (Applied Biosystems, USA).
Molecular cloning of full-length cDNA
Total RNA from spleen cells was reverse transcribed using Improm II (Promega, USA). The full-length cDNA was then generated by long distance PCR using elongase amplification technology (Invitrogen, USA) with primers designed from the proximal part of 5'- and the distal part of 3'-RACE products: 5'-ggtatggtccctccagaagc-3' (forward) and 5'-tcaggcctgggcttcagtcg-3' (reverse). The following cycling parameters were applied: 5 min at 94°C, then 35 cycles including: (i) 30 s at 94°C, (ii) 45 s at 58°C, and (iii) 3 min 30 s at 68°C, followed by a final extension at 68°C for 10 min. Resulting PCR products were then processed for sequencing as aforementioned for the RACE products. The CD11a cDNA sequence was deduced from sequences obtained from seven independent clones. Sequence data have been deposited at GenBank under accession nos. DQ013285 and DQ013284.
Bioinformatics
Primers design was performed with Netprimer [32] and Primer 3 [33]. Nucleotidic sequence and identity analyses were carried out using respectively Chromas v.2.21 [34] and BLAST programs [35]. Alignment of amino acids sequences was drawn by GeneDoc v.2.6.002 [36] following the blosum 62 matrix. SignalP v.2.0.b2 [37] and NetNGlyc v.1.0 [38] provided peptide signal and N-glycosylation sites prediction, respectively. The secondary structures were resolved by the GOR secondary structure prediction method version IV [39].
List of abbreviations
aa, amino acid ; Bo, bovine; CD, cluster of differenciation ; CR, complement receptor ; DCBM, divalent-cation binding motif ; Hu, human; ICAM, intercellular adhesion molecule; kan, kanamycin ; LFA, lymphocyte function-associated antigen; MIDAS, metal-ion dependent adhesion site; Mu, murine; Ov, ovine ; Po, porcine ; RACE, rapid amplification of cDNA ends ; TCR, T-cell receptor.
Authors' contributions
PVB carried out cloning, sequencing, sequence alignment and drafting of the manuscript. TF, LZ and AT participated in the design of the study and helped to structure analysis. DD conceived the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Authors are grateful to Prof M. Georges for giving free access to all the facilities of the laboratory of molecular genetics. Philippe Vanden Bergh is the recipient of a studentship from the "Fonds pour la formation à la Recherche dans l'Industrie et l'Agriculture", rue d'Egmont 5, B-1000 Bruxelles. Thomas Fett and Laurent Zecchinon are supported by the belgian federal services for public health and security of the food chain and environment, grant S-6107. We thank Hélène Vandegaart for its technical assistance.
==== Refs
Springer TA Adhesion receptors of the immune system Nature 1990 346 425 434 1974032 10.1038/346425a0
Dib K Beta 2 integrin signaling in leukocytes Front Biosci 2000 5 D438 451 10762594
Humphries MJ Symonds EJ Mould AP Mapping functional residues onto integrin crystal structures Curr Opin Struct Biol 2003 13 236 243 12727518 10.1016/S0959-440X(03)00035-6
Gahmberg CG Valmu L Fagerholm S Kotovuori P Ihanus E Tian L Pessa-Morikawa T Leukocyte integrins and inflammation Cell Mol Life Sci 1998 54 549 555 9676574 10.1007/s000180050183
Arnaout MA Leukocyte adhesion molecules deficiency: its structural basis, pathophysiology and implications for modulating the inflammatory response Immunol Rev 1990 114 145 180 1973407
Trowald-Wigh G Hakansson L Johannisson A Norrgren L Hard af Segerstad C Leucocyte adhesion protein deficiency in Irish setter dogs Vet Immunol Immunopathol 1992 32 261 280 1352926 10.1016/0165-2427(92)90050-Z
Nagahata H Kehrli ME JrMurata H Okada H Noda H Kociba GJ Neurtrophil function and pathologic findings in Holstein calves with leukocyte adhesion deficiency Am J Vet Res 1994 55 40 48 7908182
Marlin SD Springer TA Purified intercellular adhesion molecule-1 (ICAM-1) is a ligand for lymphocyte function-associated antigen 1 (LFA-1) Cell 1987 51 813 819 3315233 10.1016/0092-8674(87)90104-8
Simmons D Makgoba MW Seed B ICAM, an adhesion ligand of LFA-1, is homologous to the neural cell adhesion molecule NCAM Nature 1988 331 624 627 3340213 10.1038/331624a0
Staunton DE Dustin ML Springer TA Functional cloning of ICAM-2, a cell adhesion ligand for LFA-1 homologous to ICAM-1 Nature 1989 339 61 64 2497351 10.1038/339061a0
Fawcett J Holness CL Needham LA Turley H Gatter KC Mason DY Simmons DL Molecular cloning of ICAM-3, a third ligand for LFA-1, constitutively expressed on resting leukocytes Nature 1992 360 481 484 1448173 10.1038/360481a0
Ihanus E Uotila L Toivanen A Stefanidakis M Bailly P Cartron JP Gahmberg CG Characterization of ICAM-4 binding to the I domains of the CD11a/CD18 and CD11b/CD18 leukocyte integrins Eur J Biochem 2003 270 1710 1723 12694184 10.1046/j.1432-1033.2003.03528.x
Lally ET Kieba IR Sato A Green CL Rosenbloom J Korostoff J Wang JF Shenker BJ Ortlepp S Robinson MK Billings PC RTX toxins recognize a beta2 integrin on the surface of human target cells J Biol Chem 1997 272 30463 30469 9374538 10.1074/jbc.272.48.30463
Zecchinon L Fett T Desmecht D How Mannheimia haemolytica defeats host defence through a kiss of death mechanism Vet Res 2005 36 133 156 15720968 10.1051/vetres:2004065
Bosse JT Janson H Sheehan BJ Beddek AJ Rycroft AN Kroll JS Langford PR Actinobacillus pleuropneumoniae: pathobiology and pathogenesis of infection Microbes Infect 2002 4 225 235 11880056 10.1016/S1286-4579(01)01534-9
Losinger WC Economic impacts of reduced pork production associated with the diagnosis of Actinobacillus pleuropneumoniae on grower/finisher swine operations in the United States Prev Vet Med 2005 68 181 193 15820115 10.1016/j.prevetmed.2004.12.004
Lee J-K Schook LB Rutherford MS Molecular cloning and characterization of the porcine CD18 leukocyte adhesion molecule Xenotransplantation 1996 3 222 230
Colombatti A Bonaldo P The superfamily of proteins with von Willebrand factor type A-like domains: one theme common to components of extracellular matrix, hemostasis, cellular adhesion, and defense mechanisms Blood 1991 77 2305 2315 2039815
Qu A Leahy DJ Crystal structure of the I-domain from the CD11a/CD18 (LFA-1, alpha L beta 2) integrin Proc Natl Acad Sci USA 1995 92 10277 10281 7479767
Mould AP Humphries MJ Regulation of integrin function through conformational complexity: not simply a knee-jerk reaction? Curr Opin Cell Biol 2004 16 544 551 15363805 10.1016/j.ceb.2004.07.003
Larson RS Corbi AL Berman L Springer T Primary structure of the leukocyte function-associated molecule-1 alpha subunit: an integrin with an embedded domain defining a protein superfamily J Cell Biol 1989 108 703 712 2537322 10.1083/jcb.108.2.703
Rothlein R Springer TA The requirement for lymphocyte function-associated antigen 1 in homotypic leukocyte adhesion stimulated by phorbol ester J Exp Med 1986 163 1132 1149 3517218 10.1084/jem.163.5.1132
Pardi R Bossi G Inverardi L Rovida E Bender JR Conserved regions in the cytoplasmic domains of the leukocyte integrin alpha L beta 2 are involved in endoplasmic reticulum retention, dimerization, and cytoskeletal association J Immunol 1995 155 1252 1263 7636193
Fett T Zecchinon L Baise E Desmecht D Cloning and characterization of the primary structure of the sheep lymphocyte function-associated antigen-1 alpha subunit Mol Immunol 2005 42 1503 1508 15950743 10.1016/j.molimm.2005.01.005
Kaufmann Y Tseng E Springer TA Cloning of the murine lymphocyte function-associated molecule-1 alpha-subunit and its expression in COS cells J Immunol 1991 147 369 374 2051027
Fett T Zecchinon L Baise E Desmecht D The bovine (Bos taurus) CD11a-encoding cDNA: molecular cloning, characterisation and comparison with the human and murine glycoproteins Gene 2004 325 97 101 14697514 10.1016/j.gene.2003.09.043
Dustin ML Springer TA T-cell receptor cross-linking transiently stimulates adhesiveness through LFA-1 Nature 1989 341 619 624 2477710 10.1038/341619a0
Vitte J Pierres A Benoliel AM Bongrand P Direct quantification of the modulation of interaction between cell- or surface-bound LFA-1 and ICAM-1 Leukoc Biol 2004 76 594 602 10.1189/jlb.0204077
Kim M Carman CV Springer TA Bidirectional transmembrane signaling by cytoplasmic domain separation in integrins Science 2003 301 1720 1725 14500982 10.1126/science.1084174
Hibbs ML Xu H Stacker SA Springer TA Regulation of adhesion of ICAM-1 by the cytoplasmic domain of LFA-1 integrin beta subunit Science 1991 251 1611 1613 1672776
Salas A Shimaoka M Kogan AN Harwood C von Andrian UH Springer TA Rolling adhesion through an extended conformation of integrin alphaLbeta2 and relation to alpha I and beta I-like domain interaction Immunity 2004 20 393 406 15084269 10.1016/S1074-7613(04)00082-2
Netprimer
Rozen S Skaletsky HJ Primer3 on the WWW for general users and for biologist programmers Bioinformatics Methods and Protocols: Methods in Molecular Biology 2000 Totowa, Humana Press 365 386
Chromas v.2.21
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Nicholas B Karl B Nicholas P Hugh B Jr GeneDoc : a tool for editing and annotating multiple sequence alignments
Nielsen H Engelbrecht J Brunak S von Heijne G Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites Protein Eng 1997 10 1 6 9051728 10.1093/protein/10.1.1
Jensen LJ Gupta R Blom N Devos D Tamames J Kesmir C Nielsen H Staerfeldt HH Rapacki K Workman C Andersen CA Knudsen S Krogh A Valencia A Brunak S Prediction of human protein function from post-translational modifications and localization features J Mol Biol 2002 319 1257 1265 12079362 10.1016/S0022-2836(02)00379-0
Garnier J Gibrat JF Robson B GOR method for predicting protein secondary structure from amino acid sequence Methods Enzymol 1996 266 540 553 8743705
|
16216120
|
PMC1276801
|
CC BY
|
2021-01-04 16:29:46
|
no
|
BMC Vet Res. 2005 Oct 10; 1:5
|
utf-8
|
BMC Vet Res
| 2,005 |
10.1186/1746-6148-1-5
|
oa_comm
|
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-341625091310.1186/1476-7120-3-34ResearchDobutamine stress echocardiography in healthy adult male rats Plante Eric [email protected] Dominic [email protected] Marie-Claude [email protected] Élise [email protected] Jacques [email protected] Marie [email protected] Groupe de Recherche sur les Valvulopathies, Centre de Recherche Hôpital Laval, Institut de cardiologie de Québec, Université Laval, Québec, Canada2005 26 10 2005 3 34 34 19 7 2005 26 10 2005 Copyright © 2005 Plante et al; licensee BioMed Central Ltd.2005Plante et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Dobutamine stress echocardiography is used to investigate a wide variety of heart diseases in humans. Dobutamine stress echocardiography has also been used in animal models of heart disease despite the facts that the normal response of healthy rat hearts to this type of pharmacological stress testing is unknown. This study was performed to assess this normal response.
Methods
15 normal adult male Wistar rats were evaluated. Increasing doses of dobutamine were infused intravenously under continuous imaging of the heart by a 12 MHz ultrasound probe.
Results
Dobutamine stress echocardiography reduced gradually LV diastolic and systolic dimensions. Ejection fraction increased by a mean of +24% vs. baseline. Heart rate increased progressively without reaching a plateau. Changes in LV dimensions and ejection fraction reached a plateau after a mean of 4 minutes at a constant infusion rate.
Conclusion
DSE can be easily performed in rats. The normal response is an increase in heart rate and ejection fraction and a decrease in LV dimensions. A plateau in echocardiographic measurements is obtained after 4 minutes of a constant infusion rate in most animals.
Dobutaminestress echocardiographyratanimal modelstress
==== Body
Introduction
Dobutamine stress echocardiography (DSE) is commonly used in clinical practice to investigate patients with a wide variety of cardiovascular diseases. DSE can be used for many purposes such as to seek myocardial ischemia, evaluate valvular diseases, measure myocardial viability and assess myocardial contractile reserve [1-8]. Improvements in cardiac ultrasound imaging devices in recent years have allowed investigators to use echocardiography in small animals with cardiac diseases such as rats and mice and obtain images of very good quality. Consequently, research groups are now routinely using cardiac ultrasound imaging not only to investigate large animals (such as dogs and pigs) but also small rodents. Dobutamine stress imaging has also emerged in those small animals. Investigators working with small animal models of cardiac diseases have been using dobutamine stimulation mostly to assess myocardial contractile reserve. Accurate interpretation of a stress test requires that the response of normal subjects to that test is well documented. However, there is currently no available data on the normal response to dobutamine stimulation of the hearts of rats and we do not know if they respond to DSE as humans do. Therefore, the present study was designed to assess the normal response of rat hearts to DSE.
Methods
Animals
15 adult male Wistar rats (400–450 g) were investigated. This protocol was approved by the Université Laval's animal protection committee and was consistent with the recommendations of the Canadian Council on animal care.
Dobutamine Stress Echocardiography
The animals were sedated with isoflurane (2.5%, inhaled). The femoral vein was dissected and cannulated for the intravenous dobutamine infusion. The femoral artery was also cannulated to measure arterial blood pressure. The thorax was shaved and the animal was put on a dorsal decubitus position. Imaging of the heart was done using a 12 MHz ultrasound probe and a Sonos 5500 echographer (Philips Medical Imaging, Andover, MA). Imaging depth was adjusted for each animal to obtain the largest possible 2D image including the whole left ventricle in the imaging sector. Harmonic imaging was used for all the studies with optimal adjustment of gain to ensure the best quality of images and adequate delineation of endocardial borders. Images were stored on magneto-optical disks for off-line analysis. Baseline measurements were obtained under isofurane anesthesia before the beginning of dobutamine infusion and continuous imaging was done during the dobutamine test. Results were compiled at every minute during the infusion.
2 different dobutamine infusion protocols were tested
Protocol #1
Dobutamine was infused intravenously at a constant rate of 10 μg/kg/min until a plateau was reached in the left ventricular (LV) ejection fraction for at least 2 minutes (n = 10 rats). Similar doses of dobutamine infusion has been used by others to evaluate myocardial contractile in rats [9].
Protocol #2
Progressively increasing doses of dobutamine (5, 10 and 20 μg/kg/min) were infused in this protocol (n = 5 rats). Each dose was infused for 4 minutes before imaging and increasing to the next step. This protocol of progressively increasing doses is closer to DSE protocols used in clinical practice and has been used by other investigators [10] using other imaging techniques.
Left ventricular dimensions and septal and posterior wall thicknesses were obtained in the parasternal long axis view using M-mode imaging. Relative wall thickness (RWT) was calculated as the ratio of the sum of the wall thicknesses (septal and posterior) to the end-diastolic diameter. Ejection fraction was calculated using the formula of Quinones et al. [11].
Statistical analysis
Results are presented as mean ± SEM unless specified otherwise. Paired Student t tests were used to compare the results at baseline with those at the end of dobutamine infusion. One-way analysis of variance was performed to compare serial data. Statistical significance was set at a p value of 0.05 or less using post-hoc Tukey's test. Data and statistical analysis were performed using GraphPad Prism version 4.02 for Windows, GraphPad Software (San Diego California USA).
Results
Feasibility
Dobutamine stress echocardiography was easily performed in all animals and well tolerated. All animals survived the procedure and we did not encounter any complications.
Hemodynamic response (fig. 1 and 5a)
Figure 1 Hemodynamic response to dobutamine infusion in rats (protocol #1). A: heart rate, B: systolic blood pressure (expressed as % of maximal response) and C: diastolic blood pressure (expressed as % of maximal response). *: p < 0.05 vs. baseline. Data is presented are mean ± SEM.
Figure 5 Dose-response effects of dobutamine infusion on the LV dimensions and ejection fraction (protocol #2). A: Heart rate, B LV diameters, C: septal wall, (SW). D posterior wall, (PW) E: Ejection fraction (EF) and F: relative wall thickness (RWT). Closed circles: diastolic measurements; Open circles: systolic measurements*: p > 0.05 vs. baseline. Data are presented as mean ± SEM.
Protocol #1
Heart rate and blood pressure response to dobutamine infusion are displayed in figure 1. There was a constant increase in heart rate (Baseline: 349 ± 12 bpm vs. Dobutamine: 424 ± 4 bpm, p < 0.01) in all the animals during the infusion of dobutamine (fig. 1a). Systolic blood pressure sharply increased after only 1 minute of infusion and reached a plateau rapidly after (fig. 1b). We observed a trend towards a mild increase in diastolic blood pressure after 2 minutes of infusion but this did not reach statistical significance and diastolic pressure remained comparable to baseline for the rest of the protocol (fig 1c).
Protocol #2
Animals in protocol #2 responded in a similar fashion to protocol #1 with a progressive increase in heart rate with each step (fig 5a), and a plateau in systolic and diastolic pressure (not shown).
ECG monitoring
ECG was continuously monitored in 5 animals. All animals progressively developed sinus tachycardia (see above) with dobutamine stimulation as expected. We did not notice any ST segment change nor any form of ventricular or supra-ventricular arythmia.
LV dimensions and ejection fraction (fig. 2, 3, 4)
Figure 2 Left ventricular dimensions and ejection fraction during dobutamine infusion (protocol #1). A: End-diastolic diameter B: end-systolic diameter C: ejection fraction. *: p < 0.05 vs. baseline. Data are presented as mean ± SEM.
Figure 3 Representative M-mode long axis view of the left ventricle at before (t = 0) and after 6 minutes (t = 6) of dobutamine infusion.
Figure 4 Left ventricular wall thickness and relative wall thickness during dobutamine infusion. A: septal wall (SW), B: posterior wall (PW) and C: relative wall thickness (RWT). Closed circles: diastolic measurements; Open circles: systolic measurements. *: p < 0.05 vs. baseline. Data are presented as mean ± SEM.
Protocol #1
Dobutamine infusion reduced LV dimensions as shown in fig 2 and 3. End-diastolic dimensions (fig 2a) progressively decreased compared to baseline, reaching a plateau after 4 minutes (Baseline: 8.5 ± 0.14 mm vs. Dobutamine: 7.9 ± 0.28 mm, p < 0.01). The effects of dobutamine were more prominent on end-systolic dimensions (fig. 2b) with a measured decrease from 4.1 ± 0.24 mm to 2.6 ± 0.27 mm (p < 0.01). As for diastolic dimensions, a plateau was reached in end-systolic dimensions after a mean of 4 minutes of infusion. Ejection fraction behaved similarly with an increase from 74 ± 3.1% at baseline to 84 ± 3.6% with dobutamine (fig. 2c). The left ventricular cavity transiently remodeled in a concentric fashion during the dobutamine infusion as depicted by the progressive increase in relative wall thickness (fig. 4c). Diastolic septal and posterior wall thickness were unaffected by dobutamine whereas systolic thickening increased under dobutamine stimulation as expected under an inotropic stimulus (fig. 4a and 4b). Significant increases in systolic thickening was observed only after a mean of 3 minutes of dobutamine infusion and remained constant afterwards.
Protocol #2
Findings in protocol #2 were similar to protocol #1 with a progressive decrease in diastolic and systolic dimensions (fig 5b), increase in systolic thickening (fig 5c–d), increase in ejection fraction (fig 5e) and acute concentric remodeling (fig 5f).
Discussion
In this study, we demonstrate that dobutamine stress echocardiography is easily feasible in rats and that rat hearts respond to dobutamine infusion in a way that shares strong similarities with humans. Dobutamine infusion in rats is safe, quick and yields important information about the animal's heart response to an adrenergic stress.
Evaluating the response of the animal heart to stress in vivo is a quite challenging task. Imaging of the heart on a live, non sedated small animal such as a rat is virtually impossible. Exercise imaging would be even more complex. The explanted heart can be stimulated ex vivo in a perfusion setting but the results obtained may not reflect the reactions of the heart in the live animal. Moreover, this type of experimentation does not allow serial evaluations of the heart for obvious reasons.
Dobutamine stress echocardiography allows non-invasive, serial evaluation of the heart in the live animal. DSE is widely used in humans for the evaluation of left ventricular function, contractile reserve, ischemia and viability [1-8]. DSE has proven its usefulness in the investigation of a wide range of cardiac diseases such as cardiomyopathies, valvular diseases, heart failure and ischemic heart disease. The normal response in humans is an increase in heart rate, cardiac output and myocardial contractility. Left ventricular dimensions normally decrease under dobutamine stimulation as a result of the changes in preload and afterload induced by dobutamine. We found similar responses in our normal rats.
Dobutamine stress echocardiography has been used by other investigators using animal models of heart disease but mostly in large animals such as pigs and dogs [12-24]. Little data is available in smaller animals such as mice and rats. When DSE was used in those small animals in previous studies, collected data consisted mostly of fractional shortening or short axis cross sectional area changes or invasive intracardiac measures of dP/dT [25-29]. The global echocardiographic response of the heart to dobutamine stimulation in rats has never been described. Our study is the first to report the normal expected dynamic changes in healthy animals. We believe that such data is essential to analyze the results obtained in sick animals.
We have also shown that a plateau in echocardiographic parameters was obtained in most animals after 4 minutes of a constant infusion rate, some animals needing up to 6 minutes. Investigators using dobutamine stimulation should be aware of these delays to avoid underestimations of the effects of the drug on the left ventricle. This is of critical importance especially if left ventricular dimensions are used to calculate fractional shortening for the evaluation of contractile reserve.
Study limitations
The results of this study apply only to normal young adult male Wistar rats. The response of females, youngsters, older rats or other rodents to DSE needs to be assessed. Contractile reserve can be easily assessed in our protocol using a dose of 10 μg/kg/min of dobutamine as shown by the significant increase in wall thickening and ejection fraction but higher doses may need to be used to detect ischemia in animal models of ischemic heart disease.
Conclusion
Dobutamine stress echocardiography can be easily performed in rats and can be used to evaluate left ventricular function in response to adrenergic stress. Sufficient time should be allowed for the echographic parameters to stabilize before any echographic measurements are made during dobutamine stimulation in the rat. Investigators using DSE should be aware of the normal response of the control group of their animal model to this type of stress testing.
Acknowledgements
This work was supported by operating grants to Dr Couet and Arsenault from the Canadian Institutes of Health Research (MOP-61818), the Heart and Stroke Foundation of Canada and the Quebec Heart Institute Corporation.
==== Refs
Lualdi JC Douglas PS Echocardiography for the assessment of myocardial viability J Am Soc Echocardiogr 1997 10 772 780 9339433
Krahwinkel W Ketteler T Godke J Wolfertz J Ulbricht LJ Krakau I Gulker H Dobutamine stress echocardiography Eur Heart J 1997 18 D9 15 9183605
Pellikka PA Roger VL Oh JK Miller FA Seward JB Tajik AJ Stress echocardiography. Part II. Dobutamine stress echocardiography: techniques, implementation, clinical applications, and correlations Mayo Clin Proc 1995 70 16 27 7808046
Krenning BJ Geleijnse ML Poldermans D Roelandt JR Methodological analysis of diagnostic dobutamine stress echocardiography studies Echocardiography 2004 21 725 736 15546374 10.1111/j.0742-2822.2004.03161.x
Wu WC Bhavsar JH Aziz GF Sadaniantz A An overview of stress echocardiography in the study of patients with dilated or hypertrophic cardiomyopathy Echocardiography 2004 21 467 475 15209731 10.1111/j.0742-2822.2004.03083.x
Wu WC Ireland LA Sadaniantz A Evaluation of aortic valve disorders using stress echocardiography Echocardiography 2004 21 459 466 15209730 10.1111/j.0742-2822.2004.t01-1-03082.x
Wu WC Aziz GF Sadaniantz A The use of stress echocardiography in the assessment of mitral valvular disease Echocardiography 2004 21 451 458 15209729 10.1111/j.0742-2822.2004.03081.x
Picano E Stress echocardiography Expert Rev Cardiovasc Ther 2004 2 77 88 15038415 10.1586/14779072.2.1.77
Sullebarger JT D'Ambra PM Clark LC Thanikarry L Fontanet HL Effect of Digoxin on Ventricular Remodeling and Responsiveness of beta-Adrenoceptors in Chronic Volume Overload J Cardiovasc Pharmacol Ther 1998 3 281 290 10684510
Croteau E Benard F Bentourkia M Rousseau J Paquette M Lecomte R Quantitative myocardial perfusion and coronary reserve in rats with 13N-ammonia and small animal PET: impact of anesthesia and pharmacologic stress agents J Nucl Med 2004 45 1924 30 15534064
Quinones MA Waggoner AD Reduto LA Nelson JG Young JB Winters WL JrRibeiro LG Miller RR A new, simplified and accurate method for determining ejection fraction with two-dimensional echocardiography Circulation 1981 64 744 53 7273375
McFalls EO Murad B Haspel HC Marx D Sikora J Ward HB Myocardial glucose uptake after dobutamine stress in chronic hibernating swine myocardium J Nucl Cardiol 2003 10 385 394 12900743 10.1016/S1071-3581(03)00431-8
Handke M Heinrichs G Magosaki E Lutter G Bode C Geibel A Three-dimensional echocardiographic determination of cardiac output at rest and under dobutamine stress: comparison with thermodilution measurements in the ischemic pig model Echocardiography 2003 20 47 55 12848697 10.1046/j.1540-8175.2003.00006.x
Weidemann F Dommke C Bijnens B Claus P D'hooge J Mertens P Verbeken E Maes A Van de WF De SI Defining the transmurality of a chronic myocardial infarction by ultrasonic strain-rate imaging: implications for identifying intramural viability: an experimental study Circulation 2003 107 883 888 12591760 10.1161/01.CIR.0000050146.66577.4B
Hammond HK McKirnan MD Effects of dobutamine and arbutamine on regional myocardial function in a porcine model of myocardial ischemia J Am Coll Cardiol 1994 23 475 482 8294703
Bin JP Le DE Jayaweera AR Coggins MP Wei K Kaul S Direct effects of dobutamine on the coronary microcirculation: comparison with adenosine using myocardial contrast echocardiography J Am Soc Echocardiogr 2003 16 871 879 12878997 10.1067/S0894-7317(03)00423-1
Le DE Bin JP Coggins MP Wei K Lindner JR Kaul S Relation between myocardial oxygen consumption and myocardial blood volume: a study using myocardial contrast echocardiography J Am Soc Echocardiogr 2002 15 857 863 12221400 10.1067/mje.2002.121275
McEntee K Amory H Clercx C Soyeur D Michaux C Vanhaeverbeek O Jacqmot O Henroteaux M Physiologic response to dobutamine infusion during cardiac stress testing of dogs Am J Vet Res 1998 59 1160 1165 9736396
Iwanaga Y Hoshijima M Gu Y Iwatate M Dieterle T Ikeda Y Date MO Chrast J Matsuzaki M Peterson KL Chronic phospholamban inhibition prevents progressive cardiac dysfunction and pathological remodeling after infarction in rats J Clin Invest 2004 113 727 736 14991071 10.1172/JCI200418716
Sun LS Takuma S Lui R Homma S The effect of maternal cocaine exposure on neonatal rat cardiac function Anesth Analg 2003 97 878 882 12933420 10.1213/01.ANE.0000074792.87311.DB
Abraham TP Laskowski C Zhan WZ Belohlavek M Martin EA Greenleaf JF Sieck GC Myocardial contractility by strain echocardiography: comparison with physiological measurements in an in vitro model Am J Physiol Heart Circ Physiol 2003 285 H2599 H2604 12907429
Southworth R Garlick PB Dobutamine responsiveness, PET mismatch, and lack of necrosis in low-flow ischemia: is this hibernation in the isolated rat heart? Am J Physiol Heart Circ Physiol 2003 285 H316 H324 12637355
Sullebarger JT D'Ambra PM Clark LC Thanikarry L Fontanet HL Effect of Digoxin on Ventricular Remodeling and Responsiveness of beta-Adrenoceptors in Chronic Volume Overload J Cardiovasc Pharmacol Ther 1998 3 281 290 10684510
Krenz M Robbins J Impact of beta-myosin heavy chain expression on cardiac function during stress J Am Coll Cardiol 2004 44 2390 2397 15607403 10.1016/j.jacc.2004.09.044
Broberg CS Pantely GA Barber BJ Mack GK Lee K Thigpen T Davis LE Sahn D Hohimer AR Validation of the myocardial performance index by echocardiography in mice: a noninvasive measure of left ventricular function J Am Soc Echocardiogr 2003 16 814 823 12878990 10.1067/S0894-7317(03)00399-7
Nemoto S DeFreitas G Mann DL Carabello BA Effects of changes in left ventricular contractility on indexes of contractility in mice Am J Physiol Heart Circ Physiol 2002 283 H2504 H2510 12427596
Williams RV Lorenz JN Witt SA Hellard DT Khoury PR Kimball TR End-systolic stress-velocity and pressure-dimension relationships by transthoracic echocardiography in mice Am J Physiol 1998 274 H1828 H1835 9612396
Hoit BD Khan ZU Pawloski-Dahm CM Walsh RA In vivo determination of left ventricular wall stress-shortening relationship in normal mice Am J Physiol 1997 272 H1047 H1052 9124412
Tanaka N Dalton N Mao L Rockman HA Peterson KL Gottshall KR Hunter JJ Chien KR Ross J Jr Transthoracic echocardiography in models of cardiac disease in the mouse Circulation 1996 94 1109 1117 8790053
|
16250913
|
PMC1276802
|
CC BY
|
2021-01-04 16:38:32
|
no
|
Cardiovasc Ultrasound. 2005 Oct 26; 3:34
|
utf-8
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-34
|
oa_comm
|
==== Front
Cell Commun SignalCell communication and signaling : CCS1478-811XBioMed Central London 1478-811X-3-111620737210.1186/1478-811X-3-11ResearchEffect of connective tissue growth factor (CCN2/CTGF) on proliferation and differentiation of mouse periodontal ligament-derived cells Asano Masahiro [email protected] Satoshi [email protected] Tohru [email protected] Takashi [email protected] Tomoichiro [email protected] Gen [email protected] Kazumi [email protected] Tomosada [email protected] Yoji [email protected] Masaharu [email protected] Department of Biochemistry and Molecular Dentistry, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama, Japan2 Department of Periodontal Science, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama, Japan3 Department of Oral Functional Anatomy, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama, Japan2005 5 10 2005 3 11 11 22 6 2005 5 10 2005 Copyright © 2005 Asano et al; licensee BioMed Central Ltd.2005Asano et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
CCN2/CTGF is known to be involved in tooth germ development and periodontal tissue remodeling, as well as in mesenchymal tissue development and regeneration. In this present study, we investigated the roles of CCN2/CTGF in the proliferation and differentiation of periodontal ligament cells (murine periodontal ligament-derived cell line: MPL) in vitro.
Results
In cell cultures of MPL, the mRNA expression of the CCN2/CTGF gene was stronger in sparse cultures than in confluent ones and was significantly enhanced by TGF-β. The addition of recombinant CCN2/CTGF (rCCN2) to MPL cultures stimulated DNA synthesis and cell growth in a dose-dependent manner. Moreover, rCCN2 addition also enhanced the mRNA expression of alkaline phosphatase (ALPase), type I collagen, and periostin, the latter of which is considered to be a specific marker of the periosteum and periodontium; whereas it showed little effect on the mRNA expression of typical osteoblastic markers, e.g., osteopontin and osteocalcin. Finally, rCCN2/CTGF also stimulated ALPase activity and collagen synthesis.
Conclusion
These results taken together suggest important roles of CCN2/CTGF in the development and regeneration of periodontal tissue including the periodontal ligament.
==== Body
Background
Connective tissue growth factor (CCN2/CTGF) is a cysteine-rich protein with a molecular weight of 36–38 kDa [1-3] and belongs to the CCN family, which stands for CTGF, CEF10/Cyr61, and Nov [4]. This family now consists of 6 distinct members [1-3], including additional members that had been named WISP1/ELM1, WISP2/COP-1, and WISP3. These proteins retain structural similarity to one another, characterized by an N-terminal secretory signal, followed by 4 modules, with the exception of COP-1, which lacks the C-terminal module [1-4]. As such, CCN proteins share 30–50% over all amino acid sequence homology and 40–60% nucleotide sequence homology with each other [1-4]. The members of this family exhibit functional diversity, being involved in such cellular processes as adhesion, migration, proliferation and differentiation, and in biological processes such as angiogenesis and chondrogenesis [1-3].
CCN2/CTGF was discovered as a chemotactic and mitogenic factor for fibroblast-like cells in vitro and so was named accordingly [5]. It was also found in skin fibroblasts in vivo as a repairing growth factor [6]. Of note, transforming growth factor (TGF)-β was shown to act in tissue during normal wound repair [6,7] and fibrotic disorders of the skin [8,9], and in internal organs and tissues [10-12]. CCN2/CTGF was also shown to be closely linked to the expression of TGF-β in these situations [6].
CCN2/CTGF increased the expression of extracellular matrix molecules such as type I collagen, fibronectin, and integrin [13], and was overexpressed in fibroblasts in the dermis of patients with scleroderma [14] or other fibrotic disorders [15-18]. These findings suggest that CCN2/CTGF plays an important role in cell proliferation and matrix synthesis in connective tissue.
Critical roles of CCN2/CTGF in other mesenchymal tissues have been uncovered as well. Our previous studies revealed that ccn2/ctgf was highly expressed in hypertrophic chondrocytes in growth cartilage and human chondrocytic HCS-2/8 cells [19]. CCN2/CTGF actually stimulated the proliferation and differentiation of chondrocytes in vitro [20] and induced articular cartilage regeneration in vivo [21]. Importantly, CCN2/CTGF also promotes the proliferation, migration, and tube formation of vascular endothelial cells in vitro and angiogenesis in vivo [22,23]. Such biological effects of CCN2/CTGF were further confirmed by the abnormal phenotypic change in the growth plate of CCN2/CTGF knockout mice. Moreover, roles of CCN2/CTGF in the proliferation and differentiation of osteoblasts have been also indicated [24]. In a recent study, regulated expression of CCN2/CTGF in developing tooth germs was reported [25,26]. During odontogenesis, CTGF/CCN2 gene expression was clearly observed not only in the dental mesenchyme but also in the dental epithelium up to the stage of enamel secretion [25,26]. Therefore, CCN2/CTGF is now regarded as a common regulator of the development of the 2 hard tissues in vertebrates – tooth and bone.
The periodontal ligament is a fibrous tissue located between these 2 mineralized hard tissues, and plays an important role in anchoring tooth and alveolar bone and in maintaining their structural integrity; and it is also involved in the periodontal regeneration process. The ligament consists of a number of different cell types, such as fibroblasts, osteoblasts, osteoclasts, cementoblasts, and their precursors, and so on. In particular, fibroblasts in the periodontal ligament are considered to be multi-potential cells, or heterogeneous cell populations, which have the capacity to differentiate into other types of cells. In this respect, the periodontal ligament fibroblasts supposedly play a key role in the formation, maintenance, and regeneration of this connective tissue.
Therefore, in the present study, we focused on these periodontal ligament fibroblasts by evaluating the effects of CCN2/CTGF on the in vitro biological phenotypes of a mouse periodontal ligament cell line (MPL).
Results
Expression of ccn2/ctgf mRNA in different growth phases in MPL cell cultures
Initially, the expression of ccn2/ctgf mRNA in different phases of MPL growth was analyzed by in situ hybridization analysis. Positive signals were significantly observed in sparse cultures, but not in confluent ones. However, stronger mRNA expression was induced by the addition of TGF-β, even in cultures in the confluent state (Fig. 1).
Figure 1 Expression of CCN2/CTGF mRNA in sparse and confluent MPL cultures (in situ hybridization) with or without TGF-β stimulation. MPL cultures in the sparsity phase (A and B) and confluence phase (C and D) were treated with TGF-β (B and D) or PBS (control: A and C). The cells were then cultured for 12 hrs, fixed, and hybridized with an antisense riboprobe for ccn2/ctgf mRNA. CCN2/CTGF mRNA was distinctly expressed in the cells in the sparse (A), but not in those in the confluent (C) state. However, the addition of TGF-β enhanced the level of CCN2/CTGF mRNA strongly, not only in the sparse cultures (B), but also in the confluent ones (D).
Since strong expression of ccn2/ctgf mRNA was detected in sparse cultures of MPL by in situ hybridization analysis, we subsequently evaluated the expression level in different growth phases by semi-quantitative RT-PCR. The steady-state mRNA level in the sparse cultures of MPL cells was approximately 3 times higher than that in the confluent ones (Fig. 2). These results suggest the involvement of CCN2/CTGF and TGF-β in MPL cell proliferation on demand during the regeneration of periodontal ligament.
Figure 2 RT-PCR analysis of the expression of CCN2/CTGF mRNA in sparse (S) and confluent (C) MPL cultures. Total RNA was purified from MPL cultures under sparse and confluent conditions, and 1 μg in each sample was reverse transcribed and used for PCR. The expression of CCN2/CTGF mRNA was stronger under the sparse than the confluent condition as evaluated in the exponential phase of the amplification (25 cycles). Levels of GAPDH mRNA, used as an internal control, were almost the same among these samples.
Effects of rCCN2/CTGF on DNA synthesis and cell proliferation
Based on the findings above, we next evaluated the effects of exogenous rCCN2/CTGF on MPL cell growth. As shown in Fig. 3A, rCCN2/CTGF stimulated DNA synthesis in the cells in a dose-dependent manner, as evaluated by [3H] thymidine incorporation. Significant effects were observed at 1 ng/ml and higher concentrations of rCCN2/CTGF. Similarly, the addition of rCCN2/CTGF increased the cell number under standard cell culture conditions after 3 days of treatment (Fig. 3B). The maximum effect produced a 1.2-fold increase over the control.
Figure 3 (A) Effect of rCCN2/CTGF on DNA synthesis in MPL cells. MPL cells were inoculated at a density of 5 × 103/well in 96-multi-well plates and cultured in α-MEM with 10% FBS for 2 days. Then, the culture medium was changed to α-MEM with 0.5% FBS, and the cells were incubated with various concentrations of rCCN2/CTGF for 24 hrs and labeled with [3H] thymidine for the last 4 hrs. The incorporated radioactivity was determined by a liquid scintillation counter. Values represent the averages ± SD. Data were computed with the results of 2 independent series of experiments with multiple sample numbers. Asterisks denote statistically significant difference from the vehicle-treated control. (B) Effect of rCCN2/CTGF on the proliferation of MPL cells. MPL cells were inoculated at a density of 2 × 103/well in 96-multi-well plates and cultured in α-MEM with 10% FBS for 24 hrs. Then, the medium was changed to α-MEM with 1% FBS, and the cells were cultured with various concentrations of rCCN2/CTGF for 3 days. The cell number was computed by using Tetra-Color One as specified by the manufacturer. Points and bars represent the averages and SD, respectively. Asterisks indicate significant differences between vehicle-treated control and rCCN2/CTGF-treated samples at the significance level of **p < 0.01 or *p < 0.05.
Effects of rCCN2/CTGF on expression of genes related to the differentiated phenotypes of periodontal ligament (PDL) cells
In order to examine the effects of rCCN2/CTGF on gene expression of known mesenchymal differentiation markers, we conducted RT-PCR analysis. The mRNA expression of ALPase and type I collagen, which are considered to be phenotypic markers of periodontal ligament (PDL) cells as well as of osteoblasts, was up-regulated 24 hrs after the addition of 100 ng/ml of rCCN2/CTGF (ALPase: about 2.8-fold, type I collagen: about 1.5-fold; Fig. 4). Moreover, the expression of periostin, which is a marker of periosteal cells as well as of PDL cells, was also up-regulated approximately 2.5-fold 24 hrs after the addition of rCCN2/CTGF. On the other hand, the addition of rCCN2/CTGF had little effect on the mRNA expression of bone-specific phenotypic marker genes encoding osteopontin and osteocalcin (data not shown). The amount of GAPDH transcript, used as an internal control, was almost the same with or without rCCN2/CTGF.
Figure 4 Effects of rCCN2/CTGF on mRNA expression of osteoblast and PDL-related markers in MPL cells. Confluent cultures of MPL cells were treated with rCCN2/CTGF (100 ng/ml, closed bars) or vehicle (dotted bars) for 24 hrs. Total RNA was purified, and 1 μg of each sample was reverse-transcribed and used for PCR. Alkaline phosphatase (ALPase), type I collagen, and periostin mRNAs were shown to be up-regulated clearly by the addition of rCCN2/CTGF under exponential conditions (25, 30, and 25 cycles, respectively) for amplification. The expression of GAPDH, used for an internal control, was almost the same among these samples.
Effect of rCCN2/CTGF on ALPase Activity displayed by MPL cells
In addition to the examination of the mRNA level, we also investigated the effect of rCCN2/CTGF on ALPase activity displayed by MPL cells. As shown in Fig. 5, the enzyme activity increased in a dose-dependent manner 48 hrs after the addition of rCCN2/CTGF. A significant effect was observed at a dose as low as 1 ng/ml, reaching a plateau at 100 ng/ml.
Figure 5 Effects of rCCN2/CTGF on ALPase activity from MPL cells. When MPL cells reached confluence, the culture medium was replaced with α-MEM containing 1% FBS, and the cells were then cultured with various concentrations of rCCN2/CTGF. After 48 hrs, the cell layers were collected, and ALPase activity was determined as described in "Materials and Methods." Values represent the averages ± SD of 3 separate experiments. Asterisks denote statistically significant differences from the vehicle-treated control at the significance level of *p < 0.01.
Effect of rCCN2/CTGF on collagen synthesis by MPL cells
Since the predominant component of the extracellular matrix of the periodontal ligament is type I collagen, we further examined the effect of rCCN2/CTGF on collagen synthesis by MPL cells. As shown in Fig. 6, both total protein synthesis and collagen synthesis were increased by the addition of rCCN2/CTGF in a dose-dependent manner. However, the collagen-specific increase in [3H]-proline incorporation as estimated by collagenase digestion was significantly higher than that of the total increase. Therefore, CCN2/CTGF was indicated to promote the expression of PDL cell phenotypes, as well as overall protein synthesis, which might be the outcome of stimulated cell proliferation.
Figure 6 Effect of rCCN2/CTGF on the synthesis of collagen and total protein in MPL cells. Various concentrations of rCCN2/CTGF were added to confluent cultures of MPL cells, and the cells were cultured for 12 hrs. Then, the cells were labeled with [3H] proline for another 12 hrs, after which their cell layers were collected for analysis. The radioactivities of [3H] proline incorporated into total nascent proteins and collagenase-digestable portions were measured as described in "Materials and Methods." Slashed and closed boxes indicate the collagen and total protein synthesis, respectively. Values represent the mean average ± SD (n = 2). Asterisks denote statistically significant differences (**p < 0.01, *p < 0.05) from the vehicle-treated control, except the double asterisks specifying a significant difference between the stimulation levels of total protein and collagen synthesis (as indicated by the bracket). Data were computed with the results of 2 independent series of experiments with multiple sample numbers.
Discussion
Active roles of CCN2/CTGF in tooth development and remodeling of periodontal tissues have been suggested by several reports that described the stage-specific expression of ccn2/ctgf with a particular spatial distribution in corresponding tissues [25,26]. In view of these reports, CCN2/CTGF is suggested to have some physiological function, playing a role in the organization and maintenance of periodontal tissues.
TGF-β, a major stimulator of the ccn2/ctgf gene, is expressed broadly in rat molar periodontal tissue in late embryos and after birth [36,37]. CCN2/CTGF expression was induced by TGF-β, but not by PDGF, EGF, or bFGF in cultured human foreskin fibroblasts [6]; and in a wound healing model, TGF-β and CCN2/CTGF mRNAs were coordinately expressed [6]. Therefore, as CCN2/CTGF has been shown to be closely linked to the expression of TGF-β, the CCN2/CTGF gene might be induced by TGF-β in periodontal tissue. Indeed, interaction between dental epithelium and mesenchyme was required for CCN2/CTGF induction, which could be substituted by exogenous TGF-β [25]. In our present study, TGF-β actually induced ccn2/ctgf expression in MPL cells and might be involved in the growth phase-dependent ccn2/ctgf expression. Therefore, interplay of TGF-β and CCN2/CTGF during tooth development is indicated as well.
The periodontal ligament lies between 2 hard tissues, alveolar bone and cementum, and consists of various kinds of cells. Owing to this specific situation, the major constituent cells – periodontal ligament fibroblasts – have unique characteristics unlike other fibroblasts. Periodontal ligament fibroblasts express calcification-related genes, and show ALPase activity. Although their functional significance is unclear, these phenotypes are thought to be critically important for fulfilling the specific missions of PDL cells.
As was shown in Fig. 1, positive signals for CCN2/CTGF mRNA in sparse MPL cultures were distinctly detected by in situ hybridization analysis, but were not clearly seen in confluent cells. We next evaluated the proliferative effect of rCCN2/CTGF on MPL cells. Regarding DNA synthesis, the incorporation of [3H] thymidine was increased in a dose-dependent manner by rCCN2/CTGF (Fig. 3A), and cell growth was stimulated in the same manner (Fig. 3B). These results suggest that CCN2/CTGF acts as an autocrine and/or paracrine cell proliferation-inducing factor. In addition, in our experiments, the stimulatory effect of rCCN2/CTGF on DNA synthesis was lost when MPL cells were cultured with α-MEM supplemented with 10% FBS, suggesting that more potent growth factors than rCCN2/CTGF are present in FBS.
In our RT-PCR investigation, the mRNA expression of ALPase and of type I collagen was up-regulated 24 hrs after the stimulation with rCCN2/CTGF (Fig. 4). In addition, rCCN2/CTGF stimulated the ALPase activity (Fig. 5). In the course of the calcification of hard tissues, including that of MPL cells in cell culture, the ALPase level is first up-regulated, the expression levels of osteopontin and osteocalcin are then up-regulated, and finally calcification occurs [38]. However, the gene expression of osteopontin and osteocalcin was not up regulated by CCN2/CTGF. Since we previously evaluated the cell biological effects of CCN2/CTGF on osteoblastic cells as well, the effects of CCN2/CTGF on the differentiation markers are comparatively summarized in Table 1. Such restricted expression of calcification-related genes may represent the specific phenotype of PDL cells, which discriminate them from osteoblasts. It should be also noted that the expression of osf-2/cbfa1, a master gene of osteoblastic differentiation, was not detectable in MPL cells. The observed effects of CCN2/CTGF to promote the proper differentiated phenotype of each type of the cells represent the characteristics of CCN2/CTGF as a modulator of extracellular signalling network.
Table 1 Expression profiles of the phenotypic markers in MPL versus osteoblastic cells
Cells Basal osf-2/cbfa1 Induction by CCN2/CTGF
ALP opn ocn periostin
MPL - + - - +
MC3T3-E1* + + + + N.D.
*Nishida et al. [24]
As for periodontal ligament fibroblasts, there is no clear differentiation marker that is specifically characteristic of this type of cell [39]. Thus, osteoblast-related markers have usually been used, although the ligament is an elastic tissue. However, in 1995, periostin cDNA was cloned from the MC3T3-E1 cell line by Takeshita et.al. [40]. They reported its stimulatory effect on cell-cell attachment and migration of MC3T3-E1 cells. Later, its limited expression was detected in both the periosteum and periodontium of teeth [41]. It was also reported that its strong expression in the periosteum was due to that in putative pre-osteoblasts and that this expression decreased as these pre-osteoblasts underwent maturation [40], indicating that periostin gene expression is a differentiation marker for immature, osteoblast progenitor cells. Gene expression of periostin is also detected in periodontium of teeth [41], which is also an elastic membranous tissue close to calcified-tissues, as is periostium, thus indicating that periostin gene expression is a marker of not only the periosteum but also of the periodontium. In other words, periostin is a marker of cells that are present close to calcified tissues and have potential to calcify but retain characteristics of an elastic, membranous tissue. Therefore, our finding that CCN2/CTGF potently stimulated periostin gene expression suggests that CCN2/CTGF stimulates the differentiation of these cells that constitute elastic, membranous tissues in between or close to calcified-tissues. Periostin is thought to mediate tissue remodeling by modifying the constitution of collagen fibers, while CCN2/CTGF induces collagen synthesis (Fig. 6) and binds to gelatin in vitro [21]. As such, periostin itself may collaborate with CCN2/CTGF molecule in the remodeling of periodontal ligament.
As well as at the mRNA expression level, rCCN2/CTGF also stimulated collagen synthesis (Fig. 6). The turnover of collagen in the periodontal ligament is believed to be controlled by the balance between collagen synthesis and degradation. In this respect, up-regulation of collagen synthesis by rCCN2/CTGF may be accompanied by the dynamic control of gene expression of matrix metalloproteinases (MMPs) or tissue inhibitors of matrix metalloprotainases (TIMPs) for the remodeling of periodontal ligament [42]. It should be noted that induction of particular MMPs by CCN2/CTGF was reported to occur in vascular endothelial cells. Therefore, CCN2/CTGF may play a central role in promoting the remodeling of periodontal ligament through direct and indirect actions.
According to a previous report [25], CCN2/CTGF expression was observed in dental laminas, invaginating epithelium, and condensing mesenchyme at the bud stage. Afterwards, strong expression was observed in the enamel knot and preameloblasts; and CCN2/CTGF molecules were also detected in stratum intermedium and underlying dental mesenchyme, as well as in the dental epithelium. Additionally, during experimental tooth movement, CCN2/CTGF was highly expressed in osteoblasts around the periodontal ligament [42]. In this study, we confirmed the ccn2/ctgf expression and cell biological effects of CCN2/CTGF in vitro, by using MPL cell cultures. Overall, our results support the in vivo findings mentioned above, together suggesting critical roles of CCN2/CTGF in the development and remodeling of dental tissues.
Conclusion
In this study, expression of CCN2/CTGF in MPL cells was confirmed in vitro. Utilizing this system, we uncovered the potential of CCN2/CTGF to promote the growth and differentiation of MPL cells. These findings indicate the critical roles of CCN2/CTGF in periodontal tissue development and strongly suggest the utility of CCN2/CTGF in periodontal tissue regeneration, as already proven in articular cartilage.
Methods
Cells and Cell culture
The mouse periodontal ligament cell line (MPL) was established from periodontal ligament explants from extracted mandibular molars of a BALB/c mouse [27] and maintained as described [28]. The cells were grown in alpha-modified Eagle's medium (α-MEM; Nissui, Tokyo, Japan) supplemented with 10% heat-inactivated fetal bovine serum (FBS; JRH Biosciences, Lenexa, KS, USA), and kanamycin (100 μg/ml) at 37°C in a humidified atmosphere of 5% CO2 in air.
Production of recombinant connective tissue growth factor (rCCN2/CTGF)
Human ccn2/ctgf cDNA containing the entire coding region was inserted into the plasmid expression vector pcDNA3.1(-) (Invitrogen, Carlsbad, CA, USA), and the vector was used to transform HeLa cells. The conditioned medium of the HeLa cells was concentrated and purified by heparin-affinity column and anti-CTGF/Hcs24 affinity-column chromatography. Purity was determined by immunoblotting and silver staining of SDS-PAGE gels, and highly purified fractions were used for experiments. The biological activity of rCCN2/CTGF was almost the same as that of the factor purified from a human chondrosarcoma cell line (HCS-2/8), as determined by their biological activity toward chondrocytes.
Estimation of DNA synthesis
The mitogenic effect of rCCN2/CTGF was assessed by measuring [3H] thymidine (Amersham Bioscience Corp., Piscataway, NJ; Specific activity 74 TBq/nmol) incorporation into MPL cells [29,30]. The cells were labeled with [3H] thymidine (4.8 MBq/ml in DMEM) for 4 hrs, then washed with phosphate-buffered saline (PBS), and collected in 100 μl of PBS containing 0.25% trypsin (w/v) and 0.02% EDTA (w/v). They were next harvested onto glass-fiber paper filters (WALLAC, Turku, Finland) with a semi-automatic microharvester (TOMTEC, Yamaguchi, Japan), washed with PBS, and treated subsequently with 5% trichloroacetic acid (TCA). The radioactivity was determined with a Micro Beta PLUS (WALLAC).
Estimation of cell proliferation
The effect of rCCN2/CTGF on cell proliferation was assessed by using Tetracolor-One (Seikagaku Corp., Tokyo, Japan) as instructed by the manufacturer.
Northern blotting
Ten micrograms of total RNA was electrophoresed in a formaldehyde agarose gel and subsequently blotted onto a nylon membrane. For hybridization probes, CCN2/CTGF and CCN1/Cyr61 cDNA fragments were prepared by RT-PCR with pairs of specific primers. Primers specific for CCN2/CTGF [27] and CCN1/Cyr61 [28] were described previously. These PCR products were radiolabeled and used for hybridization as described earlier [20].
Reverse transcription-polymerase chain reaction (RT-PCR)
By using ISOGEN (Nippon Gene, Tokyo, Japan)[31], we extracted total cellular RNA from MPL cell cultures treated with rCCN2/CTGF (100 ng/ml) or vehicle; and before RT-PCR, the RNA was treated with RNase-free DNase I. Total RNA (1 μg) was reverse-transcribed by using an RNA PCR kit (Perkin Elmer, Branchburg, NJ), and the cDNA was used as a template to amplify fibroblast or osteoblast-related marker genes and murine CCN2/CTGF and GAPDH genes by PCR using Taq GOLD polymerase (Perkin-Elmer). Primer sequences employed and PCR conditions were as follow:
murine glyceraldehyde 3-phosphate dehydrogenase (GAPDH), sense 5'-CACCATGGAGAAGGCCGGGG-3' and antisense 5'-GACGGACACATTGGGGGTAG-3'(418 bp); type I collagen, sense 5'-TCTCCACTCTTCTAGGTCCT-3' and antisense 5'-TTGGGTCATTTCCACATGC-3'(250 bp); osteopontin, sense 5'-ACACTTTCACTCCAATCGTCC-3' and antisense 5'-TGCCCTTTCCGTTGTTGTCC-3'(239 bp); osteocalcin, sense 5'-TCTGACAAACCTTCATGTCC-3' and antisense 5'-AAATAGTGATACCGTAGATGCG-3'(198 bp); periostin, sense 5'-TGGTTCCTCTCCTGCCCTTA-3' and antisense 5'-ACCATGCCGTGTTTCAGGTC-3'(556 bp); ALPase, sense 5'-GCCCTCTCCAAGACATATA-3' and antisense 5'-CCATGATCACGTCGATATCC-3'(372 bp); Osf-2/Cbfa-1, sense 5'-GAGGGCACAAGTTCTATCTGGA-3' and antisense 5'-GGTGGTCCGCGATGATGTC-3' (385 bp); murine CCN2/CTGF, sense 5'-GGTAAGGTCCGATTCCTACCAGG-3' and antisense 5'-CTAGAAAGGTGCAAACATGTAAC-3'(120 bp).
After pre-denaturation for 7 min at 94°C, the chain reaction was repeated for 25 cycles to maintain exponential conditions for amplifications (30 cycles for murine osteopontin). Each cycle consisted of denaturation at 94°C for 1 min, with annealing and polymerization at 60°C for 2 min. Ten-micro liter samples were analyzed on 2% agarose gels and stained with ethidium bromide along with a 100 bp-ladder molecular size marker.
Alkaline phosphatase (ALPase) activity
The effect of rCCN2/CTGF on ALPase activity was assayed according to the procedure of Majeska et.al. [32]. The cells were homogenized in 0.5 ml of 0.9% NaCl and 0.2% Triton X-100 with a Polytron at 4°C, and centrifuged for 15 min at 12,000 × g. ALPase activity in the supernatant was measured by using p-nitro phenyl phosphate (p-NP) as a substrate. The supernatant was mixed with 0.5 M Tris-HCL buffer (pH9.0) containing 0.5 mM p-NP and 0.5 mM MgCl2. The sample was then incubated at 37°C for 30 min, and the reaction was stopped by the addition of a 0.25 volume of 1N NaOH. Hydrolysis of p-NP was monitored as the change in absorbance at 415 nm with a spectrometer (Amersham Bioscience Corp.). The protein concentration was determined by using the BCA protein assay system (Pierce, Rockford, IL). The activity was defined in units, 1 unit being the enzyme activity hydrolyzing 1 nmol of p-NP per 1 mg protein in 30 min.
Collagen synthesis
The effect of rCCN2/CTGF on collagen synthesis was assayed by measuring the incorporation of [3H] proline (Amersham, 1.7 TBq/nmol) in collagenase-digestable proteins [33-35]. When MPL cells reached confluence, they were labeled with [3H] proline (37 MBq/ml) for 12 hrs and then collected into 1 ml of 50 mM Tris (pH7.2) containing 0.2% Triton X-100 and 1 mM phenylmethylsulfonyl fluoride (PMSF). Half of the solubilized homogenate was digested with purified bacterial collagenase for 4 hrs at 37°C, and the other half was incubated with the vehicle (control). The radioactivity in the suspended protein was measured with a liquid scintillation counter.
In situ hybridization
For generation of the antisense probe, a 120-bp fragment of murine ccn2/ctgf cDNA corresponding to a non-coding region was subcloned into the pGEM-T plasmid (Promega, Madison, WI, USA), and riboprobes were synthesized by using a DIG RNA labelling kit (Roche, Mannheim, Germany) or 35S-UTP(30TBq/nmol, Amersham Biosciences Corp.) following the manufacturer's instructions. Sense probes were used as negative controls. After in vitro transcription, the labeling reaction mixture was treated with DNase I, and the probes were precipitated with ethanol, and then dissolved in 0.1% diethyl pyrocarbonate (DEPC)-treated double distilled water containing 0.01% dithiothreitol.
MPL cells cultured on glass slides were washed with PBS and fixed with 4% paraformaldehyde (PFA)- 0.1% sodium-cacodylate at 4°C for 2 hrs. The samples were washed with PBS, and digestion with 1 μg/ml proteinase K was then performed at room temperature for 10 min. After having been washed with PBS, the samples were treated with 2 mg/ml of glycine-PBS at room temperature for 10 min, and immersed in 0.1 M triethanolamine-HCl (pH 8.0). Thereafter acetic acid-anhydride (0.25% concentration) was added dropwise over a 5-min period, and incubation was then continued for another 15 min. The slides were rinsed with 4 × standard saline-citrate (SSC) for 10 min twice, and immersed in pre-hybridization buffer (3 × SSC and 50% formamide) at 70°C for 30 min for blocking.
Hybridization was performed with a hybridization mixture (50% formamide, 3 × SSC, 10% dextran sulfate, 1 μg/ml of tRNA, 1 μg/ml of sonicated salmon sperm DNA, and 1 μg/ml BSA) containing 0.4 μg/ml riboprobes at 55°C for 16 hrs in a humidified box. The slides were washed 3 times for 20 min each time with 2 × SSC and 50% formamide at 55°C, while being shaken at 90 strokes/min to remove the excess riboprobe; and then they were incubated at 37°C for 30 min with 20 μg/ml of RNase A in NTE buffer [0.5 mM NaCl, 10 mM Tris-HCl (pH 8.0) and 1 mM EDTA]. After having been rinsed in NTE buffer at 37°C for 5 min, the slides were washed 3 times for 20 min each time with 0.5 × SSC at 37°C and incubated at room temperature for 30 min in TBS buffer [100 mM Tris-HCl (pH 9.5), 100 mM NaCl and 150 mM NaCl] containing 1% normal goat serum. Next, the sections were incubated with 1/500-diluted alkaline phosphatase (AP)-conjugated sheep anti-digoxigenin antibody at room temperature for 1 hr. To visualize digoxigenin-labeled riboprobe-tissue mRNA complexes, we incubated the sections in a freshly prepared solution of nitro blue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3-indolylphosphate (BCIP) in an alkaline buffer [100 mM Tris-HCl (pH9.5), 100 mM NaCl and 50 mM MgCl2] containing 1 mM levamisole in a dark room at room temperature. Finally, the slides were washed in distilled water 3 times and mounted in aqueous mounting material.
Statistical methods
The statistical difference between treated or control samples was assessed with Student's t-test. All experiments were repeated at least twice, and similar results were obtained.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MA performed molecular biological studies and prepared most of the data. SK rearranged and drafted the manuscript. T Nakanishi arranged the constitution of the work. T Nishida designed a significant portion of the experiments. TY and TS established and participated in the in situ hybridization analysis. GY and KO collaborated in performing most of the cell biological analyses. YM provided general support. MT conceived and designed the study and drafted the manuscript.
Acknowledgements
This work was supported in part by Grants-in-Aid for Scientific Research (C) (to T.N.) and (S) (to M.T.) and Exploratory Research (to M.T.) from the Ministry of Education, Science, Sports, and Culture of Japan and Japan Society for the Promotion of Science (JSPS), by Grants-in-Aid from the Ministry of Health and Welfare (to T.N., M.T.), and by grants from the Naito Foundation (to M.T.), the Sumitomo Foundation (to M.T.), and the Foundation for Growth Science in Japan (to M.T.). The authors wish to thank Dr. Takako Hattori for helpful suggestions, and Ms. Yuki Nonami and Kazumi Ohyama for their technical and secretarial assistance.
==== Refs
Takigawa M CTGF/Hcs24 as a multifunctional growth factor for fibroblasts, chondrocytes and vascular endothelial cells Drug News Perspect 2003 16 11 21 12682668 10.1358/dnp.2003.16.1.829302
Takigawa M Nakanishi T Kubota S Nishida T The role of CTGF/Hcs24/ecogenin in skeletal growth control J Cell Physiol 2003 194 256 266 12548546 10.1002/jcp.10206
Brigstock DR The connective tissue growth factor/cysteine-rich 61/nephroblastoma over expressed (CCN) family Endocr Rev 1999 20 189 206 10204117 10.1210/er.20.2.189
Perbal B CCN proteins: multifunctional signalling regulators Lancet 2004 363 62 64 14723997 10.1016/S0140-6736(03)15172-0
Bradham DM Igarashi A Potter RL Grotendorst GR Connective tissue growth factor: a cysteine-rich mitogen secreted by human vascular endothelial cells is related to the SRC-induced immediate early gene product CEF-10 J Cell Biol 1991 114 1285 1294 1654338 10.1083/jcb.114.6.1285
Igarashi A Okochi H Bradham DM Grotendorst GR Regulation of connective tissue growth factor gene expression in human skin fibroblasts and during wound repair Mol Biol Cell 1993 4 637 645 8374172
Levine JH Noses HL Gold LI Nanney LB Spatial and temporal patterns of immunoreactive transforming growth factor beta1, beta2, beta3 during excisional wound repair Am J Pathol 1993 143 368 380 8342593
Kulozik M Hogg A Lannkat-Buttgereit B Krieg T Co-localization of TGF-β and type I collagen mRNA in tissue sections of patients with systemic sclerosis J Clin Invest 1990 85 917 922 1697606
Peltonen J Kahari L Jaakkola S Kahari V-M Varga J Uitto J Jimenez SA Evaluation of TGF-β and type I procollagen gene expression in fibrotic skin diseases by in situ hybridization J Invest Dermatol 1990 94 365 371 2307855 10.1111/1523-1747.ep12874491
Nagy P Schaff Z Lapis K Immunohistochemical detection of TGF-β1 in fibrotic liver disease Hepatology 1991 14 269 273 1713566 10.1016/0270-9139(91)91414-V
Kagami S Border WA Ruoslahti E Noble NA Coordinated expression of beta 1 integrins and transforming growth factor-beta induced matrix proteins in glomerulonephritis Lab Invest 1993 69 68 76 8331901
Bahadori L Milider J Gold L Botney M Active macrophage associated TGF-β co-localizes with type I procollagen gene expression in atherosclerotic human pulmonary arteries Am J Pathol 1995 146 1140 1149 7747808
Frazier KS Williams S Kothepelli D Klapper H Groendorst GR Stimulation of fibroblast cell growth, matrix production, and granulation tissue formation by connective tissue growth factor J Invest Dermatol 1996 107 404 411 8751978 10.1111/1523-1747.ep12363389
Igarashi A Nashiro K Kikuchi K Sato S Ihn H Grotendorst GR Takehara K Significant correlation between connective tissue growth factor gene expression and skin sclerosis in tissue sections from patients with systemic sclerosis J Invest Dermatol 1995 105 280 284 7636314 10.1111/1523-1747.ep12318465
Igarashi A Nashiro K Kikuchi K Sato S Ihn H Fujimoto M Grotendorst GR Takehara K Connective tissue growth factor gene expression in tissue sections from localized scleroderma, keloid, and other fibrotic skin disorders J Invest Dermatol 1996 106 729 733 8618012 10.1111/1523-1747.ep12345771
Oemer BS Werner A Garnier JM Do DD Godoy N Nauck M Marz W Rupp J Pech M Luscher TF Human connective tissue growth factor is expressed in advanced athelosclerotic lesions Circulation 1997 95 831 839 9054739
Igarahi A Hayashi N Nashiro K Takehara K Differential expression of connective tissue growth factor gene in cutaneous fibrohistiocytic and vascular tumors J Cutan Pathol 1998 25 143 148 9550312
Ito Y Aten J Bende RJ Oemer BS Rabelink TJ Weening JJ Goldschmeding R Expression of connective tissue growth factor in human renal fibrosis Kidney Int 1998 53 853 861 9551391 10.1111/j.1523-1755.1998.00820.x
Nakanishi T Kimura Y Tamura T Ichikawa H Yamaai Y Sugimoto T Takigawa M Cloning of mRNA preferentially expressed in chondrocytes by differential display-PCR from a human chondrocytic cell line that is identical with connective tissue growth factor (CTGF) mRNA Biochem Biophys Res Commun 1997 234 206 210 9168990 10.1006/bbrc.1997.6528
Nakanishi T Nishida T Shimo T Kobayashi K Kubo T Tamatani T Tezuka K Takigawa M Effects of CTGF/Hcs24, a product of a hypertrophic chondrocyte-specific gene, on the proliferation and differentiation of chondrocytes in culture Endocrinology 2000 141 263 273 10.1210/en.141.1.264
Nishida T Kubota S Kojima S Kuboki T Nakao K Kushibiki T Tabata Y Takigawa M Regeneration of defects in articular cartilage in rat knee joints by CCN2 (connective tissue growth factor) J Bone Miner Res 2004 19 1308 1319 15231019
Shimo T Nakanishi T Kimura Y Nishida T Ishizeki K Matsumura T Takigawa M Inhibition of endogenous expression of connective tissue growth factor by its antisense oligonucleotide and antisense RNA suppresses proliferation and migration of vascular endothelial cells J Biochem 1998 124 130 140 9644255
Shimo T Nakanishi T Nishida T Asano M Kanyama M Kuboki T Tamatani T Tezuka K Takemura M Matsumura T Takigawa M Connective tissue growth factor induces the proliferation, migration, and tube formation of vascular endothelial cells in vitro, and angiogenesis in vivo J Biochem 1999 126 137 145 10393331
Nishida T Nakanishi T Asano M Shimo T Takigawa M Effects of CTGF/Hcs24, a hypertrophic chondrocyte-specific gene product, on the proliferation and differentiation of osteoblastic cells in vitro J Cell Physiol 2000 184 197 206 10867644 10.1002/1097-4652(200008)184:2<197::AID-JCP7>3.0.CO;2-R
Shimo T Wu Billings PC Piddington R Rosenbloom J Pacifici M Koyama E Expression, gene regulation, and roles of Fisp12/CTGF in developing tooth germs Dev Dyn 2002 224 267 278 12112457 10.1002/dvdy.10109
Yamaai T Nakanishi T Asano M Nawachi K Yosimichi G Ohyama K Komori T Sugimoto T Takigawa M Gene expression of connective tissue growth factor (CTGF/CCN2) in calcifying tissue of normal and cbfa-1-null mutant in late stage of mouse embryonic development J Bone Miner Metab 2005 15981023
Tasaka S Establishment from mouse periodontal ligament of an immortalized cell line (MPDL-22) expressing physiological differentiation phenotype J Osaka Univ Dent Sch 1996 41 118 135
Tsuboi Y Nakanishi T Takano-Yamamoto T Miyamoto M Yamashiro T Takigawa M Mitogenic effects of neurotrophins on a periodontal ligament cell line J Dent Res 2001 80 881 886 11379889
Fasler S Aversa G Terr A Thestrup-Pedersen K de Vries J Yssel H Peptide-induced allergy in allergen-specific human Th2 cells results in lack of cytokine production and B cell help for IgE synthesis. Reversal by IL-2, not by Il-4 or IL-13 J Immunol 1995 155 4199 4206 7594575
Tortorella C Schulze-Koops H Thomas R Splawski JB Davis LS Picker LJ Lipsky PE Expression of CD45RB and CD27 identifies subsets of CD4+ memory T cells with different capacities to induce B cell differentiation J Immunol 1995 155 149 162 7541412
Chomczynski P Sacchi N Single-step method of RNA isolation by acid guanidium thiocyanate-phenol-chloroform extraction Anal Biochem 1987 162 156 159 2440339 10.1016/0003-2697(87)90021-2
Majeska RJ Rodan GA Alkaline phosphatase inhibition by parathyroid hormone and isoproterenol in a clonal rat osteosarcama cell line. Possible mediation by cyclic AMP Calcif Tissue Int 1982 34 59 66 6279255
Peterkofsky B diegelmann RF Use of a mixture of proteinase-free collagenases for the specific assay of radioactive collagen in the presence of other proteins Biochemistry 1971 6 988 994 4323854 10.1021/bi00782a009
Diegelmann RF Peterkofsky B Collagen biosynthesis during connective tissue development in chick embryo Dev Biol 1972 28 443 453 4339519 10.1016/0012-1606(72)90028-0
Kream BE Tetradis S LaFrancis D Fall PM Feyen JH Raisz LG Modulation of the effects of glucocorticoids on collagen synthesis in fetal rat calvariae by insulin-like growth factor binding protein-2 J Bone Miner Res 1997 12 889 895 9169346
D'Souza RN Happonen RP Flanders KC Butler WT Histochemical localization of transforming growth factor-β1 in developing rat molars using antibodies to different epitopes J Biol Buccale 1990 18 299 306 1710211
Gao J Symons AL Bartold PM Expression of transforming growth factor-beta1(TGF-β1) in the developing periodontium of rats J Dent Res 1998 77 1708 1716 9759668
Aubin JE Triffitt JT Bilezikian JP, Raisz LG, Rodan GA Mesenchymal stem cells and osteoblast differentiation Principals of Bone Biology 2002 2 Academic Press New York 59 81
Kasugai S Todescan R JrNagata T Yao KL Butler WT Sodek JJ Expression of bone matrix proteins associated with mineralized tissue formation by adult rat bone marrow cells in vitro: inductive effects of dexamethasone on the osteoblastic phenotype J Cell Physiol 1991 147 111 120 2037618 10.1002/jcp.1041470115
Takeshita S Kikuno R Tezuka K Amann E Osteoblastic-specific factor 2: cloning of a putative bone adhesion protein with homology with the insect protein fasciclin I Biochem J 1993 294 271 278 8363580
Horiuchi K Amizuka N Takeshita S Takamatsu H Katsuura M Ozawa H Toyama Y Bonewald LF Kudo A Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor β J Bone Miner Res 1999 14 1239 1249 10404027
Yamashiro T Fukunaga T Kobashi N Kamioka H Nakanishi T Takigawa M Takano-Yamamoto T Mechanical stimulation induces CTGF expression in rat osteocytes J Dent Res 2001 80 461 465 11332533
|
16207372
|
PMC1276803
|
CC BY
|
2021-01-04 16:39:15
|
no
|
Cell Commun Signal. 2005 Oct 5; 3:11
|
utf-8
|
Cell Commun Signal
| 2,005 |
10.1186/1478-811X-3-11
|
oa_comm
|
==== Front
Clin Pract Epidemiol Ment HealthClinical Practice and Epidemiology in Mental Health : CP & EMH1745-0179BioMed Central 1745-0179-1-201621265610.1186/1745-0179-1-20ReviewClinical use of coping in affective disorder, a critical review of the literature Christensen Maj Vinberg [email protected] Lars Vedel [email protected] Department of Psychiatry, Rigshospitalet, University Hospital of Copenhagen, Blegdamsvej 9, DK-2100 Copenhagen, Denmark2005 7 10 2005 1 20 20 20 7 2005 7 10 2005 Copyright ©2005 Christensen and Kessing; licensee BioMed Central Ltd.2005Christensen and Kessing; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Background
The relationship between life stressors, coping and affective disorder is interesting when predicting onset of a affective disorder and relapse of mood episodes.
Methods
A litteratur review of cross-sectional and longitudinal studies concerning coping and affective disorder in adults including a Medline and Embase search was conducted.
Results
11 cross-sectional studies and 17 longitudinal studies concerning affective disorder and coping were found, among these, two studies include patients with bipolar disorder exclusively. Only four studies elucidate whether emotion-oriented and/or avoidance coping styles are associated with a higher risk of developing affective disorder, so this hypothesis remains unclear. Most studies shows that emotion-oriented and avoidance coping strategies are associated with relapse of depressive episodes. Conversely, problem-focused and task-oriented coping seem to be associated with a good outcome.
Conclusion
There is a gap between coping theory and clinical use of coping and the clinical relevance of coping is, though promising, still unclear. In future research it is recommended to concentrate on development of a semi-structured interview combining coping style, life events and personality traits.
Affective Disorder, Bipolar DisorderDepressive Disorder, Manic-depressive Illness and Coping.
==== Body
Introduction
Psychosocial stressors may precipitate depression [1,2] and presumably also mania [3]. However, most persons exposed to stressful events do not develop psychiatric impairments [4]. This raises the question: why do some people experience an affective disorder in relation to a stressful life event and others do not? The answer is complex; it involves genetic loading, personality, prior experiences, parental style, social network and probably how individuals deal or cope with stress.
Coping and stress are aspects of life and coping style plays an important role in individual well-being. As formulated by H. Selye "that although we cannot avoid stress as long as we live, we can learn a great deal about how to keep its damaging side effects "distress" to a minimum" [5]. Freud's interest in defence mechanisms had a close connection to what today is named coping [6]. He believed some defence mechanisms to be healthier or less regressed than others in the same way that some coping styles seem to be associated with a healthier outcome. Synonyms for coping may be mastery, adaptation or behavioural style. Coping is defined in various ways: "any response to external life strains that serves to prevent, avoid or control emotional distress" [7], or as "constantly changing cognitive and behavioural efforts to manage specific external and/or internal demands that are appraised as taxing or exceeding the resources of the person" [8].
The failure of coping to deal adaptively with stress may lead to mental and physical illness. In general, people who rely more on approach coping tend to adapt better to life stressors and experience fewer psychological problems [9]. Approach and task-oriented coping are strategies involving problem solving, seeking information and attempts to alter the situation [10]. Conversely, avoidance and emotion-oriented coping strategies seem to be associated with psychological distress [11]. Avoidance coping describes activities aimed at avoiding the stressful situation and involves denial, wishful thinking and withdrawal [12]. Emotion-oriented coping describes emotional reactions that are self-orientated in order to reduce stress. These reactions involve emotional responses (individuals blaming themselves for being too emotional, becoming angry-, or tense) [13] and ruminative responses defined as: behaviours and thoughts that focus attention on depressive symptoms and on the implication of these symptoms (individuals thinking how tired they feel and why they get depressed, why others do not) [10].
Many studies of coping examine the relationship between coping and somatic illness. Research on the relation of coping and depressive disorder has attracted attention, but in clinical psychiatry there is a gap between clinical practice and research [14,15]. With changing treatment approaches for depressive and bipolar disorders such as interpersonal psychotherapy and cognitive-behavioural therapy coping may be important as a target for treatment. Therefore clinical studies involving coping and depressive and bipolar disorders in adults are reviewed in order to investigate two hypotheses; Are avoidance and emotion-oriented coping strategies associated with an increased risk of developing an affective disorder and with a higher risk of relapse of mood episodes?
Methods
Electronic searching in Embase and Medline using the combination of the following terms: Mood Disorders or Bipolar Disorder or Depression (MeSH-terms) Mania or Manic-depressive Illness (Text-words) and Coping (Text-word): 5864 hits Medline, 3016 hits Embase. Studies of coping and affective disorder were selected by reading headings and abstracts. The following studies were excluded; coping and somatic illness, coping in children and adolescents and studies of coping concerning individuals as employees, parents etc. Medline Index starts in 1968. The search was ended at March 2005.
Results
As seen from the Tables, 11 cross-sectional studies and 17 follow-up studies were identified. The studies are presented in three tables according to design and in historical order. Five studies investigated the same sample [16,35,36,40,43].
Cross-sectional studies
The 11 cross-sectional studies are presented in Table 1. The cross-sectional studies showed; Firstly, a relation between emotion-oriented and avoidance coping strategies and the prevalence and symptom severity of depression. Secondly, the cross sectional studies cannot show whether these strategies were associated with first episode or with relapse of a depressive or manic episode.
Table 1 Cross-Sectional studies of the relationship between mood disorder and coping style.
Study Controlled = CON Patients No. Sex
Men % Mean
Age Measures Results
Billings [16]
1984 CON 424 outpatients 55 41 yr HDLF
RDC Coping responses directed toward problem solving and affect regulation were associated with less dysfunction. Emotional discharge responses were linked to greater dysfunction.
Rosenberg [17]
1987 CON 24 inpatients.
36 controls 48 36 yr BDI
CR Depressed patients more often reported avoidance strategies to cope and non-depressed utilize active coping techniques
McNaughton [18]
1992 CON 27 inpatients 100 42 yr RDC, HRDS
LEDS, WCC The depressed group used more emotion-focused coping than control persons.
Turner [19]
1992 26 outpatients 31 39 yr BDI
MCI Significantly positive correlation between high depression scores and emotion oriented coping. Significantly negative correlation between high depression scores and task oriented coping.
Lam [20]
1997 40 bipolar patients 43 44 yr MRC, IQ
MA, BDI
CPI Patients' level of social functioning was related to their level of insight and how well they coped with prodromes of mania and whether they could detect prodromes of depression.
Dekker [21]
1999 * 248 outpatients major depression. 42 Ca 35 yr HDRS
SCL-90
QLDS
UCL The more depressed patients used passive reaction pattern and avoidance significantly more often than less depressive patients. Passive reaction was the most consistent predictor of high depression score.
Schouws [22]
2001 211 outpatients with major depression. 42 34 yr DSM-III
HDRS, UCL
SCL-90, QLDS Active approach and seeking social support associated with a higher quality of life. No gender differences in coping. Avoidance coping was related to higher severity of depression.
Vossler [23]
2001 CON 41 in and out patients, 41 control persons 53 42 yr SCID
F-SOZU
FKV-LIS Depressive patients reported more social stress and used ineffective coping strategies and wishful thinking more often than the control persons.
Ravindran [24]
2002 CON 229 dysthymic, major depressive. 44 controls 39 41 yr HAM-D, BDI
MADRS, CGI, CGIS, DHUS Among depressive patients the severity of illness was associated with emotion focused coping, whereas control-persons favoured the use of cognitive strategies.
Lam [25]
2003 109 outpatients unipolar 45 44 yr SCID, HRSD, RSQ
IIP, SPC Rumination was associated with higher levels of depression. Distraction was associated with lower level of depression.
McWilliams [26]
2003 298 outpatients major depression. 48 43 yr CISS
NEO-FFI
BDI, STAI-S Task-oriented and social coping were associated negatively and emotion-oriented coping was associated positively with high scores on depression and anxiety.
* Used a comparison group from another study.
BDI Beck Depression Inventory. Depression Scale. CGIS Clinical Global Impressions Scale. CIDI Composite International Diagnostic Interview-Short Form.
CISS Coping Inventory of Stressful Situations. CPI Coping with Prodromes Interview. CR Coping Response. CSS Coping Strategies Scale.
DSM-IIII Diagnostic and Statistical Questionnaire Manual of Mental Disorders fourth edition. DCS 8 8 Different Coping Strategies. DHUS Daily Hassles and Uplifts Scales.
HPS Hypomanic Personality Scale. HAM-A Hamilton Anxiety Scale. HDLF The Health and Daily Living Form. HRSD Hamilton Rating Scale for Depression.
IIP Inventory of Interpersonal Problems. IQ The Insight Questionnaire. LE Life Events over the past year. LEDS Life Events and Difficulties Schedule.
MADRS Montgomery Asberg Depression Rating Scale. MAS The Mania Scale. MCI The Multidimensional Coping Inventory. MRC Social Performance Schedule.
NEO-FFI NEO Five Factor Inventory. QLDS Quality of Life Depression Scale. RCD Research Diagnostic Criteria. RL Recent Life events and Chronic Stress.
RSQ Nolen-Hoeksema Response Styles Questionnaire. SCID Structured Clinical Interview for DSM-IV SCL-90 Symptom Check List of 90 items.
SPS Social Performance Scale. STAI-S. The State Anxiety Subscale of the Spielberger State-Trait Anxiety Inventory. UCL Utrecht Coping List. UCLA Loneliness Scale. WCC The Ways of Coping Checklist.
Short-term studies
The seven short-term studies are presented in table 2. Short-term studies are here defined as studies with a three- to six-month follow-up period. The short-term studies indicated: firstly, in four of the seven studies that emotion-oriented and ruminative coping style were state dependent [27,29,31,32], secondly, in four studies these coping styles were related to the severity of the depressive symptoms [27,29-31], thirdly, three studies showed that recovery from depression were associated with changes in coping style [27,29,31], accordingly, recovered patients relied less on inappropriate emotion-focused coping strategies and finally, two studies showed that emotion-oriented coping predicted the presence of depression at follow-up [28,30]. Conversely one study showed that ruminative coping might be associated with a good outcome in depressed patients [33].
Table 2 Short-term (3–6 months] studies of the relation between coping and mood disorder.
Study Controlled = CON Follow-up month Patients No. Sex % Men Mean age Measures Results
Parker [27] 1982, CON 4 95 from a non-clinical group 33 38 yr BCG
EF Disturbances in antidepressive behaviours were more likely to be a consequence of rather than an antecedent of depression
Parker [28] 1986 4 43 depressed
66 volunteers 17
39 32 yr
42 yr BDI, PSE
CQ, ZDS Higher self-consolation scores predicted less improvement. Those who scored low on three subscales self-consolation, distraction and socialization had better improvement
Schussler [29], 1992 2 40 depressed patients 45 yr DS
ADV-L Found difficulty in distinguish between symptoms of depression and certain coping behaviours [e.g. withdrawal]
Ravindran [30] 1995 CON 6 17 dysthymia 17 depressive and 18 controls. 49 38 yr MADRS, CSS
HAM-A, CGIS
DHUS, UCLA Recovery from depression was associated with change in coping style, so that patients relied less on inappropriate emotion-focused coping strategies.
Kuehner [31] 1999 4 52 unipolar 42 44 yr SCAN, PSE-10
IDD, RSQ A diagnosis of depression was associated with rumination. Baseline rumination predicted follow-up levels of depression.
Uehara [32] 2002 4 36 depressed, 13 anxiety 39 39 yr CISS, HRSD
SAS, DSM-III Task-oriented coping was influenced by depression. Emotion-oriented coping was influenced by anxiety. State and treatment phase affected coping measurement
Yamada [33] 2003 6 105 depressed patients 42 44 yr COALA
GAS
HRDS Patients with a good outcome at 6 months used significantly more rumination while patients with a poor outcome used significantly more dangerous activity
ADV-L Die Liste Antidepressiver Verhaltensweisen von Hautzinger. BDI Beck Depression Inventory. BCG Behavioural Change Questionnaire. CISS Coping Inventory for Stressful Situations.
COALA Comprehensive Assessment List for Affective Disorders. CQ Coping Questionnaire. CSS Coping Strategies Scale. DHUS Daily Hassles and Uplifts Scales.
DS Depressionsscale von v. Zerssen. DSM-IIII Diagnostic and Statistical Questionnaire Manuel of Mental Disorders fourth edition. EF Effectiveness Scale.
EHEI Early Home Enviroment Interview. GAS Global Assessment Scale. HAM-A Hamilton Anxiety Scale. HRSD Hamilton Rating Scale for Depression. IDD Inventory to Diagnose Depression.
LIFE Longitudinal Interval Follow-up Evaluation. MADRS Montgomery Asberg Depression Rating Scale. RSQ Response Style Questionnaire. SAS Self-rating Anxiety Scale.
SCAN PSE 10 Schedules for clinical assessment in Neuropsychiatry with the Present State Examination, 10 th edition. SCID Structured Clinical Interview for DSM-III. ZDS Zung Depression Scale.
Long-term studies
The 10 long-term studies are presented in table 3. Long-term studies are here defined as studies with a follow-up period longer than six months. The long-term studies indicated: firstly, that adaptive coping strategies predicted increased remission and decreased risk of relapse in half of the studies [35,37,39,41,42]. One study concluded that depressed persons did not differ in amount of problem-focused coping compared to control persons [34]; another study did not find that problem-focused coping was a predictor of post-treatment symptoms [38]. Secondly, the use of passive and ruminative coping strategies and coping by avoiding other people was found to be related to a higher risk of relapse and of greater severity of depressive symptoms at follow-up in six of the ten studies [34,35,37,39,40,42]. Thirdly, one study showed that a high level of stressors was connected with higher depression scores at follow-up [37]. Finally, the study on bipolar disorder showed that patients that used stimulating coping strategies more often had a manic episode [42]
Table 3 Long-term (over 6 months] studies of the relation between coping and mood disorder.
Study Controlled = CON Follow up year Patients No. Sex Men % Mean age Measures Results
Coyne [34] 1981, CON 1 15 depressed and 72 controls 52 55 yr HSCL, WCL 4 weeks intervals The coping of depressed persons was characterized by seeking emotional support and by wishful thinking.
Billings [35, 36] 1985, CON 1 380 unipolar 57 40 yr HDL
FES
RDC Patients at follow-up used significantly more affect regulation and less reliance on information seeking and emotional discharge, latter coping styles were associated with poorer outcome
Swindle [37] 1989, CON 4 352 unipolar 44 44 yr HDL, FES
RDC Problem solving related to less depression and greater self-esteem. Emotional discharge associated with depression.
Hoffart [38] 1993 1 21 depressed 17 depr/phobia, 23 agoraphobia. 34 41 yr SCID, WICCA
BDI, ACS
CPRS Seeking social support may be a trait dependent coping style. Problem focused coping and wishful thinking appeared as a state phenomena.
Sherbourne [39] 1995 2 604 depressed 26 46 yr DSM-III
SF-36, COD Better clinical course of depression was associated with more active and less avoidant coping styles
Moos [40] 1999, CON 10 313 unipolar 40 48 yr DSSI, HDL
RDC Patients were at risk for a chronic course if they coped with stressors by avoiding being with people.
Oldehinkel [41] 2000 3 1/2 86 from primary care 31 37 yr PSE, DSM, UCL
LEDS, SRS, ABV Predictors that expedited remission were high self-esteem and a tension reducing coping style.
Lam [42] 2001 1 1/2 40 bipolar 43 44 yr MAS, SCID
CPSI More who used stimulating coping strategies had a manic relapse. More who used passive coping strategies had a depressive relapse.
Holahan [43] 2003, CON 10 313 unipolar 40 48 yr HDL, RDC, DTC
DSSI, DP Patient who more often drank to cope at baseline had a stronger association to depressive symptoms and drinking problems.
Szadoczky [44] 2004 2 117 unipolar 25 44 yr HRSD, MMPI, WCC, SAS, LEQ No significant difference between the group of remitters and the group of non-remitters in problem-solving coping and emotion-focused coping.
ABV Amsterdams Biografische Vragenlijst. ACS Agoraphobic Cognition scale. BAI Beck Anxiety Inventory. BDI Beck Depression Inventory. COD Course of Depression.
CPRS Comprehensive Psychopathology Rating Scale. CPSI Coping with Prodomal Symptoms Interview. DIS Diagnostic Interview Schedule.
DSM-III & IV Diagnostic and Statistical Manual of Mental Disorders [third and fourth edition]. DP Drinking Problems. DTC Drinking to Cope. DSSI Depressive Symptoms Severity Index.
FES Family Environment Scale. GAS Global Assessment Scale. GDS Geriatric Depression Scale. HSCL Hopkins symptom Checklist. HDL Health and Daily Living Form.
HRSD Hamilton Rating Scale for Depression. MAS Mania Scale. LEQ Life Event Quistionnaire. PB Passive Behaviours. MMPI Minnesota Mulriphasic Inventory. PSE Present State Examination.
RDC Research Diagnostic Criteria. SADS-L Schedule for Affective Disorders and Schizophrenia-Lifetime. SAS Social Adjustment Scale. SCI Stress Coping Inventory.
SCID Structured Clinical Interview for DSM-III. SF-36 a 36-item Self-report health-related quality of life measure. Social Support and Rejection Scale. UCL Utrecht Coping List.
WCC The Ways of Coping Checklist. WWC L The revised Ways of Coping Checklist.
Results summary
Regarding the two hypotheses none of the studies were capable of answering whether certain coping strategies are predictive of affective disorder. Most studies showed that in depressive disorder avoidance and emotion-oriented coping strategies seem to be related to recurrence/relapse of depressive episodes. These coping strategies might also be associated with an increased time to recovery. Only a few studies on bipolar disorder were identified, limiting the findings to depressive disorder mainly. In the cross-sectional and short-term studies it is difficult to make a clear distinction between coping processes and symptoms, and avoidance and emotion-oriented coping styles seemed to be state related, so the results of these studies are not able to answer our hypothesis.
Discussion
Study-design
Regarding the trait or state effect, the use of cross-sectional designs and designs with a short follow-up period can be problematically because of the possibility of measuring residual depressive symptoms. The coping style of an individual may change during a depresive or manic episode, as illustrated in the short-term studies. Possibly, the emotion-oriented coping style found especially in the cross-sectional studies could be explained by a state-dependent consequence of a depressed mood. One way to determine whether the anomalies are caused by the depressive or manic episode (state model) is to include a control group. Among the 28 studies, 12 included a control group. In the follow-up studies, the lack of a control group makes it difficult to distinguish whether the changes that occurred reflect normalization of coping over time or whether the changes reflect changes in affective symptoms. The best impression of the relation between coping and state comes from the follow-up studies that used repeated measures of affective symptoms, life events and coping behaviour and include a control group [34-37,40,43]. The studies with a long follow-up time [37,40,41,43] illustrate the time-relation between coping and affective episodes: these studies support the hypothesis that emotion-oriented and avoidance coping strategies seem to be associated with increased risk of relapse. So, the controlled follow-up studies provided the most valid study design to comply with methodological problems.
Comparing results
It is difficult to compare studies involving heterogeneous samples and it is likely that selection bias influenced the results unpredictable. Most studies used clinical samples of unipolar patients. A general limitation is the lack of attention to confounding variables such as the duration and severity of the mood episodes and the number of previous episodes. Some studies investigated these parameters [21,23,28,30,39-41] and these results did not deviate substantially from those of other studies. The use of passive coping strategies predicted relapse of depression and stimulating coping strategies increased the relapse risk of mania but from the limited number of studies concerning bipolar disorder [20,42] it is not possible to elucidate the relation between coping and bipolar disorder. The little attention to research into coping mechanisms and mania was already stressed in a case report in 1990 [45] and latest by the authors in a research report from 1999 where the development and validation of the coping inventory for prodomes of mania was described [46].
Participants were registered differently according to comorbid psychopathology. In one study the most common comorbid disorders were panic disorder, dysthymia and social phobia [26]. One study concentrated on comorbid agoraphobia [39] and one on comorbid personality disorder [44]. In some studies patients with comorbid psychopathology were excluded [16,21,22,28,29]. Although the relation of coping processes and risk for substance abuse is a defined area of coping research [47], only two studies [39,43] adjusted for the possible effect of comorbid substance use as a part of a coping strategy. This may have affected the results so that some inappropriate coping strategies are not identified.
Coping measurements
As seen from the tables, the studies measured coping differently, which makes it difficult to compare across studies. The coping measures were based on self-reports and a variety of questionnaires were used without any golden standard. It is important to develop reliable and valid tools that assess how people cope with stressful situations and negative events. Therefore developing structured and easily scorable scales started in the 1960s [48] and during the 1980 and 1990s, – much research was conducted on self-reported measures of coping. Facing particularly stressful events, individuals are asked about the kind of coping behaviour(s) in which they have engaged [49]. This research used widely divergent strategies, techniques and measures; additionally some of the scales possess a variety of psychometrically drawbacks [13]. When examining the coping questionnaires, it is clear that substance use as a coping behaviour has virtually been ignored [13,27,50,51]. The two studies concerning abuse showed that depressive symptoms were significantly associated with more drinking problems [43] and that improvement in functioning and well-being was associated with less alcohol consumption [39].
As discussed, in a critical review concerning the gap between coping theory and clinical intervention research [52], when using a coping checklist and thereby reducing coping to a summary score, a lot of information will be lost. The questions seem to be to general for the results to be valid and crucial aspects of timing, sequencing and appropriateness may be lost.
Coping and life events
There is a substantial and partially causal link between life-events and depression [53] and the occurrence of major life events signals a period of increased risk of developing a depressive episode [54]; several studies measured life events [17,18,21,23,24,26,27,29,34-37,39,43,44]. A general problem in measuring coping in connection with life events is the time relation, as participants should recall a stressful situation and then reconstruct how they dealt with it. Firstly, there is recall bias and secondly, coping has a temporal aspect: an individual can cope before a stressful event occurs, while it is happening and afterwards [55]. The issue of timing is complex as there presumably may be an interaction between individual vulnerability, coping, life events and the risk of developing depression. Thus it is difficult to measure the coping process given the typical brief period between a severe life event and depression. Capturing the coping process represents a challenge [56] and it may require a more frequent monitoring of coping than is usually done in studies.
All coping measures are based on self-reports and participants are often asked in connection with hypothetical stressors e.g. the Coping Inventory for Stressful Situations [13] and it is difficult to know what kind of stressors were thought of. Many scales for coping are developed to assess overall coping strategies but it reasonable that human cope differently according to the nature of the event. Linking coping measures to recent actual stressors and attempting objective assessment of the stressful life event to avoid circularity is of main importance. The procedure of semi-structured interviews is time- and labour consuming and might be a reason for the consistent use of self-reports. The lacking use of coping in clinical assessment and treatment may be due to the rather confusing divergent coping scales. It can be rather meaningless to detect coping not knowing what kind of stressor the coping process relates to and the influence of personality traits cannot be ignored [56]. For advancing research and clinical focus on coping there is a need of developing an integrated semi-structured interview detecting personality traits, life events and coping together. Beside, a specific development of coping schedules capable of identifying prodomal symptoms of bipolar episodes as described by Wong et al. [46] would be of clinical use.
Coping and personality
Coping is closely related to personality and personality can affect coping measures and in coping as well as in personality research there is a distinction between trait and state variables. In the concept of coping responses [57], coping is seen as a response to specific stressful situations rather than a stable feature of personality. Another approach emphasizes the method of coping; whether a response entails primal cognitive or behavioural efforts [58]. One study [26] showed that less-adaptive coping strategies (emotion-oriented coping) were associated with neuroticism and depression, whereas the reverse association was found regarding adaptive coping strategies (task-orientated coping). These findings are replicated in studies of the general population [59,60]. Finally, a study [61] from Japan found that task-oriented and avoidance coping were related to extraversion and that emotion-oriented coping was related to neuroticism. Personality traits are important in the assessment of coping and it is advisable to measure these traits when dealing with coping measures; however, only three studies have done so [26,41,44].
Coping and Gender
In most of the present studies more women were included reflecting the gender differences in the prevalence of depressive disorder. In some studies gender differences were found and the general tendency was that men tend to distract themselves using active coping strategies, whereas women use strategies involving expressing emotion [16,21,25]. Other studies found no gender differences [22,30,33]; however, most studies did not take gender into consideration in the analyses. According to the hypothesis of Nolen-Hoeksema [62] the increased vulnerability of women to developing depression is related to gender differences in coping; men's response to their dysphoria is more behavioural and dampens their depressive episodes, whereas women's response to their dysphoria is more ruminative and amplifies them. Accordable to a review [63] of gender differences in depression, it is possible that men tend to distract themselves from their mood by engaging in physical or instrumental activities, whereas women are less active and ruminate over the possible causes and implications of their depression. These hypotheses are compatible with findings from other studies [64-66] the latter study involving data covering representative population samples from six European countries. Conversely, an older prospective one-year study [67] of 100 healthy community-residing men and women, in which participants were interviewed seven times at four-week intervals, found no gender differences in emotion-focused coping. Another study [68] of couples that recently had experienced at least one threatening life event that was potentially depressonigenic for both showed that women had a greater risk than men of depressive episodes following the life event. The greater risk was restricted to episodes that followed events involving children, housing or reproductive problems. Women's greater risk was only present among those couples for whom there were clear gender differences in associated roles.
Coping and age
Age differs between the studies but most studies had a broader age range. How age influences the stress and coping process is not clear: The Normative Aging Study [69], a longitudinal study examined stress, appraisal and coping in three groups' middle-aged, young-old and old-old men. The study followed 2280 men for more than 30 years. A significant overall effect of age on coping strategies: instrumental action, cognitive reframing, social support and interpersonal hostile strategies were found and all coping strategies showed linear decrease with age. The relation between age and coping is complex and there is no clear answer as to whether persons cope better or worse as they age. In a recent study, the association between life events and onset of depression and mania was not found to change throughout life [70]
Coping and medical adherence
The episodic course of affective disorders presents a challenge to the clinician and the patient, in making and fulfilling a treatment plan. Effective treatment depends on medication adherence also when treating medical patients with comorbid depression. When treating affective disorder, doctors and patients have to deal with different problems: lack of insight, symptom-free intervals, residual affective symptoms and poor social support, which can complicate long-term treatment [71]. Identification of these factors may add to increase medication adherence as seen in a study [72] that investigated medication adherence in 32 patients with bipolar I disorder. Consistent with the hypothesis that acceptance coping would be positively associated and denial coping negatively associated with adherence to mood-stabilizing medication, it was found that low levels of acceptance and a high level of denial undermined medication adherence.
Clinical implications
Coping strategies could be a target for selective prevention targeting subgroups of the population whose risk of developing an affective disorder is higher than average. Further, a way to ameliorate the course in affective disorder is to help patients to identify prodomal symptoms and individual maladaptive coping strategies and try to change these, e.g. problem training as a treatment for depression [46,73,74] or intervention designed to improve coping. More recent literature supports the utility of individual cognitive behavioural and psycho-educational approaches, particular in enhancing medication adherence, so medication and psychotherapy are not only compatible but also synergistic [75] and a strong focus in psychological treatment involves structured attempts to teach patients new coping skills.
Conclusion
The gap between coping theory and clinical use of coping remains and the clinical relevance of coping is, though promising, still unclear. It is difficult to make a clear distinction between coping processes and symptoms. The complex interaction between life stressors, coping, personality and affective disorders need to be better understood before coping behaviour and changes in coping strategies can be included more systematically in patient treatment. Primary, there is a need to develop new valid and reliable measures combining assessment of coping, life events and personality in a semi-structured interview. Long-term high-risk studies capable of detecting coping before and after an onset of an mood disorder would provide new information of the relation between coping strategies and affective disorder.
Acknowledgements
The Theodore and Vada Stanley Medical Research Institute and the Lundbeck Foundation supported the study.
==== Refs
Ellicott A Hammen C Gitlin M Brown M Jamison K Life events and the course of bipolar disorder Am J Psychiatry 1990 147 1194 98 1974746
Rudolp KD Hammen C Burge D Toward an interpersonal life-stress model of depression: The development context of stress generation Develop Psychopath 2000 12 215 34
Kessing LK Agerbo E Mortensen PB Major stressful life events and first admission with mania 2004 6 122 9 15005750
Kessler RC The effects of stressful events on depression Ann Rev Psychol 1997 48 191 214
Seley H The stress of life The McGraw-Hill Companies, Inc 1976 Revised
Bond M Gardner ST Christian J Sigal JJ Empirical study of self-rated defence styles Arch Gen Psychiatry 1983 40 333 38 6830412
Pearlin LI Schooler C The structure of coping J Health Soc Behav 1978 19 2 21 649936
Lazarus RS Folkman S Stress, Appraisal and Coping. New York 1984 Springer Publishing Company 328 33
Folkman S Lazarus RS The relationship between coping and emotion: implications for theory and research Soc Sci Med 1988 26 309 17 3279520
Ray C Lindop J Gibson S The concept of coping Psychol Med 1982 12 385 95 7100361
Suls J Fletcher B Self attention, life stress and illness: a prospective study Psychosom Med 1985 47 469 81 4059480
Folkman S Lazarus RS The relationship between coping and emotion: implications for theory and research Soc Sci Med 1988 26 309 17 3279520
Endler NS Parker JDA CISS Coping Inventory for Stressful Situations Manual second edition, Multi-Health Systems 1999
Sommerfield MR McCrae RR Stress and coping research. Methodological challenges, theoretical advances and clinical applications Am Psychol 2000 55 620 25 10892204
Lazarus RS Toward better research on stress and coping Am Psychol 2000 55 6 665 73 10892209
Billings AG Moss RH Coping, stress and social resources among adults with unipolar depression J Pers Soc Psychol 1984 46 877 91 6737198
Rosenberg SJ Peterson RA Hayes JR Coping behaviours among depressed and nondepressed medical inpatients J Psychosom Res 1987 31 653 58 3430429
McNaughton ME Patterson TL Irwin MR Grant I The relationship of life adversity, social support, and coping to hospitalisation with major depression J Nerv Ment Dis 1992 180 491 97 1500930
Turner RA King PR Tremblay PF Coping styles and depression among psychiatric outpatients Pers Individ Diff 1992 13 1145 47
Lam D Wong G Prodromes, coping strategies, insight and social functioning in bipolar affective disorders Psychol Med 1997 27 1091 1100 9300514
Dekker J Ormel J Depression and coping Eur J Psychiatry 1999 13 183 89
Schouws S Dekker J Tuynman-Qua H Kwakman H Jonghe F Relation between quality of life and coping and social behaviour in depression Eur J Psychiatry 2001 15 49 56
Vossler A Sommer G Bühler KE Haltenhof H Soziale unterstützung and krankheitsverarbeitung bei 41 patienten mit depressiven störungen Nervenheilkunde 2001 20 17 21
Ravindran AV Matheson K Griffiths J Merali Z Anisman H Stress, coping, uplifts and quality of life in subtypes of depression: a conceptual frame and emerging data J Affect Disord 2002 71 121 30 12167508
Lam D Scuck N Smith N Farmer A Checkley S Response style, interpersonal difficulties and social functioning in major depressive disorder J Affect Disord 2003 75 279 83 12880940
McWilliams AL Cox BJ Enns MW Use of the coping inventory for stressful situations in a clinically depressed sample: Factor structure, personality correlates, and prediction of distress J Clin Psychol 2003 59 4 423 37 12652635
Parker GB Brown LB Coping behaviours that mediate between life events and depression Arch Gen Psychiatry 1982 39 1386 91 7149898
Parker GB Brown LB Coping behaviours as predictors of the course of clinical depression Arch Gen Psychiatry 1986 43 561 65 3707288
Schussler G Grischke M Ruger U Coping with depressive disorders Nervenartz 1992 63 416 21
Ravindran AV Griffiths J Waddell C Anisman H Stressful life events and coping styles in relation to dysthymia and major depressive disorders. Variations associated with alleviation of symptoms following pharmacotherapy Prog Neuropsychopharmacol Biol Psychiatry 1995 19 637 53 8588062
Kuehner C Weber I Responses to depression in unipolar depressed patients: An investigation of Nolen-Hoeksema's response styles theory Psychol Med 1999 29 1323 33 10616938
Uehara T Sakado K Sato T Takizawa R Coping measurement and the state effect of depression and anxiety in psychiatric outpatients Psychopathology 2002 35 48 51 12006748
Yamada K Nagayama H Tsutiama K Kitamura T Furukawa T Coping behaviour in depressed patients: a longitudinal study Psychiatry Res 2003 121 169 77 14656451
Coyne JC Aldwin C Lazarus RS Depression and coping in stressful episodes J Abnorm Psychol 1981 90 439 47 7298992
Billings AG Moos RH Life stressors and social resources affect posttreatment outcomes among depressed patients J Abnorm Psychol 1985 94 140 53 3998282
Billings AG Moos RH Psychosocial processes of remission in unipolar depression: comparing depressed patients with matched community controls J Cons Clin Psychol 1985 53 314 25
Swindle RW Cronkite RC Moos RH Life stressors, social resources, coping and the 4-year course of unipolar depression J Abnorm Psychol 1989 4 468 77
Hoffart A Martinsen EW Coping strategies in major depression, agoraphobic and comorbid in-patients: A longitudinal study Br J Med Psychol 1993 66 143 55 8353108
Sherbourne CD Hayes RD Wells KB Personal and psychosocial risk factors for physical and mental health outcomes and course of depression among depressed patients J Consul Clin Psychol 1995 63 345 55
Moos RH Cronkite RC Symptom-based predictors of a 10-year chronic course of treated depression J Nerv Ment Dis 1999 187 360 68 10379723
Oldehinkel AJ Ormel J Neeleman J Predictors of time to remission from depression in primary care patients: Do some people benefit more from positive life change than others J Abnorm Psychol 2000 2 299 307
Lam DH Wong G Sham P Prodromes, coping strategies and course of illness in bipolar affective disorder-a naturalistic study Psychol Med 2001 31 1397 1402 11722154
Holahan CJ Moos RH Holahan CK Cronkite RC Drinking to cope and alcohol use in unipolar depression: a ten-year model J Abnorm Psychol 2003 112 159 65 12653424
Szadoczky E Rozsa S Zambori J Furedi J Predictors of 2-year outcome of major depressive episode J Affect Disord 2004 83 1 49 57 15546645
Suresh TR Srinivasan TN Claimed simulation of insanity Br J Psychiatr 1990 157 603 5
Wong G Lam D The development and validation of the coping inventory for prodomes of mania J Affect Disord 1999 53 57 65 10363667
Wills TA Hirky AE Coping and substance abuse: a theoretical model and review of the evidence Zeidner M, Endler NS, Handbook of Coping 1996 279 302
Sidle A Moos R Adams J Cady P Alto P Development of a coping scale Gen Psychiatry 1969 20 226 32
Endler NS Parker JDA A factor analytic study of coping styles and the MMPI-2 content scales J Clin Psychol 1990 49 523 27
Rosenbaum M A schedule for assessing self-control behaviours: prelaminary findings Behav Ther 1980 11 109 21
Porter LS Stone AA Zeidner M An approach to assessing daily coping 1996 Endler NS: Handbook of Coping 133 50 9147108
Coyne JC Racioppo MW Never the Twain shall meet? Closing the gap between coping research and clinical intervention research Am Psychol 2000 55 655 64 10892208
Brown GW Harris T Social origins of depression 1978 University printing house, Cambridge
Paykel ES Life events, social support and depression Acta Psychiatr Scand 1994 337 50 58
Schwarzer R Stress and coping from a social-cognitive perspective Annals New York Academy Science 1998 851 531 37
Kessler RC Price RH Wortman CB Social factors in psychopathology: stress, social support and coping processes Annu Rev Psychol 1985 36 531 72 3883893
Coyne JC Gottlieb BH The mismeasure of coping by checklist J Pers 1996 64 959 91 8956519
Moos RH Holahan CJ Dispositional and contextual perspectives on coping: toward an integrative framework J Clin Psychol 2003 12 1387 1403
Cosway R Endler NS Sadler AJ Deary IJ The coping for stressful situations; Factorial structure and associations with personality traits and psychological health J Applied Biobehav Res 2000 5 121 43
Saklofske DH Kelly IW Coping and personality Psychol Rep 1995 77 481 82 8559873
Uehara T Sakado K Sato T Takizawa R Relationship between stress coping and personality in patients with major depressive disorder Psychother Psychosom 1999 68 26 30 9873239
Nolen Hoeksema SH Sex differences in unipolar depression: evidence and theory Psychol Bull 1987 101 259 82 3562707
Piccinelli M Wilkinson G Gender differences in depression: Critical review Br J Psychiatry 2000 177 486 92 11102321
Wilhelm K Parker G Sex differences in depressiogenic risk factors and coping strategies in a socially homogeneous group Acta Psychiat Scand 1993 88 205 11 8249654
Bouhuys AL Geerts E Gordijn CM Gender-specific mechanisms associated with outcome of depression: perception of emotions, coping and interpersonal functioning Psychiatr Res 1999 85 247 61
Angst J Gamma A Gastpar M Lépine JP Mendlewicz J Tylee A Gender differences in depression Eur Arch Psychiatr Clin Neuroscience 2000 252 201 9
Folkman S Lazarus RS An analysis of coping in a middle-aged community sample J Health Soc Behav 1980 21 219 39 7410799
Nazroo JY Edwards AC Brown GW Gender differences in the onset of depression following a shared life event: a study of couples Psychol Med 1997 27 9 19 9122312
Aldwin CM Sutton KJ Chiara G Spir A Age differences in stress, coping, and appraisal: findings from the normative aging study J Gerontol 1996 51 177 88
Kessing LK Agerbo E Mortensen PB Does the impact of major stressful life events on the risk of developing depression change throughout life? Psychol Med 2003 33 1177 84 14580072
Schuman C Lenz G Berghöfer A Müller-Oerlinghausen B Non-adherence with long-term prophylaxis: a 6-year naturalistic follow-up study of affectively ill patients Psychiatr Res 1999 89 247 57
Greenhause WJ Meyer B Johnson SH Coping and medication adherence in bipolar disorder J Affect Disord 2000 59 237 41 10854641
Jané-Llopis E Hosman C Jenkins R Anderson P Predictors of efficacy in depression prevention programmes Br J Psychiatry 2003 183 384 97 14594912
Fava GA Kellner R Prodomal symptoms in affective disorders Am J Psychiatry 1991 148 823 30 2053620
Craighead WE Miklowitz DJ Psychosocial interventions for bipolar disorder J Clin Psychiatry 2000 61 58 64 11153813
|
16212656
|
PMC1276804
|
CC BY
|
2021-01-04 18:01:02
|
no
|
Clin Pract Epidemiol Ment Health. 2005 Oct 7; 1:20
|
utf-8
|
Clin Pract Epidemiol Ment Health
| 2,005 |
10.1186/1745-0179-1-20
|
oa_comm
|
==== Front
Global HealthGlobalization and Health1744-8603BioMed Central London 1744-8603-1-151620970310.1186/1744-8603-1-15DebateAssessing the impact of the Australia-United States Free Trade Agreement on Australian and global medicines policy Faunce Thomas [email protected] Evan [email protected] David [email protected] Peter [email protected] Andrew [email protected] Brita [email protected] Warwick [email protected] Globalisation and Health Project, Centre for Governance of Knowledge and Development Regulatory Institutions Network Australian National University, Acton, Canberra ACT, Australia2 Medical School and Law Faculty, The Australian National University, Acton, Canberra, ACT Australia3 Newcastle Institute of Public Health, University of Newcastle, Newcastle, New South Wales, Australia4 Clinical Pharmacology, School of Medical Practice and Population Health, University of Newcastle, Newcastle, New South Wales, Australia5 Centre for Regulation and Market Analysis, University of South Australia, Adelaide, South Australia, Australia2005 6 10 2005 1 15 15 21 7 2005 6 10 2005 Copyright © 2005 Faunce et al; licensee BioMed Central Ltd.2005Faunce et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
On 1 January 2005, a controversial trade agreement entered into force between Australia and the United States. Though heralded by the parties as facilitating the removal of barriers to free trade (in ways not achievable in multilateral fora), it also contained many trade-restricting intellectual property provisions and others uniquely related to altering pharmaceutical regulation and public health policy in Australia. The latter appear to have particularly focused on the world-respected process of federal government reimbursement after expert cost-effectiveness evaluation, popularly known as the Pharmaceutical Benefits Scheme ('PBS'). It remains uncertain what sort of impacts – if any – the Australia-United States Free Trade Agreement ('AUSFTA') will have on PBS processes such as reference pricing and their important role in facilitating equitable and affordable access to essential medicines.
This is now the field of inquiry for a major three year Australian Research Council ('ARC')-funded study bringing together a team of senior researchers in regulatory theory from the Australian National University and pharmacoeconomics from the University of Newcastle. The project proposes to monitor, assess and analyse the real and potential impacts of the AUSFTA in this area, providing Australian policy-makers with continuing expertise and options.
To the extent that the AUSFTA medicines provisions may represent an
important precedent in a global strategy by industry on
cost-effectiveness evaluation of pharmaceuticals, the study will also be
of great interest to policy makers in other jurisdictions.
==== Body
Introduction
The final text of the Australia-United States Free Trade Agreement ('AUSFTA') was signed in Washington on 18 May 2004, by the Australian Trade Minister and the United States Trade Representative. On 17 November 2004, the parties exchanged notes accepting their respective implementing processes and the agreement entered into force on 1 January 2005. The AUSFTA contained numerous provisions either directly or indirectly related to medicines regulation in Australia, particularly Annex 2C of Chapter Two, Chapter Seventeen on intellectual property and Chapter Twenty One on dispute resolution.
It remains uncertain whether the AUSFTA will have either a detrimental or beneficial impact on access to medicines and the promotion and maintenance of good health in Australia. There does, however, appear to have been a substantial difference in opinion between the Parties over procedural changes that would result in Australian medicines regulation.
Throughout the negotiations, the Australian Government's position was either that the government cost-effectiveness reimbursement system, the Pharmaceutical Benefits Scheme ('PBS'), would not be included in the AUSFTA, or that if it was, it was an item of public health policy whose core components would be protected[1]. After signature, the Australian government maintained that the fundamental architecture of the PBS remained unchanged. It acknowledged commitments to make improvements to the transparency and timeliness of PBS processes. It also affirmed its reasonable expectations that, as a result of the AUSFTA, Australian citizens would benefit from faster access to new prescription medicines, that the price of medicines on the PBS would not increase and that the text of the AUSFTA made no changes to the cost-effectiveness methods used to set PBS reimbursement levels[2].
On the other hand, the Deputy US Trade Representative stated to the US Congress:
The U.S.-Australia FTA is the first to include non-tariff market access provisions to address issues in the pharmaceutical sector. Recognizing the sensitivity of this issue, we drew on studies prepared by the Australian government to propose changes that would improve transparency and the regulatory procedures for listing new drugs in Australia. Under the FTA, the United States and Australia agreed to common principles on facilitating high quality health care and continued improvements in public health, including through government support for research and development in the pharmaceutical industry. We also agreed to establish a Medicines Working Group to discuss emerging health policy issues. Australia committed to specific steps to improve the transparency, accountability and promptness of the listing process, including establishment of an independent review of listing decisions[3].
Representatives of the multinational brand-name pharmaceutical industry, including its regional organisation Medicines Australia, claimed that there was no basis to claims that the US wanted the PBS dismantled[4]. They argued that the regulatory changes required by these areas of the AUSFTA would (a) help redress an alleged current undervaluing of pharmaceutical 'innovation' in Australian pricing arrangements and (b) stimulate locally-based research and development, as well as the local, mostly generic, pharmaceutical industry[5]. They asserted the negotiated modifications would make Australia's regulatory system more oriented to the global market pressures on industry, more responsible in its approach to intellectual property rights and so more attractive to private investment, resulting in a net welfare benefit[6].
Others, however, have pointed to US legislation requiring that nation's negotiators to seek in the AUSFTA provisions facilitating the "elimination of government measures such as price controls and reference pricing which deny full market access for United States [pharmaceutical] products"[7]. The Australian Senate Select Committee on the AUSFTA concluded:
While no single one of the specific commitments will create immediate and measurable price rises for the PBS, the new measures may well over time alter the bargaining power between the PBS and pharmaceutical companies. This may have long term ramifications that are not in the interest of Australian consumers[8].
Concern has been expressed about AUSFTA provisions with the potential to encourage higher medicines prices in Australia. These include provisions in chapter 17 (Intellectual Property) that expand the obligations of the Trade Related Intellectual Property Rights ('TRIPS') agreement by prohibiting parallel importation, restricting compulsory licensing to "national emergencies of extreme urgency," prohibiting generic manufacturers exporting to a patent-expired market when a domestic patent exists and increasing data exclusivity protections[9].
A significant additional worry for these commentators was article 17.10.4. For the first time in Australia, this linked generic regulatory market approval on quality and safety grounds with the patent status of the relevant brand name product[10]. This Hatch-Waxman-type provision was felt to risk brand name manufacturers "evergreening" soon-to-expire pharmaceutical patents, as had occurred after comparable regulations were introduced in jurisdictions such as the US and Canada[11]. The academic, community and parliamentary concern in Australia was so great on this issue, that it resulted in the Australian government passing "anti-evergreening" amendments to its AUSFTA implementing legislation. These imposed a $A10 million penalty for a bad faith challenge by a brand name manufacturer of a generic notification certificate under the new s26B of the Therapeutic Goods Act 1989 (Cth). They also allowed cost recovery in such circumstances by the Australian government[12].
Provisions in Annex 2C(1) emphasising the need for increased government recognition of pharmaceutical "innovation" and "research and development" were likewise viewed by such critics as having the potential to encourage brand name industry lobbying. This could potentially weaken, in the long term, the capacity of Australia's Pharmaceutical Benefits Advisory Committee ('PBAC') to reject, on clinical and cost effectiveness grounds, new medicines from inclusion in the government's PBS positive reimbursement list, or to reference their reimbursement price against older products with equivalent efficacy but much reduced price[13].
In this paper, we present a rationale and outline a draft plan for a three-year study, funded by the Australian Research Council ('ARC'), which will examine the impact of the AUSFTA on a range of regulatory, public health and industrial interests involved with access to medicines in Australia. An important initial point to make is that we consider the AUSFTA is best researched as a component of an ongoing process of interaction with Australia's medicines policy by the global pharmaceutical industry. This trade agreement should be viewed, in other words, either as a catalyst that may enhance the speed of regulatory change, or a tangible manifestation of industry lobbying principles that, till now, may have been more implicit. It would be misleading, in any event, to investigate the AUSFTA's potential impacts on Australian medicines policy in isolation of demonstrable long-term corporate strategies.
Some central issues our study will examine include to what extent the AUSFTA requires, facilitates, or is likely to result in, changes to Australia's generic pharmaceutical industry, as well as its PBS cost-effectiveness system of pharmaceutical regulation. We also aim to consider relevant net welfare gain or loss; whether the Australian community will get the same value-for-dollar spent on medicines, either through Commonwealth government reimbursement, hospital or patient purchase.
