A pattern of greater divergence in mating traits between sister-species pairs with overlapping ranges than between allopatric species pairs is expected if reinforcement commonly contributes to speciation. Few large-scale comparative analyses have addressed this prediction, especially for genital form. Here, we show that penial morphology follows the predicted pattern in 40 robustly identified sister-species pairs in the marine gastropod subfamily Littorininae. Further work is needed to exclude other processes that may contribute to genital divergence between sympatric species, but the clear pattern we observe strongly suggests a role for genital form in reproductive isolation in this large clade.
When genetically differentiated populations are in contact, selection may operate against interbreeding either because inter-population matings are directly costly or because the resulting hybrid offspring have reduced fitness. This may result in increased prezygotic reproductive isolation between populations and, perhaps, lead to speciation (Dobzhansky, 1937; Servedio & Noor, 2003). One possible signature of this process of reinforcement is reproductive character displacement (RCD): a pattern of greater divergence in sexual traits between sympatric than between allopatric populations of a pair of interacting species (e.g. Höbel & Gerhardt, 2003). A comparable difference between sister-species pairs with overlapping ranges (which we refer to as ‘sympatric’ pairs, for simplicity) and allopatric sister-species pairs has also been used as evidence for reinforcement, most notably in Drosophila (Coyne & Orr, 1989; Coyne & Orr, 1997), and has the advantage that it can be applied to large clades to ask how commonly the pattern occurs. However, these patterns may have several possible origins. They may evolve through reinforcement during speciation, where selection favours divergence in mating signals and preferences that act to increase reproductive isolation in the area of range overlap. They may evolve after speciation, where reproductive interference and direct costs of interspecific matings may produce similar divergence. It is also possible for the pattern to be generated without selection specifically acting against inter-population matings, for example, if range overlap on secondary contact is only possible between populations with divergent mate recognition systems (Noor, 2003; Servedio & Noor, 2003).
Strong evidence for the pattern of RCD in individual taxa requires comparison of multiple populations in both allopatry and sympatry. Good individual examples exist, suggesting that natural selection on mate recognition traits can produce divergence between hybridizing populations or sister species (e.g. Noor, 1995; Sætre et al., 1997; Rundle & Schluter, 1998; Higgie et al., 2000; Cooley et al., 2001; Jiggins et al., 2001; Höbel & Gerhardt, 2003; Smadja & Ganem, 2005). However, large comparative studies are needed to assess the frequency with which interactions between co-occurring sister species contribute to the evolution of mate recognition and potentially to speciation. Ideally, such studies should examine many sister-species pairs, comparing allopatric pairs to sympatric pairs. Few such comparisons have been conducted, but the classic study of Drosophila (Coyne & Orr, 1989; Coyne & Orr, 1997), a recent study of the plant genus Mimulus (Grossenbacher & Whittall, 2011) and a small number of other studies (Van Der Niet et al., 2006; Lc Gac & Giraud, 2008) have revealed a pattern of greater divergence or assortative mating between sympatric sister-species pairs, whereas a study of the land snail genus Xerocrassa did not (Sauer & Hausdorf, 2009). It is difficult to be sure whether this pattern is due to reinforcement or other processes. In the Drosophila case, the fact that only prezygotic isolation and not post-zygotic isolation is enhanced in sympatry argues strongly in favour of reinforcement (Coyne & Orr, 1989; Coyne & Orr, 1997), as do concordant asymmetries in pre- and post-zygotic isolation between reciprocal crosses (Yukilevich, 2012). Nevertheless, revealing the pattern is an important first step, which can be followed by investigations of process.
