We investigated the importance of sexual selection in facilitating speciation in a land snail radiation on Crete. We used differences in the genitalia of the Cretan Xerocrassa species as potential indices of sexual selection. First, we rejected the hypothesis that differences in the genitalia of the Xerocrassa species can be explained by genetic drift using coalescent simulations based on a mitochondrial gene tree. Second, we showed that there is no evidence for the hypothesis that the differences in the genitalia can be explained by natural selection against hybrids under the assumption that this is more likely in geographically overlapping species pairs and clades. Third, we showed that there is a positive scaling between male spermatophore-producing organs and female spermatophore-receiving organs indicating sexual coevolution. The spermatophore enables the sperm to escape from the female gametolytic organ. Thus, the coevolution might be a consequence of sexual conflict or cryptic female choice. Finally, we showed that the evolution of differences in the length of the flagellum that forms the tail of the spermatophore is concentrated toward the tips of the tree indicating that it is involved in speciation. If speciation is facilitated by sexual selection, niches may remain conserved and nonadaptive radiation may result.
We investigated whether sexual selection was associated with the radiation of the hermaphroditic land snail genus Xerocrassa on Crete (Hausdorf and Sauer 2009). There are 11 Xerocrassa species on Crete, 10 of which are endemic. The Xerocrassa species are xerophilic and live in open, dry habitats. During the summer they aestivate under stones and bushes or, more rarely, attached to the vegetation. At the beginning of the rainy season at the end of September or the beginning of October they became active and mate. At the end of the winter, the adults usually die. The Xerocrassa species feed on decaying plants. Differential adaptation of the Cretan species to different habitats could not be ascertained. Thus, the Xerocrassa radiation might be a nonadaptive radiation as has also been suggested for other land snail radiations on Crete (Gittenberger 1991; Parmakelis et al. 2005). Most Cretan Xerocrassa species can be distinguished by characters of the genitalia. Differences in the genitalia are often a product of sexual selection (Eberhard 1985; Arnqvist 1998; Eberhard 2001; Hosken and Stockley 2004). Some recent studies indicate that the evolution of the genitalia is also strongly influenced by sexual selection and male–female counter-adaptation in hermaphroditic gastropods (Koene and Schulenburg 2005; Beese et al. 2006, 2009; Anthes et al. 2008). The lack of adaptation to different niches and the fact that several of the Cretan Xerocrassa species can be distinguished only by characters of the genitalia imply that sexual selection might have driven the Xerocrassa radiation on Crete.
The genitalia of Xerocrassa are shown in Figure 1. Mating lasts 1–2 h. During mating, the sperm are transported from the vesicula seminalis in the hermaphroditic duct, where it is stored, through the spermoviduct and the vas deferens into the epiphallus. The epiphallus forms the broader anterior part of the spermatophore (Fig. 2) containing most of the sperm. The narrow tail of the spermatophore is formed by the blind ending flagellum. Both, sperm container and tail are furnished with hook-like structures that point toward the anterior end so that they impede and delay the transfer of the spermatophore. Copulation is reciprocal and after simultaneous intromission of the penis into the vagina of the mating partner each snail transfers the spermatophore into the partner's bursa copulatrix, the female gametolytic organ. In the bursa copulatrix, the spermatophore of the mating partner with the vast majority of sperm (99.98% in Cornu aspersum; Rogers and Chase 2001) is digested. Sperm have to actively swim out via the tail of the spermatophore to avoid digestion (Lind 1973). Sperm are most successful at reaching the spermathecae when the tail of the spermatophore is protruding into the vagina. Thus, a lengthening of the flagellum might increase paternity success. Koene and Schulenburg (2005) found correlations between the length of the flagellum and the spermatophore-receiving organ in helicoid land snails indicating coevolution probably as a result of counter-adaptation between male and female reproductive organs that should secure control over fertilization. Moreover, most helicoid land snails possess a dart apparatus. In C. aspersum, it has been shown that the donor of the spermatophore can influence the partner's bursa copulatrix with an allohormone from the mucus glands of the dart apparatus to enhance paternity (Koene and Chase 1998; Chase and Blanchard 2006). The allohormone reconfigures the spermatophore-receiving organ of the mating partner in such a manner as to allow more of the donated sperm to escape digestion, precede to the spermathecal storage sacs, and fertilize eggs. Most helicoid land snails can transfer the mucus produced by the glands of the dart apparatus with a calcareous dart into the hemolymph of the mating partner. The Xerocrassa species possess a vestigial dart apparatus with glands, but without darts. It is not clear whether the snails apply the substance produced by the mucus glands onto the body surface or directly onto the inner surface of the copulatory organs of the partner during mating.
