Introduction

One of the main concerns of population genetics is to understand how genetic variation is distributed within and among populations. One important factor is the breeding system, which mediates the way genes are transmitted across generations (e.g. Jarne & Städler, 1995). Hermaphroditic animals provide opportunities to study the evolution of breeding systems, in particular selfing vs. outcrossing. Pulmonate gastropods are all hermaphrodites (Heller, 1993) and show a wide variety of breeding systems, involving obligate outcrossing, obligate selfing, parthenogenesis and mixed breeding systems [for a review see Selander & Ochman (1983)]. Jarne (1995) showed that the genetic variability, as estimated by the mean number of alleles per locus, and gene diversity are reduced in selfing pulmonates when compared to outcrossing species. Experimental studies also showed that many gastropods are either almost complete selfers or complete outcrossers [e.g. Jarne (1995)]. A few basommatophoran freshwater snails exhibit a mixed mating system with much interindividual variation in the selfing rate (Jarne & Städler, 1995 and references therein), but in these species selfing rates are high, while little is known about the geographical distribution of selfing and outcrossing (Viard et al., 1997).

Previous allozyme studies on North American and Irish populations of the slug subgenus Carinarion Hesse 1926 have suggested that its three species, viz. Arion (Carinarion) fasciatus (Nilsson 1823), A. (C.) circumscriptus Johnston 1828 and A. (C.) silvaticus Lohmander 1937 each consist of one or two homozygous multilocus genotypes (MLGs) as a result of a uniparental breeding system (McCracken & Selander, 1980; Foltz et al., 1982). This led Selander & Ochman (1983) to claim that the species strictly reproduce by selfing. Recently, Backeljau et al. (1997) and Jordaens et al. (1998) reported that in NW Europe A. fasciatus, A. silvaticus and A. circumscriptus are not single MLGs but complexes of many co-occurring MLGs, thus confirming a uniparental reproductive mode.

However, the fact that (1) at least the copulation of A. fasciatus has been described once (Gerhardt, 1935, 2) the occurrence of spermatophores is reported regularly in Carinarion (e.g. Jordaens et al., 1996) and (3) nine heterozygote individuals have been found in two Belgian (one A. silvaticus and two A. circumscriptus), one Polish (one A. fasciatus) and one German population (five A. silvaticus) (in 884 specimens screened for seven polymorphic loci), may imply that outcrossing between MLGs, though very rare, is not impossible (Jordaens et al., 1996; Backeljau et al., 1997; Jordaens et al., 1998). Based on a limited number of additional observations in an Austrian population, Jordaens et al. (1996) and Backeljau et al. (1997) even tentatively suggested a geographical pattern in the prevalence of outcrossing. Interestingly, several heterozygotes possibly represented interspecific hybrids (Jordaens et al., 1996; Backeljau et al., 1997), indicating that the species may not be reliably separated by their MLGs alone.

Against this background, the present contribution aims at: (1) estimating the amount of species-specific differentiation in outcrossing Carinarion populations; (2) investigating geographical patterns in the prevalence of outcrossing in Carinarion; and (3) comparing the genetic variability (i.e. mean number of alleles per locus and gene diversities) of selfing and outcrossing populations. To this end, we surveyed allozyme variation in a large number of native Carinarion populations from N-NW and central Europe.

Materials and methods

Sampling and electrophoresis

Vertical polyacrylamide gel electrophoresis (PAGE) was used to assess allozyme variation in 45 Carinarion populations (Table 1). The total sampling area ranged from south Sweden in the north to the Czech Republic, Austria and Switzerland in the south. All three species are common in this area except for A. fasciatus which has not been recorded from Belgium and The Netherlands. Each sample consisted of slugs collected from a few m2 only, to avoid Wahlund effects. Species were identified following Lohmander (1937) and Waldén (1955). Eight polymorphic enzyme loci were surveyed: esterase Q (EsQ, EC 3.1.1.1), fumarate hydratase (Fumh, EC 4.2.1.2), alanine aminotransferase (Alat, EC 2.6.1.2), phosphoglucomutase (Pgm, EC 2.6.1.1), lactate dehydrogenase (Ldh, EC 1.1.1.47) and three leucylalanine aminopeptidase loci (Pep, EC 3.4.11). Sample preparation and experimental procedures followed Backeljau (1987) and Backeljau et al. (1997). Staining recipes were adapted from Harris & Hopkinson (1976).

