Appendix S1.Supplementary Details on the ABC Analyses

Supplementary materials

Appendix S1.Supplementary details on the ABC analyses

Mutation model.Coalescent simulations assume a mutation model for each type of loci. The mutation model for microsatellite loci was a generalized stepwise-mutation (GSM) model (Estoup et al. 2002) with two parameters: a mean mutation rate () and mean of the geometric distribution for the length, in repeat numbers, of mutation events () drawn from uniform prior distributions (: [10-3 – 10-4] and : [0.1-0.3], seeTable S1). We accounted for variation in µmic and P among loci by drawing their individual values from a gamma distribution (Table S1). These settings allowed for large mutation rate variance across loci (i.e. range of 10−5 to 10−2). We also considered mutations inserting or deleting a single nucleotide in the microsatellite sequence.

We used jModelTest 2.1 (Darriba et al. 2012) to identify the best substitution model and estimate its parameter. The best mutation model describing the mtDNA sequence evolution was a HKY+I+G model (Hasegawa et al. 1985) with a proportion of constant sites of 86.5%, and a shape of the gamma distribution of mutations among sites equal to 0.63 (Table S1). We assumed a per-site and per-generation mutation rate ranging uniformly between 1 × 10−7 and 1 × 10−5, as found in the literature (Alter & Palumbi 2009; Fontaine et al. 2010).

Summary statistics. Overall, 78 summary statistics describing within and among population genetic diversity were calculated with DIYABC. Within population statistics for microsatellite loci included the mean number of alleles per locus, expected heterozygosity, allele size variance, MGW statistic of Garza & Williamson across loci (Garza & Williamson 2001). Between population statistics for microsatellites included the mean number of alleles between two populations, FST(Weir & Cockerham 1984), shared allele distance (Chakraborty & Jin 1993), and (δµ)2 Goldstein’s distance (Goldstein et al. 1995). For the mtDNA data, the descriptive statistics within populations include the number of segregating sites, the mean pairwise differences and its variance, Tajima’s D, and the number of private segregating sites. Statistics computed between groups were the mean of within sample pairwise differences, mean of between sample pairwise differences, and HST between two samples (Hudson et al. 1992).

Figure S1.Flow chart about inference of population history using ABC in program DIYABC.

Table S1.Model specification, prior distributions for demographic parameters and mutation model parameters for the ABC analysis (Figure 2).

Demographic Parameter / Type / Prior
N1 / N / UN~[1 – 4,000]
N2 / N / UN~[1 – 4,000]
N3 / N / UN~[1 – 15,000]
N4 / N / UN~[1 – 10,000]
Na / N / UN~[10 – 10,000]
t1 (≤ t3) / T / UN~[10 – 5,000]
t2 (≤ t3) / T / UN~[10 – 5,000]
t3 (≤ t4) / T / UN~[100 – 5,000]
t4 / T / UN~[100 – 5,000]
Tr / T / UN~[10 – 5,000]
dB / T / UN~[5 – 15]
Nbn / N / UN~[1 – 50]
Microsatellites mutation parameter / GSM(40 steps allowed)
/ UN~[1 x 10-4 – 1 x10-3]
Gµmic / GA~[1 x 10-5, 1 x 10-2, , 2]
/ UN~[1 x 10-1, 3 x 10-1]
GP / GA~[1 x10-2, 9 x 10-1, , 2]
SNI / LU~[1 x10-8, 1 x10-5]
GSNI / GA~[1 x10-9, 1 x10-4, SNI, 2]
MtDNAmutation parameter / HKY (p-inv: 86.5, α: 0.634)
µseq / UN~[1 x 10-7, 1 x 10-5]
K1 / UN~[0.050, 20]

Type of parameters: (N) effective population size, (T) time of the event in generation. Uniform distribution (UN) with 2 parameters: min and max; Gamma distribution (GA) with 4 parameters: min, max, mean and shape; Log-Uniform (LU) distribution with 2 parameters: min and max. See Figure 2for the demographic parameters of each model tested. The mutation model parameters for the microsatellite loci were the mean mutation rate (µmic), the parameter determining the shape of the gamma distribution of individual loci mutation rate (P), and the Single NucleotideInsertion rate (SNI). The Mt-DNA mutation model was a HKY with two variable parameters, the per-site and generation mutation rate (µseq) and the transition/transversion ratio (K1) parameter, and two fixed parameters, the proportion of constant sites (p-inv.), and the shape of the Gamma distribution of mutations among sites (α).

