Post-copulatory sexual selection has been proposed to drive the rapid evolution of reproductive proteins, and, more recently, to increase genome-wide mutation rates. Comparisons of rates of molecular evolution between lineages with different levels of female multiple mating represent a promising, but under-utilized, approach for testing the effects of sperm competition on sequence evolution. Here, I use comparisons between primate species with divergent mating systems to examine the effects of sperm competition on reproductive protein evolution, as well as on sex-averaged mutation rates. Rates of nonsynonymous substitution are higher for testis-specific genes along the chimpanzee lineage in comparison to the human lineage, consistent with expectations. However, the data reported here do not allow firm conclusions concerning the effects of mating system on genome-wide mutation rates, with different results obtained from different species pairs. Ultimately, comparative studies encompassing a range of mating systems and other life history traits will be required to make broad generalizations concerning the genomic effects of sperm competition.

Post-copulatory sexual selection is thought to drive the evolution of traits involved in competition over fertilization opportunities (sperm competition) and in female control over sperm usage (cryptic female choice), just as pre-copulatory sexual selection is responsible for the evolution of a wide array of showy display traits used in obtaining mates. For example, post-copulatory sexual selection has been invoked to explain co-evolution between sperm length and female sperm storage organ size in flies (Pitnick et al. 2003), and the correlation between male testis size and female multiple mating rate in primates (first described by Short 1979).

A number of authors have suggested that post-copulatory sexual selection may be an important determinant of patterns of molecular evolution, most notably at the protein level. Sexual selection is generally thought to contribute to the rapid evolution of reproductive proteins in a wide variety of taxa (reviewed in Clark et al. 2006; Panhuis et al. 2006; Turner and Hoekstra 2008). In flies, for example, genes expressed in the testis or in the male accessory glands (which produce important ejaculatory proteins) have a higher average rate of nonsynonymous substitution, and are more likely to experience positive selection, than are most other groups of genes (Haerty et al. 2007). Functional data indicate roles for some of these genes in sperm storage, sperm competition, and the control of post-mating reproductive behaviors, suggesting that sexual selection drives their rapid evolution.

Recently, it has been suggested that sperm competition may also have consequences for variation in the genome-wide mutation rate (Møller and Cuervo 2003; Whitlock and Agrawal 2009). It is by now widely appreciated that mutation rates are typically higher in the male germline than in the female germline (Makova and Li 2002; Ellegren 2007; Bachtrog 2008), as originally suggested by Haldane (1937). This sex difference is probably due to the extra rounds of cell division, and hence DNA replication, required for the production of sperm in comparison to eggs. A simple extension of this logic leads to the hypothesis that taxa with a higher intensity of sperm competition will experience relatively higher mutation rates in males: If males produce more sperm as an adaptation to sperm competition, then extra cell divisions will be required for higher sperm production in species with intense sperm competition. All else being equal, increased mutation rates in males should lead to an increase in the sex-averaged mutation rate of the entire nuclear genome.

A widespread correlation between the intensity of sperm competition and testis size in animals lends credence to the idea that increased sperm production is a common adaptation to post-copulatory sexual selection (e.g., Short 1979; Kenagy and Trombulak 1986; Jennions and Passmore 1993; Gage 1994). Nonetheless, data addressing the relationship between mating system and sex-averaged mutation rates are limited. In birds (Bartosch-Harlid et al. 2003) and primates (Presgraves and Yi 2009), the ratio of male:female mutation rates α does appear to vary systematically with mating system. In both of these taxonomic groups, α is higher in species where sperm competition is more prevalent, as estimated from rates of divergence on the autosomes and sex chromosomes. Nonetheless, this intriguing observation does not in itself address the sex-averaged mutation rate between species with different levels of sperm competition—only differences between males and females within a species.

