Genomic Signatures of Sexual Conflict

Abstract

Sexual conflict is a specific class of intergenomic conflict that describes the reciprocal sex-specific fitness costs generated by antagonistic reproductive interactions. The potential for sexual conflict is an inherent property of having a shared genome between the sexes and, therefore, is an extreme form of an environment-dependent fitness effect. In this way, many of the predictions from environment-dependent selection can be used to formulate expected patterns of genome evolution under sexual conflict. However, the pleiotropic and transmission constraints inherent to having alleles move across sex-specific backgrounds from generation to generation further modulate the anticipated signatures of selection. We outline methods for detecting candidate sexual conflict loci both across and within populations. Additionally, we consider the ability of genome scans to identify sexually antagonistic loci by modeling allele frequency changes within males and females due to a single generation of selection. In particular, we highlight the need to integrate genotype, phenotype, and functional information to truly distinguish sexual conflict from other forms of sexual differentiation.

genome evolution, interlocus sexual conflict, intersexual genetic covariance, intralocus sexual conflict, linkage disequilibrium, pleiotropy

Introduction

Reproducing optimally is in the evolutionary interest of both sexes, yet rarely is that optimum realized. Since sexual selection acts differentially on each sex, reproductive trait values are often skewed in the favor of one sex to the detriment of the other. This asymmetry in the realization of each sex’s reproductive interests due to conflicts over mating and parental efforts is classically described as sexual conflict (Trivers 1972; Parker 1979). Focusing within the process of mating, sexual conflict can be more specifically described as the negative fitness consequences generated as a result of different reproductive trait optima between the sexes. For example, Bateman (1948) first characterized mating rate as a sexual conflict trait by showing that fecundity measured as a function of mating success differs between males and females in Drosophila melanogaster, such that the fitness of males increases monotonically with increased mating rate, whereas females have an optimum mating rate beyond which total reproductive fitness decreases. Following Bateman, studies across a variety of taxa have shown that polyandrous females incur a cost of mating in the form of decreased fecundity or lifespan relative to monogamous females (Fowler and Partridge 1989; Rice 1996; Watson et al. 1998; Pitnick et al. 2001). Sexual conflict also occurs as part of post-insemination male ejaculate-female reproductive tract dynamics (Hollis et al. 2015; McDonough et al. 2016; Wilburn and Swanson 2016). In particular, seminal fluid proteins can alter female physiology and behavior and contribute to decreased female lifespan (Poiani 2006; Chapman 2011; Sirot et al. 2015). Additionally, sexual conflict can occur at the point of sperm-egg fusion due to fertilization rate differences (Swanson et al. 2001; Clark et al. 2009; Pujolar and Pogson 2011). Collectively, these studies demonstrate the fitness costs of sexual selection pushing males and females in opposing directions of phenotype space.

While sexual conflict is defined at the phenotypic level, the potential for conflict is a genomic property. The alleles that promote the optimal reproductive success of a given sex may have a positive, neutral, or negative impact on the other sex. In the case of negative effects, the subset of sexually differentiated genes with opposing fitness effects in males and females can broadly be categorized as intergenomic conflicts, of which sexual conflict is a special case (Rice and Holland 1997). If males and females had separate genomes, selection could optimize each for male- and female-specific attributes. However, males and females share a common genome and therefore can be viewed as representing an extreme form of polyphenism (Box 1). As selection acts to optimize the reproductive success of one sex, there is a response in the other sex, creating a genomic tug-of-war such that different alleles within the same genome are favored depending on the sexual environment in which they find themselves. Yet in each generation, sex unites the genome and maintains the opportunity for sexual conflict. The resulting functional and evolutionary consequences depend on the balance between pleiotropy and linkage of the genes underlying the conflict trait. Male and female beneficial alleles contributing to a conflict trait can exist at different loci, resulting in interlocus sexual conflict (Rice and Holland 1997; Parker 2006). A single locus can also exhibit sexually antagonistic pleiotropy, whereby different alleles have opposing effects on male and female fitness. If this pleiotropy affects the optimal reproductive success of each sex, then the genetic basis of conflict is referred to as intralocus sexual conflict (Parker 1979, 2006).

Box 1. Sex as a Polyphenism

A polyphenism is an association of a genome with two or more discrete phenotypes across different environments (Suzuki and Nijhout 2006; Simpson et al. 2011). Therefore, polyphenisms are an extreme and discrete form of phenotypic plasticity. For example, in a butterfly color polyphenism, there is an association between the gene expression driving each phenotype and the season in which that phenotype is seen. As shown in the figure below, gene expression is regulated such that it is optimized for each environment and any incorrect expression signifies an environmental mismatch. Similarly, the sexes represent a polyphenism: gene expression is associated with a male or female phenotype based on the sexual environment (Reuter et al. 2017; Mank 2017). However, unlike other polyphenisms, the sexes must interact each generation to reproduce. Males and females are likely to require different genetic functions simply to operate as separate sexually differentiated individuals. These differences by themselves do not necessarily constitute conflict. Only when expression selected for a function in one sexual background has a negative effect on the other does conflict arise. Similarly, not all potentially antagonistic differentiation generated by sex-specific function can be said to be sexual conflict. Gene expression that leads to sex-specific fitness effects is classified as sexual conflict only when those functional effects influence the mating and fertilization success of the individual expressing those genes.

