Introduction

Genes related to sex and reproduction are among the more rapidly evolving groups of genes (Singh and Kulathinal 2000; Swanson and Vacquier 2002a, b). This trend was first discovered in Drosophila, where it was found that proteins in the ovary and testes were more highly diverged between species compared to proteins found in the head (Civetta and Singh 1995) and, subsequently, that the genes encoding these proteins were also highly diverged (Civetta and Singh 1998). The rapid evolution of reproduction-related genes has since been reported in many other taxa, including Chlamydomonas (Ferris et al. 1997), diatoms (Armbrust and Galindo 2001), gastropods (Swanson and Vacquier 1995; Hellberg et al. 2000), and mammals (Wyckoff et al. 2000; Swanson et al. 2001a, 2003; Torgerson et al. 2002). For example, in mammals, it was shown that genes expressed in sperm cells evolve faster than genes expressed in a variety of other tissues (Torgerson et al. 2002) and that fertilization proteins in general are evolving rapidly in both males and females (Swanson et al. 2003).

The mechanism driving rapid evolutionary change in reproductive genes likely resides in Darwin’s theory of sexual selection, where individuals must compete for a limited number of mates. Strong selection for traits that give an individual an advantage over others can drive the rapid evolution of such reproductive traits, and may often result in positive selection (Singh and Kulathinal 2000; Swanson and Vacquier 2002a, b). In fact, positive selection has been implicated in driving the rapid adaptive evolution of many reproduction-related genes in mammals, including those involved in both male (Wyckoff et al. 2000; Torgerson et al. 2002; Swanson et al. 2003) and female (Swanson et al. 2001a, 2003, 2004) reproduction.

There are several factors that may influence the degree to which sexual selection can drive the rapid evolution of reproductive genes. One such factor may be a gene’s level of dispensability, as selective constraints under natural selection may restrict evolutionary change in order to preserve an essential function. However, there is some controversy as to whether essential genes evolve faster than nonessential genes. There is evidence on one side to suggest that essential genes evolve more slowly than nonessential genes in viruses (Bull et al. 2003), bacteria (Jordan et al. 2002), and yeast (Hirsh and Fraser 2001; Zhang and He 2005). In humans, oncogenes and tumor suppressor genes have a significant overlap with essential homologous genes in the mouse and were found to be under stronger selective constraints than non-disease-related genes (Thomas et al. 2003), whereas Smith and Eyre-Walker (2003) found that disease genes in general evolve faster under lower selective constraint than nondisease genes (which may contain lethal genes). Other studies have suggested that the essential nature of a gene does not influence the evolutionary rate as significantly as other factors do, such as gene duplication (Yang et al. 2003), directional selection (Hurst and Smith 1999), and rate of expression (Pál et al. 2003).

A second factor that may contribute to rapid evolutionary change under sexual selection could be the breadth of gene expression and degree of specialization. Coulthart and Singh (1988) found that genes expressed in both Drosophila testes and Drosophila accessory glands had lower rates of evolution than those expressed in only one tissue. Pál et al. (2001) found that highly expressed genes in yeast evolve more slowly than genes expressed at lower levels, and several studies have shown that genes expressed in a greater number of tissues tend to evolve more slowly than tissue-specific genes (Duret and Mouchiroud 2000; Zhang and Li 2003; Winter et al. 2004). It may be that increased functional constraint at a greater number of sites results in lowering the rate of evolution. It is then possible that if reproduction-related genes were to become more specialized as opposed to maintaining a more ubiquitous function, rapid evolution via sexual selection could be more easily enabled.

In the present study we compared the levels of divergence between human and mouse orthologous genes, as well as the actual number of genes that are essential to male and female fertility. Genes were classified as essential to fertility according to mouse knockout mutation phenotypes as reviewed by Matzuk and Lamb (2002). We found that genes with an essential function in male or female fertility evolve faster than lethal genes, except in the case when genes are essential to fertility in both sexes. We also found that a greater number of genes affect male sterility over female sterility and that fertility genes tend to specialize to one sex rather than both. Our data provide evidence that the rapid evolution of sex- and reproduction-related genes is facilitated through an increased specialization of function and that dispensability is not the major factor determining their evolutionary rate.

