Effects of chromosomal rearrangements on human-chimpanzee molecular evolution

Transcription

1 Genomics 84 (2004) Short communication Effects of chromosomal rearrangements on human-chimpanzee molecular evolution Eric J. Vallender a,b, Bruce T. Lahn a, * a Department of Human Genetics, Howard Hughes Medical Institute, University of Chicago, Chicago, Illinois 60637, United States b Committee on Genetics, University of Chicago, Chicago, Illinois 60637, United States Received 28 April 2004; accepted 12 July 2004 Available online 13 August Abstract Many chromosomes are rearranged between humans and chimpanzees while others remain colinear. It was recently observed, based on over 100 genes, that the rates of protein evolution are substantially higher on rearranged than on colinear chromosomes during humanchimpanzee evolution. This finding led to the conclusion, since debated in the literature, that chromosomal rearrangements had played a key role in human-chimpanzee speciation. Here we re-examine this important conclusion by employing larger a data set (over 7000 genes), as well as alternative analyses. We show that the higher rates of protein evolution on rearranged chromosomes observed in the earlier study are not reproduced by our survey of the larger data set. We further show that the conclusion of the earlier study is likely confounded by two factors introduced by the relatively limited sample size: (1) nonuniform distribution of genes in the genome, and (2) stochastic noise in substitution rates inherent to short lineages such as the human-chimpanzee lineage. Our results offer a general cautionary note on the importance of controlling for hidden factors in studies involving bioinformatic surveys. D 2004 Elsevier Inc. All rights reserved. It has long been proposed that chromosomal rearrangements may facilitate speciation by erecting reproductive barriers between subpopulations carrying chromosomal rearrangements relative to each other [1]. This notion, known as chromosomal speciation, is consistent with several experimental observations [2,3], though it has yet to be definitively proven by empirical evidence. More recently, the idea of chromosomal speciation was further extended at the theoretical level to include some testable hypotheses [4]. Most notably, it was hypothesized that positive selection should operate more intensely on rearranged than on colinear chromosomes during chromosomal speciation. This is because positive selection drives the fixation of genetic incompatibilities inherent during speciation, and these incompatibilities might be more likely to accumulate on rearranged chromosomes where cross-population introgression is restricted. * Corresponding author. Fax: (773) address: blahn@bsd.uchicago.edu (B.T. Lahn). Several large-scale chromosomal rearrangements exist between humans and chimpanzees, but the significance of these rearrangements in human chimpanzee molecular evolution is unclear [5]. A recent study exploited these rearrangements to test the aforementioned hypothesis [6]. By comparing 115 pairs of human/chimpanzee orthologs, the authors found that the ratio of nonsynonymous substitution rate (K a ) to synonymous substitution rate (K s ) a measure for the pace of protein evolution scaled to mutation rate is much higher in rearranged than in colinear chromosomes. Additionally, the authors found that a disproportionate number of genes with K a /K s ratios greater than 1, traditionally considered as evidence for position selection, were located on rearranged chromosomes relative to colinear ones. They concluded that, consistent with the hypothesis from earlier theoretical work [4], rearranged chromosomes had indeed experienced notably heightened positive selection during human chimpanzee speciation. In addition to providing empirical support for the theory of chromosomal speciation, this study and its companion commentary postulated that chromosomal rear /$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi: /j.ygeno

2 758 E.J. Vallender, B.T. Lahn / Genomics 84 (2004) rangements played a very prominent (and perhaps even causal) role in human chimpanzee speciation [6,7]. The companion commentary also put forward the intriguing notion that humans and chimpanzees might have continued to hybridize for about 3 million years after they first began to diverge into separate species some 6 million years ago [7]. These conclusions were subsequently reexamined by several investigators. Some felt that human chimpanzee hybridization for such a prolonged period seemed unlikely [8]. Others wondered that the genes used in the study might just happened to be skewed toward rapidly evolving genes on rearranged chromosomes for reasons unrelated to speciation [9]. However, the authors of the original study argued on multiple grounds that these concerns were insufficient to counter the validity of their study (for detail, see [10]). All parties did agree, however, that a much larger data set was needed to settle the debate. Given that the conclusions of the Navarro and Barton study have important ramifications for the understanding of speciation in general and human origins in particular, we sought to reexamine them by employing a more comprehensive data set and alternative analytical approaches. A recent effort to systematically sequence expressedsequence tags (ESTs) in the chimpanzee led to the