Males and females express a range of unbiased, sex-biased, sex-limited, and sex-linked genes that allow each sex to function properly. While unbiased gene expression may capture the early stages of sexual conflict, the sex-related forms of expression may represent varying degrees of sexual conflict (Connallon and Knowles 2005; Grath and Parsch 2016). Sex-biased genes have skewed expression patterns such that when the expression levels for male and female genes are plotted against each other (as shown below), a negative correlation exists. These patterns show an association with a sex at a given time but are not necessarily a fixed property of a gene and can change based on the external environment (Grath and Parsch 2016). Such a correlation is expected for intergenomic conflicts such as sexual conflict, though not exclusive to this class. Further, genes that are inherently sexually dimorphic in expression, such as those expressed in gonadal tissues, have a negative intersexual correlation despite there being no necessary link to a mating interaction over which there is conflict. Thus, just as gene expression can be associated with a physical environment, so are sex-biased genes associated with a sexual environment. Sex-limited expression represents an extreme form of sex-biased expression, where there is no correlation between male and female gene expression levels (as shown below). Such an expression state can represent a resolution to genomic conflicts, including sexual conflict. Similarly, by linking gene expression with the heterogametic sex chromosome, sexual conflict can be resolved through sex-limited expression. Here, the simplest expression profile is shown with sex-limited expression of the heterogametic chromosome (W) and complete dosage compensation of the homogametic chromosome (Z) between the sexes. However, dosage compensation can be widely variable among taxa, and therefore the homogametic sex chromosome may show a range of expression patterns (Mank 2009; Bachtrog et al. 2011). In all cases, caution must be taken in inferring past sexual conflict as other evolutionary dynamics, such as anisogamy, parental care, and imprinting, can also result in the evolution of sex-limited expression.

Sexual conflict studies have predominantly focused on characterizing conflict traits, so naturally, the search for the underlying genes has followed a forward genetics approach (Wilkinson et al. 2015). Thus far, a handful of genes underlying post-insemination conflict traits have been identified in this manner (Chapman et al. 1995; Swanson and Vacquier 1997; Wigby and Chapman 2005; Clark et al. 2009). Here, the conflict traits are identified during a stage of mating in which the interaction driving sexual conflict is explicit and in a manner in which these traits can then be linked with the genome. However, since such studies necessarily exist on a gene-by-gene basis, they do not address the broader role of sexual conflict in genome evolution.

Reverse genetic approaches are gaining popularity for identifying sexually antagonistic loci. For instance, one potential approach is to focus on very rapidly evolving genes, as these are often related to reproduction (Clark et al. 2006) or immunity (Sackton et al. 2007). This logic, however, is somewhat circular: reproductive proteins are considered rapidly evolving primarily because a number of independent studies have reported them to be so. A more common approach is to quantify gene expression differences between the sexes (Box 1). Although such studies have certainly identified numerous transcriptional sexual dimorphisms (Yang et al. 2006; Innocenti and Morrow 2010; Baker et al. 2011; Viguerie et al. 2012; Bohne et al. 2014), sexual differentiation does not equate to sexual conflict. The genes that enhance male survival at the expense of female survival represent an intersexual trade-off (sexual antagonism), but differential gene expression by itself cannot be taken as evidence for sexual conflict, sensu stricto (Box 1). Therefore, the sex-specific effects of genes alone cannot be taken as primary evidence that those genes underlie a conflict trait.

The genomics revolution has transformed both our conceptual understanding of how evolutionary processes affect the genome as well as our ability to detect the signatures of these processes. Genomic data have the potential to address key questions about the evolutionary importance of sexual conflict, including how many loci are involved in sexual conflict in natural systems? How many loci contribute to a particular conflict trait? How important is sexual conflict relative to other evolutionary processes in shaping genome evolution? Currently, we lack a null expectation to accurately assess genomic data and distinguish true sexual conflict from signatures of sexual differentiation and intergenomic conflict. Perhaps more important, as we will argue, is a general absence of a direct link between genomic variation and reproductive function in many approaches. Such a relationship is critical for distinguishing evolution generated by sexual conflict from more general forms of sexual differentiation. In the following sections, we synthesize our understanding of sexual conflict as a driver of genomic evolution with a specific focus on hypotheses of genomic signatures of sexual conflict, methods for their detection, and the functional information needed to tie genomic evolution to conflict traits.

