Elevated Rates of Nonsynonymous Substitution in Island Birds

Elevated Rates of Nonsynonymous Substitution in Island Birds Kevin P. Johnson* and Jon Seger* *Department of Biology, University of Utah, Salt Lake City, Utah; and Illinois Natural History Survey, Champaign, Illinois Slightly deleterious mutations are expected to fix at relatively higher rates in small populations than in large populations. Support for this prediction of the nearly-neutral theory of molecular evolution comes from many cases in which lineages inferred to differ in long-term average population size have different rates of nonsynonymous substitution. However, in most of these cases, the lineages differ in many other ways as well, leaving open the possibility that some factor other than population size might have caused the difference in substitution rates. We compared synonymous and nonsynonymous substitutions in the mitochondrial cyt b and ND2 genes of nine closely related island and mainland lineages of ducks and doves. We assumed that island taxa had smaller average population sizes than those of their mainland sister taxa for most of the time since they were established. In all nine cases, more nonsynonymous substitutions occurred on the island branch, but synonymous substitutions showed no significant bias. As in previous comparisons of this kind, the lineages with smaller populations might differ in other respects that tend to increase rates of nonsynonymous substitution, but here such differences are expected to be slight owing to the relatively recent origins of the island taxa. An examination of changes to apparently preferred and unpreferred synonymous codons revealed no consistent difference between island and mainland lineages Introduction A slightly deleterious mutation has a reasonable chance of drifting to fixation in a population smaller than roughly the reciprocal of the mutation s selective disadvantage (N e 1/s), but not in a population much larger than this (Kimura 1962, 1983). If a substantial fraction of nonsynonymous nucleotide mutations are slightly deleterious, then the rate of amino acid fixation should be higher in small populations than in large ones (Ohta 1973, 1992; Tachida 1991, 1996; Araki and Tachida 1997). The fixation of slightly deleterious mutations could be a significant feature of evolution, with implications for many aspects of biology (Li 1987; Gillespie 1991; Charlesworth, Morgan, and Charlesworth 1993; Kondrashov 1995; Lande 1998). To the degree that this expectation is satisfied, slightly deleterious mutations are inferred to be abundant, and many amino acid replacements are inferred to be slightly deleterious, in all but the largest populations. This nearly-neutral model of molecular evolution enjoys broad support because its assumptions are plausible and because it has passed many empirical tests (see below). However, these tests all suffer from one or more of the following problems: (1) only one phylogenetically independent comparison is made between a large-population lineage and a related small-population lineage; (2) the lineages have been separated so long that they have come to differ in many respects other than population size; and (3) the evidence for long-term population size differences between the lineages being contrasted is indirect or otherwise subject to doubt. Studies of mitochondrial sequence variation within and between species of animals often find a relative excess of low-frequency nonsynonymous polymorphisms Key words: deleterious mutations, island populations, mitochondrial genes, cyt b, ND2, nearly-neutral model, nonsynonymous substitutions. Address for correspondence and reprints: Kevin P. Johnson, Illinois Natural History Survey, 607 East Peabody Drive, Champaign, Illinois 61820-6970. E-mail: kjohnson@inhs.uiuc.edu. Mol. Biol. Evol. 18(5):874 881. 2001 2001 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038 within species (Nachman et al. 1996; Rand and Kann 1996, 1998; Hasegawa, Cao, and Yang 1998; Nachman 1998; Wise, Sraml, and Easteal 1998; Fry 1999). This pattern is usually interpreted as evidence that slightly deleterious nonsynonymous mutations frequently occur and drift at low frequencies but seldom reach high frequencies or become fixed, owing to selection against them (Parsons et al. 1997; Hasegawa, Cao, and Yang 1998). However, some of these nonsynonymous mitochondrial mutations do become fixed. The interesting but difficult question is, are many of these fixations deleterious? Several studies have compared a lineage of species with presumably small effective population sizes to a lineage with presumably larger population sizes and have found an increased rate of molecular evolution or of nonsynonymous substitution (or a higher K A /K S ratio) in the group assumed to have small population sizes (e.g., DeSalle and Templeton 1988; Easteal and Collet 1994; Ohta 1995). In each of these studies, only one comparison is made, so even where the observed difference is statistically significant, it amounts to only one data point with respect to a test of the general hypothesis that rates of nonsynonymous fixation will increase in small populations. One system where multiple comparisons can be made is in lineages of endosymbiotic bacteria. Endosymbiotic bacteria can be assumed to have smaller effective population sizes than their free-living relatives, and free-living forms have frequently given rise to endosymbiotic lineages (Wernegreen and Moran 1999). These lineages tend to show increased rates of nonsynonymous substitution, altered patterns of synonymous codon usage bias, and relatively destabilized 16S rrna secondary structures as compared with their free-living relatives (Lambert and Moran 1998; Wernegreen and Moran 1999). These patterns suggest that deleterious mutations have drifted to fixation at relatively high rates in endosymbiotic lineages. However, the role of population size cannot be cleanly distinguished from the roles of other factors that may also differ between endosym- 874

