• Christine Grossen,

    1. Department of Ecology & Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland
    2. Institute of Evolutionary Biology and Environmental Studies (IEU), University of Zürich, CH-8057 Zürich, Switzerland
    3. E-mail:
    Search for more papers by this author
  • Samuel Neuenschwander,

    1. Department of Ecology & Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland
    2. Vital-IT, Swiss Institute of Bioinformatics, University of Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author
  • Nicolas Perrin

    1. Department of Ecology & Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland
    Search for more papers by this author


Recombination arrest between X and Y chromosomes, driven by sexually antagonistic genes, is expected to induce their progressive differentiation. However, in contrast to birds and mammals (which display the predicted pattern), most cold-blooded vertebrates have homomorphic sex chromosomes. Two main hypotheses have been proposed to account for this, namely high turnover rates of sex-determining systems and occasional XY recombination. Using individual-based simulations, we formalize the evolution of XY recombination (here mediated by sex reversal; the “fountain-of-youth” model) under the contrasting forces of sexually antagonistic selection and deleterious mutations. The shift between the domains of elimination and accumulation occurs at much lower selection coefficients for the Y than for the X. In the absence of dosage compensation, mildly deleterious mutations accumulating on the Y depress male fitness, thereby providing incentives for XY recombination. Under our settings, this occurs via “demasculinization” of the Y, allowing recombination in XY (sex-reversed) females. As we also show, this generates a conflict with the X, which coevolves to oppose sex reversal. The resulting rare events of XY sex reversal are enough to purge the Y from its load of deleterious mutations. Our results support the “fountain of youth” as a plausible mechanism to account for the maintenance of sex-chromosome homomorphy.

According to dominant models of sex chromosome evolution (Ohno 1967; Bull 1983; Rice 1996), the initial step in the life of a sex chromosome is set by an autosomal mutation (or gene duplication) that interacts with the sex-determining cascade, such that heterozygotes develop into one sex (e.g., XY males), while homozygotes develop into the other (XX females).

Sexually antagonistic mutations should soon accrue in the vicinity of this gene, benefiting from linkage disequilibrium with the sex-determining locus (Fisher 1931; Rice 1987). Cosegregation with sex will then be improved through reduced recombination in the heterogametic sex (Nei 1969; Charlesworth and Charlesworth 1980; Bull 1983; Rice 1987). This nonrecombining sex-determining region (SDR) might later expand along the chromosome, as additional sexually antagonistic mutations appear (Rice 1996). Note that, when sex chromosomes evolve from hermaphroditism, two mutations are required (one to suppress the male function, the other to suppress the female function), so that recombination arrest is also selected to prevent the production of neuter individuals (Charleworth et al. 2005).

As a side effect of recombination arrest, however, this nascent SDR will also accumulate deleterious mutations. The drastic reduction in effective size of the Y chromosome (one third of the X, one quarter of an autosome), further accentuated by sexual selection and Hill–Robertson effects (including background selection and hitchhiking effects), will enhance drift, counteracting the fixation of beneficial mutations and the elimination of mildly deleterious ones (reviewed in Charlesworth and Charlesworth 2000; Bachtrog 2006). Lethal recessive mutations will also accumulate (being protected from purging by permanent heterozygosity), as already observed on young sex chromosomes in fish (e.g., Haskins et al. 1970) and amphibians (Miura et al. 2012). Muller's ratchet will condemn this proto Y to a progressive decay (Muller 1964; Felsenstein 1974; Charlesworth 1978; Charlesworth and Charlesworth 1997; Gordo and Charlesworth 2000b).

The above scenario convincingly accounts for the highly decayed Y and W chromosomes found in mammals and birds respectively, where evolutionary strata also testify to the progressive expansion of the nonrecombining segment (Lahn and Page 1999; Lawson-Handley et al. 2004). After some 170 million years of differentiation, the human Y only codes for 27 distinct proteins, as compared to the 1669 genes found on the X (Skaletsky et al. 2003; Livernois et al. 2012). In sharp contrast with this pattern, however, most cold-blooded vertebrates present homomorphic sex chromosomes, with no apparent sign of decay or differentiation (e.g., Devlin and Nagahama 2002; Eggert 2004).

