Molecular Ecology (2013) 22, 1003–1018
doi: 10.1111/mec.12161
Contrasting patterns of genome-wide polymorphism in the native and invasive range of the marine mollusc Crepidula fornicata AUT,*† MARION BALLENGHIEN,*†‡§ F L O R E N T I N E R I Q U E T , * † C L A I R E D A G U I N - T H I EB ER IQUE VIARD*† N I C O L A S B I E R N E § ¶ and F R ED *UPMC Univ Paris 06, Team Diversity and Connectivity in Coastal Marine Landscapes, UMR 7144, Station Biologique de Roscoff, 29680 Roscoff, France, †CNRS, Laboratory Adaptation and Diversity in the Marine Environment, UMR 7144, Station Biologique de Roscoff, 29680 Roscoff, France, ‡Universite Montpellier 2, Place Eugene Bataillon, 34095 Montpellier, France, §CNRS, Institut des Sciences de l’Evolution, Laboratory Phylogenie et Evolution moleculaire, UMR 5554, Universite Montpellier 2, Place Eugene Bataillon, 34095 Montpellier Cedex 05, France, ¶CNRS, Institut des Sciences de l’Evolution, UMR 5554, Station Mediterraneenne de l’Environnement Littoral, 2 rue des Chantiers, 34200 Sete, France
Abstract Selection processes are believed to be an important evolutionary driver behind the successful establishment of nonindigenous species, for instance through adaptation for invasiveness (e.g. dispersal mechanisms and reproductive allocation). However, evidence supporting this assumption is still scarce. Genome scans have often identified loci with atypical patterns of genetic differentiation (i.e. outliers) indicative of selection processes. Using microsatellite- and AFLP-based genome scans, we looked for evidence of selection following the introduction of the mollusc Crepidula fornicata. Native to the northwestern Atlantic, this gastropod has become an emblematic invader since its introduction during the 19th and 20th centuries in the northeastern Atlantic and northeastern Pacific. We examined 683 individuals from seven native and 15 introduced populations spanning the latitudinal introduction and native ranges of the species. Our results confirmed the previously documented high genetic diversity in native and introduced populations with little genetic structure between the two ranges, a pattern typical of marine invaders. Analysing 344 loci, no outliers were detected between the introduced and native populations or in the introduced range. The genomic sampling may have been insufficient to reveal selection especially if it acts on traits determined by a few genes. Eight outliers were, however, identified within the native range, underlining a genetic singularity congruent with a well-known biogeographical break along the Florida. Our results call into question the relevance of AFLP genome scans in detecting adaptation on the timescale of biological invasions: genome scans often reveal long-term adaptation involving numerous genes throughout the genome but seem less effective in detecting recent adaptation from pre-existing variation on polygenic traits. This study advocates other methods to detect selection effects during biological invasions—for example on phenotypic traits, although genome scans may remain useful for elucidating introduction histories. Keywords: adaptation, microsatellites
AFLP,
biological
invasion,
Crepidula
fornicata,
genome
Received 10 August 2012; revision received 15 October 2012; accepted 29 October 2012
Correspondence: Florentine Riquet, Fax: +33298292324; E-mail:
[email protected]; Frederique Viard, Fax: +33298292324; E-mail:
[email protected] © 2013 Blackwell Publishing Ltd
scan,
1004 F . R I Q U E T E T A L .
Introduction Natural selection is involved in adaptive responses to a variety of environmental changes driven by natural or anthropogenic modifications (e.g. climate change, habitat alterations and fisheries). However, studying selection is difficult. Distinguishing the effects of selection from those of other evolutionary forces can be challenging because selection does not frequently leave distinguishable footprints in the genome (Li et al. 2012). For instance, gene surfing (i.e. alleles that reach high frequencies just by chance during population expansion) has often been interpreted as evidence of selection events, although it is just a sign of genetic drift (Excoffier & Ray 2008). In addition, even when local selection maintains phenotypic differences, it does not necessarily produce genetic differentiation at all of the genes underlying the trait (Le Corre & Kremer 2012), and those that do show genetic structure are expected to affect neutral variation on a very small chromosomal scale (Charlesworth et al. 1997; Bierne 2010). Finally, adaptation can be locally impeded or even offset by gene flow (i.e. ‘gene swamping’, Lenormand 2002), a pattern exemplified in many marine species. Large population sizes and the long-lived dispersive phases of many marine species mean that populations are only weakly genetically structured (i.e. connected by high gene flow), opposing local adaptation (Nielsen et al. 2009). For example, selection has not been demonstrated in contrasted environments in the highly dispersive periwinkle Echinolittorina hawaiiensis (Tice & Carlon 2011). Nevertheless, selection is known to play a key role in maintaining adaptive polymorphism in dispersive marine species (e.g. Koehn et al. 1980; Wilding et al. 2001; Bradbury et al. 2010). In this context, invasive marine invertebrates make interesting case studies. Nonindigenous species successfully survive, reproduce and disperse into new environments characterized by different ecological conditions from those prevailing in the native range. However, introduced phenotypes may not be the best-adapted to the new habitat, and adaptive changes following introduction are thought to be an important process behind invasion success (Stockwell et al. 2003; Sax et al. 2007). Selection of advantageous traits may be rapid, conferring a competitive advantage to the introduced species over the native species within a few generations (Reznick & Ghalambor 2001; Prentis et al. 2008). Identifying such adaptive changes can help elucidate the reasons for the successful establishment of nonindigenous species, a central issue in the study of biological invasions (Sax et al. 2007). Selection processes may be particularly effective in marine invasive species, which generally display large populations and a high level of genetic diversity in introduced populations (e.g. Simon-Bouhet et al.
2006; Roman & Darling 2007 for a review). This high genetic diversity may increase the likelihood of encountering adapted genotypes, creating new genetic combinations and limiting the effect of genetic drift as well as favouring selection on pre-existing genetic variation (Stockwell et al. 2003; Facon et al. 2006; Barrett & Schluter 2008). We here tested for the occurrence of postintroduction selection processes by looking for genomic regions shaped by selection in one of the most emblematic invasive molluscs, the slipper limpet Crepidula fornicata. This long-lived species (living up to 10 years) displays a typical marine bentho-pelagic life cycle alternating a 2- to 7-week free-floating larval stage and a sessile adult stage. It inhabits shallow bays and estuaries. Native to the northwestern Atlantic and present from Canada to Mexico, this marine gastropod was first introduced in Great Britain at the end of the 19th century and beginning of the 20th century, probably with spats of the American oyster Crassostrea virginica (Hoagland 1985). The species expanded during the 20th century along the coasts of Europe (France, Belgium, the Netherlands and Germany). In addition, massive accidental introductions occurred in the 1970s along the French Atlantic coasts and in the Mediterranean Sea with the intentional introduction of the Pacific oyster Crassostrea gigas from the northeastern Pacific (Puget Sound), where the Pacific oyster and the slipper limpet had previously been introduced (in the 1930s; Blanchard 1997; Hoagland 1985). Today, C. fornicata is well established in the northeastern Pacific and in Europe (mainly from Norway to Spain). Previous studies of C. fornicata have documented high genetic polymorphism in both native and introduced populations (Hoagland 1985; Dupont et al. 2003), concurring with its complex introduction history. However, the previous studies failed to cover both the full native and introduced ranges of C. fornicata and only a limited number of allozyme markers were used, preventing efficient detection of selection footprints. Changes in selection regime between the native and invasive ranges may be first expected because of differences in species assemblages between the two ranges (e.g. release from natural enemies and new trophic interactions, Colautti et al. 2004; Dunn 2009; Strayer & Hillebrand 2012). For instance, Le Cam & Viard (2011) documented a major infestation of introduced populations of C. fornicata by a native shell borer Cliona celata. Adaptation for invasiveness may also occur, for instance through selection on traits involved in dispersal ability (e.g. pelagic larval duration) or reproductive assurance (e.g. an increased number of individuals per mating groups in C. fornicata, Le Cam et al. 2009). In addition, in its invasive range, C. fornicata is found across a large latitudinal gradient encompassing several marine ecoregions and biogeographical provinces (Boreal, © 2013 Blackwell Publishing Ltd