Materials and Methods

In an extensive review, Matzuk and Lamb (2002) generated a list of over 200 mouse knockout mutations in genes that affect fertility and classified them as to whether mutations affected male fertility, female fertility, or fertility in both sexes. They also summarized the reproductive phenotype and classified the knockout as resulting in either subfertility or complete sterility. In this study, we considered knockout genes that result in complete sterility as being essential fertility genes, and those that result in reduced fertility (or subfertility) as being nonessential fertility genes. We selected genes that fell into six categories from those of Matzuk and Lamb (2002): knockout genes causing (1) male sterility, (2) male subfertility, (3) female sterility, (4) female subfertility, (5) both male and female sterility, and (6) both male and female subfertility. From the initial list of 202 fertility genes, 33 were discarded because the knockout phenotypes were not consistent with those six defined categories, including knockouts producing a lethal (n = 18) or fertile (n = 4) phenotype, no phenotype listed (n = 6), an increase in fertility (n = 1), or phenotypes resulting from intercrossing (n = 2) or by an insertional mutation (n = 2). In addition to these, eight genes were removed from the dataset due to a different degree of fertility disruption between males and females or because of a varying phenotype within one sex; for example, if the knockout mutation resulted in sterility in the male and subfertility in the female or caused both sterility and subfertility in the female depending on the individual. The divergence of seven of the eight excluded fertility genes was not significantly different from that of the overall dataset in terms of both nonsynonymous and synonymous divergence (p = 0.53 and p = 0.20, respectively). We also retrieved a list of 1978 lethal knockout mouse mutations from the Mouse Knockout and Mutation Database at BioMedNet (http://www.bmn.com) and the alignments of 7642 human and mouse orthologous genes from Clark et al. (2003) to represent a genomic sample.

Homologous human genes were retrieved from HomoloGene at the National Center for Biotechnology Information (http://www. ncbi.nlm.nih.gov), using a pairwise reciprocal best hits criterion of percentage nucleotide sequence identity. Human and mouse orthologous genes were aligned using ClustalW version 1.83 (Thompson et al. 1994), and measures of divergence (nonsynonymous and synonymous nucleotide substitutions per site) were calculated in all datasets according to Nei and Gojobori (1986) using the program SNAP (Korber 2000), available at http://www.hiv.lanl.gov/content/hiv-db/SNAP/WEBSNAP/SNAP.html. Sequences were discarded for comparisons of divergence if the gene names were ambiguous (a total of six fertility genes) or if there were no orthologous human genes listed (a total of five fertility genes). These genes were not included in comparisons of divergence so as not to bias a comparison with lethal genes but were included in comparisons of fertility gene numbers. Of these 11 discarded fertility genes, we were able to find human orthologues for 9: 4 of the 6 genes with ambiguous names by searching on the alias name and all 5 genes with no HomoloGene entry by using syntenic regions. The nonsynonymous and synonymous divergences of the nine excluded fertility genes were not significantly different from those of the final dataset (p = 0.38 and p = 0.56, respectively). Our overall dataset included 161 fertility genes with measures of divergence for 150 orthologous genes affecting fertility and 803 orthologous genes affecting viability.

Mean divergence at both nonsynonymous (Ka) and synonymous (Ks) sites was compared using Student’s t-test assuming unequal variance. We also compared the ratios of nonsynonymous-to-synonymous nucleotide substitutions per site (Ka/Ks) as an estimate of the average selective constraints acting on a gene. When Ka/Ks < 1 it indicates that the gene is under selective constraint, whereas when Ka/Ks = 1 selective constraints are completely relaxed, and when Ka/Ks > 1 the gene is presumed to be under positive selection. Comparisons were made between the level of divergence and selective constraint between genes that are essential or nonessential to fertility, genes affecting fertility or viability, genes affecting male or female fertility, and genes affecting the fertility of one or both sexes. We also compared the number of genes affecting male or female fertility and the number of genes that have become specialized to fertility in one sex or both sexes to a binomial distribution.