Here, I investigate the effects of sperm competition on rates of molecular evolution using comparisons between primates with different mating systems. Primates are an ideal taxonomic group with which to address the effects of sperm competition on rates of molecular evolution, since their mating systems are well characterized, and due to the availability of extensive sequence information. I find increased rates of nonsynonymous substitution in testes-specific genes in a polyandrous lineage (chimpanzee) in comparison to a less polyandrous lineage (human), consistent with the hypothesis that post-copulatory sexual selection drives reproductive protein evolution. By contrast, in three comparisons between species pairs with contrasting mating systems, patterns of synonymous substitution are inconsistent with respect to mating system, suggesting a need for deeper taxonomic sampling.

Materials and Methods


To examine rates of nonsynonymous substitution for reproductive tract genes, I estimated rates of molecular evolution in the species triplet ((Chimpanzee, Human), Macaque), where macaque is an outgroup and chimpanzees have a higher rate of female remating than do humans (Dixson 1998). Alignments for 10,376 nuclear genes for this triplet were obtained from Gibbs et al. (2007). For each gene, rates of nonsynonymous substitution dN were estimated using HyPhy (Pond et al. 2005), under the Goldman and Yang (1994) model of codon evolution. dN was estimated under two models: One allowing branch-specific dN, and a constrained model whereby the human and chimpanzee lineages have the same dN. For each gene, I calculated the difference in nonsynonymous substitution rates between the chimpanzee and the human lineage under the branch-specific model; this difference will be referred to as δdN. A positive value of δdN indicates a higher rate of nonsynonymous substitution along the chimpanzee lineage.

According to the hypothesis that post-copulatory sexual selection drives rates of reproductive protein evolution, mean δdN should be greater than zero for reproductive genes, but not for other loci. To test this hypothesis, I used reproductive tract-specific expression as a proxy for reproductive function. I obtained expression data from Jongeneel et al. 2005, wherein deep sequencing of transcripts was conducted on 32 healthy adult tissues from humans. Expression profiles and chimpanzee/human/macaque sequence alignments were available for 6003 genes in total. For each gene, the tissue-specificity index τ (Yanai et al. 2005) was calculated as follows:


where N is the number of tissues and x is the set of expression values normalized by the maximum expression. τ ranges from 0 (no specificity) to 1 (absolutely tissue specific). I considered a gene to be tissue specific if its τ was in the top quartile of all 6003 values of τ, which corresponds to τ > 0.9581. According to this criterion, 321, 37, and 29 genes were specific to the testis, prostate, and uterus, respectively. A linear model was used to compare δdN between genes expressed specifically in reproductive tissues (testes, ovaries, prostate) and all other tissues, via a call to the lm function in R version 2.9.1 (R Core Development Team 2009).


To compare mutation rates between polyandrous and monandrous lineages, I used rates of synonymous substitution, following on a suggestion by Whitlock and Agrawal (2009). Synonymous substitutions should mirror mutational processes, assuming that most synonymous mutations are neutral or nearly so. According to the hypothesis that post-copulatory sexual selection drives mutation rates, all nuclear loci should experience a higher rate of mutation—i.e., the effects of mating system should not be limited to reproductive tract genes. As such, I was not limited to using whole genome sequence data (in contrast to the previously described analysis, wherein reproductive tract-specific genes were compared to genes elsewhere in the genome). I used three species triplets to examine rates of synonymous substitution: ((Chimpanzee, Human), Macaque); ((Squirrel monkey, owl monkey), Human); and ((Macaque, Colobus), Human), with the more promiscuous species listed first and the outgroup last. A total of 10,364 alignments for nuclear genes for the chimp/human/macaque trio were obtained from Gibbs et al. (2007). Sequences for 30 genes for the Macaque/Colobus/Human triplet, and for 27 genes for the remaining triplet, were obtained from GenBank and from Prasad et al. (2008) and Hurle et al. (2007). GenBank accession numbers are available in the Supporting Information. Mitochondrial gene sequences were obtained from the OGRe database (Jameson et al. 2003; http://drake.physics.mcmaster.ca/ogre/). Alignments were performed using the ClustalW algorithm, as implemented by Mega3.1 (Kumar et al. 2004). Alignment lengths and numbers of synonymous sites for all genes are given in the online Supporting Information.

