Unequal contribution of native South African phylogeographic lineages to the invasion of the African clawed frog, Xenopus laevis, in Europe

Due to both deliberate and accidental introductions, invasive African Clawed Frog (Xenopus laevis) populations have become established worldwide. In this study, we investigate the geographic origins of invasive X. laevis populations in France and Portugal using the phylogeographic structure of X. laevis in its native South African range. In total, 80 individuals from the whole area known to be invaded in France and Portugal were analysed for two mitochondrial and three nuclear genes, allowing a comparison with 185 specimens from the native range. Our results show that native phylogeographic lineages have contributed differently to invasive European X. laevis populations. In Portugal, genetic and historical data suggest a single colonization event involving a small number of individuals from the south-western Cape region in South Africa. In contrast, French invasive X. laevis encompass two distinct native phylogeographic lineages, i.e., one from the south-western Cape region and one from the northern regions of South Africa. The French X. laevis population is the first example of a X. laevis invasion involving multiple lineages. Moreover, the lack of population structure based on nuclear DNA suggests a potential role for admixture within the invasive French population.


INTRODUCTION
Reconstructing the invasion history and dynamics of invasive species is crucial for understanding biological invasions and for developing appropriate management strategies (Sakai et al., 2001;Lee & Gelembiuk, 2008;Prentis et al., 2009). Moreover, exploring patterns of population genetic variation and evolutionary processes may be key to infer the invasive potential of invasive alien species (Sakai et al., 2001;Lee & Gelembiuk, 2008). In this study, we focus on the African clawed frog, Xenopus laevis (Daudin, 1802), which is indigenous to Southern Africa, including South Africa up to Malawi (i.e., X. laevis sensu stricto Furman et al., 2015). Due to deliberate and accidental introductions from laboratories and pet suppliers, invasive X. laevis populations have become established in Asia, Europe, North America and South America (Tinsley & McCoid, 1996;Lobos & Measey, 2002;Crayon, 2005;Fouquet & Measey, 2006;Faraone et al., 2008;Measey et al., 2012;Peralta-García, Valdez-Villavicencio & Galina-Tessaro, 2014). Invasive X. laevis populations have negative impacts on local biota by reducing the occurrence of reproduction (Lillo, Faraone & Lo Valvo, 2010) and increasing predation pressures on native prey organisms (Lafferty & Page, 1997;Faraone et al., 2008;Measey et al., 2015). Lobos & Measey (2002) also suggested that X. laevis might have indirect impacts on the aquatic system such as increasing water turbidity and nutrient release. Finally, the spread of Batrachochytrium dendrobatis, which causes the amphibian skin disease chytridiomycosis and negatively impacts amphibian populations (Berger et al., 1998;Lips et al., 2006;Skerratt et al., 2007;Voyles et al., 2009;Crawford, Lips & Bermingham, 2010), has been linked to invasive amphibian species such as X. laevis which are often asymptotic carriers (Weldon et al., 2004); however, this link has yet to be proven. Measey et al. (2012) suggested that the global potential invasiveness of X. laevis has been severely underestimated and that it is likely that X. laevis will expand its present colonized area. For example in Europe, introduced X. laevis populations are currently established in France (Fouquet, 2001), Portugal (Rebelo et al., 2010) and Italy (Sicily) (Lillo et al., 2005), though the predicted suitable climate space for X. laevis covers over one million km 2 making this a species of European concern (Measey et al., 2012).