Results

A complete list of the mouse knockout genes used in this study, along with their associated phenotype and level of divergence between orthologous human sequences, is available as supplementary material. Overall we found that genes important to fertility are more highly diverged at nonsynonymous sites than genes important to viability (mean Ka = 0.103 vs 0.064; p = 4.5 × 10−5). However, when a gene has an essential fertility function in both the male and the female, there is no significant difference in the average nonsynonymous mutation rate compared to that of lethal genes (mean Ka = 0.078 vs 0.064; p = 0.22). The average nonsynonymous and synonymous divergences for all classes of genes are summarized in Fig. 1. There is no significant difference in the nonsynonymous divergence of fertility genes in males versus females (Table 1) or between fertility genes affecting one sex compared to both (Table 2). We also found no significant difference in the synonymous mutation rates between fertility and viability genes (mean Ks = 0.548 vs 0.555; p = 0.68) and between different categories of fertility genes (Fig. 1).

Figure 1
figure 1

The average nonsynonymous (above; Ka) and synonymous (below; Ks) divergence between human and mouse orthologues for genes important to fertility in the male, the female, or both sexes. The white bars represent nonessential fertility genes where knockout mutations result in subfertility, the gray bars represent essential fertility genes where knockout mutations result in sterility, and the black bars represent essential viability genes where knockout mutations are lethal and a genomic sample of orthologous genes from Clark et al. (2003). Error bars represent twice the standard error.

Full size image
Table 1 A comparison of the average evolutionary rate of genes affecting male or female fertility
Full size table
Table 2 A comparison of the average evolutionary rate of genes affecting the fertility of one or both sexes
Full size table

We compared our data to those of Clark et al. (2003) to represent a genomic sample of 7642 human and mouse orthologous genes, as the authors have shown their data to be 97% consistent with 1527 entries in HomoloGene. Compared to the genomic sample, fertility genes have an overall significantly higher nonsynonymous nucleotide divergence (p < 0.001), however, this is not the case in lethal genes, where there is no significant difference between average values of Ka compared to the genomic sample (p = 0.82). On the other hand, synonymous nucleotide divergence is significantly higher in the genomic sample compared to both fertility (p < 0.001) and lethal (p < 0.001) genes.

We compared ratios of Ka/Ks to give an indication of the level of selective constraint acting on each category of fertility genes (Fig. 2). Selective constraints are significantly more relaxed on fertility genes relative to both lethal genes (mean Ka/Ks = 0.19 vs 0.13; p = 0.005) and the genomic sample (mean Ka/Ks = 0.19 vs 0.11; p < 0.001), but genes essential to fertility in both genders are on average under selective constraints similar to those for both lethal genes (mean Ka/Ks = 0.14 vs 0.13; p = 0.77) and the genomic sample (mean Ka/Ks = 0.14 vs 0.11; p = 0.07). We also find no significant difference in Ka/Ks between males and females for both essential and nonessential fertility genes (Table 1) or between fertility genes affecting one sex and both sexes (Table 2). Selective constraints appear to be reduced in lethal genes compared to the genomic sample (mean Ka/Ks = 0.13 vs 0.11; p = 0.006), however, this may reflect differences in the synonymous mutation rate, as measures of Ka were not significantly different between lethal genes and the genomic sample.

Figure 2
figure 2

The average ratio of nonsynonymous-to-synonymous nucleotide substitution rates (Ka/Ks) for human and mouse orthologous genes that are important to fertility in the male, the female, or both sexes. The white bars represent nonessential fertility genes where knockout mutations result in subfertility, the gray bars represent essential fertility genes where knockout mutations result in sterility, and the black bars represent essential viability genes where knockout mutations are lethal and a genomic sample of orthologous genes from Clark et al. (2003). Error bars represent twice the standard error.

Full size image

Of 161 knockout genes that affect mouse fertility, we found that 57 result in male sterility, and of these, at least 45 genes are associated with defects in spermatogenesis or in the mature sperm. However, female sterility was found to have no obvious bias in knockout phenotype as the males did, with 28 genes affecting anything from reproductive organs, oogenesis, estrous cycle, ovulation, implantation, and corpus luteum formation to parturition. We found a total of 31 genes that are essential to fertility in both genders, of which 8 appear to have a distinct phenotype between males and females.