Two main causes may account for this contrast. On one hand, newly arising master sex-determining genes might regularly replace older ones, so that sex chromosomes have no time to decay (Schartl 2004; Volff et al. 2007). This “high-turnover model” has received strong empirical support from investigations on fish (e.g., Woram 2003; Tanaka et al. 2007; Cnaani et al. 2008; Ross et al. 2009), amphibians (Miura 2008; Stöck et al. 2011a), and reptiles (Ezaz et al. 2006).

On the other hand, homomorphy might also stem from occasional XY recombination, as recently documented in amphibians (Stöck et al. 2011b): Several European treefrog species (Hyla arborea group) inherited the same pair of sex chromosomes from a common ancestor, 5–7 million years ago. However, despite the absence of male recombination, their Y chromosomes show no sign of decay or differentiation (X and Y sequences cluster by species, not by gametolog), pointing to occasional XY recombination in all lineages. This process may also account for the large overlaps in X- and Y-allelic frequencies at sex-linked markers found in natural populations of amphibians, despite absence of male recombination (e.g., Berset-Braendli et al. 2007; Alho et al. 2010; Matsuba et al. 2010). Note that these two causes are not exclusive, and more likely interdependent: turnover might actually be facilitated by XY recombination (van Doorn and Kirkpatrick 2007, 2010).

XY recombination might occur either in XY males, or in sex-reversed XY females (Perrin 2009). Sex is easily reversed by temperature in many fish, amphibians, and reptiles, at values likely to be occasionally met in nature (e.g., Wallace and Wallace 2000; Baroiller et al. 2009). Sex reversal has been documented from natural populations in a range of species (e.g., Crew 1921; Witschi 1929; Aida 1936; Kawamura and Nishioka 1977; Nagler et al. 2001; Matsuba et al. 2008). As recombination patterns often depend on phenotypic rather than genotypic sex (e.g., Wallace et al. 1997; Kondo et al. 2001; Lynn et al. 2005; Campos-Ramos et al. 2009; Matsuba et al. 2010), X and Y are expected to recombine in sex-reversed XY females. This “fountain of youth” should prevent the Y from accumulating deleterious mutations (Perrin 2009), hence contributing to the widespread sex-chromosome homomorphy observed in many cold-blooded vertebrates.

Both sex reversal and XY recombination might occur simply as side products of a physiological dysfunction (namely, a failure of the mechanisms that control sex determination or male recombination). Sexually antagonistic selection, however, should oppose such failures: Individual fitness will be negatively affected if male- and female-benefiting alleles are expressed in the wrong sex.

Alternatively, sex reversal and XY recombination might actually be favored as a way to purge the load of deleterious mutations accumulating on the Y chromosome. In absence of dosage compensation, this load is expected to impose a heavy burden on male fitness. Rare events of recombination are sufficient to counteract Muller's ratchet (Maynard Smith 1978; Charlesworth et al. 1993), and might thus be selected as a way to rejuvenate decaying sex chromosomes.

In the present article, we use individual-based simulations to investigate the interplay between sexually antagonistic selection and deleterious mutations on the evolution of XY recombination. As already pointed out, X and Y chromosomes might recombine either in XY males or in sex-reversed XY females. We might have chosen to model either the evolution of male recombination rate, or that of sex-determination and sex-reversal mechanisms. We chose the latter, being interested in further investigating the “fountain-of-youth” hypothesis, but readers should keep in mind that similar investigations and results might be reached by modeling the former. We also warn that our simulations, although inspired by the Hyla and Rana models (which provide evidence for XY recombination in absence of male recombination), are not intended to provide quantitative predictions for any specific system. They are only aimed at formalizing our argument (to check whether evolutionarily stable rates of XY recombination might indeed differ from zero), and investigating which parameters (in terms of intensity of sexually antagonistic selection, strength and dominance of deleterious mutations, or effective population size) are expected to affect this process.