Within its native range, X. laevis has a wide geographical distribution in which it occupies a variety of natural, as well as manmade waterbodies (Measey, 2004). Native X. laevis populations are distributed from winter rainfall regions in the south-western Cape region to summer rainfall regions in the north; and from sea level to nearly 3,000 m in Lesotho (Measey, 2004). Furthermore, several physiological and behavioural traits enable X. laevis to cope with, dehydration, high levels of salinity, starvation and anoxic conditions (reviewed in Measey et al., 2012). Throughout the native range of X. laevis significant population differentiation has been observed based on both mitochondrial (mtDNA) and nuclear DNA (nDNA) sequences (Grohovaz, Harley & Fabian, 1996;Measey & Channing, 2003;Du Preez et al., 2009;Furman et al., 2015). As such, Furman et al. (2015) identified four phylogeographic lineages within South Africa based on mtDNA and nDNA: (1) south-western Cape (SA1-SA2), (2) Beaufort West (SA4), (3) Niewoudtville (SA7) and (4) northern South Africa (Kimberley, Victoria West, Potchefstroom, SA5) (Fig. 1). The latter two clades are geographically separated by the coastal regions due to the Great Escarpment i.e., a plateau edge running parallel with the South African coast and separating the inland plateau from the coastal plains (Grab, 2010). An admixture zone was observed around Laingsburg (SA3) between south-western Cape (SA1-SA2) and Beaufort West (SA4). Laingsburg (SA3) and Beaufort West (SA4) are both located within the lowland part of the Great Karoo (700-800 m above sea level) which is separated from the south-western Cape (SA1-SA2) by the Cape Fold Mountains (Du Preez et al., 2009;Furman et al., 2015) and from the north by the Great Escarpment (Fig. 1). Another admixture zone was suggested between Niewoudtville (SA7) and the south-western Cape populations (SA1) around Vredendal (Measey & Channing, 2003). In its native range X. laevis also displays substantial phenotypic population differentiation. Du Preez et al. (2009), for example, showed that male X. laevis from south of the Cape Fold Mountains were longer and heavier than males from north of the Cape Fold Mountains.
Given its wide native geographical and ecological ranges and high population genetic diversity, it is important to identify the source areas and population(s) of invasive X. laevis populations if one aims to understand the invasion history and invasive potential of this species (Sakai et al., 2001;Lee, 2002;Dlugosch & Parker, 2008;D'Amen, Zimmermann & Pearman, 2013). The origin of invasive X. laevis populations in Sicily and Chile has already been explored using DNA markers (Lillo et al., 2013;Lobos et al., 2014). Similarly, the origin of specimens from animal suppliers i.e., Xenopus-1 Inc. (Dexter, MI, USA) and Xenopus Express (Brooksville, FL, USA), has been assessed (Du Preez et al., 2009). These studies support the assumption that export of X. laevis specimens for laboratory use mainly stemmed from the south-western Cape region in South Africa (Tinsley & McCoid, 1996;Weldon, De Villiers & Du Preez, 2007), although other source areas cannot be excluded, especially for older introduced stocks (Du Preez et al., 2009;L Van Sittert & J Measey, 2015, unpublished data).
The present paper aims to unravel to which extent single or multiple native phylogeographic lineages have contributed to the invasive populations in France and Portugal by comparing mtDNA and nDNA data sampled across the invaded, as well as the native range. In France, it is assumed that X. laevis was introduced at Bouillé-Saint-Paul (Deux Sèvres) from a nearby breeding facility where X. laevis was bred from the 1950s until 1996 (Fouquet, 2001;Fouquet & Measey, 2006). From that breeding facility X. laevis may have escaped repeatedly and was probably released when the facility was definitively closed in 1996 (Measey et al., 2012). Currently, French X. laevis populations occupy an area of approximately 200 km 2 near the city of Saumur (Maine-et-Loire) (Fig. 1). The introduction of X. laevis in Portugal is assumed to be accidental, caused in 1979 by the inundation of a basement of a research institute at Oeiras along river Laje, about 20 km west of Lisbon (Rebelo et al., 2010;Measey et al., 2012). Nowadays, several populations are found in two tributaries of river Tagus, i.e., river Laje and river Barcarena ( Fig. 1) (Rebelo et al., 2010).