We propose to investigate these questions empirically (and provide a sound structure for the gradual acquisition of suitable data). This will be done first by identifying, with the assistance of qualitative interviews, actual or likely AUSFTA-associated changes to the structure and process of Australia's PBAC, as well as the marketing processes, development and sector competitiveness of generic pharmaceutical manufacturers in Australia. This aspect of the study will also review the legitimacy of such actual or proposed alterations by examining the history of Australia's PBS as a social justice measure designed to ensure universal access to essential medicines. We will also review such proposed changes for coherence with basic norms of bioethics, domestic law and international human rights. We shall then attempt to determine their actual or potential impact on a range of indicators including drug prices expenditure and affordability, drug availability and equity of access.
We hope that publishing an outline of our proposed study will further encourage policy discussion, facilitate collaborations and provide a template for governments of other countries planning to enter such agreements. Although much of the detail of the AUSFTA is specific to Australia, there are important elements likely to be relevant to future trade agreements involving the US or other countries that have a major vested interest in the production, export and rent generation associated with patented medicines. These include whether the strengthening of pharmaceutical intellectual property protection and weakening of medicines clinical and cost-effectiveness evaluation and/or reference pricing, necessarily involves a weakening of a nation's social and economic fabric, or the capacity of its population to age well and age productively.
Background: Australia's PBS in the Context of the AUSFTA
Australia's pharmaceutical sector is dominated by the operation of the federally funded PBS, which, after a process of clinical and cost-effectiveness evaluation contributing recommendations to price negotiation, provides reimbursement (currently approximately 75%) for around 80% of the prescription medicines used in Australia[15]. The PBS does not restrict market access, but facilitates maximisation of sales volume for listed products. In developing relevant price indices, our study will also take into account AUSFTA impacts on prices for hospital-used medicines (which can be calculated from the PBS reimbursement price less the minimum safety net value) and predict expenditure on medicines costed under the co-payment level.
Central to our analysis of the impact of the AUSFTA on medicines in Australia, however, will be an evaluation of its effect on the PBS. Australia's PBS was established as a free formulary of essential drugs after the Second World War by the Curtin-Chifley federal administrations[16]. It was a social justice measure designed to ensure that all Australian citizens gained access to affordable, essential medicines. Legislation to create the PBS had to survive two High Court challenges and required a successful Constitutional referendum[17]. Successive Commonwealth governments used and built upon the 1940s enactments, before a conservative party enacted the National Health Act 1953 (Cth) ('National Health Act')[18]. This is an extremely important point, that will be focused on by ARC research scholar Warwick Neville. The PBS is one of the few examples of public health policy in Australia's history that appears to have an unequivocal democratic mandate. Such an historical-jurisprudential perspective on the social justice aspects of the PBS will be a unique and distinctive feature of our analysis.
The modern PBS revolves around Part VII section 85 of the National Health Act 1953 (Cth). Section 101(4) of this Act states that the relevant Minister, upon the advice of the PBAC (with secretariat support from the Pharmaceutical Benefits Branch of the Department of Health and Ageing ('DOHA'), may declare a pharmaceutical listed on the PBS and so subject to a level of government reimbursement (except for a patient co-payment) that has been negotiated by the Pharmaceutical Benefits Pricing Authority ('PBPA'). Prior to the AUSFTA, PBS listing was required to only occur after the Therapeutic Goods Administration ('TGA') and the Australian Drug Evaluation Committee ('ADEC') had approved the relevant pharmaceutical's safety and efficacy[19]. The AUSFTA, through implementing amendments to the Therapeutic Goods Act 1989 (Cth), has already produced, as mentioned, a requirement that a market-entering generic manufacturer provide evidence that no counteracting patent is claimed, or that the brand name owner has been notified.
Central to the PBAC's clinical and cost-effectiveness evaluation is section 101(3) of the National Health Act. This requires the PBAC to base its recommendation on: 'the effectiveness and cost of therapy involving the use of the drug, preparation or class, including by comparing the effectiveness and cost [emphasis added] of that therapy with that of alternative therapies, whether or not involving the use of other drugs or preparations.' The section goes on to state that if the product is 'substantially more costly' than the selected comparator in its class it shall not be recommended by the PBAC for PBS listing 'unless...[it] provides a significant improvement in efficacy or reduction of toxicity over the alternative therapy or therapies.'
The PBS system that has evolved under this section is a variant of pharmaceutical reference pricing[20]. The pharmaceutical manufacturer's ('sponsor's') submission to the PBAC nominates a disease indication (and relevant subsets involving patient characteristics) as well as a listing price supposedly based on the pharmaceutical company's assessment of the best relevant available data on clinical effect against a comparator. The comparator is generally the drug most prescribed on the PBS for the same indication, but may be the standard medical (non-drug) treatment. Pharmaceutical companies tend to prefer comparisons against the most expensive drug with the best 'head-to-head' data, rather than the compound that is most pharmacologically similar[21]. Part of our initial task will be to determine what aspects, if any, of this important public health-related process could be subjected to pressure emerging from the AUSFTA. It will be an interesting threshold question to ascertain to what extent members of the PBAC were aware of, or consulted in, the development of the AUSFTA articles relevant to the PBS.
The PBAC's expert reviewers evaluate whether any of the assumptions in the submission are unjustifiable and create simulations to assess the incremental cost-effective ratio (the additional cost for an additional beneficial effect, or Quality of Life Years ('QALY') gained)[22]. The reports of these experts ('pink pages') are then passed back to be reviewed by the PBAC, along with an industry response to them ('blue pages') and the summary from the Economic Sub-Committee ('ESC') in the 'green pages.' The process is designed to take six weeks and follows guidelines set out on the PBS website. The extent to which such guidelines alter as a result of the AUSFTA and what impact this has on the PBAC process will be another aspect of our study.
The PBAC may ask the pharmaceutical manufacturer for additional information, but has no legal power to compel its production, even if not covered by 'commercial-in-confidence' protections. The pharmaceutical manufacturer may be claiming a price premium because of a claimed additional benefit (that is improved effectiveness, better adverse event profile or delivery system) conferred by the new product over its therapeutic rivals. One hypothesis we hope to test is that brand-name pharmaceutical manufacturers may use Annex 2C(1) of the AUSFTA to seek price premiums for alleged "innovation" as a separate issue from cost-effectiveness. If the relevant evidence submitted does not clearly establish clinical and cost effectiveness, then a process of cost minimization is undertaken[23]. The extent to which members of the PBAC are aware of any broader, strategic agenda of the brand name pharmaceutical industry in making individual submissions (for example on reference pricing), will also be an aspect of our research in this area.
Once a decision is made to list the drug, the PBPA then evaluates the requested price against an international benchmark price for drugs in that class. Thus, under the PBS system, the members of the PBAC and ESC use pharmacoeconomic analysis to determine the community value of a new drug against an agreed comparator therapy, while the monopsony bargaining power of the PBPA is used to counter the increasingly prolonged and wide potential for monopoly rents accorded to brand name pharmaceutical patent holders[24]. For brand name manufacturers, the process of bringing a new patented product to the Australian market at a higher price than currently available medicines may be lengthy, expensive and uncertain. However PBS listing provides a secure foothold in a substantial market, particularly (due to reference pricing) for the generic pharmaceutical industry[25].
The predominant norms underlying the current Australian system are cost-containment, efficiency and equity. When evaluated against such standards the PBS has performed well[26]. The considerable monopsony bargaining power it offers to government has resulted in lower overall prices for the medicines listed on the PBS. Lacking a capped budget and dependent upon prescribers following the approved guidelines, the PBS process, however, was viewed, prior to the AUSFTA, as primarily involved with quality use of medicines, rather than government cost containment[27].
Brand-name manufacturers have argued that PBS-type systems involving reference pricing, create a regulatory environment hostile to investment and innovation. Specifically, they claim, as they have in other jurisdictions, that reference pricing makes it progressively more difficult for innovative brand-name pharmaceuticals to enter the market at a price sufficient to recoup the cost of research and development. The low prices achieved through such tactics have also been alleged by such manufacturers to reduce the potential for locally based industry expansion and to risk eventual precipitation of a withdrawal of international manufacturers from the sector. They claim that Australia can only achieve the low price it currently commands for innovation by opportunistically "free-riding" on the research and development spending of developed nations, such as the US. The US Department of Commerce has recently produced a report criticising medicines price controls in OECD countries, which applies the same arguments to those jurisdictions[28]. These claims will need to be tested against available evidence.
Medicines-Related Provisions of the AUSFTA
The AUSFTA is one of a number of recent bilateral agreements sought by the US in its strategy of negotiating stronger intellectual property rights ('IPRs') with smaller trading partners. So-called "TRIPS-Plus" IPRs, seen by the US as essential for the protection of the monopoly rent it draws from intangible assets such as pharmaceutical patents, are an important feature of the AUSFTA. The AUSFTA medicines provisions also arise from an intention on the part of the US pharmaceutical industry, through the USTR, to modify government evaluative structures and processes to increase communication between industry and regulators while strengthening the financing of innovation through research and development. The extent to which this aim succeeded is debateable.
The AUSFTA, to summarise, contains approximately fifty provisions in four areas relevant to Australia's pharmaceutical sector: Annex 2C (Pharmaceuticals); the side-letters between the Australian Trade Minister and US Trade Ambassador; Chapter 17 (Intellectual Property Rights); and Chapter 21 (Dispute Resolution Procedures)[29].
Annex 2C(1) begins by articulating one overarching principle – that the parties to the agreement are '...committed to facilitating high quality health care and continued improvements in public health for their nationals.' It then mentions four subsidiary principles:
a) [recognising] the important role played by innovative pharmaceutical products in delivering high quality health care;
b) [recognising] the importance of research and development in the pharmaceutical industry and of appropriate government support, including through intellectual property protection and other policies;
c) the need to promote timely and affordable access to innovative pharmaceuticals through transparent, expeditious, and accountable procedures, without impeding a Party's ability to apply appropriate standards of quality, safety, and efficacy; and
d) the need to recognize the value of innovative pharmaceuticals through the operation of competitive markets or by adopting or maintaining procedures that appropriately value the objectively demonstrated therapeutic significance of a pharmaceutical.
Other subsections of Annex 2C, and the associated exchange of letters, relate to increased opportunities for a manufacturer to interact with regulators. This includes Australia providing manufacturers with an opportunity for hearings before the PBAC during the application for PBS listing; providing an opportunity for an independent review process following a negative PBAC price determination; the creation of a Medicines Working Group with health officials from each country in dialogue about aspects of Australia's regulatory mechanisms; an ongoing dialogue between the TGA and the US Food and Drug Administration on the issue of making pharmaceutical innovation 'quickly available'; and finally, Annex 2C(5) permits a pharmaceutical manufacturer to disseminate information about pharmaceutical innovation via the Internet.
Chapter 17 of the AUSFTA includes intellectual property provisions aimed specifically at Australia's pharmaceutical sector. Parallel importation, as already mentioned, is prohibited; compulsory licensing of pharmaceuticals is restricted to a standard more stringent than that applying in TRIPS ("national emergencies of extreme urgency"); generic production of domestic-patented drugs for export to jurisdictions where patents have already expired is prevented. Data exclusivity is extended, as is patent protection where there have been delays in issuing marketing approval[30]. Article 17.10.4 of the AUSFTA requires the Australian TGA to "prevent" marketing approval for a generic product whenever any type of patent is "claimed" for brand-named drug. Australian implementing legislation, however, while creating the required notification process, has imposed penalties for "evergreening" which resemble similar provisions in both the US and Canada[11].
Under the dispute resolution Chapter 21, a panel of three nominated trade lawyers will have the power to interpret compliance with obligations in the AUSFTA. Article 21.2(c) contains what is known in international trade law as a non-violation nullification of benefits ('NVNB') article[31]. Such articles allow dispute resolution proceedings to be commenced where only the spirit of the treaty had been broken, or more technically, legitimate expectations have been nullified [32]. Australia may be able to use this provision to argue that its legitimate expectation was that the AUSFTA would lead to no amendment of the National Health Act 1953 (Cth) and in particular to the mechanism therein of cost-effectiveness pricing of pharmaceuticals.
Potential impacts of the AUSFTA on Medicines in Australia
On the face of it, and as argued by the Australian government, the provisions of the AUSFTA represent procedural changes rather than substantive reform to current regulatory arrangements. In any event, there negotiations have seen principles such as recognition of pharmaceutical innovation set here in a unique public health context. Within the principles of Annex 2C(1), for example, 'innovation' is linked with high quality health care, 'affordability', 'accountability' and 'objectively demonstrated therapeutic significance'. This linkage is arguably reflective of the current Australian approach of defining innovation with regard to its comparative therapeutic value, that is its clinical and cost-effectiveness. Further, Annex 2C(1) commits both parties to promoting 'affordable' access to innovative drugs and to a recognition of innovation that may involve either competitive markets (that is, a market not dominated by monopolistic patents) or procedures that appropriately value the objectively demonstrated therapeutic significance of a pharmaceutical (such as, but not specifically referring to, the system under Australia's PBS).
However, the interpretive principles of Annex 2C(1), do not specifically refer to the PBS (unlike those in Annex 2C(2) on transparency). They are also sufficiently vague to allow considerable scope in interpreting what obligations they create. Despite its apparent centrality to the lobbying agenda of the brand-name pharmaceutical industry, 'innovative' is not defined explicitly (here or anywhere else in the text of the AUSFTA) and the precise obligations created by a requirement for stronger recognition of pharmaceutical 'innovation' are not clear. This possibly deliberate lack of clarity extends to other provisions of Annex 2C, such as the creation on an 'independent review process' and the 'Medicines Working Group'. The crucial concept of transparency, is also not unambiguously defined in the AUSFTA.
The problem, which should not be understated, is that where such differences in interpretation are structured into the agreement, disputes about expectations and obligations are only postponed, rather than resolved[33]. It may be important to consider, therefore, the effect of Annex 2C(1) framing such obligations on governments to recognise pharmaceutical "innovation" and the research and development necessary for it within the overarching obligation of industry to objectively prove the contribution of such products to overall public health. It could also be relevant to study whether these changes facilitate policy proposals with the capacity to diminish equity of access to essential medicines (contrary to the National Medicines Policy), leading to reduced health outcomes for elderly citizens and those reliant on such therapies for quality of life and productivity. Such proposals could include patient co-payment rises, means-tested co-payments, medicines savings accounts, changes to reference pricing, to the pricing of generic pharmaceuticals, or a diminution of the capacity of the PBAC to make cost-effectiveness recommendations[14].
If the principles and provisions of Annex 2C represent the 'spirit' of the AUSFTA regarding pharmaceuticals – it will be important to research to what extent that Australia could satisfactorily meet that spirit and continue to apply pharmaceutical reference pricing, or prohibit direct-to-consumer advertising. Policy suggestions that we could research here include the creation of a pharmaceutical "innovation" prize system outside the PBS.
The provisions in each area of the AUSFTA articulate with the provisions in other areas. Should the US determine that the spirit of Annex 2C is not being met, it is highly plausible that it could seek redress by invoking the NVNB clause in Chapter 21. In this context, an important component of our research will be to examine whether the new ss26C and D of the Therapeutic Goods Act 1989 (Cth) are a "dead-letter," as some have suggested (due to reasonable exceptions grounds and the uncertain incentives for Australian generic manufacturers to bring such actions), or whether these amendments may play an important role in clarifying Australia's legitimate expectations in this area, for the purposes of a subsequent NVNB trade dispute action under article 21.2(c).
The overall significance here is that future PBAC decisions not to list 'innovative' new drugs from US companies (because they were judged not cost-effective) will be made in the shadow of possible US trade retaliation in important areas such as manufacturing and agriculture[34]. What effect such a shadow might have on the deliberative processes of Australian regulators is difficult to predict and indicators of such pressure may need to be established.
Thus, while this version of the AUSFTA (a supervising committee under chapter 21 may recommend changes) does not ostensibly seek to modify the basic architecture of the PBS, it appears to give greater representation and greater weight to the needs of the private sector. It is not obvious in the wording of the provisions how the AUSFTA will achieve the mooted benefits and avoid the possible risks to the PBS. This lack of clarity has generated considerable uncertainty and much criticism. The regulatory changes required by the AUSFTA, to strengthen intellectual property protection for example, could increase the reward to manufacturers for innovation, but could also substantially reduce government capacity to apply price controls in the pharmaceutical sector. Without the bargaining authority afforded by expert cost-effectiveness evaluation and reference pricing, government capacity to sustain historically desirable and medically/socially acceptable sector outcomes is far from certain. One hypothesis is that the AUSFTA may result in increased industry investment, but perhaps only at the cost of reduced equity of access.
Framework for Evaluation of the AUSFTA's Medicines Impacts
Few studies have made direct measurements of the effects of trade agreements on access to medicines. The evidence for the putative benefits of stronger pharmaceutical patent rights – increased drug innovation, development of local drug research and development capacity and enhanced overall welfare of an introducing nation, is ambivalent. Some evidence suggests such strengthened intellectual property monopolies can stimulate invention, at least in encouraging incremental improvements by originators[35]. Other evidence suggests they freeze out future radical inventors[36]. Increased pharmaceutical intellectual property protection appears to have little positive impact on the level of local medicinal research and development[37]. A number of reasonably well-structured research projects conducted by competent scholars, in fact, have failed to find credible evidence that stronger intellectual property rights stimulates local pharmaceutical innovation[41]. Such rights, on the other hand, may substantially increase medicines prices in the countries that introduce it[38]. They can can also produce major changes in the local pharmaceutical industry that do not favour cheaper generic products[39]. This appears likely to produce major adverse health impacts for disadvantaged sectors of the population[40].
Overall, the welfare effects of global patent protection appear to be asymmetrical with the welfare in the inventing country rising with the extension of patent protection, while that of the introducing country falls by a proportionately greater amount[42]. The evidence rather suggests that the impact of strengthened intellectual property protection in developed nations such as Australia depends on the extent to which government regulation facilitates the continuance of generic pharmaceutical competition[43]. To what extent the viability of a nation's generic pharmaceutical industry should be resolved by the operation of market forces or lobbying from the brand name industry are major policy questions in this area[44].
While often not directly measurable, change in the character of Australian pharmaceutical regulation may be observable in how Australia's regulatory system materially and normatively responds to the changes required by the AUSFTA. We plan to observe the impact of the AUSFTA on key impact points of the regulation and governance of Australia's pharmaceutical sector and to track the effects of associated changes on drug expenditure, industry activity and medicine utilisation and affordability.
Many potential impacts are unlikely to be immediately observable; more probably it will be years before some changes are manifest. It is also probable that there will be sequelae, benign or otherwise, that are unable to be predicted. Our study, consequentially, will be ongoing with a wide focus, but provisional and responsive to what emerges as significant over time. Our broad interests – pharmacoeconomic, legal, public health, regulatory and socio-political – will be drawn on for relevant methodological and theoretical options for collecting and analysing qualitative and quantitative data from a range of primary and secondary sources.
Our plan on the quantitative side of the Project, is for the research to move from simply identifying association between the AUSFTA and changes in the price and supply of pharmaceuticals in Australia, to the more useful objective of assigning causation. The weakness in attempting to model alternative policy scenarios relates to the amount and quality of empiric data available and acceptance of the underlying assumptions. Emphasis, rather, will be on the ARC research scholar Andrew Searles constructing a theoretical framework for viewing the impact of the AUSFTA on the price and supply characteristics of pharmaceuticals in Australia drawing upon established economic theory, particularly in relation to public goods. We will also investigate the validity of the economic assumptions underpinning the AUSFTA medicines provisions. A pharmaceutical price index will be constructed taking into account the potential performance of its formulae under both axiomatic and economic approaches.
Regulation and Governance
We aim to identify changes to regulatory structure and process, particularly the application of PBS reference pricing, associated with interpretation of the AUSFTA provisions. Material changes range from the possible – amendments to or repeal of relevant legislation; to the probable – changes to the processes and relationships within and between, the TGA, the PBAC (and its subcommittees) and the PBPA. We will also investigate the impact of the AUSFTA on the normative order applied in the deliberative processes. Changes, for example, might involve modifying the PBAC its evaluative process in ways that are mutually beneficial (transparency) or less so (reward of innovation taking precedence over cost-effectiveness). The extent to which impacts relates to core social justice principles such as equity of access to essential medicines will be a major focus, as mentioned, of the work of ARC research scholar Warwick Neville.
Quantitative and qualitative methods will be used longitudinally to observe for such changes and associated outcomes. These could include variation in the number of actual and possible listings, number of rejections of asking price (introductory price and price readjustments) and changes to the number of applications that include non-clinical claims concerning price. The number and type of proposals aired in the media by government and industry concerning pharmaceutical regulation will also be studied. The extent to which such proposals rely on adequate economic or other research, or other forms of justification, may also be investigated.
One hypothesis is that the AUSFTA may lead to changes in the type of medicines-related provisions included in subsequent bilateral trade agreements. Our research in this context will also explore the extent to which "innovation" and/or cost-effectiveness evaluation of new pharmaceuticals can be considered global public goods and what type of long term strategies can be developed to enhance their rational development. The project, for example, may consider the regulatory and fiscal advantages of bilateral trade deals, such as the China-Australia Free Trade Agreement, including a medicines cost-effectiveness working committee to facilitate exchange of pharmacoeconomic expertise[45]. This, or proposals for a multilateral treaty on the same topic, in part, may provide an economic and social justice balance to the potential impact of the AUSFTA on the PBS processes[46]. Creation of a medicines cost-effectiveness treaty, or related committees in bilateral trade agreements, could be a policy change that promotes quality use of medicines in all nations so involved[47]. Qualitative methods of the Project in this area will include key stakeholder interviews and a follow-up questionnaire.
Industry Activity
We will examine the effects of the AUSFTA on the activity and returns of originator and generic manufacturers, including changes in profitability ratios, increases in employment and changes to Australia's pharmaceutical balance of trade. For originators, relevant indicators would include changes to monopoly rent for pharmaceutical patent holders; the number of applications to the TGA and PBS for listing of innovative patented products; changes to investment in research and development; and changes to expenditure on promotion and marketing. For generic manufacturers, we plan to observe for changes to the number of applications for marketing approval and changes to the timing of generic entry.
Drug Expenditure
The Project will observe for changes in Federal and State government pharmaceutical expenditure associated with AUSFTA provisions – increased patent protection leading to delays in generic entry for example. This will include calculating the opportunity cost for other health areas of increased Federal and State hospital expenditure on innovative medicines. Direct and indirect (changes to over-the-counter drugs) price effects will be monitored. This may involve observing the pricing trends of strategically selected brand name products nearing patent expiration and the rate, price and number of relevant generic market entrants compared against expected results. As mentioned, for this outcome component the ARC research scholar Andrew Searles will develop an Australian pharmaceutical price index ('PPI') in an Excel spreadsheet, its design allowing the user to define subgroups of medicines.
Drug Availability
We will examine whether AUSFTA required changes result in an increased availability of innovative drugs, faster access to subsidies for new prescription medicines and changes in the mix of generic and brand name drugs in the Australian market. This may also involve an independent expert evaluation of the innovative aspects of new drugs and post-marketing surveillance of associated treatment outcomes.
Drug Utilisation and Affordability
We will observe for changes to overall drug utilisation and for changes to the use of newer innovative drugs compared to existing therapies on the PBS. We will also observe for changes to out-of pocket patient charges, for example increases in prescription co-payments, linked to the increased PBS expenditure associated with use of innovative medicines. Australian medicine users will be surveyed for changes to affordability of medicines following increases in out-of-pocket cost.
Conclusion
As in other industrialised countries, regulation of the Australian pharmaceutical sector is an uneasy contingent system, its character reflecting the normative strength of private, relative to public, policy imperatives. With the recently implemented bilateral AUSFTA, however, the balance of such power may shift. The numerous material changes required by the AUSFTA in the pharmaceutical sector potentially aggregate to create a regulatory environment more attuned to encouraging private investment and profit-making. One hypothesis, in this context, is that the AUSFTA represents a normative shift that may dramatically change not only the character of Australian regulation but the protection of global public goods involved with assuring the general community obtains value for its pharmaceutical expenditure. It is uncertain to what degree a public health scheme with strong social justice traditions, like the PBS, can operate credibly under such changed conditions. The experience we detect is likely to be relevant to other countries with highly subsidised health care systems contemplating trade agreements with countries that have strong patent-protected pharmaceutical industries.
Plausibly the provisions of the AUSFTA represent minor procedural adjustments and leave the PBS intact. Some AUSFTA articles, however, clearly intend to create an expectation that distinct changes, to current Australian rather than US, pharmaceutical regulatory arrangements, will occur. Determining what alterations occur, and with what public health outcomes, will require careful empirical observation and theoretical analysis.
In summary, the potential exists for the AUSFTA to reshape the character of Australia's regulatory system concerning medicines- from a public good to a private rights-oriented system. Should the AUSFTA precipitate such a normative shift (particularly one away from scientific cost-effectiveness evaluation of pharmaceuticals) the regulatory implications are likely to be profound and resonate beyond Australia to impact on the health care sectors of other nations.
Funding statement
'The Impact of International Trade Agreements on the Regulation and Provision of Medicines in Australia' is a three-year study funded by the Australia Research Council's Discovery Grant (project DP0556635), the Australian National University and the University of Newcastle. The Chief Investigators are Dr Thomas Faunce (Project Director), Professor Peter Drahos and Professor David Henry. Australian Research Council funded PhD scholars under the Project are Andrew Searles and Warwick Neville.
Competing interests
The author(s) declare that they have no competing interests.
==== Refs
Mark Vaile Australian Trade minister Interview on ABC Radio Am program March 4 2003 last accessed 12 Sept 2005
Australian Government. Department of Foreign Affairs. The Australia-United States Free Trade Agreement: Pharmaceutical benefits Scheme (PBS) Outcomes last accessed 11 September 2005
Deputy US Trade representative J Shiner. Testimony Before the Full Committee of the House Committee on Ways and Means. June 16, 2004 last accessed 12 Sept 2005
Schneeman K Chief Executive of Medicines Australia-US Free Trade Agreement Business Group. PBS Off the Table. 27 August 2003 last accessed 12 Sept 2005
Senator Kyl Chairman of the Subcommittee on Health Care, Joint Hearing by the Subcommittees on Health Care and International Trade of the US Senate Committee on Finance April 27 2004 accessed 5.10.04
Hayes S Director of Strategic Relations at Medicines Australia. 'The Triumph is in the Text.' Presentation at The Australian APEC Centre Conference Report US Free Trade Agreement: New Opportunities and Impacts 1–2 March 2004 Canberra last accessed 11 Sept 2005
Trade Act 2002 19 USCS §3801–3802 107–210, δ2102(b)(8)(D) Medicare Prescription Drug Improvement and Modernization Act 2003 21 USCS conference agreement
Australian Senate Select Committee on the Free Trade Agreement between Australia and the United States of America. Final Report 2004 Commonwealth of Australia 119 para 4.62
Faunce TA Testimony before the Senate Select Committee on the Free Trade Agreement between Australia and the United States. Transcripts of Evidence 5 May and 21 June 2004 last accessed 11 Sept 2005
Outterson K 'Agony in the Antipodes. The Generic Drug Provisions of the Australia-US free Trade Agreement' Journal of Generic Medicines 2005 2 316 326
Faunce TA The Awful Truth About Evergreening The Age August 7 2004 last accessed 12 Sept 2005
US Free Trade Agreement Implementation Act 2004 (Cth) adding a new s26B and s26C and s26D to the Therapeutic Goods Act 1989 (Cth)
Drahos P Lokuge B Faunce T Goddard M Henry D 'Pharmaceuticals, Intellectual Property and Free Trade: The Case of the US-Australia Free Trade Agreement' Prometheus 2004 22 243 257 10.1080/0810902042000255705
Harvey KJ Faunce TA Lokuge B Drahos P 'Will the Australia-United States Free Trade Agreement Undermine the Pharmaceutical Benefits Scheme?' Medical Journal of Australia 2004 181 256 259 15347273
Department of Health and Ageing 'Frequently asked questions' via the internet. Accessed 29 June 2005
Sloan C A History of the Pharmaceutical Benefits Scheme 1947–1992 1995 Department of Human Services and Health, Canberra
AG Vic (ex rel Dale and ors) v Cth (the Pharmaceutical Benefits Case) 1945 71 CLR 237 Federal Council of the British Medical Association in Australia v Cth (1949) 79 CLR 201. Section 51xxiiiA inserted by Constitutional Referendum in 1946
Hunter T 'Pharmaceutical Benefits Legislation' The Economic Record 1965 41 412 425
Hill S Henry D Pekarsky B Mitchell A 'Economic evaluation of pharmaceuticals: what are reasonable standards for clinical evidence – the Australian experience.' British Journal of Clinical Pharmacology 1997 44 421 5 9384457 10.1046/j.1365-2125.1997.t01-1-00597.x
Henry DA Hill SR 'Assessing new health technologies: lessons to be learnt from drugs.' Medical Journal of Australia 1999 171 554 6 10816710
Birkett DJ Mitchell AS McManus P 'A Cost-Effectiveness Approach to Drug Subsidy and Pricing in Australia' Health Affairs 2001 20 104 114 11585158 10.1377/hlthaff.20.3.104
Mitchell A Henry D 'Is the drug 'cost-effective'?' Australian Prescriber 1992 15 74 5
Johannesson M Henry DA 'The Australian Guidelines for subsidisation of pharmaceuticals: the road to cost-effective prescribing?' PharmacoEconomics 1992 2 355 62 422-6 10147048
Harvey K Murray M Gardner H 'Medicinal Drug Policy' 1995 2 The Politics of Health: The Australian Experience 238
Henry DA 'Should Indirect Costs and Benefits be included?' PharmacoEconomics 1992 1 462 467 10147025
Hill SR Mitchell AS Henry DA 'Problems with the Interpretation of Pharmacoeconomic Analyses: A Review of Submissions to the Australian Pharmaceutical Benefits Scheme.' Journal of the American Medical Association 2000 283 2116 21 10791503 10.1001/jama.283.16.2116
Australian Government. Productivity Commission 2003, International Pharmaceutical Price Differences Research Report last accessed 11 September 2005
Drummond MF Bloom BS Freund DA Evans D Henry DA Dittus R 'Peer Review: Prescription Policy in Australia.' Health Affairs 1992 11 191 206 1483639 10.1377/hlthaff.11.4.191
US Department of Commerce, International Trade Administration Pharmaceutical Price Controls in OECD Countries Implications for US Consumers, Pricing, Research and Development and Innovation 2004 Washington DC
Australian Government, Department of Foreign affairs and Trade Australia-United States Free Trade Agreement Full text, reports, fact sheets, etc 2004
Faunce TA 'FTA and the PBS:The Senate's Crucial Role' Australian Policy Online 26 May 2005 last accessed 1 Oct 2005
Faunce TA Lokuge B 'Sting in the Tail of the FTA.' last accessed 3 Oct 2005
Trachtman JP 'The Domain of WTO Dispute Resolution' Harvard Int L J 1999 40 333 346
Faunce TA Bennett B, Tomossy GF 'Global Intellectual Property Protection of Innovative Pharmaceuticals: Challenges for Bioethics and Health Law' Springer Dordrecht 2005 Globalisation and helath: Challenges for Bioethics and Health Law 87 107
Faunce TA 'Time to Gear Up for the Generic v Brand Name Drugs Showdown' Sydney Morning Herald 7 February 2005 last accessed 1 Oct 2005
O'Donoghue T Scotchmer S Thisse J-F 'Patent Breadth, Patent Life and the Pace of Technological Progress' J of Economics and Management 1998 7 1 32
Lerner J 150 Years of Patent Protection, Harvard MA, Harvard Business School Working Paper 2001
Deardorff A 'Welfare Effects of Global Patent Protection' Economica 1992 59 35 51
Nogues J 'Patents and Pharmaceutical Drugs; Understanding the Pressures on Developing Countries World Bank Country Economics Department 1990
La Croix S Kawaura A 'Product Patent Reform and its Impact on Korea's Pharmaceutical Industry' Int Economics J 1996 10 109
Chaudhuri S Goldberg PK Jia P The Effects of Extending Intellectual Property Rights Protection to Developing Countries: A Case Study of the Indian Pharmaceutical Market 2003 Working Paper 10159 National Bureau of Economic Research
Branstetter LG Maskus KE, Reichman JH 'Do Stronger Patents Induce More Local innovation?' International Public Goods and Transfer of Technology Under a Globalized Intellectual Property Regime 2005 Cambridge University Press. Cambridge 309 320
Abbott FM Maskus KE, Reichman JH 'Managing the Hydra: The Herculean Task of Ensuring Acess to Essential Medicines' International Public Goods and Transfer of Technology Under a Globalized Intellectual Property Regime 2005 Cambridge University Press. Cambridge 393 424
Sakakihara M Bransteeter L 'Do Stronger Patents Induce More Innovation. Evidence From 1988 Japanese Patent law Reforms' NZEB Working paper 1999
Pazderka B 'Patent Protection and Pharmaceutical Research and Development Spending in Canada' Canadian Public Policy 1999 23 29 64
Faunce TA Submission P5 to Parliament of Australia, Senate Inquiry into Australia's Relationship with China 25 March 2005 last accessed 20 Sept 2005
Faunce TA "Cashing in on our pharmaceutical expertise: A cost-effectiveness committee in the China-Australia FTA' Canberra Times 11 April 2005 last accessed 1 Oct 2005
Faunce TA 'Pharmaceutical Development, Reference Pricing Systems and Intellectual Property' Medicine and Society 2005 18 5 8 Chinese language publication
|
16209703
|
PMC1276805
|
CC BY
|
2021-01-04 16:39:02
|
no
|
Global Health. 2005 Oct 6; 1:15
|
utf-8
|
Global Health
| 2,005 |
10.1186/1744-8603-1-15
|
oa_comm
|
==== Front
Global HealthGlobalization and Health1744-8603BioMed Central London 1744-8603-1-151620970310.1186/1744-8603-1-15DebateAssessing the impact of the Australia-United States Free Trade Agreement on Australian and global medicines policy Faunce Thomas [email protected] Evan [email protected] David [email protected] Peter [email protected] Andrew [email protected] Brita [email protected] Warwick [email protected] Globalisation and Health Project, Centre for Governance of Knowledge and Development Regulatory Institutions Network Australian National University, Acton, Canberra ACT, Australia2 Medical School and Law Faculty, The Australian National University, Acton, Canberra, ACT Australia3 Newcastle Institute of Public Health, University of Newcastle, Newcastle, New South Wales, Australia4 Clinical Pharmacology, School of Medical Practice and Population Health, University of Newcastle, Newcastle, New South Wales, Australia5 Centre for Regulation and Market Analysis, University of South Australia, Adelaide, South Australia, Australia2005 6 10 2005 1 15 15 21 7 2005 6 10 2005 Copyright © 2005 Faunce et al; licensee BioMed Central Ltd.2005Faunce et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
On 1 January 2005, a controversial trade agreement entered into force between Australia and the United States. Though heralded by the parties as facilitating the removal of barriers to free trade (in ways not achievable in multilateral fora), it also contained many trade-restricting intellectual property provisions and others uniquely related to altering pharmaceutical regulation and public health policy in Australia. The latter appear to have particularly focused on the world-respected process of federal government reimbursement after expert cost-effectiveness evaluation, popularly known as the Pharmaceutical Benefits Scheme ('PBS'). It remains uncertain what sort of impacts – if any – the Australia-United States Free Trade Agreement ('AUSFTA') will have on PBS processes such as reference pricing and their important role in facilitating equitable and affordable access to essential medicines.
This is now the field of inquiry for a major three year Australian Research Council ('ARC')-funded study bringing together a team of senior researchers in regulatory theory from the Australian National University and pharmacoeconomics from the University of Newcastle. The project proposes to monitor, assess and analyse the real and potential impacts of the AUSFTA in this area, providing Australian policy-makers with continuing expertise and options.
To the extent that the AUSFTA medicines provisions may represent an
important precedent in a global strategy by industry on
cost-effectiveness evaluation of pharmaceuticals, the study will also be
of great interest to policy makers in other jurisdictions.
==== Body
Introduction
The final text of the Australia-United States Free Trade Agreement ('AUSFTA') was signed in Washington on 18 May 2004, by the Australian Trade Minister and the United States Trade Representative. On 17 November 2004, the parties exchanged notes accepting their respective implementing processes and the agreement entered into force on 1 January 2005. The AUSFTA contained numerous provisions either directly or indirectly related to medicines regulation in Australia, particularly Annex 2C of Chapter Two, Chapter Seventeen on intellectual property and Chapter Twenty One on dispute resolution.
It remains uncertain whether the AUSFTA will have either a detrimental or beneficial impact on access to medicines and the promotion and maintenance of good health in Australia. There does, however, appear to have been a substantial difference in opinion between the Parties over procedural changes that would result in Australian medicines regulation.
Throughout the negotiations, the Australian Government's position was either that the government cost-effectiveness reimbursement system, the Pharmaceutical Benefits Scheme ('PBS'), would not be included in the AUSFTA, or that if it was, it was an item of public health policy whose core components would be protected[1]. After signature, the Australian government maintained that the fundamental architecture of the PBS remained unchanged. It acknowledged commitments to make improvements to the transparency and timeliness of PBS processes. It also affirmed its reasonable expectations that, as a result of the AUSFTA, Australian citizens would benefit from faster access to new prescription medicines, that the price of medicines on the PBS would not increase and that the text of the AUSFTA made no changes to the cost-effectiveness methods used to set PBS reimbursement levels[2].
On the other hand, the Deputy US Trade Representative stated to the US Congress:
The U.S.-Australia FTA is the first to include non-tariff market access provisions to address issues in the pharmaceutical sector. Recognizing the sensitivity of this issue, we drew on studies prepared by the Australian government to propose changes that would improve transparency and the regulatory procedures for listing new drugs in Australia. Under the FTA, the United States and Australia agreed to common principles on facilitating high quality health care and continued improvements in public health, including through government support for research and development in the pharmaceutical industry. We also agreed to establish a Medicines Working Group to discuss emerging health policy issues. Australia committed to specific steps to improve the transparency, accountability and promptness of the listing process, including establishment of an independent review of listing decisions[3].
Representatives of the multinational brand-name pharmaceutical industry, including its regional organisation Medicines Australia, claimed that there was no basis to claims that the US wanted the PBS dismantled[4]. They argued that the regulatory changes required by these areas of the AUSFTA would (a) help redress an alleged current undervaluing of pharmaceutical 'innovation' in Australian pricing arrangements and (b) stimulate locally-based research and development, as well as the local, mostly generic, pharmaceutical industry[5]. They asserted the negotiated modifications would make Australia's regulatory system more oriented to the global market pressures on industry, more responsible in its approach to intellectual property rights and so more attractive to private investment, resulting in a net welfare benefit[6].
Others, however, have pointed to US legislation requiring that nation's negotiators to seek in the AUSFTA provisions facilitating the "elimination of government measures such as price controls and reference pricing which deny full market access for United States [pharmaceutical] products"[7]. The Australian Senate Select Committee on the AUSFTA concluded:
While no single one of the specific commitments will create immediate and measurable price rises for the PBS, the new measures may well over time alter the bargaining power between the PBS and pharmaceutical companies. This may have long term ramifications that are not in the interest of Australian consumers[8].
Concern has been expressed about AUSFTA provisions with the potential to encourage higher medicines prices in Australia. These include provisions in chapter 17 (Intellectual Property) that expand the obligations of the Trade Related Intellectual Property Rights ('TRIPS') agreement by prohibiting parallel importation, restricting compulsory licensing to "national emergencies of extreme urgency," prohibiting generic manufacturers exporting to a patent-expired market when a domestic patent exists and increasing data exclusivity protections[9].
A significant additional worry for these commentators was article 17.10.4. For the first time in Australia, this linked generic regulatory market approval on quality and safety grounds with the patent status of the relevant brand name product[10]. This Hatch-Waxman-type provision was felt to risk brand name manufacturers "evergreening" soon-to-expire pharmaceutical patents, as had occurred after comparable regulations were introduced in jurisdictions such as the US and Canada[11]. The academic, community and parliamentary concern in Australia was so great on this issue, that it resulted in the Australian government passing "anti-evergreening" amendments to its AUSFTA implementing legislation. These imposed a $A10 million penalty for a bad faith challenge by a brand name manufacturer of a generic notification certificate under the new s26B of the Therapeutic Goods Act 1989 (Cth). They also allowed cost recovery in such circumstances by the Australian government[12].
Provisions in Annex 2C(1) emphasising the need for increased government recognition of pharmaceutical "innovation" and "research and development" were likewise viewed by such critics as having the potential to encourage brand name industry lobbying. This could potentially weaken, in the long term, the capacity of Australia's Pharmaceutical Benefits Advisory Committee ('PBAC') to reject, on clinical and cost effectiveness grounds, new medicines from inclusion in the government's PBS positive reimbursement list, or to reference their reimbursement price against older products with equivalent efficacy but much reduced price[13].
In this paper, we present a rationale and outline a draft plan for a three-year study, funded by the Australian Research Council ('ARC'), which will examine the impact of the AUSFTA on a range of regulatory, public health and industrial interests involved with access to medicines in Australia. An important initial point to make is that we consider the AUSFTA is best researched as a component of an ongoing process of interaction with Australia's medicines policy by the global pharmaceutical industry. This trade agreement should be viewed, in other words, either as a catalyst that may enhance the speed of regulatory change, or a tangible manifestation of industry lobbying principles that, till now, may have been more implicit. It would be misleading, in any event, to investigate the AUSFTA's potential impacts on Australian medicines policy in isolation of demonstrable long-term corporate strategies.
Some central issues our study will examine include to what extent the AUSFTA requires, facilitates, or is likely to result in, changes to Australia's generic pharmaceutical industry, as well as its PBS cost-effectiveness system of pharmaceutical regulation. We also aim to consider relevant net welfare gain or loss; whether the Australian community will get the same value-for-dollar spent on medicines, either through Commonwealth government reimbursement, hospital or patient purchase.
We propose to investigate these questions empirically (and provide a sound structure for the gradual acquisition of suitable data). This will be done first by identifying, with the assistance of qualitative interviews, actual or likely AUSFTA-associated changes to the structure and process of Australia's PBAC, as well as the marketing processes, development and sector competitiveness of generic pharmaceutical manufacturers in Australia. This aspect of the study will also review the legitimacy of such actual or proposed alterations by examining the history of Australia's PBS as a social justice measure designed to ensure universal access to essential medicines. We will also review such proposed changes for coherence with basic norms of bioethics, domestic law and international human rights. We shall then attempt to determine their actual or potential impact on a range of indicators including drug prices expenditure and affordability, drug availability and equity of access.
We hope that publishing an outline of our proposed study will further encourage policy discussion, facilitate collaborations and provide a template for governments of other countries planning to enter such agreements. Although much of the detail of the AUSFTA is specific to Australia, there are important elements likely to be relevant to future trade agreements involving the US or other countries that have a major vested interest in the production, export and rent generation associated with patented medicines. These include whether the strengthening of pharmaceutical intellectual property protection and weakening of medicines clinical and cost-effectiveness evaluation and/or reference pricing, necessarily involves a weakening of a nation's social and economic fabric, or the capacity of its population to age well and age productively.
Background: Australia's PBS in the Context of the AUSFTA
Australia's pharmaceutical sector is dominated by the operation of the federally funded PBS, which, after a process of clinical and cost-effectiveness evaluation contributing recommendations to price negotiation, provides reimbursement (currently approximately 75%) for around 80% of the prescription medicines used in Australia[15]. The PBS does not restrict market access, but facilitates maximisation of sales volume for listed products. In developing relevant price indices, our study will also take into account AUSFTA impacts on prices for hospital-used medicines (which can be calculated from the PBS reimbursement price less the minimum safety net value) and predict expenditure on medicines costed under the co-payment level.
Central to our analysis of the impact of the AUSFTA on medicines in Australia, however, will be an evaluation of its effect on the PBS. Australia's PBS was established as a free formulary of essential drugs after the Second World War by the Curtin-Chifley federal administrations[16]. It was a social justice measure designed to ensure that all Australian citizens gained access to affordable, essential medicines. Legislation to create the PBS had to survive two High Court challenges and required a successful Constitutional referendum[17]. Successive Commonwealth governments used and built upon the 1940s enactments, before a conservative party enacted the National Health Act 1953 (Cth) ('National Health Act')[18]. This is an extremely important point, that will be focused on by ARC research scholar Warwick Neville. The PBS is one of the few examples of public health policy in Australia's history that appears to have an unequivocal democratic mandate. Such an historical-jurisprudential perspective on the social justice aspects of the PBS will be a unique and distinctive feature of our analysis.
The modern PBS revolves around Part VII section 85 of the National Health Act 1953 (Cth). Section 101(4) of this Act states that the relevant Minister, upon the advice of the PBAC (with secretariat support from the Pharmaceutical Benefits Branch of the Department of Health and Ageing ('DOHA'), may declare a pharmaceutical listed on the PBS and so subject to a level of government reimbursement (except for a patient co-payment) that has been negotiated by the Pharmaceutical Benefits Pricing Authority ('PBPA'). Prior to the AUSFTA, PBS listing was required to only occur after the Therapeutic Goods Administration ('TGA') and the Australian Drug Evaluation Committee ('ADEC') had approved the relevant pharmaceutical's safety and efficacy[19]. The AUSFTA, through implementing amendments to the Therapeutic Goods Act 1989 (Cth), has already produced, as mentioned, a requirement that a market-entering generic manufacturer provide evidence that no counteracting patent is claimed, or that the brand name owner has been notified.
Central to the PBAC's clinical and cost-effectiveness evaluation is section 101(3) of the National Health Act. This requires the PBAC to base its recommendation on: 'the effectiveness and cost of therapy involving the use of the drug, preparation or class, including by comparing the effectiveness and cost [emphasis added] of that therapy with that of alternative therapies, whether or not involving the use of other drugs or preparations.' The section goes on to state that if the product is 'substantially more costly' than the selected comparator in its class it shall not be recommended by the PBAC for PBS listing 'unless...[it] provides a significant improvement in efficacy or reduction of toxicity over the alternative therapy or therapies.'
The PBS system that has evolved under this section is a variant of pharmaceutical reference pricing[20]. The pharmaceutical manufacturer's ('sponsor's') submission to the PBAC nominates a disease indication (and relevant subsets involving patient characteristics) as well as a listing price supposedly based on the pharmaceutical company's assessment of the best relevant available data on clinical effect against a comparator. The comparator is generally the drug most prescribed on the PBS for the same indication, but may be the standard medical (non-drug) treatment. Pharmaceutical companies tend to prefer comparisons against the most expensive drug with the best 'head-to-head' data, rather than the compound that is most pharmacologically similar[21]. Part of our initial task will be to determine what aspects, if any, of this important public health-related process could be subjected to pressure emerging from the AUSFTA. It will be an interesting threshold question to ascertain to what extent members of the PBAC were aware of, or consulted in, the development of the AUSFTA articles relevant to the PBS.
The PBAC's expert reviewers evaluate whether any of the assumptions in the submission are unjustifiable and create simulations to assess the incremental cost-effective ratio (the additional cost for an additional beneficial effect, or Quality of Life Years ('QALY') gained)[22]. The reports of these experts ('pink pages') are then passed back to be reviewed by the PBAC, along with an industry response to them ('blue pages') and the summary from the Economic Sub-Committee ('ESC') in the 'green pages.' The process is designed to take six weeks and follows guidelines set out on the PBS website. The extent to which such guidelines alter as a result of the AUSFTA and what impact this has on the PBAC process will be another aspect of our study.
The PBAC may ask the pharmaceutical manufacturer for additional information, but has no legal power to compel its production, even if not covered by 'commercial-in-confidence' protections. The pharmaceutical manufacturer may be claiming a price premium because of a claimed additional benefit (that is improved effectiveness, better adverse event profile or delivery system) conferred by the new product over its therapeutic rivals. One hypothesis we hope to test is that brand-name pharmaceutical manufacturers may use Annex 2C(1) of the AUSFTA to seek price premiums for alleged "innovation" as a separate issue from cost-effectiveness. If the relevant evidence submitted does not clearly establish clinical and cost effectiveness, then a process of cost minimization is undertaken[23]. The extent to which members of the PBAC are aware of any broader, strategic agenda of the brand name pharmaceutical industry in making individual submissions (for example on reference pricing), will also be an aspect of our research in this area.
Once a decision is made to list the drug, the PBPA then evaluates the requested price against an international benchmark price for drugs in that class. Thus, under the PBS system, the members of the PBAC and ESC use pharmacoeconomic analysis to determine the community value of a new drug against an agreed comparator therapy, while the monopsony bargaining power of the PBPA is used to counter the increasingly prolonged and wide potential for monopoly rents accorded to brand name pharmaceutical patent holders[24]. For brand name manufacturers, the process of bringing a new patented product to the Australian market at a higher price than currently available medicines may be lengthy, expensive and uncertain. However PBS listing provides a secure foothold in a substantial market, particularly (due to reference pricing) for the generic pharmaceutical industry[25].
The predominant norms underlying the current Australian system are cost-containment, efficiency and equity. When evaluated against such standards the PBS has performed well[26]. The considerable monopsony bargaining power it offers to government has resulted in lower overall prices for the medicines listed on the PBS. Lacking a capped budget and dependent upon prescribers following the approved guidelines, the PBS process, however, was viewed, prior to the AUSFTA, as primarily involved with quality use of medicines, rather than government cost containment[27].
Brand-name manufacturers have argued that PBS-type systems involving reference pricing, create a regulatory environment hostile to investment and innovation. Specifically, they claim, as they have in other jurisdictions, that reference pricing makes it progressively more difficult for innovative brand-name pharmaceuticals to enter the market at a price sufficient to recoup the cost of research and development. The low prices achieved through such tactics have also been alleged by such manufacturers to reduce the potential for locally based industry expansion and to risk eventual precipitation of a withdrawal of international manufacturers from the sector. They claim that Australia can only achieve the low price it currently commands for innovation by opportunistically "free-riding" on the research and development spending of developed nations, such as the US. The US Department of Commerce has recently produced a report criticising medicines price controls in OECD countries, which applies the same arguments to those jurisdictions[28]. These claims will need to be tested against available evidence.
Medicines-Related Provisions of the AUSFTA
The AUSFTA is one of a number of recent bilateral agreements sought by the US in its strategy of negotiating stronger intellectual property rights ('IPRs') with smaller trading partners. So-called "TRIPS-Plus" IPRs, seen by the US as essential for the protection of the monopoly rent it draws from intangible assets such as pharmaceutical patents, are an important feature of the AUSFTA. The AUSFTA medicines provisions also arise from an intention on the part of the US pharmaceutical industry, through the USTR, to modify government evaluative structures and processes to increase communication between industry and regulators while strengthening the financing of innovation through research and development. The extent to which this aim succeeded is debateable.
The AUSFTA, to summarise, contains approximately fifty provisions in four areas relevant to Australia's pharmaceutical sector: Annex 2C (Pharmaceuticals); the side-letters between the Australian Trade Minister and US Trade Ambassador; Chapter 17 (Intellectual Property Rights); and Chapter 21 (Dispute Resolution Procedures)[29].
Annex 2C(1) begins by articulating one overarching principle – that the parties to the agreement are '...committed to facilitating high quality health care and continued improvements in public health for their nationals.' It then mentions four subsidiary principles:
a) [recognising] the important role played by innovative pharmaceutical products in delivering high quality health care;
b) [recognising] the importance of research and development in the pharmaceutical industry and of appropriate government support, including through intellectual property protection and other policies;
c) the need to promote timely and affordable access to innovative pharmaceuticals through transparent, expeditious, and accountable procedures, without impeding a Party's ability to apply appropriate standards of quality, safety, and efficacy; and
d) the need to recognize the value of innovative pharmaceuticals through the operation of competitive markets or by adopting or maintaining procedures that appropriately value the objectively demonstrated therapeutic significance of a pharmaceutical.
Other subsections of Annex 2C, and the associated exchange of letters, relate to increased opportunities for a manufacturer to interact with regulators. This includes Australia providing manufacturers with an opportunity for hearings before the PBAC during the application for PBS listing; providing an opportunity for an independent review process following a negative PBAC price determination; the creation of a Medicines Working Group with health officials from each country in dialogue about aspects of Australia's regulatory mechanisms; an ongoing dialogue between the TGA and the US Food and Drug Administration on the issue of making pharmaceutical innovation 'quickly available'; and finally, Annex 2C(5) permits a pharmaceutical manufacturer to disseminate information about pharmaceutical innovation via the Internet.
Chapter 17 of the AUSFTA includes intellectual property provisions aimed specifically at Australia's pharmaceutical sector. Parallel importation, as already mentioned, is prohibited; compulsory licensing of pharmaceuticals is restricted to a standard more stringent than that applying in TRIPS ("national emergencies of extreme urgency"); generic production of domestic-patented drugs for export to jurisdictions where patents have already expired is prevented. Data exclusivity is extended, as is patent protection where there have been delays in issuing marketing approval[30]. Article 17.10.4 of the AUSFTA requires the Australian TGA to "prevent" marketing approval for a generic product whenever any type of patent is "claimed" for brand-named drug. Australian implementing legislation, however, while creating the required notification process, has imposed penalties for "evergreening" which resemble similar provisions in both the US and Canada[11].
Under the dispute resolution Chapter 21, a panel of three nominated trade lawyers will have the power to interpret compliance with obligations in the AUSFTA. Article 21.2(c) contains what is known in international trade law as a non-violation nullification of benefits ('NVNB') article[31]. Such articles allow dispute resolution proceedings to be commenced where only the spirit of the treaty had been broken, or more technically, legitimate expectations have been nullified [32]. Australia may be able to use this provision to argue that its legitimate expectation was that the AUSFTA would lead to no amendment of the National Health Act 1953 (Cth) and in particular to the mechanism therein of cost-effectiveness pricing of pharmaceuticals.
Potential impacts of the AUSFTA on Medicines in Australia
On the face of it, and as argued by the Australian government, the provisions of the AUSFTA represent procedural changes rather than substantive reform to current regulatory arrangements. In any event, there negotiations have seen principles such as recognition of pharmaceutical innovation set here in a unique public health context. Within the principles of Annex 2C(1), for example, 'innovation' is linked with high quality health care, 'affordability', 'accountability' and 'objectively demonstrated therapeutic significance'. This linkage is arguably reflective of the current Australian approach of defining innovation with regard to its comparative therapeutic value, that is its clinical and cost-effectiveness. Further, Annex 2C(1) commits both parties to promoting 'affordable' access to innovative drugs and to a recognition of innovation that may involve either competitive markets (that is, a market not dominated by monopolistic patents) or procedures that appropriately value the objectively demonstrated therapeutic significance of a pharmaceutical (such as, but not specifically referring to, the system under Australia's PBS).
However, the interpretive principles of Annex 2C(1), do not specifically refer to the PBS (unlike those in Annex 2C(2) on transparency). They are also sufficiently vague to allow considerable scope in interpreting what obligations they create. Despite its apparent centrality to the lobbying agenda of the brand-name pharmaceutical industry, 'innovative' is not defined explicitly (here or anywhere else in the text of the AUSFTA) and the precise obligations created by a requirement for stronger recognition of pharmaceutical 'innovation' are not clear. This possibly deliberate lack of clarity extends to other provisions of Annex 2C, such as the creation on an 'independent review process' and the 'Medicines Working Group'. The crucial concept of transparency, is also not unambiguously defined in the AUSFTA.
The problem, which should not be understated, is that where such differences in interpretation are structured into the agreement, disputes about expectations and obligations are only postponed, rather than resolved[33]. It may be important to consider, therefore, the effect of Annex 2C(1) framing such obligations on governments to recognise pharmaceutical "innovation" and the research and development necessary for it within the overarching obligation of industry to objectively prove the contribution of such products to overall public health. It could also be relevant to study whether these changes facilitate policy proposals with the capacity to diminish equity of access to essential medicines (contrary to the National Medicines Policy), leading to reduced health outcomes for elderly citizens and those reliant on such therapies for quality of life and productivity. Such proposals could include patient co-payment rises, means-tested co-payments, medicines savings accounts, changes to reference pricing, to the pricing of generic pharmaceuticals, or a diminution of the capacity of the PBAC to make cost-effectiveness recommendations[14].
If the principles and provisions of Annex 2C represent the 'spirit' of the AUSFTA regarding pharmaceuticals – it will be important to research to what extent that Australia could satisfactorily meet that spirit and continue to apply pharmaceutical reference pricing, or prohibit direct-to-consumer advertising. Policy suggestions that we could research here include the creation of a pharmaceutical "innovation" prize system outside the PBS.
The provisions in each area of the AUSFTA articulate with the provisions in other areas. Should the US determine that the spirit of Annex 2C is not being met, it is highly plausible that it could seek redress by invoking the NVNB clause in Chapter 21. In this context, an important component of our research will be to examine whether the new ss26C and D of the Therapeutic Goods Act 1989 (Cth) are a "dead-letter," as some have suggested (due to reasonable exceptions grounds and the uncertain incentives for Australian generic manufacturers to bring such actions), or whether these amendments may play an important role in clarifying Australia's legitimate expectations in this area, for the purposes of a subsequent NVNB trade dispute action under article 21.2(c).
The overall significance here is that future PBAC decisions not to list 'innovative' new drugs from US companies (because they were judged not cost-effective) will be made in the shadow of possible US trade retaliation in important areas such as manufacturing and agriculture[34]. What effect such a shadow might have on the deliberative processes of Australian regulators is difficult to predict and indicators of such pressure may need to be established.
Thus, while this version of the AUSFTA (a supervising committee under chapter 21 may recommend changes) does not ostensibly seek to modify the basic architecture of the PBS, it appears to give greater representation and greater weight to the needs of the private sector. It is not obvious in the wording of the provisions how the AUSFTA will achieve the mooted benefits and avoid the possible risks to the PBS. This lack of clarity has generated considerable uncertainty and much criticism. The regulatory changes required by the AUSFTA, to strengthen intellectual property protection for example, could increase the reward to manufacturers for innovation, but could also substantially reduce government capacity to apply price controls in the pharmaceutical sector. Without the bargaining authority afforded by expert cost-effectiveness evaluation and reference pricing, government capacity to sustain historically desirable and medically/socially acceptable sector outcomes is far from certain. One hypothesis is that the AUSFTA may result in increased industry investment, but perhaps only at the cost of reduced equity of access.
Framework for Evaluation of the AUSFTA's Medicines Impacts
Few studies have made direct measurements of the effects of trade agreements on access to medicines. The evidence for the putative benefits of stronger pharmaceutical patent rights – increased drug innovation, development of local drug research and development capacity and enhanced overall welfare of an introducing nation, is ambivalent. Some evidence suggests such strengthened intellectual property monopolies can stimulate invention, at least in encouraging incremental improvements by originators[35]. Other evidence suggests they freeze out future radical inventors[36]. Increased pharmaceutical intellectual property protection appears to have little positive impact on the level of local medicinal research and development[37]. A number of reasonably well-structured research projects conducted by competent scholars, in fact, have failed to find credible evidence that stronger intellectual property rights stimulates local pharmaceutical innovation[41]. Such rights, on the other hand, may substantially increase medicines prices in the countries that introduce it[38]. They can can also produce major changes in the local pharmaceutical industry that do not favour cheaper generic products[39]. This appears likely to produce major adverse health impacts for disadvantaged sectors of the population[40].
Overall, the welfare effects of global patent protection appear to be asymmetrical with the welfare in the inventing country rising with the extension of patent protection, while that of the introducing country falls by a proportionately greater amount[42]. The evidence rather suggests that the impact of strengthened intellectual property protection in developed nations such as Australia depends on the extent to which government regulation facilitates the continuance of generic pharmaceutical competition[43]. To what extent the viability of a nation's generic pharmaceutical industry should be resolved by the operation of market forces or lobbying from the brand name industry are major policy questions in this area[44].
While often not directly measurable, change in the character of Australian pharmaceutical regulation may be observable in how Australia's regulatory system materially and normatively responds to the changes required by the AUSFTA. We plan to observe the impact of the AUSFTA on key impact points of the regulation and governance of Australia's pharmaceutical sector and to track the effects of associated changes on drug expenditure, industry activity and medicine utilisation and affordability.
Many potential impacts are unlikely to be immediately observable; more probably it will be years before some changes are manifest. It is also probable that there will be sequelae, benign or otherwise, that are unable to be predicted. Our study, consequentially, will be ongoing with a wide focus, but provisional and responsive to what emerges as significant over time. Our broad interests – pharmacoeconomic, legal, public health, regulatory and socio-political – will be drawn on for relevant methodological and theoretical options for collecting and analysing qualitative and quantitative data from a range of primary and secondary sources.
Our plan on the quantitative side of the Project, is for the research to move from simply identifying association between the AUSFTA and changes in the price and supply of pharmaceuticals in Australia, to the more useful objective of assigning causation. The weakness in attempting to model alternative policy scenarios relates to the amount and quality of empiric data available and acceptance of the underlying assumptions. Emphasis, rather, will be on the ARC research scholar Andrew Searles constructing a theoretical framework for viewing the impact of the AUSFTA on the price and supply characteristics of pharmaceuticals in Australia drawing upon established economic theory, particularly in relation to public goods. We will also investigate the validity of the economic assumptions underpinning the AUSFTA medicines provisions. A pharmaceutical price index will be constructed taking into account the potential performance of its formulae under both axiomatic and economic approaches.
Regulation and Governance
We aim to identify changes to regulatory structure and process, particularly the application of PBS reference pricing, associated with interpretation of the AUSFTA provisions. Material changes range from the possible – amendments to or repeal of relevant legislation; to the probable – changes to the processes and relationships within and between, the TGA, the PBAC (and its subcommittees) and the PBPA. We will also investigate the impact of the AUSFTA on the normative order applied in the deliberative processes. Changes, for example, might involve modifying the PBAC its evaluative process in ways that are mutually beneficial (transparency) or less so (reward of innovation taking precedence over cost-effectiveness). The extent to which impacts relates to core social justice principles such as equity of access to essential medicines will be a major focus, as mentioned, of the work of ARC research scholar Warwick Neville.
Quantitative and qualitative methods will be used longitudinally to observe for such changes and associated outcomes. These could include variation in the number of actual and possible listings, number of rejections of asking price (introductory price and price readjustments) and changes to the number of applications that include non-clinical claims concerning price. The number and type of proposals aired in the media by government and industry concerning pharmaceutical regulation will also be studied. The extent to which such proposals rely on adequate economic or other research, or other forms of justification, may also be investigated.
One hypothesis is that the AUSFTA may lead to changes in the type of medicines-related provisions included in subsequent bilateral trade agreements. Our research in this context will also explore the extent to which "innovation" and/or cost-effectiveness evaluation of new pharmaceuticals can be considered global public goods and what type of long term strategies can be developed to enhance their rational development. The project, for example, may consider the regulatory and fiscal advantages of bilateral trade deals, such as the China-Australia Free Trade Agreement, including a medicines cost-effectiveness working committee to facilitate exchange of pharmacoeconomic expertise[45]. This, or proposals for a multilateral treaty on the same topic, in part, may provide an economic and social justice balance to the potential impact of the AUSFTA on the PBS processes[46]. Creation of a medicines cost-effectiveness treaty, or related committees in bilateral trade agreements, could be a policy change that promotes quality use of medicines in all nations so involved[47]. Qualitative methods of the Project in this area will include key stakeholder interviews and a follow-up questionnaire.
Industry Activity
We will examine the effects of the AUSFTA on the activity and returns of originator and generic manufacturers, including changes in profitability ratios, increases in employment and changes to Australia's pharmaceutical balance of trade. For originators, relevant indicators would include changes to monopoly rent for pharmaceutical patent holders; the number of applications to the TGA and PBS for listing of innovative patented products; changes to investment in research and development; and changes to expenditure on promotion and marketing. For generic manufacturers, we plan to observe for changes to the number of applications for marketing approval and changes to the timing of generic entry.
Drug Expenditure
The Project will observe for changes in Federal and State government pharmaceutical expenditure associated with AUSFTA provisions – increased patent protection leading to delays in generic entry for example. This will include calculating the opportunity cost for other health areas of increased Federal and State hospital expenditure on innovative medicines. Direct and indirect (changes to over-the-counter drugs) price effects will be monitored. This may involve observing the pricing trends of strategically selected brand name products nearing patent expiration and the rate, price and number of relevant generic market entrants compared against expected results. As mentioned, for this outcome component the ARC research scholar Andrew Searles will develop an Australian pharmaceutical price index ('PPI') in an Excel spreadsheet, its design allowing the user to define subgroups of medicines.
Drug Availability
We will examine whether AUSFTA required changes result in an increased availability of innovative drugs, faster access to subsidies for new prescription medicines and changes in the mix of generic and brand name drugs in the Australian market. This may also involve an independent expert evaluation of the innovative aspects of new drugs and post-marketing surveillance of associated treatment outcomes.
Drug Utilisation and Affordability
We will observe for changes to overall drug utilisation and for changes to the use of newer innovative drugs compared to existing therapies on the PBS. We will also observe for changes to out-of pocket patient charges, for example increases in prescription co-payments, linked to the increased PBS expenditure associated with use of innovative medicines. Australian medicine users will be surveyed for changes to affordability of medicines following increases in out-of-pocket cost.
Conclusion
As in other industrialised countries, regulation of the Australian pharmaceutical sector is an uneasy contingent system, its character reflecting the normative strength of private, relative to public, policy imperatives. With the recently implemented bilateral AUSFTA, however, the balance of such power may shift. The numerous material changes required by the AUSFTA in the pharmaceutical sector potentially aggregate to create a regulatory environment more attuned to encouraging private investment and profit-making. One hypothesis, in this context, is that the AUSFTA represents a normative shift that may dramatically change not only the character of Australian regulation but the protection of global public goods involved with assuring the general community obtains value for its pharmaceutical expenditure. It is uncertain to what degree a public health scheme with strong social justice traditions, like the PBS, can operate credibly under such changed conditions. The experience we detect is likely to be relevant to other countries with highly subsidised health care systems contemplating trade agreements with countries that have strong patent-protected pharmaceutical industries.
Plausibly the provisions of the AUSFTA represent minor procedural adjustments and leave the PBS intact. Some AUSFTA articles, however, clearly intend to create an expectation that distinct changes, to current Australian rather than US, pharmaceutical regulatory arrangements, will occur. Determining what alterations occur, and with what public health outcomes, will require careful empirical observation and theoretical analysis.
In summary, the potential exists for the AUSFTA to reshape the character of Australia's regulatory system concerning medicines- from a public good to a private rights-oriented system. Should the AUSFTA precipitate such a normative shift (particularly one away from scientific cost-effectiveness evaluation of pharmaceuticals) the regulatory implications are likely to be profound and resonate beyond Australia to impact on the health care sectors of other nations.
Funding statement
'The Impact of International Trade Agreements on the Regulation and Provision of Medicines in Australia' is a three-year study funded by the Australia Research Council's Discovery Grant (project DP0556635), the Australian National University and the University of Newcastle. The Chief Investigators are Dr Thomas Faunce (Project Director), Professor Peter Drahos and Professor David Henry. Australian Research Council funded PhD scholars under the Project are Andrew Searles and Warwick Neville.
Competing interests
The author(s) declare that they have no competing interests.
==== Refs
Mark Vaile Australian Trade minister Interview on ABC Radio Am program March 4 2003 last accessed 12 Sept 2005
Australian Government. Department of Foreign Affairs. The Australia-United States Free Trade Agreement: Pharmaceutical benefits Scheme (PBS) Outcomes last accessed 11 September 2005
Deputy US Trade representative J Shiner. Testimony Before the Full Committee of the House Committee on Ways and Means. June 16, 2004 last accessed 12 Sept 2005
Schneeman K Chief Executive of Medicines Australia-US Free Trade Agreement Business Group. PBS Off the Table. 27 August 2003 last accessed 12 Sept 2005
Senator Kyl Chairman of the Subcommittee on Health Care, Joint Hearing by the Subcommittees on Health Care and International Trade of the US Senate Committee on Finance April 27 2004 accessed 5.10.04
Hayes S Director of Strategic Relations at Medicines Australia. 'The Triumph is in the Text.' Presentation at The Australian APEC Centre Conference Report US Free Trade Agreement: New Opportunities and Impacts 1–2 March 2004 Canberra last accessed 11 Sept 2005
Trade Act 2002 19 USCS §3801–3802 107–210, δ2102(b)(8)(D) Medicare Prescription Drug Improvement and Modernization Act 2003 21 USCS conference agreement
Australian Senate Select Committee on the Free Trade Agreement between Australia and the United States of America. Final Report 2004 Commonwealth of Australia 119 para 4.62
Faunce TA Testimony before the Senate Select Committee on the Free Trade Agreement between Australia and the United States. Transcripts of Evidence 5 May and 21 June 2004 last accessed 11 Sept 2005
Outterson K 'Agony in the Antipodes. The Generic Drug Provisions of the Australia-US free Trade Agreement' Journal of Generic Medicines 2005 2 316 326
Faunce TA The Awful Truth About Evergreening The Age August 7 2004 last accessed 12 Sept 2005
US Free Trade Agreement Implementation Act 2004 (Cth) adding a new s26B and s26C and s26D to the Therapeutic Goods Act 1989 (Cth)
Drahos P Lokuge B Faunce T Goddard M Henry D 'Pharmaceuticals, Intellectual Property and Free Trade: The Case of the US-Australia Free Trade Agreement' Prometheus 2004 22 243 257 10.1080/0810902042000255705
Harvey KJ Faunce TA Lokuge B Drahos P 'Will the Australia-United States Free Trade Agreement Undermine the Pharmaceutical Benefits Scheme?' Medical Journal of Australia 2004 181 256 259 15347273
Department of Health and Ageing 'Frequently asked questions' via the internet. Accessed 29 June 2005
Sloan C A History of the Pharmaceutical Benefits Scheme 1947–1992 1995 Department of Human Services and Health, Canberra
AG Vic (ex rel Dale and ors) v Cth (the Pharmaceutical Benefits Case) 1945 71 CLR 237 Federal Council of the British Medical Association in Australia v Cth (1949) 79 CLR 201. Section 51xxiiiA inserted by Constitutional Referendum in 1946
Hunter T 'Pharmaceutical Benefits Legislation' The Economic Record 1965 41 412 425
Hill S Henry D Pekarsky B Mitchell A 'Economic evaluation of pharmaceuticals: what are reasonable standards for clinical evidence – the Australian experience.' British Journal of Clinical Pharmacology 1997 44 421 5 9384457 10.1046/j.1365-2125.1997.t01-1-00597.x
Henry DA Hill SR 'Assessing new health technologies: lessons to be learnt from drugs.' Medical Journal of Australia 1999 171 554 6 10816710
Birkett DJ Mitchell AS McManus P 'A Cost-Effectiveness Approach to Drug Subsidy and Pricing in Australia' Health Affairs 2001 20 104 114 11585158 10.1377/hlthaff.20.3.104
Mitchell A Henry D 'Is the drug 'cost-effective'?' Australian Prescriber 1992 15 74 5
Johannesson M Henry DA 'The Australian Guidelines for subsidisation of pharmaceuticals: the road to cost-effective prescribing?' PharmacoEconomics 1992 2 355 62 422-6 10147048
Harvey K Murray M Gardner H 'Medicinal Drug Policy' 1995 2 The Politics of Health: The Australian Experience 238
Henry DA 'Should Indirect Costs and Benefits be included?' PharmacoEconomics 1992 1 462 467 10147025
Hill SR Mitchell AS Henry DA 'Problems with the Interpretation of Pharmacoeconomic Analyses: A Review of Submissions to the Australian Pharmaceutical Benefits Scheme.' Journal of the American Medical Association 2000 283 2116 21 10791503 10.1001/jama.283.16.2116
Australian Government. Productivity Commission 2003, International Pharmaceutical Price Differences Research Report last accessed 11 September 2005
Drummond MF Bloom BS Freund DA Evans D Henry DA Dittus R 'Peer Review: Prescription Policy in Australia.' Health Affairs 1992 11 191 206 1483639 10.1377/hlthaff.11.4.191
US Department of Commerce, International Trade Administration Pharmaceutical Price Controls in OECD Countries Implications for US Consumers, Pricing, Research and Development and Innovation 2004 Washington DC
Australian Government, Department of Foreign affairs and Trade Australia-United States Free Trade Agreement Full text, reports, fact sheets, etc 2004
Faunce TA 'FTA and the PBS:The Senate's Crucial Role' Australian Policy Online 26 May 2005 last accessed 1 Oct 2005
Faunce TA Lokuge B 'Sting in the Tail of the FTA.' last accessed 3 Oct 2005
Trachtman JP 'The Domain of WTO Dispute Resolution' Harvard Int L J 1999 40 333 346
Faunce TA Bennett B, Tomossy GF 'Global Intellectual Property Protection of Innovative Pharmaceuticals: Challenges for Bioethics and Health Law' Springer Dordrecht 2005 Globalisation and helath: Challenges for Bioethics and Health Law 87 107
Faunce TA 'Time to Gear Up for the Generic v Brand Name Drugs Showdown' Sydney Morning Herald 7 February 2005 last accessed 1 Oct 2005
O'Donoghue T Scotchmer S Thisse J-F 'Patent Breadth, Patent Life and the Pace of Technological Progress' J of Economics and Management 1998 7 1 32
Lerner J 150 Years of Patent Protection, Harvard MA, Harvard Business School Working Paper 2001
Deardorff A 'Welfare Effects of Global Patent Protection' Economica 1992 59 35 51
Nogues J 'Patents and Pharmaceutical Drugs; Understanding the Pressures on Developing Countries World Bank Country Economics Department 1990
La Croix S Kawaura A 'Product Patent Reform and its Impact on Korea's Pharmaceutical Industry' Int Economics J 1996 10 109
Chaudhuri S Goldberg PK Jia P The Effects of Extending Intellectual Property Rights Protection to Developing Countries: A Case Study of the Indian Pharmaceutical Market 2003 Working Paper 10159 National Bureau of Economic Research
Branstetter LG Maskus KE, Reichman JH 'Do Stronger Patents Induce More Local innovation?' International Public Goods and Transfer of Technology Under a Globalized Intellectual Property Regime 2005 Cambridge University Press. Cambridge 309 320
Abbott FM Maskus KE, Reichman JH 'Managing the Hydra: The Herculean Task of Ensuring Acess to Essential Medicines' International Public Goods and Transfer of Technology Under a Globalized Intellectual Property Regime 2005 Cambridge University Press. Cambridge 393 424
Sakakihara M Bransteeter L 'Do Stronger Patents Induce More Innovation. Evidence From 1988 Japanese Patent law Reforms' NZEB Working paper 1999
Pazderka B 'Patent Protection and Pharmaceutical Research and Development Spending in Canada' Canadian Public Policy 1999 23 29 64
Faunce TA Submission P5 to Parliament of Australia, Senate Inquiry into Australia's Relationship with China 25 March 2005 last accessed 20 Sept 2005
Faunce TA "Cashing in on our pharmaceutical expertise: A cost-effectiveness committee in the China-Australia FTA' Canberra Times 11 April 2005 last accessed 1 Oct 2005
Faunce TA 'Pharmaceutical Development, Reference Pricing Systems and Intellectual Property' Medicine and Society 2005 18 5 8 Chinese language publication
|
0
|
PMC1276806
|
CC BY
|
2021-01-04 16:36:49
|
no
|
Harm Reduct J. 2005 Oct 11; 2:20
|
latin-1
|
Harm Reduct J
| 2,005 |
10.1186/1477-7517-2-20
|
oa_comm
|
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-611621266810.1186/1477-7525-3-61ResearchPatients' perspectives on how idiopathic pulmonary fibrosis affects the quality of their lives Swigris Jeffrey J [email protected] Anita L [email protected] Michael K [email protected] Sandra R [email protected] Division of Pulmonary Medicine, Interstitial Lung Disease Program, National Jewish Medical and Research Center, Denver, CO, USA2 University of California San Francisco, San Francisco, CA, USA3 VA Palo Alto Health Care System, Palo Alto, CA, USA4 Palo Alto Medical Foundation Research Institute, Palo Alto Medical Foundation, Palo Alto, CA, USA5 Division of Pulmonary and Critical Care Medicine, Stanford University, Stanford, CA, USA2005 7 10 2005 3 61 61 31 8 2005 7 10 2005 Copyright © 2005 Swigris et al; licensee BioMed Central Ltd.2005Swigris et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Idiopathic pulmonary fibrosis (IPF) is a debilitating lung disease with a survival of only three to five years from the time of diagnosis. Due to a paucity of studies, large gaps remain in our understanding of how IPF affects the quality of patients' lives. In only one other study did investigators ask patients directly for their perspectives on this topic. Further, currently there is no disease-specific instrument to measure health-related quality of life (HRQL) in patients with IPF. A carefully constructed measurement instrument, sensitive to underlying change, is needed for use in clinical trials and longitudinal studies of patients with IPF. Before developing such an instrument, researchers must improve their understanding of the relevant effects of IPF on patients' lives. On a broader scale, to provide the best care for people with IPF, clinicians must appreciate – from patients' perspectives – how this disease affects various aspects of their lives.
Methods
We used focus groups and individual in-depth interviews with 20 IPF patients to collect their perspectives on how IPF affects their lives (with a focus on the quality of their lives). We then analyzed these perspectives and organized them into a conceptual framework for describing HRQL in patients with IPF. Next, we examined how well certain existing measurement instruments – which have been administered to IPF patients in prior studies – covered the domains and topics our patients identified.
Results
In our framework, we identified 12 primary domains: symptoms, IPF therapy, sleep, exhaustion, forethought, employment and finances, dependence, family, sexual relations, social participation, mental and spiritual well-being, mortality. Each domain is composed of several topics, which describe how IPF affects patients' lives. When we compared the content of our conceptual framework with the existing instruments, we found the coverage of the existing instruments to be inadequate for several reasons, including they may tap general areas of QOL or HRQL but not some areas that appear to be most directly affected by IPF, and they include items that are relevant to symptoms and effects of other respiratory diseases but not IPF.
Conclusion
Collecting patients' perspectives and developing an organized inventory of the relevant effects of IPF on patients' lives provides valuable information for improving our understanding of the impact of this disease on patients and their loved ones. We believe our findings will help alert clinicians and researchers to IPF patients' experiences and concerns. Based on the comparison or our conceptual framework with the content of four existing instruments, it would appear that developing an IPF-specific measurement instrument is justified. Our conceptual framework for describing health-related quality of life in patients with IPF lays a solid foundation for constructing such an instrument.
Pulmonary fibrosisinterstitial lung diseasequality of lifeHealth-related quality of lifequalitative.
==== Body
Background
Idiopathic pulmonary fibrosis (IPF) is the most common of the idiopathic interstitial pneumonias. It is thought to affect about 30 persons per 100,000 people in the general population and perhaps as many as 175 per 100,000 people 75 years and older [1,2]. Breathlessness and irritating dry cough, often refractory to anti-tussive therapy, are classic symptoms. The disease course of IPF is somewhat variable, but patients commonly suffer progressive decline in lung function that culminates in respiratory failure and death. Median survival ranges from three to five years from the time of diagnosis [3-5]. Conventional IPF pharmacotherapy, which includes corticosteroids (i.e., prednisone) in combination with an immunosuppressive agent (e.g., azathioprine or cyclophosphamide), is largely ineffective, fraught with adverse effects, and often requires frequent laboratory monitoring [6].
The definition for quality of life (QOL) that we use in this manuscript refers to an individual's "holistic" evaluation of satisfaction with his own life [7]. Health-related quality of life (HRQL) incorporates the subjectively perceived impact of one's health – including aspects of well-being (or lack thereof) in the physical, mental, emotional, social, and spiritual facets of life – on life domains of perceived importance. Currently, there is no instrument that is claimed to be appropriate for evaluating HRQL specifically in patients with IPF. A handful of studies [8-13], using existing generic or non-IPF respiratory-specific measurement instruments, have assessed QOL or HRQL in patients with IPF, but none of these instruments have been shown to be sensitive to disease progression or to treatment effects (although detecting the latter presupposes the existence of effective treatment, which unfortunately is not currently available in the case of IPF). The generic measurement instruments that have been used with IPF patients include the World Health Organization 100-Item Instrument (WHOQOL-100) [14], the Quality of Well-Being Scale (QWB) [15], and the Medical Outcomes Study Short-Form 36-Item Instrument (SF-36) [16] – none of which were specifically designed for patients with devastating illnesses like IPF. Further, in a cross-sectional study of 50 patients with various interstitial lung diseases, including 33 with IPF, Chang and her colleagues [8] found that the QWB's content and scaling made it incapable of distinguishing patients with varying degrees of IPF severity.
The respiratory-specific measures administered to IPF patients in prior studies included the Chronic Respiratory Questionnaire (CRQ) [17] and St. George's Respiratory Questionnaire (SGRQ) [18] – both of which were designed specifically for patients with obstructive lung diseases (e.g., asthma, chronic bronchitis, and emphysema) [8-13]. Idiopathic pulmonary fibrosis lies within a completely different spectrum of diseases from these; its physiologic hallmark is ventilatory restriction, not obstruction. The symptoms of IPF are different from the symptoms of obstructive diseases, which cause wheezing, productive cough, and attacks of disease activity: IPF does not cause wheezing, its cough is typically not productive, and IPF symptoms are not episodic. In the same study mentioned above, Chang and her co-investigators [8] noted that the original form of the CRQ underestimated the negative impact of breathlessness on their IPF patients' quality of life because that instrument allowed patients to rate their dyspnea during self-identified activities. Patients whose activity level had become increasingly restricted were rating their breathlessness while performing less taxing activities; hence, the true impact of breathlessness was not reflected in their scores.
In only one study, which enrolled ten patients, did investigators directly ask patients for their perspectives on how IPF impacted their lives [11]. However, the ultimate aim of that study was to assess the relevance of two measurement instruments for patients with IPF. The investigators did not intend to inventory the myriad effects of IPF on the quality of patients' lives or to develop a conceptual framework for describing quality of life in patients with IPF. Such objectives require study in a more systematic manner and on a somewhat larger scale. Further, additional study is required to properly assess the adequacy of current instruments for evaluating patients with IPF, and, if the instruments are found to be inadequate, to provide a basis for developing a more appropriate instrument to serve this purpose.
We conducted the present study to achieve four objectives: 1) to gain further insight into the effects of IPF on patients' lives – from their own perspectives, 2) to organize those effects into a structured conceptual framework, 3) to examine whether four existing instruments adequately cover the elements of this framework, and 4) to examine the extent to which the four existing instruments include items that may be irrelevant to patients with IPF – or that may not serve well the purpose of evaluating the impact of IPF progression and the effects of disease treatment on IPF patients' quality of life.
Methods
We recruited from the general pulmonary and Interstitial Lung Disease Clinics at Stanford University patients with IPF and invited them to participate in focus group meetings or individual in-depth interviews to discuss how IPF affects their lives. Such qualitative data are often collected using multiple methods, including focus groups, key informant interviews, expert opinion, and clinical observation. To improve data capture, we decided to collect patients' perspectives using two formats – focus groups and interviews. An interview accommodates those patients who are unwilling or unable to participate in a focus group. For example, in some cases, patients are willing to freely share information in an individual interview that they might not share in a group setting, while a focus group allows participants to react and respond to the other members and to build on ideas raised during the session. We selected a heterogeneous sample to capture the views of patients in each stage of disease and with varying times since diagnosis. To identify the number of patients to enroll in the study, we used the process of "sampling to redundancy"; in other words, we conducted interviews and focus groups with different patients until no new themes or effects emerged [19]. We ended up with 20 patients in our sample. For each participant, the diagnosis of IPF was confirmed using currently accepted criteria [1]. Patients with other causes of lung fibrosis were not eligible.
To ensure that we addressed all major categories (dimensions) of general quality of life, two investigators (JS and SW) developed a brief set of questions, based loosely on Flanagan's Quality of Life Scale, which were used in the focus groups and interviews [20]. One (JS) or two (JS and SW) investigators took notes and moderated the focus groups and interviews, which were conducted between September 2003 and February 2004. Each of the three focus groups lasted approximately two hours. Each of the five individual interviews lasted approximately one hour. Groups and interviews were audio-taped and transcribed. The study protocol was approved by the Stanford University Institutional Review Board, and all the participants provided informed, written consent prior to enrollment.
Analysis
In the first step of the analysis, we divided the transcripts into individual text units, defined as identifiable segments of continuous speech, ranging in size from phrases to entire paragraphs that identified some effect of IPF on an individual's life. Using NVivo qualitative analysis software (QSR International Pty. Ltd.), we formed sub-categories by clustering identical text units, or ones that addressed essentially the same concept. We then grouped similar sub-categories to form primary conceptual categories (domains of IPF-related quality of life). Thus, the domains include sub-categories that are all distinct from each other and that are comprehensive of all the unique effects mentioned by patients in the groups and interviews.
We then compared the topics and specific item content of the WHOQOL-100, SF-36, CRQ, and SGRQ (measurement instruments administered to IPF patients in previous studies) with the domains that we identified in our analysis. In this process, we examined how well these instruments reflected the identified effects of IPF on patients' lives and whether they contained items that were – based on our patients' perceptions – either not relevant or of questionable utility for assessing change due to disease progression or in response to therapy for IPF.
Results
Demographic data and selected clinical characteristics of the participants are presented in Table 1. Our sample consisted of 13 men than 7 women. Most patients had their diagnoses confirmed by surgical lung biopsy, and all but one patient was taking at least one medication specifically for IPF.
Table 1 Demographic and clinical characteristics of IPF patients
Characteristic Distribution
Gender
male 13
female 7
Age 67 yrs (44–82 yrs)*
Years since diagnosis 1.8 yrs (0.67–11 yrs)*
Mode of diagnosis
Via surgical biopsy 14
Via clinical criteria 6
Supplemental oxygen use
No use 6
Use with exertion and sleep 4
Continuous use 10
Comorbid conditions
Cured prostate cancer 1
Stable coronary artery disease 1
Chronic obstructive pulmonary disease 1
Paroxysmal atrial fibrillation 1
*Data presented as median and (range).
Our analysis produced a conceptual framework consisting of 12 domains. The definitions of these domains are presented in Table 2, and a narrative summary of some of the qualitative data substantiating each domain is provided in this section.
Table 2 Definitions of domains in conceptual framework for describing HRQL in Patients with IPF
Domain Definition
1. Symptoms Amount, severity, and impact of cough and breathlessness; impact of symptoms on physical functioning
2. IPF Therapy Feelings toward medications and impact of medications on physical and mental health; supplemental oxygen use and impact on quality of life
3. Sleep Quality and quantity of sleep; impact of sleep disturbance
4. Exhaustion Lack of energy; feeling exhausted; impact of energy/exhaustion on quality of life
5. Forethought Need to plan and prepare for activities before undertaking them; others' appreciation for the amount of planning and preparation required; impact of need for forethought on quality of life
6. Employment and Finances Effects on employment status and financial security
7. Dependence Need to rely/depend on other people; need to ask for help; fear of being a burden
8 Family Impact of disease on family and relationships with family members
9. Sexual Relations Limitations on sexual activity and sexuality; impact of impaired sexual relations on quality of life
10. Social Participation and Leisure Activities Impact on functioning in relationships, social interactions; social isolation; attention to use of time
11. Mental and Spiritual Well-being Psychological effects including fear, worry, anxiety; problems concentrating/focusing; effects on spirituality/spiritual self
12. Mortality Feelings about death and dying; thoughts on mortality; impact on quality of life
Symptoms
Not surprisingly, symptoms (shortness of breath and cough) were mentioned as significant impairments to overall quality of life. Patients in the earlier stages of the disease were less breathless than those in the later stages. The latter patients noted that shortness of breath was extremely distressing, curtailed all physical activity, and made "even brushing my teeth an exertion". Participants performed physical activity of any kind less often and less intensely because of breathlessness. They noted having to "pause for at least five minutes just to catch my breath" while performing even simple tasks. They were breathless "carrying groceries...carrying anything", taking a shower, bending at the waist, and stooping. Cough was also very bothersome. It was frequently described as "dry and non-productive" or "hacking" and "occurring when I talk for long periods". Several participants mentioned having "a nagging desire to cough constantly" and "never feeling relieved after coughing".
IPF Therapy
Because of the extremely low likelihood of a sustained beneficial response, and because of the high rates of bothersome side effects, most patients perceived conventional IPF therapies as being difficult to tolerate and, in many ways, "worse than the disease itself". Those patients who stopped taking certain conventional therapies were "glad to get rid of [those] medications". Nearly every participant voiced a willingness to be a "medical guinea pig" by taking novel experimental therapies for IPF. Patients who used supplemental oxygen felt "tied to the hoses" that supplied it. Having to fill their car trunks with oxygen tanks and experiencing various types of distress due to this visible indicator of their disability limited their willingness and ability to leave their homes, to participate in social activities, or travel.
Sleep
Perceptions of how IPF impacted sleep ranged from no effect to nightly disturbance. Participants reported occasionally being entirely unable to fall asleep at night or being awoken in the middle of the night because of their cough.
Exhaustion
Participants experienced low energy and feelings of exhaustion, which were distinct from the sensation of breathlessness. Exhaustion and "overwhelming fatigue" were very prominent and "as bothersome as breathlessness". Many noted a "consistent lack of energy", an ongoing "gradual decline in energy", and a need to "economize energy" during the day. Because of low energy, they "rest up, do part of the chore, and rest up again" to accomplish many of the things they want or need to do. Several patients mentioned feeling "completely wiped out" at the end of a normal day. Many mentioned that even their creative energy was low.
Forethought
Participants described a need to "analyze every activity" before starting it and noted that IPF forced them to plan every activity throughout the day. They pored over excursions away from home: "Does this restaurant have a ramp [making pulling an oxygen tank easier]?", "How far away is the parking lot from the door?". They mapped out their routes through the grocery store to get needed items without unnecessary exertion. One woman described making "dry runs" – scouting out the driving distance, parking lots, and entrances to buildings – a day prior to her engagements.
Employment and finances
In terms of their occupations or jobs, participants fell into three categories. They either: (1) had already retired prior to being diagnosed with IPF, (2) could not retire because their medical costs were so great, or (3) were disabled or lost [their] job/career because of IPF. Some of the patients who were still working felt the need to conceal their chronic illness from business colleagues, because the patients believed it made them "appear weak." One man summed up many of the patients' fears about financial insecurity by saying that "in terms of using up my finances, continuing living is a real concern of mine." In general, participants did not want to exhaust family savings on their medical care.
Dependence
The near certainty of disease progression made participants sad and fearful, especially of "becoming more dependent on loved ones". Most said that "the least satisfying aspect of my life is not being as independent as I once was", and many noted that having IPF caused them a "loss of privacy" due to the need for assistance. Having someone else assist with bathing was incredibly worrisome for them. All of them (young and old) relied on someone or something just to "get by" (e.g., hand-rails in the shower, raised toilet seats). They were very frustrated by this and even more so by the likelihood that they will "become a physical or financial burden to family members".
Family
Most participants mentioned an increased appreciation for the relationships they had with family members and the love and support that family members gave them throughout the course of their disease. Many said that the most satisfying part of their current lives was family. Some, however, mentioned that IPF strained their relationship with their spouse or significant loved ones. The limitations that IPF imposed caused many couples to completely change and rearrange their lifestyles; this was extremely frustrating, saddening, and stress-inducing. Many participants also found positive aspects of having IPF, including that it gave them the motivation and opportunity to spend more quality time with their family members.
Sexual relations
Patients experienced decreased libido and a substantial curtailment of sexual activity, mostly due to diminished physical stamina. Several mentioned that their sexual partners were hesitant to engage in sexual activity or refrained from sexual activity altogether because of concerns about the patient over-exerting him- or herself. Many mentioned feeling less sexually attractive or desirable because of having IPF; this was a particularly common concern of patients using supplemental oxygen.
Social participation
Most participants curtailed their social participation in engagements involving crowds of people for fear of "catching something [a respiratory illness that might lead to their demise]". Most patients went out to eat, to the theater, or to other social events much less frequently than before being diagnosed with IPF. Those in the later stages of the disease stayed in their homes almost exclusively. Many patients felt the need to try to hide the fact that they had a chronic illness when they were in public. The subjects were generally "more discriminating" with how they spent their time. This often translated into having difficulty "keeping up with certain relationships." Several patients felt like friends could not understand all that living with IPF entails. For nearly every patient (including those not yet needing supplemental oxygen), travel was considered extremely burdensome or, quite reluctantly, abandoned altogether.
Mental and spiritual well-being
Several participants mentioned feeling sad, mainly in "anticipation of a decline in function". They commonly reported fear, worry, anxiety and panic and related these emotions to having IPF. IPF had the effect of "turning life upside-down" and causing them to "readjust life goals" and "refocus their lives". Many mentioned that IPF had become the "focal point" of their lives (and their spouses' and family members' lives as well). Some had difficulty with activities that required cognition or concentration. The effect of IPF on participants' spiritual well-being was generally positive; many of them became more contemplative, reflective, and had a stronger sense of their spiritual selves.
Mortality
Not surprisingly, given the grave prognosis of IPF, the disease forced patients to "face reality", "recognize their mortality", and realize that " [they are] on the course of expiration". Participants yearned for more attention to end-of-life issues from the medical community. In general, they felt insecure about the dying process, afraid their symptoms would not be controlled toward the end of their lives and that the experience of death would be that of conscious suffocation. However, more than the dying process itself, many patients feared that they would be living a "worthless existence" toward the end of their lives. Many voiced concerns that they had several things that they wanted to do before dying. They wanted to get their "affairs in order" and most had "many preparations" to make in this regard.
Comparison of the identified effects of IPF on quality of life with the content of existing instruments
Table 3 shows the numbers of items on each of the two non-IPF respiratory-specific measurement instruments (the CRQ-SAS and SGRQ) and the two generic instruments (the SF-36 and WHOQOL-100) which relate, in any way, to the domains that our participants identified as relevant to patients with IPF. The CRQ-SAS and SF-36 appear to have very limited coverage of the domains relevant to patients with IPF. Neither have items that focus on therapy, sleep, forethought, employment and finances, dependence, sexual relations, or mortality. Both of these instruments have items that address symptoms; however the scope of the CRQ-SAS is limited to assessing how breathless a respondent becomes while performing certain activities, and only one item on this instrument mentions cough. Certain items on the SF-36 ask about how physical health affects activities; however, there are no items that mention breathlessness or cough, which would make it impossible to differentiate their individual effects. Pain, which is also a domain on the SF-36, was not mentioned as part of our patients' disease experience. The wording of some items in the symptom domain of the SGRQ makes them irrelevant to patients with IPF; for example, they mention "wheezing", "attacks of chest trouble", or "chest condition" – effects and descriptions not identified by our participants. In addition, the SGRQ has no items that address sleep, dependence, family, sexual relations, or mortality. Patients in the current study identified multiple effects of IPF that were not covered at all on the WHOQOL-100, including several effects in our symptoms, therapy, forethought, dependence, and mortality domains.
Table 3 Comparison of the content/domain coverage of the CRQ-SAS, SF-36, SGRQ, and WHOQOL-100 with the domains identified in our conceptual framework
Number of Items
Domains in conceptual framework CRQ-SAS k = 20† SF-36 k = 36† SGRQ k = 50† WHOQOL-100 k = 100†
1. IPF Symptoms 5‡ 14§ 33 ∥ 0
2. IPF Therapy 0 0 4** 0
3. Sleep 0 0 0 4
4. Exhaustion 4 4¶ 2 4¶
5. Forethought 0 0 2 0
6. Employment and Finances 0 0 1 8
7. Dependence 0 0 0 0
8. Family 0 1 0 3
9. Sexual Relations 0 0 0 4
10. Social Participation 0 2 2 9
11. Mental and Spiritual Well-being 11 8 6 12
12. Mortality 0 0 0 0
†These columns show how each item from the CRQ-SAS, SF-36, SGRQ, and WHOQOL-100 maps onto the domains identified in this study; ‡One item mentions cough; §These items address the impact of physical health on activities but do not specifically address symptoms (e.g., breathlessness and cough); ∥Two items are about wheezing and two are about "attacks of chest trouble", making them irrelevant for patients with IPF; ¶These items ask about energy level and fatigue, but none of them pertain to exhaustion; **Items do not pertain specifically to IPF therapy; CRQ-SAS = Chronic Respiratory Questionnaire Self-Administered Standardized Format; SF-36 = Medical Outcomes Study Short-Form 36-item Instrument; SGRQ = St. George's Respiratory Questionnaire; WHOQOL-100 = World Health Organization 100-item Quality of Life Instrument.
Discussion
In this study, we identified specific effects of IPF on patients' quality of life, by using patients' own perspectives. We grouped these specific effects into 12 conceptual categories, which compose both our conceptual framework of HRQL in IPF and might constitute provisional domains for a disease-specific measure. By eliciting patients' perspectives, we also have identified the reasons why existing generic and non-IPF respiratory disease-specific instruments are less than ideal for measuring QOL or HRQL in patients with IPF. An appropriate instrument must include items relevant to IPF patients and must tap important IPF-specific effects that are not captured well (or, in some cases, at all) by existing instruments.
According to patients, IPF significantly impairs quality of life. Symptoms of breathlessness and cough are extremely bothersome and limit physical activity, social participation, travel, and sexual relations. Fatigue, or more precisely, exhaustion is another effect of IPF that patients mention as occurring frequently and negatively impacting their lives. Interestingly, our patients were careful to make the distinction between breathlessness and low energy or exhaustion; they perceived the difference between the two quite clearly.
Most of our patients had to rearrange their lives quite extensively because of the effects of IPF. They had to take more time to prepare for the day, they used a lot of mental energy examining tasks to determine if they could complete them, and they were fearful of the impending need to depend on other people. Like many other patients with chronic or life-threatening illnesses, our patients took great comfort in realizing the love and support of their family members. While many patients recognized this positive aspect that living with IPF had on their relationships with their spouses and family members, several patients mentioned how the effects of this disease caused a great deal of tension between them and their loved ones. Not surprisingly, patients said they sometimes felt like a burden to other people, or they felt lazy because they were unable to do certain things (e.g., chores around the house).
Living with IPF also made patients reflect on their lives and their emotional selves. They were forced to think about things that they didn't necessarily want to think about (e.g., their own mortality and the effects that would have on loved ones left behind). Regarding death, patients wanted assurance that their symptoms would be controlled, that their passing would be peaceful, and that the dying process would occur on their own terms.
In the only study, other than the present one, that directly assessed IPF patients' perspectives, De Vries and colleagues [11] conducted three focus groups with a total of 10 IPF patients to assess the disease's impact on patient quality of life and to discuss the SGRQ and the WHOQOL-100. Their patients emphasized the physical limitations imposed by IPF and viewed the fatigue and social isolation caused by IPF as "serious problems". Other general areas perceived to be negatively affected by IPF included mobility, leisure activities, social relations, and working capacity – all areas included within the domains we have identified. Many of the basic findings of De Vries's and our study are consistent. However, perhaps because of the larger number of patients in the present study, or perhaps because of the somewhat more systematic and detailed analyses that we used, we identified additional effects of IPF that were not previously reported. Both De Vries's patients and (by inference) ours felt that the SGRQ did not adequately capture their disease experience, and the reasons are apparent when one inventories the SGRQ items in comparison with the effects we identified – as we have presented in Table 3.
In their study, De Vries and colleagues concluded that the WHOQOL-100 was well-suited to measuring QOL in patients with IPF, and that development of a disease-specific instrument for IPF was unnecessary [11]. We would agree that the WHOQOL-100 provides a useful measurement tool for many purposes, particularly if one is interested in comparisons across healthy populations or in those with a variety of health problems, rather than comparisons within a specific disease population. However, if the purpose is to measure changes in quality of life that may be associated with different stages of IPF, or that may be associated with different treatments of IPF, we would argue that because it is a generic instrument, the WHOQOL-100 is likely to have limited value. In fact, IPF patients in the study by De Vries and colleagues suggested that the WHOQOL-100 did not place sufficient emphasis on breathlessness, depression and social relationships. In addition, while the investigators stated that the WHOQOL-100 includes every general aspect of life that their patients mentioned, they did not detail the number of items that their patients found completely irrelevant nor how well patients believed that the purportedly relevant items captured their IPF-related circumstances.
To be useful and valid for a particular purpose in a given population, an instrument must not only tap relevant domains; even more importantly, the domains must be represented by relevant items that are in the correct range to adequately assess the population under study, and the instrument must possess the psychometric properties that substantiate its use in that population. There is no evidence to date that the WHOQOL-100 possesses the requisite sensitivity among IPF patients. That reason alone would render premature the conclusion that the WHOQOL-100 is adequate for studies in this population.
We would argue further that there is little reason for confidence that any of the existing measurement instruments that have been used in patients with IPF would be sufficiently sensitive for the purposes we are interested in – or that they would be more sensitive than an instrument specifically designed to address the concerns of patients with IPF. The instruments that we examined in this study focus on some aspects of disease that are not relevant to patients with IPF, they define or operationalize important domains in ways that make them less relevant for IPF, they miss potentially important domains altogether, their scales may be out of the range necessary to reflect the experiences of patients with IPF, and the steps between response options may be too great to capture important differences among IPF patients – or within the same IPF patient over time. All of these features tend to detract from the face validity of these instruments for IPF patients. If these instruments were used to evaluate IPF patients over time, their inclusion of items that cover less relevant topics may introduce variance into patients' scores that would tend to obscure, rather than reveal, changes in quality of life specifically associated with IPF or its therapy. Further, their omission of items on more relevant dimensions means that these instruments would not be able to reflect changes on those aspects at all.
While our study enrolled only 20 patients from one center, our sample included patients of varying age, with a broad range of disease duration, and representing the full spectrum of IPF disease. Some patients were diagnosed a short time before their focus group or interview, others were listed for lung transplantation at the time of the study, and some were in hospice care. While no study is absolutely free of bias, we attempted to minimize it by allowing the themes and items to emerge from the data.
Conclusion
In this study, we conducted focus groups and in-depth interviews with a heterogeneous sample of IPF patients to identify specific effects of IPF on patients' lives. We used these patients' perspectives to develop a comprehensive conceptual framework for describing HRQL in this population. We identified 12 primary domains and numerous sub-categories and specific effects of IPF on the quality of patients' lives. We then examined how well four existing instruments covered these identified topics and found that there were several gaps and insufficiencies in these instruments' abilities to capture the effects of IPF on the quality of patients' lives. We suggest that a more appropriate instrument to measure HRQL in patients with IPF is needed and would be of great value.
List of abbreviations
CRQ-SAS Chronic Respiratory Questionnaire Self-Administered Standardized Format
HRQL health-related quality of life
IPF idiopathic pulmonary fibrosis
QOL quality of life
SF-36 Medical Outcomes Study Short-Form 36-item Instrument
SGRQ St. George's Respiratory Questionnaire
WHOQOL-100 World Health Organization 100-item Quality of Life Instrument
Acknowledgements
Dr. Swigris received support for this research from an NIH training grant (T32 HL07948-01A1). Dr. Gould received an Advanced Research Career Development Award from the VA Health Services Research and Development Service. The views expressed in this article are those of the authors and not necessarily the views of the Department of Veterans Affairs.
==== Refs
Joint Statement of the American Thoracic Society and the European Respiratory Society: Idiopathic pulmonary fibrosis: diagnosis and treatment. Am J Respir Crit Care Med 2000 161 646 664 10673212
Coultas DB Zumwalt RE Black WC Sobonya RE The epidemiology of interstitial lung disease Am J Respir Crit Care Med 1994 150 967 972 7921471
Bjoraker JA Ryu JH Edwin MK Meyers JL Tazelaar HD Schroeder DR Offord KP Prognostic significance of histopathologic subsets in idiopathic pulmonary fibrosis Am J Respir Crit Care Med 1998 157 199 203 9445300
Daniil ZD Gilchrist FC Nicholson AG Hansell DM Harris J Colby TV du Bois RM A histologic pattern of nonspecific interstitial pneumonia is associated with a better prognosis than usual interstitial pneumonia in patients with cryptogenic fibrosing alveolitis Am J Respir Crit Care Med 1999 160 899 905 10471616
Nicholson AG Colby TV du Bois RM Hansell DM Wells AU The prognostic significance of the histologic pattern of interstitial pneumonia in patients presenting with the clinical entity of cryptogenic fibrosing alveolitis Am J Respir Crit Care Med 2000 162 2213 2217 11112140
Flaherty KR Toews GB Lynch JP Kazerooni EA Gross BH Strawderman RL Hariharan K Flint A Martinez FJ Steroids in idiopathic pulmonary fibrosis: a prospective assessment of adverse reactions, response to therapy, and survival Am J Med 2001 110 278 282 11239846 10.1016/S0002-9343(00)00711-7
Curtis JR Martin DP Martin TR Patient-assessed health outcomes in chronic lung disease: What are they, how do they help us, and where do we go from here? Am J Respir Crit Care Med 1997 156 1032 1039 9351600
Chang JA Curtis R Patrick DL Raghu G Assessment of health-related quality of life in patients with interstitial lung disease Chest 1999 116 1175 1182 10559073 10.1378/chest.116.5.1175
Clark M Cooper B Singh S Cooper M Carr A Hubbard R A survey of nocturnal hypoxaemia and health related quality of life in patients with cryptogenic fibrosing alveolitis Thorax 2001 56 482 486 11359966 10.1136/thorax.56.6.482
De Vries J Kessels BL Drent M Quality of life of idiopathic pulmonary fibrosis patients Eur Respir J 2001 17 954 961 11488332 10.1183/09031936.01.17509540
De Vries J Seebregts A Drent M Assessing health status and quality of life in idiopathic pulmonary fibrosis: which measure should be used? Respir Med 2000 94 273 278 10783939 10.1053/rmed.1999.0736
Martinez JAB Martinez TY Galhardo FPL Pereira CAC Dyspnea scales as a measure of health-related quality of life in patients with idiopathic pulmonary fibrosis Med Sci Mon 2002 8 CR405 CR410
Martinez TY Pereira CA dos Santos ML Ciconelli RM Guimaraes SM Martinez JAB Evaluation of the short-form 36-item questionnaire to measure health-related quality of life in patients with idiopathic pulmonary fibrosis Chest 2000 117 1627 1632 10858394 10.1378/chest.117.6.1627
Group TWHOQOL The world health organization quality of life assessment (WHOQOL): development and general psychometric properties Soc Sci Med 1998 46 1569 1585 9672396 10.1016/S0277-9536(98)00009-4
Kaplan RM Bush JW Berry CC Health status: types of validity and the index of well-being Health Serv Res 1976 11 478 507 1030700
Ware Jr. JE Sherbourne CD The MOS 36-item short-form health survey (SF-36). I. Coceptual framework and item selection Med Care 1992 30 473 483 1593914
Guyatt GH Berman LB Townsend M Pugsley SO Chambers LW A measure of quality of life for clinical trials of chronic lung disease Thorax 1987 42 773 778 3321537
Jones PW Quirk FH Baveystock CM The St. George's Respiratory Questionnaire Respir Med 1991 85 25 31 1759018
Streiner DL Norman GR Health Measurement Scales: A practical guide to their development and use 1995 New York , Oxford University Press
Flanagan JC A research approach to improving our quality of life Am Psychol 1978 33 138 147 10.1037//0003-066X.33.2.138
|
16212668
|
PMC1276807
|
CC BY
|
2021-01-04 16:38:13
|
no
|
Health Qual Life Outcomes. 2005 Oct 7; 3:61
|
utf-8
|
Health Qual Life Outcomes
| 2,005 |
10.1186/1477-7525-3-61
|
oa_comm
|
==== Front
Hum Resour HealthHuman Resources for Health1478-4491BioMed Central London 1478-4491-3-101622344710.1186/1478-4491-3-10MethodologyValidating a work group climate assessment tool for improving the performance of public health organizations Perry Cary [email protected] Nancy [email protected] Greg [email protected] Allison [email protected] Joan [email protected] Management & Leadership Program, Management Sciences for Health, Cambridge, Massachusetts, USA2 Wellesley Centers for Women, Wellesley College, Wellesley, Massachusetts, USA2005 13 10 2005 3 10 10 21 12 2004 13 10 2005 Copyright © 2005 Perry et al; licensee BioMed Central Ltd.2005Perry et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
This article describes the validation of an instrument to measure work group climate in public health organizations in developing countries. The instrument, the Work Group Climate Assessment Tool (WCA), was applied in Brazil, Mozambique, and Guinea to assess the intermediate outcomes of a program to develop leadership for performance improvement. Data were collected from 305 individuals in 42 work groups, who completed a self-administered questionnaire.
Methods
The WCA was initially validated using Cronbach's alpha reliability coefficient and exploratory factor analysis. This article presents the results of a second validation study to refine the initial analyses to account for nested data, to provide item-level psychometrics, and to establish construct validity. Analyses included eigenvalue decomposition analysis, confirmatory factor analysis, and validity and reliability analyses.
Results
This study confirmed the validity and reliability of the WCA across work groups with different demographic characteristics (gender, education, management level, and geographical location). The study showed that there is agreement between the theoretical construct of work climate and the items in the WCA tool across different populations. The WCA captures a single perception of climate rather than individual sub-scales of clarity, support, and challenge.
Conclusion
The WCA is useful for comparing the climates of different work groups, tracking the changes in climate in a single work group over time, or examining differences among individuals' perceptions of their work group climate. Application of the WCA before and after a leadership development process can help work groups hold a discussion about current climate and select a target for improvement. The WCA provides work groups with a tool to take ownership of their own group climate through a process that is simple and objective and that protects individual confidentiality.
==== Body
Background
This article describes the validation of an instrument to measure work group climate in public health organizations in developing countries. In light of decentralizing health care systems and the urgent need to scale up services to combat HIV/AIDS, tuberculosis, and malaria, it is critical that providers of technical assistance have valid tools to help build institutional capacity in both public-sector and nongovernmental organizations. Health care managers in developing countries need a simple, inexpensive tool that is objective and applicable to small work groups and that can become part of the team's own self-evaluation process. The goal of improving work group climate is to strengthen organizational performance and improve health service delivery.
The development of the Work Group Climate Assessment Tool was carried out between 2002 and 2004 by the Management & Leadership (M&L) Program, a five-year cooperative agreement between the United States Agency for International Development (USAID) and Management Sciences for Health (MSH). MSH is a nonprofit organization with headquarters in Cambridge, Massachusetts. M&L works with ministries of health, national and international programs, and nongovernmental organizations in 27 developing countries to strengthen the leadership skills of health personnel and the management systems that are essential to deliver high-quality health services.
A positive work group climate is a primary outcome of a leadership development process aimed at improving the performance of managers and their work groups. This hypothesis is based on evidence that leadership and management practices that provide employees with clarity, support, and challenge contribute to a positive work climate. A positive work climate leads to and sustains employee motivation and high performance by liberating "discretionary effort," or the level of extra effort that employees exert above and beyond job expectations [1].
Organizational climate and culture
The terms "organizational climate" and "organizational culture" are sometimes treated as different concepts because they arise out of distinct theoretical traditions, but they have many overlapping elements. Stringer explains climate as a subset of organizational culture [2]. Culture applies to the deeply rooted value systems inherent in all organizations and is difficult to change [3]. Organizational strategy, the external environment, organizational arrangements and historical forces all affect the context and milieu within which a work group operates. These "cultural" influences develop outside the work group and are beyond the direct control of the work group manager. Burke describes the relationship between climate and culture in a series of papers that discuss the modification of organizational culture at British Airways [4,5]. Burke posits that changes to climate are more achievable than changes in culture, because climate is associated with the "transactional level of human behaviour – the everyday interactions and exchanges" [4].
Thus, while every organization has an organizational culture, each work group (or team) within the organization has its particular climate. Tagiuri defines organizational climate as a "quality of the internal environment of an organization that (a) is experienced by its members, (b) influences their behaviour and (c) can be described in terms of the values of a particular set of characteristics (or attributes) of the organization" [6]. Burke [4] stresses that leadership, mission, strategy and organizational culture have an organization-wide focus, whereas climate is experienced and created in the work group or team.
A work group's climate may be similar to or different from the overall organizational climate. High-performing work groups sometimes operate in organizations troubled by declining funding or inadequate leadership at the senior level. The leadership and management practices of a manager can create a positive work climate and strong results within a work group, even if an organization's climate is less than optimal. Regardless of a manager's level, his or her efforts to improve the work group's climate can contribute to strong employee performance and results.
Work group climate and performance
Positive work climate has been identified in a variety of environments as a driver of performance. According to the business literature, there is a positive correlation between climate and performance and also between climate and financial results. "Organizational climate is not the only driver of performance. Economic conditions and competitive dynamics matter enormously. But our analysis suggests that climate accounts for nearly a third of the results" [1].
While many of the factors such as an organization's history, culture, organizational strategies and structures are outside the control of the work group manager, the manager is uniquely placed to influence work climate within the work group. "What the boss of a work group does is the most important determinant of climate. The boss's behavior drives climate, which arouses motivation. And aroused motivation is a major driver of bottom-line performance" [2].
The work group manager, therefore, has a substantial impact on the development of work climate and the productivity of the work group. This relationship was clearly identified by a research project conducted at Harvard Business School in 1968. The project studied the relationship between motivation and organizational climate and reviewed the impact of different leadership styles on three evenly matched teams working on the same production project. The researchers demonstrated that leadership styles affected both the development of work group climate and the productivity of the three teams [7].
The impact of work climate is not restricted to the commercial sector. Research in the health and education fields supports the conclusions from the business literature. For instance, in a study of Canadian staff nurses, Laschinger, Finegan and Shamian describe the relationship between empowerment, job satisfaction, and commitment [8]. A positive work climate creates an environment conducive to the development of trust and empowerment, which in turn leads to high-quality patient care [8].
Positive climate has also been demonstrated to drive success in schools operated by the Department of Education and Employment in the United Kingdom. "Our research demonstrates a significant link between classroom climate and student academic progress. . . to the degree that teachers can develop skills and characteristics that impact climate, so they can hope to more effectively motivate and engage their students" [9].
Over the past decade a number of instruments have been developed to help organizations measure organizational climate, predominantly in US-based organizations in the private sector. However, these tools are often proprietary (and therefore expensive) or complicated and lengthy to administer. Surveys of organizational climate such as the PAHO Organizational Climate Instrument [10], and the Gallup Q12 [11] do not conform to the needs of small work groups. The 80-item PAHO Organizational Climate Instrument is extremely lengthy, and like the Q12, measures individual employees' satisfaction, rather than perception of the overall climate in a work group.
Methods
Description of the Work Group Climate Assessment Tool
The Work Group Climate Assessment Tool (WCA) is a self-administered assessment form originally consisting of 14 items: 12 that correspond to three sub-dimensions of climate – clarity, support and challenge – and 2 items that capture perceptions of productivity and quality. These sub-dimensions and the individual items are based on the work of George Litwin and Robert Stringer, who pioneered the study of climate in corporate environments [2,7].
The WCA is designed to measure climate among intact teams or work groups in the health sector of developing countries. (An intact team is defined as a group of individuals who work together regularly at the same work site, whether in a central or regional office or a health facility.) The WCA is the first assessment tool that has been developed for this purpose. It is intended to measure climate in work groups at any level of an organization. To date, the WCA has been used to measure work group climate before and after the M&L Leading for Performance Improvement Program that was conducted in five sites – Egypt, Mozambique, Brazil, Guinea, and Kenya – and via a virtual distance-learning program for leadership development.
The WCA is divided into two sections. The first section includes 12 items, which were mapped to the three hypothesized sub-dimensions mentioned above of clarity, support and challenge. The items for the original WCA were the following:
1. We are recognized for individual contributions.
2. We have a common purpose.
3. We have the resources we need to do our jobs well.
4. We develop our skills and knowledge.
5. We have a plan which guides our activities.
6. We strive to improve our performance.
7. We understand each other's capabilities.
8. We are clear about what is expected in our work.
9. We seek to understand the needs of our clients.
10. We participate in the decisions of our work group.
11. We take pride in our work.
12. We readily adapt to new circumstances.
The second section, items 13 and 14, relates to perceptions of productivity and quality, which are defined for the respondent on the assessment form:
13. Our work group is known for quality work.
14. Our work group is productive.
To apply the survey, all members of the work group (both managerial and staff) complete the assessment form. Each team member rates each item. The scores are then tabulated across all respondents, and results for each item and an overall climate score for items 1–12 are calculated for the team as a whole. Results for items 13 and 14 are calculated separately.
At the conclusion of the leadership program, the WCA is applied again among all team members. The post-intervention scores are again calculated for the team as a whole and then compared to the baseline team scores and targets to determine the amount of change produced by the intervention vis-à-vis the anticipated results (climate targets).
Initial validation of the WCA
M&L tested the WCA for face validity throughout 2002–2003 with counterparts in Brazil and Nicaragua as well as with several teams working on the M&L Program in Cambridge. In addition, the WCA was used to collect baseline and follow-up data among participants in the Leading for Performance Improvement Programs in Egypt and Guinea and participants in the Virtual Leadership Development Program in Latin America. The WCA was translated into Portuguese, Spanish and French and pre-tested in the different countries to make sure that it was appropriate for use across cultures. Based on feedback from the field tests, the WCA form and instructions were refined, and ultimately published in 2003 [12]. The expectation was that publishing the tool would allow M&L leadership programs to test and refine the tool in preparation for a full validation study and peer-review.
Based on the work of Litwin and Stringer [7], the original tool incorporated a measure of the importance of each item to the respondent in order to assist teams to prioritize the sub-dimensions that needed more attention. Managers participating in the pre-testing of the instrument, however, tended to rate the importance of all items quite high, and therefore the importance measure was not useful for determining the teams' priorities. Part of the purpose of the current study was to determine if the importance column could be eliminated without compromising the statistical validity of the tool.
Using the data collected during the field tests, M&L examined certain aspects of the tool's validity and reliability. For example, data from Brazil suggest that the tool has discriminant validity. The WCA was applied with three groups of managers in Brazil: one group was in the state of Ceará and had undergone extensive leadership training over a period of five years, while the other two groups were in states that had only begun to participate in leadership training. The spread in mean scores from a high mean in Ceará to much lower means in the other two states suggests the tool can discriminate between high- and low-performing work groups.
In terms of reliability, a Cronbach's alpha of 0.87 was calculated on 122 cases of WCA data collected in Latin America and Egypt. The coefficient alpha suggests that the items in the WCA have a high level of internal consistency. Results of an initial factor analysis conducted on the same data indicate that the assessment items load on a set of three to four factors. However, additional analysis on a larger data set was necessary to determine whether the identified factors relate to the hypothesized sub-dimensions of clarity, challenge and support.
While the WCA was initially validated using Cronbach's alpha reliability coefficient and exploratory factor analysis, a subsequent validation study was necessary to refine these analyses to account for nested data, provide item-level psychometrics and establish construct validity. This paper presents the results of this second validation study. Study participants were a purposive sample of present and past recipients of M&L technical assistance to strengthen management and leadership in the public health sector in developing countries. Participants came from ministries of health in Mozambique and Guinea, the Secretariat of Health for the State of Ceará, Brazil, and Brazilian public health laboratories. The participants represented a wide variety of positions, including central-level ministry staff, district-level managers, hospital administrators, laboratory technicians and clinic personnel.