Materials and Methods


The SDR was assumed to comprise one sex-determining gene (SD), one sexually antagonistic gene (SA), and 100 additional functional genes, potentially mutating to deleterious forms (Fig. 1B). Sex was modeled as a threshold trait, underlain by a liability trait (A), such that offspring with a liability exceeding a threshold (arbitrarily set to zero) developed into males, otherwise into females (Fig. 1A). Individual liability value was a quantitative trait, computed as the sum of the two allelic values at the sex-determining locus (161 possible allelic values ranging from −8 to +8, step 0.1) plus an environmental effect, randomly sampled from a normal distribution with average 0 and variance inline image. We implemented multiple alleles with mutations among them, to investigate the evolution of rate of sex reversal, expected to be higher for smaller (absolute) allelic values and larger environmental variance (Fig. 1A; see also Grossen et al. 2011).

Figure 1.

(A) Sex determination in a quantitative-genetics framework. The amount of sex factor (vertical axis) depends both on genotype (XX or XY) and on environmental variance (normal distribution around genotypic mean). Sex reversals are expected for small (absolute) allelic values and large environmental variance. Arrows indicate selective pressure for allelic values at the sex determiner to diverge (plain arrows, due to sexually antagonistic genes) or to converge (dashed arrows, due to the accumulating load of deleterious mutations). (B) Male and female linkage map of the model nonrecombining sex-determining region (SDR), in cM units. No recombination occurs in males (map length = 0 cM). In phenotypic females, the sex-determining gene (SD) was arbitrarily positioned at the end of the chromosomes (0 cM) and the sex antagonistic gene 30 cM away from it. One hundred functional genes, potentially mutating to deleterious forms (potentially deleterious loci) were evenly distributed every cM.

The sexually antagonistic trait B affected adult fitness in a sex-specific way, according to the sigmoid functions (Fig. S1):


where M is the lower asymptote (minimum fitness value). Hence male fitness increased from M to unity with increasing B values, while female fitness decreased from unity to M (Fig. S1). The trait value (B) was obtained from the sum of the two allelic values at the SA locus plus an environmental effect, randomly sampled from a normal distribution with average 0 and variance inline image. We allowed a large range of allelic values (from −6 to +6, step one), to investigate the fine-scale evolution of this trait under the opposite forces of sexually antagonistic selection (expected to maximize the contrast between males and females) and recombination (expected to counterselect such contrasts). Sexually antagonistic traits (e.g., color or size) are a priori continuous traits, even though often reduced a posteriori (i.e., by natural or sexual selection) to small sets of discrete values.

Fitness was also affected by deleterious mutations hitting the functional genes spread on the sex chromosomes. Offspring viability (v) was computed as the product over loci


where vi= 1, 1 −s, or 1 −hs depending on whether the locus i was homozygous wild type, homozygous mutant, or heterozygous, respectively. The selection and dominance coefficients s and h were varied throughout simulation sets (but held constant within sets; see values below). Following Higgins and Lynch (2001), we defined h as a negative function of s: inline image which implies that mutations with large effects are much more recessive than mutations with small effects. For a subset of parameter values, we also ran simulations with h set to either 0.05 or 1, independent of s.

The SDR only recombined in females (independent of genotype), with functional genes regularly distributed one cM apart from each other (hence, SDR total map length was 100 cM in females, vs. 0 cM in males). The sex-determining locus was arbitrarily localized at one extremity of the region, and the sexually antagonistic locus 30 cM away (Fig. 1B).