Genital form is an essential character for identification of species in many animal taxa. The species-specific forms of male genitalia demonstrate that they evolve rapidly (Eberhard, 1985) and suggest that they are important characters in maintaining reproductive isolation among closely related species (Dufour, 1844). Rapid genital divergence may be the result of post-copulatory sexual selection within populations (Arnqvist, 1998; Arnqvist et al., 2000), and this may lead to speciation as an incidental consequence of divergence. However, genital divergence may have a more direct role in speciation, through reproductive interference or reinforcement. As an example of reproductive interference, in beetles (Ohomopterus), it has been demonstrated that differences in genital characteristics impose a direct cost for heterospecific mating, with the consequence that selection favours the evolution of mate choice in areas of sympatry (Sota & Kubota, 1998; Usami et al., 2006). As genital form commonly influences pre- or post-copulatory mate choice, it could alternatively be subject to reinforcement, potentially contributing to its rapid evolution and species specificity. Either reinforcement or reproductive interference between sister species are predicted to generate RCD and the large-scale pattern of greater divergence between sympatric sister-species pairs, but to date no large-scale comparative study based on a robust phylogeny has tested for this pattern in animal genitalia.
In the marine gastropod family Littorinidae, the form of the elaborate penis is often species specific, leading to speculation about its role in precopulatory recognition (Reid, 1986, 1996). The molecular phylogeny, taxonomy, geographical distributions and ages of divergence of a worldwide clade of 152 littorinid species (subfamily Littorininae) are known in detail (97% complete; Reid et al., 2012). This is, therefore, an ideal model system for a large-scale test of the prediction that interaction in areas of range overlap results in greater genital divergence in sympatric than in allopatric sister-species pairs.
Materials and methods
We selected all robustly identified (Bayesian posterior probability = 100% for all but four cases) sister-species pairs from molecular phylogenies of littorinid genera with complete (or almost so) species sampling (Table S1); unresolved polychotomies and deeper divergences within the phylogeny were excluded, so that no species was used more than once. Range overlap between species in these pairs was determined from published distribution maps and records (see Table S1). We followed the same criteria as Coyne & Orr (1989) in order to classify sister-species pairs: a pair was identified as sympatric if both species were collected from the same locality and designated allopatric if they showed no such range overlap. The distributions of littoral organisms that occur on continental margins or island archipelagos are not well represented by area polygons. Detailed locality records for littorinids are available worldwide (e.g. Reid, 1986, 1996, 1999, 2007). Proportional overlap was therefore calculated as the number of localities at which both species occurred in sympatry, divided by the number of localities from which the more narrowly distributed species was recorded, expressed as a percentage (as done by Williams & Reid, 2004).
We selected 18 traits that could be scored reliably on the highly variable male genitalia of littorinids (Fig. 1, Table S2). These traits were then scored from camera lucida drawings made by DGR for five individuals from each of the 80 chosen species, using individuals from two to five different locations (apart from seven species where only a single location was available and two species where only three individuals could be scored). Shell height was also measured as a proxy for overall size.
The trait matrix was used to construct a similarity matrix between individuals, which was then reduced to a between-species similarity matrix by taking means over individual comparisons. The similarity metric used was Euclidean for continuously variable traits, ‘ecological’ for counts or continuous traits where zero represents absence, simple matching for unordered categorical traits and Jaccard for presence/absence traits (see Table S2 for details). Similarities were calculated for four data sets: with either zero or missing value for traits EFGHPQR and with the observed data for continuous variables replaced by the residuals from a regression on shell height. Unless otherwise stated, the results are presented for the data set with zeros and after size correction. All calculations were made in genstat v.10 (VSN International, Hemel Hempstead, UK).
To account for the time since separation of sister species, we used the age of the relevant node estimated by Reid et al. (2012). This measure was available for 38 of the 40 sister-species pairs (Table S1).
