The Effects of Sex-Biased Gene Expression and X-Linkage on Rates of Sequence Evolution in Drosophila

The Effects of Sex-Biased Gene Expression and X-Linkage on Rates of Sequence Evolution in Drosophila Abstract A faster rate of adaptive evolution of X-linked genes compared with autosomal genes (the faster-X effect) can be caused by the fixation of recessive or partially recessive advantageous mutations. This effect should be largest for advantageous mutations that affect only male fitness, and least for mutations that affect only female fitness. We tested these predictions in Drosophila melanogaster by using coding and functionally significant noncoding sequences of genes with different levels of sex-biased expression. Consistent with theory, nonsynonymous substitutions in most male-biased and unbiased genes show faster adaptive evolution on the X. However, genes with very low recombination rates do not show such an effect, possibly as a consequence of Hill–Robertson interference. Contrary to expectation, there was a substantial faster-X effect for female-biased genes. After correcting for recombination rate differences, however, female-biased genes did not show a faster X-effect. Similar analyses of noncoding UTRs and long introns showed a faster-X effect for all groups of genes, other than introns of female-biased genes. Given the strong evidence that deleterious mutations are mostly recessive or partially recessive, we would expect a slower rate of evolution of X-linked genes for slightly deleterious mutations that become fixed by genetic drift. Surprisingly, we found little evidence for this after correcting for recombination rate, implying that weakly deleterious mutations are mostly close to being semidominant. This is consistent with evidence from polymorphism data, which we use to test how models of selection that assume semidominance with no sex-specific fitness effects may bias estimates of purifying selection. Drosophila, faster-X effect, gene expression, natural selection, sex bias Introduction The differences between the modes of inheritance of genes on the X chromosome and the autosomes are expected to affect their patterns of variation and evolution (reviewed by Meisel and Connallon 2013). Under suitable conditions with respect to the levels of dominance of new mutations, sex differences in selection coefficients, recombination rates, and effective population size, theory shows that X-linked genes can evolve under positive selection more rapidly than autosomal genes—the so-called “faster-X effect.” This reflects the fact that the hemizygosity of the X chromosome in males (in species with male heterogamety) means that positive selection can sometimes be more effective on new X-linked mutations than on autosomal mutations with equivalent effects on fitness (Haldane 1924; Charlesworth et al. 1987; Meisel and Connallon 2013). This is, however, only likely to happen when evolutionary change depends on new mutations; fixation of variants present as standing variation tends to favor a slower rate of evolution of X-linked genes (Charlesworth et al. 1987; Orr and Betancourt 2001; Meisel and Connallon 2013). An important consequence of the faster-X effect is that X (or Z) chromosomes are likely to play a disproportionate role in speciation and evolutionary divergence (Orr and Coyne 1989; Presgraves 2008). Faster-X evolution from new mutations is facilitated by partial recessivity of beneficial mutations (dominance coefficients less than one-half) (Charlesworth et al. 1987); and by a higher ratio of the effective population size of X-linked to autosomal genes (NeX/NeA) than the value of three-quarters expected under the standard Wright–Fisher model, which assumes equal variances in reproductive success in males and females (Vicoso and Charlesworth 2009; Meisel and Connallon 2013). Genes that only affect male fitness are more likely to exhibit a faster-X effect than genes that affect the fitnesses for both sexes. In contrast, genes that affect only female fitness will not experience a faster-X effect unless NeX/NeA> ¾ (Charlesworth et al. 1987; Vicoso and Charlesworth 2009; Mank et al. 2010; Meisel and Connallon 2013). With sexually antagonistic effects of mutations on fitness, a faster-X effect can arise with recessive or partially recessive mutations that increase male fitness at the expense of female fitness, or dominant or partially dominant mutations that increase female fitness at the expense of male fitness (Rice 1984; Charlesworth et al. 1987). These theoretical expectations are supported by recent studies of D. melanogaster and its close relatives, using whole genome sequence comparisons between species and resequencing of multiple individuals of D. melanogaster. These studies showed that there is a faster rate of protein sequence evolution on the X chromosome (X) than on the autosomes (A), and that this is attributable to a greater incidence of fixations of selectively favorable nonsynonymous mutations on X compared with A (Langley et al. 2012; Mackay et al. 2012; Hu et al. 2013; Campos et al. 2014; Ávila et al. 2015). In a previous study (Ávila et al. 2015), we tested the above predictions of the theory concerning sex-specific fitness effects by using genes with different levels of sex-biased expression (unbiased, male- and female-biased) in Drosophila melanogaster, and estimating the extent of adaptive evolution of non-synonymous mutations. We detected a faster-X effect for all categories of gene expression, including female-biased genes. The sex bias in gene expression for each gene was used as a proxy for male/female fitness, assuming that unbiased genes are likely to have similar fitness effects on both sexes (or to be sexually antagonistic), male-biased genes to affect male fitness more than female fitness, and vice-versa for female-biased genes. A faster-X effect was found for female-biased genes, even after adjusting for the effects of differences in recombination rate between X and A that reflects differences in Ne caused by selection at linked sites; such effects were detected by Campos et al. (2013) (see their fig. 2). However, Ávila et al. (2015) used a meta-analysis of Drosophila gene expression, most of which included whole-fly expression data. A FDR cut-off of 20% was used to classify genes as sex-biased. With this liberal criterion, “female-biased” X-linked genes could still have significant expression levels in males, and hence be exposed to positive selection on the hemizygous X. Here, we use a much more stringent criterion for sex-biased gene expression, to ensure that female-biased genes can have little effect on male fitness. Faster-X divergence for gene expression has been reported in Drosophila (Kayserili et al. 2012; Meisel et al. 2012), possibly driven by adaptive and recessive cis-regulatory mutations that affect gene expression levels (Coolon et al. 2015). A theoretical framework for how adaptive cis-regulatory mutations could be recessive was proposed by Meisel et al. (2012). Previous work has shown that many noncoding regions are functional and experience both negative and positive selection (Andolfatto 2005; Haddrill et al. 2005; Halligan and Keightley 2006; Halligan et al. 2013). In addition, a faster-X effect for noncoding regions has been described in Drosophila (Begun et al. 2007; Mackay et al. 2012; Hu et al. 2013), which could be related to the faster X-effect on gene expression. We have therefore extended the Ávila et al. (2015) analysis to noncoding regions, including 5′ UTRs, 3′ UTRs and long introns (introns shorter than 100 bp appear to experience little selective constraint [Halligan and Keightley 2006; Parsch et al. 2010]). We also tested for the existence of a slower-X effect for deleterious mutations. Much evidence suggests that mutations are partially recessive in their effects on fitness (Dobzhansky and Wright 1941; Muller 1950; Crow and Simmons 1983; García-Dorado and Caballero 2000; Manna et al. 2011). Purifying selection is thus usually expected to be stronger on the X than A, and a lower rate of fixation of slightly deleterious mutations is expected for X than for A, unless the X/A ratio of Ne is < ¾ (Mank et al. 2010; Vicoso and Charlesworth 2009). However, there is only weak empirical evidence for a slower-X (or Z) effect (Carneiro et al. 2012; Hvilsom et al. 2012; Veeramah et al. 2014; Rousselle et al. 2016). We tested for this effect by estimating the proportion of slightly deleterious mutations fixed on X and A in a comparison of a pair of closely related species. Overall, we did not find any substantial differences between X and A, suggesting that slightly deleterious mutations may be close to being semidominant. This was consistent with the results of comparisons between X and A of nucleotide site diversities at functional sites. Results Rates of Adaptive Evolution on X versus A A total of 7,853 D. melanogaster genes were available for this study (see supplementary table 1, Supplementary Material online). As described in the Materials and Methods, we used a data set on germline gene expression to classify genes as male-biased, female-biased and unbiased, with a much more stringent criterion for sex-biased expression than that of Ávila et al. (2015). For this data set, the largest group was that for unbiased A genes (N = 4,529), and the smallest group was that for female-biased X genes (N = 64). The distribution of the three categories of expression between A and X was significantly different (χ2 = 23.37, P < 0.01), with a slight excess on the X of female-biased (20.9%) and unbiased (16.8%) genes, and a deficit of male-biased (12.9%) genes; based on the number of X-linked genes (1250/7853), approximately 16% of genes in a given class would be expected to be on the X chromosome. In order to minimize the possibility that a gene that shows biased gene expression in the germline of one sex may be expressed in the soma of the opposite sex, we also used gene expression levels in heads to quantify sex specificity even more stringently, as described in the Materials and Methods. After doing this, our sex-biased data set was considerably smaller, with only 39 female-biased genes, but the distribution of genes among X and A was almost identical to that for the germline data set (unbiased, 16.3% on the X; male-biased, 12.9% on the X; female-biased, 20.8% on the X). For all three categories in the germline-only data set (unbiased, male- and female-biased), genes on the X showed a higher rate of adaptive nonsynonymous site evolution as measured by ωa, the rate of substitution of positively selected mutations for a given class of functional sites relative to that for synonymous mutations (fig. 1 and supplementary table 1, Supplementary Material online). UTRs and long introns also showed a faster-X effect, although this was not significant for female-biased genes in long introns (fig. 1 and supplementary table 1, Supplementary Material online). Fig. 1. View largeDownload slide Plots of ωa values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased, and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with larger ωa value for A than X. Fig. 1. View largeDownload slide Plots of ωa values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased, and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with larger ωa value for A than X. Use of the more conservative data set that used expression data on heads gave results that were similar to those for the germline-only data sets, with the sole exception of female-biased 3′ UTRs (supplementary table 2, Supplementary Material online). Overall, the results of Ávila et al. (2015) were confirmed by the present analysis, suggesting that factors other than dominance level may be involved in the faster-X effect, at least for female-biased genes. As seen in previous studies (Baines et al. 2008; Parsch and Ellegren 2013; Ávila et al. 2014; Kousathanas et al. 2014; Ávila et al. 2015), male-biased genes have the fastest rates of adaptive evolution, for both X and A (fig. 1 and supplementary tables 1 and 2, Supplementary Material online). Factors Potentially Influencing the Faster-X Effect Narrower breadth and lower levels of gene expression are known to be associated with increased rates of sequence evolution (Larracuente et al. 2008; Meisel et al. 2012). Using the criteria described in the Materials and Methods, we found that, overall, X genes had a narrower breadth of gene expression than A genes (P = 0.02). This suggests that ωa could be higher for the X because of a difference in breadth of gene expression. However, we observed no significant differences in expression breadth between X and A within the three pairs of groups of interest (unbiased, male-biased and female-biased genes) (supplementary table 3, Supplementary Material online). We also found no difference in average expression levels between X and A (P = 0.58). When we compared X and A within the three pairs of groups of interest, we again found no significant differences (supplementary table 3, Supplementary Material online). Thus, neither breadth of expression nor average expression seem to be involved in the higher rate of adaptive evolution of female-biased genes for X compared with A. The rate of adaptive sequence evolution is also known to be positively correlated with the rate of crossing over per Megabase (Campos et al. 2014; Charlesworth and Campos 2014; Castellano et al. 2016). As was found in previous studies (Comeron et al. 2012; Campos et al. 2013), the X chromosome had a significantly higher effective rate of crossing over than the autosomes: mean for A = 1.14 cM/Mb, with 95% confidence interval (CI) (1.13, 1.16), and mean for X = 1.96 cM/Mb (CI: 1.86, 1.95) (this statistic corrects for the differential transmission of X and A through females and males: see Materials and Methods for details). Campos et al. (2013) showed that NeX/NeA (estimated from the X/A ratio of 4-fold degenerate site diversities) for X and A genes with comparable effective recombination rates was close to ¾, except for genes in regions with very low rates of crossing over. Comparisons of X and A genes in regions with similar effective recombination rates should therefore largely remove any effects of departures of NeX/NeA from ¾, which might cause a faster-X effect for female-biased genes (see the introduction). To adjust for differences in effective recombination rates between X and A, we divided the genes for both X and A into four bins of recombination rates, using the germline data set, because it had a larger sample size for male- and female-biased genes; as shown above, the initial results were similar for both data sets. The effective recombination rates for A and X genes did not overlap for rates above ∼2 cM/Mb, so that only the first two bins for X overlap those for A (see fig. 2). For nonsynonymous mutations, the confidence intervals for ωa for the bins of unbiased and male-biased X genes with the lowest recombination rate overlapped those for A (fig. 2 and supplementary table 4, Supplementary Material online). However, the remaining X bins had significantly higher ωa values than the corresponding A bins, including the second X bin with a similar effective recombination rate to the fourth bin of autosomal genes (proportion of ωa values greater on the A: P = 0 and P = 0.016 for unbiased and male-biased genes, respectively). In contrast, female-biased genes (F) with similar effective recombination rates did not show a faster-X effect (fig. 2 and supplementary table 4, Supplementary Material online; 2nd A bin vs. 1st X bin, P = 0.349; 4th A bin vs. 2nd X bin, P = 0.92). Due to the small sample size of the first two XF bins that overlap