Two main morphotypes have been discriminated visually. They correspond to the forms described by Pfenninger et al. 42 : a round and broad form with deep ridges (round form, R) and a narrow form with closely spaced ridges (saddle form, S) (see 39 for pictures of both forms). Both forms occur syntopically in the Hollands Diep (Rhine-Meuse delta), in the River Meuse (Belgium, Table 3) and in the Rivers Seine and Rhine (Table 4). The individuals from the River Rhône can be subdivided into two colour groups of form R 36 , viz. one with dark shells whose interior surface is white and purple (as in form R sampled elsewhere), the other with lighter shells whose interior is white-yellow (light form, Rlc). Form Rlc from the Rhône has also been recorded in the River Gardon (France, see 39 ). Measurements were made on 429 specimens of Corbicula from all sampled localities. The measurements provided in the literature for the lectotypes of C. fluminea Müller and C. fluminalis Müller (collection of Universitetets Zoologisk Museum of Copenhagen 20 ) and the type specimen of C. leana Prime (original description of the species in 26 ) were also included in the analysis. The two main forms identified a priori (R and S) could be distinguished on the basis of the Principal Components Analysis (PCA) conducted on the three studied ratios (H/L, H/W and L/W, see Methods) (Figure 1). Form Rlc from the Rhône was morphometrically close to form R (Figure 1). We therefore conducted a new PCA on these two forms exclusively (results not shown). This analysis still showed an overlap between the two forms but suggested that specimens of form Rlc have a significantly lower H/L ratio (analysis of variance, p < 0.01). However, our study included only 14 specimens of form Rlc and a more extensive sampling is needed to test the differences with form R. The morphological distinction between forms R and Rlc is mainly based on colour, especially of the inner shell. Moreover, we noticed that juvenile individuals of form Rlc have an orange spot on their umbo (the dorsal shell protrusion) which is not observed for juveniles of form R (Pigneur LM & Marescaux J, personal observation).

The lectotype of C. fluminea clustered with individuals of morphotypes R and Rlc on the biplot. The type specimen of C. leana was found in the margin of the group composed of individuals of form R only. The lectotype of C. fluminalis was found within the group of specimens of form S (Figure 1). The two individuals (Mo1 and Se26) showing a cytonuclear mismatch (see COI and cyt b sequences analysis below) were morphologically part of the R-Rlc group (Figure 1).

The sperm morphology of the three morphotypes of Corbicula (forms R, S and Rlc) was investigated because the biflagellate sperm type has been recognized as a diagnostic marker of androgenetic reproduction in this genus 15 16 18 . The analysis revealed biflagellate spermatozoa in the three morphotypes of Corbicula (forms R, S and Rlc). These three forms thus seem to reproduce through androgenesis like the invasive lineages found in America 18 .

COI and cyt b sequences were obtained respectively from 215 and 41 specimens from Meuse, Seine, Rhine and Rhône (Tables 3, 4). The analysis revealed only three haplotypes each corresponding to one of the three morphotypes. All individuals of a same morphotype exhibited exactly the same haplotype (except for 2 individuals, see below). Sequence comparisons between the three COI haplotypes revealed 22 variable sites distributed throughout the sequenced fragment (681 bp) (Alignment provided in Additional file 1). For cyt b, only 10 sites out of 326 were variable (Additional file 2). Since our cyt b sequences were relatively short (326 bp) and a lower number of sequences for this gene are available for Corbicula, we have conducted the sequence analysis and the phylogenetic study with COI only.

One haplotype (COI haplotype 1) was found in all our sampled populations. This haplotype 1 was exclusively found in form R and its nucleotide sequence is identical to European C. fluminea (haplotype I, 40 ) and haplotype H2 41 , the American form A 18 34 , the Japanese C. leana 34 and the Asian haplotype FW5 31 (Tables 1, 2). We will refer to this COI haplotype as haplotype FW5 which was a common code used in Park & Kim 31 for all individuals of Corbicula with this COI sequence.

The second haplotype (COI haplotype 2) was found in all specimens of form S and in two specimens of form R (Table 2). Haplotype 2 is identical to the European "C. fluminea-like" COI haplotype IV from the Rhône 40 and haplotype H4 found in form S in the Rhine 42 and to the American form C 18 (Tables 1, 2). We will refer to this haplotype as haplotype FW17 as in Park & Kim 31 . Haplotype FW17 has not been found in any Asian population 31 .