S E L E C T I O N A N D I N V A S I O N I N C . F O R N I C A T A 1005 Lusitanian and Mediterranean Sea; Spalding et al. 2007) characterized by different set of abiotic conditions and community assemblages probably to exert different selective pressures. For example, the annual temperature mean at 10 m in depth varied from 8.7 °C in Tjarno (Sweden) to 16.1 °C in Sete (France, Mediterranean Sea) in the European introduction range (Levitus data, Fig. S1, Supporting information). Temperature has been shown to be an important factor for the reproduction and settlement success of C. fornicata (Bohn et al. 2012 and references herein). Such latitudinal gradient and correlated environmental changes are also documented in the American native range spanning over cold temperate, warm temperate and tropical northwestern Atlantic regions (Spalding et al. 2007); for instance, annual temperature mean is varying with latitude from 7.3 °C in Nahant (MA) to 24.4 °C in Longboat Key (FL). One increasingly used approach to detect selection effects is the use of a large number of markers spread throughout the genome (i.e. the genome scan approach; Storz 2005) to identify loci that display levels of population differentiation higher than those under neutral expectations. These so-called outlier loci are directly under selection or, more likely, linked with a locus under selection (Bierne et al. 2011). Outlier loci have often been identified along latitudinal gradients characterized by noticeable environmental and climatic variations (e.g. Bonin et al. 2006). In nonmodel species for which genome data are not available (e.g. C. fornicata), the Amplified Fragment Length Polymorphism (AFLP) technique is a fast and cost-effective method to provide the required number of markers over the whole genome (Bonin et al. 2006). In this study, we developed a genome scan approach using 344 polymorphic loci (17 microsatellite and 327 AFLP markers) to compare the genetic diversity of native populations with populations sampled in two distinct areas of introduction of C. fornicata. We had three objectives: (i) to confirm the previously suggested scenario of introduction (i.e. large propagule pool associated with a large genetic diversity), (ii) to examine the possible presence of postintroduction selection events by comparing native vs. invasive populations and (iii) to compare the results of a genome scan conducted in the native range—where adaptation has evolved on a long term—with a genome scan conducted in the invaded range—where adaptation can only have evolved on a short term.
Pierce (FL, USA), which was sampled in April 2010. Seven populations were sampled in the native range of the species (along the eastern coast of North America) and 15 in introduced ranges: one from the Pacific coast of the USA and 14 from Europe (North Sea, Channel, Atlantic coast and Mediterranean Sea; Fig. 1, Table 1). This sampling is representative of the latitudinal distribution and correlated environmental gradients (see Fig. S1, Supporting information for an example with temperature data) in the native and invasive range of C. fornicata. An average of 32 individuals per population was examined. Individuals were either frozen at 80 °C or preserved in 96% ethanol for genetic analyses. Genomic DNA was extracted from adults using either the phenol–chloroform method or using the DNeasy Tissue Kit (Qiagen) following the manufacturer’s protocol.
Microsatellite and AFLP genotyping Microsatellite genotyping. Individuals were genotyped at 17 microsatellite loci: 12 loci were EST-SSRs (Riquet et al. 2011), whereas five loci were not derived from an EST library but isolated from anonymous nuclear libraries [CfH7 (Dupont et al. 2006), Cf8 (Proestou 2006), CfCA4, CfCA2 (Dupont & Viard 2003) and E4 (details in Riquet et al. 2011)]. Loci were amplified by
Population samples Number Native range Nahant Fairhaven Long Island Somers Point Chesapeake Fort Pierce Longboat Key
1 2 3 4 5 6 7
Intoduced ranges Mud Bay
8
Tjarno Limfjord Sylt Island Yerseke Lawrenny Portsmouth Canvey Island Gravelines Port-en-Bessin Bay of Morlaix Bourgneuf Fouras Arcachon Sète
9 10 11 12 13 14 15 16 17 18 19 20 21 22
Materials and methods Sample and DNA collection Twenty-two populations of Crepidula fornicata were sampled in 2001–2002, except for the one from Fort © 2013 Blackwell Publishing Ltd
Fig. 1 Distribution (black line) and sampling locations of Crepidula fornicata: native populations are numbered 1–7 (black dots), and introduced populations are numbered 8–22 (white dots).