We investigated whether the differences in the genitalia of the Cretan Xerocrassa species are the result of sexual selection and whether the changes in the genitalia were associated with speciation. First, we tried to exclude two alternative causes of changes in the genitalia, namely genetic drift and natural selection against hybrids. We tested whether differences in the genitalia of the Xerocrassa species can be explained by genetic drift using coalescent simulations based on a mitochondrial gene tree. We tested whether the differences in the genitalia can be explained by natural selection against hybrids under the assumption that this is more likely in geographically overlapping species pairs and clades. Then, we investigated whether changes in male spermatophore-producing organs and female spermatophore-receiving organs are correlated indicating sexual coevolution. Finally, we investigated whether the changes in the genitalia facilitated speciation using randomizations of phylogenetic independent contrasts across the species tree.
Materials and Methods
Xerocrassa specimens were sampled at about 500 localities across Crete in July/August and September/October 2004 and September/October 2005. Mitochondrial cytochrom oxidase subunit 1 (cox1/COI) sequences were determined for 122 Xerocrassa specimens covering all 11 species distinguished by Hausdorf and Sauer (2009) and all regions of Crete and two Trochoidea species used as outgroups.
Measurements of the shell and the parts of the genitalia of the Cretan Xerocrassa species were taken from Hausdorf and Sauer (2009). In addition, the bursa copulatrix has been measured in a subsample (n= 63). The small “gradilis” form of Xerocrassa cretica has not been considered in the calculations, because its distribution on Crete is restricted to a few square kilometers whereas the typical large form is distributed across the whole island.
Shell measurements were taken from digital photographs using the program analySIS Pro version 3.2 (Olympus Soft Imaging Solutions, Münster, Germany) or with an ocular micrometer. The measurements of parts of the genitalia were taken with an ocular micrometer and were usually repeated once. Penis, epiphallus, flagellum, vagina, and bursa copulatrix (Fig. 1) were used for the analyses, because they are copulatory organs or organs that are involved in spermatophore production or spermatophore uptake and might be affected by sexual selection and because species-specific differences were observed in these parts (Hausdorf and Sauer 2009).
Body size has been measured as shell volume calculated as the mean of the volume of a cone and of a cylinder based on the small diameter of the shell and shell height (1/2 × (1/4 π× small diameter2× height + 1/12 π× small diameter2× height)), the best approximation for shell volume according to J. Heller (pers. comm.).
Usually, total genomic DNA was extracted from tissue samples of the foot preserved in 100% isopropanol following the protocol proposed by Sokolov (2000) with slight modifications. Tissue samples were minced and incubated in 1 mL lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM EDTA, 1% sodium dodecyl sulphate, 0.24 mg/mL proteinase K) at 56°C for 120 min or until complete digestion. Then, 100 μL saturated KCl was added to the lysate, gently mixed, and incubated for 5 min on ice. The samples were centrifuged at 13,000 rpm for 15 min. The supernatant was brought to a clean tube and 1 mL of ice-cold ethanol (100%) and 50 μL 3M sodium acetate were added. Precipitation of DNA took place at −20°C overnight. The samples were centrifuged at 13,000 rpm for 15 min and the pellet was washed in 70% ethanol and air-dried. The pellet was resuspended in 80 μL of Tris-HCl pH 8.7. In some cases tissue samples were minced and incubated in a Chelex extraction solution (5% Chelex 100 (Bio-Rad, München, Germany), 5 mM DTT and 40 μg/mL proteinase K), for 60–90 min at 56°C, centrifuged at a low speed and the supernatant was directly used for a polymerase chain reaction (PCR).