The participants completed self-administered questionnaires anonymously in a group setting in each participating organization in May 2004. The survey contained two sections: The first consisted of the original 12 climate items from the WCA, the two productivity and quality items, and nine additional items generated to increase the item pool for measurement refinement. The second section consisted of 24 items from one section of the Stringer Organizational Climate Survey [7]. Participants rated each item on a Likert scale, where 1 = not at all, 2 = to a small degree, 3 = to a moderate degree, 4 = to a great degree and 5 = to a very great degree.
The Stringer survey served as the gold standard for this study. This instrument, which has been used repeatedly since 1968, was validated through studies that showed its association with objective measures of organizational climate in corporate settings in the United States [2].
The analysis data set consisted of data from 305 individuals in 42 work groups: Brazil (21 work groups, 182 employees), Mozambique (18 work groups, 97 employees), and Guinea (3 work groups, 26 employees). With 42 work group sites, we had a statistical power of 0.87 to detect validation correlations as low as 0.20 at the standard significance value of 0.05. Values outside the admissible range for a given variable were reassigned as missing. Four cases were omitted from the analysis sample due to miskeyed data.
Analyses included eigenvalue decomposition analysis, confirmatory factor analysis (which included tests for measurement noninvariance by gender, management status and educational level), and construct validity and reliability analyses of the WCA. Once a final set of items had been selected, reliability coefficients for both the work group and individual employee levels of analysis were computed.
Results and discussion
Description of participants
Among participants from Brazil and Mozambique, women represented 62% of the respondents, consistent with trends of female participation in the health field. In these two samples, 59 respondents were managers (22%). Individuals had been employed in the health field for up to 40 years (mean [M] = 17.7, standard deviation [SD] = 9.7) and in the rated organization for up to 37 years (M = 11.5, SD = 9.2). Demographic information was not available for the three Guinean work groups.
Item weighting
Respondents to the WCA survey rate each item twice: first according to how the item was currently performed in the work group ("actual performance") and second according to the item's perceived importance to the work group. Models using the "actual performance" ratings were compared to those weighted by "importance" ratings. (A weighted score is the value assigned to the "actual performance" of a given item multiplied by the value of the "importance" assigned to that item.) It was found that in the importance-weighted models the factor structure was stronger, due to improved distributional characteristics of the individual items. As a result, all subsequent models in the study used importance-weighted scores for the WCA items. However, use of the WCA in the field suggests that the weighting of actual scores by importance scores is not easily understood by health managers and is not useful for prioritizing actions to improve climate. To determine if the weighting could be eliminated, we correlated mean individual scores from the "actual performance" ratings and those same scores weighted by the "importance" rating. This correlation was extremely high (R = 0.83, p < 0.01), indicating that the importance column could be eliminated without compromising the validity of the tool.
Eigenvalue decomposition analysis
The variance/covariance matrix of the 21 items from the WCA were submitted to an eigenvalue decomposition analysis, and the resulting eigenvalues were plotted (Fig. 1). (Items 13–14, measuring productivity and quality, were not included in the model and were analysed separately, since they are considered outputs of a well-functioning work group rather than components of climate.) The resulting scree plot clearly shows a unidimensional structure. Scree plots based on eigenvalues obtained from each of the three separate countries (not shown) also indicated a clearly unidimensional construct. Initially, models were conducted by disaggregating the items into the three hypothesized sub-dimensions: clarity, challenge, and support. While the factor loadings were strong for all three sub-dimensions, the factor intercorrelations were very high (0.81 to 0.95), and modification indices suggested that there was considerable cross-loading of items across factors. Because of a lack of factor discrimination and because the scree plot (Fig. 1) shows evidence of a single dimension, all subsequent models contain a single WCA factor.
Figure 1 Scree plot of eigenvalues (importance-weighted WCA items) for the full sample.
Confirmatory factor analyses
An initial confirmatory factor model with the full set of 21 importance-weighted items was fitted to adjust for the clustering in the data. In this model and all subsequent models, missing data were handled by means of a maximum likelihood estimation technique, which minimizes bias due to listwise deletion [13]. Factor loadings from this model ranged from 0.37 to 0.68. One item had a factor loading below the standard 0.40 cutoff level [14] so it was omitted from further analysis models.
Noninvariance, or differential item functioning, across various individual characteristics (gender, management status and educational level) was tested by means of a series of multiple group confirmatory factor models in which factor loadings were constrained to be equal across groups. Using a cutoff of 1.00 and above for model-generated modification indices, we identified six items showing evidence of noninvariance across gender, three items showing noninvariance across management status, and three items showing noninvariance across educational levels. (Because there were only 16 respondents with only a primary school educational level, too few to allow estimation, we compared the groups that had finished secondary school [n = 81] and those who had completed university education [n = 160]).
A two-level confirmatory factor model was fitted to 12 importance-weighted WCA items that had at least a 0.40 factor loading in the previous models and that showed no evidence of noninvariance across gender, management status, or educational level. At the work group level of the model, the factor structure of employees' ratings, aggregated to the work group level, was examined for consistency across the 42 work groups. At the individual level of the model, the factor structure of the ratings was examined for consistency across employees within each work group. The factor loadings were constrained to be equal across these two levels of analysis so that a parallel interpretation of work group climate was maintained when considering each level of analysis.
At the work group level of the model, there was too little variability in three items to adequately model their contribution to the overall work group climate factor. A fourth item returned a lower than acceptable factor loading (< 0.40). The final model was based on the remaining eight items that showed strong and consistent psychometric properties throughout the preceding series of models. These items and their associated factor loadings are given in Table 1. There was significant variability in the work group climate factor at both the work group (σ = 3.54, p < 0.001) and individual (σ = 14.12, p < 0.001) levels of analysis, indicating that variability in both work group characteristics and individual employee characteristics may influence overall ratings of work group climate. Rather than eliminate the effects of individual variation, its presence was statistically modeled and the variance partialed into variability that was due to differences between work groups and within the work group across individuals.
Table 1 Standardized factor loadings for the final eight WCA items, by level of analysis
Individual level Work group level
4. We feel our work is important.a 0.62 0.82
7. We strive to achieve successful outcomes.a 0.67 0.90
8. We have a plan which guides our activities.b 0.47 0.73
9. We pay attention to how well we are working together.a 0.51 0.77
11. We understand each other's capabilities.b 0.49 0.79
14. We seek to understand the needs of our clients.b 0.66 0.93
15. We understand the relevance of the job of each member in our group.a 0.54 0.92
17. We take pride in our work.b 0.61 0.93
Notes: a Additional items included in the model
b Original WCA items
Validation and reliability analyses
Once the factor model had been finalized, correlations between the WCA and a composite of the Stringer items were estimated at both the individual and work group levels. At the work group level, this correlation was extremely high (R = 0.93, p < 0.001), indicating that the eight-item WCA scale captured the same construct as the 24-item Stringer scale in differentiating the work group climate quality across work groups. The correlation at the individual level was moderate (R = 0.48, p < 0.001), indicating that the WCA scale captures a construct that is similar to but different from that captured by the Stringer scale in differentiating among employees' perceptions of climate within the same work group. The ways in which the WCA differs from the Stringer scale in assessing individual perceptions of the same work group climate require further study to be fully understood.
A reliability analysis of the data on individuals was conducted, without adjusting for the clustered design. The internal consistency in the 8 WCA items across individuals was good (α = 0.81). Since this scale is most commonly used to assess the work group itself, the data were restructured and a reliability analysis using the work group as the level of analysis was done. The internal consistency of the eight items across work groups (aggregated across employees within a work group) was also good (α = 0.86). Our final models contained eight items. Considering only the 42 work groups, this number gives us a ratio of more than five observations per factor, satisfying a common rule of thumb in psychometrics. Additionally, since the estimates of these items at the 42 sites draw from data on 305 individuals, these estimates are very stable, reliable and unbiased.
Conclusion
The results of this study confirmed the construct validity and reliability of a revised version of the WCA tool across work groups with different demographic characteristics (gender, education, management level and geographical location). This study shows that there is agreement between the theoretical construct of work climate and the items in the WCA tool across different populations. Of the 21 items tested, eight were selected that conferred the greatest measurement power for the smallest number of questions. These items showed the least variance across the different groups and the strongest psychometric properties. The internal consistency of the eight-item WCA was high across work groups, indicating that the individual items in the instrument are associated with each other and all appear to be measuring the same underlying construct. Finally, the eight items selected for the final model correlated well with the 24 items from the Stringer gold standard instrument, indicating that the WCA scale captures the same underlying construct as the Stringer scale in differentiating the climates of work groups.
The WCA was designed to measure three sub-dimensions of climate: clarity, support and challenge. However, study results indicated the WCA items do not discriminate between sub-dimensions of work group climate but rather capture a single perception of climate. While the terms clarity, support and challenge are not measured as sub-dimensions of climate by the WCA tool, they work together to define the construct of the overall quality of work group climate.
Based on these analyses, we recommend that the analysis and feedback of WCA scores to work groups be revised. First, we recommend that the importance column in the tool be eliminated to simplify it further. Second, we recommend that feedback to work groups be provided based on average individual and work group scores and on patterns across the work group (all high scores, all low scores, or a variable pattern of some members reporting positive feelings about climate in the work group while others are reporting more negative feelings). The aspects of clarity, support and challenge are useful to help work groups to discuss their climate, but because the study did not confirm that they exist as separate subscales, scores on individual items or sub-dimensions should not be compared over time.
Once the individual and work group composite scores have been obtained, comparisons can be made between work groups in an organization, between pre- and post-test assessments of the same work group, or between a single work group and a pre-determined value of climate quality that serves as a benchmark. In addition, disparities in the experience of climate within a work group can be assessed by comparing individuals' scores within the work group and tracking changes over time in individuals' perceptions of climate.
Work groups in developing countries can benefit from using the WCA tool as a basis for discussing and improving their climate. Applying the WCA is a simple and rapid exercise that provides a work group with a set of objective scores it can use to select a target for improving climate and to identify strategies for meeting that target. The manager and members of a work group should review their baseline scores (both individual and group) and discuss what may cause the patterns in the data by reflecting on the leading and managing practices used by the group. This process allows the work group's members to decide how they will work together to create a more positive climate.
The WCA also serves as a monitoring tool; once a work group has implemented actions to improve climate and has reapplied the tool, members can compare baseline and follow-up scores to determine what progress they have made in changing their climate. A detailed guide to facilitating the use of the WCA in the field as well as a tabulation sheet, guidelines for analysing and using results, and the tool itself can be downloaded from the MSH website [15].
The results of this study suggest that the WCA is a simple, reliable, and valid instrument for measuring climate among work groups in the health sector of developing countries. This tool is an important contribution to programs working to strengthen the performance of managers and their work groups in order to improve the delivery of health services in developing countries.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
These authors contributed equally to this work: Cary Perry, Nancy LeMay, Greg Rodway and Allison Tracy. Cary Perry, Nancy LeMay and Greg Rodway participated in the design of the study, analysis of the data and drafting of the manuscript. Allison Tracy performed the statistical analyses and drafted the results section. Joan Galer conceived of the study and helped develop the instrument. All authors read and approved the final manuscript.
Acknowledgements
Funding for this article was provided by the Office of Population and Reproductive Health, Bureau for Global Health, United States Agency for International Development, under the Management & Leadership Program, award number HRN-A-00-00-00014-00. This Program financed the article-processing charge. The opinions expressed herein are those of the authors and do not necessarily reflect the views of USAID.
We gratefully acknowledge the contribution of Robert Stringer, who granted written permission for MSH to use his Organizational Climate Survey for this study.
==== Refs
Goleman D Leadership that gets results Harvard Business Review 2000 78 90
Stringer R Leadership and Organizational Climate 2002 Upper Saddle River, Prentice Hall
Denison D What is the difference between organizational culture and organizational climate?: A native's point of view on a decade of paradigm wars Academy of Management Review 1996 21 619ff
Burke W Organizational Development: A Process of Learning and Changing 1993 Reading, Addison-Wesley
Goodstein L Burke W Creating successful organizational change Organizational Dynamics 1991 19 5 17 10.1016/0090-2616(91)90050-J
Tagiuri R Tagiuri R, Litwin GH The concept of organizational climate Organizational Climate: Explorations of a Concept 1968 Cambridge, Harvard University
Litwin G Stringer R Motivation and Organizational Climate 1968 Boston, Harvard University Press
Laschinger H Finegan J Shamian J The impact of workplace empowerment, organizational trust on staff nurses' work satisfaction and organizational commitment Health Care Manage Rev 2001 26 7 23 11482178
Hay Group Research into teacher effectiveness: a report by Hay/McBer for the Department for Education and Employment – June 2000
Marín JM Melgar A Castaño C PAHO Organizational Climate Instrument Teoría y Técnicas de Desarrollo Organizacional [Theories and Techniques of Organizational Development] Technical Documents Series no PSDCG-T-10 1990 3 Proyecto Subregional de Desarrollo de la Capacidad Gerencial de los Servicios de Salud de Centroamérica y Panamá. Guatemala: Pan American Health Organization
The Gallup Organization Princeton NJ Gallup Q12 1999 The Gallup Organization
Management Sciences for Health Creating a climate that motivates staff and improves performance The Manager 11 1 22
Schafer JL Analysis of Incomplete Multivariate Data 1997 London, Chapman & Hall
Gorsuch RL Factor Analysis 1983 2 London, Lawrence Erlbaum Associates
Management Sciences for Health Workgroup climate assessment (WCA) tool and guide for facilitators The Health Manager's Toolkit 2005 Boston, MSH
|
16223447
|
PMC1276808
|
CC BY
|
2021-01-04 16:37:43
|
no
|
Hum Resour Health. 2005 Oct 13; 3:10
|
utf-8
|
Hum Resour Health
| 2,005 |
10.1186/1478-4491-3-10
|
oa_comm
|
==== Front
Hum Resour HealthHuman Resources for Health1478-4491BioMed Central London 1478-4491-3-91621267210.1186/1478-4491-3-9ResearchIranian staff nurses' views of their productivity and human resource factors improving and impeding it: a qualitative study Nayeri Nahid Dehghan [email protected] Ali Akbar [email protected] Mahvash [email protected] Fazlollah [email protected] Faculty of Nursing, Tehran University of Medical Sciences, Tehran, Iran2 Faculty of Medicine, Tarbyat Modarres University, Tehran, Iran2005 8 10 2005 3 9 9 13 7 2005 8 10 2005 Copyright © 2005 Nayeri et al; licensee BioMed Central Ltd.2005Nayeri et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Nurses, as the largest human resource element of health care systems, have a major role in providing ongoing, high-quality care to patients. Productivity is a significant indicator of professional development within any professional group, including nurses. The human resource element has been identified as the most important factor affecting productivity. This research aimed to explore nurses' perceptions and experiences of productivity and human resource factors improving or impeding it.
Method
A qualitative approach was used to obtain rich data; open, semi-structured interviews were also conducted. The sampling was based on the maximum variant approach; data analysis was carried out by content analysis, with the constant comparative method.
Results
Participants indicated that human resources issues are the most important factor in promoting or impeding their productivity. They suggested that the factors influencing effectiveness of human resource elements include: systematic evaluation of staff numbers; a sound selection process based on verifiable criteria; provision of an adequate staffing level throughout the year; full involvement of the ward sister in the process of admitting patients; and sound communication within the care team. Paying attention to these factors creates a suitable background for improved productivity and decreases negative impacts of human resource shortages, whereas ignoring or interfering with them would result in lowering of nurses' productivity.
Conclusion
Participants maintained that satisfactory human resources can improve nurses' productivity and the quality of care they provide; thereby fulfilling the core objective of the health care system.
==== Body
Background
In most health care organizations, nurses are the largest work group and play a major role in the organization's success. Hence nurses' productivity affects an organization's success by influencing organizational total factor productivity (TFP) [1]. Health care organizations cannot succeed without productive nursing staff [2]. But recent studies suggest that nurses no longer feel their work is valued and are concerned with their productivity [3].
Although nurses are concerned about declining levels of effective care and productivity [4], staff productivity rarely has been assessed within the health care organization of Iran and little is known about factors that affect nurses' productivity. The complexity associated with the impact of human resources on nurses' productivity is one of the factors least studied.
Managers' efforts to promote productivity have been mostly ineffective and have resulted in too many changes and personnel frustration without improving patient care [5]. Assessing nurses' viewpoint on productivity should be the first step towards improving nurses' productivity.
Theoretical and empirical approaches to productivity
Productivity is defined as the ratio of outputs to inputs or as the relationship between inputs and outputs. McConnell suggested that comparing output to input was similar to comparing apples with oranges, because the two are often so different [6].
In light of traditional economic definitions of productivity as the numeric cost ratio of outputs to inputs, administrators in health care have focused on quantifying nurses' work in economic input-output terms [7]. Within the discipline of nursing, some viewed nurse productivity as a measure of the efficiency with which the input of nursing tasks and other labourers' tasks, materials and equipment were converted into goods and services delivered within the health care system [3,7].
Also, nursing productivity was defined as equilibrium between demand for and supply of services and managing cost structure of a system by integration of financial and clinical processes and providing good care in a cost-effective manner. Some researchers have defined productivity as managing staffing (depending on census and patient acuity) to meet budgeted standards [6]. Others view nurse productivity as the ratio of output produced compared with resources consumed and proposed that nurse productivity can be measured at the hospital unit, departmental or divisional levels [7,8].
Employee productivity was also measured by hours of care per day and salary dollars per procedure. Sometimes acuity also was factored in. Productivity was also measured by budgetary standards set by organizations, or through community or national norms. Recently, in organizational redesign strategies, productivity measures have been set by measuring the time necessary to do each job, and incorporating acuity standards [5].
Another measure of nurse productivity that has become prevalent and is used by many hospitals is the popular "hours per patient day" (HPPD) unit of measure for nurse productivity [7,9]. Nursing care hours have been further delineated into direct and indirect care hours, and sometimes have focused on costs [6].
Despite that term nursing productivity is an old concept used daily by nurse executives, there is little clarity and consistency in how it is used [6]. Also, these different definitions of productivity lead to measurements that cannot be compared [5]. In a nursing context, productivity has been used exclusively in terms of resources used – based on economic theories of productivity- while certain significant outcomes of nursing practice have not been considered [10,11]. In addition, the emphasis of economic theories of productivity is on measuring administrative systems rather than on what has been accomplished [11]. In business literature and research, productivity has usually been discussed in terms of hypothetical variables that could improve the outcome.
For instance, researchers (1993) reported that employees with higher levels of job satisfaction and skills directly related to their jobs had significantly higher productivity ratings than their co-workers [12]. Another study revealed that practices such as performance appraisal had a strong effect on productivity [13]. In addition, training programmes for new employees increased their productivity [14]. A number of researchers (McCloskey and McCain 1988, McNeese Smith 1995, 1996) have assessed factors affecting nurses' productivity [15-17], while Grosskopf, Margaritis and Valdmanis (2001) evaluated effects of teaching on hospital productivity [18]. Hall (2003) assessed the contribution of knowledge and skill, and factors such as organizational trust and commitment on nursing productivity [3]. Curtin (1995) described how patient classification could be used to improve staff productivity [11]. However a concept analysis done by Holcomb, Hoffart and Fox (2002) revealed the complexity of the concept and its measurement [6]. Even though it is essential to understand care providers' views about productivity, only one study (McNeese Smith 2001) examined in detail how nurses viewed their own productivity.
However, no studies have examined nurses' productivity in Iran; findings of studies undertaken in other countries may not be directly applicable to the Iranian context due to significant cultural and socioeconomic differences.
Identifying factors that affect productivity from the nurses' viewpoint leads to an understanding of how best to improve productivity. Therefore, this study aimed to highlight nurses' perception of their own productivity and factors affecting it. It attempted to address questions such as "What are nurses' perceptions of their own productivity?" and "Which factors improve or diminish nurses' productivity?"
To address the former questions a qualitative methodology was adopted. It was considered that this was the most suitable approach to address the complexities of the subject of this study, which a quantitative approach may fail to fully elucidate. A qualitative approach was used as an important and essential first step in understanding the participants' insiders' view and their perceptions. This method resulted in a set of themes and categories about the perceived productivity and human resource elements aiding or impeding nurses' productivity.
Method
Semi-structured interviews were used to gather data covering nurses' perception of their own productivity and factors influencing it. This approach provides understanding of any phenomena at a deeper and richer level that could not be reached through a quantitative approach [19]. Content analysis has been applied to transcriptions and interviews for the purpose of understanding human behaviours within various contexts, as it is an unobtrusive means of analysing interactions and provides insight into complex models of human thought [21].
In the current study, the content analysis method was used to identify categories and subcategories in participants' descriptions. Qualitative content analysis elicits contextual meaning in context through the development of emerging themes. This method consisted of identifying, coding and summarizing the concepts and themes, consistent with established qualitative data analysis methods [20]. Also this is a method that uses a set of categorization procedures for making valid and replicable inferences from data (text or images) to their context. Researchers analysed the presence, meanings and relationships between words and concepts, and subsequently made inferences about the messages within the texts.
This method was particularly suitable for productivity, as the outcome of social interactions lends itself to content analysis of the core aspects of social interaction. Furthermore, this method allowed a closeness to text that can alternate between specific categories and relationships. This method enabled us to explore rich data and allowed an interpretive understanding of what was going on.
Sampling and data gathering
Sampling was targeted based on a set of predetermined criteria. For instance, in the current study, the researchers made preliminary sampling decisions to recruit staff with a minimum of five years' nursing experience working in hospitals affiliated to Tehran University. It was considered that the participants would have sufficient work experience to enable them to analyse factors affecting their productivity and its process. The sampling was based on the maximum variant approach.
Sampling started with a nurse with 29 years' experience and then extended to other nurses, managers or supervisors in the same teaching hospital or others. The setting was education hospitals affiliated to Tehran Medical University.
In this study data collection and analysis proceeded concurrently with the development of themes related to the reality of the nurses' productivity. Sampling continued until saturation was reached: this was when no new categories or subcategories emerged [22].
Participants and interviews
Initially, the researcher contacted each of the potential participants and explained the objectives of the study. If they agreed to take part in the research, interview time and setting were arranged. Semi-structured interviews, based on clear guidelines, were usually carried out in a rest room on the ward. Participants were initially asked to describe a typical working day. They were subsequently asked about their perception of nurses' productivity and asked to highlight experiences where they had felt productive as a nurse and identify factors that had a positive or negative impact on their productivity. Each interview was approached individually, guided by participants' responses while covering the core research questions.
The selection criteria were that they be staff members with more than five years of nursing experience who worked full-time in hospitals covered by the Ministry of Health and Medical Education in Tehran, Iran. Any nurse who met this qualification was considered a potential participant.
A total of 26 participants in various positions were interviewed. Interview sessions lasted from 30 minutes to 70 minutes. Interviewing continued on all shifts, over a three-month period, until the information from the last four interviews with new participants was becoming repetitious and thus saturation of the data seemed to have been achieved.
Ethical considerations
This study formed a part of a PhD thesis in Tehran Medical Science University. The ethics committee of the University reviewed and corroborated its ethical considerations. All the participants were informed of the purpose and design of the study, and that their participation was voluntary and would be treated with confidentiality. Participants were asked to sign a form confirming informed consent prior to taking part in the study. Audiotaped, semi-structured interviews were conducted in private rooms and solely by the researcher. In addition, permission was obtained from hospital directors and head nurses for the nurses to be interviewed in their work setting. They were assured that participants could cease participating in the study at any point.
Data analysis
All interviews were audiotaped. Analysis of data was based on overall impression, reading and re-reading of transcripts. Analysis consisted of identifying, coding and summarizing the concepts and themes, consistent with established qualitative data analysis methods [20]. Data were analysed by the constant comparative analysis method and coding process. Tapes were transcribed and data were broken down into discrete parts and codes were noted next to phrases, words or comments in the text. Then codes were sorted into categories and redefined into further focused categories.
All data were grouped within the categories according to their "fit". When required, the researcher returned to the field to extend categories. For instance, some of the interviews and codes indicated that the education process has a significant impact on the group's productivity, and subsequent sampling focused on this factor. Also, some of the interviews and codes indicated that nurse managers – especially supervisors and matrons – have a significant impact on the group's productivity through adopting appropriate qualitative and quantitative strategies to address staffing issues. In addition, they have valuable experience about group productivity. They worked as nurses in the wards before becoming managers. Throughout the process the researcher investigated the causal conditions arising from the data.
Validity and reliability
To maintain trustworthiness of the conclusions, Lincoln and Guba's four criteria were used. This in quantitative research is equivalent to empirical positivistic criteria (validity and reliability). The four trustworthiness criteria are credibility, confirmability, transferability and dependability [22].
Credibility was enhanced through prolonged engagement with participants, the achievement of saturation, sampling approach (maximum variant sampling). In particular, participants' revision of coding – member check – supported credibility. After data analysis participants were provided a complete transcript of their coded interviews with emergent themes. They verified whether the codes and themes matched their answers. Maximum variant sampling also validated the confirmability of data.
Also reliability of study results was further enhanced when other researchers in the field were asked to review the interviews and the coding process. They were three expert supervisors and two other faculty members experienced in qualitative research. They checked a large number of the transcripts of interviews and coded them. There was close agreement (90% or higher) between the resultant coding achieved by a number of researchers.
Furthermore, the results were discussed with other nurses who did not take part in the research but who confirmed the soundness, fitness and transferability of the results. This confirmed transferability of results.
Results
Data saturation was reached with 26 participants. Twenty-four of the participants were female, 30 to 56 years of age. All of them had bachelor's or master's degrees. They had between 5 and 29 years of work experience with a mean of 17 years. Participants were clinical nurses, head nurses, supervisors, nurse managers and nursing educators.
The findings of this study indicate that human resources is the most important factor affecting nurses' productivity. Other factors included managers' role, organizational structure and culture, social factors, financial security and the nature of the nursing work undertaken.
In this paper, we have discussed two categories – human resources and productivity – that emerged from content analysis. The human resource category included several subcategories. Categories and subcategories of nurse productivity that emerged are reported, together with certain extracts from the interviews to highlight nurses' experiences and the manner in which they have has been communicated.
Definitions of productivity
Participants defined their productivity from different perspectives. Although participants used different terms and occupied various positions, they were referring to the same concept and themes that were highlighted in almost all interviews. For instance, while some used the term "usefulness" others chose terms such as "effectiveness", "being efficient", "providing high quality care" and "being at the bedside and providing good care" to refer to productivity. We classify these words into two categories: the quantitative aspect, which is equivalent to efficiency, and a qualitative feature that is equivalent to effectiveness.
Most of the participants referred to productivity as a qualitative feature, while some of them implied both quantitative and qualitative aspects. In certain cases, participants referred to productivity based on prerequisites or outcomes relating to productivity. However, the majority of the participants in the current study considered productivity as equivalent to "helpfulness, effectiveness for patients and providing high quality care". Therefore, productivity in these nurses' opinion means to be effective for the patients, and that they feel they are highly productive when "they focus on taking care of patients". Table 1 shows productivity from the Iranian nurses' viewpoint.
Table 1 Productivity from the Iranian nurses' viewpoint
Quantity Quality Prerequisites Outcome
To be efficient To be effective Commitment to the organization Patient satisfaction
To use time properly Providing high-quality care To be conscientious Personnel satisfaction
Accomplishing all the necessary tasks To be useful to patients To be responsible Organization profit
Doing work correctly Being at the patient's bedside (accessibility for patients) To be careful
Working with high ability Doing the right work To be aware
Working with minimum facilities Solving patients' problems
One of the participants stated: "In my opinion, our productivity is increased when we can provide our clients skilled care as well as providing a condition in which the patients are treated and possibly regain their full health" (a primary-care nurse). A staff nurse said: "I feel productive when giving good care and meeting their needs ". Another nurse said she felt a sense of productivity "when my patients feel satisfied at the end of the shift".
Most of the participants believed that nurses do not have high productivity and cited a number of reasons for this; others stated that under present conditions nurses could not be expected to do a lot more. One of the participants said: "I think considering what we studied and what is expected of us, nurses' productivity is about 20 %"(a head nurse). She added that nurses' energy was spent mostly on administrative tasks, which has an adverse impact on their productivity and effectiveness for patients.
Thus from the participants' viewpoint, existing productivity was affected by the lack of resources, especially the shortage of experienced personnel, as well as having to perform non-nursing administrative tasks.
Nurses in this study also believed that lack of equipment, shortage of personnel and disproportionate nurse-to-patient ratio have negative effects on productivity. For example, one participant said: "Considering the existing facilities and very high numbers of patients, nurses could even be considered to have high productivity levels compared with other professionals, but if compared with developed countries, nurses' productivity in Iran is poor" (a nurse with 20 years of experience). Overall a range of factors – especially shortage of nurses, workload and inexperienced personnel – all have a negative impact on nurses' productivity.
Nurses believed that they use only 15–20% of their productivity as they have to perform certain duties that they felt they should not perform. That is to say, 80% of their productivity is wasted, as it is not directed to patient care.
One of the participants with extensive clinical, management and nursing education experience, referred to unfavourable conditions for nurses, saying: "Productivity promotion requires certain conditions of which even the minimum do not exist for Iranian nurses. Even though the number of nurses is probably 30%–50% of the minimum required, the expectations are very high." Finally participants believed that most of the factors that hinder their productivity can be altered by managers through modification of human resource elements.
Human resources
The participants most frequently referred to human resource elements, and believed them to be the most important factor affecting productivity. Although the impacts of such a factors are widely known by participants – managers and nurses – alike, they believed that human resource impediments are still widely prevalent in all spheres of their work. The subcategories of human resources are: personnel needs and nurse staffing, staff expertise and experience, work coordination and teamwork, and effects of present staffing.
Personnel needs and nurses staffing
In the participants' view, one of the factors facilitating achievement of higher productivity is an appropriate staffing level. They believe this should be assessed in the context of nurses' functions and role within the system. It is possible to evaluate the appropriate staffing requirements only if nursing workload, patient needs and their required care and the level of non-nursing administrative tasks are fully considered. Participants stated that all the latter factors need to be considered, if the correct staffing levels are to be achieved and personnel shortage and excessive workloads are to be avoided. A supervisor said: "nursing workload is not assessed systematically; decision-makers only consider routine tasks. Our hospitals put additional demands on nursing staff from direct patient care to voluminous clinical and medical documentation... but these are not assessed when personnel needs are estimated for wards".
While one of the factors facilitating achievement of higher productivity is appropriate staffing level, participants believed that in most cases the wards are facing a shortage of professional and assistant nurses. As one surgical nurse said: "We are usually two nurses for 35 patients, without doubt we cannot supply a high-quality care and to be productive". Another nurse said: "We can only give the drugs and monitor vital signs ... a routine work, we cannot provide effective care because there is so much work". One of the general nurses said: "I don't feel productive when it is too busy and work overload, and I can't help all my patients". Another nurse said: "I know that we don't have the time to provide all the care correctly".
Participants believed this shortage is in turn due to a non-systematic approach, which fails to fully take into account factors such as physical structure of the ward, facilities, nurses' experience and their skills, nature of the care patients require and admission procedures, as well as circumstances of other hospital services such as the pharmacy and catering section. Participants stated that all of these factors need to be considered if the correct staffing levels are to be achieved and personnel shortage and excessive workloads are to be avoided.
For instance, one of the participants highlighted the effect of the physical structure of the unit, which has an impact on the number of required personnel. She stated "This unit has a very long corridor. If you are at one end of the unit and need something, you have to walk a long distance to reach the nursing station and this affects the number of required personnel" (chief nurse of a surgery ward).
Nurses cited numerous experiences highlighting shortcomings for patients and the nursing staff alike resulting from shortages of human resources. For instance, one said: "There have been situations in which we have had three ICU cases, three cases of bedsores and patients needing tube feeding and suctioning. Sometimes we have to spend one hour on a patient's dressing. In these circumstances, how can nurses provide adequate care as and when needed? When one has to complete work that requires four members of staff, this invariably gives rise to immense stress". Another participant said: "The pressure of work and overloading of responsibilities made nurses so busy that they seldom have enough time to be productive – that is, we cannot provide excellent care".
Personnel shortage has also resulted in patients' care being delegated to non-nursing personnel and the patients' family, who probably lack sufficient skills. Finally this will result in a decline in productivity and quality of care and cause patients' dissatisfaction. One participant said: "We give patients' care to their family because we cannot meet all the patients' needs...we seldom have enough time ... we are very busy".
In participants' views, lack of staff replacement cover in cases where personnel are on leave or when staff have completed their Free Education Compensation Period (FECP) as well as periods when no FECP persons are referred to hospitals, are other factors that increase staff work pressures. Furthermore, a reluctance of FECP personnel to work in centres with no NET is another parameter that exacerbates hospital staff shortages and increases the workload for the existing staff.
Delegating nurses to nursing roles not in line with their qualifications or nurses' working in other sections restricts appropriate human resource distribution, resulting in low nursing productivity and inappropriate patient care. One nurse manager said: "Our NETs work everywhere (pharmacy, etc.) except as a nurse on the hospital ward, but when assessing the staffing levels within the wards, they are included in the nursing statistics".
Selection procedures
As nursing aims to serve the public it needs active and motivated personnel. Correct selection of nurses and nursing students can promote nurses' productivity, according to respondents. In the participants' view, selection of capable nurses is possible only through setting proper selection procedures and adhering to them in order to ensure nurses are selected based on merit. One of the supervisors said: "Employment regulations should be closely adhered to, but here the people who perform the selection process are by no means capable."
Staff expertise and experience
Participants cited nursing personnel's skill and experience as one of the factors resulting in productivity improvement. Also, new-staff orientation and assigning newly qualified staff to work alongside experienced personnel are two important elements aiding nursing team productivity, in the participants' viewpoint. Nevertheless participants, especially managers, have pointed out the shortcoming in this respect and said they believe that staff lack of knowledge is a significant barrier to being productive and providing a high quality of care. They cited application of knowledge and skill in special wards as the sign of effective care. They believed they were more productive in such wards for this reason alone.
Participants, especially those in management positions, argued that employing recently graduated staff could diminish the overall system's quality because of inadequate experience. One participant said: "All shifts were covered by newly graduated personnel who are novices and inexperienced. It takes them so long to adjust and learn clinical work, and the patient suffers the consequences of the shortcomings" (a manager of a paediatrics hospital).
Participants believed that hiring newly graduated staff who have little experience and skill and a high ratio of these staff compared to experienced personnel, as well as frequent personnel turnover, are among factors that lead to low overall skill and experience level within the ward. They believed these factors contribute to increased pressure on the nurses and managers, and ultimately lead to complications for the patients. One participant commented: "It is not only the number of staff that has an impact on productivity, but the skill and experience of the personnel also play an important role"(a staff nurse).
Participants claimed that permanent employment was a significant positive factor contributing to increased skill, experience and a more responsible approach towards patients, and in the organization as a whole leads to improved productivity.
Work coordination and teamwork
In the participants' viewpoint, not only is a systematic assessment of human resource requirements lacking, but one of the factors impeding productivity and effective care is a lack of coordination among different workgroups. Un-systematic patient admission creates an imbalance in the nurse/patient ratio within the ward. Informants believed that this is caused by a lack of harmony between the physicians and hospital managers as well as uncoordinated patient transfer from other hospitals and emergency centres. One participant said: "As a result of uncoordinated patient transfer from other hospitals, our ICU beds are 110% occupied; while our staff are sufficient for 80%–90% occupancy" (a supervisor of infant care).
In addition to the number of personnel, informants believe that appropriate communication between team members is an important factor contributing to their productivity. Effective communication can improve group performance and reduce the effect of excessive workload caused by personnel shortage on tolerance levels and coordination within the ward, resulting in higher productivity. Participants frequently referred to their close relationships with managers, which they believed resulted in higher productivity. For instance, one participant said: "There should be strong relationships between the personnel in a trusting environment. This would enable staff to discuss and address any possible errors at an early stage to avoid further damage" (a chief nurse of a unit). Another nurse said: "We worked under difficult conditions within the emergency unit but sound working relations between personnel and in particular with the unit manager enabled us to work effectively." However, participants believed shortage of staff and a heavy workload can have a negative effect on interpersonal relations.
Effects of present staffing
Participants believed that shortage of nursing staff has a negative impact on both the productivity of the nurses and the effectiveness of care. Moreover, this leads to certain significant care procedures' being overlooked and increased errors and necessitates involving unsuitable staff or patients' friends or family members in the care process.
Personnel shortage hinders effective management processes such as comprehensive personnel performance appraisal, which is a prerequisite of improving productivity and patients' care. In this context, a head nurse commented: "If I criticize the quality of nursing care they respond by pointing out that there is merely one nurse for every 20 patients. They obviously have a point, and I have to overlook certain shortcomings".
Staff shortage impedes nurses' productivity directly and indirectly and creates a difficulty for both patients and nurses. Participants stated that staff not practising learnt principles, not providing sound nursing care, overlooking patients' education and effective communication with patients, not assessing the patients' condition and not solving patients' problems are all attributable to staff shortage. In addition, participants have stated that providing merely routine patient care devoid of attention to individual needs, increased risk of practical errors and sometimes even being unaware of the presence of certain patients' within the ward until their discharge are all due to personnel shortage and unsuitable human resources.
The informants believed that personnel shortage not only adversely effects patient care, but also over a period of time excessive work pressure and undertaking a range of duties results in nurses' loss of knowledge and motivation, exhaustion, burnout, severe stress and eventually leads to staff leaving the profession. One of the participants said "It gives us a sense of burnout... masses of things you constantly feel you should be doing, looking after the patient, you have to do the documentation. It is a high pressure. I don't sense I am productive; I think I am burned-out, tired".
However all the participants believed that appropriate staffing level is the most important factor facilitating their productivity. When asked how to improve productivity, all the participants indicated that provision of human resources to a high standard in all nursing ranks, as well as proper personnel planning and organization, would enhance productivity. One said: "One way to ensure that care is accomplished properly and completely is to have adequate staff" (first staff of a ward). Also, participants believed that most of the factors that hinder their productivity can be altered by managers through modification of human resource elements.
Ultimately participants believed that achieving increased productivity requires certain conditions, but that the present circumstances lack even minimum criteria and than an inappropriate human resources pattern is a major obstacle in this respect. One said: "When talking about productivity of a group, certain basic conditions should be in place before they could be expected to perform to the required standard. Currently only one nurse is assigned to care for five patients in ICU whereas they allocate three nurses for two or three beds in other countries" (a participant with clinical and management experience).
Discussion
In this study, nurses described productivity as "being effective for the patients" and said they feel they have high productivity when "focused on taking care of patients".
The finding of this research has been presented in a diagram (Fig. 1). The concepts within this diagram include systematic evaluation of staff numbers required; appropriate, adequate and regulated selection process; supplying staff of different ranks; patient admission coordinated with the head nurse; and appropriate working relations within teams. Adhering to these themes on the whole creates the potential for increased productivity, which in turn reduces negative impact of inappropriate human resource provision.
Figure 1 Nurses' productivity and human resources.
The findings of this study indicated that productivity from nurses' perspective differs from the adopted models suggested by industry and economy, which focus on output rather than outcome. The output is amount produced from input. This is a quantitative fragment but outcome is all the results of a work process.
The key point here is that although productivity from an administrative organizational perspective is represented by cost-effectiveness, nurses often assess their own productivity by quality of care provided to patients. This fundamental difference between nurses' perception and that of managers results in different assessments. In this study, nurses believe that to improve productivity there should be adequate and qualified human resources, while organizations consider increased productivity in terms of limiting permanent employment, imposing compulsory overtime with the least financial benefit to workers, hiring FECP staff and decreased cost.
Even though productivity has so far been considered by organizations and managers from a quantitative perspective (efficiency), nurses often emphasized the qualitative dimension and their own effectiveness. Participants consider productivity in the context of the whole care process and its outcome. This gap reflects the different perspectives of those who decide nursing budgets and often are not nurses themselves and that of nurses, who are in direct contact with the patients. Although for an organization's managers the logic of cost-efficiency prevails, for nurses providing high-quality care is the major goal that not only results in patients' self-sufficiency and their return to full health but promotes the profile of the nursing profession as a whole.
Accordance to results of this research, management experts believe that methodology of staffing is a systematic process that is used to evaluate the exact numbers and type of personnel needed to provide the standard care in an organization [23]. Researchers believed that although many staffing studies use various methods to quantify the volume of tasks that nurses must provide for patients, few consider the impact of non-staffing elements on nurses. They did not consider nursing work context as a system. Nurses do not care for patients in isolation but within complex organizations that make great demands on them [4]. Analysis of nurses' work structure based on 100 000 individual observations in two hospitals over a seven-year period revealed that about two-thirds of nurses' time was spent on indirect care and merely a third was dedicated to direct care [8].
Also, researchers believed that nurses' workload is dictated by factors such as nursing standards, nurses' experience and skill, organizational strategies and procedures, available equipment and the activities of other members within the health care team [24]. These are according to results of this research.
Former studies have shown that patterns and personnel combinations as well as skill combinations and education levels of the personnel positively affect productivity [25] and the fact that a higher ratio of experienced nurses decreases hospital operational costs [26]. Researchers believe that chief executives and their management team can influence productivity by hiring skilled, industrious and capable people [27]. Hall (2003) concluded that the future of nursing depends on the knowledge and skill of the nursing workforce and the availability of adequate numbers of nurses for the health care sector [3]. Such sentiments agree with findings of this research that staff expertise and work experience positively affect productivity.
Hall (2003) has stated that since individuals differ in intelligence, skills, motivation and personality it is essential that organizations determine what qualities they are seeking when recruiting new staff, in order to ensure that they can function effectively within the organization [3]. According to the results of research, Huber (2000) also believed that planning and staffing levels affect nurses' career and work conditions, workload, personal life and morale. Also she confirmed that security and quality of care to the clients are affected by staffing [28].
Work coordination and teamwork is one of the themes that emerged from the data. Houser (2003) demonstrated that effective teamwork and greater expertise can have a major influence on the outcome of patient care [4]. Curtin's research findings showed the relationship between teamwork and prevention of complications [11]. Also, previous research suggested that the quality of communication among disciplines is important for achievement of positive outcomes [29].
Another finding of this study was that nurses cannot consider themselves productive enough as a result of human resource shortages and that they are concerned about the quality of care provided to patients. Smith's findings (2001) also confirmed the latter results [5].
Conclusion
This study revealed nurses' views of productivity and human resource factors improving and impeding it. Nurses are fully aware of the importance of the qualitative dimension of productivity and factors contributing to it. The goal of productivity is to promote an adequate and satisfactory level of nursing care acceptable by patients, nurses and physicians [30]. If these findings can contribute to nurses' and managers' improved productivity, not only the patients will benefit but the nurses' quality of work life will also improve. The other feature of productivity improvement is its positive effect on attitude of patients, personnel, managers, nurses' job satisfaction, improved morale and budget control.
Managers who are aware of nurses' viewpoints would be able to create conditions to achieve higher productivity levels. The findings are also beneficial to managers, enabling them to identify and promote productive work practices among clinical staff.
Although a wide range of participants' views has been studied in this research, the small number of informants may limit the generalizability of the results.
As health care organizations have a dynamic nature, further studies of productivity trends and the changing nature of its contributing factors is strongly desirable. Furthermore, it is recommended that comparable studies be carried out in different parts of the country, together with quantitative studies to reinforce the studies' results.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NDN: Initiation and design of the research, collection and analysis of the data and writing the paper.
AAN: Main supervision of the project, and editorial revision of draft papers.
MAH: Supervision of the project, co-analysis of the data and editorial revision of draft papers.
FA: Supervision of the project, co-analysis of the data and editorial revision of draft papers.
Acknowledgements
The authors would like to acknowledge the assistance of all the participants who contributed in this research.
The authors also gratefully acknowledge the very helpful critiques of Dr Judith Spiers (member of the International Institute of Qualitative Methodology, University of Alberta) and Dr Mohsen Adib Hajbagheri (Associate Professor, Faculty of Nursing, Kashan University of Medical Sciences).
We are very grateful to Professor Pat Hayes (Editor of Clinical Nursing Research Journal) for her useful comments.
==== Refs
Eastaugh SR Hospital Nurse Productivity Journal of Health Care Finance 2002 29 14 22 12199492
F Helmer TSJD Pictures of Performance: The Key to Improved Nursing Productivity Health Care Management Review 1988 13 65 70 3230011
Hall LM Nursing Intellectual Capital: A Theoretical Approach for Analysing Nursing Productivity Nursing Economics 2003 21 14 19 12632713
Houser J A Model for Evaluating the Context of Nursing Care Delivery he Journal of Nursing Administration (JONA) 2003 33 39 47 10.1097/00005110-200301000-00008
Mc-Neese-Smith DK Staff Nurse View of Their Productivity and Nonproductivity Health care Management Review 2001 26 7 19
BR Holcomb NHMF Nursing Theory and Concept Development or Analysis, Defining and Measuring Nursing Productivity: A Concept Analysis and Pilot Study Journal of Advanced Nursing 2002 38 378 386 11985689 10.1046/j.1365-2648.2002.02200.x
Moody RC Nurse Productivity Measures for the 21st Century Health Care Management Review 2004 29 98 106 15192982
J W Williamson JJ Understanding, Evaluating and Improving Productivity Nursing Nursing Management 1988 19 49 52
K Beverley BK Manage productivity with HPPD Nursing Management 2003 34 47
O Vincent NR Measuring Productivity: Outcome versuss Output Nursing Management 1989 20
Curtin LL Nursing Productivity from Data to Definition Nursing Management 1995 26 25 32 7731592
P Varca MJV The Relationship of Ability and Satisfaction to Job Performance Applied Psychology: An International Review 1993 42 265 275
Daley D Productivity Tools and Employee Incentives: Availability and Associations among North Carolina Municipalities Public Productivity & Management Review 1993 17 161 171
Bartel AP Productivity Gains from the Implementation of Employee Training Programs Industrial Relations 1994 33 411 425
JC McCloskey BMC Variables Related to Job Performance Image 1988 20 203 207
McNeese-Smith DK Increasing Employee Productivity, Job Satisfaction, and Organizational Commitment Hospital and Health Services Administration 1996 41 160 175
McNeese-Smith DK Job Satisfaction, Productivity, and Organizational Commitment Journal of Nursing Administration 1995 25 17 26 7674041
S Grosskopf DMVV The eFFects of teaching on hospital productivity Socio-Economic Planning Sciences 2001 35 189 204 10.1016/S0038-0121(01)00006-4
A Strauss JC Basics of Qualitative Research: Grounded Theory Procedures and Techniques 1994 Newbury Park , Sage Publications
writing guides content analysis http://writingcolostateedu/indexcfm
J Rugkasa BKJS Anxious Adults vs. Cool Children: Children’s Views on Smoking and Addiction
Soc Sci Med 2001 53 593 602 11478539 10.1016/S0277-9536(00)00367-1
D F Polit CTBBPH Essentials of Nursing Research 2001 5
philadephia , Lippincott Company
R C Swansburg RJS Introductory Management and Leadership For Nurses 2000 2 Boston , Jones and Bartlett Publisher
Nursing Professional Advisory Working Group of the JPPC MIS: Nursing Resource Consumption 1997
S D Jordan Nursing Productivity in Rural Hospital Nursing Management 1994 25 58 62
J R Bloom JAAANB Nurse Staffing Patterns and Hospital Efficiency In The United States SocSci Med 1997 44 147 155 10.1016/S0277-9536(96)00063-9
Quality compensation incentives work when system fosters quality Health Care Strategic Management 2001 19 15
Huber D Leadership and Nursing Care Management 2000 2 Philadelphia , Saunders Company
L M Hall DDISRS A Study of the Impact of Nursing Staff Mix Models and Organizational Change Strategies on Patient, System and Nurse Outcomes 2001 University of Toronto,Faculty of Nursing
Walker DA Bottom-Line Approach to Nurse Staffing Nursing Management 1996 27 31 32 8932099
|
16212672
|
PMC1276809
|
CC BY
|
2021-01-04 16:37:43
|
no
|
Hum Resour Health. 2005 Oct 8; 3:9
|
utf-8
|
Hum Resour Health
| 2,005 |
10.1186/1478-4491-3-9
|
oa_comm
|
==== Front
Int J Health GeogrInternational Journal of Health Geographics1476-072XBioMed Central London 1476-072X-4-251624204310.1186/1476-072X-4-25ResearchGeographical variation of cerebrovascular disease in New York State: the correlation with income Han Daikwon [email protected] Shannon S [email protected] Peter A [email protected] Frederick E [email protected] Department of Social and Preventive Medicine, University at Buffalo, Buffalo, NY 14214 USA2 Department of Geography and National Center for Geographic Information and Analysis, University at Buffalo, Buffalo, NY 14261 USA3 Department of Neurology and Jacobs Neurological Institute, University at Buffalo, Buffalo, NY 14203 USA4 Department of Biostatistics, University at Buffalo, Buffalo, NY 14214 USA2005 21 10 2005 4 25 25 29 6 2005 21 10 2005 Copyright © 2005 Han et al; licensee BioMed Central Ltd.2005Han et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Income is known to be associated with cerebrovascular disease; however, little is known about the more detailed relationship between cerebrovascular disease and income. We examined the hypothesis that the geographical distribution of cerebrovascular disease in New York State may be predicted by a nonlinear model using income as a surrogate socioeconomic risk factor.
Results
We used spatial clustering methods to identify areas with high and low prevalence of cerebrovascular disease at the ZIP code level after smoothing rates and correcting for edge effects; geographic locations of high and low clusters of cerebrovascular disease in New York State were identified with and without income adjustment. To examine effects of income, we calculated the excess number of cases using a non-linear regression with cerebrovascular disease rates taken as the dependent variable and income and income squared taken as independent variables. The resulting regression equation was: excess rate = 32.075 - 1.22*10-4(income) + 8.068*10-10(income2), and both income and income squared variables were significant at the 0.01 level. When income was included as a covariate in the non-linear regression, the number and size of clusters of high cerebrovascular disease prevalence decreased. Some 87 ZIP codes exceeded the critical value of the local statistic yielding a relative risk of 1.2. The majority of low cerebrovascular disease prevalence geographic clusters disappeared when the non-linear income effect was included. For linear regression, the excess rate of cerebrovascular disease falls with income; each $10,000 increase in median income of each ZIP code resulted in an average reduction of 3.83 observed cases. The significant nonlinear effect indicates a lessening of this income effect with increasing income.
Conclusion
Income is a non-linear predictor of excess cerebrovascular disease rates, with both low and high observed cerebrovascular disease rate areas associated with higher income. Income alone explains a significant amount of the geographical variance in cerebrovascular disease across New York State since both high and low clusters of cerebrovascular disease dissipate or disappear with income adjustment. Geographical modeling, including non-linear effects of income, may allow for better identification of other non-traditional risk factors.
==== Body
Background
Cerebrovascular disease disproportionately affects certain areas of the United States, including many areas within New York State [1]. Western New York, specifically, has excessive rates of cerebrovascular disease whereas other areas of New York State are seemingly under-affected. In recent years, researchers from the Department of Neurology at the State University of New York at Buffalo have analyzed age-, gender-, and race-adjusted cerebrovascular disease hospitalization data throughout New York State, focusing mostly at the regional and county levels. These analyses demonstrated that there are rates that are higher than those in other regions, such as Western New York State, and these data suggested that these geographic differences cannot be fully attributed to age, gender, and/or race.
Epidemiologic studies, including our own, that have explored causes of cerebrovascular disease among various populations, have historically focused on identifying associations between vascular disease and traditional risk factors. This research is well-founded since cerebrovascular disease, the third leading cause of death in the US, has long been associated with such traditional risk factors as hypertension, diabetes, elevated cholesterol, obesity, and tobacco use [2-4]. However, traditional risk factors and demographic characteristics such as age, gender, and race explain only some of the observed variance in vascular disease rates. In recent years non-traditional risk factors, such as income and education, have emerged as predictors of cerebrovascular disease. Unlike traditional risk factor prevalence, which is largely difficult to measure, socioeconomic risk factor prevalence is often known.
A number of studies have explored the relationships between vascular disease and socioeconomic risk factors and have identified associations [5-8]. Recent literature has shown that socioeconomic factors, and specifically income, are more determinative of vascular health than traditional risk factors and may, in fact, be the best predictors of vascular disease [6]. Socioeconomic factors have been shown to contribute directly to behavioral causes of vascular disease where studies have indicated a greater propensity for behavioral risk factors among persons who had not completed high school and among those who are unemployed or are employed in unskilled or low-paid positions [9]. A myriad of studies have focused on the many ways in which income can adversely affect health and have demonstrated evidence of an association between income and cerebrovascular disease [10-12]. Shi et al. maintain that income inequality and stroke mortality are related in that income inequality affects psychosocial factors that exacerbate stroke risk factors [10]. Some theories regarding this association contend that disparities in income and social status create appreciable strain that ultimately impairs one's health [13,14].
In addition, there is considerable geographic variations in cerebrovascular disease mortality across various geographical scales; in the US and world, and even within New York State [1,15,16]. Disease mapping and cluster analyses have been used in addressing public health concerns, especially when one is interested in identifying spatial patterns of disease [17,18] Furthermore, geographic clustering analyses can be used as exploratory tools to identify areas of elevated disease risk, and thus to provide hypotheses on causal relationships of disease mechanism and some clues for unknown etiologic studies. For example, recent studies of geographic clustering of breast cancer incidence and mortality rates in the Northeastern US found significant spatio-temporal clustering of breast cancer in the Northeast [19,20].
There is cause to believe that income, is contributing to the heretofore unexplained variance in disease rates carried by certain areas of New York State. We questioned if the rates generated during our previous work would change once income was accounted for. We utilized a new geographic analysis method to examine clustering of cerebrovascular disease and the non-linear effects of income, one that has not been previously applied in cerebrovascular disease and socioeconomic risk data analysis in New York State.
The purpose of this study was to 1) pursue a cross-disciplinary, innovative approach to identifying a significant nontraditional socioeconomic risk factor, 2) correlate this socioeconomic risk factor with the prevalence of cerebrovascular events at the ZIP code level in New York State in the year 2000, and 3) apply income-adjusted geographic clustering analyses to identify geographic patterns of cerebrovascular disease correlated with income within ZIP codes in New York State. We hypothesized that the novel approach of geographical cluster analysis, together with nonlinear regression that associated spatial distributions of income with cerebrovascular disease spatial distributions would enhance the power to predict event rates as compared to traditional risk factors. Such factors would have the power to facilitate the identification of high-risk ZIP codes or groups of ZIP codes for direct interventions, as well as low-risk ZIP codes or groups of ZIP codes for further exploration.
Results
Geographic clustering of cerebrovascular disease in New York State
Figure 1, geographic clustering of cerebrovascular disease in New York State, shows the locations of clusters of significantly raised prevalence; (a) areas with local statistics greater than 3.85 are shown in dark red as statistically significant clusters that exceed the critical value, and (b) areas with local statistics greater than the commonly used threshold of 2.5 and less than 3.85 are highlighted in light red. The maximum local statistic was 9.39, and it was obtained in the center of the Buffalo-Niagara cluster (ZIP code 14224 in Erie County). One hundred ZIP code areas throughout New York State exceed the critical value; these areas contain 7,935 observed cases and 6,478.9 expected cases, yielding a relative risk of 1.23.
Figure 1 Clustering of cerebrovascular disease in New York State. Shown are map of New York State by zip code boundaries; areas of high prevalence of cerebrovascular disease are presented in red (local statistics over 3.85 and between 2.5 and 3.85), while areas of low prevalence are in blue (under -3.85 and between -2.5 and -3.85).
We also identified locations of significantly low prevalence; (a) areas with local statistics less than -3.85 are depicted in dark blue, and (b) areas with local statistics between -3.85 and -2.5 are shown in light blue. The minimum local statistic was -8.42 (corresponding to ZIP code 14892 in Chemung County). 153 ZIP code areas have local statistics below the critical value; these areas contain 5,412 observed cases and 7,161.3 expected cases, yielding a relative risk of 0.76. We also tested different kernel bandwidths, ranging from σ = 0.6 to σ = 2.5. These correspond to searching for clusters of different sizes; best results (in the form of highest local statistics) were obtained with σ = 1.0.
Effects of income
To determine whether the differences between the observed and expected prevalence of cerebrovascular disease could be attributed to income, we performed a nonlinear regression; the age-adjusted cerebrovascular disease rate was taken as the dependent variable, and income and income squared were taken as independent variables. Although income has been previously noted as a risk factor, we also tested the hypothesis of a nonlinear effect of income. The resulting regression equation was:
y = 32.075 - 1.22*10-4 (income) + 8.068*10-10 (income2)
where y is the predicted age adjusted cerebrovascular disease rate, per 10,000 population. The value of r2 is 0.045, but this is significantly different from zero, given the large number of ZIP codes examined. In addition, both the income and income squared variables were significant at the 0.01 level. As expected, the excess number of cases declines with increasing income. Each $10,000 increase in a ZIP code area's median income results in an average reduction in the age adjusted cerebrovascular disease rate of 1.22 per 10,000 individuals. In addition, the significant nonlinear effect indicates a lessening of the income effect with increasing income. For example, an increase in income from $20,000 to $30,000 results in a decrease in the cerebrovascular disease rate, on average, from 29.96 to 29.14 cases per 10,000 population while an increase from $60,000 to $70,000 results in a smaller decline – from 27.66 to 27.49 cases per 10,000 individuals. The age-adjusted rate is a minimum at an income level of $75,000; for ZIP codes with median income levels above this, the rate, on average, begins to increase.
Clustering analysis with income adjustment
Using the standardized residuals from the regression analysis, we identified geographic clusters of cerebrovascular disease in New York State with income adjustment. In this case the clusters in Figure 2 indicate areas of raised prevalence, once the effects of both age and income have been accounted for. The locations of high prevalence, with local statistics greater than the critical value of 3.85 are in dark red, and areas with local statistics greater than 2.5 and less than 3.85 are depicted in light red. There are two such clusters – one on Long Island and one in the Buffalo-Niagara area. The maximum local statistic was 7.57, and was obtained in the center of the Long Island cluster (ZIP code 11704). 87 ZIP code areas exceed the critical value; these contain 7,418 observed cases and 6,197.7 expected cases, yielding a relative risk of 1.2. Note that once income is included as a covariate, the number and size of clusters of high prevalence decreases; thirteen fewer ZIP code areas were significant after income adjustment.
Figure 2 Clustering of cerebrovascular disease with income adjustment. Shown are map of New York State by zip code boundaries; areas of high prevalence of cerebrovascular disease with income adjustment are indicated in red (local statistics over 3.85 and between 2.5 and 3.85), while areas of low prevalence with income adjustment are in blue (under -3.85 and between -2.5 and -3.85).
We also identified the geographic locations of low prevalence after taking into account the non-linear effects of income. Again those areas with local statistics less than -3.85 in dark blue, and areas with local statistics between -3.85 and -2.5 in light blue are shown in the figure. The minimum local statistic was -4.42 (ZIP code 10069 in Westchester County). Thirteen ZIP code areas had local statistics below the critical value; these contained 917 observed cases and 1,262.3 expected cases, yielding a relative risk of 0.73. The majority of low prevalence areas in Figure 2 disappeared with income adjustment, indicating that many of the areas of low prevalence are explainable by income. Both the areas of high and low cerebrovascular disease prevalence identified are generally high income areas (the mean of the median incomes in the 87 high prevalence areas is $62,515, while the mean of the median incomes in the 13 low prevalence areas is $64,012), when compared to the New York State mean of the median income of $46,678. The resulting curves are presented in Figure 3. The scatterplot shows variation of cerebrovascular disease rates with income; shown are resulting non-linear curves (top), and the curve with scatter plot of observed values (bottom). There are some ZIP codes with zero observed cerebrovascular disease events.
Figure 3 Relationships between cerebrovascular disease rates and income. Shown are resulting non-linear curves (top), and the curve with scatter plot of observed values (bottom).
Discussion and conclusion
We found that income was statistically significantly associated with cerebrovascular disease prevalence after taking into account age. After adjusting for income, the relative risk of having cerebrovascular disease for residents of the Buffalo-Niagara and Long Island regions was 1.2 times greater than for residents in other areas. For residents of Westchester County, the relative risk for cerebrovascular disease was 0.73, inferring a protective effect of residence in that area after adjusting for age and income. The magnitude of the income effect in the nonlinear regression equation is small in part because the effects of income on prevalence are being averaged over the large number of individuals who live in each ZIP code area.
Initial analyses conducted with age but not income adjustment during this study corroborated findings from earlier works produced by our group. Of the one hundred ZIP code areas throughout New York that exceeded the critical value in the high clustering analyses without income, many of the ZIP codes are within western part, north-central part of New York State, and Long Island regions. The maximum local statistic of 9.39 was found in the center of the Buffalo-Niagara cluster. Also, in the analyses performed without income adjustment, highly concentrated areas of low clustering were found in the Finger Lakes area, namely Chemung County. The minimum local statistic was -8.42. This result is not unexpected since it substantiates previous findings as well.
Our analyses demonstrated that a number of areas of high and low disease prevalence of cerebrovascular disease are explainable by income when it is included as a covariate since the majority of clusters were absent when income adjustment was applied during nonlinear regression analysis. This supports our conviction that income, in fact, is a strong predictor of cerebrovascular disease. However, high clusters in the Buffalo-Niagara and Long Island regions remain above both the 3.85 and 2.5 critical values, as described earlier. As well, a significant low clustering area remains in the Hudson Valley region indicating that other factors are moderating low rates of cerebrovascular disease in this region.
Excess numbers of cerebrovascular disease cases decline with increasing income. Each $10,000 increase in a ZIP code's median income corresponds to a decrease in the rate of 1.22 per 10,000 population. This nonlinear effect weakens with increasing income and above about $75,600, the rate begins to increase slowly with income. The finding of high cerebrovascular disease prevalence in high income areas is unusual; one would expect this relationship to be inverse. Our study analyzed morbidity and not mortality data. Therefore, it is possible that residents in higher income areas survive longer with cerebrovascular disease than do those in lower income areas and deaths are not as prevalent.
There may be some bias related to spatial mismatch, since we have used zip-code level hospitalization data and ZCTA-level population and income data in our analysis. The US Census Bureau recently developed a new statistical entity, called ZCTA, to represent the United State postal service-defined zip code areas in a more cohesive way. Those ZCTA may be different from traditional zip codes, even though the ZCTA code equals the zip code in most cases [21]. We are well aware of the potential mismatch issue between the use of ZCTA and zip codes, especially in applying socio-economic data from the Census [22]. Unfortunately, we could not find any empirical study that validates this issue of spatial mismatch. As stated in the method section, our analyses were generally restricted to those areas that are found in both ZCTA and zip code tables. However, we were not able to assess the extent and scope of spatial mismatch, and the effects of such mismatch on the outcomes in this study.
Our study did not distinguish between types of cerebrovascular disease and therefore it is not known if income has a greater effect when correlated with one type of cerebrovascular disease than another. We do not know whether income is a causal factor or only a precipitating factor of cerebrovascular disease since we did not analyze individual-level data, nor did we adjust for other potential confounders. Based on our findings, it will be of great interest to further examine geographic distributions of traditional risk factors and non-traditional risk factors, such as education levels, occupation, measures of community deprivation, and environmental pollutants, to determine their contribution to geographic variations or clustering of cerebrovascular disease in New York State. In addition, it may be useful to examine other variables such as race and ethnicity to explore potential roles and relationships with those non-traditional risk factors.
In summary, income is a nonlinear predictor of cerebrovascular disease. Income alone explains a significant amount of the geographical variance in cerebrovascular disease across New York. These associations were observed after taking into account age. These findings support the contention that cerebrovascular disease cases are susceptible to the influence of socioeconomic factors, notably income. Where clusters failed to disappear, further analysis is indicated to determine what factors are implicated. Additional analysis may also be conducted to further explain the relationship with income. We suspect that a number of factors affect this relationship, including access to and utilization of care, and treatment patterns. These geographic analyses of multiple variables at the ZIP code level allow researchers to determine more precisely where disease events are occurring, along with the causative factors. Further analyses in other geographical scales, such as census tract level, other than at the ZIP code level may ensure these findings. This evidence-based information is necessary in order to affect public policy and isolate small areas such as ZIP codes or groups of ZIP codes for direct health interventions.
Materials and methods
We obtained the Administratively Releasable (ADREL) inpatient hospitalization dataset for New York State from the Statewide Planning and Research Cooperative System (SPARCS) at the New York State Department of Health. Observed prevalence of cerebrovascular disease was extracted from the SPARCS inpatient dataset by ZIP code according to codes listed within "cerebrovascular disease" in the International Classification of Diseases (ICD). Income variables were extracted from US Bureau of the Census 2000 ZIP code level data files. For mapping purposes, ZIP code tabulation area (ZCTA) boundaries were also obtained from the US Bureau of the Census. Several ZCTAs were excluded for purposes of this study since they corresponded to hydrographic features such as lakes, parks, or forested lands. The final merged dataset that was prepared for geographic clustering analysis contained about 1600 ZCTAs, after restricting to those areas that are found in both ZCTA and zip code tables.
The cerebrovascular disease hospitalization rates were calculated using the principal diagnosis code issued at discharge that is included in each individual record within the SPARCS dataset. The following inpatient records from the SPARCS dataset were eliminated from the analyses: a) patients who lived out of state (N = 33,930), and b) patients who were discharged to another acute care hospital (N = 55,320). Note that the above numbers are not necessarily mutually exclusive. The ICD-9-CM codes used to determine cerebrovascular disease included ICD-9-CM code 430.00 through 438.99, cerebrovascular disease. The Census dataset provided population counts by gender and race in five-year age increments for each of the 1600 ZIP codes that had recorded populations. These five-year age groups were collapsed into the following 11 age groupings: 0–24, 25–34, 35–44, 45–54, 55–59, 60–64, 65–69, 70–74, 75–79, 80–84 and 85+. The 11 age groupings determined were appropriate for analysis of cerebrovascular disease; they were chosen to be wide enough to include a reasonably large population in each group, and they were narrow enough that the hospitalization rates would not vary too much within each age grouping.
Using age-specific population data from the Census, the age-adjusted expected number of cerebrovascular disease events was determined for each ZIP code using the indirect method of standardization. Age-adjustment allowed for a comparison without the influence of differences in how much older one population was than another. Several steps were required to obtain the age-adjusted hospitalization rates applying the indirect method of standardization [23]. The age-specific hospitalization rate by ZIP code for cerebrovascular disease was obtained first, where the numerator (age-specific counts of hospitalizations) was obtained from the SPARCS dataset, and the denominator (the age-specific population) was obtained from the US Census dataset. The SPARCS age-specific rate was multiplied by the age-specific Census counts of the population in each ZIP code to calculate the age-specific expected number of hospitalizations in the ZIP code. The total number of hospitalizations expected within each ZIP code was calculated by adding the expected number of hospitalizations across all age groups in the ZIP code.
The standardized rate (SR) for each ZIP code was then calculated as the ratio of the total number of hospitalizations observed in the ZIP code (O) divided by the total number of hospitalizations expected in the ZIP code (E), and the standard error of SR was calculated by applying the formula SE2(SR) = O/(E2). The percentage of excess risk (R) was calculated as SR-1.0, with R having the same standard error as SR. The 95% confidence interval for R was calculated as the interval from R-1.96SE to R+1.96SE.
Spatial clustering methods
To test for the existence of geographic clusters of cerebrovascular disease exhibiting significantly higher or lower observations than could be expected upon the basis of age-structure, we used the statistical test suggested by Rogerson [24]. This approach has an attractive feature where the multiple testing associated with examining many ZIP codes is accounted for (otherwise, one might find "significant" areas of raised prevalence, but only because so many different geographic areas were being examined). In this sense it is similar to the spatial scan statistic popularized by Kulldorff [25]. We first transformed the observed and expected number of cases into a standardized, normal score for each ZIP code. Specifically, the quantity will have an approximate normal distribution, with mean 0 and variance 1, where O is the observed number of cases, and E is the expected number of cases in ZIP code i. To optimize the detection of geographic clusters of a given size, these standardized scores need to be smoothed, by calculating for each ZIP code a z-score (zi) that is a weighted sum of the scores in the geographic neighborhood of the ZIP code:
where the weights are large near the ZIP code, and get smaller with distance:
and where dij is the distance between the centroids of ZIP code areas i and j, and σ is a parameter indicating how quickly the weights change with distance (this can be thought of, approximately, as the radius of the neighborhood around each ZIP code).
The zi scores are known as local statistics, and they also each have a normal distribution with mean 0 and variance 1. If the null hypothesis of no geographic clustering is true, 95% of the time a map of the z scores will have a maximum that is no larger than:
where A is the number of subareas (1600). In our case, we used σ = 1, corresponding to defining, for each ZIP code, a neighborhood of approximately one ZIP code area in each direction. This yields a critical value of z* = 3.85.
An additional step is taken to correct for edge effects before carrying out the weighting described above (areas near the edge of the map must be treated differently, since they do not have as many neighboring ZIP code areas to carry out the smoothing). We began by overlaying a square grid containing lattice points at intervals equal to 5.5 miles (which is the median distance between ZIP code centroids) onto the study area. We then created additional, hypothetical ZIP code centroids around the border of New York State, and assigned them hypothetical, standardized scores, in keeping with the null hypothesis that there was no raised prevalence in these hypothetical locations.
We conducted a similar analysis of geographic clustering after adjusting for income in each ZIP code area. In this case, a regression analysis was carried out by first assuming a quadratic relationship between the excess number of cerebrovascular cases in a ZIP code area and income. We then used the standardized regression residuals (which has a normal distribution with mean 0 and variance 1) as input into the geographic clustering analysis. These geographic clustering analyses were carried out using S-Plus and exported into ArcView GIS for visualization.
Authors' contributions
DH and SSC performed statistical and spatial clustering analysis, and wrote the manuscript. PAR and FEM provided critical review and input. DH, SSC, PAR and FEM participated in the design of the study, participated in interpretation, as well as in data acquisition efforts. All authors read and approved this manuscript.
Acknowledgements
This work was supported in part by NIH grant 1R01 ES09816-01.
==== Refs
New York State Department of Health The Burden of Cardiovascular Disease in New York: Mortality, Prevalence, Risk Factors, Costs, and Selected Populations 2003 Albany, NY
American Heart Association Heart Disease and Stroke Statistics – 2005 Update 2004 Dallas, TX: American Heart Association
Cooper R Cutler J Desvigne-Nickens P Fortmann SP Friedman L Havlik R Hogelin G Marler J McGovern P Morosco G Mosca L Pearson T Stamler J Stryer D Thom T Trends and disparities in coronary heart disease, stroke, and other cardiovascular diseases in the United States: findings of the national conference on cardiovascular disease prevention Circulation 2000 102 3137 3147 11120707
Benjamin EJ Smith SC Cooper RS Hill MN Luepker RV Task force #1-magnitude of the prevention problem: opportunities and challenges J Am Coll Cardiol 2002 40 588 603 12204489 10.1016/S0735-1097(02)02082-X
Kaplan GA Keil JE Socioeconomic factors and cardiovascular disease: a review of the literature Circulation 1993 88 1973 1998 8403348
Raphael D Farrell S Income inequality and cardiovascular disease in North America: shifting the paradigm Harvard Health Policy Review 2002 3
Lawlor DA Bedford C Taylor M Ebrahim S Geographical variation in cardiovascular disease, risk factors, and their control in older women: British Women's Heart and Health Study J Epidemiol Community Health 2003 57 134 40 12540690 10.1136/jech.57.2.134
Lynch JW Everson SA Kaplan GA Salonen R Salonen JT Does low socioeconomic status potentiate the effects of heightened cardiovascular responses to stress on the progression of carotid atherosclerosis? Am J Public Health 1998 88 389 94 9518969
Choiniere R Lafontaine P Edwards A Distribution of cardiovascular disease risk factors by socioeconomic status among Canadian adults CMAJ 2000 162 S13 S24 10813023
Shi L Macinko J Starfield B Xu J Politzer R Primary care, income inequality, and stroke mortality in the United States: longitudinal analysis, 1985–1995 Stroke 2003 34 1958 1964 12843344 10.1161/01.STR.0000082380.80444.A9
Kaplan GA Pamuk ER Lynch JW Cohen RD Balfour JL Inequality in income and mortality in the United States: analysis of mortality and potential pathways Br Med J 1996 312 999 1003 8616393
Kawachi I Kennedy BP Lochner K Prothrow-Stith D Social capital, income inequality, and mortality Am J Public Health 1997 87 1491 1498 9314802
Wilkinson R Income inequality, social cohesion, and health: clarifying the theory: a reply to Muntaner and Lynch Int J Health Serv 1999 29 525 543 10450545 10.2190/3QXP-4N6T-N0QG-ECXP
Marmot M Wilkinson R Psychosocial and material pathways in the relation between income and health: a response to Lynch, et al Br Med J 2001 322 1233 1236 11358781
Casper ML Barnett E Williams GI Halverson JA Braham VE Greenlund KJ Atlas of Stroke Mortality: Racial, Ethnic, and Geographic Disparities in the United States 2003 Atlanta, GA: Department of Health and Human Services, Centers for Disease Control and Prevention
Mackay J Mensah G Atlas of Heart Disease and Stroke 2004 Geneva, World Health Organization
Olsen SF Martuzzi M Elliott P Cluster analysis and disease mapping-why, when, and how? A step-by-step guide Br Med J 1996 313 863 866 8870578
Wartenberg D Investigating disease clusters: why, when and how? Journal of the Royal Statistical Society A 2001 164 13 22 10.1111/1467-985X.00181
Han D Rogerson PA Khan O, Skinner R Application of a GIS-based statistical method to assess spatio-temporal changes in breast cancer clustering in the northeastern United States GIS and Health Applications 2002 New York, NY: Idea Group 114 125
Kulldorff M Feuer EJ Miller BA Freedman LS Breast cancer clusters in the United States: a geographic analysis Am J Epidemiol 1997 146 161 170 9230778
US Census Bureau Zip code tabulation area (ZCTA)
Krieger N Waterman P Chen JT Soobader MJ Subramanian SV Carson R Zip code caveat: bias due to spatiotemporal mismatches between zip codes and US census-defined geographic areas-the Public Health Disparities Geocoding Project Am J Public Health 2002 92 1100 1102 12084688
Pagano M Gauvreau K Principles of Biostatistics 1993 Belmont, CA: Duxbury Press
Rogerson PA Statistical method for the detection of geographic clustering Geogr Anal 2001 33 215 227
Kulldorff M Nagarwalla N Spatial disease clusters: detection and inference Stat Med 1995 14 799 810 7644860
|
16242043
|
PMC1276810
|
CC BY
|
2021-01-04 16:39:06
|
no
|
Int J Health Geogr. 2005 Oct 21; 4:25
|
utf-8
|
Int J Health Geogr
| 2,005 |
10.1186/1476-072X-4-25
|
oa_comm
|
==== Front
J Inflamm (Lond)Journal of Inflammation (London, England)1476-9255BioMed Central London 1476-9255-2-111624203210.1186/1476-9255-2-11ResearchEarly relief of osteoarthritis symptoms with a natural mineral supplement and a herbomineral combination: A randomized controlled trial [ISRCTN38432711] Miller Mark JS [email protected] Komal [email protected] Sameer [email protected] Vidyanand [email protected] Jayesh [email protected] Ramesh [email protected] Anil [email protected] Hemant [email protected] Himanshu [email protected] Paul [email protected] Jayesh [email protected] Center for Cardiovascular Sciences, Albany Medical College, Albany, New York, USA2 Vedic Lifesciences, Mumbai, India3 Ashok Raut Orthopedic Hospital, Virarwest, India4 Bhagwati Hospital, Mumbai, India5 K.J. Somaiya Medical College & Hospital, Mumbai, India6 Santerra Pharmaceuticals, LLC, Raleigh, North Carolina, USA2005 21 10 2005 2 11 11 7 3 2005 21 10 2005 Copyright © 2005 Miller et al; licensee BioMed Central Ltd.2005Miller et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
This study was designed to determine if a natural mineral supplement, sierrasil, alone and in combination with a cat's claw extract (Uncaria guianensis), vincaria, has therapeutic potential in mild to moderate osteoarthritis of the knee.
Methods
Patients (n = 107) with mild to moderate osteoarthritis of the knee were randomly assigned to one of 4 groups; high dose sierrasil (3 g/day), low dose sierrasil (2 g/day), low dose sierrasil (2 g/day) + cat's claw extract (100 mg/day) or placebo, administered for 8 weeks. Treatment was double blinded. Primary efficacy variables were WOMAC scores (A, B, C and total). Visual analog score (VAS) for pain, consumption of rescue medication (paracetamol), and tolerability were secondary variables. Safety measures included vital signs and laboratory-based assays.
Results
Ninety-one of the 107 patients successfully completed the protocol. All four groups showed improvement in WOMAC and VAS scores after 8 weeks (p < 0.001), in all 3 groups receiving sierrasil the magnitude of benefits were greater vs. placebo (WOMAC Total 38–43% vs. 27%) but this was not statistically significant. In reference to baseline values sierrasil treated groups had a considerably faster onset of benefits. Placebo-treated individuals failed to show significant benefits at 4 weeks (11% reduction in total WOMAC). In contrast, after 1 or 2 weeks of therapy all the sierrasil groups displayed significant reductions in WOMAC scores (p < 0.05) and at week 4 displayed a 38–43% improvement. VAS was significantly improved at 4 weeks in all groups (p < 0.001) but was significantly greater in all sierrasil groups compared to placebo (p < 0.05). Rescue medication use was 28-23% lower in the herbomineral combination and high dose sierrasil groups although not statistically different from placebo (P = 0.101 and P = 0.193, respectively). Tolerability was good for all groups, no serious adverse events were noted and safety parameters remained unchanged.
Conclusion
The natural mineral supplement, sierrasil alone and in combination with a cat's claw extract, improved joint health and function within 1–2 weeks of treatment but significant benefits over placebo were not sustained, possibly due to rescue medication masking. Sierrasil may offer an alternative therapy in subjects with joint pain and dysfunction.
==== Body
Background
Osteoarthritis is a painful and debilitating joint condition that affects hundreds of millions worldwide [1]. Despite the prevalence of the disease, current therapeutic options are not optimal. Pharmaceutical approaches to disease management include the popular non-steroidal anti-inflammatory class (NSAIDs) which block cyclo-oxygenase (COX). NSAIDS provide symptomatic relief [2] but do not abrogate the underlying disease process. Moreover, therapy has recently been dominated by the COX-2 specific class, which was designed to reduce unwarranted renal and gastrointestinal side-effects associated with non-specific COX inhibitors [3]. However, recent studies have revealed heightened cardiovascular risks [4,5], which may limit the application of the COX-2 specific class of NSAIDs.
Use of complimentary medicine as an alternative therapeutic approach is common in conditions associated with pain and discomfort and especially when local traditions support such approaches. Examples of complimentary medicines that have been proposed to offer benefits for osteoarthritis include acupuncture [6], nutraceuticals [7-9] and botanicals [10,11]. Additionally, various forms of physical therapy offer established non-pharmacologic benefits [12].
Of the nutraceutical approaches one of the best known and widely used is glucosamine (alone or in combination with chondroitin). Glucosamine and chondroitin are structural elements of cartilage and matrix. The therapeutic approach centers on the assumption that ingestion of large amounts of these matrix elements will assist in the replacement of the comparable material that is lost as a result of the catabolic process associated with inflammation. Recently, it has been suggested that the absorption of oral glucosamine is not sufficient for the chondroitin production and cartilage deposition [13]. There is limited evidence for a direct anti-inflammatory action of glucosamine [14]. Nevertheless by design it is not an approach that influences the disease process directly, but rather is thought to maintain cartilage architecture in the face of ongoing catabolic pathways. Hence, not surprisingly, the onset of action in those individuals for whom it does provide relief is commonly on the order of months after treatment initiation [15-17]. Studies as to the efficacy of glucosamine and chondroitin have produced variable results [18-21] suggesting that the benefits of this approach may have limitations.
Botanicals, especially those with redox-based actions are promising in the treatment of chronic inflammation because of their inherent disease modifying characteristics. Green tea catechins, especially epigallocatechin gallate (EGCG), have been shown to limit human cartilage degradation in vitro [22,23] and maintain joint architecture in an animal model [24]. This anti-inflammatory action is thought to be the result of inhibition of transcriptional events, particularly prevention of NF-κB activation by cytokines and oxidants. NF-κB is a critical transcription factor in chronic inflammation and is a desirable target for new therapeutics, including pharmaceutical development, as it regulates numerous genes that contribute to the inflammatory process [25-27]. Of particular note for joints NF-κB regulates the production of matrix metalloproteases (MMPs) by chondrocytes [28]. During inflammation or injury chondrocytes release MMPs, which in turn degrade the cartilage matrix, releasing glycosaminoglycans and eventually, glucosamine. Haqqi and colleagues have elegantly demonstrated how EGCG can prevent MMP formation and cartilage degradation in explants of human cartilage stimulated by the pro-inflammatory cytokine, IL-1β [22] and reviewed the use of botanicals, in general, as arthritis therapeutics [11].
Sierrasil® is a natural mineral product derived from the Sierra Mountains in the USA that has a cultural history of use to treat joint pain. We have recently demonstrated in the Haqqi model of human cartilage explants, that extracts of sierrasil reduced cartilage degradation in response to IL-1β, as well as nitric oxide production secondary to the induction of inducible nitric oxide synthase [29,30]. Increased chondrocyte production of nitric oxide is associated with catabolic activities and inhibitors of inducible nitric oxide synthase display anti-inflammatory properties [23,30-35]. While an action of sierrasil was not directly assessed on transcriptional events nevertheless, that is a likely target based on its ability to suppress IL-1β mediated events in human cartilage.
In the human cartilage explant study with sierrasil, co-administration of an extract of the botanical cat's claw (Uncaria guianensis, Vincaria®) complimented these actions. Indeed, this Uncaria guianensis extract is a remarkably potent inhibitor of NF-κB activity and tumor necrosis factor (TNFα) production [36-38], with clear cytoprotective actions [39] and has already successfully completed a placebo controlled trial for osteoarthritis of the knee [10]. Cat's claw is a vine indigenous to the Amazon Rainforest, that has a long history of use for joint pain and chronic inflammation [40].
Given that sierrasil was able to limit human cartilage degradation induced by IL-1β in vitro [29], it was determined that further study in human osteoarthritis was warranted. Further, it was postulated that based on the acute in vitro observations on cartilage protection and suppression of inflammatory events sierrasil may result in a clinical response that was expedient. Given the complimentary actions in vitro with cat's claw, a combination therapy was also evaluated. The goal was to determine if these approaches offered benefits in a safe manner, over the course of two months of treatment with a focus on rapid induction of benefits.
Methods
Research Design
This randomized, double-blind, placebo-controlled multi-center trial was approved by the Institutional Ethics Committee of K.J. Somaiya Medical College & Hospital, Mumbai, India and was in compliance with the Helsinki Declaration.
Participants
Subjects were recruited from three centers – two private arthritis clinics and the K.J. Somaiya Medical College & Hospital, in Mumbai, India. Inclusion criteria were ambulatory, adult patients of either sex and greater than 20 years of age. Mild to moderate osteoarthritis was determined by radiological examination and ARA functional class II or III, and Kellgren Lawrence classification for knee osteoarthritis grade II or Grade III, and a baseline functional assessment of overall pain of at least 50 mm on a 100 mm Visual Analog Scale.
Exclusion criteria were osteoarthritis of grade I or IV (Kellgren Lawrence or ARA functional class), existence of other forms of arthritis, arthroscopy of either knee within the past year, administration of intra-articular steroids within the past 3 months or hyaluronic acid in the last 9 months, pregnancy or lactating women or women not taking adequate contraceptive measures, presence of any concomitant unstable disease or abnormality of any clinically relevant laboratory test, evidence of severe renal or hematologic disease or severe cardiac insufficiency, moderate to severe neuropathy, and unwillingness to come to regular follow-up visits for the length of the study.
A Fixed Allocation Randomization procedure was used to assign interventions to the participants with a pre-specified probability. Randomization was done in blocks of four, related to the number of treatment groups, and the total possibilities were 24. Treatment codes were held in a sealed manner by investigators in case of a serious adverse event. The allocation sequence was generated by the consulting statistician and subjects were assigned treatment according to this algorithm until subject 96 (total possibilities) and then the same cycle was repeated. Allocation of treatment according to this method was assigned by authors (SK, KM) along with monitoring of trial supplies, blindness, and adverse event monitoring. Research coordinators however, were not involved in any trial related activities to avoid bias.
For ease of presentation the 4 subject groups are given the following descriptors:
Group A – High dose sierrasil
Group B – Low dose sierrasil
Group C – Low dose sierrasil plus cat's claw extract
Group D – Placebo
A total of 107 subjects were recruited based on a planned recruitment of 25 subjects per group (total of 100). This recruiting estimate was based on results obtained in a trial which demonstrated efficacy with the cat's claw extract alone [10].
The sequence of the study involved an assessment of vital signs, radiological assessment of the affected knee and laboratory tests one week before baseline. Laboratory tests included: complete blood count, erythrocyte sedimentation rate, SGPT, serum creatinine, urine pregnancy test. At baseline the following were recorded – vital signs, WOMAC score (A, B, C and total) and VAS. This was repeated at weeks 1, 2, 4, 6, and 8 after treatment initiation along with monitoring of compliance and adverse event monitoring. At the conclusion of the study at week 8, laboratory tests were repeated.
Patients were considered drop-outs from the study if they went more than 4 days without medication or failed to report within the fifth day of a scheduled visit. Protocol deviation was characterized by not attending a scheduled visit by more than one day, skipping medication for an entire day, consuming other medications without consulting the investigator and not bringing the rescue medication or study medication bottles at the time of a scheduled visit.
Patients were not allowed to consume any NSAIDs, analgesics, cartilage supplements, calcium supplements, steroids, or other agents that may affect the outcomes of the study other than the rescue medication. Any medication taken by subjects for two months prior to the inclusion of the study, and whose intake was stabilized, was permitted and monitoring that dosing of these medications was not changed for the duration of the investigation.
Treatments
The duration of treatment was 8 weeks, administered as 2 capsules twice a day taken orally with meals. The rescue medication was paracetamol (acetaminophen) as a single tablet of 500 mg, at a dosage that was not to exceed 4 tablets a day.
Sierrasil® is a 100% natural composite, yet novel composite mineral (SM317) containing silicate minerals of calcium, magnesium, potassium, sodium and aluminum, among others. Sierrasil contains primarily 45.0% SiO2 (silica), 9.5% aluminum (as aluminum silicate), 5.9% iron (in several mineral forms), 1.3% calcium (in various forms), and other trace elements (Table 1). Sierrasil alone was administered in two treatment groups. The high dose group received a total of 3 g/day, and the low dose group 2 g/day. In addition, there was a separate treatment group where low dose sierrasil (2 g/day) was combined with a cat's claw extract, vincaria® (Uncaria guianensis) at a dose of 100 mg/day. The relative components of major mineral sources in sierrasil are noted in Table 1, along with the expected daily intake at the doses of 2 and 3 grams.
Table 1 Mineral composition of Sierrasil and the expected intake at the low (2 g/day) and high (3 g/day) doses. Mineral bioavailability in the various forms has not been established but data on acid liberalization suggest that for some minerals the dose available for absorption may be as low as < 0.1%.
Metal assays (ICP-MS) Total Intake at 2 g/day dose Total Intake at 3 g/day dose
Aluminum 188 mg/day 282 mg/day
Antimony < 0.006 mg/day < 0.009 mg/day
Arsenic 0.026 mg/day 0.035 mg/day
Barium 2.0 mg/day 2.9 mg/day
Beryllium < 0.002 mg/day < 0.003 mg/day
Bismuth < 0.06 mg/day < 0.09 mg/day
Cadmium 0.0036 mg/day 0.0054 mg/day
Calcium 26 mg/day 40 mg/day
Carbon 0.60 mg/day 0.90 mg/day
Chromium 0.040 mg/day 0.060 mg/day
Chromium (VI) < 0.00048 mg/day < 0.00072 mg/day
Cobalt 0.017 mg/day 0.026 mg/day
Copper 0.074 mg/day 0.11 mg/day
Iron 119 mg/day 178 mg/day
Lead 0.013 mg/day 0.020 mg/day
Lithium 0.02 mg/day 0.03 mg/day
Magnesium 11 mg/day 17 mg/day
Manganese 0.29 mg/day 0.44 mg/day
Mercury 0.0012 mg/day 0.0017 mg/day
Molybdenum 0.002 mg/day 0.003 mg/day
Nickel 0.027 mg/day 0.040 mg/day
Phosphorous 3.8 mg/day 5.7 mg/day
Potassium 20 mg/day 30 mg/day
Selenium 0.0030 mg/day 0.0045 mg/day
Silica 899 mg/day 1349 mg/day
Silver 0.015 mg/day 0.022 mg/day
Sodium < 12 mg/day < 18 mg/day
Strontium 2.1 mg/day 3.2 mg/day
Sulphur 30 mg/day 45 mg/day
Thorium < 0.010 mg/day < 0.015 mg/day
Tin < 0.02 mg/day < 0.03 mg/day
Titanium 10 mg/day 15 mg/day
Uranium < 0.12 mg/day < 0.18 mg/day
Vanadium 0.34 mg/day 0.51 mg/day
Zinc 0.082 mg/day 0.12 mg/day
Zirconium 0.23 mg/day 0.35 mg/day
Cat's claw is an Amazonian vine whose bark has a long history of ethnomedical use for treating inflammation [38,40]. Commonly two species are used Uncaria tomentosa and Uncaria guianensis, and the Vincaria® extract uses the Uncaria guianensis species, which was chosen as it is has been shown to be more potent as an anti-inflammatory agent [38], and is the only source of cat's claw with documented benefits in osteoarthritis [10]. The vincaria extract, which is a water based extraction, is devoid of oxindole alkaloids [38], present in Uncaria tomentosa that are thought to be immune enhancing [41,42], and therefore counter-productive for this application.
Sierrasil (SM317) was obtained from the manufacturer, Sierra Mountain Minerals, Inc. Bozeman, MT, USA and Vincaria cat's claw (RN180) was obtained from Rainforest Nutritionals, Inc., Raleigh, NC, USA . Sierrasil and vincaria are produced under GMP conditions. Vincaria is standardized to a maximum level of oxindole alkaloids of 0.3 mg/g and an IC50 for TNF αinhibition of at least 1 ug/ml. Lactose was used as the placebo and as a filler for treatment group capsules to ensure that all treatments were uniform in size and color. Uniformity was maintained for all four treatment groups in terms of bottle filling, labeling, and packaging. Treatments were packaged in gelatin capsules and packed in wide mouthed white opaque plastic bottles with screw caps in a clean room. Investigators were provided with blinding chits having patient codes along with their treatment group (alphabetical code). In the case of a serious adverse event investigators were instructed to inform the monitors and then only unblind the treatment group of the subject in order to address needful treatment.
Primary Efficacy Variable
The Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index is a disease-specific self-administered, health status measure that is widely accepted as reflective of osteoarthritis disease activity. The original index consists of 24 questions (5 pain, 2 stiffness and 17 physical functions). Individual question responses are assigned a score of between 0 (none) to 4 (extreme) and summed to form a score ranging from 0 (best) to 96 (worst). There are three sections to the WOMAC Score. Section A deals with the amount of pain (5 questions). Section B addresses the amount of joint stiffness (2 questions). Section C addresses aspects of physical function (17 questions).
In section C of the WOMAC Score 5 questions were addedfor a total of 22 questions. This score was then normalized to produce a total WOMAC score in a range of 0–100. Details of all questions included, per section, are found in Table 2. The addition of questions to the Section C component of WOMAC, which primarily addresses physical function, may be skewed the WOMAC total score away from pain and stiffness. Nevertheless, the pain and stiffness components were assessed individually, as well as collectively, to ensure that specific benefits could be ascertained as well as a modified global assessment.
Table 2 Modified WOMAC Questionnaire. Subjects responded with the following numerical assessment: 0 = none; 1 = slight; 2 = moderate; 3 = severe; 4 = extreme
WOMAC A: Pain on
- Walking
- Stair climbing
- Nocturnal
- Rest
- Weight bearing
WOMAC B: Stiffness
- In morning
- Stiffness occurring during the day
WOMAC C: Level of difficulty performing the following functions:
- Descending stairs
- Ascending stairs
- Rising from sitting
- Standing
- Bending to the floor
- Walking on flat
- Getting in/out of a car
- Going shopping
- Putting on socks-Rising from bed
- Taking off socks
- Lying in bed
- Getting in/out of bath
- Sitting
- Getting on/off toilet
- Heavy domestic duties
- Light domestic duties
- While working
- Sitting cross legged
- While cycling
- While driving vehicle
- While praying.
Secondary Efficacy Variables
VAS – Visual Analog Score uses a 100 mm linear measure of pain status, with 0 representing no pain and 100 being unbearable pain. Patients marked on the linear scale the relevant amount of pain they were experiencing and the value was noted by the investigator.
Recovery, as assessed by investigator and physician, was characterized by 5 categories: Excellent – complete relief of symptoms; Good – partial relief of symptoms; Fair – minimal relief of symptoms; poor – no relief of symptoms; very poor – worsening of symptoms.
Tolerability was assessed by 3 categories. Good – no side effects; Fair – mild to moderate side effects; poor – severe side effects and withdrawal of therapy. Measurements of recovery and tolerability were performed at the end of the protocol (week 8).
Use of rescue medication, paracetamol (acetominophen) was addressed as a measure of both pain management and efficacy. The amount of rescue medication was only assessed in terms of total use at the conclusion of the study period. Rescue medication use was not assessed sequentially along with other variables.
Data Quality Assurance
All investigators were informed of ICH-GCP guidelines, the quality of data and study execution was monitored by individuals independent of subject contact and treatment assessment.
Statistical Analysis
Data was analyzed by an independent statistician using the following tests – ANOVA, paired and unpaired t tests, Bonferroni, Chi Square, Friedman and Wilcoxan tests as appropriate. The following software was used: SPSS 11.5, PEPI, EPI INFO 2000 and MS Excel. Statistical significance was taken at the 95% level (p < 0.05). Results are expressed as the mean ± SEM.
Results
Patient Randomization
Patient randomization was effective. Subject age was comparable in all groups (Table 3). Gender was predominantly female 73/107 (68.2%) vs. males 34/107 (31.8%). At entry 70% of subjects had ARAF-Class II and 30% ARAF-Class III. A similar distribution was observed for the Kellgren Lawrence Criteria for diagnosis, with 79% Grade 2 and 21% Grade 3 indicating that the majority of subjects had mild osteoarthritis of the knee. When the four treatment groups were compared there was no statistical difference in disease activity on entry – for ARAF-Class, Chi square = 0.586, p = 0.900; Kellgren Lawrence Criteria, Chi square = 0.690, p = 0.876.
Table 3 Baseline characteristics of treatment groups
Placebo Group A Group B Group C
N completed (started) 23 (29) 20 (25) 22 (24) 26 (29)
Age (years) 51.3 ± 1.0 50.5 ± 1.7 52.3 ± 1.8 52.1 ± 1.6
Gender (F: M %) 72:28 76:24 75:25 52:48
ARA Functional Class II/III, % 69 : 31 76 : 24 67 : 33 69 : 31
Kellgren Lawrence Criteria Grade 2/3, % 76 : 24 83 : 17 83 : 17 82 : 18
WOMAC A 8.7 ± 1.0 9.1 ± 1.2 8.0 ± 1.1 8.7 ± 1.0
WOMAC B 3.4 ± 0.5 3.5 ± 0.5 3.3 ± 0.4 3.3 ± 0.4
WOMAC C 36.4 ± 3.9 38.4 ± 4.0 34.7 ± 4.2 32.8 ± 4.3
WOMAC Total 48.6 ± 5.2 51.1 ± 6.4 46.4 ± 5.6 44.8 ± 5.5
VAS 72.9 ± 2.0 72.9 ± 3.0 70.1 ± 2.5 70.7 ± 2.4
Data expressed as a % of total or mean ± sem. Group A = High dose sierrasil, Group B = Low dose sierrasil, Group C = Low dose sierrasil + cat's claw extract. None of the entry assessments were significantly different amongst the four randomized groups.
Primary Efficacy Variable – WOMAC
Baseline disease activity was comparable in all 4 treatment groups (Table 3). While all four groups displayed an improvement from baseline for WOMAC values (subsets and total) over the course of 8 weeks of treatment (p < 0.001), including placebo, the magnitude of these benefits were greater in all 3 sierrasil groups (A, B and C). It is of note that after 4 weeks of treatment the placebo treated group (D) was not significantly different from baseline for any of the WOMAC scores, yet for groups A, B and C there was a marked improvement over baseline (p < 0.001). Indeed, statistically significant benefits were evident with one week of treatment for many of the tests in sierrasil treated subjects (Tables 4, 5, 6, 7).
Table 4 Sequential WOMAC A Scores – Pain
Group Baseline Week 1 Week 2 Week 4 Week 6 Week 8
Placebo 8.7 ± 1.0 8.9 ± 1.1 8.5 ± 1.2 8.4 ± 1.2 6.8 ± 1.1 * 6.3 ± 1.2 *
A 9.1 ± 1.2 8.1 ± 1.2 * 7.5 ± 1.2 ** 6.5 ± 1.2 ** 6.4 ± 1.2 ** 6.3 ± 1.1 **
B 8.0 ± 1.1 7.5 ± 1.1 6.9 ± 1.1 ** 5.7 ± 1.0 ** 5.6 ± 1.1 ** 5.0 ± 0.9 **
C 8.7 ± 1.0 8.0 ± 1.0b 6.9 ± 0.9 *** 5.6 ± 0.8 *** 5.0 ± 0.8 *** 4.3 ± 0.7 ***
Data expressed as mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. All groups displayed a significant change over the 8 weeks of the study (p < 0.001, ANOVA). Using the Wilcoxan Matched-Pairs Signed-Rank test the following descriptors reflect statistical significance from baseline; * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 5 Sequential WOMAC B Scores – Joint Stiffness
Group Baseline Week 1 Week 2 Week 4 Week 6 Week 8
Placebo 3.4 ± 0.5 3.6 ± 0.5 3.1 ± 0.5 3.0 ± 0.5 2.6 ± 0.5 * 2.4 ± 0.5 *
A 3.5 ± 0.5 3.0 ± 0.5 * 2.9 ± 0.6 * 2.4 ± 0.5 ** 2.4 ± 0.5 ** 2.3 ± 0.5 **
B 3.3 ± 0.4 2.9 ± 0.5 * 2.6 ± 0.5 ** 2.1 ± 0.4 ** 2.2 ± 0.5 ** 1.9 ± 0.3 **
C 3.3 ± 0.4 3.2 ± 0.4 3.0 ± 0.4 2.1 ± 0.3 *** 2.0 ± 0.3 *** 1.6 ± 0.4 ***
Data expressed as mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. All groups displayed a significant change over the 8 weeks of the study (p < 0.001, ANOVA). Using the Wilcoxan Matched-Pairs Signed-Rank test the following descriptors reflect statistical significance from baseline; * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 6 Sequential WOMAC C Scores – Physical Function
Group Baseline Week 1 Week 2 Week 4 Week 6 Week 8
Placebo 36.4 ± 3.9 37.0 ± 4.3 35.0 ± 4.4 32.0 ± 4.3 29.5 ± 4.5 26.6 ± 4.6 **
A 38.4 ± 4.9 32.4 ± 5.3 31.0 ± 5.1 ** 26.1 ± 4.6 *** 26.3 ± 4.6 *** 23.1 ± 4.6 ***
B 34.7 ± 4.2 31.3 ± 4.4 * 28.7 ± 4.4 ** 24.5 ± 3.8 *** 22.8 ± 3.9 *** 21.3 ± 2.9 ***
C 32.8 ± 4.3 31.1 ± 4.0 29.1 ± 4.0 * 24.3 ± 3.4 *** 21.7 ± 3.5 *** 19.7 ± 3.4 ***
Data expressed as mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. All groups displayed a significant change over the 8 weeks of the study (p < 0.001, ANOVA). Using the Wilcoxan Matched-Pairs Signed-Rank test the following descriptors reflect statistical significance from baseline; * p < 0.05, ** p < 0.01, *** p < 0.001.
Table 7 Sequential WOMAC Total – Summary
Group Baseline Week 1 Week 2 Week 4 Week 6 Week 8
Placebo 48.6 ± 5.2 49.0 ± 5.8 46.6 ± 5.9 43.5 ± 5.8 38.8 ± 5.9 * 35.3 ± 6.3 **
A 51.5 ± 6.4 43.5 ± 6.9 41.4 ± 6.7 ** 35.0 ± 6.2 *** 36.2 ± 6.2 *** 31.8 ± 6.1 ***
B 46.4 ± 5.6 42.0 ± 5.9 * 38.6 ± 6.0 *** 32.6 ± 5.2 *** 31.5 ± 5.3 *** 28.5 ± 3.9 ***
C 44.8 ± 5.5 42.3 ± 5.2 39.0 ± 5.1 ** 32.0 ± 4.4 *** 28.7 ± 4.5 *** 25.6 ± 4.4 ***
Data expressed as mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. All groups displayed a significant change over the 8 weeks of the study (p < 0.001, ANOVA). Using the Wilcoxan Matched-Pairs Signed-Rank test the following descriptors reflect statistical significance from baseline; * p < 0.05, ** p < 0.01, ** p < 0.001. Note the p values for groups A and C approached significance at week one (p = 0.051 and 0.053).
Specifically, for group A (high dose) WSA, WSB and WST were significantly improved at week 1 (p < 0.05). At week 2 for group A, all WOMAC scores were significantly improved (p < 0.05). For group B (low dose), WSB, WSC, and WST were significantly improved at week 1 (p < 0.05). After 2 weeks of therapy in group B, all WOMAC scores were significantly improved (p < 0.01). For group C (low dose sierrasil + cat's claw extract) WSA was significantly improved from baseline (p < 0.01) and by week 2 all WOMAC scores except WSB were improved (p < 0.05).
These early benefits are readily apparent in Table 4, 5, 6, and Figures 1, 2, 3, 4 which displays the percentage changes from baseline for WOMAC A, WOMAC B, WOMAC C, and WOMAC total respectively. However, when the test groups were compared to placebo, the magnitude of these differences were not significant for the majority of time points (Friedman test).
Figure 1 Sequential changes in WOMAC A (pain) scores expressed as a percentage of baseline values. Placebo (blue, n = 22), high dose sierrasil (green, n = 20), low dose sierrasil (red, n = 21) and low dose sierrasil + cat's claw extract (orange, n = 25) all demonstrated a time dependent improvement in the pain indices of WOMAC A. However, in the placebo group this was delayed until the last month of the study. All sierrasil treated groups displayed a faster onset of action.
Figure 2 Sequential changes in WOMAC B (stiffness) scores expressed as a percentage of baseline values. Placebo (blue, n = 22), high dose sierrasil (green, n = 20), low dose sierrasil (red, n = 21) and low dose sierrasil + cat's claw extract (orange, n = 25) displayed a time dependent improvement in WOMAC B scores, measuring stiffness, over the course of the study. A trend for a faster onset of improvement was evident in all sierrasil treated groups when compared to placebo controls.
Figure 3 Sequential changes in WOMAC C (physical activity) scores expressed as a percentage of baseline values. Placebo (blue, n = 22), high dose sierrasil (green, n = 20), low dose sierrasil (red, n = 21) and low dose sierrasil + cat's claw extract (orange, n = 25) displayed a time dependent improvement in physical activity and function scores (WOMAC C). A trend for a faster onset of benefits was evident in all sierrasil treated groups versus placebo.
Figure 4 Sequential changes in WOMAC total scores expressed as a percentage of baseline values. Placebo (blue, n = 22), high dose sierrasil (green, n = 20), low dose sierrasil (red, n = 21) and low dose sierrasil + cat's claw extract (orange, n = 25) displayed a time dependent improvement in total WOMAC Scores. There was a trend for a faster onset of action in all sierrasil treated groups when compared to placebo responses.
Secondary Efficacy Variables – VAS
The VAS, a pain assessment, was significantly improved in all 4 treatment groups at the conclusion of the study (week 8, p < 0.001), Table 8. The placebo group (D) also displayed a significant improvement in VAS at 4 weeks as did groups A-C (P < 0.01) although, the magnitude of VAS reductions in these sierrasil groups (A-C) at week 4 were significantly greater than placebo (p < 0.05). However, the greater response in groups A-C vs. placebo, was not statistically significant at week 8 or other time points. Changes in VAS expressed as a percentage of change from baseline are shown in Figure 5.
Table 8 Sequential VAS – Pain Index
Group Baseline Week 1 Week 2 Week 4 Week 6 Week 8
Placebo 72.9 ± 2.1 71.4 ± 2.7 68.4 ± 3.2 63.9 ± 3.3 ** 58.6 ± 3.9 ** 52.0 ± 4.7 ***
A 72.9 ± 3.0 66.8 ± 3.3 62.5 ± 4.7 ** 51.2 ± 5.0 *** 49.7 ± 5.1 *** 42.7 ± 5.2 ***
B 70.1 ± 2.5 65.8 ± 3.2 58.9 ± 3.8 ** 52.0 ± 4.7*** 51.9 ± 5.0 *** 44.6 ± 4.9 ***
C 70.7 ± 2.4 65.0 ± 2.6 60.7 ± 2.9 ** 52.7 ± 2.5 *** 47.2 ± 3.4 *** 43.4 ± 3.8 ***
Data expressed as mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. All groups displayed a significant change over the 8 weeks of the study (p < 0.001, ANOVA). Using a Wilcoxan matched pairs Signed-Rank test from baseline; * p < 0.05, ** p < 0.01, *** p < 0.001.
Figure 5 Sequential changes in VAS (pain) expressed as a percentage of baseline values. Placebo (blue, n = 22), high dose sierrasil (green, n = 20), low dose sierrasil (red, n = 21) and low dose sierrasil + cat's claw extract (orange, n = 25) displayed a time dependent improvement in VAS scores for pain. However, there was a trend for a faster onset of benefits in the sierrasil treated groups compared to placebo.
Secondary Efficacy Variable – Rescue Medication
Paracetamol was used as a rescue medication with dosing limited to 4 × 500 mg per day. There were no significant differences in the use of paracetamol in the 4 treatment groups (Figure 6), although some trends are clear. In group A, there was a 23% lower use of paracetamol (p = 0.193 v placebo). For group C there was a 28% reduction in paracetamol use, which approached significance when compared to placebo (p = 0.101) or low dose sierrasil alone (p < 0.055). The consumption of rescue medication was not assessed at sequential timepoints. Furthermore, the trend for greater consumption of rescue medication in the placebo and low dose sierrasil groups may have masked the ability to determine significant differences between groups in terms of the magnitude of pain and discomfort relief.
Figure 6 Consumption of rescue medication (paracetamol) for the study duration. Results for placebo (blue column, n = 19), high dose sierrasil (green column, n = 19), low dose sierrasil (red column, n = 20) and low dose sierrasil + cat's claw extract (orange column, n = 23) groups are expressed as mean ± sem. There was no significant difference in rescue medication use between the various groups, although a trend for less paracetamol consumption was evident in both the high dose sierrasil and low dose + cat's claw groups which approached significance (p = 0.119 and p = 0.101 vs placebo respectively).
Secondary Efficacy Variables – Global Assessments and Tolerance
Two global assessments to treatment were determined – patient self assessment and the investigator's assessment. For both perspectives 49% responded that the response was good to excellent, however there was no statistical difference between treatment groups Patient's response: Chi square = 8.118, p = 0.776.
Investigator's response: Chi square = 4.336, p = 0.888
Tolerance was reported as "good" by 94/95 subjects. The one subject that reported a poor tolerance received high dose sierrasil (group A).
Safety Variables – Laboratory
All laboratory tests were unchanged from baseline to week 8 for all groups with the following exceptions.
(1) Group A hemoglobin was increased from 12.0 ± 0.2 to 12.5 ± 0.3, p < 0.05
(2) Group B lymphocyte levels increased from 30.3 ± 1.8 to 34.3 ± 1.8, p < 0.05
These changes, while statistically significant, were not considered as indicative of an adverse response based on values observed in the other groups. A summary of these safety variables is depicted in Table 9.
Table 9 Laboratory-based evaluations of safety
Test Placebo Wk. 0 Placebo Wk. 8 A Wk. 0 A Wk. 8 B Wk. 0 B Wk. 8 C Wk. 0 C Wk. 8
Neu 62.0 ± 1.7 64.2 ± 2.1 59.3 ± 2.4 58.6 ± 2.3 64.7 ± 1.6 61.5 ± 1.6 61.6 ± 1.5 63.0 ± 1.7
Bas 0 0 0 0 0 0 0 0.3 ± 0.3
Eos 3.6 ± 0.8 3.6 ± 0.6 4.2 ± 0.5 3.7 ± 0.9 4.0 ± 0.6 4.8 ± 1.5 3.9 ± 0.5 4.4 ± 0.5
Lym 34.0 ± 2.0 31.7 ± 2.5 35.7 ± 2.5 35.2 ± 1.8 30.3 ± 1.8 34.3 ± 1.8* 33.4 ± 1.7 31.8 ± 1.8
Mon 0.8 ± 0.3 0.6 ± 0.2 1.0 ± 0.2 0.8 ± 0.3 0.9 ± 0.2 0.7 ± 0.2 1.1 ± 0.2 0.7 ± 0.2
WBC 8052 ± 417 8260 ± 499 7994 ± 514 8037 ± 655 8750 ± 476 7857 ± 478 8160 ± 308 7815 ± 459
RBC 3.98 ± 0.09 4.11 ± 0.11 4.26 ± 0.09 4.41 ± 0.12 4.33 ± 0.11 4.31 ± 0.10 4.39 ± 0.12 4.30 ± 0.10
HB 12.0 ± 0.2 12.4 ± 0.4 12.0 ± 0.2 12.5 ± 0.3* 12.2 ± 0.3 12.1 ± 0.3 12.3 ± .2 12.2 ± 0.3
ESR 29.0 ± 3.6 32.4 ± 6.0 35.0 ± 3.5 30.5 ± 4.8 31.7 ± 4.2 40.7 ± 6.2 26.5 ± 3.6 22.2 ± 3.2
SGPT 28.8 ± 1.9 27.7 ± 2.4 23.6 ± 1.9 29.3 ± 4.5 24.5 ± 1.6 25.6 ± 1.7 24.9 ± 1.1 26.4 ± 3.3
CRE 0.98 ± 0.04 0.97 ± 0.05 1.4 ± 0.4 0.9 ± 0.04 1.7 ± 0.7 1.0 ± 0.03 1.67 ± 0.69 1.01 ± 0.05
Data expressed as a mean ± sem. Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22.
Neu = neutrophils %, Bas = basophils %, Eos = eosinophils %, Lym = lymphocytes %, Mon = monocytes %, WBC = white blood cells per mm3, RBC = red blood cells 106 × mm3, HB = hemoglobin gm/dl, ESR = erythrocyte sedimentation rate mm, SGPT = IU/L, CRE = creatinine mg/dl. Using unpaired t test, the following descriptors describe a significant difference from baseline, * p < 0.05.
Safety Variables – Vital Signs
Blood pressure (systolic and diastolic), respiration rate and pulse rate were vital signs that were monitored. Using ANOVA there was no change in these variables from screening for the study, at the 5 intermediate evaluation points and the study's conclusion at 8 weeks, with the following exceptions.
(1) Pulse rate declined in Group B (p < 0.05) from a screening value of 81.4 ± 1.5 to 76.2 ± 1.2 beats per minute at week 8.
(2) Respiration rate declined in Group C (p < 0.05) from a screening value of 19.3 ± 0.6 to 18.9 ± 0.4 breaths per minute
Data for baseline vital sign values and again at week 8 are depicted in Table 10. Data for screening and the intervening assessments at weeks 1, 2, 4 and 6 are not included for simplicity.
Table 10 Vital Signs
Vital Sign Placebo Wk 0 Placebo Wk 8 A Wk. 0 A Wk. 8 B Wk. 0 B Wk. 8 C Wk. 0 C Wk. 8
Pulse Rate 80.3 ± 1.3 76.9 ± 1.1 79.1 ± 1.3 74.7 ± 0.9 80.7 ± 1.3 76.1 ± 1.2 78.6 ± 0.9 76.6 ± 0.9
Systolic BP 128 ± 1 130 ± 2 129 ± 2 130 ± 3 132 ± 2 129 ± 3 * 129 ± 1 128 ± 2
Diastolic BP 83 ± 1 86 ± 1 82 ± 1 82 ± 1 83 ± 1 82 ± 1 80 ± 1 80 ± 2
Respiration Rate 19.5 ± 0.7 19.9 ± 0.7 19.2 ± 0.6 18.9 ± 0.4 19.7 ± 0.7 19.4 ± 0.8 20.2 ± 0.5 18.5 ± 0.5 *
Data expressed as a mean ± sem. Units of measure were: Pulse rate (beats per min), systolic and diastolic blood pressure (mm Hg), respiration rate (breaths per minute). Group A = High dose sierrasil (n = 20), Group B = Low dose sierrasil (n = 21), Group C = Low dose sierrasil + cat's claw extract (n = 25), placebo n = 22. Pulse rate and respiration rate are given as the rate per minute and blood pressure is expressed in mm Hg. Using unpaired t test, the following descriptors describe a significant difference from baseline, *p < 0.05.
Discussion
The purpose of this study was to determine if the natural mineral supplement, sierrasil, would relieve the symptoms of mild to moderate osteoarthritis of the knee in a safe manner. The approach included two doses of the mineral supplement to encompass the anecdotal clinical experience, and to evaluate the inclusion of a botanical extract, cat's claw (vincaria), which had previously been shown to be effective in treating osteoarthritis [10]. Using a randomized, double-blind placebo-controlled multi-center design, it is clear that this mineral supplement is indeed safe. Treatments were efficacious, particularly compared to baseline conditions, but there were clear difficulties in determining a sustained disassociation from placebo. In all sierrasil treated groups there was a significantly faster onset of benefits from initial values (evident from week 1 to 2) compared to placebo (first evident at week 6) but at the conclusion of the study differences between groups was not significant.
While it is of interest that the sierrasil that provided early relief of symptoms the inability to establish sustained significant differences from placebo poses limitations on interpretation. In part this dilemma is the result of the small study group size in this preliminary clinical evaluation. Additionally, an unexpected sharp improvement in primary and secondary assessments in the placebo group at weeks 6 & 8 contributed to the study's limitations. While clearly not in the instructions, subjects may have had expectations that all potential treatments in the randomized protocol would provide benefits and this may account for the placebo response. However, if this were the case one may expect that this placebo effect would be continuous as opposed to an exaggerated response that was observed in the last month of a two month study.
Another important consideration is a potential masking effect of rescue medication use. Rescue medication use was greater in placebo and low dose sierrasil groups, and this may have masked differences between the positive benefits related to treatment and placebo. Total consumption of rescue medication was determined and so it is not possible to link changes in rescue medication to perceived changes in disease activity on a weekly or monthly basis.
With the sierrasil test groups significant reductions in the baseline values of the primary efficacy variable, WOMAC, were evident as early as week 1 with steady improvements with continued administration (Figs. 1, 2, 3, 4). In contrast, placebo treated subjects did not report significant benefits from baseline until week 6. This early onset of benefits is not inconsistent with the in vitro studies demonstrating the protection of human cartilage degradation induced by IL-1β, which was prevented by acute exposure to sierrasil [25]. These human cartilage explant studies also demonstrated that the activation of nitric oxide production, a catabolic pathway [30-32], was attenuated by sierrasil. However, the present study does not directly assess whether protection of against cartilage degradation was associated with the therapies, nor is it likely that a substantial change in joint architecture would occur in this timeframe.
Of note, co-administration of the Uncaria guianensis extract, vincaria, was also chondroprotective in vitro [29] and associated with a rapid onset of benefits in osteoarthritis as noted in a separate double blind placebo controlled study [10]. Cat's claw has considerable data demonstrating that it is an effective inhibitor of transcription via NF-κB [36] and this formulation is a quite potent inhibitor of tumor necrosis factor [37,38]. Thus, both sierrasil and vincaria have the potential to act as disease modifying agents in osteoarthritis although only safety and symptomatic relief were the focal issues of this preliminary clinical study.
As this mineral supplement is relatively unknown it was important to evaluate safety as well as efficacy. In this 2 month study there were no changes in various clinical and laboratory measures of safety. The study design included an evaluation of mineral supplement dose (2 vs. 3 g/day) and the herbomineral combination. The reason for evaluating these somewhat similar doses reflects the anecdotal clinical experience with these doses, which brought suggestions that the higher dose necessitated a more rigorous assessment. There was little difference between these groups although a better defined week 1 and 2 responses with the high dose were evident. The herbomineral combination produced a greater percentage reduction in WOMAC scores and both groups were associated with reduced consumption of rescue medications. This suggests that these approaches provided additional value but a definitive statistical difference to advocate a higher dose or the herbomineral combination was not achieved. However, with a greater subject enrolment these trends would likely have reached significance.
VAS, as an index of pain, was responsive to both treatment as well as placebo. While there were significant differences at 1 month between mineral supplement treatments and placebo this was not statistically evident at 2 months. Indeed there was a trend for an exaggerated placebo effect for many efficacy variables from week 4 to week 8.
VAS is regarded as being a less sensitive index of disease activity than the WOMAC scores, which assess pain as well as stiffness and physical activity/function. The similar trends for earlier symptomatic relief in sierrasil treated subjects in the all three WOMAC subsets, as well as VAS pain, within the first month was noticeable and readily distinguishable from placebo treated individuals (Figs 1, 2, 3, 4, 5). Indeed, the similar trends in these three treatment groups when taken together suggest that this therapeutic approach has a clear early onset of benefits.
The mechanisms by which this natural mineral supplement achieves these actions and benefits is unclear. The study in human explants indicates that it may affect transcriptional events, as indicated by the reduced production of nitric oxide in response to IL-1β. Increased production of nitric oxide under these circumstances is attributed to the expression of inducible nitric oxide synthase [31,35]. A decade ago we defined that this nitric oxide isoform promoted chronic inflammation [34,35] and inhibitors alleviate numerous inflammatory conditions including arthritis [30].
The protection of matrix degradation in response to IL-1 by sierrasil may reflect a reduction in matrix metalloprotease (MMP) production (also transcriptionally regulated in response to IL-1β) or perhaps a direct interference in the activity of MMPs. There is no evidence for the later but is speculated based on the mineral/metal content of sierrasil and the requirement of metals as catalysts in MMPs. Studies by the manufacturer on the acid liberalization of minerals from sierrasil suggest that the majority of the minerals within sierrasil are tightly bound and are not freely bioavailable (data not shown). Additionally, while the ingestion of supplemental minerals may alter the basal nutritional status of the subjects the literature does not provide a clear link between a nutrition-based action of minerals and an effective anti-arthritic therapy [43]. It is speculated that a micromineral action on gene expression or direct interaction with MMPs may be an alternative action to consider.