The number of genes affecting male fertility is significantly higher than the number of genes affecting female fertility compared to a binomial distribution (p = 1.2 × 10−4; Table 3). However, this is only the case for essential fertility genes, and the number of nonessential fertility genes is not significantly different between the sexes. The number of genes that are important to fertility in one sex is also significantly greater than the number of genes that have an important function in both sexes (Table 4). We found that in both essential and nonessential fertility genes there were more than three times as many genes affecting one sex rather than both.

Table 3 A comparison of the numbers of genes affecting fertility in the male or female
Full size table
Table 4 A comparison of the number of genes affecting fertility in one or both sexes
Full size table

Discussion

Divergence of Essential and Nonessential Fertility Genes

In an evolutionary framework, genes essential to reproduction are just as important to an organism’s fitness as are genes essential to viability. Failure to reproduce will reduce an individual’s fitness to zero, as does the failure to survive to a reproductive age. Therefore high fitness costs may be associated with deleterious mutations that occur in genes affecting fertility. However, in general, genes essential to fertility are more diverged than lethal genes and a genomic sample, with the exception of genes that are essential to fertility in both genders. The lower divergence of genes essential to fertility in both sexes suggests that this class of fertility genes evolves more slowly because of increased fitness costs when both sexes are affected, leading to higher selective constraints. However, this is probably not the most important factor affecting the evolutionary rates of fertility genes for three reasons. First, there is no significant difference in the evolutionary rates of genes affecting the fertility of one or both sexes, whereas we may have expected a difference due to higher fitness costs when both sexes are affected. Second, we find that the nonsynonymous divergence of nonessential fertility genes is not significantly different from the divergence of essential fertility genes when compared in both the male and the female. Therefore genes that are essential to fertility in one sex are on average evolving just as fast as those that are nonessential. Third, genes essential and nonessential to either male or female fertility are significantly more diverged between human and mouse compared to lethal genes or a genomic sample, even though genes affecting the fertility in one sex may have a reduced marginal fitness due to fertility effects in their offspring of the opposite sex. We therefore suggest that the evolutionary rate of sex-specific fertility genes is not directly related to their degree of dispensability and that sex- and reproduction-related genes do not evolve rapidly simply due to a less essential nature.

Although the topic is controversial (see Introduction), there are several studies suggesting that the correlation between dispensability and rate of evolution is generally weak. Pál et al. (2003) performed a re-analysis of the yeast and worm data of Hirsh and Fraser (2001) and suggested a stronger correlation between gene expression level and dispensability, rather than with the evolutionary rate. Yang et al. (2003) also found no correlation between dispensability and rate of evolution when duplicate genes were removed from the data, suggesting that gene copy number is a more important factor affecting the evolutionary rate of genes. Krylov et al. (2003) investigated several factors affecting the rate of evolution and found that the propensity for gene loss and the number of protein interactions were more strongly correlated with dispensability than the rate of evolution. Hurst and Smith (1999) found a significant difference in the divergence of essential and nonessential genes in mice but noticed that nonessential genes contained a disproportionately high number of genes involved in the immune response. After correcting for this bias, they similarly found no significant difference between the evolutionary rates of essential and nonessential genes and, therefore, attributed their difference to forces of directional selection and coevolution in host–parasite interactions rather than the degree of dispensability.

Male vs Female Divergence

The rapid evolution of reproductive traits has until recently centered on male reproduction, perhaps due to the more commonly observed and elaborate morphological and behavioral reproductive traits in the male, where sexual selection may manifest more strongly (Bateman 1948). However, several studies have reported that reproductive genes in the female can also evolve at a high rate. For example, in abalone the egg protein VERL evolves rapidly under positive selection (Galindo et al. 2003), and in mammals several zona pellucida genes are also evolving rapidly under positive selection (Swanson et al. 2001a). Although direct comparisons have rarely been made between the divergence of male and that of female reproductive genes, a comparison has been made in Drosophila with respect to testes- and ovary-expressed genes. Using two-dimensional protein electrophoresis, Civetta and Singh (1995) showed that testes proteins are more diverged than ovary proteins, findings that have more recently been confirmed at the DNA sequence level (Jagadeeshan and Singh (2005). Contrary to this, we find that genes important to fertility in males and females are evolving at a similar rate, suggesting a different evolutionary pattern in mammals. However, this discrepancy is more likely due to the nature of the genes in each study, as our dataset contained genes that affect fertility rather than genes that are expressed in the testes or ovaries. Our study included genes where the knockout phenotype is known to have an adverse affect on fertility and, therefore, may not have included genes that enhance reproductive ability, for example, in the way that rapidly evolving accessory gland proteins do in Drosophila (Wolfner 2002). It is possible that genes enhancing reproductive potential are more abundant in the male and are more diverged than those in the female.