Genomics of Sexual Conflict: Selection toward Multiple and Moving Fitness Optima

Interlocus Sexual Conflict: Coevolution and Linkage Disequilibrium

Interlocus sexual conflict, in which multiple loci contribute to the conflict trait, creates an intersexual genetic covariance and drives coevolution between the sexes (Gavrilets and Waxman 2002; Gavrilets and Hayashi 2006; Hayashi et al. 2007). These dynamics are analogous to those proposed by the Red Queen Hypothesis, where biotic interactions drive rapid, continuous evolutionary change (Van Valen 1973). Such interactions cause various modes of coevolution—fluctuating, escalatory, and chase away—based on the action of selection (Brockhurst et al. 2014). In the case of sexual conflict, antagonistic male-female interactions are the self-driving force behind coevolution, and the coevolutionary trajectory of the sexes can follow any of these Red Queen modes. Specifically, fluctuating Red Queen dynamics result from frequency-dependent selection, such that the sexes coevolve in a time-lagged, matching fashion. Here, cyclic coevolution is predicted as long as genetic variation for the conflict trait is maintained (Haygood 2004). Conversely, escalatory Red Queen dynamics are characterized by an “arms race” between the sexes, whereby directional selection drives continuous coevolution (Gavrilets 2000; Gavrilets et al. 2001; Gavrilets and Waxman 2002; Hayashi et al. 2007; Pennell et al. 2016). Similarly, chase-away Red Queen dynamics are also driven by directional selection and therefore can lead to continuous coevolution. However, here the driving antagonistic interaction is male exploitation countered by female resistance, such that males must overcome the ever-increasing female resistance threshold to mate (Holland and Rice 1998). Chase-away dynamics in particular create the opportunity for allelic diversification and assortative mating based on genotype fitness (Gavrilets and Waxman 2002; Gavrilets and Hayashi 2006; Hayashi et al. 2007). Unlike traditional Red Queen scenarios, such as host-parasite interactions, interlocus sexual conflict occurs within a shared genome, and thus the evolutionary dynamics are shaped by the degree of sex-specific expression and pleiotropy at each locus contributing to the conflict trait.

When interlocus sexual conflict persists over long timescales, recurrent positive selection can result in accelerated protein evolution. Using a comparative molecular evolution framework based on this expectation, studies have investigated the evolutionary dynamics of interlocus sexual conflict by estimating the ratio of nonsynonymous to synonymous substitutions (ω-ratio) for genes assumed to be involved in sexual conflict so as to determine if the proteins are adaptively coevolving (Yang 1998; Clark and Aquadro 2010). This approach has been particularly well-suited for studying potential sperm-egg fusion conflicts in broadcast spawning marine invertebrates (reviewed in Swanson and Vacquier 2002; Clark et al. 2006; Vacquier and Swanson 2011). Here, polyspermy (the fertilization of a single egg by multiple sperm) drives coevolution between sperm and egg recognition proteins with signatures of escalatory (Clark et al. 2009) and chase-away Red Queen dynamics (Levitan 2006; Manier and Palumbi 2008; Pujolar and Pogson 2011; Sunday and Hart 2013). Elevated rates of evolution in seminal fluid proteins from Drosophila provide another example in which the rate of divergence of fertilization success-related proteins suggests a role in sexual conflict (Begun et al. 2000; Wolfner 2002; Wagstaff 2005). These studies exemplify some of the most powerful information we have on the Red Queen Hypothesis on a macroevolutionary scale.

However, rapid evolution by itself cannot be taken as sole evidence for directional selection via sexual conflict. With a view toward seeking to understand broad patterns of genome evolution, it should be noted that the rapid turnover of amino acid residues may result from adaptive evolution at multiple amino acid residues, but this expectation need not be the case: reduced effective population sizes at loci under selection reduce the efficacy of purifying selection, and thus the observed amino acid substitutions may instead be effectively neutral or mildly deleterious (Smith and Haigh 1974; Vitti et al. 2013; Dapper and Wade 2016). This may be especially important for the evolution of genes on sex chromosomes, which have smaller effective population sizes than autosomes because of their hemizygous state (Bachtrog et al. 2011). However, sex-specific selection can cause deviations from the expected effective population sizes of sex chromosomes due to the faster-X effect (Vicoso and Charlesworth 2009; Mank et al. 2010), making sex-linked genes a potentially interesting subset of genes for such macroevolutionary interlocus sexual conflict studies. Moreover, divergence between populations must be driven by Red Queen dynamics occurring within populations and can therefore be more precisely described by taking advantage of within-population polymorphism at the whole-genome scale (Wilkinson et al. 2015).