Taxon sampling
In total 80 individuals from 32 localities, covering the known area invaded by X. laevis in France (FR) and Portugal (PT) were captured. For comparison with previous work of Lillo et al. (2013) we included two specimens from Sicily (IT, provided by Francesco Lillo). Within the native range of X. laevis, 21 specimens were sampled. Seven specimens came from two localities in the south-western Cape (SA1) i.e., Cape of Good Hope nature reserve and a single dam at Jonkershoek, the location from which animals were shipped for international trade from the 1940s to 1970s (L Van Sittert & J Measey, 2015, unpublished data

DNA amplification and sequencing
Invasive specimens were euthanized with a lethal injection of sodium pentobarbital. Muscle tissue was dissected from invasive specimens, while native wild caught specimens were toe-clipped to obtain tissue samples. Genomic DNA was extracted by means of a NucleoSpin R tissue kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's protocol. Five genomic DNA fragments (circa (ca.) 2,040 bp) representing two mitochondrial and three nuclear gene fragments were amplified using PCR. These genes were selected in order to enable comparison with previously published work on native (Measey & Channing, 2003;Du Preez et al., 2009;Bewick, Anderson & Evans, 2011;Furman et al., 2015) and invasive X. laevis specimens (Lillo et al., 2013;Lobos et al., 2014). Fragments of the mitochondrial cytochrome b gene (Cytb; ca. 280 bp) and 16S ribosomal DNA (16S; ca. 800 bp) were amplified and sequenced with the primer pairs Cytb I /Cytb II and 16Sc-L/16Sd-H (Kessing et al., 1989;Evans et al., 2003). Fragments of the nuclear protein coding genes arginine methyltransferase 6 (Prmt6; ca. 666 bp), androgen receptor isoform α (AR; ca. 402 bp) and microtubule associated serine/threonine kinase-like protein (Mastl; ca. 539 bp) were amplified and sequenced with the primer pairs Exon4_for1/Exon4_rev2, XLAR_for_40/XLAR_rev_431 and Exon13_fora/Exon13_reva (Bewick, Anderson & Evans, 2011). The nuclear primer pairs are assumed to be paralog-specific, hence only amplifying one pair of alleles (see Bewick, Anderson & Evans, 2011). PCR amplifications were run with the conditions reported in Online Resource 3. PCR products were purified with FastAP TM thermosensitive alkaline phosphatase in combination with exonuclease I and subsequently sequenced in both directions. Nucleotide sequences were assembled and edited in CodonCode Aligner (CodonCode Corporation, Dedham, MA, USA). They were aligned together with corresponding well-documented sequences of South African X. laevis available from GenBank using the ClustalW algorithm (Thompson, Higgins & Gibson, 1994) in MEGA v. 6 (Tamura et al., 2013). For comparison, the two Cytb haplotypes found by Lobos et al. (2014) in Chilean (CL) invasive X. laevis populations were included in the Cytb alignment. Cytb and 16S gene fragments obtained by Lillo et al. (2013) were not included in the current alignments as they only partially overlapped due to the use of different primers. Nuclear sequences were converted into haplotypes using the PHASE algorithm (Stephens, Smith & Donnelly, 2001;Stephens & Donnelly, 2003) implemented in DnaSP v. 5 (Librado & Rozas, 2009). All new sequence data were deposited in Genbank (accession numbers in Online Resource 1: https://figshare.com/s/44b71de473fc11e5829d06ec4b8d1f61).

Genetic diversity
The following indices of genetic diversity were estimated using DnaSP v. 5 (Librado & Rozas, 2009): numbers of different haplotypes (h), the number of segregating sites (S), nucleotide diversity (π, i.e., the average number of nucleotide differences per site between two sequences) and haplotype/allelic diversity (Hd, i.e., the probability that two haplotypes drawn uniformly at random from a population are not the same).