in recombination with AF, we also grouped them into a single bin (21 genes); however, there was no significant difference from the whole AF data set (P = 0.622, ωa XF = 0.057 vs. ωaAF = 0.063). Fig. 2. View largeDownload slide Plots of ωa against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωa values. Because of the small number of female-biased genes, only three bins were used for long introns. Fig. 2. View largeDownload slide Plots of ωa against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωa values. Because of the small number of female-biased genes, only three bins were used for long introns. We carried out similar analyses for noncoding sequences (fig. 2 and supplementary table 4, Supplementary Material online). For unbiased and male-biased genes, the patterns for UTRs and long introns were similar to those for nonsynonymous sites, such that ωa for the X bin with the lowest recombination rate did not differ significantly from that for A genes with similar recombination rates (P > 0.05). The remaining bins for UTRs and long introns of unbiased and male-biased genes showed a faster-X effect, including the 2nd X bin that overlaps in recombination rate with the 4th A bin (P = 0). For UTRs of female-biased genes, there was a faster-X effect across all bins (P < 0.05). This result was stronger for 5′ UTRs (supplementary fig. 1, Supplementary Material online; 2nd A bin—1st X bin, P = 0.056; 4th A bin—2nd X bin, P = 0.008) but was also seen with 3′ UTRs (supplementary fig. 1, Supplementary Material online; 2nd A bin—1st X bin, P = 0.014; 4th A bin—2nd X bin, P = 0.179). For female-biased genes, there was a faster-X effect only for long introns in bins with a higher effective recombination rate on the X (fig. 2; 2nd A bin—1st X bin, P = 0.91; note that, due to the small sample size, there is only one X bin that overlaps with A in recombination rate). Purifying Selection on X and A We also performed several tests to compare the strength of purifying selection on mutations in X and A genes, since the exposure of mutations to selection in the hemizygous state in males is expected to result in larger selection coefficients for rare, deleterious X-linked mutations relative to comparable autosomal mutations, with the exception of mutations with male-limited effects and a dominance coefficient h ≥ 2/3, as shown in the overview of relevant population genetics theory in the Supplementary Material online. The theoretical results imply that parameters such as the rates of fixation of weakly deleterious mutations, and the ratios of diversity levels at sites under selection to those at neutral sites, are dependent on the product of the strength of selection and the effective population size (Ne). The interpretation of the measures described below is therefore complicated by possible differences in Nebetween X and A, especially as Ne may be affected by hitchhiking and hence by the rate of recombination for the genomic region in question, as discussed above (see also Vicoso and Charlesworth 2009; Charlesworth and Campos 2014). The first measure of the efficacy of purifying selection that we used was the estimate of the rate of nonadaptive substitutions at functional sites, relative to the rate for synonymous substitutions (ωna). For nonsynonymous sites, this was consistently lower for X than for A for all expression categories (fig. 3 and supplementary table 5, Supplementary Material online), suggesting a larger product of effective population size (Ne) and strength of selection against slightly deleterious mutations for X compared with A. The X—A differences were significant for the combined UTR data set for all categories of gene expression, as well as for 5′ UTRs, 3′ UTRs and introns of unbiased genes, for 5′ UTRs of female-biased genes, and for 3′UTRs of male-biased genes. Fig. 3. View largeDownload slide Plots of ωna values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with an ωna value larger for X than A. Fig. 3. View largeDownload slide Plots of ωna values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with an ωna value larger for X than A. However, ωna makes several assumptions that could affect the results (see the Discussion), especially the use of a model of selection that assumes semidominance with no sex-specific fitness effects (Eyre-Walker and Keightley 2009). We therefore used diversity statistics to estimate the efficacy of purifying selection; where appropriate, we applied correction factors to take into account possible differences in Ne between the X and A (see the theoretical section of the Supplementary Material online). For the diversity data for the chromosomes as a whole, except for very low recombination regions, our previous work showed that the effective population size for the X chromosome, NeX, is approximately equal to that for the autosomes, NeA, when X and A regions with similar effective recombination rates are compared (Campos et al. 2013). For semidominant mutations with h = ½ and sufficiently strong selection (Nes ≫ 1), diversity for functional sites on the X (πF) is expected to be approximately three-quarters of that for A, assuming that these mutations are sufficiently strongly selected that Nes ≫ 1; it will be less than this for partially recessive unbiased genes and male-biased genes. When NeX = NeA, the same should apply to the ratio of the mean pairwise nucleotide site diversity at putatively functional sites to the mean synonymous site diversity (πF/πS), assuming the latter to be close to neutral; this statistic has the advantage of reducing variability among different classes of genes caused by differences in their mutation rates. Tables 1 and 2 show the mean values and 95% bootstrap confidence intervals for πF and πF/πS, respectively, for each type of sequence and chromosome. For nonsynonymous sites, especially for πF/πS for unbiased genes, there is a clear indication of stronger selective constraints on X chromosomal genes; the same pattern is also noticeable for UTRs and introns in unbiased genes. When we applied a correction factor of 4/3 to the X, πF and πF/πSfor X were similar to, or larger than, the values for A (tables 1 and 2), so that there was no evidence for partial recessivity of deleterious mutations. As expected, πF/πS is larger for UTRs and introns than nonsynonymous sites. It is interesting to note that πF/πS for nonsynymous sites is larger for male-biased and female-biased than unbiased genes, paralleling what is seen for ωna (fig. 3 and supplementary table 5, Supplementary Material online), and suggesting that genes with sex-biased expression have lower levels of selective constraints. This pattern is not seen for UTRs and introns. Table 1. πF Values for All Expression Categories and Functional Classes. Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.00122 (0.00118, 0.00126)  0.00109 (0.00101, 0.00118)  0.00145 (0.00135, 0.00157)  1    Male  0.00161 (0.00153, 0.00171)  0.00161 (0.00139, 0.00181)  0.00214 (0.00185, 0.00242)  1    Female  0.00171 (0.00153, 0.00192)  0.00159 (0.00121, 0.00203)  0.00212 (0.00161, 0.00271)  0.919  5′ and 3´ UTRs  Unbiased  0.00370 (0.00362, 0.00379)  0.00359 (0.00343, 0.00375)  0.00479 (0.00457, 0.00500)  1    Male  0.00419 (0.00404, 0.00433)  0.00425 (0.00383, 0.00466)  0.00567 (0.00510, 0.00621)  1    Female  0.00394 (0.00359, 0.00435)  0.00393 (0.00338, 0.00453)  0.00524 (0.00450, 0.00604)  0.999  5′ UTRs  Unbiased  0.00400 (0.00389, 0.00412)  0.00374 (0.00350, 0.00398)  0.00499 (0.00467, 0.00531)  1    Male  0.00419 (0.00398, 0.00441)  0.00445 (0.00399, 0.00499)  0.00593 (0.00532, 0.00665)  1    Female  0.00453 (0.00395, 0.00522)  0.00389 (0.00308, 0.00472)  0.00518 (0.00411, 0.00630)  0.849  3´ UTRs  Unbiased  0.00350 (0.00341, 0.00359)  0.00351 (0.00331, 0.00368)  0.00467 (0.00442, 0.00491)  1    Male  0.00418 (0.00399, 0.00437)  0.00413 (0.00360, 0.00467)  0.00551 (0.00480, 0.00622)  1    Female  0.00356 (0.00316, 0.00400)  0.00392 (0.00314, 0.00475)  0.00523 (0.00419, 0.00634)  0.995  Introns  Unbiased  0.00470 (0.00459, 0.00481)  0.00472 (0.00455, 0.00490)  0.00630 (0.00607, 0.00653)  1    Male  0.00474 (0.00458, 0.00490)  0.00482 (0.00437, 0.00531)  0.00643 (0.00583, 0.00708)  1    Female  0.00444 (0.00406, 0.00488)  0.00458 (0.00413, 0.00497)  0.00611 (0.00550, 0.00663)  1  Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.00122 (0.00118, 0.00126)  0.00109 (0.00101, 0.00118)  0.00145 (0.00135, 0.00157)  1    Male  0.00161 (0.00153, 0.00171)  0.00161 (0.00139, 0.00181)  0.00214 (0.00185, 0.00242)  1    Female  0.00171 (0.00153, 0.00192)  0.00159 (0.00121, 0.00203)  0.00212 (0.00161, 0.00271)  0.919  5′ and 3´ UTRs  Unbiased  0.00370 (0.00362, 0.00379)  0.00359 (0.00343, 0.00375)  0.00479 (0.00457, 0.00500)  1    Male  0.00419 (0.00404, 0.00433)  0.00425 (0.00383, 0.00466)  0.00567 (0.00510, 0.00621)  1    Female  0.00394 (0.00359, 0.00435)  0.00393 (0.00338, 0.00453)  0.00524 (0.00450, 0.00604)  0.999  5′ UTRs  Unbiased  0.00400 (0.00389, 0.00412)  0.00374 (0.00350, 0.00398)  0.00499 (0.00467, 0.00531)  1    Male  0.00419 (0.00398, 0.00441)  0.00445 (0.00399, 0.00499)  0.00593 (0.00532, 0.00665)  1    Female  0.00453 (0.00395, 0.00522)  0.00389 (0.00308, 0.00472)  0.00518 (0.00411, 0.00630)  0.849  3´ UTRs  Unbiased  0.00350 (0.00341, 0.00359)  0.00351 (0.00331, 0.00368)  0.00467 (0.00442, 0.00491)  1    Male  0.00418 (0.00399, 0.00437)  0.00413 (0.00360, 0.00467)  0.00551 (0.00480, 0.00622)  1    Female  0.00356 (0.00316, 0.00400)  0.00392 (0.00314, 0.00475)  0.00523 (0.00419, 0.00634)  0.995  Introns  Unbiased  0.00470 (0.00459, 0.00481)  0.00472 (0.00455, 0.00490)  0.00630 (0.00607, 0.00653)  1    Male  0.00474 (0.00458, 0.00490)  0.00482 (0.00437, 0.00531)  0.00643 (0.00583, 0.00708)  1    Female  0.00444 (0.00406, 0.00488)  0.00458 (0.00413, 0.00497)  0.00611 (0.00550, 0.00663)  1  Note.—πFfor the X shown in the fifth column was corrected by multiplying by 4/3. (95% bootstrap confidence intervals in brackets.) P, Proportion of bootstraps in which the value of πF for X (after multiplication by 4/3) was greater than that for A. Table 2. πF/πS Values for All Categories and Gene Features. Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.0882 (0.0851, 0.0912)  0.0695 (0.0644, 0.0753)  0.0927 (0.0859, 0.1004)  0.866    Male  0.0997 (0.0938, 0.1062)  0.0933 (0.0789, 0.1091)  0.1244 (0.1053, 0.1455)  0.997    Female  0.1082 (0.0935, 0.1254)  0.0827 (0.0596, 0.1107)  0.1103 (0.0794, 0.1476)  0.52  5′ and 3′ UTR  Unbiased  0.2732 (0.2661, 0.2810)  0.2335 (0.2213, 0.2451)  0.3113 (0.2950, 0.3268)  1    Male  0.2581 (0.2462, 0.2702)  0.2492 (0.2246, 0.2740)  0.3322 (0.2995, 0.3653)  1    Female  0.2530 (0.2221, 0.2857)  0.2141 (0.1719, 0.2683)  0.2854 (0.2292, 0.3577)  0.8  5′ UTRs  Unbiased  0.2970 (0.2875, 0.3065)  0.2425 (0.2263, 0.2595)  0.3233 (0.3017, 0.3459)  0.985    Male  0.2571 (0.2420, 0.2728)  0.2626 (0.2292, 0.2998)  0.3501 (0.3055, 0.3998)  1    Female  0.2916 (0.2501, 0.3413)  0.2083 (0.1539, 0.2724)  0.2778 (0.2052, 0.3632)  0.384  3′ UTRs  Unbiased  0.2590 (0.2504, 0.2669)  0.2278 (0.2139, 0.2426)  0.3038 (0.2852, 0.3235)  1    Male  0.2559 (0.2418, 0.2699)  0.2429 (0.2141, 0.2728)  0.3239 (0.2855, 0.3638)  0.999    Female  0.2289 (0.1974, 0.2625)  0.2125 (0.1667, 0.2682)  0.2833 (0.2223, 0.3576)  0.915  Introns  Unbiased  0.3331 (0.3207, 0.3456)  0.3087 (0.2912, 0.3258)  0.4115 (0.3883, 0.4345)  1    Male  0.2818 (0.2663, 0.2981)  0.2887 (0.2513, 0.3334)  0.3849 (0.3351, 0.4446)  1    Female  0.2436 (0.2129, 0.2791)  0.2471 (0.2079, 0.3052)  0.3294 (0.2772, 0.4069)  0.997  Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.0882 (0.0851, 0.0912)  0.0695 (0.0644, 0.0753)  0.0927 (0.0859, 0.1004)  0.866    Male  0.0997 (0.0938, 0.1062)  0.0933 (0.0789, 0.1091)  0.1244 (0.1053, 0.1455)  0.997    Female  0.1082 (0.0935, 0.1254)  0.0827 (0.0596, 0.1107)  0.1103 (0.0794, 0.1476)  0.52  5′ and 3′ UTR  Unbiased  0.2732 (0.2661, 0.2810)  0.2335 (0.2213, 0.2451)  0.3113 (0.2950, 0.3268)  1    Male  0.2581 (0.2462, 0.2702)  0.2492 (0.2246, 0.2740)  0.3322 (0.2995, 0.3653)  1    Female  0.2530 (0.2221, 0.2857)  0.2141 (0.1719, 0.2683)  0.2854 (0.2292, 0.3577)  0.8  5′ UTRs  Unbiased  0.2970 (0.2875, 0.3065)  0.2425 (0.2263, 0.2595)  0.3233 (0.3017, 0.3459)  0.985    Male  0.2571 (0.2420, 0.2728)  0.2626 (0.2292, 0.2998)  0.3501 (0.3055, 0.3998)  1    Female  0.2916 (0.2501, 0.3413)  0.2083 (0.1539, 0.2724)  0.2778 (0.2052, 0.3632)  0.384  3′ UTRs  Unbiased  0.2590 (0.2504, 0.2669)  0.2278 (0.2139, 0.2426)  0.3038 (0.2852, 0.3235)  1    Male  0.2559 (0.2418, 0.2699)  0.2429 (0.2141, 0.2728)  0.3239 (0.2855, 0.3638)  0.999    Female  0.2289 (0.1974, 0.2625)  0.2125 (0.1667, 0.2682)  0.2833 (0.2223, 0.3576)  0.915  Introns  Unbiased  0.3331 (0.3207, 0.3456)  0.3087 (0.2912, 0.3258)  0.4115 (0.3883, 0.4345)  1    Male  0.2818 (0.2663, 0.2981)  0.2887 (0.2513, 0.3334)  0.3849 (0.3351, 0.4446)  1    Female  0.2436 (0.2129, 0.2791)  0.2471 (0.2079, 0.3052)  0.3294 (0.2772, 0.4069)  0.997  Note.—πF/πSfor the X shown in the fifth column was corrected by multiplying by 4/3. (95% bootstrap confidence intervals in brackets.) P, Proportion of bootstraps in which the value of πF for X (after multiplication by 4/3) was greater than that for A. We also examined how the higher overall effective recombination rate of the X might affect these statistics, by increasing Ne as a result of weaker hitchhiking effects. For all types of functional sequence, there was no difference in ωna between X and A for regions that overlap in their effective recombination rates (fig. 4), suggesting that the difference in effective population sizes between X and A cancels the X-A differences in selection coefficients against new deleterious mutations (see the theoretical section of the Supplementary Material online). The same pattern was found for πF/πS (supplementary fig. 2 and table 4, Supplementary Material online). Since our previous work showed that the ratio NeX/NeA was approximately ¾ in overlap regions other than those with low recombination rates (Campos et al. 2013), we did not use a correction factor for πF/πS for the overlap regions studied here, because this difference in Ne exactly cancels the difference in strength of selection on the hypothesis of semidominance. But after applying the 4/3 correction, πF for the overlap regions was either higher for X than for A or similar to A (supplementary fig. 3, Supplementary Material online). Again, there was no evidence for significant departures from semidominance for slightly deleterious mutations. Fig. 4. View largeDownload slide Plots of ωna against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωna values. Because of the low number of female-biased genes, only three bins were used for long introns. Fig. 4. View largeDownload slide Plots of ωna against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωna values. Because of the low number of female-biased genes, only three bins were used for long introns. Discussion Under suitable conditions on the levels of dominance of the fitness effects of beneficial mutations, there should be a faster rate of fixation of new mutations on the X chromosome than the autosomes—the faster-X effect (Rice 1984; Charlesworth et al. 1987; Meisel and Connallon 2013). Genes expressed only in males are the most likely to show such an effect, whereas it should not be seen for genes expressed only in females, unless the effective population size for the X is higher than that for the autosomes. Consistent with these theoretical predictions, the results presented above show that substitutions in most types of functionally significant sequences in male-biased and unbiased genes (nonsynonymous sites, UTRs, and