Mitochondrial Substitutions in Island Birds 875 bionts and their free-living relatives. For example, selection might be weaker in endosymbionts than in freeliving bacteria (at least with respect to the genes that have been studied), and endosymbionts might have lower rates of recombination. More generally, the ecologies of endosymbionts differ greatly from those of free-living bacteria, and endosymbionts are highly modified for their peculiar ways of life, so it is hard to know how far the all else equal assumption can safely be pushed in these cases. There is also evidence of selection for codon usage biases (Akashi 1995; Xia 1996; Rand and Kann 1998; Ballard 2000) in some taxa. Synonymous substitutions to unpreferred codons (codons translated more slowly or less accurately than preferred codons) should be only slightly deleterious compared with most amino acid changes. Thus, more changes to unpreferred codons should accumulate in small populations than in large ones (Wernegreen and Moran 1999). Akashi (1995) found a higher substitution rate to unpreferred codons in Drosophila melanogaster nuclear genes than in Drosophila simulans. Drosophila melanogaster is inferred to have a smaller long-term effective population size on the basis of the relative levels of standing genetic variation in the two species, not on the basis of direct estimates of population size. One potential solution to many of these problems is to compare species that live on islands with their sister taxa on mainlands (e.g., Llopart and Aguadé 1999). Island habitats are always limited in area, so island populations are limited in size. Of course, mainland populations may fluctuate in size, but those that are currently much larger than their island-endemic relatives are likely to have been larger for much of the recent history of the species pair. Island endemics arise frequently from mainland ancestors in many taxonomic groups, so large numbers of phylogenetically independent comparisons can be made. Birds are nearly ideal in this respect because they frequently colonize islands, and two large mitochondrial sequence data sets recently generated for phylogenetic purposes contain numerous island-mainland sister pairs. Here, we describe a set of phylogenetically independent contrasts of synonymous and nonsynonymous substitution in closely related, ecologically similar lineage pairs where one member of the pair is an island endemic and therefore almost certain to have had a limited population size (generally 1,000 individuals) for much of its recent history. We also compare the relative ratios of apparently preferred and unpreferred codon usage in island and mainland taxa. The data set included seven island-versus-mainland comparisons within the dabbling ducks (genus Anas) and two such comparisons within doves (genus Zenaida). Materials and Methods Parsimony and maximum-likelihood analyses were performed on published mitochondrial cytochrome b (cyt b) and NADH dehydrogenase subunit 2 (ND2) gene sequences for dabbling ducks (genus Anas) (Johnson and Sorenson 1998, 1999) and doves (genus Zenaida) (Johnson and Clayton 2000a). The ND2 sequences are complete (1,038 bp) and the cyt b sequences are nearly complete (1,041 bp), so the concatenated alignments were 2,079 bp long. In the maximum-parsimony (MP) analysis, we used the published trees to infer synonymous and nonsynonymous substitutions on branches of interest. These branches are all so short that multiple nucleotide substitutions are expected (and observed) to be rare at individual sites and even within codons. Because the vast majority of codons contain at most one substitution, we estimated the number of nonsynonymous substitutions on each branch as the number of inferred amino acid substitutions, and we estimated the number of synonymous substitutions as the difference between the inferred numbers of nucleotide and amino acid substitutions on the branch. In the maximum-likelihood (ML) analysis, we used DNAML, version 3.6 (Felsenstein 1989), to estimate both the trees and the sequences at internal nodes under a model with empirical base frequencies, a transitiontransversion ratio of 4, and four categories of sites evolving at relative rates of 1, 4, 12, and 20. Sites were assigned to categories by reference to the numbers of substitutions estimated in a parsimony analysis. Both for ducks and for doves, the ML tree was topologically similar to the MP tree, and the trees were identical with respect to the comparisons of interest. We used the method of Li (1993) and Pamilo and Bianchi (1993) to estimate per-site synonymous and nonsynonymous substitutions (K S, K A ) along each branch from the inferred last common ancestor. For the ML tree, we also estimated the numbers of preferred and unpreferred codon changes from an inferred ancestor to both island