Page 14 - Colliard_ali_2010

Colliard et al. BMC Evolutionary Biology 2010, 10:232 Page 14 of 16
http://www.biomedcentral.com/1471-2148/10/232

DNA extraction and microsatellite data generation Isolation by distance was investigated with Mantel
DNA was extracted using the Qiagen DNeasy™kit. tests (FSTAT) by regressing pairwise F ST /(1-F ST ) against
Microsatellite primer (Bcal μ10) described by Rowe the natural logarithm-transformed Euclidean geographic
et al. [58] and six pairs of microsatellites (BaturaC203, distances [64], for B. balearicus and B. siculus indepen-
C205, C218, C223, D105, D5) developed by Colliard et dently, selecting localities with ten or more samples.
al. [24] were selected for analysis, due to their level of Two sets of analyses were performed, including or not
polymorphism and applicability in both species. PCR- the three populations at the contact zone which showed
amplification, electrophoresis and allele scoring were mitochondrial introgression. A principal component
performed as described in [24]. Multiplexing of PCR analysis was conducted with PCAGEN [28] to visualize
products was performed. For D105 and D5 we used a pairwise differentiation among populations (F ST ), with
mixture in a 1:3 ratio; Bcal μ10 was 5% diluted and 1000 randomizations of genotypes to test for signifi-
mixed with C223 and C203 products in a 1:2:2 ratio. cance of axes.
The third multiplexing involved products of C218 and In order to better determine potential signs of intro-
C205 in a ratio of 1:2. gression between the two species, we used the Bayesian
clustering algorithm of STRUCTURE v. 2.2 [25]. We
Mitotyping, sequencing and analyses of mitochondrial used the admixture model and allowed for correlated
DNA and nuclear DNA allele frequencies between populations, as recommended
We used a mitotyping approach described in Colliard by the authors for cases of subtle population structure.
et al. [24] to detect the mtDNA haplotype group for all We tested a range of cluster numbers (K) from 1 to the
individuals. We also sequenced 577 bp of the mitochon- number of localities per analysis, plus an additional
drial control region (D-loop) [as described in [18]] in three to enable us to potentially infer subtle structure.
5
162 individuals from across the study area (Figure 2). Each run, replicated 10 times, consisted of 10 itera-
4
An intron of alpha-Tropomyosine, situated between tions, after a “burn-in” of 10 . To infer which K best fits
exons 5 and 6 was amplified, cloned and sequenced as our data, we applied the ad hoc ΔK statistic developed
described by Stöck et al. [23]. We used homologous by Evanno et al. [26]. We performed three analyses: (1)
sequences from 21 individuals from throughout the sam- all Sicilian localities plus the lab-cross individuals; (2) all
pling area plus one individual from a museum collection B. siculus localities, and (3) all B. balearicus localities.
coming from a locality near pop. 12. Four of these 22
green toads came from the contact zone with assignment Identification of hybrids
values varying from 68% to 92% to either parental species. Four alternative approaches were used to identify
All mitochondrial and nuclear sequences were sub- hybrids. First, we used the cluster assignment value
mitted to GenBank (Acc.-Numbers HM852594- (STRUCTURE, parameters as above, K = 2), considering
HM852744, in part from [23]). Maximum likelihood as hybrids all individuals with assignment values below
(ML) phylogenies of mitochondrial and nuclear 90%. Second, we performed two analyses using NEW-
sequence alignment were generated using PhyML ver- HYBRIDS v.1.1 Beta3 [27] to assign Sicilian toads to
sion 2.4.5 [59] and using HKY models according to genotypic classes (parental, F 1 ,F 2 , backcrosses). The
MrModeltest [60]. In each case, we choose a BioNJ as a method computes Bayesian posterior probability that an
starting tree, and optimized topology, branch length and individual belongs to each of these different hybrid
rate parameters. Other parameters were used as defaults classes while simultaneously estimating allelic frequen-
of the program. We generated bootstrap values based cies for parental species. Runs were repeated several
on 1000 resampled data sets. times with varying lengths of the “burn-in” and number
of sweeps, as recommended in the program manual.
Genotype data analyses The first analysis was based on all Sicilian individuals.
To exclude genotyping errors due to null alleles, stutter- The second focused on populations where hybrids were
ing and allelic drop-out, Micro-Checker v.2.2.3 [61] was shown to occur or were likely to do so (pop. 12 to 16
used. Genotypic linkage disequilibrium between each and 18), with addition of few pre-assigned non-hybrid
pair of loci per population was tested using ARLEQUIN individuals from “pure” populations (pop. 9 to 11 and
v. 3.0 [62]. Significance was adjusted for multiple tests 17), using the z option as recommended by the authors
using Bonferroni corrections. Estimation of heterozygote [27]. Third, we identified individuals as hybrids if they
deficits and its significance was assessed using FSTAT v. were assigned by STRUCTURE to one clade but con-
2.9.4 [63]. For each microsatellite marker and species, tained mtDNA from the other (= cyto-nuclear disequili-
we estimated genetic diversity by calculating number of brium). Finally, we compared the microsatellite allele
alleles (Na), and within-sample expected heterozygosity composition of Sicilian B. siculus and B. balearicus
(Hs) (Additional file 1). populations far apart from the contact zone (i.e.,