long introns) tend to exhibit faster rates of substitution of beneficial mutations for X genes compared with A genes (figs. 1 and 2). However, our analyses also showed that genomic regions with very low rates of recombination do not show a faster-X effect (fig. 2 and supplementary table 4, Supplementary Material online), consistent with previous observations that showed a lack of adaptive evolution at nonsynonymous sites in very low recombination regions of the D. melanogaster genome (Campos et al. 2014). This observation was interpreted in terms of strong Hill–Robertson interference in such regions (Hill and Robertson 1966; Felsenstein 1974), whereby both positive and negative selection at closely linked sites reduce the rate of fixation of beneficial mutations (Campos et al. 2014; Charlesworth and Campos 2014; Castellano et al. 2016). The influence of recombination rate on adaptive evolution caused by Hill–Robertson interference is also probably responsible for the fact that the faster-X effect seen for nonsynonymous and long intron substitutions in female-biased genes disappears after correcting for their recombination rates—for these categories, only genes with a higher effective recombination rate on the X chromosome than the autosomes show a faster-X effect. Thus, the faster-X effect that we observed previously for nonsynonymous mutations in female-biased genes (Ávila et al. 2015) is probably due to an overall higher recombination rate for these genes, which reduces their exposure to Hill–Robertson interference. As has been found previously for many different systems (Baines et al. 2008; Parsch and Ellegren 2013; Ávila et al. 2014; Kousathanas et al. 2014; Ávila et al. 2015), male-biased genes exhibited higher rates of adaptive evolution at nonsynonymous sites compared with unbiased and female-biased genes, for both X and A genes. Such an effect was not observed, however, for UTRs and long introns (fig. 1 and supplementary table 1, Supplementary Material online). It is unclear why such a difference between types of sequence should occur, and it casts doubt on the frequently expressed idea that sexual selection causes faster rates of evolution of male-biased genes (Parsch and Ellegren 2013), since this is just as likely to affect regulatory mutations as amino-acid mutations. We also found evidence for weaker selective constraints on the protein sequences of genes with sex-biased expression, but not on UTRs or long introns. The extent of between-species divergence in the level of gene expression has been also found to be greater for X than A in Drosophila (Kayserili et al. 2012; Llopart 2012; Meisel et al. 2012), suggesting faster-X adaptive evolution of cis-regulatory sequences (Kayserili et al. 2012; Meisel et al. 2012; Coolon et al. 2015). This hypothesis is consistent with the observation of faster-X divergence of noncoding sequences in Drosophila (Hu et al. 2013); that study did not, however, analyze the contribution of positively selected substitutions to differences between X and autosomes. Our results provide quantitative estimates of the extent of X–A differences in the rates of fixation due to positive selection (figs. 1 and 2 and supplementary table 1, Supplementary Material online), and demonstrate an adaptive faster-X effect for UTRs and long introns, consistent with the results of Begun et al. (2007) and Mackay et al. (2012). The faster-X effect for noncoding regions may well be related to the faster X-effect for gene expression. However, the faster-X effect that was found for the UTRs of female-biased genes (figs. 1 and 2 and supplementary table 1, Supplementary Material online) is not predicted by the standard theoretical models, and is not associated with a difference in recombination rate; furthermore, no faster-X effect for gene expression in genes with expression biased towards female-limited reproductive tissues was found by Meisel et al. (2012). Differences between X and A in traits other than male hemizygosity and recombination rate must be causing this pattern, possibly reflecting the specific biological functions of the small group of genes involved. A much-debated question has been the relative contributions of protein-coding sequences versus noncoding sequences to adaptive evolution (e.g., Carroll 2000; Hoekstra and Coyne 2007; Stern and Orgogozo 2008). Our analyses showed that mutations in functional noncoding regions showed considerably higher rates of adaptive evolution than nonsynonymus mutations. These results are consistent with previous work on Drosophila using both McDonald–Kreitman type approaches (Andolfatto 2005; Begun et al. 2007; Mackay et al. 2012; Campos et al. 2017) and selective sweep signals (Elyashiv et al. 2016). One explanation is that noncoding regions are less constrained by purifying selection than nonsynonymous sites (ωna is higher for noncoding regions than for nonsynonymous sites [figs. 3 and 4 and supplementary table 5, Supplementary Material online]), possibly because they are subject to fewer pleiotropic restrictions (Orr 1998; Carroll 2000; Orr 2000), making it easier for a new mutation to confer a selective advantage. These results are similar to the recent finding of higher rates of positive selection observed in genes that are less constrained by purifying selection (Campos et al. 2017). Higher rates of adaptive sequence evolution in noncoding regions compared with protein sequences have been also found in other systems, such as mice (Halligan et al. 2013) and the flowering plant Capsella (Williamson et al. 2014). We found evidence for differences in the strength of purifying selection between X and A, at least for unbiased genes, when examining the overall patterns of diversity for the two chromosomes (tables 1 and 2). Such a pattern is expected for sufficiently strongly selected deleterious mutations (Nes ≫ 1) with dominance coefficients h ≤ ½ (see the theoretical section of the Supplementary Material online). There is much evidence to suggest that h for deleterious mutations is mostly nonzero but < ½, that is, mutations are partially recessive in their effects on fitness (Dobzhansky and Wright 1941; Muller 1950; Crow and Simmons 1983; García-Dorado and Caballero 2000; Manna et al. 2011). However, our analyses of diversities at functional sites, πF, (table 1 and supplementary fig. 3, Supplementary Material online) and their ratios to synonymous site diversities, πF/πS, (table 2 and supplementary fig. 2, Supplementary Material online) showed no evidence that h was substantially less than ½. The same conclusion was suggested by comparisons of ωna between X and A (figs. 3 and 4 and supplementary table 5, Supplementary Material online). It is important to note, however, that the estimates of statistics such as ωna and α (the proportion of between-species substitutions in a functional category that were fixed by positive selection) could be influenced by the dominance coefficients of deleterious mutations. The DFE-alpha method for estimating these quantities (Eyre-Walker and Keightley 2009) predicts the contribution of slightly deleterious mutations to variation and evolution using a population genetic model of autosomal loci subject to deleterious mutations with h = ½. This raises the question of the reliability of parameter estimates from DFE-alpha when h is < ½, for which the selection equations differ substantially from those with semidominance, except for mutations with no fitness effects on males (see the theoretical section of the Supplementary Material online). Using an approach similar to DFE-alpha, Veeramah et al. (2014) showed that, except for very low h values (h < 0.2), the dominance coefficient has only a small effect on the relative values of α estimates for X and A for human data (see fig. 2 of Veeramah et al. 2014). Their figure 2 suggests that the assumption of h = ½, as was made here, should not lead to serious problems inferences concerning the faster-X effect for adaptive mutations. This conclusion is strengthened by the analyses of purifying selection just described, which suggested that h is close to ½ for deleterious mutations (see tables 1 and 2). Similarly, recessive effects of deleterious variants cannot explain the observed pattern of higher codon usage on the X (see Campos et al. 2013). The comparisons between X and A of the corrected values of πF and πF/πS(tables 1 and 2) yielded many cases in which the X–A difference was significantly larger than that predicted on the hypothesis of strong selection with h = ½. This anomaly is larger for introns and UTRs. If there is a sizeable contribution from weakly selected mutations with Nes ≤ 1, the expected X/A ratio of πF/πSwill be greater than 3/4; in the limit of all mutations being neutral, it will approach 1 when NeX = NeA, as is the case for the two chromosomes as a whole. Multiplication by 4/3 would then be an overcorrection. Consistent with this hypothesis, inspection of the DFEs of the different sequence classes suggests that a considerably greater fraction of noncoding sequence mutations are weakly selected (15–26% mutations have Nes in the range 0–1), than nonsynonymous sites (< 7% with Nes 0–1)—see supplementary table 6, Supplementary Material online. Further theoretical predictions that take stochastic effects and the DFE parameters into account will be needed to test this hypothesis in detail. Materials and Methods We used the same data set as in Ávila et al. (2015), consisting of polymorphism data from 17 haploid genomes from the Gikongoro (Rwanda) population of D. melanogaster provided by the Drosophila Population Genomics Project 2 (Pool et al. 2012). As outgroup material, we used sequences from D. yakuba and D. simulans obtained from the D. melanogaster–D. simulans–D. yakuba gene alignments of Hu et al. (2013). The coding sequence data were filtered and analyzed as described in Campos et al. (2014). In brief, the 17 D. melanogaster genomes were obtained from the DPGP2 data set (http://www.dpgp.org/dpgp2/DPGP2.html). We mapped and aligned the reads to the reference sequence r5.33, available on Flybase [http://flybase.org/] and last accessed December 25, 2012, using BWA (Li and Durbin 2009). We used the Genome Analysis Toolkit (GATK, last accessed December 20, 2012), to do multisample SNP calling (DePristo et al. 2011). From the multisample VCF file, we made a consensus sequence FASTA file for each individual using a custom Perl script. Using the coding sequence coordinates, we extracted their sequences and made FASTA alignments including the D. yakuba and D. simulans outgroups. We excluded 225 genes located in the autosomal heterochromatic regions and on chromosome four, where crossing over is absent (Campos et al. 2012, 2014). We calculated pairwise nucleotide diversity (π, Tajima 1983) at putatively functional sites (πF) and neutral synonymous site diversity (πS). πF/πSwas calculated as the ratio of the mean of πF to the mean of πS for the set of genes in question. As functional sites we considered nonsynonymous sites, 5′ and 3′ UTRs, and introns longer than 100 bp—shorter introns are under less constraint (Halligan and Keightley 2006). For the analyses of UTRs and introns, we masked any coding sequences and exons, respectively, and excluded any sequence with no outgroup information. To assign sites as synonymous and nonsynonymous and to estimate the nonsynonymous divergence and synonymous divergences, KA and KS, we used the method of Comeron (1995)—see Campos et al. 2014 for further details. To obtain the noncoding divergence data, we applied a Kimura two-parameter correction (Kimura 1980). We used the outgroup D. yakuba for coding sites and UTRs; and D. simulans for long introns. Female crossing over rates for these genes, in terms of centiMorgans per megabase, were obtained from Comeron et al. (2012). These rates were multiplied by two-thirds for X-linked genes and by one-half for autosomal genes, respectively. This procedure provides estimates of “effective recombination rates,” taking into account the absence of recombination between homologous chromosomes in males, and the fact that X-linked genes and autosomal genes spend two-thirds and one-half of their time in females, respectively. This allows comparisons between X-linked and autosomal genes that experience the same rates of recombination as far as evolutionary processes are concerned (Campos et al. 2014). In order to classify genes as unbiased, male- or female-biased, we used the tissue-specific expression values (in RPKM) of Brown et al. (2014). To obtain female expression values, we used female ovaries; for male expression values we used the average of gene expression in testes and accessory glands of male flies. In a first step, we obtained a sex bias ratio (SBR) value as: log 2 (male value/female value). We classified genes with a SBR > 1 as male-biased; genes with SBR < –1 as female-biased, and genes with SBR between –1 and 1 as unbiased. This is equivalent to using a 2-fold expression difference between the sexes as the criterion for sex bias. In a second step, male and female expression values were ranked, and the 15th percentiles of their distributions were selected. If a gene was categorized as male-biased in the first step, and showed expression in females (ovary) of less than the female 15th percentile value, the gene was ultimately labeled as male-biased, and vice-versa for female bias. The aim of the second step was to ensure that biased genes have a very low level of expression in the opposite sex. Genes were then sorted into six categories, based on chromosomal location (autosomal vs. X-linked) and the sex-biased expression categories described above. Our final data set consisted of 6,603 autosomal genes and 1,250 genes. A potential reason for a faster-X effect for female-biased genes (which is contrary to theoretical predictions) is that such genes may exhibit some expression in males in somatic tissues, allowing positive selection to act when the gene is in a hemizygous state, even though the gene has been classed as female-biased according to the gonad-based sex bias ratio. To deal with this problem, we made a second data set using gene expression in both heads and germline to define sex bias. To be classed as male-biased in expression, a gene must then meet two conditions. First, the gene must be expressed in the male gonad and head at a value greater than the male 15th percentile values for both the relevant tissues. Second, expression in the female ovary and head must be lower than the 15th percentile expression values for both tissues, respectively; complementary criteria were used to classify a gene as female-biased. The breadth of expression of a gene across different tissues is known to affect its rates of protein sequence evolution (Larracuente et al. 2008), and thus needs to be taken into account when comparing X and A. We measured this by τ, the proportion of tissues in which a gene is found to be expressed. τ ranges from 0 (for broadly expressed genes) to 1 (for narrowly expressed genes) (Yanai et al. 2005). Similarly, because levels of gene expression affect protein evolution (Larracuente et al. 2008), we calculated level of gene expression as the average expression across tissues for each gene. We used DFE-alpha (Eyre-Walker and Keightley 2009) to estimate ωa, the ratio of the rate of adaptive substitutions per basepair relative to neutral substitutions for nonsynonymous sites, UTR sites and sites in long introns. This method attempts to correct for the contribution of slightly deleterious mutations to polymorphism and divergence and the impact of recent demographic change. We used D. yakuba (KS ≈ 0.25) as an outgroup for CDS and UTRs, and D. simulans (KS ≈ 0.12) for long introns. We used a demographic model where the population at initial size N1 (set to 100 in the DFE-alpha program) experienced a step change to N2 at n generations in the past. A preliminary analysis for the whole data set showed that a constant population size model gave a poor fit to the data, consistent with the results of Pool et al. (2012) that show that D. melanogaster African populations have apparently experienced a population expansion. To calculate confidence intervals we generated replicate bootstrap estimates for each group separately by resampling genes 1,000 times within a given group, and running DFE-alpha for each bootstrap. In order to test for a significant difference in a given statistic between a pair of categories, such as X and A, we calculated the proportion of its bootstrap values that were larger for one category than the other. Acknowledgments This work was supported by Grant RPG-2015-2033 from the Leverhulme Trust (to B.C.). The authors are grateful to the Drosophila Population Genomics Project (DPGP), for making these data available. They thank editors Krushnamegh Kunte and Deepa Agashe for inviting us to contribute to the special issue on Genetics of Adaptation. They also thank three reviewers for constructive comments and suggestions. References Andolfatto P. 2005. Adaptive evolution of non-coding DNA in Drosophila. Nature  437( 7062): 1149– 1152. http://dx.doi.org/10.1038/nature04107 Google Scholar CrossRef Search ADS PubMed  Ávila V, Campos JL, Charlesworth B. 2015. The effects of sex-biased gene expression and X-linkage on rates of adaptive protein sequence evolution in Drosophila. Biol Lett . 