and mainland sister taxa. In avian mtdna, codons ending in A or C are much more common than those ending in T or G. This preference matches a general pattern in the anticodons of the trnas encoded in the avian mitochondrion (Desjardins and Morais 1990). All anticodons end in U or G (matching A and C in the codon) except the anticodon for methionine, which ends in C. Thus, the most common third-position bases correspond to the base composition of the anticodons. We defined preference as those bases encoded by anticodons (A or C) and unpreferred codons as those ending in T or G. For methionine-encoding codons, we considered T and G to be the preferred codon, but the results were unchanged if A or C was considered to be preferred for all codons. We classified each formally recognized species or subspecies as occurring on a mainland or an island (table 1 lists the geographic distributions for each of the island taxa). For each island taxon, we identified its sister taxon for comparison. Some of the comparisons involved terminal sister taxa, but in others the island taxon was sister to a clade containing more than one mainland taxon. In these cases, we used the average total branch length of the several mainland taxa to the common ancestor with the island taxon as the branch length for comparison. (These branch lengths are not statistically

876 Johnson and Seger Table 1 Differences in Synonymous and Nonsynonymous Substitutions Between Island and Mainland Sister Taxa ISLAND TAXON (distribution) Anas luzonica (Philippines)... Anas superciliosa (Pacific Islands)... Anas laysanensis (Laysan Island)... Anas melleri (Madagascar)... MAINLAND COMPARISON Avg. Anas zonorhyncha, Anas poecilorhyncha, Anas platyrhynchos 1 Avg. Anas diazi, Anas rubripes, Anas fulvigula, A. platyrhynchos 2, A. zonorhyncha, A. poecilorhyncha, A. platyrhynchos 1 Avg. A. diazi, A. rubripes, A. fulvigula, A. platyrhynchos 2, A. zonorhyncha, A. poecilorhyncha, A. platyrhynchos 1 Avg. A. diazi, A. rubripes, A. fulvigula, A. platyrhynchos 2, A. zonorhyncha, A. poecilorhyncha, A. platyrhynchos 1 Synonymous DIFFERENCE (island mainland) Nonsynonymous P K A /K S unpreferred 5.66 1.0 0.06 0.117 9.14 4.0 0.14 0.117 2.14 1.0 0.06 0.116 0.14 4.0 0.22 0.211 Anas bernieri (Madagascar)... Avg. Anas gibberifrons, Anas castanea 11.0 3.0 0.04 0.163 Anas bahamensis bahamensis (Caribbean Islands)... Anas bahamensis rubrirostris 1.0 1.0 0.0 1.0 Avg. Anas aucklandica aucklandica (Auckland Island) and Anas aucklandica nesiotis (Campbell Island)... Anas chlorotis 9.5 3.0 0.17 0.510 Zenaida graysoni (Socorro Island)... Zenaida macroura 1.5 2.0 0.19 0.536 Zenaida galapagoensis (Galapagos Island)... Avg. Z. macroura, Zenaida auriculata 40.3 2.5 0.03 0.020 independent, but if the phylogeny had been less complete, any one of them would have been sufficient as a mainland branch for comparison with the island branch.) Multiple substitutions at a single nucleotide position may be recovered more readily in a clade with several terminal taxa (Fitch and Bruschi 1987; Sanderson 1990), so any bias resulting from this effect should make our test conservative with respect to the alternative hypothesis, because it will tend to increase the estimated lengths of mainland branches. However, as mentioned above, any such bias should be very small owing to the low levels of divergence between the sequences being compared. In one case, two sister taxa on very small islands (Anas aucklandica aucklandica and Anas aucklandica nesiotis on Auckland Island and Campbell Island, respectively) were sister to the brown teal (Anas chlorotis) from New Zealand, which is a very large island. The brown teal undoubtedly has a much larger effective population size than the species on the very small Auckland and Campbell Islands. Thus, we used this more terminal sister comparison rather than the one that contrasts these three related island taxa with a very large and much more distantly related clade that includes most other dabbling ducks, although we consider this alternative higher-level contrast briefly in the discussion. In this case, we used the average values for the Auckland and Campbell Island teals. Within dabbling ducks, we identified seven islandversus-mainland comparisons, and within doves, we identified two such comparisons. From the parsimony analysis, we calculated the difference between synonymous and nonsynonymous substitutions between the island taxa and the mainland taxa (island branch length minus mainland branch length). From the ML analysis, we calculated the difference in K A /K S (from terminal taxa to their reconstructed last common ancestor). Under the null hypothesis that each kind of fixation (synonymous and nonsynonymous) occurs at equal rates in island and mainland sister lineages, all of these differences should be distributed symmetrically around means of 0 and, in particular, they should be equally likely to be positive or negative. Under the alternative (nearly-neutral) hypothesis, the nonsynonymous substitution and K A /K S differences should be positive more often than negative. To examine codon preference, we calculated the fraction of inferred substitutions that were unpreferred (P unpreferred) and subtracted the mainland ratio from the island ratio, with the expectation that under the nearly-neutral model this difference would be positive. We used sign tests (binomial probabilities) to evaluate these hypotheses.

