Sex and differentiation: population genetic divergence and sexual dimorphism in Mexican goodeid fish


Mike Ritchie, Environmental & Evolutionary Biology, Dyers Brae House, University of St Andrews, St Andrews, Fife, KY16 9TH, UK.
Tel.: +44 1334 463495; fax: 01334463500; e-mail:


Genetic differentiation arises due to the interaction between natural and sexual selection, migration and genetic drift. A potential role of sexual selection in speciation has received much interest, although comparative studies are inconsistent in finding supporting evidence. A poorly tested prediction is that species subject to a higher intensity of sexual selection should show greater genetic differentiation amongst populations because females from these populations should be more choosy in mate choice. The Goodeinae is a group of endemic Mexican fishes in which female choice has driven some species to be morphologically sexually dimorphic, whereas others are relatively monomorphic. Here, we measured population divergence, using microsatellite loci, within four goodeid species which show contrasting levels of sexual dimorphism. We found higher levels of differentiation between populations of the more dimorphic species, implying less gene flow between populations. We also found evidence of higher levels of genetic differences between the sexes within populations of the dimorphic species, consistent with greater dispersal in males. Adjusted for geographic distance, the mean FST for the dimorphic species is 0.25 compared with 0.16 for the less dimorphic species. We conclude that population differentiation is accelerated in more sexually dimorphic species, and that comparative phylogeography may provide a more powerful approach to detecting processes, such as an influence of sexual selection on differentiation, than broad-scale comparative studies.


Genetic differentiation between populations arises due to the interaction between natural and sexual selection, migration and genetic drift. At neutral loci, the observed levels of population differentiation are usually assumed to represent an equilibrium between mutation, migration and genetic drift (Wright, 1931). Variation in levels of neutral intraspecific differentiation will therefore largely reflect differences in dispersal and population size, although population structure will also be influenced by species-specific responses to biogeographic history. Recent (Pleistocene and younger) changes in distributions will affect whether species are in drift–migration equilibrium, and species may vary in their approach to this equilibrium (Hewitt, 1999, 2004; Bailey et al., 2007). A potential role for sexual selection in population differentiation has received much recent interest. It has long been supposed that species in which sexual selection is strong might evolve sexual isolation more rapidly (Darwin, 1871; Lande, 1982; Gavrilets, 2000). Some detailed studies of closely related species support a role for sexual behaviour in divergence (e.g. Gray & Cade, 2000; Boake, 2005; Mendelson & Shaw, 2005). However, carefully controlled comparative studies of this are ambiguous (e.g. compare Arnqvist et al., 2000 with Gage et al., 2002), and it has been argued that this attractive idea lacks clear supporting evidence (Panhuis et al., 2001).

A rarely tested prediction of a potential role of sexual selection in speciation is that species subject to a greater intensity of sexual selection should show greater levels of differentiation amongst populations. Rapid evolution of signal–receiver communication systems will reduce gene flow between populations, for example, a novel song in populations of an Amazonian frog is associated with greater genetic differentiation between populations (Boul et al., 2007). Coevolutionary models of sexual selection (e.g. Lande, 1982) predict that divergence in sexual behaviours will lead to discrimination against males from other populations, reducing gene flow for a given level of dispersal. However, sexual conflict complicates these predictions. Where there is conflict over the outcome of mating encounters the ‘winning’ sex will disproportionately influence gene flow – for example, if males can counteract female mating preferences, gene flow may be greater than when female preferences are more important to the outcome of mating encounters (Parker & Partridge, 1998; Partridge & Parker, 1999), but other models predict that antagonistic coevolution will also lead to consistent discrimination against alien males (Gavrilets et al., 2001).