11( 4): 20150117. Google Scholar CrossRef Search ADS PubMed  Ávila V, Marion de Procé S, Campos JL, Borthwick H, Charlesworth B, Betancourt AJ. 2014. Faster-X effects in two Drosophila lineages. Genome Biol Evol . 6( 10): 2968– 2982. Google Scholar CrossRef Search ADS PubMed  Baines JF, Sawyer SA, Hartl DL, Parsch J. 2008. Effects of X-linkage and sex-biased gene expression on the rate of adaptive protein evolution in Drosophila. Mol Biol Evol . 25( 8): 1639– 1650. http://dx.doi.org/10.1093/molbev/msn111 Google Scholar CrossRef Search ADS PubMed  Begun DJ, Holloway AK, Stevens K, Hillier LW, Poh Y-P, Hahn MW, Nista PM, Jones CD, Kern AD, Dewey CN, et al.   2007. Population genomics: whole-genome analysis of polymorphism and divergence in Drosophila simulans. PLoS Biol . 5( 11): e310. Google Scholar CrossRef Search ADS PubMed  Brown JB, Boley N, Eisman R, May GE, Stoiber MH, Duff MO, Booth BW, Wen J, Park S, Suzuki AM, et al.   2014. Diversity and dynamics of the Drosophila transcriptome. Nature  512( 7515): 393– 399. Google Scholar CrossRef Search ADS PubMed  Campos JL, Charlesworth B, Haddrill PR. 2012. Molecular evolution in nonrecombining regions of the Drosophila melanogaster genome. Genome Biol Evol . 4( 3): 278– 288. http://dx.doi.org/10.1093/gbe/evs010 Google Scholar CrossRef Search ADS PubMed  Campos JL, Zeng K, Parker DJ, Charlesworth B, Haddrill PR. 2013. Codon usage bias and effective population sizes on the X chromosome versus the autosomes in Drosophila melanogaster. Mol Biol Evol . 30( 4): 811– 823. Google Scholar CrossRef Search ADS PubMed  Campos JL, Halligan DL, Haddrill PR, Charlesworth B. 2014. The relation between recombination rate and patterns of molecular evolution and variation in Drosophila melanogaster. Mol Biol Evol . 31( 4): 1010– 1028. Google Scholar CrossRef Search ADS PubMed  Campos JL, Zhao L, Charlesworth B. 2017. Estimating the parameters of background selection and selective sweeps in Drosophila in the presence of gene conversion. Proc Natl Acad Sci U S A . 114( 24): E4762– E47771. Google Scholar CrossRef Search ADS PubMed  Carneiro M, Albert FW, Melo-Ferreira J, Galtier N, Gayral P, Blanco-Aguiar JA, Villafuerte R, Nachman MW, Ferrand N. 2012. Evidence for widespread positive and purifying selection across the European rabbit (Oryctolagus cuniculus) genome. Mol Biol Evol . 29( 7): 1837– 1849. Google Scholar CrossRef Search ADS PubMed  Carroll SB. 2000. Endless forms: the evolution of gene regulation and morphological diversity. Cell  101( 6): 577– 580. http://dx.doi.org/10.1016/S0092-8674(00)80868-5 Google Scholar CrossRef Search ADS PubMed  Castellano D, Coronado-Zamora M, Campos JL, Barbadilla A, Eyre-Walker A. 2016. Adaptive evolution is substantially impeded by Hill–Robertson interference in Drosophila. Mol Biol Evol . 33( 2): 442– 455. Google Scholar CrossRef Search ADS PubMed  Charlesworth B, Campos JL. 2014. The relations between recombination rate and patterns of molecular variation and evolution in Drosophila. Annu Rev Genet . 48: 383– 403. http://dx.doi.org/10.1146/annurev-genet-120213-092525 Google Scholar CrossRef Search ADS PubMed  Charlesworth B, Coyne J, Barton N. 1987. The relative rates of evolution of sex chromosomes and autosomes. Am Nat . 130( 1): 113– 146. Google Scholar CrossRef Search ADS   Comeron JM. 1995. A method for estimating the numbers of synonymous and nonsynonymous substitutions per site. J Mol Evol . 41( 6): 1152– 1159. Google Scholar CrossRef Search ADS PubMed  Comeron JM, Ratnappan R, Bailin S. 2012. The many landscapes of recombination in Drosophila melanogaster. PLoS Genet . 8( 10): e1002905. Google Scholar CrossRef Search ADS PubMed  Coolon JD, Stevenson KR, McManus CJ, Yang B, Graveley BR, Wittkopp PJ. 2015. Molecular mechanisms and evolutionary processes contributing to accelerated divergence of gene expression on the Drosophila X chromosome. Mol Biol Evol . 32( 10): 2605– 2615. Google Scholar CrossRef Search ADS PubMed  Crow J, Simmons M. 1983. The mutation load in Drosophila. In: Ashburner M, Carson HL, Thompson JN, editors. The genetics and biology of drosophila , Vol. 3c, pp. 1– 35. London: Academic Press. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C, Philippakis AA, del Angel G, Rivas MA, Hanna M, et al.   2011. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet . 43( 5): 491– 498. http://dx.doi.org/10.1038/ng.806 Google Scholar CrossRef Search ADS PubMed  Dobzhansky T, Wright S. 1941. Genetics of natural populations. V. Relations between mutation rate and accumulation of lethals in populations of Drosophila pseudoobscura. Genetics  26( 1): 23– 51. Google Scholar PubMed  Elyashiv E, Sattath S, Hu TT, Strutsovsky A, McVicker G, Andolfatto P, Coop G, Sella G. 2016. A genomic map of the effects of linked selection in Drosophila. PLoS Genet . 12( 8): e1006130. Google Scholar CrossRef Search ADS PubMed  Eyre-Walker A, Keightley PD. 2009. Estimating the rate of adaptive molecular evolution in the presence of slightly deleterious mutations and population size change. Mol Biol Evol . 26( 9): 2097– 2108. http://dx.doi.org/10.1093/molbev/msp119 Google Scholar CrossRef Search ADS PubMed  Felsenstein J. 1974. The evolutionary advantage of recombination. Genetics  78( 2): 737– 756. Google Scholar PubMed  García-Dorado A, Caballero A. 2000. On the average coefficient of dominance of deleterious spontaneous mutations. Genetics  155( 4): 1991– 2001. Google Scholar PubMed  Haddrill PR, Thornton KR, Charlesworth B, Andolfatto P. 2005. Multilocus patterns of nucleotide variability and the demographic and selection history of Drosophila melanogaster populations. Genome Res . 15( 6): 790– 799. Google Scholar CrossRef Search ADS PubMed  Haldane JBS. 1924. A mathematical theory of natural and artificial selection. Trans Camb Philos Soc . 23: 19– 41. Halligan DL, Keightley PD. 2006. Ubiquitous selective constraints in the Drosophila genome revealed by a genome-wide interspecies comparison. Genome Res . 16( 7): 875– 884. http://dx.doi.org/10.1101/gr.5022906 Google Scholar CrossRef Search ADS PubMed  Halligan DL, Kousathanas A, Ness RW, Harr B, Eöry L, Keane TM, Adams DJ, Keightley PD. 2013. Contributions of protein-coding and regulatory change to adaptive molecular evolution in murid rodents. PLoS Genet . 9( 12): e1003995. Google Scholar CrossRef Search ADS PubMed  Hill WG, Robertson A. 1966. The effect of linkage on limits to artificial selection. Genet Res . 8( 3): 269– 294. http://dx.doi.org/10.1017/S0016672300010156 Google Scholar CrossRef Search ADS PubMed  Hoekstra HE, Coyne JA. 2007. The locus of evolution: evo-devo and the genetics of adaptation. Evolution  61( 5): 995– 1016. http://dx.doi.org/10.1111/j.1558-5646.2007.00105.x Google Scholar CrossRef Search ADS PubMed  Hu TT, Eisen MB, Thornton KR, Andolfatto P. 2013. A second-generation assembly of the Drosophila simulans genome provides new insights into patterns of lineage-specific divergence. Genome Res . 23( 1): 89– 98. http://dx.doi.org/10.1101/gr.141689.112 Google Scholar CrossRef Search ADS PubMed  Hvilsom C, Qian Y, Bataillon T, Li Y, Mailund T, Sallé B, Carlsen F, Li R, Zheng H, Jiang T, et al.   2012. Extensive X-linked adaptive evolution in central chimpanzees. Proc Natl Acad Sci U S A.  109( 6): 2054– 2059. Google Scholar CrossRef Search ADS PubMed  Kayserili MA, Gerrard DT, Tomancak P, Kalinka AT. 2012. An excess of gene expression divergence on the X chromosome in Drosophila embryos: implications for the faster-X hypothesis. PLoS Genet . 8( 12): e1003200. Google Scholar CrossRef Search ADS PubMed  Kimura M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J Mol Evol . 16( 2): 111– 120. http://dx.doi.org/10.1007/BF01731581 Google Scholar CrossRef Search ADS PubMed  Kousathanas A, Halligan DL, Keightley PD. 2014. Faster-X adaptive protein evolution in house mice. Genetics  196( 4): 1131– 1143. http://dx.doi.org/10.1534/genetics.113.158246 Google Scholar CrossRef Search ADS PubMed  Langley CH, Stevens K, Cardeno C, Lee YCG, Schrider DR, Pool JE, Langley SA, Suarez C, Corbett-Detig RB, Kolaczkowski B, et al.   2012. Genomic variation in natural populations of Drosophila melanogaster. Genetics  192( 2): 533– 598. Google Scholar CrossRef Search ADS PubMed  Larracuente AM, Sackton TB, Greenberg AJ, Wong A, Singh ND, Sturgill D, Zhang Y, Oliver B, Clark AG. 2008. Evolution of protein-coding genes in Drosophila. Trends Genet . 24( 3): 114– 123. Google Scholar CrossRef Search ADS PubMed  Li H, Durbin R. 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics  25( 14): 1754– 1760. http://dx.doi.org/10.1093/bioinformatics/btp324 Google Scholar CrossRef Search ADS PubMed  Llopart A. 2012. The rapid evolution of X-linked male-biased gene expression and the large-X effect in Drosophila yakuba, D. santomea, and their hybrids. Mol Biol Evol . 29( 12): 3873– 3886. http://dx.doi.org/10.1093/molbev/mss190 Google Scholar CrossRef Search ADS PubMed  Mackay TFC, Richards S, Stone EA, Barbadilla A, Ayroles JF, Zhu D, Casillas S, Han Y, Magwire MM, Cridland JM, et al.   2012. The Drosophila melanogaster genetic reference panel. Nature  482( 7384): 173– 178. Google Scholar CrossRef Search ADS PubMed  Mank JE, Vicoso B, Berlin S, Charlesworth B. 2010. Effective population size and the Faster-X effect: empirical results and their interpretation. Evolution  64( 3): 663– 674. http://dx.doi.org/10.1111/j.1558-5646.2009.00853.x Google Scholar CrossRef Search ADS PubMed  Manna F, Martin G, Lenormand T. 2011. Fitness landscapes: an alternative theory for the dominance of mutation. Genetics  189( 3): 923– 937. http://dx.doi.org/10.1534/genetics.111.132944 Google Scholar CrossRef Search ADS PubMed  Meisel RP, Connallon T. 2013. The faster-X effect: integrating theory and data. Trends Genet . 29( 9): 537– 544. http://dx.doi.org/10.1016/j.tig.2013.05.009 Google Scholar CrossRef Search ADS PubMed  Meisel RP, Malone JH, Clark AG. 2012. Disentangling the relationship between sex-biased gene expression and X-linkage. Genome Res . 22( 7): 1255– 1265. http://dx.doi.org/10.1101/gr.132100.111 Google Scholar CrossRef Search ADS PubMed  Muller HJ. 1950. Our load of mutations. Am J Hum Genet . 2( 2): 111– 176. Google Scholar PubMed  Orr HA. 1998. The population genetics of adaptation: The distribution of factors fixed during adaptive evolution. Evolution  52( 4): 935– 949. http://dx.doi.org/10.1111/j.1558-5646.1998.tb01823.x Google Scholar CrossRef Search ADS PubMed  Orr HA. 2000. Adaptation and the cost of complexity. Evolution  54( 1): 13– 20. http://dx.doi.org/10.1111/j.0014-3820.2000.tb00002.x Google Scholar CrossRef Search ADS PubMed  Orr HA, Betancourt A. 2001. Haldane’s sieve and adaptation from the standing genetic variation. Genetics  157( 2): 875– 884. Google Scholar PubMed  Orr HA, Coyne JA. 1989. The genetics of postzygotic isolation in the Drosophila virilis group. Genetics  121( 3): 527– 537. Google Scholar PubMed  Parsch J, Ellegren H. 2013. The evolutionary causes and consequences of sex-biased gene expression. Nat Rev Genet . 14( 2): 83– 87. http://dx.doi.org/10.1038/nrg3376 Google Scholar CrossRef Search ADS PubMed  Parsch J, Novozhilov S, Saminadin-Peter SS, Wong KM, Andolfatto P. 2010. On the utility of short intron sequences as a reference for the detection of positive and negative selection in Drosophila. Mol Biol Evol . 27( 6): 1226– 1234. Google Scholar CrossRef Search ADS PubMed  Pool JE, Corbett-Detig RB, Sugino RP, Stevens KA, Cardeno CM, Crepeau MW, Duchen P, Emerson JJ, Saelao P, Begun DJ, et al.   2012. Population genomics of sub-Saharan Drosophila melanogaster: African diversity and non-African admixture. PLoS Genet . 8( 12): e1003080. Google Scholar CrossRef Search ADS PubMed  Presgraves DC. 2008. Sex chromosomes and speciation in Drosophila. Trends Genet . 24( 7): 336– 343. http://dx.doi.org/10.1016/j.tig.2008.04.007 Google Scholar CrossRef Search ADS PubMed  Rice WR. 1984. Sex chromosomes and the evolution of sexual dimorphism. Evolution  38( 4): 735– 742. http://dx.doi.org/10.1111/j.1558-5646.1984.tb00346.x Google Scholar CrossRef Search ADS PubMed  Rousselle M, Faivre N, Ballenghien M, Galtier N, Nabholz B. 2016. Hemizygosity enhances purifying selection: lack of Fast-Z evolution in two Satyrine butterflies. Genome Biol Evol . 8( 10): 3108– 3119. Google Scholar CrossRef Search ADS PubMed  Stern DL, Orgogozo V. 2008. The loci of evolution: how predictable is genetic evolution?. Evolution  62( 9): 2155– 2177. http://dx.doi.org/10.1111/j.1558-5646.2008.00450.x Google Scholar CrossRef Search ADS PubMed  Veeramah KR, Gutenkunst RN, Woerner AE, Watkins JC, Hammer MF. 2014. Evidence for increased levels of positive and negative selection on the X chromosome versus autosomes in humans. Mol Biol Evol . 31( 9): 2267– 2282. Google Scholar CrossRef Search ADS PubMed  Vicoso B, Charlesworth B. 2009. Effective population size and the faster-X effect: an extended model. Evolution  63( 9): 2413– 2426. http://dx.doi.org/10.1111/j.1558-5646.2009.00719.x Google Scholar CrossRef Search ADS PubMed  Williamson RJ, Josephs EB, Platts AE, Hazzouri KM, Haudry A, Blanchette M, Wright SI. 2014. Evidence for widespread positive and negative selection in coding and conserved noncoding regions of Capsella grandiflora. PLoS Genet . 10( 9): e1004622. Google Scholar CrossRef Search ADS PubMed  Yanai I, Benjamin H, Shmoish M, Chalifa-Caspi V, Shklar M, Ophir R, Bar-Even A, Horn-Saban S, Safran M, Domany E, et al.   2005. Genome-wide midrange transcription profiles reveal expression level relationships in human tissue specification. Bioinformatics  21( 5): 650– 659. Google Scholar CrossRef Search ADS PubMed  © The Author 2017. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Molecular Biology and Evolution Oxford University Press

The Effects of Sex-Biased Gene Expression and X-Linkage on Rates of Sequence Evolution in Drosophila

Loading next page...
 
/lp/ou_press/the-effects-of-sex-biased-gene-expression-and-x-linkage-on-rates-of-Bvte0vT38P
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com
ISSN
0737-4038
eISSN
1537-1719
D.O.I.