Mitochondrial Substitutions in Island Birds 877 FIG. 1. Phylogeny of dabbling ducks (Anas) from Johnson and Sorenson (1999), showing inferred numbers of (a) synonymous and (b) nonsynonymous substitutions in the cyt b and ND2 genes. Branch lengths are proportional to the numbers of reconstructed substitutions, which are also indicated by numbers. The names of island taxa are shown in boldface type. Both trees are rooted on a large composite outgroup (not shown). Results Within the dabbling ducks, island lineages showed more amino acid (nonsynonymous) substitutions and higher K A /K S ratios in all seven comparisons (fig. 1 and table 1). Island lineages also tended to show more synonymous change (six of seven cases). Within the doves, the two island lineages showed more nonsynonymous substitutions, but fewer synonymous substitutions, than their mainland sisters (fig. 2 and table 1). Thus, nine out of nine island lineages showed increased nonsynonymous substitution, and the null hypothesis was rejected (P 0.004, two-tailed; P 0.002, one-tailed), but synonymous substitutions showed no significant bias (P 0.5, two-tailed; P 0.25, one-tailed). The fraction of synonymous substitutions to unpreferred codons was higher for island lineages in just four of the nine comparisons (P 0.9, table 1). Both the dove comparisons and two of the more distantly related duck comparisons were in the direction predicted if unpreferred codons were slightly deleterious. However, even if one excludes all but one of the comparisons involving the mallard group (see table 1), four cases in the predicted direction is not significant (P 0.10). Discussion Island taxa, with small effective population sizes, are expected to show increased rates of nonsynonymous substitution under the nearly-neutral model of molecular evolution (Ohta 1973, 1992). Island ducks and doves consistently showed the expected increase in nonsynonymous substitutions. This result seems unlikely to be an artifact of the reconstruction method, because it is robust to both MP and ML reconstructions, and it is evident in two distantly related and ecologically dissimilar groups of birds. Several previous empirical tests of the effect of population size on the rate of nonsynonymous substitution have involved just one or a few comparisons (DeSalle and Templeton 1988; Easteal and Collet 1994). Llopart and Aguadé (1999) found an elevated rate of substitution to unpreferred synonymous codons in the highly conserved RpII215 gene of Drosophila guanche (endemic to the Canary Islands), and they found a smaller, nonsignificant increase in Drosophila madeirensis (endemic to Madeira). Lambert and Moran (1998) found three to six (depending on phylogenetic resolution) independent origins of a less stable 16S rrna secondary structure in bacterial endosymbionts than in free-living relatives. Here, we found an increased rate of replacement substitution in nine island-mainland comparisons. In four comparisons we were forced to use the same mainland taxa owing to the structures of the trees. However, even if we considered only the seven independent comparisons that can be made using nonoverlapping branches (matched-pair comparisons; Clayton and Cotgreave 1994), the result remained significant (P 0.016, two-tailed; P 0.008, one-tailed). The population sizes of birds (especially ducks) are relatively well documented. Most of the island taxa considered here have current population sizes ranging from a few tens of individuals up to a few thousands, and one species (Zenaida graysoni) is extinct in the wild (Jehle and Parkes 1982). Population size estimates for the

878 Johnson and Seger FIG. 2. Phylogeny of doves (Zenaida) from Johnson and Clayton (2000a), showing inferred numbers of (a) synonymous and (b) nonsynonymous substitutions in the cyt b and ND2 genes. Branch lengths are proportional to the numbers of reconstructed substitutions, which are also indicated by numbers. The names of island taxa are shown in boldface type. Both trees are rooted on a composite outgroup (not shown). mainland species generally are on the order of several million individuals. It is unlikely that the population sizes of the island species have often been much larger than they are today, owing to the limited amounts of suitable habitat on these islands. While the population sizes of mainland species may fluctuate, they seem likely to have been large for much of their recent histories. Most of the comparisons considered here involved species with