Samples and molecular analysis
Sixteen natural populations (N = 474) were sampled among the three phylogeographic units of D. labrax distribution area: Atlantic Ocean (AO), WM, and EM basins (Table 1). Most samples have been already analyzed in previous studies (Table 1). Nineteen microsatellite and one minisatellite markers were used, classified into two classes of markers: gene-associated loci (hereafter GAL) and gene-independent loci (GIL). A marker was classified as a GAL when it was situated either inside the span of an annotated gene or at <2 kilobases (kb) from one extremity of a coding sequence (CDS). The limit of 2 kb was retained because it generally encompasses the proximal promoter of most genes in vertebrates. Proximal promoters are known to be under functional constraints and, together with their closeness from CDS, mini- and microsatellite loci they contain may have greater probability to be influenced by selection (e.g., Li et al. 2004; Gemayel et al. 2010). A symmetrical size limit of 2 kb was also retained after the 3′-end of the CDS while – as far as we know – documented functional constraints are less reported on this side of genes. Those limits are, however, operating limits to classify GAL and GIL; they do not mean that loci located farther from CDS cannot be impacted by selection (and vice versa). The GAL comprises one mini- and three microsatellite markers that have been specifically designed for their association with growth hormone (GH: μGH and mGH), somatolactin (SL: μSL), and insulin-like growth factor (IGF-1: μIGF-1; Quéré et al. 2010; description and conditions herein) and seven microsatellite loci from two simple sequence repeat (SSR)-enriched genomic libraries (Tsigenopoulos et al. 2003; Chistiakov et al. 2004) that appeared to be close to CDS afterward. The GIL class contains nine loci originating from the same SSR-enriched genomic libraries than GAL (Table 2).
Table 1. Samples used in this study
|Basin||Abbreviation||Population||Country||Number of individuals||Geographic location||Origin of sample|
|AO (31)||BB||Bay of Biscay||France||31||45°44′19″N||1°25′11″O||Fritsch et al. (2007)|
|WM (365)||GOU||La Goulette||Tunisia||30||36°50′06″N||10°19′59″E||Guinand et al. (2008)|
|ICH||Ichkeul||Tunisia||49||37°10′21″N||9°39′55″E||Bahri-Sfar et al. (2000)|
|MAR||Marsala||Italia||24||37°48′09″N||12°25′11″E||Naciri et al. (1999)|
|ANN||Annaba||Algeria||24||36°54′47″N||7°46′36″E||Naciri et al. (1999)|
|SBD||Sabaudia||Italia||16||41°19′36″N||13°59′30″E||Lemaire et al. (2000)|
|FIUM||Fiumicino||Italia||22||41°51′36″N||12°70′02″E||Lemaire et al. (2000)|
|OR||Etang de l'Or||France||54||43°34′35″N||4°01′40″E||This study|
|SET||Sète||France||38||43°23′06″N||3°42′19″E||Guinand et al. (2008)|
|MUR||Murcia||Spain||48||37°51′50″N||0°43′31″O||Lemaire et al. (2005)|