10.1093/molbev/msx317
Publisher site
See Article on Publisher Site

Abstract

Abstract A faster rate of adaptive evolution of X-linked genes compared with autosomal genes (the faster-X effect) can be caused by the fixation of recessive or partially recessive advantageous mutations. This effect should be largest for advantageous mutations that affect only male fitness, and least for mutations that affect only female fitness. We tested these predictions in Drosophila melanogaster by using coding and functionally significant noncoding sequences of genes with different levels of sex-biased expression. Consistent with theory, nonsynonymous substitutions in most male-biased and unbiased genes show faster adaptive evolution on the X. However, genes with very low recombination rates do not show such an effect, possibly as a consequence of Hill–Robertson interference. Contrary to expectation, there was a substantial faster-X effect for female-biased genes. After correcting for recombination rate differences, however, female-biased genes did not show a faster X-effect. Similar analyses of noncoding UTRs and long introns showed a faster-X effect for all groups of genes, other than introns of female-biased genes. Given the strong evidence that deleterious mutations are mostly recessive or partially recessive, we would expect a slower rate of evolution of X-linked genes for slightly deleterious mutations that become fixed by genetic drift. Surprisingly, we found little evidence for this after correcting for recombination rate, implying that weakly deleterious mutations are mostly close to being semidominant. This is consistent with evidence from polymorphism data, which we use to test how models of selection that assume semidominance with no sex-specific fitness effects may bias estimates of purifying selection. Drosophila, faster-X effect, gene expression, natural selection, sex bias Introduction The differences between the modes of inheritance of genes on the X chromosome and the autosomes are expected to affect their patterns of variation and evolution (reviewed by Meisel and Connallon 2013). Under suitable conditions with respect to the levels of dominance of new mutations, sex differences in selection coefficients, recombination rates, and effective population size, theory shows that X-linked genes can evolve under positive selection more rapidly than autosomal genes—the so-called “faster-X effect.” This reflects the fact that the hemizygosity of the X chromosome in males (in species with male heterogamety) means that positive selection can sometimes be more effective on new X-linked mutations than on autosomal mutations with equivalent effects on fitness (Haldane 1924; Charlesworth et al. 1987; Meisel and Connallon 2013). This is, however, only likely to happen when evolutionary change depends on new mutations; fixation of variants present as standing variation tends to favor a slower rate of evolution of X-linked genes (Charlesworth et al. 1987; Orr and Betancourt 2001; Meisel and Connallon 2013). An important consequence of the faster-X effect is that X (or Z) chromosomes are likely to play a disproportionate role in speciation and evolutionary divergence (Orr and Coyne 1989; Presgraves 2008). Faster-X evolution from new mutations is facilitated by partial recessivity of beneficial mutations (dominance coefficients less than one-half) (Charlesworth et al. 1987); and by a higher ratio of the effective population size of X-linked to autosomal genes (NeX/NeA) than the value of three-quarters expected under the standard Wright–Fisher model, which assumes equal variances in reproductive success in males and females (Vicoso and Charlesworth 2009; Meisel and Connallon 2013). Genes that only affect male fitness are more likely to exhibit a faster-X effect than genes that affect the fitnesses for both sexes. In contrast, genes that affect only female fitness will not experience a faster-X effect unless NeX/NeA> ¾ (Charlesworth et al. 1987; Vicoso and Charlesworth 2009; Mank et al. 2010; Meisel and Connallon 2013). With sexually antagonistic effects of mutations on fitness, a faster-X effect can arise with recessive or partially recessive mutations that increase male fitness at the expense of female fitness, or dominant or partially dominant mutations that increase female fitness at the expense of male fitness (Rice 1984; Charlesworth et al. 1987). These theoretical expectations are supported by recent studies of D. melanogaster and its close relatives, using whole genome sequence comparisons between species and resequencing of multiple individuals of D. melanogaster. These studies showed that there is a faster rate of protein sequence evolution on the X chromosome (X) than on the autosomes (A), and that this is attributable to a greater incidence of fixations of selectively favorable nonsynonymous mutations on X compared with A (Langley et al. 2012; Mackay et al. 2012; Hu et al. 2013; Campos et al. 2014; Ávila et al. 2015). In a previous study (Ávila et al. 2015), we tested the above predictions of the theory concerning sex-specific fitness effects by using genes with different levels of sex-biased expression (unbiased, male- and female-biased) in Drosophila melanogaster, and estimating the extent of adaptive evolution of non-synonymous mutations. We detected a faster-X effect for all categories of gene expression, including female-biased genes. The sex bias in gene expression for each gene was used as a proxy for male/female fitness, assuming that unbiased genes are likely to have similar fitness effects on both sexes (or to be sexually antagonistic), male-biased genes to affect male fitness more than female fitness, and vice-versa for female-biased genes. A faster-X effect was found for female-biased genes, even after adjusting for the effects of differences in recombination rate between X and A that reflects differences in Ne caused by selection at linked sites; such effects were detected by Campos et al. (2013) (see their fig. 2). However, Ávila et al. (2015) used a meta-analysis of Drosophila gene expression, most of which included whole-fly expression data. A FDR cut-off of 20% was used to classify genes as sex-biased. With this liberal criterion, “female-biased” X-linked genes could still have significant expression levels in males, and hence be exposed to positive selection on the hemizygous X. Here, we use a much more stringent criterion for sex-biased gene expression, to ensure that female-biased genes can have little effect on male fitness. Faster-X divergence for gene expression has been reported in Drosophila (Kayserili et al. 2012; Meisel et al. 2012), possibly driven by adaptive and recessive cis-regulatory mutations that affect gene expression levels (Coolon et al. 2015). A theoretical framework for how adaptive cis-regulatory mutations could be recessive was proposed by Meisel et al. (2012). Previous work has shown that many noncoding regions are functional and experience both negative and positive selection (Andolfatto 2005; Haddrill et al. 2005; Halligan and Keightley 2006; Halligan et al. 2013). In addition, a faster-X effect for noncoding regions has been described in Drosophila (Begun et al. 2007; Mackay et al. 2012; Hu et al. 2013), which could be related to the faster X-effect on gene expression. We have therefore extended the Ávila et al. (2015) analysis to noncoding regions, including 5′ UTRs, 3′ UTRs and long introns (introns shorter than 100 bp appear to experience little selective constraint [Halligan and Keightley 2006; Parsch et al. 2010]). We also tested for the existence of a slower-X effect for deleterious mutations. Much evidence suggests that mutations are partially recessive in their effects on fitness (Dobzhansky and Wright 1941; Muller 1950; Crow and Simmons 1983; García-Dorado and Caballero 2000; Manna et al. 2011). Purifying selection is thus usually expected to be stronger on the X than A, and a lower rate of fixation of slightly deleterious mutations is expected for X than for A, unless the X/A ratio of Ne is < ¾ (Mank et al. 2010; Vicoso and Charlesworth 2009). However, there is only weak empirical evidence for a slower-X (or Z) effect (Carneiro et al. 2012; Hvilsom et al. 2012; Veeramah et al. 2014; Rousselle et al. 2016). We tested for this effect by estimating the proportion of slightly deleterious mutations fixed on X and A in a comparison of a pair of closely related species. Overall, we did not find any substantial differences between X and A, suggesting that slightly deleterious mutations may be close to being semidominant. This was consistent with the results of comparisons between X and A of nucleotide site diversities at functional sites. Results Rates of Adaptive Evolution on X versus A A total of 7,853 D. melanogaster genes were available for this study (see supplementary table 1, Supplementary Material online). As described in the Materials and Methods, we used a data set on germline gene expression to classify genes as male-biased, female-biased and unbiased, with a much more stringent criterion for sex-biased expression than that of Ávila et al. (2015). For this data set, the largest group was that for unbiased A genes (N = 4,529), and the smallest group was that for female-biased X genes (N = 64). The distribution of the three categories of expression between A and X was significantly different (χ2 = 23.37, P < 0.01), with a slight excess on the X of female-biased (20.9%) and unbiased (16.8%) genes, and a deficit of male-biased (12.9%) genes; based on the number of X-linked genes (1250/7853), approximately 16% of genes in a given class would be expected to be on the X chromosome. In order to minimize the possibility that a gene that shows biased gene expression in the germline of one sex may be expressed in the soma of the opposite sex, we also used gene expression levels in heads to quantify sex specificity even more stringently, as described in the Materials and Methods. After doing this, our sex-biased data set was considerably smaller, with only 39 female-biased genes, but the distribution of genes among X and A was almost identical to that for the germline data set (unbiased, 16.3% on the X; male-biased, 12.9% on the X; female-biased, 20.8% on the X). For all three categories in the germline-only data set (unbiased, male- and female-biased), genes on the X showed a higher rate of adaptive nonsynonymous site evolution as measured by ωa, the rate of substitution of positively selected mutations for a given class of functional sites relative to that for synonymous mutations (fig. 1 and supplementary table 1, Supplementary Material online). UTRs and long introns also showed a faster-X effect, although this was not significant for female-biased genes in long introns (fig. 1 and supplementary table 1, Supplementary Material online). Fig. 1. View largeDownload slide Plots of ωa values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased, and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with larger ωa value for A than X. Fig. 1. View largeDownload slide Plots of ωa values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased, and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with larger ωa value for A than X. Use of the more conservative data set that used expression data on heads gave results that were similar to those for the germline-only data sets, with the sole exception of female-biased 3′ UTRs (supplementary table 2, Supplementary Material online). Overall, the results of Ávila et al. (2015) were confirmed by the present analysis, suggesting that factors other than dominance level may be involved in the faster-X effect, at least for female-biased genes. As seen in previous studies (Baines et al. 2008; Parsch and Ellegren 2013; Ávila et al. 2014; Kousathanas et al. 2014; Ávila et al. 2015), male-biased genes have the fastest rates of adaptive evolution, for both X and A (fig. 1 and supplementary tables 1 and 2, Supplementary Material online). Factors Potentially Influencing the Faster-X Effect Narrower breadth and lower levels of gene expression are known to be associated with increased rates of sequence evolution (Larracuente et al. 2008; Meisel et al. 2012). Using the criteria described in the Materials and Methods, we found that, overall, X genes had a narrower breadth of gene expression than A genes (P = 0.02). This suggests that ωa could be higher for the X because of a difference in breadth of gene expression. However, we observed no significant differences in expression breadth between X and A within the three pairs of groups of interest (unbiased, male-biased and female-biased genes) (supplementary table 3, Supplementary Material online). We also found no difference in average expression levels between X and A (P = 0.58). When we compared X and A within the three pairs of groups of interest, we again found no significant differences (supplementary table 3, Supplementary Material online). Thus, neither breadth of expression nor average expression seem to be involved in the higher rate of adaptive evolution of female-biased genes for X compared with A. The rate of adaptive sequence evolution is also known to be positively correlated with the rate of crossing over per Megabase (Campos et al. 2014; Charlesworth and Campos 2014; Castellano et al. 2016). As was found in previous studies (Comeron et al. 2012; Campos et al. 2013), the X chromosome had a significantly higher effective rate of crossing over than the autosomes: mean for A = 1.14 cM/Mb, with 95% confidence interval (CI) (1.13, 1.16), and mean for X = 1.96 cM/Mb (CI: 1.86, 1.95) (this statistic corrects for the differential transmission of X and A through females and males: see Materials and Methods for details). Campos et al. (2013) showed that NeX/NeA (estimated from the X/A ratio of 4-fold degenerate site diversities) for X and A genes with comparable effective recombination rates was close to ¾, except for genes in regions with very low rates of crossing over. Comparisons of X and A genes in regions with similar effective recombination rates should therefore largely remove any effects of departures of NeX/NeA from ¾, which might cause a faster-X effect for female-biased genes (see the introduction). To adjust for differences in effective recombination rates between X and A, we divided the genes for both X and A into four bins of recombination rates, using the germline data set, because it had a larger sample size for male- and female-biased genes; as shown above, the initial results were similar for both data sets. The effective recombination rates for A and X genes did not overlap for rates above ∼2 cM/Mb, so that only the first two bins for X overlap those for A (see fig. 2). For nonsynonymous mutations, the confidence intervals for ωa for the bins of unbiased and male-biased X genes with the lowest recombination rate overlapped those for A (fig. 2 and supplementary table 4, Supplementary Material online). However, the remaining X bins had significantly higher ωa values than the corresponding A bins, including the second X bin with a similar effective recombination rate to the fourth bin of autosomal genes (proportion of ωa values greater on the A: P = 0 and P = 0.016 for unbiased and male-biased genes, respectively). In contrast, female-biased genes (F) with similar effective recombination rates did not show a faster-X effect (fig. 2 and supplementary table 4, Supplementary Material online; 2nd A bin vs. 1st X bin, P = 0.349; 4th A bin vs. 2nd X bin, P = 0.92). Due to the small sample size of the first two XF bins that overlap in recombination with AF, we also grouped them into a single bin (21 genes); however, there was no significant difference from the whole AF