A critical issue is how to place these findings in a therapeutic perspective. Osteoarthritic patients, and their healthcare providers, are deeply concerned with the recent documentation of an increased risk for cardiovascular disease and stroke with COX-2 inhibitors [4,5], as well as the significant gastro-intestinal, renal complications and premature deaths associated with non-selective COX inhibitors [3]. With the appreciation that the NSAID class provides symptomatic relief rather that abrogating the disease process [2], there is a great need for alternatives.
Despite the potential of redox-based botanicals as anti-inflammatory agents [11], the most commonly used natural product approach to the management of osteoarthritis is glucosamine and chondroitin [43]. While early studies suggested excellent albeit slow-onset responsiveness [15-17], more recent studies have suggested that the benefits offered by glucosamine and chondroitin may have limitations or be variable in nature [13,18-21]. One recent study suggested that continued use of glucosamine may be unwarranted, even if there were initial benefits [44]. Clarification as to therapeutic potential of this approach is likely to occur when the results of the Glucosamine Arthritis Intervention Trial, with 1588 subjects from 13 centers randomized to 5 treatment groups.
While sierrasil was noted to produce early benefits but it was not clearly established that sustained actions were indistinguishable from placebo. Similar issues have been raised with glucosamine and chondroitin, as well as cetylated fatty acids, where there is some evidence that after 2 months of therapy there are reductions in pain and flexibility but no benefits related to physical function/activity [9].
Conclusion
In summary, the natural mineral supplement sierrasil, alone and in combination with an extract of the Amazonian medicinal plant, cat's claw (Uncaria guianensis), provide a rapid relief of osteoarthritis symptoms. The benefits were evident within a week, and associated with an excellent safety profile. However, a lack of discrimination between treatment and placebo, especially with sustained administration, limits the assessment sierrasil's role in the treatment of osteoarthritis. Nevertheless, as alternative approaches to the management of osteoarthritis are desirable, sierrasil may offer a valuable option for some subjects.
Competing interests
MJSM has an equity interest in Santerra Pharmaceuticals, LLC & Rainforest Nutritionals, Inc.
PB has an equity interest in Santerra Pharmaceuticals, LLC & Rainforest Nutritionals, Inc.
Authors' contributions
MJSM assisted in study design, data analysis and manuscript preparation.
VR contributed to patient recruitment, evaluation and study execution
JG contributed to patient recruitment, evaluation and study execution
RD contributed to patient recruitment, evaluation and study execution
KM assisted in data management, study design monitoring and site communication, laboratory monitoring and manuscript drafting.
SK contributed to study design, GCP monitoring and document planning
AS contributed to site selection and monitoring, protocol compliance and laboratory monitoring and control
HT contributed to protocol and case report form writing.
HP contributed to study design, project planning, quality assurance and manuscript preparation
PB contributed to study design and manuscript preparation
JC contributed to study design, project planning, quality assurance and manuscript preparation.
==== Refs
Murray CJL Lopez AD The Global Burden of Disease 1996 Boston: Harvard University Press
Kraan PM Berg WM Anabolic and destructive mediators in osteoarthritis Curr Opin Nutr Metab Care 2000 3 205 211 10.1097/00075197-200005000-00007
Dickman A Ellershaw J NSAIDs: gastroprotection or selective COX-2 inhibitor? Palliat Med 2004 18 275 86 15198117 10.1191/0269216304pm894fd
Graham DJ Campen D Hui R Spence M Cheetham C Levy G Shoor S Ray WA Risk of acute myocardial infarction and sudden cardiac death in patients treated with cyclo-oxygenase 2 selective and non-selective non-steroidal anti-inflammatory drugs: nested case-control study Lancet 2005 365 475 81 15705456
Kimmel SE Berlin JA Reilly M Jaskowiak J Kishel L Chittams J Strom BL Patients exposed to rofecoxib and celecoxib have different odds of nonfatal myocardial infarction Ann Intern Med 2005 142 157 64 15684203
Berman BM Lao L Langenberg P Lee WL Gilpin AM Hochberg MC Effectiveness of acupuncture as adjunctive therapy in osteoarthritis of the knee: a randomized, controlled trial Ann Intern Med 2004 141 901 10 15611487
Noack W Fischer M Forster KK Rovarti LC Setnikar A Glucosamine sulfate in osteoarthritis of the knee Osteoarthritis Cartilage 1994 2 61 69 11548225 10.1016/S1063-4584(05)80006-8
Walter-Bone K 'Natural remedies' in the treatment of osteoarthritis Drugs Aging 2003 20 517 526 12749749
Hesslink R Armstrong D Nagendra MV Sreevatsan S Barathur R Cetylated fatty acids improve knee function in patients with osteoarthritis J Rheumatol 2002 29 1708 1712 12180734
Piscoya J Rodriuez Z Bustamante SA Okuhama NN Miller MJS Sandoval M Efficacy and safety of freeze-dried cat's claw in osteoarthritis of the knee: mechanisms of action of the species Uncaria guianensis Inflamm Res 2001 50 442 448 11603848
Ahmed S Anuntiyo J Malemud CJ Haqqi TM Biological basis for the use of botanicals in osteoarthritis & rheumatoid arthritis: a review Evidence-Based Complimentary Alternative Medicine
Baker K McAlindon T Exercise for knee osteoarthritis Curr Opin Rheumatol 2000 12 456 463 10990187 10.1097/00002281-200009000-00020
Biggee BA Blinn CM McAlindon TE Nuite M Silbert JE Low levels of human serum glucosamine after ingestion of glucosamine sulphate relative to capability for peripheral effectiveness Ann Rheumatol 2005
Hua J Sakamoto K Nagaoka I Inhibitory actions of glucosamine, a therapeutic agent for osteoarthritis, on the functions of neutrophils J Leukoc Biol 2002 71 632 640 11927650
Pujalte JM Llavore EP Ylescupidez FR Double-blind clinical evaluation of oral glucosamine sulphate in the basic treatment of oseoarthrosis Curr Med Res Opin 1980 7 110 114 7002479
Muller-Fassbender H Bach GI Haase W Rovarti LC Setnikar I Glucosamine sulfate compared to ibuprofen in osteoarthritis of the knee Osteoarthritis Cartilage 1994 44 75 80
Lopez Vas A Double-blind clinical evaluation of the relative efficacy of ibuprofen and glucosamine sulphate in the management of osteoarthrosis of the knee in out-patients Curr Med Res Opin 1982 8 145 149 7128187
McAlindon T Why are clinical trials of glucosamine no longer uniformly positive? Rheum Dis Clin North Am 2003 29 789 801 14603583 10.1016/S0889-857X(03)00064-4
Reginster JY Deroisy R Rovati LC Lee RL Lejeune E Bruyere O Long-term effects of glucosamine sulphate on osteoarthritis progression: a randomised, placebo-controlled clinical trial Lancet 2001 357 251 256 11214126 10.1016/S0140-6736(00)03610-2
Pavelka K Gatterova J Olejarova M Machacek S Giacovelli C Rovarti LC Glucosamine sulphate use and delay of progression of knee osteoarthritis: a 3-year, randomized, placebo-controlled, double-blind study Arch Intern Med 2002 162 2113 2123 12374520 10.1001/archinte.162.18.2113
McAlindon TE Formica M LaValley M Lehmer M Kabbara K Effectiveness of glucosamine for symptoms of knee osteoarthritis: results from an internet-based randomized double-blind controlled trial Am J Med 2004 117 643 649 15501201 10.1016/j.amjmed.2004.06.023
Singh R Ahmed S Malemud CJ Goldberg VM Haqqi TM Epigallocatechin-3-gallate selectively inhibits interleukin-1β-induced activation of mitogen activated protein kinase subgroup c-Jun N-terminal kinase in human osteoarthritis chondrocytes J Orthopaedic Res 2003 21 102 109 10.1016/S0736-0266(02)00089-X
Ahmed S Rahman A Hasnain A Lalonde M Goldberg VM Haqqi TM Green tea polyphenol epigallocatechin-3-gallate inhibits the IL-1β-induced activity and expression of cyclo-oxygenase-2 and nitric oxide synthase-2 activity in human chondrocytes Free Radical Biology Med 2002 33 1097 1105 10.1016/S0891-5849(02)01004-3
Haqqi TM Anthony DD Gupta S Ahmed N Lee M-S Kumar GK Mukhtar H Prevention of collagen induced arthritis in mice by a phenolic fraction of green tea Proc Natl Acad Sci USA 1999 96 4525 4529 10.1073/pnas.96.8.4524
Marok R Winyard PG Coumbe A Kus ML Gaffney K Blades S Mapp PI Morris CJ Blake DR Kaltschmidt C Baeuerle PA Activation of the nuclear transcription factor-κB in human inflamed synovium Arthritis Rheum 1996 39 583 591 8630106
Winyard PG Blake DR Antioxidants, redox-regulated transcription factors Adv Pharmacol 1997 18 403 421 8895818
Lohmander LS What can we do about osteoarthritis? Arthritis Res 2000 2 95 100 11094419 10.1186/ar74
Ahmed S Rahman A Hasnain A Goldberg VM Haqqi TM Pheny N-ter-butylnitrone down-regulates interleukin-1β-stimulated matrix metalloproteinase-13 gene expression in human chondrocytes: suppression of c-Jun NH2-terminal kinase, p38-mitogen-activated protein kinase and activating protein-1 J Pharmacol Exptl Ther 2003 305 981 988 12626640 10.1124/jpet.102.048611
Miller MJS Ahmed S Bobrowski P Haqqi TM Suppression of human cartilage degradation and chondrocyte activation by a unique mineral supplement (sierrasil™) and a cat's claw extract, vincaria® J Amer Nutr Assoc 2004 7 32 39
Connor JR Manning PT Settle SL Moore WM Jerome GM Webber RK Tjoeng FS Currie MG Suppression of adjuvant-induced arthritis by selective inhibition of inducible nitric oxide synthase Eur J Pharmacol 1995 273 15 24 7537678 10.1016/0014-2999(94)00672-T
Amin R Abramson SB The role of nitric oxide in articular cartilage breakdown in osteoarthritis Curr Opin Rheumatol 1998 10 263 268 9608331
Murrell GA Jang D Williams RJ Nitric oxide activates metalloproteinase enzymes in articular cartilage Biochem Biophys Res Commun 1995 206 15 21 7529496 10.1006/bbrc.1995.1003
Clancy RM Amin AR Abramson SB The role of nitric oxide in inflammation and immunity Arthritis Rheum 1998 41 1141 1151 9663469 10.1002/1529-0131(199807)41:7<1141::AID-ART2>3.0.CO;2-S
Miller MJS Sadowska-Krowicka H Chotinaruemol S Kakkis JL Clark DA Amelioration of chonic ilieitis by nitric oxide synthase inhibition J Pharmacol Exptl Ther 1993 264 11 16 7678645
Miller MJS Thompson JH Zhang XJ Sadowska-Krowicka H Kakkis JL Munshi UK Sandoval M Rossi JE Eloby-Childress S Beckman JS Ye YZ Roddi CP Manning PT Currie MG Clark DA Role of inducible nitric oxide synthase expression and peroxynitrite formation in guinea pig ilieitis Gastroenterology 1995 109 1475 1483 7557128 10.1016/0016-5085(95)90633-9
Sandoval-Chacon M Thompson JH Liu X Mannick EE Sadowska-Krowicka H Charbonnet R Clark DA Miller MJS Anti-inflammatory actions of cat's claw: the role of NF-κB Alimentary Pharmacol Ther 1998 12 1279 1289 10.1046/j.1365-2036.1998.00424.x
Sandoval M Charbonnet RM Okuhama N Roberts J Krenova Z Trentacosti AM Miller MJS Cat's claw inhibits TNFα production and scavenges free radicals: role in cytoprotection Free Radical Biol Med 2000 29 71 78 10962207 10.1016/S0891-5849(00)00327-0
Sandoval M Okuhama NN Zhang X-J Condezo LA Lao J Angeles FM Bobrowski P Miller MJS Anti-inflammatory and antioxidant activities of cat's claw (Uncaria tomentosa and Uncaria guianensis) are independent of their alkaloid content Phytomedicine 2002 9 325 337 12120814 10.1078/0944-7113-00117
Miller MJS Angeles FM Reuter BK Bobrowski P Sandoval M Dietary antioxidants protect gut epithelial cells from oxidant induced apoptosis BMC Compl Altern Med 2001 1 11 10.1186/1472-6882-1-11
Obregan LE Cat's claw, Uncaria genus Botanical, Chemical and Pharmacological Studies of Uncaria tomentosa (Wiild) DC (Rubiaceae) and Uncaria guianensis (Aubl) Gmel 1995 Instituto de Fitotherapia Americano. Lima
Wurm M Kacani L Laus G Keplinger K Dierich MP Pentacyclic oxindole alkaloids from Uncaria tomentosa induce human endothelial cells to release a lymphocyte-proliferation-regulatory factor Planta Medica 1998 64 701 704 9933988
Lemaire L Assinewe V Cano P Awang DVC Arnason JT Stimulation of interleukin-1 and -6 production in alveolar macrophages by the neotropical liana, Uncaria tomentosa (Una de gato) J Ethnopharmacol 1999 64 109 115 10197746 10.1016/S0378-8741(98)00113-5
McAlindon TE Biggee BA Nutritional factors and osteoarthritis: recent developments Curr Opin Rheumatol 2005 17 647 652 16093847 10.1097/01.bor.0000175461.57749.46
Cibere J Kopec JA Thorne A Singer J Canvin J Robinson DB Pope J Hong P Grant E Esdaile JM Randomized, double-blind, placebo-controlled glucosamine discontinuation trial in knee osteoarthritis Arthritis Rheumat 2004 51 738 745 15478160 10.1002/art.20697
|
16242032
|
PMC1276811
|
CC BY
|
2021-01-04 16:36:23
|
no
|
J Inflamm (Lond). 2005 Oct 21; 2:11
|
utf-8
|
J Inflamm (Lond)
| 2,005 |
10.1186/1476-9255-2-11
|
oa_comm
|
==== Front
J NeuroinflammationJournal of Neuroinflammation1742-2094BioMed Central London 1742-2094-2-231623231810.1186/1742-2094-2-23ResearchEarly correlation of microglial activation with enhanced tumor necrosis factor-alpha and monocyte chemoattractant protein-1 expression specifically within the entorhinal cortex of triple transgenic Alzheimer's disease mice Janelsins Michelle C [email protected] Michael A [email protected] Salvatore [email protected] Frank M [email protected] Howard J [email protected] William J [email protected] Department of Neurology, University of Rochester School of Medicine and Dentistry, Rochester, New York 14642, USA2 Department of Microbiology and Immunology, University of Rochester School of Medicine and Dentistry, Rochester, New York 14642, USA3 Center for Aging and Developmental Biology, University of Rochester School of Medicine and Dentistry, Rochester, New York 14642, USA4 Department of Neurobiology and Behavior, University of California, Irvine, California 92697, USA2005 18 10 2005 2 23 23 15 9 2005 18 10 2005 Copyright © 2005 Janelsins et al; licensee BioMed Central Ltd.2005Janelsins et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Alzheimer's disease is a complex neurodegenerative disorder characterized pathologically by a temporal and spatial progression of beta-amyloid (Aβ) deposition, neurofibrillary tangle formation, and synaptic degeneration. Inflammatory processes have been implicated in initiating and/or propagating AD-associated pathology within the brain, as inflammatory cytokine expression and other markers of inflammation are pronounced in individuals with AD pathology. The current study examines whether inflammatory processes are evident early in the disease process in the 3xTg-AD mouse model and if regional differences in inflammatory profiles exist.
Methods
Coronal brain sections were used to identify Aβ in 2, 3, and 6-month 3xTg-AD and non-transgenic control mice. Quantitative real-time RT-PCR was performed on microdissected entorhinal cortex and hippocampus tissue of 2, 3, and 6-month 3xTg-AD and non-transgenic mice. Microglial/macrophage cell numbers were quantified using unbiased stereology in 3xTg-AD and non-transgenic entorhinal cortex and hippocampus containing sections.
Results
We observed human Aβ deposition at 3 months in 3xTg-AD mice which is enhanced by 6 months of age. Interestingly, we observed a 14.8-fold up-regulation of TNF-α and 10.8-fold up-regulation of MCP-1 in the entorhinal cortex of 3xTg-AD mice but no change was detected over time in the hippocampus or in either region of non-transgenic mice. Additionally, this increase correlated with a specific increase in F4/80-positive microglia and macrophages in 3xTg-AD entorhinal cortex.
Conclusion
Our data provide evidence for early induction of inflammatory processes in a model that develops amyloid and neurofibrillary tangle pathology. Additionally, our results link inflammatory processes within the entorhinal cortex, which represents one of the earliest AD-affected brain regions.
NeuroinflammationAlzheimer's diseasebeta-amyloidpro-inflammatory moleculemicroglia3xTg-ADTNF-αMCP-1
==== Body
Background
Alzheimer's disease (AD) is an age-related neurodegenerative disorder associated with progressive functional decline, dementia and neuronal loss. Demographics make evident that the prevalence of AD will increase substantially over the coming decades. Patients initially exhibit an inability to assimilate new information and as the disease progresses, both declarative and nondeclarative memory become ever more profoundly impaired [1]. The pervasive societal and economic burden created by this debilitating disease should provide sufficient incentive for the development of new natural history-modifying therapeutic approaches. However, because the mechanistic underpinnings of AD are incompletely understood, the clinical disease spectrum broad, and the neuropathological features of its initiation and progression limited, the development of such potential disease modifying therapies has been relatively limited.
The pathological hallmarks of the AD brain include extracellular proteinaceous deposits (plaques), composed largely of amyloid beta (Aβ) peptides, and intraneuronal neurofibrillary tangles (NFTs), which are characterized by excessive phosphorylation of tau protein. Other AD-related histopathologic features include, but are not limited to, astrogliosis, microglial activation, and reduction of synaptic integrity. These features appear to arise in a region- and time-dependent manner (reviewed in [2]). Amyloid pathology evolves in stages: early involvement is anatomically circumscribed to the basal neocortex, most often within poorly myelinated temporal areas; progression involves adjacent neocortical areas, the hippocampal formation, perforant path inclusive of its coursing through the subiculum and termination within the molecular layers of the dentate gyrus, and; finally the process involves all cortical areas [3]. Neurofibrillary tangle pathology is also progressive: Initially involving projection neurons with somata in the transentorhinal region, tangles then extend to the entorhinal region proper typically in the absence of amyloid deposition. Subsequent progression to the hippocampus and temporal proneocortex, and then association neocortex, followed by superiolateral spread and ultimately extending to primary neocortical areas [4-6]. Moreover, individuals diagnosed with mild cognitive impairment, a forme fruste of AD, display decreased entorhinal and hippocampal volume, primarily associated with diminished neuron number as compared to non-cognitively impaired controls [7-10]. These data suggest that the entorhinal cortex and hippocampus are selectively vulnerable early during the disease process.
Gaining an enhanced understanding of why these brain regions are specifically susceptible to neurodegeneration in the context of AD and elucidating the mechanisms underlying these disease processes has been the subject of intensive investigation over the past several decades. Attention has been focused upon synaptic dysfunction, due to the previously observed diminution of cholinergic synapse density and overall synapse numbers during early stages of AD [11,12]. Additionally, mouse models overexpressing human amyloid precursor protein (APP), the protein from which pathogenic Aβ peptides are proteolytically derived, exhibit decreased synaptic function antecedent to plaque deposition [13], thereby further implicating disrupted synaptic function in early stages of AD pathogenesis.
Inflammatory processes, marked by activated microglia and astrocytes in the post-mortem AD brain some of which co-localize to plaques and tangles, have long been hypothesized to contribute to AD pathogenesis [14]. The role that this response plays in the disease process, especially during pre-symptomatic stages, is not well defined. There exist multiple means by which inflammatory processes can affect neurons and potentially synaptic function in AD. Cytokines have been shown to be expressed in response to Aβ generation and a subset of these molecules have demonstrated neurotoxic activities [15-17]. Such observations imply these inflammatory molecules may serve to mechanistically link the elaboration of pathological hallmarks and synaptic dysfunction. We hypothesized that inflammation plays a role early during the disease process, at a time when synaptic dysfunction and early cognitive deficits first become evident. Disease-related inflammatory contributors to synaptic dysfunction found in early AD have long been debated, but such studies have been hampered by the lack of age-matched, early-stage human post-mortem tissue samples as well as AD-relevant animal models. In the present study, we sought to determine the temporal and region-specific expression of inflammatory molecules, previously implicated in late-stage AD, in the context of a mouse model that develops amyloid and tau pathology. A triple-transgenic model of AD (3xTg-AD) has recently been created that harbors three disease-relevant genetic alterations: a human Presenilin M146V knock-in mutation (PS1M146V), human amyloid precursor protein Swedish mutation (APPswe), and the human tauP301L mutation. These mice develop plaques and tangles in a spatial and time-dependent manner similar to pathological hallmarks observed in the brains of AD-afflicted individuals [18,19]. Most notably, this is the first animal model developed to date which facilitates the study of inflammation in the context of both amyloid and tau pathology. We performed region-specific quantitative transcript analyses and unbiased stereological counting to correlate regional and temporal changes in inflammatory molecule expression profiles to alterations in inflammatory cell numbers and AD-related pathologies. Our findings further implicate inflammatory processes as playing a role early during the disease process, and that regional differences exist that may elucidate why particular brain regions are more susceptible to AD-related disease mechanisms.
Materials and methods
Strains of mice
Triple transgenic (3xTg-AD) mice were created as previously described [18,19]. Age-matched 2, 3, and 6 month-old male mice were used in all studies (n = 6 per experimental group for biochemical assays, n = 4 per experimental group for quantitative stereological studies). Age-matched male C57BL/6 mice were used as non-transgenic controls in all experiments. All animal housing and procedures were performed in compliance with guidelines established by the University Committee of Animal Resources at the University of Rochester.
Quantitative real-time PCR analysis of pro-inflammatory molecules from brain-derived RNA
RNA was isolated from microdissected hippocampus- or entorhinal cortex-enriched tissue from 2, 3, and 6 month-old 3xTg-AD and non-transgenic mice with TRIzol solution (Invitrogen, Carlsbad, CA). RNA was treated with RQ DNAse I (Promega, Madison, WI) to selectively degrade any contaminating genomic DNA, followed by phenol:chloroform extraction and ethanol precipitation. One microgram of total RNA was reverse transcribed using Applied Biosystems High-Capacity cDNA Archive Kit. An aliquot of cDNA (100 ng) was used to assess presence of 23 inflammatory targets per mouse, and was analyzed in a standard PE7900HT quantitative PCR reaction using a Taqman Assay on Demand primer probe sets in Microfluidic cards (Applied Biosystems, Foster City, CA) and 100 μL MasterMix containing HotStart DNA polymerase (Eurogentec, Belgium). 18s RNA served as the control to which all samples were normalized (Applied Biosystems, Foster City, CA). We further analyzed the data using the ΔΔCT method, normalizing the 3 and 6 month-old 3xTg-AD and control mouse samples to the 2 month-old 3xTg-AD and non-transgenic samples, respectively.
Quantitative histochemical analysis of macrophages and microglia in brains of 3xTg-AD and non-transgenic mice
Age-matched 3xTg-AD and non-transgenic mice were sacrificed and processed with 4% paraformaldehyde (PFA)/PB trans-cardiac perfusions; brains were removed and post-fixed overnight with 4% PFA/PB. Sequentially, brains were transferred to 20% sucrose in PBS overnight and then 30% sucrose where they remained until sectioning. Brains were sectioned coronally (30 μm) on a sliding microtome, and stored in cryoprotectant until used for immunohistochemistry.
Sections were washed four times for 3 min. each in PB to remove cyroprotectant. To quench endogenous peroxidase activity, sections were incubated for 25 min. with 3% H2O2 (Sigma). Sections were mounted onto slides and allowed to dry. Slides were incubated in 0.15 M PB + 0.4% Triton-X100 for 5 min. at room temperature (RT; 22°C) to permeabilize the tissue. Then slides were incubated with blocking solution containing 3% normal goat serum, 3% bovine serum albumin, and 0.4% Triton-X 100 in 0.15 M PB for 1 hr. Slides were incubated with rat monoclonal anti-F4/80 antibody (Serotec, 1:100) overnight in blocking solution. Next, slides were washed eight times for 3 min. each with 0.15 M PB prior to incubation with Vectastain biotinylated goat anti-immunoglobulin (Vector Laboratories, Burlingame, CA) for 2 hrs. at RT. Excessive secondary antibody was washed in 0.15 M PB and incubated with A and B reagents (Vector Laboratories, Burlingame, CA) to conjugate HRP. Slides were developed using a DAB peroxidase kit, according to manufacturer's instructions for nickel enhancement (Vector Laboratories, Burlingame, CA).
Positively stained F4/80-expressing cells were visualized using an Olympus AX-70 microscope equipped with a motorized stage (Olympus, Melville, NY) and the MCID 6.0 Elite Imaging Software (Imaging Research, Inc.). Sections were tiled under 4× magnification. Five equal sections of entorhinal cortex and seven equal sections of hippocampus from each mouse (4 mice total) per timepoint were analyzed. Fifty percent of the defined region of interest in the entorhinal cortex or hippocampus was assessed, under 60× magnification. The coordinates from which sections were chosen for the entorhinal cortex were 2.92 mm to 4.04 mm posterior from Bregma. The sections counted in the hippocampus were from 1.70 mm to 3.40 mm posterior from Bregma.
Qualitative immunohistochemical analysis of amyloid deposition in 3xTg-AD and non-transgenic mice
Sections were washed three times for 5 min. each, then twice for 30 min. in PBS to remove cryoprotectant. To quench endogenous peroxidase activity, sections were incubated with 3% H2O2 and 3% methanol for 25 min. Sections were then washed twice for 5 min. each with PBS, followed by epitope retrieval treatment with 90% formic acid for 5 min. at RT. Next, sections were washed twice for 5 min. each with PBS. Tissue was permeabilized with PBS + 0.1% PBS/Triton-X 100. Sections were then incubated for 1 hr at RT with PBS + 0.1% PBS/Triton-X 100 + 10% normal goat serum. Sections were incubated overnight at 4°C with primary 6E10 antibody (Signet, 1:1000) in PBS 0.1% PBS/Triton-X 100 + 1% normal goat serum. Samples were washed twice for 10 min. each with PBS + 0.1% Triton-X 100 + 1% normal goat serum prior to addition of secondary antibody. The mouse HRP ABC kit was used according to manufacturer's protocol (Vector Laboratories, Burlingame, CA). Excessive secondary antibody was washed in PBS and developed using a DAB peroxidase kit, according to manufacturer's instructions for nickel enhancement (Vector Laboratories, Burlingame, CA) and mounted on slides.
Results
Temporal progression of intracellular Aβ accumulation in 3xTg-AD mice
Inflammatory processes have been intimately associated with classic AD pathology in the post-mortem human brain, where evidence of astrogliosis and activated microglia in the vicinity of amyloid plaques has been readily observed [20]. Implication of inflammatory mediators in early pathogenic events during pre-symptomatic stages of AD, however, has not been clearly defined at present due to limited availability of early-stage human clinical samples and a lack of animal models that faithfully recapitulate the human disorder. The recently characterized triple-transgenic AD mouse (3xTg-AD) presently represents the most advanced animal model available in that it harbors three AD-relevant genetic alterations, which result in spatial distribution and progression of amyloid and tau pathologies strikingly similar to human AD [18,19]. To clarify the role of inflammatory processes early during disease progression, we initially assessed the age-dependent accretion of human Aβ in the entorhinal cortex and hippocampus of 3xTg-AD mice, as many posit accumulating Aβ acts as a likely early trigger of AD-related inflammatory processes [21]. The entorhinal cortex and hippocampus were the regions chosen because of the abundance of evidence implicating these regions in the earliest stages of disease [6,7,10,22]. Coronal sections from 2, 3, and 6-month old 3xTg-AD and non-transgenic mice were immunohistochemically stained with 6E10 antibody to assess extent of intracellular and extracellular human Aβ deposition. Immunohistochemical analyses revealed intracellular human Aβ staining at the 3-month time-point whereas Aβ is not detectable at 2 months of age. The number of Aβ immuno-positive cells increased by 6 months of age and the intensity of individual cell staining was significantly enhanced in 3xTg-AD mice (Fig. 1A). Extracellular accumulations or amyloid plaque-like deposits were not observed at any of these early ages. Non-transgenic mice did not exhibit Aβ staining, further confirming that the 6E10 antibody was specific for human Aβ in 3xTg-AD mice (Fig. 1B).
Figure 1 Intracellular Aβ appears at 3 months and is enhanced by 6 months in 3xTg-AD mice. Coronal brain sections from 2, 3 and 6 month-old 3xTg-AD and non-transgenic mice were stained with human APP/Aβ-specific 6E10 antibody and developed using DAB. Panel A illustrates that the brains of 2 month-old 3xTg-AD mice are pre-pathologic, while at 3 months, hAPP/Aβ can be readily detected in both the entorhinal cortex and hippocampus of 3xTg-AD mice. By 6 months of age, 3xTg-AD mice exhibit further enhanced deposition of hAβ in both regions. Panel B identifies sections of entorhinal cortex and hippocampus from non-transgenic mice, which are not immunohistochemically positive for endogenous mouse Aβ, therefore indicating that the 6E10 antibody specifically detects transgene-driven expression of hAPP/Aβ in 3xTg-AD mice. The scale bars depict 50 μm. The insets represent 60× magnification.
Pro-inflammatory transcript profiling of 3xTg-AD and non-transgenic mouse entorhinal cortex and hippocampus reveals temporal and spatial expression of TNF-α and MCP-1
We predicted that if inflammation was involved at the earliest stages of the disease process, we would observe the coordinate expression of immunomodulatory molecules between 3 and 6 months of age, when intracellular Aβ begins to accumulate in the entorhinal cortex and hippocampus of 3xTg-AD mice. To this end, we selected a set of target inflammatory molecules that have been implicated in inflammatory responses within the central nervous system, including cytokines, chemokines, cell adhesion molecules, T cell markers and immune-related enzymes (Table 1). Many of these markers have been implicated in late-stage AD and possess the potential to influence early pathogenic processes within the regions affected early in AD. Quantitative real-time RT-PCR was performed to determine levels of these targets in microdissected entorhinal cortex and hippocampus tissue of 2, 3, and 6 month-old 3xTg-AD mice. Age-matched non-transgenic mouse samples derived from identical regions were employed as controls (n = 6 per genotype per time point). Surprisingly, we detected a 14.8-fold up-regulation of TNF-α, a pro-inflammatory modulator, and 10.8-fold increase of the chemokine MCP-1 mRNA in the entorhinal cortex of 6 month-old 3xTg-AD mice versus the 2 month-old animals (Table 2). Levels of both pro-inflammatory molecules are also slightly elevated in the 3-month 3xTg-AD entorhinal cortex, although not reaching statistical significance as compared to 2 month-old counterparts. This trending increase of TNF-α and MCP-1 transcript levels at 3 months of age correlates with the initial appearance of human transgene-derived Aβ in 3xTg-AD mice. Conversely, no detectable changes were observed in any of the assessed transcriptional targets in cDNA pools generated from hippocampal RNA samples at any of the time-points (Table 3), even though intracellular human Aβ was readily detectable within this brain region (Fig. 1A). It is remarkable that the TNF-α and MCP-1 transcript response is specific to cells resident to the entorhinal cortex, suggesting that aspects of the cellular environment may be responsible for differential inflammatory outcomes in these two disease-affected brain regions.
Table 1 Proinflammatory markers investigated in the temporal and spatial progression of early AD pathogenesis. Immune cell molecules/inflammatory markers were assessed from RNA isolated from entorhinal cortex and hippocampus tissue of 2, 3, and 6 month-old 3xTg-AD and non-transgenic mice by qRT-PCR using Applied Biosystems Microfluidic Cards.
Immune Marker Major Functions
C3 complement protein, binds to pathogenic structures
CCL2 (MCP-1) chemokine, promotes extravasation, activates macrophages, promotes Th2 immunity
CCL3 chemokine, promotes extravasation, antiviral defense, promotes Th1 immunity
Fractalkine chemokine, involved in brain inflammation, endothelial adhesion
IP10 chemokine, antiangiogenic, promotes Th1 immmunity
TNF-α cytokine, proinflammatory, attracts innate immune cells, activates macrophages
TGF-β cytokine, inhibits cell growth
IL-2 cytokine, T cell growth factor
IL-6 cytokine, B and T cell growth and differentiation
IL-8 cytokine, secreted by macrophage(predominately in response to bacterial infection), recruits innate and adaptive immune cells
IL-1α cytokine, macrophage and T cell activation
IL-1β cytokine, macrophage and T cell activation
IL-12a cytokine, activates NK cells, induces CD4 differentiation to Th1 cell
ICAM 1 intercellular adhesion molecule present on endothelial cells, binds LFA-1 and Mac-1 (Cd11b)
VCAM 1 adhesion molecule present on endothelial cells, binds VLA-4 integrin
CD4 cell surface marker for TH1 and TH2 T cells, coreceptor for MHC II
CD8 cell surface marker on cytotoxic T cells, coreceoptor for MHC I
CD80 cell surface marker, T cell/antigen presenting cell costimulation
CD86 cell surface marker, T cell/antigen presenting cell costimulation
Ptgs1 Cyclooxygenase type 1 (Cox-1)
Ptgs2 Cyclooxygenase type 2 (Cox-2)
Caspase 3 late-stage molecule involved in apoptosis
Table 2 TNF-α and MCP-1 mRNA levels are selectively elevated in the entorhinal cortex of 3xTg-AD mice prior to overt amyloid plaque pathology. Total RNA was purified from microdissected entorhinal cortex from 2, 3, and 6 month-old 3xTg-AD and non-transgenic control mice. cDNA was generated and subjected to Applied Biosystems Microfluidic Card analysis, a high-throughput quantitative RT-PCR technology that facilitates the simultaneous quantitation of 23 inflammation-related transcriptional targets. Of the panel of transcripts analyzed, only TNF-α and MCP-1 transcript levels were significantly enhanced by 6 months of age specifically within the entorhinal cortex of 3xTg-AD mice (n = 6/group). These cytokine transcripts were unchanged in the entorhinal cortex of non-transgenic mice at all time-points analyzed. *p < 0.0005 when compared to the 2 month timepoint. Proinflammatory transcript expression in 3xTg-AD and non-transgenic mice in the entorhinal cortex
3xTg-AD Non-Transgenic
3 months 6 months 3 months 6 months
Marker Fold Change (Relative to 2 months) Fold Change (Relative to 2 months) Fold Change (Relative to 2 months) Fold Change (Relative to 2 months)
C3 0.357 +/- 0.258 0.615 +/- 0.477 0.94 +/- 0.622 4.432 +/- 9.424
MCP-1 (CCL2) 5.079 +/- 4.826 10.796* +/- 3.298 1.459 +/- 1.230 2.616 +/- 5.184
CCL3 0.317 +/- 0.279 0.521 +/- 0.388 0.754 +/- 0.394 0.414 +/- 0.170
Fractalkine 0.853 +/- 0.215 0.858 +/- 0.333 1.584 +/- 0.568 0.989 +/- 0.311
IP10 1.291 +/- 0.911 1.922 +/- 1.028 1.157 +/- 0.507 1.393 +/- 1.062
TNF-α 5.299 +/- 4.580 14.822* +/- 5.618 1.215 +/- 1.793 1.702 +/- 1.916
TGF-β 1.149 +/- 0.181 1.092 +/- 0.527 1.054 +/- 0.178 0.922 +/- 0.312
IL-2 1.292 +/- 0.830 1.900 +/- 2.225 0.811 +/- 0.218 1.459 +/- 0.740
IL-6 2.216 +/- 2.921 4.900 +/- 5.817 1.632 +/- 1.604 3.340 +/- 4.726
IL-8 0.110 +/- 0.075 0.190 +/- 0.180 0.350 +/- 0.234 0.424 +/- 0.418
IL-1α 0.932 +/- 0.243 1.371 +/- 0.687 0.788 +/- 0.290 0.568 +/- 0.285
IL-1β 0.413 +/- 0.478 0.899 +/- 0.925 1.528 +/- 1.502 0.893 +/- 0.859
IL-12α 0.402 +/- 0.182 0.444 +/- 0.346 15.159 +/- 22.054 13.181 +/- 13.192
ICAM 1 1.000 +/- 0.317 1.215 +/- 0.567 1.029 +/- 0.076 0.733 +/- 0.202
VCAM 1 1.031 +/- 0.067 0.969 +/- 0.399 0.890 +/- 0.137 0.690 +/- 0.337
CD4 3.149 +/- 2.363 2.191 +/- 1.772 0.549 +/- 0.488 0.605 +/- 0.440
CD8 0.876 +/- 0.603 2.190 +/- 1.978 0.329 +/- 0.433 9.954 +/- 17.454
CD80 0.888 +/- 0.158 0.812 +/- 0.460 0.822 +/- 0.473 0.714 +/- 0.451
CD86 0.839 +/- 0.237 0.843 +/- 0.394 1.002 +/- 0.207 0.628 +/- 0.237
Ptgs1 1.066 +/- 0.225 1.270 +/- 0.609 1.165 +/- 0.232 0.966 +/- 0.323
Ptgs2 0.725 +/- 0.265 0.538 +/- 0.261 2.105 +/- 1.357 1.369 +/- 0.485
Caspase 3 0.955 +/- 0.171 0.900 +/- 0.424 0.884 +/- 0.173 0.741 +/- 0.366
VEGF 0.887 +/- 0.214 0.811 +/- 0.266 0.959 +/- 0.229 0.974 +/- 0.618
*p < 0.0005, student t test:two sample equal variance
Table 3 Inflammation-related transcript levels remain stable in the hippocampus of 2, 3, and 6 month-old 3xTg-AD and control mice. Total RNA was purified from microdissected hippocampus from 2, 3, and 6 month-old 3xTg-AD and non-transgenic control mice. cDNA was generated and subjected to Applied Biosystems Microfluidic Card analysis of 23 inflammation-related transcriptional targets. None of the transcriptional targets assessed exhibited altered expression at any of the assessed time-points. Proinflammatory transcript expression in 3xTg-AD and non-transgenic mice in the hippocampus
3xTg-AD Non-Transgenic
3 months 6 months 3 months 6 months
Marker Fold Change (Relative to 2 months) Fold Change (Relative to 2 months) Fold Change (Relative to 2 months) Fold Change (Relative to 2 months)
C3 1.181 +/- 0.855 1.006 +/- 0.473 0.754 +/- 0.385 1.123 +/- 0.734
MCP-1 (CCL2) 0.554 +/- 0.72 0.204 +/- 0.122 1.864 +/- 0.756 1.392 +/- 0.709
CCL3 0.621 +/- 0.353 0.571 +/- 0.339 0.664 +/- 0.305 0.638 +/- 0.369
Fractalkine 0.619 +/- 0.136 0.718 +/- 0.256 0.973 +/- 0.086 0.919 +/- 0.240
IP10 0.347 +/- 0.154 0.388 +/- 0.246 1.185 +/- 0.552 0.796 +/- 0.481
TNF-α 0.238 +/- 0.278 0.216 +/- 0.195 0.899 +/- 0.432 0.461 +/- 0.348
TGF-β 0.683 +/- 0.187 0.793 +/- 0.138 1.117 +/- 0.178 0.935 +/- 0.468
IL-2 0.650 +/- 0.077 0.637 +/- 0.055 0.360 +/- 0.228 0.462 +/- 0.447
IL-6 0.373 +/- 0.215 0.531 +/- 0.524 0.719 +/- 0.320 0.396 +/- 0.506
IL-8 0.180 +/- 0.228 0.198 +/- 0.170 0.292 +/- 0.202 0.206 +/- 0.105
IL-1α 0.602 +/- 0.111 0.838 +/- 0.229 0.766 +/- 0.431 0.825 +/- 0.552
IL-1β 0.449 +/- 0.687 0.517 +/- 0.885 1.489 +/- 1.711 1.672 +/- 1.570
IL-12α 1.597 +/- 2.736 3.329 +/- 7.112 0.362 +/- 0.364 0.678 +/- 0.747
ICAM 1 0.771 +/- 0.174 0.756 +/- 0.162 1.075 +/- 0.163 0.941 +/- 0.324
VCAM 1 0.730 +/- 0.175 0.738 +/- 0.173 0.943 +/- 0.191 0.833 +/- 0.337
CD4 0.603 +/- 0.751 0.681 +/- 0.822 1.258 +/- 0.750 1.190 +/- 0.847
CD8 0.534 +/- 0.484 0.943 +/- 1.047 1.127 +/- 1.011 10.462 +/- 10.77
CD80 0.464 +/- 0.353 0.349 +/- 0.249 1.116 +/- 0.496 1.117 +/- 1.079
CD86 0.614 +/- 0.126 0.614 +/- 0.163 0.893 +/- 0.112 0.686 +/- 0.307
Ptgs1 0.771 +/- 0.153 0.913 +/- 0.262 1.013 +/- 0.105 1.025 +/- 0.287
Ptgs2 0.649 +/- 0.309 0.936 +/- 0.351 1.060 +/- 0.092 0.868 +/- 0.370
Caspase 3 0.556 +/- 0.145 0.569 +/- 0.119 0.830 +/- 0.187 0.697 +/- 0.243
VEGF 0.728 +/- 0.196 0.723 +/- 0.169 0.819 +/- 0.111 0.809 +/- 0.257
Increased microglial/macrophage numbers in the entorhinal cortex correlates with enhanced TNF-α and MCP-1 expression
Our finding that specific TNF-α and MCP-1 transcript expression within the entorhinal cortex of 3xTg-AD mice indicates that a regional difference exists between the entorhinal cortex and hippocampus that would elaborate a time-dependent increase in the intensity of inflammatory processes. This difference appears to be independent of solely intracellular human Aβ accumulation since transgene expression is extant in both regions at 3 and 6 months. Because microglia and macrophages represent likely candidate(s) of TNF-α and MCP-1 production [23,24], we assessed the total number of cells in the entorhinal cortex and hippocampus of transgenic and non-transgenic mice to determine if differences exist that may explain regional differences of inflammatory cytokine expression profiles. Coronal sections containing the entorhinal cortex and hippocampus of 2 and 6-month 3xTg-AD and control mice were immunohistochemically stained with anti-F4/80 antibody to identify resident microglia and macrophages. Microscopic analysis revealed that the microglial phenotype in the entorhinal cortex of 3xTg-AD mice was that of a highly activated state as shown by enhanced staining and cellular morphology at 6 months of age (Fig. 2A). Minimal differences in microglial activation state were apparent in identical regions in non-transgenic control animals (Fig. 2A). To quantify cell number in this region, we performed unbiased stereology on 3xTg-AD and control entorhinal cortex sections. We detected a significant increase in the total number of F4/80-positive cells in the entorhinal cortex of 6 month-old 3xTg-AD mice as compared to 2 month-old counterparts (Fig. 2B). There was no change in the number of F4/80-positive cells in the entorhinal cortex of control mice, indicating that the increase in microglial cell number in transgenic mice was not due to normative aging. Assessment of F4/80-positive cells in the hippocampus revealed no detectable alteration between 2 and 6 month-old 3xTg-AD and control mice in terms of microglial activation status (Fig. 3A). Quantification using unbiased stereology indicated no significant differences in F4/80-positive cell numbers between the 2 and 6-month time-points (Fig. 3B).
Figure 2 The 3xTg-AD entorhinal cortex harbors an increased number of macrophages/microglia at 6 months of age. Coronal brain sections from 2 and 6 month-old 3xTg-AD and control mice were stained with anti-F4/80 antibody and developed using DAB. (A) Qualitative image analysis reveals activation of F4/80-expressing macrophages and microglia specifically in the entorhinal cortex of 3xTg-AD mice at 6 months of age. (B) Unbiased quantitative stereologic analyses were performed on the entorhinal cortex to derive the total number of F4/80-positive cells. Error bars indicate standard deviation. N = 4 per genotype per time point. "*" indicates p < 0.008. The scale bar represents 50 μm.
Figure 3 The 3xTg-AD hippocampus does not have an increased number of macrophages/microglia at 6 months of age. Coronal brain sections from 2 and 6 month-old 3xTg-AD and control mice were stained with the anti-F4/80 antibody and developed using DAB. (A) Qualitative analyses shows little change in activation of microglia and macrophages in the hippocampus of 3xTg-AD or non-transgenic mice over time. (B) Unbiased stereology was performed on the hippocampal formation to determine total number of F4/80-positive cells. Error bars indicate standard deviation. N = 4 per genotype per time point. The scale bar represents 50 μm.
Discussion
Dissecting the role that inflammation plays early in AD is challenging, as AD is a complex chronic disorder with varying pathologic sequelae from which the underlying causative mechanisms are unknown. Activation of microglia and astrocytes, and the presence of many inflammatory mediators, including cytokines, chemokines and complement proteins have been only identified in the post-mortem AD brain in the vicinity of senile plaques and NFTs [20,25]. This observation leads one to question if inflammation is involved early during the course of AD and, if so, how does it contribute to pathogenesis? Understanding the earliest events is of utmost importance, as inflammation may represent a viable therapeutic target of AD. Interestingly, retrospective studies assessing the effects of non-steroidal anti-inflammatory drugs (NSAIDs) on nondemented individuals have shown decreased risk of developing AD when these individuals utilized NSAIDs for prolonged periods of time [26,27]. Our studies aimed to identify the earliest period during which inflammatory processes initiate in the 3xTg-AD mouse model [19]. Our results illustrate 3 main points: 1) Inflammatory processes precede significant extracellular amyloid plaque deposition in the 3xTg-AD brain, substantiated by increased TNF-α and MCP-1 transcript levels, coincident temporally with the production of intracellular Aβ accumulation. 2) The expression of these molecules is spatially localized to the entorhinal cortex but not hippocampus at the early time-points assessed. 3) There is a marked increase in the number of microglia and macrophages in the entorhinal cortex that correlates with when TNF-α and MCP-1 transcript levels are significantly up-regulated.
In the late-stage AD brain, it has been shown that inflammatory molecules are produced primarily by microglia and astrocytes as they respond to plaques and neuronal damage [17]. Our finding of increased TNF-α and MCP-1 expression prior to significant plaque deposition in 3xTg-AD mice, which occurs extensively at 12 months [19], may represent a contributory role between inflammatory processes and early AD pathogenesis. Precisely how these molecules impart effects in 3xTg-AD mice at this early time-point is not certain; however, recent evidence has suggested that TNF-α and MCP-1, as well as other pro-inflammatory molecules may play a role in inhibiting microglial phagocytosis of fibrillar Aβ in vitro [28]. Likewise, increased inflammatory responses and subsequent secretion of cytokines, in particular, IL-1β, may play an important role in tau phosphorylation in the 3xTg-AD model [29]. In this study, activation of microglia by LPS only affected the tau pathology via cdk5/p25 activation, but not the amyloid pathology, further highlighting the potential pathophysiological changes that can be induced by inflammation in AD. Certainly, inflammatory mediators have been implicated as being both protective and exacerbating, depending on the model system and the levels of cytokine present [30].
TNF-α can be expressed by astrocytes, microglia and neurons in response to various stimuli in the CNS [17]. Initially, TNF-α is an innate mediator, promoting chemokine and cytokine expression and extravasation of other immune cells. One possible mechanism that may implicate TNF-α in contributing to AD pathogenesis is evidence that it can increase Aβ peptide production [31]. Additionally, inflammatory molecule signaling may cause increased cleavage of APP by the γ-secretase complex, whereby TNF-α, IL-1β, and IFN-γ have been shown to enhance production of Aβ peptides via a γ-secretase-dependent mechanism in vitro. Moreover, antagonizing TNFR1 signaling can lead to diminished γ-secretase activity [32]. Further evidence supporting pathogenic effects of TNF-α-mediated signaling is TNFR1 and TRADD, a TNF receptor adaptor protein that allows for NF-κB and JNK activation, are both increased in AD tissue and APPswe mice. This increase is correlative with TUNEL-positive neurons in primary hippocampal cultures [33]. Collectively, these observations suggest TNF-α contributes to aberrant APP processing and initiation of pro-apoptotic pathways.
MCP-1 is a chemokine that is expressed by microglia and astrocytes that facilitates extravasation of immune cells expressing its cognate receptor, CCR2, to cross the blood brain barrier and guides them to the site of damage. As with TNF-α, the role of MCP-1 in AD pathophysiology is uncertain. A recent study of APPswe/CCL2 (MCP-1) bigenic mice showed increased diffuse Aβ deposition, as compared to APPswe mice at 14 months of age. Since changes were not observed in APP processing, the authors concluded that MCP-1 overexpression in APPswe mice correlated with diminished clearance of Aβ [34]. Overall, it is interesting that of the 23 immunomodulatory markers assessed in our study, TNF-α and MCP-1 were the only two that changed significantly over time, possibly signifying their importance during nascent stages of AD pathogenesis. Perhaps, the other inflammatory targets are triggered at later stages of the disease in response to further neurodegenerative events.
Exogenously applied Aβ can trigger the expression of cytokines in vitro and when injected directly into the mouse brain [24,35]. However, the fascinating result in our study is that expression of TNF-α and MCP-1 was detected specifically within the entorhinal cortex and not the hippocampus, despite the fact that immunocytochemically detectable intraneuronal Aβ increased over time in both brain regions. Additionally, although not statistically significant, TNF-α and MCP-1 transcript levels were elevated at 3 months of age in 3xTg-AD mouse entorhinal cortex, and were increased to statistical significance by 6 months of age suggesting a state of chronic up-regulation and positive-feedback for the expression of both of these inflammatory molecules. Therefore, we are unable to conclude, as detected by the methodology employed in this study, that intracellular Aβ accumulation is the sole contributing factor promoting TNF-α and MCP-1 transcript expression specifically in the entorhinal cortex. It is possible that these molecules are neuronally expressed within the entorhinal cortex because they are inherently more sensitive to accumulating Aβ, as other neurons were shown previously to express both of these molecules during times of stress [17]. Another possibility is that the entorhinal cortex elaborates inflammatory molecule expression owing to the structure of Aβ elaborated. For example, oligomeric Aβ is posited to be the more toxic structural intermediate arising during Aβ fibrillogenesis, and this form has been shown to readily induce cytokine expression in vitro [36]. Regional differences in intracellular and extracellular oligomeric Aβ profiles in vivo may, therefore, account for the regional specificity of cytokine/chemokine expression and microglial activation that we observed in the 3xTg-AD mice. Conversely, the regional elaboration of TNF-α and MCP-1 may occur via an Aβ-independent mechanism and caused by an environmental stimulus, such as region-selective oxidative stress. Practico and colleagues demonstrated that lipid peroxidation in the APPswe brain can occur prior to Aβ deposition in APPswe mice [37], suggesting that disruption of cellular membrane homeostasis could also contribute to inflammatory molecule induction, and perhaps, regional differences in lipid peroxidation profiles are responsible for our region-specific observations in 3xTg-AD mice.
The significant increase in the number of microglia and macrophages in the entorhinal cortex from 2 to 6 months of age in 3xTg-AD mice is coincident with the increase of TNF-α and MCP-1. We believe this could be due to microglial proliferation, activation of the resting resident population of brain microglia and macrophages and/or recruitment of peripheral macrophage-like cells (F4/80-positive) from outside the brain. Macrophages express CCR2 and thus, are capable of responding to a compromised entorhinal cortex via chemotaxis. Whether the observed increase in F4/80+ cell number indicates a homeostatic or pathologic response is not clear. APPswe/CCL2 mice demonstrate enhanced microglial numbers that are concurrent with increased extracellular Aβ deposition, that the authors postulate is due to an inability to effectively clear Aβ [34]. This may relate partially to the increased ApoE levels observed in APPswe/CCL2 mice produced by microglia and macrophages. If a similar mechanism is at play in the 3xTg-AD mouse model, this finding suggests a pathogenic role for these cells in initiating degeneration within the entorhinal cortex.
In summary, our results indicate a potential early role for inflammatory processes in the temporal and spatial evolution of AD pathogenesis. Because TNF-α and MCP-1 are produced specifically within the entorhinal cortex where human AD has been shown to arise, these molecules are likely playing an instrumental role in disease perpetuation. This work provides insight into the involvement of TNF-α and MCP-1 mediated inflammation in the temporal and spatial progression of early AD pathogenic events and may potentially herald new therapeutic targets. Our use of the 3xTg-AD model to assess these early events is unique, as all previous studies have examined inflammatory processes in the context of either amyloid or tau pathology, but not both. This transgenic mouse allows us to directly examine the dynamic interplay of inflammation, Aβ pathology, and tau dysfunction. Modulating TNF-α and MCP-1 function in future studies will elucidate how these inflammatory mediators influence the severity and progression of AD-related pathology and synaptic dysfunction.
List of abbreviations
AD: Alzheimer's disease
APP: Amyloid precursor protein
APPswe: Amyloid precursor protein Swedish mutation
PS1: Presenilin 1
Aβ : Beta-amyloid
TNF-α : Tumor necrosis factor-alpha
MCP-1: monocyte chemoattractant protein-1
Tg: Transgenic
PBS: Phosphate-buffered saline
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MCJ carried out the quantitative real-time PCR, 6E10 immunohistochemistry, F4/80 immunohistochemistry, stereology, experimental analysis and data interpretation, and prepared the manuscript. MAM performed tissue microdissection, brain sectioning, and 6E10 immunohistochemistry. SO and FML conceived the design of and generated the 3xTg-AD mouse model. HJF and WJB conceived the design of the study, aided in the preparation of the manuscript, and provided critical analysis of the manuscript.
Acknowledgements
The authors wish to thank Dr. Linda Callahan for immunohistochemistry, microscopy, and stereological advice. We also thank Landa Prifti for animal care and husbandry. This work was supported by NIH F31NS049995 to MCJ, NIH R01AG023593 to WJB and NIH R01AG020204 to HJF.
==== Refs
Selkoe DJ Alzheimer's disease is a synaptic failure Science 2002 298 789 791 12399581 10.1126/science.1074069
Nestor PJ Scheltens P Hodges JR Advances in the early detection of Alzheimer's disease Nat Med 2004 10 Suppl S34 41 15298007 10.1038/nrn1433
Braak H Braak E Neuropathological stageing of Alzheimer-related changes Acta Neuropathol (Berl) 1991 82 239 259 1759558 10.1007/BF00308809
Haroutunian V Purohit DP Perl DP Marin D Khan K Lantz M Davis KL Mohs RC Neurofibrillary tangles in nondemented elderly subjects and mild Alzheimer disease Arch Neurol 1999 56 713 718 10369312 10.1001/archneur.56.6.713
Haroutunian V Perl DP Purohit DP Marin D Khan K Lantz M Davis KL Mohs RC Regional distribution of neuritic plaques in the nondemented elderly and subjects with very mild Alzheimer disease Arch Neurol 1998 55 1185 1191 9740112 10.1001/archneur.55.9.1185
Braak H Braak E Staging of Alzheimer's disease-related neurofibrillary changes Neurobiol Aging 1995 16 271 8; discussion 278-84 7566337 10.1016/0197-4580(95)00021-6
Pennanen C Kivipelto M Tuomainen S Hartikainen P Hanninen T Laakso MP Hallikainen M Vanhanen M Nissinen A Helkala EL Vainio P Vanninen R Partanen K Soininen H Hippocampus and entorhinal cortex in mild cognitive impairment and early AD Neurobiol Aging 2004 25 303 310 15123335 10.1016/S0197-4580(03)00084-8
Kordower JH Chu Y Stebbins GT DeKosky ST Cochran EJ Bennett D Mufson EJ Loss and atrophy of layer II entorhinal cortex neurons in elderly people with mild cognitive impairment Ann Neurol 2001 49 202 213 11220740 10.1002/1531-8249(20010201)49:2<202::AID-ANA40>3.0.CO;2-3
Gomez-Isla T Price JL McKeel DWJ Morris JC Growdon JH Hyman BT Profound loss of layer II entorhinal cortex neurons occurs in very mild Alzheimer's disease J Neurosci 1996 16 4491 4500 8699259
deToledo-Morrell L Stoub TR Bulgakova M Wilson RS Bennett DA Leurgans S Wuu J Turner DA MRI-derived entorhinal volume is a good predictor of conversion from MCI to AD Neurobiol Aging 2004 25 1197 1203 15312965 10.1016/j.neurobiolaging.2003.12.007
Burns A Whitehouse P Arendt T Forsti H Alzheimer's disease in senile dementia: loss of neurones in the basal forebrain. Whitehouse, P., Price, D., Struble, R., Clarke, A., Coyle, J. and Delong, M. Science (1982), 215, 1237-1239 Int J Geriatr Psychiatry 1997 12 7 10 9050416 10.1002/(SICI)1099-1166(199701)12:1<7::AID-GPS484>3.0.CO;2-P
Davies CA Mann DM Sumpter PQ Yates PO A quantitative morphometric analysis of the neuronal and synaptic content of the frontal and temporal cortex in patients with Alzheimer's disease J Neurol Sci 1987 78 151 164 3572454 10.1016/0022-510X(87)90057-8
Hsia AY Masliah E McConlogue L Yu GQ Tatsuno G Hu K Kholodenko D Malenka RC Nicoll RA Mucke L Plaque-independent disruption of neural circuits in Alzheimer's disease mouse models Proc Natl Acad Sci U S A 1999 96 3228 3233 10077666 10.1073/pnas.96.6.3228
McGeer EG McGeer PL Inflammatory processes in Alzheimer's disease Prog Neuropsychopharmacol Biol Psychiatry 2003 27 741 749 12921904 10.1016/S0278-5846(03)00124-6
Li M Pisalyaput K Galvan M Tenner AJ Macrophage colony stimulatory factor and interferon-gamma trigger distinct mechanisms for augmentation of beta-amyloid-induced microglia-mediated neurotoxicity J Neurochem 2004 91 623 633 15485493 10.1111/j.1471-4159.2004.02765.x
Strohmeyer R Rogers J Molecular and cellular mediators of Alzheimer's disease inflammation J Alzheimers Dis 2001 3 131 157 12214082
Hanisch UK Microglia as a source and target of cytokines Glia 2002 40 140 155 12379902 10.1002/glia.10161
Oddo S Caccamo A Kitazawa M Tseng BP LaFerla FM Amyloid deposition precedes tangle formation in a triple transgenic model of Alzheimer's disease Neurobiol Aging 2003 24 1063 1070 14643377 10.1016/j.neurobiolaging.2003.08.012
Oddo S Caccamo A Shepherd JD Murphy MP Golde TE Kayed R Metherate R Mattson MP Akbari Y LaFerla FM Triple-transgenic model of Alzheimer's disease with plaques and tangles: intracellular Abeta and synaptic dysfunction Neuron 2003 39 409 421 12895417 10.1016/S0896-6273(03)00434-3
Vehmas AK Kawas CH Stewart WF Troncoso JC Immune reactive cells in senile plaques and cognitive decline in Alzheimer's disease Neurobiol Aging 2003 24 321 331 12498966 10.1016/S0197-4580(02)00090-8
Strohmeyer R Kovelowski CJ Mastroeni D Leonard B Grover A Rogers J Microglial responses to amyloid beta peptide opsonization and indomethacin treatment J Neuroinflammation 2005 2 18 16111494 10.1186/1742-2094-2-18
Braak H Braak E Staging of Alzheimer-related cortical destruction Int Psychogeriatr 1997 9 Suppl 1 257 61; discussion 269-72 9447446 10.1017/S1041610297004973
Streit WJ Microglia and Alzheimer's disease pathogenesis J Neurosci Res 2004 77 1 8 15197750 10.1002/jnr.20093
Tuppo EE Arias HR The role of inflammation in Alzheimer's disease Int J Biochem Cell Biol 2005 37 289 305 15474976 10.1016/j.biocel.2004.07.009
Afagh A Cummings BJ Cribbs DH Cotman CW Tenner AJ Localization and cell association of C1q in Alzheimer's disease brain Exp Neurol 1996 138 22 32 8593893 10.1006/exnr.1996.0043
in t' Veld BA Ruitenberg A Hofman A Launer LJ van Duijn CM Stijnen T Breteler MM Stricker BH Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease N Engl J Med 2001 345 1515 1521 11794217 10.1056/NEJMoa010178
Stewart WF Kawas C Corrada M Metter EJ Risk of Alzheimer's disease and duration of NSAID use Neurology 1997 48 626 632 9065537
Koenigsknecht-Talboo J Landreth GE Microglial phagocytosis induced by fibrillar beta-amyloid and IgGs are differentially regulated by proinflammatory cytokines J Neurosci 2005 25 8240 8249 16148231 10.1523/JNEUROSCI.1808-05.2005
Kitazawa M Oddo S Yamasaki TR Green KN LaFerla FM Lipopolysaccharide-induced inflammation exacerbates tau pathology by a cyclin-dependent kinase 5-mediated pathway in a transgenic model of Alzheimer's disease J Neurosci 2005 25 8843 8853 16192374 10.1523/JNEUROSCI.2868-05.2005
Zhao B Schwartz JP Involvement of cytokines in normal CNS development and neurological diseases: recent progress and perspectives J Neurosci Res 1998 52 7 16 9556025 10.1002/(SICI)1097-4547(19980401)52:1<7::AID-JNR2>3.0.CO;2-I
Blasko I Veerhuis R Stampfer-Kountchev M Saurwein-Teissl M Eikelenboom P Grubeck-Loebenstein B Costimulatory effects of interferon-gamma and interleukin-1beta or tumor necrosis factor alpha on the synthesis of Abeta1-40 and Abeta1-42 by human astrocytes Neurobiol Dis 2000 7 682 689 11114266 10.1006/nbdi.2000.0321
Liao YF Wang BJ Cheng HT Kuo LH Wolfe MS Tumor necrosis factor-alpha, interleukin-1beta, and interferon-gamma stimulate gamma-secretase-mediated cleavage of amyloid precursor protein through a JNK-dependent MAPK pathway J Biol Chem 2004 279 49523 49532 15347683 10.1074/jbc.M402034200
Del Villar K Miller CA Down-regulation of DENN/MADD, a TNF receptor binding protein, correlates with neuronal cell death in Alzheimer's disease brain and hippocampal neurons Proc Natl Acad Sci U S A 2004 101 4210 4215 15007167 10.1073/pnas.0307349101
Yamamoto M Horiba M Buescher JL Huang D Gendelman HE Ransohoff RM Ikezu T Overexpression of monocyte chemotactic protein-1/CCL2 in beta-amyloid precursor protein transgenic mice show accelerated diffuse beta-amyloid deposition Am J Pathol 2005 166 1475 1485 15855647
Perry RT Collins JS Wiener H Acton R Go RC The role of TNF and its receptors in Alzheimer's disease Neurobiol Aging 2001 22 873 883 11754994 10.1016/S0197-4580(01)00291-3
Lindberg C Selenica ML Westlind-Danielsson A Schultzberg M beta-Amyloid Protein Structure Determines the Nature of Cytokine Release From Rat Microglia J Mol Neurosci 2005 27 1 12 16055942 10.1385/JMN:27:1:001
Pratico D Uryu K Leight S Trojanoswki JQ Lee VM Increased lipid peroxidation precedes amyloid plaque formation in an animal model of Alzheimer amyloidosis J Neurosci 2001 21 4183 4187 11404403
|
16232318
|
PMC1276812
|
CC BY
|
2021-01-04 16:38:22
|
no
|
J Neuroinflammation. 2005 Oct 18; 2:23
|
utf-8
|
J Neuroinflammation
| 2,005 |
10.1186/1742-2094-2-23
|
oa_comm
|
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-501621612710.1186/1475-2875-4-50OpinionHost microsatellite alleles in malaria predisposition? Gaikwad Sonali [email protected] Richa [email protected] Nirbhay [email protected] Rajni [email protected] VK [email protected] Central Forensic Science Laboratory, National DNA Analysis Centre, 30 Gorachand Road, Kolkata-700014, West Bengal, India2 Malaria Research Institute, Department of Molecular Microbiology and Immunology, John Hopkins Bloomberg School of Public Health, E5144, 15N.Wolfe Street, Baltimore, MD 21205, USA2005 10 10 2005 4 50 50 12 6 2005 10 10 2005 Copyright © 2005 Gaikwad et al; licensee BioMed Central Ltd.2005Gaikwad et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Malaria is a serious, sometimes fatal, disease caused by Plasmodium infection of human red blood cells. The host-parasite co-evolutionary processes are well understood by the association of coding variations such as G6PD, Duffy blood group receptor, HLA, and beta-globin gene variants with malaria resistance. The profound genetic diversity in host is attributed to polymorphic microsatellites loci. The microsatellite alleles in bacterial species are known to have aided their survival in fatal environmental conditions. The fascinating question is whether microsatellites are genomic cushion in the human genome to combat disease stress and has cause-effect relationships with infections.
Presentation of the hypothesis
It is hypothesized that repeat units or alleles of microsatellites TH01 and D5S818, located in close proximity to beta-globin gene and immune regulatory region in human play a role in malaria predisposition. Association of alleles at aforesaid microsatellites with malaria infection was analysed. To overrule the false association in unrecognized population stratification, structure analysis and AMOVA were performed among the sampled groups.
Testing of hypothesis
Associations of microsatellite alleles with malaria infection were verified using recombination rate, Chi-square, and powerful likelihood tests. Further investigation of population genetic structure, and AMOVA was done to rule out the confounding effects of population stratification in interpretation of association studies.
Implication of the hypothesis
Lower recombination rate (θ) between microsatellites and genes implicated in host fitness; positive association between alleles -13 (D5S818), 9 (TH01) and strong susceptibility to Plasmodium falciparum; and alleles-12 (D5S818) and 6 (TH01) rendering resistance to human host were evident. The interesting fact emerging from the study was that while predisposition to malaria was a prehistoric attribute, among TH01 alleles; evolution of resistant allele-6 was a recent phenomenon, which could conceivably be driven by infection related selective forces. The host's microsatellite allelic associations with malaria infection were valid in the light of low genetic variance between sampled groups and no population stratification.
==== Body
Background
Infectious diseases influence or respond to levels of host genetic variation [1]. As a part of the co-evolutionary process, pathogens had also acquired factors for mitigation and longer survival in the host milieu [2]. Plasmodium parasite causes infection of human red blood corpuscles causing malaria. Of its four species, Plasmodium falciparum infection is the leading cause of mortality, but the changes in environment and human demography have altered the host-parasite interactions that have subsequently affected the disease spectrum [3]. Malaria provides the paradigm for a disease that has shaped the human genome through natural selection of protective genetic traits. Many molecular studies have documented the associations of human's coding gene polymorphisms such as the haemoglobin variants (Hb E, Hb C, Hb S, α- and β-thalassaemia), G6PD, membrane receptor (Duffy protein), blood group proteins, HLA (HLA-B53, DRB1*1302) and other immune regulatory region with malaria resistance [4,5]. However, human DNA harbours enormous diversity, maintained by neutral polymorphisms such as microsatellites (STR), which are the recombination 'hotspots' owing to repetitive length sequences. It is known that microsatellites present within Opa genes of prokaryotic bacterial species, namely, Neisseria gonorrhoeae, Hemophilus influenzae aid their survival in fatal environmental conditions [6]. However, it would be fascinating to explore whether high allelic diversity of microsatellite regions in human genome are genomic cushions that has cause-effect relationship aiding species survival against environmental stresses including large number of infections. This paper conjecture is tested and verified at microsatellite loci namely D5S818 and TH01, present in close proximity to human genes implicated in resistance (reproductive fitness) to malaria.
Methods
Investigations began after the approval from our Institutional Review Committee and the written consent of patients or the parents of the patients. Blood samples were collected from the patients including clinically positive for P. falciparum (n = 105) and Plasmodium vivax (n = 67) infections. The sickle cell anaemic (n = 72 represents 49 different families) and β-thalassaemic (n = 150 represents 100 different families) patients with no incidence of malarial infection were also sampled. In order to have an overall profile for resistance and susceptibility to malaria, normal healthy individuals (n = 174) with no personal history of Plasmodium infection and haemoglobin disorder were also studied. The sample collection was restricted to the eastern zone of India.
DNA was isolated from lymphocytes of collected blood specimens following organic extraction protocol [7]. Nested-PCR was used to detect the presence of malaria parasites based on its small subunit ribosomal RNA gene [8]. ARMS-PCR and PCR-RFLP methods were employed to diagnose and confirm the carriers of β-thalassaemia and sickle cell anaemia respectively [9,10]. A total fifteen microsatellite loci including TH01 and D5S818 were amplified following user's manual and reagents supplied with PowerPlex 16® System (Promega Corporation, Madison, USA). These were subsequently detected on 6% Long Ranger®acrylamide gel mounted on ABI 377 DNA Sequencer (Applied Biosystems, Foster City, CA). Genotypes were determined using Genescan™ (version 3.7) and PowerTyper™(version 3.7) software packages.
At first, an admixture model assuming correlated allele frequency between groups for inferring population structure implemented in structure program version 2.0 was employed using genotype data of 13 unlinked markers [11]. Analysis of molecular variance (AMOVA) was carried out on genotype data of unlinked markers between two groups using Arlequin [12]. The proposed hypothesis of this paper was tested by various statistical computations. Recombination rate (θ) as an index of linkage between two loci [13] was computed between the host's 5q31-33 immune regulatory region controlling P. falciparum infection level [14] and D5S818 microastellite locus; and between β-globin gene(11p15.5) and TH01 (11p15.5) microastellite locus. The region from β-globin gene to TH01 marker spans 3.83 cM (1 Mb = 1.168 cM). Allele frequencies at each locus in different patients and healthy individuals were estimated by gene count. Likelihood-based tests for allelic associations were performed using the established model [15]. Age of the microsatellite alleles was also calculated [16,17].
Results and discussion
Microsatellite alleles play a significant role in adaptation of bacteria and perhaps higher organisms to their ever-changing environments. On the downside, they have also been linked to human diseases, production of bile pigment bilirubin and neurotransmitters [6]. Genetic epidemiologists are interested in exploring whether the high allelic diversity in human host has association with large number of infections including malaria or it may be due to population stratification.
The evaluation of model-based clustering method showed no evidence of population substructure between sampled groups, which in turn has implication in detecting valid association between a putative candidate allele and malaria disease (Figure 1). Analysis of molecular variance (AMOVA) shows that most of the variance in the sampled population is attributable to within population group variation (98.5%) of the variance and between population group variation is less than 1.5%. The fixation index, FST for the whole sample is only 0.0146. The computed recombination rate (θ) between host's 5q31-33 and D5S818 equals 0.4; and between β-globin gene and TH01 was 0.06. These values were lower than 0.5 (cut off) indicating linkage between studied microsatellite markers and genes owing to small genetic distance (cM) as compared to earlier reports [18,19] that clearly depicted significant associations between widely separated loci (17–19 cM) on the same chromosome. The studied tetrameric microsatellite loci manifest varying frequencies of co-dominant alleles in global human populations. Although they are located within blocks of genes subject to selection, alleles numbering 7 to 16 at D5S818 and 4 to 9, 9.3, 10–11 and 13.3 at TH01 are considered selectively neutral. These alleles differentiated in frequency, age and evolutionary history because on arising they move through population either via genetic drift or being associated with positively selected marker, which may offers evolutionary benefits to the host. Allele frequencies at D5S818 locus in P. falciparum malaria patients and healthy individuals is presented in Figure 2, which showed the highest occurrence of allele-13 in P. falciparum patients and allele-12 in healthy individuals respectively. The distribution pattern in case of TH01 (Figure 3) showed allele-9 to be highly frequent in malaria positive isolates while allele-6 was predominant in the normal group with no clinical history of malaria. The Chi-square (χ2) test was applied to check the disproportionate distribution of alleles across two loci in patients and normal groups. Table 1 shows that the estimated values were significantly (P < 0.01) higher than the theoretical table value (9.210), thereby indicating unequal and significant differences in microsatellite allele distribution (frequencies) between P. falciparum patients and controls indicate linkage disequilibrium between closely mapped locus and the disease locus. However, weak associations between microsatellite alleles and P. vivax infection as well as haemoglobin variants (sickle cell anaemia and β-thalassaemia carriers) were observed. These results might be due to weak and mild malarial infection by P. vivax parasite and innate resistance to malaria by haemoglobin disorders. Therefore, the weakly associated samples were excluded from further analysis. Further, statistical analysis employed Likelihood-test, which considered microsatellite loci as bi-allelic markers (associated and non-associated alleles) that are represented in 2 × 2 contingency table. Here, an allele with the higher χ2 value than the table value (at P < 0.01) was taken to be associated. In P. falciparum infected group, the likelihood estimate of allele-13 at D5S818 was found to be significantly greater than allele-12 (see Table 1), and that of allele-9 at locus TH01 was found to be greater than allele-6 (non-associated allele with malaria) indicating a positive association of allele-9 with the disease expression.
Figure 1 Triangle plot showing estimates of membership coefficient (Q) for each individual by sampled groups, analysed under admixture model, assuming correlated allele frequencies.
Figure 2 Distribution of allele – 12, 13 and others at microsatellite locus D5S818 in positive (+) and negative (-) samples of P. falciparum Malaria.
Figure 3 Distribution of allele – 6, 9 and others at microsatellite locus TH01 in positive (+) and negative (-) samples of P. falciparum malaria.
Table 1 Chi square and Likelihood estimates of alleles – 12, 13 at microsatellite D5S818; alleles – 6 and 9 at microsatellite TH01 markers (Significance level = 1%) and malaria susceptibility and resistance.
Malaria parasite Chi-Square values Likelihood-test estimates
D5S818
Normal (allele-12) Disease (allele-13) Associated with infection (allele-13) Non-associated with infection (allele-12)
P. falciparum 184.65 109.04 3.56 2.22 × 10-4
TH01
Normal (allele-6) Disease (allele-9) Associated with infection (allele-9) Non-associated with infection (allele-6)
P. falciparum 140.542 128.922 1.75 9.61 × 10-2
The prehistoric imprint of host susceptibility to malaria disease was demonstrated by allele-9 (TH01) and allele-13 (D5S818) because these alleles are estimated to be older (age) than allele – 6 and allele-12. These findings further strengthen the role of host microsatellite alleles in malarial susceptibility as well as disease resistance since people with one version of the gene tend to be resistant to malaria while the other variant make the host susceptible. Furthermore, allele distribution patterns at locus TH01 in worldwide populations revealed high frequencies of allele-6 than allele-9 in European populations from Germany, Italy, and Slovakia [20], probably because of zero incidence of fatal falciparum malaria and negligible selection pressure. On the other hand, allele-9 in Indian populations [21] suggests primeval imprint of malaria susceptibility and genetic drift. Further, haemoglobin mutations evolved as presumed protective mechanisms against the infection.
Conclusion
A large number of STR loci dispersed in human genome led to postulation that they act as genomic cushion to combat stress of microbial infections. The study provided significant evidence that an individuals' genotype (microsatellite alleles) is a product of host interaction with P. falciparum infection. Thus, positive associations between allele-6 (TH01), allele-12 (D5S818) and malaria-negative (healthy) individuals demonstrate their role in genetic resistance whereas allele-9 (TH01) and allele-13 (D5S818) in P. falciparum infected individuals illustrate genetic predisposition towards the disease. Furthermore, predominance of allele-9 at TH01 (AATG repeat unit) in the individuals inhabiting malaria endemic area suggests genetic predisposition towards malaria as an archaic phenomenon. The host reproductive fitness has evolved during recent times in the form of resistant traits viz., haemoglobin variants (sickle cell mutation, β-thalassaemia) and allele-6 (164 bp) at TH01 by deletion of 12 bp from allele-9 (176 bp); both have significantly altered the parasite dynamics. The above findings are further supported by younger ages of microsatellite alleles associated with malaria resistance. A more recent discovery in a low-risk malarial zone shows total absence of allele-9, predominance of lower repeat units at TH01 and lack of manifestation of malaria in Jarawa (an aboriginal tribal population of Andaman and Nicobar Island). However, the non-tribal inhabitants (migrants) of this island have higher incidence of malaria, thus substantiating further implication of allele-9 in malaria susceptibility (Kashyap et al. unpublished data). The above findings are valid in the absence of population genetic structure.
Authors' contributions
VKK is the main investigator who has conceptualized the entire study. RA and SG performed the experiments, collated the data, performed statistical analysis and drafted the manuscript. NK and RT are the co-investigators who reviewed and improved the general presentation of this manuscript. All authors read and approved the final manuscript
Acknowledgements
We are thankful to the Bureau of Police Research and Development, Ministry of Home Affairs, Government of India, for financial support to carry out this research work. We also acknowledge all the blood donors and the co-operation of Drs. S.L Kate (Pune), R. Wankhede (Nagpur), R.K Nath, D.K Ganguly, B. Daripa, S.S.N Rao (Jamshedpur), D. GhoshDastidar and Ms. Neeta Sarkar (Kolkata) in the collection of blood. Ms. Seema Singh provided support in preparation of the manuscript.
==== Refs
Haldane JBS The rate of mutation of human genes Hereditas 1949 35 267 273
Hastings IM Paget-McNicol S Saul A Can mutations and selection explain virulence in human P. falciparum infections? Malar J 2004 3 2 14992697 10.1186/1475-2875-3-2
Jobling MA Hurles M Tyler-Smith C Health implications of our evolutionary heritage Human Evolutionary Genetics Origins, peoples and Disease 2004 New York: Garland Science, Taylor and Francis group 439 471
Mourant AE Kopec AC Domaniewska-Sobczak K Blood groups and Diseases 1978 Oxford: Oxford University Press
Hill AV The immunogenetics of resistance to malaria Proc Assoc Am Physicians 1999 111 272 277 10417733 10.1046/j.1525-1381.1999.99234.x
Moxon ER Wills C DNA microsatellites: agents of evolution? Sci Am 1999 280 94 99 9891422
Sambrook J Fritsch EF Maniatis T Molecular cloning: a laboratorymanual 1989 2 Cold Spring Harbor: Cold Spring Harbor Laboratory Press
Snounou G Viriyakosol S Zhu XP Jarra W Pinheiro L Rosario VE Thaithong S Brown KN High sensitivity of detection of human malaria parasites by the use of nested polymerase chain reaction Mol Biochem Parasitol 1993 61 315 320 8264734 10.1016/0166-6851(93)90077-B
Varawalla NY Old JM Sarkar R Venkatesan R Weatherall DJ The spectrum of β thalassemia mutations on the Indian subcontinent: the basis for prenatal diagnosis Br J Haematol 1991 78 242 247 2064964
Kashyap VK Ashma R Microsatellite diversity in HbS carrier and normal individuals of tribal populations of malaria infested regions Int Hum Genet 2003 3 197 203
Pritchard JK Stephens M Donnelly P Inference of population structure using multilocus genotype data Genetics 2000 155 945 959 10835412
Schneider S Roessli D Excoffier L Arlequin ver. 2.000 A software for population genetics data analysis 2000 Genetics and Biometry Laboratory, University of Geneva, Switzerland
Haldane JBS The combination of linkage values and the calculation of distance between the loci of linked factors J Genet 1919 8 299 309
Rihet P Traore Y Abel L Aucan C Traore-Leroux T Fumoux F Malaria in humans: Plasmodium falciparum blood infection levels are linked to chromosome 5q31-q33 Am J Hum Genet 1998 63 498 505 9683598 10.1086/301967
Terwilliger JD A powerful likelihood method for the analysis of linkage disequilibrium between trait loci and one or more polymorphic marker loci Am J Hum Genet 1995 56 777 787 7887434
Kimura M Ohta T The age of a neutral mutant persisting in a finite population Genetics 1973 75 199 212 4762875
Slatkin M Rannala B Estimating allele age Annu Rev Genomics Hum Genet 2000 1 225 249 11701630 10.1146/annurev.genom.1.1.225
Payseur BA Nachman MW Microsatellite variation and recombination rate in the human genome Genetics 2000 156 1285 1298 11063702
Zapata C Rodriguez S Visedo G Sacristan F Spectrum of nonrandomassociation between microstellite loci on human chromosome 11p15 Genetics 2001 158 1235 1251 11454771
The distribution of human DNA-PCR polymorphisms
Deka R Shriver MD Yu LM Heidreich EM Jin L Zhong Y McGarvey ST Agrawal SS Bunker CH Miki T Hundrieser J Yin SJ Raskin S Barrantes R Ferrell RE Chakraborty R Genetic variation at twenty-three microsatellite loci in sixteen human population J Genet 1999 78 99 121
|
16216127
|
PMC1276813
|
CC BY
|
2021-01-04 16:37:30
|
no
|
Malar J. 2005 Oct 10; 4:50
|
utf-8
|
Malar J
| 2,005 |
10.1186/1475-2875-4-50
|
oa_comm
|
==== Front
Nutr Metab (Lond)Nutrition & Metabolism1743-7075BioMed Central London 1743-7075-2-301624204210.1186/1743-7075-2-30ReviewTargeting energy metabolism in brain cancer: review and hypothesis Seyfried Thomas N [email protected] Purna [email protected] Biology Department, Boston College, Chestnut Hill, MA 02467, USA2005 21 10 2005 2 30 30 22 8 2005 21 10 2005 Copyright © 2005 Seyfried and Mukherjee; licensee BioMed Central Ltd.2005Seyfried and Mukherjee; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Malignant brain tumors are a significant health problem in children and adults and are often unmanageable. As a metabolic disorder involving the dysregulation of glycolysis and respiration, malignant brain cancer is potentially manageable through changes in metabolic environment. A radically different approach to brain cancer management is proposed that combines metabolic control analysis with the evolutionarily conserved capacity of normal cells to survive extreme shifts in physiological environment. In contrast to malignant brain tumors that are largely dependent on glycolysis for energy, normal neurons and glia readily transition to ketone bodies (β-hydroxybutyrate) for energy in vivo when glucose levels are reduced. The bioenergetic transition from glucose to ketone bodies metabolically targets brain tumors through integrated anti-inflammatory, anti-angiogenic, and pro-apoptotic mechanisms. The approach focuses more on the genomic flexibility of normal cells than on the genomic defects of tumor cells and is supported from recent studies in orthotopic mouse brain tumor models and in human pediatric astrocytoma treated with dietary energy restriction and the ketogenic diet.
gliomavascularitycaloric restrictionketone bodiesmetabolic control analysisangiogenesisapoptosisinflammationWarburg
==== Body
Introduction
The world-wide incidence of malignant brain tumors may be increasing in both children and the elderly [1-4]. Regardless of these ominous findings, the standard therapies for malignant gliomas (surgical resection and radiation) are basically the same today as they have been for over five decades [5-7]. While these therapies may retard glioma growth over the short term, they can facilitate glioma recurrence and enhance growth rate over the longer term through alterations in morphogenetic fields [8,9]. Chemotherapy has little long-term benefit on most malignant gliomas and is often associated with adverse effects that diminish the length or quality of life [7,10]. The therapeutic targeting of brain tumor-associated mutations may also be problematic as most tumor mutations arise as epiphenomena of tissue disorganization and their relationship to causality is uncertain [8,11,12]. Despite modest gains in survival with temozolomide chemotherapy, few things are more certain in the brain tumor field than the impotence of most current therapies [2,7,13]. Hence, new approaches are needed that can provide long-term management of malignant brain tumors while permitting a decent quality of life.
Metabolic Control Analysis
Metabolic control analysis evaluates the degree of flux in metabolic pathways and can be used to analyze and treat complex diseases [14-16]. The approach is based on findings that compensatory genetic and biochemical pathways regulate the bioenergetic potential of cells and ultimately the phenotype [14,15,17]. As rate-controlling enzymatic steps in biochemical pathways are dependent on the metabolic environment of the physiological system, the management of disease phenotype depends more on the flux of the entire system than on the expression of any specific gene or enzyme alone [16-19]. In other words, complex disease phenotypes can be managed through self-organizing networks that display system wide dynamics involving glycolysis and respiration. Global manipulations of these metabolic networks can restore orderly adaptive behavior to widely disordered states involving complex gene-environmental interactions [15,17,20,21].
As abnormal energy metabolism and biological chaos are characteristics of brain tumors [8,22-24], the general principles of metabolic control analysis can be effective for brain cancer management. This hypothesis is based on the known differences in energy metabolism between normal and neoplastic brain cells. As long as brain tumors are provided a physiological environment conducive for their glycolytic energy needs, they will survive; when this environment is restricted or abruptly changed, they will either growth arrest or perish. Here we describe how new therapeutic approaches, which lower circulating glucose and elevate ketone bodies (acetoacetate and β-hydroxybutyrate), target brain tumors while enhancing the metabolic efficiency of normal neurons and glia.
Energy Metabolism in Normal Brain Cells
To manage brain cancer through metabolic targeting it is necessary to consider energy metabolism in the normal orthotopic tissue. Figure 1 illustrates some of the metabolic pathways discussed here. Under normal physiological conditions, the mature brain derives almost all of its energy from the aerobic oxidation of glucose [15,25,26]. The glucose transporter, GLUT-1, is enriched in the brain capillary endothelial cells and mediates the facilitated diffusion of glucose through the blood brain barrier. Most of the glucose is metabolized to pyruvate, which enters the mitochondria of neurons and glia and is converted to acetyl-CoA before entering the TCA cycle. Only about 13% of glycolytic pyruvate is converted to lactate under normal conditions [26]. Fatty acids are attached to lipoproteins and do not pass the blood brain barrier as fuel substrates, though octanoate may be an exception [26-28]. It is also unlikely that lactate is used for energy metabolism in adult brain, but this remains somewhat controversial [26,29-31]. Hence glucose is the primary, if not exclusive, brain metabolic fuel under normal physiological conditions.
Figure 1 Perspectives on the metabolic management of brain cancer through a dietary reduction of glucose and elevation of ketone bodies. A dietary reduction in circulating glucose will increase ketone utilization for energy in normal neurons and glia. This will induce an energy transition from glycolysis to respiration. Cancer cells however, may be unable to transition from glucose to ketones due to alterations in mitochondrial structure or function (dashed lines). The double slash indicates a disconnection between glycolysis and respiration according to the Warburg hypothesis. Abbreviations: GLUT-1 (glucose transporter), MCT-1 (monocarboxylate transporter), SCOT (succinyl-CoA-acetoacetate-CoA transferase), β-OHB (β-hydroxybutyrate), β-HBDH (β-hydroxybutyrate dehydrogenase).
While glucose is the preferred energy substrate, neurons and glia will metabolize ketones for energy under fasting-induced reductions of blood glucose. This is a conserved physiological adaptation to prolonged food restriction and evolved to enhance survival and maintain adequate brain function while sparing proteins [15,27,32-34]. Ketone bodies, consisting of acetoacetate, and β-hydroxybutyrate (β-OHB) are derived from fat catabolism in the liver and their concentration in blood is inversely related to that of glucose [21,35-37]. Ketone bodies are transported into the brain through the blood-brain barrier monocarboxylic transporters (MCTs), whose expression is regulated in part by circulating ketone and glucose levels [30,31,34,38-41]. β-OHB is the predominant blood ketone body and is rapidly oxidized to acetyl-CoA in the mitochondria through an enzymatic series involving 3-hydroxybutyrate dehydrogenase, SCOT (succinyl-CoA-acetoacetate-CoA transferase), and mitochondrial acetyl-CoA thiolase [33,34,42,43]. Acetone is a non-enzymatic byproduct of ketone body synthesis and is largely excreted in the urine or exhaled from the lungs [34,44].
Although the levels of glucose and ketones in brain are proportional to their levels in blood [34,45], the adult brain does not usually metabolize ketone bodies for energy unless blood glucose levels are reduced [36]. Therapeutic efficacy of the ketogenic diet is best when coupled to dietary energy restriction under which conditions circulating glucose levels are gradually reduced in conjunction with elevations of ketone bodies [15,21]. The emphasis here is on the term "gradual", as ketone bodies cannot be used for energy following acute hypoglycemia [46]. The situation is different, however, for in vitro preparations where neuronal glial interactions are disrupted and the blood brain barrier is absent [47,48]. The gradual transition from glucose to ketone bodies for energy in vivo requires a flexible genome for the coordinated integration of multiple metabolic networks according to principles of metabolic control analysis [15,36].
Ketone bodies are more energetically efficient than either pyruvate or fatty acids because they are more reduced (greater hydrogen/carbon ratio) than pyruvate and do not uncouple the mitochondrial proton gradient as occurs with fatty acid metabolism [14]. In contrast to glucose, ketone bodies by-pass cytoplasmic glycolysis and directly enter the mitochondria where they are oxidized to acetyl-CoA [44,49]. The amount of acetyl-CoA formed from ketone body metabolism is also greater than that formed from glucose metabolism [50]. This increases TCA cycle metabolites (from citrate to α-ketoglutarate) while reducing the mitochondrial NAD couple, [NAD+]/[NADH], and increasing the mitochondrial Q couple [Q]/[QH2] [14,50]. The difference between these couples increases the redox span between the NADH dehydrogenase complex (site I), and the CoQH2-cytochrome C reductase (site III) thus enhancing the mitochondrial proton gradient [14]. This enhances the energy available from the hydrolysis of ATP, ΔG'ATP, the cell's key energy reserve generated through the mitochondrial Fl ATPase [14,16]. Remarkably, the ketone body-induced increase in the ΔG'ATP is also accomplished using less oxygen [48,50]. These and other findings led Veech to designate ketone bodies as a "super fuel" [14].
Ketones and Free Radicals
In addition to increasing ATP production while reducing oxygen consumption, ketone body metabolism can also reduce production of damaging free radicals [14,16,48]. The semiquinone of Q, the half reduced form, spontaneously reacts with oxygen and is the major source of mitochondrial free radical generation [14,51]. Oxidation of the Q couple reduces the amount of the semiquinone form thus decreasing superoxide production [14]. Since the cytosolic free NADP+/NADPH concentration couple is in near equilibrium with the glutathione couple, ketone body metabolism will increase the reduced form of glutathione thus facilitating destruction of hydrogen peroxide [14]. The reduction of free radicals through ketone body metabolism will also reduce tissue inflammation provoked by reactive oxygen species. Thus, ketone bodies are not only a more efficient metabolic fuel than glucose, but also possess anti-inflammatory potential.
Energy Metabolism in Brain Tumors
In contrast to normal brain that oxidizes glucose as well as ketone bodies for energy, malignant brain tumors from either humans or animal models lack metabolic flexibility and are largely dependent on glucose for energy [21,23,52-58]. Enhanced glycolysis produces excess lactic acid that can return to the tumor as glucose through the Cori cycle [59] (Figure 1). Although some neural tumors metabolize ketone bodies, this metabolism is largely for lipid synthesis rather than for energy production [60,61]. Many tumors also have reduced activity of SCOT, the rate-controlling step for utilizing β-OHB as a respiratory fuel [42,62,63]. Consistent with these observations, we recently found that β-OHB could rescue normal mouse astrocytes under low glucose conditions, but could not rescue mouse astrocytoma cells [20]. Although glutamine may provide energy to some non-neural tumors, glutamine stimulates glycolysis in C6 rat glioma cells and may not serve as a direct respiratory fuel [64]. Hence brain tumors, like most malignant tumors, depend heavily on glucose and glycolysis for their metabolic energy and are generally unable to metabolize β-OHB for energy.
In addition to glycolytic dependence, most tumors including brain tumors, express abnormalities in the number and function of their mitochondria [65,66]. Such abnormalities would prevent the bioenergetic utilization of ketone bodies, which require functional mitochondria for their oxidation [47]. Warburg originally emphasized that the high glycolytic rate of tumors resulted from diminished or disturbed respiration [67,68]. While most cells die from damaged respiration, those cells that can enhance and modify their anaerobic glycolysis in response to respiratory damage will survive and form tumors [67,68]. Later studies in a variety of neural and non-neural tumor systems showed that these respiratory disturbances involve abnormalities in TCA cycle components, alterations in electron transport, and deficiencies in oxidative phosphorylation [23,55,69-71]. Mitochondrial DNA mutations, however, may not be involved [12]. Structural defects of the inner mitochondrial membrane, that would reduce or dissipate the proton motive gradient, could also prevent normal ATP production despite the appearance of oxidative metabolism, i.e., oxygen consumption and CO2 production [70,72,73]. Considered together, these findings indicate that brain tumors suffer from reduced respiratory capacity coupled to an increased glycolysis and lactic acid production.
Although persistent aerobic glycolysis (glycolysis in the presence of oxygen) or defects in the Pasteur effect (reduction of glycolysis in the presence of oxygen) are characteristics of many tumors, Warburg considered these phenomena too labile or too dependent on environmental conditions to be reliable indicators of tumor metabolism [68]. Rather, he emphasized the importance of defects in the coordination of glycolysis with respiration. The latency between tumor initiation and progression was considered the period necessary to disconnect respiration from glycolysis [68]. Recent studies in human glioblastoma cells suggest that this disconnect involves activation of the Akt oncogene rendering cancer cells dependent on aerobic glycolysis for continued growth and survival [74]. This is consistent with the Warburg hypothesis that the increased glycolysis of tumor cells occurs gradually in order to compensate for respiratory failure [68]. In contrast to normal brain cells, in which glycolysis and respiration are tightly coupled, tumor cells are defective in their ability to integrate energy metabolism between glycolysis and respiration [71]. It is these defects that will render brain tumor cells vulnerable to metabolic targeting through metabolic control analysis. Support for this possibility comes from studies with the ketogenic diet and dietary energy restriction.
Dietary Energy and Brain Cancer
The Ketogenic Diet
In 1995, Nebeling and coworkers attempted the first nutritional metabolic therapy for human malignant brain cancer using the ketogenic diet [75]. The ketogenic diet (KD) is a high fat low carbohydrate diet that has been used for decades as an effective therapy for refractory seizures in children [15,76,77]. The objective of the Nebeling study was to shift the prime substrate for energy metabolism from glucose to ketone bodies in order to disrupt tumor metabolism while maintaining the nutritional status of patients [75].
The patients in this landmark clinical study included two female children with nonresectable advanced grade brain tumors (anaplastic astrocytoma stage IV, and cerebellar astrocytoma stage III) [75]. Measurable tumor remained in both subjects following extensive radiation and chemotherapy. Although severe life threatening adverse effects occurred from the radiation and chemotherapy, both children responded remarkably well to the KD and experienced long-term tumor management without further chemo or radiation therapy [75]. Indeed, one of the patients is still alive and well at the time of this writing (Nebeling, personal communication). Positron Emission Tomography with fluro-deoxy-glucose (FDG-PET) also showed a 21.8% reduction in glucose uptake at the tumor site in both subjects on the KD [75].
Despite the logic of these studies and the dramatic findings, no further human studies or clinical trials have been conducted on the therapeutic efficacy of the KD for brain cancer. The reason for this is not clear but may reflect a preference of the major Brain Tumor Consortia for using "hand-me-down" drug therapies from other cancer studies rather than exploring more effective biological or non-chemotherapeutic approaches [7]. This is unfortunate as our recent findings in brain tumor animal models show that the therapeutic potential of the restricted KD, involving reduced glucose and elevated β-OHB, is likely to be greater than that for any current brain tumor chemotherapy [20]. Moreover, the KD would eliminate or reduce the need for adjuvant anticonvulsant and steroidal medications for brain tumor patients as the KD has antiepileptic and anticonvulsant effects and, when restricted in caloric intake, will naturally elevate circulating glucocorticoid levels [15,77-80]. These findings suggest that the KD would be an effective multifactorial diet therapy for malignant brain cancer and should be considered as an alternative therapeutic option.
Dietary Energy Restriction
The findings of the Nebeling group were recently confirmed in a series of orthotopic mouse brain tumor models treated with the KD and dietary energy restriction [21,37,81,82]. As with the KD, dietary energy restriction (DR) reduces glucose and elevates ketone bodies. Indeed, the DR-induced inhibition of brain tumor growth is directly correlated with reduced levels of glucose and elevated levels of ketone bodies [20,21]. This energy transition contributes to the powerful anti-angiogenic effects of DR [81,82]. As a natural dietary therapy, DR improves health, prevents tumor formation, and reduces inflammation [83-88]. DR also improves mitochondrial respiratory function and glutathione redox state in normal cells [87,89]. Thus, DR naturally inhibits glycolysis and tumor growth while enhancing the health and vitality of normal cells and tissues.
The anti-angiogenic effects of DR arise from reduced tumor energy metabolism [21,37,81,82]. This is important since the angiogenic properties of most human gliomas are closely linked to metabolic activity [90]. Previous studies showed that the antitumor effects of DR result from caloric restriction per se and not from the restriction of any specific dietary component such as proteins, vitamins, minerals, fats, or carbohydrates [21,88,91,92]. DR or fasting can also reduce cerebral blood flow and oxygen consumption that would further stress brain tumor cells already weakened from reduced glucose levels [33]. Besides reducing angiogenesis, DR also increases brain tumor apoptosis [81,82]. The proapoptotic effects of DR occur in large part from reduced glycolytic energy that most tumors rely upon for growth.
A reduction in glycolytic energy would reduce lactate levels and hydroxyl radical production. This is important since lactate and hydroxyl radicals enhance tumor inflammation as well as cytokine production (tumor necrosis factor α, interleukin-6, and -1β) through glial activation (microglia and astroglia) [93,94]. DR also reduces inflammation and the inflammatory properties of macrophages, while enhancing their phagocytic activities [86,95]. An uncoupling of the detrimental inflammatory properties of tumor associated macrophages from their beneficial phagocytic properties (to remove tumor cell corpses) is considered essential for the eventual management of brain cancer [8]. Hence diet therapies, which lower glucose availability and elevate ketone bodies, can reduce brain tumor growth through integrated anti-inflammatory, anti-angiogenic and pro-apoptotic mechanisms.
Metabolic Control of Brain Cancer: An Evolutionary Perspective
Based on the differences in energy metabolism between normal brain cells and brain tumor cells, we propose a radically different approach to brain cancer management that combines metabolic control analysis with the evolutionarily conserved capacity of normal cells to survive extreme shifts in physiological environment. The adaptation to environmental extremes is conserved within the genome according to the ecological instability theory of Potts [96]. Consequently, only those cells with a flexible genome will be capable of surviving abrupt changes in metabolic landscape. Cells with genomic defects, which would limit flexibility, should be less adaptable to metabolic stress and therefore vulnerable to elimination through principles of metabolic control analysis. This strategy focuses more on the genetic capabilities of normal cells than on the genetic defects of tumor cells.
As a metabolic disorder involving the dysregulation of glycolysis and respiration, brain cancer is potentially manageable through abrupt changes in metabolic environment. Significant cellular energy is used to maintain the activity of transmembrane ion pumps (the Na+, K+-ATPase and Ca2+ and Mg2+ ATPases) [29,97,98] (Figure 1). The amount of energy needed to maintain pump function is also greater than that needed for mitosis, which is largely dependent on glycolytic energy [70,97]. Despite differences in membrane potential, most cells have a constant ΔG' of about -57 kJ/mol [14]. According to Veech this is "the still point in the turning world" and if cells cannot maintain this useable ATP they lose potassium, gain sodium and calcium, swell, and eventually die [16]. In other words, regardless of whether the cell is a normal neuron, a glial cell, or even a transformed tumor cell, its survival depends on maintaining an adequate ΔG' of ATP hydrolysis. Tumor cells with limited genomic flexibility should therefore be less capable than normal cells in utilizing alternative energy substrates to maintain their ΔG' of ATP hydrolysis.
The energy used to maintain pump function and cell viability in normal brain cells comes from either glycolysis or aerobic respiration [29,99]. In the case of C6 glioma cells and most brain tumors for that mater, this energy is mostly derived from glycolysis [29,99]. This would then render brain tumor cells vulnerable to reductions in circulating glucose levels as these mutant cells would have difficulty oxidizing alternative fuels (ketone bodies) through respiration. While some brain tumor cells may survive through up-regulation of their glucose transporters [100,101], most will either perish or reduce their growth. Direct support for this hypothesis comes from our recent findings in experimental mouse brain tumor models and from those of Nebeling and co-workers in human pediatric astrocytoma [20,37,75,81].
The widely held notion that brain tumor cells are somehow hardy or tough and resistant to death (programmed or nonprogrammed) may be a misconception. How can brain tumor cells, or any tumor cell for that matter, that have multiple types and kinds of genetic mutations be more fit and hardy than normal cells that possess a flexible genome and can easily transition between glycolysis and respiration for energy maintenance? The notion that tumor cells are more versatile than nontumor cells is also illogical in the context of evolutionary biology. While knowledge of tumor-associated mutations and genomic instability is of considerable academic interest, this information has produced no new or effective clinical therapies for brain tumors. Regardless of when or how genomic defects become involved in the initiation or progression of brain tumors, these defects can be exploited for the metabolic destruction of the tumor.
Normal cells evolved to survive extremes in metabolic and physiological environment due largely to oscillatory changes in the physical environment [96]. Moreover, the ability to adapt to extreme environmental stress is retained within the normal flexible genome [19,102]. It is this flexibility that allows normal brain cells to transition from glucose to β-OHB for energy under reduced energy availability. Due to accumulated nuclear genetic mutations and respiratory defects, brain tumor cells will be less adaptable than normal cells to abrupt changes in metabolic environment and can be either destroyed outright or isolated metabolically from normal cells. Hence, the genomic and metabolic flexibility of normal brain cells can be used to target indirectly the genetically defective and less metabolically flexible brain tumor cells.
According to this hypothesis, a novel strategy can be suggested for human brain tumor management that enhances the respiratory potential of normal brain cells while metabolically targeting the tumor cells. The approach would involve a sequential series of therapeutic steps and should be effective against any primary or secondary brain tumor regardless of cell of origin, anatomical location, or histological grade. Step one would lower circulating glucose levels and elevate circulating β-OHB levels through diet therapies or ketone body supplementation. Glucose ranges between 3.0–3.5 mM (55–65 mg/dl) and β-OHB ranges between 4–5 mM should be effective for tumor management in most patients. The conditions for these parameters have been described previously for children and adults and can be adjusted on a case-by-case basis [34,35,75,103]. These values are also well within normal physiological ranges and will have antiangiogenic and proapoptotic effects causing metabolic isolation and significant tumor shrinkage that can be assessed from imaging analysis. Reduced glucose and elevated ketones could also antagonize tumor cachexia as previously mentioned [82].
Step two would involve surgical resection if necessary. Smaller brain tumors with reduced vascularity and clearly circumscribed boundaries should be easier to resect than larger brain tumors with poorly circumscribed boundaries and extensive vascularization [104]. The diet therapy could also be adjusted following surgery to facilitate healing and to maintain metabolic pressure on any surviving tumor cells.
Finally, step three could involve the use of either conventional or novel targeted therapies. While these therapeutic approaches might have little if any long-term benefit on malignant brain tumor management if used initially, they could be highly effective following the step one strategy after the tumor cells are weakened and metabolically isolated from the physiologically strengthened normal brain cells. Moreover, glycolysis inhibitors that would adversely affect normal cells might also be more effective following the first two steps of the proposed therapeutic strategy [105-109]. It is also possible that carefully executed weight cycling strategies could maintain metabolic pressure on surviving tumor cells and facilitate their eradication or growth retardation [110]. In addition to global diet therapies, more specific amino acid restrictions could also be effective in eliminating surviving tumor cells [111].
The objective of this new brain tumor therapeutic approach is to consistently change the physiological and metabolic environment of the tumor and the host. Only those cells with a normal flexible genome, honed through millions of years of environmental forcing and variability selection, are expected to survive extreme shifts in metabolic environment. Indeed, extreme conditions of survival and fitness will test the limits of a cell population's persistence in any given location over time [96]. While some brain tumor cells might survive under one restricted environment or another, it is unlikely that all tumor cells will survive all restricted environments. In other words, it is the theory of Potts applied with sustained pressure to the entire population of normal and neoplastic brain cells. We predict that this therapeutic approach will be more successful than current approaches because it is based on the principles of evolutionary biology and metabolic control analysis.
Abbreviations
KD, ketogenic diet; SCOT, succinyl-CoA-acetoacetate-CoA transferase; β-OHB, β-hydroxybutyrate; TCA, tricarboxylic acid.
Acknowledgements
This work was supported by grants from the NIH (HD39722 and CA102135) and from the American Institute of Cancer Research. The author would like to thank George Cahill, Richard Veech, and Linda Nebeling for comments and Michael Kiebish and John Mantis for technical assistance.
==== Refs
Kaiser J No meeting of minds on childhood cancer Science 1999 286 1832 1834 10610570 10.1126/science.286.5446.1832
Lowry JK Snyder JJ Lowry PW Brain tumors in the elderly: recent trends in a Minnesota cohort study Arch Neurol 1998 55 922 928 9678309 10.1001/archneur.55.7.922
Kaatsch P Rickert CH Kuhl J Schuz J Michaelis J Population-based epidemiologic data on brain tumors in German children Cancer 2001 92 3155 3164 11753995 10.1002/1097-0142(20011215)92:12<3155::AID-CNCR10158>3.0.CO;2-C
Jukich PJ McCarthy BJ Surawicz TS Freels S Davis FG Trends in incidence of primary brain tumors in the United States, 1985–1994 Neuro-oncol 2001 3 141 151 11465394 10.1215/S1522851700000557
Zimmerman HM The nature of gliomas as revealed by animal experimentation Amer J Pathol 1955 31 1 29 13218129
Chang SM Parney IF Huang W Anderson FA JrAsher AL Bernstein M Lillehei KO Brem H Berger MS Laws ER Patterns of care for adults with newly diagnosed malignant glioma Jama 2005 293 557 564 15687310 10.1001/jama.293.5.557
Fisher PG Buffler PA Malignant gliomas in 2005: where to GO from here? Jama 2005 293 615 617 15687318 10.1001/jama.293.5.615
Seyfried TN Perspectives on brain tumor formation involving macrophages, glia, and neural stem cells Perspect Biol Med 2001 44 263 282 11370160
Sonnenschein C Soto AM The Society of Cells: Cancer and the Control of Cell Proliferation 1999 New York: Springer-Verlag
Shapiro WR Current therapy for brain tumors: back to the future Arch Neurol 1999 56 429 432 10199330 10.1001/archneur.56.4.429
Sonnenschein C Soto AM Somatic mutation theory of carcinogenesis: why it should be dropped and replaced Mol Carcinog 2000 29 205 211 11170258 10.1002/1098-2744(200012)29:4<205::AID-MC1002>3.0.CO;2-W
Kiebish MA Seyfried TN Absence of pathogenic mitochondrial DNA mutations in mouse brain tumors BMC Cancer 2005 5 102 16105171 10.1186/1471-2407-5-102
Stupp R Mason WP van den Bent MJ Weller M Fisher B Taphoorn MJ Belanger K Brandes AA Marosi C Bogdahn U Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma N Engl J Med 2005 352 987 996 15758009 10.1056/NEJMoa043330
Veech RL The therapeutic implications of ketone bodies: the effects of ketone bodies in pathological conditions: ketosis, ketogenic diet, redox states, insulin resistance, and mitochondrial metabolism Prostaglandins Leukot Essent Fatty Acids 2004 70 309 319 14769489 10.1016/j.plefa.2003.09.007
Greene AE Todorova MT Seyfried TN Perspectives on the metabolic management of epilepsy through dietary reduction of glucose and elevation of ketone bodies J Neurochem 2003 86 529 537 12859666 10.1046/j.1471-4159.2003.01862.x
Veech RL Metabolic control analysis of ketone and insulin action: Implications for phenotyping of disease and design of therapy 2002
Strohman R Maneuvering in the complex path from genotype to phenotype Science 2002 296 701 703 11976445 10.1126/science.1070534
Kacser H Burns JA The molecular basis of dominance Genetics 1981 97 639 666 7297851
Greenspan RJ The flexible genome Nat Rev Genet 2001 2 383 387 11331904 10.1038/35072018
Seyfried TN Mukherjee P Adams E Mulroony T Abate LE Metabolic Control of Brain Cancer: Role of Glucose and Ketone Bodies Proc Amer Assoc Cancer Res 2005 46 1147
Seyfried TN Sanderson TM El-Abbadi MM McGowan R Mukherjee P Role of glucose and ketone bodies in the metabolic control of experimental brain cancer Br J Cancer 2003 89 1375 1382 14520474 10.1038/sj.bjc.6601269
Kirsch WM Schulz Q Van Buskirk J Nakane P Anaerobic energy metabolism in brain tumors Prog Exp Tumor Res 1972 17 163 191 4343199
Mangiardi JR Yodice P Metabolism of the malignant astrocytoma Neurosurgery 1990 26 1 19 2403663 10.1097/00006123-199001000-00001
Lichtor T Dohrmann GJ Respiratory patterns in human brain tumors Neurosurgery 1986 19 896 899 3027607
Lowry OH Passonneau JV Hasselberger FX Shulz DW Effect of ischemia on known substrates and cofactors of the glycolytic pathway in brain J Biol Chem 1964 239 18 30 14114842
Clarke DD Sokoloff L Siegel GJ, Agranoff BW, Albers RW, Fisher SK, Uhler MD Circulation and energy metabolism in the brain Basic Neurochemistry 1999 sixth New York: Lippincott-Raven 637 669
Cahill GF Jr Starvation in man N Engl J Med 1970 282 668 675 4915800
Ebert D Haller RG Walton ME Energy contribution of octanoate to intact rat brain metabolism measured by 13C nuclear magnetic resonance spectroscopy J Neurosci 2003 23 5928 5935 12843297
Allen NJ Karadottir R Attwell D A preferential role for glycolysis in preventing the anoxic depolarization of rat hippocampal area CA1 pyramidal cells J Neurosci 2005 25 848 859 15673665 10.1523/JNEUROSCI.4157-04.2005
Pellerin L Pellegri G Martin JL Magistretti PJ Expression of monocarboxylate transporter mRNAs in mouse brain: support for a distinct role of lactate as an energy substrate for the neonatal vs. adult brain Proc Natl Acad Sci U S A 1998 95 3990 3995 9520480 10.1073/pnas.95.7.3990
Pellerin L Bergersen LH Halestrap AP Pierre K Cellular and subcellular distribution of monocarboxylate transporters in cultured brain cells and in the adult brain J Neurosci Res 2005 79 55 64 15573400 10.1002/jnr.20307
Owen OE Morgan AP Kemp HG Sullivan JM Herrera MG Cahill GF Jr Brain metabolism during fasting J Clin Invest 1967 46 1589 1595 6061736
VanItallie TB Nufert TH Ketones: metabolism's ugly duckling Nutr Rev 2003 61 327 341 14604265 10.1301/nr.2003.oct.327-341
Morris AA Cerebral ketone body metabolism J Inherit Metab Dis 2005 28 109 121 15877199 10.1007/s10545-005-5518-0
Cahill GF JrVeech RL Ketoacids? Good medicine? Trans Am Clin Climatol Assoc 2003 114 149 161 discussion 162–143 12813917
Mantis JG Centeno NA Todorova MT McGowan R Seyfried TN Management of multifactorial idiopathic epilepsy in EL mice with caloric restriction and the ketogenic diet: role of glucose and ketone bodies Nutr Metab (Lond) 2004 1 11 15507133 10.1186/1743-7075-1-11
Seyfried TN Mukherjee P Meadows GG Anti-Angiogenic and Pro-Apoptotic Effects of Dietary Restriction in Experimental Brain Cancer: Role of Glucose and Ketone Bodies Integration/Interaction of Oncologic Growth 2005 15 2 New York: Kluwer Academic
Leino RL Gerhart DZ Duelli R Enerson BE Drewes LR Diet-induced ketosis increases monocarboxylate transporter (MCT1) levels in rat brain Neurochem Int 2001 38 519 527 11248400 10.1016/S0197-0186(00)00102-9
Koehler-Stec EM Simpson IA Vannucci SJ Landschulz KT Landschulz WH Monocarboxylate transporter expression in mouse brain Am J Physiol 1998 275 E516 524 9725820
Zhang F Vannucci SJ Philp NJ Simpson IA Monocarboxylate transporter expression in the spontaneous hypertensive rat: effect of stroke J Neurosci Res 2005 79 139 145 15578721 10.1002/jnr.20312
Pan JW de Graaf RA Rothman DL Hetherington HP 13C-[2,4]-b-hydroxybutyrate metabolism in human brain J Neurochem 2002 81 45
Fredericks M Ramsey RB 3-Oxo acid coenzyme A transferase activity in brain and tumors of the nervous system J Neurochem 1978 31 1529 1531 298319
Bhagavan NV Medical Biochemistry 2002 Fourth New York: Harcourt
Veech RL Chance B Kashiwaya Y Lardy HA Cahill GF Jr Ketone bodies, potential therapeutic uses IUBMB Life 2001 51 241 247 11569918
Laterra J Keep R Betz AL Goldstein GW Siegel GJ, Agranoff BW, Albers RW, Fisher SK, Uhler MD Blood-brain-cerebrospinal fluid barriers Basic Neurochemistry: Molecular, Cellular and Medical Aspects 1999 sixth New York: Lippincott-Raven 671 689
Fanelli C Di Vincenzo A Modarelli F Lepore M Ciofetta M Epifano L Pampanelli S Brunetti P Bolli GB Post-hypoglycaemic hyperketonaemia does not contribute to brain metabolism during insulin-induced hypoglycaemia in humans Diabetologia 1993 36 1191 1197 8270135 10.1007/BF00401065
Kashiwaya Y Takeshima T Mori N Nakashima K Clarke K Veech RL D-beta-hydroxybutyrate protects neurons in models of Alzheimer's and Parkinson's disease Proc Natl Acad Sci U S A 2000 97 5440 5444 10805800 10.1073/pnas.97.10.5440
Masuda R Monahan JW Kashiwaya Y D-beta-hydroxybutyrate is neuroprotective against hypoxia in serum-free hippocampal primary cultures J Neurosci Res 2005 80 501 509 15825191 10.1002/jnr.20464
Nehlig A Pereira de Vasconcelos A Glucose and ketone body utilization by the brain of neonatal rats Prog Neurobiol 1993 40 163 221 8430212 10.1016/0301-0082(93)90022-K
Sato K Kashiwaya Y Keon CA Tsuchiya N King MT Radda GK Chance B Clarke K Veech RL Insulin, ketone bodies, and mitochondrial energy transduction Faseb J 1995 9 651 658 7768357
Chance B Sies H Boveris A Hydroperoxide metabolism in mammalian organs Physiol Rev 1979 59 527 605 37532
Rhodes CG Wise RJ Gibbs JM Frackowiak RS Hatazawa J Palmer AJ Thomas DG Jones T In vivo disturbance of the oxidative metabolism of glucose in human cerebral gliomas Ann Neurol 1983 14 614 626 6606389 10.1002/ana.410140604
Nagamatsu S Nakamichi Y Inoue N Inoue M Nishino H Sawa H Rat C6 glioma cell growth is related to glucose transport and metabolism Biochem J 1996 319 477 482 8912684
Roslin M Henriksson R Bergstrom P Ungerstedt U Bergenheim AT Baseline levels of glucose metabolites, glutamate and glycerol in malignant glioma assessed by stereotactic microdialysis J Neurooncol 2003 61 151 160 12622454 10.1023/A:1022106910017
Floridi A Paggi MG Fanciulli M Modulation of glycolysis in neuroepithelial tumors J Neurosurg Sci 1989 33 55 64 2674359
Galarraga J Loreck DJ Graham JF DeLaPaz RL Smith BH Hallgren D Cummins CJ Glucose metabolism in human gliomas: correspondence of in situ and in vitro metabolic rates and altered energy metabolism Metab Brain Dis 1986 1 279 291 3508247 10.1007/BF00999357
Mies G Paschen W Ebhardt G Hossmann KA Relationship between of blood flow, glucose metabolism, protein synthesis, glucose and ATP content in experimentally-induced glioma (RG1 2.2) of rat brain J Neurooncol 1990 9 17 28 2213113 10.1007/BF00167064
Oudard S Boitier E Miccoli L Rousset S Dutrillaux B Poupon MF Gliomas are driven by glycolysis: putative roles of hexokinase, oxidative phosphorylation and mitochondrial ultrastructure Anticancer Res 1997 17 1903 1911 9216643
Tisdale MJ Biology of cachexia J Natl Cancer Inst 1997 89 1763 1773 9392617 10.1093/jnci/89.23.1763
Patel MS Russell JJ Gershman H Ketone-body metabolism in glioma and neuroblastoma cells Proc Natl Acad Sci U S A 1981 78 7214 7218 6118869
Roeder LM Poduslo SE Tildon JT Utilization of ketone bodies and glucose by established neural cell lines J Neurosci Res 1982 8 671 682 7161845 10.1002/jnr.490080412
Sawai M Yashiro M Nishiguchi Y Ohira M Hirakawa K Growth-inhibitory effects of the ketone body, monoacetoacetin, on human gastric cancer cells with succinyl-CoA: 3-oxoacid CoA-transferase (SCOT) deficiency Anticancer Res 2004 24 2213 2217 15330163
Tisdale MJ Role of acetoacetyl-CoA synthetase in acetoacetate utilization by tumor cells Cancer Biochem Biophys 1984 7 101 107 6147189
Portais JC Voisin P Merle M Canioni P Glucose and glutamine metabolism in C6 glioma cells studied by carbon 13 NMR Biochimie 1996 78 155 164 8831946 10.1016/0300-9084(96)89500-9
Meixensberger J Herting B Roggendorf W Reichmann H Metabolic patterns in malignant gliomas J Neurooncol 1995 24 153 161 7562002 10.1007/BF01078485
Pedersen PL Tumor mitochondria and the bioenergetics of cancer cells Prog Exp Tumor Res 1978 22 190 274 149996
Warburg O The Metabolism of Tumours 1931 New York: Richard R. Smith
Warburg O On the origin of cancer cells Science 1956 123 309 314 13298683
Cuezva JM Chen G Alonso AM Isidore A Misek DE Hanash SM Beer DG The bioenergetic signature of lung adenocarcinomas is a molecular marker of cancer diagnosis and prognosis Carcinogenesis 2004 25 1157 1163 14963017 10.1093/carcin/bgh113
Aisenberg AC The Glycolysis and Respiration of Tumors 1961 New York: Academic Press
Ikezaki K Black KL Conklin SG Becker DP Histochemical evaluation of energy metabolism in rat glioma Neurol Res 1992 14 289 293 1360622
Schlame M Rua D Greenberg ML The biosynthesis and functional role of cardiolipin Prog Lipid Res 2000 39 257 288 10799718 10.1016/S0163-7827(00)00005-9
Canuto RA Biocca ME Muzio G Dianzani MU Fatty acid composition of phospholipids in mitochondria and microsomes during diethylnitrosamine carcinogenesis in rat liver Cell Biochem Funct 1989 7 11 19 2752533 10.1002/cbf.290070104
Elstrom RL Bauer DE Buzzai M Karnauskas R Harris MH Plas DR Zhuang H Cinalli RM Alavi A Rudin CM Akt stimulates aerobic glycolysis in cancer cells Cancer Res 2004 64 3892 3899 15172999 10.1158/0008-5472.CAN-03-2904
Nebeling LC Miraldi F Shurin SB Lerner E Effects of a ketogenic diet on tumor metabolism and nutritional status in pediatric oncology patients: two case reports J Am Coll Nutr 1995 14 202 208 7790697
Stafstrom CE Bough KJ The ketogenic diet for the treatment of epilepsy: a challenge for nutritional neuroscientists Nutr Neurosci 2003 6 67 79 12722982 10.1080/1028415031000084427
Freeman JM Freeman JB Kelly MT The Ketogenic Diet: A Treatment for Epilepsy 2000 third New York: Demos
Stewart JW Koehler K Jackson W Hawley J Wang W Au A Myers R Birt DF Prevention of mouse skin tumor promotion by dietary energy restriction requires an intact adrenal gland and glucocorticoid supplementation restores inhibition Carcinogenesis 2005 26 1077 1084 15746164 10.1093/carcin/bgi051
Zhu Z Jiang W Thompson HJ Mechanisms by which energy restriction inhibits rat mammary carcinogenesis: in vivo effects of corticosterone on cell cycle machinery in mammary carcinomas Carcinogenesis 2003 24 1225 1231 12807724 10.1093/carcin/bgg077
Patel NV Finch CE The glucocorticoid paradox of caloric restriction in slowing brain aging Neurobiol Aging 2002 23 707 717 12392776 10.1016/S0197-4580(02)00017-9
Mukherjee P Abate LE Seyfried TN Antiangiogenic and proapoptotic effects of dietary restriction on experimental mouse and human brain tumors Clin Cancer Res 2004 10 5622 5629 15328205 10.1158/1078-0432.CCR-04-0308
Mukherjee P El-Abbadi MM Kasperzyk JL Ranes MK Seyfried TN Dietary restriction reduces angiogenesis and growth in an orthotopic mouse brain tumour model Br J Cancer 2002 86 1615 1621 12085212 10.1038/sj.bjc.6600298
Duan W Lee J Guo Z Mattson MP Dietary restriction stimulates BDNF production in the brain and thereby protects neurons against excitotoxic injury J Mol Neurosci 2001 16 1 12 11345515 10.1385/JMN:16:1:1
Birt DF Yaktine A Duysen E Glucocorticoid mediation of dietary energy restriction inhibition of mouse skin carcinogenesis J Nutr 1999 129 571S 574S 10064335
Spindler SR Rapid and reversible induction of the longevity, anticancer and genomic effects of caloric restriction Mech Ageing Dev 2005 15927235
Chung HY Kim HJ Kim KW Choi JS Yu BP Molecular inflammation hypothesis of aging based on the anti-aging mechanism of calorie restriction Microsc Res Tech 2002 59 264 272 12424787 10.1002/jemt.10203
Weindruch R Walford RL The retardation of aging and disease by dietary restriction 1988 Springfield, IL: Thomas 3346517
Tannenbaum A Homburger F Nutrition and cancer Physiopathology of Cancer 1959 NY: Paul B. Hober 517 562
Rebrin I Kamzalov S Sohal RS Effects of age and caloric restriction on glutathione redox state in mice Free Radic Biol Med 2003 35 626 635 12957655 10.1016/S0891-5849(03)00388-5
Aronen HJ Pardo FS Kennedy DN Belliveau JW Packard SD Hsu DW Hochberg FH Fischman AJ Rosen BR High microvascular blood volume is associated with high glucose uptake and tumor angiogenesis in human gliomas Clin Cancer Res 2000 6 2189 2200 10873068
Tannenbaum A The genesis and growth of tumors: II. Effects of caloric restriction per se Cancer Res 1942 2 460 467
Mukherjee P Sotnikov AV Mangian HJ Zhou JR Visek WJ Clinton SK Energy intake and prostate tumor growth, angiogenesis, and vascular endothelial growth factor expression J Natl Cancer Inst 1999 91 512 523 10088621 10.1093/jnci/91.6.512
Pennathur S Ido Y Heller JI Byun J Danda R Pergola P Williamson JR Heinecke JW Reactive carbonyls and polyunsaturated fatty acids produce a hydroxyl radical-like species: A potential pathway for oxidative damage of retinal proteins in diabetes J Biol Chem 2005 15855169
Andersson AK Ronnback L Hansson E Lactate induces tumour necrosis factor-alpha, interleukin-6 and interleukin-1beta release in microglial- and astroglial-enriched primary cultures J Neurochem 2005 93 1327 1333 15934951 10.1111/j.1471-4159.2005.03132.x
Dong W Selgrade MK Gilmour IM Lange RW Park P Luster MI Kari FW Altered alveolar macrophage function in calorie-restricted rats Am J Respir Cell Mol Biol 1998 19 462 469 9730874
Potts R Humanity's Descent: The Consequences of Ecological Instability 1996 New York: William Morrow & Co., Inc
Harold FM The Vital Force: A Study of Bioenergetics 1986 New York: W. H. Freeman
Albers RW Siegel GJ Siegel GJ, Agranoff BW, Albers RW, Fisher SK, Uhler MD Membrane transport Basic Neurochemistry 1999 sixth New York: Lippincott-Raven 95 118
Lipton P Robacker K Glycolysis and brain function: [K+]o stimulation of protein synthesis and K+ uptake require glycolysis Fed Proc 1983 42 2875 2880 6136420
Nishioka T Oda Y Seino Y Yamamoto T Inagaki N Yano H Imura H Shigemoto R Kikuchi H Distribution of the glucose transporters in human brain tumors Cancer Res 1992 52 3972 3979 1617673
Boado RJ Black KL Pardridge WM Gene expression of GLUT3 and GLUT1 glucose transporters in human brain tumors Brain Res Mol Brain Res 1994 27 51 57 7877454 10.1016/0169-328X(94)90183-X
Strohman R Thermodynamics – old laws in medicine and complex disease Nat Biotechnol 2003 21 477 479 12721560 10.1038/nbt0503-477
Nebeling LC Lerner E Implementing a ketogenic diet based on medium-chain triglyceride oil in pediatric patients with cancer J Am Diet Assoc 1995 95 693 697 7759747 10.1016/S0002-8223(95)00189-1
Covarrubias DJ Rosen BR Lev MH Dynamic magnetic resonance perfusion imaging of brain tumors Oncologist 2004 9 528 537 15477637 10.1634/theoncologist.9-5-528
Ko YH Smith BL Wang Y Pomper MG Rini DA Torbenson MS Hullihen J Pedersen PL Advanced cancers: eradication in all cases using 3-bromopyruvate therapy to deplete ATP Biochem Biophys Res Commun 2004 324 269 275 15465013 10.1016/j.bbrc.2004.09.047
Fearon KC Nutritional pharmacology in the treatment of neoplastic disease Baillieres Clin Gastroenterol 1988 2 941 949 3072982 10.1016/0950-3528(88)90043-7
Sokoloff B Eddy WH Saelof CC Beach J Glucose antagonists in experimental cancer Arch Pathol
Magee BA Potezny N Rofe AM Conyers RA The inhibition of malignant cell growth by ketone bodies Aust J Exp Biol Med Sci 1979 57 529 539 548019
Landau BR Laszlo J Stengle J Burk D Certain metabolic and pharmacologic effects in cancer patients given infusions of 2-deoxy-D-glucose J Natl Cancer Inst 1958 21 485 494 13576102
Cleary MP Jacobson MK Phillips FC Getzin SC Grande JP Maihle NJ Weight-cycling decreases incidence and increases latency of mammary tumors to a greater extent than does chronic caloric restriction in mouse mammary tumor virus-transforming growth factor-alpha female mice Cancer Epidemiol Biomarkers Prev 2002 11 836 843 12223427
Meadows GG Fu Y-M Meadows GG Dietary restriction of specific amino acids modulates tumor and host interactions Integration/Interaction of Oncologic Growth 2005 15 2 New York: Kluwer Academic
|
16242042
|
PMC1276814
|
CC BY
|
2021-01-04 16:37:45
|
no
|
Nutr Metab (Lond). 2005 Oct 21; 2:30
|
utf-8
|
Nutr Metab (Lond)
| 2,005 |
10.1186/1743-7075-2-30
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-571624889510.1186/1477-7827-3-57ResearchRegulation of soluble vascular endothelial growth factor receptor (sFlt-1/sVEGFR-1) expression and release in endothelial cells by human follicular fluid and granulosa cells Gruemmer Ruth [email protected] Karin [email protected] Daniela [email protected] Herbert A [email protected] Joseph [email protected] Institute of Anatomy, University Hospital, Essen, Germany2 Clinic of Gynecological Endocrinology and Reproductive Medicine, RWTH Aachen, Germany3 Dep. Gene Regulation and Differentiation, GBF Braunschweig, Germany2005 25 10 2005 3 57 57 18 8 2005 25 10 2005 Copyright © 2005 Gruemmer et al; licensee BioMed Central Ltd.2005Gruemmer et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
During the female reproductive cycle, follicular development and corpus luteum formation crucially depend on the fast generation of new blood vessels. The importance of granulosa cells and follicular fluid in controlling this angiogenesis is still not completely understood. Vascular endothelial growth factor (VEGF) produced by granulosa cells and secreted into the follicular fluid plays an essential role in this process. On the other hand, soluble VEGF receptor-1 (sFlt-1) produced by endothelial cells acts as a negative modulator for the bioavailability of VEGF. However, the regulation of sFlt-1 production remains to be determined.