identification of over 7000 human chimpanzee orthologous pairs [11]. These genes constitute a significant fraction of all the genes in the genome, and may therefore offer a more representative glimpse at genome-wide trends. Combining these genes with previously identified chimpanzee genes resulted in a data set of 7355 human chimpanzee orthologous pairs. Roughly half of these (3689) were located on rearranged chromosomes and half (3666) on colinear chromosomes, in line with the breakdown of the earlier report. We performed analysis on this large data set using the same methodologies as that employed by the original study. We found that the average K a /K s was 0.26 for the rearranged chromosomes and 0.27 for the colinear chromosomes, each with a standard error of approximately Thus, our finding differs from the original study, in that the average K a /K s in our data set is statistically indistinguishable between rearranged and colinear chromosomes. We next examined the distribution of K a /K s values in this large data set. We found that rearranged chromosomes and colinear chromosomes assumed distributions that are essentially identical either by visual inspection (Fig. 1), or by statistical tests ( p = 1 by Mann-Whitney s U test). Rearranged chromosomes in fact have a slight deficit of genes with K a /K s greater than 1 as compared to colinear chromosomes, even though the difference is not statistically significant. The above observations are at odds with the previous study, and imply that a complete human chimpanzee genome comparison, once it becomes possible, might not bear out the conclusion of the previous study. We suspected that the relatively small sample size of the original study was the underlying reason for our inability to reproduce its finding. But it was unclear as to what confounding factors were introduced by a small sample size. Here, we further investigate two likely possibilities. The first is that, among the relatively small number of genes sampled in the original study, rearranged chromosomes happen to contain more rapidly evolving genes due to the nonuniform distribution of these genes in the genome. It is well established that genes are not randomly sprinkled across the genome. Related members of gene families are often found in tight clusters. Genes of unrelated sequences but similar expression patterns also have a strong tendency to cluster [12,13]. Moreover, the genome appears to be composed of large blocks, where K s values of genes tend to be similar within a block but different between blocks [14]. Such mosaic structures of the genome can create considerable nonuniformity in the distribution of different types of genes. This in turn can result in patterns reflecting local rather than genomic trends when the sample size is limited. The second possible confounding factor is stochastic noise in K s. In calculating the K a /K s ratio, K s of a given gene is used as a proxy for its neutral mutation rate. However, when dealing with short lineages, such as the human-to-chimpanzee lineage, the number of synonymous mutations that fix in a gene can undergo significant stochastic fluctuation around the true underlying mutation rate of that gene. As a result, K s becomes a rather imprecise proxy of neutral mutation rate in short lineages. Since K s is used as the denominator in K a /K s calculation, stochastic variations in K s are further Fig. 1. Distributions of human chimpanzee K a /K s values for the large data set of 7355 genes. Genes are divided into those on chromosomes rearranged between human and chimpanzee (solid bars), and those on chromosomes colinear between the two species (open bars).

3 E.J. Vallender, B.T. Lahn / Genomics 84 (2004) amplified in the resulting K a /K s ratios. For example, a gene of average size and typical mutation rate should have about 3 synonymous substitutions between human and chimpanzee (given the ~1.2% human chimpanzee divergence). But the observed number of synonymous substitutions for that gene can easily be a number between 0 and 6 due to stochastic variation (even though the expected number of synonymous changes under the gene s true mutation rate is 3). Consequently, the K a /K s ratio of that gene can be underestimated by a few fold or overestimated by infinite times. The dramatic tendency for overestimating K a /K s of individual genes can lead to erroneous overestimation of the average K a /K s for a group of genes; and the extent of overestimation can differ from group to group, especially in the case of small groups. To address these possibilities, we compiled a data set of 227 pairs of human/chimpanzee autosomal orthologs from public databases. This data set subsumes all 115 genes used in the earlier study [6], and also includes additional genes to facilitate our subsequent analyses using the Old World monkey orthologs as outgroups (see later). In keeping with the earlier study, we excluded 11 MHC-related genes because of possible polymorphisms predating human chimpanzee speciation, as well as 19 genes that had identical coding sequences in human and chimpanzee. Of the remaining 197 genes, 90 reside on chromosomes rearranged between human and chimpanzee, and 107 on chromosomes colinear between the two species. We calculated K a /K s ratios of these genes, using the same methodologies as the original study [6]. Excluding genes with K s of 