1006 F . R I Q U E T E T A L . Table 1 Genetic diversity of the population samples (named after the locality in which they were sampled) for 17 microsatellite markers and 327 AFLP markers Genetic diversity estimates based on 17 microsatellite markers Population Native range Nahant Fairhaven Long Island Somers Point Chesapeake Fort Pierce Longboat Key Introduced ranges Mud Bay Tjarno Limfjord Sylt Island Yerseke Lawrenny Portsmouth Canvey Island Gravelines Port-en-Bessin Bay of Morlaix Bourgneuf Fouras Arcachon Sete
Country—State
USA—Massachussetts USA—Massachussetts USA—New York USA—New Jersey USA—Virginia USA—Florida—East coast USA—Florida—West coast USA—Washington Sweden Denmark Germany The Netherlands United Kingdom United Kingdom United Kingdom France France France France France France France
Genetic diversity estimates based on AFLP markers
N
Nall
Ar
He
FIS*
N
PLP
Hj
Br
202 32 23 32 31 13 53 18 481 32 32 30 34 33 32 35 32 32 31 32 30 32 32 32
24.47 10.94 9.65 12.65 11.76 8.06 14.47 9.18 22.47 11.00 10.94 9.41 10.35 9.76 10.65 12.29 11.35 11.29 10.41 10.35 10.47 10.12 10.12 9.94
7.28 6.89 7.07 7.56 7.44 7.40 7.28 7.32 6.88 7.05 7.02 6.67 6.65 6.59 6.86 7.25 7.23 7.15 6.81 6.92 6.96 6.68 6.62 6.74
0.656 0.630 0.647 0.652 0.649 0.633 0.640 0.639 0.644 0.645 0.669 0.640 0.635 0.645 0.654 0.645 0.646 0.647 0.606 0.635 0.651 0.627 0.627 0.661
0.168 0.137 0.160 0.154 0.181 0.099 0.124 0.201 0.154 0.159 0.075 0.195 0.196 0.234 0.200 0.151 0.172 0.119 0.185 0.146 0.146 0.072 0.070 0.154
151 27 23 31 12 13 29 16 428 29 20 32 28 32 27 30 28 31 24 28 25 32 31 31
30.3 30.3 28.7 30.9 30.6 28.1 30.3 33.0 31.7 31.8 32.7 28.4 33.6 30.0 28.1 34.3 30.6 33.3 27.5 31.5 33.3 31.5 34.3 34.3
0.102 0.099 0.101 0.101 0.124 0.114 0.101 0.111 0.102 0.102 0.109 0.100 0.107 0.105 0.104 0.116 0.108 0.101 0.092 0.104 0.123 0.105 0.107 0.100
1.38 1.34 1.36 1.36 1.43 1.40 1.35 1.41 1.37 1.35 1.38 1.34 1.37 1.36 1.36 1.41 1.38 1.37 1.30 1.37 1.44 1.39 1.38 1.36
N: number of individuals genotyped, Nall: average number of alleles per locus, Ar: allelic richness, He: expected heterozygosity, FIS: fixation index, PLP: proportion of polymorphic loci (with 95% criteria), Hj: Nei’s gene diversity and Br: band richness. For geographical locations of sampled populations, see Fig. 1. *P-values associated with HWE < 0.0001 in each case.