The Goodeinae is a group of endemic Mexican live-bearing fish consisting of around 36 extant species. Comparative evidence suggests that they have radiated since the Miocene with a predominantly allopatric mode of speciation, but with frequent range changes due to volcanism and river piracy (the capture of river drainages due to changes in course) (Webb et al., 2004; Ritchie et al., 2005). All goodeines are live-bearers, males lack a gonopodium and females lack sperm-storage organs, so sexual selection probably acts most strongly on premating behaviours. Males inseminate females via a modified anal fin following successful courtship, and this requires female cooperation (Bisazza, 1997), so the mating system tends to be dominated by female choice (although female preferences can sometimes be exploited by males, Macías Garcia & Ramirez, 2005). Male courtship includes species-specific display behaviour usually involving highly modified fins, which are sexually dimorphic in shape and often colourful (Fig. 1). In studies of several species in the group there is evidence of female preference for colour patterns (Moyaho et al., 2005), colourful fin structures (Macías Garcia & Ramirez, 2005), fin size and associated body shape (Macías Garcia et al., 1994) and courtship displays (Macías Garcia & Saborío, 2004). Levels of morphological sexual dimorphism vary greatly within the group (Ritchie et al., 2005), with some species showing little differentiation in body shape other than size differences, but in others males show very dimorphic fins. Detailed studies of the dimorphic species Girardinichthys multiradiatus have shown that enlarged male fins are a target of female preference and are a handicap to males as their exaggerated morphology leads to higher levels of predation by snakes (Macías Garcia et al., 1994, 1998). Furthermore, differences between populations in the extent of dimorphism and fin elaboration predict assortative mating in the laboratory (González Zuarth & Macías Garcia, 2006).

Figure 1.

 The four species sampled. Goodea atripinnis and Zoogonecticus quitzeoensis are examples of sexually monomorphic species, whereas Xenotoca melanosoma and Characodon lateralis are more dimorphic.

To examine patterns of speciation in the goodeines, Ritchie et al. (2005) quantified morphological measures of sexual dimorphism amongst 25 goodeine species and found that inferred speciation rates (from an mtDNA phylogeny) were not significantly greater in more dimorphic lineages (although they were somewhat increased). As a whole, the clade showed evidence of a Miocene radiation, which coincides with high levels of tectonic rearrangements in the Mesa Central and a subsequent slowdown in speciation rates. Here, we examine an alternative prediction of increased genetic differentiation between populations of more sexually dimorphic species. We measured population divergence, using microsatellite loci, within four goodeine species which were chosen because they show contrasting levels of sexual dimorphism. We sampled four populations of each species. We found higher levels of population differentiation (FST) in the more dimorphic species, implying lower gene flow between populations. We also found higher levels of sex-specific differentiation in one of the dimorphic species, consistent with greater dispersal in males.

Materials and methods


Fin clips were collected from anaesthetized fishes in the summers of 1999 and 2000. We obtained 15 individuals of each sex at four separate sites for each of four species (Fig. 1). As representatives of dimorphic species, we chose Xenotoca melanosoma (which shows extreme fin size dimorphism) and Characodon lateralis (which shows marked colour dimorphism). As monomorphic species, we chose Goodea atripinnis (monomorphic in colour, although males have larger dorsal fins than females) and Zoogeneticus quitzeoensis (relatively little colour or fin dimorphism). Utraviolet can influence mate choice in Goodeines (Macías Garcia & Burt de Perera, 2002). We only have data on UV reflectance for Z. quitzeoensis and C. lateralis and these species are not dimorphic for UV (C. Macías Garcia, pers. obs.). Distances used in the analyses were measured ‘as the crow flies’. This is a conservative measure, as it is more likely to underestimate barriers to dispersal across large heterogeneous areas such as those inhabited by the monomorphic species (G. atripinnis and Z. quitzeoensis), than within the smaller and more homogeneous watersheds occupied by the sexually dimorphic species (C. lateralis and X. melanosoma).