Table S2. Model choice procedure and ABC performance analysis for step a in Figure 2.

SC1 / SC2 / SC3 / SC4 / SC5 / SC6 / SC7 / SC8 / SC9 / SC10 / SC11
Post. Pr. / 3.1% / 64.7% / 27.5% / 0.0% / 0.1% / 0.1% / 0.0% / 1.4% / 2.0% / 0.1% / 1.2%
95%CI / [0.0 – 8.6] / [62.6 – 66.7] / [24.0 – 31.0] / [0 – 5.7] / [0 – 5.8] / [0 – 5.8] / [0 – 5.7] / [0 – 7.0] / [0 – 9.0] / [0 – 5.7] / [0 – 6.8]
Confidence in SC selection
SC1 / SC2 / SC3 / SC4 / SC5 / SC6 / SC7 / SC8 / SC9 / SC10 / SC11
D1 / 60.6% / 9.4%* / 5.6% / 4.2% / 6.8% / 8.8% / 3.0% / 5.8% / 6.6% / 9.6% / 7.6%
D2 / 3.2%† / 68.6% / 22.4%† / 0.2%† / 0.2%† / 0.2%† / 0.0%† / 4.8%† / 2.2%† / 1.4%† / 0.2%†
D3 / 1.2% / 16.8%* / 66.0% / 0.0% / 0.2% / 0.0% / 0.0% / 1.2% / 3.4% / 0.4% / 0.2%
D4 / 2.0% / 0.0%* / 0.2% / 70.6% / 11.0% / 0.2% / 0.2% / 0.2% / 6.4% / 2.8% / 1.6%
D5 / 1.8% / 0.0%* / 0.0% / 12.4% / 68.0% / 0.0% / 0.0% / 0.2% / 1.4% / 3.8% / 4.6%
D6 / 4.2% / 0.4%* / 0.0% / 0.2% / 0.4% / 62.8% / 18.0% / 2.4% / 1.0% / 6.8% / 5.0%
D7 / 0.8% / 0.0%* / 0.0% / 0.0% / 0.0% / 9.0% / 70.6% / 2.4% / 0.0% / 0.2% / 1.2%
D8 / 1.6% / 2.0%* / 1.8% / 0.2% / 0.0% / 0.6% / 3.6% / 55.8% / 14.2% / 7.4% / 0.2%
D9 / 3.0% / 1.8%* / 2.6% / 3.0% / 1.0% / 0.2% / 0.2% / 13.2% / 51.4% / 6.4% / 0.2%
D10 / 12.8% / 0.2%* / 1.0% / 3.0% / 6.0% / 8.8% / 1.8% / 13.0% / 12.0% / 60.2% / 1.6%
D11 / 8.8% / 0.8%* / 0.4% / 6.2% / 6.4% / 9.4% / 2.6% / 1.0% / 1.4% / 1.0% / 77.6%
Model check (number of outlying statistics)
P < 0.05 / 12 / 8 / 7 / 15 / 14 / 13 / 13 / 11 / 11 / 13 / 16
P < 0.01 / 2 / 1 / 2 / 5 / 2 / 3 / 7 / 3 / 1 / 5 / 6
P < 0.001 / 2 / 1 / 1 / 3 / 3 / 2 / 4 / 2 / 2 / 3 / 2

Table S3. Model choice procedure and ABC performance analysis for step b in Figure 2.

Model selection / SC1 / SC2 / SC3 / SC4
Post. Pr. / 0.7% / 15.7% / 29.2% / 54.4%
95%CI / [0.0 – 2.5] / [14.2–17.3] / [27.1–31.3] / [53.0–55.7]
Confidence in model choice
SC1 / SC2 / SC3 / SC4
D1 / 95.60% / 11.00% / 13.40% / 13.6%*
D2 / 2.00% / 37.40% / 17.20% / 22.4%*
D3 / 1.40% / 19.20% / 44.80% / 22.4%*
D4 / 1.0%† / 32.4%† / 24.6%† / 41.60%
Model check (number of outlying statistics)
P < 0.05 / 15 / 8 / 10 / 7
P < 0.01 / 3 / 1 / 0 / 1
P < 0.001 / 2 / 1 / 2 / 1

Table S4.Model choice procedure and ABC performance analysis for step c in Figure 2.