mitochondrial sequence divergences of 5%, corresponding to separations of less than 3 Myr (Klicka and Zink 1997), and some of the comparisons were even closer than this. We therefore expected few undetected multiple substitutions. Indeed, some separations are so recent that few substitutions have occurred at all (see fig. 1), and in none of the nine island-mainland comparisons was the difference in nonsynonymous substitutions between island and mainland branches statistically significant for that comparison alone (although several approach significance; P 0.0625). However, all nine comparisons showed differences in the direction predicted by the nearly-neutral theory, and this consistent bias was highly significant. (If one comparison had gone the other way, then the one-tailed significance level would have dropped to P 0.02). Thus, the pattern is one of modest but predictable increases in the rate of nonsynonymous substitution, beginning (apparently) soon after a lineage becomes isolated on an island. The relatively young ages of these comparisons suggest that it is possible to detect an increased rate of slightly deleterious substitution over timescales much shorter than those of the comparisons between endosymbiotic and free-living bacteria (tens to hundreds of millions of years). In principle, most of the excess nonsynonymous substitutions in island populations might have occurred during bottlenecks shortly after the island populations were founded. In most cases, we cannot say whether the substitutions on an island branch occurred early, late, or all along the branch. However, the ducks A. a. aucklandica and A. a. nesiotis are insular sister taxa, and replacement substitutions appear both on the ancestral (presumably insular) branch that unites them and on the terminal branches after the two subspecies diverge (see fig. 1b). This case would seem to support the inference that replacement substitutions may continue to accumulate after an island lineage is established, at least if the island and the resulting population size are very small. The brown teal (A. chlorotis) of New Zealand could be classified as an island species, along with its sisters A. a. aucklandica and A. a. nesiotis. The phylogenetically weighted nonsynonymous length of the branch that includes these three taxa is 5.75 (calculated as 4 0.5(0 (1 0.5(2 3))); see fig. 1b), and the phylogenetically weighted branch length for all mainland ducks in the large sister taxon (Anas capensis through Anas platyrhynchos 1) is 4.19. Thus, even in this high-order contrast, the insular branch shows more nonsynonymous substitutions than the mainland branch. The brown teal alone (four nonsynonymous substitutions) does not show a longer branch than the average of its mainland sisters (4.19), but it does show a higher K A /K S ratio (0.038 vs. an average of 0.025 for 10 mainland species distributed evenly throughout its large sister clade), owing to a somewhat lower rate of synonymous substitution on its branch. Two features of vertebrate mitochondrial genomes may make them especially likely to suffer deleterious substitutions: they do not recombine, and they experience relatively high mutation rates (Muller 1964; Felsenstein 1974; Lynch 1996, 1997; Lynch and Blanchard 1998). In birds as in other vertebrates, mitochondrial mutation rates appear to be much higher than nuclear mutation rates (Prychitko and Moore 1997; Johnson and Clayton 2000b). Thus, although deleterious mutations appear to accumulate on short timescales in the mitochondrial genomes of island birds, this does not imply that similar excess accumulations would necessarily be seen in their nuclear genomes, which recombine and experience much lower mutation rates. Although it seems reasonable to interpret the increased rate of replacement substitution in island birds as evidence of the predicted population size effect on the rate of accumulation of deleterious mutations, there