For each gene analyzed for a given species trio, branch-specific rates of synonymous substitution dS were estimated under the Goldman-Yang model of codon evolution (Goldman and Yang 1994). The difference in synonymous substitution rates δds was calculated as dSpolyandrousdSmonandrous, which will be designated δdS. The hypothesis under consideration states that, on average, δds should be greater than 0, indicating higher substitution rates in the polyandrous lineage. One-sided, one sample t-tests were used for each trio to determine whether average δds was greater than 0.



Previous studies have found evidence for positive selection on individual sperm and seminal fluid proteins in primates (Clark and Swanson 2005; Hamm et al. 2007), and rates of testis protein evolution are elevated on average in some mammals (Turner et al. 2008) and in Drosophila (Haerty et al. 2007). Consistent with these findings, average dN for testis genes in the chimpanzee/human/macaque species trio is significantly higher than average dN for nonreproductive tract genes or for prostate or uterus-specific genes (t= 15.2, P < 0.0001; Fig. 1A). Prostate- and uterus-specific genes showed a slight increase (prostate) or decrease (uterus) in average dN compared to nonreproductive tract-specific genes, but these differences were not significant (Prostate: t= 1.02, P= 0.31; Uterus: t=−0.78, P= 0.43).

Figure 1.

(A) Rapid evolution of testis proteins in the chimpanzee/human/macaque species trio. The rate of nonsynonymous substitution dN was estimated for 6003 genes, with sample sizes for each tissue indicated. Here, dN was constrained to be equal along the chimpanzee and human lineages. Error bars represent ±1 SEM. (B) Elevated dN along the polyandrous chimpanzee lineage. Branch-specific estimates of dN were obtained for the chimpanzee and human lineages, and their difference δdN was calculated. Positive δdN indicates an increased dN along the chimpanzee branch. Error bars represent ±1 SEM. **=P < 0.001.

Comparisons of evolutionary rates between lineages with different mating systems should provide deeper insight into the role of post-copulatory sexual selection in driving reproductive protein evolution (Jensen-Seaman and Li 2003; Kingan et al. 2003; Dorus et al. 2004; Herlyn and Zischler 2007; Kelleher et al. 2007; Wagstaff and Begun 2007; Ramm et al. 2008; Walters and Harrison 2008). If sperm competition is responsible for the observed elevation of dN for testis-specific genes in the chimpanzee/human/macaque species trio, then dN should be higher for testis genes along the chimpanzee lineage than along the human lineage, given higher female remating rates among female chimpanzees. Consistent with this prediction, I find that the difference in branch-specific dN between the chimpanzee and human lineages (δdN) is significantly greater than 0 (t= 3.3, P < 0.001; Fig. 1B). By contrast, δdN is not significantly different from 0 for nonreproductive genes, or for genes expressed in other reproductive tissues (Prostate: t= 1.06, P= 0.29; Uterus: t= 0.61, P= 0.54; Fig. 1B). Thus, these data support the hypothesis that sperm competition drives the rapid evolution of testis proteins.

Ideally, it would be informative to compare levels of positive selection on reproductive proteins in the chimpanzee and human lineages. However, divergence between these two species is low, such that there is little power to detect positive selection—recent genome-scale analyses find evidence for positive selection on only 10 genes along the human lineage, and on eighteen genes along the chimpanzee lineage (Kosiol et al. 2008). Of the 18 genes inferred by Kosiol et al. (2008) to have been subject to positive selection along the chimpanzee lineage, two (scml1 and C14orf178) have testis-biased expression (see also Wu and Su 2004). None of the 10 genes showing evidence for positive selection along the human lineage shows preferential expression in reproductive tissues.


While the rapid evolution of reproductive genes has been well-documented, only a few studies have investigated the effects of sexual selection on genome-wide mutation rates (Møller and Cuervo 2003; Nadeau et al. 2007). I estimated branch-specific rates of synonymous substitution at multiple nuclear loci for three species triplets, to determine whether mating system influences rates of mutation. I found no evidence that δdS was greater than 0 for nuclear genes for two of the species triplets (Fig. 2; Table 1), contrary to the initial hypothesis. By contrast, for the squirrel monkey/owl monkey species pair, δdS was significantly greater than 0 (mean δdS= 0.48; one-sided t-test t= 2.76, P= 0.0052). Thus, the available data do not provide strong evidence either for or against an effect of mating system on mutation rates.