SC1 / SC2 / SC3 / SC4 / SC5
Post. Pr. / 35.7% / 36.7% / 17.7% / 5.5% / 4.4%
95%CI / [35.0 – 36.5] / [36.0 – 37.4] / [17.1 – 18.3] / [5.0 – 6.0] / [3.9 – 4.9]
Confidence in model choice
SC1 / SC2 / SC3 / SC4 / SC5
D1 / 51.4% / 36.4%*,†† / 33.0%†† / 26.0%†† / 25.8%††
D2 / 21.6%†,** / 27.0% / 17.0%† / 12.6%† / 11.6%†
D3 / 11.2%** / 14.6%* / 30.4% / 8.8% / 12.4%
D4 / 10.8%** / 14.4%* / 11.0% / 34.4% / 20.0%
D5 / 5.0%** / 7.6%* / 8.6% / 18.2% / 30.2%
Model check (number of outlying statistics)
P < 0.05 / 7 / 7 / 2 / 7 / 5
P < 0.01 / 1 / 1 / 3 / 1 / 1
P < 0.001 / 1 / 1 / 0 / 1 / 1

D – proportion of case in which the simulation-based model choice procedure was able to select a scenario as the most probable with non-overlapping confidence intervals of the posterior probabilities of each scenario.* Type-I or α-error rate for SC2. † Type-II or β-error rate and 1- Σβi provides the power of the model choice procedurefor SC2.** Type-I or α-error rate for SC1. †† Type-II or β-error rate and 1- Σβi provides the power of the model choice procedurefor SC1.

Table S5.

Model check procedure for the step c in Figure 2.

Statistics / Observed / Prob. (Ssimul. < Sobs.)
SC1 / SC2 / SC3 / SC4 / SC5
Microsatellites
MGW_3 / 0.7616 / 0.0285 (*) / 0.0190 (*) / 0.0055 (**) / 0.0165 (*) / 0.0145 (*)
FST_1&2 / 0.059 / 0.917 / 0.9665 (*) / 0.454 / 0.916 / 0.6405
MtDNA
NHA_1 / 4 / 0.0030 (**) / 0.0070 (**) / 0.0055 (**) / 0.0015 (**) / 0.0035 (**)
NHA_2 / 5 / 0.0170 (*) / 0.0260 (*) / 0.0765 / 0.0285 (*) / 0.052
MPD_1 / 0.8818 / 0.0420 (*) / 0.0705 / 0.0635 / 0.0440 (*) / 0.0445 (*)
DTA_1 / -1.5873 / 0.0190 (*) / 0.0300 (*) / 0.0230 (*) / 0.0195 (*) / 0.0190 (*)
MNS_1 / 4.8333 / 0.0305 (*) / 0.0460 (*) / 0.0470 (*) / 0.0255 (*) / 0.0395 (*)
NH2_1&2 / 6 / 0.0000 (***) / 0.0005 (***) / 0.0040 (**) / 0.0000 (***) / 0.0005 (***)
NH2_1&4 / 19 / 0.0385 (*) / 0.0420 (*) / 0.0605 / 0.0350 (*) / 0.0450 (*)
HST_1&2 / 0.4054 / 0.9530 (*) / 0.9670 (*) / 0.907 / 0.9525 (*) / 0.9315

Evolutionary scenarios SC1 to SC5 are represented in Figure 2c. The probability Prob.(Ssimul.Sobs.) given for each summary statistic was calculated from 1,000 pseudo-observed datasets simulated from the posterior distributions of parameters obtained under the focused scenario. Corresponding tail-area probabilities (p-values) were obtained as Prob. (Ssimul.Sobs.) and 1.0 - Prob. (Ssimul.Sobs.) for Prob.(Ssimul.Sobs.)≤0.5 and > 0.5, respectively (*, **, *** = tail-area probability < 0.05, < 0.01 and < 0.001, respectively). In addition to the statistics used during the model choice procedure, the model check procedure used also the two sample statistics including the mean genetic diversity, mean size variance, and the classification index for microsatellite loci.For mtDNA the model check procedure used also within sample statistics including the number of haplotype, mean number of the rarest nucleotide at segregating sites and its variance, and the two sample statistics comprising the number of haplotypes and number of segregating sites. Only significant summary statistics for at least one scenario are shown. Abbreviations for the summary statistics are as follows: Mean Garcia-Williamson index (MGW), FST-statistics (FST); number of mtDNA haplotypes (NHA), mean pairwise differences (MPD); Tajima’s D (DTA); mean number of the rarest nucleotide at segregating sites (MNS); the number of distinct haplotypes in two pooled samples (NH2); HST value between populations. Numbers following each statistics refer to the population(s) considered following Figure 2.