polymerase chain reactions (PCRs) following protocols detailed in the study by Riquet et al. (2011). Amplification products were separated by electrophoresis on an ABI 31309l DNA sequencer (Applied Biosystems), and alleles were scored using GeneMapper® v. 4.0 (Applied Biosystems). A part of the whole data set (11.2%) was genotyped twice with the same results. AFLP genotyping. The AFLP protocol from the study by Vos et al. (1995) was adapted to Crepidula fornicata following Bonin et al. (2005) with minor modifications. Briefly, genomic DNA (between 50 and 150 ng) was digested 3 h at 37 °C with restriction endonucleases MseI and EcoRI (5 U, New England Biolabs, NEB). EcoRI and MseI double-stranded adapters were ligated to restriction fragments with T4 DNA ligase (5 U, NEB) overnight at 16 °C. Preselective amplification reactions were carried out with two primer pairs (EcoRI+A/MseI+A and EcoRI+C/MseI+C). For the selective amplification, four primer pairs (Table 2) were selected of 62 tested, ranging from 50 to 600 bp.
They yielded polymorphic peaks with high signal above noise, thus maximizing scoring peaks with reliability. The four selective PCR products were pooled, and 2 lL was electrophoresed on an ABI 31309l DNA sequencer (Applied Biosystems) with 0.3 lL of GeneScan 600LIZ Size Standard (Applied Biosystems) and 9.7 lL of deionized formamide. Negative controls at each step of the AFLP protocol were included to detect any contamination or artefactual peaks. More than 12% of the samples were randomly chosen and genotyped twice for each primer combination to test for the repeatability of the AFLP profiles, as recommended by Bonin et al. (2005) and as required for error rate estimation in AFLPSCORE (Whitlock et al. 2008; see below). AFLP profiles were carefully examined with GENEMAPPER v.4.0 (Applied Biosystems). Peak heights (in relative fluorescence units, RFU) for all individuals for each primer combination were exported into the R script AFLPSCORE v. 1.4 (Whitlock et al. 2008). Replicate samples were used to optimize the AFLPSCORE threshold (i.e. threshold that minimizes © 2013 Blackwell Publishing Ltd
S E L E C T I O N A N D I N V A S I O N I N C . F O R N I C A T A 1007 Table 2 AFLP marker analysis: primer pair combination (with ABI dyes used) and criteria for retaining AFLP loci for data analyses
ABI Dyes NED PET FAM VIC
Selective EcoRI Primers (5′–3′)
Selective MseI Primers (5′–3′)
GACTGCGTAC CAATTCACG GACTGCGTAC CAATTCAAC GACTGCGTAC CAATTCCTG GACTGCGTAC CAATTCCTC
GATGAGTCCT GAGTAAAGC GATGAGTCCT GAGTAAACC GATGAGTCCT GAGTAACAG GATGAGTCCT GAGTAACTC
Number of markers used as polymorphic markers
% of markers used for genotyping
Nrep (%)
Initial markers
Mismatch error %
e1.0
e
606
12.96
187
1.26
10.73
0.02
110
58.82
594
12.81
96
2.96
8.13
0.07
37
38.54
597
12.57
101
2.48
8.75
0.13
85
84.16
597
12.39
116
1.72
7.86
0.14
95
81.90
Total SD
12.68 0.25
500
2.11 0.76
8.87 1.30
0.09 0.05
327
65.40
N
0.1
The number of individuals (N) and the percentage of replicate samples (Nrep) are indicated with the number of peaks initially analysed (initial markers), the error rate estimation (mismatch error, e1.0 and e0.1) and the number and proportion of polymorphic markers used for further analyses.
genotyping error and maximizes the number of valuable markers) and to determine repeatability rates for the AFLP markers (% of markers that are consistently scored). Markers probably to contribute high error rates to the data were discarded, and the threshold that gave the lowest error rates for each primer combination was used to determine the final genotype for all individuals.