|EM (98)||SYR||Syria||Syria||32||35°34′50″N||35°31′36″E||This study|
|CYP||Cyprus||Cyprus||15||34°33′03″N||32°58′36″E||Boutet et al. (2008)|
Table 2. Microsatellite loci used in this study, including size ranges, GenBank accession numbers, fluorochrome labeling, and PCR primers
|Locus||Category||Annotation||Motif||GenBank Accession||Size range (bp)a||Fluorochrome|| ||Primers (5′3′)|
| DLA0041 ||GIL||–||(TC)8(AC)21AA(AC)3AA(AC)3|| DQ363864 ||167–201||FAM||F||AAAAGGAACAGCCCTCCAC|
| DLA0044 ||GAL||Unknown||(TC)18TTTT(TC)6CTCC|| DQ363867 ||105–129||HEX||F||TCCGCTCCGCACCGAGTGAC|
| DLA0051 ||GAL||MAPK3c||(GT)16|| DQ363874 ||149–181||ROX||F||AGGTTCTTGGCCTGGGAATC|
| DLA0060 ||GAL||Bestrophin 3d||(CA)12(TA)3AA(CA)2|| DQ363883 ||119–131||FAM||F||GAGAGTTCATCCTGTTCGCTC|
| DLA0061 ||GIL||–||(TG)14|| DQ363884 ||153–167||FAM||F||AAAGGCCAGTGAAACTCATGT|
| DLA0066 ||GILb||–||(AG)22|| DQ363887 ||135–161||PET||F||GTTGACCGGAGTCCTAGC|
| DLA0068 ||GIL||–||(CA)7CGCACG(CA)3|| DQ363889 ||247–266||NED||F||CAACACCTGTTCCTCTGAACC|
| DLA0070 ||GAL||Unknown||(AC)30|| DQ363891 ||126–155||VIC||F||TCTGCTTGCATCTGTGGAAT|
| DLA0073 ||GILb||–||(CT)36|| DQ363894 ||157–181||NED||F||CATGACTTCATGTGCTAATGTCC|
| DLA0075 ||GIL||–||(CA)15|| DQ363896 ||180–186||PET||F||CACATACACAAGCTTAACCC|
| DLA0078 ||GAL||MAN2A1e||(AG)29|| DQ363899 ||223–243||VIC||F||AAGACTGGACCTCTGGAGACC|
| DLA0081 ||GALb||PPP2R5Af||(CA)16|| DQ363902 ||202–222||PET||F||GACGAAGACTTCAGACGAGCTAT|
| DLA0086 ||GIL||–||(AC)26|| DQ363907 ||191–205||FAM||F||GCTAGAGGATTCATGTCGCTT|
| DLA0089 ||GAL||LLGL1g||(GT)15|| DQ363910 ||127–135||NED||F||ACGAGTAATGAGGACCCA|
| DLA0096 ||GILb||–||(TG)16|| EF471091 ||256–270||ROX||F||AACTTAGTGAAGTAACTTGTGGCAA|
| DLA0097 ||GIL||–||(GT)4GC(GT)2GC(GT)13|| EF471092 ||216–240||HEX||F||GCTGCAGGAGTGTGAGAGG|
Markers from SSR libraries were amplified in two multiplexes PCR (Table 2). For the first multiplex, amplification conditions have been reported in Guinand et al. (2008). The second multiplex was optimized for a final volume of 10 μL with 10× Taq buffer, 30 mmol/L MgCl2, 2.7 mmol/L dNTP, 10 ng of genomic DNA and 4 μmol/L of each primer. Reactions were performed on a PTC-200 (MJ Research, St Bruno, Québec, Canada) as follows: an initial denaturation at 95°C for 3 min, 35 cycles at 94°C for 45 s, annealing at the 58°C for 45 s, and 72°C for 45 s, followed by a final elongation at 72°C for 10 min. Genotyping was performed on ABI PRISM 3130xl or 3700 DNA Analyser (Life Technologies, St Aubin, France), using 5′-labeled reverse primers and the GeneScan® TM-500 LIZ Size Standard (Life Technologies) as internal size standard.
Figure 1. Map of sampling locations; samples are abbreviated as in Table 1. The colored symbols for samples belonging to each metapopulation will be used in other figures. Location of the Almeria-Oran Front and the Siculo-Tunisian Strait are reported.