Individual-based simulations were run with a modified version of the program quantiNemo 1.0.3 (Neuenschwander et al. 2008), using a simple life cycle with nonoverlapping generations. At reproduction, offspring were produced in numbers matching the carrying capacity (N= 100, 1000, or 10,000), and randomly allocated to individual mothers and fathers (corresponding to a promiscuous mating system), with probabilities given by sex-specific fitness values (eq. 1a and 1b). Subsequent offspring survival depended on their load of deleterious mutations (eq. 2).

A first set of simulations was aimed at investigating the effect of sex reversal and recombination on the accumulation of deleterious mutations. Deleterious mutations (s= 0.025 or 0.0125) occurred at rate μD= 10−4 (neglecting reverse mutations to wild-type alleles) in initially mutation-free individuals. Alleles at the sex-determining locus were set to X =−2 and Y = 6, producing genotypic values −4 and +4 for XX and XY individuals, respectively. Different rates of sex reversal and recombination were generated via environmental variance (inline image was varied from 0 to 2, steps 0.2). The sexually antagonistic trait (B) was allowed to evolve, under sexually antagonistic selection (M= 0.5) or not (M= 1), with mutations (μSA= 10−4) uniformly distributed over the range −6 to +6 (KAM mutation model). We furthermore investigated in more detail a system with purely genetic sex determination (GSD), that is, with complete absence of sex reversal (environmental variance set to inline image= 0). Sexually antagonistic selection was varied by changing the lower fitness asymptote from M= 0 to 0.75 (step 0.25), and the accumulation of deleterious mutations by varying the selection coefficient from s= 0 to 0.8 (logarithmic scale: s= 0.8/2n, where n= {0,1,2 … 12,∞}).

The second set of simulations was aimed at investigating the evolution of recombination rate under the opposing forces of sexually antagonistic selection and deleterious mutations. The sex-determining locus was allowed to mutate (rate μSD= 10−4, truncated normal distribution of mutation steps with variance 1.8). Initial allelic values were either set to X =−2 and Y = 6, or drawn from a uniform distribution from −8 to +8. Environmental variance was set to inline image= 1. Sexually antagonistic selective pressure was varied from M= 0 to 0.75 (step 0.25) and the strength of deleterious mutations from s= 0 to 0.8 (logarithmic scale: s= 0.8/2n, where n= {0,1,2 … 12,∞}). Simulations were run for all possible combinations of parameters (fully factorial design).



In the first set of simulations, the proportion of loci fixing deleterious mutations progressively increased with time (Fig. 2), at a pace set by the rate and strength of deleterious mutations, chromosomal effective population sizes (NX or NY) and recombination rates (determined by environmental variance). Recombination was negligible for inline image≤ 0.4, so that the dynamics of deleterious mutations on the X and the Y were effectively decoupled. As a result, deleterious mutations accumulated on the Y (Fig. 3A), strongly affecting the survival of male offspring (Fig. 3B). For inline image > 0.4, some sex reversal and recombination occurred, at rates increasing with environmental variance. The mutation load on the Y was thereby much purged, reaching values similar to the X at inline image= 2 (Fig. 3A). As a result, the survival of male offspring was also much improved (Fig. 3B). One XY female every few generations was sufficient to purge most of the mutation load on the Y. For N= 1000, M= 1, and s= 0.0125, for instance, one sex-reversal event every five generations (inline image= 1.4) lowered the proportion of deleterious mutations fixed on the Y down to 1–2% (Figs. 2 and 3A).

Figure 2.

Proportion of loci on the Y that fixed deleterious mutations (vertical axis) as a function of time (horizontal axis). Recombination rate is a function of environmental variance (inline image). Mutations accumulate in absence of recombination (small inline image values), but are purged when sex reversal allows recombination with the X (large inline image values). Dynamics also depend on whether sexually antagonistic selection occurs (white boxes, M= 0.5) or not (gray boxes, M= 1). Other parameter values: N= 1000, s= 0.0125, and μ= 10−4. Box plots with median are delimited by 25th and 75th percentiles, and whiskers represent the 5th and 95th percentiles.

Figure 3.