Methods
We analyzed the influence of human follicular fluid obtained from FSH-stimulated women as well as of human granulosa cell conditioned medium on sFlt-1 production in and release from human umbilical vein endothelial cells (HUVEC) in vitro. Soluble Flt-1 gene expression was determined by RT-PCR analysis, amount of sFlt-1-protein was quantified by Sandwich-ELISA.
Results
Human follicular fluid as well as granulosa cell-conditioned medium significantly inhibit the production of sFlt-1 by endothelial cells on a posttranscriptional level. Treatment of cultured granulosa cells with either hCG or FSH had not impact on the production of sFlt-1 inhibiting factors. We further present data suggesting that this as yet unknown sFlt-1 regulating factor secreted by granulosa cells is not heat-sensitive, not steroidal, and it is of low molecular mass (< 1000 Da).
Conclusion
We provide strong support that follicular fluid and granulosa cells control VEGF availability by down regulation of the soluble antagonist sFlt-1 leading to an increase of free, bioactive VEGF for maximal induction of vessel growth in the ovary.
==== Body
Background
Angiogenesis is a rare process in normal adult organs predominantly occurring during wound healing and tumor growth. However, under physiological conditions it plays an important role in the female reproductive tract with regard to follicular development, corpus luteum formation, and uterine endometrial proliferation during the menstrual cycle [1,2]. Here, the cyclic corpus luteum of the ovary is the organ with the strongest physiological angiogenesis [3,4]. Defects in ovarian angiogenesis may contribute to a variety of disorders including anovulation and infertility, pregnancy loss, ovarian hyperstimulation syndrome, and ovarian neoplasms [5-7].
During follicular growth, angiogenesis is restricted to the theca cell layer. After ovulation, however, massive angiogenesis occurs and new blood vessels penetrate the basement membrane of the follicle invading the growing corpus luteum [8]. The establishment of such a complex capillary network requires precise timing. Angiogenesis depends on a balance between positive and negative endothelial regulators [9]. Among the many endothelial regulators, vascular endothelial growth factor (VEGF) has been characterized as the most potent promoter of angiogenesis. This key regulator acts specifically on endothelial cells by stimulating cell growth, differentiation, migration and permeability [10,11]. VEGF, especially the isoforms VEGF-A121 and VEGF-A165, are produced by human granulosa cells [5,12-15], and VEGF-dependent angiogenesis is essential for corpus luteum development [16]. VEGF expression in granulosa cells can be increased by gonadotropins (FSH, hCG) [17,18]. The biological activity of VEGF is mediated by two tyrosine kinase family receptors that are located on endothelial cells (VEGFR-1 = flt-1, VEGFR-2 = KDR) [19]. Binding of VEGF to either of the receptors induces autophosphorylation and signal transduction. Besides these transmembrane receptors, a soluble receptor (sFlt-1) is generated in endothelial cells by differential splicing of the VEGFR-1 mRNA [20,21]. Soluble Flt-1 retains full VEGF binding potency and acts as an inhibitor of VEGF bioactivity by sequestering VEGF, thus reducing the ligand binding to transmembrane and signalling receptors [22,23]. It plays a pivotal role in the generation of vascular diseases like pre-eclampsia or intra-uterine growth retardation [24], and could be linked to the ovarian response to stimulation protocols [25].
The regulation of sFlt-1 production in endothelial cells remains to be determined. In the ovary, it could be demonstrated in high amounts in follicular fluid aspirated during IVF oocyte retrieval. The function of the avascular granulosa cell layer in controlling angiogenesis in the vicinity of developing follicles is still a matter of discussion. In the present study we analyzed the influence of human follicular fluid as well as of human granulosa cell conditioned medium on Flt-1 production and release by human endothelial cells.
Materials and methods
Follicular fluid
Follicular fluid was obtained by follicular aspiration from 15 FSH-treated women (34.6 ± 5.6 years) undergoing oocyte retrieval for in vitro fertilization (IVF/ICSI) at the Department of Gynecological Endocrinology at the University Hospital Aachen, Germany. The experimental design was approved by the local ethical committee of the University Hospital Aachen (# EK 2008). Written informed consent was obtained from patients individually. IVF/ICSI was performed due to tubal occlusion (7), andrological (6) or idiopathic (2) reasons. Ovarian stimulation and oocyte retrieval were performed as described previously [26]. After removal of oocytes, follicular fluids of individual patients were pooled, and centrifuged at 500 g for 5 min. Supernatants were frozen at -20°C until further processing.
Follicular fluid was partially purified by using two successive ultrafiltration units with different molecular weight cut offs (MWCOs). To prevent clogging of filters, a 100000 Da ultra-filter (Millipore GmbH, Eschborn, Germany) was used as a pre-filter. The first flow through was consecutively treated with a 1000 Da filter unit (Pall Life Sciences, Ann Arbor, MI). Depending on the follicular fluid's viscosity centrifugation steps were carried out both times for 4–6 hours at 4400 g and 4°C. The final flow through was then incubated either directly with HUVECs or treated further as follows: heating for 10 minutes at 96°C, or dialysis with 100 Da MWCO membranes (Spectrum Laboratories Inc., Rancho Dominguez, CA) equilibrated against two changes of a two hundredfold volume PBS over a 6 hour period at 4°C. In another experiment the partially purified follicular fluid was stirred with dextran-coated charcoal (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) at a concentration of 10 mg/ml for 30 minutes at room temperature to absorb free steroids and fatty acids. Any precipitated material or charcoal was removed by centrifugation for 5 minutes at 13 600 g and 4°C.
Granulosa cell culture
Human granulosa cells were obtained by follicular aspiration from FSH-treated women as described before [26]. Cells were plated at a density of 5 × 105 cells/well in 6-well dishes and cultured in M199 Earle's Medium supplemented with 10% FCS, 2 mM L-glutamine, and 1% Pen/Strep (all Biochrom, Berlin, Germany) at 37°C in 95% air-5% CO2 humidified environment. In one experimental group, medium was additionally supplemented with FSH (Gonal F, Serono, 100 ng/ml), in another experimental group with hCG (Pregnesin, Serono, 1 IU / ml). Cell culture medium was changed after 24 hours and was then harvested after 4 days of culturing and stored at -20°C until further use.
Endothelial cell culture
Human umbilical vein endothelial cells (HUVEC) were isolated from umbilical cords and cultured as described previously [26]. For each experiment, HUVECs of 3–5 umbilical veins were pooled and seeded in 6- or 12-well plates. At confluence after two days of culturing cells were incubated either with culture medium alone, with culture medium containing 30% follicular fluid (in those experiments with constant concentration of follicular fluid) or with 30% GC-conditioned medium, respectively. After up to 4 days of incubation at 37°C culture medium was collected, centrifuged and supernatant was frozen at -20°C, endothelial cells were harvested and frozen at -20°C until RNA-preparation. For each experimental design HUVECs were separately incubated with follicular fluid or GC-conditioned medium, respectively, of at least 5 different patients.
Proliferation studies
Human umbilical vein endothelial cells were seeded into 24-well-plates with a density of 50000 cells per well. Cells were incubated with culture medium containing 30% follicular fluid for up to 4 days or with culture medium alone as a control. Growth rates were determined in nine experiments with follicular fluid of nine different patients using an electronic Coulter counter (CASY 1, Schärfe System, Reutlingen, Germany). Each probe has been measured twice.
Quantification of sFlt-1 protein
Total sFlt-1 concentration in HUVEC supernatant was quantified with a specific enzyme-linked immunosorbent assay (ELISA) (RELIATech, Braunschweig, Germany [22] and BMS268, Bender MedSystems GmbH, Vienna, Austria). Intra- and inter-assay co-efficiencies were CV<10% and CV<20%, respectively. Soluble Flt-1 ELISA analyses were performed in duplicate for each probe.
RT-PCR
Isolation of total RNA from eutopic endometrial tissues as well as from endometrial tissue grown in nude mice was performed using the RNeasy Minikit® (Qiagen, Hilden, Germany) according to the manufactures instructions. The concentration of RNA was determined spectrophotometrically and the RNA was stored at -80°C until use. Reverse transcription of RNA from HUVECs treated with follicular fluid or granulosa cell conditioned medium was carried out as described previously [26]. Briefly, two micrograms of total RNA were digested with DNase I (Invitrogen, Karlsruhe, Germany) and transcribed into cDNA by reverse transcription with M-MLV Reverse Transcriptase (Invitrogen, Karlsruhe, Germany) using an oligo (dT)16 primer in a total volume of 50 μl. Reverse transcription was performed for 60 min at 37°C in a thermocycler (Biometra, Göttingen, Germany) followed by 10 min at 90°C. 4 μl of the RT-reaction were used for PCR experiments. The following primers were used: sFlt-1 forward 5'-GCACCTTGGTTGTGGCTGAC-3'; sFlt-1 reverse 5'-AATGTTTTACATTACTTTGTGTGG-3' (product size 510 bp), β-actin forward 5'-ACCAACTGGGACGACATGGAGAAAA-3', β-actin reverse 5'-TACGGCCAGAGGCGTACAGGGATAG-3' (product size 214 bp). Amplifications were run in 50 μl volume using BioTherm Taq polymerase (Genecraft, Muenster, Germany) for 35 amplification cycles of 30 sec denaturation at 94°C, 45 sec annealing at 62°C and 30 sec elongation at 72°C. The PCR amplification was followed by a 10 minute final extension at 72°C. The conditions were chosen so that the sflt-1 cDNA as well as the control β-actin cDNA were in the exponential phase of amplification and did not reach a plateau at the end of the amplification protocol. The generated PCR amplification products were electrophoresed on a 2% agarose gel and detected by ethidiumbromide staining. PCR products were normalized to β-actin by densitometric analysis using a Gel imager (Intas, Goettingen, Germany) and were relatively quantified (Gelscan Professional V4.0).
Statistical analysis
Statistical analysis was performed using the non-parametric Mann-Whitney test. The level of significance was set at p < 0.05.
Results
Regulation of endothelial cell proliferation and sFlt-1 secretion by follicular fluid
Incubation of HUVECs with culture medium containing 30% follicular fluid led to a significant increase in cell number compared to controls on day 3 and 4 of culturing (Fig. 1). Analyzing secretion of sFlt-1 by endothelial cells, HUVECs secreted 14.63 ± 1.36 ng sFlt-1 per ml into the culture medium during 24 hours of monolayer culture (Fig. 2). After 3 days of culture the amount of secreted sFlt-1 accumulated to 37.2 ± 5.2 ng/ml. This increase in sFlt-1 -1 was inhibited by the presence of 30% follicular fluid in the culture medium (Fig. 2). The inhibition of sFlt-1 production was proven to show a dose-effect as it was dependent on the concentration of follicular fluid in the culture medium. The amount of sFlt-1 decreased with increasing concentration of follicular fluid, showing a significant inhibition of sFlt-1 production at a concentration of 30% (Fig. 3).
Figure 1 Proliferation of HUVECs incubated with culture medium containing 30% follicular fluid (+FF) or incubated with culture medium alone (control) for up to 4 days. From day 3 onwards a significant increase in cell number can be observed for those endothelial cells treated with follicular fluid compared to controls. * = p < 0.05.
Figure 2 Amount of sFlt-1 in the culture supernatant of HUVECs treated with medium only (grey bars) or with medium containing 30% human follicular fluid (FF, black bars) for up to 3 days. From day 2 onwards a significant increase in sFlt-1 content can be observed for those endothelial cells treated with medium only but not for those incubated with medium containing follicular fluid. * = p < 0.05.
Figure 3 Amount of sFlt-1 in the culture supernatant of HUVECs after 4 days of incubation with medium containing different concentrations of human follicular fluid. Amount of sFlt-1 decreases with increasing concentrations of follicular fluid showing a significant inhibition of sFlt-1 production at a concentration of 30%. * = p < 0.05.
To exclude an unspecific role of FCS on sFlt-1 production in HUVECs, culture medium containing FCS had been either untreated, heat inactivated or FCS was omitted. There was no measurable effect of either of these controls on sFlt-1 production of endothelial cells (Fig. 4). To analyze the follicular fluid in regard to factors possibly responsible for sFlt-1 regulation, follicular fluid has been partially purified by ultrafiltration. The resulting flow through containing only molecules smaller than 1000 Dalton still significantly inhibited sFlt-1 secretion by endothelial cells. Moreover, this effect could not be prevented by exposure of the flow through to heat or charcoal treatment (Fig. 4).
Figure 4 Quantification of sFlt-1 in endothelial cell culture supernatant. Neither heat inactivation of culture medium containing FCS (C2) nor absence of FCS (C3) has a measurable effect on sFlt-1 production of endothelial cells compared to incubation with untreated control medium containing FCS (C1). The significant inhibitory effect of untreated human follicular fluid (FF) on sFlt-1 production is maintained after ultra filtration leaving only molecules smaller than 1000 Dalton. This inhibition is not prevented neither by heat inactivation nor by charcoal treatment of the follicular fluid-flow through. All follicular fluids have been added to the culture medium at a concentration of 30%. * = p < 0.05
Transcription of sFlt-1 is not regulated by follicular fluid
RT-PCR analysis of HUVECs incubated with medium containing 30% follicular fluid revealed a slight but not significant decrease in sFlt-1 expression compared to endothelial cells treated with medium only (Fig. 5), pointing to a presumable posttranscriptional regulatory mechanism of this soluble receptor by follicular fluid.
Figure 5 RT-PCR of mRNA of HUVECs incubated with medium only (C) or with medium containing 30% follicular fluid (+FF). Densitometric evaluation revealed no significant difference in mRNA-expression of sFlt-1 between these two experimental groups at p < 0.05.
Regulation of sFlt-1 secretion by granulosa cell-conditioned medium
This inhibition of sFlt-1 production by endothelial cells was also obtained by incubation of HUVECs with granulosa cell-conditioned medium. At the concentration of 30% granulosa cell-conditioned medium a significant inhibition of sFlt-1 production could be observed after 4 days of culturing (Fig. 6). This inhibition was not influenced by incubation of the granulosa cell culture either with FCS or hCG. In general, however, treatment with granulosa cell-conditioned medium resulted in a lesser inhibition of sFlt-1 production compared to the inhibition by corresponding follicular fluid of the same patients (Fig. 6).
Figure 6 Amount of sFlt-1 in the culture supernatant of HUVECs incubated for 4 days with medium containing 30% granulosa cell (GC)-conditioned medium. Addition of GC-conditioned medium leads to a significant inhibition of sFlt-1 production of HUVECs (*). Treatment of granulosa cells with FCS or hCG during conditioning of medium has no influence on this inhibition. Incubation with follicular fluid (FF) of the same patients leads to a significant reduced production of sFlt-1 compared to controls as well as to treatment with GC-conditioned medium (**). p < 0.05.
Discussion
In the present study we demonstrate that human follicular fluid significantly increases proliferation and inhibits the production of the VEGF antagonist sFlt-1 in endothelial cells. It is proven that granulosa cells produce large amounts of VEGF which plays the key role in corpus luteum angiogenesis in vivo [16]. The biological activity of VEGF depends on the availability of this protein to its transmembrane receptors Flt-1 (VEGFR-1) and KDR (VEGFR-2). Soluble Flt-1 is secreted by endothelial cells [20] and acts as a receptor antagonist by sequestering free VEGF, thus repressing angiogenesis mediated by VEGF [22,23]. Soluble Flt-1 may modulate VEGF activity in physiological as well as in patho-physiological angiogenesis in the female reproductive tract. It is secreted by the placenta and released into the maternal circulation during pregnancy [27], and it could be shown that preeclampsia is associated with increased levels of sFlt-1 [24,28-30]. In addition, it could be shown that excess in sFlt-1 goes in parallel with poor response to gonadotropins in stimulation protocols due to the decreased availability of bioactive VEGF [25]. An influence of sFlt-1 on physiological angiogenesis in the ovary was shown by Ferrara and co-workers [16] who reported that treatment with truncated sFlt-1 receptors resulted in complete suppression of corpus luteum angiogenesis in a rat model.
The role of the avascular granulosa cell layer in controlling angiogenesis in the developing follicles is still not clarified. It has been shown before that human follicular fluid contains angiogenic factors such as basic fibroblast growth factors [31], angiogenin [32] and VEGF [33] and that follicular fluid has the capacity to induce angiogenesis [34]. As demonstrated in the present study, the inhibition of sFlt-1 by factors secreted by granulosa cells, thus increasing the amount of free, bioactive VEGF, may represent another regulatory level involved in the angiogenic cascade. Though it is known that soluble receptor protein is generated by alternative splicing of Flt-1 pre-mRNA [35] and exogenous sFlt-1 can dramatically inhibit biological actions of VEGF, there is still little information about physiological functions of sFlt-1 or the mechanisms controlling its biosynthesis. We demonstrated here that factors secreted by granulosa cells and which are contained in follicular fluid inhibit sFlt-1 production, and that this regulation mainly occurs on a posttranscriptional level. This is concordant with our former microarray analyses showing that sFlt-1 mRNA was not amongst the genes significantly regulated by follicular fluid in endothelial cells [26]. A post-transcriptional control of sFlt-1 expression could also be demonstrated by Huckle and Roche [36] using cleavage-polyadenylation mutants of the mouse Flt-1 intron 13. Thus, an opportunity exists for granulosa cells to regulate sFlt-1 protein expression at a widely constant rate of sFlt-1 gene transcription. However, in cyclic human endometrium regulation of sFlt-1 could be demonstrated on a transcriptional level [37] pointing to different regulatory pathways in different organs. The weaker effect of granulosa cell conditioned medium on inhibition of sFlt-1 compared to follicular fluid could be due to a lower concentration of the inhibitory factor in the granulosa cell conditioned medium since this medium was harvested already after 4 days. Though FSH as well as LH / hCG application are able to increase VEGF expression in human granulosa cells [17,18], treatment of cultured luteal granulosa cells either with hCG or FSH had no effect on the inhibition of sFlt-1 secretion. Redmer and co-workers [38] showed that luteal cells secrete a non-steroidal factor which stimulates migration of endothelial cells and that pre-treatment of the luteal cultures with hCG or FSH had no effect on the endothelial cell migration stimulating activity of luteal cell conditioned media. Since the granulosa cells have been exposed to FSH in vivo, a regulatory effect of this hormone can not be excluded. However, depletion of FSH in vitro had no effect on the production of this factor. The nature of the inhibitory factor(s) secreted by granulosa cells is still unknown. As demonstrated here, the factor is of low molecular mass (< 1000 Da) and heat inactivation does not change the inhibitory effect on sFlt-1 secretion. In addition, a steroidal nature can be excluded since this effect is not be eliminated by charcoal treatment.
A complex cascade of events regulated on a transcriptional as well as post-transcriptional level might contribute to the initiation, progression, morphogenesis and regression of blood vessels in the ovary. We have shown previously that after the LH surge, yet before ovulation, granulosa cells secrete factors resulting in destabilization of vessel walls by down-regulating of fibulin-5 and elastin and up-regulation of angiopoietin-2 [26]. In addition, granulosa cells may improve the sprouting of new vessels by down-regulation of the soluble antagonist sFlt-1 thus enhancing the amount of bioactive VEGF. This suppression of sFlt-1 immediately before corpus luteum formation may act to rapidly promote and fine tune angiogenesis in this ephemeral organ. The release in the ovary could be one factor which helps to explain the temporal and spatial discrepancy between the high expression of VEGF in the granulosa cells and the restriction of angiogenesis to the thecal layer in the pre-ovulatory follicle. Disruption of this balance between VEGF and sFlt-1 may result in a disturbed physiological state or various pathological conditions. Understanding the underlying molecular mechanisms which regulate the complex angiogenic cascade is a major challenge with implications for the understanding of blood vessel growth and regression in reproductive biology as well as in pathological conditions.
Authors' contributions
RG participated in the design of the study, carried out cell culture experiments, performed RT-PCR, performed the statistical analysis and drafted the manuscript. KM and DB carried out part of the cell culture experiments. HAW carried out protein quantification by ELISA. HAW and JN contributed to the conception and design of the study and revised the draft critically. All authors read and approved the final manuscript.
Acknowledgements
This work has been supported by the Deutsche Forschungsgemeinschaft (DFG), grant GR 1138/8-1/2 to R.G. and J.N.
==== Refs
Findlay JK Angiogenesis in reproductive tissues J Endocrinol 1986 111 357 366 2433380
Reynolds LP Killilea SD Redmer DA Angiogenesis in the female reproductive system FASEB J 1992 6 886 892 1371260
Augustin HG Vascular morphogenesis in the ovary Baillieres Best Pract Res Clin Obstet Gynaecol 2000 14 867 882 11141338 10.1053/beog.2000.0132
Fraser HM Wulff C Angiogenesis in the primate ovary Reprod Fertil Develop 2001 13 557 566 10.1071/RD01055
Neulen J Yan Z Raczek S Weindel K Keck C Weich HA Marme D Breckwoldt M Human chorionic gonadotropin-dependent expression of vascular endothelial growth factor/vascular permeability factor in human granulosa cells: importance in ovarian hyperstimulation syndrome J Clin Endocrinol Metabol 1995 80 1967 1971 10.1210/jc.80.6.1967
Geva E Jaffe RB Role of vascular endothelial growth factor in ovarian physiology and pathology Fertil Steril 2000 74 429 438 10973633 10.1016/S0015-0282(00)00670-1
Abulafia O Sherer DM Angiogenesis of the ovary Am J Obstet Gynecol 2000 182 240 246 10649185
Cavender JL Murdoch WJ Morphological studies of the microcirculatory system of periovulatory ovine follicles Biol Reprod 1988 39 989 997 3207814
Hanahan K Folkman J Patterns and emerging mechanisms of the angiogenic switch during tumorogenesis Cell 1996 86 33 364 10.1016/S0092-8674(00)80108-7
Risau W Mechanisms of angiogenesis Nature 1997 386 671 674 9109485 10.1038/386671a0
Ferrara N The role of VEGF in the regulation of physiological and pathological angiogenesis EXS 2005 94 209 231 15617481
Yan Z Weich HA Bernart W Breckwoldt M Neulen J Vascular endothelial growth factor (VEGF) messenger ribonucleic acid (mRNA) expression in luteinized human granulosa cells in vitro J Clin Endocrinol Metabol 1993 77 1723 1725 10.1210/jc.77.6.1723
Laitinen M Ristimaki A Honkasalo M Narko K Paavonen K Ritvos O Differential hormonal regulation of vascular endothelial growth factors VEGF, VEGF-B, and VEGF-C messenger ribonucleic acid levels in cultured human granulosa-luteal cells Endocrinol 1997 138 4748 4756 10.1210/en.138.11.4748
Otani N Minami S Yamoto M Shikone T Otani H Nishiyama R Otani T Nakano R The vascular endothelial growth factor/fms-like tyrosine kinase system in human ovary during the menstrual cycle and early pregnancy J Clin Endocrinol Metabol 1999 84 3845 3851 10.1210/jc.84.10.3845
Kamat BR Brown LF Manseau EJ Senger DR Dvorak HF Expression of vascular permeability factor/vascular endothelial growth factor by human granulosa and theca lutein cells. Role in corpus luteum development Am J Pathol 1995 146 157 165 7531945
Ferrara N Chen H Davis-Smyth T Gerber HP Nguyen TN Peers D Chisholm V Hillan KJ Schwall RH Vascular endothelial growth factor is essential for corpus luteum angiogenesis Nature Med 1998 4 336 340 9500609 10.1038/nm0398-336
Christenson LK Stouffer RL Follicle-stimulating hormone and luteinizing hormone/chorionic gonadotropin stimulation of vascular endothelial growth factor production by macaque granulosa cells from pre- and periovulatory follicles J Clin Endocrinol Metab 1997 82 2135 2142 9215284 10.1210/jc.82.7.2135
Neulen J Raczek S Pogorzelski M Grunwald K Yeo TK Dvorak HF Weich HA Breckwoldt M Secretion of vascular endothelial growth factor/vascular permeability factor from human luteinized granulosa cells is human chorionic gonadotrophin dependent Mol Hum Reprod 1998 4 203 206 9570265 10.1093/molehr/4.3.203
Ferrara N Davis-Smyth T The biology of vascular endothelial growth factor Endocrine Rev 1997 18 4 25 9034784 10.1210/er.18.1.4
Kendall RL Thomas KA Inhibition of vascular endothelial cell growth factor activity by an endogenously encoded soluble receptor Proc Natl Acad Sci USA 1993 90 10705 10709 8248162
Kendall RL Wang G Thomas KA Identification of a natural soluble form of the vascular endothelial growth factor receptor, FLT-1, and its heterodimerization with KDR Biochem Biophys Res Commun 1996 226 324 328 8806634 10.1006/bbrc.1996.1355
Hornig C Weich HA Soluble VEGF receptors Angiogenesis 1999 3 33 39 14517442 10.1023/A:1009033017809
Hornig C Barleon B Ahmad S Vuorela P Ahmed A Weich HA Release and complex formation of soluble VEGFR-1 from endothelial cells and biological fluids Lab Invest 2000 80 443 454 10780661
McKeeman GC Ardill JE Caldwell CM Hunter AJ McClure N Soluble vascular endothelial growth factor receptor-1 (sFlt-1) is increased throughout gestation in patients who have preeclampsia develop Am J Obstet Gynecol 2004 191 1240 1246 15507947 10.1016/j.ajog.2004.03.004
Neulen J Wenzel D Hornig C Wunsch E Weissenborn U Grunwald K Buttner R Weich H Poor responder-high responder: the importance of soluble vascular endothelial growth factor receptor 1 in ovarian stimulation protocols Hum Reprod 2001 16 621 626 11278207 10.1093/humrep/16.4.621
Gruemmer R Klein-Hitpass L Neulen J Regulation of gene expression in endothelial cells: the role of human follicular fluid J Mol Endocrinol 2005 34 37 46 15691876 10.1677/jme.1.01589
Clark DE Smith SK He Y Day KA Licence DR Corps AN Lammoglia R Charnock-Jones S A vascular endothelial growth factor antagonist is produced by the human placenta and released into the maternal circulation Biol Reprod 1998 59 1540 1548 9828203
Vuorela P Helske S Hornig C Alitalo K Weich H Halmesmaki E Amniotic fluid – soluble vascular endothelial growth factor receptor-1 in preeclampsia Obstet Gynecol 2000 95 353 357 10711543 10.1016/S0029-7844(99)00565-7
Dimitrakova ED Dimitrakov JD Karumanchi SG Pehlivanov BK Milchev NP Dimitrakov DI Placental soluble fms-like tyrosine-kinase-1 (sFlt-1) in pregnant women with preeclampsia Folia Med (Plovdiv) 2004 46 19 21 15362808
Levine RJ Maynard SE Qian C Lim K-H England LJ Yu KF Schisterman EF Thadhani R Sachs BP Epsein FH Sibai BM Suhhatme VP Karumanchi SA Circulating angiogenic factors and the risk of preeclampsia N Engl J Med 2004 350 672 683 14764923 10.1056/NEJMoa031884
Seli E Zeyneloglu HB Senturk LM Bahtiyar OM Olive DL Arici A Basic fibroblast growth factor: peritoneal and follicular fluid levels and its effect on early embryonic development Fertil Steril 1998 69 1145 1148 9627307 10.1016/S0015-0282(98)00074-0
Koga K Osuga Y Tsutsumi O Momoeda M Suenaga A Kugu K Fujiwara T Takai Y Yano T Taketani Y Evidence for the Presence of Angiogenin in Human Follicular Fluid and the Up-Regulation of Its Production by Human Chorionic Gonadotropin and Hypoxia J Clin Endocrinol Metab 2000 85 3352 3335 10999833 10.1210/jc.85.9.3352
Krasnow JS Berga SL Guzick DS Zeleznik AJ Yeo KT Vascular permeability factor and vascular endothelial growth factor in ovarian hyperstimulation syndrome: a preliminary report Fertil Steril 1996 65 552 555 8774285
Frederick JL Shimanuki T diZerega GS Initiation of angiogenesis by human follicular fluid Science 1984 224 389 390 6200930
He Y Smith SK Day KA Clark DE Licence DR Charnock-Jones DS Alternative splicing of vascular endothelial growth factor (VEGF)-R1 (FLT-1) pre-mRNA is important for the regulation of VEGF activity Mol Endocrinol 1999 13 537 545 10194760 10.1210/me.13.4.537
Huckle WR Roche RI Post-Transcriptional Control of Expression of sFlt-1, an Endogenous Inhibitor of Vascular Endothelial Growth Factor J Cell Biochem 2004 93 120 132 15352169 10.1002/jcb.20142
Krussel JS Casan EM Raga F Hirchenhain J Wen Y Huang HY Bielfeld P Polan M Expression of mRNA for vascular endothelial growth factor transmembranous receptors Flt1 and KDR, and the soluble receptor sFlt in cycling human endometrium Mol Hum Reprod 1999 5 452 458 10338368 10.1093/molehr/5.5.452
Redmer DA Rone JD Goodman AL Evidence for a non-steroidal angiotropic factor from the primate corpus luteum: stimulation of endothelial cell migration in vitro Proc Soc Exp Biol Med 1985 179 136 140 3991594
|
16248895
|
PMC1276815
|
CC BY
|
2021-01-04 16:37:13
|
no
|
Reprod Biol Endocrinol. 2005 Oct 25; 3:57
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-57
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-581625091210.1186/1477-7827-3-58ResearchInfluences of dietary soy isoflavones on metabolism but not nociception and stress hormone responses in ovariectomized female rats Bu Lihong [email protected] Kenneth DR [email protected] Edwin D [email protected] Physiology and Developmental Biology Department and Neuroscience Center, Brigham Young University, Provo, Utah, 84602, USA2 Department of Pediatrics, Children's Hospital Medical Center, Cincinnati, OH, 45229, USA2005 26 10 2005 3 58 58 15 8 2005 26 10 2005 Copyright © 2005 Bu et al; licensee BioMed Central Ltd.2005Bu et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Isoflavones, the most abundant phytoestrogens in soy foods, are structurally similar to 17beta-estradiol. Few studies have examined the nociception and stress hormone responses after consumption of soy isoflavones.
Methods
In this study, ovariectomized (OVX) female Long-Evans rats were fed either an isoflavone-rich diet (Phyto-600) or an isoflavone-free diet (Phyto-free). We examined the effects of soy isoflavones on metabolism by measuring body weights, food/water intake, adipose tissue weights as well as serum leptin levels. Also, circulating isoflavone levels were quantified. During chemically induced estrous, nociceptive thresholds were recorded. Then, the animals were subjected to a stressor and stress hormone levels were quantified.
Results
Body weights were significantly lower in Phyto-600 fed rats compared to Phyto-free values within one week and during long-term consumption of soy isoflavones. Correspondingly, Phyto-600 fed animals displayed significantly less adipose deposition and lower serum leptin levels than Phyto-free values. However, rats on the Phyto-600 diet displayed greater food/water intake compared to Phyto-free levels. No changes in thermal pain threshold or stress hormone levels (ACTH and corticosterone) were observed after activation of the hypothalamic-pituitary-adrenal (HPA) stress axis.
Conclusion
In summary, these data show that consumption of soy isoflavones 1) increases metabolism, demonstrated by significantly decreased body weights, adipose tissue deposition and leptin levels, but 2) does not alter nociception or stress hormone responses, as indexed by thermal pain threshold, serum corticosterone and ACTH levels in chemically-induced estrous OVX rats.
==== Body
Background
Phytoestrogens are naturally occurring, plant derived, non-steroid molecules that are structurally similar to 17beta-estradiol [1]. Of all the phytoestrogens, soy-derived isoflavones are the most abundant in rodent and human diets and the most studied in both animal and clinical research. Dietary soy isoflavones exist as active aglycones (daidzein and genistein) and inactive glucosides (mainly daidzin and genistin). When consumed, glucosides are hydrolyzed by intestinal glucosidases, which release the aglycones, daidzein and genistein. Daidzein can be further metabolized to a potent and abundant molecule in rodents, equol [2]. The structural similarity between isoflavones and 17beta-estradiol enables isoflavones to exert moderate estrogenic or antiestrogenic properties via mammalian estrogen receptors (ER). It is well established that genistein has a greater affinity for ER beta than ER alpha [3]. Recently, equol has been reported to be anti-androgenic by blocking 5 alpha-dihydrotestosterone (DHT) in the circulation and presumably within cells [4]. Moreover, equol appears to bind ER beta > ER alpha, similar to that of genistein [2]. The influences of equol are likely to be substantial due to the high level of equol (1,000~2,500 ng/ml) compared to 17beta-estradiol (10~100 pg/ml) in the circulation of rodents [4], that consume a soy-based diet.
Of all the studies examining the effects of soy isoflavones, most investigative attention has been on age-related diseases and hormone-dependent cancers [1,3,5-9]. However, few studies have examined the influence of soy isoflavones on metabolism, nociception and stress hormone responses. It is known from our previous studies that dietary soy isoflavones significantly increase food/water intake while at the same time significantly decreasing body weight in intact male and female rats [4,10]. In reference to pain thresholds, in one study, dietary soy consumption suppressed neuropathic pain in rats after partial sciatic nerve ligation [11]. Also, it has been reported that genistein and daidzein decreased cortisol synthesis by suppressing the activity of P450c21 in cultured adrenal cortical cells [12]. Phytoestrogens have received increased research interest as an alternative to estrogen replacement therapy due to their selective estrogen receptor modulator (SERM)-activity [1]. However, effects of phytoestrogens on nociception and the hypothalamic-pituitary-adrenal (HPA) stress axis during steroid replacement therapy have not been investigated in vivo in previous studies. In this study, we used OVX Long-Evans rats to mimic the surgically postmenopausal condition and studied the effects of dietary soy isoflavones on metabolism by measuring body weight, food/water intake, adipose tissue deposition as well as serum leptin levels. Then, steroid replacement therapy was reproduced in OVX rats with the administration of a steroid regimen (estrogen then progesterone) to induce an LH surge and estrus. Nociception and stress hormone responses were studied by examining thermal pain thresholds, serum corticosterone and ACTH levels during the chemically-induced estrous state.
Methods
Animals
OVX Long-Evans female rats at 50 days old were purchased from Charles River Laboratories (Wilmington, MA, USA). These animals were ovariectomized prior to shipping. They were caged individually and housed on a light/dark schedule (lights on 0600–1900 h) in the Brigham Young University Bio-Ag vivarium. The animals and methods of this study were approved by the Institute of Animal Care and Use Committee (IACUC) at Brigham Young University.
Treatment-Diets
The diet from the supplier (Ziegler Bros., NIH07, Gardner, PA, USA) contained approximately 200 ppm isoflavones. Upon arrival the animals were allowed ad libitium access to water and either a commercially available diet with high isoflavone levels (Harlan Teklad Rodent Diet 8604, Madison, WI, USA) containing approximately 600 ppm of soy isoflavones (referred to hereafter as the Phyto-600 diet), or a custom diet (Ziegler Bros., Sterol free diet, Gardner, PA, USA) containing approximately 10–15 ppm of soy isoflavones (referred to hereafter as the Phyto-free diet) [13,14]. The composition of these diets is described in detail elsewhere [14]. The diets were balanced and matched for equivalent percentage content of protein, carbohydrate, fat, amino acids, vitamins and minerals, etc. Circulating serum isoflavone levels from rats maintained on these diets have been reported previously by our laboratory using GC/MS analysis [14] and are reported here for OVX rats. All the rats were sacrificed at 94 days of age after approximately 6 weeks on the diet treatments.
Weight Measurements
Body weights were measured on a Metter 1200 balance (St. Louis, MO, USA) at 50, 58 and 93 days of age, when the animals were delivered, one week and six weeks after consuming the treatment diets, respectively. White (in the abdominopelvic cavity surrounding reproductive tissues) and brown adipose (inter-scapular) tissues were dissected and weighed on a Sartorious balance (Brinkman Inst. Co., Westbury, NY, USA). Food intake was measured on a Metter 1200 balance and water intake was measured in drinking tubes for three consecutive days at the age of 85 days. For each animal, the average food and water intake per day was calculated.
Pain Threshold Level and Stress Response
Two days before pain threshold was measured, all OVX rats received a subcutaneous injection of 0.1 mg estradiol benzoate (EB) in 0.1 ml oil at 1200 h. A subcutaneous injection of 1.0 mg progesterone in 0.1 ml oil was administered 42 hours after the EB injection [15,16]. Six hours after the progesterone injection, pain threshold levels were measured (during 1200~1230 h or during the steroid-induced LH surge) [17] using a hot-plate analgesic meter (Columbus Instruments, Columbus, Ohio, USA). Serum was collected at 40 hours after EB injection and 6 hours after the progesterone injection for the detection of LH levels with a kit from Diagnostic Systems Laboratories (Webster, Texas, USA). After the body weight of each animal was measured, one animal was placed on the hot-plate apparatus (uniform plate temperature was set at 50°C) and the time in seconds was recorded until the animal licked its back paw (latency to paw lick). Immediately following the hot-plate test, each animal was subjected to restraint, illumination and heat stress for 5 minutes, as previously performed in our laboratory [18]. In general, the animals were placed in a 150 mm (l) × 60 mm (w) × 40 mm (h) Plexiglas tube. The tube was under the illumination of two 150-W flood-lamps (generating 2200 lm/m2) and the intratube temperature reached 31–34°C during the stress interval.
Stress Hormones and Leptin RIA
Immediately after the exposure to the stress protocol (above), each rat was sacrificed, trunk blood collected, serum prepared and stored at -20°C until assayed for hormone levels. Serum corticosterone levels were assayed with a kit from Diagnostic Products Corporation (Los Angeles, CA, USA). Serum ACTH concentrations were quantified with a kit from Diagnostic Systems Laboratories (Webster, Texas, USA). Serum leptin levels were determined by a kit from Linco Res. Inc. (St. Charles, MO, USA). All samples were run in duplicate in the respective assays with internal control samples and the intra-assay coefficients of variance were 7% (corticosterone), 9% (ACTH) and 10% (leptin).
Statistical Analysis
All the data were expressed as Mean ± SEM and analyzed by the Statistical Analysis System (SAS). The data were tested by the two-sample Student's t-test, or where appropriate by repeated measures based mixed-model analysis and considered significantly different at p < 0.05.
Results
Circulating serum isoflavones levels by diet treatment
Figure 1 illustrates the levels and types of isoflavones from 94 day-old OVX rats fed the Phyto-600 versus the Phyto-free diet from 50 days of age. The rats fed on the Phyto-600 diet displayed significantly higher levels of daidzein, genistein, equol (a known metabolite of daidzein) and total isoflavone compared to animals on the Phyto-free diet. Equol was the major circulating isoflavones, whereas, daidzein and genistein represented lower percentages of the total circulating isoflavones.
Figure 1 Circulating serum soy isoflavone levels of 94-day-old OVX Long-Evans rats in ng/ml fed either an isoflavone-rich (600) or an isoflavone-free (free) diet. The analysis was performed by GC/MS. The Phyto-free fed animals displayed significantly lower daidzein, genistein, equol, or total isoflavone levels compared to the Phyto-600 values (*).
Effects on metabolism of dietary soy isoflavones
Body weights were significantly lower in Phyto-600 fed OVX rats compared to Phyto-free group within one week and after long-term (approximately six week) consumption of soy isoflavones [Figure 2]. Interestingly, the Phyto-600 fed OVX rats displayed significantly higher food and water intake (36.3 ± 1.2 g and 45.0 ± 1.6 ml) compared to the Phyto-free fed animals (23.3 ± 0.8 g and 36.3 ± 1.2 ml) at approximately 85 days of age [Figure 3].
Figure 2 Effects of dietary soy isoflavones on body weight in OVX Long-Evans rats, fed either an isoflavone-rich (600) or an isoflavone-free (free) diet. At 58 and 93 days old, 600 body weights (*) were significantly lower compared to free fed OVX rats.
Figure 3 Effects of dietary soy isoflavones on average food and water intake per day in 85 day-old OVX Female Long-Evans Rats. Rats fed an isoflavone-rich (600) diet displayed significantly less (*) food (A) and water intake (B) compared to rats fed an isoflavone-free (free) diet.
Exposure to dietary soy isoflavones significantly decreased white and brown adipose tissue weights of OVX rats [Figure 4]. White adipose tissue deposition in the Phyto-600 rats (9.4 ± 0.5 g) was about 50% less than that in the Phyto-free group (19.1 ± 1.9 g). Correspondingly, serum leptin levels of Phyto-600 fed OVX rats were significantly lower compared to Phyto-free values [6.1 ± 0.9 vs. 8.6 ± 0.6, ng/ml ± SEM, Figure 5].
Figure 4 Effects of dietary soy isoflavones on white adipose tissue (WAT) and brown adipose tissue (BAT) weights from 94 day-old OVX female Long-Evans rats. White adipose tissue and brown adipose tissue weights were significantly lower in rats fed the 600 diet (*) compared to the free fed animals.
Figure 5 Effects of dietary soy isoflavones on serum leptin levels. The 600 fed animals displayed significantly lower leptin levels (*) compared to the free fed animals, a finding consistent with the white adipose tissue weight results (by diet treatment) shown in figure 4.
Pain threshold
To evaluate the effects of dietary soy isoflavones on nociception during the sequential injection of steroid-induced LH surge, the thermal pain threshold was measured as the latency to paw lick (time in seconds on the 50°C hot plate until licking the back paw). The LH surge was confirmed to be induced with sequentially injection of EB and progesterone by measuring the LH levels in serum [40 hours after EB injection: Phyto-600 values = 4.7 ± 0.8 vs. Phyto-free values = 4.5 ± 0.4; 6 hours after progesterone injection: Phyto-600 values = 21.1 ± 1.4 vs. Phyto-free values = 21.9 ± 1.6; ng/ml ± SEM; data not shown graphically]. No significant difference was observed between the diet treatment groups [Phyto-600 values = 44.5 ± 6.7 vs. Phyto-free values = 45.4 ± 5.3; second ± SEM, Figure 6] for the pain threshold parameter (latency to paw lick).
Figure 6 Effects of dietary soy isoflavones on nociception. The 600 fed animals displayed similar thermal pain threshold, latencies to paw lick [in seconds (s)] on the hot plate, to the free fed animals.
Stress hormones
Immediately after the pain threshold test, the animals were restrained for 5 minutes (see methods above) and the stress hormones (corticosterone and ACTH) were determined from serum samples, collected immediately after the stress interval. No significant changes of either corticosterone [Phyto-600 values = 85.8 ± 10.0 vs. Phyto-free values = 87.3 ± 6.5; μg/dl ± SEM; Figure 7] or ACTH [Phyto-600 values = 505 ± 134 vs. Phyto-free values = 418 ± 58; pg/ml ± SEM; Figure 7] were detected with the consumption of soy isoflavones during the chemically induced LH surge.
Figure 7 Effects of dietary soy isoflavones on stress hormones from 94 day-old OVX Long-Evans rats after activation of the hypothalamic-pituitary-adrenal (HPA) stress axis during chemically induced estrus. No significant changes on corticosterone (A) or ACTH (B) were observed between the Phyto-600 and Phyto-free group.
Discussion
The effects of dietary phytoestrogens on metabolism, nociception and stress response were studied in the postmenopausal condition, mimicked by using 94 day-old OVX Long-Evans rats. The present data showed that consumption of soy isoflavones increases metabolism, demonstrated by significantly decreased body weight, adipose tissue deposition and leptin levels. But during a chemically induced LH surge, dietary soy isoflavones do not alter nociception or stress hormone responses, as indexed by thermal pain threshold, serum corticosterone and ACTH levels.
We have previously reported that a variety of behavioral, metabolic and neuroendocrine parameters are influenced by consumption of soy isoflavones in intact animals [4,10,14]. The increased metabolism observed in this study from OVX rats is in agreement with our previous findings in males, showing significant decreases in body weight, white and brown adipose tissue weight and serum leptin levels [19,20]. Notably, the decrease of body weight in OVX rats, in the present study, was significant within about one week's consumption of the Phyto-600 diet. It is reported that genistein has the ability to increase lipolysis in isolated rat adipocytes and decrease lipogenesis in white adipose tissue in OVX rats [21,22]. This may account, at least in part, for the reduced body weights in Phyto-600 fed OVX animals. Additionally, it has been shown that locomotor activity is increased in Phyto-600 fed animals, along with increases in thyroid hormone levels [14,20]. This may play a role in further reducing body weights, even though the food/water intake of these animals is significantly greater than Phyto-free fed animals, which is in accord with other studies in OVX animals [23]. Since leptin is synthesized and secreted from white adipose tissue, it is not surprising to find the significantly reduced leptin levels in Phyto-600 animals compared to Phyto-free group. Consequently, the higher hypothalamic NPY levels, as a result of negative feedback from lower leptin levels, stimulate feeding behavior [20]. The finding of decreased brown adipose tissue weight is similar to the results from another study in which male rats consumed soy isoflavones [20]. However, the biological effects of this decrease are not clear. Hence, further studies are necessary to determine the mechanisms of how metabolic levels are increased by the consumption of soy isoflavones.
The use of natural remedies is omnipresent in addressing women's health issues, especially in light of the negative risk factors associated with estrogen replacement therapy in postmenopausal women [1,5,24,25]. Some recent studies suggest that soy consumption (especially via dietary supplements) is effective in treating symptoms of peri- and postmenopause [26]. In this regard, OVX rats were administered a steroid regimen (estrogen then progesterone) to induce an LH surge and estrus, in order to reproduce, in part, steroid replacement therapy [27]. It has been shown that responsiveness to noxious stimuli change after gonadal steroid treatment and during the estrous cycle [28]. There is evidence that LHRH may interact with central opioid systems causing an increased sensitivity to nociceptive stimulation (hyperalgesia) and reduction of the antinociceptive effect of morphine in female rats [29]. Thus, using this steroid paradigm in the present study, induced LH surges were stimulated in all the OVX rats and thus one might predict that pain sensitivities would increase. However, no significant difference was observed between Phyto-600 vs. Phyto-free fed OVX animals in the present pain threshold experiment, and our recorded time (paw lick) latencies were within range of previous pain threshold studies, but higher than those recorded in intact male rats (Phyto-600 values = 25.94 ± 3.1 vs. Phyto-free values = 24.14 ± 2.5; second ± SEM) [30]. While prior examination of this parameter is scarce in relation to soy consumption, a recent report did not show a reduction in thermal pain in soy fed animals [11], a finding similar to that of our present results.
It was reported that genistein and daidzein decrease cortisol synthesis by suppressing the activity of P450c21 in cultured adrenal cortical cells [12], our data showed similar post-stress corticosterone levels between Phyto-600 and Phyto-free fed OVX female rats. The lack of detecting any difference in corticosterone levels, in the present in vivo study compared to prior in vitro results [12], is most likely due to the adrenal cortex being maximally stimulated by the high ACTH levels in both the Phyto-600 and Phyto-free groups [31].
In intact male rats, results obtained after similar stressor, ACTH levels were significantly greater in the Phyto-600 vs. the Phyto-free fed group, signifying that dietary soy isoflavones may increase the hypothalamic-pituitary stress response [30]. The finding of no differences in ACTH levels, in the present OVX rats, suggests that gender differences exist in the biological actions of dietary soy isoflavones, especially in relationship to the stress response where females typically display significantly higher stress hormone levels compared to males [32]. Finally, future time-dependent stress response studies may determine whether consumption of soy isoflavones reduce glucocorticoid release from the adrenal cortex.
Conclusion
The present in vivo study shows that consumption of a commercially available soy-based rodent diet increases metabolism, demonstrated by significantly decreased body weights, adipose tissue deposition and leptin levels in OVX female rats. However, in female rats studied here, nociception and stress hormone responses, as indexed by thermal pain threshold, serum corticosterone and ACTH levels, were not influenced by consumption of soy isoflavones. However, whether the effects are due to soy protein or to differences in isoflavones is unknown. Further research is warranted due to the pervasive biological actions of dietary soy isoflavones altering physiology and behavior.
Acknowledgements
This work was supported, in part, by grants from the USDA (2002-00798 and 2004-01811; EDL), and The BYU Dean's Graduate Fellowship in Neuroscience (LB).
==== Refs
Knight DC Eden JA A review of the clinical effects of phytoestrogens Obstet Gynecol 1996 87 897 904 8677131
Setchell KDR Brown NM Lydeking-Olsen E The clinical importance of the metabolite equol – a clue to the effectiveness of soy and its isoflavones J Nutr 2002 132 3577 3584 12468591
Kuiper GG Lemmen JG Carlsson B Corton JC Safe SH van der Saag PT ven der Burg B Gustafusson JA Interaction of estrogenic chemicals and phytoestrogens with estrogen receptor beta Endocrinology 1998 139 4252 4263 9751507 10.1210/en.139.10.4252
Lephart ED Setchell KD Handa RJ Lund TD Behavioral effects of endocrine-disrupting substances: phytoestrogens ILAR J 2004 45 443 454 15454683
Murkies AL Wilcox G Davis SR Clinical Review 92: Phytoestrogens J Clin Endocrinol Metab 1998 83 297 303 9467531 10.1210/jc.83.2.297
Setchell KDR Phytoestrogens: the biochemistry, physiology, and implications for human health of soy isoflavones Am J Clin Nutr 1998 68 1333S 1346S 9848496
Whitten PL Patisaul HB Cross-species and interassay comparisons of phytoestrogen action Environ Health Perspect 2001 109 5 20 11250801
Adlercreutz H Mazur W Phyto-oestrogens and Western diseases Ann Med 1997 29 95 120 9187225
Setchell KDR Cassidy A Dietary isoflavones: biological effects and relevance to human health J Nutr 1999 129 758S 767S 10082786
Lephart ED Adlercreutz H Lund TD Dietary soy phytoestrogens effects on brain structure and aromatase in Long-Evans rats Neuro Report 2001 12 3451 3455
Shir Y Campbell JN Raja SN Seltzer Z The correlation between dietary soy phytoestrogens and neuropathic pain behavior in rats after partial denervation Anesth Analg 2002 94 421 426 11812712 10.1097/00000539-200202000-00037
Mesiano S Katz SL Lee JY Jaffe RB Phytoestrogens alter adrenocortical function: genistein and daidzein suppress glucocorticoid and stimulate androgen production by cultured adrenal cortical cells J Clin Endocrinol Metab 1999 84 2443 2448 10404819 10.1210/jc.84.7.2443
Miksicek RJ Commonly occurring plant flavonoids have estrogenic activity Mol Pharmacol 1993 44 37 43 8341277
Weber KS Setchell KDR Stocco DM Lephart ED Dietary soy-phytoestrogens decrease testosterone levels and prostate weight; without altering LH; prostate 5α-reductase or testicular StAR levels in adult male Sprague-Dawley rats J Endocrinol 2001 170 591 599 11524239 10.1677/joe.0.1700591
Caligaris L Astrada JJ Taleisnik S Biphasic effect of progesterone on the release of gonadotropin in rats Endocrinology 1971 89 331 337 5557987
Nagle CA Rosner JM Plasma norepinephrine during the rat estrous cycle and after progesterone treatment to the ovariectomized estrogen-primed rat Neuroendocrinology 1976 22 89 96 1035969
Rhees RW Kirk BA Sephton S Lephart ED Effects of Prenatal Testosterone on Sexual Behavior, Reproductive Morphology and LH Secretion in the Female Rat Dev Neurosci 1997 19 430 437 9323463
Lephart ED Watson MA Jacobson NA Rhees RW Ladle DR Calbindin-D28k is Regulated by Adrenal Steroids in Hypothalamic Tissue During Perinatal Development Dev Brain Res 1997 100 117 121 9174253 10.1016/S0165-3806(97)00022-9
Niswender KD Schwartz MW Insulin and leptin revisited: adiposity signals with overlapping physiological and intracellular signalling capabilities Front Neuroendocrinol 2003 24 1 10 12609497 10.1016/S0091-3022(02)00105-X
Lephart ED Porter JP Lund TD Bu L Setchell KDR Ramoz G Crowley CR Dietary soy isoflavones alter regulatory behaviors, metabolic hormones, and neuroendocrine function in Long-Evans male rats Nutri Metab 2004 1 16 10.1186/1743-7075-1-16
Szkudelska K Nogowski L Szkudelska T Genistein affects lipogenesis and lipolysis in isolated rat adipocytes J Steroid Biochem Mole Biol 2000 75 265 271 10.1016/S0960-0760(00)00172-2
Nogowski L Mackowiak P Kandulska K Szkudelski T Nowak KW Genistein-induced changes in lipid metabolism of ovariectomized rats Ann Nutr Metab 1998 42 360 366 9895424 10.1159/000012756
Pan Y Anthony MS Binns M Clarkson TB A comparison of oral micronized estradiol with soy phytoestrogen effects on tail skin temperature of ovariectomized rats Menpause 2001 8 171 174 10.1097/00042192-200105000-00005
La Vecchia C Brinton LA McTiernan A Cancer risk in menopausal women Best Pract Res Clin Obstet Gynaecol 2002 16 293 307 12099664 10.1053/beog.2002.0283
Newton KM Buist DS Keenan NL Anderson LA LaCroix AZ Use of alternative therapies for menopause symptoms: results of a population-based survey Obstet Gynecol 2002 100 18 25 12100799 10.1016/S0029-7844(02)02005-7
Scambia G Mango D Signorile PG Anselmi Angeli RA Palena C Gallo D Bombardelli E Morazzoni P Riva A Mancuso S Clinical effects of a standardized soy extract in postmenopausal women: a pilot study Menopause 2000 7 105 11 10746892
Gudmundsson A Goodman B Lent S Barczi S Grace A Boyle L Ershler WB Carnes M Effects of estrogen replacement therapy on the circadian rhythms of serum cortisol and body temperature in postmenopausal women Exp Gerontol 1999 34 810 818 10.1016/S0531-5565(99)00044-3
Ratka A Simplins JW Effects of estradiol and progesterone on the sensitivity to pain and on morphine-induced antinociception in female rats Horm Behav 1991 25 217 228 2066081 10.1016/0018-506X(91)90052-J
Ratka A Simpkins JW A modulatory role for luteinizing hormone-releasing hormone in nociceptive responses of female rats Endocrinology 1990 127 667 673 2197079
Lephart ED Galindo E Bu L Stress (HPA axis) and pain response in male rats exposed lifelong to high vs. low phytoestrogen diets Neurosci Lett 2003 342 65 68 12727319 10.1016/S0304-3940(03)00262-3
Meaney MJ Aitken DH Sharma S Viau V Basal ACTH, corticosterone and corticosterone-binding globulin levels over the diurnal cycle, and age-related changes in hippocampal type I and type II corticosteroid receptor binding capacity in young and aged, handled and nonhandled rats Neuroendocrinology 1992 5 204 213 1320217
Rivier C Gender, sex steroids, corticotrophin-releasing factor, nitric oxide, and the HPA response to stress Pharmacol Biochem Behav 1999 64 739 751 10593197 10.1016/S0091-3057(99)00148-3
|
16250912
|
PMC1276816
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Oct 26; 3:58
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-58
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-1161621910710.1186/1465-9921-6-116ResearchRespiratory epithelial cells require Toll-like receptor 4 for induction of Human β-defensin 2 by Lipopolysaccharide MacRedmond Ruth [email protected] Catherine [email protected] Clifford C [email protected] Noel [email protected]'Neill Shane [email protected] Department of Respiratory Research, Royal College of Surgeons in Ireland, Beaumont Hospital, Dublin 9, Ireland2005 12 10 2005 6 1 116 116 27 4 2005 12 10 2005 Copyright © 2005 MacRedmond et al; licensee BioMed Central Ltd.2005MacRedmond et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The respiratory epithelium is a major portal of entry for pathogens and employs innate defense mechanisms to prevent colonization and infection. Induced expression of human β-defensin 2 (HBD2) represents a direct response by the epithelium to potential infection. Here we provide evidence for the critical role of Toll-like receptor 4 (TLR4) in lipopolysaccharide (LPS)-induced HBD2 expression by human A549 epithelial cells.
Methods
Using RTPCR, fluorescence microscopy, ELISA and luciferase reporter gene assays we quantified interleukin-8, TLR4 and HBD2 expression in unstimulated or agonist-treated A549 and/or HEK293 cells. We also assessed the effect of over expressing wild type and/or mutant TLR4, MyD88 and/or Mal transgenes on LPS-induced HBD2 expression in these cells.
Results
We demonstrate that A549 cells express TLR4 on their surface and respond directly to Pseudomonas LPS with increased HBD2 gene and protein expression. These effects are blocked by a TLR4 neutralizing antibody or functionally inactive TLR4, MyD88 and/or Mal transgenes. We further implicate TLR4 in LPS-induced HBD2 production by demonstrating HBD2 expression in LPS non-responsive HEK293 cells transfected with a TLR4 expression plasmid.
Conclusion
This data defines an additional role for TLR4 in the host defense in the lung.
Airway epitheliumToll-like Receptor 4LipopolysaccharideHuman β-defensin 2.
==== Body
Introduction
The lung represents the largest epithelial surface in the body and is a major portal of entry for pathogenic microorganisms. It employs a number of efficient defense mechanisms to eliminate airborne pathogens encountered in breathing, including the specific innate and adaptive immune responses, which represent a dynamic interaction of host and pathogen. Lipopolysaccharide (LPS) is an important antigenic component of Gram-negative bacteria, and is a potent stimulus to local and systemic immune responses. The human receptor for LPS is Toll-like-receptor 4 (TLR4) [1].
TLRs are a family of pattern recognition receptors whose pivotal importance in orchestrating the innate immune response is widely accepted. Binding of ligand activates a signaling cascade involving TRAF6, IKKs and I-κBs, culminating in NF-κB translocation to the nucleus [1]. NF-κB regulates the inducible expression of cytokines, chemokines, adhesion molecules and acute phase proteins which activate cellular immune responses [2]. TLR signaling pathways arise from intracytoplasmic Toll/IL-1 receptor (TIR) domains, which are conserved among TLRs and TIR domain-containing adaptor proteins such as MyD88, Mal/TIRAP and TRIF/TICAM-1. These adaptor proteins confer specificity on TLR signaling, with Mal specifically involved in MyD88-dependent signaling via TLR2 and TLR4, and TRIF in the MyD88-independent TLR3- and TLR4- signaling [3]
The mammalian innate immune system produces a variety of anti-microbial peptides (AMPs) as part of its host defense repertoire. The defensins are a broadly dispersed group of AMPs, and are classified according to their molecular structure into three distinct families: the α-, β- and the θ-defensins. Unlike α-defensins, which are produced mainly by neutrophils, β-defensins are produced directly by epithelial cells, and combat infection both through direct microbicidal action and by modulation of cell-mediated immunity [4-7]. To date, four human β-defensins (HBD) have been identified (HBD1-4), although genomic studies suggest more have yet to be discovered [8,9]. In contrast to HBD1, which is constitutively and stably expressed, HBD2 expression is induced in response to infective stimuli, including Gram-negative and, less potently, Gram-positive bacteria or their components or to proinflammatory stimuli including tumor necrosis factor α (TNFα) and interleukin-1β (IL-1β) in vitro [10,11].
Like other defensins, HBD2 has a broad spectrum of antimicrobial activity, displaying potent microbicidal activity against many Gram-negative bacteria and less potent bacteriostatic activity against Gram-positive bacteria [11]. It has recently been demonstrated that activation of TLR2 by bacterial lipoprotein results in up regulation of HBD2 in tracheobronchial epithelium [12]. LPS and Gram-negative bacteria such as mucoid P. aeruginosa are a more potent stimulus for HBD2 production, which in turn has anti-bacterial activity predominantly against Gram-negative bacteria. Colonisation and infection due to Gram-negative bacteria are important in many pulmonary diseases including severe COPD [13] and Cystic Fibrosis [14]. Production of HBD2 by respiratory epithelium is an important component of host defense against Gram-negative organisms, and understanding of the signaling pathways involved may further our understanding of and guide future therapeutic strategies in these diseases.
Cultured intestinal epithelial cells have been shown to produce HBD2 in response to LPS following transfection with TLR4 and MD2 [15]. Although CD-14 is known to be critical to LPS-induced HBD2 production in airway epithelium [16], the role of TLR4 in transcriptional regulation of HBD2 in respiratory epithelium has not been established. Indeed, the importance of the respiratory epithelium in the innate immune response to LPS has been called into question by some recent publications [17,18]. In this study we demonstrate TLR4 expression in A549 pulmonary epithelial cells and production of HBD2 in response to LPS. We examine the effect of modulation of TLR4 by receptor blockade and expression of a dominant negative TLR4 construct on induced expression of HBD2. We show that LPS-unresponsive HEK293 cells can produce HBD2 in response to LPS following transfection with TLR4 and MD2 transgenes and demonstrate that the adaptor proteins MyD88 and Mal are involved in transcriptional regulation of HBD2 in response to LPS.
Methods
Cell lines and culture
The human embryonic kidney cell line, HEK293, (ECACC-85120602) was obtained from the European Collection of Cell Cultures. Cells were cultured at 37°C in 5% CO2 in Eagle's minimal essential medium (EMEM, Biowhittaker) supplemented with 10% fetal calf serum (FCS), 1% L-glutamine, 1% penicillin/streptomycin, 1% NEAA (Gibco-BRL). The type II-like human lung epithelial cell line A549, (European Collection of Cell Cultures, Porton Down, UK) were cultured in Ham's F12 (Gibco-BRL), 10% FCS, 1% penicillin/streptomycin. Prior to agonist treatment, cells were washed with serum-free EMEM/F12 and placed under serum-free conditions or in serum containing 1% FCS for LPS stimulation experiments, including control conditions.
Reverse Transcription (RT)-PCR
RNA isolation and cDNA synthesis were performed as previously described [19]. The integrity of RNA extraction and cDNA synthesis was verified by PCR by measuring the amounts of GAPDH cDNA in each sample using GAPDH-specific primers to generate a 211 bp product. PCR reactions were performed using standard conditions [19] with 50 pmol each of gene-specific primers (Table 1). After an initial step of 95°C for 5 min, thermocycling conditions were 35 cycles of 95°C for 30 sec, 55°C (TLR4 or CD14) or 58°C (MD2, Mal and GAPDH) for 30 sec and 72°C for 1 min per kb, followed by a final extension step of 72°C for 10 min. The more abundant GAPDH was amplified using 25 cycles. Control PCR reactions using an RNA template failed to generate any products. Products were resolved on 1.5% TBE agarose gels containing 0.5 μg/ml ethidium bromide (Sigma) and images were captured using the GeneGenius Gel Documentation and Analysis System (Syngene), analysed by densitometry and compared in a semi quantitative manner using ImageMaster® TotalLab Software (Amersham Pharmacia, Amersham, UK). The ratio of PCR fragment intensities of HBD2 relative to GAPDH was determined. All expression values were verified by at least two independent RT-PCRs.
Table 1 Gene (Accession No.) Primers (5'-3') Bases Product Size
TLR4 (NM_003266)
F AGATGGGGCATATCAGAGC 569–587 481 bpa
R GTCCATCGTTTGGTTCTGG 1068–1050
CD14 (NM_000877)
F TACTCCCGCCTCAAGGAA 459–476 197 bp
R GCTTGGGCAATGCTCAGT 655–638
MD2 (NM_000877)
F GCAACTCATCCGATGCA 95–112 225 bp
R CATCAGATCCTCGGCAAA 319–302
HBD2 (NM_AF071216)
F GGTATAGGCGATCCTGTTACC TGC 2688–2709 202 bp
R TCATGGCTTTTTGCAGCA TTTTGTTC 4542–4567
GAPDH (BC004109)
F AACTCTGGTAAAGTGGAT 122–138 211 bp
R TACTCAGCGCCAGCATCG 333–316
a bp, base pairs
IL-8 Production
Cells (1 × 105) were left untreated or stimulated with LPS from Pseudomonas aeruginosa 01, human recombinant TNF-α or IL-1β (R&D systems). IL-8 protein concentrations in the cell supernatants were determined by sandwich ELISA (R & D Systems, U.K.). All assays were performed in duplicate or triplicate a minimum of three times.
Preparation of membrane and cytosolic protein extracts
Cells were suspended in 1 ml ice-cold PBS and pelleted by centrifugation at 10,000 rpm for 5 min at 4°C. Supernatant was removed and the pellet resuspended in 100 μl hypotonic buffer A (5 mM Tris (pH 6.8), 2 mM EDTA, leupeptin 5 μg/ml, pepstatin 0.7 μg/ml, benzamidine 5 μg/ml, PMSF 1 mM)(Sigma, Ireland). Cellular components were separated by ultracentrifugation at 55000 rpm × 20 min at 4°C. Supernatant, which constituted cytosolic and nuclear fractions, was removed and stored at -20°C. The pellet consisting of the membrane fraction was resuspended in hypotonic buffer B (20 mM Tris HCl pH 6.8%, 150 mM NaCl, 10 mM EDTA, 1 mM EGTA, 1% Triton X100, leupeptin 5 μg/ml, pepstatin 0.7 μg/ml, benzamidine 5 μg/ml, PMSF 1 mM) by forcing the pellet through a 22 G needle 5–8 times. Protein concentrations of extracts were determined by the method of Bradford, and stored at -20°C.
Western Blot analysis
Extracts (10 μg of protein) were separated by electrophoresis on 10% SDS-polyacrylamide gels and transferred to nitrocellulose, blocked with 0.2% I-Block (Tropix, MA) and PBS containing 0.1% Tween-20 (Sigma, Ireland). TLR4 protein was detected using rabbit anti-TLR4 (sc-10741 Santa Cruz Biotechnology, diluted 1:200), horse-radish peroxidase-conjugated anti-rabbit IgG (Tropix, MA) and chemiluminescent LumiGlo Reagent A and Peroxide Reagent B (New England Biolabs) according to the manufacturer's instructions.
Laser Scanning Cytometry
For analysis by laser cytometry, 5 × 104cells/well were seeded in 8 well chamber slides (Nunc). Slides were fixed in methanol (AnalaR) for 5 minutes and labelled with goat anti-HBD-2 antibody (a gift from Dr Paul McCray, University of Iowa, Iowa City, IA, USA) 1:100 dilution) or isotype control (1:1000 dilution of 1 mg/ml stock of isotype goat IgG) and incubated at 4°C in the dark for 30 min. Slides were washed three times in PBS and probed with 1:10 dilution of anti-goat IgG fluorescein isothiocyanate (FITC, Dako, Glostrup, Denmark) at 4°C in the dark for 30 min. The washing step was repeated and cells were permeabilised in a 1:1 ratio of permeabilisation solution (Dako) and 0.2 μg/ml solution of propidium iodide (PI, Molecular Probes, Leiden, The Netherlands). Slides were washed in PBS and HBD-2 expression was quantified on a CompuCyte laser scanning cytometer (CompuCyte, Cambridge, MA, USA). Cellular fluorescence of at least 5 × 103 cells was measured by laser scanning cytometry. Fluorescence excitation was provided by a 488 nm laser line. Orange (PE) and green (FITC) fluorescence were measured at 588 +/- 10 nm or 530 +/- 20 nm, respectively. The threshold contour was set on scatter or orange to detect all cells as appropriate. Artificially contoured debris was gated out based on contour size. Aggregated cells were gated out using an algorithm in the LSC software that finds and marks multiple cells. Individual TLR4 or HBD-2-expressing cells were identified and quantified using CompuCyte software on the basis of integrated green fluorescence reflecting binding of anti-goat FITC antibody.
Transfection and reporter gene studies
Cells (1.5 × 105) were transfected with plasmid DNA (dominant negative (Δ) Mal (Mal P/H), ΔMyD88, ΔTLR4, MD2, wild type TLR4 or CD4-Toll plasmid) and/or 1 μg HBD2 promoter-linked luciferase reporter plasmid [20] using TransFast (Promega) according to the manufacturer's instructions. ΔMyD88 contains only a functional TIR domain and lacks the death domain required for downstream signaling, while Mal P/H is a dominant negative version of Mal with a proline to histidine point mutation in box 2 of the TIR domain. ΔTLR4 lacks an intracytoplasmic signaling domain and CD4-Toll is a constitutively active chimera of the extracellular domain of CD4 fused to the transmembrane and cytosolic domains of TLR4 [21]. Uniform transfection efficiencies were achieved by measuring expression from co-transfected luciferase or β-galactosidase expression plasmids, as appropriate. initially optimizing transfection conditions using a constitutive luciferase expression vector, pGL3-control (Promega). In all experiments, equal amounts of empty vector plasmid DNA was used in control cells, such that the total amount of transfected DNA remained constant. After 48 h cells were lysed with Reporter Lysis Buffer (1×) (Promega) (300 μl/well), protein concentrations were determined, and reporter gene activity was quantified by luminometry (Wallac Victor2, 1420 multilabel counter, Turku, Finland) using the Promega luciferase assay system according to the manufacturer's instructions.
Statistical analysis
Data were analyzed with GraphPad Prism 3.0 software package (GraphPad Software, San Diego, CA). Results are expressed as mean ± S.E. and were compared by Mann-Whitney test. Differences were considered significant when the P value was ≤ 0.05.
Results
A549 cells respond to LPS with production of IL-8 and up regulation of HBD2
Signaling via TLR4 by LPS activates NF-κB and upregulates a variety of pro-inflammatory genes, including IL-8. We investigated LPS responsiveness in A549 and HEK293 cells using IL-8 protein production as a surrogate of LPS responsiveness. Figure 1A shows that A549 cells dose-dependently induced IL-8 protein expression in response to stimulation with LPS to levels similar to those induced by IL-1β. In contrast HEK293 cells failed to induce IL-8 expression in response to LPS, but did respond to IL-1β or TNFα stimulation with increased IL-8 expression.
Figure 1 LPS-induced IL-8 and HBD2 expression in A549 and HEK293 cells. (A) A549 or HEK293 cells (3 × 105/ml) were left untreated (control) or stimulated for 24 h with LPS (10 or 50 μg/ml), IL-1β (100 ng/ml) or TNFα (10 ng/ml). Levels of IL-8 in supernatants were measured by ELISA and values are expressed as ng/ml. Assays were performed in duplicate a minimum of three times. Values are expressed as mean+/- S.E. (n = 3). (B) Total RNA was extracted from A549 or HEK293 cells, reverse transcribed into cDNA and used as a template in PCR reactions using HBD2 gene-specific primers. Products were electrophoresed in 1.5% TBE agarose gels containing 0.5 μg/ml ethidium bromide and visualized under UV. + represents positive control PCR reaction and lanes 1–4 represent untreated cells and cells stimulated for 24 hours with LPS (10 μg/ml), IL-1β (100 ng/ml) or TNFα (10 ng/ml), respectively. Gels are representative of three independent experiments using cultures from different time points.
We next investigated the effect of LPS stimulation on HBD2 gene expression in both cells lines. Compared to untreated cells, LPS induced HBD2 expression at both 10 and 50 μg/ml in A549 cells (Figure 1B). Densitometric analysis quantified these increases to be 2- and 3.6-fold, respectively. IL-1β was used as a positive control and increased HBD2 expression over 30-fold. There was no response to LPS in the HEK293 cells however both IL-1β and TNFα did up regulate HBD2 by a factor of 1.8 and 1.5 respectively.
Airway epithelial cells express TLR4 on the cell surface
We characterized the HEK293 cell line to determine whether it lacked a critical factor for LPS responsiveness. RTPCR revealed that the HEK293 cells, similar to the A549 cells, express TLR4, Mal, and CD14 mRNA but unlike the A549 cells do not express MD2 mRNA (Figure 2A). MD2 is a secreted protein who's interaction with LPS and CD-14 is necessary for the cellular response to LPS [22]
Figure 2 Characterization of A549 and HEK293 cell lines. (A) Duplicate samples of total RNA was extracted from 1 × 106 HEK293 (HEK) and A549 cells, reverse transcribed into cDNA and used as a template in PCR reactions using TLR4, Mal, CD14, MD2 and GAPDH gene-specific primers. Products were electrophoresed in 1.5% TBE agarose gels containing 0.5 μg/ml ethidium bromide and visualized under UV. (B) Western blot analysis of membrane (mem) and cytosolic (cyt) extracts (10 μg) from A549 and HEK293 cells probed with an anti-TLR4 antibody. Data are representative of three separate experiments. (C) For fluorescence microscopy, A549 cells (2 × 104) were grown in chamber slides, Fc-blocked and labelled with anti-TLR4 (clear) or isotype control antibodies (solid) and fluorophore-conjugated detection antibodies. TLR4 expression was quantified by laser scanning cytometry.
We performed western immunoblotting of cytosolic and membrane fractions from HEK293 and A549 cells to detect TLR4 protein expression. Figure 2B shows that TLR4 was present in both fractions from the A549 cells but was not evident in membrane fractions isolated from HEK293 cells. Next we quantified cell surface expression of TLR4 on A549 by fluorescence microscopy. In accord with the findings of Monick et al [17] but in contrast to Guillot, [23] we detected TLR4 on the surface of A549 cells (Figure 2C).
LPS-induced HBD2 gene and protein expression in A549 cells requires TLR4
Next the role of TLR4 in LPS-induced regulation of HBD2 (Figure 3A) in A549 cells was investigated. Compared to untreated cells (lane 1), LPS increased HBD2 expression at 24 h (lane 5). This effect was blocked by pre-treatment with a TLR4 neutralizing antibody (eBioscience, Clone HTA125, Cat. No. 16–9917) (lane 6). An isotype control antibody had no effect (data not shown). A similar effect was demonstrated at protein level using laser scanning cytometry. Stimulation with LPS resulted in a small but statistically significant increase in HBD2 protein above control (P < 0.03). This effect is inhibited by pretreatment with the TLR4 blocking antibody (P < 0.03).
Figure 3 LPS-induced HBD2 expression in A549 cells requires TLR4. A459 cells were incubated with an isotype control or anti-TLR4 neutralizing antibody (Anti-TLR4 mAB 5 μg/ml, 30 min) then, (A) left untreated or stimulated with LPS (10 μg) for 4 or 24 hours. Total RNA was extracted, reverse transcribed into cDNA and used as a template in PCR reactions using HBD2 gene-specific primers. Products were electrophoresed in 1.5% TBE agarose gels containing 0.5 μg/ml ethidium bromide and visualized under UV. Gels are representative of three independent experiments or (B) left untreated or stimulated with LPS (10 μg) for 24 hours, Fc-blocked and labeled with anti-HBD2 (solid) or isotype control antibodies (clear) and fluorophore-conjugated detection antibodies. HBD2 expression was quantified by laser scanning cytometry, as described, and data from three experiments is presented. HBD2 expression is expressed as Mean Channel Fluorescence (MCF) + SEM. (* P < 0.05 vs control, † P < 0.05 vs control + LPS).
Dominant negative TLR4 inhibits LPS-induced HBD2 expression in A549 cells
In order to further demonstrate the functional role of TLR4 in LPS-induced HBD2 expression, we examined the effect of functionally ablating TLR4 by over expression of a functionally inactive ΔTLR4 construct in A549 cells. Because the upregulation of HBD2 by LPS as measured by semi-quantitative RTPCR was small, albeit statistically significant, we further quantified the effect by luciferase activity of a co-transfected HBD2 promoter-linked luciferase construct (Figure 4). LPS increased both HBD2 mRNA expression (Figure 4A) (P ≤ 0.05) and HBD2 promoter activity (Figure 4B) (P ≤ 0.03) however over expression of ΔTLR4 significantly inhibited both effects (P ≤ 0.05 and 0.03 compared to LPS-treated empty vector-transfected cells, respectively).
Figure 4 ΔTLR4 inhibits LPS-induced HBD2 expression in A549 cells. A549 cells (1.5 × 105) were transfected with pcDNA3 (empty vector) or a ΔTLR4 expression plasmid. 24 h post transfection, cells were left untreated or stimulated with LPS (10 μg/ml) for 24 h. (A) Total RNA was extracted, reverse transcribed into cDNA and used as a template in semi-quantitative PCR reactions using HBD2 and GAPDH gene-specific primers. HBD2 expression was given an arbitrary value of 1 in control cells. Data are expressed as mean +/- S.E. and are obtained from three experiments (* P < 0.05 vs control, † P < 0.05 vs control + LPS). (B) Duplicate experiments were performed in cells cotransfected with a HBD2 promoter-linked luciferase reporter plasmid. Cells were lysed and reporter gene activity was quantified by luminometry. Data are expressed as HBD2 luciferase activity (n = 3). (* P < 0.05 vs control, † P < 0.05 vs control + LPS)
TLR4/MD2 transgene expression confers LPS-responsiveness on HEK293 cells
Having established the inhibitory effect of neutralizing TLR4 in A549 cells by receptor blockade and over-expression of a dominant negative construct, we next determined whether HEK293 cells could be rendered responsive to LPS by expression of functional TLR4 and MD2. Expression of CD4/Toll (a constitutively active TLR4 chimera [21]) and MD2 resulted in significant induction of the HBD2 promoter (Figure 5A).
Figure 5 TLR4/MD2 transgene expression confers LPS-responsiveness on HEK293 cells. HEK293 cells (1.5 × 105) were cotransfected with MD-2 and CD4/Toll or full-length TLR4 expression plasmids and a HBD2 promoter-linked luciferase reporter gene. Equal amounts of the corresponding empty vector were transfected into control cells such that the total amount of transfected DNA remained constant. Uniform transfection efficiencies were confirmed using a β-galactosidase reporter plasmid. 24 hours post transfection, cells were left untreated or stimulated with LPS (10 μg/ml) then lysed and reporter gene activity was quantified by luminometry. Data are expressed as HBD2 luciferase activity. Assays were performed in duplicate and are representative of at least three separate experiments. (* P < 0.05, ** P < 0.005, *** P < 0.001 vs control).
Stimulation of HEK293 cells with LPS has no effect on HBD2 promoter activity (Figure 5B). However transfection with MD2 and TLR4 transgenes resulted in significant up regulation of HBD2 promoter activity (P ≤ 0.005 compared to LPS-treated empty vector-transfected cells), an effect that was further significantly augmented by stimulation with LPS (P ≤ 0.05 compared to untreated MD2/TLR4-transfected cells).
Expression of dominant negative MyD88 and Mal constructs inhibits LPS-stimulated HBD2 expression
Having demonstrated that LPS up regulates HBD2 expression via TLR4, we went on to elucidate the pathway by which the signal is transduced to the nucleus. Figure 6 shows that transfection of dominant negative constructs of the TIR domain containing adaptor proteins MyD88 and Mal alone or in combination dose-dependently inhibited HBD2 expression in response to LPS. Technical restrictions of performing transfection experiments in chamber slides precluded measurement of HBD2 protein in this part of the study. We would intuitively expect a similar qualitative effect in HBD2 mRNA and protein expression, as demonstrated in Figure 3.
Figure 6 ΔMyD88 and ΔMal inhibit LPS-induced HBD2 expression in A549 cells. A549 cells (1.5 × 105) were transfected with pCDNA3.1 or pDC304 (empty vectors), ΔMyD88 or Mal P/H expression plasmids as indicated. 24 h post transfection, cells were stimulated with LPS (10 μg/ml) for 24 h. HBD2 expression was measured by semi-quantitative RTPCR. Expression in LPS-treated cells was ascribed a value of 100 %. Data shown are mean+/- S.E. (n = 3). Regarding amounts of transfected DNA, ΔMyD88 + and ++ represent 100 and 200 ng of dominant negative MyD88 plasmid DNA respectively while Mal + and ++ represent 50 and 100 ng of dominant negative Mal plasmid DNA respectively. (* P < 0.05, ** P < 0.01, *** P < 0.005 vs control + LPS)
Discussion
Induced expression of HBD2 in response to infective and pro-inflammatory stimuli represents an immediate and dynamic response by the host epithelium to potential infection, and the mechanism by which this occurs has been the subject of much recent investigation. Here we provide evidence for a critical role for TLR4 in LPS-induced HBD2 expression in airway epithelial cells. The data show that A549 cells respond to LPS with increased HBD2 gene and protein expression and that these effects can be blocked by a TLR4 neutralizing antibody or transfection with functionally inactive TLR4, MyD88 or Mal transgenes. We further implicate TLR4 in LPS-induced HBD2 expression by demonstrating that expression of functional TLR4 and MD2 (a co-factor which is unique to and necessary for TLR4 activity) in HEK293 cells confers LPS responsiveness to these cells, which can lead to HBD2 induction.
There is some discrepancy in the literature regarding surface expression of TLR4 in A549 cells. Guillot and colleagues reported that TLR4 is not expressed on the surface of A549 cells, but is compartmentalized to the intracellular compartment [23], while Monick et al. demonstrated low level surface expression on the same cells [17]. Our data showed low-level surface expression of TLR4 by LSC, and TLR4 was also detectable in membrane fractions by Western blotting. Further evidence of surface expression comes from the ability to block the receptor with anti-TLR4 monoclonal antibody. Surface expression of TLR4 has been demonstrated on other respiratory epithelial cells including human bronchial epithelial cells [24] and human airway cells in primary culture, where it was found in a more basolateral distribution. The findings of higher levels of TLR4 in the cytoplasmic fraction is also relevant, as internalisation of LPS in A549 cells has been reported as early as 4 hours after LPS challenge [25], and this internalisation modulated expression of ICAM-1 and TNF. It remains unclear whether LPS-TLR4 co-localisation intracellularly activates the same signalling cascade as at the cell membrane.
Similar controversy exists regarding the LPS responsiveness of A549 cells. Previous studies have suggested that A549 cells were hyporesponsive to LPS, at doses of up to 100 μg/ml [26-28]. This was not the case in our study. A549 cells responded to 10 μg/ml LPS with significant up regulation of both HBD2 and IL-8. One potential reason for this difference is the type of LPS used. Previous studies used E. coli LPS. P. aeruginosa is an important respiratory pathogen, particularly in patients with lung diseases such as cystic fibrosis (CF) [14]. E. coli, in contrast, is more important in the gastrointestinal and genitourinary tract, and as an important cause of septic shock [29]. Mucoid strains of Pseudomonas have been demonstrated to induce HBD2 in respiratory epithelia including A549 cells [10]. For these reasons, we used Pseudomonas LPS in our study. CD-14 is a glycoprotein which, together with TLR4 and LPS Binding Protein (LBP), forms the LPS signaling complex, and exists in membrane bound and soluble (in serum) forms. Soluble CD14 is required for LPS signaling in A549 cells [30], and it is therefore important that LPS stimulation is performed in the presence of serum, as in our study.
Different types of LPS differ in their ability to stimulate cells. Structural differences in LPS, most commonly in the O-polysaccharide chain [31], may result in different biological properties. E. coli LPS is highly toxic in its ability to propagate the systemic inflammatory response syndrome, principally through activation of monocytes. P. aeruginosa LPS differs from E. coli LPS both in the O-polysaccharide side chain and in the Lipid A component, and stimulates significantly less endotoxic effect [32], but induces sustained airway inflammation in a number of chronic lung diseases including CF and diffuse panbronchiolitis [33]. Pseudomonas LPS has been shown to be significantly more potent than LPS from a number of different strains of E. coli in its ability to stimulate IL-8 and granulocyte colony-stimulating factor (G-CSF) from respiratory epithelial cells, including A549 cells [34]. The reason for this is not clear, but structural differences may determine its processing by the TLR4/MD2 complex, either directly or through its interaction with the host plasma membrane [35,36]. Basolateral expression of TLR4 has been reported in pulmonary epithelial cells [37], and this is particularly interesting given that Pseudomonas elastase has been shown to increase epithelial permeability by its effect on tight junctions [38], thereby potentially increasing access of Pseudomonas LPS to the basally expressed TLR4.
Previous work by Becker et al [16] in primary human tracheobroncial cells shows a clear increase in HBD2 protein in response to LPS by western blotting, although another recent study in primary airway epithelial cells reports low expression of MD2 limiting LPS responsiveness [18]. MD2 was induced in response to pro-inflammatory cytokines and bacterial products [18], while MD2 expression in A549 cells is enhanced along with TLR4 following infection with RSV [17]. While differences speak to the limitations of using a cultured cell model, as well as to the variable responses of cultured primary cells, they also reinforce the critical importance of both TLR4 and MD2 in the signalling pathway. A549 cells clearly expressed MD2 and TLR4 and responded to Pseudomonas LPS in our hands. LPS stimulation resulted in a 5-fold increase in HBD2 promoter linked activity. The increase in HBD2 protein, though statistically significant, was small in absolute terms raising questions regarding the physiological significance of LPS-induced epithelial derived HBD2 production.
Indirect activation of epithelial cells by proinflammatory cytokines released by stimulated alveolar macrophages may be more important at lower concentrations of LPS [26], while direct stimulation of the epithelial cells may become important when bacterial load is high, where TLR4 and MD2 expression is enhanced, or following internalisation of LPS. The kinetics of the epithelial response in this and other studies [16,20] may also be relevant, with direct stimulation of the epithelial cells providing a slower but potentially more sustained release of HBD2 than that induced by inflammatory mediators. Indeed, recent studies of cellular cross-talk between epithelial cells and mononuclear cells suggests co-localisation of these cells may result in a more pronounced immune response in vivo [25]. Further work using primary cell culture and co-culture with immune cells mimicking physiologic conditions is required to clarify the relative contributions of these cells to HBD2 production in vivo.
The mechanism by which LPS induces HBD2 expression in respiratory epithelium has not been previously reported. Indirect evidence for involvement of a TLR came first from Diamond and colleagues [39] in 1996, who described CD-14 dependent LPS induced production of an anti-microbial peptide in bovine tracheal epithelium, and later from Becker and colleagues, who showed that HBD2 up regulation in response to LPS in human tracheobronchial cells was CD14 dependent (33). The GPI-linked CD14 receptor lacks a cytosolic domain and must interact with another receptor to transduce its signal to the nucleus [40]. Transcriptional regulation of HBD2 in response to LPS has also been shown to involve NF-κB [41], but the signaling pathway upstream of NF-κB has not been elucidated. Until recently, TLRs appeared to share a common signaling pathway downstream of their TIR domain. It is now known that individual TLRs utilize different adaptor proteins for signaling, thus conferring biological specificity in their response to activation by individual ligands [42,43]. MyD88 is involved in signaling from all TLRs with the exception of TLR3, whilst Mal is known to have a role in TLR2 and TLR4 intracellular signaling. Our data implicate both MyD88 and Mal in LPS-induced HBD2 expression and clearly demonstrates the critical role of TLR4 in LPS induction of HBD2, thus defining another role for TLR4 in pulmonary host defense. Along with the previously well-defined functions of TLR4 in induction of a large number of cytokines, chemokines and adhesion molecules that activate phagocytosis and the adaptive immune responses, activation of TLR4 in the epithelium results in a direct microbicidal response via production of a potent anti-microbial peptide.
Modulation of TLR4 expression in respiratory epithelium may critically affect the production of HBD2. Risk factors for respiratory tract infections include increased age and smoking [44], both of which may affect TLR4 expression. TLR4 expression in macrophages is reduced in aged mice, who have a blunted cytokine response to LPS [45]. Cigarette smoke has also been demonstrated to reduce LPS responsiveness in alveolar macrophages [46]. As LPS is an active component of cigarette smoke [47], and down regulation of TLR4 expression by LPS in cigarette smoke, otherwise known as LPS tolerance, may result in impaired HBD2 production in response to Gram-negative pathogens, facilitating colonization and infection. This is the subject of ongoing work in our laboratory. Similarly, therapies for septic shock aimed at inhibiting LPS signaling by blockade of the TLR4 receptor [48] may result in increased susceptibility to nosocomial Gram-negative pneumonia through impaired HBD2 induction.
HBD2 is an important component of host defense in the lung. This study defines a critical role for TLR4 in induced expression of HBD2 in response to LPS and highlights the potential effect of modulation of TLR4 and the accessory proteins MyD88 and Mal expression on pulmonary production of this potent anti-microbial peptide. We also provide important information regarding the cellular responses of A549 cells, a cultured cell model which is widely used in studies of host defense.
Acknowledgements
We are grateful to Dr Andrew Bowie and Dr Aisling Dunne, Dept of Biochemistry, Trinity College Dublin, for providing the Mal P/H plasmid, TLR4 plasmid and MD2 plasmid, Dr M. Muzio, Department of Immunology and Cell Biology, Instituto di Ricerche Farmacologiche Mario Negri, Milano, Italy for providing ΔMyD88 and ΔTLR4 plasmids, Dr C. A. Janeway, Yale University School of Medicine, New Haven, CT 06520, USA, for providing CD4-Toll plasmid, and Dr Paul McCray, University of Iowa, Iowa City, IA, USA for providing anti-HBD2 antibody. This work was funded by research grants from Enterprise Ireland (SC/2001/104), TheHealth Research Board (RP/2004/146), The Program for Research in Third Level Institutes (PRTLI) administered by The Higher Education Authority, and The Royal College of Surgeons in Ireland.
==== Refs
Bowie A O'Neill LA The interleukin-1 receptor/Toll-like receptor superfamily: signal generators for pro-inflammatory interleukins and microbial products J Leukoc Biol 2000 67 508 514 10770283
May MJ Ghosh S Signal transduction through NF-kappa B Immunol Today 1998 19 80 88 9509763 10.1016/S0167-5699(97)01197-3
Yamamoto M Takeda K Akira S TIR domain-containing adaptors define the specificity of TLR signaling Mol Immunol 2004 40 861 868 14698224 10.1016/j.molimm.2003.10.006
Yang D Chertov O Bykovskaia SN Chen Q Buffo MJ Shogan J Anderson M Schroder JM Wang JM Howard OM Oppenheim JJ Beta-defensins: linking innate and adaptive immunity through dendritic and T cell CCR6 Science 1999 286 525 528 10521347 10.1126/science.286.5439.525
Ganz T Antimicrobial polypeptides in host defense of the respiratory tract J Clin Invest 2002 109 693 697 11901174 10.1172/JCI200215218
Bals R Hiemstra PS Innate immunity in the lung: how epithelial cells fight against respiratory pathogens Eur Respir J 2004 23 327 333 14979512 10.1183/09031936.03.00098803
Ganz T Defensins: antimicrobial peptides of innate immunity Nat Rev Immunol 2003 3 710 720 12949495 10.1038/nri1180
Raj PA Dentino AR Current status of defensins and their role in innate and adaptive immunity FEMS Microbiol Lett 2002 206 9 18 11786250 10.1016/S0378-1097(01)00496-7
Schutte BC Mitros JP Bartlett JA Walters JD Jia HP Welsh MJ Casavant TL McCray PBJ Discovery of five conserved beta -defensin gene clusters using a computational search strategy Proc Natl Acad Sci U S A 2002 99 2129 2133 11854508 10.1073/pnas.042692699
Harder J Meyer-Hoffert U Teran LM Schwichtenberg L Bartels J Maune S Schroder JM Mucoid Pseudomonas aeruginosa, TNF-alpha, and IL-1beta, but not IL-6, induce human beta-defensin-2 in respiratory epithelia Am J Respir Cell Mol Biol 2000 22 714 721 10837369
Singh PK Jia HP Wiles K Hesselberth J Liu L Conway BA Greenberg EP Valore EV Welsh MJ Ganz T Tack BF McCray PBJ Production of beta-defensins by human airway epithelia Proc Natl Acad Sci U S A 1998 95 14961 14966 9843998 10.1073/pnas.95.25.14961
Hertz CJ Wu Q Porter EM Zhang YJ Weismuller KH Godowski PJ Ganz T Randell SH Modlin RL Activation of Toll-like receptor 2 on human tracheobronchial epithelial cells induces the antimicrobial peptide human beta defensin-2 J Immunol 2003 171 6820 6826 14662888
Sethi S Murphy TF Bacterial infection in chronic obstructive pulmonary disease in 2000: a state-of-the-art review Clin Microbiol Rev 2001 14 336 363 11292642 10.1128/CMR.14.2.336-363.2001
Parad RB Gerard CJ Zurakowski D Nichols DP Pier GB Pulmonary outcome in cystic fibrosis is influenced primarily by mucoid Pseudomonas aeruginosa infection and immune status and only modestly by genotype Infect Immun 1999 67 4744 4750 10456926
Vora P Youdim A Thomas LS Fukata M Tesfay SY Lukasek K Michelsen KS Wada A Hirayama T Arditi M Abreu MT Beta-defensin-2 expression is regulated by TLR signaling in intestinal epithelial cells J Immunol 2004 173 5398 5405 15494486
Becker MN Diamond G Verghese MW Randell SH CD14-dependent lipopolysaccharide-induced beta-defensin-2 expression in human tracheobronchial epithelium J Biol Chem 2000 275 29731 29736 10882713 10.1074/jbc.M000184200
Monick MM Yarovinsky TO Powers LS Butler NS Carter AB Gudmundsson G Hunninghake GW Respiratory syncytial virus up-regulates TLR4 and sensitizes airway epithelial cells to endotoxin J Biol Chem 2003 278 53035 53044 14565959 10.1074/jbc.M308093200
Jia HP Kline JN Penisten A Apicella MA Gioannini TL Weiss J McCray PBJ Endotoxin responsiveness of human airway epithelia is limited by low expression of MD-2 Am J Physiol Lung Cell Mol Physiol 2004 287 L428 37 15121639 10.1152/ajplung.00377.2003
Greene C Lowe G Taggart C Gallagher P McElvaney N O'Neill S Tumor necrosis factor-alpha-converting enzyme: its role in community-acquired pneumonia J Infect Dis 2002 186 1790 1796 12447765 10.1086/345799
Griffin S Taggart CC Greene CM O'Neill S McElvaney NG Neutrophil elastase up-regulates human beta-defensin-2 expression in human bronchial epithelial cells FEBS Lett 2003 546 233 236 12832046 10.1016/S0014-5793(03)00577-5
Frantz S Kobzik L Kim YD Fukazawa R Medzhitov R Lee RT Kelly RA Toll4 (TLR4) expression in cardiac myocytes in normal and failing myocardium J Clin Invest 1999 104 271 280 10430608
Re F Strominger JL Separate functional domains of human MD-2 mediate toll-like receptor 4-binding and lipopolysaccharide responsiveness J Immunol 2003 171 5272 5276 14607928
Guillot L Medjane S Le-Barillec K Balloy V Danel C Chignard M Si-Tahar M Response of human pulmonary epithelial cells to lipopolysaccharide involves Toll-like receptor 4 (TLR4)-dependent signaling pathways: evidence for an intracellular compartmentalization of TLR4 J Biol Chem 2004 279 2712 2718 14600154 10.1074/jbc.M305790200
Devaney JM Greene CM Taggart CC Carroll TP O'Neill SJ McElvaney NG Neutrophil elastase up-regulates interleukin-8 via toll-like receptor 4 FEBS Lett 2003 544 129 132 12782302 10.1016/S0014-5793(03)00482-4
Lee JH Del Sorbo L Uhlig S Porro GA Whitehead T Voglis S Liu M Slutsky AS Zhang H Intercellular adhesion molecule-1 mediates cellular cross-talk between parenchymal and immune cells after lipopolysaccharide neutralization J Immunol 2004 172 608 616 14688373
Tsutsumi-Ishii Y Nagaoka I Modulation of human beta-defensin-2 transcription in pulmonary epithelial cells by lipopolysaccharide-stimulated mononuclear phagocytes via proinflammatory cytokine production J Immunol 2003 170 4226 4236 12682256
Keicho N Elliott WM Hogg JC Hayashi S Adenovirus E1A upregulates interleukin-8 expression induced by endotoxin in pulmonary epithelial cells Am J Physiol 1997 272 L1046 52 9227502
Krakauer T Stimulant-dependent modulation of cytokines and chemokines by airway epithelial cells: cross talk between pulmonary epithelial and peripheral blood mononuclear cells Clin Diagn Lab Immunol 2002 9 126 131 11777841 10.1128/CDLI.9.1.126-131.2002
Kaper JB Nataro JP Mobley HL Pathogenic Escherichia coli Nat Rev Microbiol 2004 2 123 140 15040260 10.1038/nrmicro818
Schulz C Farkas L Wolf K Kratzel K Eissner G Pfeifer M Differences in LPS-induced activation of bronchial epithelial cells (BEAS-2B) and type II-like pneumocytes (A-549) Scand J Immunol 2002 56 294 302 12193231 10.1046/j.1365-3083.2002.01137.x
Erridge C Bennett-Guerrero E Poxton IR Structure and function of lipopolysaccharides Microbes Infect 2002 4 837 851 12270731 10.1016/S1286-4579(02)01604-0
Kulshin VA Zahringer U Lindner B Jager KE Dmitriev BA Rietschel ET Structural characterization of the lipid A component of Pseudomonas aeruginosa wild-type and rough mutant lipopolysaccharides Eur J Biochem 1991 198 697 704 1904818 10.1111/j.1432-1033.1991.tb16069.x
Lyczak JB Cannon CL Pier GB Establishment of Pseudomonas aeruginosa infection: lessons from a versatile opportunist Microbes Infect 2000 2 1051 1060 10967285 10.1016/S1286-4579(00)01259-4
Koyama S Sato E Nomura H Kubo K Miura M Yamashita T Nagai S Izumi T The potential of various lipopolysaccharides to release IL-8 and G-CSF Am J Physiol Lung Cell Mol Physiol 2000 278 L658 66 10749742
Wurfel MM Wright SD Lipopolysaccharide-binding protein and soluble CD14 transfer lipopolysaccharide to phospholipid bilayers: preferential interaction with particular classes of lipid J Immunol 1997 158 3925 3934 9103463
Viriyakosol S Tobias PS Kitchens RL Kirkland TN MD-2 binds to bacterial lipopolysaccharide J Biol Chem 2001 276 38044 38051 11500507
Muir A Soong G Sokol S Reddy B Gomez M Van Heeckeren A Prince A Toll like receptors in normal and cystic fibrosis airway epithelial cells Am J Respir Cell Mol Biol 2003 14656745
Azghani AO Miller EJ Peterson BT Virulence factors from Pseudomonas aeruginosa increase lung epithelial permeability Lung 2000 178 261 269 11147311 10.1007/s004080000031
Diamond G Russell JP Bevins CL Inducible expression of an antibiotic peptide gene in lipopolysaccharide-challenged tracheal epithelial cells Proc Natl Acad Sci U S A 1996 93 5156 5160 8643545 10.1073/pnas.93.10.5156
Haziot A Chen S Ferrero E Low MG Silber R Goyert SM The monocyte differentiation antigen, CD14, is anchored to the cell membrane by a phosphatidylinositol linkage J Immunol 1988 141 547 552 3385210
Tsutsumi-Ishii Y Nagaoka I NF-kappa B-mediated transcriptional regulation of human beta-defensin-2 gene following lipopolysaccharide stimulation J Leukoc Biol 2002 71 154 162 11781391
O'Neill LA Dunne A Edjeback M Gray P Jefferies C Wietek C Mal and MyD88: adapter proteins involved in signal transduction by Toll-like receptors J Endotoxin Res 2003 9 55 59 12691620 10.1179/096805103125001351
Takeda K Akira S TLR signaling pathways Semin Immunol 2004 16 3 9 14751757 10.1016/j.smim.2003.10.003
Graham NM The epidemiology of acute respiratory infections in children and adults: a global perspective Epidemiol Rev 1990 12 149 178 2286216
Renshaw M Rockwell J Engleman C Gewirtz A Katz J Sambhara S Cutting edge: impaired Toll-like receptor expression and function in aging J Immunol 2002 169 4697 4701 12391175
Ohta T Yamashita N Maruyama M Sugiyama E Kobayashi M Cigarette smoking decreases interleukin-8 secretion by human alveolar macrophages Respir Med 1998 92 922 927 10070565 10.1016/S0954-6111(98)90191-3
Hasday JD Bascom R Costa JJ Fitzgerald T Dubin W Bacterial endotoxin is an active component of cigarette smoke Chest 1999 115 829 835 10084499 10.1378/chest.115.3.829
Iwami KI Matsuguchi T Masuda A Kikuchi T Musikacharoen T Yoshikai Y Cutting edge: naturally occurring soluble form of mouse Toll-like receptor 4 inhibits lipopolysaccharide signaling J Immunol 2000 165 6682 6686 11120784
|
16219107
|
PMC1276817
|
CC BY
|
2021-01-04 16:36:26
|
no
|
Respir Res. 2005 Oct 12; 6(1):116
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-116
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-1201624202910.1186/1465-9921-6-120ResearchEthnic variations in incidence of asthma episodes in England & Wales:national study of 502,482 patients in primary care Netuveli Gopalakrishnan [email protected] Brian [email protected] Aziz [email protected] Department of Primary Care & Social Medicine, Imperial College London, London W6 8RP, UK2 School of Humanities, King's College London, Strand, London WC2R 2LS, UK3 Division of Community Health Sciences: GP Section, University of Edinburgh, 20, West Richmond Street, Edinburgh EH8 9DX, UK2005 21 10 2005 6 1 120 120 7 7 2005 21 10 2005 Copyright © 2005 Netuveli et al; licensee BioMed Central Ltd.2005Netuveli et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent studies have demonstrated marked international variations in the prevalence of asthma, but less is known about ethnic variations in asthma epidemiology within individual countries and in particular the impact of migration on risk of developing asthma. Recent within country comparisons have however revealed that despite originating from areas of the world with a low risk for developing asthma, South Asian and Afro-Caribbean people in the UK are significantly (3× and 2× respectively) more likely to be admitted to hospital for asthma related problems than Whites.
Methods
Using data from the Fourth National Study of Morbidity Statistics in General Practice, a one-percent broadly representative prospective cohort study of consultations in general practice, we investigated ethnic variations in incident asthma consultations (defined as new or first consultations), and compared consultation rates between those born inside and outside the UK (migrant status). Logistic regression models were used to examine the combined effects of ethnicity and migration on asthma incident consultations.
Results
Results showed significantly lower new/first asthma consultation rates for Whites than for each of the ethnic minority groups studied (mean age-adjusted consultation rates per 1000 patient-years: Whites 26.4 (95%CI 26.4, 26.4); South Asians 30.4 (95%CI 30.3, 30.5); Afro-Caribbeans 35.1 (95%CI 34.9, 35.3); and Others 27.8 (27.7, 28.0). Within each of these ethnic groups, those born outside of the UK showed consistently lower rates of incident asthma consultations. Modelling the combined effects of ethnic and migrant status revealed that UK-born South Asians and Afro-Caribbeans experienced comparable risks for incident GP consultations for asthma to UK-born Whites. Non-UK born Whites however experienced reduced risks (adjusted OR 0.82, 95%CI 0.69, 0.97) whilst non-UK born South Asians experienced increased risks (adjusted OR 1.33, 95%CI 1.04, 1.70) compared to UK-born Whites.
Conclusion
These findings strongly suggest that ethnicity and migration have significant and independent effects on asthma incidence. The known poorer asthma outcomes in UK South Asians and Afro-Caribbeans may in part be explained by the offspring of migrants experiencing an increased risk of developing asthma when compared to UK-born Whites. This is the first study to find heterogeneity for incident asthma consultations in Whites by migrant status.
==== Body
Background
The International Study of Asthma and Allergies in Childhood (ISAAC) and European Community and Respiratory Health Survey (ECRHS) have revealed marked international variations in the prevalence of asthma, with populations living in economically-developed countries experiencing the highest prevalence rates [1,2]. The precise reasons underpinning these variations are poorly understood, but point in particular to the possible importance of differing exposures to environmental risk factors.
In contrast to marked international variations in the prevalence of asthma, a recent systematic review and meta-analysis of epidemiological studies within the UK has found that despite originating from low risk areas internationally, South Asians and Afro-Caribbeans experience significantly poorer asthma outcomes than do Whites [3]. Possible reasons for these poorer outcomes could include differences in asthma incidence, severity, management and/or health seeking behaviour between ethnic groups [4-8].
In this study, we investigated possible ethnic variations in incidence of asthma episodes and in addition explored the impact of migration on risk of developing asthma. We hypothesised that the incidence of asthma episodes would vary between ethnic groups with those born in low risk regions experiencing a lower incidence than those born in the UK (a very high risk region).
Whilst some of these results have previously been presented in summary format in our systematic review and meta-analysis of ethnic variations in the epidemiology and outcomes of asthma in UK minority ethnic groups [3], the background, rationale, methods and detailed results have never previously been reported in the peer-review literature.
Methods
Study sample
Our study sample consisted of patients included in the Fourth National Study of Morbidity Statistics in General Practice (MSGP4), a year long prospective cohort study during the period September 1991 – August 1992 of more than half a million patients registered with 60 general practices in England and Wales [9]. The study sample was a broadly representative one-percent sample of the general population of England and Wales.
Definitions and measures
Ethnicity was coded using 1991 census categories [10]. Because of small numbers in some ethnic groups, we re-categorised these data into four broader ethnic groups: 'Whites', 'South Asians' 'Afro-Caribbeans' and 'Others'. In doing so, we combined Indians, Pakistanis and Bangladeshis into 'South Asians', while assigning Chinese to the 'Others' group. Immigrants were defined as all those born outside of the UK irrespective of nationality or ethnicity.
A 'consultation' was defined as a face-to-face encounter between a patient and a member of the practice clinical staff leading to at least one Read code diagnosis. Read codes were mapped onto the ninth revision of the International Classification of Diseases (ICD-9) [11]. If the diagnostic code was for asthma (ICD-9 493), we defined this as being an asthma consultation. We marked asthma consultations as 'first' if the patient had never before consulted for asthma during their life time and 'new' if the patient had previously consulted for asthma, but not in the last 28 days. All those consultations marked 'first' by definition mark 'new' episode consultations for asthma and hence, by considering 'first' and 'new' consultations together we were able to examine the incidence of asthma episodes. However, not all registered patients consulted during the MSGP4 study period and some patients consulted more than once. Furthermore, due to births, deaths and patient mobility, the denominator was not constant over the study period. We therefore calculated the 'patient years at risk' by dividing the number of days for which patients were registered with a practice during the 12-months study period by the number of days in the year (366 in 1991–92).
As general explanatory variables, we used age, sex, Registrar General's social class, dichotomised as non-manual (Classes I, II, IIINM) and manual (Classes IIIM, IV, V), urban, and current smoking status which was assessed by a single question ('smoked in the previous week').
Statistical analysis
The incidence rates for asthma episodes were calculated using episodes defined as new or first consultations per person years at risk. Ethnic minority groups in the UK are known to be significantly younger than the White majority population [12]; we therefore adjusted for age by standardising against the age structure of the total sample. As a sensitivity analysis, we also examined age-specific rates.
We used logistic regression modelling to examine for the effects of ethnicity and migrant status on asthma incidence. As explanatory variables, we used a categorical variable for ethnicity (Whites, South Asians, Afro-Caribbeans and Others) and a binary variable to represent whether subjects were born inside/outside the UK. Three models, containing ethnicity alone, immigrant status alone and ethnicity and immigrant status together, were fitted. The dependent variable in our regression models was the presence or absence of a new or first consultation for asthma episodes. The possibility of confounding and/or effect modification was assessed using the following covariates: age, sex, social class, house ownership, urban area and current smoking status.
We undertook a sensitivity analysis to test the hypothesis that data on ethnicity were missing more in non-White ethnic groups (selection bias) using the Heckman procedure [13]. Briefly, we created a binary variable denoting absence of ethnicity information and used age, sex, manual social class and smoking status to model this variable. This selection model was fitted together with the substantive model using new/first asthma episode consultations as the dependant variable and ethnicity and migration status as predictor variables. The parameter of interest was the correlation between error terms in the substantive model and the selection model which might indicate significant selection bias in the sample. All analyses were undertaken using Stata Version 7 [14].
Results
Information on ethnicity was available for 82.7% (n = 415,528) of subjects, those with missing data experiencing significantly lower new/first asthma episode consultation rates than the overall sample (Table 1). Of these, 2.4% (n = 9,981) belonged to non-White ethnic minority groups.
Table 1 Sample description and new/first consultations for asthma episodes (incidence) by ethnicity and place of birth
Ethnicity Country of birth Total number of patients (%) Mean age, years (sd) Asthma
Total consultations/1000 patient years New/first Consultations/1000 patient years
Unadjusted Unadjusted Age-adjusted
Mean Mean Mean 95%CI
Whites UK 393,300 (78.3) 36 (23) 35.7 27.1 26.5 26.2 to 26.8
Non-UK 12,247 (2.4) 43 (21) 29.3 19.8 24.9 23.3 to 26.4
Total 405,547 (80.7) 37 (23) 35.5 26.9 26.4 26.4 to 26.4
South Asians UK 2,187 (0.4) 12 (11) 54.5 46.3 34.7 26.0 to 43.5
Non-UK 3,501 (0.7) 38 (16) 27.6 24.6 23.7 20.2 to 27.3
Total 5,688 (1.1) 28 (19) 37.9 32.9 30.4 30.3 to 30.5
Afro-Caribbeans UK 1,428 (0.3) 19 (15) 61.2 52.3 39.2 31.5 to 46.8
Non-UK 1,080 (0.2) 40 (16) 28.0 22.0 20.4 14.2 to 26.5
Total 2,508 (0.5) 28 (19) 47.0 39.3 35.1 34.9 to 35.3
Others UK 932 (0.2) 13 (13) 73.0 38.2 30.9 18.6 to 43.1
Non-UK 853 (0.2) 32 (16) 27.8 27.8 30.3 23.2 to 37.4
Total 1,785 (0.4) 22 (18) 51.9 33.4 27.8 27.7 to 28.0
Missing 86,954 (17.3) 36 (21) 16.4 11.8 13.4 12.9 to 13.8
Total 502,482 (100) 36 (23) 35.7 23.7 24.8 24.3 to 25.2
Age-standardised analysis revealed that new/first asthma episode consultation rate for Whites was significantly lower than for South Asians, Afro-Caribbeans and Others. When these ethnic groups were subdivided into those born inside and outside of the UK, those born outside of the UK were found to experience significantly lower rates of new/first episode consultations for asthma (Table 1).
Unadjusted models for ethnicity showed that non-White ethnic minorities had greater risk of experiencing a new/first asthma episode consultation than did Whites; however after adjusting for important confounders only South Asians were found to be at increased risk of new/first asthma episode consultations (OR 1.33, 95%CI 1.06, 1.67) (Table 2).