Although these species do not represent the extremes of sexual dimorphism within the Goodeinae, they do represent different levels of sexual dimorphism. Ritchie et al. (2005) quantified sexual dimorphism as the Mahalanobis distance between sexes in a multivariate morphological analysis, not involving colour, of 25 species (samples from one population of G. atripinnis were omitted as they included some females with an aberrant fin morphology). The more dimorphic species are where male morphology diverges from the female in traits, especially dorsal and ventral fins, which are usually targets of female mate preferences (Macías Garcia et al., 1994; González Zuarth & Macías Garcia, 2006). The morphological scores used did not take account of sexually dimorphic colours, which also influence mate choice (Macías Garcia & Burt de Perera, 2002; Moyaho et al., 2004; Macías Garcia & Ramirez, 2005), but colours were considered when choosing the four species to be analysed here. On morphology alone, G. atripinnis had a dimorphism Z-score of −0.446 and lay in the bottom 30% of the distribution, Z. quitzeoensis was less dimorphic (Z = −0.732, bottom 20%), C. lateralis was in the top 30% (Z = 0.472) and X. melanosoma (the most dimorphic species sampled here) was in the top 25% (Z = 0.715). Population samples were determined by availability and it was not possible to match the geographic distribution of samples due to the high degree of endemism of many species (Miller et al., 2005). The dimorphic species were sampled from a less widespread geographic region, due to the natural variation in the species’ ranges, and C. lateralis has a geographically disjunct range, although the distances between our samples were similar to those of X. melanosoma (Fig. 2). The range size differences are conservative for a hypothesized relationship between sexual dimorphism and population differentiation because the samples from the dimorphic species were from closer proximity than the monomorphic species. Fin samples were stored in 90% ethanol in the field for subsequent DNA extraction and analysis.

Figure 2.

 Sample locations. Geographic location of sampled populations, plus approximate distributions of each species.

Marker scoring and analysis

A series of microsatellite loci have been developed from various species in this group (Boto & Doadrio, 2003; Hamill et al., 2007) but none specifically for the four species studied here. Hamill et al. (2007) studied the cross-utility of these loci in the group as a whole. There is no strong evidence for ascertainment bias; genetic distance between source and target species does not predict overall amplification success or deviation from Hardy–Weinberg equilibrium (estimated as FIS). There is evidence that a greater distance does increase the likelihood of monomorphism, but not significantly. For this study, we successfully amplified five to seven unlinked loci (chosen for good cross-amplification) in all samples and scored variability on a CEQ 8000 capillary array system sequencer (Beckman Coulter, Fullerton, CA, USA). Some loci were monomorphic, resulting in a sample of five or six variable loci in each species. These were XC18, XC25, CA12, IW196 and Zt1.9 in C. lateralis, XC18, XC25, AS2, AS5 and IW196 (IW193 was monomorphic) in X. melanosoma, XC18, XC25, CA12, AS2, AS5 and Zt1.9 (IW196 was almost monomorphic) in Z. quitzeoensis and XC25, CA12, AS5, Zt1.9 and Zt1.6 (Zt1.2 was monomorphic) in G. atripinnis. Details of each locus, amplification parameters and scoring are as in Hamill et al. (2007).

Genepop (Raymond & Rousset, 1995) was used to carry out HW tests for each locus independently in each population, a total of 84 tests. After strict Bonferroni correction (P = 0.0006), there were five significant departures, all of heterozygote deficiency. Relaxing the critical alpha to 0.003 (0.05/16 populations) results in seven significant exceptions. Several included the XC18 locus, and excluding this from all species still resulted in five exceptions under a strict Bonferroni correction, and a final sample size of four or five loci per species. We conclude that these probably represent populations with moderate frequencies of null alleles; they were not seen across loci in populations, as would be expected from a Wahlund effect (the exceptions were each unique – i.e. only seen for one locus in one population), and were not more likely to be seen in either type of species. Patterns described in the results below are unchanged if we include or exclude XC18 or data for the individual exceptions to HW.

We used FST (also calculated with Genepop) to estimate population differentiation rather than RST because many of the loci do not conform to the step-wise model (several have imperfect repeats). Furthermore, simulation studies imply that FST is more conservative and superior when few loci are scored (Gaggiotti et al., 1999). Because sex-biased dispersal is an important component of the role of sexual selection in isolation, we tested for sex-specific levels of differentiation using FIS and assignment approaches (Goudet et al., 2002) implemented in FSTAT (Goudet, 1995) (P values are from 1000 replicates). These are one-tailed tests, with the expectation that males disperse more (Goudet et al., 2002).