(A) Proportions of deleterious mutations fixed on X (gray boxes, bottom lines around zero) and on Y (white boxes) after T= 100,000 generations, and (B) corresponding survival rates for male (white) and female offspring (gray, small dark boxes around 0.98), as a function of environmental variance inline image (horizontal axis). Mutations accumulate on the Y and depress male offspring survival when inline image is too small to allow sex reversal and recombination. Quantitative values also depend on whether sexually antagonistic selection occurs (right panels, M= 0.5) or not (left panels, M= 1), as well as on population size (N= 1000 or 10,000). Other parameter values s= 0.0125 and μ= 10−4. Box plots as in Figure 2.

In the absence of sexually antagonistic selection (M set to 1), allelic values at the sexually antagonistic locus evolved neutrally around an average of zero, with a large variance. With M set to 0.5, by contrast, these alleles diverged markedly (close to +6 and −2 for the Y- and X-linked alleles, respectively; data not shown). This affected the purging process by depressing fitness in sex-reversed XY females (due to the male allele at the sexually antagonistic locus), thereby lowering the opportunity for X-Y recombination. Purging on the Y was thus less efficient (Fig. 2, white boxes; Fig. 3A, right panels), resulting in lower survival of male offspring (Fig. 3B, right panels).

The more detailed simulations of a pure GSD system (inline image= 0) allowed better characterization of the crucial role played by different strengths (s) of deleterious mutations (Fig. 4, left panels), together with effective population size (N). Selection is expected to counteract drift for sN values exceeding one (Kimura 1983). Accordingly, increasing s induced a shift from a regime of accumulation to one of elimination (Fig. 4B, left panel). Offspring survival thus first decreased with increasing s (due to the increasing deleterious effect of accumulated mutations), then reincreased when reaching the domain of elimination (Fig. 4C, left panel). The shift from accumulation to elimination occurred at much higher s values for the Y than for the X (e.g., at s= 0.025 for Y vs. s= 0.003 for X, for N= 1000), due to its lower effective size and absence of recombination. This induced discrepancies between male and female survival curves at intermediate s values (Fig. 4C, left panel), providing incentives for the Y to recombine with the X.

Figure 4.

(A) Allelic values at the sex-determining locus on the X (orange) and Y chromosomes (green), (B) proportions of deleterious mutations fixed on X (orange boxes) and Y (green boxes), and (C) survival rate of female (red boxes) and male offspring (blue boxes) at T= 100,000 generations, for varying strength of deleterious mutation selection (s= 0–0.8, logarithmic scale on horizontal axis with s= 0.8/2n, where n= {0,1,2, … 12,∞}, μSA= 10−4, μD= 10−4). Left panels: pure GSD (inline image= 0 and μSD= 0). The shift from accumulation to elimination occurs at lower s values for the X than for the Y, and also for large than for small populations (N= 10,000 vs. 1000). Right panels: SD evolvable (inline image= 1 and μSD= 10−4). The X and Y alleles converge toward the threshold at intermediate s values, which induces sex reversal and recombination, improving male offspring survival (compare with left panels). This trend is stronger in larger populations (N= 10,000 vs. 1000) and at weaker sexually antagonistic selection (M= 0.75 vs. 0). Box plots as in Figure 2. See Figure S2 for a version that also provides results for N= 100 and M= 0.25, 0.5.


In the second set of simulations, alleles at the SD locus were allowed to evolve. Values after 100,000 generations are provided in Figure 4A (right panel) as a function of the selection coefficient s, for different intensities of sexually antagonistic selection (M= 0 and 0.75) and population sizes (N= 1000 or 10,000; values for N= 100 and M= 0.25, 0.5 are provided in Fig. S2). At both large and small values of s, the Y allele at the SD locus evolved close to the maximal possible value (e.g., 6.8 and 7.4 at N= 1000 and 10,000 respectively, for s= 0.8), with the X allele coevolving to values ensuring equal sex ratios (−2.7 and −2.8 respectively; Fig. 4A, right panel). At intermediate s values, by contrast, SD alleles took values closer to the threshold, allowing sex reversal and recombination to occur occasionally (Fig. S3). The shift in SD values (and increase in recombination rates) was stronger for weak sexually antagonistic selection (Fig. 4A, right panel, M= 0.75 vs. 0). The effect of recombination on offspring survival is best seen by comparing left with right panels in Figure 4C (pure GSD vs. evolvable SD): while female survival was only marginally affected, male survival was largely improved by sex reversal, mostly at low sexually antagonistic selection (M= 0.75).