Table 2 Logistic regression models for relationship between ethnicity, immigrant status and risk of new/first consultation for asthma episodes
Odds ratios for new/first consultation for asthma
Unadjusted Adjusted*
OR 95% CI OR 95% CI
Ethnicity
Whites 1 1
South Asians 1.21 1.04 to 1.41 1.33 1.06 to 1.67
Afro-Caribbeans 1.44 1.17 to 1.78 1.04 0.73 to 1.48
Others 1.64 1.30 to 2.07 1.12 0.68 to 1.83
Immigrant status
Born outside UK 0.75 0.67 to 0.83 0.93 0.82 to 1.06
Ethnicity and immigrant status
White, UK-born 1 1
White, non-UK born 0.71 0.62 to 0.81 0.82 0.69 to 0.97
South Asian, UK-born 1.71 1.39 to 2.10 1.48 0.81 to 2.70
South Asian, non-UK born 0.89 0.71 to 1.11 1.33 1.04 to 1.70
Afro-Caribbean, UK-born 1.94 1.52 to 2.46 1.30 0.79 to 2.13
Afro-Caribbean, non-UK born 0.78 0.51 to 1.19 0.96 0.57 to 1.60
Other, UK-born 2.27 1.72 to 2.99 1.47 0.65 to 3.37
Other, non-UK born 0.95 0.61 to 1.46 0.80 0.40 to 1.61
*Adjusted for age, sex, social class, house ownership, urban and smoking status
Unadjusted models for migrant status showed that for each of the ethnic groups those born outside the UK showed reduced risk of incident asthma episodes in primary care compared with those born in the UK, and this relationship was retained after adjusting for age differences except in the Others group (Table 1), but was no longer found in any of the groups when adjusted for differences in age, sex, social class, house ownership, urban rural location and smoking status (OR 0.93, 95%CI 0.82, 1.06) (see Table 2).
When ethnicity and migrant status were considered together, there were no significant differences in odds ratios for incident asthma episodes between UK-born Whites, UK-born South Asians and (all) Afro-Caribbeans (Table 2). However, non-UK born Whites had a reduced risk of new/first asthma episode consultation (OR 0.82, 95%CI 0.69, 0.97) compared with UK born Whites, whereas non-UK born South Asians experienced an increased risk (OR 1.33, 95%CI 1.04, 1.70) compared with UK born Whites.
Sensitivity analysis using Heckman sample selection modelling revealed no evidence of sample selection bias (p = 0.40).
Discussion
This large national study has found that UK ethnic minority groups experience significantly higher age-adjusted new/first episode consultation rates for asthma than the White majority population, this suggesting significant variation in asthma incidence between ethnic groups. Taken together with our finding of significantly lower age-adjusted rates of incident episodes in those (including Whites) born outside the UK and the known patterns of migration to the UK, these findings reinforce the likely importance of early life environmental exposures in influencing risk of developing, and subsequent outcomes for, one of the commonest chronic disorders worldwide.
Table 1 reveals that the higher incident consultation rates in minority ethnic groups is driven by those born inside the UK; being born outside of the UK is associated with lower asthma consultation rates. There are three main explanations which could account for these findings: differing demographies of the populations being compared, a genuinely increased risk resulting from social (environmental) exposures in those born in the UK and differences in health seeking behaviour between groups. Comparison of the unadjusted and age-adjusted rates in Table 1 shows that population age structure is an important consideration, but that not all of the differences are accounted for by age. Turning to Table 2, the unadjusted analyses reveal that ethncitiy and migration status are both important, with UK born subjects behaving very differently from non-UK born subjects except for the heterogeneous Other group. These findings are unlikely to be the result simply of differing age or other demographic factors modelled in the analysis, as the adjusted point estimates demonstrate that these factors do not explain the entire effect.
Ethnicity has been established as an important variable in asthma epidemiology; but ethnic identity is not homogenous (often due to inadequate definitions) [15]. We have shown that migrant status in three of the main UK ethnic groups – Whites, South Asians and Afro-Caribbeans – contributes significantly to the heterogeneity of incident consultation for asthma in primary care, a finding that strengthens the possibility that there is both a spatial and a temporal effect on asthma incidence in ethnic groups.
Importantly, our findings contrast with those of the European Community Respiratory Health Survey study, which found no significant differences in utilisation of health services by migrants, and no pattern in the prevalence of asthma symptoms after taking account of asthma prevalence in the regions of origin and in the host country [16].
The limitations of our study need to be appreciated. No data on ethnicity were recorded on 17.3% of subjects consulting for asthma in the MSGP4 who were therefore excluded from our analysis. The excluded subjects had lower rates of new/first episodes of asthma than those on whom ethnicity data were available (Table 1), which could represent an important source of bias. The results of the sensitivity analysis, however, make it unlikely that ethnicity information was systematically more commonly withheld by non-White and/or non-UK born subjects (but caution is in order because ethnic minorities constituted 2.4% of the population sample in our study compared to 6% reported in the 1991 census) [17]. Also of potential relevance is that using MSGP4 data to estimate asthma episode incidence involves use of a surrogate, as new consultations for asthma refer to practice episodes rather than patient episodes; therefore some patients coded as having a new/first episode of asthma may already have had asthma diagnosed in a previous practice. It is however important to appreciate that in creating MSGP4 data, a very diligent edit process is used, enabling MSGP4 authors to maintain that "all first and new episode rates indicate incidence of a condition or group of conditions" [9]. Another potential limitation is that such incidence data relate to episodes rather than to persons, but such episode incidence data are acceptable for the purposes of comparing different groups based on ethnicity and migrant status. It s also important to acknowledge that in using MSGP4 consultations to elicit asthma episodes our results could be confounded by behavioural factors that influence consultations. In defence of our study it should be pointed out that we used only consultations for asthma as diagnosed by the clinician and within each ethnic group the impact of migrant status remains similar (and this was also true for the total number of consultations for asthma (data not shown)). One should also take into account limitations in the generalisability of the MSGP4 practice sample: participating practices in the survey were not entirely representative of the distribution of practices in England and Wales in 1991. Despite this, there is good agreement on asthma consultation rates between MSGP4 and asthma diagnosis rates in the General Practice Research Database [18]. An important limitation in our study is the lack of information on length of UK residence, as immigrant asthma prevalence is reported to become closer to that in the host country population with increasing length of residence [19-21].
The strengths of this dataset are however substantial and include its prospective cohort design, the broadly representative large sample size, and the availability of data on a range of potential confounders and effect modifiers in the relationship between ethnicity, migration and asthma consultations. For example, ethnic minorities in the UK are on the whole significantly younger than the White majority population, but we were able to standardise for this when calculating new/first asthma episode consultation rates thus allowing meaningful comparisons to be made across ethnic groups.
Results which echo our findings have been reported in other countries. It has been shown that those born outside of western countries are at reduced risk of asthma-related hospital admissions [22,23]. In Germany, greater cultural adaptation has been shown to be associated with increased prevalence of asthma in Turkish children [24], these findings also pointing to the importance of ethnic and migration related factors. There is some evidence that such factors might be lifestyle related; while South Asian children born in Africa, who have a more westernised lifestyle, had asthma prevalence similar to South Asian children born in the UK [25], South Asian children born in South Asia had a lower prevalence [26].
Our study found significant differences based on migration status. It is possible that migrants bring with them the same pattern of morbidity and health service utilisation as they experience in their country of origin. However, interpreting this finding is difficult since there are several possible types of explanation, including genetic, immunity from early exposures in native country, lack of exposure to factors in the host country, and selection through the healthy migrant effect [27-29].
Conclusion
This large national study suggests that poorer asthma outcomes may in part be explained by the offspring of South Asian and Afro-Caribbean migrants to the UK experiencing an increased risk of developing asthma compared to UK-born Whites.
Abbreviations
ECRHS – European Community and Respiratory Health Survey
ISAAC – International Study of Asthma and Allergies in Childhood
MSGP4 – 4th Morbidity Survey in General Practice
UK – United Kingdom
Conflict of interest
The author(s) declare that they have no competing interests.
Contributorship
GN and AS conceived this study and together with BH formulated the study protocol. GN performed data analysis. All three authors were involved in interpreting results and writing the paper. GN is guarantor.
Funding
GN was employed on a grant from Asthma UK (01/018) when he undertook this work.
Ethics approval
Not required
Acknowledgements
We thank Professor Raj Bhopal for his constructive comments on an earlier draft of this manuscript.
==== Refs
ISAAC Steering Committee World-wide variation in prevalence of symptoms of asthma, allergic rhinoconjunctivitis, and atopic eczema: ISAAC Lancet 1998 351 1225 32 9643741 10.1016/S0140-6736(97)07302-9
Janson C Anto J Burney P Chinn S de Marco R Heinrich J European Community Respiratory Health Survey II. The European Community Respiratory Health Survey: what are the main results so far? Eur Respir J 2000 18 598 611 11589359 10.1183/09031936.01.00205801
Netuveli G Hurwitz B Levy M Fletcher M Barnes G Durham SR Sheikh A Ethnic variations in UK asthma frequency, morbidity and health-service use: a systematic review and meta-analysis Lancet 2005 365 312 17 15664226
Lung Asthma Information Agency Factsheet 2001/1 2001 London St George's Hospital Medical School
Jackson SHD Bannan LT Beevers DG Ethnic differences in respiratory diseases Postgrad Med J 1981 57 777 778
Gilthorpe MS Lay-Yee R Wilson RC Walters S Griffiths RK Bedi R Variations in hospitalization rates for asthma among black and minority ethnic communities Resp Med 1998 92 642 648 10.1016/S0954-6111(98)90511-X
Gillam SJ Jarman B White P Law R Ethnic differences in consultation rates in urban general practice BMJ 1989 299 953 957 2508951
Moudgil H Honeybourne D Differences in asthma management between white European and Indian subcontinent ethnic groups living in socio-economically deprived areas in the Birmingham (UK) conurbation Thorax 1998 53 490 494 9713449
McCormick D Fleming D Charlton J Morbidity statistics from general practice Fourth national study, 1991/1992 1995 London: HMSO
Coleman D Salt J eds Ethnicity in the 1991 Census, Volume One: Demographic characteristics of the ethnic minority populations 1996 London, HMSO
American Medical Association International Classification of Diseases (9th Revision – Clinical Modification) 1999 Chicago: American Medical Association
Nazroo J The health of Britain's ethnic minorities 1997 Policy Studies Institute, London
Heckman JJ Sample selection bias as a specification error Econometrica 1979 47 153 161
Stata Corporation Stata reference manual: Release 70 2001 Stata Corporation, College Station, TX
Senior PA Bhopal R Ethnicity as a variable in epidemiological research BMJ 1994 309 327 330 8086873
Tobias A Soriano JB Chinn S Anto JM Sunyer J Burney P Symptoms of asthma, bronchial responsiveness and atopy in immigrants and emigrants in Europe Eur Respir J 2001 18 459 465 11589342 10.1183/09031936.01.00026501
Census 1991 Accessed 04/01/05
Hansell A Hollowell J Nichols T McNiece R Strachan D Use of the General Practice Research Database (GPRD) for respiratory epidemiology: a comparison with the 4th Morbidity Survey in General Practice (MSGP4) Thorax 1999 54 413 419 10212105
Ormerod LP Myers P Prescott RJ Prevalence of asthma and 'probable' Asian population in Blackburn, UK Resp Med 1999 93 16 20 10.1016/S0954-6111(99)90071-9
Powell CVE Nolan TM Carlin JB Bennett CM Johnson PDR Respiratory symptoms and duration of residence in immigrant teenagers living in Melbourne, Australia Arch Dis Child 1999 81 159 162 10490527
Holguin F Mannino DM Anto JM Mott J Ford ES Teague GW Redd SC Romieu I Country of birth as a risk factor for asthma among Mexican Americans Am J Respir Crit Care Med 2005 171 103 108 15516539 10.1164/rccm.200402-143OC
Hjern A Haglund B Bremberg S Ringback-Weitoft1 G Social adversity, migration and hospital admissions for childhood asthma in Sweden Acta Pediatr 1999 88 1107 1112 10.1080/08035259950168171
Hjern A Rasmussen F Johansson M Aberg N Migration and atopic disorder in Swedish conscripts Pediatr Allergy Immunol 1999 10 209 215 10565562 10.1034/j.1399-3038.1999.00014.x
Gruber C Illi S Pleith A Sommerfeld C Wahn U Cultural adaptation is associated with atopy and wheezing among children of Turkish origin living in Germany Clin Exp All 2002 32 526 531 10.1046/j.0954-7894.2002.01331.x
Smith JM Harding LK Cumming G The changing prevalence of asthma in schoolchildren Clin Aller 1971 1 57 61
Smith JM The prevalence of asthma and wheezing in children Br J Dis Chest 1976 70 73 77 952726 10.1016/0007-0971(76)90009-7
Kasl SV Berkman L Health consequences of the experience of migration Ann Rev Pub Hlth 1983 4 69 90 10.1146/annurev.pu.04.050183.000441
Thomas SL Thomas SDM Displacement and health British Medical Bulletin 2004 69 115 127 15226201 10.1093/bmb/ldh009
Strachan DP Sheikh A Kuh D, Ben-Shlomo Y A life course approach to respiratory and allergic diseases A life course approach to chronic disease epidemiology 2004 2 Oxford: Oxford University Press 240 259
|
16242029
|
PMC1276818
|
CC BY
|
2021-01-04 16:36:26
|
no
|
Respir Res. 2005 Oct 21; 6(1):120
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-120
|
oa_comm
|
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-2-431623616510.1186/1742-4682-2-43ResearchModeling the signaling endosome hypothesis: Why a drive to the nucleus is better than a (random) walk Howe Charles L [email protected] Departments of Neuroscience and Neurology, Mayo Clinic College of Medicine, Guggenheim 442-C, 200 1st Street SW, Rochester, MN 55905, USA2005 19 10 2005 2 43 43 1 9 2005 19 10 2005 Copyright © 2005 Howe; licensee BioMed Central Ltd.2005Howe; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Information transfer from the plasma membrane to the nucleus is a universal cell biological property. Such information is generally encoded in the form of post-translationally modified protein messengers. Textbook signaling models typically depend upon the diffusion of molecular signals from the site of initiation at the plasma membrane to the site of effector function within the nucleus. However, such models fail to consider several critical constraints placed upon diffusion by the cellular milieu, including the likelihood of signal termination by dephosphorylation. In contrast, signaling associated with retrogradely transported membrane-bounded organelles such as endosomes provides a dephosphorylation-resistant mechanism for the vectorial transmission of molecular signals. We explore the relative efficiencies of signal diffusion versus retrograde transport of signaling endosomes.
Results
Using large-scale Monte Carlo simulations of diffusing STAT-3 molecules coupled with probabilistic modeling of dephosphorylation kinetics we found that predicted theoretical measures of STAT-3 diffusion likely overestimate the effective range of this signal. Compared to the inherently nucleus-directed movement of retrogradely transported signaling endosomes, diffusion of STAT-3 becomes less efficient at information transfer in spatial domains greater than 200 nanometers from the plasma membrane.
Conclusion
Our model suggests that cells might utilize two distinct information transmission paradigms: 1) fast local signaling via diffusion over spatial domains on the order of less than 200 nanometers; 2) long-distance signaling via information packets associated with the cytoskeletal transport apparatus. Our model supports previous observations suggesting that the signaling endosome hypothesis is a subset of a more general hypothesis that the most efficient mechanism for intracellular signaling-at-a-distance involves the association of signaling molecules with molecular motors that move along the cytoskeleton. Importantly, however, cytoskeletal association of membrane-bounded complexes containing ligand-occupied transmembrane receptors and downstream effector molecules provides the ability to regenerate signals at any point along the transmission path. We conclude that signaling endosomes provide unique information transmission properties relevant to all cell architectures, and we propose that the majority of relevant information transmitted from the plasma membrane to the nucleus will be found in association with organelles of endocytic origin.
==== Body
Background
The transmission of signals from the extracellular surface of the plasma membrane to the nucleus is a complex process that involves a large repertoire of trafficking-related and signal-transducing proteins. A highly dynamic and carefully orchestrated series of molecular events has evolved to ensure that signals emanating from outside the cell are communicated to the nuclear transcriptional apparatus with fidelity and signal integrity. The classic model for the execution of this molecular symphony is a cascade of protein:protein interactions resulting in the spread of an amplified wave of protein phosphorylation that eventually culminates in a cadence of transcription factor activity. For example, as illustrated in Figure 1, epidermal growth factor (EGF) binds to it receptor tyrosine kinase (EGFR) on the surface of a cell, resulting in the transmission of a wave of tyrosine, serine, and threonine phosphorylation events that leads to the activation and nuclear translocation of several transcription factors, including STAT-3 (signal transducer and activator of transcription-3) and ERK1/2 (extracellular signal-related kinase-1/2; also known as mitogen-activated protein kinase, MAPK). This cascading wave model depends inherently upon the notion that activated transcription factors diffuse through the cytoplasm, enter the nucleus, and execute a program of transcriptional activation. Conceptually, this model is easy to grasp – but does it accurately reflect the biology and the physical constraints of cellular architecture? The answer appears to be "No", as a significant body of work over the past decades has challenged the fundamental validity of the diffusion model [1-3] and has offered elegant alternative models for the transmission of intracellular signals [4,5].
Figure 1 Simplified diagram showing the activation of STAT-3 and Erk1/2 downstream from EGF binding to EGFR. In the general model of signal transduction, the cascading chain of phosphorylation events culminating in activation of transcription factors such as STAT-3 and Erk1/2 depends upon the diffusion of these molecules from the site of signal initiation at the plasma membrane to the site of transcriptional regulation within the nucleus.
Neurons exhibit a unique architecture that places severe physical limitations on the possible mechanisms for translocation of signals. As shown in Figure 2A, projection neurons extend axons into target fields over distances that dwarf the dimensions of the cell body. And yet, the Neurotrophic Factor Hypothesis of neurodevelopment requires that target-derived soluble trophic factors induce signals in the presynaptic terminal of axons that result in transcriptional and translational changes in the nucleus and neuronal cell body (Figure 2B) [6]. While it is possible that a signal generated at the plasma membrane of the presynaptic terminal diffuses along the length of the axon in order to elicit an effect at the nucleus – it is not at all probable [5]. For some projection neurons the length of the axon is five orders of magnitude greater than the diameter of the neuron cell body, and the axoplasm therefore constitutes 1000-fold more volume than the cytoplasm of an average cell. The Signaling Endosome Hypothesis posits that an active, directed process of signal transmission is required to overcome the physical constraints of axonal distances and volumes [7]. Specifically, this hypothesis states that the most efficient mechanism for signaling-at-a-distance involves the packaging of a secreted growth factor signal into a discrete, coherent, membrane-bounded organelle that is moved along the length of the axon via a cytoskeleton-based transport machine (Figure 3) [7]. Indeed, a substantial body of research supports the signaling endosome hypothesis within the context of neurotrophin signaling in neurons [8-12]. However, while the unique geometry of neurons provides a teleological basis for the existence of signaling endosomes, it is far more interesting to posit that the signaling endosome hypothesis represents a general biological mechanism for signal transduction and signal compartmentalization [4]. Such a generalized hypothesis might state that the most efficient mechanism for communicating signals from the plasma membrane to the nucleus is the compartmentalization of signal transducers into quantal endocytic membrane-associated signaling packets that are retrogradely transported along microtubules through the cytoplasm. By utilizing the intrinsic directionality and nucleus-directed organization of the cellular microtubule network, signaling endosomes provide a noise-resistant mechanism for the vectorial transport of plasma membrane-derived signals to the nucleus.
Figure 2 A) Neurons throughout the nervous system send axonal projections over distances ranging from microns to meters. For large or anatomically specialized animals such as the giraffe or the whale, more than 5 meters may separate the neuron cell body from the distal axon terminal. B) During development, neurons establish trophic interactions with target tissues. As an organism develops, the strength and maintenance of these trophic interactions determine whether neurons survive or die. Soluble protein trophic factors released by the target tissue (1) bind to transmembrane receptors on the presynaptic axon terminal (2), inducing receptor activation and the induction of intracellular signaling cascades (3). These signals must travel from the site of initiation to the distant cell body (4) in order to enter the nucleus and elicit transcriptional changes that determine the survival of the cell. This long-distance information transfer is a universal theme in neurodevelopment.
Figure 3 The signaling endosome hypothesis of long-distance axonal signal transmission. Soluble protein trophic factors released by the target (1) bind to transmembrane receptors on the presynaptic axon terminal (2), inducing receptor activation and internalization via clathrin-coated membranes or other endocytic structures (3). These endocytic vesicles give rise to transport endosomes that bear the receptor and associated signaling molecules as well as molecular motors (shown in turquoise) (4) that utilize microtubules (shown in pink) within the axon to carry the endosome toward the cell body (5). Upon arrival at the neuron cell body the endosome-associated signals may either initiate additional local signals or may directly translocate (6) into the nucleus to elicit transcriptional changes (7).
A number of findings support the concept that signaling from internal cellular membranes is a general phenomenon that is relevant to understanding receptor tyrosine kinase signaling in many cellular systems. For example, EGFR, as discussed above, is internalized via clathrin-coated vesicles following EGF-binding and receptor activation [13-15]. In the past, trafficking through this compartment was considered part of a normal degradative process that removes activated receptors from the plasma membrane and thereby truncates and controls downstream signaling [16]. But while this certainly remains a critical function of endocytosis, recent experiments demonstrate that EGFR remains phosphorylated and active following internalization [17], and that downstream signaling partners such as Ras colocalize with these internalized, endosome-associated receptors [18-23]. Moreover, the signals emanating from these internalized EGFR are biologically meaningful, as cell survival is directly supported by such signaling [24]. Likewise, Bild and colleagues recently observed that STAT-3 signaling initiated by EGFR activation localized to endocytic vesicles that moved from the plasma membrane to the nucleus, and they found that inhibition of EGFR endocytosis prevented STAT-3 nuclear translocation and abrogated STAT-3-mediated gene transcription [25]. However, while evidence supports the existence of signaling endosomes, it does not rule out simultaneous diffusion-based signal transduction.
We have previously provided evidence that neurotrophin-induced Erk1/2 signaling from retrogradely transported endosomes is more efficient than diffusion over distances ranging from 1.3 microns to 13 microns [7]. We also suggested that the phosphorylation signal associated with signaling endosomes is regenerative [7], consistent with our previous observations regarding the characterization of purified signaling endosomes from neurotrophin-stimulated cells [26]. Figure 4 provides additional analysis in support of the regenerative capacity of signaling endosomes. Such signal regeneration is in stark contrast to the terminal dephosphorylation experienced by diffusing signal transducers, and is a key element in favor of the signaling endosome hypothesis [4,7]. However, our previous observations depended upon the comparison of the Einstein-Stokes diffusion equation-derived root-mean-square effective distance for Erk1/2 and the average transport velocity for nerve growth factor [7]. Such a comparison overlooks a critical feature of signaling endosome transport and a critical failure of diffusion: directionality. Diffusion is inherently directionless, while the movement of signaling endosomes along microtubules is inherently directional and vectorial (see Figure 5 "The Microtubular Highway"). Likewise, simple modeling of the root-mean-square effective diffusion distance against transport velocity ignores dephosphorylation and the regenerative capacity of endosome-associated signals. Herein, we report that brute-force Monte Carlo (random walk) simulations of STAT-3 diffusion and dephosphorylation kinetics indicates that facilitated transport of endosomal-based signals is more efficient than diffusion over even very small cellular distances. Therefore, we conclude that signaling from endosomes represents a general biological principle relevant to all cell types and to all signal transduction pathways.
Figure 4 Growth factor receptors are internalized into clathrin-coated vesicles (CCVs) following ligand binding and receptor activation (1–5). These CCVs are uncoated (6) and mature into early endosomes (EE) (7) that may serve as transport endosomes [48]. The concentration of growth factor in transport endosomes is high enough to guarantee effectively 100% receptor occupancy. Hence, if the endosome-associated receptor encounters a phosphatase, the phosphorylation signal is rapidly regenerated.
Figure 5 The Microtubular Highway. Evidence of the directionality of dynein-mediated retrograde transport.
Results and discussion
Assumptions – Transport Velocity
For modeling, a dynein-based transport rate of 5 microns per second is assumed, based on a report by Kikushima and colleagues [27]. This value was used for ease of calculation: with a cell radius of 7.5 microns and a nuclear radius of 2.5 microns, a 5 μm per second transport rate moves the signaling endosome from the plasma membrane to the nucleus in one second. Actual transport rates likely range from 1–10 μm per second in cytosol or axoplasm [7].
Assumptions – Diffusion Coefficient
The crystal structure of STAT-3B [28], deposited in the Protein Data Bank as PDB 1BG1 [29], indicates unit cell dimensions of 17.4 × 17.4 × 7.9 nm. With the caveat that this structure is bound to an 18-base nucleic acid, the volume of a STAT-3B molecule is 2400 nm3. Assuming a spherical molecule, STAT-3B therefore has a molecular radius of approximately 8 nm. Likewise, the molecular weight of STAT-3 is 100000 Daltons, and therefore one molecule of STAT-3 weighs 1.7 × 10-19 g. The Einstein-Stokes equation for the coefficient of diffusion is:
D = (1/8)(k·T)/(π·γ·η)
where k is Boltzmann's constant, T is absolute temperature in degrees Kelvin, γ is the radius of the molecule, and η is the viscosity of an isotropic medium. The viscosity of axoplasm is approximately 5 centipoise [30], a value that also approximates cytoplasm [31,32]. Hence,
k = 1.3805 × 10-20 m2·g·(1/(s2·K))
T = 310 K
γ = 8 × 10-9 m
m = 1.7 × 10-19 g
η = 5 g/(m·s)
Therefore, the coefficient of diffusion for a molecule of STAT-3 is:
D = 4.3 μm2 per second
Likewise, the instantaneous velocity vx , the step length δ, and the step rate τ, were derived as:
vx = ((k·T)/m)0.5 = 5 m/s
δ = (1/4)(k·T)/(vx·π·γ·η) = 1.7 × 10-12 m
τ = vx /δ = 2.9 × 1012 sec-1
It is important to note that our mass estimation may substantially underestimate the actual mass of the functional STAT-3 molecular complex, described by Sehgal and colleagues as two populations with masses ranging from 200–400 kDa ("Statosome I") to 1–2 MDa ("Statosome II") [33,34]. Such a massive molecular complex certainly has important biological implications for STAT-3 diffusion. However, because no crystal structure exists for these higher molecular weight statosomes from which to calculate the molecular radius, and in order to calculate the "best-case scenario" for effective diffusion distance, we have calculated the STAT-3 diffusion coefficient on the basis of a 100 kDa monomeric molecule. The actual diffusion coefficient for STAT-3 may be 30% of the value calculated above (assuming 2 MDa mass and a four-fold increase in molecular radius to account for molecular packing of the statosome) and the root-mean-square displacement may be 50% of the value calculated below. The impact of these variables awaits further investigation.
Assumptions – Diffusion Modeling
We modeled diffusion using a random walk algorithm in two dimensions. The choice of dimensionality was constrained by the intensive computational burden associated with three-dimensional algorithms, as discussed below (see Methods). At every iteration of the random walk two pseudo-random numbers (see Methods) were generated and used to determine the direction of movement in the x-y plane. Using the instantaneous velocity vx , the step length δ, and the step rate τ, defined above, we conclude that a diffusing molecule of STAT-3 will randomly walk 3 × 1012 steps per second, and each step will be 1.7 × 10-12 meters long. Thus, the root-mean-square displacement for STAT-3 diffusion in one second is 2.9 μm. The random walk was modeled on one second of biological time using a loop of 3 × 1012 iterations. During each iteration the molecule randomly moved ± 1.7 × 10-12 meters in the x-plane and ± 1.7 × 10-12 meters in the y-plane.
Assumptions – Dephosphorylation Kinetics
The decay of a phospho-protein is an exponential function mapped between the plasma membrane and the nucleus [5,35]:
α2 = (Kp )(L2/D)
And the probability function for dephosphorylation is:
p(x)/p(m) = (eαx – e-αx)/x(eα – e-α)
Where α is a dimensionless measure of dephosphorylation probability, Kp is the first-order rate constant for the activity of the relevant phosphatase, L is the cell diameter, D is the diffusion coefficient, x is the distance from the cell center, and m is the distance from cell center to plasma membrane normalized to a value of one. α scales such that for α = 10, half of all phospho-molecules become dephosphorylated within approximately 0.075 units of distance from the plasma membrane to the cell center (e.g. 750 nm for a cell with 10 μm radius) [5]. In general, Kp , the first-order rate constant of phosphatase activity, varies between 0.1 per second and 10 per second [4,35-37]. For our model Kp = 5 was assumed, yielding α = 8.1.
With regard to an estimate of enzymatic activity relevant to dephosphorylation of STAT-3, Todd and colleagues report a second-order rate constant of 40000/M·s for dephosphorylation of Erk1/2 [38], which gives:
kcat /km = 40000/M·s
Furthermore, Denu and colleagues report that diphosphosphorylated Erk1/2 peptides exhibit km values of approximately 100 μM in vitro [39]. Therefore:
kcat = 4/s
Since kcat measures the number of substrate molecules turned over per enzyme per second, a kcat of 4 per second means that, on average, each molecule of enzyme (phosphatase) converts (dephosphorylates) 4 substrate molecules every second. Assuming a degree of molecular similarity between Erk dephosphorylation and STAT-3 dephosphorylation, and for ease of calculation, we set kcat = 5 per second. It is important to note that this assumption may not be valid, but has been necessarily adopted in the absence of better biophysical data in order to illustrate the potential circumscription of diffusion by dephosphorylation.
Assumptions – Dephosphorylation Modeling
The random walk employed for modeling STAT-3 diffusion depends upon the massively iterative generation of random numbers to describe the movement of the walking molecule in two-dimensional space. Since significant computational time was already invested in our diffusion calculations for the generation of extremely long period pseudo-random numbers, we opted to model STAT-3 dephosphorylation as a stochastic event using the following logic: for any given randomly walking molecule, the probability of encountering a phosphatase is independent of both all other molecules and all other steps in the walk. Therefore, during one second of biological time, equivalent to 3 × 1012 steps in the random walk, and assuming that kcat = 5 dephosphorylations per second, there will be 1.67 × 10-12 dephosphorylation events per step. This can be effectively modeled as a probability test by generating a pseudo-random number on (0,1) at each step of the random walk and asking whether this number is less than 1.67 × 10-12. If the test is positive, the molecule is considered to be "dephosphorylated" and the random walk is truncated. High-speed modeling of time to dephosphorylation for a large number of molecules (i.e. in the absence of the random walk) led to a probability function that matched the equations described by Kholodenko [5].
Results – Diffusion-only Model
Figure 6 shows the result of 12 random walks plotted in two-dimensional space and compared to the pathlength of a signaling endosome transported on microtubules. For these simulations, 500 milliseconds of biological time were modeled, resulting in the transport of the signaling endosome over 2.5 μm. The random walks were simulated using only the diffusion coefficient criteria (i.e. no dephosphorylation modeling) over the same time window. This figure illustrates the tremendous variability in the path vector for each of the diffusing particles. While not unexpected or surprising, Figure 6 offers graphic evidence that the model is working appropriately. Average pathlength analysis is discussed below.
Figure 6 Representative trajectories for 12 random walk simulations using only diffusion criteria (red and blue lines), compared to the movement of a signaling endosome within the same 500 millisecond time frame (green line). Parameters: 15 μm cell diameter, 5 μm nucleus diameter, 37°C, 500 msec, coefficient of diffusion as described in the text. Arrows along the plasma membrane surface denote the sites of signal initiation.
Results – Diffusion and Dephosphorylation Model
Figure 7 shows the result of 22 random walks modeled over one second of biological time incorporating both the diffusion coefficient criteria and the dephosphorylation probability criteria. Again, the random walks are compared to the pathlength for the transported signaling endosome, which in this case moves across the entire 5 μm distance separating the plasma membrane and the nucleus. As with Figure 6, there is a large amount of variability in the diffusion paths, but it is clear that the incorporation of dephosphorylation into the model substantially truncates the effective distance over which a diffusing molecule of STAT-3 travels. As discussed above, with α = 8.1, 50% of all phosphorylated molecules should be dephosphorylated within 0.1 distance units of the plasma membrane. For our model, this means that 50% of phospho-STAT-3 molecules should be inactivated within 750 nm of the plasma membrane (α = 8.1; x = 0.9 for p = 0.5; radius = 7.5 μm; hence x = 6.75 μm, or 750 nm from the plasma membrane). Likewise, only 15% of phosphorylated STAT-3 molecules remain active at a distance half-way between the cell center and the plasma membrane, and, assuming a nucleus of 2.5 μm radius in a cell with 7.5 μm radius, fewer than 4% of phosphorylated molecules will cross the entire distance. Our random walk incorporating the dephosphorylation probability model captures the salient features of the expected dephosphorylation kinetics.
Figure 7 Representative trajectories for 22 random walk simulations using both diffusion and dephosphorylation criteria (red and blue lines), compared to the movement of a signaling endosome within the same 1 second time frame (green line). Parameters: 15 μm cell diameter, 5 μm nucleus diameter, 37°C, 1 sec, coefficient of diffusion and dephosphorylation probability as described in the text. Arrows along the plasma membrane surface denote the sites of signal initiation.
Results – Endpoint Analysis of Both Models
Finally, Figure 8 illustrates the endpoint analysis for 100 diffusion-only random walks and 100 diffusion plus dephosphorylation walks. It should be noted that each random walk required, on average, more than 48 hours of dedicated processor time. For this analysis, the final coordinate of each diffusing molecule was used to calculate a vector for the random walk (i.e. distance and direction from point of origin). Of the 200 vectors calculated under both models, no diffusing molecule intersected the nuclear membrane within the computed timeframe. In contrast, for the one second computations incorporating both diffusion and dephosphorylation, the retrogradely transported signaling endosome reaches the nucleus with the STAT-3 phosphorylation state intact. Finally, the observed root-mean-square displacement for the 100 dephosphorylation model random walks was 0.96 μm ± 0.1 μm, or less than 20% of the distance from the plasma membrane to the nucleus. As calculated above using only the step length and step rate derived from the coefficient of diffusion parameters, the predicted root-mean-square displacement for STAT-3 is 2.9 μm. Thus, the observed effective distance for a phosphorylated STAT-3 molecule is one-third of the predicted distance, indicating that our previously published analysis substantially overestimated the range over which diffusion efficiently transmits intracellular information.
Figure 8 Endpoint analysis of 100 diffusion-only random walks and 100 diffusion plus dephosphorylation random walks. Black lines represent vectors calculated by the final random walk point for each simulation, compared to the distance covered by a retrogradely transported signaling endosome in the same amount of time (green lines). The blue line represents the averaged vector for 100 diffusion-only random walks, while the red line depicts the averaged vector for 100 diffusion plus dephosphorylation simulations.
Predictions
Using the observed root-mean-square displacement after one second of biological time to establish an adjustment factor (33% of predicted), and assuming that the relationship between observed and predicted values is linear through time, we generated the plots shown in Figure 9. Figure 9A shows that the signaling endosome becomes more efficient at transmitting information from the plasma membrane over distances greater than 2 microns (greater than 400 milliseconds of biological time) using the predicted root-mean-square displacement values for comparison. However, using the adjusted root-mean-square displacement values for comparison, the signaling endosome is more efficient than diffusion within 200 nanometers from the plasma membrane (within 40 milliseconds of biological time) (Figure 9B). Therefore, our model predicts that the facilitated retrograde transport of signaling endosomes is a more efficient mechanism of information transfer from the plasma membrane to the nucleus, and is, in fact, more efficient for the transmission of phosphorylated STAT-3 signals over any distance greater than only 200 nanometers.
Figure 9 A and B) Diffusion modeling incorporating dephosphorylation kinetics indicates substantial truncation of the root-mean-square (r.m.s.) displacement for STAT-3 diffusion (dashed red line compared to solid red line). This has the effect of reducing the crossing point at which signaling endosome transport (solid blue line) overcomes diffusion (ca. 2 μm for theoretical r.m.s. vs. transport reduced to ca. 200 nm for adjusted r.m.s. vs. transport). B shows same data as A at higher Y-axis magnification.
Caveats and Future Directions
The signaling endosome retrograde transport rate utilized in our model may overestimate the actual transport velocity, especially as an average across the entire lifetime of the endosome-associated signal. The rate we modeled did not account for the kinetics of endocytosis or of vesicle loading onto the microtubule network. Our previous observations suggested transport velocities that ranged from 5.6 μm per second to 0.56 μm per second [7], but experiments addressing real transport rates for a variety of signaling molecules are required to improve our model. On the other hand, while we potentially overestimated the retrograde transport rate for the signaling endosome, we also very likely overestimated the size of the effective diffusion domain due to the two-dimensional restrictions of our current model. While the cytoskeletal transport of the signaling endosome is inherently a dimensionally-restricted vectorial event, diffusion within the cell most certainly occurs in three dimensions. Our current model predicts a three-fold reduction in the actual root-mean-square displacement for STAT-3 as compared to the predicted displacement using a two-dimensional random walk model, and we predict that a model incorporating three dimensions will exhibit even greater curtailment of the effective spatial domain for diffusion. However, the addition of a third dimension to the random walk simulations substantially increases computational demand, and therefore this analysis awaits either a more efficient algorithm or more computer time. Our current and future goals are to parallelize the random walk algorithm in order to perform massively parallel diffusion simulations in three dimensions.
Conclusion
Molecular diffusion obviously benefits from the extremely high molecular velocities of single particles moving in a vacuum. For gases and other very small molecules and under conditions of low viscosity or high temperature, diffusion is extremely fast and far-ranging. However, within the context of biological molecules and biological viscosities, diffusion is vastly circumscribed [1-3,40]. Despite the limitations imposed by biological parameters, diffusion at first glance still appears to be a viable mechanism for the transmission of information through cytoplasm. In fact, the "textbook" conception of signal transduction depends upon the free diffusion of signaling molecules. However, closer scrutiny finds several faults in the diffusion model [1]. For example, diffusion is certainly directionless – even within the context of a bounded space such as the cell, the majority of molecular motions taken by a diffusing molecule are non-productive with regard to movement of signals toward a target (such as the nucleus). Likewise, a diffusing molecular signal is a ready target for interaction with and truncation by cytoplasmic phosphatases. Certainly, the effective range over which a diffusing signal maintains informational integrity depends upon the concentration and activity of equally randomly diffusing phosphatases, but it also seems likely that cells maintain levels of phosphatase sufficient to prevent run-away signal transduction [41,42]. Thus, diffusion of information is limited by both lack of direction and inevitable signal elimination. In distinct contrast, the retrograde movement of quantal signaling units capable of regenerating the information content of the original stimulus is inherently vectorial. Therefore, signaling endosomes, despite an overall lower transport velocity compared to diffusion velocities, exhibit characteristics of an optimized information transmission system. We previously sought to determine the effective range over which Erk1/2 signaling endosomes exhibited greater efficiency than diffusing Erk1/2 molecules [7]. This work relied upon the direct comparison of the root-mean-square displacement for phosphorylated Erk1/2 with the retrograde transport velocity of neurotrophin-induced signaling endosomes. In an effort to refine this model we incorporated in our present study the additional element of dephosphorylation kinetics. Thus our current model addresses both the non-vectorial nature of diffusion and the inherent susceptibility to signal truncation by interaction with cellular phosphatases. Using an iterative random walk modeling scheme we determined that the root-mean-square displacement predicted by the coefficient of diffusion for STAT-3 overestimated the root-mean-square displacement observed in our simulations by a factor of 3. Incorporating this scaling factor into the equation for root-mean-square displacement through time, we found that signaling endosomes become more effective at the transmission of information when the distance from the plasma membrane exceeds 200 nanometers. This observation suggests that any cellular situation that requires the transmission of information in the form of phosphorylated signaling molecules over distances in excess of 200 nanometers would benefit from the packaging of such signals into quantal, cytoskeleton-associated signaling packets such as signaling endosomes.
Our model suggests that cells utilize two distinct information transmission paradigms: 1) fast local signaling via diffusion over spatial domains on the order of less than 200 nanometers; 2) long-distance (>200 nanometers) signaling via information packets associated with the cytoskeletal transport apparatus. Moreover, while we have focused explicitly on the role of signaling endosomes derived from the internalization of plasma membrane receptor tyrosine kinases and associated downstream signaling partners, our model suggests that any signal that must move from the outer reaches of the cytoplasm to the perinuclear region would benefit from an association with the retrograde transport machine. For example, transcription factors may associate directly with molecular motors and chaperone proteins that protect them from dephosphorylation in a nonvesiculated manner that takes advantage of directional retrograde transport in the absence of a plasma-membrane-derived organelle. Such a mechanism was recently proposed for the transport of soluble (i.e. non-membrane-associated) activated Erk1/2 within injured axons [43]. Thus, our model supports previous observations suggesting that the signaling endosome hypothesis is a subset of a more general hypothesis that the most efficient mechanism for intracellular signaling-at-a-distance involves the association of signaling molecules with molecular motors that move along the cytoskeleton [4]. The additional benefit provided by the cytoskeletal association of membrane-bounded complexes that package a ligand-bound transmembrane receptor with downstream effector molecules is the ability to regenerate the signal at any point along the transmission path [7]. We conclude that signaling endosomes provide unique information transmission properties relevant to all cell architectures, and we propose that the majority of relevant information transmitted from the plasma membrane to the nucleus will be found in association with organelles of endocytic origin.
Methods
Pseudo-Random Number Generation
It should be self-evident that "built-in" pseudo-random number generators (RNGs) available in the majority of operating systems and programming languages are essentially useless for large-scale Monte Carlo simulations [44]. However, during our initial efforts to optimize the processing time for the one-second simulations we experimented with several common RNGs; all failed to exhibit sufficiently long periods, a failure that was manifested in an initial period of random walking followed by capture in a continuously repeating cyclical path. We also experimented with an implementation of the Mersenne Twister algorithm, which exhibited a robust period (theoretically 219937-1) and computational demand comparable to many other standard RNGs [45]. However, our final optimized diffusion-only code utilized a multiply-with-carry RNG (MWC) described by George Marsaglia [44,46,47]. The MWC algorithm generates extremely long-period pseudo-random numbers on [0,1], and we utilized this very efficient RNG for Boolean testing of step direction in two dimensions. For the combined diffusion and dephosphorylation models, we used the Mersenne Twister modified to generate pseudo-random numbers on (0,1) for the probabilistic determination of a dephosphorylation event and the MWC algorithm for step direction determination.
Hardware
We utilized a variety of platforms for development, testing, and implementation of the diffusion models, including the IBM Power4 p690 supercomputer (running AIX 5.2) and the SGI Altix 3700 supercomputer (running SGI Advanced Linux 3.4) at the University of Minnesota Supercomputing Institute. The serial models described above were primarily implemented on a single processor Intel P4 3.0 GHz machine running Red Hat Linux 9.0. The IBM Power4, the SGI Altix 3700, and a dual processor Xeon 3.0 GHz Nocona box running Red Hat Enterprise Linux 3.0 were used for development and testing of parallel implementations. Total wallclock time on all platforms currently exceeds 10000 hours.
Software
All algorithms were coded in C and compiled with gcc or xlc (serial implementations) or with pgcc, xlc, or icc (OpenMP parallel implementations). Our first diffusion model efforts required more than one week of dedicated processing time per walk; after several rounds of code optimization we could obtain one second of simulated time in approximately 48 hours on the Power4 architecture and the Pentium 4 architecture described above.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The author contributed to all phases of the work.
Acknowledgements
The author thanks the University of Minnesota Supercomputing Institute (MSI) for access to the IBM Power4 pSeries 690 and to the SGI Altix supercomputers. The author also thanks Dr. Birali Runesha of the MSI for technical assistance. This work was supported by Donald and Frances Herdrich and by grant RG3636 from the National Multiple Sclerosis Society.
==== Refs
Agutter PS Malone PC Wheatley DN Intracellular transport mechanisms: a critique of diffusion theory J Theor Biol 1995 176 261 272 7475114 10.1006/jtbi.1995.0196
Agutter PS Malone PC Wheatley DN Diffusion theory in biology: a relic of mechanistic materialism J Hist Biol 2000 33 71 111 11624418 10.1023/A:1004745516972
Agutter PS Wheatley DN Random walks and cell size Bioessays 2000 22 1018 1023 11056478 10.1002/1521-1878(200011)22:11<1018::AID-BIES8>3.0.CO;2-Y
Kholodenko BN MAP kinase cascade signaling and endocytic trafficking: a marriage of convenience? Trends Cell Biol 2002 12 173 177 11978536 10.1016/S0962-8924(02)02251-1
Kholodenko BN Four-dimensional organization of protein kinase signaling cascades: the roles of diffusion, endocytosis and molecular motors J Exp Biol 2003 206 2073 2082 12756289 10.1242/jeb.00298
Sofroniew MV Howe CL Mobley WC Nerve growth factor signaling, neuroprotection, and neural repair Annu Rev Neurosci 2001 24 1217 1281 11520933 10.1146/annurev.neuro.24.1.1217
Howe CL Mobley WC Signaling endosome hypothesis: A cellular mechanism for long distance communication J Neurobiol 2004 58 207 216 14704953 10.1002/neu.10323
Miaczynska M Pelkmans L Zerial M Not just a sink: endosomes in control of signal transduction Curr Opin Cell Biol 2004 16 400 406 15261672 10.1016/j.ceb.2004.06.005
Guzik BW Goldstein LS Microtubule-dependent transport in neurons: steps towards an understanding of regulation, function and dysfunction Curr Opin Cell Biol 2004 16 443 450 15261678 10.1016/j.ceb.2004.06.002
Zweifel LS Kuruvilla R Ginty DD Functions and mechanisms of retrograde neurotrophin signalling Nat Rev Neurosci 2005 6 615 625 16062170 10.1038/nrn1727
Weible MW Hendry IA What is the importance of multivesicular bodies in retrograde axonal transport in vivo? J Neurobiol 2004 58 230 243 14704955 10.1002/neu.10318
Howe CL Mobley WC Long-distance retrograde neurotrophic signaling Curr Opin Neurobiol 2005 15 40 48 15721743 10.1016/j.conb.2005.01.010
Carpenter G The EGF receptor: a nexus for trafficking and signaling Bioessays 2000 22 697 707 10918300 10.1002/1521-1878(200008)22:8<697::AID-BIES3>3.0.CO;2-1
Sorkin A Internalization of the epidermal growth factor receptor: role in signalling Biochem Soc Trans 2001 29 480 484 11498013 10.1042/BST0290480
Sorkin A Von Zastrow M Signal transduction and endocytosis: close encounters of many kinds Nat Rev Mol Cell Biol 2002 3 600 614 12154371 10.1038/nrm883
Le Roy C Wrana JL Clathrin- and non-clathrin-mediated endocytic regulation of cell signalling Nat Rev Mol Cell Biol 2005 6 112 126 15687999 10.1038/nrm1571
Sorkin A McClure M Huang F Carter R Interaction of EGF receptor and grb2 in living cells visualized by fluorescence resonance energy transfer (FRET) microscopy Curr Biol 2000 10 1395 1398 11084343 10.1016/S0960-9822(00)00785-5
Wiley HS Burke PM Regulation of receptor tyrosine kinase signaling by endocytic trafficking Traffic 2001 2 12 18 11208164 10.1034/j.1600-0854.2001.020103.x
Roy S Wyse B Hancock JF H-Ras signaling and K-Ras signaling are differentially dependent on endocytosis Mol Cell Biol 2002 22 5128 5140 12077341 10.1128/MCB.22.14.5128-5140.2002
Pol A Calvo M Enrich C Isolated endosomes from quiescent rat liver contain the signal transduction machinery. Differential distribution of activated Raf-1 and Mek in the endocytic compartment FEBS Lett 1998 441 34 38 9877160 10.1016/S0014-5793(98)01517-8
Jiang X Sorkin A Coordinated traffic of Grb2 and Ras during epidermal growth factor receptor endocytosis visualized in living cells Mol Biol Cell 2002 13 1522 1535 12006650 10.1091/mbc.01-11-0552
Chiu VK Bivona T Hach A Sajous JB Silletti J Wiener H Johnson RL Cox AD Philips MR Ras signalling on the endoplasmic reticulum and the Golgi Nat Cell Biol 2002 4 343 350 11988737
Burke P Schooler K Wiley HS Regulation of epidermal growth factor receptor signaling by endocytosis and intracellular trafficking Mol Biol Cell 2001 12 1897 1910 11408594
Wang Y Pennock S Chen X Wang Z Endosomal signaling of epidermal growth factor receptor stimulates signal transduction pathways leading to cell survival Mol Cell Biol 2002 22 7279 7290 12242303 10.1128/MCB.22.20.7279-7290.2002
Bild AH Turkson J Jove R Cytoplasmic transport of Stat3 by receptor-mediated endocytosis Embo J 2002 21 3255 3263 12093727 10.1093/emboj/cdf351
Howe CL Valletta JS Rusnak AS Mobley WC NGF signaling from clathrin-coated vesicles: evidence that signaling endosomes serve as a platform for the Ras-MAPK pathway Neuron 2001 32 801 814 11738027 10.1016/S0896-6273(01)00526-8
Kikushima K Yagi T Kamiya R Slow ADP-dependent acceleration of microtubule translocation produced by an axonemal dynein FEBS Lett 2004 563 119 122 15063734 10.1016/S0014-5793(04)00278-9
Becker S Groner B Muller CW Three-dimensional structure of the Stat3beta homodimer bound to DNA Nature 1998 394 145 151 9671298 10.1038/28980
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucleic Acids Res 2000 28 235 242 10592235 10.1093/nar/28.1.235
Haak RA Kleinhans FW Ochs S The viscosity of mammalian nerve axoplasm measured by electron spin resonance J Physiol 1976 263 115 137 65468
Luby-Phelps K Cytoarchitecture and physical properties of cytoplasm: volume, viscosity, diffusion, intracellular surface area Int Rev Cytol 2000 192 189 221 10553280
Luby-Phelps K Lanni F Taylor DL The submicroscopic properties of cytoplasm as a determinant of cellular function Annu Rev Biophys Biophys Chem 1988 17 369 396 3293592 10.1146/annurev.bb.17.060188.002101
Guo GG Patel K Kumar V Shah M Fried VA Etlinger JD Sehgal PB Association of the chaperone glucose-regulated protein 58 (GRP58/ER-60/ERp57) with Stat3 in cytosol and plasma membrane complexes J Interferon Cytokine Res 2002 22 555 563 12060494 10.1089/10799900252982034
Ndubuisi MI Guo GG Fried VA Etlinger JD Sehgal PB Cellular physiology of STAT3: Where's the cytoplasmic monomer? J Biol Chem 1999 274 25499 25509 10464281 10.1074/jbc.274.36.25499
Kholodenko BN Brown GC Hoek JB Diffusion control of protein phosphorylation in signal transduction pathways Biochem J 2000 350 Pt 3 901 907 10970807 10.1042/0264-6021:3500901
Haugh JM Lauffenburger DA Physical modulation of intracellular signaling processes by locational regulation Biophys J 1997 72 2014 2031 9129805
Zhao Y Zhang ZY The mechanism of dephosphorylation of extracellular signal-regulated kinase 2 by mitogen-activated protein kinase phosphatase 3 J Biol Chem 2001 276 32382 32391 11432864 10.1074/jbc.M103369200
Todd JL Tanner KG Denu JM Extracellular regulated kinases (ERK) 1 and ERK2 are authentic substrates for the dual-specificity protein-tyrosine phosphatase VHR. A novel role in down-regulating the ERK pathway J Biol Chem 1999 274 13271 13280 10224087 10.1074/jbc.274.19.13271
Denu JM Zhou G Wu L Zhao R Yuvaniyama J Saper MA Dixon JE The purification and characterization of a human dual-specific protein tyrosine phosphatase J Biol Chem 1995 270 3796 3803 7876121 10.1074/jbc.270.45.26782
Wheatley DN Diffusion theory, the cell and the synapse Biosystems 1998 45 151 163 9544405 10.1016/S0303-2647(97)00073-7
Hunter T Signaling--2000 and beyond Cell 2000 100 113 127 10647936 10.1016/S0092-8674(00)81688-8
Blume-Jensen P Hunter T Oncogenic kinase signalling Nature 2001 411 355 365 11357143 10.1038/35077225
Perlson E Hanz S Ben-Yaakov K Segal-Ruder Y Seger R Fainzilber M Vimentin-dependent spatial translocation of an activated MAP kinase in injured nerve Neuron 2005 45 715 726 15748847 10.1016/j.neuron.2005.01.023
L'Ecuyer P Gentle JE, Haerdle W and Mori Y Random Number Generation Handbook of Computational Statistics 2004 , Springer-Verlag 35 70
Matsumoto M Nishimura T Mersenne Twister: A 623-dimensionally equidistributed uniform pseudo-random number generator ACM Transactions on Modeling and Computer Simulation 1998 8 3 30 10.1145/272991.272995
L'Ecuyer P Blouin F Couture R A search for good multiple recursive random number generators ACM Transactions on Modeling and Computer Simulation 1993 3 87 98 10.1145/169702.169698
Goresky M Klapper A Efficient multiply-with-carry random number generators with maximal period ACM Transactions on Modeling and Computer Simulation 2003 13 1 12 10.1145/945511.945514
Delcroix JD Valletta JS Wu C Hunt SV Kowal AS Mobley WC NGF Signaling in sensory neurons: evidence that early endosomes carry NGF retrograde signals Neuron 2003 39 69 84 12848933 10.1016/S0896-6273(03)00397-0
Valetti C Wetzel DM Schrader M Hasbani MJ Gill SR Kreis TE Schroer TA Role of dynactin in endocytic traffic: effects of dynamitin overexpression and colocalization with CLIP-170 Mol Biol Cell 1999 10 4107 4120 10588646
Quintyne NJ Gill SR Eckley DM Crego CL Compton DA Schroer TA Dynactin is required for microtubule anchoring at centrosomes J Cell Biol 1999 147 321 334 10525538 10.1083/jcb.147.2.321
King SJ Brown CL Maier KC Quintyne NJ Schroer TA Analysis of the dynein-dynactin interaction in vitro and in vivo Mol Biol Cell 2003 14 5089 5097 14565986 10.1091/mbc.E03-01-0025
Burkhardt JK Echeverri CJ Nilsson T Vallee RB Overexpression of the dynamitin (p50) subunit of the dynactin complex disrupts dynein-dependent maintenance of membrane organelle distribution J Cell Biol 1997 139 469 484 9334349 10.1083/jcb.139.2.469
Lakadamyali M Rust MJ Babcock HP Zhuang X Visualizing infection of individual influenza viruses Proc Natl Acad Sci U S A 2003 100 9280 9285 12883000 10.1073/pnas.0832269100
Heerssen HM Pazyra MF Segal RA Dynein motors transport activated Trks to promote survival of target-dependent neurons Nat Neurosci 2004 7 596 604 15122257 10.1038/nn1242
Watson FL Heerssen HM Moheban DB Lin MZ Sauvageot CM Bhattacharyya A Pomeroy SL Segal RA Rapid nuclear responses to target-derived neurotrophins require retrograde transport of ligand-receptor complex J Neurosci 1999 19 7889 7900 10479691
Yano H Lee FS Kong H Chuang J Arevalo J Perez P Sung C Chao MV Association of Trk neurotrophin receptors with components of the cytoplasmic dynein motor J Neurosci 2001 21 RC125 11157096
Bhattacharyya A Watson FL Pomeroy SL Zhang YZ Stiles CD Segal RA High-resolution imaging demonstrates dynein-based vesicular transport of activated Trk receptors J Neurobiol 2002 51 302 312 12150505 10.1002/neu.10062
Ehlers MD Kaplan DR Price DL Koliatsos VE NGF-stimulated retrograde transport of trkA in the mammalian nervous system J Cell Biol 1995 130 149 156 7540615 10.1083/jcb.130.1.149
|
16236165
|
PMC1276819
|
CC BY
|
2021-01-04 16:39:26
|
no
|
Theor Biol Med Model. 2005 Oct 19; 2:43
|
utf-8
|
Theor Biol Med Model
| 2,005 |
10.1186/1742-4682-2-43
|
oa_comm
|
==== Front
J Transl MedJournal of Translational Medicine1479-5876BioMed Central London 1479-5876-3-371619755310.1186/1479-5876-3-37ResearchInterleukin-1β induced vascular permeability is dependent on induction of endothelial Tissue Factor (TF) activity Puhlmann Markus [email protected] David M [email protected] Jeffrey M [email protected] Nancy M [email protected] Ewa M [email protected] H Richard [email protected] Surgery Branch, National Cancer Institute, National Institutes of Health, Bethesda, Maryland, 20892, USA2005 30 9 2005 3 37 37 15 7 2005 30 9 2005 Copyright © 2005 Puhlmann et al; licensee BioMed Central Ltd.2005Puhlmann et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
IL-1β is a pleotropic cytokine that may mediate increased procoagulant activity and permeability in endothelial tissue during inflammatory conditions. The procoagulant effects of IL-1β are mediated through induction of tissue factor (TF) but its alterations on vascular permeability are not well characterized. We found that IL-1β induced a rapid and dose-dependent increase in TF activity in human umbilical vein endothelial cells (ECs) under routine culture conditions. However, IL-1β caused a rapid and marked increase in permeability across confluent EC monolayers using a two-compartment in vitro model only in the presence of factor VIII-deficient plasma that was completely abrogated by neutralizing anti-TF antibody pre-treatment. In vitro permeability was associated with loss of EC surface expression of VE-cadherin and contraction of F-actin cytoskeletal elements that resulted in EC intercellular gap formation. These data demonstrate that IL-1β induces marked changes in permeability across activated endothelium via a TF dependent mechanism and suggest that modulation of TF activity may represent a strategy to treat various acute and chronic inflammatory conditions mediated by this cytokine.
videomicroscopyinflammationneovasculatureendothelial tissue
==== Body
Introduction
IL-1β is a pleotropic 17.5 kD cytokine, secreted primarily by monocytes and macrophages, that mediates the pathophysiology of various acute and chronic inflammatory conditions. High levels of circulating IL-1 have been identified in experimental models of endotoxic shock and acute bacterial infection and increased gene expression of IL-1 has been identified in tissues at sites of experimentally induced inflammation [1,2]. Clinically, high levels of circulating IL-1 have been identified in patients with acute bacterial infections and elevated levels of IL-1 have been detected in the diseased articular tissues of patients with chronic rheumatoid arthritis [3,4]. In experimental models of endotoxin or E. coli induced shock, immune complex colitis, cancer progression, cachexia, and non-specific inflammation, IL-1 blockade significantly ameliorates the pathophysiological host response in these conditions [5-9]. Of note, administration of recombinant IL-1ra is used clinically to ameliorate the symptoms of severe rheumatoid arthritis [10].
Although IL-1 receptors are widely expressed on different cell types, the protein appears to exert many of its physiological actions through effects on endothelial tissue. IL-1 induces endothelial cell (EC) production of cytokines such as IL-8, IL-6 and TNF, multiple cytokine receptors, adhesion molecules, growth factors, matrix metalloproteinases, and coagulation factors such as tissue factor, fibrinogen, urokinase plasminogen activator, type 1,2 plasminogen activator and protease nexin 1 [11-14].
Two sentinel homeostatic functions of vasculature are initiation of coagulation via the extrinsic clotting cascade and regulation of vascular permeability. Increased vascular permeability most likely represents the initial event in pathological or reparative inflammation or angiogenesis by allowing efflux of plasma proteins into interstitium to serve as a provisional matrix for circulating inflammatory cells or activated endothelium [15]. IL-1 is known to induce a procoagulant phenotype on endothelial tissue primarily through induction of TF [16,17] and has been shown to alter endothelial cell monolayer permeability in in vitro models [18,19]. However, its effects on vascular permeability in vivo have not been consistent and the mechanisms for these effects are not known [20-22]. For example, in lapine models, systemic recombinant IL-1 has been shown to induce shock and significant pulmonary vascular injury manifested by marked perivascular pulmonary edema [20,23]. However, in rodent or guinea pig models others have not demonstrated any independent effects of aerosolized or systemically administered IL-1 on pulmonary vascular permeability or injury [21,22].
The present studies were performed to characterize the inflammatory properties of IL-1 as mediated by alterations in permeability across endothelial monolayers in vitro. The data indicate that IL-1 induces both procoagulant effects and permeability via a single TF dependent mechanism and suggest that TF may represent a novel target for acute or chronic inflammatory conditions mediated by IL-1.
Materials and methods
Cell culture
Human umbilical vein ECs (hUVECs) were obtained from Clonetics (Clonetics #CC 2519) and propagated at 37°C in a 5% CO2 incubator in EGM-2 media (Clonetics #CC-3156 and #CC-4176) according to the manufacturer's instructions but without addition of the VEGF-supplement. Cells in the experiments were passaged for a maximum of 5 generations. MC-38, a non-metastatic methylcholanthrene-induced murine adenocarcinoma was maintained in Dulbecco's modified Eagle's medium (DMEM) (Biofluids, Rockville, MD) supplemented with 10% heat-inactivated fetal calf serum, 2 mM glutamine and 50 U/mL penicillin/streptomycin.
In vitro permeability assay
In-vitro permeability was quantitated by spectrophotometric measurement of Evans Blue-bound albumin across functional hUVEC monolayers using a modified 2-compartment chamber model as previously described in detail. Briefly, hUVECs were plated (8 × 105/ well) in 1 μm PET Transwell filter inserts (Falcon #35 3102) using EGM-2 (as described above). On day 3, cells were washed and treated with EBM-2 (Clonetics #CC-3156) with 2% BSA with or without IL-1β with various concentrations as indicated. Except during time course experiments, cytokine exposure was kept constant at 2 hours (at 37°C, 5% CO2). Cells were briefly washed with PBS/ Ca2+ and Mg2+. Fresh EBM-2 including 1% Factor VIII-deficient plasma (George-King Biochemical) (where indicated) was added and incubated for an additional 1 hour at 37°C, 5% CO2. In experiments evaluating TF-blocking antibodies, designated wells were incubated first with a 1:100 dilution of antibody (American Diagnostica, #4509) and PBS/1% BSA. Inserts were washed with PBS/ Ca2+ and Mg2+ before adding 1.5 mL of Evans Blue (EB) bound to 0.1% BSA in PBS in the upper chamber. Two mL PBS/ Ca2+ and Mg2+ was added to the lower chamber in which absorbance of EB was determined after 1 hr using a spectrophotometer (620 nm). Experiments were performed in triplicates and repeated multiple times.
Tissue factor assay
Tissue factor activity was determined using a 1-stage coagulation assay. HUVECs were plated (8 × 105/ well) in a 6-well plate and incubated for 3 days. Cells were washed with PBS and treated with various concentrations of IL-1β (in EBM-2/ 2% BSA) for 2 hours at 37°C, 5% CO2. After this incubation, HUVECs were washed with sterile PBS followed by incubation for 10 min at 37°C in 1 ml TRIS (25 mM, pH 8). Plates were transferred to a freezer (-80°C) for at least 1 hour. After thawing the plates at 37°C, cells were harvested using a mechanical cell scraper and centrifuged in an Eppendorf table centrifuge at 14,000 rpm for 5 min. Cell pellets were suspended in 120 μL of assay buffer (25 mM TRIS/ PBS/0.1% BSA) and immediately assayed.
Tissue factor activity was measured using a 1-step coagulation assay (Amelung microcoagulation analyzer) by addition of 100 μL factor VIII-deficient human plasma (George-King Biomedical). Clotting time was determined at 37°C and measurement started after addition of 100 μL of 30 nM CaCl2 (Sigma). The reference curve for TF was obtained by reconstituting various concentrations of lipidated recombinant hTF (American Diagnostica, # 4500 L/B2) according to the manufacturer's recommendation and allowed for detection limits between 0.395 ng/mL to 25 ng/mL (7 – 470 pM/mL).
Immunohistochemistry
HUVECs were grown to confluence on fibronectin coated two well culture slides (BIOCOAT Fibronectin, Becton Dickinson, # 354628) and exposed to IL-1 as described for the permeability assays. After fixation with 4% formaldehyde (RT, 10 min), cells were permeabilized with 1% Triton X for 10 min. A 1:200 dilution of a polyclonal VE-cadherin antibody (Santa Cruz Biotechnology, #sc-6458) was used as primary antibody and incubated at RT for 1 hour. After several washes with PBS, cells were labeled with Alexa Fluor 488 donkey anti-goat IgG (Molecular Probes, #A-11055) for 1 hour at RT. After repeated washes, the F-actin cytoskeleton was stained with Texas Red labeled phalloidin (Molecular Probes, #T-7471) for 40 min at room temperature. Slides were mounted and sealed using Gel/Mount (Biomeda, #M01). Photomicrographs were obtained with a Zeiss Axiovert 200 M inverted fluorescent microscope (Zeiss, Germany) at 400 × magnification.
Results
TF induction and permeability in endothelial cells by IL-1β in vitro
TF induction in ECs was assessed under basal culture conditions using a single-stage coagulation assay after a 2-hour exposure to various concentrations of IL-1β. At concentrations of 100 pg/mL and higher, TF activity was markedly increased (Figure 1a). There was little TF activity measured at concentrations of IL-1β below 100 pg/mL and no further increase in TF activity was observed above 1 ng/ml. The effects of IL-1β on in vitro vascular permeability were assessed using a two-compartment model as described. There was only a slight increase in permeability above background (PBS control) after a 2-hour exposure to 0.1 or 1.0 ng/mL of IL-1β under basal culture conditions and exposure of ECs to factor VIII-deficient plasma alone had no effect on permeability (Figure 1b). However, at IL-1 doses that resulted in increased TF activity, IL-1 β caused a marked increase in permeability in the presence of factor VIII-deficient plasma (Figure 1b). Figure 2 shows a dose-dependent increase in EC permeability after a 2-hour incubation with IL-1β in the presence of factor VIII-deficient plasma. Permeability increased after only a 15-minute exposure to IL-1β compared to PBS control treated ECs and continued to increase with longer incubation times up to 90 minutes. However, incubation times of 2 hours or longer did not result in further increases in permeability (data not shown).
Figure 1 Increase in endothelial surface TF activity in response to IL-1β and dependence of IL-1β induced vascular permeability on tissue factor expression. Confluent EC monolayers were incubated for 3 hours with increasing concentrations of IL-1β and a single step coagulation assay performed to measure TF activity. A rapid increase of TF activity was observed at IL-1β of 100 pg/ml to rapidly plateau at 1 ng/ml or above (a). At concentrations of 0.1 ng and 1.0 ng/ ml, IL-1β strongly induced vascular permeability only in the presence of Factor VIII-deficient plasma and there was a modest increase in vascular permeability without plasma (b). All experiments were conducted in triplicate.
Figure 2 Dose dependent increase in EC monolayer permeability by 1L-1β. ECs were exposed to increasing concentrations of cytokine as described. All experiments were done in the presence of factor VIII deficient plasma.
To test whether TF induction on ECs was responsible for the alterations in permeability by IL-1 β, experiments using pre-treatment with a neutralizing anti-TF antibody were performed. Anti-TF antibody had no independent effect on permeability. However, IL-1β induced flux of Evans Blue-bound albumin across EC monolayers was completely abrogated in the presence of TF antibody (Figure 3). Together, these data demonstrated that IL-1β induces a rapid, dose dependent increase in permeability across EC monolayers via a TF dependent mechanism.
Figure 3 After stimulation of EC monolayers with 0.1 or 1.0 ng/ mL, IL-1β, cells were incubated with or without blocking TF-antibodies before addition of Factor VIII-deficient plasma. While permeability increased substantially in the absence of TF-blocking antibodies it could be complete abrogated by blocking TF receptors (c) demonstrating the necessity of endothelial TF activity and activation of the extrinsic coagulation cascade to induce permeability. Results were conducted in triplicate.
Activation of the extrinsic coagulation cascade by IL-1β leads to conformational changes of the EC cytoskeleton and loss of VE-cadherin complexes
To characterize the EC morphological changes that occur with IL-1β induced permeability we performed immunohistochemistry studies on ECs under conditions that resulted in increased permeability in-vitro. Figure 4 illustrates the effects of IL-1β on VE-cadherin, the major intercellular EC adhesion molecule, and F-actin cytoskeletal staining on EC monolayers. Untreated control confluent EC monolayers had homogenous VE-cadherin staining along cellular junctions with diffuse unorganized cytosolic F-actin and abundant localization adjacent to cell membranes. In the presence of factor VIII-deficient plasma alone or IL-1β alone, the EC monolayers remained intact with uniform well established intercellular tight junctions reflected by VE-cadherin and F-actin staining comparable to untreated cells. Incubation of EC monolayers with IL-1β and factor VIII-deficient plasma demonstrated markedly diminished VE-cadherin staining and associated F-actin cytoskeletal alignment particularly in areas of intercellular gap formation (Figure 4 arrows).
Figure 4 Effect on IL-1β on intercellular VE-cadherin complexes on ECs is shown. ECs were grown to confluence on fibronectin coated glass slides. Cells were stained for VE-cadherin and F-actin and visualized using an inverted fluorescence microscope. Untreated ECs form a functional monolayer without cellular gaps (control). Fluorescein labeled VE-cadherin antibodies demarcated individual cell borders whereas a Texas Red F-actin antibodies illustrate a relaxed cytoskeleton. Factor VIII-deficient plasma or IL-1 treatment alone showed similar staining patterns as untreated control cells. However, when used in combination there were large intracellular gaps seen (white arrows) in association with downregulation of VE-cadherin and the F-actin cytoskeletal filaments appear contracted and aligned in a spindle fashion.
Discussion
IL-1β exerts many of its proinflammatory effects through modulation of vascular permeability. Our data indicate that induction of tissue factor on the surface of ECs plays a central role in the regulation of vascular permeability by IL-1. Exposed endothelial surface TF, in the presence of plasma, forms complexes with coagulation factor VIIa and activates the extrinsic coagulation cascade. Local thrombin production most likely results in EC intercellular gap formation associated with decreased VE-cadherin and contraction of cytosolic F-actin fibers. We noted that in the absence of coagulation factors, IL-1β induces negligible permeability and that in the presence of factor VIII-deficient plasma, neutralizing TF antibodies completely abrogated IL-1 induced permeability. Taken together, these data indicate that IL-1 induction of TF on ECs represents a common initiating cellular event that results in both a procoagulant and permeable phenotype. Production of IL-1 at sites of acute or chronic inflammation may result in local alterations in vascular permeability mediated through the TF pathway and, to that end, modulation of TF activity may be a useful strategy to control the host inflammatory response (Figure 5).
Figure 5 Schematic of proposed mechanism of IL-1 induced vascular permeability. Local IL-1 production results in tissue factor upregulation on EC's. In the presence of extrinsic clotting cascade factors local thrombin production activates the EC thrombin receptor and results in VE-cadherin disassembly and contraction of cellular F-actin elements. The resulting intercellular gap formation allows efflux of plasma, proteins, and inflammatory cells into the interstitium.
We have previously shown that TNF shares these properties in an in vitro system highlighting the know widely overlapping biological activities of these two cytokines [24,25] particularly on endothelial tissue [26]. The acute loss of VE-cadherin complexes, contraction of F-actin cytoskeletal fibers, and the formation of large intercellular gaps due to TNF or IL-1 indicates that these morphological alterations are secondary to physiological EC responses rather than non-specific cell injury. EC monolayer integrity in this model is restored after the inflammatory stimuli is removed [27]. The effects are only observed in the presence of plasma, and hence, with activation of the extrinsic coagulation cascade and thrombin production.
It is interesting that the selective procoagulant and permeability effects of TNF, a cytokine with broadly overlapping biological activities with IL-1, on tumor neovasculature have been applied in the clinical treatment of cancer. When high dose TNF is administered via an isolated organ perfusion there is increased local permeability followed by selective obliteration of the neovasculature [28]. These selective effects on tumor neovasculature are thought to be important in augmenting delivery of chemotherapy, such as melphalan, to the tumor bed [29]. Our data would suggest that IL-1 might be a suitable alternative to TNF in this capacity because of the similar vascular effects and its limited toxicity compared to the other.
In summary, these experiments demonstrate that IL-1 mediates potent permeability effects on endothelial tissue that is mediated through EC surface expression of TF. To that end, TF activity may represent a novel target to ameliorate pathological inflammatory conditions mediated by this cytokine.
==== Refs
Zentella A Manogue K Cerami A The role of cachectin/TNF and other cytokines in sepsis. Bacterial endotoxins: cytokine mediators and new therapies for sepsis 1991 New York: Wiley-Liss, Inc 9 24 1924435
Brandtzaeg P Waage A Mollnes TE Oktedalen O Kierulf P Severe human septic shock involves more than tumor necrosis factor. Bacterial Endotoxins: cytokine mediators and new therapies for sepsis 1991 New York: Wiley-Liss, Inc 25 42 1924430
Ghivizzani SC Kang R Georgescu HI Lechman ER Jaffurs D Engle JM Watkins SC Tindal MH Suchanek MK McKenzie LR Evans CH Robbins PD Constitutive intra-articular expression of human IL-1 beta following gene transfer to rabbit synovium produces all major pathologies of human rheumatoid arthritis J Immunol 1997 159 3604 12 9317160
Pettipher ER Higgs GA Henderson B Interleukin 1 induces leukocyte infiltration and cartilage proteoglycan degradation in the synovial joint Proc Natl Acad Sci USA 1986 83 8749 53 3490671
Mulcare RJ Solis A Fortner JG Isolation and perfusion of the liver for cancer chemotherapy J Surg Res 1973 15 87 95 4725220 10.1016/0022-4804(73)90147-9
Gershenwald JE Fong Y Fahey TJI Calvano SE Chgizzonite R Kilian PL Lowry SF Moldawer LL Interleukin 1 receptor blockade attenuates the host inflammatory response Proc Natl Acad Sci USA 1990 87 4966 70 1695006
Fischer E Marano MA van Zee KJ Rock CS Hawes AS Thompson WA DeForge L Kenney JS Remick DG Bloedow DC Thompson RC Lowry SF Moldawer LL Interleukin-1 receptor blockade improves survival and hemodynamic performance in Escherichia coli septic shock, but fails to alter host responses to sublethal endotoxemia J Clin Invest 1992 89 1551 7 1533231
Cominelli F Nast CC Clark BD Schindler R Lierena R Eysselein VE Thompson RC Dinarello CA Interleukin-1 (IL-1) gene expression, synthesis, and effect of specific IL-1 receptor blockade in rabbit immune complex colitis J Clin Invest 1990 86 972 80 2168444
Alexander HR Doherty GM Venzon DJ Merino MJ Fraker DL Norton JA Recombinant interleukin-1 receptor antagonist (IL-1ra): Effective therapy against gram-negative sepsis in rats Surgery 1992 112 188 94 1386478
Jiang Y Genant HK Watt I Cobby M Bresnihan B Aitchison R McCabe D A multicenter, double-blind, dose-ranging, randomized, placebo-controlled study of recombinant human interleukin-1 receptor antagonist in patients with rheumatoid arthritis: radiologic progression and correlation of Genant and Larsen scores Arthritis Rheum 2000 43 1001 9 10817552 10.1002/1529-0131(200005)43:5<1001::AID-ANR7>3.0.CO;2-P
Bevilacqua MP Pober JS Majeau GR Cotran RS Gimbrone MA Jr Interleukin 1 (IL-1) induces biosynthesis and cell surface expression of procoagulant activity in human vascular endothelial cells J Exp Med 1984 160 618 23 6332168 10.1084/jem.160.2.618
Dias S Choy M Alitalo K Rafii S Vascular endothelial growth factor (VEGF)-C signaling through FLT-4 (VEGFR-3) mediates leukemic cell proliferation, survival, and resistance to chemotherapy Blood 2002 99 2179 84 11877295 10.1182/blood.V99.6.2179
Gloor SM Weber A Adachi N Frei K Interleukin-1 modulates protein tyrosine phosphatase activity and permeability of brain endothelial cells Biochem Biophys Res Commun 1997 239 804 9 9367850 10.1006/bbrc.1997.7557
Lebovic DI Bentzien F Chao VA Garrett EN Meng YG Taylor RN Induction of an angiogenic phenotype in endometriotic stromal cell cultures by interleukin-1beta Mol Hum Reprod 2000 6 269 75 10694276 10.1093/molehr/6.3.269
Carmeliet P Mechanisms of angiogenesis and arteriogenesis Nature Med 2000 6 389 95 10742145 10.1038/74651
Osnes LT Westvik AB Joo GB Okkenhaug C Kierulf P Inhibition of IL-1 induced tissue factor (TF) synthesis and procoagulant activity (PCA) in purified human monocytes by IL-4, IL-10 and IL-13 Cytokine 1996 8 822 7 9047078 10.1006/cyto.1996.0110
Schwager I Jungi TW Effect of human recombinant cytokines on the induction of macrophage procoagulant activity Blood 1994 83 152 60 8274733
Didier N Romero IA Creminon C Wijkhuisen A Grassi J Mabondzo A Secretion of interleukin-1beta by astrocytes mediates endothelin-1 and tumour necrosis factor-alpha effects on human brain microvascular endothelial cell permeability J Neurochem 2003 86 246 54 12807444 10.1046/j.1471-4159.2003.01829.x
Nooteboom A van der Linden CJ Hendriks T Tumor necrosis factor-alpha and interleukin-1beta mediate endothelial permeability induced by lipopolysaccharide-stimulated whole blood Crit Care Med 2002 30 2063 8 12352042 10.1097/00003246-200209000-00019
Goldblum SE Yoneda K Cohen DA McClain CJ Provocation of pulmonary vascular endothelial injury in rabbits by human recombinant interleukin-1 beta Infect Immun 1988 56 2255 63 3261716
Schulman CI Wright JK Nwariaku F Sarosi G Turnage RH The effect of tumor necrosis factor-alpha on microvascular permeability in an isolated, perfused lung Shock 2002 18 75 81 12095138 10.1097/00024382-200207000-00014
Reynolds AM Holmes MD Scicchitano R Interleukin-1beta and tumour necrosis factor-alpha increase microvascular leakage in the guinea pig trachea Respirology 2002 7 23 8 11896897 10.1046/j.1440-1843.2002.00362.x
Okusawa S Gelfand JA Ikejima T Connolly RJ Dinarello CA Interleukin-1 induces a shock like state in rabbits J Clin Invest 1988 81 1162 72 3258319
Dejana E Bertocchi F Bortolami MC Regonesi A Tonta A Breviario F Giavazzi R Interleukin 1 promotes tumor cell adhesion to cultured human endothelial cells J Clin Invest 1988 82 1466 70 3262629
Sato N Goto T Haranka K Satomi N Nariuchi H Mano-Hirano Y Sawasaki Y Action of Tumor Necrosis Factor on Cultered Vascular Endothelial Cells: Morphologic Modulation, Growth Inhibiton, and Cytoxicity J Natl Cancer Inst 1986 76 1113 21 3458948
Zhao B Stavchansky SA Bowden RA Bowman PD Effect of interleukin-1beta and tumor necrosis factor-alpha on gene expression in human endothelial cells Am J Physiol Cell Physiol 2003 284 C1577 C1583 12734110
Friedl J Puhlmann M Bartlett DL Libutti SK Turner EN Gnant MF Alexander HR Induction of permeability across endothelial cell monolayers by tumor necrosis factor (TNF) occurs via a tissue factor-dependent mechanism: relationship between the procoagulant and permeability effects of TNF Blood 2002 100 1334 9 12149215
Alexander HR Feldman AL Steven A, Rosenberg MD Tumor necrosis factor: Basic principles and clinical application in systemic and regional cancer treatment Biologic Therapy of Cancer 2000 Philadelphia: Lippincott 174 93
Olieman AFT van Ginkel RJ Hoekstra HJ Mooyaart EL Molenaar WM Koops HS Angiographic response of locally advanced soft-tissue sarcoma following hyperthermic isolated limb perfusion with tumor necrosis factor Ann Surg Oncol 1997 4 64 9 8985519
|
16197553
|
PMC1276820
|
CC BY
|
2021-01-04 16:39:27
|
no
|
J Transl Med. 2005 Sep 30; 3:37
|
utf-8
|
J Transl Med
| 2,005 |
10.1186/1479-5876-3-37
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-651620738310.1186/1477-7819-3-65Case ReportMalignant schwannoma of the upper mediastinum originating from the vagus nerve Shoji Fumihiro [email protected] Riichiroh [email protected] Tatsuro [email protected] Hiroshi [email protected] Kenichi [email protected] Yukito [email protected] Department of Thoracic Oncology, Kyushu Cancer Center, 3-1-1, Notame, Minami-ku, Fukuoka 811-1395, Japan2 Department of Pathology, Kyushu Cancer Center, 3-1-1, Notame, Minami-ku, Fukuoka 811-1395, Japan2005 6 10 2005 3 65 65 3 6 2005 6 10 2005 Copyright © 2005 Shoji et al; licensee BioMed Central Ltd.2005Shoji et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Malignant schwannoma of the upper mediastinum originating from the vagus nerve is extremely rare.
Case presentation
A 46-year-old female was admitted for a left cervical mass which was associated with both hoarseness and Horner's syndrome. Chest computed tomography showed a mass extending from the left upper mediastinum to the left supraclavicular area. A fine needle aspiration cytological examination suggested primary lung cancer stage IIIB large cell carcinoma. After administering induction chemo-radiotherapy, a complete surgical resection was performed. The tumor was found to involve both the left vagus nerve and the left sympathetic nerve. Histological examination of the resected specimen revealed the tumor to be malignant schwannoma.
Conclusion
Despite incorrect preoperative diagnosis, the multimodality treatment administered in this case, including induction chemo-radiotherapy and surgery, proved to be effective.
==== Body
Background
According to a collected series of 2399 cases of mediastinal tumors reported in the literature [1], 496 cases (20.7%) were of neurogenic tumors, and most of them occurred in the posterior mediastinum. Neurogenic tumors can be divided into two groups depending on their origin: those that arise from the nerve sheath and those that arise from nerve cells. The majority of the tumors of nerve sheath origin in adults are either benign schwannomas or neurofibromas, and they usually arise from either an intercostal nerve or a sympathetic nerve. Intrathoracic schwannoma originating from the vagus nerve, is extremely rare.
Case presentation
A 46-year-old female with symptoms of hoarseness and Horner's syndrome presented with a left cervical mass that was diagnosed to be undifferentiated carcinoma based on the findings of aspiration cytology (Figure 1). The patient's chest computed tomography (CT) findings showed a mass measuring 5.0 cm in size spreading from the left upper mediastinum to the left supraclavicular area, which pressed against both the trachea and esophagus and it seemed to involve the left common carotid artery (Figure 2). Based on these findings, and cytology findings a clinical diagnosis of stage IIIB (T4N3M0) non-small cell lung cancer (NSCLC) originating from the apex of the left lung involving both the mediastinum and the supraclavicular lymph nodes was made [2]. The patient received concurrent chemo-radiotherapy (cisplatin 80 mg/m2 for days 8 and 36 + UFT 400 mg/m2, both on days 1–14 and on days 29–42 plus radiotherapy, 2 Gy/day on days 1–20 for a total of 40 Gy) [3]. After this treatment regimen, the tumor size decreased by 35.0%. Thereafter, the patient underwent a surgical resection though a median sternotomy with a combined resection of the left clavicle. During the operation, an encapsulated tumor was detected in the mediastinum. Although the tumor was easily ablated from the left common carotid artery, it involved both the left vagus and the sympathetic nerves. As a result, both nerves had to be sacrificed in order to achieve a complete resection of the tumor. Grossly, the tumor was in continuity of the vagus nerve was whitish in color and oval shaped measuring 5 × 3 cm in diameter (Figure 3). Both cytological and histological examinations revealed 1) Continuity between the vagus nerve and tumor was seen, while the no continuity between the tumor and the sympathetic nerve was found. 2) The findings of aspiration cytology of the tumor diagnosing it to be undifferentiated carcinoma before the treatment included an atypical spindle cell. (Figure 1). 3) The predominantly tumor consisted of necrotic tissue and a few viable atypical spindle cells (Figure 4A) which were positive for S-100 protein (Figure 4B). As a result, the tumor was considered to arise from the left vagus nerve while invading the left sympathetic ganglion, and was therefore diagnosed it to be a malignant schwannoma. At present, the patient has survived for about 2 years since operation without any recurrence.
Figure 1 Aspiration cytology of the tumor before treatment showing scattered atypical spindle cells (arrows) (Giemsa staining, high power view x400).
Figure 2 Chest CT showed a mass, measuring 5.0 cm in size in the upper mediastinum.
Figure 3 Macroscopic findings of the tumor. The tumor showing the continuity of the vagus nerve (arrow head) was whitish in color and oval shaped while measuring 5 × 3 cm in diameter.
Discussion
Tumors of vagus nerve origin are observed in about 2% of all neurogenic tumors of the mediastinum [4], however, no instace of malignant schwannoma was reported in this review. To our knowledge, only a few such cases have been previously reported [5-7]. As a result, malignant schwannoma originating from the vagus nerve is therefore considered to be extremely rare.
Malignant peripheral nerve sheath tumors (MPNST) including malignant schwannoma are the malignant variants of schwannomas and neurofibromas. Although the 5-year survival rates have been reported to be up to 75 % in MPNST's patients, MPNST often advance locally and can also occasionally metastasize to the lung or other organs [8]. Therefore, in addition to a complete surgical resection, adjuvant therapy is usually advocated. However, in an adjuvant setting, the efficacy of chemotherapy or radiotherapy appears to provide little additional benefit [9,10]. We previously reported concurrent chemo-radiotherapy with UFT plus cisplatin as an induction treatment followed by a surgical resection for patients with marginally resectable stage IIIB NSCLC to be both a feasible and promising treatment [3]. Since we preoperatively considered the disease to be marginally resectable stage IIIB NSCLC, we performed concurrent induction chemo-radiotherapy followed by surgery. As a result, this multimodality treatment proved to be effective and the patient is now doing well without any recurrence.
We initially misdiagnosed this patient's disease to be non-small cell lung cancer. The reason for this was partly due to the cytological findings which indicated undifferentiated carcinoma. In general, an exact diagnosis cannot always be made based on the findings of aspiration cytology alone. The second reason for a misdiagnosis in this case was due to the patient's symptoms which included hoarseness and Horner's syndrome. Apical lung cancer involving both the vagus and the sympathetic nerve is occasionally observed. However, to the best of our knowledge, the present case is considered to be the first case demonstrating malignant schwannoma of the vagus nerve involving the sympathetic nerve.
Conclusion
Malignant schwannoma of the upper mediastinum arising from the vagus nerve is rare. The multimodality treatment administered in this case, including induction chemo-radiotherapy and surgery, proved to be effective.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FS: conceived of the study, participated in its design and coordination and drafted the manuscript.
RM and TO: carried out the literature search and helped in drafting the manuscript.
HW: participated in the design of the study and helped in drafting the manuscript.
KN: performed histological examination and provided photographs.
YI: shaped the idea for the study, coordinated the study and edited the manuscript.
All authors read and approved the final manuscript.
Figure 4 Histological findings of the tumor. 4A) The tumor was found to consist of a few of viable atypical spindle cells with hyperchromatic nuclei (Hematoxilin-eosin staining x200) 4B). Tumor cells were positive for S-100 protein (immunohistochemical staining, original magnification: X200).
Acknowledgements
We thank Mr. Brian Quinn for critical comments on the manuscript.
Written consent was obtained from the patient for the publication of this case.
==== Refs
Davis RD Oldham HN Sabiston DC Primary cysts and neoplasms of the mediastinum: recent changes in clinical presentation, methods of diagnosis, management, and results Ann Thorac Surg 1987 44 229 237 2820323
Travis WD Colby TV Corrin B World Health Organization International Histological Classification of Tumors. Histological typing of Lung and Pleural Tumors, 3 rd ed 1999 Berlin: Springer Verlag
Ichinose Y Fukuyama Y Asoh H Ushijima C Okamoto T Ikeda J Okamoto J Sakai M Induction chemoradiotherapy and surgical resection for selected stage IIIB non-small cell lung cancer Ann Thorac Surg 2003 76 1810 1814 14667588 10.1016/S0003-4975(03)01075-0
Sugio K Inoue T Inoue K Tateishi M Ishida T Sugimachi K Neurogenic tumors of the mediastinum originated from the vagus nerve Eur J Surg Oncol 1995 21 214 216 7720904 10.1016/S0748-7983(95)90798-X
Maebeya S Miyoshi S Fujiwara K Sekii H Suzuka T Yoshimasu T Naito Y Nishino E Malignant schwannoma of the intrathoracic vagus nerve:report of a case Surg Today 1993 23 1078 1080 8118122 10.1007/BF00309097
Yano T Hara N Ichinose Y Maeda K Yokoyama H Ohta M An intrathoracic vagus nerve schwannoma invading the trachea Surg Today 1993 23 1113 1115 8118129 10.1007/BF00309105
Suzuki H Yamaguchi Y Kimura H Shiba M Baba M Hiroshima K Malignant mediastinal schwannoma associated with von Recklinghausen's disease J Jpn Assn Thorac Surg 1996 44 864 868
Marchevsky AM Mediastinal tumors of peripheral nerve system origin Semin Diagn Pathol 1999 16 65 78 10355655
Lodding P Kindblom LG Angervall L Epithelioid malignant schwannoma. A study of 14 cases Virchows Arch 1986 409 A 433 451
Fukai I Masaoka A Yamakawa Y Niwa H Eimoto T Mediastinal malignant epithelioid schwannoma Chest 1995 108 574 575 7634904
|
16207383
|
PMC1276821
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Oct 6; 3:65
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-65
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-661621267110.1186/1477-7819-3-66Case ReportRetroperitoneal inflammatory myofibroblastic tumor Attili Suresh VS [email protected] C Rama [email protected] Dadhich K [email protected] Poonamalle P [email protected] Clementeena [email protected] G [email protected] Department of Medical Oncology, Kidwai Memorial Institute of Oncology, Bangalore, India2 Department of SurgicalOncology, Kidwai Memorial Institute of Oncology, Bangalore, India3 Department of Pathology, Kidwai Memorial Institute of Oncology, Bangalore, India2005 8 10 2005 3 66 66 12 7 2005 8 10 2005 Copyright © 2005 Attili et al; licensee BioMed Central Ltd.2005Attili et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Inflammatory myofibroblastic tumor (IMT) is a neoplasm of unknown etiology occurring at various sites. By definition, it is composed of spindle cells (myofibroblasts) with variable inflammatory component, hence the name is IMT.
Case presentation
The present case is of a 46 years old woman presented with a history of flank pain, abdominal mass and intermittent hematuria for last 6 months. The initial diagnosis was kept as renal cell carcinoma. Finally, it turned out to be a case of retroperitoneal IMT. The patient was managed by complete surgical resection of the tumor.
Conclusion
IMT is a rare neoplasm of uncertain biological potential. Complete surgical resection remains the mainstay of the treatment.
==== Body
Introduction
Inflammatory myofibroblastic tumor (IMT) is a relatively rare neoplasm. The outlook of this disease has changed with time from a benign reactive process to a malignant neoplasm, based on the multiple case reports demonstrating recurrent and constant clonal genetic alterations [1-5]. There are three main histological patterns: nodular fasciitis-like, fibrous histiocytoma-like, and desmoid or scar tissue-type. Though morphologically similar, they encompass a spectrum of entities with varied etiology, ranging from reactive/regenerative proliferations to low-grade neoplasms with a risk of local recurrence, and metastatic potential [3]. The commonest site of IMT is lung with a few case reports from extra pulmonary sites [6]. In the genitourinary tract, it most commonly occurs in the bladder. However it rarely originates in the kidney, renal pelvis, and ureter [4]. In the English literature only six cases of retroperitoneal IMT were reported [6-8]
Case presentation
A 46-years-old woman presented with history of flank pain, abdominal lump and intermittent hematuria of 6 months duration. She was diagnosed as renal cell carcinoma (RCC) elsewhere and referred. The investigations at our hospital revealed normal hematological and biochemical parameters. The urine microscopy showed deformed RBC. The computerized tomography (CT) scan of the abdomen showed a large irregular well defined heterogeneous lesion occupying the left hypochondrium, lumbar and supra umbilical regions, measuring approximately 16 × 13.6 × 12.1 cm, and showed presence of predominant cystic portions with no obvious calcifications (figure 1 &2). The mass was contiguously involving the upper half of the kidney. Fine needle aspiration cytology (FNAC) of the mass was suggestive of sarcomatoid renal cell carcinoma. The metastatic work-up including bone scan, and CT scan of the thorax were essentially normal. With a provisional diagnosis of renal cell carcinoma surgery was planned. Intra operatively a huge mass was observed in the retroperitoneal region arising from the left kidney and attempts to separate the mass from the kidney were unsuccessful. Hence, left nephrectomy with adrenalectomy was performed with complete excision of the tumor mass. Histopathology showed large encapsulated mass 16 × 17 × 12 cm attached to kidney on its posterior aspect. The tumor showed a predominantly spindle cell pattern consisting of cellular fascicles showing mild atypia, with a diffuse sprinkling of lymphocytes and plasma cells and a few lymphoid aggregates. Only an occasional mitosis was identified in these areas. Hypocellular areas with sparse cells and a collagenous stroma were also present. One area of the tumor showed features of malignant transformation, with sheets of bi and multinucleated and multilobed cells in a background of atypical spindle cells with interspersed inflammatory cells (figure 3). Mitoses in this area were 3–4/high power field and focal necrosis was present. Immunohistochemisty for desmin, smooth muscle actin (SMA), epithelial membrane antigen (EMA), cytokeratin (CK), CD68, HMB45 and ALK were performed. Desmin showed diffuse positivity, while SMA, EMA and CD68 showed focal positivity. CK, HMB45 and ALK were negative. A diagnosis of inflammatory myofibroblastic tumor with malignant transformation was made.
Figure 1 CT abdomen of the patient showing large irregular heterogeneous lesion measuring approximately 16 × 13.6 × 12.1 cm with predominant cystic portions without calcifications.
Figure 2 contrast enhanced CT abdomen of the patient showing the same lesion as in fig. 1 which is non enhancing.
Figure 3 Spindle cells with interspersed lymphocytes and plasma cells. Inset shows pleomorphic, multilobed tumor cells (Hematoxylin and Eosin × 400).
Discussion
IMT is a relatively rare neoplasm. It was previously referred as plasma cell granuloma, or inflammatory pseudo tumor (IPT). It has long been debated regarding the origin of IMT whether it was truly neoplastic or a post inflammatory process. The proposed etiologies included Epstein Barr virus (EBV), Human herpes virus (HHV8), and over expression of interleukin IL-6. Though other diseases like Kaposi's sarcoma and Castleman's diseases also have similar etiologies, molecular transcription form of open reading frame (ORF) -16, K13, 72 expressed in IMT are not expressed in the aforesaid diseases [9]. Moreover the recent research suggest that IMT is probably a neoplasm rather than a post-inflammatory process because of cytogenetic clonality, recurrent involvement of chromosomal region 2p23, occasional aggressive local behavior and metastasis of the tumor [1,2,4,5].
It is more common in children without sex predilection, though extra pulmonary forms are more common in adult females. The patients often present with fever of unknown origin and other vague nonspecific symptoms. Usually it has a benign course and in most of the cases it is a slow growing, locally confined tumor with less metastatic potential. However, there are some predictors for aggressive behavior and metastatic potential of IMT which include presence of ganglion like cells, cellular atypia, aneuploidy, and p53 over expression [11,12].
The tumor commonly occurs in lung. Extra pulmonary IMT are rare. In the largest series of 84 cases of extra pulmonary IMT, only four retroperitoneal IMT were reported [6] The medline search revealed only six cases of exclusive retroperitoneal IMT (excluding renal and pancreatic IMT). All the cases had flank pain as in our case. However hematuria which was observed in our case was found only in one case [8]. Other common features like fever, weight loss were not observed in our case. We initially approached the case as renal cell carcinoma due to the presence of classical triad of pain, lump and hematuria along with typical CT scan findings, and to our surprise the final diagnosis came as IMT. It would be possible on frozen section to differentiate this tumor from renal cell carcinoma. However the area showing malignant transformation could be confused with a sarcomatoid renal cell carcinoma. It is also difficult on frozen section to differentiate this tumor from an inflammatory process and this difficulty would also apply to assessment of margins of resection. So we feel that the total excision was more appropriate as the tumor is known for more local recurrences than distant metastasis. Preoperative or intra operative FNAC would be adequate to differentiate this lesion from renal cell carcinoma. However if the pleomorphic area is sampled it could lead to confusion with a sarcomatoid RCC. Therefore, avoiding of this situation is difficult and whenever the suggestion is IMT, a complete excision is advisable rather than going for organ preserving surgery.
Though IMT will not make its place in the common differential diagnosis of retroperitoneal masses, it should be kept in mind as one of the possibilities. Typically the IMT is characterized by the expression of vimentin, smooth muscle actin, and cytokeratins, corresponded to that of myofibroblasts along with other inflammatory markers [10]. In the present case, the IHC is positive for EMA and desmin, which are smooth muscle markers and CD-68 is an inflammatory marker, suggesting the tumor to be IMT. Little data is available in the management of IMT owing to its rarity. Surgery remains the main stay of treatment. Though radiotherapy [12], immunosuppression and chemotherapy (cisplatin, doxorubicin & methotrexate) [13] have been tried as an adjunct to surgery, however, no definitive benefit was demonstrated by any of these modalities [14]. In the present case as the tumor was completely removed and the metastasis is rare in IMT, the patient was offered no adjuvant therapy. The patient remained asymptomatic even after 1 year of follow-up without any evidence of disease. We expect long term survival in this case, as all the reported cases of retroperitoneal IMT had long disease-free survival [8].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SVSA have been involved in drafting the manuscript & revising it critically for important intellectual content
RCC acquisition of data, or analysis and interpretation of data
HKD revising it critically for important intellectual content
PPB revising it critically for important intellectual content and given final approval of the version to be published
CRR interpretation of data
GA have been involved in drafting the manuscript & revising it critically for important intellectual content
Acknowledgements
Written consent was obtained from the patient for publication of this case report
==== Refs
Pungpapong S Geiger XJ Raimondo M Inflammatory myofibroblastic tumor presenting as a pancreatic mass: a case report and review of the literature JOP 2004 5 360 367 15365204
Coffin CM Dehner LP Meis-Kindblom JM Inflammatory myofibroblastic tumor, inflammatory fibrosarcoma, and related lesions: an historical review with differential diagnostic considerations Semin Diagn Pathol 1998 15 102 110 9606802
Freeman A Geddes N Munson P Joseph J Ramani P Sandison A Fisher C Parkinson MC Anaplastic lymphoma kinase (ALK 1) staining and molecular analysis in inflammatory myofibroblastic tumors of the bladder: a preliminary clinicopathological study of nine cases and review of the literature Mod Pathol 2004 17 765 771 15105807 10.1038/modpathol.3800078
Kapusta LR Weiss MA Ramsay J Lopez-Beltran A Srigley JR Inflammatory myofibroblastic tumors of the kidney: a clinicopathologic and immunohistochemical study of 12 cases Am J Surg Pathol 2003 27 658 666 12717250 10.1097/00000478-200305000-00009
Biselli R Boldrini R Ferlini C Boglino C Inserra A Bosman C Myofibroblastic tumors: neoplasias with divergent behavior. Ultrastructural and flow cytometric analysis Pathol Res Pract 1999 195 619 632 10507082
Coffin CM Watterson J Priest JR Dehner LP Extrapulmonary inflammatory myofibroblastic tumor (inflammatory pseudotumor). A clinic pathologic and immuno-histochemical study of 84 cases Am J Surg Pathol 1995 19 859 872 7611533
Esmer-Sanchez D Rangel D Inflammatory pseudotumor of the retroperitoneum Rev Gastroenterol Mex 2002 67 97 99 12214342
Tambo M Kondo H Kitauchi T Hirayama A Cho M Fujimoto K Yoshida K Ozono S Hirao Y Yamada E Ichijima K A case of inflammatory myofibroblastic tumor of the retroperitoneum Hinyokika Kiyo 2003 49 273 276 12822456
Gomez-Roman JJ Sanchez-Velasco P Ocejo-Vinyals G Hernandez-Nieto E Leyva-Cobian F Val-Bernal JF Human herpesvirus-8 genes are expressed in pulmonary inflammatory myofibroblastic tumor (inflammatory pseudotumor) Am J Surg Pathol 2001 25 624 629 11342774 10.1097/00000478-200105000-00009
Sastre-Garau X Couturier J Derre J Aurias A Klijanienko J Lagace R Inflammatory myofibroblastic tumour (inflammatory pseudotumour) of the breast. Clinicopathological and genetic analysis of a case with evidence for clonality J Pathol 2002 196 97 102 11748648 10.1002/path.1004
Hussong JW Brown M Perkins SL Dehner LP Coffin CM Comparison of DNA ploidy, histological and immunohistochemical findings with clinical outcome in inflammatory myofibroblastic tumors Mod Pathol 1999 12 279 286 10102613
Imperato JP Folkman J Sagerman RH Cassady JR Treatment of plasma cell granuloma with radiation therapy: a report of two cases and a review of the literature Cancer 1986 57 2127 2129 3697912
Dishop MK Warner BW Dehner LP Kriss VM Greenwood MF Geil JD Moscow JA Successful treatment of inflammatory myofibroblastic tumor with malignant transformation by surgical resection and chemotherapy J Pediatr Hematol 2003 25 153 158 10.1097/00043426-200302000-00014
Hagenstad CT Kilpatrick SE Pettenati MJ Savage PD Inflammatory myofibroblastic tumor with bone marrow involvement. A case report and review of the literature Arch Pathol Lab Med 2003 127 865 877 12823044
|
16212671
|
PMC1276822
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Oct 8; 3:66
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-66
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-681622344510.1186/1477-7819-3-68ResearchExpression of Placenta growth factor (PlGF) in non-Small cell Lung cancer (NSCLC) and the clinical and prognostic significance Zhang Lijian [email protected] Jinfeng [email protected] Yang [email protected] Robert E [email protected] Wen G [email protected] Department of Surgery, Peking University School of Oncology, No. 52 Fu-cheng Road, Haidian District, Beijing, China2 Department of Cell Biology, Health Science Center, Peking University, No. 38 Xueyuan Rd, Hai Dian District, Beijing China3 Metastasis & Angiogenesis Research Group, Wales College of Medicine, Cardiff University, Cardiff, UK2005 13 10 2005 3 68 68 7 7 2005 13 10 2005 Copyright © 2005 Zhang et al; licensee BioMed Central Ltd.2005Zhang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Placenta growth factor (PlGF) is a member of the vascular endothelial growth factor (VEGF) family. Over-expression of PlGF is known to be associated with pathological angiogenesis. This study examined PlGF expression at protein and message levels in non-small cell lung cancer (NSCLC), in which no reports on the significance of PlGF expression is available to date.
Patients and methods
We used immunohistochemistry to assess the PlGF protein and correlated PlGF with microvessel density (MVD), as well as clinical outcome in patients with NSCLC tumours (n = 91). In addition, we applied a real time quantitative PCR assay using SYBR Green chemistry to measure PlGF mRNA in normal lung tissues and NSCLC tumours.
Results
PlGF was positively stained mainly in cytoplasm of lung cancer cells. High level staining of PlGF was found in 38.5% NSCLC patients. A high level of MVD in NSCLC was found in 42.9% of cases. Tumours with high level and low level PlGF staining had a significantly different MVD (26.69 vs. 20.79, respectively, p = 0.003). Using both univariate and multivariate analyses, PlGF was found to be an independent prognostic factor. Real time PCR analysis revealed that PlGF mRNA was higher in the cancer tissue than normal tissue (0.95 ± 0.19 vs. 0.57 ± 0.24; p < 0.005) and that PlGF mRNA was significant higher in III-IV stage patients than in I-II stage patients (1.03 ± 0.20 vs. 0.80 ± 0.17; p = 0.011).
Conclusion
PlGF expression is significantly more in NSCLC tumour tissues than in matched normal tissues. It has a significant positive association with MVD and is an independent factor for NSCLC patients. PlGF may have a pivotal role in NSCLC development and disease progression.
==== Body
Introduction
Angiogenesis is essential for a solid tumour to grow beyond 1–3 mm in diameter [1]. It also is a significant predictive factor for prognosis in patients with solid tumours [2,3]. Amongst the numerous angiogenic factors, VEGF is the most powerful and most extensively studied. VEGF belongs to a protein family, within which Placental growth factor (PlGF) is a member (other members include VEGF-B, -C and D). PlGF is a secreted, disulfide-linked dimeric glycoprotein originally cloned from a cDNA library of term placenta [4]. PlGF shares 53% of similarity in its overall amino acid (aa) residues with VEGF. The biological functions of VEGF and PlGF are similar, including stimulation of the growth of vascular endothelial cells [5]. As a result of alternative splicing of the primary PlGF transcript, PlGF has at least three isoforms, PlGF-1 (PlGF149), PlGF-2 (PlGF170) and PlGF-3 (PlGF221) [4]. In cells co-expressing VEGF and PlGF mRNA, a heterodimeric VEGF/PlGF protein has been detected [6,7]. VEGF/PlGF heterodimer has been shown to promote capillary growth in vivo [7].
PlGF is known to specifically bind with Flt-1. VEGFR2/KDR/Flk-1 and VEGFR1/Flt-1 are the two main receptors of VEGF during the embryonic vascular development [8,9]. Flk-1 primarily mediates VEGF signal transduction and biological responses [10]. In addition to acting as the receptor for VEGF and PlGF, Flt-1 is a special receptor for VEGF-B. It has been shown that VEGF and PlGF can induce transcription factors FosB and c-Fos mRNA expression, indicating the possibility that these factors may play a role in the biological responses mediated by PlGF and Flt-1 [9]. The protein and message for PlGF can be detected in endothelial and epithelial cells and have been found in a few tumours [7,10].
To our knowledge, there has been no report on the significance of PlGF expression with clinical outcome of patients with lung cancer, including non-small cell lung cancer (NSCLC). In order to ascertain the clinical significance of PlGF expression in human non-small cell lung cancer, we analysed the expression pattern of PlGF using both immunohistochemical method and real time quantitative PCR and attempted to establish if a relationship existed between PlGF and MVD, and subsequently between PlGF and the predicted prognosis.
Patients and methods
Patients and samples
A total of 91 patients with non-small cell lung cancer, who attended Beijing Cancer Hospital from July 2000 to August 2003, were included. None of the patients received any neoadjuvant therapy prior to operation. Histological types of the lung cancer included squamous carcinoma, adenocarcinoma, large cell carcinoma, squamous adenocarcinoma and alveolar carcinoma, pathologically (table 1). No other previous or concomitant primary cancer was present. Clinico-pathological characteristics were defined according to the TNM criteria of the UICC (11) (table 1). Slides were reviewed and evaluated by two independent researchers. Clinico-pathologic factors, such as age, sex, histological types of tumours, tumour cell grade, TNM stage, vessel embolism, lymph node metastasis, were reviewed and stored in the patients' database. Patients were followed up from the day of operation to August 2004 as the end of follow-up. The follow-up intervals were calculated as survival intervals after surgery.
Table 1 The correlation between PlGF expression and the clinical pathological factors.
Variable Case(n and %) PlGF staining pattern P value (x2)
Low-staining (n =) High-staining (n =)
Sex
Male 63 (69.2%) 39 24 1.000
Female 28 (30.8%) 17 11
Age
≤60 43 (47.3%) 29 14 0.290
>60 48 (52.7%) 27 21
Histological type
Squamous carcinoma 42 (46.2%) 29 13 0.167
Adenocarcinoma 33 (36.3%) 20 13
Adenosquamous ca. 6 (6.6%) 3 3
Large cell carcinoma 2 (2.2%) 2 0
Carcinoid 2 (2.2%) 0 2
Alveolar carcinoma 6 (6.6%) 2 4
Grade of differentiation
Poor 11 (12.1%) 9 2 0.288
Moderate 48 (51.7%) 27 21
Well 32 (35.2%) 20 12
Tumour stage
T1 9 (9.9%) 7 2 0.643
T2 60 (65.9%) 37 23
T3 19 (20.9%) 10 9
T4 3 (3.3%) 2 1
Nodal status
N (-) 49 (53.8%) 31 18 0.829
N (+) 42 (46.2%) 25 17
Vessel cancer embolus
V (-) 67 (73.6%) 42 25 0.808
V (+) 24 (28.4%) 14 10
TNM stage
I 40 (43.9%) 28 12 0.173
II 20 (21.9%) 11 9
IIIa 28 (30.7%) 17 11
IIIb 1 (1.1%) 0 1
IV 2 (2.2%) 0 2
A separate collection of tissue samples from 21 primary non-small cell lung cancers were used for mRNA based analysis. These tumours were resected surgically from patients at the Clinical Oncology School of Peking University from 2002 to 2003 and were saved in the Tissue Bank of Peking University Oncology School. The patients consisted of 13 men and 8 women, with a mean age of 56.2 ± 6.4 years. The histological type of lung cancer was classified based according to the World Health Organization classification [12]. Tumour staging was performed according to the TNM staging criteria of the UICC [11]. The tumour specimens included 10 squamous cell carcinomas, 8 adenocarcinomas, and 3 undifferentiated carcinomas. Tumour staging was I in 3 cases, II in 7 cases, IIIA or IIIB in 10 cases, and IV in 1 case. Immediately after surgery, tumour samples and surrounding normal lung tissues (more than 5 cm away from the tumour margin) were placed in liquid nitrogen and stored frozen at -80°C for RNA extraction and the RT-PCR.
Materials
The goat polyclonal antibody of PlGF and a mouse monoclonal antibody of β-actin was purchased from Santa Cruz Biotechnology (Santa Cruz, California, USA). The mouse monoclonal antibody of CD31 was purchased from Beijing Zhongshan Biotechnology Co. Ltd (Beijing, China). The biotin conjugated anti-goat IgG, anti-mouse IgG antibodies were purchased from Sigma (Poole, Dorset, England, UK). The Horse Radish Peroxidase (HRP) conjugated anti-goat IgG, anti-mouse IgG antibodies were obtained from Sigma (Poole, Dorset, England, UK). The Target Retrieval Solution was purchased from DAKO Corp. (Beijing, China). RNA extraction and reverse transcription kits and PCR mix were purchased from Bio-Rad Corp. (Beijing, China). Primers were synthesized by BioAsia Corporation (Shanghai, China). PCR reaction was carried out using the iCycler iQ™ system (Bio-Rad). The working stock solution of SYBR Green is 1:100 (Bio-Rad).
PlGF staining and microvessel counting
The paraffin-embedded tissue sections of 91 patients were cut at 4 microns and mounted on polylysin-coated glass slides for immunohistochemistry. Briefly, deparaffinized sections were heated (60°C 1 hour). Antigen retrieval was performed by heating the samples without boiling in Target Retrieval Solution (DAKO Corp.), pH 6.70 (200 ml) in a microwave oven for 10 min. After endogenous peroxidase was blocked with 3% hydrogen peroxide in the section, each section was incubated with non-immunised horse serum (Sigma) for 15 minutes, in order to block the non-specific antigen site.
The immunohistochemical staining procedure was performed according to the protocol of the DAKO Corp. The primary anti-PlGF antibodies were used at a dilution 1:100 from the stock. The primary CD31 antibody was used at working dilution of 1:100. The specificity of anti-PlGF antibody was documented elsewhere [13]. Following incubation at 4°C overnight, the sections were extensively washed and then incubated with link antibodies (Sigma). Following more washing, bound antibodies were linked to avidin-biotin-peroxidase according to manufacturer's instruction (Dako Corp.). The staining was completed by developing colour using the DAB (diaminobenzidine tetrahydrochloride) solution for 5 min. The slides were counterstained with Mayers Haematoxylin Blue in 0.3% ammonia. For negative controls, sections were stained in the same manner, except that the primary antibody was absent from the solution.
PlGF staining in lung cancer cells was independently assessed by two observers using a modification of the system of grading the relative intensity of immunoreactivity for the respective antibodies [14]. PlGF immunohistochemical staining of a tissue sample was graded as either low level expression or high level. High-level staining represented uniformly intense immunoreactivity; low level staining represented patchy and weak or negative immunoreactivity.
MVD was evaluated as previously described [15]. After screening the areas with intense neovasularized spots at low power field (×100), microvessels in the area with the highest number of discrete microvessels were counted in a ×400 field. Three separate intense neovascularized areas were assessed for each spot, and the mean was calculated as MVD of each tumour evaluated. The MVD level were graded as low level with MVD number lesser than 26, while the high level with MVD number over than 26. This was based on the pilot analysis of the microvessel density according to pathological grade, in that 26 yielded a clear division between the groups.
Generation of cDNA from NSCLC tissue and normal tissue and RT-PCR
RNA was extracted from tumour and the normal surrounding tissues in RNA extraction buffer according to the manufacturer's protocol. The concentration of RNA was measured with a spectrophotometer. Reverse transcription was performed from 1 μg of total RNA using oligo dt primer according to the manufacture's instructions. Conventional PCR primers were designed using Primer 3 , to allow amplification of regions that have no overlap with other known genes and span at least one intron. Primers were synthesized by BioAsia Corporation (Shanghai, China): Primer sequences for PlGF were 5'ACGTGGAGCTGACGTTCTCT'3 and 5'CAGCAGGAGTCACTGAAGAG'3 and for GAPDH 5'AGGTCGGAGTCAACGGATTTG'3 and 5'GTGATGGCATGGACTGTGGT'3. Conventional PCR was performed using cDNA from tissues together with the PCR master mix using respective primers. The reaction conditions were: 95°C 5 min, 94°C for 1 min, 55°C for 45 s, 72°C for 30 s and a final extension phase at 72°C for 7 min for 40 cycles. The PCR products were separated on a 2% agarose gel and stained with 5 μl ethidium bromide prior to examination under UV light and photographs taken.
Generation of standard for real time PCR analysis
The PCR product from the above reaction was gel-excised and purified using gel purification kit (Tianwei Corp, Beijing). It was subsequently quantified on a gel with lambda molecular weight standards and in a spectrophotometer. The number of copies of target template was calculated. The DNA sample was serially diluted to yield a concentration range between 102 to 108 copies, which was subsequently used as the internal standard. This was finally prepared in elution buffer, aliquoted and stored at -80°C until use.
Real Time quantitative RT-PCR
The iCycler iQ™ system incorporates a gradient thermocycler and a 96-channel optical unit. SYBR Green as a DNA-dye can affinity dsDNA was used to detect the PCR product of PlGF and GAPDH. The melting point, optimal conditions and the specificity of the reaction were first determined using a standard procedure [17]. The working stock solution of SYBR Green is 1:100 (Bio-Rad). Quantitative PCR was carried out in 96-well plate with 10 pmol forward and reverse primers, and the working solution SYBR green, using a customer PCR master mix, with the following conditions: 95°C for 5 minutes, followed by 40 cycles at 95°C for 1 minute, 55°C for 45 seconds, 72°C for 30 second. Every assay included test cDNA samples, 10-fold serial dilutions of the standard qualification, and controls (no template), as we previously reported [16]. The copy number of each transcript was calculated as the relative copy number normalised by GAPDH copy number.
Statistical analysis
Patients were divided into two groups: those with high-level PlGF staining and those with low-level staining. Chi-square analysis was used to test the association of PlGF expression level with standard pathological variables. Clinical pathological parameters and PlGF expression status were correlated with survival time in both univariate and multivariate analyses. Variables included in univariate analysis were gender, TNM stage, grade of differentiation, vessel cancer embolus, lymph nodal status and use of postoperative adjuvant therapy, MVD and PlGF. Variables included in multivariate analysis were MVD, PlGF expression status, gender, TNM stage, grade of differentiation, vessel cancer embolus, lymphatic nodal status and use of postoperative adjuvant therapy. The log-rank test was used to test equality across categorical factors in univariate analysis, and the level of significance was set at p ≤ 0.05 based on two sided test. The multivariate analysis was performed using Cox proportional hazards model, and the level of significance was set at p ≤ 0.05 based on a two sided test.
Paired-samples analysis was used (Student's t-test) to determine whether the difference of PlGF mRNA expression level observed between matched cancer tissue and the normal tissue. Independent-samples analysis was used (Student's t-test) to determine whether the differences observed between PlGF expression level in NSCLC tissue and the clinico-pathological characteristics. Statistical tests were performed using the software SPSS 10.0 (SPSS Inc., Chicago).
Results
Immunohistochemical analysis and the localization of PlGF in lung cancer
Lung cancer cells stained positively for PlGF. As shown in figure 1A, 1B, adenocarcinoma cells showed strong and diffuse cytoplasmic staining of PlGF. Squamous cancer cells of the lung also displayed strong and positive cytoplasmic staining of PlGF as shown in figure 2A, 2B. Compared with normal tissues of the lung as shown in figure 1E, 1F and 2E, 2F, the negative staining in tumour tissues was shown in figure 1C, 1D and 2C, 2D.
Figure 1 Immunohistochemical staining for PlGF in adenocarcinoma of lung. A and B showed strong diffuse cytoplasmic staining of PlGF in adenocarcinoma of lung. C and D showed the negative staining of PlGF in adenocarinoma of lung. E and F showed the negative staining status of PlGF in normal alveolar (Original magnification is ×100).
Figure 2 Immunohistochemical staining for PlGF in squamous cell carcinoma of the lung. A and B showed strong diffuse cytoplasmic staining of PlGF in squamous carcinoma of the lung. C and D showed negative staining of PlGF in squamous carcinoma of lung. E and F showed the negative staining status of PlGF in alveolar. (Original magnification ×400)
PlGF and the clinical correlation
The immunohistochemical staining results of PlGF in NSCLC were shown in table 1. High level expression of PlGF in NSCLC was found in 35 (38.5%) cases. Among all the available information, no significant correlation was seen (p > 0.05).
Micro-vessel Density in lung cancer and its correlation with PlGF
The micro-vessel endothelial cells stained positively for CD31. Endothelial cells showed strong and diffuse cytoplasmic staining of CD31, The correlation between MVD and clinico-pathologic features in NSCLC were shown in table 2. Low level of MVD in NSCLC was found in 52 (57.1%) cases. Amongst the available clinical parameters, no clinical factors showed significant correlation with CD31 as shown in table 2 (p > 0.05).
Table 2 Microvessel Density counting vs. clinico-pathological features in the complete series (n = 91)
Variable Cases (n and %) MVD pattern P value (x2)
Low MVD (n =) High MVD (n =)
Sex
Male 63 (69.2%) 37 26 0.654
Female 28 (30.8%) 15 13
Age
≤60 43 (47.3%) 24 19 0.835
>60 48 (52.7%) 28 20
Histological type
Squamous carcinoma 42 (46.2%) 24 18 0.312
Adenocarcinoma 33 (36.3%) 20 13
Adenosquamous carcinoma 6 (6.6%) 2 4
Large cell carcinoma 2 (2.2%) 2 0
Carcinoid 2 (2.2%) 0 2
Alveolar carcinoma 6 (6.6%) 4 2
Grade of differentiation
Poor 11 (12.1%) 7 4 0.345
Moderate 48 (52.7%) 24 24
Well 32 (35.2%) 21 11
Tumour stage
T1 9 (9.9%) 5 4 0.988
T2 60 (65.9%) 34 26
T3 19 (20.9%) 11 8
T4 3 (3.3%) 2 1
Nodal status
N (-) 49 (53.8%) 29 20 0.678
N (+) 42 (46.2%) 23 19
Vessel cancer embolus
V (-) 67 (73.6%) 39 28 0.812
V (+) 24 (26.4%) 13 11
TNM stage
I 40 (44.0%) 27 13 0.389
II 20 (22.0%) 10 10
IIIa 28 (30.8%) 14 14
IIIb 1 (1.1%) 0 1
IV 2 (2.2%) 1 1
35 lung NSCLC tumours that had high-level expression of PlGF had a mean MVD at 26.69 and the standard deviation (SD) was 8.89. While in 56 NSCLS tumour which had low-level staining of PlGF, the mean MVD ± SD was 20.79 ± 8.82. The difference between the two groups was highly significant, p = 0.003.