Table 1 List of populations, numbers of specimens and numbers of MLGs (in parentheses) of A. fasciatus, A. silvaticus and A. circumscriptus, Nei’s (1978) gene diversity (Hexp) and observed heterozygosity (Hobs). †These populations were analysed by Jordaens et al. (1998)
Full size table

Data analysis

The program BIOSYS-2 [updated version of BIOSYS-1 of Swofford & Selander (1981)] was used to calculate allele frequencies, Nei’s (1978) unbiased expected heterozygosities (i.e. gene diversity Hexp) and observed heterozygosities (Hobs). Tests for genotypic (i.e. linkage) disequilibrium (LD) and Hardy–Weinberg equilibrium (HWE) were performed with the program GENEPOP v. 3 [updated version of GENEPOP v. 1.2 of Raymond & Rousset (1995)]. LD was tested for each population separately. HWE was evaluated with a probability test, which corresponds to the ‘exact HW test’ of Weir (1996). For all tests a sequential Bonferroni correction was applied to correct for multiple testing (Rice, 1989). For each population, GENEPOP v. 3 was used to estimate FIS values according to Weir & Cockerham (1984) as a measure of heterozygote deviation. We estimated the selfing rate S using the classical relationship FIS=S/(2 − S) (e.g. Viard et al., 1997). Tests for HWE and LD and the calculation of S were restricted to populations with more than 20 individuals to avoid problems of small sample sizes.

Nei’s (1978) genetic distance (D) was calculated between MLGs and used for UPGMA clustering and nonmetric multidimensional scaling (NMDS) with the NTSYS program v. 1.80 (Rohlf, 1993).

Results

Species differentiation

The species-specific alleles of A. fasciatus (i.e. Pgm-185), A. silvaticus (i.e. Alat100) and A. circumscriptus (i.e. EsQ100 and Fumh100) reported by Backeljau et al. (1997) in material from NW Europe, were now also found in other species, so that none of the species is characterized by species-specific alleles. This also makes A. fasciatus and A. circumscriptus more polymorphic than hitherto reported, so that at least A. silvaticus and A. fasciatus show comparable gene diversities (Table 1).

Combined with the studies of Backeljau et al. (1997) and Jordaens et al. (1998), a total of 46 different MLGs are now recorded in Carinarion: 13 in A. fasciatus, 24 in A. silvaticus and nine in A. circumscriptus. Two MLGs were shared by A. fasciatus and A. silvaticus. Due to ties (e.g. Backeljau et al., 1996), 11 alternative UPGMA topologies were found. In all trees, A. fasciatus and A. silvaticus appeared ‘paraphyletic’ (Fig. 1). In three topologies also A. circumscriptus appeared ‘paraphyletic’. The average interstrain genetic distance and the NMDS plot (stress:0.37) (Fig. 2) suggested a close relationship between A. fasciatus and A. silvaticus genotypes (average D=0.21 ± 0.10, range: 0.00–0.53) and a more distinct position of A. circumscriptus genotypes (A. fasciatus–A.circumscriptus: average D=0.28±0.08, range: 0.06–0.53; A. silvaticus–A. circumscriptus: average D=0.31 ± 0.10, range: 0.13–0.64). Intraspecific genetic distances between strains were comparable for the three species (A. fasciatus: average D=0.15 ± 0.07, range: 0.06–0.35; A. silvaticus: average D=0.18 ± 0.08, range: 0.04–0.44; A. circumscriptus: average D=0.13 ± 0.06).

Fig. 1
figure 1

UPGMA dendrogram of Nei’s (1978) genetic distances between 46 Carinarion MLGs: A. fasciatus (F1–F13), A. silvaticus (S1–S24) and A. circumscriptus (C1–C9).

Full size image
Fig. 2
figure 2

Two-dimensional projection of Nei’s (1978) distances between 46 Carinarion MLGs as obtained by NMDS: A. fasciatus (F1–F13), A. silvaticus (S1–S24) and A. circumscriptus (C1–C9).