Interestingly, both Y and X alleles converged toward the threshold (Fig. 4A, right panel), but the latter to a lower extent. As a result, sex-reversed XY females were twice as frequent as sex-reversed XX males (Fig. S3). These rates also depended on population size: After 100,000 generations, XX sex reversal occurred at rates 0.03, 2.8 × 10−4, and 5.5 × 10−5 for N= 100, 1000, and 10,000 respectively, as compared to 0.06, 4.8 × 10−4, and 11 × 10−5 for XY sex reversals (average over M and s values). These rates also decreased as sexually antagonistic selection increased (from M= 0.75 to M= 0, Fig. S3). Medians were even lower, due to the asymmetric distribution of equilibrium rates.

Outcomes were qualitatively similar when h was set to either 0.05 or 1, independent of s (performed for N= 1000 and 10,000, M= 0.75) and when the allelic values at the sex determiner were initially drawn from a uniform distribution (data not shown).


From our results, low rates of sex reversal and X-Y recombination allow purging Y chromosomes from the load of accumulating deleterious mutations (as also recently shown by Karhunen 2011 for a single sex-linked deleterious mutation). Moreover, our simulations suggest that the benefits from purging may outweigh the sexually antagonistic selection that normally drives recombination arrest, inducing evolutionarily stable rates of XY recombination. Although such rates were quite low under our settings, we recall that our simulations were not aimed at providing specific quantitative predictions. Rather we aimed at testing whether the processes investigated here have the potential to contribute to the patterns of sex chromosomes homomorphy widespread among cold-blooded vertebrates, and identifying key parameters influencing the system.


One crucial player in the game is the mutation model for deleterious mutations, including the selection and (for diploid genomes) dominance coefficients. We tested a range of s values (from 0 to 0.8) covering the full range of possible dynamics, from accumulation (at small s) to elimination (at high s). Mutations with large selection coefficients presumably have lower mutation rates than implemented here, but this is inconsequential given that they were eliminated anyway. We assumed s constant for any given set of simulations, to identify values that were critical in determining elimination or accumulation under our settings (e.g., Fig. 4). Similar simulation studies (e.g., Gordo and Campos 2008; Jaquiéry et al. 2009) found no qualitative effect of drawing s from exponential, log-normal or gamma distributions (rather than keeping it constant). Also in line with similar modeling studies (e.g., Higgins and Lynch 2001; Jaquiéry et al. 2009), we assumed a negative relationship between h and s, which conforms to empirical evidence (e.g., Deng and Lynch 1996; Phadnis and Fry 2005; Agrawal and Whitlock 2011). We furthermore also tested h values set to 0.05 or 1 (independent of s), with qualitatively similar results (data not shown).

The shift from accumulation to elimination occurred at lower s values for the X than for the Y chromosome (Fig. 4B), due to the threefold difference in effective size, and to the action of Muller's ratchet (both triggered by the absence of recombination). Slightly deleterious mutations (which might represent the most frequent category; Haddrill et al. 2010, reviewed in Lynch et al. 1999 and Eyre-Walker and Keightley 2007) thus accumulated much faster on the Y than on the X. This contrast would certainly have been even stronger if we had implemented sex differences in mating strategies (polygyny would further lower the Y effective size) or in mutation rates (higher in males; reviewed in Hedrick 2007). Our settings were thus conservative in this respect.