Population differentiation

F ST increased with geographic distance for each species (Fig. 3). Mantel tests of isolation by distance were not significant for any species, but these are weak tests with only four populations sampled per species (six comparisons). Although the sexually dimorphic species were sampled from more close geographic localities, they show a greater degree of population differentiation when geographic distance is accounted for as a covariate. In a sequential general linear model across all data with distance as a covariate, mono- vs. dimorphism as a factor (dimorphism) and species nested within dimorphism, FST is dependent on dimorphism (F1,18 = 5.7, approx. P = 0.028). From the ancova, the least squares mean FST for dimorphic species is 0.25 compared with 0.16 for monomorphic species. The slopes of the regressions against distance are similar [F1,18 = 0.96, NS (P =0.34)] but the lines are displaced in the dimorphic species. Probability values from a linear model are potentially exaggerated because data points are not independent. A partial Mantel test (Manly, 1991), implemented in FSTAT, with geographic distance and dimorphism (dimorphic vs. monomorphic) as predictor variables gave partial correlation coefficients of 0.117 (P = 0.028 from 10 000 randomizations) and 0.475 (P =0.019) for distance and dimorphism respectively.

Figure 3.

 Relative genetic differentiation. FST vs. geographic distance for the four species. The dimorphic species are in black and the monomorphic in grey, the dashed regression lines go with the open points.

Sex-biased dispersal

We tested each species for sex bias in dispersal by analyses of FIS and assignment tests (Table 1). If one sex disperses more than the other, then a population will contain a mixture of immigrants and residents of the dispersing sex, whereas the other sex will be more uniform in genotype. This can be detected by a larger FIS in the dispersing sex. An alternate method of detecting biased dispersal is an assignment test where the probabilities of individuals of each sex belonging to the total group are compared. Both tests were significant for the sexually dimorphic C. lateralis (and the FIS test was marginal for X. melanosoma), but neither was for the monomorphic species.

Table 1.   Sex-biased dispersal.
Species Goodea atripinnis Zoogeneticus quitzeoensis Characodon lateralis Xenotoca melanosoma
  1. Tests for sex-biased dispersal for each of the goodeine species studied.

F IS (males)0.0460.0740.1320.008
F IS (females)0.0560.116−0.0890.006
P (FIS)NS (0.61)NS (0.79)0.0010.056
P (equal assignment)NS (0.60)NS (0.29)0.017NS (0.26)


We chose four species of goodeids that contrast in levels of sexual dimorphism and examined the extent of genetic differentiation amongst four populations of each. The more sexually dimorphic species showed greater genetic differentiation, with much larger values of FST than monomorphic species after adjustment for geographic distance, as predicted if stronger female-biased sexual selection promoted population differentiation. Sexual dimorphism in goodeines is a good proxy for the extent of female preference driven sexual selection (see Introduction), and there are few other obvious differences between the species studied here. These are fairly closely related, having diverged since the Miocene and largely share their biogeographic history, ecological niches and most aspects of reproductive biology (all are matrotrophic live-bearers). Phylogenetically, none are sibling species, and C. lateralis is something of an outlier amongst these species, lying in a distinct basal clade of the group (Webb et al., 2004). Characodon is also a geographic outlier (Fig. 2) and therefore may differ in other aspects of its biology, including biogeography, yet C. lateralis shows a pattern of variation similar to X. melanosoma (Fig. 3). There was a bias in the geographic sampling of the populations, with populations of monomorphic species typically coming from wider geographic areas than dimorphic species (the average distance between pairs of populations was 35 km for dimorphic species, vs. 175 km for monomorphic), which is a reflection of the size of the species’ geographic ranges, the monomorphic species are amongst the most widespread of all goodeids (together with Xenotoca variata).

Under a standard migration–drift equilibrium model, we would predict that isolation by distance, which contributes to overall differentiation amongst populations (Fig. 3), would have led to less differentiation within the more dimorphic species due to their closer proximity, yet this was not observed. Furthermore, most pairs of monomorphic populations occupied different watersheds whereas none of the dimorphic population pairs did, yet FST was less in the monomorphic species. It is surprising how little influence the occupation of different watersheds has on the levels of differentiation seen, although river piracy is known to occur in this region (Webb et al., 2004). There have been few studies of dispersal in goodeids but White & Turner (1985) found evidence for isolation by distance within a single lake in G. atripinnis so dispersal is limited. Unfortunately, there are no direct estimates of dispersal of goodeines available. The mean standard lengths of the species are 32.22 (mm) for C. lateralis, 49.9 for X. melanosoma, 76.28 for G. atripinnis and 29.56 for Z. quitzeoensis, so our results could not arise simply due to size-related differences in dispersal ability amongst species (the dimorphic species are the largest and smallest sampled). In general, smaller, lighter and more actively courting males would be expected to move more than females in fish species. There is evidence for this in studies of direct movement via tracking in guppies (Croft et al., 2003) or mark–recapture in salmonids (Hutchings & Gerber, 2002). Genetic studies support this in salmonids (Bekkevold et al., 2004) but not guppies (Russell et al., 2004), at least over the relatively small geographic scale over which guppy schools were sampled. We have genetic evidence for male-biased dispersal in C. lateralis, where males show greater FIS and are less likely to be correctly placed to population by assignment tests (Table 1). There is also a greater FIS amongst males in X. melanosoma, although this is not quite significant.