Microsatellite and AFLP data analyses Genetic diversity. For microsatellites, allele frequencies, the average number of alleles (Nall) and allelic richness (Ar, i.e. the expected number of alleles corrected for sampling size, based on a rarefaction method) were estimated for each sample using FSTAT 2.9.3 (Goudet 1995). We used GENEPOP 4.1 software (Rousset 2008) to estimate expected heterozygosity (He), fixation index (FIS) and to test for departure from Hardy–Weinberg equilibrium in each population (10 000 dememorization steps, 500 batches and 5000 iterations per batch). To adjust the P-values for multiple tests, we computed the q-values using the QVALUE package in the R software (Storey 2002). For AFLPs, estimates of allelic frequencies were computed according to a Bayesian method with nonuniform prior distribution of allele frequencies (Zhivotovsky 1999) using the software AFLPSURV (Vekemans 2002). For each sample, allele frequencies were used to calculate the proportion of polymorphic loci (PLP, i.e. loci with allele frequencies from 5% to 95%), as well as unbiased estimates of expected heterozygosity (Hj or Nei’s gene diversity). Band richness (Br), an analogue of allelic richness, was © 2013 Blackwell Publishing Ltd
computed on the AFLP data using the software (Coart et al. 2005).
AFLPDIV
Genetic population structure. The genetic structure between sampled populations was computed by calculating an estimate of FST (Weir & Cockerham 1984) using GENEPOP 4.1 for microsatellites. Exact tests for population differentiation (10 000 dememorization steps, 500 batches and 5000 iterations per batch) were carried out to test for differences in allele distributions among native and introduced populations and between pairwise populations. For AFLPs, FST estimates were calculated and tested based on a permutation procedure (10 000 permutations) using AFLP-SURV. The overall genetic structure among population samples was depicted using a correspondence analysis (CA) computed on allelic frequencies with GENETIX software (Belkhiret al. 1996–2004) for microsatellites and the R package ADEGENET 1.3–4 (Jombart & Ahmed 2011) for AFLPs. As recommended by Jombart et al. (2009), we excluded alleles with frequencies lower than 5% (microsatellites) or markers with one of the two alleles displaying frequencies lower than 5% (AFLPs). Principal component analysis (PCA) was also conducted using PCAGEN software (http://www2.unil.ch/popgen/softwares/pcagen.htm) for microsatellites and the R package ADEGENET 1.3–4 (Jombart & Ahmed 2011) for AFLPs. Genetic structure was also investigated by analysing both microsatellite and AFLP data with an individualbased Bayesian clustering implemented in the software STRUCTURE 2.3.3 (Pritchard et al. 2000). All analyses were performed using the University of Oslo Bioportal (https://www.bioportal.uio.no/). For each value of K
1008 F . R I Q U E T E T A L . (ranging from 1 to 25), we ran 30 replicate chains of 150 000 Markov Chain Monte Carlo (MCMC) iterations and discarded the 50 000 burn-in iterations. An admixture model with correlated allele frequencies was applied without using a priori information on population origin. The most likely number of clusters was ascertained using the method proposed by Evanno et al. (2005) that determines the maximum value of DK, a measure for the second-order rate of change in the likelihood of K, using the online software STRUCTURE HARVESTER (Earl & vonHoldt 2011). Infiles produced by STRUCTURE HARVESTER were used in CLUMPP (Jakobsson & Rosenberg 2007) to determine individual assignment probabilities that best matched with all replicate runs. We used DISTRUCT (Rosenberg 2004) to visualize the individuals’ assignment to the different clusters. Outlier detection. To identify outliers among all population samples and between each pair of samples (22 populations and 231 comparisons), we used two approaches. Both are methods that are based on the identification of loci showing genetic differentiation (FST-based estimates) significantly different from the estimated genome average. For the first approach, we used Beaumont & Nichols’s (1996) method, using the software LOSITAN (Antao et al. 2008) (for codominant microsatellite markers) and DFDIST (for dominant AFLP markers). Briefly, for each marker, this method uses simulations to estimate the distribution of FST according to the degree of heterozygosity under neutral expectations. Loci that do not fit neutral expectations (i.e. unusually high or low FST values) are identified as outliers. In this study, the trimmed FST (the highest and lowest 30% of empirical FST values are removed) was used as the targeted FST value in simulations, as recommended by Caballero et al. (2008). A total of 50 000 loci were simulated to generate FST values, and different values for the mutation rate l were initially tested for the simulations (103–106) with identical results. The false discovery rate was controlled, and q-values were computed using the QVALUE R package (Storey 2002). For the second approach, we used the software BAYESCAN, which implements the method developed by Foll & Gaggiotti (2008). It uses a Bayesian method and a logistic regression model to estimate directly the posterior probability (i.e. posterior odds) that a given locus is subject to selection. Briefly, locus population FST values are decomposed as a linear combination of a locusspecific component (a) and a population-specific component (b). A locus is assumed to be under selection when a is necessary to explain the observed pattern of diversity. For each locus, a reversible jump MCMC explores two alternative models (with or without a) and estimates their relative posterior probabilities. The
Jeffrey Interpretation scale (Jeffreys 1961) was used to identify the outliers (Foll & Gaggiotti 2008). Outliers found in both methods were examined in detail.