Full size image

Variation in breeding system

Significant LD was found in 24 tests (out of 65 possible tests) at a 5% significance level. Twenty significant values were recorded in populations with high selfing rates. Yet, after applying the sequential Bonferroni procedure, only two tests (between Pgm-1 and Alat and between Est and Fumh in S2) remained significant.

In N-NW European Carinarion populations (Belgium and Sweden) no heterozygotes were observed (i.e. FIS=1 in all cases), except for a single Ldh heterozygote in St.-Antonius (B6) (FIS=0.960, Table 2; Jordaens et al., 1998) resulting in a mean FIS value of 0.996 and high selfing rates (Table 2; Fig. 3). There were no indications for the presence of null alleles. In many situations genotype frequencies deviated significantly from HWE (Table 2). In central Europe (other countries), however, heterozygous individuals were found in 11 out of 23 multi-MLG populations and in one single-MLG population (i.e. A3) (Table 1; Fig. 3), yielding a wide range of FIS values <1 (0.553 ± 0.377) and lower selfing rates (Table 2; Fig. 3). At many loci genotype proportions did not deviate significantly from HWE (Table 2). Moreover, genotype proportions at the Pgm-1 locus never deviated from HWE expectations in populations with heterozygotes. In contrast, Alat and Pep genotype proportions sometimes deviated significantly from HWE expectations in populations with heterozygotes (Alat in A1, A2, A5, Pep-1 in A1 and A5 and Pep-2 in P5 and A7).

Table 2 Allele frequencies, FIS estimates, selfing rates (S) and exact p-values for HWE (*P < 10–6) of populations with more than 20 individuals. Pep-3 was monomorphic in all populations $: FIS not calculated because one of the two alleles is represented only once; see Raymond & Rousset, 1995)
Full size table
Fig. 3
figure 3

Inferred selfing rates (solid sections) and outcrossing rates (open sections) in Carinarion populations with more than 20 individuals.

Full size image

Genetic variability

Hexp, Hobs and the number of MLGs in each population are given in Table 1. Neither the mean number of alleles per locus (Mann–Whitney U-test: all populations P=0.89; populations with n ≥ 20 P=0.59), nor gene diversity (Mann–Whitney U-test: all populations P=0.06; populations with n ≥ 20 P=0.13) were higher in outcrossing populations (i.e. populations with heterozygotes) compared to selfing populations (i.e. populations without heterozygotes). Compared to Backeljau et al. (1997) and Jordaens et al. (1998), we found two new alleles in central Europe, i.e. Pep-1115 and Pgm-1110 (Table 2).

Discussion

Taxonomic implications

Our study shows that the species specificity of the allozyme markers reported by Backeljau et al. (1997) is no longer valid if material from central Europe is included. The occurrence of outcrossing in some populations, the lack of species-specific allozyme markers, the ‘paraphyly’ of A. fasciatus and A. silvaticus and the observation that two MLGs are shared by A. fasciatus and A. silvaticus provide new support to Backeljau et al.’s (1997) questioning of the species status of at least A. fasciatus and A. silvaticus. Therefore, we suggest that Carinarion would be considered as a single species which in N-NW Europe may be divided into a number of MLGs which in some areas hybridize.

Distribution of genetic variance and breeding system

The high FIS values in N-NW European Carinarion populations result from the near absence of heterozygotes and thus confirm earlier reports of high selfing rates in these populations (Foltz et al., 1982; Backeljau et al., 1997; Jordaens et al., 1998). Population densities in many N-NW European Carinarion populations are low (personal observation; Jordaens et al., 1998) and thus we may expect a strong influence of genetic drift on the genetic structuring of these populations. The genetic consequences of genetic drift may be amplified by selfing, which may explain the low variability in some populations (e.g. Njiokou et al., 1993; Jarne, 1995; Viard et al., 1997).

Nevertheless, our results contradict the claim of Selander & Ochman (1983) that Carinarion strictly reproduces by self-fertilization, since the occurrence of heterozygous individuals in central Europe may indicate some degree of outcrossing. Outcrossing may even be the more common reproductive mode in this area since several loci do not deviate significantly form HWE expectations, at least in A. fasciatus, although significant departures from HWE may suggest some degree of selfing in these populations too.