Clinical outcome and the prognostic value of variables
Univariate analysis of the impact of histological types and PlGF expression status on prognosis was presented in table 3. Longer survival time was found to be significantly correlated with the following factors: low TNM stag (p = 0.0005), no lymph node metastasis (mean survival time, 41.04 months vs. 31.25 months, p = 0.0051), low level of MVD (mean survival time, 39.96 months vs 31.85 months, p = 0.0434), and PlGF low-level expression (mean survival time, 40.68 months vs 27.76 months, p = 0.0028).
Table 3 Potential prognostic factors using univariate analysis
Characteristics Patients (n =) Mean survival (mths) P Value*
Gender
Male 63 36.42 (31.77–41.08) 0.8860
Female 28 37.08 (30.73–43.44)
TNM stage
I 40 41.61 (37.35–45.87) 0.0005
II 20 38.44 (31.18–45.70)
IIIa 28 30.31 (22.76–37.86)
IIIb 1 6.00 (6.00-6.00)
IV 2 13.50 (3.11–23.89)
Grade of differentiation
Well 32 37.43 (31.17–43.69) 0.4984
Moderate 48 34.76 (29.86–39.67)
Poor 11 39.73 (28.34–51.12)
Nodal status
N (-) 49 41.04 (36.64–45.45) 0.0051
N (+) 42 31.25 (25.64–36.86)
Vessel cancer embolus
V (-) 67 37.50 (33.41–41.58)
V (+) 24 33.89 (25.75–42.03)
Postoperative adjuvant therapy
No 26 37.47 (29.70–45.24) 0.7225
Chemotherapy or Radiation 65 36.03 (31.57–40.49)
MVD
High MVD 39 31.85 (25.68–38.03) 0.0434
Low MVD 52 39.96 (35.48–44.44)
PlGF staining
Low expression 56 40.68 (36.70–44.66) 0.0028
Over expression 35 27.76 (21.17–34.34)
* Log-Rank Test
A multivariate prognostic analysis based on the Cox proportional hazard model was performed to test the independent value of each parameter predicting overall survival. Of the all the factors analysed (gender, tumour TNM stage, grade of differentiation, vascular emboli of cancer cells, nodal status and use of postoperative adjuvant therapy, MVD level, and PlGF), TNM stage and high-level PlGF were two independent prognostic factors, which were the best general prognostic indicators (p = 0.007 and p = 0.011, respectively) (table 4). Furthermore the relative risk of TNM stage and high-level PlGF expression were 1.311 and 2.738 respectively (95%CI, 1.076–1.597 vs 1.269–6.102).
Table 4 The prognostic value of PlGF staining evaluated on the basis of multivariate analysis in the COX model (Backward: Wald) for 91 NSCLC cases.
Variable Regression Coeff. (B) Standard error (SE) Wald value P value OR (EXP) 95% CI for OR
PlGF expression 1.024 0.401 6.532 0.011 2.738 (1.269–6.102)
TNM staging 0.271 0.101 7.201 0.007 1.311 (1.076–1.597)
Figure 3A shows the Kaplan-Meier survival curve, based on PlGF expression status. The cumulative survival time for patients with PlGF low-expression (n = 56) was significantly longer than the patients with high-level PlGF (n = 35) (p = 0.0028). Figure 3B shows the Kaplan-Meier survival curve for 91 NSCLC patients based on lymphatic node metastasis. The survival difference between lymph node negative patients (n = 49) and positive patients (n = 42) was significant (P = 0.0051). Figure 3C shows the Kaplan-Meier survival curve for 91 NSCLC patients based on MVD status. The survival difference between low level of MVD (n = 52) and high level MVD patients (n = 43) was significant (p = 0.0434).
Figure 3 Overall survival of NSCLC patients based on PlGF (left), nodal status (middle) and MVD (right). Left: PlGF and the overall survival. A, PlGF low expression tumours (n = 56). B, PlGF over expression tumours (n = 35). The survival curves are significantly separated and the patients with PlGF low expression have long survival time than those with over expression (p = 0.0028, log-rank test). Middle: Overall survival based on nodal status of NSCLC patients (n = 91). The difference between nodal negative patients (A, n = 49) and nodal positive patients (B, n = 42) is significant (p = 0.0051, log-rank test). Right: Overall survival based on MVD status of NSCLC patients (n = 91). The difference between low level of MVD (A, n = 52) and high level MVD patients (B, n = 43) is significant (p = 0.0434, log-rank test).
PlGF mRNA expression level in NSCLC and normal lung tissue assessed by real time RT-PCR
The amplification plot and the melting curve confirmed that the amplification products for all three molecules were specific (data not shown). We then employed the analysis tool for both PlGF and GAPDH. The PlGF mRNA (copy number of PlGF mRNA/copy number of GAPDH mRNA) is shown here as the relative copy number. PlGF mRNA expression and was detected in all 21 (100%) paired lung cancer and non-tumorous lung tissue samples by real time RT-PCR. The relative copy number for PlGF in 21 samples of lung cancer tissue ranged from 0.52 to 1.19, with a mean ± standard deviation (SD) of 0.95 ± 0.20, while the corresponding values in the matched non tumours lung tissue ranged from 0.21 to 1.00, with a mean ± SD of 0.57 ± 0.24. In 17 of 21 (81%) of patients, tumours displayed higher PlGF than the matched normal tissues. Only in 19% (4/21) of patients, healthy tissues had higher levels than tumour tissues. PlGF mRNA expression in lung cancer tissue was significantly higher than in the matched non tumorous lung tissue (95% CI: 0.2382~0.5321, p < 0.005, paired test).
Relationship between PlGF mRNA expression and clinicopathologic variables
Table 5 shows the relationship between PlGF mRNA expression level and the clinico-pathological features. PlGF mRNA was significantly higher in III-IV stage patients than in stage I-II patients (1.03 ± 0.20 vs. 0.80 ± 0.17; p = 0.011, independent t test). The difference between PlGF in different tumour size was not significant (p > 0.05). The relationship between PlGF mRNA expression and sex, histological type, lymph node status was otherwise not statistically significant.
Table 5 Relationship between relative levels of PlGF mRNA (PlGF/G3PDH) and clinico-pathologic characteristics
Variables No. PlGF mRNAa (mean ± SD) P value
Sex
Male 13 0.8760 ± 0.2468 0.261
Female 8 0.9881 ± 0.1461
Histologic type
Squamous 10 0.9706 ± 0.2153 0.307
Non squamous 11 0.8714 ± 0.2177
Stage
I–II 10 0.7979 ± 0.1744 0.011
III–IV 11 1.0284 ± 0.1977
Tumour status
T1 6 0.7939 ± 0.1656 0.096
T2–4 15 0.9686 ± 0.2195
Lymph node status
N0 8 0.8231 ± 0.1640 0.116
N1–3 13 0.9775 ± 0.2303
Tissue
Tumour 21 0.9187 ± 0.2171 <0.005*
Adjacent normal tissue 21 0.5652 ± 0.2406
aPlGF mRNA expression derived from real-time quantitative RT-PCR: PlGF/GAPDH.
*p value derived from paired t test; others derived from independent t test.
Discussion
PlGF is a member of the VEGF family and is known to be a powerful angiogenic factor, like its family members VEGF. Although PlGF has been studies in a number of clinical tumour types, little is known with regard to this factor in lung cancer, particularly non-small cell lung cancer (NSCLC). The current study investigated the expression of PlGF, at the protein level and the messenger RNA level and whether it had a bearing on clinical outcome in patients with NSCLC.
Firstly, the current study has demonstrated that PlGF can be found, at the protein level and mRNA level, in lung cancer cells. Furthermore, PlGF protein immunostaining is primarily seen in the cytoplasmic region of the cells, which is in accordance with literature reports. Stromal cells and endothelial cells displayed little staining. Secondly, PlGF is significantly linked to MVD. High levels of PlGF are significantly linked to high MVD.
The angiogenic role of PlGF is interesting to observe. PlGF expression is restricted to the placenta tissue in normal physiological condition [18] and has been indicated to play an important role in the angiogenic process. There is evidence to suggest that up-regulation of PlGF and VEGFR-1 expression can stimulate the response of vascular endothelial cells to VEGF and enhance angiogenesis in pathological disorders [19]. Although the precise mechanism by which PlGF regulates angiogenesis is unclear, some leads have been suggested. Up-regulation of PlGF can displace VEGF from Flt-1 and will make more VEGF to bind and activate Flk-1 [20]. PlGF activates Flt-1 and will lead to intermolecular transphosphorylation of Flk-1. This can enhance phosphorylation levels of Flk-1 on tyrosine residues. Furthermore, VEGF/PlGF heterodimer can activate and transmit angiogenic signals through the Flk-1/Flt-1 heterodimer receptor complex.
In the current study, a high level of PlGF was seen 38.46% of all the cases. The significant correlation between PlGF and MVD further indicates that PlGF expression level is associated with MVD, and potentially with angiogenesis in human lung cancer. We, and others, have previously reported that MVD was an influence factor to predict the prognosis of patients with non-small cell lung cancer Furthermore, both univariate and multivariate analyses revealed that PlGF is an independent prognostic factor. Taken together, it is concluded that PlGF is an important factor that can influence the angiogenesis process in NSCLC, and that the levels of PlGF reflect MVD, which might be a predictive factor for NSCLC patients' prognosis.
The current study has used quantitative real time PCR, to determine the levels of PlGF gene transcript. It has been shown that PlGF mRNA expression was detected in all NSCLC tissues and the matched normal tissues. The levels of PlGF transcript in these tissues varied from weak to strong. In 17 pairs of lung cancer and normal tissue samples, PlGF mRNA expression level in cancers was significant higher than the normal tissue. There has been no report about the PlGF mRNA expression level in NSCLC, although limited reports have studied mRNA expression of VEGF, a family member of PlGF, in NSCLC. Dolrini et al reported VEGF mRNA expression in cancer and normal tissue samples from 22 patients with the method of quantitative competitive real time PCR [21]. They detected VEGF mRNA in 18 samples of healthy tissue and all samples of tumour tissue and found that VEGF expression was higher in the tumour tissue than in the matched healthy tissue in 17 cases [21]. The current study has also supported by a recent report to show that lung cancer cell lines expressed PlGF, although it was indicated that SCLC tissues had a higher levels of PlGF staining than NSCLC [22]. Collectively, we suggest that PlGF is a factor that has strong prognostic value in NSCLC as shown in the present study and possibly in SCLC as shown in a smaller scale study [22]. A larger scale study will undoubtedly further clarify the connection.
The other interesting finding from the current study is that PlGF mRNA expression level has no significant difference with sex, histological type, tumour size, and lymph node status. However, advanced NSCLC tumours (stage III-IV) have higher levels of PlGF expression than the early stage NSCLC (I-II). This is somewhat in contrast with the immunohistochemical assessment, in that little statistical difference was observed between different stages (table 1). This is interesting and perhaps reflects the following: (1) the sensitivity of different assays. Quantitative analysis of gene transcript is a highly sensitive technique which can detect minute amount of genetic material. The technique also allows detection of very high levels of expression. Immunohistochemical analysis and the assessment based on colorimetric staining are less sensitive and sometime subjective in nature. (2) The correlation between protein and mRNA. Although a good relationship between mRNA and protein translation in the cells have been well documented, this may not be exactly translated at the tissue level. Thus, each technique has its advantage, i.e. Q-PCR for sensitivity and quantitative nature and IHC for its ability to identify protein and the location of proteins.
In summary, the current study has shown that the angiogenic factor, PlGF is ubiquitously in lung tissues and is predominantly a cytoplasmic protein in lung epithelial and lung cancer cells. PlGF expression is significantly higher in cancer tissue than in normal tissue, and is positively correlated with tumour stage and tumour size. This indicates that PlGF may have some role in tumour progression and that blocking/targeting PlGF expression may have promising therapeutic future in NSCLC.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LJZ: sample collection and preparation, study design, and patients follow-up;
JFC: sample processing and conducting experimental analysis;
YK: Experimental design and preparation of the manuscript;
REM: study design and manuscript preparation;
WGJ: study design, data analysis, and manuscript preparation.
Acknowledgements
Cancer Research Wales and Cancer Research, UK are acknowledged for their support.
==== Refs
Folkman J Shing Y Angiogenesis J Biol Chem 1992 267 10931 10934 1375931
Maeda T Matsumura S Hiranuma H Jikko A Furukawa S Ishida T Fuchihata H Expression of vascular endothelial growth factor in human oral squamous cell carcinoma: its association with tumour progression and p53 gene status J Clin Path 1998 51 771 775 10023341
Tanigawa N Amaya H Matsumura M Shimomatsuya T Horiuchi T Muraoka R Iki M Extent of tumour vascularization correlates with prognosis and haematogenous metastasis in gastric carcinomas Cancer Res 1996 56 2671 2676 8653715
Maglione D Guerriero V Viglietto G Ferraro MG Aprelikova O Alitalo K Del Vecchio S Lei KJ Chou JY Persico MG Two alternative mRNAs coding for the angiogenic factor, placenta growth factor (PlGF), are transcribed from a single gene of chromosome 14 Oncogene 1993 8 925 931 7681160
Shibuya M Role of VEGF-Flt receptor system in normal and tumour angiogenesis Advances Cancer Res 1995 67 281 316 8571818
DiSalvo J Bayne ML Conn G Kwok PW Trivedi PG Soderman DD Palisi TM Sullivan KA Thomas KA Purification and characterization of a naturally occurring vascular endothelial growth factor/placenta growth factor heterodimer J Biol Chem 1995 270 7717 7723 7706320 10.1074/jbc.270.13.7717
Cao Y Chen H Zhou L Chiang MK Anand-Apte B Weatherbee JA Wang Y Fang F Flanagan JG Tsang ML Heterodimers of placenta growth factor/vascular endothelial growth factor J Biol Chem 1996 271 3154 3162 8621715 10.1074/jbc.271.6.3154
Fong GH Rossant J Gertsenstein M Breitman ML Role of the Flt-1 receptor tyrosine kinase in regulating the assembly of vascular endothelium Nature 1995 376 66 70 7596436 10.1038/376066a0
Shalaby F Rossant J Yamaguchi TP Gertsenstein M Wu XF Breitman ML Schuh AC Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice Nature 1995 376 62 66 7596435 10.1038/376062a0
Hauser S Weich HA A heparin-binding form of placental growth factor (PlGF-2) is expressed in human umbilical vein endothelial cells and in placenta Growth Factor 1993 9 259 268
Sobin LH Fleming ID TNM Classification of Malignant Tumours, fifth edition. Union Internationale Contre le Cancer and the American Joint Committee on Cancer Cancer 1997 80 1803 1804 9351551 10.1002/(SICI)1097-0142(19971101)80:9<1803::AID-CNCR16>3.0.CO;2-9
Travis WD WHO histological typing of lung and pleural tumours 1999 Geneva, WHO
Parr C Cai J Boulton M Mansel RE Jiang WG PlGF and its receptor in human breast cancer Eur J Cancer 2005
Kumazaki K Nakayama M Suehara N Wada Y Expression of vascular endothelial growth factor, and their receptors Flt-1 and KDR in human placenta under pathologic conditions Hum Pathol 2002 33 1069 1077 Erratum in: Hum Pathol 2002, 33(12):1244 12454810 10.1053/hupa.2002.129420
Mineo TC Ambrogi V Baldi A Rabitti C Bollero P Vincenzi B Tonini G Prognostic impact of VEGF, CD31, CD34 and CD105 expression and tumour vessel invasion after radical surgery for IB-IIA non-small cell lung cancer J Clin Pathol 2004 57 591 597 15166262 10.1136/jcp.2003.013508
Jiang WG Watkins G Martin TA Sampson J Lee-Jones L Watkins G Douglas-Jones A Mokbel K Mansel RE The expression of TSC genes in human breast cancer Eur J Cancer 2005 41 1628 1636 15951164 10.1016/j.ejca.2005.03.023
Ririe KM Rasmussen RP Wittwer CT Product differentiation by analysis of DNA melting curves during the polymerase chain reaction Anal Biochem 1997 245 154 160 9056205 10.1006/abio.1996.9916
Cao Y Ji WR Qi P Rosin A Cao Y Placenta growth factor: identification and characterization of a nove isoform generated by RNA alternative splicing Biochem Biophys Res Commun 1997 235 493 498 9207183 10.1006/bbrc.1997.6813
Carmeliet P Moons L Luttun A Vincenti V Compernolle V De Mol M Wu Y Bono F Devy L Beck H Scholz D Acker T DiPalma T Dewerchin M Noel A Stalmans I Barra A Blacher S Vandendriessche T Ponten A Eriksson U Plate KH Foidart JM Schaper W Charnock-Jones DS Hicklin DJ Herbert JM Collen D Persico MG Synergism between vascular endothelial growth factor and placenta growth factor contributes to angiogenesis and plasma extravasation in pathological conditions Nat Med 2001 7 575 583 11329059 10.1038/87904
Park JE Chen HH Winer J Houck KA Ferrara N Placenta growth factor. Potentiation of vascular endothelial growth factor bioactivity, in vivo and in vitro, and high anity binding to Flt-1 but not to Flk-1/KDR J Biol Chem 1994 269 25646 25654 7929268
Boldrini L Calcinai A Silvestri V Basolo F Lucchi M Mussi A Angeletti CA Bevilacqua G Fontanini G Quantitation by competitive PCR assay of vascular endothelial growth factor in non-small cell lung carcinoma Int J Oncol 1999 14 161 168 9863024
Woo IS Park MJ Byun JH Hong YS Lee KS Park YS Lee JA Park YI Ahn HK Expression of placental growth factor gene in lung cancer Tumour Biol 2004 25 1 6 15192305 10.1159/000077716
|
16223445
|
PMC1276823
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Oct 13; 3:68
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-68
|
oa_comm
|
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-591625091810.1186/1475-925X-4-59ResearchAxial stent strut angle influences wall shear stress after stent implantation: analysis using 3D computational fluid dynamics models of stent foreshortening LaDisa John F [email protected] Lars E [email protected] Douglas A [email protected] David C [email protected] Judy R [email protected] Paul S [email protected] Department of Pediatrics (Division of Cardiology), Stanford University, Palo Alto, California, USA2 Department of Anesthesiology, the Medical College of Wisconsin and the Clement J. Zablocki Veterans Affairs Medical Center, Milwaukee, Wisconsin, USA3 Department of Medicine (Division of Cardiovascular Diseases), the Medical College of Wisconsin and the Clement J. Zablocki Veterans Affairs Medical Center, Milwaukee, Wisconsin, USA4 Department of Pharmacology and Toxicology, the Medical College of Wisconsin and the Clement J. Zablocki Veterans Affairs Medical Center, Milwaukee, Wisconsin, USA5 Department of Biomedical Engineering, Marquette University, Milwaukee, Wisconsin, USA2005 26 10 2005 4 59 59 22 8 2005 26 10 2005 Copyright © 2005 LaDisa et al; licensee BioMed Central Ltd.2005LaDisa et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
The success of vascular stents in the restoration of blood flow is limited by restenosis. Recent data generated from computational fluid dynamics (CFD) models suggest that the vascular geometry created by an implanted stent causes local alterations in wall shear stress (WSS) that are associated with neointimal hyperplasia (NH). Foreshortening is a potential limitation of stent design that may affect stent performance and the rate of restenosis. The angle created between axially aligned stent struts and the principal direction of blood flow varies with the degree to which the stent foreshortens after implantation.
Methods
In the current investigation, we tested the hypothesis that stent foreshortening adversely influences the distribution of WSS and WSS gradients using time-dependent 3D CFD simulations of normal arteries based on canine coronary artery measurements of diameter and blood flow. WSS and WSS gradients were calculated using conventional techniques in ideal (16 mm) and progressively foreshortened (14 and 12 mm) stented computational vessels.
Results
Stent foreshortening increased the intrastrut area of the luminal surface exposed to low WSS and elevated spatial WSS gradients. Progressive degrees of stent foreshortening were also associated with strut misalignment relative to the direction of blood flow as indicated by analysis of near-wall velocity vectors.
Conclusion
The current results suggest that foreshortening may predispose the stented vessel to a higher risk of neointimal hyperplasia.
CFDrestenosisneointimal hyperplasiaatherosclerosisstent geometry
==== Body
Introduction
Stents are commonly used to treat coronary and peripheral vascular stenoses, but restenosis after implantation remains a persistent problem. The mechanisms responsible for restenosis have yet to be clearly elucidated[1,7,14]. Many stent designs are currently used in the clinical setting, and each of these designs has unique geometric and mechanical properties[5,13,36]. A correlation between stent geometry and the severity of subsequent neointimal hyperplasia has been described in animal models and humans[14,35,37,44]. We recently demonstrated that the number, width and thickness of stent struts and the deployment diameter and scaffolding created by an implanted stent substantially affects the area of the stented zone subjected to low wall shear stress (WSS) using time-dependent 3D computational fluid dynamics modeling[23,24]. We have also recently shown that these areas of low WSS predict subsequent development of neointimal hyperplasia in vivo[25].
Foreshortening (the difference between the desired and actual stent length after deployment) is a potential limitation of stent design that may affect stent performance and the rate of restenosis. The deployment of all types of stents is associated with deformation of struts about specific vertices within each stent mesh. Foreshortening varies with stent morphology because of differential deformation of interconnected struts during deployment[2,40]. The angle created between axially aligned stent struts and the principal direction of blood flow varies with the degree to which the stent foreshortens after implantation. Thus, theoretically deleterious variations in WSS distribution may occur in stent designs that demonstrate substantial degrees of foreshortening. A previous study demonstrated that strut angle influences cellular adhesion in vitro[12]. These data suggest that changes in strut angle resulting from foreshortening contribute to an altered flow environment that may be conducive to restenosis. In the current investigation, we tested the hypothesis that foreshortening adversely influences the distribution of WSS and WSS gradients previously implicated in neointimal hyperplasia using 3D computational fluid dynamics models.
Methods
Construction of stented computational vessels
A custom automated geometric construction and mesh generation algorithm was used to construct idealized computational arteries containing a slotted-tube stent embedded within a normal unstented computational vessel based on blood flow and diameter measurements obtained from canine left anterior descending coronary arteries[21]. The algorithm allows for the alteration of several geometric parameters including the stent length and diameter relative to that of the unstented portions of the computational vessel. The thickness and width of all stent struts were 0.096 and 0.197 mm, respectively. All computational vessels were composed of structured hexahedral control volumes arranged in a three-domain butterfly design that exploited symmetric stent and vessel properties to model one-fourth of the computational vessel. Three time-dependent simulations were performed in total. Computational vessels were created consisting of 8 axial and circumferential repeating strut sections that were deployed using a stent-to-artery diameter ratio of 1.2 to 1 (Figure 1). The diameter of the unstented portions of the computational vessels for all simulations was 2.74 mm. The computational vessel within the stented region conformed to the geometry of the implanted stent as previously documented[3,7,11,31]. The length of the computational stents were 12,14 or 16 mm where 16 mm represents the ideal stent length after implantation and the 14 and 12 mm stents represent progressive degrees of foreshortening, respectively. The amount of foreshortening used in the current models (12 to 25%) is consistent with stent foreshortening observed clinically[6,16]. The stent strut angles with respect to the primary direction of blood flow were 151°, 146° and 141°, for the 16, 14 and 12 mm stents, respectively. Maintaining the same number of axial and circumferential repeating strut units throughout each simulation facilitated the study of flow disturbances produced by this stent strut orientation angle with respect to the primary direction of blood flow (Figure 2).
Figure 1 Computational vessels implanted with 12, 14 or 16 mm stents consisting of 8 axial and circumferential repeating strut sections that were deployed using a stent-to-artery diameter convention of 1.2 to 1. The diameter of the native vessel for all simulations was 2.74 mm. The computational vessel implanted with the 16 mm stent was designated as the ideal stent length after implantation and the 14 and 12 mm stents represent progressive degrees of foreshortening.
Figure 2 Schematic illustration demonstrating measurement of the stent strut angle with respect to the primary direction of blood flow.
Computational model simulations
Simulations were performed using the commercially available software package CFD-ACE (CRDRC; Huntsville, AL, ). This software uses a finite volume approach to solve the Navier-Stokes equations at the center of each hexahedral control volume. Theoretical arteries were subjected to a blood flow velocity waveform obtained from a canine coronary artery under normal resting conditions (Figure 3). A plug flow velocity profile was imposed at the inlet of each vessel and additional length was added to all arteries to allow for fully developed flow[10,23]. An initial axial velocity of 0.105 m/s was imposed within the fluid domain at the start of each simulation to avoid a "cold start" and increase the likelihood of convergence. A zero pressure boundary condition was imposed at the outlet of the computational vessels. Simulations were conducted using a backward Euler temporal differencing method to investigate time-dependent changes in indices of WSS within the stented portion of each vessel. Computational simulations were conducted assuming a Newtonian, incompressible fluid with a density of 1.06 g/cm3 and viscosity of 3.7 cP. The Reynolds and Womersley numbers for all of the simulations were approximately 105 and 2.91, respectively.
Figure 3 Blood flow velocity waveform measured in the proximal portion of a canine left anterior descending coronary artery and used for the time-dependent simulations conducted in the current investigation.
Determination of wall shear stress indices
Wall shear stress was determined as the product of viscosity and shear rate. A detailed discussion of this calculation is presented elsewhere[24]. Briefly, the CFD-ACE flow solver calculates shear rate during incompressible flow using the second invariant of the strain rate tensor. Thus, shear rate () was determined as where u, v, and w are the x, y and z components of velocity vector, u, respectively.
This definition accounts for pure shear as well as extensional or elongational deformation in the flow domain.
Spatial wall shear stress gradients (WSSG) were calculated during post-processing as discussed previously[15,23]. WSSG was used to quantify the influence of non-uniform hemodynamic forces on adjacent intravascular cells. Previous studies have suggested that this spatial WSS inhomogeneity may correlate with the location of neointimal hyperplasia in vivo[4,38,39] as was recently observed following chronic stent implantation into rabbit iliac arteries[25]. Frictional forces that act predominantly in axial and circumferential directions were most likely to cause expansion of intracellular gaps and disrupt intracellular junctions[28]. As a result, WSSG was calculated as , where τw is the WSS in the axial (z) and circumferential (θ) directions, respectively. WSSG observed overlying stent struts were excluded from the current analysis because these areas do not acutely contain biologically active tissue immediately after acute stent implantation.
Quantification of simulation results
The threshold for comparing distributions of low WSS between simulations was established at 5 dynes/cm2 for comparison to previous work[23,24] and because vascular regions subjected to WSS below this value have been shown to strongly correlate with sites of intimal thickening[17,18]. Regions of low WSS were also expressed as percentages of the stent area within intrastrut regions in the proximal, middle and distal portions of the stent in order to determine perturbations produced by foreshortening and stent strut orientation. Near-wall velocity vectors were also visualized at spatial locations in the proximal, middle and distal portions of the stent to observe the behavior of blood flow in these regions.
WSSG have also been used previously to examine the hypothesis that normally confluent cells react to nonuniform distributions of WSS in a way that promotes neointimal hyperplasia[9,32,38]. The percentage of the vessel wall subjected to WSSG values above 20 dynes/cm3 was quantified and compared between simulations in the current investigation. WSSG of this order of magnitude previously correlated with areas of neointimal hyperplasia in the toe region of an end-to-side arterial anastomosis[19,27,32].
Results
Mesh and time-step independence
Multiple simulations were performed to investigate spatial mesh independence. The spatial mesh density was four times greater in stented as compared to unstented regions of the computational vessels resulting in approximately 250,000 nodes per quarter of each computational artery. Results were considered to be spatially independent of the computational mesh when differences in the distributions of WSS were less than 6% between successive mesh densities[20,30]. Simulations were performed on a Dell Optiplex GX270 2.4 GHz workstation with 2 Gbyte of RAM that enabled convergence of simulations at a rate of approximately 25 time-steps per day. Time-step independence was examined by subjecting computational vessels to the coronary artery blood flow velocity waveform illustrated in Figure 3 using time-step increments of 10.9, 8.0 or 5.4 ms. Three consecutive cardiac cycles were allowed to reach simulation convergence, and distributions of WSS were compared at equivalent points during each cardiac cycle and between waveform permutations. The results demonstrated that a single cardiac cycle was sufficient to allow for the evolution of initial conditions, and simulation results became periodic shortly after the first cardiac cycle. A time step increment of 8.0 ms was sufficient to resolve temporal distributions of WSS within the stented and unstented regions of each vessel.
Simulation results
Indices of WSS corresponding to the mean blood flow velocity during the cardiac cycle are shown in Table 1. Lower WSS was observed within the stented region of all simulations, and stagnation zones occurred around stent struts. The percentage of intrastrut area exposed to distributions of WSS < 5 dynes/cm2 was greatest in the proximal portion of each stent independent of length and strut orientation (Table 1). These intrastrut areas of low WSS were least pronounced in the distal portion of the stents. The percentage of intrastrut area exposed to WSS < 5 dynes/cm2 within a particular intrastrut region decreased as the length of the stent increased. Conversely, the total area of the stent exposed to WSS < 5 dynes/cm2 increased as the stent length increased. However, when normalized by the length of the stent, the percentage of the computational vessel subjected to WSS < 5 dynes/cm2 modestly increased with the degree of foreshortening (0.78 as compared to 0.71 mm2 for 16 and 12 mm stents, respectively).
Table 1 Stent properties and indices of wall shear stress
Stent Length (mm) 16 14 12
Luminal surface area covered by the stent (%) 24 22 19
Stent strut angle with respect to primary flow direction (°) 151 146 141
Proximal Intrastrut WSS < 5 dynes/cm2 (%) 48 52 55
Middle Intrastrut WSS < 5 dynes/cm2 (%) 18 46 50
Distal Intrastrut WSS < 5 dynes/cm2 (%) 12 21 48
Total area exposed to WSS < 5 dynes/cm2 (mm2) 117 111 96.8
Total area exposed to WSSG > 20 dynes/cm3 (mm2) 89.3 72.5 67.4
Normalized area exposed to WSS < 5 dynes/cm2 0.71 0.77 0.78
Normalized area exposed to WSSG > 20 dynes/cm3 0.54 0.50 0.54
WSSGmax (dynes/cm3) 248 293 395
Abbreviations: WSS = wall shear stress; WSSG = wall shear stress gradient
The total area of the computational vessel subjected to WSSG greater than 20 dynes/cm3 increased as the stent length increased (67.4 versus 89.3 dynes/cm3 for 12 and 16 mm stents, respectively). In addition, the maximum WSSG observed throughout the stented region decreased as stent length increased (395, 293 and 248 dynes/cm3 for the 12, 14 and 16 mm stents, respectively). When normalized by the length of the stent, the portion of vessels subjected to elevated WSSG was similar for each computational vessel simulation (Table 1).
Value-weighted near-wall velocity vectors in the first proximal, middle and last distal repeating stent unit resulting from the unique strut orientation of each of the stents are illustrated in Figure 4. Reductions in the size of the vectors in the proximal as compared to the middle region resulted from the increase in luminal diameter within the stented region and account for the pronounced low WSS found in the proximal portion of the computational stent in all simulations. Adjacent blood layers appear to converge before entering each repeating stent unit and diverge after entering each axial diamond. This divergence accounted for reductions in WSS adjacent to stent struts. Figure 4 also demonstrates that the 12 mm stent foreshortening simulation produced large deviations in value-weighted near-wall velocity vectors adjacent to stent struts, in contrast to the observations with computational stents deployed to an ideal length of 16 mm.
Figure 4 Value-weighted near-wall velocity vectors in the proximal, middle and distal portions of the stent resulting from the unique strut orientation angles corresponding to ideal (16 mm) and progressively foreshortened stents (14 and 12 mm).
Time-dependent alterations in the spatial distributions of WSS throughout the cardiac cycle are illustrated in Figure 5. Stagnation regions observed adjacent to stent struts developed immediately before blood flow deceleration at the local and global maxima of the flow waveform (rows 2, 4 and 6). These stagnation regions were more pronounced in the 12 and 14 mm computational stents as compared to the 16 mm simulation.
Figure 5 Time-dependent alterations in spatial wall shear stress throughout the cardiac cycle in ideal stents (16 mm) and those with progressive degrees of foreshortening (14 and 12 mm).
Discussion
Increasing evidence suggests that the geometry and physical properties of a deployed stent may influence the local blood flow environment and alter distributions of WSS after implantation, thereby rendering certain areas of the vessel wall more susceptible to neointimal hyperplasia and subsequent restenosis[11,29,31,43]. Our laboratory has previously demonstrated that the number, width and thickness of strut struts, the stent-to-artery deployment diameter, and stent scaffolding influence the area of a vessel exposed to low WSS and elevated spatial and temporal WSSG using computational fluid dynamics models generated from measured in vivo data[20,23,24]. These results suggested that stent geometry profoundly affects fluid dynamics that correlate with the spatial and temporal localization of neointimal hyperplasia as we recently reported using a combined approach of chronic iliac artery stent implantation in vivo, microfocal x-ray CT imaging, computational fluid dynamics and vascular histology[22,25,26].
The current results confirm and extend our previous observations and further demonstrate the importance of stent geometry on intravascular fluid dynamics. The current findings with 3D computational fluid dynamics modeling demonstrate that stent foreshortening increases the area of the luminal surface exposed to low WSS and elevated spatial WSSG. These data strongly suggest that stent foreshortening may precipitate more pronounced development of neointimal hyperplasia as compared to a stent deployed to its ideal length. This hypothesis remains to be tested in vivo, however. The angle created between axially aligned stent struts and the principal direction of blood flow varied with the deployed stent length. The current results using computational foreshortened stents and those deployed to their ideal length demonstrate that struts of the foreshortened stents were progressively misaligned with the direction of blood flow. This strut misalignment within the flow domain also substantially influenced the behavior of near-wall velocity vectors and contributed to the observed increases in deleterious distributions of WSS and WSSG in foreshortened as compared to ideal computational stent architecture. Stent foreshortening effectively increases the number of struts per unit length of a vessel, and we have previously demonstrated that such an increase in strut number is associated with greater exposure of the luminal surface to low WSS and elevated WSSG[23]. These previous and current observations are consistent with the intuitive notion that optimal stent design should assure that the vast majority of struts are aligned more parallel to blood flow in order to minimize changes in laminar blood flow characteristics and WSS distributions.
Although the normalized luminal area exposed to WSS < 5 dynes/cm2 was similar for each of the stents computationally modeled in the current investigation, it is important to consider which of the parameters reported in Table 1 are most likely to influence the local cellular response and be associated with neointimal hyperplasia. The table clearly indicates that the angle of stent struts with respect to the primary flow direction strongly influences the intrastrut area of the vessel wall exposed to low WSS. Cells in these pronounced areas of low WSS are more likely to participate in the neointimal response. Normalizing the total area of the vessel exposed to low WSS by the length provides a less sensitive indication of the local impact on cells within the stented region that are likely to contribute to neointimal hyperplasia.
The current results require interpretation within the constraints of several potential limitations. Stents were implanted in idealized computational representations of healthy vessels and the results would likely differ if the study was performed using computational models of vascular disease. From a clinical perspective, stent diameter typically increases with the degree of foreshortening. Such an increase in stent diameter would be expected to intrinsically produce a reduction in blood velocity, and hence WSS, within the stented region. We conducted the current investigation using a consistent diameter within the stented region of each computational vessel in order to isolate the influence of the incident strut angle on indices of WSS independent of changes in vessel diameter. As a result, the current results may actually underestimate the relative amount of the computational vessels subjected to low WSS and elevated spatial WSSG as a result of this observed clinical increase in diameter often associated with foreshortening. The current investigation was conducted using a rigid-wall approximation and a simplified outlet boundary condition. Previous studies have demonstrated that implantation of 16 mm slotted-tube stents into canine epicardial coronary arteries in vivo reduced vessel compliance to zero within the stented region[21] and that wall deformability does not greatly alter the velocity field under normal conditions[34]. Nevertheless, the results may differ from those observed clinically as a result of arterial compliance proximal and distal to the stented region and the ability of the distal vasculature to vasodilate in response to local metabolic need, thus altering the upstream resistance through the stented segment.
The potential clinical applicability of the current results in lieu of the widespread use of drug-eluting stents to treat vascular stenoses also deserves comment. Recent reviews suggested that drug-eluting stents may simply delay restenosis, but do not facilitate healing of the intima that is damaged by prior vascular disease or the stent implantation procedure. Thus, restenosis rates with drug-eluting stent may ultimately be similar to bare metal stents[8,33,41,42]. Moreover, this drug-eluting technology may not be applicable to all patient populations and locations within the arterial vasculature[8,33]. The current results further support the hypothesis that local flow patterns created by the stent should also be considered during stent design so as to minimize indices of fluid dynamics implicated in neointimal hyperplasia and to avoid the use of exogenous cytotoxic or antiproliferative agents.
Conclusion
In summary, the current results using 3D computational fluid dynamics modeling indicate that foreshortening-induced orientation of stent struts with respect to the direction of blood flow is associated with increases in the area of the vessel subjected to low WSS and high WSSG, factors that have been previously correlated with subsequent neointimal hyperplasia in vivo. The change in incident strut angle with respect to the predominant flow direction was modest between modeled foreshortened as compared to ideal computational stents (141 to 151°), but increasing the degree of foreshortening reduced the axial alignment angle of stent struts and produced greater intrastrut areas of low WSS and elevated WSSG. Stent geometries that successfully restore distal perfusion but produce the least disruption to the native flow environment may ultimately provide maximum chronic vessel patency.
Authors' contributions
JFL planned and conducted the experiments from which the blood flow and diameter measurements used in the current investigation were obtained, created the automated geometric construction and mesh generation algorithm, formulated fluid dynamics models, conducted simulations and wrote drafts of the manuscript. LEO assisted with the formulation of fluid dynamics models, planning and completion of simulations, and helped with data analysis. DAH assisted in the design of the computational investigation, formulated fluid dynamics models, and helped with data analysis. JRK assisted in data analysis and critical revisions of manuscript. DCW also assisted in data analysis and critical revisions of the manuscript. PSP critically reviewed the design of the computational study and experiments from which the measurements used in the current investigation were obtained, fluid dynamics theory, and analysis of results and also provided multiple critical revisions of several drafts of the manuscript, including the final submitted manuscript. All authors read and agreed to the submission of the manuscript in its current form.
Acknowledgements
The authors would like to recognize David A. Schwabe and John P. Tessmer for their experimental support, Mary Lorence-Hanke for assistance in the preparation of this manuscript (Department of Anesthesiology, Medical College of Wisconsin), and Max Imas (Olin Engineering Center, Marquette University) and Juan Santiago and Rajiv Bharadwaj (Microfluidics Laboratory, Stanford University Department of Mechanical Engineering) for their technical and computational support. This work was supported in part by a Sigma Xi Grants-in-Aid of Research Award (to Dr. LaDisa) and by National Institutes of Health grants HL-03690 (to Dr. Kersten), HL-063705 (to Dr. Kersten), HL-054820 (to Dr. Warltier), and GM-008377 (to Dr. Warltier) from the United States Public Health Service, Bethesda, Maryland.
==== Refs
Bennett MR O'Sullivan M Mechanisms of angioplasty and stent restenosis: implications for design of rational therapy Pharmacology & Therapeutics 2001 91 149 166 11728607 10.1016/S0163-7258(01)00153-X
Farb A Weber DK Kolodgie FD Burke AP Virmani R Morphological predictors of restenosis after coronary stenting in humans Circulation 2002 105 2974 2980 12081990 10.1161/01.CIR.0000019071.72887.BD
Kastrati A Mehilli J Dirschinger J Pache J Ulm K Schuhlen H Seyfarth M Schmitt C Blasini R Neumann FJ Schomig A Restenosis after coronary placement of various stent types Am J Cardiol 2001 87 34 39 11137830 10.1016/S0002-9149(00)01268-6
Duda SH Wiskirchen J Tepe G Bitzer M Kaulich TW Stoeckel D Claussen CD Physical properties of endovascular stents: an experimental comparison Journal of Vascular and Interventional Radiology 2000 11 645 654 10834499
Kalmar G Hubner F Voelker W Hutzenlaub J Teubner J Poerner T Suselbeck T Borggrefe M Haase KK Radial force and wall apposition of balloon-expandable vascular stents in eccentric stenoses: an in vitro evaluation in a curved vessel model Journal of Vascular and Interventional Radiology 2002 13 499 508 11997358
Roguin A Beyar R beStent--the serpentine balloon expandable stent: review of mechanical properties and clinical experience Artificial Organs 1998 22 243 249 9527286 10.1046/j.1525-1594.1998.06120.x
Rogers C Edelman ER Endovascular stent design dictates experimental restenosis and thrombosis Circulation 1995 91 2995 3001 7796511
Schulz C Herrmann RA Beilharz C Pasquantonio J Alt E Coronary stent symmetry and vascular injury determine experimental restenosis Heart 2000 83 462 467 10722552 10.1136/heart.83.4.462
Yoshitomi Y Kojima S Yano M Sugi T Matsumoto Y Saotome M Tanaka K Endo M Kuramochi M Does stent design affect probability of restenosis? A randomized trial comparing Multilink stents with GFX stents Am Heart J 2001 142 445 451 11526357 10.1067/mhj.2001.117321
LaDisa JFJ Olson LE Guler I Hettrick DA Audi SH Kersten JR Warltier DC Pagel PS Stent design properties and deployment ratio influence indexes of wall shear stress: a three-dimensional computational fluid dynamics investigation within a normal artery J Appl Physiol 2004 97 424 430 14766776 10.1152/japplphysiol.01329.2003
LaDisa JFJ Olson LE Guler I Hettrick DA Kersten JR Warltier DC Pagel PS Circumferential vascular deformation after stent implantation alters wall shear stress evaluated using time-dependent 3D computational fluid dynamics models Journal of Applied Physiology 2005 98 947 957 15531564 10.1152/japplphysiol.00872.2004
LaDisa JFJ Olson LE Molthen RC Hettrick DA Pratt PF Hardel MD Kersten JR Warltier DC Pagel PS Alterations in wall shear stress predict sites of neointimal hyperplasia after stent implantation in rabbit iliac arteries Am J Physiol Heart Circ Physiol 2005 288 H2465 75 15653759 10.1152/ajpheart.01107.2004
Verstandig AG Bloom AI Sasson T Haviv YS Rubinger D Shortening and migration of Wallstents after stenting of central venous stenoses in hemodialysis patients Cardiovascular and Interventional Radiology 2003 26 58 64 12522643 10.1007/s00270-002-1953-6
Choussat R Black AJ Bossi I Joseph T Fajadet J Marco J Long-term clinical outcome after endoluminal reconstruction of diffusely degenerated saphenous vein grafts with less-shortening wallstents Journal of the American College of Cardiology 2000 36 387 394 10933347 10.1016/S0735-1097(00)00724-5
Hamuro M Palmaz JC Sprague EA Fuss C Luo J Influence of stent edge angle on endothelialization in an in vitro model Journal of Vascular and Interventional Radiology 2001 12 607 611 11340140
LaDisa JFJ Hettrick DA Olson LE Guler I Gross ER Kress TT Kersten JR Warltier DC Pagel PS Coronary stent implantation alters coronary artery hemodynamics and wall shear stress during maximal vasodilation J Appl Physiol 2002 93 1939 1946 12391052
Danenberg HD Welt FG Walker M Seifert P Toegel GS Edelman ER Systemic inflammation induced by lipopolysaccharide increases neointimal formation after balloon and stent injury in rabbits Circulation 2002 105 2917 2922 12070123 10.1161/01.CIR.0000018168.15904.BB
Garasic JM Edelman ER Squire JC Seifert P Williams MS Rogers C Stent and artery geometry determine intimal thickening independent of arterial injury Circulation 2000 101 812 818 10683357
Newman VS Berry JL Routh WD Ferrario CM Dean RH Effects of vascular stent surface area and hemodynamics on intimal thickening J Vasc Interv Radiol 1996 7 387 393 8761820
Kreutzer J Rome JJ Open-cell design stents in congenital heart disease: a comparison of IntraStent vs. Palmaz stents Catheter Cardiovasc Interv 2002 56 400 409 12112897 10.1002/ccd.10180
Ebeid MR Balloon expandable stents for coarctation of the aorta: review of current status and technical considerations Images in Pediatric Cardiology 2003 15 25 41
Fox RW McDonald AT Introduction to Fluid Mechanics 1992 4 New York, John Wiley & Sons, Inc. 322 1601824
Kleinstreuer C Hyun S Buchanan JR Longest PW Archie JP Truskey GA Hemodynamic parameters and early intimal thickening in branching blood vessels Crit Rev Biomed Eng 2001 29 1 64 11321642
DePaola N Gimbrone MAJ Davies PF Dewey CF Vascular endothelium responds to fluid shear stress gradients Arterioscler Thromb 1992 12 1254 1257 1420084
Tada S Tarbell JM Flow through internal elastic lamina affects shear stress on smooth muscle cells (3D simulations) American Journal of Physiology 2002 282 H576 584 11788405
Tardy Y Resnick N Nagel T Gimbrone MA Dewey CF Shear stress gradients remodel endothelial monolayers in vitro via a cell proliferation-migration-loss cycle Arterioscler Thromb Vasc Biol 1997 17 3102 3106 9409299
Lei M Kleinstreuer C Truskey GA A focal stress gradient-dependent mass transfer mechanism for atherogenesis in branching arteries Med Eng Phys 1996 18 326 332 8963477 10.1016/1350-4533(95)00045-3
Ku DN Giddens DP Zarins CK Glagov S Pulsatile flow and atherosclerosis in the human carotid bifurcation. Positive correlation between plaque location and low oscillating shear stress Arteriosclerosis 1985 5 293 302 3994585
Ku DN Blood flow in arteries Annu Rev Fluid Mech 1997 29 399 434 10.1146/annurev.fluid.29.1.399
Finol EA Amon CH Blood flow in abdominal aortic aneurysms: pulsatile flow hemodynamics J Biomech Eng 2001 123 474 484 11601733 10.1115/1.1395573
Ojha M Spatial and temporal variations of wall shear stress within an end-to-side arterial anastomosis model J Biomech 1993 26 1377 1388 8308043 10.1016/0021-9290(93)90089-W
Kute SM Vorp DA The effect of proximal artery flow on the hemodynamics at the distal anastomosis of a vascular bypass graft: computational study J Biomech Eng 2001 123 277 283 11476372 10.1115/1.1374203
Lei M Kleinstreuer C Archie JPJ Geometric design improvements for femoral graft-artery junctions mitigating restenosis J Biomech 1996 29 1605 1614 8945659 10.1016/0021-9290(96)83730-4
Myers JG Moore JA Ojha M Johnston KW Ethier CR Factors influencing blood flow patterns in the human right coronary artery Annals of Biomedical Engineering 2001 29 109 120 11284665 10.1114/1.1349703
LaDisa JFJ Guler I Olson LE Hettrick DA Kersten JR Warltier DC Pagel PS Three-dimensional computational fluid dynamics modeling of alterations in coronary wall shear stress produced by stent implantation. Ann Biomed Eng 2003 31 972 980 12918912 10.1114/1.1588654
Murata T Hiro T Fujii T Yasumoto K Murashige A Kohno M Yamada J Miura T Matsuzaki M Impact of the cross-sectional geometry of the post-deployment coronary stent on in-stent neointimal hyperplasia: an intravascular ultrasound study Circ J 2002 66 489 493 12030346 10.1253/circj.66.489
Wentzel JJ Whelan DM van der Giessen WJ van Beusekom HM Andhyiswara I Serruys PW Slager CJ Krams R Coronary stent implantation changes 3-D vessel geometry and 3-D shear stress distribution Journal of Biomechanics 2000 33 1287 1295 10899339 10.1016/S0021-9290(00)00066-X
LaDisa JFJ Meier HT Olson LE Kersten JR Warltier DC Pagel PS Antegrade iliac artery stent implantation for the temporal and spatial examination of stent-induced neointimal hyperplasia and alterations in regional fluid dynamics Journal of Pharmacological and Toxicological Methods 2005 51 115 121 15767204 10.1016/j.vascn.2004.08.001
LaDisa JFJ Olson LE Ropella KM Molthen RC Haworth ST Kersten JR Warltier DC Pagel PS Microfocal x-ray computed tomography post-processing operations for optimizing reconstruction volumes of stented arteries during 3D computational fluid dynamics modeling Computer Methods and Programs in Biomedicine 2005 79 121 134 15894401 10.1016/j.cmpb.2005.03.005
Perktold K Rappitsch G Computer simulation of local blood flow and vessel mechanics in a compliant carotid artery bifurcation model Journal of Biomechanics 1995 28 845 856 7657682 10.1016/0021-9290(95)95273-8
Finn AV Palacios IF Kastrati A Gold HK Drug-eluting stents for diabetes mellitus: a rush to judgment? J Am Coll Cardiol 2005 45 479 483 15708689 10.1016/j.jacc.2004.10.060
Palmaz JC Intravascular stents in the last and the next 10 years J Endovasc Ther 2004 Suppl 2 II200 206 15760261
Virmani R Farb A Guagliumi G Kolodgie FD Drug-eluting stents: caution and concerns for long-term outcome Coronary Artery Disease 2004 15 313 318 15346088 10.1097/00019501-200409000-00003
Virmani R Kolodgie FD Farb A Lafont A Drug eluting stents: are human and animal studies comparable? Heart 2003 89 133 138 12527658 10.1136/heart.89.2.133
|
16250918
|
PMC1276824
|
CC BY
|
2021-01-04 16:37:35
|
no
|
Biomed Eng Online. 2005 Oct 26; 4:59
|
utf-8
|
Biomed Eng Online
| 2,005 |
10.1186/1475-925X-4-59
|
oa_comm
|
==== Front
CoughCough (London, England)1745-9974BioMed Central London 1745-9974-1-31627092310.1186/1745-9974-1-3MethodologyEvaluation of an ambulatory system for the quantification of cough frequency in patients with chronic obstructive pulmonary disease Coyle Michael A [email protected] Desmond B [email protected] Linda S [email protected] Michael L [email protected] Brett K [email protected] David W [email protected] Michael G [email protected] Physiology Program, Harvard School of Public Health, Boston, MA, USA2 VivoMetrics, Inc., Ventura, CA, USA3 GlaxoSmithKline, Respiratory and Inflammation Centre of Excellence for Drug Discovery Research Triangle Park, NC, USA4 GlaxoSmithKline, Respiratory and Inflammation Centre of Excellence for Drug Discovery Stevenage, UK5 Community Research, Inc., Cincinnati, OH, USA6 Department of Psychiatry, Indiana University School of Medicine, Indianapolis, IN, USA2005 4 8 2005 1 3 3 25 4 2005 4 8 2005 Copyright © 2005 Coyle et al; licensee BioMed Central Ltd.2005Coyle et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
To date, methods used to assess cough have been primarily subjective, and only broadly reflect the impact of chronic cough and/or chronic cough therapies on quality of life. Objective assessment of cough has been attempted, but early techniques were neither ambulatory nor feasible for long-term data collection. We evaluated a novel ambulatory cardio-respiratory monitoring system with an integrated unidirectional, contact microphone, and report here the results from a study of patients with COPD who were videotaped in a quasi-controlled environment for 24 continuous hours while wearing the ambulatory system.
Methods
Eight patients with a documented history of COPD with ten or more years of smoking (6 women; age 57.4 ± 11.8 yrs.; percent predicted FEV1 49.6 ± 13.7%) who complained of cough were evaluated in a clinical research unit equipped with video and sound recording capabilities. All patients wore the LifeShirt® system (LS) while undergoing simultaneous video (with sound) surveillance. Video data were visually inspected and annotated to indicate all cough events. Raw physiologic data records were visually inspected by technicians who remained blinded to the video data. Cough events from LS were analyzed quantitatively with a specialized software algorithm to identify cough. The output of the software algorithm was compared to video records on a per event basis in order to determine the validity of the LS system to detect cough in COPD patients.
Results
Video surveillance identified a total of 3,645 coughs, while LS identified 3,363 coughs during the same period. The median cough rate per patient was 21.3 coughs·hr-1 (Range: 10.1 cghs·hr-1 – 59.9 cghs·hr-1). The overall accuracy of the LS system was 99.0%. Overall sensitivity and specificity of LS, when compared to video surveillance, were 0.781 and 0.996, respectively, while positive- and negative-predictive values were 0.846 and 0.994. There was very good agreement between the LS system and video (kappa = 0.807).
Conclusion
The LS system demonstrated a high level of accuracy and agreement when compared to video surveillance for the measurement of cough in patients with COPD.
==== Body
Background
The most frequent complaint for which patients seek treatment from primary care physicians in the United States is cough [1]. Type, frequency and diurnal changes of cough may be criteria for differential diagnosis, therapeutic efficacy, and a gauge for the progression of chronic disease. Historically, cough evaluation has been difficult and of limited clinical value due to a lack of surveillance tools to assess cough frequency completely and its impact on health-related quality of life (HRQL).
To date, methods used to assess cough have been primarily subjective, and only broadly reflect the impact of chronic cough and/or chronic cough therapies on quality of life [2-5]. These methods have been unable to offer substantial information related to the minimal reduction in cough frequency necessary to achieve a significant improvement in HQRL. Objective assessment of cough has been attempted, but these techniques were neither ambulatory nor feasible for long-term data collection [6-8]. Other systems have evaluated sound to quantify cough frequency and intensity with moderate success [9,10], but have been limited in their effectiveness outside the laboratory and requires labor intensive analysis and interpretation [11-15].
We evaluated a novel ambulatory cardio-respiratory monitoring system with an integrated unidirectional, contact microphone, and report here the results from a study of patients with COPD who were videotaped in a quasi-controlled environment for 24 continuous hours while wearing the ambulatory system.
Methods
Subjects
Eight subjects with chronic obstructive pulmonary disease (COPD) who complained of cough as a prominent symptom (e.g., ten or greater self reported bouts of cough per day) were recruited for the study. Subjects were men and women over the age of 40 who had a documented medical history of COPD and a smoking history of ≥ 10 years with chronic productive cough. Patient characteristics can be found in Table 1. Patients were excluded from the study if, upon screening, (1) it was determined from patient medical history that cough could be due to other known causes such as gastro-esophageal reflux, asthma, or any anatomical abnormalities of the upper respiratory tract, and/or (2) if patients were using prescribed or over the counter anti-tussive medications within 24-hours of the start of the study.
Table 1 Patient characteristics. Values are means ± SD; Ht = height; Wt = weight; BMI = Body mass index; %FEV1 = % predicted forced expiratory volume in one second; n = 8 (6 women)
Variable
Age (yrs) 57.4 ± 11.8
Ht (cm) 165.4 ± 7.2
Wt (kg) 76.1 ± 14.4
BMI (kg/m2) 27.8 ± 4.7
%FEV1 49.6 ± 13.7
The protocol was approved by an independent ethical review board (Western IRB, 3535 7th Avenue SW, Olympia, WA, USA, 98502) and all patients received a verbal and written description of the study and gave informed consent prior to participation. All data were collected under the medical supervision of board certified pulmonologists.
Instrumentation and monitoring
LifeShirt® System
Patients were fitted with the wearable LifeShirt® system (LS, VivoMetrics, Inc., Ventura, CA, USA), which incorporates respiratory inductance plethysmography (RIP) for the non-invasive measurement of volume and timing ventilatory variables and has been described elsewhere [15-22]. The system also incorporates a unidirectional contact microphone, a single channel ECG, and a centrally located, 3-axis accelerometer. Data were processed and stored on a compact flash card that was housed within the recorder unit. Patients were invited to wear the LS system for a maximum of 24 hours.
Video surveillance
Patients spent the testing period in an assigned room where the video monitoring equipment was installed. Patients were monitored via video recorder camera (low-lux) with unidirectional free-air microphone for the duration of the testing period. The video data stream was synchronized to the LS data stream by the coordination of the device clocks. The LS recorder has an on-board electronic diary which creates an event time stamp in the LS software data stream which was referenced to the video data time display to determine the beginning of the recording period. Patients were allowed free range of the research facility and were permitted to watch television, use the telephone, dine, take breaks and sleep.
Data analysis and statistics
Raw physiologic data records were uploaded to a centralized data center and were visually inspected for quality by technicians. 94.1% of the data were interpretable and available for comparison to video. Specialized software (VivoLogic®, VivoMetrics, Inc., Ventura, CA, USA) was used to view the LS data and a proprietary algorithm housed within the software was used to identify cough from the physiologic recordings. LS data were visually inspected by two independent reviewers who remained blinded to the video data. Each noted the time (hour, minute and second) of each cough. These data were captured into a spreadsheet. Cough events (hour, minute, second, millisecond) identified by the LS software were exported into a separate spreadsheet. A practical extraction and report language (PERL) script was written to temporally align the two data streams so that the output from each device could be compared for agreement on an event by event basis.
To summarize the validity and reliability of the ambulatory system to detect cough under several conditions, six validation and agreement measures were used including, sensitivity (SN), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), accuracy and kappa [23] were calculated relative to video rating. The PPV is the probability that a patient coughed, if the system judged the respiratory event as a cough. Likewise, the NPV is the probability that the patient did not cough, given that the system did not judge the event as a cough. Accuracy is the proportion of all correct tests. The method used to calculate the confidence intervals was the Wilson score method without continuity correction [24], which has been previously shown to exhibit a logit scale symmetry property with consequent log scale symmetry for certain derived intervals [25].
Results
A satisfactory fit of the available standard sizes of the respiratory inductance plethysmography (RIP) garment was achieved in all patients and the system was well tolerated during the recording period. Figure 1 and Figure 2 depict a representative recording of a single cough during quiet breathing and during a series of coughs close together, respectively. Figure 3 is a representative recording of speech.
Figure 1 Representative recording of a single cough followed by a throat clear during quiet breathing. VT = tidal volume; Fb = breathing frequency; Mic = contact microphone output; SE = sound envelope; HR = heart rate; ECG = electrocardiograph tracing; Posture = body position defined as upright, supine, right decubitis and left decubitis; Cough = cough output from algorithm. The shaded bar contains the cough event. Cough is indicated by a single solid line at the end of the breath that contains the cough. Note the change in the posture from supine to upright to supine immediately following the cough. Entire duration of depicted recording is 1-min and 1-sec.
Figure 2 Representative recording of coughing during sleep. VT = tidal volume; Fb = breathing frequency; Mic = contact microphone output; SE = sound envelope; HR = heart rate; ECG = electrocardiograph tracing; Posture = body position defined as upright, supine, right decubitis and left decubitis; Cough = cough output from LS algorithm. The first shaded bar contains the cough bout. Ten coughs occurred during the 9-sec bout. Each cough is indicated by a single solid line at the end of the breath that contains the cough. Note that the cough bout was followed by a 15-sec apnea (second shaded bar). Entire duration of depicted recording is 1-min and 23-sec.
Figure 3 Representative recording of talking and laughing. VT = tidal volume; Fb = breathing frequency; Mic = contact microphone output; SE = sound envelope; HR = heart rate; ECG = electrocardiograph tracing; Posture = upright. The shaded bar contains a burst of laughing. Entire duration of depicted recording is 1-min and 37-sec.
Patients were invited to be observed for a maximum of 24 hours. A total of 109 hours of simultaneous recordings of video and LS were obtained. Of that time, 73.9 hours were observed during the day and 34.7 hours were observed during the night. During the recording period, the total number of coughs documented by video surveillance was 3,645. The LS system reported 3,363 coughs during the same time period. The median cough rate per patient was 21.3 coughs·hr-1 (Range: 10.1 cghs·hr-1 – 59.9 cghs·hr-1).
Table 2 provides performance summaries for the LS system to detect cough during for night vs. day and for low and high respiration rates, respectively. Patients were assigned to low vs. high respiratory rate depending on whether the rate was below or above the median breathing frequency (median Fb = 21 br·min-1). The system was highly accurate in identifying cough as a respiratory event during night or day. Accuracy during the night was 99.4%, while accuracy during the day was 98.8% for a difference = 0.53%. The specificities and negative predictive values are considered 'excellent' by the criteria proposed by Byrt (1996)[26]. Sensitivities, positive predictive values and kappa can be considered 'very good' by the same criteria. Likewise, the performance summaries for the system between high or low respiration rates were remarkably similar. Accuracy, specificities, and negative predictive values were 'excellent' and sensitivities, positive predictive values and kappa were 'very good' [26].
Table 2 Validation and agreement statistics (with 95% confidence intervals) for the LifeShirt system during day & night and at low & high respiration rates. Values are calculated values for the sensitivity (SN), specificity (SP), positive predictive-value (PPV), negative predictive-value (NPV), accuracy (ACC) and the kappa statistic. Values in parentheses are the 95% confidence intervals. All values are for LS system compared to video documentation of cough events; * p-value < 0.0001 for night vs. day comparisons of period for SN, SP, NPV and ACC; ¶p-value < 0.0001 for night vs. day comparisons of respiratory rate for SN, SP, PPV, NPV; ‡ p-value = 0.004 for day vs. night comparisons of respiratory rate for ACC. Day period defined as 0600–1800; Night period defined as 1800–0600. Patients were assigned to the low or high respiratory rate group based on whether their mean Fb was below or above the median Fb (median = 21 br·min-1).
Period Respiratory Rate
Combined Day Night Low High
SN 78.1 (76.7, 79.4) 76.7 (75.1, 78.2) 82.7 (80.0, 85.1)* 69.5 (66.3, 72.6) 80.6 (79.0, 82.0)¶
SP 99.6 (99.5, 99.6) 99.6 (99.5, 99.6) 99.7 (99.7, 99.8)* 99.5 (99.5, 99.6) 99.7 (99.6, 99.7) ¶
PPV 84.6 (83.3, 85.8) 84.5 (83.0, 85.8) 85.0 (82.3, 87.3) 69.8 (66.5, 72.8) 89.4 (88.1, 90.5) ¶
NPV 99.4 (99.3, 99.4) 99.3 (99.2, 99.3) 99.7 (99.6, 99.7)* 99.5 (99.5, 99.6) 99.3 (99.2, 99.3) ¶
ACC 99.0 (99.0, 99.1) 98.8 (98.8, 98.9) 99.4 (99.3, 99.5)* 99.1 (99.0, 99.2) 99.0 (98.9, 99.0)‡
Kappa 80.7 (79.7, 81.7) 79.8 (78.6, 81.0) 83.5 (81.5, 85.5) 69.2 (66.6, 71.7) 84.2 (83.1, 85.3)
Discussion
We report validity and reliability statistics for a novel ambulatory system to evaluate its capability to detect cough and demonstrate a high level of agreement and accuracy when compared to video surveillance for cough over an extended period. The system was well-tolerated and allowed for free movement throughout the monitoring facility. The system continuously monitors several cardio-pulmonary-activity variables, which allowed us to evaluate ventilatory strategies associated with coughing, which is one of its novel features.
The patient population in this study coughed with great frequency, which reflects the fact these were COPD patients who had a primary complaint of cough. At screening, patients were asked if they coughed ten times per day or more. Although all of the patients met this requirement, they had difficulty recalling how many cough bouts per day they experienced. As such, we did not predict that this population would cough with such a high frequency and, although the number of coughs was higher than anticipated, it was within the range of what has been reported previously [12].
Agreement between the LS system and video surveillance was excellent. Interestingly, we did observe that the nocturnal validation and agreement statistics, as well as differences between low and high respiratory rates, were statistically significantly different, although they differed only slightly in magnitude. These small, yet statistically significant, difference likely reflect the influence of the respiratory events (e.g., VT and Fb) sample size during the recording period on the statistical power for these comparisons and is not clinically significant.
Objective cough assessment has been attempted on numerous previous occasions [6-8,10,12,14]. Until now, a robust, accurate ambulatory system has failed to emerge. This is likely due to the fact that previous systems have attempted to identify cough with a single physiologic signal (e.g., sound). Sound-based technologies have been the primary means of cough assessment due to the audible sound that is generated during a cough [28]. These systems, however, are susceptible to a high false positive rate when ambient noise is prominent and are unable to distinguish cough-like sounds (e.g., throat clearing) from true cough. Hsu et al. (1994) [12] augmented sound analyses with concomitant intercostal electromyography (EMG) analyses and evaluated various clinical populations (e.g., normal controls, stable asthmatics and patients with daily, persistent & non-productive cough) and concluded that their system may be useful in the assessment of antitussive therapies. Hsu et al., however, did not present evidence of agreement by comparing their results to a reference standard.
There were three limitations to the study. First, the sample size was small. Sample size was limited by available resources to review the vast amount of video and LS data. Second, women (6/8) were over represented in the study, which was due to an inauspicious baseline imbalance. Third, the data were collected in a clinical setting due to the requirement for video monitoring equipment. However, patients were not confined to any one space and ambulated, spoke on the phone, dined and performed additional activities of daily living.
A substantial challenge in this study was the choice of a reference standard with which to evaluate the novel device. We chose video based on the fact that the source document (video) could be reviewed during the adjudication process if there was uncertainty with respect to the occurrence of a cough. Scoring the video in duplicate was an arduous task which likely increased the possibility of human error due to fatigue and it is possible that some coughs were missed. Events that were missed by both reviewers would not have been adjudicated, but identified by the LS system and therefore would have been inappropriately scored as a false-positive. Thus, these results are a conservative estimate of LS capabilities and may underestimate the predictive power of the device.
Conclusion
We report data from a novel, ambulatory, multi-signal device that shows a high level of agreement and accuracy when compared to video/audio surveillance over an extended period and confirm its potential in the evaluation of antitussive therapies. The availability of a valid, robust ambulatory tool for quantifying cough will enable the determination of the minimal required reduction in cough to maximally improve patient HQRL, and open up a broad array of research questions both specific to cough and wherein cough may be an important covariate, comorbidity, or confounding factor.
Competing interests
Financial Disclosure: MAC and DBK were employed by VivoMetrics during the course of the study. MGW consults for VivoMetrics. LSH, MLW, BKH are employed by GlaxoSmithKline. Supported by a grant from GlaxoSmithKline.
==== Refs
Schappert SM National ambulatory medical care survey: summary Vital Health Stat 1993 230 1 20
Birring SS Prudon B Carr AJ Singh SJ Morgan MD Pavord ID Development of a symptom specific health status measure for patients with chronic cough: Leicester Cough Questionnaire (LCQ) Thorax 2003 58 339 343 12668799 10.1136/thorax.58.4.339
French CL Irwin RS Curley FJ Krikorian CJ Impact of chronic cough on quality of life Arch Intern Med 1998 158 1657 1661 9701100 10.1001/archinte.158.15.1657
Irwin RS French CT Fletcher KE Quality of life in coughers Pulm Pharmacol Ther 2002 15 283 286 12099779 10.1006/pupt.2002.0347
French CT Irwin RS Fletcher KE Adams TM Evaluation of a cough-specific quality-of-life questionnaire Chest 2002 121 1123 1131 11948042 10.1378/chest.121.4.1123
Barach AL Bickerman HA Beck GJ Clinical and physiological studies on the use of metacortandracin in respiratory disease.1.Bronchial asthma DC 1955 27 515 522
Bickerman HA Barach AL The experimental production of cough in human subjects induced by citric acid aerosols. Peliminary studies in the evaluation of antitussive agents Am J Med Sci 1954 228 156 163 13180508
Prime FJ The assessment of antitusive drugs in man Br Med J 1961 1 1149 1151 13738034
Woolf CR Rosenberg A The cough suppressant effect of heroin and codeine: a controlled clinical study Can Med Assoc J 1962 87 810 814 13985171
Sevelius H Colmore JP Objective assessment of antitussive agents in patients with chronic cough J New Drugs 1966 6 216 233 5338653
Chang AB Phelan PD Robertson CF Newman RG Sawyer SM Frequency and perception of cough severity J Paediatr Child Health 2001 37 142 145 11328468 10.1046/j.1440-1754.2001.00608.x
Hsu JY Stone RA Logan-Sinclair RB Worsdell M Busst CM Chung KF Coughing frequency in patients with persistent cough: assessment using a 24 hour ambulatory recorder Eur Respir J 1994 7 1246 1253 7925902 10.1183/09031936.94.07071246
Dalmasso F Isnardi E Sudaro L Bellantoni R Bioacoustins of cough during bronchial inhalation challenge (BIC) with methacholine Eur Resp J 2001 18 135S
Subburaj S Parvez L Rajagopalan TG Methods of recording and analysing cough sounds Pulm Pharmacol 1996 9 269 279 9232663 10.1006/pulp.1996.0035
Tobin MJ Chadha TS Jenouri G Birch SJ Gazeroglu HB Sackner MA Breathing patterns. 1. Normal subjects Chest 1983 84 202 205 6872603
Tobin MJ Chadha TS Jenouri G Birch SJ Gazeroglu HB Sackner MA Breathing patterns. 2. Diseased subjects Chest 1983 84 286 294 6884104
Chadha TS Schneider AW Birch S Jenouri G Sackner MA Breathing pattern during induced bronchoconstriction Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 1984 56 1053 1059
Sackner MA Gonzalez HF Jenouri G Rodriguez M Effects of abdominal and thoracic breathing on breathing pattern components in normal subjects and in patients with chronic obstructive pulmonary disease American Review of Respiratory Disease 1984 130 584 587 6486557
Tobin MJ Guenther SM Perez W Mador MJ Accuracy of the respiratory inductive plethysmograph during loaded breathing Journal of Applied Physiology 1987 62 497 505 3558209
Cantineau JP Escourrou P Sartene R Gaultier C Goldman M Accuracy of respiratory inductive plethysmography during wakefulness and sleep in patients with obstructive sleep apnea Chest 1992 102 1145 1151 1395758
Carter GS Coyle MA Mendelson WB Validity of a portable cardio-respiratory system to collect data in the home environment in patients with obstructive sleep apnea Sleep and Hypnosis 2004 6 85 92
Clarenbach CF Senn O Brack T Bloch KE Monitoring of ventilation during exercise by a novel portable respiratory inductive plethysmograph Chest, In Print 2005
Greenhalgh T How to read a paper. Papers that report diagnostic or screening tests BMJ 1997 315 540 543 9329312
Wilson EB Probable inference, the law of succession, and statistical inference J Am Stat Assoc 1927 22 209 212
Newcombe RG Two-sided confidence intervals for the single proportion: comparison of seven methods Stat Med 1998 17 857 872 9595616 10.1002/(SICI)1097-0258(19980430)17:8<857::AID-SIM777>3.0.CO;2-E
Byrt T How good is that agreement? (Letter to editor) Epidemiology 1996 7 561 8862998
|
16270923
|
PMC1277005
|
CC BY
|
2021-01-04 16:37:44
|
no
|
Cough. 2005 Aug 4; 1:3
|
utf-8
|
Cough
| 2,005 |
10.1186/1745-9974-1-3
|
oa_comm
|
==== Front
CoughCough (London, England)1745-9974BioMed Central London 1745-9974-1-41627092810.1186/1745-9974-1-4ResearchEffect of stroke location on the laryngeal cough reflex and pneumonia risk Addington W Robert [email protected] Robert E [email protected] John G [email protected] Kamel [email protected] Brevard Rehabilitation Medicine, 200 Ocean Avenue, Suite 201; Melbourne Beach, Florida, 32951, USA2 Chair, Department of Anatomy, Kansas City University of Medicine and Biosciences, 1750 Independence Avenue, Kansas City, Missouri, 64106, USA3 Emeritus Professor, University of London, 116 Pepys Road, London SW20 8NY, UK4 Chair, Department of Mathematics and Statistics, University of Missouri–Kansas City; Kansas City, Missouri, USA2005 4 8 2005 1 4 4 1 7 2005 4 8 2005 Copyright © 2005 Addington et al; licensee BioMed Central Ltd.2005Addington et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The purpose of this study was to evaluate the risk of developing pneumonia in acute stroke patients comparing the early anatomical stroke location and laryngeal cough reflex (LCR) testing.
Methods
A prospective study of 818 consecutive acute stroke patients utilizing a reflex cough test (RCT), which assesses the neurological status of the LCR compared to magnetic resonance imaging or computerized tomography for stroke location and subsequent pneumonia outcome. Stroke diagnosis and stroke location were made by a neurologist and clinical radiologist, respectively; both were blinded to the RCT results.
Results
Brainstem (p-value < .007) and cerebral strokes (p-value < .005) correlated with the RCT results and pneumonia outcome. Of the 818 patients, 35 (4.3%) developed pneumonia. Of the 736 (90%) patients who had a normal RCT, 26 (3.5%) developed pneumonia, and of the 82 (10%) patients with an abnormal RCT, 9 (11%) developed pneumonia despite preventive interventions (p-value < .005). The RCT had no serious adverse events.
Conclusion
The RCT acted as a reflex hammer or percussor of the LCR and neurological airway protection and indicated pneumonia risk. Despite stroke location, patients may exhibit "brainstem shock," a global neurological condition involving a transient or permanent impairment of respiratory drive, reticular activating system or LCR. Recovery of these functions may indicate emergence from brainstem shock, and help predict morbidity and mortality outcome.
coughreflexlaryngealstrokepneumoniabrainstem shock
==== Body
Background
The laryngeal cough reflex (LCR) protects the supraglottic larynx from significant aspiration of food or fluids during inspiration or pharyngeal spillage during swallowing [1]. The reflex cough test (RCT), using nebulized tartaric acid solution, provides an effective stimulus to the receptors in the supraglottic mucosa, and, like a reflex hammer or percussor, triggers a cascade of neurological activity in both craniospinal nerves and the central nervous system. The vagus nerve mediates the afferent component of the LCR. Tartaric acid-induced cough stimulates rapid adapting receptors (RARs) in the supraglottic region of the larynx and sensory impulses are conveyed to the medulla via the middle ramus of the internal branch of the superior laryngeal nerve (ibSLN) and vagus nerve [2-6]. The fibers of the ibSLN terminate on neurons near the nucleus tractus solitarius (NTS) of the medulla. The central connections of the reflex and voluntary cough circuits have been reviewed,[7,8] Although the central connections of reflex cough are unclear, research suggests a rapid latency [5]. Central processing of the cough reflex quickly sets off a cascade of synchronized central and peripheral responses involving the nucleus ambiguus, retroambigualis, phrenic nucleus, and medial motor cell column which project to the vagus, phrenic, intercostal and thoracoabdominal nerves, respectively [9].
Human studies have indicated the clinical implications of central and peripheral lesions of the cough system. The LCR may be impaired in individuals who have a transient (e.g., post-general anesthesia) or permanent (e.g., post-stroke, cervical cord trauma, Parkinson's disease, amyotrophic lateral sclerosis) neurological event, which may affect the afferent, central or efferent components of the LCR [10-18]. The purpose of this study was to determine the effect of identifying the initial radiological anatomical stroke location on the laryngeal cough reflex test result and the relationship to the subsequent risk of developing pneumonia.
Methods
This was a clinical prospective study of 818 consecutive patients during a three year period of time, who were admitted to an acute rehabilitation hospital with a primary diagnosis of acute stroke. Stroke diagnosis was made by a neurologist and the stroke location was determined by a neuroradiologist, both were blinded to the RCT findings. In this study, stroke location was noted according to computer tomography (CT) or magnetic resonance imaging (MRI) results as reported by the radiologist in the clinical setting. Stroke locations were categorized as: cerebral, brainstem, multiple CNS infarcts, basal ganglion, cerebellar, or location not specified by MRI or CT report.
Upon admission to the acute rehabilitation hospital, all patients were tested with the RCT, as the first component of a standard bedside swallow examination. The RCT (Pneumoflex Systems, Inc., Melbourne Beach, FL) comprised a 20% solution of pharmaceutical grade L-tartaric acid dissolved in sterile 0.15 M NaCl solution and inhaled from a Bennett Twin Nebulizer (3012-60 ml, Puritan-Bennett Company, Carlsbad, CA). During the inhalation, the subject's nose was pinched closed. The nebulizer output was 0.2 ml/min [1,3,19-23]. The RCT was administered at bedside by a either a respiratory therapist or speech pathologist. The subject was asked to exhale, then insert the mouthpiece, and take a sharp, deep inhalation. Leakage around the mouthpiece and "puffing" the nebulizer were not considered effective inhalations. The test was administered by a either a speech pathologist or respiratory therapist at bedside and required less than five minutes to complete. The expected result of a normal RCT was an immediate series of forceful coughs, which are primarily expiratory "airway clearing" in character. A normal finding indicated a normal function of the LCR, vagus nerve, and a neurologically protected airway. If the subject had a normal RCT, additional inhalations of the RCT were not performed. The expected result of an abnormal RCT was represented by an absence of coughing, or a diminished (weak) coughing, or coughing not immediately after administration of the test stimulus. An abnormal finding indicated dysfunction of the LCR, vagus nerve or the reflex cough system, and a neurologically unprotected airway. The RCT was terminated when the subject either elicited a cough or failed to cough after three valid inhalations. The subjects were then treated clinically based on the RCT findings. The previously published RCT algorithm was followed for subsequent feeding strategies such as restricted diet, nothing by mouth (NPO) or nutritional support via percutaneous endoscopic gastrostomy (PEG) [1]. These treatment strategies were noted for all patients.
Subjects were monitored for the development of pneumonia during their hospital stay of approximately one month. Pneumonia was diagnosed as a subject having respiratory symptoms with either temperature greater than 101°, leukocytosis, or both, and an infiltrate confirmed by chest x-ray. Adverse events of the RCT were collected for all subjects.
Data Analysis
Statistics were generated using SPSS 10.0.5. Subjects who had either a normal or abnormal RCT finding were statistically compared as to gender, age, and length of stay in the acute care setting. The principal endpoint for the study was the development of pneumonia among acute stroke patients. This endpoint is binary. An appropriate test of significance for this situation was the Fisher's exact test with the Null Hypothesis stating that among the acute stroke patients there is no significant difference in the development of pneumonia, regardless of the stroke location, between patients that had a normal RCT and those patients that had an abnormal RCT. In addition to a test of significance, it was important to determine a 95 percent confidence interval for p1 - p2, where p1 was the proportion of acute stroke patients that developed pneumonia and had an abnormal RCT, and p2 was the proportion of acute stroke patients that developed pneumonia and had a normal RCT. For completeness, the odds in favor of not developing pneumonia among the patients who had an abnormal RCT were compared to the odds in favor of not developing pneumonia among the patients who had a normal RCT. As part of a power analysis, there are standard formulae for determining sample sizes for the comparison of the proportions. The level of significance, power of the test and the proportions were evaluated. The RCT results, stroke location and pneumonia outcome were crosstabulated utilizing the Chi-Square test.
The issue of sensitivity and specificity of the RCT in determining pneumonia risk, though obviously important, is not appropriate to assess in this study because the interventions, guided by the results of the RCT, can effect pneumonia outcome [1,22,24].
Results
The mean age of these patients was 73.69 ± 10.44, and included 426 males and 392 females. The patients in this acute stroke population included more than 59 overall comorbidies. The mean length of stay in the acute care and rehabilitation hospitals was 7.7 ± 7.7 and 31.2 ± 18.4 days, respectively. Analysis of gender, age, and length of stay in the acute care setting indicated that there were no epidemiological differences between subjects who had either a normal or abnormal RCT finding.
The principal endpoint for the study is the development of pneumonia. Among the 818 acute stroke patients, 736 (90%) patients had a normal RCT, of which 26 patients (3.5%) developed pneumonia (Table 1). Eighty-two (10%) patients had an abnormal RCT, defined as weak or absent. Of the abnormal RCT group, 69 (84%) patients had a weak RCT, of which 7 (10%) developed pneumonia. Thirteen (16%) patients had absent RCT and 2 (15%) developed pneumonia. A significant difference for pneumonia outcome was found (p-value < .005) (Table 2). The 95 percent confidence interval for p1 - p2 was (.04, .11), respectively. This two-sided 95 percent confidence interval clearly showed that p1 is greater than p2. The proportion of acute stroke patients that developed pneumonia and had an abnormal RCT was significantly greater than the proportion of acute stroke patients that developed pneumonia and had a normal RCT. In fact, 3.5% of the patients with a normal RCT versus 11% of the patients with an abnormal (weak or absent) RCT developed pneumonia.
Table 1 Reflex Cough Test × Pneumonia in Rehabilitation Crosstabulation
Pneumonia in Rehabilitation Total
Reflex Cough Test Yes No
Normal 26 710 736
Abnormal
Weak 7 62 69
Absent 2 11 13
Total 35 783 818
Table 2 Chi-Square Tests on the Effects of RCT on Pneumonia Outcome
Value df Asymp. Sig. (2-sided) Exact Sig. (2-sided) Exact Sig. (1-sided)
Pearson Chi-Square 9.980 1 .002
Continuity Correction 8.245 1 .004
Likelihood Ratio 7.428 1 .006
Fisher's Exact Test .005 .005
Linear-by-Linear Association 9.967 1 .002
N of Valid Cases 818
a Computed only for a 2×2 table
b 1 cells (25.0%) have expected count less than 5. The minimum expected count is 3.51.
The odds in favor of not developing pneumonia among the patients who had an abnormal RCT were compared to the odds in favor of not developing pneumonia among the patients who had a normal RCT. The odds ratio test indicated that the odds in favor of not developing pneumonia for acute stroke patients with an abnormal RCT were significantly smaller than the odds in favor of not developing pneumonia for acute stroke patients that had a normal RCT. In fact, the ratio of the odds was .297, which was significantly smaller than 1, and a 95 percent confidence interval for the odds ratio was (.134, .658). When the level of significance is fixed at 0.05, the power of a two-sided test is 80 percent.
Stroke location, RCT results and pneumonia outcomes are shown in Table 3. Crosstabulation of the RCT results, pneumonia in rehabilitation and brainstem infarcts was significant at identifying the risk of developing pneumonia (p = .007) as was the crosstabulation for cerebral hemispheric infarcts (p = .005). Basal ganglionic, cerebellar, multiple infarcts or stroke location not specified by CT or MRI did not correlate with RCT results and predicting the development of pneumonia. Data analysis for basal ganglion and cerebellar infarcts were not crosstabulated because none of these patients developed pneumonia in rehabilitation.
Table 3 Stroke Location, Reflex Cough Test (RCT) Results and Pneumonia Outcome
RCT × Pneumonia in Rehabilitation × Brainstem Infarct Crosstabulation and Chi-Square Test
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 1 55 56
Abnormal 3 6 9
Total 4 61 65
Fisher's Exact Test p = .007
RCT × Pneumonia in Rehabilitation × Cerebral Infarct Crosstabulation and Chi-Square Test
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 9 355 364
Abnormal 5 31 36
Total 14 386 400
Fisher's Exact Test p = .005
RCT × Pneumonia in Rehabilitation × Stroke Location Not Specified Crosstabulation and Chi-Square Test
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 13 210 223
Abnormal 1 27 28
Total 14 237 251
Fisher's Exact Test p = .522
RCT × Pneumonia in Rehabilitation × Multiple CNS Infarcts Crosstabulation
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 3 48 51
Abnormal 0 4 4
Total 3 52 55
Fisher's Exact Test p = .794
Reflex Cough Test × Pneumonia in Rehabilitation × Basal Ganglion Infarcts Crosstabulation
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 0 24 24
Abnormal 0 3 3
Total 0 27 27
Reflex Cough Test × Pneumonia in Rehabilitation × Cerebellar Infarcts Crosstabulation
Reflex Cough Test Pneumonia in Rehabilitation Total
Yes No
Normal 0 18 18
Abnormal 0 2 2
Total 0 20 20
Thirty-two (3.91%) of the 818 patients had a percutaneous endoscopic gastrostomy (PEG) while in rehabilitation (Table 4). Ten PEGs were placed while in rehab, and 15 PEGs were removed in rehabilitation. Seven (0.9%) of the 818 patients received a modified barium swallow (MBS) examination.
Table 4 Crosstabulation of Pneumonia in Rehabilitation, Reflex Cough Test, Stroke Location and PEG
Pneumonia in Rehabilitation Reflex Cough Test Location of Stroke Percutaneous Endoscopic Gastrostomy Total
Yes No
No Normal Cerebral Infarct 6 349 355
Basal ganglia 3 21 24
Brainstem infarct 5 50 55
Cerebellar infarct 2 16 18
Multiple Infarctions 2 46 48
Location not determined 1 209 210
Abnormal Cerebral Infarct 8 23 31
Basal ganglia 1 2 3
Brainstem infarct 1 5 6
Cerebellar infarct 0 2 2
Multiple Infarctions 1 3 4
Location not determined 0 27 27
Yes Normal Cerebral Infarct 1 8 9
Brainstem infarct 1 0 1
Multiple Infarctions 0 3 3
Location not determined 0 13 13
Abnormal Cerebral Infarct 0 5 5
Brainstem infarct 0 3 3
Location not determined 0 1 1
Total 32 786 818
Seventeen patients (2.1%) were transferred to acute care from rehabilitation, 15 had a normal RCT and 3 of these patients developed pneumonia in rehabilitation. Two of the transferred patients had an abnormal RCT and neither developed pneumonia in rehabilitation. Two of the 818 patients died in rehabilitation. One patient had a left cerebral hemispheric infarct, and died of complications secondary to cancer. The other patient had a middle cerebral artery infarct, and died four days after admission to the rehabilitation hospital due to ongoing complications secondary to pneumonia acquired in acute care.
In this study, there were no serious adverse medical sequelae from RCT administration. In the 82 stroke patients who had an abnormal RCT, there was no statistical correlation for comorbidities such as congestive heart failure, diabetes mellitus, chronic obstructive pulmonary disease, or patients who had been intubated.
Discussion
Nebulized tartaric acid appears to be an effective, specific and safe stimulus to laryngeal receptors and testing neurological airway protection. This concentration of nebulized tartaric acid-induced cough has been used in a number of cough sensitivity studies involving normal subjects, smokers and asthmatics without causing bronchoconstriction [19-21,25,26]. Our studies on stroke subjects and other patients with neurological impairment use a single breath inhalation protocol similar to Choudry and Fuller [27]. When the internal branch of the superior laryngeal nerve was completely, bilaterally anesthetized, the LCR was transiently absent in normal subjects and they could tidal breathe nebulized tartaric acid without eliciting laryngeal or tracheobronchial cough [4]. Testing the neurological function of the LCR may help indicate those patients who are at risk of respiratory complications such as pneumonia.
This study reported a significant relationship among RCT results, pneumonia risk and both brainstem and cerebral strokes. Although all subjects had a primary diagnostic code of stroke and the initial stroke location was determined and described by a radiologist in the clinical setting, it is not always possible to determine the extent of the neurological deficits by the location of the infarct emergently using MRI or CT. This study, using the present examination techniques for identifying neurological deficits in the emergency setting showed that subjects, who had a subsequent brainstem or cerebral hemispheric infarct identified by CT or MRI and a subsequently impaired LCR, were at risk of developing pneumonia. Indeed, a clinician in the emergency room presently could not test the status of a patient's involuntary neurological airway protection, i.e., LCR, before making a decision to place a nasogastric (NG) tube or administer food, fluid or medications orally. The use of a NG tube in the acute stroke setting, without the knowledge of the neurological status of the LCR may be a significant contributing factor for the development of respiratory complications such as pneumonia, ventilator necessitation or death [28].
In the present study brainstem infarcts and cerebral hemispheric infarcts correlated with a significant risk of developing pneumonia, although basal ganglionic, cerebellar, multiple infarcts or stroke location not specified by CT or MRI did not correlate with RCT results and predicting pneumonia risk. None of the patients who had a basal ganglion and cerebellar infarct developed pneumonia in the acute rehabilitation setting. Nakagawa and coworkers reported that the incidence of pneumonia was significantly higher and the latency of swallowing response to the onset of was also longer in patients, who had either unilateral or bilateral basal ganglia infarcts than in patients with no infarct on CT [29]. However, they were studying the swallow reflex and silent aspiration during sleep in long term care patients and did not evaluate the LCR. Since the LCR and swallowing are separate neurological events, both must be evaluated separately. Although swallowing function is often assessed in hospital settings, testing the LCR is not presently performed although information as to the integrity of this vital, airway protective reflex would be helpful in patient management [1,22].
The numerous comorbidities of this acute stroke population along with the RCT results, stroke location, and pneumonia outcome data suggest the need to test all patients who might have an acute neurological impairment. The overall incidence of an abnormal RCT, in the rehabilitation setting day 4 or 5 post onset, was about 10% regardless of stroke location. The incidence of abnormal tests on acute stroke presentation in the emergency room would be higher. Patients with basal ganglionic, cerebellar, or infarct location unspecified and an abnormal RCT are not necessarily false positives, since the RCT results identified pneumonia risk and the need for appropriate intervention.
Although the central connections of the LCR are not clear in humans, the LCR probably has reciprocal connections with supratentorial areas that modulate or modify the LCR. Assuming adequate cognition and laryngeal sensorium, we are aware when the LCR is triggered by a noxious laryngeal stimulus–suggesting projections to the cerebrum. In humans a neurological interrelationship between the brainstem mediated LCR and supratentorial influences has been reported and an amygdalo-hypothalamo-reticular pathway has been suggested [30].
Cerebral hemispheric infarcts may, through mechanisms that are unclear at this time, suppress the LCR circuitry. Perhaps, moderate to large cerebral hemispheric lesions may result in neurotransmitter or neurophysiologic circuitry disruption or a downward pressure and/or mass effect secondary to cerebral edema, which could have an adverse effect upon these vital brainstem functions by interruption of descending facilitatory supratentorial pathways, as in spinal shock and general anesthesia. This condition is different from isolated brainstem lesions such alternating hemiplegias, lateral medullary syndrome, pontocerebellar angle syndrome or other bulbar lesions.
Suppression of the LCR tends to support our clinical observations that many cerebral hemispheric stroke patients, who show a transient or permanent impairment of the LCR, may have a condition we refer to as "brainstem shock." Brainstem shock may be defined as a global neurological condition involving a transient or permanent impairment of one or more of the following vital functions: the reticular activating system, respiratory drive, or the LCR. Frequently patients with large or small hemispheric strokes, bleeds or infarcts, may present unconscious with a depressed reticular activating system, or require intubation secondary to a depressed respiratory drive. Although presently not clinically tested in the emergency room or intensive care units, airway protection may also be impaired at this stage, and the initial emergency radiological examination may not give adequate information regarding the patient's neurological status. Patients in brainstem shock may recover reticular activating system function, respiratory drive, or neurological airway protection at different rates, similar to recovery from general anesthesia, and may be an important predictor of morbidity and mortality.
More detailed brain mapping of the lesion is not generally feasible, or available in the clinical setting. Such technologies might elucidate the connections between cerebral and brainstem structures associated with reflex cough, and would pose an interesting study. Although cough may be measured using more sophisticated techniques in the laboratory, such as electrophysiological or plethysmography, we feel that the method for grading cough, used in this study, is appropriate for the clinical setting as part of a comprehensive neurological examination.
Conclusion
Although instrumented exams of the CNS using MRI or CT may be an important component of a neurological evaluation, they cannot adequately assess vital neurological functions such as respiratory centers in the reticular formation, consciousness or airway protective reflexes such as the LCR. Neurologists are familiar with the neurologically impaired patient, who has an unremarkable brain imaging study or one in which the stroke location is not specified. Although these results rarely mitigate the primary diagnosis of stroke, clinicians must still assess the status of these vital functions. Objective assessment of respiratory function and clinical evaluation of consciousness are commonly performed. However, bedside testing of the LCR is not currently available, yet its status plays an important role when the clinician must initiate a strategy for food, fluids and medications that is safe for the patient. The RCT may be helpful for identifying a change in neurological status or progressing cerebral edema in emergent stroke patients before identification is possible with imaging exams, thus assist in directing urgent care. The RCT examination may be helpful in assessing recovery of airway protection following extubation or general anesthesia and would be important further research for patient care.
The most powerful finding in this study is a normal RCT. In acute neurological patients, without confounding structural head and neck conditions that may prevent physical closure of the larynx during swallowing, a normal RCT reliably provides the opportunity to safely and aggressively approach emergency procedures such as NG tube placement, and the administration of food, fluids and medications in the acute setting. Further research needs to be done on other neuropathophysiological conditions for those patients who have an abnormal RCT. Further research on the neurological condition of brainstem shock in acute neurological conditions needs to be performed.
Competing interests
Although none of the authors has been financially compensated for the research associated with this research, a commercial party with a financial interest in the reflex cough test may confer a financial benefit upon one or more of the authors. The reflex cough test (Pneumoflex®) of the laryngeal cough reflex is patented and trademarked by Pneumoflex Systems, LLC, Melbourne Beach, Florida. Pneumoflex® has not been used commercially in the past or present. Pneumoflex Systems, LLC, is pursuing FDA and EU approval. Use of this technique in the health care system requires regulatory approval.
Authors' contributions
RES and WRA conceived the study and drafted and revised the manuscript, WRA collected all subject data, JW helped draft and revise the manuscript, and KR performed the statistical analysis and wrote the appropriate section. All of the authors contributed to drafting the original and revised manuscripts, and have granted final approval of this published version.
==== Refs
Addington WR Stephens RE Gilliland KA Assessing the laryngeal cough reflex and the risk of developing pneumonia after stroke: An interhospital comparison Stroke 1999 30 1203 1207 10356100
Stephens RE Wendel KH Addington WR Anatomy of the internal branch of the superior laryngeal nerve Clin Anat 1999 12 79 83 10089032 10.1002/(SICI)1098-2353(1999)12:2<79::AID-CA1>3.0.CO;2-W
Addington WR Stephens RE Gilliland KA Miller SP Tartaric acid-induced cough and the superior laryngeal nerve evoked potential Am J Phys Med Rehabil 1998 77 523 526 9862540 10.1097/00002060-199811000-00014
Addington WR Stephens RE Goulding RE Anesthesia for the superior laryngeal nerves and tartaric acid-induced cough Arch Phys Med Rehabil 1999 80 1584 1586 10597810 10.1016/S0003-9993(99)90334-9
Addington WR Stephens RE Widdicombe JG Ockey RR Anderson JW Miller SP Electrophysiologic latency to the external obliques of the laryngeal cough expiration reflex in humans Am J Phys Med Rehabil 2003 82 370 373 12704276 10.1097/00002060-200305000-00010
Stephens RE Addington WR Wendel KH Anatomical measurements of the laryngeal cough expiratory reflex confirming the latency to the external abdominal oblique muscle in humans (Submitted February 2005) 2005
Pantaleo T Bongianni F Mutolo D Central nervous mechanisms of cough Pulm Pharmacol Ther 2002 15 227 233 12099769 10.1006/pupt.2002.0358
Lee PC Cotterill-Jones C Eccles R Voluntary control of cough Pulm Pharmacol Ther 2002 15 317 320 12099785 10.1006/pupt.2002.0365
Miller AJ Cirone D Low level potentiation of the brain stem laryngeal reflex Brain Res Bull 1976 1 385 391 184887 10.1016/0361-9230(76)90032-0
Behera D Das S Dash RJ Jindal SK Cough reflex threshold in diabetes mellitus with and without autonomic neuropathy Respiration 1995 62 263 268 8560093
Fontana GA Pantaleo T Lavorini F Benvenuti F Gangemi S Defective motor control of coughing in Parkinson's disease Am J Respir Crit Care Med 1998 158 458 464 9700121
Hadjikoutis S Wiles CM Eccles R Cough in motor neuron disease: a review of mechanisms Qjm 1999 92 487 494 10627867 10.1093/qjmed/92.9.487
Honda Y Takishima T [Respiratory functions of the upper airway with special reference to physiological implications of respiratory disease] Nihon Kyobu Shikkan Gakkai Zasshi 1990 28 3 4 2355683
Kobayashi H Hoshino M Okayama K Sekizawa K Sasaki H Swallowing and cough reflexes after onset of stroke Chest 1994 105 1623 8181386
Morice AH Turley AJ Linton TK Human ACE gene polymorphism and distilled water induced cough Thorax 1997 52 111 113 9059468
Niimi A Matsumoto H Ueda T Takemura M Suzuki K Tanaka E Chin K Mishima M Amitani R Impaired cough reflex in patients with recurrent pneumonia Thorax 2003 58 152 153 12554900 10.1136/thorax.58.2.152
Viguera M Diakum TA Shelsky R Casals P Cochs J Fauli A [Efficacy of topical administration of lidocaine through a Malinckrodt Hi-Lo Jet tube in lessening cough during recovery from general anesthesia] Rev Esp Anestesiol Reanim 1992 39 316 318 1410755
Wong CH Morice AH Cough threshold in patients with chronic obstructive pulmonary disease Thorax 1999 54 62 64 10343635
Fujimura M Sakamoto S Kamio Y Matsuda T Cough receptor sensitivity and bronchial responsiveness in normal and asthmatic subjects Eur Respir J 1992 5 291 295 1572440
Fujimura M Sakamoto S Kamio Y Saito M Miyake Y Yasui M Matsuda T Cough threshold to inhaled tartaric acid and bronchial responsiveness to methacholine in patients with asthma and sino-bronchial syndrome Intern Med 1992 31 17 21 1568037
Fujimura M Sakamoto S Kamio Y Matsuda T Sex difference in the inhaled tartaric acid cough threshold in non-atopic healthy subjects Thorax 1990 45 633 634 2402729
Addington WR Stephens RE Gilliland KA Rodriguez M Assessing the laryngeal cough reflex and the risk of developing pneumonia after stroke Arch Phys Med Rehabil 1999 80 150 154 10025488 10.1016/S0003-9993(99)90112-0
Addington WR Stephens RE Ockey RR Kann D Rodriguez M A new aspiration screening test to assess the need for modified barium swallow study (Abstract) Arch Phys Med Rehabil 1995 76 1040
AHCPR Summary, evidence report/technology assessment: Number 8, Diagnosis and treatment of swallowing disorders (dysphagia) in acute-care stroke patients 1999 Rockville, MD , United States Department of Health and Human Services 50 58
Sakamoto S Fujimura M Kamio Y Saito M Yasui M Miyake Y Matsuda T [Relationship between cough threshold to inhaled tartaric acid and sex, smoking and atopy in humans] Nihon Kyobu Shikkan Gakkai Zasshi 1990 28 1478 1481 2290233
Fujimura M Sakamoto S Kamio Y Matsuda T Effects of methacholine induced bronchoconstriction and procaterol induced bronchodilation on cough receptor sensitivity to inhaled capsaicin and tartaric acid Thorax 1992 47 441 445 1386691
Choudry NB Fuller RW Sensitivity of the cough reflex in patients with chronic cough Eur Respir J 1992 5 296 300 1572441
Dziewas R Ritter M Schilling M Konrad C Oelenberg S Nabavi DG Stogbauer F Ringelstein EB Ludemann P Pneumonia in acute stroke patients fed by nasogastric tube J Neurol Neurosurg Psychiatry 2004 75 852 856 15145999 10.1136/jnnp.2003.019075
Nakagawa T Sekizawa K Arai H Kikuchi R Manabe K Sasaki H High incidence of pneumonia in elderly patients with basal ganglia infarction Arch Intern Med 1997 157 321 324 9040300 10.1001/archinte.157.3.321
Stiller KR Ambridge MA Bowman PK Inability to cough voluntarily following a left cerebrovascular accident Aust N Z J Med 1991 21 78 79 2036083
|
16270928
|
PMC1277006
|
CC BY
|
2021-01-04 16:37:44
|
no
|
Cough. 2005 Aug 4; 1:4
|
utf-8
|
Cough
| 2,005 |
10.1186/1745-9974-1-4
|
oa_comm
|
==== Front
CoughCough (London, England)1745-9974BioMed Central London 1745-9974-1-51627093210.1186/1745-9974-1-5ResearchChange in bronchial responsiveness and cough reflex sensitivity in patients with cough variant asthma: effect of inhaled corticosteroids Fujimura Masaki [email protected] Johsuke [email protected] Shigeharu [email protected] Respiratory Medicine, Cellular Transplantation Biology, Kanazawa University Graduate School of Medicine, Kanazawa, Japan2005 25 8 2005 1 5 5 5 4 2005 25 8 2005 Copyright © 2005 Fujimura et al; licensee BioMed Central Ltd.2005Fujimura et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cough variant asthma (CVA) is a cause of chronic cough and a precursor of typical asthma. We retrospectively examined the longitudinal change in bronchial responsiveness and cough reflex sensitivity in CVA patients with respect to the effect of long-term inhaled corticosteroids (ICS).
Methods
Provocative concentration of methacholine causing a 20% fall in forced expiratory volume in one second (PC20-FEV1) and provocative concentration of capsaicin eliciting 5 or more coughs (C5) were measured before treatment and during a follow up period following relief of cough (median; 2.0 (range; 0.5 to 8.0) years after the initial visit) in a total of 20 patients with CVA (7 males and 13 females, mean ± SD age of 49.9 ± 12.9 years).
Results
Three of 8 patients not taking long-term ICS developed typical asthma compared to none of 12 patients taking ICS (p = 0.0171). PC20-FEV1 significantly (p < 0.0001) increased from 1.80 (GSEM, 1.35) to 10.7 (GSEM, 1.63) mg/ml in patients taking ICS but did not change in patients not taking ICS [2.10 (GSEM, 1.47) compared to 2.13 (GSEM, 1.52) mg/ml]. Cough threshold did not change in patients whether taking or not taking ICS.
Conclusion
Long-term ICS reduces bronchial hyperresponsiveness in CVA as recognized in typical asthma. Cough reflex sensitivity is not involved in the mechanism of cough in CVA.
==== Body
Background
Cough variant asthma is a well-known cause of chronic non-productive cough as well as gastroesophageal reflux-associated cough and post-nasal drip-induced cough [1].
Pathophysiological features of cough variant asthma [2] appear to be similar to typical asthma, with mildly increased bronchial responsiveness and eosinophilic inflammation of central and peripheral airways, and a cough responsive to bronchodilator therapy [3]. It is, however, controversial whether cough reflex sensitivity contributes to the cough in CVA [4-7].
Johnson [8] reported that a significant proportion of patients diagnosed with cough variant asthma eventually develops wheezing, sometimes severe enough to require continuous bronchodilator therapy. Corrao et al. [3] reported that 2 of 6 patients with cough variant asthma began wheezing within 18 months of completing the study. Braman [9] restudied 16 patients diagnosed with cough variant asthma 3 to 5 years previously, and found that 37% of these patients manifested intermittent wheezing during the study period. Therefore, as nearly 30% of cough variant asthma patients have been demonstrated to develop typical asthma, cough variant asthma has been recognized as a precursor of typical asthma.
In our previous study [4], long-term inhaled corticosteroids (ICS) prevented the development of typical asthma from cough variant asthma. In another of our studies [5], longitudinal decline in pulmonary function in cough variant asthma was not different from that in healthy subjects and inhaled corticosteroids had no effect on the pulmonary function decline in cough variant asthma. However, it is unknown 1) whether bronchial responsiveness and cough reflex sensitivity change after relief of cough, 2) whether inhaled corticosteroids have an beneficial effect on bronchial responsiveness and cough reflex sensitivity, and 3) whether bronchial responsiveness increases after onset of typical asthma.
Although some researchers [6] reported that cough reflex sensitivity was increased in patients with cough variant asthma, our series of studies [4,5,7] have clearly demonstrated that cough reflex sensitivity is within normal limits in cough variant asthma as well as in stable typical asthma [10]. Cough reflex sensitivity is entirely independent of bronchial responsiveness [11] and bronchomotor tone [12]. Furthermore, cough reflex sensitivity does not change immediately after a patient's cough is completely relieved on therapy within 2 months [7]. Thus, abnormal cough reflex sensitivity is not considered to be essential in cough variant asthma.
We examined longitudinal changes in bronchial responsiveness and cough reflex sensitivity and influence of ICS on both responses in patients with cough variant asthma. Bronchial responsiveness to methacholine and cough reflex sensitivity to inhaled capsaicin were measured at least two times; at the initial visit and during the follow up period after relief of cough on treatment.
Methods
Twenty patients with cough variant asthma as a single cause of chronic cough (median age 54 years, 7 men and 13 women), who had undertaken spirometry, bronchial reversibility test, methacholine provocation test, capsaicin cough provocation test, measurements of peripheral blood eosinophil count, serum total IgE and specific IgE to common allergens, and induced sputum eosinophil count at presentation, were followed up with special emphasis on typical asthma onset during 6 months or more (median 5 years, range 0.5 – 14) (Table 1). Spirometry and methacholine provocation test were repeated during the follow up period after their cough was completely relieved on the treatment.
Table 1 Clinical parameters in cough variant asthma patients with and without inhaled corticosteroids
Without ICS With ICS P value Total
Age (years) 53.3 ± 14.3 47.6 ± 12.0 0.3495 49.9 ± 12.9
Gender (male/female) 2/6 5/7 0.4439 7/13
Intreval of methcholine provocations (years) 2.7 ± 1.0 3.4 ± 2.8 0.5172 3.1 ± 2.2
2.9 (1.1–4.0)* 2.0 (0.5–8.0)* 2.0 (0.5–8.0)*
Duration of illness (months) 41.5 ± 51.9 23.6 ± 31.4 0.3466 30.8 ± 40.5
10.0 (2.0–120.0)* 12.5 (2.0–108.0)* 12.0 (2.0–120)*
Cough threshold (μM) 11.1 (1.63)** 6.2 (1.59)** 0.4163 7.8 (1.40)**
PC20-FEV1 (mg/ml) 2.13 (1.52) 1.80 (1.35) 0.7464 1.93 (1.27)
FVC (% predicted) 105.4 ± 14.3 103.1 ± 19.1 0.7765 104.0 ± 17.0
FEV1 (% predicted) 97.4 ± 15.2 93.2 ± 16.4 0.5763 94.9 ± 15.7
FEV1/FVC (%) 73.3 ± 6.9 78.1 ± 6.5 0.1318 76.2 ± 6.9
*; median (range), **; geometric mean (geometric standard error of the mean).
When the cough resolved on treatment with bronchodilators and/or inhaled and/or oral corticosteroids, we informed each patient that cough variant asthma is a precursor of typical asthma and induction of long-term inhaled corticosteroids (ICS) is desirable because the long-term therapy is recommended by many asthma guidelines in typical asthma even if the disease severity is mild. Long-term treatment with ICS was accepted and taken by 12 patients but not by the other 8 patients.
The diagnosis of cough variant asthma was made according to the following criteria proposed by Japanese Cough Research Society [13], excluding a criterion of cough reflex sensitivity within normal limits:
1) Isolated chronic non-productive cough lasting more than 8 weeks
2) Absence of a history of wheezing or dyspnea, and no adventitious lung sounds on physical examination
3) Absence of post-nasal drip to account for the cough
4) Forced expiratory volume in one second (FEV1), forced vital capacity (FVC), and FEV1/FVC ratio within normal limits (FEV1 ≥80% of predicted value, FVC ≥80% of predicted value, and FEV1/FVC ratio ≥70%)
5) Presence of bronchial hyperresponsiveness (provocative concentration of methacholine causing a 20% fall in FEV1 (PC20-FEV1) <10 mg/mL)
6) Relief of cough with bronchodilator therapy
7) No abnormal findings indicative of cough aetiology on chest roentgenogram
All patients with cough variant asthma had been successfully treated with bronchodilators and/or corticosteroids, without use of other medications such as proton pump inhibitors and histamine H1-antagonists. Thus, cough variant asthma was the single cause of chronic cough in all patients studied.
The efficacy of bronchodilator therapy described above was assessed according to the following criteria:
1) "Excellent" when cough totally resolved
2) "Good" when sleep and daytime quality of life were improved
3) "Fairly good" when severity and frequency of cough were somewhat decreased
4) "Poor" when cough was unchanged
An assessment of "Excellent" or "Good" was judged as effective.
Pulmonary function, cough reflex sensitivity, and bronchial responsiveness were measured in that order within two weeks of the first visit (Table 1), and then repeated during the follow up period after each patient's cough was completely relieved by the treatment. FVC, FEV1 and flow-volume curves were measured using a dry wedge spirometer (Chestac 11, Chest Co., Ltd., Tokyo, Japan). Spirometry was performed and evaluated according to the ATS criteria [14]. Capsaicin cough threshold (C5), a concentration of capsaicin solution eliciting 5 or more coughs, was measured as an index of cough reflex sensitivity [7,10-12]. A provocative concentration of methacholine causing a 20% or greater fall in FEV1 from prechallenge values (PC20-FEV1) was measured as an index of non-specific bronchial responsiveness [15].
The onset of typical asthma was defined as wheezing and/or dyspnoeic attack responding to bronchodilator therapy.
Data analysis
Data excluding PC20-FEV1 and C5 were presented as mean ± standard deviation (SD). PC20-FEV1 and C5 were expressed as geometric mean value with geometric standard error of the mean. Differences between groups were determined by parametric one-way analysis of variance (ANOVA) or the χ2 test. Changes within group were assessed using the paired t test. PC20-FEV1 and C5 were analyzed using logarithmically transformed values. A p value of 0.05 or less was considered significant.
Results
Typical asthma onset was recognized in 3 (37.5%) of 8 patients not taking long-term inhaled corticosteroids (ICS) and none of 12 patients taking ICS. The prevalence of asthma onset was significantly (p = 0.0214) different between the groups. Details of the 3 patients who developed typical asthma are shown in Fig. 1.
Figure 1 Longitudinal change in bronchial responsiveness in 3 patients with cough variant asthma who developed typical asthma while did not taking inhaled corticosteroids. PC20-FEV1, provocative concentration of methacholine causing a 20% or greater fall in forced expiratory volume in 1 second (FEV1), was determined by a mouth tidal breathing method. Bronchial responsiveness was not obviously increased following onset of typical asthma. ICS, inhaled corticosteroids. Arrows indicate onset of typical asthma.
Clinical parameters at initial presentation are summarized in Table 1. Median interval between the first and the second measurement of bronchial responsiveness was 2.9 years (range 1.1–4.0, mean (SD) 2.7 (1.0)) in the 8 patients not taking long-term ICS and 2.0 years (range 0.5–8.0, mean (SD) 3.4 (2.8)) in the 12 patients taking long-term ICS. The follow-up period was not significantly different between the groups. Age, gender, duration from onset of cough to presentation, capsaicin cough threshold, PC20-FEV1, FVC, FEV1 and FEV1/FVC ratio were not different between the two groups.
PC20-FEV1 significantly (p < 0.0001) increased by 5.9 (GSEM, 1.40) times from 1.80 (GSEM, 1.35) to 10.7 (GSEM, 1.63) mg/ml in patients taking ICS, but did not change in patients not taking ICS [by 0.97 (GSEM, 1.17) times from 2.13 (GSEM, 1.52) to 2.10 (GSEM, 1.47) mg/ml] (Fig. 2). PC20-FEV1 did not significantly change in the 3 patients who developed typical asthma [from 0.68 (GSEM, 1.38) to 0.89 (GSEM, 1.81) mg/ml] (Fig. 1, Fig. 2) or in 5 patients who did not develop typical asthma while they were not taking ICS [from 4.23 (GSEM, 1.46) to 3.52 (GSEM, 1.42) mg/ml] (Fig. 2). Capsaicin cough threshold (Fig. 3) or FEV1 (Fig. 4) did not change in patients whether taking or not taking ICS.
Figure 2 Longitudinal change in bronchial responsiveness in patients with cough variant asthma taking or not taking long-term inhaled corticosteroids. Closed triangles indicate patients developing typical asthma. ***p < 0.0001.
Figure 3 Longitudinal change in cough reflex sensitivity in patients with cough variant asthma taking or not taking long-term inhaled corticosteroids. Closed triangles indicate patients developing typical asthma.
Figure 4 Longitudinal change in forced expiratory volume in one second (FEV1) in patients with cough variant asthma taking or not taking long-term inhaled corticosteroids. Closed triangles indicate patients developing typical asthma.
Change in PC20-FEV1 by long-term treatment with ICS did not correlate with duration from the onset of cough to the induction of ICS treatment (r = 0.265, P = 0.4045) (Fig. 5) or duration of ICS treatment (r = 0.009, p = 0.9774) (Fig. 6).
Figure 5 Relationship between duration of illness before induction of inhaled corticosteroids and degree of improvement of bronchial hyperresponsiveness in patients with cough variant asthma taking long-term inhaled corticosteroids.
Figure 6 Relationship between duration of inhaled corticosteroid treatment and degree of improvement of bronchial hyperresponsiveness in patients with cough variant asthma taking long-term inhaled corticosteroids.
Discussion
Cough variant asthma was first described by Glauser [16]. The only presenting symptom is isolated chronic cough responsive to bronchodilator therapy. The cough can occur for many years as an extremely annoying symptom interfering with work, sleep, and quality of life. Recognition of cough variant asthma is clinically important because bronchodilator therapy is an effective antitussive in cough variant asthma. Bronchodilators usually exert no antitussive effect in other causes of isolated chronic cough [17] such as post-nasal drip-induced cough, gastroesophageal reflux-associated cough [17], and atopic cough [4,5].
Nearly 30% of cough variant asthma patients eventually develop wheezing, sometimes severe enough to require continuous treatment with bronchodilators [3-5]. In this study, wheezing was recognized in none of 12 patients taking long-term inhaled corticosteroid (ICS) therapy and in 3 of 8 patients without ICS therapy. This result confirms our previous investigation [4] that the typical asthma onset rate was significantly lower in patients receiving ICS therapy, suggesting the utility of long-term ICS as an intervention against typical asthma onset from cough variant asthma.
The present study clearly showed that bronchial responsiveness did not change after relief of cough without use of ICS, and long-tem ICS attenuated bronchial responsiveness to inhaled methacholine in patients with cough variant asthma, probably resulting in prevention of development of typical asthma from cough variant asthma. There were only 3 patients developing typical asthma whose bronchial responsiveness was more increased among the patients not taking ICS and was not obviously increased after the asthma onset as shown in Fig. 1. These findings suggest that an increased bronchial responsiveness at presentation may be a risk factor for asthma development from cough variant asthma whereas further increase in bronchial responsiveness may not be necessary for the asthma onset. It is unclear why only coughing occurs and additional wheezing appears without change in bronchial hyperresponsiveness in this eosinophilic airway disorder.
It has been shown that early induction of ICS within 2 years following asthma onset is beneficial to achieve both control of symptom and improvement of pulmonary function and bronchial responsiveness in asthma [18]. Although number of patients taking long-term ICS was small in this study, there was no significant influence of duration of illness before induction of ICS on the degree of improvement of bronchial responsiveness. Niimi et al [19] have shown that airway remodelling exists but the extent is smaller in cough variant asthma than in typical asthma. This is likely to be responsible for the lack of influence of ill duration on effect of ICS on bronchial responsiveness. Further studies are needed to clarify this issue.
Although other researchers have reported that cough reflex sensitivity was heightened and recovered to a normal level following successful treatments of cough variant asthma [20-23], it should be recognized that cough reflex sensitivity is entirely independent of bronchial responsiveness [11] or bronchomotor tone [12], and that it is within normal limits in stable typical asthma [10]. We previously showed that 14 of 64 non-asthmatic healthy subjects (21.9%) had bronchial hyperresponsiveness when PC20-FEV1 of 10 mg/ml or less was defined as bronchial hyperresponsiveness [24]. In another of our studies [11], a C5 of 1.95 μM or less, 3.9 μM or less, and 7.8 μM or less was seen in 4 (5.6%), 14 (19.7%), and 31 (43.7%) of 71 non-asthmatic healthy subjects, respectively. Considering the proportion of subjects with bronchial hyperresponsiveness, it is considered that a C5 of 3.9 μM or less to be defined as cough reflex hypersensitivity. Thus, in the present study, cough reflex sensitivity was judged to be increased at initial presentation in 2 of 8 patients (25%) not taking ICS and 6 of 12 patients (50%) receiving ICS. These findings are not consistent with our previous findings that cough reflex sensitivity was within normal limits in cough variant asthma [4,5,7,10]. Nevertheless cough reflex sensitivity did not change after relief of cough despite use of ICS in the present study, confirming our previous findings that cough reflex sensitivity did not change following successful treatment of cough variant asthma [7]. Taken together, it can be concluded that cough reflex sensitivity is not involved in the mechanism of cough in cough variant asthma even when it is increased. In other words increased cough reflex sensitivity is not a primary feature of cough variant asthma and ICS does not affect the sensitivity. We do not know why eosinophilic airway inflammation does increase cough reflex sensitivity in atopic cough but not in cough variant asthma. Precise interaction between eosinophilic airway inflammation and cough reflex sensitivity should be disclosed by future studies.
Early induction of ICS within 2 years following asthma onset has been shown to be beneficial in attenuating bronchial hyperresponsiveness as well as achieving both control of symptom and improvement of pulmonary function [18]. In this study, the degree of reduction of bronchial hyperresponsiveness with ICS did not correlate with the duration between onset of cough and induction of ICS. It is not consist with the above-mentioned result on asthma [18]. A possible explanation of this discrepancy may be that airway remodelling increasing bronchial responsiveness such as subepithelial fibrosis and smooth muscle hypertrophy does not develop or become more severe as the duration of illness is longer, while thickening of subepithelial layer has been demonstrated in cough variant asthma [19]. This possibility needs to be clarified in future studies.
Conclusion
The present retrospective study showed that bronchial hyperresponsiveness and cough reflex sensitivity did not change after relief of cough when ICS therapy was not taken in patients with cough variant asthma. A median of 2 years ICS treatment attenuated bronchial hyperresponsiveness, but not cough reflex sensitivity. Bronchial responsiveness did not further increase after onset of typical asthma in 3 patients not taking ICS. These findings suggest that long-term ICS treatment may prevent onset of typical asthma from cough variant asthma by reducing bronchial hyperresponsiveness, and that cough reflex sensitivity is not involved in mechanism of cough in cough variant asthma. Further studies including randomized placebo-controlled studies are needed to confirm the preventive effect of long-term ICS on typical asthma onset from cough variant asthma.
List of abbreviations
ANOVA = analysis of variance, C5 = provocative concentration of capsaicin eliciting 5 or more coughs, CVA = cough variant asthma, FEV1 = forced expiratory volume in one second, FVC = forced vital capacity, GSEM = geometric standard error of the mean, ICS = inhaled corticosteroids, PC20-FEV1 = provocative concentration of methacholine causing a 20% fall in forced expiratory volume in one second, SD = standard deviation,.
Acknowledgements
This study was supported in part by a grant-in-aid for Scientific Research from the Ministry of Education, Science and Culture (14570546) by the Japanese Government.