mtDNA analysis
Phylogenetic relationships among invasive and native X. laevis mtDNA sequences were reconstructed with Bayesian inference (BI) and Maximum Parsimony (MP). Three alignments were created: two separate Cytb and 16S alignments including only unique alleles and a concatenated Cytb-16S alignment including only unique haplotypes. In all analyses, X. gilli was used as outgroup. Parsimony informative sites were calculated with DnaSP v. 5 (Librado & Rozas, 2009). BI analyses were performed using MrBayes v. 3.2.4 (Ronquist & Huelsenbeck, 2003). Cytb, 16S and Cytb-16S alignments were analysed under a general time reversible (GTR) model with all model parameters estimated from the data and a proportion of invariant sites (+I) as selected by jModeltest v. 2.1.7 (Posada, 2008). BI analyses were run with two different Metropolis-coupled Markov chains for 10 million generations with sampling every 1,000th generation. The average standard deviations of split frequencies, the potential scale reduction factors and the plots of likelihood versus generation were evaluated to ensure convergence based on the log file and using Tracer v1.6 (Rambaut et al., 2013). A total of 25% of the trees were discarded as burn-in and posterior probabilities were calculated for each split from the remaining set of trees. MP trees were estimated with a neighbour joining tree as starting tree using the Phangorn package (Schliep, 2011) in R 3.1.1 (R Development Core Team, 2014). A set of most-parsimonious trees was generated using the parsimony ratchet (Nixon, 1999) with nearest neighbour interchange rearrangement, 10,000 ratchet iterations and up to maximum 10 rounds. Parsimony bootstrap values were obtained with 1,000 bootstrap replicates using the bootstrap.phyDat function in R 3.1.1. All topologies were visualized and edited in respectively Figtree v. 1.4.2 (Rambaut, 2014) and TreeGraph2 v. 2.4.0-456 beta (Stöver & Müller, 2010). Differentiation between native geographical areas and invaded regions was quantified using pairwise Fst values computed based upon nucleotide pairwise distances and tested for significance with 999 permutations using Arlequin v. 3.5.3.1 (Excoffier & Lischer, 2010). The population comparison was done solely for the 16S alignment as this dataset included sequences of most individuals (n = 179) and representatives of all populations.

Autosomal analysis
The minimum number of recombination events within alleles were estimated in DnaSP v.5 (Hudson & Kaplan, 1985;Librado & Rozas, 2009). Given putative recombination, medianjoining networks (MJN) (Bandelt, Forster & Rohl, 1999) were constructed in PopART v.1.7.2 (PopART, 2015) for each nuclear locus separately to illustrate the mutational differences between alleles and their distribution among invasive and native populations (Mardulyn, 2012). Pairwise Fst values were computed based on the allele frequencies to show to what extent alleles are differently sorted among populations (e.g., where the same alleles are found in different populations but with different frequencies). Native populations were defined based upon the geographical regions mentioned in 'Taxon sampling.' Individuals from France and Portugal respectively were treated as two populations. Pairwise Fst values were computed from the allele frequencies with 999 permutations to generate the probability that a random value would be greater than or equal to the observed data using GenAlEx v. 6.5 (Peakall & Smouse, 2012). Individuals with missing data for more than one gene were excluded from the analysis. In order to visualize population differentiation, the pairwise Fst matrix was subsequently used as a distance matrix for a principal coordinates analysis (PCoA) in GenAlEx v.6.5.

RESULTS mtDNA
Mitochondrial alignments were constructed for Cytb and 16S involving 101 and 181 individuals respectively. The concatenated Cytb-16S alignment involved 64 individuals. None of the alignments showed gaps. The Cytb alignment involved16 unique alleles of X. laevis with 42 variable positions of which 34 were parsimony informative. No Cytb sequences were obtained of specimens from the regions SA2, SA3 and SA4. In contrast, all geographic regions were represented in the 16S alignment, which involved 15 alleles of X. laevis with 25 variable positions of which 17 were parsimony informative. The concatenated Cytb-16S alignment involved seven unique X. laevis haplotypes with 43 variable positions of which 39 were parsimony informative (Online Resource 2: https://figshare.com/s/44b71de473fc11e5829d06ec4b8d1f61). The concatenated alignment comprised only specimens from the invaded regions PT, IT and FR and the native regions SA1 and SA6. The BI and MP trees for Cytb and 16S are shown in Fig. 2. These analyses all resolved consistently three well-supported clades comprising X. laevis alleles from the geographic regions (1) SA5, SA6 and FR, (2) SA7 and (3) SA1, PT, FR, IT. Additionally, the phylogeny generated from the 16S data strongly supported (1) SA1, SA2, SA3, FR, PT and IT and (2) SA3-SA4 