data set (P = 0.622, ωa XF = 0.057 vs. ωaAF = 0.063). Fig. 2. View largeDownload slide Plots of ωa against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωa values. Because of the small number of female-biased genes, only three bins were used for long introns. Fig. 2. View largeDownload slide Plots of ωa against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωa values. Because of the small number of female-biased genes, only three bins were used for long introns. We carried out similar analyses for noncoding sequences (fig. 2 and supplementary table 4, Supplementary Material online). For unbiased and male-biased genes, the patterns for UTRs and long introns were similar to those for nonsynonymous sites, such that ωa for the X bin with the lowest recombination rate did not differ significantly from that for A genes with similar recombination rates (P > 0.05). The remaining bins for UTRs and long introns of unbiased and male-biased genes showed a faster-X effect, including the 2nd X bin that overlaps in recombination rate with the 4th A bin (P = 0). For UTRs of female-biased genes, there was a faster-X effect across all bins (P < 0.05). This result was stronger for 5′ UTRs (supplementary fig. 1, Supplementary Material online; 2nd A bin—1st X bin, P = 0.056; 4th A bin—2nd X bin, P = 0.008) but was also seen with 3′ UTRs (supplementary fig. 1, Supplementary Material online; 2nd A bin—1st X bin, P = 0.014; 4th A bin—2nd X bin, P = 0.179). For female-biased genes, there was a faster-X effect only for long introns in bins with a higher effective recombination rate on the X (fig. 2; 2nd A bin—1st X bin, P = 0.91; note that, due to the small sample size, there is only one X bin that overlaps with A in recombination rate). Purifying Selection on X and A We also performed several tests to compare the strength of purifying selection on mutations in X and A genes, since the exposure of mutations to selection in the hemizygous state in males is expected to result in larger selection coefficients for rare, deleterious X-linked mutations relative to comparable autosomal mutations, with the exception of mutations with male-limited effects and a dominance coefficient h ≥ 2/3, as shown in the overview of relevant population genetics theory in the Supplementary Material online. The theoretical results imply that parameters such as the rates of fixation of weakly deleterious mutations, and the ratios of diversity levels at sites under selection to those at neutral sites, are dependent on the product of the strength of selection and the effective population size (Ne). The interpretation of the measures described below is therefore complicated by possible differences in Nebetween X and A, especially as Ne may be affected by hitchhiking and hence by the rate of recombination for the genomic region in question, as discussed above (see also Vicoso and Charlesworth 2009; Charlesworth and Campos 2014). The first measure of the efficacy of purifying selection that we used was the estimate of the rate of nonadaptive substitutions at functional sites, relative to the rate for synonymous substitutions (ωna). For nonsynonymous sites, this was consistently lower for X than for A for all expression categories (fig. 3 and supplementary table 5, Supplementary Material online), suggesting a larger product of effective population size (Ne) and strength of selection against slightly deleterious mutations for X compared with A. The X—A differences were significant for the combined UTR data set for all categories of gene expression, as well as for 5′ UTRs, 3′ UTRs and introns of unbiased genes, for 5′ UTRs of female-biased genes, and for 3′UTRs of male-biased genes. Fig. 3. View largeDownload slide Plots of ωna values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with an ωna value larger for X than A. Fig. 3. View largeDownload slide Plots of ωna values for all categories and gene features, using germline expression levels to classify genes as unbiased, male-biased and female-biased (vertical bars represent 95% bootstrap confidence intervals). Asterisks show significant (P < 0.05) differences between A and X values, calculated as the proportion of bootstraps with an ωna value larger for X than A. However, ωna makes several assumptions that could affect the results (see the Discussion), especially the use of a model of selection that assumes semidominance with no sex-specific fitness effects (Eyre-Walker and Keightley 2009). We therefore used diversity statistics to estimate the efficacy of purifying selection; where appropriate, we applied correction factors to take into account possible differences in Ne between the X and A (see the theoretical section of the Supplementary Material online). For the diversity data for the chromosomes as a whole, except for very low recombination regions, our previous work showed that the effective population size for the X chromosome, NeX, is approximately equal to that for the autosomes, NeA, when X and A regions with similar effective recombination rates are compared (Campos et al. 2013). For semidominant mutations with h = ½ and sufficiently strong selection (Nes ≫ 1), diversity for functional sites on the X (πF) is expected to be approximately three-quarters of that for A, assuming that these mutations are sufficiently strongly selected that Nes ≫ 1; it will be less than this for partially recessive unbiased genes and male-biased genes. When NeX = NeA, the same should apply to the ratio of the mean pairwise nucleotide site diversity at putatively functional sites to the mean synonymous site diversity (πF/πS), assuming the latter to be close to neutral; this statistic has the advantage of reducing variability among different classes of genes caused by differences in their mutation rates. Tables 1 and 2 show the mean values and 95% bootstrap confidence intervals for πF and πF/πS, respectively, for each type of sequence and chromosome. For nonsynonymous sites, especially for πF/πS for unbiased genes, there is a clear indication of stronger selective constraints on X chromosomal genes; the same pattern is also noticeable for UTRs and introns in unbiased genes. When we applied a correction factor of 4/3 to the X, πF and πF/πSfor X were similar to, or larger than, the values for A (tables 1 and 2), so that there was no evidence for partial recessivity of deleterious mutations. As expected, πF/πS is larger for UTRs and introns than nonsynonymous sites. It is interesting to note that πF/πS for nonsynymous sites is larger for male-biased and female-biased than unbiased genes, paralleling what is seen for ωna (fig. 3 and supplementary table 5, Supplementary Material online), and suggesting that genes with sex-biased expression have lower levels of selective constraints. This pattern is not seen for UTRs and introns. Table 1. πF Values for All Expression Categories and Functional Classes. Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.00122 (0.00118, 0.00126)  0.00109 (0.00101, 0.00118)  0.00145 (0.00135, 0.00157)  1    Male  0.00161 (0.00153, 0.00171)  0.00161 (0.00139, 0.00181)  0.00214 (0.00185, 0.00242)  1    Female  0.00171 (0.00153, 0.00192)  0.00159 (0.00121, 0.00203)  0.00212 (0.00161, 0.00271)  0.919  5′ and 3´ UTRs  Unbiased  0.00370 (0.00362, 0.00379)  0.00359 (0.00343, 0.00375)  0.00479 (0.00457, 0.00500)  1    Male  0.00419 (0.00404, 0.00433)  0.00425 (0.00383, 0.00466)  0.00567 (0.00510, 0.00621)  1    Female  0.00394 (0.00359, 0.00435)  0.00393 (0.00338, 0.00453)  0.00524 (0.00450, 0.00604)  0.999  5′ UTRs  Unbiased  0.00400 (0.00389, 0.00412)  0.00374 (0.00350, 0.00398)  0.00499 (0.00467, 0.00531)  1    Male  0.00419 (0.00398, 0.00441)  0.00445 (0.00399, 0.00499)  0.00593 (0.00532, 0.00665)  1    Female  0.00453 (0.00395, 0.00522)  0.00389 (0.00308, 0.00472)  0.00518 (0.00411, 0.00630)  0.849  3´ UTRs  Unbiased  0.00350 (0.00341, 0.00359)  0.00351 (0.00331, 0.00368)  0.00467 (0.00442, 0.00491)  1    Male  0.00418 (0.00399, 0.00437)  0.00413 (0.00360, 0.00467)  0.00551 (0.00480, 0.00622)  1    Female  0.00356 (0.00316, 0.00400)  0.00392 (0.00314, 0.00475)  0.00523 (0.00419, 0.00634)  0.995  Introns  Unbiased  0.00470 (0.00459, 0.00481)  0.00472 (0.00455, 0.00490)  0.00630 (0.00607, 0.00653)  1    Male  0.00474 (0.00458, 0.00490)  0.00482 (0.00437, 0.00531)  0.00643 (0.00583, 0.00708)  1    Female  0.00444 (0.00406, 0.00488)  0.00458 (0.00413, 0.00497)  0.00611 (0.00550, 0.00663)  1  Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.00122 (0.00118, 0.00126)  0.00109 (0.00101, 0.00118)  0.00145 (0.00135, 0.00157)  1    Male  0.00161 (0.00153, 0.00171)  0.00161 (0.00139, 0.00181)  0.00214 (0.00185, 0.00242)  1    Female  0.00171 (0.00153, 0.00192)  0.00159 (0.00121, 0.00203)  0.00212 (0.00161, 0.00271)  0.919  5′ and 3´ UTRs  Unbiased  0.00370 (0.00362, 0.00379)  0.00359 (0.00343, 0.00375)  0.00479 (0.00457, 0.00500)  1    Male  0.00419 (0.00404, 0.00433)  0.00425 (0.00383, 0.00466)  0.00567 (0.00510, 0.00621)  1    Female  0.00394 (0.00359, 0.00435)  0.00393 (0.00338, 0.00453)  0.00524 (0.00450, 0.00604)  0.999  5′ UTRs  Unbiased  0.00400 (0.00389, 0.00412)  0.00374 (0.00350, 0.00398)  0.00499 (0.00467, 0.00531)  1    Male  0.00419 (0.00398, 0.00441)  0.00445 (0.00399, 0.00499)  0.00593 (0.00532, 0.00665)  1    Female  0.00453 (0.00395, 0.00522)  0.00389 (0.00308, 0.00472)  0.00518 (0.00411, 0.00630)  0.849  3´ UTRs  Unbiased  0.00350 (0.00341, 0.00359)  0.00351 (0.00331, 0.00368)  0.00467 (0.00442, 0.00491)  1    Male  0.00418 (0.00399, 0.00437)  0.00413 (0.00360, 0.00467)  0.00551 (0.00480, 0.00622)  1    Female  0.00356 (0.00316, 0.00400)  0.00392 (0.00314, 0.00475)  0.00523 (0.00419, 0.00634)  0.995  Introns  Unbiased  0.00470 (0.00459, 0.00481)  0.00472 (0.00455, 0.00490)  0.00630 (0.00607, 0.00653)  1    Male  0.00474 (0.00458, 0.00490)  0.00482 (0.00437, 0.00531)  0.00643 (0.00583, 0.00708)  1    Female  0.00444 (0.00406, 0.00488)  0.00458 (0.00413, 0.00497)  0.00611 (0.00550, 0.00663)  1  Note.—πFfor the X shown in the fifth column was corrected by multiplying by 4/3. (95% bootstrap confidence intervals in brackets.) P, Proportion of bootstraps in which the value of πF for X (after multiplication by 4/3) was greater than that for A. Table 2. πF/πS Values for All Categories and Gene Features. Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.0882 (0.0851, 0.0912)  0.0695 (0.0644, 0.0753)  0.0927 (0.0859, 0.1004)  0.866    Male  0.0997 (0.0938, 0.1062)  0.0933 (0.0789, 0.1091)  0.1244 (0.1053, 0.1455)  0.997    Female  0.1082 (0.0935, 0.1254)  0.0827 (0.0596, 0.1107)  0.1103 (0.0794, 0.1476)  0.52  5′ and 3′ UTR  Unbiased  0.2732 (0.2661, 0.2810)  0.2335 (0.2213, 0.2451)  0.3113 (0.2950, 0.3268)  1    Male  0.2581 (0.2462, 0.2702)  0.2492 (0.2246, 0.2740)  0.3322 (0.2995, 0.3653)  1    Female  0.2530 (0.2221, 0.2857)  0.2141 (0.1719, 0.2683)  0.2854 (0.2292, 0.3577)  0.8  5′ UTRs  Unbiased  0.2970 (0.2875, 0.3065)  0.2425 (0.2263, 0.2595)  0.3233 (0.3017, 0.3459)  0.985    Male  0.2571 (0.2420, 0.2728)  0.2626 (0.2292, 0.2998)  0.3501 (0.3055, 0.3998)  1    Female  0.2916 (0.2501, 0.3413)  0.2083 (0.1539, 0.2724)  0.2778 (0.2052, 0.3632)  0.384  3′ UTRs  Unbiased  0.2590 (0.2504, 0.2669)  0.2278 (0.2139, 0.2426)  0.3038 (0.2852, 0.3235)  1    Male  0.2559 (0.2418, 0.2699)  0.2429 (0.2141, 0.2728)  0.3239 (0.2855, 0.3638)  0.999    Female  0.2289 (0.1974, 0.2625)  0.2125 (0.1667, 0.2682)  0.2833 (0.2223, 0.3576)  0.915  Introns  Unbiased  0.3331 (0.3207, 0.3456)  0.3087 (0.2912, 0.3258)  0.4115 (0.3883, 0.4345)  1    Male  0.2818 (0.2663, 0.2981)  0.2887 (0.2513, 0.3334)  0.3849 (0.3351, 0.4446)  1    Female  0.2436 (0.2129, 0.2791)  0.2471 (0.2079, 0.3052)  0.3294 (0.2772, 0.4069)  0.997  Functional Class  Expression Bias  A  X  Corrected X Value  P  Nonsynonymous  Unbiased  0.0882 (0.0851, 0.0912)  0.0695 (0.0644, 0.0753)  0.0927 (0.0859, 0.1004)  0.866    Male  0.0997 (0.0938, 0.1062)  0.0933 (0.0789, 0.1091)  0.1244 (0.1053, 0.1455)  0.997    Female  0.1082 (0.0935, 0.1254)  0.0827 (0.0596, 0.1107)  0.1103 (0.0794, 0.1476)  0.52  5′ and 3′ UTR  Unbiased  0.2732 (0.2661, 0.2810)  0.2335 (0.2213, 0.2451)  0.3113 (0.2950, 0.3268)  1    Male  0.2581 (0.2462, 0.2702)  0.2492 (0.2246, 0.2740)  0.3322 (0.2995, 0.3653)  1    Female  0.2530 (0.2221, 0.2857)  0.2141 (0.1719, 0.2683)  0.2854 (0.2292, 0.3577)  0.8  5′ UTRs  Unbiased  0.2970 (0.2875, 0.3065)  0.2425 (0.2263, 0.2595)  0.3233 (0.3017, 0.3459)  0.985    Male  0.2571 (0.2420, 0.2728)  0.2626 (0.2292, 0.2998)  0.3501 (0.3055, 0.3998)  1    Female  0.2916 (0.2501, 0.3413)  0.2083 (0.1539, 0.2724)  0.2778 (0.2052, 0.3632)  0.384  3′ UTRs  Unbiased  0.2590 (0.2504, 0.2669)  0.2278 (0.2139, 0.2426)  0.3038 (0.2852, 0.3235)  1    Male  0.2559 (0.2418, 0.2699)  0.2429 (0.2141, 0.2728)  0.3239 (0.2855, 0.3638)  0.999    Female  0.2289 (0.1974, 0.2625)  0.2125 (0.1667, 0.2682)  0.2833 (0.2223, 0.3576)  0.915  Introns  Unbiased  0.3331 (0.3207, 0.3456)  0.3087 (0.2912, 0.3258)  0.4115 (0.3883, 0.4345)  1    Male  0.2818 (0.2663, 0.2981)  0.2887 (0.2513, 0.3334)  0.3849 (0.3351, 0.4446)  1    Female  0.2436 (0.2129, 0.2791)  0.2471 (0.2079, 0.3052)  0.3294 (0.2772, 0.4069)  0.997  Note.—πF/πSfor the X shown in the fifth column was corrected by multiplying by 4/3. (95% bootstrap confidence intervals in brackets.) P, Proportion of bootstraps in which the value of πF for X (after multiplication by 4/3) was greater than that for A. We also examined how the higher overall effective recombination rate of the X might affect these statistics, by increasing Ne as a result of weaker hitchhiking effects. For all types of functional sequence, there was no difference in ωna between X and A for regions that overlap in their effective recombination rates (fig. 4), suggesting that the difference in effective population sizes between X and A cancels the X-A differences in selection coefficients against new deleterious mutations (see the theoretical section of the Supplementary Material online). The same pattern was found for πF/πS (supplementary fig. 2 and table 4, Supplementary Material online). Since our previous work showed that the ratio NeX/NeA was approximately ¾ in overlap regions other than those with low recombination rates (Campos et al. 2013), we did not use a correction factor for πF/πS for the overlap regions studied here, because this difference in Ne exactly cancels the difference in strength of selection on the hypothesis of semidominance. But after applying the 4/3 correction, πF for the overlap regions was either higher for X than for A or similar to A (supplementary fig. 3, Supplementary Material online). Again, there was no evidence for significant departures from semidominance for slightly deleterious mutations. Fig. 4. View largeDownload slide Plots of ωna against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωna values. Because of the low number of female-biased genes, only three bins were used for long introns. Fig. 4. View largeDownload slide Plots of ωna against the mean effective recombination rates for bins of genes, for the three categories of functional sequences. Confidence intervals were obtained by bootstrapping. Lines joining the bins are quadratic fits for the four ωna values. Because of the low number of female-biased genes, only three bins were used for long introns. Discussion Under suitable conditions on the levels of dominance of the fitness effects of beneficial mutations, there should be a faster rate of fixation of new mutations on the X chromosome than the autosomes—the faster-X effect (Rice 1984; Charlesworth et al. 1987; Meisel and Connallon 2013). Genes expressed only in males are the most likely to show such an effect, whereas it should not be seen for genes expressed only in females, unless