Interlocus sexual conflict impacts genomic variation within a population through two related mechanisms. First, Red Queen dynamics affect patterns of polymorphism and haplotype structure. In particular, escalatory and chase-away Red Queen dynamics have the potential to produce classical signatures of persistent directional selection, namely selective sweeps (Figure 1). Specifically, selective sweeps reduce nucleotide variation near the locus under selection because the advantageous allele only exists on limited number of genetic backgrounds. Any rare variants physically linked to the selected allele will also hitchhike to higher frequency and skew the site frequency spectrum (SFS) toward intermediate and high-frequency-derived variants, creating “U-shaped” spectra (Figure 1) (Nielsen et al. 2005; Smith and Haigh 1974). Until recombination can break the linkage between the selected allele and nearby variants, an in-progress or recently completed sweep will also impact local haplotype structure: selection reduces the total number of observed haplotypes and extends the length of the haplotype containing the selected allele (Sabeti et al. 2002, 2007). Under fluctuating Red Queen dynamics, frequency-dependent selection may further affect patterns of variation as selective sweeps remain incomplete and even reverse direction. Like positive selection, frequency-dependent selection can skew the SFS toward high-frequency-derived variants (Figure 1) (Huerta-Sanchez et al. 2008). If genetic variation is sampled while selected alleles are at intermediate frequency, however, frequency-dependent selection may produce excesses of intermediate frequency polymorphisms that are physically linked to the sites under selection (Charlesworth 2006). This type of constant selection will also maintain multiple distinct haplotypes associated with alternative alleles. Depending on the strength of selection and local recombination rates, this perturbation of haplotype variation may be extensive enough to be detectable as a signature of selection. Note, however, that these variant patterns are not unique to sexual conflict, and thus distinguishing between natural selection and sexual conflict will be a challenge without additional functional information.

Figure 1.

The effects of sexual conflict on polymorphism and linkage and methods for their detection. Interlocus and intralocus sexual conflict represent specific instances of broader categories of selection. Trajectories of conflict-associated alleles are shown through time, with unique alleles shown in shades of gray. Representative shifts in the SFS surrounding conflict alleles are shown below their respective trajectories. Solid lines represent the neutral expectation based on Watterson (1975). Effects on linkage and methods for detection are summarized below and in the text.

A second and perhaps more specific effect of interlocus sexual conflict is the generation of linkage disequilibrium among male- and female-beneficial loci due to positive assortative mating generated by fertilization success (Gavrilets and Waxman 2002; Patten et al. 2010; Tomaiuolo and Levitan 2010). Unlike the patterns of sequence variation discussed above, this pattern is the direct result of a shared genome between the sexes. Such a pattern of linkage disequilibrium is especially the case with chase away or escalatory Red Queen dynamics, as only high fitness males can overcome female resistance and successfully fertilize high fitness females. Offspring from these matings therefore contain both male- and female-beneficial alleles at a higher frequency than would be expected under random mating, and if male- and female-beneficial loci occur on the same chromosome, sex-beneficial alleles will initially be found in repulsion—female- and male-beneficial alleles will reside on maternal and paternal chromosomes, respectively. Recombination in the offspring will promote positive linkage disequilibrium by creating physical linkage between sex-beneficial alleles. This pattern of linkage disequilibrium should in principle be detectable within populations, although we are unaware of any studies that have used an approach based on this prediction to study sexual conflict.

The process of linking sex-beneficial alleles is also thought to underlie the evolution of sex chromosomes (reviewed in Bachtrog et al. 2011; Mank et al. 2014; Kirkpatrick 2017). Sex chromosomes are predicted to be hotspots of sexual conflict (Rice 1984) due to their transmission dynamics and their different effective population sizes relative to autosomes. In particular, the build-up of sexually antagonistic loci causes recombination suppression between the sex chromosomes and creates patterns of antagonistic divergence over time (Charlesworth 1991; Wright et al. 2016). By taking advantage of high quality genomic data, recent studies across multiple taxa support this accumulation of sexually antagonistic loci on sex chromosomes (Nam et al. 2015; Lucotte et al. 2016; Wright et al. 2017). Thus, a qualitatively different approach to studying interlocus sexual conflict would be to focus on sex chromosome—particularly neo-sex chromosome—evolution and the accumulation of sexually antagonistic loci. However, despite this a priori expectation, potential conflict loci must still be linked with a conflict trait as not all genes physically linked to the sex-determining locus will necessarily be sexually antagonistic.

For a variety of reasons, the loci involved in interlocus sexual conflict are also likely to demonstrate sex-biased or sex-limited expression. For example, as in classic cases of interlocus sexual conflict, if the genes involved are expressed only in the reproductive tract, then they will by definition have sex-limited expression (Swanson and Vacquier 1997; Kamei 2003; Findlay et al. 2014). In general, sex-limited expression has the potential to alleviate the negative pleiotropic effects of a sexual environment-gene expression mismatch (Box 1). However, the resolution of intralocus sexual conflict, which often results in sex-limited expression, has the potential to generate or exaggerate interlocus sexual conflict (Pennell and Morrow 2013; Berger and Martinossi-Allibert 2016; Pennell et al. 2016). Additionally, genes whose expression is already sex-limited, such has genes functioning in the development of the reproductive tract or secondary sexual characteristics, can still be involved in interlocus sexual conflict even though these genes are already evolving in a sex-specific manner. Indeed, if mating dynamics promote conflict between the sexes, then genes already involved in reproduction and whose expression is sex-limited may be in fact preferentially recruited into conflict.