Number of Genes Affecting Male and Female Fertility

The difference between male and female fertility genes does not lie in their level of divergence but, rather, in the number of genes responsible. Relative to females, a greater number of genes affecting male fertility in mammals is consistent with several studies in Drosophila that have found the same trend. A classic study by Lindsley and Lifschytz (1972) using isofemale lines demonstrated higher rates of male sterility in homozygous lines rather than female sterility or sterility in both sexes. More recently, studies have shown that there are a greater number of hybrid sterility factors in males relative to females (Hollacher and Wu 1996; True et al. 1996) and that the rate of accumulation of hybrid sterility factors is much higher in the male than the female (Tao and Hartl 2003). Moreover, male hybrid sterility appears to result from the effects of multiple genes (Sawamura et al. 2004a, b), in contrast to evidence that single major genes may be responsible for female sterility (Sawamura et al. 2004b).

One reason for there being a greater number of genes affecting male fertility is that functional divergence is being driven by stronger sexual selection in the male than in the female. This is supported by the finding of a relatively high number of genes expressed specifically in the testes, similar to the number of genes expressed specifically in the brain of Drosophila (Andrews et al. 2000). However, there are several examples of genes involved in female reproduction that are subject to positive selection (Swanson et al. 2001a, 2003, 2004; Galindo et al. 2003), suggesting that sexual selection may also be an important driver of female reproductive evolution. A second possibility is that spermatogenesis may require a greater number of more specialized genes. While the initial stages of gametogenesis appear to be very similar between males and females, and could potentially utilize the same genes, perhaps the production of a fully motile gamete requires a greater number of specialized genes. In support of the second idea, we find that the majority of genes affecting gametogenesis in both sexes are a result of a disruption in meiosis. However, following meiosis, a complex sequence of events must occur during postmeiotic differentiation of male gametes allowing them to become mobile and fertilize the egg. In Drosophila it was found that 81% of male sterile mutations affected spermatogenesis, and of these, 68% affected spermatid differentiation in the later stages of spermatogenesis (Wakimoto et al. 2004). Their findings appear to be very similar to that in mammals, as 79% (45/57) of male sterile knockout mutations affect spermatogenesis in the mouse. Our data support the hypothesis that a greater number of genes affect male fertility because of an increased requirement for specialized genes in the later stages of spermatogenesis. However, it cannot distinguish whether this is due to stronger selection for functional divergence or whether there is a greater genetic demand to producing a mobile gamete.

In turn, if there are a greater number of genes involved in spermatogenesis, this process becomes much more vulnerable to genetic perturbation. Although studies in Drosophila have demonstrated the sensitivity of male fertility, they have also shown that genes involved in male reproduction evolve rapidly. Therefore the sensitivity of male fertility may not be the result of an increased sensitivity of individual genes to mutation but, rather, of the increased number of targets available (Singh and Kulathinal 2005). Our data suggest that this may also be the case in mammals, as genes essential to male fertility are evolving faster and with lower selective constraints than lethal genes and a genomic sample, and because there are a disproportionately high number of genes that are specialized to male fertility.

