RESULTS
The four genetic distance/similarity matrices are presented in Supplementary Material 2. They are the basis for constructing the population dendrograms. Six of the ten different dendrograms exhibit the same topology of two main branches, where each is divided into two clusters. Cluster 1, henceforth the Eastern Atlantic cluster (EA), consists of Biarritz, the Canaries and Madeira; cluster 2, the Western Mediterranean cluster (WM), consists of Bizerte, Málaga and Melilla. Clusters 1 and 2 both belong to the same main group. Cluster 3, the Mid-Mediterranean cluster (MM), consists of Bastia, Birżebbuġa and Pantelleria; Cluster 4, the Eastern Mediterranean cluster (EM), consists of Bodrum, Dubrovnik, Fažana, Larnaca and Rethymno. Clusters 3 and 4 both belong to the same main group. Two representatives of this configuration are given in Figs. 2 and 3. We checked the robustness of the topology presented in this group of dendrograms by repeating the Ward’s analysis of Fig. 2, 14 times – each time omitting a different location (thus remaining with 13 locations). Except for the omitted location, the remaining structure of the four above-mentioned clusters did not change. Further demonstration of this division into four clusters is displayed by the PCoA analysis in Fig. 4. The two axes account for 86.5% of the variance.
Two of the ten different dendrograms exhibit a slightly different topology of two main branches. One consists of the Eastern Atlantic cluster only, whereas the other is divided into two sub-branches. One of these sub-branches consists of the Eastern Mediterranean cluster and the other is further divided into two clusters – the Western Mediterranean and the Mid-Mediterranean clusters. Thus, the setup of the four above-mentioned clusters also exists in these two dendrograms. A representative of this configuration is given in Fig. 5.
The remaining two dendrograms place Pantelleria in the Western Mediterranean cluster, next to the geographically nearby Bizerte. A representative of this configuration is given in Fig. 6.
The mean number of different alleles per position, the mean expected heterozygosity measures and the mean percentage of polymorphic positions in each population are presented in Table 2.
The distribution of the number of alleles per position is not statistically different from a normal distribution (Kolmogorov-Smirnov test, p = 0.364). In accordance with the results of the multi-variate analysis, as presented by the genetic dendrograms, we divide the 14 populations into four groups: EA (the Eastern Atlantic cluster), WM (the Western Mediterranean cluster), MM (the Mid-Mediterranean cluster) and EM (the Eastern Mediterranean cluster). The means (± se) of each group are: 1.361 ± 0.030, 1.782 ± 0.042, 1.857 ± 0.039 and 1.942 ± 0.020, for EA, WM, MM and EM, respectively (see Fig. 7a). Using a one-way ANOVA for testing the differences in the mean number of alleles per position between these four groups, we getF 3,10 = 68.849, p < 0.001. Post-hoc: EA vs. WM p < 0.001; EA vs. MM p< 0.001; EA vs. EM p < 0.001; WM vs. MMp = 0.841; WM vs. EM p = 0.020; MM vs. EM p = 0.408.
The distribution of the expected heterozygosity is not statistically different from a normal distribution (Kolmogorov-Smirnov test, p= 0.464). The means (± se) of each group are: 0.069 ± 0.008, 0.154 ± 0.007, 0.205 ± 0.005 and 0.230 ± 0.003, for EA, WM, MM and EM, respectively (see Fig. 7b). Using a one-way ANOVA for testing the differences in expected heterozygosity between these four sub-clusters, we get F 3,10 = 169.899, p < 0.001. Post-hoc: EA vs. WM p < 0.001; EA vs. MMp < 0.001; EA vs. EM p < 0.001; WM vs. MM p = 0.001; WM vs. EM p < 0.001; MM vs. EMp = 0.041
The distribution of the percentage of polymorphic positions is not statistically different from a normal distribution (Kolmogorov-Smirnov test, p = 0.295). The means (± se) of each group are: 33.9% ± 2.7%, 67.2% ± 3.0%, 72.6% ± 3.4% and 77.5% ± 1.62%, for EA, WM, MM and EM, respectively (see Fig. 7c). Using a one-way ANOVA for testing the differences in the percentage of polymorphic positions between these four groups, we get F 3,10 = 57.722, p< 0.001. Post-hoc: EA vs. WM p < 0.001; EA vs. MM p < 0.001; EA vs. EM p < 0.001; WM vs. MM p >0.999; WM vs. EM p = 0.084; MM vs. EM p >0.999.
Focusing more on the differentiation between the populations of the Eastern and the Western Mediterranean basins, we compared the number of exclusive alleles in each. Since the number of exclusive alleles is influenced by the number of locations in a basin, for a balanced comparison we considered only three of the five locations in the East (i.e., Bodrum, Larnaca and Rethymno), and compared them to the three locations in the West (i.e., Bizerte, Málaga and Melilla). In the East, we counted a total of 15 exclusive alleles (that is, 10 alleles of EF1 and 5 alleles of NaKA that are present in the East, but not in the West), compared to only 3 alleles (3 of EF1 and none of NaKA) that are exclusive to the West. Using the exact binomial test, with equal number of exclusive alleles in each basin as the null hypothesis, we obtained ap -value of 0.008.