The detection of candidate loci involved in interlocus sexual conflict can take advantage of existing tools developed to interrogate patterns of genomic variation and gene expression, and distinguishing sexual conflict from other potential causes of within-genome antagonisms should ideally involve both approaches (Figure 1). In particular, incorporating sex-biased or sex-limited gene expression data could help distinguish candidate interlocus sexual conflict loci from loci under positive natural selection (Cheng and Kirkpatrick 2016). Multiple algorithms are available that identify signatures of selection from genome-wide single nucleotide polymorphism (SNP) data, generated either through whole-genome shotgun sequencing or reduced-representation approaches (Nielsen 2005; Schrider and Kern 2016). To scan for sex-biased or sex-limited expression across the entire transcriptome, RNA-seq studies are becoming commonplace, even in nonmodel organisms. Since linkage disequilibrium between alleles at multiple interacting loci may be an important signature of interlocus sexual conflict, it will be beneficial to use sequencing protocols that retain genotype—and ideally haplotype—information for all individuals in a population. This presents a problem especially for studies using small organisms or very large sample sizes, where pooling of multiple individuals is common. Luckily, DNA library preparation methods are rapidly improving for small amounts of starting DNA, and molecular barcoding now allows multiplexing with tens of thousands of unique barcodes. Certainly, these methods will continue to improve in the coming years.

Intralocus Sexual Conflict: Pleiotropy and Balancing Selection

Intralocus sexual conflict is a specific case of antagonistic pleiotropy, with the pleiotropic effects being dependent on shared effects across different sexual environments (Box 1). Thus, intralocus sexual conflict is similar to an environment-dependent fitness trade-off, where sex is the environment. Such trade-offs will create a sex-by-genotype interaction and would be expected to drive balancing selection (Connallon and Clark 2014). Further, this form of antagonistic pleiotropy should prevent alleles with differential sex-specific fitness effects from reaching fixation, thus maintaining sexual conflict (Arnqvist 2011). However, few predictions and little empirical evidence exist on the genomic consequences of intralocus sexual conflict.

Effects of intralocus sexual conflict on genomic variation should be similar to those predicted from more classical environment-dependent fitness trade-offs that produce balanced polymorphisms (Charlesworth et al. 1997, 2003; Lenormand 2002). Over long timescales, balancing selection can extend the expected residence time of an allele within a population far beyond that for neutral or positively selected sites, leading to increased polymorphism in linked genomic regions. Persistent balancing selection should therefore be detectable in between-population or between-species comparisons. Ideally, studies of intralocus conflict would focus on ongoing conflict within a mating population, when the phenotypic and fitness consequences of conflict can be directly measured. But, this level of study may make the genomic signatures of sexual conflict more difficult to detect.

In a general scenario of an environment-dependent fitness trade-off, selection generates genetic divergence among chromosomes carrying alternative alleles (Lenormand 2002). At the locus under selection, maintenance of variation depends on selection overriding the force of migration between environments. In flanking genomic regions, divergence extends out from the selected locus as a function of these forces and the recombination rate. In other words, larger selection coefficients and lower migration and recombination rates result in more extensive divergence among chromosomes.

A key difference between intralocus sexual conflict and other models of environment-dependent fitness trade-offs arises when we consider the migration rate. In most, limited migration among environments allows for the selective maintenance of alternative alleles and produces signatures of divergence at and around selected loci (Slatkin 1987; Lenormand 2002; Yeaman and Whitlock 2011; Roesti et al. 2015). Under sexual conflict, however, the different environments are males and females. Thus, complete outcrossing among selective environments in each generation enforces an extremely high effective “migration” rate. Nevertheless, recent models suggest that intralocus sexual conflict can promote and maintain divergence between recombining sex chromosomes when conflict loci are in linkage disequilibrium with the sex-determining region (Kirkpatrick and Guerrero 2014). Genetic variation can also be maintained on autosomal loci if sexually antagonistic selection coupled with assortative mating is sufficiently strong (Arnqvist 2011). However, a more general understanding of how intralocus sexual conflict impacts genomic variation is currently lacking. The shared genome between the sexes likely provides a potent barrier to allelic divergence, as recombination will rapidly reduce divergence among chromosomes carrying alternative alleles. Even if intralocus sexual conflict remains unresolved for long periods of time, differentiation among chromosomes carrying alternative alleles is likely to be extremely localized unless recombination rates are very low. Therefore, while a single instance of intralocus sexual conflict may be a weak force structuring genomic variation, opposing selection in males and females may still result in allele frequency differences between the sexes at a conflict locus. Open questions remain about how many loci are involved in these conflicts and what their combined effects on genomic variation may be.