Figure 2.

Effects of mating system on rates of synonymous substitution dS. Branch-specific estimates of dS were estimated for multi-locus datasets (sample sizes indicated) for the chimpanzee/human/macaque, macaque/colobus/human, and squirrel monkey/owl monkey/human trios. δdS, the difference in dS between the polyandrous and monandrous lineages, is positive when dS is higher for the polyandrous branch. Error bars represent ±1 SEM. **=P < 0.01.

Table 1.  Tests for deviations from δdS= 0.
Species pairGene setnAlignment length#syn sites Mean δdSP-value
Chimpanzee/humanNuclear10,37614,389,0953,375,338.3  8.6×10−50.31
Macaque/colobusNuclear3050,19311,221.4 −0.120.88
Squirrel monkey/Owl monkeyNuclear2743,5189401.61 0.480.0052
Chimpanzee/humanMitochondrial1311,4063030.4 −0.100.20
Macaque/colobusMitochondrial1311,4063010.3 −0.080.10
Squirrel monkey/Owl monkeyMitochondrial1210,8812876.3 −0.070.28

If male mutation rates are increased due to multiple mating, then a concomitant reduction in mutation rate in females might account for the lack of a strong effect of mating system on sex-averaged mutation rates. To test this hypothesis, I calculated δdS for the mitochondrial genome for each species triplet. Since mitochondria are maternally inherited, this analysis should reflect molecular evolution uninfluenced by males. I again found no effect of mating system on the rate of synonymous substitution (Table 1), although for each of the species triplets δdS was negative, as expected if female mutation rates are lower in multiply mating lineages. It should also be noted that the molecular machinery used to replicate mitochondrial DNA is substantially different from that used for nuclear DNA, such that rates of mtDNA evolution may not accurately reflect patterns of nuclear gene evolution.


There is substantial evidence from a wide range of taxa that genes expressed in reproductive tissues tend to evolve rapidly, with positive selection contributing to this pattern (e.g., Panhuis et al. 2006; Turner and Hoekstra 2008). While the phenomenon of rapid reproductive tract protein evolution is consistent with sexual selection acting on these molecules, it may also be explained by other evolutionary processes, such as immune interactions (e.g., Ramm et al. 2008; Wong et al. 2008). Other approaches, such as detailed functional studies, experimental evolution, and comparative analyses should help to disentangle the effects of sexual selection on reproductive proteins.

A number of previous studies have compared rates of evolution of reproductive proteins between taxa with different mating systems (e.g., Jensen-Seaman and Li 2003; Kingan et al. 2003; Dorus et al. 2004; Herlyn and Zischler 2007; Kelleher et al. 2007; Nadeau et al. 2007; Wagstaff and Begun 2007; Carnahan and Jensen-Seaman 2008; Ramm et al. 2008; Turner et al. 2008). Most of these studies have focused on a small (<10) number of genes, and evidence for an effect of mating system on patterns of molecular evolution is mixed. Several genes, notably the mating plug protein SEMG2 in primates (Jensen-Seaman and Li 2003; Kingan et al. 2003; Dorus et al. 2004; Carnahan and Jensen-Seaman 2008; Ramm et al. 2008; but see Hurle et al. 2007), and the seminal-vesicle derived protein Svs2 in rodents (Ramm et al. 2008), show evidence for positive selection in multiply mating lineages and/or reduced constraint in monandrous lineages, but this pattern is by no means shared by all seminal proteins (Ramm et al. 2008). On a larger scale, studies comparing reproductive protein evolution in clades of Drosophila with differing levels of multiple mating have suggested that higher remating rates drive the diversification of reproductive tract proteins in both males (Wagstaff and Begun 2007) and females (Kelleher et al. 2007). A recent study in Heliconius butterflies, by contrast, found the reverse pattern, with higher rates of substitution in a monandrous lineage compared to a polyandrous lineage (Walters 2009).