Specialization of Fertility Genes

A high proportion of fertility genes appears to have become specialized to one sex rather than having an important function in both sexes. Evidence from previous studies has suggested that a higher degree of specialization can lead to faster rates of evolution. Duret and Mouchiroud (2000) compared the expression profiles of homologous human and mouse genes and found a negative correlation between nonsynonymous divergence and the number of tissues in which a gene is expressed. Similarly, Zhang and Li (2003) found that housekeeping genes that are expressed in multiple tissues evolve more slowly and have lower values of Ka/Ks than tissue-specific genes, and Winter et al. (2004) report a positive correlation between rates of evolution and tissue specificity. When a gene is expressed in more than one tissue it could be performing either a single ubiquitous function or various multiple functions, which in both cases could increase selective constraint and decrease evolutionary rates. For example, if a gene has a ubiquitous function and is expressed in a large number of tissues, it has a greater chance of interacting with more proteins, and genes with a greater number of protein interactions tend to have lower rates of evolution (Fraser et al. 2002). Similarly, if a gene performs multiple functions, it may encounter a greater diversity of interacting proteins, and if a larger proportion of the gene is involved in these functions, then the evolutionary rate will also be reduced (Fraser et al. 2002).

Studies on gene duplication have also indicated that specialization can result in faster evolution. When a gene becomes duplicated, selective constraints may become relaxed on one or both of the copies, as only one of the copies may be required to maintain ancestral function (Force et al. 1999). For example, in Drosophila, two testes-specific α4 proteasome subunits are under lower selective constraints relative to the more ubiquitous and somatic functioning α4 subunit (Torgerson and Singh 2004). Comparative studies have also shown that the evolutionary rate of a gene often increases following gene duplication (Van de Peer et al. 2001; Zhang et al. 2003) and that duplicated genes often develop tissue-specific patterns of expression (Lynch and Force 2000). Therefore duplicated genes that develop narrow tissue expression patterns could be evolving faster due to lower selective constraints resulting from fewer functional constraints.

However, we cannot assume that because a gene is essential to both male and female fertility, it is performing more than one specialized function. Of 31 genes essential to both male and female fertility, there are only 8 that appear to have a different knockout phenotype in males and females. The majority of genes affecting both sexes result in hypogonadism or meiotic defects in both males and females, which may suggest a similar function in both sexes. However, because we find that the majority of fertility genes have become specialized to one sex, then even if a gene performs the same function in each sex, it has a higher probability of interacting with a greater number of sex-specific proteins. Proteins that interact with one another generally evolve at similar rates due to coevolutionary processes, and may evolve more slowly due to functional constraints at a higher proportion of sites (Fraser et al. 2002). Therefore even if a gene has the same function in both males and females, the possible interaction with a greater number of sex-specific proteins may act to lower the evolutionary rate of fertility genes affecting both sexes. Overall our data support the hypothesis that the specialization of fertility genes to one sex facilitates a faster rate of evolution under pressures from sexual selection.

Conclusions

Genes related to sex and reproduction are in many cases under unique evolutionary pressure from sexual selection acting on reproductive traits. Evolution can occur through natural selection when not all individuals in a population or species survive, leaving the ones that are better adapted to their environment. However, evolution can also occur through sexual selection when not all individuals in a population reproduce at an equal rate. A strong competition between individuals of the same sex for reproductive opportunities can be a strong driving force behind evolutionary change.

In many Drosophila species the female will mate multiple times, and males have evolved a complex mixture of accessory gland proteins to help with sperm competitive ability (Wolfner 2002). The genes encoding these accessory gland proteins are among the most rapidly evolving genes in Drosophila (Swanson et al. 2001b), which may result from strong sexual selection acting on these genes. Signatures of positive selection have been found in a variety of reproductive proteins in mammals, which is likely to be driven by forces of sexual selection. For example, in a comparison of humans and Old World monkeys several male reproductive genes appear to be under positive selection (Wyckoff et al. 2000). In particular, fertilization proteins in mammals (Swanson and Vacquier 2003), including both sperm proteins (Torgerson et al. 2002) and egg proteins (Swanson et al. 2001a), are often found to be under positive selection despite having a presumably important reproductive function.

We have shown that the evolutionary rate of fertility genes is not merely a function of their degree of dispensability, supporting the theory that selection is driving the rapid evolution of reproductive traits. Moreover, we suggest that the specialization of fertility genes to male or female functions can facilitate rapid evolution under selection, particularly for genes that have become specialized to the male. Rapid evolution in fertility genes does not appear to be a consequence of a less essential function, suggesting that sexual selection is a more important factor increasing the evolutionary rate of sex- and reproduction-related genes.