Genomic Detection of Sex-specific Selection

Given that the sexes share alleles, the detection of candidate loci involved in intralocus sexual conflict is somewhat more complicated than for interlocus conflict. Methods based on the SFS may be able to detect conflict loci through the expected excess of intermediate frequency polymorphisms (Figure 1). But as noted above, this effect will be highly localized unless conflict alleles are very young. The skew in the SFS will also closely resemble other forms of balancing selection, making it difficult to identify sexual conflict as the causative process. Recently, genome scan approaches have been adapted to find sexually differentiated SNPs by identifying allele frequency differences between the sexes (Cheng and Kirkpatrick 2016; Lucotte et al. 2016; Flanagan and Jones 2017). These approaches are appealing because they use established population genetic statistics, such as F_ST, to identify sexually antagonistic loci. Specifically, a male-female F_ST measures the change in allele frequency due to opposing selection between the sexes. This signature can be caused either by sexual conflict over reproductive fitness or sex-specific viability effects. In the latter case, sexually differentiated loci could be linked to sexual conflict in that they affect the optimal reproductive interest of each sex. However, sex-specific viability effects can also result simply from natural selection acting differentially on males and females. A male-female F_ST alone cannot distinguish between these selective processes, and therefore candidate genes would still need functional verification. Coupling these regions of differentiation to sex-specific expression does not necessarily distinguish among these possibilities as sexual differentiation does not equal to sexual conflict. Since the sexes for the most part share the same genome and reproduce each generation, any allele frequency difference between males and females only reflects a single generation of selection, regardless of the source of selection. This single generation limitation makes the feasibility of a male-female genome scan suspect.

To investigate this possibility more deeply, we developed a population genetic model to quantify the change in allele frequency at a single locus after a single generation of sexually antagonistic selection. Specifically, we examined the change in frequency of an allele separately within males and females to account for the sex-specific effects of selection. In particular, we focused on a male-beneficial, female-antagonistic allele. Using these sex-conditional allele frequency changes, we calculated the difference in allele frequency between the sexes (similar to calculating a Fisher’s exact test) and male-female F_ST. The complete model is outlined in Box 2.

Box 2. Population Genetic Model of Intralocus Sexual Conflict

Can whole genome scans identify sexually antagonistic loci by comparing allele frequency differences within males and females? We consider this possibility by developing a simple population genetic model of the change in allele frequency between the sexes due to a single generation of sexually antagonistic selection. Specifically, we use a framework that allows us to estimate sex-conditional allele frequency changes such that selection can act differentially on an allele depending on the sexual environment (following Kidwell et al. 1977). This approach is actually fairly straightforward because transmission dynamics do not have to be considered. Instead, only the within-generation changes in genotype frequencies within male- and female-specific pools need to be tracked.

Consider a locus segregating for alleles A and a that are sex-specific beneficial: allele A is female-beneficial and allele a is male-beneficial. We represent sexually antagonistic selection as the cost of having the allele favored in the other sex. The relative fitnesses of the genotypes AA, Aa, and aa in males are thus $1-s_f::1-h_1{s}_f::1$, where s_f is the cost of a male possessing the female-favorable allele and h₁ is the dominance coefficient in males. The relative fitnesses of the genotypes AA, Aa, and aa in females are similarly $1::1-h_2{s}_m::1-s_m$, where s_m is the cost of having the male allele, and h₂ is the dominance coefficient in females.

Since sexual selection is typically stronger in males, we quantified the sex-conditional change in allele frequency of the male-beneficial allele. We can describe the frequency of the male-beneficial allele in males after a single generation of selection (q_m) as the relative frequency of a-containing male genotypes after selection divided by mean fitness of males the generation before selection:

 

$$q_m=\frac{q^2+pq(1-h_1{s}_f)}{p^2(1-s_f)+2pq(1-h_1{s}_f)+q^2}.$$

Here, q is the starting frequency of a, and p = 1 − q. Similarly, the frequency of the male-beneficial allele in females after a single generation of selection (q_f) is the frequency of the allele in females divided by the mean fitness of females:

 

$$q_f = \frac{q^2(1 - s_m) +pq(1 - h_2{s}_m)}{p^2 + 2pq(1 - h_2{s}_m) + q^2(1 - s_m)}.$$

The magnitude of change in allele frequency is simply the difference in allele frequencies between the sexes, such that:

 

$$\Delta q = q_m- q_f.$$

This can be translated into the familiar F_ST statistic (Wright 1931; Cheng and Kirkpatrick 2016) as:

 

$$F_{ST}= \frac{\Delta q^2}{4pq}.$$

As outlined in the main text, we considered the effects of dominance for five intralocus sexual conflict scenarios: additive beneficial and conflict allele effects (h₁ = h₂ = 0.5), conflict allele dominance (h₁ = h₂ = 1), beneficial allele dominance (h₁ = h₂ = 0), female-beneficial allele dominance (h₁ = 1, h₂ = 0), and male-beneficial allele dominance (h₁ = 0, h₂ = 1) (Figures 2 and 3). For simplicity, we used a symmetrical cost of selection between the sexes (s_m = s_f) when looking at dominance effects. We also examined the effects of asymmetrical selection costs between the sexes, focusing on the additive dominance case (h₁ = h₂ = 0.5). A haploid version of the model is qualitatively similar to the diploid additive case (data not shown).