Results Level of genetic diversity in the native and introduced ranges Table 1 provides the estimates of genetic diversity for each population, based on microsatellite and AFLP markers. All 17 microsatellite markers were polymorphic with 2–84 alleles per locus. The number of alleles per population sample ranged from 8.06 (Chesapeake) to 14.47 (Fort Pierce). Allelic richness (Ar) was quite stable, ranging from 6.59 (Yerseke) to 7.56 (Long Island). The same held for He, ranging from 0.606 (Porten-Bessin) to 0.669 (Tjarno). Departures from Hardy– Weinberg equilibrium were found in each population mainly due to a deficiency in heterozygotes at eight loci. Because these heterozygote deficiencies may be partly due to null alleles, all analyses (diversity, structure and outlier analyses) were performed with and without these eight loci, and results were identical. For AFLPs, a total of 327 polymorphic markers were selected, ranging from 50 to 500 bp. The number of markers varied among primer pairs, yielding from 37 to 110 polymorphic AFLP markers (Table 2). Of the 683 individuals genotyped, individuals that were not scored for the four primer combinations were excluded from the data set (no missing values). The average mismatch error rate for the four primer combinations was 2.11 0.76, the average e1.0 error rate was 8.87 1.30, and e0.1 was 0.09 0.05 (Table 2). This data set is the best trade-off between the number of markers and their quality (Holland et al. 2008; Whitlock et al. 2008). The values of PLP, Hj and Br were similar across samples: they ranged from 27.5 (Port-en-Bessin) to 34.3 (Portsmouth, Arcachon, Sete), from 0.092 (Port-en-Bessin) to 0.124 (Somers Point) and from 1.30 (Port-en-Bessin) to 1.44 (Bourgneuf), respectively. The gene diversity of native and introduced populations was not significantly different for either type of marker (group comparison with FSTAT software: P = 0.99 and P = 0.85 for AFLPs and microsatellites, respectively).
Genetic structure among and between native and introduced populations Genetic structure was estimated using 17 microsatellite markers (data set 1), 327 AFLP markers (data set 2) or 319 AFLP markers, excluding the identified outliers (see below; data set 3). The exact test revealed a significant difference in allele frequencies among the 22 sampled © 2013 Blackwell Publishing Ltd
S E L E C T I O N A N D I N V A S I O N I N C . F O R N I C A T A 1009 populations using any of the three data sets (‘All population samples’ in Table 3). In all data sets, the genetic structure among samples from the native range was higher than the genetic structure between the native and introduced ranges. In the native range, overall significant genetic structure was mainly due to the genetic differentiation of the Longboat Key (LBK) sample, located at the southern edge of the native range: lower FST estimates were observed when LBK was discarded (Table 3). Many pairwise comparisons revealed significant differences, but the only consistent results were observed with LBK, showing significant genetic differences with all of the 21 other populations (Table S1, Supporting information). As shown in Table 3 (Table S1, Supporting information), the genetic structure estimated with 327 AFLP markers was higher than with the AFLP set excluding outliers, especially when LBK was involved. The eight outliers were thus the loci showing the most clearly the genetic differentiation of the LBK sample. In addition, low genetic structure was observed among introduced populations. The correspondence analysis (CA) output is shown in Fig. 2 for microsatellites. The results are not provided for AFLPs as