the effective population size for the X is higher than that for the autosomes. Consistent with these theoretical predictions, the results presented above show that substitutions in most types of functionally significant sequences in male-biased and unbiased genes (nonsynonymous sites, UTRs, and long introns) tend to exhibit faster rates of substitution of beneficial mutations for X genes compared with A genes (figs. 1 and 2). However, our analyses also showed that genomic regions with very low rates of recombination do not show a faster-X effect (fig. 2 and supplementary table 4, Supplementary Material online), consistent with previous observations that showed a lack of adaptive evolution at nonsynonymous sites in very low recombination regions of the D. melanogaster genome (Campos et al. 2014). This observation was interpreted in terms of strong Hill–Robertson interference in such regions (Hill and Robertson 1966; Felsenstein 1974), whereby both positive and negative selection at closely linked sites reduce the rate of fixation of beneficial mutations (Campos et al. 2014; Charlesworth and Campos 2014; Castellano et al. 2016). The influence of recombination rate on adaptive evolution caused by Hill–Robertson interference is also probably responsible for the fact that the faster-X effect seen for nonsynonymous and long intron substitutions in female-biased genes disappears after correcting for their recombination rates—for these categories, only genes with a higher effective recombination rate on the X chromosome than the autosomes show a faster-X effect. Thus, the faster-X effect that we observed previously for nonsynonymous mutations in female-biased genes (Ávila et al. 2015) is probably due to an overall higher recombination rate for these genes, which reduces their exposure to Hill–Robertson interference. As has been found previously for many different systems (Baines et al. 2008; Parsch and Ellegren 2013; Ávila et al. 2014; Kousathanas et al. 2014; Ávila et al. 2015), male-biased genes exhibited higher rates of adaptive evolution at nonsynonymous sites compared with unbiased and female-biased genes, for both X and A genes. Such an effect was not observed, however, for UTRs and long introns (fig. 1 and supplementary table 1, Supplementary Material online). It is unclear why such a difference between types of sequence should occur, and it casts doubt on the frequently expressed idea that sexual selection causes faster rates of evolution of male-biased genes (Parsch and Ellegren 2013), since this is just as likely to affect regulatory mutations as amino-acid mutations. We also found evidence for weaker selective constraints on the protein sequences of genes with sex-biased expression, but not on UTRs or long introns. The extent of between-species divergence in the level of gene expression has been also found to be greater for X than A in Drosophila (Kayserili et al. 2012; Llopart 2012; Meisel et al. 2012), suggesting faster-X adaptive evolution of cis-regulatory sequences (Kayserili et al. 2012; Meisel et al. 2012; Coolon et al. 2015). This hypothesis is consistent with the observation of faster-X divergence of noncoding sequences in Drosophila (Hu et al. 2013); that study did not, however, analyze the contribution of positively selected substitutions to differences between X and autosomes. Our results provide quantitative estimates of the extent of X–A differences in the rates of fixation due to positive selection (figs. 1 and 2 and supplementary table 1, Supplementary Material online), and demonstrate an adaptive faster-X effect for UTRs and long introns, consistent with the results of Begun et al. (2007) and Mackay et al. (2012). The faster-X effect for noncoding regions may well be related to the faster X-effect for gene expression. However, the faster-X effect that was found for the UTRs of female-biased genes (figs. 1 and 2 and supplementary table 1, Supplementary Material online) is not predicted by the standard theoretical models, and is not associated with a difference in recombination rate; furthermore, no faster-X effect for gene expression in genes with expression biased towards female-limited reproductive tissues was found by Meisel et al. (2012). Differences between X and A in traits other than male hemizygosity and recombination rate must be causing this pattern, possibly reflecting the specific biological functions of the small group of genes involved. A much-debated question has been the relative contributions of protein-coding sequences versus noncoding sequences to adaptive evolution (e.g., Carroll 2000; Hoekstra and Coyne 2007; Stern and Orgogozo 2008). Our analyses showed that mutations in functional noncoding regions showed considerably higher rates of adaptive evolution than nonsynonymus mutations. These results are consistent with previous work on Drosophila using both McDonald–Kreitman type approaches (Andolfatto 2005; Begun et al. 2007; Mackay et al. 2012; Campos et al. 2017) and selective sweep signals (Elyashiv et al. 2016). One explanation is that noncoding regions are less constrained by purifying selection than nonsynonymous sites (ωna is higher for noncoding regions than for nonsynonymous sites [figs. 3 and 4 and supplementary table 5, Supplementary Material online]), possibly because they are subject to fewer pleiotropic restrictions (Orr 1998; Carroll 2000; Orr 2000), making it easier for a new mutation to confer a selective advantage. These results are similar to the recent finding of higher rates of positive selection observed in genes that are less constrained by purifying selection (Campos et al. 2017). Higher rates of adaptive sequence evolution in noncoding regions compared with protein sequences have been also found in other systems, such as mice (Halligan et al. 2013) and the flowering plant Capsella (Williamson et al. 2014). We found evidence for differences in the strength of purifying selection between X and A, at least for unbiased genes, when examining the overall patterns of diversity for the two chromosomes (tables 1 and 2). Such a pattern is expected for sufficiently strongly selected deleterious mutations (Nes ≫ 1) with dominance coefficients h ≤ ½ (see the theoretical section of the Supplementary Material online). There is much evidence to suggest that h for deleterious mutations is mostly nonzero but < ½, that is, mutations are partially recessive in their effects on fitness (Dobzhansky and Wright 1941; Muller 1950; Crow and Simmons 1983; García-Dorado and Caballero 2000; Manna et al. 2011). However, our analyses of diversities at functional sites, πF, (table 1 and supplementary fig. 3, Supplementary Material online) and their ratios to synonymous site diversities, πF/πS, (table 2 and supplementary fig. 2, Supplementary Material online) showed no evidence that h was substantially less than ½. The same conclusion was suggested by comparisons of ωna between X and A (figs. 3 and 4 and supplementary table 5, Supplementary Material online). It is important to note, however, that the estimates of statistics such as ωna and α (the proportion of between-species substitutions in a functional category that were fixed by positive selection) could be influenced by the dominance coefficients of deleterious mutations. The DFE-alpha method for estimating these quantities (Eyre-Walker and Keightley 2009) predicts the contribution of slightly deleterious mutations to variation and evolution using a population genetic model of autosomal loci subject to deleterious mutations with h = ½. This raises the question of the reliability of parameter estimates from DFE-alpha when h is < ½, for which the selection equations differ substantially from those with semidominance, except for mutations with no fitness effects on males (see the theoretical section of the Supplementary Material online). Using an approach similar to DFE-alpha, Veeramah et al. (2014) showed that, except for very low h values (h < 0.2), the dominance coefficient has only a small effect on the relative values of α estimates for X and A for human data (see fig. 2 of Veeramah et al. 2014). Their figure 2 suggests that the assumption of h = ½, as was made here, should not lead to serious problems inferences concerning the faster-X effect for adaptive mutations. This conclusion is strengthened by the analyses of purifying selection just described, which suggested that h is close to ½ for deleterious mutations (see tables 1 and 2). Similarly, recessive effects of deleterious variants cannot explain the observed pattern of higher codon usage on the X (see Campos et al. 2013). The comparisons between X and A of the corrected values of πF and πF/πS(tables 1 and 2) yielded many cases in which the X–A difference was significantly larger than that predicted on the hypothesis of strong selection with h = ½. This anomaly is larger for introns and UTRs. If there is a sizeable contribution from weakly selected mutations with Nes ≤ 1, the expected X/A ratio of πF/πSwill be greater than 3/4; in the limit of all mutations being neutral, it will approach 1 when NeX = NeA, as is the case for the two chromosomes as a whole. Multiplication by 4/3 would then be an overcorrection. Consistent with this hypothesis, inspection of the DFEs of the different sequence classes suggests that a considerably greater fraction of noncoding sequence mutations are weakly selected (15–26% mutations have Nes in the range 0–1), than nonsynonymous sites (< 7% with Nes 0–1)—see supplementary table 6, Supplementary Material online. Further theoretical predictions that take stochastic effects and the DFE parameters into account will be needed to test this hypothesis in detail. Materials and Methods We used the same data set as in Ávila et al. (2015), consisting of polymorphism data from 17 haploid genomes from the Gikongoro (Rwanda) population of D. melanogaster provided by the Drosophila Population Genomics Project 2 (Pool et al. 2012). As outgroup material, we used sequences from D. yakuba and D. simulans obtained from the D. melanogaster–D. simulans–D. yakuba gene alignments of Hu et al. (2013). The coding sequence data were filtered and analyzed as described in Campos et al. (2014). In brief, the 17 D. melanogaster genomes were obtained from the DPGP2 data set (http://www.dpgp.org/dpgp2/DPGP2.html). We mapped and aligned the reads to the reference sequence r5.33, available on Flybase [http://flybase.org/] and last accessed December 25, 2012, using BWA (Li and Durbin 2009). We used the Genome Analysis Toolkit (GATK, last accessed December 20, 2012), to do multisample SNP calling (DePristo et al. 2011). From the multisample VCF file, we made a consensus sequence FASTA file for each individual using a custom Perl script. Using the coding sequence coordinates, we extracted their sequences and made FASTA alignments including the D. yakuba and D. simulans outgroups. We excluded 225 genes located in the autosomal heterochromatic regions and on chromosome four, where crossing over is absent (Campos et al. 2012, 2014). We calculated pairwise nucleotide diversity (π, Tajima 1983) at putatively functional sites (πF) and neutral synonymous site diversity (πS). πF/πSwas calculated as the ratio of the mean of πF to the mean of πS for the set of genes in question. As functional sites we considered nonsynonymous sites, 5′ and 3′ UTRs, and introns longer than 100 bp—shorter introns are under less constraint (Halligan and Keightley 2006). For the analyses of UTRs and introns, we masked any coding sequences and exons, respectively, and excluded any sequence with no outgroup information. To assign sites as synonymous and nonsynonymous and to estimate the nonsynonymous divergence and synonymous divergences, KA and KS, we used the method of Comeron (1995)—see Campos et al. 2014 for further details. To obtain the noncoding divergence data, we applied a Kimura two-parameter correction (Kimura 1980). We used the outgroup D. yakuba for coding sites and UTRs; and D. simulans for long introns. Female crossing over rates for these genes, in terms of centiMorgans per megabase, were obtained from Comeron et al. (2012). These rates were multiplied by two-thirds for X-linked genes and by one-half for autosomal genes, respectively. This procedure provides estimates of “effective recombination rates,” taking into account the absence of recombination between homologous chromosomes in males, and the fact that X-linked genes and autosomal genes spend two-thirds and one-half of their time in females, respectively. This allows comparisons between X-linked and autosomal genes that experience the same rates of recombination as far as evolutionary processes are concerned (Campos et al. 2014). In order to classify genes as unbiased, male- or female-biased, we used the tissue-specific expression values (in RPKM) of Brown et al. (2014). To obtain female expression values, we used female ovaries; for male expression values we used the average of gene expression in testes and accessory glands of male flies. In a first step, we obtained a sex bias ratio (SBR) value as: log 2 (male value/female value). We classified genes with a SBR > 1 as male-biased; genes with SBR < –1 as female-biased, and genes with SBR between –1 and 1 as unbiased. This is equivalent to using a 2-fold expression difference between the sexes as the criterion for sex bias. In a second step, male and female expression values were ranked, and the 15th percentiles of their distributions were selected. If a gene was categorized as male-biased in the first step, and showed expression in females (ovary) of less than the female 15th percentile value, the gene was ultimately labeled as male-biased, and vice-versa for female bias. The aim of the second step was to ensure that biased genes have a very low level of expression in the opposite sex. Genes were then sorted into six categories, based on chromosomal location (autosomal vs. X-linked) and the sex-biased expression categories described above. Our final data set consisted of 6,603 autosomal genes and 1,250 genes. A potential reason for a faster-X effect for female-biased genes (which is contrary to theoretical predictions) is that such genes may exhibit some expression in males in somatic tissues, allowing positive selection to act when the gene is in a hemizygous state, even though the gene has been classed as female-biased according to the gonad-based sex bias ratio. To deal with this problem, we made a second data set using gene expression in both heads and germline to define sex bias. To be classed as male-biased in expression, a gene must then meet two conditions. First, the gene must be expressed in the male gonad and head at a value greater than the male 15th percentile values for both the relevant tissues. Second, expression in the female ovary and head must be lower than the 15th percentile expression values for both tissues, respectively; complementary criteria were used to classify a gene as female-biased. The breadth of expression of a gene across different tissues is known to affect its rates of protein sequence evolution (Larracuente et al. 2008), and thus needs to be taken into account when comparing X and A. We measured this by τ, the proportion of tissues in which a gene is found to be expressed. τ ranges from 0 (for broadly expressed genes) to 1 (for narrowly expressed genes) (Yanai et al. 2005). Similarly, because levels of gene expression affect protein evolution (Larracuente et al. 2008), we calculated level of gene expression as the average expression across tissues for each gene. We used DFE-alpha (Eyre-Walker and Keightley 2009) to estimate ωa, the ratio of the rate of adaptive substitutions per basepair relative to neutral substitutions for nonsynonymous sites, UTR sites and sites in long introns. This method attempts to correct for the contribution of slightly deleterious mutations to polymorphism and divergence and the impact of recent demographic change. We used D. yakuba (KS ≈ 0.25) as an outgroup for CDS and UTRs, and D. simulans (KS ≈ 0.12) for long introns. We used a demographic model where the population at initial size N1 (set to 100 in the DFE-alpha program) experienced a step change to N2 at n generations in the past. A preliminary analysis for the whole data set showed that a constant population size model gave a poor fit to the data, consistent with the results of Pool et al. (2012) that show that D. melanogaster African populations have apparently experienced a population expansion. To calculate confidence intervals we generated replicate bootstrap estimates for each group separately by resampling genes 1,000 times within a given group, and running DFE-alpha for each bootstrap. In order to test for a significant difference in a given statistic between a pair of categories, such as X and A, we calculated the proportion of its bootstrap values that were larger for one category than the other. Acknowledgments This work was supported by Grant RPG-2015-2033 from the Leverhulme Trust (to B.C.). The authors are grateful to the Drosophila Population Genomics Project (DPGP), for making these data available. They thank editors Krushnamegh Kunte and Deepa Agashe for inviting us to contribute to the special issue on Genetics of Adaptation. They also thank three reviewers for constructive comments and suggestions. References Andolfatto P. 2005. Adaptive evolution of non-coding DNA in Drosophila. Nature  437( 7062): 1149– 1152. http://dx.doi.org/10.1038/nature04107 Google Scholar CrossRef Search ADS PubMed  Ávila V, Campos JL, Charlesworth B. 2015. The effects of sex-biased gene expression and X-linkage on rates of adaptive protein sequence evolution in Drosophila. Biol Lett . 