DNA AMPLIFICATION AND SEQUENCING
Fragments of the cytochrom oxidase subunit 1 (cox1) gene were amplified using a PCR with the following primers LCO1490 5′-GGTCAACAAATCATAAAGATATTGG-3′ and HCO2198 5′-TACTTCAGGGTGACCAAAAAATCA-3′ (modified from Folmer et al. 1994). Amplifications were performed in 25 μL volumes containing 2.5 μL 10× amplification buffer (peqlab, Erlangen, Germany), 2.5 mM MgCl2, 0.2 mM each dNTP (peqlab, Erlangen, Germany), 1μL of each primer (10 pmol), 1.5 units Taq DNA polymerase (peqlab, Erlangen, Germany), 2.5 μL enhancer solution P (peqlab), and template DNA (usually 2.5 μL undiluted DNA extract). The cox1 fragments were amplified under the following conditions: an initial denaturing at 94°C for 2 min, 35 cycles of PCR (94°C for 30 s, 50°C for 30 s, 72°C for 30 s), and a final extension at 72°C for 5 min. A negative control (no template) was included in each amplification run. The PCR products were purified using QIAquick PCR purification Kit (Qiagen, Hilden, Germany). Both strands of the amplified fragments were directly cycle-sequenced using the amplification primers and a DNA Sequencing Kit (Applied Biosystems, Darmstadt, Germany) and electrophoresed with an automated DNA sequencer.
Forward and reverse sequences were assembled using ChromasPro version 1.33 (Technelysium, Tewantin, Australia). The sequences were aligned with the CLUSTAL W algorithm (Thompson et al. 1994) as implemented in MEGA version 4.0 (Tamura et al. 2007) with the default settings. The sequences analyzed in this article have been deposited in GenBank under the accession numbers FJ627054-FJ627177. The used alignment is available at TreeBASE (http://www.treebase.org, accession number S2413).
Models of sequence evolution for the maximum-likelihood analyses were chosen using ModelTest version 3.7 (Posada and Crandall 1998) based on the Akaike information criterion (AIC). Maximum-likelihood analyses were conducted with Treefinder (Jobb et al. 2004; Jobb 2007). Confidence values were computed by bootstrapping (100 replications; Felsenstein 1985).
An ultrametric species tree is necessary for some of the following analyses for which relative datings of the speciation events are required. The species tree has been constructed from the cox1 gene tree using a modification of the shallowest divergence clustering method (Maddison and Knowles 2006) that is based on the observation that the order of interspecific coalescences provides a high probability of consistency with the actual species history (Takahata 1989). We determined the shallowest interspecific divergences in an ultrametric gene tree to estimate the order of interspecific coalescences and to get relative datings of these events. The ultrametric tree has been constructed using the penalized-likelihood method with the truncated Newton algorithm implemented in r8s version 1.71 (Sanderson 2002). Zero-length branches had to be collapsed prior to the estimation of the ultrametric tree with r8s. The deepest node has been arbitrarily calibrated with 1.
TEST FOR DIFFERENTIATION BY GENETIC DRIFT
We used the approach proposed by Masta and Maddison (2002) to test whether the divergence of traits can be explained by genetic drift or whether it must have been accelerated by selection. The probability of fixation of phenotypic traits under genetic drift is estimated using coalescent-based simulations of neutrally evolving mitochondrial haplotypes. For this test, we did not consider X. cretica and the species pair X. subvariegata–Xerocrassa grabusana. These taxa diverged much earlier than the other species (Figs. 3 and 4) so that they do not provide power in distinguishing between drift and selectively driven divergence, because species that diverged more than 4N generations (where N is the population size of each sister taxon) ago are expected to be monophyletic with high probability (Neigel and Avise 1986) irrespective of whether selection was involved. All other Cretan Xerocrassa diverged in a comparatively short period of time. All Cretan Xerocrassa species except the species pair X. amphiconus–X. siderensis that cannot be distinguished by genital characters show fixed differences in genital characters (Hausdorf and Sauer 2009). Thus, we considered the species pair X. amphiconus–X. siderensis to be one unit for this test.