We observed a discrepancy between the mitochondrial lineage and morphology for two individuals of form R. The first one was a single individual from the River Meuse at Liège-Monsin (individual Mo1) and the second one is a single individual from the Seine (individual Se26). Both individuals had the morphotype R but exhibited the COI haplotype 2 (found in form S). The same results were obtained for cyt b haplotypes. We sequenced both genes again on new tissue samples from these two individuals and obtained the same results.

Our third haplotype (COI haplotype 3) was exclusively found in the individuals of form Rlc from the Rhône while form R from this same site had the haplotype 1 (FW5) (Table 2). Renard et al. 40 found COI haplotype IV in the five individuals studied from the Rhône; we did not find this haplotype IV in the new samples we obtained from River Rhône. Our haplotype 3 from the Rhône is identical to the sequence of C. fluminalis (haplotype V) from Renard et al. 40 (Table 2). Therefore, there is a discordance between our haplotypes and those obtained by Renard et al. 40 for form S. The same discrepancy was also reported by Pfenninger et al. 42 (Table 2). Our haplotype 3 is identical to the haplotype FW4 of the Indonesian C. subplanata 31 50 (Table 1). We will here refer to haplotype FW4 as in Park & Kim 31 . This haplotype FW4 clusters with the American haplotype B1 (identities: 636/637 bp) and C. fluminea from Korea (identities: 613/614 bp) (Figure 2).

Our analysis yielded only three haplotypes for all studied individuals among the four sampled European rivers. Surprisingly, Pfenninger et al. 42 recorded a larger number of COI haplotypes, especially in the River Rhine. We aligned all of these haplotypes and observed that some sequences contain several Ns or are shorter than most published sequences. We therefore included only some representative haplotypes of that study in our Maximum Likelihood (ML) and Bayesian Inference (BI) analyses.

In total, 57 COI sequences of Corbicula available in GenBank, including those three sequences from the present study, were included in the phylogenetic analysis (Table 1). Out of these 57 sequences, 47 COI haplotypes were defined (Additional file 1). From the 714 positions (562 excluding gaps or missing data) including the outgroup, 158 sites were variable (28%). Only 13% of the sites remained variable when we excluded the outgroup, the estuarine lineages (see Figure 2) and C. madagascariensis.

Based on the lower value of Akaike information criterion (AIC), the GTR+γ model was selected for phylogenetic analysis. Neocorbicula limosa was used as an outgroup. ML analysis and BI gave a similar topology. The phylogenetic tree (Figure 2) shows two divergent clades: an "estuarine clade" (Asian C. fluminalis and C. japonica) and a "freshwater clade". Within this freshwater clade, C. madagascariensis is a sister-taxon to all other freshwater lineages of Corbicula. In this latter group, there is no clear geographic clustering of haplotypes of Corbicula, highlighting their cosmopolitan distribution. However the phylogenetic relationships between the freshwater lineages of Corbicula remain unresolved due to the low support (bootstrap and posterior probabilities) obtained for most clades. The three haplotypes of our study (referred to as FW4, FW5 and FW17 as in Park & Kim 31 ) are each associated with a distinct freshwater subclade. Within each of these subclades, there are mostly representatives of Corbicula with biflagellate sperm (see Table 1) suggesting that these lineages reproduce through androgenesis 19 . However the clade containing haplotype FW4 also includes C. linduensis which has monoflagellate sperm 17 , suggesting that this latter reproduces sexually.

Nuclear markers were mainly used in this study to investigate the observed mismatches between mtDNA and nrDNA/morphotype. For all eleven polymorphic micosatellite loci, the two individuals with morphotype R and COI haplotype 2 (FW17) of form S (individuals Mo1 and Se26) exhibited exactly the same microsatellite genotype as the form R individuals. Therefore, both individuals clearly belong to the form R lineage at the morphological and nuclear levels but have the form S mitochondrial haplotype.

The same eleven microsatellite loci were also used to analyse several West European individuals of Corbicula belonging to the three different morphotypes (Table 5). For microsatellite locus ClB03, forms R and S present the same alleles (heterozygous; 2 alleles) while for the loci ClA02, ClB11 and C01 forms S and Rlc have the same allele size. Two loci (ClA01 and ClA03) did not amplify in form S, though they did amplify in the two specimens of form R with mtDNA COI haplotype 2 (as in form S). Two other loci (ClD06 and ClE01) did not amplify in form Rlc. The remaining four microsatellite loci amplified in all three forms and revealed differences in allele size between form R, S and Rlc. Since the microsatellite library was constructed for form R 51 , it is not surprising that, due to the genetic divergence between the three forms, some loci did not cross-amplify in form S or Rlc (Table 5).