as well-supported X. laevis clades. BI and MP trees for the concatenated data (Online Resource 2) showed two highly supported clades comprising haplotypes from (1)   nDNA AR sequences (n = 121) showed eight polymorphic positions of which seven revealed heterozygous genotypes. The maximum number of heterozygous positions observed within a single AR sequence was three but that was only observed in two individuals. The two most frequent AR alleles i.e., hap_1 and hap_2 differed by two nonsynonymous substitutions. The Mastl alignment (n = 180) showed 37 polymorphic positions and 108 individuals were heterozygous at 3 ± 2 positions on average. The Prmt6 alignment (n = 176) showed 20 polymorphic positions and 107 individuals were heterozygous at on average 2 ± 1 positions. MJN revealed high allelic diversity at Mast1 and Prmt6 and low allelic diversity at AR (Table 1). In the MJN, the most frequent alleles were often found in different geographical regions (Fig. 4). When combining the allelic frequency information of the three nDNA genes, pairwise population Fst values among all geographic regions were significant except between SA5 and SA6 (Online Resource 5: https://figshare.com/s/44b71de473fc11e5829d06ec4b8d1f61). The first two PCoA-axes explained 75% of the total variance from the pairwise Fst values. There was a clear separation among SA7 and the remnant populations along the first PCoA axis (explaining 28% of the total variance). Negative values along the second PCoA axis (explaining 47% of the total variance) were linked to the northernmost up to the north-eastern native populations (SA4-SA7), while positive values were linked to the invasive (Fr, PT) and southernmost  native populations (SA1-SA3) (Fig. 3). Native populations SA5 and SA6 clustered together (Fig. 3).

South Africa
Haplotype diversity of nDNA and mtDNA within the native range of X. laevis was on average 0.60 ± 0.20 and 0.45 ± 0.18 respectively. Mitochondrial nucleotide diversity within the native regions was either rather low (π ≤ 0.0016: SA1, SA2, SA4, SA6) or high (π ≥ 0.0059: SA3, SA5, SA7). As mentioned previously, mtDNA and nDNA variation was geographically structured within South Africa (Fig. 2, Online Resource 4 and 5) (Furman et al., 2015). The monophyly of the sequences from northern South African regions (SA5) and the newly sampled sites at Rooikrantz Dam in the Eastern Cape (SA6) was consistently strongly supported by both Cytb and 16S (Fig. 2). Furthermore, the pairwise population Fst values based on nDNA and on 16S separately were not significant between these two regions (Online Resource 4 and 5: https://figshare.com/s/44b71de473fc11e5829d06ec4b8d1f61).

France
Two distinct alleles were observed in the introduced French population for both Cytb and 16S (Table 1, Fig. 2). Moreover, these alleles were highly similar or identical to alleles found in two geographically non-overlapping native regions in South Africa, i.e., the northern South Africa and Rooikrantz Dam region (SA5-SA6) and the south-western Cape region (SA1-SA3) (Fig. 2). The most abundant Cytb allele in France (i.e., hap_16; 86% of specimens) was very similar to Cytb sequences in northern South Africa and Rooikrantz Dam populations (SA5-SA6) with on average four nucleotide differences (Fig. 2). Similarly, 86% of the French individuals had a 16S allele (i.e., hap_12) identical to a 16S sequence only found in northern South Africa and Rooikrantz Dam region (SA5-SA6) (Fig. 2). The concatenated Cytb-16S dataset comprised 38 French individuals which represented two different haplotypes. One haplotype was identical to a haplotype found in SA1 (i.e., hap_12_2) and the other haplotype was only found in France but highly similar to haplotypes found in SA6 (i.e., hap_16_12; Online Resource 2). MtDNA haplotype diversity for each marker was low within the French population (Hd ≤ 0.251, Table 1). Pairwise Fst value based on 16S among SA6 and France was not significant (Fst = 0.049; Online Resource 4). Haplotype diversities for nDNA ranged from 0.50 for AR up to 0.80 for Mastl and Prmt6 (Table 1). France comprised 13 Mastl and 16 Prmt6 alleles with seven Mastl and 10 Prmt6 alleles being hitherto only found in France. nDNA allele frequencies in France were most similar to those in native populations from Laingsburg in the south-western Western Cape Province (SA3, Fig. 3, Online Resource 4: https://figshare.com/s/44b71de473fc11e5829d06ec4b8d1f61). No significant difference was observed when comparing nDNA allele frequencies among individuals representing the two different mitochondrial groups (Fst = 0; p = 0.421), however sample sizes were low i.e., 49 and 8 individuals representing northern South Africa-Rooikrantz Dam mtDNA (SA5-SA6) and south-western Cape mtDNA (SA1-SA2) respectively. In comparison with the native geographic regions, mean nucleotide diversity across all loci in France was higher (π = 0.009 ± 0.008 versus π = 0.0039 ± 0.0035) while French haplotype diversity across all loci was comparable (Hd = 0.528 ± 0.287 versus Hd = 0.558 ± 0.323).