The method assumes that nuclear genes code for the phenotypic characters of interest and tests whether the fixation of phenotypes is likely to occur under the assumption of genetic drift. The approach consists of five steps: (1) it is shown that the lack of monophyly of the morphologically defined species units in the mitochondrial gene tree is consistent with a pattern of incomplete lineage sorting and not the result of a poor resolution of the mitochondrial gene tree, (2) it is shown that the used cox1 haplotypes evolve neutrally, (3) a measure of incomplete lineage sorting among species, namely s of Slatkin and Maddison (1989), is calculated for the reconstructed mitochondrial gene tree, (4) this s is compared against s values from simulated gene trees assuming a star phylogeny to estimate time since divergence, and (5) nuclear gene trees in populations with the estimated divergence times are simulated to determine the probability of fixation of differences in nuclear-encoded phenotypic characters under the assumption of neutrality.
To show that the lack of monophyly of the morphologically defined species in the mitochondrial gene tree is not the result of a poor resolution of the mitochondrial gene tree, we calculated the maximum-likelihood tree under the constraint that the sequences of each recognized species form a clade using the “resolve multifurcations” option of Treefinder. Then, we investigated whether this tree can be rejected in comparison with the unconstrained maximum-likelihood tree by applying the approximately unbiased test (Shimodaira 2002) implemented in Treefinder.
Second, we used the McDonald and Kreitman (1991) test implemented in DnaSP version 4.10.9 (Rozas et al. 2003) to determine whether the cox1 haplotypes evolve neutrally. We compared the ratio of nonsynonymous to synonymous polymorphisms in cox1 within the seven recently diverged Cretan Xerocrassa species used for the genetic drift analysis and within the three most basal Xerocrassa species (X. cretica, X. subvariegata, Xerocrassa grabusana) with the ratio of nonsynonymous to synonymous polymorphisms fixed between these groups. These two ratios should be the same if the gene is evolving in accordance with the neutral theory of Kimura (1983).
Third, an imaginary character state was assigned to each mitochondrial haplotype representing the species unit to which it belongs. The number of parsimony steps s in this imaginary character in the mitochondrial gene tree was calculated to assess incompleteness of lineage sorting among the seven species units that show fixed differences in genital characters. Larger s values indicate greater levels of incomplete lineage sorting and suggest a more recent divergence of species.
Fourth, gene trees were simulated to estimate the upper 95% confidence limit for the number of generations since population divergence that would be expected to give the observed s value for the mitochondrial gene tree. Ten thousand gene trees were generated with the program MESQUITE (Maddison and Maddison 2006) with an effective population size Ne of each species equal to 10,000 (probably lower than would be realistic for the widespread species, but all calculations are scaled by Ne). Generations since isolation (branch length of the population tree) were estimated, conservatively, as the greatest time that would yield at least a 5% probability of producing a gene tree with an s as high as or higher than that observed (the longer the time, the lower the expected s).
Finally, the estimated lengths of branches in the species tree were divided by two (equivalent to multiplying the population size by 2) to consider nuclear genes with two times the population size of mitochondrial genes (because the examined snails are hermaphrodites and, thus, all individuals of a population can pass mitochondrial DNA on to the next generation). Coalescent simulations with these branch lengths were run to estimate the probability of complete fixation of genital differences (s= 6 for seven species units) by genetic drift.