Surprisingly, for each tested microsatellite locus, zero polymorphism was observed between individuals within each morphotype originating from different West-European populations. Only three distinct genotypes (genotypes 1, 2 and 3 for forms R, S and Rlc respectively) were identified. Bayesian clustering conducted using the STRUCTURE software revealed distinct categories, with the highest and most reliable probabilities for K = 3 (Additional file 3). This confirmed the observed differences between the three forms and the presence of only three invasive haplotypes/genotypes of Corbicula in Western Europe (Table 5).

These results are indicative of an overall very low genetic divergence within European Corbicula and could be related to their clonal reproductive mode.

The dark-coloured form R individuals we found correspond morphologically to the description of C. fluminea in 36 41 42 . Our morphological analysis did not provide a strong taxonomic assignment although the type specimen of C. leana was plotted near the group including form R. The sequence of haplotype 1, found in all studied individuals of form R in Europe (except the two "mismatches"), was identical to Asian haplotype FW5. Those results are congruent with those of Siripattrawan et al. 34 and Hedtke et al. 19 which concluded that populations of American form A (and therefore the European form R) are derived from populations of C. leana (haplotype FW5) from Japan. On the basis of these COI results, the European form R would not correspond to C. fluminea sensu stricto as previously assumed. Regarding the genotyping results, all 27 individuals of form R had one specific nuclear genotype (genotype 1) for the 11 tested loci.

The individuals of form S (Meuse, Seine and Rhine rivers) correspond morphologically to Corbicula fluminalis as defined by Renard et al. 40 and Korniushin 41 . Our morphological analysis, through the PCA, also showed similarity between the European form S and the lectotype of C. fluminalis from Mesopotamia. The COI haplotype 2 found in form S was identical to haplotype FW17 and to American haplotype C1 (of form C). Interestingly Park & Kim 31 already highlighted that no Asian haplotype corresponds to haplotype FW17. Genetic and morphological data of specimens from the native area of C. fluminalis (Middle East, Caucasus and Central Asia) are needed to confirm the taxonomic assignment of the European form S/American form C. Our results also suggest that form S does not correspond to C. fluminalis sensu Park & Kim 31 . This Chinese, estuarine C. fluminalis was found in the same mitochondrial clade as C. japonica (Figure 2). Korniushin 41 already reconsidered this form as C. cf. japonica on the basis of anatomy and reproductive features. The genotyping results showed that all 23 European individuals of Corbicula of form S had the same, distinct nuclear genotype (genotype 2) for the 9 microsatellite loci that amplified in form S although these nuclear markers were developed for form R.

We obtained new samples from the Rhône and our analysis revealed the presence of two distinct COI lineages in this river: (i) the haplotype 1 (FW5) in individuals of form R and (ii) the haplotype 3 (FW4) found in specimens of form Rlc exclusively. The European light form Rlc had the same COI haplotype as the Indonesian C. subplanata, and was closely related to the American form B and Corbicula fluminea from Asia 19 34 . Glaubrecht et al. 10 17 already made the assumption that C. subplanata would be conspecific with the widespread C. fluminea. Hedtke et al. 19 concluded that the American form B would be derived from populations of the Asian lineage C. fluminea. As in form S, 9 microsatellite loci amplified in European form Rlc and revealed a third distinct genotype (genotype 3), being identical in all tested individuals of form Rlc.

The overall low genetic divergence between and within the three forms and the possibility of a peculiar hybridization through androgenesis between forms (see below) suggest that invasive lineages of Corbicula may represent a polymorphic species complex. Therefore we propose to use the nomenclature "form R", "form S" and "form Rlc" for the three West-European lineages of Corbicula.