Sicily
The Sicilian individuals were identical to native individuals from SA1 for Cytb and from SA1-SA3 for 16S (Fig. 1). Only one concatenated Cytb-16S haplotype was found which was unique to Sicily and highly similar to haplotypes found in FR, PT and SA1. Sicilian AR and Prmt6 alleles were identical to alleles occurring in the native range regions SA1-SA3. Cytb and 16S sequences of Sicilian individuals sampled by Lillo et al. (2013) were not included in the current analyses as they involved different gene regions. However, the mtDNA sequences in the present study were identical to those of Lillo et al. (2013) in the overlapping gene regions (Cytb: ∼242 bp, 16S: ∼374 bp).

DISCUSSION
Since the 1930s, X. laevis has successfully invaded extensive areas worldwide particularly due to its popularity for laboratory use and pet trade (Tinsley, Loumont & Kobel, 1996;Gurdon & Hopwood, 2000;Lobos & Measey, 2002;Crayon, 2005;Faraone et al., 2008;Measey et al., 2012;Herrel & Van Der Meijden, 2014). Reconstructing the invasion history of X. laevis is pivotal to understand the invasion biology and range dynamics of this species and as such, might be critical for developing future management strategies (Sakai et al., 2001;Lee & Gelembiuk, 2008;Prentis et al., 2009). The identification of source populations is particularly relevant when there is extensive population differentiation within the native range because, as well phenotypic as genotypic traits of colonizing individuals might influence the invasion process (Sakai et al., 2001;Lee & Gelembiuk, 2008). Here, the origin of invasive French and Portuguese X. laevis specimens was investigated using DNA sequence data.
The genetic diversity of the invasive Portuguese populations was lower than across the native range of X. laevis in South Africa, but the Portuguese sequences were very similar or even identical to those of native individuals from the south-western Cape region in South Africa. This suggest (1) that the Portuguese populations may be derived from the latter and (2) that the Portuguese population derives from a single colonization event involving a small number of individuals, most likely stemming from one and the same source population in the south-western Cape region. This is in line with what one would expect from the historical data which attributes the introduction of X. laevis in Oeiras (Portugal) to a single accidental flood of a basement of a research institute in 1979 (Rebelo et al., 2010). Analogous to the Portuguese samples, the identical sequences shared by the Sicilian (Lillo et al., 2013), Chilean (Lobos et al., 2014) and south-western Cape populations samples support the idea that X. laevis was imported in Sicily and Chile from wild populations in the south-western Western Cape Province. Indeed, the export of X. laevis from the Western Cape Province is well documented, especially between 1940 and 1974 when there was only one official supplier in South Africa, i.e., Jonkershoek Fish Hatchery (Weldon, De Villiers & Du Preez, 2007).
