Although sexual antagonism may have played a role in forming some sex chromosome systems, there appears to be little empirical or theoretical justification in assuming that it is the driving force in all cases of sex chromosome evolution. In many species, sex chromosomes have diverged in size and shape through the accumulation of mutations in regions of suppressed recombination. It is commonly assumed that recombination is suppressed in sex chromosomes due to selection to resolve sexually antagonistic pleiotropy. However, the requirement for a sex chromosome-specific mechanism for suppressing recombination is questionable, since more general models of recombination suppression on autosomes also appear to be applicable to sex chromosomes. Direct tests of the predictions of the sexual antagonism hypothesis offer only limited support in specific sex chromosome systems and circumstantial evidence remains open to interpretation.
Sex chromosomes with large non-recombining regions have evolved independently in organisms belonging to a wide range of plant and animal groups, including bryophytes, flowering plants, vertebrates and invertebrates 1. Once recombination between a pair of homologous chromosomes has ceased, they are free to follow separate evolutionary trajectories, independently accumulating deletions, insertions, duplications and rearrangements. These changes can alter the relative length, gene content and chromatin structure of the sex chromosomes. Hence, suppression of homologous recombination plays a key role in the divergence in shape (heteromorphy) that characterises many sex chromosome pairs 2, 3.
The non-recombining regions of sex chromosomes appear to be produced by a range of different processes, acting singly or in concert. In mammals, for example, recombination between the sex chromosomes appears to have been suppressed through a combination of inversions, translocations and suppression of crossing over by local modifiers 4–6. In mammals, birds and some flowering plants, genetic divergence between the sex chromosomes decreases with distance from the sex-determination locus, with discontinuities producing evolutionary strata 5, 7, 8. Although strata may be caused by variation in the rate of gene conversion along the sex chromosomes 4, they are usually interpreted as evidence that recombination has been restricted in several stages 5, 6. This suggests that non-recombining regions have gradually expanded through cumulative recombination-suppressing events.
In contrast to sex chromosomes, the genome's remaining chromosomes (known as autosomes) are rarely heteromorphic. This apparent association between the function of sex determination and suppression of recombination has led to the suggestion that restricted recombination in sex chromosomes evolves due to selection to resolve sexually antagonistic genetic interactions 2, 9–11. This sexual antagonism hypothesis has assumed the status of conventional wisdom, to the extent that, in many recent publications, sexual antagonism is the only hypothesis suggested to account for restricted recombination in sex chromosomes 1, 6, 12–22. However, empirical support for the sexual antagonism hypothesis is equivocal 1. The aim of this review is to evaluate the evidence for sexual antagonism as a cause of restricted recombination in sex chromosomes, to suggest alternative hypotheses and to highlight areas that require further research.
This review begins with a description of the sexual antagonism hypothesis for restricted recombination in sex chromosomes. This is followed by a discussion of more general hypotheses for suppression of recombination, originally proposed to explain the spread of inversions on autosomes. These can also be applied to sex chromosomes, providing alternative hypotheses to sexual antagonism. Empirical evidence for sexual antagonism is then reviewed, beginning with direct tests of the sexual antagonism hypothesis' predictions. Circumstantial evidence for and against the role of sexual antagonism in causing the initial establishment and subsequent expansion of non-recombining regions in sex chromosomes is also discussed.
The sexual antagonism hypothesis
The sexual antagonism hypothesis for restricted recombination in sex chromosomes was originally proposed by Sir Ronald Fisher in 1931 10 to explain the overrepresentation of genes for male ornamentation in close linkage to the sex-determination locus of the guppy Poecilia reticulata23. In this and many other species, males and females differ in morphology and behaviour and so natural selection upon traits associated with survival and reproduction is expected to favour different optima in the two sexes. However, responses to selection are often correlated between males and females, with the result that opposing selection pressures reduce the average fitness of both sexes 24, 25. One cause of these deleterious intersexual genetic correlations is sexually antagonistic pleiotropy 26–28. This occurs when an allele that is beneficial when expressed in one sex is detrimental when expressed in the other. For example, in P. reticulata conspicuous colour patterns increase the risk of predation in both sexes while apparently increasing the reproductive success of males only 29, 30. Hence alleles producing conspicuous colour patterns produce a net benefit to males and a net detriment to females; they are sexually antagonistic.