The sequence of haplotype 3 (FW4) we found in form Rlc (Rhône) was identical to the C. fluminalis - haplotype V and not to the expected haplotype IV of the Rhône found by Renard et al. 40 . This result of Renard et al. 40 is contrasting with our data and those of Pfenninger et al. 42 . Moreover, Renard et al. 40 included 6 individuals of the C. fluminalis form (form S) from the Meuse (Netherlands) but their sequence (COI haplotype V) does not correspond to the one we found (haplotype 2) in our sampling of morphotype S and it does not correspond to haplotype H4 from form S published by Pfenninger et al. 42 (see Table 2). We do not have any evidence of temporal changes in the studied populations and we therefore hypothesize a mismatch between the published sequences and their associated names/populations in the paper of Renard et al. 40 . More specifically, the COI sequence assigned to C. fluminalis (haplotype V) in their publication actually belongs to the individuals from the Rhône (their third unidentified lineage) and corresponds to our form Rlc (also from the Rhône), while their sequence of haplotype IV (Rhône) corresponds to their specimens of C. fluminalis as Pfenninger et al. 42 and this study demonstrated. In Table 2, we present a summary of the incongruence and a comparison of European, American and Asian COI haplotypes of Corbicula.

In Europe and North America, most individuals of Corbicula with the same morphotype belong to the same mitochondrial lineage ( 18 19 42 , this study). In addition, in this study we have preliminary indications that the same Corbicula morphotype also belongs to a single genotype although more individuals should be investigated. However, we detected a discrepancy between the mitochondrial lineage and, the morphology and the nuclear lineage of two specimens of form R, one from the Meuse and one from the Seine, two of the three rivers where we found forms R and S in sympatry. These individuals with morphotype R exhibit the mitochondrial haplotype of form S (for both cyt b and COI) but all polymorphic microsatellite markers identified a form R nuclear DNA origin (genotype 1) for both specimens. In the absence of recent nuclear exchange between the morphotypes, the observed mismatch could be explained by double uniparental inheritance (DUI), population polymorphism in the ancestor of the species or egg parasitism between the species 19 . Hedtke et al. 19 did not find any evidence of DUI in individuals presenting a mtDNA/morphotype mismatch. The hypothesis of a polymorphic ancestor was rejected on the basis of their phylogenetic analysis 19 . Hedtke et al. 19 therefore concluded that this observed discrepancy could only be the consequence of an androgenetic "egg capture": after fertilization between the unreduced sperm of one haplotype and the egg of a distinct haplotype, the maternal nuclear DNA from the egg is extruded. The descendant therefore contain paternal nuclear DNA while they retain the maternal mitochondrial DNA, resulting in a cytonuclear mismatch as observed here for two individuals.

A similar cytonuclear mismatch was observed in some American populations between forms A and B and between forms B and C living in sympatry 18 19 . Discrepancies between nuclear and mitochondrial phylogenies have already been reported for Corbicula 42 44 . According to our mitochondrial phylogeny and our microsatellite data, the disjunction found in our two singular individuals would imply a spermatozoon of lineage R that "parasitized" an egg of lineage S.

In freshwater ecosystems, Corbicula fluminea sensu lato is considered one of the most invasive species 3 because of its large geographic dispersal and invasive behaviour 3 20 24 . The invasive success of lineages of Corbicula is attributed to their following biological traits: rapid growth and maturation, high fecundity, excellent dispersal capacities and association with human activities 3 . However, their unique reproductive mode might also play a role in the invasiveness of lineages of Corbicula while this has never been highlighted. Our phylogenetic analysis (Figure 2) demonstrates that the "freshwater clade" includes both sexual dioecious and androgenetic lineages. The sexual lineages are mostly restricted to limited Asian areas, e.g. C. sandai which is endemic to Lake Biwa (Japan) or C. matannensis which is endemic to Indonesia. Moreover, it has been shown that, on a limited spatial scale, the lacustrine lineages of Corbicula from Sulawesi (Indonesia) form well-defined genetic clades 50 while identical or closely related haplotypes of the androgenetic lineages are found over large geographic distances, both in the native and invaded regions ( 31 50 , this study). This emphasizes that androgenesis seems to favour the invasive success of these clams. More specifically, androgenetic Corbicula are hermaphrodites and capable of self-fertilization. Therefore, a single individual may easily found a new population if the conditions are favourable. This is one of the main advantages of asexual as compared to sexual lineages 52 .