Results and Discussion
Molecular phylogenies for the littorinid genera Littorina, Littoraria, Austrolittorina, Tectarius, Afrolittorina, Echinolittorina, Peasiella and Mainwaringia (Reid & Geller, 1997; Williams et al., 2003; Williams & Reid, 2004; Reid et al., 2006; Williams & Duda, 2008; Reid et al., 2010, 2012) reveal 40 strongly supported sister-species pairs. Sixteen of these species pairs have overlapping geographical distributions (Table S1). The overall similarity in penial morphology was significantly greater for nonoverlapping than for overlapping pairs (Mann–Whitney U = 84.0, P = 0.001). Overall similarity was significantly correlated with the extent of range overlap (Spearman r = −0.37, P = 0.008; Fig. 2), although the correlation was not significant when species pairs with overlapping ranges were considered alone (r = 0.03). The pattern was not dependent on any single penial trait from the 18 traits analysed (leaving out one trait at a time, −0.41 < r < −0.31). It was not dependent on whether absence of a character was treated as a zero (r = −0.37) or as a missing value (r = −0.36), or on whether characters were corrected for shell size (r = −0.35 when size correction was removed for analyses with either zeroes or missing values). We conclude that there is a robust pattern of greater divergence in the form of the penis between sister-species pairs with overlapping ranges than between pairs with allopatric ranges.
Four species pairs, all with overlapping distributions, show the most divergent penis morphologies (similarity < 0.8). Three of these belong to the genus Littoraria and one to Littorina, genera that show the most diverse penial shapes in the subfamily (Reid, 1986, 1996). All the sister-species pairs used here are closely similar in overall morphology, and these four pairs do not stand out as any more divergent than the others in, for example, shell form (see Supporting information and references therein). Penial similarity was not significantly correlated with age since divergence (r = −0.12; Fig. S1), and so the relationship between similarity and range overlap was barely affected when controlled for age using Spearman partial correlation (r = −0.42). The correlation remained when a species pair with an unusually large genetic distance (Mainwaringia rhizophila/leithii; no age estimate available) was removed (r = −0.39). Rates of both molecular and morphological evolution are extremely high in the genus Mainwaringia and believed to be connected with its unique hermaphrodite reproduction (Reid et al., 2012).
The geographical distributions of most littorinid species are well characterized but, nevertheless, it is possible that range overlap does occur for the ten species pairs that occupy the same coastline but have never been recorded in sympatry (Table S1). Alternatively, range overlap between these pairs may have occurred in the past. Therefore, we compared penial similarity between three groups: those on different coastlines, those on the same coastline (i.e. contiguous coastline or islands within known pelagic larval dispersal distance of 2100 km; Reid, 2007) but with no observed overlap and those species pairs with observed range overlap. Median similarities differed between these groups (Kruskal–Wallis H = 8.89, P = 0.012: median similarities 0.951, 0.960 and 0.876, respectively), but the ‘potential overlap’ and ‘no overlap’ groups did not differ significantly (Mann–Whitney U = 70.0, P = 0.50).
The correlation between range overlap and age of divergence between sister-species pairs was not significant (r = 0.05, P = 0.78). A positive relationship would have suggested that speciation was typically allopatric in this family (Barraclough & Vogler, 2000), implying that range overlap between sister pairs is a result of secondary contact, which has not yet been achieved in the potential overlap group. Allopatric speciation does, nevertheless, appear to be the dominant mode of divergence in tropical littorinids with planktotrophic development, which make up the majority of the family (Williams & Reid, 2004; Reid et al., 2010, 2012). Therefore, we can speculate that the observed genital divergence may have occurred after secondary contact as a result of either reinforcement or reproductive interference between species, depending on whether reproductive isolation was complete at the time of contact or not (Butlin, 1987). Other mechanisms are also possible (Noor, 2003), particularly the filtering process originally suggested by Templeton (1981), where range overlap is only possible between populations that happen to be sufficiently diverged in mating signals to avoid costly hybridization. This process cannot be excluded at present, but it does predict both variation in genital form among populations within species and abutting ranges for species with very similar genitalia. In fact, the ranges of sister species are not known to abut; they are either separated by clear biogeographical boundaries (e.g. lack of habitat or open water more than 2100 km in width, the maximum dispersal distance) or they overlap (Williams & Reid, 2004; Reid et al., 2010, 2012). Intraspecific variation in penial form has been stated to be low (e.g. Reid, 1986, 1996, 2007), but is the subject of continuing investigation.