Our tests for biased dispersal imply greater movement in males of the two sexually dimorphic species, yet these are the species with relatively greater genetic differentiation, or less gene flow, between populations. Our results suggest that males are moving between populations of the dimorphic species, but reproducing less successfully. This supports the mechanism by which sexual isolation can influence speciation – faster signal-preference trait coevolution. If male traits have diverged more, and females are more choosy, then effective gene flow will be less. Game theory models of sexual conflict have shown that this can be countered if males are ‘ahead’ in sexual conflict and can overcome female preferences (Parker & Partridge, 1998; Partridge & Parker, 1999). In goodeines, copulation requires female cooperation, so female preference probably predominates the mating system, and differences in the degree of sexual dimorphism and fin shape predict assortative mating between populations of a related species (González Zuarth & Macías Garcia, 2006). Males can exploit female preferences but this is not stable, with female preference ultimately dominating the mating system within these species (Macías Garcia & Ramirez, 2005). In other species, comparative studies of water striders shows that male–female antagonistic evolution is dynamic with either gender apparently taking the lead in the evolution of genitalia in this group, the mating system of which superficially seems to be dominated by males (Arnqvist & Rowe, 2002). Recent studies of the melanogaster group of Drosophila genitalia imply male coercion dominates this group (Jagadeeshan & Singh, 2006). The mating system of dung flies seems to be dominated by male coercion, but under experimental evolution, females from populations in which sexual conflict was more effective evolved greater discrimination against males from other populations (Martin & Hosken, 2003). Hence, there is comparative and experimental evidence that the evolution of female preference gives rise to discrimination against foreign males, which may be causing the more limited gene flow between the more dimorphic species studied here

Ritchie et al. (2005) used a comparative approach to assess the role of sexual selection in speciation across the whole Goodeinae. They found that the time to speciation, inferred from branching points in a molecular phylogeny, were somewhat shorter in more sexually dimorphic species, but this was not statistically significant. Here we have found that genetic differentiation between populations is greater in more dimorphic species. It might well be the case that studies of differentiation within or between very closely related species is a more powerful approach to detecting an influence of sexual selection on diversification (although it could also be that Ritchie et al.’s failure to incorporate colour into their measure of dimorphism compromised this analysis). Over a large taxonomic group differential extinction can confound detecting patterns of evolution, for example, more dimorphic species may be more likely to go extinct (Doherty et al., 2003), acting against the likelihood of detecting more species in more dimorphic clades. Comparative studies have been inconsistent in their ability to detect a role for sexual selection (Morrow et al., 2003). Comparative phylogeography may be a more powerful approach allowing the detection of species-specific effects superimposed on a relatively shared phylogeographic history. Detailed analyses of both behavioural and genetic variation within species or species complexes are relatively rare (e.g.Tilley et al., 1990; Mendelson & Shaw, 2005; Boul et al., 2007) yet potentially provide greater resolution.

We conclude that population differentiation is accelerated in more sexually dimorphic species, and that comparative phylogeography may provide a more powerful approach to detecting processes such as an influence of sexual selection on differentiation than broad-scale comparative studies.


Fieldwork in Mexico was greatly aided by E. Avila Luna, J. M. Artigas Azas, E. Smart and O. Dominguez. The work was funded by the NERC, UK. Dave Kidd and Emma Smart helped at various stages of the project and Tanya Sneddon helped in the laboratory.