In theory, an autosomal sexually antagonistic allele can increase in frequency only if its benefit to one sex is greater than its detriment to the other 9, 31. However, a sex-linked sexually antagonistic allele can increase even when its detriment in one sex is several times greater than its benefit in the other 11. This applies even when linkage between the sexually antagonistic locus and sex-determination locus is incomplete, generating selection for translocation of sexually antagonistic genes from autosomes to sex chromosomes 9, 11. However, as long as linkage remains incomplete, recombinant individuals are produced bearing sexually antagonistic alleles that are deleterious to members of their sex. This generates positive selection upon mutations that suppress recombination between sexually antagonistic loci and the sex-determination locus. Linkage is thereby increased between the sex locus and more distant loci, leading to recruitment of additional sexually antagonistic alleles further from the sex-determination locus and hence to selection for further suppression of recombination. This produces a ‘genetic chain reaction’ 11 that expands the region of restricted recombination along the length of the sex chromosomes (Fig. 1).
These models indicate that sexually antagonistic alleles are more likely to spread if they are tightly linked to a sex-determination locus. These models also indicate that a mutation suppressing recombination between a sex-determination locus and a second locus is more likely to spread if the second locus contains sexually antagonistic alleles. However, to infer from this that the non-recombining regions of sex chromosomes spread due to sexual antagonism assumes that suppression of recombination on sex chromosomes always results from selection acting directly upon the sex-determination locus to reduce its rate of recombination with other loci. If the sex-determination locus is allowed to play a passive role in the initial spread of the non-recombining region then alternative hypotheses are possible.
Alternative hypotheses for the evolution of non-recombining regions
Three alternative mechanisms are proposed to explain the spread of non-recombining regions on sex chromosomes. The first is that mutations or rearrangements causing suppression of recombination on sex chromosomes spread by genetic drift 32. The second is that suppression of recombination is selected because it prevents homozygosity of deleterious recessive genes, particularly in populations with inbreeding 33. A third hypothesis is that non-recombining regions spread due to selection against recombination within a supergene formed by two loosely linked genes with a positive epistatic effect upon fitness.
The capacity of these mechanisms to generate non-recombining regions in autosomes has been explored in the large body of theoretical research modelling the spread and maintenance of polymorphic chromosomal rearrangements such as inversions, centric fusions and reciprocal translocations 34. These chromosomal rearrangements tend to reduce recombination when heterozygous due to reduced pairing and crossing over and to selection against recombinant gametes that contain meiotic abnormalities 35. In most such theoretical studies, it is the recombination-suppressing property of chromosomal rearrangements that is modelled as the character under selection 34. The models can therefore be extended to encompass the spread and maintenance of any non-recombining region, whether it originated through a rearrangement or not. Furthermore, there is evidence that rearrangements, particularly inversions, are often responsible for initiating suppression of recombination between sex chromosomes. This appears to be the case in 10 of 34 neo-tropical fish species 36, in spiny eels 37, in Dysdercus bugs 38 and in humans 6.
The hypothesis that chromosomal rearrangements invade through genetic drift in small populations 39 receives some empirical support from studies of inversions in water beetles 40 and blackflies 41, which show no deviation from Hardy-Weinberg equilibrium, suggesting that they are selectively neutral. Although it is theoretically plausible for a chromosomal rearrangement to become fixed in a small population by drift, it is more challenging to model the spread of the rearrangement beyond a single, isolated population 42. Attempts to do so have invoked metapopulation dynamics in which populations carrying the rearrangement colonise habitat patches left vacant by local extinctions 42 or colonise new areas of habitat generated by climate change 43.
The hypothesis that chromosomal rearrangements spread through selection to prevent homozygosity of deleterious recessive genes at multiple loci was proposed by Charlesworth and Wall 33. Their models demonstrate that, in populations with moderate levels of inbreeding, selection to maintain heterozygosity at two loci can favour the spread of neo-sex chromosomes generated by centric fusions or reciprocal translocations.
The third (supergene) hypothesis was originally suggested by Dobzhansky 44, 45 to explain the spread of inversions in Drosophila. In this case, alleles at two loci contained within a non-recombining region have a joint effect upon fitness that is greater than the sum of their effects when either allele is present in the absence of the other (the two alleles are said to form a co-adapted gene complex or ‘supergene’). A classic example of suppression of recombination between autosomal loci containing co-adapted alleles is supergene control of Batesian mimicry in the butterfly Papilio dardanus46. In this case, fitness epistasis occurs because specific combinations of linked alleles allow different individuals to accurately mimic different toxic model species. Fitness epistasis, in the extreme guise of balanced lethal alleles, also appears to underlie the formation of large, non-recombining regions in the heteromorphic autosomes of Triturus newts (discussed in more detail below).