Our model focused on Muller's ratchet as the main cause of decay. Accordingly, deleterious mutations accumulated mostly by drift at individual loci subject to unidirectional mutations. More sophisticated models, comprising sequence information and reverse mutations (which may slow down Muller's ratchet; Kaiser and Charlesworth 2010), have been developed to investigate detailed interactions between the several processes likely to affect this decay (including selective sweeps and background selection), and to quantitatively estimate specific parameters for model species (e.g., Gordo and Charlesworth 2001; Gordo et al. 2002; Bachtrog and Gordo 2004; Bachtrog 2008; Kaiser and Charlesworth 2009, 2010). As emphasized before, our aim was not to provide quantitative estimates for any model species, but only to qualitatively investigate the interplay between deleterious mutations and sexually antagonistic selection in driving XY recombination.

Note, however, that the linkage between deleterious and sexually antagonistic mutations within the SDR also provides under our settings some potential for background selection and Hill–Robertson effects (which have been shown to play an important role in the decay for large nonrecombining regions; Kaiser and Charlesworth 2009). Also, contrasting with the above studies, we chose to simulate sexual diploids instead of asexual haploids, which might significantly affect the evolution of initially homomorphic sex chromosomes, because mutations accumulating on the X necessarily influence the dynamics on the Y (and vice versa).

Even though Muller's ratchet is mostly efficient at low effective population size, it may still significantly affect processes at larger sizes (in the order of hundreds of thousands), provided that deleterious effects are small enough (s << 0.02; Gordo and Charlesworth 2000a). We focused here on relatively small populations (N from 100 to 10,000), targeting vertebrate systems with small effective sizes and highly structured populations. In Hyla, effective population sizes have been shown to roughly correspond to breeding population sizes (Broquet et al. 2009), which, from our own experience with this species, fit well within the range investigated.

Population size affected our simulation outcomes partly through intrinsic demographic effects (e.g., leading to extinctions in small populations; Fig. S2), and partly in interaction with s, because drift strongly opposes selection at low N values (e.g., Kimura 1983). As a result, the shift from accumulation to elimination occurred at lower s values in larger populations (Fig. 4B, N= 1000 vs. 10,000).


The processes of sex chromosome decay simulated here may largely account for the sex-chromosome heteromorphy documented in mammals, where deleterious mutations accumulate on the nonrecombining Y, but are largely purged from the X (reviewed in Graves 2006). It is worth noting, however, that mammals also evolved specific mechanisms to counteract potential detrimental effects of such decay. These include dosage compensation, which makes up for the loss of gene activity on the Y chromosome (reviewed in Mank 2009), and YY gene conversion from palindromic sequences (Rozen et al. 2003; Skaletsky et al. 2003; Hughes et al. 2010), which counteracts the decay of genes essential for male fitness (Marais et al. 2010).

In absence of such adaptations, the accumulating mutation load is expected to severely impact male fitness. Fitness costs mostly accrue at intermediate s values (mildly deleterious mutations; Fig. 4C), because too-strongly deleterious mutations are eliminated, while too-weak effects are not deleterious enough to significantly impact male fitness. For the same reasons, mildly deleterious mutations impacting autosomes have been shown to have the strongest impact on extinction probability of populations (e.g., Gabriel et al. 1993, Higgins and Lynch 2001). In our simulations, the demographic effects induced by the accumulation of such mildly deleterious mutations were strong enough to drive extinctions at small population sizes (N= 100; Fig. S2).


As a result, incentives for X-Y recombination were highest at intermediate s values (mildly deleterious mutations, corresponding to a minimum in male fitness), in accordance with previous simulation studies that also showed a maximal advantage of sex and recombination at intermediate s values (Gordo and Campos 2008). Under our settings, X-Y recombination was realized via “demasculinization” of the Y, that is, selection for SD allelic values closer to the threshold (Fig. 4A).