In addition, our genetic analysis of European populations of Corbicula revealed an absence of polymorphism within each form. Many invasive species escape low levels of genetic diversity owing to multiple and/or repeated introduction events 53 54 . On the contrary, there is a "prevalence of genetic bottlenecks" in invasive species with several species showing a low genetic diversity but a high invasive potential 55 . This is considered the "genetic paradox of invasive species" 55 56 57 58 . In the case of Corbicula, further investigations should assess the genetic diversity of the native populations to determine whether the invasive forms have encountered genetic bottlenecks or not. Due to androgenesis several of the lineages in the native area may indeed contain a low genetic diversity. Whatever the origin of their low genetic diversity, invasive lineages of Corbicula represent one of those examples of "genetic paradox". Phenotypic plasticity, which is particularly high in molluscs, might have played an important role in the adaptation to and colonization of new environments as demonstrated in many other invasive species 59 60 61 . In other words, those three invasive clonal lineages of Corbicula, particularly form R dominating in Western Europe, could be "general purpose genotypes" or even "super-genotypes" associated with high levels of plasticity in response to fluctuating environmental conditions 62 63 .

However, a certain level of genetic diversity in Corbicula may be found in the combination of a nuclear genome of one lineage with a new mitochondrial genome of a different lineage as a result of androgenesis between lineages. This may provide a higher advantage to androgenetic lineages compared to other asexually reproducing species. Mitochondrial capture has already been pointed out as dispersal strategy in the endangered cypress tree Cupressus dupreziana. This androgenetic organism produces diploid pollen that may in turn fertilize the ovule of the related, common species Cupressus sempervirens. This latter species, then, acts as a surrogate mother 64 65 . Moreover, androgenesis allows an elevation of ploidy when there is mixing of nuclear genomes during incomplete extrusion of the maternal nuclear genome 19 45 46 .

Several studies revealed mitochondrial/morphotype discrepancies in populations of Corbicula where distinct forms live in sympatry ( 18 19 42 , this study) and hypothesized that it was the consequence of an androgenetic egg parasitism ( 18 19 , this study). We would like to highlight here that this type of mitochondrial capture could lead to inaccurate species delimitation. The mitochondrial gene tree only traces the maternal lineage and it is, in the case of androgenesis, misleading for species relationships. Until now, most phylogenetic studies in Corbicula have relied mainly or solely on mitochondrial data, but this can result in the grouping of distinct nuclear lineages in the same mitochondrial cluster. Therefore the use of an "integrative taxonomy" approach as developped here, combining nuclear and mitochondrial data along with morphology, is essential to verify the phylogenetic relationships within androgenetic taxa such as Corbicula. Moreover, we suggest that due to these peculiar hybridization events between lineages as a consequence of an androgenetic mode of reproduction, invasive European Corbicula should be considered as a polymorphic species complex including three distinct clusters. For example, the European form S and the American form C have the same COI sequence but appear very different on the basis of shell morphology compared to pictures of form C from Lee et al. 18 . The observed differences in shell shapes could be related to phenotypic plasticity, but alternatively, forms S and C could also belong to distinct nuclear lineages sharing the same mitochondrial DNA through androgenesis. The same hypothesis seems confirmed by preliminary results in the COI cluster "form B - form Rlc". These morphs can easily be distinguished: the form B from the Americas has a deep purple inner shell surface while the form Rlc from Europe has a whitish inner shell surface. The assessment of cross-amplification of microsatellite loci in American forms revealed that form B and form Rlc have distinct genotypes 51 . They therefore belong to distinct clusters at the nuclear level.

Live specimens were collected from 23 localities in the Meuse river in France, Belgium and the Netherlands (Table 3), and from one site in the Seine river in France (Table 4). Additional specimens from one locality in the Rhine and one in the Rhône were obtained (Table 4). Specimens were immediately processed or preserved in 96% ethanol prior to analysis. Individuals from the Meuse and Seine rivers were collected in fish bypasses with a handnet or directly from the river bottom with a handnet, or dug out from the sediment with an Eckman dredge.

The shell morphology of the different morphs from Meuse, Rhine, Rhône and Seine rivers was analysed. The shells of 429 individuals were measured for length (L), height (H) and width (W) with a calliper: "L" is the maximal anteroposterior dimension, "H" is the maximal dorsoventral dimension and "W" is the maximal width of the valves. We also included published measurements of the lectotypes of C. fluminea and C. fluminalis 20 and of the original description of C. leana 26 . A Principal Component Analysis (PCA) was carried out using the table of the three morphometric ratios (H/L, H/W and L/W). Calculations were performed and graphics drawn using R software with the ade4 package 66 . The function "dudi.pca" was used to apply the PCA to the correlation matrix. Scores from axis 1 were plotted against those from axis 2, each point was identified individually.