11( 4): 20150117. Google Scholar CrossRef Search ADS PubMed  Ávila V, Marion de Procé S, Campos JL, Borthwick H, Charlesworth B, Betancourt AJ. 2014. Faster-X effects in two Drosophila lineages. Genome Biol Evol . 6( 10): 2968– 2982. Google Scholar CrossRef Search ADS PubMed  Baines JF, Sawyer SA, Hartl DL, Parsch J. 2008. Effects of X-linkage and sex-biased gene expression on the rate of adaptive protein evolution in Drosophila. Mol Biol Evol . 25( 8): 1639– 1650. http://dx.doi.org/10.1093/molbev/msn111 Google Scholar CrossRef Search ADS PubMed  Begun DJ, Holloway AK, Stevens K, Hillier LW, Poh Y-P, Hahn MW, Nista PM, Jones CD, Kern AD, Dewey CN, et al.   2007. Population genomics: whole-genome analysis of polymorphism and divergence in Drosophila simulans. PLoS Biol . 5( 11): e310. Google Scholar CrossRef Search ADS PubMed  Brown JB, Boley N, Eisman R, May GE, Stoiber MH, Duff MO, Booth BW, Wen J, Park S, Suzuki AM, et al.   2014. Diversity and dynamics of the Drosophila transcriptome. Nature  512( 7515): 393– 399. Google Scholar CrossRef Search ADS PubMed  Campos JL, Charlesworth B, Haddrill PR. 2012. Molecular evolution in nonrecombining regions of the Drosophila melanogaster genome. Genome Biol Evol . 4( 3): 278– 288. http://dx.doi.org/10.1093/gbe/evs010 Google Scholar CrossRef Search ADS PubMed  Campos JL, Zeng K, Parker DJ, Charlesworth B, Haddrill PR. 2013. Codon usage bias and effective population sizes on the X chromosome versus the autosomes in Drosophila melanogaster. Mol Biol Evol . 30( 4): 811– 823. Google Scholar CrossRef Search ADS PubMed  Campos JL, Halligan DL, Haddrill PR, Charlesworth B. 2014. The relation between recombination rate and patterns of molecular evolution and variation in Drosophila melanogaster. Mol Biol Evol . 31( 4): 1010– 1028. Google Scholar CrossRef Search ADS PubMed  Campos JL, Zhao L, Charlesworth B. 2017. Estimating the parameters of background selection and selective sweeps in Drosophila in the presence of gene conversion. Proc Natl Acad Sci U S A . 114( 24): E4762– E47771. Google Scholar CrossRef Search ADS PubMed  Carneiro M, Albert FW, Melo-Ferreira J, Galtier N, Gayral P, Blanco-Aguiar JA, Villafuerte R, Nachman MW, Ferrand N. 2012. Evidence for widespread positive and purifying selection across the European rabbit (Oryctolagus cuniculus) genome. Mol Biol Evol . 29( 7): 1837– 1849. Google Scholar CrossRef Search ADS PubMed  Carroll SB. 2000. Endless forms: the evolution of gene regulation and morphological diversity. Cell  101( 6): 577– 580. http://dx.doi.org/10.1016/S0092-8674(00)80868-5 Google Scholar CrossRef Search ADS PubMed  Castellano D, Coronado-Zamora M, Campos JL, Barbadilla A, Eyre-Walker A. 2016. Adaptive evolution is substantially impeded by Hill–Robertson interference in Drosophila. Mol Biol Evol . 33( 2): 442– 455. Google Scholar CrossRef Search ADS PubMed  Charlesworth B, Campos JL. 2014. The relations between recombination rate and patterns of molecular variation and evolution in Drosophila. Annu Rev Genet . 48: 383– 403. http://dx.doi.org/10.1146/annurev-genet-120213-092525 Google Scholar CrossRef Search ADS PubMed  Charlesworth B, Coyne J, Barton N. 1987. The relative rates of evolution of sex chromosomes and autosomes. Am Nat . 130( 1): 113– 146. Google Scholar CrossRef Search ADS   Comeron JM. 1995. A method for estimating the numbers of synonymous and nonsynonymous substitutions per site. J Mol Evol . 41( 6): 1152– 1159. Google Scholar CrossRef Search ADS PubMed  Comeron JM, Ratnappan R, Bailin S. 2012. The many landscapes of recombination in Drosophila melanogaster. PLoS Genet . 8( 10): e1002905. Google Scholar CrossRef Search ADS PubMed  Coolon JD, Stevenson KR, McManus CJ, Yang B, Graveley BR, Wittkopp PJ. 2015. Molecular mechanisms and evolutionary processes contributing to accelerated divergence of gene expression on the Drosophila X chromosome. Mol Biol Evol . 32( 10): 2605– 2615. Google Scholar CrossRef Search ADS PubMed  Crow J, Simmons M. 1983. The mutation load in Drosophila. In: Ashburner M, Carson HL, Thompson JN, editors. The genetics and biology of drosophila , Vol. 3c, pp. 1– 35. London: Academic Press. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, Hartl C, Philippakis AA, del Angel G, Rivas MA, Hanna M, et al.   2011. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet . 43( 5): 491– 498. http://dx.doi.org/10.1038/ng.806 Google Scholar CrossRef Search ADS PubMed  Dobzhansky T, Wright S. 1941. Genetics of natural populations. V. Relations between mutation rate and accumulation of lethals in populations of Drosophila pseudoobscura. Genetics  26( 1): 23– 51. Google Scholar PubMed  Elyashiv E, Sattath S, Hu TT, Strutsovsky A, McVicker G, Andolfatto P, Coop G, Sella G. 2016. A genomic map of the effects of linked selection in Drosophila. PLoS Genet . 12( 8): e1006130. Google Scholar CrossRef Search ADS PubMed  Eyre-Walker A, Keightley PD. 2009. Estimating the rate of adaptive molecular evolution in the presence of slightly deleterious mutations and population size change. Mol Biol Evol . 26( 9): 2097– 2108. http://dx.doi.org/10.1093/molbev/msp119 Google Scholar CrossRef Search ADS PubMed  Felsenstein J. 1974. The evolutionary advantage of recombination. Genetics  78( 2): 737– 756. Google Scholar PubMed  García-Dorado A, Caballero A. 2000. On the average coefficient of dominance of deleterious spontaneous mutations. Genetics  155( 4): 1991– 2001. Google Scholar PubMed  Haddrill PR, Thornton KR, Charlesworth B, Andolfatto P. 2005. Multilocus patterns of nucleotide variability and the demographic and selection history of Drosophila melanogaster populations. Genome Res . 15( 6): 790– 799. Google Scholar CrossRef Search ADS PubMed  Haldane JBS. 1924. A mathematical theory of natural and artificial selection. Trans Camb Philos Soc . 23: 19– 41. Halligan DL, Keightley PD. 2006. Ubiquitous selective constraints in the Drosophila genome revealed by a genome-wide interspecies comparison. Genome Res . 16( 7): 875– 884. http://dx.doi.org/10.1101/gr.5022906 Google Scholar CrossRef Search ADS PubMed  Halligan DL, Kousathanas A, Ness RW, Harr B, Eöry L, Keane TM, Adams DJ, Keightley PD. 2013. Contributions of protein-coding and regulatory change to adaptive molecular evolution in murid rodents. PLoS Genet . 9( 12): e1003995. Google Scholar CrossRef Search ADS PubMed  Hill WG, Robertson A. 1966. The effect of linkage on limits to artificial selection. Genet Res . 8( 3): 269– 294. http://dx.doi.org/10.1017/S0016672300010156 Google Scholar CrossRef Search ADS PubMed  Hoekstra HE, Coyne JA. 2007. The locus of evolution: evo-devo and the genetics of adaptation. Evolution  61( 5): 995– 1016. http://dx.doi.org/10.1111/j.1558-5646.2007.00105.x Google Scholar CrossRef Search ADS PubMed  Hu TT, Eisen MB, Thornton KR, Andolfatto P. 2013. A second-generation assembly of the Drosophila simulans genome provides new insights into patterns of lineage-specific divergence. Genome Res . 23( 1): 89– 98. http://dx.doi.org/10.1101/gr.141689.112 Google Scholar CrossRef Search ADS PubMed  Hvilsom C, Qian Y, Bataillon T, Li Y, Mailund T, Sallé B, Carlsen F, Li R, Zheng H, Jiang T, et al.   2012. Extensive X-linked adaptive evolution in central chimpanzees. Proc Natl Acad Sci U S A.  109( 6): 2054– 2059. Google Scholar CrossRef Search ADS PubMed  Kayserili MA, Gerrard DT, Tomancak P, Kalinka AT. 2012. An excess of gene expression divergence on the X chromosome in Drosophila embryos: implications for the faster-X hypothesis. PLoS Genet . 8( 12): e1003200. Google Scholar CrossRef Search ADS PubMed  Kimura M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J Mol Evol . 16( 2): 111– 120. http://dx.doi.org/10.1007/BF01731581 Google Scholar CrossRef Search ADS PubMed  Kousathanas A, Halligan DL, Keightley PD. 2014. Faster-X adaptive protein evolution in house mice. Genetics  196( 4): 1131– 1143. http://dx.doi.org/10.1534/genetics.113.158246 Google Scholar CrossRef Search ADS PubMed  Langley CH, Stevens K, Cardeno C, Lee YCG, Schrider DR, Pool JE, Langley SA, Suarez C, Corbett-Detig RB, Kolaczkowski B, et al.   2012. Genomic variation in natural populations of Drosophila melanogaster. Genetics  192( 2): 533– 598. Google Scholar CrossRef Search ADS PubMed  Larracuente AM, Sackton TB, Greenberg AJ, Wong A, Singh ND, Sturgill D, Zhang Y, Oliver B, Clark AG. 2008. Evolution of protein-coding genes in Drosophila. Trends Genet . 24( 3): 114– 123. Google Scholar CrossRef Search ADS PubMed  Li H, Durbin R. 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics  25( 14): 1754– 1760. http://dx.doi.org/10.1093/bioinformatics/btp324 Google Scholar CrossRef Search ADS PubMed  Llopart A. 2012. The rapid evolution of X-linked male-biased gene expression and the large-X effect in Drosophila yakuba, D. santomea, and their hybrids. Mol Biol Evol . 29( 12): 3873– 3886. http://dx.doi.org/10.1093/molbev/mss190 Google Scholar CrossRef Search ADS PubMed  Mackay TFC, Richards S, Stone EA, Barbadilla A, Ayroles JF, Zhu D, Casillas S, Han Y, Magwire MM, Cridland JM, et al.   2012. The Drosophila melanogaster genetic reference panel. Nature  482( 7384): 173– 178. Google Scholar CrossRef Search ADS PubMed  Mank JE, Vicoso B, Berlin S, Charlesworth B. 2010. Effective population size and the Faster-X effect: empirical results and their interpretation. Evolution  64( 3): 663– 674. http://dx.doi.org/10.1111/j.1558-5646.2009.00853.x Google Scholar CrossRef Search ADS PubMed  Manna F, Martin G, Lenormand T. 2011. Fitness landscapes: an alternative theory for the dominance of mutation. Genetics  189( 3): 923– 937. http://dx.doi.org/10.1534/genetics.111.132944 Google Scholar CrossRef Search ADS PubMed  Meisel RP, Connallon T. 2013. The faster-X effect: integrating theory and data. Trends Genet . 29( 9): 537– 544. http://dx.doi.org/10.1016/j.tig.2013.05.009 Google Scholar CrossRef Search ADS PubMed  Meisel RP, Malone JH, Clark AG. 2012. Disentangling the relationship between sex-biased gene expression and X-linkage. Genome Res . 22( 7): 1255– 1265. http://dx.doi.org/10.1101/gr.132100.111 Google Scholar CrossRef Search ADS PubMed  Muller HJ. 1950. Our load of mutations. Am J Hum Genet . 2( 2): 111– 176. Google Scholar PubMed  Orr HA. 1998. The population genetics of adaptation: The distribution of factors fixed during adaptive evolution. Evolution  52( 4): 935– 949. http://dx.doi.org/10.1111/j.1558-5646.1998.tb01823.x Google Scholar CrossRef Search ADS PubMed  Orr HA. 2000. Adaptation and the cost of complexity. Evolution  54( 1): 13– 20. http://dx.doi.org/10.1111/j.0014-3820.2000.tb00002.x Google Scholar CrossRef Search ADS PubMed  Orr HA, Betancourt A. 2001. Haldane’s sieve and adaptation from the standing genetic variation. Genetics  157( 2): 875– 884. Google Scholar PubMed  Orr HA, Coyne JA. 1989. The genetics of postzygotic isolation in the Drosophila virilis group. Genetics  121( 3): 527– 537. Google Scholar PubMed  Parsch J, Ellegren H. 2013. The evolutionary causes and consequences of sex-biased gene expression. Nat Rev Genet . 14( 2): 83– 87. http://dx.doi.org/10.1038/nrg3376 Google Scholar CrossRef Search ADS PubMed  Parsch J, Novozhilov S, Saminadin-Peter SS, Wong KM, Andolfatto P. 2010. On the utility of short intron sequences as a reference for the detection of positive and negative selection in Drosophila. Mol Biol Evol . 27( 6): 1226– 1234. Google Scholar CrossRef Search ADS PubMed  Pool JE, Corbett-Detig RB, Sugino RP, Stevens KA, Cardeno CM, Crepeau MW, Duchen P, Emerson JJ, Saelao P, Begun DJ, et al.   2012. Population genomics of sub-Saharan Drosophila melanogaster: African diversity and non-African admixture. PLoS Genet . 8( 12): e1003080. Google Scholar CrossRef Search ADS PubMed  Presgraves DC. 2008. Sex chromosomes and speciation in Drosophila. Trends Genet . 24( 7): 336– 343. http://dx.doi.org/10.1016/j.tig.2008.04.007 Google Scholar CrossRef Search ADS PubMed  Rice WR. 1984. Sex chromosomes and the evolution of sexual dimorphism. Evolution  38( 4): 735– 742. http://dx.doi.org/10.1111/j.1558-5646.1984.tb00346.x Google Scholar CrossRef Search ADS PubMed  Rousselle M, Faivre N, Ballenghien M, Galtier N, Nabholz B. 2016. Hemizygosity enhances purifying selection: lack of Fast-Z evolution in two Satyrine butterflies. Genome Biol Evol . 8( 10): 3108– 3119. Google Scholar CrossRef Search ADS PubMed  Stern DL, Orgogozo V. 2008. The loci of evolution: how predictable is genetic evolution?. Evolution  62( 9): 2155– 2177. http://dx.doi.org/10.1111/j.1558-5646.2008.00450.x Google Scholar CrossRef Search ADS PubMed  Veeramah KR, Gutenkunst RN, Woerner AE, Watkins JC, Hammer MF. 2014. Evidence for increased levels of positive and negative selection on the X chromosome versus autosomes in humans. Mol Biol Evol . 31( 9): 2267– 2282. Google Scholar CrossRef Search ADS PubMed  Vicoso B, Charlesworth B. 2009. Effective population size and the faster-X effect: an extended model. Evolution  63( 9): 2413– 2426. http://dx.doi.org/10.1111/j.1558-5646.2009.00719.x Google Scholar CrossRef Search ADS PubMed  Williamson RJ, Josephs EB, Platts AE, Hazzouri KM, Haudry A, Blanchette M, Wright SI. 2014. Evidence for widespread positive and negative selection in coding and conserved noncoding regions of Capsella grandiflora. PLoS Genet . 10( 9): e1004622. Google Scholar CrossRef Search ADS PubMed  Yanai I, Benjamin H, Shmoish M, Chalifa-Caspi V, Shklar M, Ophir R, Bar-Even A, Horn-Saban S, Safran M, Domany E, et al.   2005. Genome-wide midrange transcription profiles reveal expression level relationships in human tissue specification. Bioinformatics  21( 5): 650– 659. Google Scholar CrossRef Search ADS PubMed  © The Author 2017. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com

Journal

Molecular Biology and EvolutionOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off