Mitochondrial Substitutions in Island Birds 879 are alternative explanations for the pattern. Perhaps most plausibly, many of these substitutions could be beneficial mutations that were carried to fixation by positive selection in novel island environments. Some of the islands do contain somewhat novel habitats, but in many cases they are similar to mainland habitats. For example, large islands such as Madagascar have wetland habitats similar to those used by related ducks on mainland Africa. Wernegreen and Moran (1999) argued that positive selection for replacement substitutions would be unlikely to affect all loci equally in endosymbiotic bacteria, and a similar argument might be made here (if less strongly). Both cyt b and ND2 individually show increased rates of replacement substitution (data not shown). Another line of evidence suggests that many or most nonsynonymous substitutions in these genes may be deleterious even in mainland populations. On average, there appears to be a slight excess of nonsynonymous substitution in terminal branches as compared with internal branches embedded more deeply in the tree. This effect can be seen by comparing the relative lengths of the same branches in figure 1a (synonymous substitutions) and figure 1b (nonsynonymous substitutions). This is counter to expectation, because rarer substitutions (e.g., nonsynonymous) are expected to be conserved and thus more readily reconstructed on deeper branches. An apparent dearth of amino acid substitutions deep in the tree (relative to synonymous substitutions) would be expected if many nonsynonymous substitutions were eventually reversed, owing to selection in favor of back-mutations. On this model, we would often fail to reconstruct slightly deleterious substitutions that occurred on deep branches but were later overprinted by advantageous reversals. To distinguish further between the deleterious and beneficial mutation hypotheses, the sequences of avian mitochondrial trna and rrna genes could be examined for evidence of reduced structural stability, as found by Lambert and Moran (1998) for bacterial endosymbionts. A second alternative is that these substitutions are indeed deleterious, but island populations experience relaxed selection. Island birds often become flightless, and in principle this could relax selection on peak metabolic performance. Some of the island ducks considered here tend to run rather than fly (e.g., Anas laysanensis and A. aucklandica), but we also find elevated rates of replacement substitution on large islands with many predators, where birds expend energy in ways similar to those of mainland taxa (e.g., Anas bernieri and Anas melleri on Madagascar). Competition for resources can be extremely intense on islands, and many island species show high levels of aggression relative to mainland species (Williams, McKinney, and Norman 1991). In fact, several island ducks have basal metabolic rates roughly 10% higher than those of closely related mainland species (Williams, McKinney, and Norman 1991). Wernegreen and Moran (1999) argue that relaxed selection (like positive selection) is also unlikely to affect all loci equally because certain loci will still be under strong constraint for function. As mentioned above, we find similar effects in cyt b and ND2 and in ducks and doves. An analysis of synonymous substitutions toward and away from codons that are hypothetically preferred (ones that match the anticodons of mitochondrially encoded trnas and tend to be most common in genes encoded on the majority strand of the mitochondrial chromosome) did not reveal the predicted relative bias toward unpreferred codons in island lineages. We see at least two possible explanations (not mutually exclusive) for our failure to find an island effect on synonymous substitutions. First, even the largest bird populations may be smaller than is necessary for effective selection against unpreferred synonymous codons. Many insect and bacterial populations are likely to be much larger than most bird populations, but the smallest could be nearly as small, giving these taxa a much wider range of population sizes than birds. On this interpretation, synonymous codon frequencies at most positions in most bird genomes would be close to their mutational equilibria. Second, there may be little selective preference among synonymous codons in the avian mitochondrial genome. We know of no direct evidence that the most common codons are in fact translated more quickly or more accurately than less common codons in the mitochondrion. Perhaps the posttranscriptional modifications of bases in and near the anticodons of mitochondrial trnas nearly equalize their performance on all of the synonymous codons they translate, especially the ones kept at the highest frequencies by mutation. On this interpretation as well, avian synonymous codon frequencies might be very near their mutational equilibria in populations of all sizes. These and other hypotheses might be tested by analyzing substitution patterns in ND6, the only protein gene