Download figure to PowerPoint
Among-sample comparisons were assessed by estimating levels of population differentiation using , an estimator of FST (Weir and Cockerham 1984). Estimates of were independently computed for three data sets (i.e., the full set of 20 loci, then GAL and GIL sets). Their significance was tested according to the permutation procedure available in GENETIX and corrected for multiple tests according to Benjamini and Yukitieli (2001) false discovery rate (FDR; Narum 2006). Distance trees were inferred from Reynolds et al. (1983) coancestry genetic distance matrix  by the Neighbor procedure of the phylogenetic package Phylip 3.6 (Felsenstein 2005). Trees were visualized using Treeview (Page 1996). Following O'Reilly et al. (2004), we explored the relationship between and Hexp to check for effects of size homoplasy. We specifically controlled whether GIL and GAL markers were differently affected. We further explored the possible role of selection in shaping differentiation in the full, GAL, and GIL data sets using neutrality tests developed by Beaumont and Nichols (1996; hereafter B&N and implemented in LOSITAN Antao et al. 2008, http://popgen.eu/soft/lositan/), by Vitalis et al. (2001; implemented in DETSEL Vitalis et al. 2003, http://www.genetix.univ-montp2.fr/detsel.html), and by Foll and Gaggiotti (2008; hereafter: F&G and implemented in BayeScan, http://www-leca.ujf-grenoble.fr/logiciels.htm). All tests were first performed among populations within each basin, then among the three basins except for DETSEL, specifically designed for pairwise comparisons then between each pair of basins. Outlier loci detected using several methods would be stronger candidate to the action of selection than loci detected once. The B&N and, when considered, the DETSEL tests were both performed according to P = 0.95 and P = 0.99 successively. F&G test considered a Log10(BF) of 1.3 and 2 (BF = Bayes Factor) corresponding to posterior probabilities of locus effects of 0.95 and 0.99, respectively.
The program MIGRATE-n v3.0 (Beerli 2008) was used to infer M, the migration rate, among basins only (M = m/μ, where m is the immigration rate per generation). Contrary to most Fst-based statistics, MIGRATE-n provides estimation of asymmetric migrations among (meta)populations, indicating if a (meta)population is a net donor or net receiver of individuals over evolutionary times. It hence provides a picture how tension zones may function. These calculations used the Brownian mutation model and the mutation rate was considered equal for all loci (μ = 10−4) even for GAL. We used coalescent maximum likelihood (ML) based on Markov Chain Monte Carlo with Hastings Metropolis importance sampling to infer the various parameters (Beerli and Felsenstein 1999, 2001). Fst estimates among basins were used as initial parameters for the estimation of Θ and M in MIGRATE-n. For each locus, the ML was run for 10 short and five long chains with 50,000- and 10,000-recorded genealogies, respectively, after discarding the first 1000 genealogies (burn-in) for each chain. One of every 20 reconstructed genealogies was sampled for both the short and long chains. We used an adaptive heating scheme with four concurrent chains. Analyses were all performed in triplicates either on the full set of loci or on the GAL and GIL sets independently.
As a complement to estimates of population differentiation and asymmetric migration rates, we computed the probability of membership of individuals to each metapopulation (i.e., individuals should have higher probability of membership in the metapopulation they were sampled). Assignments of individuals to populations and associated probabilities were inferred with the software STRUCTURE v2.3 (Pritchard et al. 2000; http://pritch.bsd.uchicago.edu/structure.html) by setting K = 3 (i.e., three basins) after verification of the most likely number of independent population clusters was K = 3 using Evanno et al.'s (2005) ΔK method (details not reported). The software used a Monte Carlo Markov Chain (MCMC) Bayesian clustering method that maximizes the within-cluster Hardy–Weinberg and linkage equilibriums. The admixture model with noncorrelated allele frequencies was used for the full, GAL, and GIL data sets. A burn-in length of 50,000 iterations and subsequent 500,000 additional MCMC iterations were carried out. Individuals were assigned to clusters based on the highest probability of membership (-value). Five replicates were independently performed, giving reproducible results. Average -values (±1 standard deviation) of individuals to each metapopulation were hence computed from those individual assignments.
Analyses using MIGRATE-n and STRUCTURE were first performed using initial sample sizes of the three basins (Table 1), but, as sample sizes were unbalanced, we also performed analyses with 31 individuals in each basin (i.e., the number of individuals from the AO basin). Results were highly reproducible and qualitatively comparable to those obtained with the full set of individuals and will not be reported.