In this study, I provide the first attempt to associate large-scale patterns of reproductive protein evolution with mating system in primates. I find that testis-specific genes have a higher rate of nonsynonymous substitution in the chimpanzee lineage, consistent with a role for sperm competition in driving reproductive protein evolution. Given that this result only applies to a single species pair, studies in additional taxa will be required to confirm this finding. In addition, the effects of mating system on reproductive genes expressed in other tissues requires further investigation. Prostate-specific and uterus-specific genes show positive δdN in this study, although this trend is not significant in either case. Evidence from previous studies suggests that seminal proteins produced in the prostate may also be subject to sexual selection—Clark and Swanson (2005), for example, found evidence for positive selection on a number of prostate-expressed (although not necessarily prostate-specific) genes. Interestingly, I find that chimpanzee/human δdN for proteins identified as present in the seminal fluid by Clark and Swanson (2005) is significantly positive (proteins identified by mass spectrometry as present in the ejaculate; n= 76, mean δdN= 0.019, P= 0.033), again consistent with an effect of mating system on reproductive protein evolution. Moreover, a positive δdN for uterine genes may suggest a role for sexual conflict in addition to sperm competition in driving variation in rates of protein evolution.

There are substantially less data concerning the effects of female multiple mating on genome-wide mutation rates. A systematic effect of mating system on mutation rates could have consequences for a wide range of evolutionary problems, including the purging of deleterious mutations by sexual selection (Whitlock and Agrawal 2009) and the paradox of the lek (Bartosch-Harlid et al. 2003; Ellegren 2007). To my knowledge, two studies have attempted to address sex-averaged mutation rates, both using data from birds. Møller and Cuervo (2003) estimated lineage-specific mutation rates from the appearance of novel mini-satellite bands in paternity studies, and found higher mutation rates in lineages with a higher probability of extra-pair paternity. Interestingly, in their study, mutation rate did not correlate with relative testis mass, suggesting that the increase in mutation rate did not derive from higher levels of sperm production. Furthermore, Nadeau et al. (2007) used a comparative approach to address the relationship between mutation rates and mating system in birds. They were unable to find a correlation between dS and dichromatism, a proxy for the intensity of sexual selection, over 5 loci.

In this study, δdS did not vary consistently with mating system. In two out of three species pairs examined, δdS was not significantly different from zero. In the squirrel monkey/owl monkey pair, by contrast, δdS was significantly greater than zero, as predicted. It is worth noting that the squirrel monkey/owl monkey pair is the most divergent of the three species pairs considered here. This greater divergence may increase power to detect subtle differences in rates of synonymous substitution. Nonetheless, the large sample size for the chimp/human contrast should allow detection of small differences in dS.

A variety of life history features, such as generation time, may also contribute to variation in dS across the primate phylogeny (e.g., Nikolaev et al. 2007). For the species pairs considered in this study, however, differences in generation time are unlikely to have confounded inferences concerning the effects of mating system on dS. Squirrel monkeys and owl monkeys both reach sexual maturity at approximately 3 years of age (Boinski 1987; Dixson 1994), so differences in generation time are unlikely to explain the observed difference in dS between these two species. In the case of human and chimpanzees, humans have a longer generation time than do chimpanzees. A longer generation time for humans would be expected to mimic the effects of sexual selection—i.e., it should lead to a higher rate of substitution in chimps, but no such effect was detected. Ultimately, strong conclusions regarding the effects of life history variation on mutation rates—including variation in mating systems—will require multi-locus sequence data from a wide range of organisms with varied mating systems and generation times. Variation in mating systems in additional taxa, e.g., nonprimate mammals (e.g., McGraw and Young 2010) as well as insects (e.g., Markow 1996), will facilitate further comparative analyses.

Associate Editor: M. Hellberg


Thanks to two anonymous reviewers for insightful comments on this manuscript, and to James Walters for discussion of his work on Heliconius reproductive proteins. Howard Rundle, Risa Sargent, Jacqueline Stepanacz, Sijmen Schoustra, and Danna Gifford provided helpful discussion. I undertook this work as an NSERC post-doctoral fellow.