Dobzhansky's 45 co-adaptation hypothesis assumes the special case of overdominance in which the inversion heterozygote is fitter than either homozygote, allowing the maintenance of a non-recombining region. The superior fitness of the inversion heterozygote advantage may be conferred by additive interactions between genes or by epistasis 39, 47. As noted by Charlesworth 48, the sexual antagonism hypothesis for suppression of recombination on sex chromosomes is actually a special case of supergene evolution. In this case, fitness epistasis occurs between alleles at the sex-determination locus and at the sexually antagonistic locus because the sexually antagonistic allele increases fitness when present in one sex and decreases fitness when present in the other. However, the models described above indicate that inversions and other suppressors of recombination can even spread in the absence of sexual antagonism.
Although it need not be involved in the initial spread of a non-recombining region, the presence of a sex-determination locus will have important consequences for the subsequent evolution of the region. Non-recombining regions that occur on non-sex chromosomes are generally small and short-lived in comparison to those of sex chromosomes 49–51. Presumably such autosomal non-recombining regions tend to be destroyed by the fixation of one karyotype and the extinction of the other, since recombination is only suppressed when the region is heterozygous (Fig. 2A).
Heterozygosity at the sex-determination locus is also required for suppression of recombination between sex chromosomes even in the absence of heteromorphy, otherwise recombination would also be suppressed in the homogametic sex. However, a sex-linked non-recombining region is prevented from spreading to fixation because frequency-dependent selection operates against the most common sex 52 (Fig. 2B). This process is equivalent to the protected polymorphism observed due to frequency-dependent selection upon certain inversions and supergenes on non-sex chromosomes 53. However, sex-linked non-recombining regions are exceptional in that the mechanism of sex determination, which is the cause of balancing selection upon sex-linked loci, tends to persist for long periods of evolutionary time. In this, it differs from other causes of balancing selection upon inversions, such as environmental gradients 54 and host-parasite co-evolution 55, which are relatively ephemeral selective agents, rarely persisting beyond the lifetime of a species. In contrast, a sex-linked inversion is held in a balanced polymorphic state until the function of sex determination shifts to another locus.
The protracted period of time during which sex-linked regions can remain non-recombining may explain why such regions tend to spread. Any mutation that suppresses recombination over a region including all or part of the original non-recombining region may, if it avoids extinction, increase in frequency until it becomes fixed relative to the sex-determining allele. However, it too will then become subject to balancing selection exerted by the sex ratio, holding it in a polymorphic state. In this way, overlapping regions of suppressed recombination can accumulate on the sex chromosomes, each increasing the size of the non-recombining region through an evolutionary ratchet mechanism (Fig. 2C). At this stage, sexual antagonism may play a role, since epistatic interactions between the sex-determination locus and other, loosely linked loci may select for expansion of the non-recombining region. However, epistatic interactions between any locus within the non-recombining region and loci loosely linked to it will have the same effect. Such interactions may result from non-sexually antagonistic supergene effects or from selection for heterozygosity in an inbreeding population. Sexual antagonism may be sufficient to produce and expand the non-recombining region but it does not appear to be necessary.
Tests of the sexual antagonism hypothesis
The sexual antagonism hypothesis offers the testable prediction that sexually antagonistic genes should accumulate in the recombining (pseudoautosomal) region (PAR) of the sex chromosomes, prior to suppression of recombination 31. In particular, recessive, sexually antagonistic genes that benefit males should be more likely to accumulate on the X chromosome because they are expressed in the hemizygous sex 31, 56. In a series of experiments on the fruit fly Drosophila melanogaster, the majority of the genome was artificially constrained to be inherited solely though males 24, 28, 57, 58. Within lines with male-limited inheritance, male fitness increased, while that of females decreased. Furthermore, the majority of the sexually antagonistic genetic variations revealed in these experiments occurs in the X chromosome 57. However, because the sexually antagonistic loci detected in these experiments occur within the region of the X chromosome that does not recombine with the Y, they may have accumulated after recombination had already been suppressed. These results therefore provide only weak support for a causal link between sexual antagonism and restriction of recombination in sex chromosomes.