0, as was done in the earlier study, the average K a /K s values for rearranged and colinear chromosomes are 0.73 and 0.46, respectively. These values are roughly consistent with the original report of 0.84 and 0.37, respectively [6]. The number of genes with K a /K s values greater than 1 are 26 and 15 for rearranged and colinear chromosomes, respectively, when genes with K s of 0 are included (these numbers in the earlier study are 20 and 6, respectively). Thus, our slightly enlarged data set is broadly in line with the original Navarro and Barton report; i.e., rearranged chromosomes are enriched for genes of high K a /K s. To investigate the first possibility, i.e., rearranged chromosomes happen to contain more rapidly evolving genes in the limited data set due to the nonuniform distribution of these genes in the genome, we created two outgroup sets of genes to see if the same K a /K s patterns could be found in the outgroup lineages that apparently did not share any of the human/chimpanzee chromosomal rearrangements. One set consisted of 125 genes (57 on rearranged chromosomes and 68 on colinear chromosomes) that also had published orthologs in Old World monkey species (OWM; Cercopithecidae). We will refer to this set of genes as the human/chimpanzee/owm three-way set. The other set consisted of 169 genes (81 on rearranged chromosomes and 88 on colinear chromosomes) that also had mouse and rat orthologs. We will refer to these genes as the human/chimpanzee/mouse/rat four-way set. In the human/chimpanzee/owm three-way set, we used parsimony to deduce gene sequences of the last human/ chimpanzee common ancestor. We then calculated K a /K s of each gene for the lineage from the human/chimpanzee ancestor to OWM (this was done by comparing the human/ chimpanzee ancestral sequence with the corresponding OWM sequence). We found that the average K a /K s ratio for this lineage was 0.67 and 0.53 for rearranged and colinear chromosomes, respectively, whereas human chimpanzee comparison for these same genes gave average K a /K s values of 0.77 and 0.46, respectively (Fig. 2A). Additionally, the distribution of K a /K s values in this lineage broadly resembles that of the human chimpanzee K a /K s distribution, in the sense that rearranged chromosomes tend to have more genes in the high K a /K s range (compare Fig. 3B with Fig. 3A). For these distributions, we used the Mann-Whitney U test to assess the significance that the distribution of the rearranged chromosomes differed from that of the colinear chromosomes, as was done in the original study. The human chimpanzee significance for these genes was 0.19 while the human chimpanzee ancestor to OWM significance was 0.21, roughly comparable. In the human/chimpanzee/ mouse/rat four-way set, the average mouse rat K a /K s is 0.36 and 0.28 for rearranged and colinear chromosomes, respectively, whereas human chimpanzee comparison for these genes gave values of 0.59 and 0.41, respectively (Fig. 2B). Furthermore, the mouse rat K a /K s distribution is similar to the human chimpanzee distribution in terms of rearranged chromosomes having more genes of high K a /K s (compare Fig. 3C with Fig. 3A). Again, significance as calculated by the Mann-Whitney U test was 0.14 for human chimpanzee and 0.16 for mouse rat. Thus, the skew toward higher K a /K s values on rearranged chromosomes in the human-tochimpanzee lineage is similarly found in two other independent lineages: the lineage from the human/chimpanzee ancestor to OWM, and the mouse-to-rat lineage. Moreover, the skew is grossly comparable in scale across these different lineages. That the skew seen in the human/chimpanzee lineage is also found in lineages that do not share the same chromosomal rearrangements argues that the skew is likely due to reasons other than human/chimpanzee chromosomal rearrangements. Perhaps the more plausible interpretation is that chromosomes rearranged between humans and chimpanzees happen to harbor a greater proportion of rapidly evolving genes among the small set of genes sampled, and that these genes have a similar tendency to evolve rapidly in other mammalian lineages as well. We next addressed the second possible cause for the higher K a /K s on rearranged chromosomes observed in the small data set, i.e., the potential contribution by the inherently high levels of stochastic noise in human chimpanzee K s. To this end, we divided human chimpanzee K a by human OWM K s, with the assumption that K s is a less noisy proxy of neutral mutation rate when calculated with a longer lineage. Applying this method to the human/

4 760 E.J. Vallender, B.T. Lahn / Genomics 84 (2004) Fig. 2. Lineage specific K a /K s ratios for the human/chimpanzee/owm three-way gene set (A), and the human/chimpanzee/mouse/rat four-way gene set (B). Standard errors derived from 100,000 bootstrap iterations are given in parentheses. Branch lengths are drawn arbitrarily. Fig. 3. Distributions of K a /K s values for the small data set of 227 genes. Genes are divided into those on chromosomes rearranged between human and chimpanzee (solid bars), and those on chromosomes colinear between the two species (open bars). (A) K a /K s distributions in the human chimpanzee comparison. (B) K a /K s distributions in the comparison between human chimpanzee ancestor and Old World monkey. (C) K a /K s distribution in the mouse rat comparison.