Overall, we find that a single generation of selection is not sufficient to change the frequency of an allele such that there is a distinct genomic signature unless selection is unrealistically strong. Although both the difference in allele frequency between the sexes and male-female F_ST increase as a function of selection (Figure 2), the actual change in allele frequency is very small. For example, under an additive effects scenario, strong sexually antagonistic selection (s = 0.1) gives a virtually undetectable male-female F_ST (F_ST = 0.0007). These values are qualitatively consistent with a similar model and empirical estimates of sex-specific F_ST in humans and Drosophila produced by Cheng and Kirkpatrick (2016). Conflict allele or beneficial allele dominance yield outcomes that are qualitatively similar as those for an additive effects scenario, although sex-specific dominance can have qualitative and quantitative effects (Figure 3). Under a scenario such as female-beneficial allele dominance, the maximum male-female F_ST with strong selection is approximately 1.5 times larger than under additive effects. Even so, this male-female F_ST still has a negligible impact on single locus detection (F_ST = 0.001). Overall there appears to be very little power to detect sex-specific differentiation within a generation, even when selection is strong (also see Cheng and Kirkpatrick 2016).

Figure 2.

The change in the male-beneficial allele frequency with additive beneficial and conflict allele effects (h₁ = h₂ = 0.5) after a single generation of sexually antagonistic selection, where the cost of selection is equal between the sexes (s_m = s_f = s). (A) The difference between male and female allele frequencies (∆q) as a function of selection for different initial values of q. (B) Male-female F_ST as a function of selection for different values of q. For both response measures, the change in allele frequency increases as the cost of selection increases. The maximum change in allele frequency is seen when the male- and female-beneficial alleles start at the same frequency (q = 0.5).

Figure 3.

The change in the male-beneficial allele frequency due to a single generation of sexually antagonistic selection is dependent on the dominance relationships between the sexes. When there is conflict allele (h₁ = h₂ = 1) or beneficial allele (h₁ = h₂ = 0) dominance in both sexes, the difference in allele frequency is maximized when the male and female-beneficial alleles start at the same frequency (q = 0.5). Additionally, there is a diminishing response to selection, such that the difference in allele frequency between the sexes forms a concave surface. Sex-specific beneficial allele dominance changes the shape and magnitude of the selection response curve. For example, when the female-beneficial allele is dominant (h₁ = 1, h₂ = 0), the response surface is shifted and stretched. Here the cost of selection is equal between the sexes (s_m = s_f = 0.1).

Since sexual selection acts differentially between the sexes, we also examined the effects of asymmetrical fitness costs between the sexes. Selection had to be at least an order of magnitude different between the sexes to detect any quantitative differences between male and female F_ST. The biological relevance of this high cost of selection in natural populations is questionable, particularly when selection is inherently linked with reproductive success. For instance, a selection cost of s = 0.1 suggests that 10% of individuals die each generation or do not contribute to the next generation due to their sexual context alone. These results suggest that the genetic load created by sexual conflict has the potential to be quite large. Similarly, if selection across the sexes is indeed this strong, then the opportunity to resolve conflict should also be strong, and we would therefore expect natural selection for alleles that counter sexual antagonistic selection. However, if multiple loci are contributing to a conflict trait, then the average cost of selection could potentially be lower (Cheng and Kirkpatrick 2016). Alternatively, this single generation framework may not capture the full effects of sex-specific selection. Our model assumes that there is uniform representation of alleles within the gamete pool at the start of each generation. However, selection on certain gamete types could skew the gamete pool such that only a subset of individuals of each sex contributes their alleles to the next generation (Kidwell et al. 1977; Arnqvist 2011). Such dynamics have the potential to drive male-female differentiation at a locus and warrant further exploration.