In contrast to the X. laevis colonization in Portugal, two distinct and divergent mtDNA lineages were detected in France. These lineages were related to two geographically nonoverlapping native regions in South Africa. A majority of the French individuals possessed mtDNA highly similar or identical to a phylogeographic lineage from northern South Africa and Rooikrantz Dam (SA5-SA6, further referred to as the northern lineage). The other French individuals had mtDNA identical to native individuals from the south-western Cape (SA1-SA2). Haplotype and sequence diversity in the French population were relatively high and comparable to those in native South African regions. The mtDNA data thus suggest that two distinct phylogeographic lineages i.e., south-western Cape and northern lineage, contributed to the invasion of X. laevis in France. Although the nDNA data are consistent with this suggestion, they solely indicate the Laingsburg region (SA3) and the southwestern Cape (SA1-SA2) as possible source areas of the French nDNA alleles. This might be explained by either a scenario of a single South African source area where south-western and northern lineages admix or a scenario in which France was invaded by individuals from several distinct South African source areas followed by admixture of the colonizing animals. In the case of the first scenario, the most likely source population based upon the nDNA data would be the Laingsburg region which is considered as an admixture zone among the south-western Cape regions SA1-SA2 and the Beaufort West region SA4 (Du Preez et al., 2009;Furman et al., 2015). However, individuals from Laingsburg do not show mtDNA of the northern populations. Hence, the occurrence of the main mitochondrial alleles (Cytb: hap_16; 16S: hap_12) in French individuals cannot be explained by this scenario. Conversely, historical data supports the second scenario in which the X. laevis invasion in France stems from different source areas as animal suppliers are known from both the south-western Western Cape province and the Eastern Cape Province (Port Elizabeth)(L Van Sittert & J Measey, 2015, unpublished data). Unfortunately, information concerning export from the latter region is limited. In contrast, there is some export data from the Western Cape province in the period that the French breeding facility was active   (Weldon, De Villiers & Du Preez, 2007). Until 1974, there was one official animal supplier i.e., Jonkershoek Fish Hatchery. Subsequently, trading was left to private enterprises for which export and collection information is hitherto unknown. Yet, since 1990, permits were licenced to four South African animal suppliers restricting the collection of X. laevis to man-made water bodies (Weldon, De Villiers & Du Preez, 2007). Around the same time, in 1989, the ownership of the French breeding facility changed (Measey et al., 2012). Taking all these circumstances together, it seems likely that during its 56 years of existence, the French breeding facility might have imported X. laevis repeatedly from different South African sources. Moreover, it seems not unlikely that secondary trading occurred within Europe or even worldwide e.g., among commercial breeding facilities and/or research institutes.
To the best of our knowledge, the French X. laevis population is the first example in which two geographically non-overlapping phylogenetic lineages participated in a X. laevis invasion. As mentioned previously, two distinct mtDNA lineages, related to on the one hand the northern regions of South Africa and on the other hand the southwestern Cape regions, are present in the French population. However, no population structure could be found in the invasive French populations based on nDNA. In contrast, within the native South African range these mtDNA lineages are significantly divergent based on nDNA. The latter observation suggests a potential role for ongoing admixture within the invasive French population. Likewise, admixture might have occurred within breeding facilities among specimens from different source populations preceding the French invasion. The combination of genetic variation from multiple phylogeographic lineages might explain the high level of genetic diversity within France, which was comparable to, or even higher than, the diversity observed in native regions. An increase in genetic diversity within invasive population relative to native populations due to multiple introductions of genetically divergent source populations has also been demonstrated in invasions of other organisms elsewhere (Novak & Mack, 1993;Kolbe et al., 2004;Kolbe et al., 2008;Lavergne & Molofsky, 2007). Moreover, intraspecific admixture among previously isolated multiple divergent genetic lineages is often suggested to play an important role in driving the success of colonising populations (Sakai et al., 2001;Lavergne & Molofsky, 2007;Kolbe et al., 2008;Lee & Gelembiuk, 2008;Rius & Darling, 2014). Yet, a species' invasiveness is not only function of the amount of genetic variation, but even more importantly of the nature of adaptive genetic variation (Dlugosch et al., 2015). Concerning X. laevis, it is clear that single introduction events even with low levels of genetic diversity have proven to be highly successful in invading non-native areas such as Chile and Sicily. However, in order to predict the species' potential range expansion the intraspecific variation should be taking into account (Pearman et al., 2010;D'Amen, Zimmermann & Pearman, 2013), especially for the French population, in which two genetically distinct and environmentally divergent South African phylogeographic lineages have contributed to the invasion. Hence, the French population offers a study system for investigating the extent to which the combination of genetic variation from divergent phylogeographic lineages might have influenced the French invasion and/or might influence its future range expansion. In sum, the genetic structure of X. laevis in its native South African range allowed us to investigate the geographic origins of invasive X. laevis populations in France and Portugal. The current analyses showed that native phylogeographic lineages are not equally represented in invasive European X. laevis populations.