TEST OF EFFECTS OF NATURAL SELECTION AGAINST HYBRIDS
There are allopatric species that are widely separated, but nevertheless differ distinctly in their genitalia. Thus, differences in the genitalia cannot be exclusively the result of natural selection against hybrids. Nevertheless, natural selection against hybrids, that is reinforcement, might contribute to the observed differences in the genitalia. We would expect effects of natural selection against hybrids especially in areas in which the ranges of two species overlap so that there is an increased chance of hybridization. We investigated the importance of natural selection against hybrids in creating differences in the genitalia by testing whether the differences are larger between species with geographically overlapping ranges than between species that are not in contact or have only slightly overlapping ranges. Following Barraclough et al. (1999), we test the prediction that the differences are larger between co-occurring species by randomizing phylogenetic independent contrasts among nodes, holding the degree of geographical overlap fixed for every node. The association between morphological changes and the degree of overlap is expressed as the sum across all nodes of the change in character Xi multiplied by the degree of geographical overlap Si at each node, . If larger morphological changes occur between geographically overlapping sister clades, the observed association between character changes and degree of overlap is expected to be larger than that under the null model of no association between morphological change and geographical overlap. Standardized phylogenetic independent contrasts between morphological characters were calculated with CAIC (Purvis and Rambaut 1995). Logarithms of the length of the penis, the epiphallus, the flagellum, the vagina, and the bursa copulatrix were standardized by the logarithm of the body size before the contrasts were calculated. The degree of geographical overlap between two species was calculated as the ratio of the area of overlap to the area of the smaller of the two ranges. Following Fitzpatrick and Turelli (2006), we calculated the nested average degree of overlap at node i that separates clades C1 and C2 as
where the double sum is over all species in the two clades, ojk denotes the degree of geographical overlap between species j and k, and njk is the number of nodes separating the two species on the tree. Geographical overlap between ranges was calculated from distribution maps based on the 10 km × 10 km UTM grid (Hausdorf and Sauer 2009).
CORRELATED EVOLUTION OF SPERMATOPHORE-PRODUCING AND SPERMATOPHORE-RECEIVING ORGANS
For the analysis of correlated evolution of male and female reproductive organs, we used the general least-squares approach (Pagel 1997, 1999) implemented in the BayesContinuous module of BayesTraits (Pagel and Meade 2007). We tested the influence of phylogeny on the correlation between the log-transformed length of the spermatophore-producing organs, the sum of the length of epiphallus and flagellum, and the log-transformed length of the bursa copulatrix (including pedunculus) by comparing a model in which the parameter lambda that estimates the extent to which character similarities match the degree of shared ancestry between species is set to 0.0 with one in which lambda is allowed to take its maximum-likelihood value. If a likelihood-ratio test comparing the log-likelihoods of the two models is significant, some sort of phylogenetic correction is required. We included log-transformed body size (as measured by the shell volume) as a covariate in the correlation analyses, because the size of the reproductive organs scales with body size.
INFLUENCE OF EVOLUTION OF GENITALIA AND BODY SIZE ON SPECIATION
Following Barraclough et al. (1999), we used randomizations to compare the observed pattern of character changes across the tree to that expected under a null model of no association with cladogenesis. If morphological change is associated with speciation, recently split species should display greater divergence than expected under the null model. In contrast, if morphological differences promote persistence and/or subsequent radiation, divergence should be greater between more distantly related lineages. Hence, we test for a concentration of changes toward either the tips or the root of the tree. The null model is implemented by randomly shuffling phylogenetic independent contrasts among branches of the tree and recording where changes occur on the tree in each trial. The pattern of variability in each character with respect to relative node age is expressed as the sum over all nodes of the amount of change occurring across each node Xi multiplied by the relative age of that node Ai . For each test, the two-tailed probability of the observed value was calculated from the null distributions obtained by 1000 randomizations.
The maximum-likelihood tree of 122 partial cox1 sequences (634 bps) of Cretan Xerocrassa species and two Trochoidea species as outgroups is shown in Figure 3. Separate models for the three codon positions as determined by ModelTest were used, because the resulting tree had a lower AIC value than the tree based on a uniform model for the complete dataset.
The mitochondrial haplotypes of six of the 11 species recognized by Hausdorf and Sauer (2009) form distinct clades in the cox1 gene tree, but the sequences of the other five species do not form monophyletic groups. In particular, sequences of the widespread Xerocrassa mesostena are paraphyletic with respect to most of the other endemic Xerocrassa species.
The mitochondrial gene tree was made ultrametric using the penalized-likelihood method with a smoothing parameter of 2 chosen as optimal based on the normalized chi-square-like cross-validation scores. The species tree shown in Figure 4 has been derived from the ultrametric mitochondrial gene tree using a modification of the shallowest divergence clustering method as described in methods.
CAN DIFFERENCES IN THE GENITALIA BE EXPLAINED BY GENETIC DRIFT?