From a practical perspective, detecting such small F_ST values is likely to be unrealistic. Nevertheless, recent studies using human data suggest that such sex-specific signatures of selection can in fact be detected at the whole-genome level (Lucotte et al. 2016). Similarly, data from pipefish found remarkably high male−female F_ST values (Flanagan and Jones 2017). The ability to detect male−female differentiation using a genome scan suggests that such strong selection costs may be more prevalent than believed (though the mechanism of selection is unknown). If true, our view that the sexes share largely the same genome must be fundamentally altered. However, these male−female tests of differentiation are subjected to false positives from many sources. Extremely large male and female sample sizes would be required not only to detect such F_ST values but also to prevent false positives resulting simply from random sampling effects. Additionally, when possible, corrections must be made for linkage to the sex chromosomes and, in particular, the sex-determining region as these can also drive spurious male−female differentiation. Such a correction may be difficult in systems where the sex-determining region or even sex chromosome is unknown, thus leading to false-positive values. We suggest that a genome-wide permutation test (as in Churchill and Doerge 1994) should be performed to determine the null expectation and distinguish false positives from true signatures of male−female divergence. While genome scan approaches may be appealing for the moment, we suggest caution in interpreting their results until further theoretical work is completed so that we more fully understand the biology underlying the statistic being measured and are able to distinguish false positives from signatures of sexually antagonistic SNPs.

Synthesis: Linking Genomic Signatures with Conflict Traits

While we can readily identify sexually differential expression patterns, not all of these necessarily correspond to sexual conflict. Since sexual conflict is defined at the phenotypic level, ultimately, we must link candidate loci identified via genetic and genomic signatures with their functional role within conflict traits. The most successful examples to date involve systems that are prima facia involved in reproductive interactions, such as sperm and seminal fluid proteins (reviewed in Swanson and Vacquier 2002; Clark et al. 2006; Vacquier and Swanson 2011). Expanding this framework to more general phenotypes will be a challenge. Several approaches can be taken to narrow down and prioritize these candidate loci (reviewed in Findlay and Swanson 2010). For instance, automated function-prediction software can verify if candidate loci can realistically be expected to be involved in a mating interaction. Unfortunately, such programs tend to be somewhat generic and have limitations to the extent of functional information that they provide (Friedberg 2006). Therefore, they should be treated as a coarse pass over data but not the only method via which sexual conflict is assigned. Coupling divergence methods with expression patterns is perhaps more informative (Harrison et al. 2015; Cheng and Kirkpatrick 2016) but still suffers from the potential weakness of conflating correlation and causation in the context of sexual conflict.

Evolutionary rate covariation (ERC) is another method that is particularly useful for prioritizing candidate interlocus sexual conflict genes (Clark and Aquadro 2010; Clark et al. 2012; Wolfe and Clark 2015). This method uses the ratio of nonsynonymous to synonymous substitutions to determine if the evolutionary rate of two proteins is correlated. Such a correlation is expected if proteins are physically interacting and subject to intermolecular coevolution (Clark et al. 2006; Clark and Aquadro 2010), if proteins are functionally related and therefore experiencing similar selective pressures (Clark and Aquadro 2010), or if gene expression patterns covary (Fraser et al. 2004; Hakes et al. 2007). By comparing the evolutionary rates for positively selected loci, potential Red Queen coevolutionary dynamics can be identified (Clark et al. 2009). While potentially powerful, this method does require a comparative genomics framework with sufficient dense phylogenetic signal.

In the end, true verification of sexual conflict necessitates experimental biology. Sexual conflict does not create a unique genomic signature—other forms of environment-specific selection, including sex itself as an environmental context, can produce the signatures outlined above. Classic molecular genetics approaches are likely to be necessary to determine the function of candidate conflict-related loci. Such approaches, once wildly unrealistic for most systems, are increasingly coming into reach with the advent of general purpose genome editing approaches (Bono et al. 2015). Only when coupled with reproductive experiments will screening for a conflict trait and the mechanism of selection be possible. This complete link between genotype, phenotype, and function is necessary to describe the complete sexual conflict pathway and should be the standard to which we strive.

Conclusion

Reproduction is a fundamental biological process, yet the complexity of this process creates the potential for antagonistic interactions to become more pronounced. Sexual conflict, in its original usage, defines the negative fitness consequences generated by the process of mating and fertilization. This requirement of a male-female reproductive interaction distinguishes sexual conflict from other forms of antagonistic selection. We urge that a precise definition of sexual conflict be used when studying evolutionary dynamics. Such specificity will become increasingly important as we move from vague descriptions of possible evolutionary patterns to identifying specific genetic loci at genomic scales. Despite an emerging ability to detect potential loci involved in sexual conflict, these signatures are largely indistinguishable from those caused by natural selection and intergenomic conflicts in general. To firmly move the field into the genomics era, we need to couple an analysis of evolutionary patterns to the sex-specific functional context of putative conflict-related loci. That the field is now poised to capitalize on these approaches promises many exciting developments and novel insights into the evolution of sexual conflict in the very near future.

Funding

This work was funded by the National Institutes of Health (training grant T32 GM007413 to KRK and R01 GM102511 to P.C.P.) and the ARCS Foundation Oregon Chapter (K.R.K.).

Acknowledgements

We are grateful to Melissa Wilson Sayres and John Logsdon for the opportunity to contribute this article as part of the symposium on the “Evolutionary Genomics of Sex”. We would like to thank three anonymous reviewers and Sean Stankowski for constructive comments on the manuscript.