5 E.J. Vallender, B.T. Lahn / Genomics 84 (2004) chimpanzee/owm three-way set, this artificial K a /K s was found to be 0.24 and 0.16 for rearranged and colinear chromosomes, respectively. The relative difference between rearranged and colinear chromosomes is now 50%, which is smaller than the 70% difference when using human chimpanzee K a /K s for the same set of genes. Thus, stochastic noise in K s also appears to have moderately augmented K a /K s ratios on rearranged chromosomes relative to colinear ones. We conclude that for the small data set, both nonuniform distribution of genes and stochastic noise in K s have contributed to the observation of elevated K a /K s ratios on rearranged chromosomes, with the former being a more prominent contributing factor. Such nonuniformity, which is particularly pronounced for small sample size, is mollified in studies using much larger data sets. In the previous debates over the role of chromosomal rearrangements in human chimpanzee speciation [8 10], all parties agreed that only the comprehensive sequencing of the chimpanzee genome could settle the issue. We have shown here that when a much larger set of genes is considered, the data fail to support this previous proposed role of chromosomal rearrangements in human chimpanzee speciation. The completion of the chimpanzee genome would ultimately bring this discussion to a final resolution. We note, however, that for many species, sequence information for a significant fraction of the genome is simply not available in the near term. Molecular evolutionists can therefore only approach their studies with limited sample size. In such cases, we suggest that it is perhaps expedient to include relevant controls and consider alternative explanations before extrapolating patterns associated with small data sets into genome-wide trends. We further note that, even if sequence information for much of the genome is available, it remains possible that a detectable pattern (such as significantly higher evolutionary rates for a subset of chromosomes) can actually be due to the genomewide nonuniform distribution of genes rather than any evolutionary properties of the underlying chromosomes (such as the occurrence of structural rearrangements). Additional analyses, such as the use of outgroup sequences, may be needed to rule out such confounding factors even in the case of large sample size. References [1] M.J.D. White, Modes of Speciation, Freeman, San Francisco, [2] M.B. Evgen ev, H. Zelentsova, H. Poluectova, G.T. Lyozin, V. Veleikodvorskaja, K.I. Pyatkov, L.A. Zhivotovsky, M.G. Kidwell, Mobile elements and chromosomal evolution in the virilis group of Drosophila, Proc. Natl. Acad. Sci. USA 97 (2000) [3] L.H. Rieseberg, C. Van Fossen, A.M. Desrochers, Hybrid speciation accompanied by genomic reorganization in wild sunflowers, Nature (1995) 375. [4] A. Navarro, N.H. Barton, Accumulating postzygotic isolation genes in parapatry: a new twist on chromosomal speciation, Evolution 57 (2003a) [5] J.J. Yunis, O. Prakash, The origin of man: a chromosomal pictorial legacy, Science 215 (1982) [6] A. Navarro, N.H. Barton, Chromosomal speciation and molecular divergence accelerated evolution in rearranged chromosomes, Science 300 (2003b) [7] L.H. Rieseberg, K. Livingstone, Chromosomal speciation in primates, Science 300 (2003) [8] E.J. Bowers, Chromosomal speciation, Science 301 (2003) author reply [9] J. Lu, W.H. Li, C.I. Wu, Comment on bchromosomal speciation and molecular divergence-accelerated evolution in rearranged chromosomesq, Science 302 (2003) 988 author reply 988. [10] A. Navarro, T. Marques-Bonet, N.H. Barton, Response to comment on bchromosomal speciation and molecular divergence-accelerated evolution in rearranged chromosomesq, Science 302 (2003) 988. [11] A.G. Clark, S. Glanowski, R. Nielsen, P.D. Thomas, A. Kejariwal, M.A. Todd, D.M. Tanenbaum, D. Civello, F. Lu, B. Murphy, S. Ferriera, G. Wang, X. Zheng, T.J. White, J.J. Sninsky, M.D. Adams, M. Cargill, Inferring nonneutral evolution from human-chimp-mouse orthologous gene trios, Science 302 (2003) [12] M.J. Lercher, A.O. Urrutia, L.D. Hurst, Clustering of housekeeping genes provides a unified model of gene order in the human genome, Nature Genet. 31 (2002) [13] E.J. Williams, L.D. Hurst, Clustering of tissue-specific genes underlies much of the similarity in rates of protein evolution of linked genes, J. Mol. Evol. 54 (2002) [14] C.M. Malcom, G.J. Wyckoff, B.T. Lahn, Genic mutation rates in mammals: local similarity, chromosomal heterogeneity, and X-versusautosome disparity, Mol. Biol. Evol. 20 (2003)