in the avian mitochondrial genome that is encoded on the minority strand. ND6 is adjacent to cyt b, so presumably it experiences a similar mutational spectrum. Its frequencies of third-position nucleotides are approximately those of the complementary bases in cyt b (i.e., with T and G at high frequencies and A and C at low frequencies) in the mitochondrial genome of the redhead duck Aythya americana (Mindell et al. 1999), consistent with the hypothesis that the third-position nucleotide distribution is determined mainly by mutational biases. However, if the mitochondrial translational system is adapted to high frequencies of A- and C-ending codons, then it seems possible that weak selection to reduce the frequencies of T- and G-ending codons in ND6 might be detectable, at least at certain critical amino acid positions in the ND6 protein (Rand and Kann 1998; Ballard 2000). In summary, rates of nonsynonymous substitution in mitochondrial protein-coding genes of island birds appear to be higher than those in close mainland relatives. This finding supports the nearly-neutral model, in which deleterious mutations accumulate by drift in small populations more rapidly than in large ones. There are conceivable alternative explanations for the pattern, but even if true in part, they seem unlikely to explain why

880 Johnson and Seger the same substitutional bias appears so consistently in recently derived, evolutionarily independent island lineages. This pattern adds to a growing body of evidence suggesting that mitochondrial genomes may be especially vulnerable to accumulation of slightly deleterious mutations. Acknowledgments We thank David Rand, Fred Adler, Joshua Cherry, Dale Clayton, Steven Proulx, and two anonymous reviewers for helpful discussions and comments on the manuscript. We thank Joe Felsenstein for the prerelease version of DNAML, version 3.6. This work was supported in part by NSF CAREER award DEB-9703003 to D. H. Clayton. LITERATURE CITED AKASHI, H. 1995. Inferring weak selection from patterns of polymorphism and divergence at silent sites in Drosophila DNA. Genetics 139:1067 1076. ARAKI, H., and H. TACHIDA. 1997. Bottleneck effect on evolutionary rate in the nearly neutral mutation model. Genetics 147:907 914. BALLARD, J. W. O. 2000. Comparative genomics of mitochondrial DNA in members of the Drosophila melanogaster subgroup. J. Mol. Evol. 51:48 63. CHARLESWORTH, D., M. T. MORGAN, and B. CHARLESWORTH. 1993. Mutation accumulation in finite populations. J. Hered. 84:321 325. CLAYTON, D. H., and P. COTGREAVE. 1994. Relationship of bill morphology to grooming behaviour in birds. Anim. Behav. 47:195 201. DESALLE, R., and A. R. TEMPLETON. 1988. Founder effects and the rate of mitochondrial DNA evolution in Hawaiian Drosophila. Evolution 42:1076 1084. DESJARDINS, P., and R. MORAIS. 1990. Sequence and gene organization of the chicken mitochondrial genome. J. Mol. Biol. 212:599 634. EASTEAL, S., and C. COLLET. 1994. Consistent variation in amino-acid substitution rate, despite uniformity of mutation rate: protein evolution in mammals is not neutral. Mol. Biol. Evol. 11:643 647. FELSENSTEIN, J. 1974. The evolutionary advantage of recombination. Genetics 78:737 756.. 1989. PHYLIP phylogeny inference package (version 3.2). Cladistics 5:164 166. FITCH, W. M., and M. BRUSCHI. 1987. The evolution of prokaryotic ferredoxins with a general method for correcting for unobserved substitutions with less branched lineages. Mol. Biol. Evol. 4:381 394. FRY, A. 1999. Mildly deleterious mutations in avian mitochondrial DNA: evidence from neutrality tests. Evolution 53:1617 1620. GILLESPIE, J. H. 1991. The causes of molecular evolution. Oxford University Press, New York. HASEGAWA, M., Y. CAO, and Z. YANG. 1998. Preponderance of slightly deleterious polymorphism in mitochondrial DNA: nonsynonymous/synonymous rate ratio is much higher within species than between species. Mol. Biol. Evol. 15: 1499 1505. JEHLE, J.R.JR., and K. C. PARKES. 1982. The status of the avifauna of the Revillagigedo Islands, Mexico. Wilson Bull. 94:1 19. JOHNSON, K. P., and D. H. CLAYTON. 2000a. A molecular phylogeny of the dove genus Zenaida: mitochondrial and nuclear DNA sequences. Condor 102:864 870.. 2000b. Nuclear and mitochondrial genes contain similar phylogenetic signal for pigeons and doves (Aves: Columbiformes). Mol. Phylogenet. Evol. 14:141 151. JOHNSON, K. P., and M. D. SORENSON. 1998. Comparing molecular evolution in two mitochondrial protein coding genes (cytochrome b and ND2) in the dabbling ducks (tribe: Anatini). Mol. Phylogenet. Evol. 10:82 94.. 