Evidence from selection experiments on the dioecious plant Silene latifolia is also equivocal. Analysis of quantitative trait loci (QTLs) for flower size indicates that the PAR of the sex chromosomes of Silene latifolia plays a disproportionate role in the control of variation in sexually dimorphic traits, including sexually antagonistic traits 25, 59. The position of these genes in the PAR shows that they accumulated prior to suppression of recombination. However, they are expressed only in one sex and so are not functionally sexually antagonistic.
In the guppy P. reticulata, some sexually antagonistic genes for conspicuous colour patterns occur in the non-recombining region of the Y chromosome, while others are only partially sex linked 60, indicating that they occur on the PAR of the sex chromosomes. Furthermore, some of these genes recombine between X and Y in small, low-predation populations but are exclusively Y-linked in large, high-predation populations where selection by predators against conspicuous colour patterns is greater 61. These findings are taken to imply that differences in Y-linkage of male colour pattern traits between guppy populations are caused by sexually antagonistic selection. However, the Y-linked colour pattern genes of guppies are frequently deleterious or lethal when homozygous, even though YY individuals otherwise show normal fertility and viability 62, 63. Suppression of recombination between conspicuous colour pattern genes on the pseudoautosomal and non-recombining regions of the Y chromosome may therefore occur due to selection to resolve non-sexual fitness epistasis rather than through sexual antagonism. Furthermore, the colour pattern genes associated with trait variation are not normally expressed in females. Selection for increased Y-linkage in large populations of P. reticulata61 is therefore difficult to explain by selection to resolve sexual antagonism at these pseudoautosomal loci because sexual antagonism has already been resolved through sex-limited expression.
A second line of research aims to test the prediction that sexual antagonism results in the accumulation on the sex chromosomes of genes with sex-limited expression. Natural selection can resolve intralocus sexual antagonism by linking the sexually antagonistic locus to a sex chromosome (sex-linked inheritance) or by down-regulating expression of the sexually antagonistic locus in the appropriate sex (sex-limited expression) 64. It follows that, with the exception of genes whose tissue specificity limits them to one sex (e.g. testis-specific genes), sex-limited genes would be sexually antagonistic if expressed in both sexes 65. Furthermore, it is suggested that sex-limited expression evolves more slowly than sex linkage 59, 64 because sex-limited expression is a ‘complex adaptation’ 64 requiring the evolution of specific regulatory sequences. This reasoning leads to the testable prediction that sex chromosomes should contain high densities of sex-limited genes in comparison to autosomes.
However, empirical evidence for the accumulation of genes with sex differences in expression on sex chromosomes is also equivocal. Profiles of genome expression indicate enrichment of female-limited genes on the X chromosome of D. melanogaster66, Drosophila simulans67 and the mouse 68 and of male-limited genes on the Z chromosome of the chicken 65. However, genes with male-biased expression are under-represented on the X chromosomes of Drosophila and Caenorhabditis elegans66, 67, 69 and on the Z chromosome of the chicken 65. Furthermore, the mosquito Anopheles gambiae shows no evidence that male- or female-limited genes are enriched on the sex chromosomes 70. Overall, the mixed results of these genomic surveys do not offer consistent support for the sexual antagonism hypothesis.
The assumption that sex linkage evolves before sex-limitation is also questionable. While there is evidence for rapid evolution of suppressed recombination in D. melanogaster71–73, comparisons of gene regulation between natural D. melanogaster strains indicate that sex-biased expression can also evolve very rapidly 74. Furthermore, transitions from bisexual to sex-limited expression need not involve complex adaptations. For example, androgen regulation of the sex-limited protein (Slp) gene of mice evolved simply through the insertion of an endogenous provirus upstream of the gene 75. Finally, other factors such as dominance, dosage compensation and X-inactivation are also predicted to influence the tendency of sex-limited genes to accumulate on sex chromosomes 76, 77.
Initial establishment of the non-recombining region
It is proposed that the evolution of dioecy from hermaphroditism requires the suppression of recombination between segregating genes for male sterility and female sterility 78. Under selection for dioecy, loci with alleles for male and female sterility are clearly subject to sexually antagonistic pleiotropy since, if both occur in the same individual, sterility results, and if neither is present, hermaphrodites are produced. This generates selection for suppression of recombination between loci containing male-sterility and female-sterility alleles, producing a non-recombining region with a sex-determination function 79.