==== Refs
Irwin RS Boulet LP Cloutier MM Fuller R Gold PM Hoffstein V Ing AJ McCool FD O'Byrne P Poe RH Prakash UB Pratter MR Rubin BK Managing cough as a defence mechanism and as a symptom: a consensus panel report of the American College of Chest Physicians Chest 1998 114 133S 181S 9725800
Niimi A Amitani R Suzuki K Tanaka E Murayama T Kuze F Eosinophilic inflammation in cough variant asthma Eur Respir J 1998 11 1064 1069 9648956 10.1183/09031936.98.11051064
Corrao WM Braman SS Irwin RS Chronic cough as the sole presenting manifestation of bronchial asthma N Engl J Med 1979 300 633 637 763286
Fujimura M Ogawa H Nishizawa Y Nishi K Comparison of atopic cough with cough variant asthma: Is atopic cough a precursor of asthma? Thorax 2003 58 14 18 12511712 10.1136/thorax.58.1.14
Fujimura M Nishizawa Y Nishitsuji M Abo M Kita T Nomura S Longitudinal decline in pulmonary function in atopic cough and cough variant asthma Clin Exp Allergy 2003 33 588 594 12752586 10.1046/j.1365-2222.2003.01658.x
Orejas GC Pascual PT Alzueta AA Cough-variant asthma. Clinical and functional characteristics. Report of 63 cases Arch Bronconeumol 1998 34 232 236 9656061
Fujimura M Kamio Y Hashimoto T Matsuda T Cough receptor sensitivity and bronchial responsiveness in patients with only chronic nonproductive cough: in view of effect of bronchodilator therapy J Asthma 1994 31 463 472 7961323
Johnson D Osborn LM Cough variant asthma: a review of the clinical literature J Asthma 1991 28 85 90 1672866
Braman SS Corrao WM Chronic cough; diagnosis and treatment Primary Care 1985 12 217 225 3848018
Fujimura M Sakamoto S Kamio Y Matsuda T Cough receptor sensitivity and bronchial responsiveness in normal and asthmatic subjects Eur Respir J 1992 5 291 295 1572440
Fujimura M Kasahara K Yasui M Myou S Ishiura Y Kamio Y Hashimoto T Matsuda T Atopy in cough sensitivity to capsaicin and bronchial responsiveness in young women Eur Respir J 1998 11 1060 1063 9648955 10.1183/09031936.98.11051060
Fujimura M Sakamoto S Kamio Y Matsuda T Effects of methacholine-induced bronchoconstriction and procaterol-induced bronchodilation on cough receptor sensitivity to inhaled capsaicin and tartaric acid Thorax 1992 47 441 445 1386691
Fujimura M Uchida Y Niimi A Fujimura M Guidelines for Diagnosis and Treatment of Chronic Cough 2003 2 Kita Media: Tokyo 1 35 (in Japanese)
American Thoracic Society Standardization of spirometry-1987 update Am Rev Respir Dis 1987 136 1285 1298 3674589
Cockcroft DW Ruffin RE Dolovich J Hargreave FE Allergen-induced increase in non-allergic bronchial reactivity Clin Allergy 1977 7 503 513 589783
Glauser FL Variant asthma Ann Allergy 1972 30 457 459 5044498
Irwin RS Corrao WM Pratter MR Chronic persistent cough in the adult: the spectrum and frequency of causes and successful outcome of specific therapy Am Rev Respir Dis 1981 123 413 417 7224353
Pauwels RA Pedersen S Busse WW Tan WC Chen YZ Ohlsson SV Ullman A Lamm CJ O'Byrne PM START Investigators Group: Early intervention with budesonide in mild persistent asthma: a randomised, double-blind trial Lancet 2003 361 1071 1076 12672309 10.1016/S0140-6736(03)12891-7
Niimi A Matsumoto H Minakuchi M Kitaichi M Amitani R Airway remodelling in cough-variant asthma Lancet 2000 356 564 565 10950236 10.1016/S0140-6736(00)02584-8
McGarvey LP Heaney LG Lawson JT Johnston BT Scally CM Ennis M Shepherd DR MacMahon J Evaluation and outcome of patients with chronic non-productive cough using a comprehensive diagnostic protocol Thorax 1998 53 738 743 10319055
Doherty MJ Mister R Pearson MG Calverley PM Capsaicin responsiveness and cough in asthma and chronic obstructive pulmonary disease Thorax 2000 55 643 649 10899239 10.1136/thorax.55.8.643
Dicpinigaitis PV Dobkin JB Reichel J Antitussive effect of the leukotriene receptor antagonist zafirlukast in subjects with cough-variant asthma J Asthma 2002 39 291 297 12095178 10.1081/JAS-120002285
Nieto L de Diego A Perpina M Compte L Garrigues V Martinez E Ponce J Cough reflex testing with inhaled capsaicin in the study of chronic cough Respir Med 2003 9 393 400 12693800 10.1053/rmed.2002.1460
Fujimura M Kamio Y Hashimoto T Matsuda T Log normal distribution of bronchial responsiveness to methacholine in normal young adults Jpn J Physiol 1993 43 541 552 8114362 10.2170/jjphysiol.43.541
|
16270932
|
PMC1277007
|
CC BY
|
2021-01-04 16:37:44
|
no
|
Cough. 2005 Aug 25; 1:5
|
utf-8
|
Cough
| 2,005 |
10.1186/1745-9974-1-5
|
oa_comm
|
==== Front
CoughCough (London, England)1745-9974BioMed Central London 1745-9974-1-61627093310.1186/1745-9974-1-6ResearchComparison of cough reflex sensitivity after an inhaled antigen challenge between actively and passively sensitized guinea pigs Hara Johsuke [email protected] Masaki [email protected] Shigeharu [email protected] Yoshitaka [email protected] Shiho [email protected] Toshiyuki [email protected] Nobuyuki [email protected] Miki [email protected] Noriyuki [email protected] Yoriko [email protected] Akihiro [email protected] Yoshihisa [email protected] Kouichi [email protected] Haruhiko [email protected] Masahide [email protected] Kazuo [email protected] Shinji [email protected] Respiratory Medicine, Cellular Transplantation Biology, Kanazawa University Graduate School of Medical Science, 13-1, Takara-machi, Kanazawa City, Ishikawa, 920-8641, Japan2005 6 9 2005 1 6 6 6 7 2005 6 9 2005 Copyright © 2005 Hara et al; licensee BioMed Central Ltd.2005Hara et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Late asthmatic response is observed following antigen challenge in actively, but not passively, sensitized guinea pigs. Although cough reflex sensitivity is increased after antigen challenge in actively sensitized guinea pigs, it is unknown whether the antigen-induced increase in cough reflex sensitivity develops in passively sensitized animals. The aim of this study was to compare the cough reflex sensitivity to inhaled capsaicin after an inhaled antigen challenge between actively and passively sensitized guinea pigs.
Methods
Measurement of number of coughs elicited by increasing concentrations of capsaicin (10-6 and 10-4 M) and bronchial responsiveness to ascending concentrations of methacholine, and analysis of bronchoalveolar lavage fluid (BALF) were separately performed 24 h after an antigen challenge in actively and passively sensitized guinea pigs.
Results
Percentage of eosinophils in BALF and bronchial responsiveness to methacholine were increased 24 h after the antigen challenge in both actively and passively sensitized animals compared with saline-challenged actively and passively sensitized animals, respectively. Absolute number of eosinophils in BALF from actively sensitized and antigen-challenged guinea pigs was significantly greater than that from passively sensitized and antigen-challenged animals. Cough response to capsaicin and concentration of substance P in BALF were increased 24 h after the antigen challenge in actively sensitized guinea pigs, but not in passively sensitized guinea pigs. Bronchial responsiveness, cough reflex sensitivity and substance P concentration and total cells in BALF were increased in actively sensitized and saline challenged guinea pigs compared with passively sensitized and saline challenged animals.
Conclusion
The results suggest that active sensitization per se increases cough reflex sensitivity accompanied by increased inflammatory cells and substance P level in BALF, and antigen challenge further increases them, while simple IgE- and/or IgG-mediated allergic reaction per se or the low intensity of eosinophil infiltration in the airway itself may not affect cough reflex sensitivity in guinea pigs.
==== Body
Background
Chronic cough is a common and distressing symptom. Eosinophilic airway disorders such as eosinophilic bronchitis without asthma [1] and atopic cough [2] are important causes of the chronic cough. In these disorders, cough reflex sensitivity is heightened while patients are coughing and becomes normal on successful treatment [3]. Knowledge of the detailed pathogenesis is needed to understand the mechanism and to develop better treatment of the disorders.
We have shown in actively sensitized guinea pigs that cough reflex sensitivity is increased 24 h after an inhaled antigen challenge, which is not mediated by bronchoconstriction [4]. Allergic reaction and cough hypersensitivity may be induced by chemical mediators such as histamine [5], prostaglandins [6], thromboxane A2 (TXA2) [4], and platelet activating factor (PAF), which are released from mast cells activated by IgE antibody and/or production of Th2 cytokines [7] such as IL-4, IL-5 and IL-13. On the other hand, simple IgE- and/or IgG-mediated allergic airway reaction occurs when passively sensitized guinea pigs are challenged with an aerosolized antigen. It is, however, unknown whether the simple IgE- and/or IgG-mediated allergic airway reaction can increase cough reflex sensitivity. To elucidate this, we compared the cough reflex sensitivity to inhaled capsaicin after an inhaled antigen challenge between actively and passively sensitized guinea pigs.
Methods
Animals
Male, albino, Hartley-strain guinea pigs were obtained from Sankyou Laboratory Service (Toyama, Japan). They were quarantined in the Animal Research Center of Kanazawa University. All the animal procedure in this study complied with the standards set out in the Guideline for the Care and Use of Laboratory Animals at the Takara – machi Campus of Kanazawa University.
Study design
In order to avoid possible interaction between capsaicin-induced cough, methacholine-induced bronchoconstriction and BALF contents, measurement of cough reflex sensitivity to inhaled capsaicin, measurement of bronchial responsiveness to inhaled methacholine and BAL were separately carried out 24 hours after an aerosolized antigen challenge in actively and passively sensitized guinea pigs.
Active sensitization and antigen challenge
Actively sensitized guinea pigs were assigned into two groups: saline challenge (A-OA/Sal) and OA challenge (A-OA/OA) groups (n = 8 for each group). Animals in A-OA/Sal group were challenged with aerosolized saline, and A-OA/OA group with aerosolized antigen. Guinea pigs weighing 200 to 220 g each were actively sensitized by the method reported by Muraki et al [8]. Animals were given an intraperitoneal administration of 2.0 mg of ovalbumin (OA) and 100 mg of aluminum hydroxide [Al(OH)3] 2 days after an intraperitoneal administration of 30 mg/kg cyclophosphamide. Three weeks later, boosting was carried out by intraperitoneal administration of 0.01 mg of OA and 100 mg of Al(OH)3. Three weeks after the boosting, actively sensitized guinea pigs were challenged with an aerosolized OA solution under spontaneous breathing at 20 min after an intraperitoneal administration of diphenhydramine (20 mg/kg) to avoid acute anaphylactic respiratory distress. Conscious guinea pigs were placed in a dual chamber plethysmograph (head chamber volume, 1520 ml) (model PMUA + SAR, Buxco Electronics, Sharon, CT). Animals were challenged with 10 mg/ml OA aerosol for 90 s (head chamber only, 0.08 ml/min output). The aerosol was generated by a Devilbiss 646 nebulizer (Devilbiss Co., Somerset, PA) operated by compressed air at 7.57 L/min (Minipon 54B-588, Origin Medical Industry Co., Ltd., Tokyo, Japan).
Passive sensitization and antigen challenge
Guinea pig homocytotropic antiserum was obtained by the method elaborated in Santives et al. [9]. Briefly, 500 μg of ovalbumin (OA) was emulsified in Freund's complete adjuvant and injected intradermally into each guinea pig at multiple sites. A booster dose was prepared and administered in the same manner 2 weeks later. Serum collected from each animal 2 weeks after the booster dose was pooled, and kept frozen until use. The antibody titre of this serum was 1:12,800, 1:6,400 and 1:512, as estimated by passive cutaneous anaphylaxis at 4 h, 24, and 7 days, respectively. Normal guinea pigs were passively sensitized with 1.0 mL/kg antiserum intraperitoneally.
Passively sensitized guinea pigs weighing 450 to 500 g were assigned into two groups: saline challenge (P-OA/Sal) and OA challenge (P-OA/OA) groups (n = 8 for each group). Animals in P-OA/Sal group were challenged with aerosolized saline, and P-OA/OA group with aerosolized antigen. One week after the passive sensitization, guinea pigs were challenged with an aerosolized OA solution (10 mg/mL) under spontaneous breathing at 20 min after an intraperitoneal administration of diphenhydramine (20 mg/kg). OA challenge to passively sensitized guinea pigs was carried out by the same method used in actively sensitized model.
Cough reflex sensitivity
Cough reflex sensitivity was measured 24 h after challenge with either OA or saline in both actively and passively sensitized guinea pigs. Each conscious guinea pig was placed in an airtight custom-built transparent plastic box consisting of a head chamber (1600 ml volume) isolated from a body chamber, and pressure in the body chamber was recorded. Coughs were detected as a change in the pressure (a rapid inspiration followed by rapid expiration). To disregard motion- and sneezing-related changes in the pressure, movements of the guinea pigs were visually monitored. Coughs were counted by a trained observer and recognized by the characteristic animal posture and the pressure transducer recordings. Increasing concentrations of capsaicin solution (10-6, 10-4 M) were inhaled for 2 min from a Devilbiss 646 nebulizer (Devilbiss Co., Somerset, PA) operated by compressed air at 1.6 l/min (Iwaki Air Pump AP-115AN, Iwaki Co., Ltd., Tokyo, Japan). The nebulizer output was 0.037 ml/min. The number of coughs was counted during a 2 min inhalation of each capsaicin solution and for additional 1 min. The total number of coughs during the 3 – min period was recorded on the inhalation of each concentration of capsaicin.
Bronchial responsiveness
Bronchial responsiveness to inhaled methacholine was measured 24 h after challenge with either OA or saline in both actively and passively sensitized guinea pigs. Guinea pigs were anesthetized by an intraperitoneal injection of 75 mg/kg of sodium pentobarbital and placed in a supine position. After the trachea was cannulated with a polyethylene tube (outside diameter, 2.5 mm; inside diameter, 2.1 mm), the animals were artificially ventilated using a small animal respirator (model 1680, Harvard Apparatus Co., Inc., South Natick, MA) adjusted to a tidal volume 10 ml/kg at a rate of 60 strokes/min. Ascending concentrations of methacholine solution (50, 100, 200, 400 μg/ml) were delivered for 20 s by an ultrasonic nebulizer (NE-U06, Omron, Kyoto, Japan) at 5 min intervals. The nebulizer generated the aerosol at a rate of 15.2 μl / min. The changes in lung resistance to insufflation, the lateral pressure of the tracheal tube (pressure at the airway opening abbreviated as Pao: cmH2O), were measured using a differential pressure transducer (model TP-603T, Nihon Koden Kogyo Co., Ltd., Tokyo, Japan). The change in Pao represents the average of the changes in pulmonary resistance (RL) and reciprocal dynamic lung compliance (1/Cdyn) [10].
Bronchoalveolar lavage (BAL)
BAL was performed 24 h after challenge with either the antigen or saline in both actively and passively sensitized guinea pigs without capsaicin or methacholine provocation. Guinea pigs were anesthetized and prepared by the same method described in the measurement of bronchial responsiveness. Through the tracheal cannula the lungs were lavaged with 10 ml of saline 2 times (total: 20 ml). The cells in BAL fluid (BALF) were stained with Turk solution and counted in duplicate in a hemocytometer (in a Burker chamber). Differential cell counts were made on a smear prepared by cytocentrifuge and stained with Wright-Giemsa.
The concentration of substance P in BALF was measured using a commercial enzyme immunoassay (EIA) kit (Cayman Chemical Company, USA). This kit is a competitive assay that provides accurate measurements of substance P with a working range of 3.9 to 500 pg/ml.
Preparation of drugs
The following chemicals were used: sodium pentobarbital (Abbott Laboratories, North Chicago, IL), methacholine (Wako Pure Chemical Ind., Osaka, Japan), diphenhydramine (Wako Pure Chemical Ind.), ovalbumin (Sigma, St. Louis, MO), Al(OH)3 (Wako Pure Chemical Ind.), dimethyl sulfoxide (Wako Pure Chemical Ind.), physiological saline (Otsuka Pharmaceutical Co., Ltd., Osaka, Japan), capsaicin (Sigma), cyclophosphamide (Shionogi Co., Ltd., Osaka, Japan).
Statistical analysis
All data are shown as mean ± standard error of the mean (SEM). Statistical differences were determined by analysis of variance (ANOVA) followed by Fisher's protected test significant differences (Statview; SAS Institute, Cary, NC, USA). A P value less than 0.05 was considered statistically significant.
Results
Cough reflex sensitivity
Fig. 1 shows the number of coughs induced by inhaled capsaicin in actively and passively sensitized guinea pigs. The number of coughs elicited by an aerosol of capsaicin (10-4 M) was significantly increased in A-OA/OA group (8.3 ± 0.9), but not in P-OA/OA group (2.3 ± 0.8), compared with each saline-challenged group (A-OA/Sal; 4.8 ± 0.6, P-OA/Sal; 1.8 ± 0.7).
Figure 1 Bronchial responsiveness
Bronchial responsiveness to inhaled methacholine in actively and passively sensitized guinea pigs are shown in Fig. 2. In the both groups, pressure at the airway opening (Pao) was dose-dependently increased by methacholine. The bronchial responsiveness in A-OA/OA (Percent increase in Pao from baseline value; 20.1 ± 16.5 %, 180.1 ± 30.5 %, 479.4 ± 89.2 %, 709.3 ± 99.8 % in 50, 100, 200, 400 μg/ml of inhaled methacholine) and P-OA/OA (51.1 ± 19.7 %, 364.7 ± 141.5 %, 637.4 ± 119.9 %, 717.2 ± 100.8 % in each concentration of methacholine) group was significantly heightened when compared with that in A-OA/Sal (5.5 ± 2.7 %, 87.2 ± 29.3 %, 182.9 ± 35.5 %, 529.1 ± 110.2 % in each concentration of methacholine) and P-OA/Sal (6.1 ± 2.8 %, 97.2 ± 61.4 %, 272.3 ± 94.5 %, 596.8 ± 64.2 % in each concentration of methacholine) group, respectively.
Figure 2 BALF analysis
The percentage of eosinophils in BALF was significantly increased in both A-OA/OA and P-OA/OA group compared with A-OA/Sal and P-OA/Sal group, respectively. The total number of cells and eosinophils in BALF collected from A-OA/OA group were significantly increased compared with those from A-OA/Sal and P-OA/OA groups. The number of eosinophils in BALF collected from P-OA/OA group was significantly increased compared with those from P-OA/Sal group. There was no significant difference in the total number of cells between P-OA/OA and P-OA/Sal groups (Table 1).
Table 1 BAL fluid cell findings 24 h after an antigen inhalation in guinea pigs.
Absolute number (cells/mL) Percentage (%)
Total cells (103) Nac (103) Neu (103) Lym (103) Eos Mac Neu Lym Eos
AP group 201.0 ± 70.9*# 73.8 ± 18.7# 3.9 ± 1.5 3.3 ± 1.7# 120.0 ± 49.8*# 39.8 ± 4.8*# 2.1 ± 0.8 1.4 ± 0.3 56.7 ± 4.4*#
AN group 123.0 ± 28.8 93.2 ± 18.7 1.3 ± 1.1 2.6 ± 1.3 26.0 ± 11.7 77.9 ± 5.4 1.1 ± 0.9 1.9 ± 0.6 19.1 ± 6.3
PP group 60.5 ± 14.5 37.5 ± 3.1 2.1 ± 0.6 1.1 ± 0.2 19.9 ± 2.4$ 61.6 ± 3.1$ 3.9 ± 1.2 1.8 ± 0.4 32.3 ± 2.9$
PN group 57.6 ± 17.5 47.0 ± 5.2 1.8 ± 1.1 0.9 ± 0.2 7.9 ± 1.6 81.1 ± 2.4 3.8 ± 2.6 1.7 ± 0.5 13.4 ± 2.4
OA; ovalbumin, Sal: saline, A-OA/OA; OA inhalation in actively sensitized animals, A-OA/Sal; saline inhalation in actively sensitized animals, P-OA/OA; OA inhalation in passively sensitized animals, P-OA/Sal; saline inhalation in passively sensitized animals, Mac; macrophages, Neu; neutrophils, Lym; lymphocytes, Eos; eosinophils.
*P < 0.01 compared with the A-OA/Sal, #P < 0.01 compared with the P-OA/OA group, $P < 0.01 compared with the P-OA/Sal group.
Fig. 3 shows the concentration of substance P in BALF. The concentration of substance P was significantly increased in A-OA/OA (15.9 ± 1.6 pg/ml) group compared with A-OA/Sal group (11.5 ± 1.2 pg/ml). The concentrations of substance P in P-OA/OA and P-OA/Sal groups were lower than 3.9 pg/ml.
Figure 3 Discussion
The present study confirmed other researchers' investigation that active sensitization per se induces airway eosinophilic inflammation and increase in cough reflex sensitivity [11] and our previous data [5] that an aerosolized antigen challenge further enhances the airway responses in actively sensitized animals. We showed for the first time that cough reflex sensitivity was unchanged following an antigen challenge in passively sensitized guinea pigs while BAL eosinophils and bronchial responsiveness to methacholine were increased compared with saline challenged animals. In addition, substance P level in BAL fluid was increased in actively sensitized guinea pigs and further increased after an antigen challenge, but the level was below that measured in passively sensitized animals in spite of antigen challenge. These findings suggest that antigen-antibody reaction in the airway is insufficient to modulate cough reflex sensitivity. In other words, airway inflammatory processes such as cell and mediator response following antigen-antibody reaction may be important in increasing cough reflex sensitivity associated with increased levels of substance P.
Although BAL eosinophils and bronchial responsiveness were increased after antigen challenge in passively sensitized guinea pigs, cough reflex sensitivity and substance P levels in BAL fluid were unchanged. Airway eosinophil infiltration may not be essential in increasing cough reflex sensitivity. We previously reported that cough reflex sensitivity was not increased in patients with cough variant asthma complaining of daily coughing [3] and stable asthmatics [12], in both of whom eosinophilic airway inflammation is characteristic. Furthermore, Minoguchi et al. [13] suggested that cough reflex sensitivity to capsaicin is not associated with eosinophilic inflammation of the airway in patients with allergic asthma because antigen challenge did not influence cough reflex sensitivity to capsaicin. On the other hand, we have shown that challenge with environmental fungal antigen causes symptomatic cough accompanied by an increase in cough reflex sensitivity in patients with atopic cough [14-18]. We do not know why antigen challenge increases cough reflex sensitivity in atopic cough, but not in asthma. At least, airway allergic reactions other than airway eosinophil infiltration may be involved in increasing cough reflex sensitivity accompanied by an increase in substance P levels in the airway. The detailed mechanism should be disclosed in further studies.
In the present study, the concentration of substance P in BALF was increased 24 h after an antigen challenge in actively sensitized guinea pigs, but not in passively sensitized animals. Substance P has been considered as an important neuropeptide in the cough reflex pathway because tachykinin antagonists partially block the cough reflex. Neutral endopeptidase (NEP) has been recognized as the major enzyme degrading substance P [19]. We previously reported that NEP activity in tracheal tissue was decreased after an antigen challenge and the antigen-induced NEP inactivation might increase the cough response to capsaicin in actively sensitized guinea pigs [20]. In this respect, it is likely that the retention of NEP activity might be responsible for the lack of development of antigen-induced cough hypersensitivity in passively sensitized guinea pigs. Further studies are needed to clarify this possibility.
The receptor for capsaicin, termed vanilloid receptor-1 (VR-1), is expressed in guinea pigs. VR-1 mediates cough induced by capsaicin [21]. An increased expression of VR-1 has also been reported in humans with chronic cough [22]. In the airway of the guinea pig, VR-1 has been shown to be activated by a decrease in pH [23]. Recently, we reported that the pH of BALF was decreased in actively sensitized guinea pigs [24]. Therefore, in the airways of actively sensitized guinea pigs, but not of passively sensitized guinea pigs, the acid environment or epithelial damage might induce an increase in the number of VR-1 contributing to the enhanced cough reflex sensitivity. This possibility should be examined in future studies.
Simple IgE- and/or IgG-mediated allergic reactions induce eosinophilic infiltration in the airway and bronchial hyperresponsiveness to methacholine, but not cough hypersensitivity to capsaicin, in passively sensitized animals. The same difference between active and passive sensitization of animals is investigated concerning the antigen-induced late asthmatic response (LAR): LAR develops in actively, but not passively, sensitized guinea pigs [25]. It is suggested that the simple IgE- and/or IgG-mediated allergic reaction cannot induce cough hypersensitivity. It is likely that complex allergic inflammatory reaction in the airway such as interaction between resident and recruited cells, mediators and cytokines is involved in the antigen-induced increase in cough reflex sensitivity as well as LAR. Future studies are required to elucidate the involvement of each possible contributor.
In conclusion, we compared the cough reflex sensitivity to inhaled capsaicin 24 h after an inhaled antigen challenge between actively and passively sensitized guinea pigs. The cough reflex sensitivity and substance P level in BALF were increased in actively sensitized guinea pigs, and further increased 24 h after an antigen challenge. On the other hand, the cough reflex sensitivity or BALF substance P level was not increased after an antigen challenge in passively sensitized animals, while bronchial hyperresponsiveness and airway eosinophilia in BAL were induced by the antigen challenge in both actively and passively sensitized animals. These results suggest that simple IgE- and/or IgG-mediated allergic reaction per se or eosinophilic infiltration in the airway itself may not affect the cough reflex sensitivity or neuropeptide metabolism in guinea pigs, and that cough reflex sensitivity and bronchial responsiveness are modulated by a different mechanism. A complex allergic reaction in the airway may be involved in the development of antigen-induced increase in cough reflex sensitivity.
Acknowledgements
This study was supported in part by a grant-in-aid for Scientific Research from the Ministry of Education, Science and Culture (14570546) by the Japanese Government.
==== Refs
Brightling CE Ward R Wardlaw AJ Pavord ID Airway inflammation, airway responsiveness and cough before and after inhaled budesonide in patients with eosinophilic bronchitis Eur Respir J 2000 15 682 686 10780759 10.1034/j.1399-3003.2000.15d10.x
Fujimura M Ogawa H Nishizawa Y Nishi K Comparison of atopic cough with cough variant asthma: is atopic cough a precursor of asthma? Thorax 2003 58 14 18 12511712 10.1136/thorax.58.1.14
Fujimura M Kamio Y Hashimoto T Matsuda T Cough receptor sensitivity and bronchial responsiveness in patients with only chronic nonproductive cough: in view of effect of bronchodilator therapy J Asthma 1994 31 463 472 7961323
Liu Q Fujimura M Tachibana H Myou S Kasahara K Yasui M Characterization of increased cough sensitivity after antigen challenge in guinea pigs Clin Exp Allergy 2001 31 474 484 11260161 10.1046/j.1365-2222.2001.00989.x
Liu Q Effect of Histamine H1 receptor antagonists on antigen-induced increase of cough sensitivity in guinea pigs. Allergology Intenational 2000 49 237 243 10.1046/j.1440-1592.2000.00189.x
Shinagawa K Kojima M Ichikawa K Hiratochi M Aoyagi S Akahane M Participation of thromboxane A(2) in the cough response in guinea-pigs: antitussive effect of ozagrel Br J Pharmacol 2000 131 266 270 10991919 10.1038/sj.bjp.0703553
Myou S Fujimura M Kurashima K Kita T Tachibana H Ishiura Y Abo M Nakao S Effects of suplatast tosilate, a new type of anti-allergic agent, on airway cough hypersensitivity induced by airway allergy in guinea-pigs Clin Exp Allergy 2001 31 1939 1944 11737047 10.1046/j.1365-2222.2001.01241.x
Muraki M Tohda Y Sugihara R Nagasaka Y Nakajima S The effect of TYB-2285 on dual phase bronchoconstriction and airway hypersensitivity in guinea-pigs actively sensitized with ovalbumin J Pharm Pharmacol 1994 46 883 886 7897592
Santives T Roska AK Hensley GT Moore VL Fink JN Abramoff P Immunologically induced lung disease in guinea pigs. A comparison of ovalbumin and pigeon serum as antigens J Allergy Clin Immunol 1976 57 582 594 932343 10.1016/0091-6749(76)90010-5
Myou S Fujimura M Kurashima K Tachibana H Watanabe K Hirose T Type 1 angiotensin II receptor antagonism reduces antigen-induced airway reactions Am J Respir Crit Care Med 2000 162 45 49 10903218
Featherstone RL Hutson PA Holgate ST Church MK Active sensitization of guinea-pig airways in vivo enhances in vivo and in vitro responsiveness Eur Respir J 1988 1 839 845 3229483
Fujimura M Sakamoto S Kamio Y Matsuda T Cough receptor sensitivity and bronchial responsiveness in normal and asthmatic subjects Eur Respir J 1992 5 291 295 1572440
Minoguchi H Minoguchi K Tanaka A Matsuo H Kihara N Adachi M Cough receptor sensitivity to capsaicin does not change after allergen bronchoprovocation in allergic asthma Thorax 2003 58 19 22 12511713 10.1136/thorax.58.1.19
Ogawa H Fujimura M Amaike S Nishiura Y Nakagawa-Yoshida K Suga M Ando M Matsuda T Seasonal chronic cough with sputum eosinophilia caused by Trichosporon cutaneum (Trichosporon asahii) Int Arch Allergy Immunol 1998 116 162 165 9652310 10.1159/000023940
Ogawa H Fujimura M Myou S Kitagawa M Matsuda T Eosinophilic tracheobronchitis with cough hypersensitivity caused by Streptomyces albus antigen Allergology Intenational 2000 49 83 87 10.1046/j.1440-1592.2000.00157.x
Ogawa H Fujimura M Tofuku Y Isolated chronic cough with sputum eosinophilia caused by Humicola fuscoatra antigen: the importance of environmental survey for fungus as an etiologic agent J Asthma 2002 39 331 336 12095183 10.1081/JAS-120002290
Ogawa H Fujimura M Tofuku Y Two cases of atopic cough successfully treated by oral cleaning with amphotericin B: relation to Basidiomycetes detected from pharyngeal swab Allergology Intenational 2004 53 193 196 10.1111/j.1440-1592.2004.00313.x
Ogawa H Fujimura M Tofuku Y Treatment of atopic cough caused by Basidiomycetes antigen with low-dose itraconazol Lung 2004 182 279 284 15742240 10.1007/s00408-004-2509-0
Sekizawa K Jia YX Ebihara T Hirose Y Hirayama Y Sasaki H Role of substance P in cough Pulm Pharmacol 1996 9 323 328 9232670 10.1006/pulp.1996.0042
Katayama N Fujimura M Ueda A Kita T Abo M Tachibana H Myou S Kurashima K Effects of carbocysteine on antigen-induced increases in cough sensitivity and bronchial responsiveness in guinea pigs J Pharmacol Exp Ther 2001 297 975 980 11356919
Jia Y McLeod RL Wang X Parra LE Egan RW Hey JA Anandamide induces cough in conscious guinea-pigs through VR1 receptors Br J Pharmacol 2002 137 831 836 12411414 10.1038/sj.bjp.0704950
Groneberg DA Niimi A Dinh QT Cosio B Hew M Fischer A Chung KF Increased expression of transient receptor potential vanilloid-1 in airway nerves of chronic cough Am J Respir Crit Care Med 2004 170 1276 1280 15447941 10.1164/rccm.200402-174OC
Kollarik M Undem BJ Mechanisms of acid-induced activation of airway afferent nerve fibres in guinea-pig J Physiol 2002 543 591 600 12205192 10.1113/jphysiol.2002.022848
Oribe Y Fujimura M Kita T Katayama N Nishitsuji M Hara J Myou S Nakao S Attenuating effect of H+K+ATPase inhibitors on airway cough hypersensitivity induced by allergic airway inflammation in guinea-pigs Clin Exp Allergy 2005 35 262 267 15784101 10.1111/j.1365-2222.2005.02158.x
Saito M Fujimura M Miyake Y Sakamoto S Yasui M Kurashima K Mastuda T [Late asthmatic response by recruitment of homocytotropic antibody from circulating blood to the airway wall in guinea pig] Arerugi 1991 40 82 6 (in Japanese) 2029224
|
16270933
|
PMC1277008
|
CC BY
|
2021-01-04 16:37:44
|
no
|
Cough. 2005 Sep 6; 1:6
|
utf-8
|
Cough
| 2,005 |
10.1186/1745-9974-1-6
|
oa_comm
|
==== Front
CoughCough (London, England)1745-9974BioMed Central London 1745-9974-1-71627093710.1186/1745-9974-1-7ReviewCough: are children really different to adults? Chang Anne B [email protected] Paediatric Respiratory and Sleep Physician, NHMRC Practitioner Fellow, Associate Professor in Paediatrics and Child Health, Dept of Respiratory Medicine, Royal Children's Hospital, Herston Rd, Brisbane, Queensland 4029, Australia2005 20 9 2005 1 7 7 6 7 2005 20 9 2005 Copyright © 2005 Chang; licensee BioMed Central Ltd.2005Chang; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Worldwide paediatricians advocate that children should be managed differently from adults. In this article, similarities and differences between children and adults related to cough are presented. Physiologically, the cough pathway is closely linked to the control of breathing (the central respiratory pattern generator). As respiratory control and associated reflexes undergo a maturation process, it is expected that the cough would likewise undergo developmental stages as well. Clinically, the 'big three' causes of chronic cough in adults (asthma, post-nasal drip and gastroesophageal reflux) are far less common causes of chronic cough in children. This has been repeatedly shown by different groups in both clinical and epidemiological studies. Therapeutically, some medications used empirically for cough in adults have little role in paediatrics. For example, anti-histamines (in particular H1 antagonists) recommended as a front-line empirical treatment of chronic cough in adults have no effect in paediatric cough. Instead it is associated with adverse reactions and toxicity. Similarly, codeine and its derivatives used widely for cough in adults are not efficacious in children and are contraindicated in young children. Corticosteroids, the other front-line empirical therapy recommended for adults, are also minimally (if at all) efficacious for treating non-specific cough in children. In summary, current data support that management guidelines for paediatric cough should be different to those in adults as the aetiological factors and treatment in children significantly differ to those in adults.
==== Body
Introduction
To health care professionals who work with them, children are clearly different to adults but this seems less obvious to some. "Children swallow just like adults", remarked an academic speech pathologist when commenting on dysphagia and cough. "Children are the same as adults. It's just the behaviour that is different", remarked another specialist. Paediatricians world-wide passionately advocate that childhood illnesses should be managed differently to adults as extrapolation of adult based data to children can result in unfavourable consequences [1,2]. This article provides an update on paediatric issues on cough and highlights the differences between adults and children that are relevant to cough.
Physiology
Central and peripheral cough pathway
The central pathway for cough is a brainstem reflex linked to control of breathing (the central respiratory pattern generator) [3], which undergoes a maturation process such that the reference values for normal respiratory rate in children are different to those in adults [4] and reaches adult values in adolescence. In early life, cough is related to primitive reflexes (laryngeal chemoreflex), that undergo maturation resulting in significant differences in swallowing between young children and adults [5]. Plasticity (modulation) of the cough reflex has been shown [3,6], although it is unknown if the young have greater plasticity (propensity to modulate or change). Like other organs directly relevant to cough (eg the systemic and mucosal immune system) [7,8] or not directly related to cough (eg the renal system), one can speculate that the cough reflex has maturational differences as well. Indeed children differ from adults in some immunological response to lipopolysaccharides [9]. Also, children, especially their neurological system, are more sensitive than adults to certain environmental exposures [10]. For example, in children, the utility of CT scans has to be balanced with the reported increased lifetime cancer mortality risk, which is age and dose dependent. Although the risk is relatively negligible, children have 10 times increased risk compared to middle aged adults [10]. Lastly, the distinct differences in respiratory physiology and neuro-physiology between young children and adults include maturational differences in airway, respiratory muscle and chest wall structure, sleep characteristics, respiratory reflexes and respiratory control [11-13].
Cortical control of cough and psychological determinants
Cough can be cortically modulated [14]. In adults, chronic cough is associated with anxiety as an independent factor [15]; such data are unavailable in children. Adults seeking medical attention are primarily self-driven but in children, parental and professional expectations influence consulting rates and prescription of medications [16-18]. Reporting of childhood respiratory symptoms is biased and parental perception of childhood cough plays an important role [19,20]. In asthma, parental psychosocial factors (in particular anxiety) were strongest predictors for emergency attendances for children whereas in adults, asthma severity factors were the risk factors [21]. In cough, use of cough medications and presentation to doctors were less likely in children with higher educated mothers [22]. Hutton and colleagues' described "parents who wanted medicine at the initial visit reported more improvement at follow-up, regardless of whether the child received drug, placebo, or no treatment" [23]. Rietveld and colleagues showed that children were more likely to cough under certain psychological settings [24,25].
Clinical evaluation of cough
What is 'normal' or expected?
'Normal' children occasionally cough as described by two studies that objectively measured cough frequency [26,27]. Normal children without a preceding upper respiratory infection in the last 4 weeks have up to 34 cough epochs per 24 hours [26]. In another study, 0–141 cough epochs/24 hours (median 10) were recorded in 'controls' (these children were considered well by parents and attending school and were age, gender and season matched [27]). Medicalisation of an otherwise common symptom can foster exaggerated anxiety about perceived disease and lead to unnecessary medical products and service [28]. Cough in this situation is termed 'expected cough'. Such data are unavailable in adults.
However, concerns of parents presenting to general practitioners for their children's cough can be extreme (fear of child dying, chest damage) [29,30]. Other parental concerns were disturbed sleep and relief of discomfort [29]. However the burden of illness on children and their family has not been well described. In contrast adult data have shown that chronic cough causes a significant burden of illness (physical and psychosocial) that is often not appreciated by physicians [20] as reflected in adult cough-QOL scores [31,32].
What is acute and what is chronic?
The utility of definitions depends on the intention of use. In adults, chronic cough is defined as cough lasting >8 weeks [33]. In children the definition of chronic cough varies from 3-weeks duration [34] to 12-weeks [35,36]. There are no studies that have clearly defined when cough should be defined chronic or persistent. As studies have shown that cough related to ARIs resolves within 1 to 3 weeks in most children [17,37] it would be logical to define chronic cough as daily cough lasting >4 weeks.
Classification of paediatric cough
Paediatric cough can be classified in several ways, based on aetiology [38], timeframe [35] and characteristic (moist vs dry). For practical reasons, guidelines based on cough duration, combined with cough quality have been developed [35]. An evidence based guideline specific for paediatrics will be published as part of the American College of Chest Physicians' Guidelines on the Management of Cough in Adults and Children [39]. The previous guidelines which stated that "the approach to managing cough in children is similar to the approach in adults" [34] was arguably inaccurate.
Unlike cough in adults, paediatric cough has also been classified into specific and non-specific cough (with an overlap) for practical reasons (figure 1). Indeed, the most common paradigm encountered in clinical paediatrics when cough is a presenting feature is the differentiation between specific and non-specific cough. Specific cough refers to cough in the presence of pointers (table 1) that suggest the presence of an underlying aetiology. A thorough history and examination to elucidate these points are necessary when assessing children with cough and in the majority of situations, specific cough aetiologies can be defined. While some of these symptoms and signs are common in adults (such as haemoptysis), others are not (such as failure to thrive). Unlike in adults, where cough characteristics has been shown to be of little diagnostic value [40], paediatricians often recognise certain cough qualities such as staccato cough (table 2). A chronic moist cough is always abnormal and represents excessive airway secretions [41]. However in a small group of children natural resolution may occur [42] and a specific paediatric diagnostic category may not be found [43]. A chronic dry cough however may represent a dry phase of an otherwise usually moist cough or airway secretions too little to influence the cough quality [41]. Chronic dry cough in the absence of specific pointers (table 1) in the history and examination is termed 'non-specific cough' or 'isolated cough', ie cough is the sole symptom. In non-specific cough, the aetiology is ill defined and we suspect that the majority are related to post viral cough and/or increased cough receptor sensitivity [44,45]. However in the majority of children, it is most likely related to a non serious aetiology [38] or may spontaneously resolve as evidenced in the placebo arms of RCTs [46-48] and cohort studies [49-51]. Thus if one assumes that the natural resolution of non-specific cough occurs in 50% of children, 85 children per study arm is required in a randomised controlled trial to detect a 50% difference between active and placebo groups, for a study powered at 90% at the 5% significance level.
Figure 1 Classification of types of cough in children (reproduced from [110]).
Table 1 Pointers to underlying aetiology i.e. presence of specific cough [39,110]
auscultatory findings
cough characteristics eg cough with choking, cough quality (table 2), cough starting from birth
cardiac abnormalities (including murmurs)
chest pain
chest wall deformity
chronic dyspnoea
daily moist or productive cough
digital clubbing
exertional dyspnoea
failure to thrive
feeding difficulties
haemoptysis
immune deficiency
neurodevelopmental abnormality
recurrent pneumonia
Table 2 Classical recognisable cough [39,110]
Barking or brassy cough Croup [252] tracheomalacia [132,134] habit cough [157,253]
Honking Psychogenic [254]
Paroxysomal (with/without whoop) Pertussis and parapertussis [123,255]
Staccato Chlamydia in infants [256]
Cough productive of casts Plastic bronchitis [257]
Symptoms
Nocturnal cough
In both adults and children, a major problem in utilising the symptom of nocturnal cough is the unreliability and inconsistency of its reporting when compared to objective measurements [52-54]. In children, however, two groups have reported that parents were able to detect change [46,54], albeit only moderately well. The ability to detect cough change was better in children with a history of troublesome recurrent cough (r = 0.52) than in children without (r = 0.38) [54]. Relationship between change in cough frequency and change in subjective scores has not been examined in adults.
Nocturnal cough is often used as a hallmark of asthma as children with asthma often report troublesome nocturnal cough [55]. However in a community based study, only a third of children with isolated nocturnal cough had an asthma-like illness [56]. To date there are no studies that have objectively documented that nocturnal cough is worse than daytime cough in children with unstable asthma. One study showed that cough frequency was higher during the day than at night in a group of children with stable asthma who were on ICS yet had elevated levels of eNO but not sputum eosinophils [57] (arguably the best marker for eosinophilic inflammation in stable asthma [58]) in schoolchildren. Whether the increased eNO is a marker of asthma instability or related to other causes of elevated nitric oxide (such as environmental pollutants) [59,60] is unknown. Nocturnal cough has been reviewed elsewhere [61].
Cough quality
Unlike adults, cough quality is associated with specific aetiology in children (table 2). Except for brassy cough and wet cough, the sensitivity and specificity of cough quality have not been defined [62]. Thus perceived cough quality by parents and clinicians may have limitations. Pertussis-like cough in children may indeed be caused by adenovirus, parainfluenza viruses, respiratory syncytial virus and Mycoplasma [63]. Children with a dry cough are more likely to naturally resolve than those with wet cough [64]. Young children rarely expectorate even when airway secretions are excessive. Hence wet/moist cough is often used interchangeably with productive cough [65,66] a term used in adults. We have recently shown the clinical validity of dry and wet/moist cough in children by scoring secretions seen during bronchoscopy [41]. In contrast, quality of cough has been shown to be of little use in adults [40,67].
Investigations
Children with specific cough usually require a variety of investigations which include chest CT, bronchoscopy, barium meal, video fluoroscopy, nuclear scans, sweat test, etc. The role of these tests for evaluation of lung disease is beyond the scope of this article as it would encompass the entire spectrum of paediatric respiratory illness. The more common problem of non-specific cough is further briefly discussed. In general investigations are rarely needed in non-specific cough.
Airway cellular assessment
Examination of cellular profile of induced sputum, a standard in some adult cough clinics, can only be performed in older children (children >6 years). The majority of children with chronic cough seen by paediatricians are in the toddler age group (1–5 years) where bronchoscopy is necessary to obtain airway cells. In contrast to adult studies, all 4 paediatric studies [51,68-70] that have examined airway cellularity in children with chronic cough have rarely found an asthma-like profile. Other than assessment of airway specimens for microbiological purposes, the use of airway cellular and inflammatory profile in children with chronic cough is currently entirely limited to supportive diagnosis and research rather than definitive diagnosis. This is in contrast to that in adults with chronic cough where some have suggested use of airway inflammatory profiles to direct therapy [71,72]. One study in children with 'cough variant asthma' (mean age 11 years) showed that those with a higher percentage (>2.5%) of eosinophils in their sputum were more likely to develop classical asthma on follow-up [73]. There was however no appropriate control group and sputum ECP was unpredictive of asthma [73].
Cough sensitivity measures
In the physiology of cough, gender differences in CRS well recognised in adults [74], are absent in children [44]. In children, CRS is instead influenced by airway calibre and age [44]. An adult type approach to CRS measurement that is reliant on a child inhaling and maintaining an open glottis during actuation of a dosimeter or during nebulisation is unreliable. Furthermore it has been shown in both adults [75,76] and children [77] that inspiratory flow (which influences lung deposition) influences CRS. Thus in children, regulation of a constant inspiratory flow is necessary for valid results [77]. Increased CRS has been found in children with recurrent cough [44], cough dominant asthma [78] and influenza infection [79]. However testing for CRS is non-diagnostic and its use is still limited to research purposes. In clinical circles, the concept of a temporal increase in CRS has been useful to explain 'expected cough'.
Use of chest and sinus CT scans
The utility of a CT scan in children has to be balanced with the reported increased lifetime cancer mortality risk [10]. The yield of ultrafast CT scans in children with chronic productive cough is 43%, where bronchiectasis was documented [80]. The yield of CT scan in evaluation of a dry cough without the presence of features in table 1 is unknown and arguably should not be performed. Lung cancers are extremely rare in children. In children, there is poor concordance in diagnostic modalities for diagnosing paranasal disease [81]. Also, a single study of paranasal sinus CT findings in children with chronic cough (>4 weeks) described that an abnormality was found in 66% [82]. However this finding has to be interpreted in the context of high rates (50%) of incidental sinus abnormality in asymptomatic children undergoing head CTs [83]. Abnormal sinus radiographs may be found in 18–82% of asymptomatic children [84]. Thus, it is arguably difficult to be confident of an objective diagnosis of nasal space disease as the cause of cough.
Flexible bronchoscopy
Indications for bronchoscopy in children with chronic cough include suspicion of airway abnormality, persistent changes on CXR, suspicion of an inhaled foreign body, evaluation of aspiration lung disease and for microbiological and lavage purposes. In these situations, cough is usually specific rather than non-specific. Bronchoscopically defined airway abnormality was present in 46.3% of children with chronic cough in a tertiary centre-based study, whereas in Callahan's [85] series, bronchoscopy assisted in diagnosis in 5.3% of children [86]. In a European series, chronic cough was the indication in 11.6% of the 1233 paediatric bronchoscopies performed [87].
Spirometry
Spirometry is valuable in the diagnosis of reversible airway obstruction in children with chronic cough. In the early studies on asthma presenting as chronic cough, abnormal baseline lung function was documented [88,89]. However spirometry is relatively insensitive [90,91] and a normal spirometry does not exclude underlying respiratory abnormality. In one study of 49 children with chronic cough, spirometry was normal in all who were able to perform the test [86].
Tests for airway hyper-responsiveness
In adults, tests for AHR are relatively easy to perform and direct AHR (methacholine, histamine) is used to exclude asthma [33]. In children (outside a research setting) testing for AHR is reliably performed only in older children (>6 years) and positive AHR especially to direct AHR challenges as an indicator of asthma has questionable validity [92,93]. Airway cellularity (sputum) in asymptomatic children with AHR was similar to children without AHR but significantly different to children with asthma [94]. In children, unlike in adults, the demonstration of AHR in a child with non-specific cough is unlikely to be helpful in predicting the later development of asthma [95] or the response to asthma medications [47]. The only RCT that examined the utility of AHR and response to inhaled salbutamol and ICS [47] found that the presence of AHR could not predict the efficacy of these therapies for cough [47]. Another study showed that AHR to hypertonic saline is significantly associated with wheeze and dyspnoea but not associated with dry cough or nocturnal cough once confounders were accounted for [96]. The older studies that equated presence of AHR in children with cough as representative of asthma were not placebo-controlled studies, confounders were not adjusted for, or used unconventional definitions of AHR [97-100]. A recent study using 6 min free running test described that exercise induced symptoms were poor predictors of bronchoconstriction [101]. However interpretation of the study is limited [102].
Other investigatory techniques
The single study on bronchial biopsies in 7 children with chronic cough described the association between early ARI and epithelial inflammation [103]. Bronchial biopsies are easily performed in adults, but are rarely performed in children except in selected centres where the procedure has been shown to be safe [104]. Airways resistance by the interrupter technique (Rint) has been used to asses values in children with cough [105] but Rint is not established in clinical practice and has problems with validity of measurements when undertaken by different investigators [106]. To date, there are no paediatric studies that have evaluated the role of NO or breath condensate in guiding management of chronic cough. Increased NO has been found in asthmatics with cough [57] but is also found in other conditions associated with cough such as environmental pollutants [60].
Outcome measures for cough-related studies
Cough severity indices, broadly divided into subjective and objective outcomes, measure different aspects of cough. In children, measures of CRS have a weak relationship with cough frequency. Subjective cough scores have a stronger and consistent relationship with cough frequency [107]. The choice of indices depends on the reason for performing the measurement [107].
Answers to questions on isolated cough are largely poorly reproducible [108] and nocturnal cough in children is unreliably reported [52,53]. The kappa value relating the chance-corrected agreement to questions on isolated cough is poor (0.02–0.57) [19,108,109] in contrast to isolated wheeze (0.7–1.0) [108]. Biased reporting of cough has been shown; parents who smoke under-report cough in their children [19]. Diary cards for cough have been validated against an objective method and children aged >6 years are better than their parents at quantifying their cough severity [54]. Cough-specific QOL questionnaires exist for adults but not for children. There is a clear need for a paediatric cough specific QOL scores, as adult QOL scores cannot be applied to children. Cough specific objective tests include ambulatory and non-ambulatory objective cough meters, CRS and cough peak flows (reviewed elsewhere) [110]. Adult type instruments require modification for use in children [111].
Aetiological factors
Although some diseases are common to both adults and children, the pattern of many respiratory illnesses in children is clearly different to adults; eg viruses associated with the common cold in adults can cause serious respiratory illnesses such as bronchiolitis and croup in previously well young children [112]. Both of these respiratory syndromes are non existent in adults. Conversely, common causes of cough and respiratory diseases in adults such as chronic bronchitis [113] and chronic obstructive pulmonary disease are not recognised diagnostic entities in paediatric respiratory literature and main textbooks [114,115]. The following highlights some of the differences between children and adults.
Cohort studies
Some hospital based clinical studies of children presenting with chronic cough have found asthma as the most common cause [116,117] but others have not [43,86]. In a prospective review of 81 children with chronic cough, none had asthma on final diagnosis [43]. In a retrospective review of 49 children with chronic cough, none of the children had asthma as the sole final diagnosis [86]. There is little doubt that the aetiology of cough would depend on the setting, selection criteria of children studied [69,86] follow-up rate [118] and depth of clinical history, examination and investigations performed. When airway profiles have been examined in children with isolated chronic cough, the studies have shown very few children with airway inflammation consistent with asthma [68-70]. Marguet and colleagues concluded that "chronic cough is not associated with the cell profiles suggestive of asthma and in isolation should not be treated with prophylactic anti-asthma drugs" [70].
Acute respiratory infections and post infections
Most coughs in early childhood are caused by viral ARIs [17,119]. In children with an ARI, 26% were still unwell 7-days after the initial consultation and 6% by day 14 [120]. Cough was however not specifically reported [120]. A systematic review on the natural history of acute cough in children aged 0–4 years in primary care reported that the majority of children improve with time but 5–10% progress to develop bronchitis and/or pneumonia [17]. Post-viral cough is a term that refers to the presence of cough after the acute viral respiratory infection. In Monto's review [121] the mean annual incidence of total respiratory illness per person year ranges from 5.0–7.95 in children aged less than 4 years to 2.4–5.02 in children aged 10–14 years [121]. A recent Australian study recorded respiratory infection/episode rates of 2.2–5.3 per person per year for children aged ≤10 years (mean duration of 5.5–6.8 days) [122]. That for adults (>20-years) was 1.7 [122].
Infections such as pertussis and mycoplasma can cause persistent cough not associated with other symptoms [123]. Pertussis should be suspected especially if the child has had a known contact with someone with pertussis even if the child is fully immunised as partial vaccine failure is an emergent problem [124]. A hospital study examined PCR and serology for pertussis in a prospective cohort of 40 children with chronic (>3 weeks) cough and found that only 5% of these children had laboratory evidence of pertussis [42]. No other published data on chronic cough have examined pertussis and mycoplasma infections with other cough etiologies. In a prospective childhood vaccine study, presence of Chlamydia pneumoniae, mycoplasma, parapertussis and pertussis were sought in children (aged 3–34 months) if a child or household member coughed for >7 days. In total, 115 aetiological agents were identified in 64% of episodes with cough [123]. The most common single agent was pertussis in 56% (median cough of 51 days), followed by Mycoplasma in 26% (cough for 23 days), Chlamydia in 17% (26 days), and parapertussis 2% [123]. Other microbial studies were not done. A factor that needs to be considered when analysing such results is determining whether the infectious agent isolated is the cause of the cough, as the percentage of asymptomatic infection can be very high (54%) [125]. In children who received the acellular pertussis vaccination, pertussis infection is clinically difficult to distinguish from diseases associated with coughing caused by other viral or bacterial infections [126].
Inhalation of foreign body
Cough is the most common symptom in some series of acute foreign material inhalation but not in others [127]. A history of a choking episode is absent in about half [128]. Presentations are usually acute [129] but chronic cough can also be the presentation of previously missed foreign body inhalation [130]. Unlike adults, a history of acute aspiration in young children has to be obtained from an adult who may not be present at the time of aspiration. Missed foreign bodies in the airways can lead to permanent lung damage [131].
Airway lesions and cough
Chronic cough is well described in children with airway lesions [132-134] and at lesser frequency in adults [135]. An adult study reported that none of 24 patients with tracheomalacia had chronic cough as a presenting symptom [135]. Gormley and colleagues described that 75% of children with tracheomalacia secondary to congenital vascular anomalies had persistent cough at presentation [134]. Other symptoms include stridor, chronic dyspnoea, recurrent respiratory infections and dysphagia [134]. How common are airway lesions in asymptomatic children is unknown and how the symptom of cough relates to airway lesions can only be postulated.
Environmental pulmonary toxic agents
In-utero tobacco smoke exposure alters respiratory control and responses [136,137], pulmonary development and physiology [138,139]. Its influence on the developing central and peripheral cough receptors, pathways and plasticity of the cough pathway [6,140] is unknown. ETS increases susceptibility to respiratory infections [141,142] causes adverse respiratory health outcomes [143] and increases coughing illnesses [144,145]. Increased ETS has also been described in cohorts of children with chronic cough compared to children without cough [69,69,143,144,146,147]. Indoor biomass combustion increases coughing illness associated with acute respiratory infections with an exposure-response effect [148]. Exposure to other ambient pollutants (particulate matter [149,150] nitrogen dioxide, gas cooking [151] etc) is also associated with increased cough in children in cross sectional [149,150] and longitudinal studies [152] especially in the presence of other respiratory illnesses such as asthma [149]. Some studies however have not shown this effect [153,154] which is likely partially related to problems with question-based epidemiological studies on isolated and nocturnal cough [14,19].
Functional respiratory disorder
Habitual cough or cough as a 'vocal tic' maybe transient or chronic and are far more commonly reported in the paediatric literature than in the adult literature [41]. In one series, psychogenic cough accounted for 10% of children with chronic cough [116]. A Swedish community study described the prevalence of chronic vocal tics was 0.3% in girls and 0.7% in boys [155]. The cough in psychogenic cough is typically thought to be absent at night. However objective cough recording in a child with psychogenic cough showed that cough during sleep does occur [156]. The typical psychogenic cough (honking cough) recognisable in children [67,157] is rare in adults [67]. In one study, 52% of those who had their cough recorded had barking (brassy, croupy) or honking cough [158]. However, brassy or croupy cough is also found in other childhood conditions associated with cough such as tracheomalacia [41].
The big three of chronic cough in adults
In adults, asthma, GORD, post-nasal drip (the big three) are said to cause upto 72–90% of chronic cough [159,160]. In contrast, there is no good data that suggest that these are common causes of chronic cough in children.
Asthma, reactive airway disease and cough in children
There is little doubt that children with asthma may present with cough. However, the majority of children with cough do not have asthma [14,69,70,161,162]. The use of isolated cough as a marker of asthma is indeed controversial with more recent evidence showing that in most children, isolated cough does not represent asthma [35,162]. Cough associated with asthma without a co-existent respiratory infection is usually dry [163]. Some medium term cohort studies on children with cough have suggested that the majority of these children eventually developed asthma [73,164] but other studies have not [49,50,165,166]. The Tuscon group showed that recurrent cough presenting early in life resolved in the majority [166]. Furthermore, these children with recurrent cough and without wheeze, had neither AHR nor atopy, and significantly differed from those with classical asthma, with or without cough [166]. Several other studies also support McKenzie's annotation [161] which highlighted the problem of over-diagnosis of asthma based on the symptom of cough alone [118]. In a prospective community study with a mean follow-up period of 3 years, 56% of children with recurrent cough aged 4–7 years later became asymptomatic; 37% reported continuing cough and 7.2% developed wheeze [49]. The proportion of children in the group who subsequently developed wheeze was similar to the asymptomatic group, who later developed wheeze on follow-up (10%) [49]. Faniran and colleagues concluded in their community based study of 1178 children that "cough variant asthma is probably a misnomer for most children in the community who have persistent cough" [118]. Thus in community settings, epidemiological studies have shown that isolated persistent cough is rarely asthma [118,161,165,167]. These data have been previously reviewed [14].
Upper airways disorders and cough in children
In adults, post-nasal drip has been reported as a common cause of cough [40]. In children, although nasal discharge and cough have been reported as the two most prominent symptoms in children with chronic sinusitis (30–120 days) [168] supportive evidence of cause and effect in children is less convincing [169]. A prospective study has shown that although sinusitis is a common condition in childhood, it is not associated with asthma or cough when the confounding factor of allergic rhinitis was removed [170]. The relationship between nasal secretions and cough is more likely linked by common aetiology (infection and/or inflammation causing both) or due to clearing of secretions reaching the larynx. Using a continuous infusion of 2.5 mls/min of distilled water into the pharynx of well adults, Nishino and colleagues demonstrated that laryngeal irritation and cough only occurred in the presence of hypercapnia (45–55 mmHg) [171] suggesting that pharyngeal secretions alone do not cause cough. Physiologically this is to be expected as the pharynx is not innervated by the vagus nerve, a necessary component of the cough reflex [172]. One study described increased extrathoracic AHR without bronchial AHR to methacholine in a group of children presenting with chronic cough [173] and other studies have linked extrathoracic AHR to sinusitis and rhinitis [174,175]. However, the repeatability and validity of extrathoracic AHR in children are ill-defined. Therapeutic approaches for allergic rhinitis have been well summarised [176].
GOR and cough in children
In adults, GORD is reported to cause up to 41% of chronic cough [177]. In non-controlled trials the improvement rate of cough by non-surgical intervention e.g. with PPI alone [178] or PPI with motility agents [179] for GORD associated cough, cough improvement rates of 86–100% have been reported [178,179]. However a systematic review found much less convincing results [180]. In children the data relating isolated cough to GORD is even far less convincing. The section on upper airway symptoms of a clinical practice guideline on the evaluation and management of children with GOR included a discussion on cough and GOR, concluded "...there is insufficient evidence and experience in children for a uniform approach to diagnosis and treatment" [181]. Cough unequivocably (RCT setting) related to acid GOR in adults has been reported to subside in 1–3 weeks [182] but such evidence is unavailable in children [180] and difficult to obtain. While GOR may be the reason for persistent cough [183,184] cough can also cause GOR [185,186]. Proof of cause and effect in children is rare [187] and it is difficult to delineate cause and effect [188]. There are limited studies which have prospectively examined causes of chronic cough in children. Those available suggest GOR is infrequently the sole cause of isolated cough in children. One prospective study of the causes of chronic cough in children found only one child with GOR out of a series of 38 [116]. A retrospective study found co-existent GOR in 4 of 49 children with chronic cough [86]. In contrast to data in adults where GOR is a frequent cause of chronic cough [159,189] there is indeed no current convincing evidence that GOR is a common cause of non-specific cough in children. Although case series have shown the link between supra-oesophageal reflux and GOR in children, there is a lack of convincing data, as Rudolph summarised "No studies have definitively demonstrated symptom improvement with medical or surgical therapy for the latter symptom presentations" [190].
Other aetiologies
Eosinophilic bronchitis and allergy
Eosinophilic bronchitis, a well described cause of chronic cough in adults [191] is not well recognised in children. 'Allergic or atopic cough' is a poorly defined condition even in adults [192]. The association between atopy and respiratory symptoms has been the subject of many epidemiological studies [193,194]. Some have described greater respiratory symptom chronicity [195] but others have not [193,194]. Inconsistent findings regarding cough and atopy are also present in the literature; reports of increased atopy (or diseases associated with atopy) in children with cough have been found in some cohort and cross sectional studies [165,196] but not in others [46,47,56,166]. Cough as a functional symptom can also be mistaken for an allergic disorder in children [197].
Medications and treatment side-effects
Chronic cough has been reported as a side effect of ACE inhibitors (2–16.7%) [198-200], inhaled ICS [201] and as a complication of chronic vagus nerve stimulation [202]. In children, cough associated with ACE inhibitors resolves within days (3–7 days) after withdrawing the medication [198,199] and may not recur when the medication is recommenced [199]. The package insert for omeprazole includes cough as an adverse event in 1.1% of adults and a single case report was recently published [203] but no reports on children were found.
Otogenic causes – Arnold's ear-cough reflex
In approximately 2.3–4.2% of people (bilateral in 0.3–2%), the auricular branch of the vagus nerve is present and the Arnold's ear-cough reflex can be elicited [204-206]. Case reports of chronic cough associated with ear canal stimulation from wax impaction and cholesteatoma have been reported [207,208]. In children, the significance of the ear reflex and cough was described as early as 1963 [209] although recently reported again [210].
Management options of non-specific cough
Cough is subject to the period-effect (spontaneous resolution of cough) [211] and thus non-placebo controlled intervention studies have to be interpreted with caution [212]. If any medications are trialled, a 'time to response' should be considered and considerations given to patient profile and setting (eg community practice vs tertiary hospital practice). The same empirical therapy (for asthma, GOR, and PND) suggested in adults [33] is largely inappropriate in children.
Physician and parental expectations
Providing parents with information on the expected time length of resolution of acute respiratory infections may reduce anxiety in parents and the need for medication use and additional consultation [120]. Appreciation of specific concerns and anxieties, and an understanding of why they present are thus important when consulting children with non-specific cough. Educational input is best done with consultation about the child's specific condition [213]. A RCT [214] examining the effect of a pamphlet and a videotape promoting the judicious use of antibiotics, found that their simple educational effort was successful in modifying parental attitudes about the judicious use of antibiotics.
Over the counter cough medications and anti-histamines
In contrast to adults where OTC medications, in particular codeine and its derivatives have been shown to be useful, systemic reviews for children have concluded that cough OTCs have little, if any, benefit in the symptomatic control of cough in children [215,216]. Moreover OTCs have significant morbidity and mortality [217,218] and are common unintentional ingestions in children aged <5 years [219]. Use of diphenhydramine is also non beneficial for symptomatic treatment of cough related to pertussis [220]. The use of steam inhalation, vitamin C, zinc, and echinacea for upper RTI has been summarised [221] with little benefit, if any, for symptomatic relief of cough for adults and children.
The efficacy of anti-histamines in relieving cough in children is minimal, if at all [222]. Thus, unlike adults, the use of anti-histamine therapy for chronic cough in children is mostly unjustified. Whether this difference between children and adults is related to atopic states is unknown. A RCT on ketotifen did not show any clinical benefit in the treatment of 113 infants and children with chronic cough and/or wheeze [223]. A systematic review of anti-histamine and nasal decongestion combinations, and anti-histamines in OTC medications has shown that these pharmaceuticals were no more likely than placebo in reducing acute cough in children [215]. The use of these medications that contain H1 receptor antagonist has to be balanced with adverse events [217,224,225] which includes reported death from toxicity in young children [217,218]. The latest published RCT (n = 100) also showed that diphenhydramine (a first generation H1-antagonist) and dextromethorphan were no different to placebo in reducing nocturnal cough or sleep disturbance in both the children and parent(s) [225]. Like other RCTs there was a significant improvement in both placebo and active arms for the cough outcomes measured [225].
Asthma therapy for cough
Old cohort studies describing that asthma therapy for that era (oral orciprenaline, salbutamol syrup [226,227] theophylline [97,227] and metaproterenol with theophylline) [88] was useful in abolishing cough included children with clinically recognisable asthma. For example, 8 of 11 children in Konig's study had cough with co-existant chest pain or dyspnea on exertion) [88] 10 of 32 children in another study had abnormal examination findings [89].
In ambulatory children with acute cough (1–10 days) with no history of asthma and a normal chest examination, oral albuterol was not effective in reducing cough frequency or duration [48]. In a meta-analysis, Smucy et al likewise concluded that "there is no evidence to support using beta2-agonists in children with acute cough and no evidence of airflow obstruction" [228]. There is only one study on use of inhaled salbutamol in chronic cough (median of 8-weeks) which also showed no benefit [47]. There is no evidence for the use of anti-cholinergics for in children with non-specific cough [229]. Use of bronchodilators must be weighed against adverse events (eg tremor, irritability [48] behaviour change, cost).
Only 2 RCTs on ICS for chronic non-specific cough in children have been published and both groups have cautioned against prolonged use of ICS [46,47]. There is no RCT on oral steroids for non-specific cough in children. In cough associated with pertussis, dexamethasone provides no significant benefit for the symptomatic relief of cough [220]. Even in children with wheeze, a RCT found that oral steroids may confer no benefit [230]. In contrast to high doses used in adults, low dose ICS has been shown to be effective in the management of the majority of childhood asthma [231-233] and there were reported significant adverse events on high doses [234,235]. Thus if a trial of asthma therapy is ever warranted, use of a moderate dose (400 mcg/day equivalent of budesonide) is suggested. This practice is however discouraged in most settings. As the earlier studies in adults and children that utilised medications for asthma for the era reported that cough related to asthma completely resolved by 2–7 days [88,89,97,236] it is recommended that reassessment is done in 2–3 weeks. Cough unresponsive to ICS should not be treated with increased doses of ICS. Cough that resolves with ICS use may be related to the period effect (spontaneous resolution) [211] or a transient effect responsive to ICS use (ICS may also impact on non-asthmatic airways with pulmonary toxicants [237]). Thus clinicians should be cognisant that the child that appears to respond to ICS does not necessarily have asthma and the child should be re-evaluated off asthma treatment.
Cromoglycate and nedocromil reduces cough associated with asthma [238,239] and in children born prematurely [240]. An open, single arm trial reported significant reduction in cough scores from 30 to 15/week after 2-weeks of treatment with nedocromil (4 mg qid) with no additional benefit in subsequent 4-weeks [241]. There are no published RCTs [242]. Leukotriene receptor antagonists have been examined in adults for cough [243] but there is no RCT data in children. Theophylline utilized in old studies [97,227] may have an effect on cough separate from its 'anti-asthmatic' properties but there are no RCTs in children [244] and theophylline has a narrow therapeutic range. Oral theophylline, but not placebo, induced complete remission in adults with ACE inhibitor related cough [245]. There is a need for RCTs examining the effectiveness of theophylline for non-specific cough in children.
Anti-microbials
The American Academy of Family Physician's guidelines discourages use of antibiotics except when rhinosinusitis and cough are present and not improving after 10 days [246]. Meta-analysis on anti-microbials for acute bronchitis (recent onset of productive cough without chronic obstructive pulmonary disease, sinusitis or pneumonia) in older children (aged >8 years) and adults showed a small benefit of 0.58 days but with significantly more adverse events [221]. In subacute cough, two paediatric RCT have shown that anti-microbials (amoxycillin/clavulanic acid [247] and erythromycin [248]) were more likely to achieve 'clinical cure' and also prevented progression of illness defined by need for antibiotics [249]. The quality of cough in both studies was not clearly defined but the secretions in both studies cultured M catarrhalis [247,248].
Cessation of ETS and other environmental toxicants
In the management of any child with cough irrespective of the aetiology, attention to exacerbation factors is encouraged. A single report was found on cessation of parental smoking as a successful form of therapy for the children's cough [250]. Behavioural counselling for smoking mothers has been shown to reduce young children's ETS exposure in both reported and objective measures of ETS [251].
Conclusion
Cough is very common and in the majority is reflective of expected childhood respiratory infections. However cough may also be representative of a significant serious disorder and all children with chronic cough should have a thorough clinical review to identify specific respiratory pointers. Physiologically, there are similarities and significant differences between adults and children. Expectedly, the aetiologies and management of cough in a child differ to those in an adult. Cough in children should be treated based on aetiology and there is no evidence for using medications for symptomatic relief of cough or for an empirical approach based on the big three adult aetiologies. The use of medications are discouraged based on current evidence and if medications are used, it is imperative that the children are reviewed within the time frame of 'time to response' and medications ceased if there is no effect. Irrespective of diagnosis, environmental influences and parental expectations should be reviewed and managed accordingly as cough impacts on the quality of life of parents and children. Children with cough should be managed differently to adults as the aetiological factors and treatment in children differ to those in adults.
Abbreviations
ACE Angiotensin converting enzyme
AHR Airway hyper-responsiveness
ARI Acute respiratory infection
CRS Cough receptor sensitivity
CXR Chest X-Ray
CT Computed Tomography
ETS Exposure to tobacco smoke
FTT Failure to thrive
GOR Gastroesophageal reflux
HRCT High resolution computed tomography of the chest
ICS Inhaled corticosteroids
OTC Over the counter
eNO exhaled nitric oxide
QOL Quality of Life
RCT Randomised controlled trial
PCR Polymerase chain reaction
Competing interests
No actual or potential conflict of interest exists.
AB Chang is funded by the Australian National Health Medical Research Council and the Royal Children's Hospital Foundation
==== Refs
Smyth RL Research with children BMJ 2001 322 1377 1378 11397728 10.1136/bmj.322.7299.1377
Sinaiko AR Daniels SR The use of short-acting nefedipine in children with hypertension: Another example of the need for comphrehensive drug tesing in children J Paediatr 2001 139 7 9 10.1067/mpd.2001.116163
Bonham AC Sekizawa S Joad JP Plasticity of central mechanisms for cough Pulm Pharmacol Ther 2004 17 453 457 15564091 10.1016/j.pupt.2004.09.008
Hay AD Schroeder K Fahey T Acute cough in children BMJ 2004 328 1062 15117793 10.1136/bmj.328.7447.1062
Thach BT Maturation and transformation of reflexes that protect the laryngeal airway from liquid aspiration from fetal to adult life Am J Med 2001 111 69S 77S 11749929 10.1016/S0002-9343(01)00860-9
Undem BJ Carr MJ Kollarik M Physiology and plasticity of putative cough fibres in the Guinea pig Pulm Pharmacol Ther 2002 15 193 198 12099763 10.1006/pupt.2002.0362
Gleeson M Cripps AW Clancy RL Modifiers of the human mucosal immune system Immunol Cell Biol 1995 73 397 404 8595916
Smart JM Suphioglu C Kemp AS Age-related T cell responses to allergens in childhood Clin Exp Allergy 2003 33 317 324 12614445 10.1046/j.1365-2222.2003.01583.x
Tulic MK Fiset PO Manuokion JJ Frankiel S Lavigne F Eidelman DH Hamid Q Roll of toll like receptor 4 in protection by bacterial lipopolysaccharide in the nasal mucosa of children but not adults Lancet 2004 363 1689 1698 15158630 10.1016/S0140-6736(04)16253-3
Brenner DJ Estimating cancer risks from pediatric CT: going from the qualitative to the quantitative Pediatr Radiol 2002 32 228 3 11956700 10.1007/s00247-002-0671-1
Nunn JF Applied Respiratory Physiology 1993 4 London: Butterworths
Polgar G Weng T The functional development of the respiratory system from the period of gestation to adulthood Am Rev Respir Dis 1979 120 625 695 384853
Haddad GG Abman SH Chernick V Basic Mechanisms of Pediatric Respiratory Disease 2002 Hamilton: BC Decker Inc
Chang AB State of the Art: Cough, cough receptors, and asthma in children Pediatr Pulmonol 1999 28 59 70 10406052 10.1002/(SICI)1099-0496(199907)28:1<59::AID-PPUL10>3.0.CO;2-Y
Ludviksdottir D Bjornsson E Janson C Boman G Habitual coughing and its associations with asthma, anxiety, and gastroesophageal reflux Chest 1996 109 1262 1268 8625678
Cockburn J Pit S Prescribing behaviour in clinical practice: patients' expectations and doctors' perceptions of patients' expectations questionnaire study BMJ 1997 315 520 523 9329308
Hay AD Wilson AD The natural history of acute cough in children aged 0 to 4 years in primary care: a systematic review Br J Gen Pract 2002 52 401 409 12014540
Little P Gould C Williamson I Warner G Gantley M Kinmonth AL Reattendance and complications in a randomised trial of prescribing strategies for sore throat: the medicalising effect of prescribing antibiotics BMJ 1997 315 350 352 9270458
Dales RE White J Bhumgara C McMullen E Parental reporting of childrens' coughing is biased Eur J Epidemiol 1997 13 541 545 9258566 10.1023/A:1007311912777
Cornford CS Why patients consult when they cough: a comparison of consulting and non-consulting patients Br J Gen Pract 1998 48 1751 1754 10198482
Mellis CM Can we reduce acute asthma attendances to hospital emergency departments? Aust NZ J Med 1997 27 275 276
Dewey CR Hawkins NS The relationship between the treatment of cough during early infancy and maternal education level, age and number of other children in the household. ALSPAC Study Team. Avon Longitudinal Study of Pregnancy and Childhood Child Care Health Dev 1998 24 217 227 9618036 10.1046/j.1365-2214.1998.00067.x
Hutton N Wilson MH Mellits ED Baumgardner R Wissow LS Bonuccelli C Holtzman NA DeAngelis C Effectiveness of an antihistamine-decongestant combination for young children with the common cold: a randomized, controlled clinical trial J Pediatr 1991 118 125 130 1670783
Rietveld S Van BI Everaerd W Psychological confounds in medical research: the example of excessive cough in asthma Behav Res Ther 2000 38 791 800 10937427 10.1016/S0005-7967(99)00099-6
Rietveld S Rijssenbeek-Nouwens LH Prins PJ Cough as the ambiguous indicator of airway obstruction in asthma J Asthma 1999 36 177 186 10227269
Munyard P Bush A How much coughing is normal? Arch Dis Child 1996 74 531 534 8758131
Chang AB Phelan PD Robertson CF Newman RG Sawyer SM Frequency and perception of cough severity J Paediatr Child Health 2001 37 142 145 11328468 10.1046/j.1440-1754.2001.00608.x
Bonaccorso SN Sturchio JL For and against: Direct to consumer advertising is medicalising normal human experience: Against BMJ 2002 324 910 11950746 10.1136/bmj.324.7342.910
Cornford CS Morgan M Ridsdale L Why do mothers consult when their children cough? Fam Pract 1993 10 193 196 8359610
Davies MJ Cane RS Ranganathan SC McKenzie SA Cough, wheeze and sleep Arch Dis Child 1998 79 465 10193266
Birring SS Prudon B Carr AJ Singh SJ Morgan MD Pavord ID Development of a symptom specific health status measure for patients with chronic cough: Leicester Cough Questionnaire (LCQ) Thorax 2003 58 339 343 12668799 10.1136/thorax.58.4.339
French CT Irwin RS Fletcher KE Adams TM Evaluation of a cough-specific quality-of-life questionnaire Chest 2002 121 1123 1131 11948042 10.1378/chest.121.4.1123
Morice AH Committee members The diagnosis and management of chronic cough Eur Respir J 2004 24 481 492 15358710 10.1183/09031936.04.00027804
Irwin RS Boulet LP Cloutier MM Fuller R Gold PM Hoffstein V Ing AJ McCool FD O'Byrne P Poe RH Managing cough as a defense mechanism and as a symptom. A consensus panel report of the American College of Chest Physicians Chest 1998 114 133S 181S 9725800
Chang AB Asher MI A review of cough in children J Asthma 2001 38 299 309 11456383 10.1081/JAS-100002296
Phelan PD Asher MI Recurrent and persistent cough in children New Ethicals Journal 1999 41 45
Hay AD Wilson A Fahey T Peters TJ The duration of acute cough in pre-school children presenting to primary care: a prospective cohort study Fam Pract 2003 20 696 705 14701895 10.1093/fampra/cmg613
Bush A Paediatric problems of cough Pulm Pharmacol Ther 2002 15 309 315 12099784 10.1006/pupt.2002.0340
Chang AB Glomb WB Guidelines for evaluating chronic cough in pediatrics Chest 2005
Mello CJ Irwin RS Curley FJ Predictive values of the character, timing, and complications of chronic cough in diagnosing its cause Archives of Internal Medicine 1996 156 997 1003 8624180 10.1001/archinte.156.9.997
Chang AB Eastburn MM Gaffney J Faoagali J Cox NC Masters IB Cough quality in children: a comparison of subjective vs. bronchoscopic findings Respiratory Research 2005 6 3 15638942 10.1186/1465-9921-6-3
Marchant JM Masters IB Chang AB Chronic cough in children – understanding the spectrum of disease Eur Respir J 2003 22 176S
Seear M Wensley D Chronic cough and wheeze in children: do they all have asthma? Eur Respir J 1997 10 342 345 9042630 10.1183/09031936.97.10020342
Chang AB Phelan PD Sawyer SM Del Brocco S Robertson CF Cough sensitivity in children with asthma, recurrent cough, and cystic fibrosis Arch Dis Child 1997 77 331 334 9389238
Chang AB Phelan PD Sawyer SM Robertson CF Airway hyperresponsiveness and cough-receptor sensitivity in children with recurrent cough Am J Respir Crit Care Med 1997 155 1935 1939 9196099
Davies MJ Fuller P Picciotto A McKenzie SA Persistent nocturnal cough: randomised controlled trial of high dose inhaled corticosteroid Arch Dis Child 1999 81 38 44 10373132
Chang AB Phelan PD Carlin J Sawyer SM Robertson CF Randomised controlled trial of inhaled salbutamol and beclomethasone for recurrent cough Arch Dis Child 1998 79 6 11 9771244
Bernard DW Goepp JG Duggan AK Serwint JR owe PC Is oral albuterol effective for acute cough in non-asthmatic children? Acta Paediatr 1999 88 465 467 10342550 10.1080/08035259950169891
Brooke AM Lambert PC Burton PR Clarke C Luyt DK Simpson H The natural history of respiratory symptoms in preschool children Am J Respir Crit Care Med 1995 52 1872 1878 8520749
Brooke AM Lambert PC Burton PR Clarke C Luyt DK Simpson H Recurrent cough: natural history and significance in infancy and early childhood Pediatr Pulmonol 1998 26 256 261 9811075 10.1002/(SICI)1099-0496(199810)26:4<256::AID-PPUL4>3.0.CO;2-O
Zimmerman B Silverman FS Tarlo SM Chapman KR Kubay JM Urch B Induced sputum: comparison of postinfectious cough with allergic asthma in children J Allergy Clin Immunol 2000 105 495 499 10719299 10.1067/mai.2000.104933
Archer LNJ Simpson H Night cough counts and diary card scores in asthma Arch Dis Child 1985 60 473 474 4015154
Falconer A Oldman C Helms P Poor agreement between reported and recorded nocturnal cough in asthma Pediatr Pulmonol 1993 15 209 211 8469572
Chang AB Newman RG Carlin J Phelan PD Robertson CF Subjective scoring of cough in children: parent-completed vs child-completed diary cards vs an objective method Eur Respir J 1998 11 462 466 9551755 10.1183/09031936.98.11020462
Meijer GG Postma DS Wempe JB Gerritsen J Knol K van Aalderen WM Frequency of nocturnal symptoms in asthmatic children attending a hospital out-patient clinic Eur Respir J 1995 8 2076 2080 8666103 10.1183/09031936.95.08122076
Ninan TK Macdonald L Russel G Persistent nocturnal cough in childhood: a population based study Arch Dis Child 1995 73 403 407 8554355
Li AM Lex C Zacharasiewicz A Wong E Erin E Hansel T Wilson NM Bush A Cough frequency in children with stable asthma: correlation with lung function, exhaled nitric oxide, and sputum eosinophil count Thorax 2003 58 974 978 14586052 10.1136/thorax.58.11.974
Green RH Brightling CE McKenna S Hargadon B Parker D Bradding P Wardlaw AJ Pavord ID Asthma exacerbations and sputum eosinophil counts: a randomised controlled trial Lancet 2002 360 1715 1721 12480423 10.1016/S0140-6736(02)11679-5
Sundblad BM Larsson BM Palmberg L Larsson K Exhaled nitric oxide and bronchial responsiveness in healthy subjects exposed to organic dust Eur Respir J 2002 20 426 431 12212977 10.1183/09031936.02.00257402
Franklin P Dingle P Stick S Raised exhaled nitric oxide in healthy children is associated with domestic formaldehyde levels Am J Respir Crit Care Med 2000 161 1757 1759 10806184
Chang AB Cough diary, electronic monitoring and significance in asthma Monitoring Asthma Lung Biology in Health and Disease 2005 London: Marcel Dekker, Inc
Springer J Geppetti P Fischer A Groneberg DA Calcitonin gene-related peptide as inflammatory mediator Pulm Pharmacol Ther 2003 16 121 130 12749828 10.1016/S1094-5539(03)00049-X
Wirsing von Konig CH Rott H Bogaerts H Schmitt HJ A serologic study of organisms possibly associated with pertussis-like coughing Pediatr Infect Dis J 1998 17 645 649 9686733 10.1097/00006454-199807000-00013
Marchant JM Masters IB Chang AB Defining paediatric chronic bronchitis Respirology 2004 9 A61
Chang AB Masel JP Boyce NC Wheaton G Torzillo PJ Non-CF bronchiectasis-clinical and HRCT evaluation Pediatr Pulmonol 2003 35 477 483 12746947 10.1002/ppul.10289
De Jongste JC Shields MD Chronic cough in children Thorax 2003 58 998 1003 14586058 10.1136/thorax.58.11.998
Riegel B Warmoth JE Middaugh SJ Kee WG Nicholson LC Melton DM Parikh DK Rosenberg JC Psychogenic cough treated with biofeedback and psychotherapy. A review and case report Am J Phys Med Rehabil 1995 74 155 158 7710731
Fitch PS Brown V Schock BC Taylor R Ennis M Shields MD Chronic cough in children: bronchoalveolar lavage findings Eur Respir J 2000 16 1109 1114 11292114 10.1034/j.1399-3003.2000.16f15.x
Gibson PG Simpson JL Chalmers AC Toneguzzi RC Wark PAB Wilson A Hensley MJ Airway eosinophilia is associated with wheeze but is uncommon in children with persistent cough and frequent chest colds Am J Respir Crit Care Med 2001 164 977 981 11587982
Marguet C Jouen Boedes F Dean TP Warner JO Bronchoalveolar cell profiles in children with asthma, infantile wheeze, chronic cough, or cystic fibrosis Am J Respir Crit Care Med 1999 159 1533 1540 10228122
Pizzichini MM Pizzichini E Parameswaran K Clelland L Efthimiadis A Dolovich J Hargreave FE Nonasthmatic chronic cough: No effect of treatment with an inhaled corticosteroid in patients without sputum eosinophilia Can Respir J 1999 6 323 330 10463960
Hargreave FE Leigh R Induced sputum, eosinophilic bronchitis, and chronic obstructive pulmonary disease Am J Respir Crit Care Med 1999 160 S53 S57 10556171
Kim CK Kim JT Kang H Yoo Y Koh YY Sputum eosinophilia in cough-variant asthma as a predictor of the subsequent development of classic asthma Clin Exp Allergy 2003 33 1409 1414 14519148 10.1046/j.1365-2222.2003.01788.x
Kastelik JA Thompson RH Aziz I Ojoo JC Redington AE Morice AH Sex-related differences in cough reflex sensitivity in patients with chronic cough Am J Respir Crit Care Med 2002 166 961 964 12359654 10.1164/rccm.2109061
Barros MJ Zammattio SL Rees PJ Effect of changes in inspiratory flow rate on cough responses to inhaled capsaicin Clin Sci 1991 81 539 542 1657502
Barros MJ Zammattio SL Rees PJ The importance of the inspiratory flow rate in the cough response to inhaled citric acid Clin Sci 1990 78 521 525 2162281
Chang AB Phelan PD Roberts RGD Robertson CF Capsaicin cough receptor sensitivity test in children Eur Respir J 1996 9 2220 2223 8947063 10.1183/09031936.96.09112220
Chang AB Phelan PD Robertson CF Cough receptor sensitivity in children with acute and non-acute asthma Thorax 1997 52 770 774 9371206
Shimizu T Mochizuki H Morikawa A Effect of influenza A virus infection on acid-induced cough response in children with asthma Eur Respir J 1997 10 71 74 9032495 10.1183/09031936.97.10010071
Coren ME Ng V Rubens M Rosenthal M Bush A The value of ultrafast computed tomography in the investigation of pediatric chest disease Pediatr Pulmonol 1998 26 389 395 9888213 10.1002/(SICI)1099-0496(199812)26:6<389::AID-PPUL3>3.0.CO;2-6
Ioannidis JP Lau J Technical report: evidence for the diagnosis and treatment of acute uncomplicated sinusitis in children: a systematic overview Pediatrics 2001 108 E57 11533375 10.1542/peds.108.3.e57
Tatli MM San I Karaoglanoglu M Paranasal sinus computed tomographic findings of children with chronic cough Int J Pediatr Otorhinolaryngol 2001 60 213 217 11551612 10.1016/S0165-5876(01)00535-3
Diament MJ Senac MO Gilsanz V Baker S Gillespie T Larsson S Prevalence of incidental paranasal sinuses opacification in pediatric patients: a CT study J Comput Assist Tomogr 1987 11 426 431 3571583
Shopfner CE Rossi JO Roentgen evaluation of the paranasal sinuses in children AJR 1973 118 176 186
Callahan CW Etiology of chronic cough in a population of children referred to a pediatric pulmonologist Journal of the American Board of Family Practice 1996 9 324 327 8884670
Thomson F Masters IB Chang AB Persistent cough in children – overuse of medications J Paediatr Child Health 2002 38 578 581 12410870 10.1046/j.1440-1754.2002.00045.x
de Blic J Marchac V Scheinmann P Complications of flexible bronchoscopy in children: prospective study of 1,328 procedures Eur Respir J 2002 20 1271 1276 12449184 10.1183/09031936.02.02072001
Konig P Hidden asthma in children Am J Dis Child 1981 135 1053 1055 7294009
Hannaway PJ Hopper GDK Cough Variant Asthma in Children JAMA 1982 247 206 208 7053462 10.1001/jama.247.2.206
Marchant JM Masel JP Dickinson FL Masters IB Chang AB Application of chest high-resolution computer tomography in young children with cystic fibrosis Pediatr Pulmonol 2001 31 24 29 11180671 10.1002/1099-0496(200101)31:1<24::AID-PPUL1003>3.0.CO;2-D
Tiddens HA Detecting early structural lung damage in cystic fibrosis Pediatr Pulmonol 2002 34 228 231 12203854 10.1002/ppul.10134
Wilson N Silverman M Silverman M Bronchial responsiveness and its measurement Childhood asthma and other wheezing disorders 1995 London: Chapman & Hall 142 174
Wilson NM Bridge P Silverman M Bronchial responsiveness and symptoms in 5–6 year old children: a comparison of direct and indirect challenge Thorax 1995 50 339 345 7785004
Pin I Radford S Kolendowicz R Airway inflammation in symptomatic and asymptomatic children with methacholine hyperresponsiveness Eur Respir J 1993 6 1249 1256 8287939
Galvez RA McLaughlin FJ Levison H The role of the methacholine challenge in children with chronic cough J Allergy Clin Immunol 1987 79 331 335 3546464 10.1016/0091-6749(87)90150-3
Strauch E Neupert T Ihorst G Van's-Gravesande KS Bohnet W Hoeldke B Karmaus W Kuehr J Bronchial hyperresponsiveness to 4.5% hypertonic saline indicates a past history of asthma-like symptoms in children Pediatr Pulmonol 2001 31 44 50 11180674 10.1002/1099-0496(200101)31:1<44::AID-PPUL1006>3.0.CO;2-2
Cloutier MM Loughlin GM Chronic cough in children: a manifestation of airway hyperreactivity Pediatrics 1981 67 6 12 7243436
Paganin F Seneterre E Chanez P Daures JP Bruel JM Michel FB Bousquet J Computed tomography of the lungs in asthma: influence of disease severity and etiology Am J Respir Crit Care Med 1996 153 110 114 8542102
de Benedictis FM Canny GJ Levison H Methacholine inhalational challenge in the evaluation of chronic cough in children J Asthma 1986 23 303 308 3546254
Nishimura H Mochizuki H Tokuyama K Morikawa A Relationship between bronchial hyperresponsiveness and development of asthma in children with chronic cough Pediatr Pulmonol 2001 31 412 418 11389572 10.1002/ppul.1068
De Baets F Bodart E Dramaix-Wilmet M Van Daele S de Bilderling G Masset S Vermeire P Michel O Exercise-induced respiratory symptoms are poor predictors of bronchoconstriction Pediatr Pulmonol 2005 39 301 305 15678500 10.1002/ppul.20185
Primhak RA Commentary on de Baets et al.: Exercise-induced respiratory symptoms are poor predictors of bronchoconstriction Pediatr Pulmonol 2005 39 299 300 15678509 10.1002/ppul.20184
Heino M Juntunen-Backman K Leijala M Rapola J Laitinen LA Bronchial epithelial inflammation in children with chronic cough after early lower respiratory tract illness Am Rev Respir Dis 1990 141 428 432 2301858
Saglani S Payne DN Nicholson AG Scallan M Haxby E Bush A The safety and quality of endobronchial biopsy in children under five years old Thorax 2003 58 1053 1057 14645972 10.1136/thorax.58.12.1053
McKenzie SA Bridge PD Healy MJ Airway resistance and atopy in preschool children with wheeze and cough Eur Respir J 2000 15 833 838 10853845 10.1034/j.1399-3003.2000.15e04.x
Klug B Nielsen KG Bisgaard H Observer variability of lung function measurements in 2–6-yr-old children Eur Respir J 2000 16 472 475 11028662 10.1034/j.1399-3003.2000.016003472.x
Chang AB Phelan PD Robertson CF Roberts RDG Sawyer SM Relationship between measurements of cough severity Arch Dis Child 2003 88 57 60 12495964 10.1136/adc.88.1.57
Brunekreef B Groot B Rijcken B Hoek G Steenbekkers A de Boer A Reproducibility of childhood respiratory symptom questions Eur Respir J 1992 5 930 935 1426200
Clifford RD Radford M Howell JB Holgate ST Prevalence of respiratory symptoms among 7 and 11 year old schoolchildren and association with asthma Arch Dis Child 1989 64 1118 1125 2782926
Chang AB Chung FK, Widdicombe JG, Boushey HA Causes, assessment and measurement in children Cough: Causes, Mechanisms and Therapy 2003 London: Blackwell Science 57 73
Corrigan DL Paton JY Pilot study of objective cough monitoring in infants Pediatr Pulmonol 2003 35 350 357 12687591 10.1002/ppul.10267
Couriel J Ellis M Infection in children Infectious diseases of the respiratory tract 1998 1 Cambridge: Cambridge University Press 406 429
Taussig LM Smith SM Blumenfeld R Chronic bronchitis in childhood: what is it? Pediatrics 1981 67 1 5 7243418
Chernick V Boat TF 1998 Philadelphia: W.B. Saunders
Taussig LM Landau LI Pediatric Respiratory Medicine 1999 St. Louis: Mosby, Inc
Holinger LD Chronic Cough in Infants and Children Laryngoscope 1986 96 316 322 3512941
Callahan CW Cough with asthma: variant or norm? Journal of Pediatrics 1996 128 440 8774523
Faniran AO Peat JK Woolcock AJ Persistent cough: is it asthma? Arch Dis Child 1998 79 411 414 10193253
Ayres JG Noah ND Fleming DM Incidence of episodes of acute asthma and acute bronchitis in general practice 1976–87 Br J Gen Pract 1993 43 361 364 8251231
Butler CC Kinnersley P Hood K Robling M Prout H Rollnick S Houston H Clinical course of acute infection of the upper respiratory tract in children: cohort study BMJ 2003 327 1088 1089 14604932 10.1136/bmj.327.7423.1088
Monto AS Studies of the community and family: acute respiratory illness and infection Epidemiol Rev 1994 16 351 373 7713184
Leder K Sinclair MI Mitakakis TZ Hellard ME Forbes A A community-based study of respiratory episodes in Melbourne, Australia Aust NZ J Public Health 2003 27 399 404
Hallander HO Gnarpe J Gnarpe H Olin P Bordetella pertussis, Bordetella parapertussis, Mycoplasma pneumoniae, Chlamydia pneumoniae and persistent cough in children Scand J Infect Dis 1999 31 281 286 10482058 10.1080/00365549950163581
Torvaldsen S Simpson JM McIntyre PB Effectiveness of pertussis vaccination in New South Wales, Australia, 1996–1998 Eur J Epidemiol 2003 18 63 69 12705625 10.1023/A:1022588118030
Schmidt SM Muller CE Mahner B Wiersbitzky SK Prevalence, rate of persistence and respiratory tract symptoms of Chlamydia pneumoniae infection in 1211 kindergarten and school age children Pediatr Infect Dis J 2002 21 758 762 12192165 10.1097/00006454-200208000-00012
Liese JG Renner C Stojanov S Belohradsky BH Clinical and epidemiological picture of B pertussis and B parapertussis infections after introduction of acellular pertussis vaccines Arch Dis Child 2003 88 684 687 12876162 10.1136/adc.88.8.684
Cataneo AJ Reibscheid SM Ruiz Junior RL Ferrari GF Foreign body in the tracheobronchial tree Clin Pediatr (Phila) 1997 36 701 706 9415838
Oguz F Citak A Unuvar E Sidal M Airway foreign bodies in childhood Int J Pediatr Otorhinolaryngol 2000 52 11 16 10699234 10.1016/S0165-5876(99)00283-9
Mu L He P Sun D Inhalation of foreign bodies in Chinese children: a review of 400 cases Laryngoscope 1991 101 657 660 2041447
Raman TS Mathew S Ravikumar Garcha PS Atelectasis in children Indian Pediatr 1998 35 429 435 10216624
Karakoc F Karadag B Akbenlioglu C Ersu R Yildizeli B Yuksel M Dagli E Foreign body aspiration: what is the outcome? Pediatr Pulmonol 2002 34 30 36 12112794 10.1002/ppul.10094
Wood RE Localised tracheomalacia or bronchomalacia in children with intractable cough J Paediatr 1997 116 404 406
Finder JD Primary bronchomalacia in infants and children J Paediatr 1997 130 59 66
Gormley PK Colreavy MP Patil N Woods AE Congenital vascular anomalies and persistent respiratory symptoms in children Int J Pediatr Otorhinolaryngol 1999 51 23 31 10622442 10.1016/S0165-5876(99)00252-9
Grathwohl KW Afifi AY Dillard TA Olson JP Heric BR Vascular rings of the thoracic aorta in adults Am Surg 1999 65 1077 1083 10551760
Lewis K Bosque E Deficient hypoxia awakening response in infants of smoking mothers: Possible relationship to sudden infant death syndrome J Pediatr 1995 127 668 669 7562300
Cella D Bullinger M Scott C Barofsky I Group vs individual approaches to understanding the clinical significance of differences or changes in quality of life Mayo Clin Proc 2002 77 384 392 11936936
Stick S Pediatric origins of adult lung disease. 1. The contribution of airway development to paediatric and adult lung disease Thorax 2000 55 587 594 10856320 10.1136/thorax.55.7.587
Le Souef PN Tobacco related lung diseases begin in childhood Thorax 2000 55 1063 1067 11083894 10.1136/thorax.55.12.1063
Joad JP Munch PA Bric JM Evans SJ Pinkerton KE Chen CY Bonham AC Passive Smoke Effects on Cough and Airways in Young Guinea Pigs: Role of Brainstem Substance P Am J Respir Crit Care Med 2004 169 499 504 14644932 10.1164/rccm.200308-1139OC
Cohen S Tyrrell DA Russell MA Jarvis MJ Smith AP Smoking, alcohol consumption, and susceptibility to the common cold Am J Public Health 1993 83 1277 1283 8363004
Wu-Williams AH Samet JM Environmental tobacco smoke: exposure-response relationships in epidemiologic studies Risk Anal 1990 10 39 48 2184474
Li JS Peat JK Xuan W Berry G Meta-analysis on the association between environmental tobacco smoke (ETS) exposure and the prevalence of lower respiratory tract infection in early childhood Pediatr Pulmonol 1999 27 5 13 10023785 10.1002/(SICI)1099-0496(199901)27:1<5::AID-PPUL3>3.0.CO;2-5
Couriel JM Passive smoking and the health of children Thorax 1994 49 731 734 8091314
Lister SM Jorm LR Parental smoking and respiratory illnesses in Australian children aged 0–4 years: ABS 1989–90 National Health Survey results Aust N Z J Public Health 1998 22 781 786 9889443
Chang AB Harrhy VA Simpson JL Masters IB Gibson PG Cough, airway inflammation and mild asthma exacerbation Arch Dis Child 2002 86 270 275 11919102 10.1136/adc.86.4.270
Charlton A Children's coughs related to parental smoking BMJ 1984 288 1647 1649 6426674
Ezzati M Kammen D Indoor air pollution from biomass combustion and acute respiratory infections in Kenya: an exposure-response study Lancet 2001 358 619 624 11530148 10.1016/S0140-6736(01)05777-4
Vedal S Petkau J White R Blair J Acute effects of ambient inhalable particles in asthmatic and nonasthmatic children Am J Respir Crit Care Med 1998 157 1034 1043 9563716
Hirsch T Weiland SK von Mutius E Safeca AF Grafe H Csaplovics E Duhme H Keil U Leupold W Inner city air pollution and respiratory health and atopy in children Eur Respir J 1999 14 669 677 10543291 10.1034/j.1399-3003.1999.14c29.x
Holscher B Heinrich J Jacob B Ritz B Wichmann HE Gas cooking, respiratory health and white blood cell counts in children Int J Hyg Environ Health 2000 203 29 37 10956587 10.1078/S1438-4639(04)70005-X
Dodge R Solomon P Moyers J Hayes C A longitudinal study of children exposed to sulfur oxides Am J Epidemiol 1985 121 720 736 4014164
Salameh PR Baldi I Brochard P Raherison C Abi SB Salamon R Respiratory symptoms in children and exposure to pesticides Eur Respir J 2003 22 507 512 14516143
Roemer W Clench-Aas J Englert N Hoek G Katsouyanni K Pekkanen J Brunekreef B Inhomogeneity in response to air pollution in European children (PEACE project) Occup Environ Med 1999 56 86 92 10448312
Khalifa N von Knorring AL Prevalence of tic disorders and Tourette syndrome in a Swedish school population Dev Med Child Neurol 2003 45 315 319 12729145 10.1017/S0012162203000598
McGarvey LP Warke TJ McNiff C Heaney LG MacMahon J Psychogenic cough in a schoolboy: Evaluation using an ambulatory cough recorder Pediatr Pulmonol 2003 36 73 75 12772228 10.1002/ppul.10291
Butani L O'Connell EJ Functional respiratory disorders Ann Allergy Asthma Immunol 1997 79 91 99 9291412
Anbar RD Hall HR Childhood habit cough treated with self-hypnosis J Pediatr 2004 144 213 217 14760264 10.1016/j.jpeds.2003.10.041
Irwin RS Corrao WM Pratter MR Chronic persistent cough in the adult: the spectrum and frequency of causes and successful outcome of specific therapy Am Rev Respir Dis 1981 123 413 417 7224353
Morice AH Chung FK, Widdicombe JG, Boushey HA Epidemiology of cough Cough: Causes, Mechanisms and Therapy 2003 London: Blackwell Science 11 16
McKenzie S Cough – but is it asthma? Arch Dis Child 1994 70 1 2 8109998
Henry RL All that coughs is not asthma [editorial] Pediatr Pulmonol 1999 28 1 2 10406043 10.1002/(SICI)1099-0496(199907)28:1<1::AID-PPUL1>3.0.CO;2-X
British Guideline on the Management of Asthma Thorax 2003 58 i1 i94i 12653493 10.1136/thorax.58.suppl_1.1i
Todokoro M Mochizuki H Tokuyama K Morikawa A Childhood cough variant asthma and its relationship to classic asthma Ann Allergy Asthma Immunol 2003 90 652 659 12839325
Powell CVE Primhak RA Stability of respiratory symptoms in unlabelled wheezy illness and nocturnal cough Arch Dis Child 1996 75 385 391 8957950
Wright AL Holberg CJ Morgan WJ Taussig L Halonen M Martinez FD Recurrent cough in childhood and its relation to asthma Am J Respir Crit Care Med 1996 153 1259 1265 8616551
Lewis HM Cough – but is it asthma? Arch Dis Child 1994 70 554 8080586
Wald ER Byers C Guerra N Casselbrant M Beste D Subacute sinusitis in children J Pediatr 1989 115 28 32 2500511
Campanella SG Asher MI Current controversies: sinus disease and the lower airways Pediatr Pulmonol 2001 31 165 172 11180693 10.1002/1099-0496(200102)31:2<165::AID-PPUL1025>3.0.CO;2-O
Lombardi E Stein RT Wright AL Morgan WJ Martinez FD The relation between physician-diagnosed sinusitis, asthma, and skin test reactivity to allergens in 8-year-old children Pediatr Pulmonol 1996 22 141 146 8893251 10.1002/(SICI)1099-0496(199609)22:3<141::AID-PPUL1>3.0.CO;2-S
Nishino T Hasegawa R Ide T Isono S Hypercapnia enhances the development of coughing during continuous infusion of water into the pharynx Am J Respir Crit Care Med 1998 157 815 821 9517596
Widdicombe JG Neurophysiology of the cough reflex Eur Respir J 1995 8 1193 1202 7589405 10.1183/09031936.95.08071193
Turktas I Dalgic N Bostanci I Cengizlier R Extrathoracic airway responsiveness in children with asthma-like symptoms, including chronic persistent cough Pediatr Pulmonol 2002 34 172 180 12203845 10.1002/ppul.10077
Bucca C Rolla G Brussino L De Rose V Bugiani M Are asthma-like symptoms due to bronchial or extrathoracic airway dysfunction? Lancet 1995 346 791 795 7674743 10.1016/S0140-6736(95)91617-2
Rolla G Colagrande P Scappaticci E Bottomicca F Magnano M Brussino L Dutto L Bucca C Damage of the pharyngeal mucosa and hyperresponsiveness of airway in sinusitis J Allergy Clin Immunol 1997 100 52 57 9257787
Fireman P Therapeutic approaches to allergic rhinitis: treating the child J Allergy Clin Immunol 2000 105 S616 S621 10856167 10.1067/mai.2000.106152
Morice AH Epidemiology of cough Pulm Pharmacol Ther 2002 15 253 259 12099774 10.1006/pupt.2002.0352
Habermann W Kiesler K Eherer A Friedrich G Short-term therapeutic trial of proton pump inhibitors in suspected extraesophageal reflux J Voice 2002 16 425 432 12395996 10.1016/S0892-1997(02)00115-7
Poe RH Kallay MC Chronic cough and gastroesophageal reflux disease: experience with specific therapy for diagnosis and treatment Chest 2003 123 679 684 12628862 10.1378/chest.123.3.679
Chang AB Lasserson T Gaffney J Connor FC Garske LA Gastro-oesophageal reflux treatment for prolonged non-specific cough in children and adults (Cochrane Review) The Cochrane Library 2005
Rudolph CD Mazur LJ Liptak GS Baker RD Boyle JT Colletti RB Gerson WT Werlin SL Guidelines for evaluation and treatment of gastroesophageal reflux in infants and children: recommendations of the North American Society for Pediatric Gastroenterology and Nutrition J Pediatr Gastroenterol Nutr 2001 32 S1 31 11525610 10.1097/00005176-200100002-00001
Ours TM Kavuru MS Schilz RJ Richter JE A prospective evaluation of esophageal testing and a double-blind, randomized study of omeprazole in a diagnostic and therapeutic algorithm for chronic cough Am J Gastroenterol 1999 94 3131 3138 10566703 10.1111/j.1572-0241.1999.01504.x
Cucchiara S Santamaria F Minella R Alfieri E Scoppa A Calabrese F Franco MT Rea B Salvia G Simultaneous prolonged recordings of proximal and distal intraesophageal pH in children with gastroesophageal reflux disease and respiratory symptoms Am J Gastroenterol 1995 90 1791 1796 7572896
Orenstein SR Orenstein DM Gastroesophageal reflux and respiratory disease in children J Pediatr 1988 112 847 858 3286854
Gastal OL Castell JA Castell DO Frequency and site of gastroesophageal reflux in patients with chest symptoms. Studies using proximal and distal pH monitoring Chest 1994 106 1793 1796 7988202
Laukka MA Cameron AJ Schei AJ Gastroesophageal reflux and chronic cough: which comes first? Journal of Clinical Gastroenterology 1994 19 100 104 7963352
Yellon RF Coticchia J Dixit S Esophageal biopsy for the diagnosis of gastroesophageal reflux-associated otolaryngologic problems in children Am J Med 2000 108 131S 138S 10718466 10.1016/S0002-9343(99)00352-6
Gilger MA Pediatric otolaryngologic manifestations of gastroesophageal reflux disease Curr Gastroenterol Rep 2003 5 247 252 12734048 10.1159/000072144
Irwin RS Madison JM The Persistently Troublesome Cough Am J Respir Crit Care Med 2002 165 1469 1474 12045118 10.1164/rccm.2110097
Rudolph CD Supraesophageal complications of gastroesophageal reflux in children: challenges in diagnosis and treatment Am J Med 2003 115 150S 156S 12928092 10.1016/S0002-9343(03)00214-6
Gibson PG Fujimura M Niimi A Eosinophilic Bronchitis: Clinical Manifestations and Implications for Treatment Thorax 2002 57 178 182 11828051 10.1136/thorax.57.2.178
McGarvey L Morice AH Atopic cough: little evidence to support a new clinical entity Thorax 2003 58 736 737 12886002 10.1136/thorax.58.8.736-a
Mertsola J Ziegler T Ruuskanen O Vanto T Koivikko A Halonen P Recurrent wheezy bronchitis and viral respiratory infections Arch Dis Child 1991 66 124 129 1847281
Clough JB Holgate ST Episodes of respiratory morbidity in children with cough and wheeze Am J Respir Crit Care Med 1994 150 48 53 8025771
Clough JB Williams JD Holgate ST Effect of atopy on the natural history of symptoms, peak expiratory flow, and bronchial responsiveness in 7- and 8-year-old children with cough and wheeze Am Rev Respir Dis 1991 143 755 760 2008988
Peat JK Salome CM Woolcock AJ Longitudinal changes in atopy during a 4-year period: relation to bronchial hyperesponsiveness and respiratory symptoms in a population sample of Australian schoolchildren J Allergy Clin Immunol 1990 85 65 74 2299108 10.1016/0091-6749(90)90223-Q
Niggemann B Functional symptoms confused with allergic disorders in children and adolescents Pediatr Allergy Immunol 2002 13 312 318 12431189 10.1034/j.1399-3038.2002.00095.x
Bianchetti MG Caflisch M Oetliker OH Cough and converting enzyme inhibitors Eur J Pediatr 1992 151 225 226 1601018 10.1007/BF01954391
von Vigier RO Mozzettini S Truttmann AC Meregalli P Ramelli GP Bianchetti MG Cough is common in children prescribed converting enzyme inhibitors Nephron 2000 84 98 10644922 10.1159/000045552
Rokicki W Borowicka E Use of converting angiotensin inhibitors in children. II. Personal experience with enalapril Wiad Lek 1997 50 85 93 9381720
Dubus JC Mely L Huiart L Marguet C Le Roux P Cough after inhalation of corticosteroids delivered from spacer devices in children with asthma Fundam Clin Pharmacol 2003 17 627 631 14703724 10.1046/j.1472-8206.2003.00191.x
Smyth MD Tubbs RS Bebin EM Grabb PA Blount JP Complications of chronic vagus nerve stimulation for epilepsy in children J Neurosurg 2003 99 500 503 12959437
Howaizi M Delafosse C Omeprazole-induced intractable cough Ann Pharmacother 2003 37 1607 1609 14565840 10.1345/aph.1D185
Tekdemir I Aslan A Elhan A A clinico-anatomic study of the auricular branch of the vagus nerve and Arnold's ear-cough reflex Surg Radiol Anat 1998 20 253 257 9787391
Bloustine S Langston L Miller T Ear-cough (Arnold's) reflex Ann Otol Rhinol Laryngol 1976 85 406 407 937970
Todisco T The oto-respiratory reflex Respiration 1982 43 354 358 7146639
Sheehy JL Lee S Chronic cough due to cholesteatoma. A case report Am J Otol 1988 9 392 3202128
Raman R Impacted ear wax – a cause for unexplained cough? Arch Otolaryngol Head Neck Surg 1986 112 679
Smith FM Jr Arnold's nerve reflex; a little known cause of cough in pediatric patients J La State Med Soc 1963 115 17 18 13989411
Jegoux F Legent F Beauvillain dM Chronic cough and ear wax Lancet 2002 360 618 12241935 10.1016/S0140-6736(02)09786-6
Evald T Munch EP Kok-Jensen A Chronic non-asthmatic cough is not affected by inhaled beclomethasone dipropionate. A controlled double blind clinical trial Allergy 1989 44 510 514 2683836
Eccles R The powerful placebo in cough studies? Pulm Pharmacol Ther 2002 15 303 308 12099783 10.1006/pupt.2002.0364
Fitzmaurice DA Written information for treating minor illness BMJ 2001 322 1193 1194 11358757 10.1136/bmj.322.7296.1193
Taylor JA Kwan-Gett TS McMahon EM Jr Effectiveness of an Educational Intervention in Modifying Parental Attitudes About Antibiotic Usage in Children Pediatrics 2003 111 e548 e554 12728108 10.1542/peds.111.5.e548
Schroeder K Fahey T Should we advise parents to administer over the counter cough medicines for acute cough? Systematic review of randomised controlled trials Arch Dis Child 2002 86 170 175 11861232 10.1136/adc.86.3.170
Schroeder K Fahey T Over-the-counter medications for acute cough in children and adults in ambulatory settings The Cochrane Database of Systematic Reviews 2004 Art. No. CD001831.pub2. DOI: 10.1002/14651858.CD001831.pub2
Kelly LF Pediatric cough and cold preparations Pediatr Rev 2004 25 115 123 15060179
Gunn VL Taha SH Liebelt EL Serwint JR Toxicity of over-the-counter cough and cold medications Pediatrics 2001 108 E52 11533370 10.1542/peds.108.3.e52
Chien C Marriott JL Ashby K Ozanne-Smith J Unintentional ingestion of over the counter medications in children less than 5 years old J Paediatr Child Health 2003 39 264 269 12755931 10.1046/j.1440-1754.2003.00148.x
Pillay V Swingler G Symptomatic treatment of the cough in whooping cough The Cochrane Database of Systematic Reviews 2003 Art. No.: CD003257. DOI: 10.1002/14651858.CD003257
Del Mar C Glasziou P Upper respiratory tract infection Clinical Evidence 2002 8 1583 1591
De Sutter AIM Lemiengre M Campbell H Mackinnon HF Antihistamines for the common cold The Cochrane Database of Systematic Reviews 2003 Art. No.: CD001267. DOI: 10.1002/14651858.CD001267
van Asperen PP McKay KO Mellis CM Loh RK Harth SC Thong YH Harris M Robertson IF Gibbeson M Rhodes L A multicentre randomized placebo-controlled double-blind study on the efficacy of Ketotifen in infants with chronic cough or wheeze J Paediatr Child Health 1992 28 442 446 1466940
Hestand HE Teske DW Diphenhydramine hydrochloride intoxication J Pediatr 1977 90 1017 1018 859051
Paul IM Yoder KE Crowell KR Shaffer ML McMillan HS Carlson LC Dilworth DA Berlin CM Jr Effect of dextromethorphan, diphenhydramine, and placebo on nocturnal cough and sleep quality for coughing children and their parents Pediatrics 2004 114 e85 e90 15231978 10.1542/peds.114.1.e85
Spelman R Two-year follow up of the management of chronic or recurrent cough in children according to an asthma protocol British Journal of General Practice 1991 41 406 409 1777294
Yahav Y Katznelson D Benzaray S Persistent cough – a forme-fruste of asthma Eur J Respir Dis 1982 63 43 46 7067755
Smucny JJ Flynn CA Becker LA Glazier RH Are beta2-agonists effective treatment for acute bronchitis or acute cough in patients without underlying pulmonary disease? A systematic review J Fam Pract 2001 50 945 951 11711010
Chang AB McKean M Morris P Inhaled anti-cholinergics for prolonged non-specific cough in children The Cochrane Database of Systematic Reviews 2003 Art. No.: CD004358.pub2. DOI: 10.1002/14651858.CD004358.pub2
Oommen A Lambert PC Grigg J Efficacy of a short course of parent-initiated oral prednisolone for viral wheeze in children aged 1–5 years: randomised controlled trial Lancet 2003 362 1433 1438 14602435 10.1016/S0140-6736(03)14685-5
Pauwels RA Pedersen S Busse WW Tan WC Chen YZ Ohlsson SV Ullman A Lamm CJ O'Byrne PM Early intervention with budesonide in mild persistent asthma: a randomised, double-blind trial Lancet 2003 361 1071 1076 12672309 10.1016/S0140-6736(03)12891-7
O'Byrne PM Barnes PJ Rodriguez-Roisin R Runnerstrom E Sandstrom T Svensson K Tattersfield A Low dose inhaled budesonide and formoterol in mild persistent asthma: the OPTIMA randomized trial Am J Respir Crit Care Med 2001 164 1392 1397 11704584
Parameswaran K O'Byrne PM Sears MR Inhaled corticosteroids for asthma: common clinical quandaries J Asthma 2003 40 107 118 12765311 10.1081/JAS-120017980
Allen DB Bielory L Derendorf H Dluhy R Colice GL Szefler SJ Inhaled corticosteroids: past lessons and future issues J Allergy Clin Immunol 2003 112 S1 40 14515117 10.1016/S0091-6749(03)01859-1
Macdessi JS Randell TL Donaghue KC Ambler GR van Asperen PP Mellis CM Adrenal crises in children treated with high-dose inhaled corticosteroids for asthma Med J Aust 2003 178 214 216 12603184
McFadden ER Exertional Dyspnea and Cough as Preludes to Acute Attacks of Bronchial Asthma N Engl J Med 1975 292 555 558 1110670
Ek A Palmberg L Larsson K The effect of fluticasone on the airway inflammatory response to organic dust Eur Respir J 2004 24 587 593 15459137 10.1183/09031936.04.00018304
Creticos PS Effects of nedocromil sodium on inflammation and symptoms in therapeutic studies Journal of Allergy & Clinical Immunology 1996 98 S143 9 8939190
Hiller EJ Milner AD Lenney W Nebulized sodium cromoglycate in young asthmatic children. Double-blind trial Arch Dis Child 1977 52 875 876 413493
Yuksel B Greenough A The effect of sodium cromoglycate on upper and lower respiratory symptoms in children born prematurely Eur J Pediatr 1993 152 615 618 8354325 10.1007/BF01954094
Chan PW Debruyne JA Inhaled nedocromil sodium for persistent cough in children Med J Malaysia 2001 56 408 413 12014758
Chang AB Marchant JM Morris P Cromones for prolonged non-specific cough in children The Cochrane Database of Systematic Reviews 2004 Art. No.: CD004436.pub2. DOI: 10.1002/14651858.CD004436.pub2
Dicpinigaitis PV Dobkin JB Reichel J Antitussive effect of the leukotriene receptor antagonist zafirlukast in subjects with cough-variant asthma J Asthma 2002 39 291 297 12095178 10.1081/JAS-120002285
Chang AB Halstead RA Petsky HL Methylxanthines for prolonged non-specific cough in children The Cochrane Database of Systematic Reviews 2005 3
Cazzola M Matera MG Liccardi G De Prisco F D'Amato G Rossi F Theophylline in the inhibition of angiotensin-converting enzyme inhibitor-induced cough Respiration 1993 60 212 215 8265877
Dowell SF Schwartz B Phillips WR Appropriate use of antibiotics for URIs in children: Part I. Otitis media and acute sinusitis. The Pediatric URI Consensus Team Am Fam Physician 1998 58 1113 1123 9787277
Gottfarb P Brauner A Children with persistent cough–outcome with treatment and role of Moraxella catarrhalis? Scand J Infect Dis 1994 26 545 551 7710536
Darelid J Lofgren S Malmvall BE Erythromycin treatment is beneficial for longstanding Moraxella catarrhalis associated cough in children Scand J Infect Dis 1993 25 323 329 8362228
Marchant JM Morris P Gaffney J Chang AB Antibiotics for prolonged moist cough in children The Cochrane Database of Systematic Reviews 2005
Brand PL Duiverman EJ Coughing and wheezing children: improvement after parents stop smoking Ned Tijdschr Geneeskd 1998 142 825 827 9623153
Hovell MF Zakarian JM Matt GE Hofstetter CR Bernert JT Pirkle J Effect of counselling mothers on their children's exposure to environmental tobacco smoke: randomised controlled trial BMJ 2000 321 337 342 10926589 10.1136/bmj.321.7257.337
Cherry JD The treatment of croup: continued controversy due to failure of recognition of historic, ecologic, etiologic and clinical perspectives J Pediatr 1979 94 352 354 762649
Wamboldt MZ Wamboldt FS Taussig LM, Landau LI Psychiatric aspects of respiratory syndromes Pediatric Respiratory Medicine 1999 St. Louis: Mosby, Inc 1222 1234
Weinberg EG 'Honking': Psychogenic cough tic in children S Afr Med J 1980 57 198 200 7361211
Jenkinson D Natural course of 500 consecutive cases of whooping cough: a general practice population study BMJ 1995 310 299 302 7866173
Schaad UB Rossi E Infantile chlamydial pneumonia – a review based on 115 cases Eur J Pediatr 1982 138 105 109 6807674 10.1007/BF00441134
Kao NL Richmond GW Cough productive of casts Ann Allergy Asthma Immunol 1996 76 231 233 8634874
|
16270937
|
PMC1277009
|
CC BY
|
2021-01-04 16:37:44
|
no
|
Cough. 2005 Sep 20; 1:7
|
utf-8
|
Cough
| 2,005 |
10.1186/1745-9974-1-7
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.