We used coalescent simulations based on the cox1 sequence data of the six most recently diverged endemic Xerocrassa species and the species pair X. amphiconus–X. siderensis (that show no fixed differences in genital characters) to test whether the fixation of differences in the genitalia between these species can be explained by genetic drift. The mitochondrial haplotypes were not completely sorted by species, suggesting relatively recent divergence of species (Fig. 3). The lack of monophyly of five of the morphologically defined species in the mitochondrial gene tree is not the result of a poor resolution of the mitochondrial gene tree, because the hypothesis that the mitochondrial haplotypes of each species form separate clades could be rejected with the approximately unbiased test (P < 0.001). Thus, the lack of monophyly of the morphologically defined species in the mitochondrial gene tree is consistent with a pattern of incomplete lineage sorting.
The cox1 sequences contained three synonymous to no nonsynonymous fixed differences between the seven recently diverged Cretan Xerocrassa species and the three most basal Xerocrassa species (X. cretica, X. subvariegata, Xerocrassa grabusana) versus 304 synonymous to nine nonsynonymous polymorphisms within these two groups. The ratio of nonsynonymous to synonymous polymorphisms within the groups is not significantly different from the ratio of nonsynonymous to synonymous polymorphisms fixed between the groups (Fisher's exact test P= 1.000). Therefore, the McDonald and Kreitman test results are consistent with neutral evolution of the cox1 gene.
There are fixed differences in genital characters between the seven investigated species units implying rapid fixation of genital differences, assumed to be encoded by nuclear genes. The minimum conceivable s value for seven species units is 6, which is equivalent to every species being fixed for a different phenotype. The observed s contrasting the seven species units in the cox1 tree was 9. The coalescent simulations indicate that s= 9 corresponds with an upper 95% confidence limit on branch lengths of 1.75N generations. An s as low as 6 occurred rarely (in 162 of 10,000 coalescent simulations) in coalescent simulations with branch lengths reduced to 0.875N generations to mimic nuclear genes. Thus, the observed fixation of differences in the genitalia among species units is highly unlikely to have arisen under neutrality, assuming the fixed differences reflect fixed differences in underlying nuclear genes. Consequently, the evolution of the genitalia is better explained by divergent selection.
CAN DIFFERENCES IN THE GENITALIA BE ASCRIBED TO NATURAL SELECTION AGAINST HYBRIDS?
The randomization tests with respect to degree of geographical overlap show that standardized phylogenetic independent contrasts in body size are significantly larger between geographically overlapping clades in the Cretan Xerocrassa radiation than expected under the null model of no association between morphological change and geographical overlap (Table 1). This indicates that competition between co-occurring species resulted in an ecological character displacement with regard to body size. On the contrary, contrasts in the length of penis, epiphallus, flagellum, vagina, and bursa copulatrix standardized by body size are not larger between geographically overlapping clades than expected under the null model (Table 1). Thus, there is no evidence for selection against hybrids that resulted in larger contrasts between geographically overlapping species.
Table 1. Positive associations between standardized contrasts in characters and the degree of geographical overlap and associations between standardized contrasts in characters and relative node age in the Cretan Xerocrassa radiation according to randomization tests. Positive signs of association of contrasts in characters with relative node age suggest that changes are concentrated toward the root and negative signs suggest that changes occur near the tips.
Geographical overlap P
Log body size
Log penis: log body size
Log epiphallus: log body size
Log flagellum: log body size
Log vagina: log body size
Log bursa copulatrix: log body size
CORRELATED EVOLUTION OF SPERMATOPHORE-PRODUCING AND SPERMATOPHORE-RECEIVING ORGANS
For the correlation between the log-transformed length of the male spermatophore-producing organs and the log-transformed length of the female bursa copulatrix the maximum-likelihood value of lambda calculated with BayesTraits (Pagel and Meade 2007) is 0.0. This indicates that the correlation evolves among species as if they were independent (Pagel 2000). Thus, a phylogenetic correction can be dispensed with. The partial correlation between the log-transformed length of the spermatophore-producing organs and the log-transformed length of the bursa copulatrix with log-transformed body size as a covariate that corrects for the scaling with body size is highly significant (Pearson coefficient R= 0.933, two-sided P < 0.001) indicating coevolution between male spermatophore-producing organs and female spermatophore-receiving organs.