In flowering plants, dioecy appears to have evolved from hermaphroditism on many separate occasions via gynodioecy or, more rarely, via androdioecy 4. The existence of these transitional states implies that alleles causing male sterility and female sterility occur at separate loci 78. This scenario also derives empirical support from the dioecious flowering plants Silene latifolia, Fragaria virginiana and Sagittaria latifolia19, 80, 81. These three plants possess heteromorphic sex chromosomes in which the non-recombining region of the Y chromosome contains separate factors for male fertility and female sterility.
In contrast, evidence to support this scenario for the initiation of sex chromosome formation is absent in animals. Although heteromorphic sex chromosomes are common in animals, separate loci containing male- and female-sterility alleles have never been detected. Furthermore, although dioecy is widespread in animals, the putative transitional stages of gynodioecy and androdioecy are rare. Gynodioecy occurs in cnidarians 82–84 but is unknown in triploblastic animals. Androdioecy is slightly more common, occurring in branchiopod crustaceans 85, rhabditid nematodes 86 and a single vertebrate species, the mangrove killifish 87, but these androdioecious species appear to have evolved from dioecious ancestors rather than from hermaphrodites 88. Androdioecy also occurs in barnacles where it is unclear whether it evolved from hermaphroditism or from dioecy 89.
In fact, few of the genetic sex-determination systems studied in dioecious animals appear to have evolved directly from hermaphroditism. In most cases a transition to genetic sex determination from environmental sex determination (ESD) or a switch from a pre-existing system of genetic sex determination appears to have occurred 90, 91. These transitions in sex determination do not appear to require the evolution of new male- and female-sterility loci. A likely explanation for this is that genetic control of sexual development in animals is regulated globally by a single gene of the doublesex-mab-3 (DM) family 92. This highly conserved system indicates an ancient homology of animal sex determination, predating the divergence of nematodes and arthropods 93. Subsequently, the various groups of animals have evolved diverse mechanisms for triggering sex determination through the addition of regulatory genes upstream of the DM regulator. These upstream regulatory genes are highly diverse, differing between closely related animal species and even within species. Whenever a new upstream regulatory gene is added to the apex of the cascade, the chromosome pair containing this new apical gene becomes the new sex chromosomes. Many switches of sex chromosome pair are therefore likely to have occurred between the initial transition from hermaphroditism to dioecy and the establishment of the current sex chromosomes of a given animal species.
Because its target is a single, global regulatory gene, there is no reason in theory why the apical sex-determination factor of an animal should not be a single gene itself. Such appears to be the case in therian mammals where the dominant Y-linked sex-determination gene Sry operates by regulating expression of the autosomal gene Sox 9 94 and hence controls the entire sex-determination cascade. The absence of Sry from non-mammalian vertebrates and from egg-laying mammals 95 indicates that this mechanism of sex determination, and hence the therian Y chromosome, was acquired relatively recently. A single, Y-linked dominant gene DMY also determines sex in the medaka fish Oryzias latipes96 but is absent from closely related species 97, indicating recent acquisition of a novel sex-determination system. In birds, the Z-linked gene DMRT1 forms a strong candidate for dosage-dependent single-gene sex determination 98, although the possibility that DMRT1 is regulated by upstream factors cannot be excluded.
The fact that a single, recently acquired sex-determination gene such as Sry can be associated with heteromorphic sex chromosomes 99 indicates that multiple, segregating sex-determination loci are not necessary for suppression of recombination to be initiated. Nonetheless, interactions with a sexually antagonistic locus might still be involved in the establishment of a new apical gene for sex determination 21. In such cases, selection for tighter linkage between the new sex-determination gene and the sexually antagonistic locus could select for suppression of recombination. However, such a relationship between sexual antagonism and switches in sex-determination system cannot be assumed because other factors may also facilitate the spread of a sex-determination usurper. For example, a new sex-determination allele may invade a dioecious population under positive selection if it counteracts a pre-existing sex ratio bias in the population 100, 101. This appears to have occurred in the wrinkled frog Rana rugosa, in which evidence from population genetics and experimental crosses suggests that sex ratio bias resulting from the hybridisation of two male-heterogametic populations drove the invasion of a new, female-heterogametic sex-determination locus 102, 103. Similarly, in populations of the female-heterogametic isopod crustacean Armadillidium vulgare, sex ratio bias caused by the bacterium Wolbachia can drive the W chromosome to extinction 104, 105. In some populations, this has selected for the spread of a dominant gene for maleness, correcting the sex ratio and rendering the populations male-heterogametic 106.