Table S6.Parameter estimationsfrom scenario SC1 and 2 combined (in Figure 2c). Mode, median and x%Quantile (Qx) are provided.

Parameter / Mode / Q2.5 / Median / Q97.5
N1-CS / 2,160 / 864 / 2,060 / 3,560
N2-CN / 1,990 / 678 / 1,960 / 3,660
N3-PA / 12,200 / 6,360 / 11,600 / 14,700
N4-PM / 4,810 / 1,500 / 4,730 / 9,200
Na / 5,000 / 766 / 5,130 / 9,560
t1 / 128 / 42 / 133 / 341
t2 / 379 / 117 / 447 / 1,130
t3 / 516 / 215 / 722 / 2,390
t4 / 2,200 / 881 / 2,810 / 4,860
/ 2.50E-04 / 1.58E-04 / 3.01E-04 / 6.60E-04
/ 2.68E-01 / 1.51E-01 / 2.53E-01 / 3.00E-01
SNI / 1.00E-08 / 1.00E-08 / 1.60E-08 / 1.50E-07
µseq / 2.32E-06 / 1.43E-06 / 2.55E-06 / 5.55E-06
K1 / 1.12E+01 / 5.47E-01 / 1.00E+01 / 1.94E+01

N is expressed in number of diploid individuals, times(t) are provided in generation before present.

Figure S2. Sampling locations and genetic group of origin for individuals included in ecological and/or morphological analyses (Ncoastal = 21 and Npelagic = 42).

Figure S3. External morphometric measurements of stranded bottlenose dolphins.

Appendix S2.Supplementary details on the preparation of the samples for stable isotope analyses.

Prior to isotopic analyses, samples were cut in small pieces (less than 1 mm) and dried at 45°C in an oven for 48h. As lipids are depleted in 13C relative to other tissue components (De Niro & Epstein 1977), they were extracted from skin samples prior to stable isotopes analyses (SIA) of carbon and nitrogen but not of sulfur. Lipid extraction had no effect on SIA of sulfur: differences between measurements with lipid extraction and without lipid extraction were less than 0.2‰, which is in the precision range of the measurements. The samples were agitated with 2 ml of cyclohexane for 1 h. Then, they were centrifuged for 10 min at 3500 tours/min and the supernatant containing lipids was discarded. This protocol was repeated until the supernatant was transparent. The sample was dried in an oven for 48 h. Subsamples were weighted (0.3–0.4 mg for carbon and nitrogen SIA and 1.0-1.3 mg for sulfur SIA) with a microbalance and packed in tin cups.Sulfur, carbon and nitrogen isotope ratios were determined by a continuous flow mass spectrometer coupled to an elemental analyser. Stable isotope values are presented in the conventional δ notation relative to IAEA-1 and IAEA-2, Vienna Pee Dee Belemnite and atmospheric N2 for δ34S, δ13C and δ15N values respectively. Typical precisions for isotopic measurements were inferior to 0.20‰. We checked that the C/N mass ratios of all the samples were below 4, thus verifying that lipid extraction was operative.

Figure S4a.Skin δ13C and δ34Svalues for coastal and pelagic bottlenose dolphins. Solid lines indicated Standard Ellipses Areas corrected for sample sizes (SEAc) and dotted lined Convex Hull Areas. Their respective areas values are given in the legend. The star indicates the possible migrant (see text).

Figure S4b. Skin δ13C and δ15N valuesfor coastal and pelagic bottlenose dolphins. Solid lines indicated Standard Ellipses Areas corrected for sample sizes (SEAc) and dotted lined Convex Hull Areas. Their respective areas values (‰²) are given in the legend. The star indicates the possible migrant (see text).