The selection pressure favouring divergence in penial morphology between sympatric populations (or species) could have arisen from direct costs of inter-population mating (Servedio, 2001), or through time-wasting or damage to the genitalia of one or the other partner, as in Ohomopterus beetles (Sota & Kubota, 1998; Usami et al., 2006). This would require some initial divergence before contact. Under the alternative scenario of reinforcement, if hybrids were less fit than parental genotypes, selection might have favoured divergence in penial morphology through the potential function of the penis in mate choice. Females may either reject males before sperm transfer or subsequently reject their sperm. This may depend on physical features of the penis or chemical signals produced by its glandular structures. Frequent remating occurs in female Littorina saxatilis (Panova et al., 2010). If this is common in littorinids, there would be ample opportunity for genitalia to evolve in this way (Arnqvist et al., 2000).
We divided the penial traits into four groups related to overall shape, branching, sperm-duct form and glandular features, then compared similarity based on each of these sets of traits between the species pairs with overlapping and nonoverlapping ranges. The shape (r = −0.41, P = 0.004) and glandular feature (r = −0.37, P = 0.011) sets showed significant effects, whereas the branch (r = −0.08, P = 0.32) and sperm duct (r = −0.13, P = 0.20) sets did not. In Littorinidae, the mucous glands that are often present on the base of the penis, sometimes on a short lateral branch, are believed to provide adhesion of the organ at copulation (Reid, 1986; Buckland-Nicks & Worthen, 1992; Reid, 1996). However, we know little about how the penis is placed within the female during copulation, or the function of any secretions from the glandular structures. Our results suggest that proper placement and perhaps chemical signalling could contribute to mate choice in littorinids.
Greater divergence in genital form in sympatry than in allopatry has been observed in other systems, including stag beetles (Kawano, 2003), land snails (Kameda et al., 2009), grasshoppers (Kawakami & Tatsuta, 2010) and millipedes (Sota & Tanabe, 2010), but not nephilid spiders (Kuntner et al., 2009). However, these studies relied on single species pairs, less robust taxonomic resolutions or were not limited to pairs of strict sister species. For example, Kameda et al. (2009) studied two species of the genus Satsuma that are sister species and found strong evidence of RCD. Sota & Tanabe (2010) analysed genital divergence within a large species complex of millipedes, but the overlapping species pairs examined were not sister taxa. Divergence between nonsister-species pairs may be important for genital evolution but is more likely to be due to reproductive interference than to reinforcement. A study including several sister-species pairs was conducted on the nephilid phylogeny, but the authors found no correlation between genital complexity and geographical distribution (Kuntner et al., 2009). A study of a small sample of littorinid sister species (genus Littorina) did not reveal a consistent pattern either (Reid, 1996). An early study on grasshoppers (Cohn & Cantrall, 1974) suggested greater divergence in genital form in sympatric than in allopatric species, but this analysis lacked a robust molecular phylogeny for the species under study. We are aware of only one comparative analysis testing for a general pattern, comparable to our study but on a much smaller clade. Sauer & Hausdorf (2009) used a phylogeny of 11 species in the land snail genus Xerocrassa in Crete to examine potential causes of rapid genital evolution. They found no evidence for character displacement, but suggested that sexual selection was the principal cause of divergence. Thus, our study is the first to use a large-scale comparative approach to demonstrate a widespread pattern of greater divergence in genital form in sympatric than in allopatric sister-species pairs. As noted above, there is compelling evidence that the primary mode of speciation in littorinids is allopatric, yet there is evidence for natural hybridization in several species pairs [e.g. Littorina arcana/saxatilis (Mikhailova et al., 2009); Echinolittorina reticulata/millegrana (Reid et al., 2006); Littoraria zebra/variegata (Reid, 1999)], possibly after secondary contact. If this hybridization results in gene flow, the case for reinforcement would be strengthened in these examples, but reproductive interference may still be the mechanism in other species pairs where reproductive isolation was complete on secondary contact. Detailed study of individual cases will be required to make these distinctions and also to understand the signalling role of the penis during mating.
JH was supported by a Marie Curie fellowship from the European Union and Professor Christer Brönmark. CS was supported by a Marie Curie fellowship from the European Union and by CNRS. RKB was supported by NERC.