Expansion of the non-recombining region
The sexual antagonism hypothesis proposes that restriction of recombination spreads deterministically along the sex chromosomes through the recruitment of sexually antagonistic loci that are progressively distant from the sex-determination locus 11. As previously stated, this mechanism is exclusive to sex chromosomes. However, expansion of non-recombining regions is occasionally observed in autosomes and such cases can demonstrate properties that are remarkably similar to those of sex chromosomes. In the newts Triturus cristatus and Triturus marmoratus, chromosome 1 is heteromorphic in all individuals of both sexes 107. Heteromorphism is maintained by overdominance, with balanced lethal alleles at two loci causing the death of all embryos that are homozygous for either heterosome 107, 108. Differences in the degree of heteromorphy between the two species and the occurrence of evolutionary strata 107, 108 suggest that the non-recombining region has expanded beyond the region containing the two lethal alleles. In this case, expansion of the non-recombining region cannot be attributed to sexual antagonism because the heterosomes occur in all individuals of both sexes. The hermaphroditic trematode Megalodiscus temperatus also possesses heteromorphic autosomes 109, demonstrating that chromosomal heteromorphy can occur even in animals without separate sexes.
Expanded regions of suppressed recombination also occur on the mating type chromosomes of some protists and fungi. Fungi differ from animals and plants in that they are isogamous and remain haploid for the majority of their life cycles. However, in many cases, each haploid individual has a mating type, genetically determined by the mating type locus, and individuals must be of opposite mating types to mate successfully. In fungi such as Neurospora tetrasperma, Microbotryum violaceum and Cryptococcus neoformans, each haploid genome contains a single copy of the mating type locus and meiotic recombination within the mating type locus is suppressed 110–112. In these fungal mating type chromosomes, recombination is suppressed over a region extending far beyond the mating type genes. Although they do not show the levels of gross heteromorphy observed in some animal and plant sex chromosomes, the occurrence of strata 111, 112 indicates that expansion of the non-recombining region in these fungal chromosomes has occurred in several stages. In the absence of distinct sexes, expansion of the non-recombining regions of fungal mating type chromosomes cannot be attributed to sexual antagonism as such, although, in theory, antagonistic processes may occur due to differences in fitness optima between fungal mating types. However, developmental, morphological and ecological differences between mating types in these fungi appear minor when compared with differences between sexes in dioecious animals and plants. The influence of antagonistic processes upon the suppression of recombination between mating type chromosomes is therefore predicted to be correspondingly minor.
Sexual antagonism provides a mechanism for the evolution of restricted recombination on sex chromosomes that is theoretically plausible. However, hypotheses based on sexual antagonism can explain restriction of recombination only in sex chromosomes. Alternative hypotheses, originally developed to explain the spread of inversions, are sufficiently general to explain restriction of recombination on sex chromosomes and autosomes.
Expansion and differentiation of the non-recombining region to produce heteromorphy is a cumulative process that requires only that non-recombining regions are maintained for extended periods of time. This is demonstrated by evidence of expansion and heteromorphy in non-recombining regions of autosomes in newts, flatworms and fungi. In sex chromosomes, selection against the more common sex maintains heterozygosity at the sex-determination locus, and hence maintains sex-linked non-recombining regions for extended time periods, irrespective of sexual antagonism. The higher incidence of heteromorphy in sex chromosomes than in autosomes does not therefore require a mechanism for suppression of recombination that is specific to sex chromosomes.
Selection experiments and genomic surveys of loci with sex-limited expression indicate that sexual antagonism is a significant evolutionary force and that sexually antagonistic genes tend to accumulate on the X chromosome. However, support for the hypothesis that sexual antagonism drives restriction of recombination in sex chromosomes is weak. Stronger tests of this hypothesis should demonstrate that the accumulation of sexually antagonistic genes occurs in regions of the sex chromosomes that have not yet ceased recombining. Future tests of the sexual antagonism hypothesis should also involve loci that are expressed in both sexes. Sex-limited genes are not a good proxy for sexually antagonistic genes because it cannot be ascertained whether sex limitation evolved before or after linkage to the sex-determination locus.
Thanks to Katy Peat, Andrew Moore, Deborah Charlesworth and three anonymous reviewers for helpful comments on this manuscript.