INFLUENCE OF EVOLUTION OF GENITALIA AND BODY SIZE ON SPECIATION
The randomization tests with respect to node age show that the distributions of the contrasts in the length of penis, epiphallus, vagina, and bursa copulatrix standardized by body size as well as in body size across the species tree do not differ from random distributions (Table 1). On the contrary, changes in the length of the flagellum standardized by body size are significantly concentrated toward the tips of the tree indicating that the evolution of differences in flagellum length is involved in speciation in the Cretan Xerocrassa radiation.
The results of our tests indicate that sexual selection affecting flagellum length is involved in speciation in the land snail genus Xerocrassa on Crete. First, we could exclude two alternative causes of the changes in the genitalia, namely genetic drift and natural selection against hybrids. Coalescent simulations based on the degree of incomplete lineage sorting in a mitochondrial gene tree and the assumption that differences in the genitalia of the Xerocrassa species as potential indices of sexual selection are nuclear-encoded showed that the fixation of these differences between species cannot be ascribed to genetic drift. Whereas 10 of the 11 Xerocrassa species on Crete can be distinguished by diagnostic differences in the genitalia, only five of the species can be distinguished by shell characters that might reflect alternative niches or predation pressures. The degree of fixation in the shell characters is similar to that in the mitochondrial DNA haplotypes. Nevertheless, the significantly larger contrasts in body size between geographically overlapping clades indicate that selection is involved in the evolution of shell size. It is possible that some other shell characters evolve mainly by genetic drift.
Randomization tests demonstrated that the differences in the genitalia are not larger between co-occurring groups than between geographically separated taxa. Thus, there is no evidence that the differences in the genitalia are the result of natural selection against hybrids that might originate especially in areas where groups co-occur. Rather, it is likely that the differences in the genitalia are a product of sexual selection as has been shown in other groups (Eberhard 1985, 2001; Arnqvist 1998; Hosken and Stockley 2004). An important role of sexual selection is also indicated by mating experiments with Cretan Xerocrassa species that demonstrated assortative mating even between populations belonging to the same species (J. Sauer and B. Hausdorf, unpubl. data) as has also been found by Baur and Baur (1992) in Arianta arbustorum, an other helicoid land snail.
The significant positive scaling of male spermatophore-producing organs and female spermatophore-receiving organs in the Cretan Xerocrassa species indicates sexual coevolution. Koene and Schulenburg (2005) have already noted a correlation between the length of the flagellum that forms the tail of the spermatophore and the spermatophore-receiving organ in helicoids land snails at a coarser taxonomic scale. The tail of the spermatophore enables sperm to leave the spermatophore after it has been deposited in the partner's bursa copulatrix, the female gametolytic organ, during copulation (Lind 1973). Sperm are most successful at leaving the spermatophore and reaching the spermathecae when the tail of the spermatophore is protruding into the vagina. Thus, a lengthening of the spermatophore-producing organs might increase paternity success. The coevolution of male spermatophore-producing organs and female spermatophore-receiving organs might be the result of an evolutionary arms race over the control of fertilization, that is, of sexual conflict (Chapman et al. 2003; Arnqvist and Rowe 2005; Bergsten and Miller 2007). Alternatively, changes in the length of the spermatophore-producing organs might be the result of cryptic female choice (Eberhard 1985, 2001) for sperm that are better in escaping sperm digestion or for larger spermatophores as nutritional nuptial gifts (Gwynne 1984; Vahed 1998) or as signals of donor quality or condition (Anthes et al. 2008) and the coevolution might be the result of the necessity to process larger spermatophores (Anthes et al. 2008). Furthermore, natural selection might counter increasing spermatophore length because of increasing predation or desiccation risk resulting from long copulation times required for the transfer of long spermatophores.