Biflagellate sperm is indicative of androgenesis in the genus Corbicula 15 16 18 . We analysed the sperm morphology of specimens of form R (Meuse, N = 4), form S (Rhine and Seine, N = 5) and form Rlc (Rhône, N = 3). To isolate and observe the spermatozoa, a protocol was adapted from several studies 67 68 69 . The individuals preserved in absolute ethanol were opened and the visceral mass, which includes the gonads, was cut. The tissue was collected in a tube and collagenase resuspended in RPMI was added (final concentration: 1 mg/l). The tissue was sheared and collagenase-RPMI mix was added to a final volume of 600 μl. The preparation was then incubated at 45°C until complete degradation. The samples were centrifuged during 3 minutes at 3000 rpm. One drop of the undernatant was put on a glass slide and observed under a light microscope at 100× magnification.

Total genomic DNA was extracted from the adductor muscles or the foot of 215 individual specimens (fresh or preserved in 96% ethanol) using the DNeasy blood & tissue kit (Qiagen) according to the manufacturer's protocol. A fragment of 710 bp of the mitochondrial cytochrome c oxidase subunit I (COI) gene and a fragment (430 bp) of the cytochrome b (cyt b) gene were amplified by Polymerase Chain Reaction (PCR) using the primers LCOI490 and HCO2198 70 and, UCYTB151F and UCYTB270R, respectively 71 . Amplifications were performed in 25 μl total volume including 0.5 μl of gDNA, 1× GoTaq Green reaction buffer (Promega), 200 μM of dNTPs (Promega), 0.5 μM of both primers and 0.1 U of GoTaq DNA polymerase (Promega). PCR conditions were: 4 min at 94°C followed by 30 cycles of 45 s at 94°C, 45 s at annealing temperature (45 °C for COI and 42°C for cyt b) and 45 s at 72°C, and then a final extension of 10 min at 72°C. PCR products were purified and sequenced with each universal primer on an automated ABI3730XL Genetic Analyzer (Macrogen Inc.). COI and cyt b sequences were deposited in GenBank [GU721082-GU721084; JF518976-JF518978]. Sequences were visualized and aligned using BioEdit 7.0.5.3 72 .

The evolutionary model for the phylogenetic analysis of COI sequences was selected using jModelTest 0.1 software 73 . Maximum-likelihood trees were constructed with PhyML 2.4.5 software and bootstrap values were obtained for 1,000 replicates 74 . A Bayesian Inference (BI) analysis was performed using MrBayes 3.1.2 75 . We ran four independent analyses with four Markov Chain Monte Carlo (MCMC) each, for 4 million generations. Trees and parameters were sampled every 100 generations. To ensure stabilization, we discarded the first 25% of sampled trees. The remaining trees were used to compute a consensus tree. The split frequency was below 0.01. For all analyses, published COI sequence of Neocorbicula limosa was used as an outgroup. A total of 54 sequences from several Corbicula lineages available on GenBank were included (origins in Table 1). Trees were visualized using MEGA4 76 .

We used eleven microsatellite loci described in Pigneur et al. 51 to analyze individuals of the three observed European morphotypes (see Table 5 for the number of tested individuals). Regarding the samples from the Meuse, Rhône and Rhine, the individuals analyzed in the present study are different from those used in Pigneur et al. 65 . For each locus, the amplification was performed in 10 μl total volume including 0.5 μl of gDNA, 1× GoTaq reaction buffer (Promega), 200 μM of dNTPs (Promega), 0.5 μM of both primers and 0.1 U of GoTaq DNA polymerase (Promega). The forward primers were 5' fluorescently labelled and fitted with a GTTTCTT PIG-tail. Amplification by PCR was performed as follows: 4 min at 94°C followed by 30 cycles: 45 s at 94°C, annealing for 45 s and 72°C for 45 s, and then a final extension of 12 min at 72°C. The fragments were analysed on an ABI 3130XL Genetic Analyzer with GeneScan-500 (LIZ) size standard (Applied Biosystems). Results were visualized using GENEMAPPER software (Applied Biosystems). Additionally, the microsatellite results were analyzed using Bayesian clustering with the STRUCTURE software version 2.0 77 to define the number of categories (K) of individuals. STRUCTURE was run for 10 repetitions of each K from 1 to 5 assuming admixture. Analysis was performed using a burn-in period of 50,000 followed by 200,000 MCMC repetitions.