S-value estimates from population genetic data assume that the observed FIS values are only influenced by the selfing rate. Of course, this may not be true, because other factors such as spatial and temporal variance in allele frequencies may also contribute to FIS values. Yet, population genetic data have the advantage over single-generation estimates, because they reflect selfing rates averaged over several generations (Jarne & Städler, 1995). It is evident that this mating-system heterogeneity across the range of the Carinarion complex would have been more convincingly demonstrated if we had performed experiments to directly estimate the outcrossing rate, via crossing experiments involving two or more MLGs. However, all our attempts to outcross Carinarion (even with animals from populations with low selfing rates) have failed, even when individuals were previously isolated for long periods, which in other gastropod species stimulates copulation (Heike Reise & Janet Leonard, personal communications).

The high selfing levels in N-NW European populations and the lower selfing levels in central European populations support Jordaens et al.’s (1996) and Backeljau et al.’s (1997) suggestion of a geographical pattern in the prevalence of outcrossing, at least in A. fasciatus. This is to our knowledge the first indication of such patterns of mating systems in hermaphroditic animals. The possible differential distribution of selfing and outcrossing Carinarion populations could be ascribed to ‘geographical self-fertilization’, somewhat similar to ‘geographical parthenogenesis’, i.e. the disjunct distribution of sexual and asexual populations (e.g. Fussey, 1984; Hughes, 1989; Enghoff, 1994). The biological meaning of this disjunct distribution of breeding systems currently remains unresolved.

The Quaternary cold periods in Europe are thought to have influenced the amount and distribution of intraspecific genetic variation in many animals and plants (Taberlet et al., 1998). The more northerly distribution of parthenogenetic populations of several species may be attributed to the high colonization ability of clones after the last glaciations (Vogel et al., 1999). Similarly, we presume that selfing Carinarion MLGs were able to (re)colonize Europe much faster than outcrossing Carinarion. This seems likely since Baur & Bengtsson (1987) and Bengtsson & Baur (1993) showed that selfing gastropods are better colonizers than outcrossers. There is a general perception that central and northern Europe were colonized by range expansion from Mediterranean refugia (viz. the Iberian peninsula, Italy and the Balkans) at the end of the last glaciation and that the northern part of Europe has been colonized primarily from the Balkan and Iberian refugia (e.g. Taberlet et al., 1998). Since Carinarion is absent from the Iberian peninsula (Castillejo, 1997), it is probable that Carinarion has (re)colonized N-NW Europe from Balkan refugia. For several small mammals Bilton et al. (1998) provided strong support for recolonization that may have occurred from glacial refugia in central Europe–western Asia. If one of these scenarios is true, then we expect also high outcrossing levels in the Balkan area. Although these models of postglacial range expansion from south or central European refugia is well supported for some temperate animals and plants, there is an alternative scenario. Some of the north European Carinarion populations may have survived the last glaciation (or several glaciations) in local, ice-free refugia and (re)colonized the rest of N-NW Europe from these refugia. However, this concept of glacial refugia in northern Europe has today lost much of its credibility (Bilton, 1994). Moreover, it will be almost impossible to distinguish between this and the former models (Tollefsrud et al., 1998). Obviously, alternative explanations, such as possible ecological and altitudinal differences between NW-European and central European populations may explain such a pattern in breeding systems as well, but the data are too preliminary to allow adequate testing of these alternatives.

Contrary to current models (e.g. Selander & Ochman, 1983; Jarne, 1995), gene diversities and mean numbers of alleles per locus in selfing and outcrossing Carinarion populations were of the same magnitudes. Indeed, although many selfing populations only contain one MLG, about 50% of the populations contain more than one MLG (Backeljau et al., 1997; Jordaens et al., 1998). Possibly, Carinarion does not fit into current models.

It is evident from our study that our knowledge is incomplete and before firm conclusion can be made, a clearer understanding of the phylogeography of Carinarion is required. Nevertheless, our data indicate that population genetic data of hermaphroditic animals may provide supplementary information for postglacial phylogeographic scenarios.