1999. Phylogeny and biogeography of dabbling ducks (genus: Anas): a comparison of molecular and morphological evidence. Auk 116:792 805. KIMURA, M. 1962. On the probability of fixation of mutant genes in a population. Genetics 47:713 719.. 1983. The neutral theory of molecular evolution. Cambridge University Press, Cambridge, England. KLICKA, J. T., and R. M. ZINK. 1997. The importance of recent ice ages in speciation: a failed paradigm. Science 277:1666 1669. KONDRASHOV, A. S. 1995. Contamination of the genome by very slightly deleterious mutations: why have we not died 100 times over? J. Theor. Biol. 175:583 594. LAMBERT, J. D., and N. A. MORAN. 1998. Deleterious mutations destabilize ribosomal RNA in endosymbiotic bacteria. Proc. Natl. Acad. Sci. USA 95:4458 4462. LANDE, R. 1998. Risk of population extinction from fixation of deleterious and reverse mutations. Genetica 102/103:21 27. LI, W. H. 1987. Models of nearly neutral mutations with particular implications for nonrandom usage of synonymous codons. J. Mol. Evol. 24:337 345.. 1993. Unbiased-estimation of the rates of synonymous and nonsynonymous substitution. J. Mol. Evol. 36:96 99. LLOPART, A., and M. AGUADÉ. 1999. Synonymous rates at the RpII215 gene of Drosophila: variation among species and across the coding region. Genetics 152:269 280. LYNCH, M. 1996. Mutation accumulation in transfer RNAs: molecular evidence for Muller s ratchet in mitochondrial genomes. Mol. Biol. Evol. 13:209 220.. 1997. Mutation accumulation in nuclear, organelle, and prokaryotic transfer RNA genes. Mol. Biol. Evol. 14: 914 925. LYNCH, M., and J. L. BLANCHARD. 1998. Deleterious mutation accumulation in organelle genomes. Genetica 102/103:29 39. MINDELL, D. P., M. D. SORENSON, D. E. DIMCHEFF, M. HAS- EGAWA, J. C. AST, and T. YURI. 1999. Interordinal relationships of birds and other reptiles based on whole mitochondrial genomes. Syst. Biol. 48:138 152. MULLER, H. J. 1964. The relation of recombination to mutational advance. Mutat. Res. 1:2 9. NACHMAN, M. W. 1998. Deleterious mutations in animal mitochondrial DNA. Genetica 102/103:61 69. NACHMAN, M. W., W. M. BROWN, M.STONEKING, and C. F. AQUADRO. 1996. Nonneutral mitochondrial DNA variation in humans and chimpanzees. Genetics 142:953 963. OHTA, T. 1973. Slightly deleterious mutant substitutions in evolution. Nature 246:96 98.. 1992. The nearly neutral theory of molecular evolution. Annu. Rev. Ecol. Syst. 23:263 286.. 1995. Synonymous and nonsynonymous substitutions in mammalian genes and the nearly neutral theory. J. Mol. Evol. 40:56 63. PAMILO, P., and N. O. BIANCHI. 1993. Evolution of the Zfx and Zfy genes: rates and interdependence between genes. Mol. Biol. Evol. 10:271 281.

Mitochondrial Substitutions in Island Birds 881 PARSONS, T. J., D. S. MUNIEC, K. SULLIVAN et al. (11 coauthors). 1997. A high observed substitution rate in the human mitochondrial DNA control region. Nat. Genet. 15: 363 367. PRYCHITKO, T. M., and W. S. MOORE. 1997. The utility of DNA sequences of an intron from the beta-fibrinogen gene in phylogenetic analysis of woodpeckers (Aves: Picidae). Mol. Phylogenet. Evol. 8:193 204. RAND, D. M., and L. M. KANN. 1996. Excess amino acid polymorphism in mitochondrial DNA: contrasts among genes from Drosophila mice and humans. Mol. Biol. Evol. 13: 735 748.. 1998. Mutation and selection at silent and replacement sites in the evolution of animal mitochondrial DNA. Genetica 102/103:393 407. SANDERSON, M. J. 1990. Estimating rates of speciation and evolution: a bias due to homoplasy. Cladistics 6:387 391. TACHIDA, H. 1991. A study on a nearly neutral mutation model in finite populations. Genetics 128:183 192.. 1996. Effects of the shape of distribution of mutant effect in nearly neutral mutation models. J. Genet. 75:33 48. WERNEGREEN, J. J., and N. A. MORAN. 1999. Evidence for genetic drift in endosymbionts (Buchnera): analysis of protein-coding genes. Mol. Biol. Evol. 16:83 97. WILLIAMS, M., F. MCKINNEY, and F. I. NORMAN. 1991. Ecological and behavioural responses of austral teal to island life. Acta XX Cong. Int. Ornithol. pp. 876 884. WISE, C. A., M. SRAML, and S. EASTEAL. 1998. Departure from neutrality at the mitochondrial NADH dehydrogenase subunit 2 gene in humans, but not in chimpanzees. Genetics 148:409 421. XIA, X. 1996. Maximizing transcription efficiency causes codon usage bias. Genetics 144:1309 1320. DAVID RAND, reviewing editor Accepted January 24, 2001