Finally, randomizations of the contrasts in genital characters across nodes in the species tree (Table 1) showed that changes in the length of the flagellum standardized by body size are significantly concentrated toward the tips of the tree. If a lock-and-key mechanism (Shapiro and Porter 1989) would have triggered the radiation, we would expect that changes in those parts of the genitalia that directly interact during copulation, namely penis and vagina, are concentrated toward the tips of the tree. This is not the case. Rather, the evolution of differences in flagellum length that are probably the result of sexual selection is involved in speciation in the Cretan Xerocrassa radiation. The length of the spermatophore is determined by the length of the flagellum and the length of the epiphallus. The epiphallus produces the broad part of the spermatophore with the sperm container, whereas the narrow tail of the spermatophore is produced by the flagellum. That only changes in the flagellum are concentrated toward the tips of the tree and not changes in the epiphallus indicates that the meaning of these changes is not to transfer a larger amount of sperm or to provide more nutritious substance, but to optimize the length of the spermatophore with a modicum of extra cost. This is better compatible with the hypothesis that the driving force of the changes is rather sexual conflict than cryptic female choice. The hook-like structures at the spermatophore (Fig. 2) that point toward the anterior end and impede and delay the transfer of the spermatophore and so enable more sperm to leave the spermatophore before digestion also indicate that sexual conflict is involved. Divergence in shape and size aspects of spermatophore morphology in allopatry might also have triggered the radiation in the land snail genus Mastus (Parmakelis et al. 2005).
In the Cretan Xerocrassa radiation, sexual selection seems to be the initial mechanism resulting in speciation. Ecological differentiation of the lineages as indicated by different body sizes is generally not associated with recent speciation (lineage splitting close to the tips of the tree), but has been achieved when clades came into contact. This is consistent with the pattern found in some Nicaraguan crater lake cichlids in which sexual selection contributes more strongly or earlier during speciation than ecological separation (Wilson et al. 2000) and with the results of the comparative analysis of Barraclough et al. (1999) who also did not find evidence for an association of speciation with ecological disparity in tiger beetles. However, separation of lineages as a result of sexual selection does not always precede ecological differentiation in radiations. Based on the distribution of ecological and morphological characteristics across the phylogeny of the cichlid fish of Lake Malawi, Danley and Kocher (2001) suggested that this radiation has proceeded in three major bursts of cladogenesis of which the first two episodes resulted in adaptation to different niches, whereas the third episode was associated with differentiation of male nuptial coloration, most likely in response to divergent sexual selection. Also studies of other taxa suggest that ecological divergence is common in the early stages of a radiation (Schluter and McPhail 1993; Schliewen et al. 1994; Losos et al. 1998; Sturmbauer 1998; Schluter 1998, 2000a,b, 2001). Although it is plausible, that niche space might be subdivided early in the history of a radiation, it is unclear why the importance of sexual selection should vary in the history of a radiation. An alternative explanation of the observed patterns might be that an appreciable fraction of the speciation events is always the result of sexual selection, but that lineages that became adapted to different niches during or after speciation have a higher chance of persistence. As the geographical pattern of body size differences in the Xerocrassa radiation indicates, differentiation in ecologically important properties is associated with sympatry. Lineages that do not differ in adaptive characteristics may become more easily extinct if they become sympatric. The differential extinction of lineages that differ only in nonadaptive characteristics will result in an apparently almost exclusive adaptive phase in the early history of a radiation and more frequent cases of speciation as a result of sexual selection toward the present.
Barraclough et al. (1999) could not establish a role of body size in interactions between North American tiger beetle species of the subgenus Cicindella (Ellipsoptera). They suggested that one reason for the lack of evidence for the importance of body size in interspecific interactions in Ellipsoptera might be that the strength and direction of species interactions may have been highly variable over time, because communities represent transient groups of species. This may explain the difference between the significant association of contrasts in body size and sympatry in the Cretan Xerocrassa radiation and the lack of evidence in the North American tiger beetles. The ranges of the Cretan land snails were probably only slightly affected by the Pleistocene glacials and Cretan land snail communities were therefore more stable than North American insect communities. Thus, the strength and direction of interactions between land snail species on Crete did not vary as much as in North American insect communities.
We are grateful to T. Barraclough for help with the randomization analyses and to L. L. Knowles, N. Anthes, and three anonymous referees for constructive criticism of an earlier version of this article. This study was funded by the priority program “Radiations—Origins of Biological Diversity” of the Deutsche Forschungsgemeinschaft (HA 2763/3-1,2).