Figures S5a to S5c. Measures of uncertainty of Bayesian ellipse areas (SEAB) based on 106 posterior draws indicating 95, 75 and 50% credibility intervals from light to dark grey respectively for a) skin δ13C andδ34S, b)skin δ34S and δ15N and c)skin δ13C and δ15N values. Black dots represent the mode of SEAB and SEAc.

Table S7. Diet composition in relative abundance (%N) and ingested biomass (%M) of coastal (N = 6) and pelagic (N=24) bottlenose dolphins. 95% confidence intervals (CI 95%) are given in parentheses.

Coastal / Pelagic
%N (CI95%) / %M (CI95%) / %N (CI95%) / %M (CI95%)
Sprattussprattus / 10.7 (0-28.9) / 0.8 (0-2.9)
Ammodytidae / 33.7 (0-78.5) / 5.2 (0-20.5) / 2.9 (0-8.5) / 0.1 (0-0.3)
Scomberscombrus / 3.6 (0.3-10.2) / 11.6 (0.7-31.6)
Trachurusspp. / 1.1 (0-5.7) / 1.1 (0-4.6) / 10.5 (4.0-20.0) / 5.4 (1.8-11.6)
Mugilidae / 35.9 (0-73.4) / 29.8 (0-63.9) / 1.1 (0.1-2.8) / 6.5 (0-19)
Sparidae / 3.2 (0-11.2) / 1.9 (0-6.5)
Dicentrarchuslabrax / 3.3 (0-15.4) / 6.9 (0-22.6) / 4.4 (0.1-13) / 5 (0.1-14.9)
Merlucciusmerluccius / 24.6 (12.9-41) / 54.6 (28.2-75.5)
Micromesistiuspoutassou / 5.4 (0-22.7) / 0.1 (0-0.4) / 18 (2.4-39.9) / 1.4 (0.4-3)
Trisopterusspp. / 5.4 (0-22.6) / 31.1 (5.8-66.9) / 10.7 (5.9-16.5) / 2.1 (0.9-4.3)
Pollachiusspp. / 1.1 (0-5.8) / 7.6 (0-24.9)
Otherfish / 3.3 (0-16.5) / 1.3 (0-4.3) / 4 (0-10.5) / 3.6 (0-9.5)
Loligospp. / 3.3 (0-7.7) / 6.1 (0-14.3) / 2.1 (0.7-4.2) / 6.7 (0.7-17.9)
Othercephalopods / 2.2 (0-10.2) / 0.1 (0-8.2) / 2 (0-5.6) / 1 (0.1-10.3)
Shrimp / 1.1 (0-4.5) / <0.1 (0-0) / 1.9 (0-5.3) / <0.1 (0-0.1)

References

Alter SE, Palumbi SR (2009) Comparing evolutionary patterns and variability in the mitochondrial control region and cytochrome b in three species of baleen whales. Journal of Molecular Evolution68, 97-111.

Chakraborty R, Jin L (1993) A unified approach to study hypervariable polymorphisms: statistical considerations of determining relatedness and population distances. EXS67, 153-175.

Darriba D, Taboada GL, Doallo R, Posada D (2012) jModelTest 2: more models, new heuristics and parallel computing. Nature Methods9, 772-772.

De Niro MJ, Epstein S (1977) Mechanism of carbon isotope fractionation associated with lipid-synthesis. Science197, 261-263.

Estoup A, Jarne P, Cornuet JM (2002) Homoplasy and mutation model at microsatellite loci and their consequences for population genetics analysis. Molecular Ecology11, 1591-1604.

Fontaine MC, Tolley KA, Michaux JR, et al.(2010) Genetic and historic evidence for climate-driven population fragmentation in a top cetacean predator: the harbour porpoises in European water. Proceedings of the Royal Society B-Biological Sciences277, 2829-2837.

Garza JC, Williamson EG (2001) Detection of reduction in population size using data from microsatellite loci. Molecular Ecology10, 305-318.

Goldstein DB, Linares AR, Cavallisforza LL, Feldman MW (1995) An evaluation of genetic distances for use with microsatellite loci. Genetics139, 463-471.

Hasegawa M, Kishino H, Yano TA (1985) Dating of the human ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution22, 160-174.

Hudson RR, Slatkin M, Maddison WP (1992) Estimation of levels of gene flow from DNA sequence data. Genetics132, 583-589.

Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution38, 1358-1370.