Interestingly, the X alleles at this locus simultaneously converged toward the threshold (Fig. 4A, right panel), as a way to oppose sex reversal and recombination. Indeed, there is a clear conflict of interest between the Y and the X, the latter spending two-thirds of its life in females where it anyway recombines. The X chromosome has little benefits to recombine with the Y, because this increases its load of deleterious mutations (in addition to disrupting epistatic interactions with sexually antagonistic alleles). Hence, this upward shift in X values is to be interpreted as a way to counteract the downward shift in the Y alleles, lowering the probability that XY individuals develop into females.

Too strong an upward shift, however, would also be costly, because the fitness of sex-reversed XX males is lowered by their female-beneficial allele at the sexually antagonistic locus. Sexually antagonistic selection is indeed bound to play an important role in this context (which might account, e.g., for the lowered fitness in sex-reversed Triturus cristatus; Wallace et al. 1999). Weaker antagonisms (M= 0.75 vs. M= 0) thus allowed more sex reversal, and thereby better purging and higher male fitness (Figs. 4B, C, and S3).

As a net result of these differential and counteracting forces, sex-reversed XY females were twice as frequent as sex-reversed XX males (Fig. S3). As this figure also shows, the rates of sex reversal at T= 100,000 generations were very low under our settings. For N= 1000, for instance, we expect about one XY female every three to four generations (median), calculated for the s value (s= 0.0125) and sexually antagonistic selection (M= 0.75) at which sex reversal was maximal. Still lower rates are actually enough to purge large parts of the load, and presumably keep the sex chromosomes homomorphic.

If confirmed under other simulation settings, this rarity may explain why sex reversal is only rarely documented in the field (e.g., Crew 1921; Witschi 1929; Aida 1936; Kawamura and Nishioka 1977; Nagler et al. 2001; Matsuba et al. 2008). In the laboratory, it is easily triggered (as reviewed in Wallace et al. 1999; Eggert 2004; Ospina-Álvarez and Piferrer 2008), but usually occurs at extreme temperatures relative to specific natural ranges. In Tilapia (Oreochromis niloticus), for instance, feminization occurs below 20°C and masculinization above 32°C (Bezault et al. 2007; Baroiller et al. 2009). In the newt T. cristatus, feminization occurs below 16°C and masculinization above 24°C (Wallace and Wallace 2000). This should ensure that sex reversal only occurs at rare occasions.


The two main hypotheses proposed to account for the widespread sex-chromosome homomorphy among cold-blooded vertebrates (high turnover and XY recombination) both received empirical support (see section Introduction). Several mechanisms have been proposed for the former (reviewed in van Doorn and Kirkpatrick 2010). Alternative mechanisms also exist for XY recombination, namely male recombination and sex reversal (the fountain of youth; Perrin 2009), which we formalize here. As our results suggest, sex reversal, which occasionally occurs in natural populations of cold-blooded vertebrates, might indeed be more than the simple side effect of a physiological dysfunction, but actually optimized by natural selection as a way to purge deleterious mutations on the Y. In absence of male recombination, we thus expect sex reversal to evolve under the opposing forces of sexually antagonistic selection and deleterious load.

Occasional male recombination (the alternative mechanism) might certainly achieve similar results, but in different ways, inducing continuously low rates of X-Y recombination (as opposed to rare bursts of recombination generated by sex reversal). The proximate mechanisms controlling recombination are poorly known, and their evolution in this context would deserve similar investigations. However, it might also be the case that, if sex reversal occurs in ectotherms anyway (e.g., as a side effect of temperature dependence in sex determination), then the ensuing XY recombination in females and purging of deleterious mutations should suppress any selective pressure for XY to recombine in males.

Associate Editor: C. Peichel


We would like to thank S. van Doorn and an anonymous reviewer for valuable comments on a previous version of the manuscript. This work was supported by a PhD Fellowship to CG from the Faculty of Biology and Medicine (University of Lausanne), and by the Swiss National Science Foundation (grants 31003A-129894 to NP and 3100A0–138180/1 to J. Goudet).