Phylogenetic analyses and population structure
The mtDNA cox1 gene fragment (661 bp in size) was sequenced for a total of 680 individuals collected from 34 locations across the northwestern Pacific (Fig. 1; Table 1). A total of 169 mtDNA cox1haplotypes were identified. Of these, 149 variable sites and 62 single-nucleotide polymorphic sites were detected, and 87 sites were parsimoniously informative. The cox1 phylogenetic trees using ML and BI methods showed identical tree topology: they showed two monophyletic, divergent lineages (termed here the northern and southern mitochondrial lineages, according to their geographic distribution) with high support values (100% BP and 1.00 BPP) (Fig. 2A). The haplotype network displayed a dumbbell shape, with 48 steps (mutations) connecting two haplogroups with a “star genealogy” (Fig. 2B). Pairwise sequence divergence between the two lineages was unexpectedly very high, ranging from 7.4%-9.2%, but there was noticeably low divergence within the northern (0.5%) and southern lineages (0.6%), respectively. Distribution of the two lineages clearly corresponds to their geographic origins: the northern lineage includes all Korean populations and the northern Japan populations (JRK, JAM, JOG, JHS, JIB, JSM, JTZ), with a distribution range mainly in the cold surface water North Pacific region. In contrast, the southern lineage is composed of all Chinese populations, the Taiwan population and the southern Japan populations (JBK, JAK, JTK, JHD, JKK), distributed northeastward along the tropical Indo-West Pacific region. In addition, six populations (JSH [8], JSS [9], JKS [11], JOI [16], JKG [17], JNG [18]) on the southeastern and western sides of the Japanese archipelago contained both lineages (Fig. 1).
In addition to the cox1 data, 31 ITS1 genotypes were obtained from 101 individuals that contained 29 polymorphic sites. Their pairwise sequence divergence was very low (0.6%) compared to mtDNA cox1 . In contrast to the cox1 results, the ML and BI tree for the ITS1 data showed only a single clade where all individuals were mixed, with no segregation between individuals representing the northern and southern mitochondrial lineages (Fig. 3A). Network analysis of the ITS1 haplotype identified two major haplogroups that were only a single step (mutation) away from each other, each forming a star-like genealogy that comprised a mixture of northern and southern mitochondrial cox1members (Fig. 3B). Unlike the deep divergence detected between the two mtDNA cox1 lineages, genotype distribution of ITS1 did not correspond to geographic location.
The NWP populations of M. virgata showed high haplotype diversity with overall h values of 0.874 for cox1 and 0.824 for ITS1 (Table 1). Nucleotide diversity (π ) and nucleotide differences (k ) for cox1 were high: 0.0417 and 27.539, respectively. In the six populations where the two divergent mtDNA lineages coexist, nucleotide diversity (π ) and nucleotide difference (k ) ranged from 0.0192 to 0.0418 and from 12.699 to 27.613, respectively, much higher than all other populations in this study (π =0.0008 to 0.0043, k =0.544 to 2.828). However, ITS1 sequences showed very low nucleotide diversity (π= 0.0026) and nucleotide differences (k= 0.909). In the SAMOVA test using the cox1 gene,M. virgata populations were subdivided into the two regional groups consistent with the results from the phylogenetic analysis: one group comprises populations of the cold-temperate zone that includes the ES, Yellow Sea and northern Pacific coast of Japan (population nos. 1-9. 19-31), and the other group is composed of populations in the subtropical-tropical zone represented by the ECS and SCS and the southern Pacific coast of Japan (population nos. 10-18, 32-34). Genetic differentiation (F ST) between populations belonging to different regions was substantial and statistically significant, ranging from 0.416 to 0.984 (Supplementary Table 1). However, F ST values between populations within the same regional group were less than 0.223, and most of the comparisons were not statistically significant. The hierarchical AMOVA analyses with locality groups inferred from the result of the SAMOVA analysis indicated that 90.54% of the total genetic variation was contributed by ‘among-groups’ variation (ФCT=0.905,P <0.001), whereas genetic variation from ‘among populations within groups’ and ‘within populations’ was 1.07% (ФSC=0.113, P <0.001) and 8.40% (ФST=0.916, P <0.001), respectively. The AMOVA analysis excluding the 6 ‘mixed’ populations (contained individuals representing both lineages) shows that most of the genetic variation was observed from ‘among groups’ (97.33 %, ФCT=0.973, P <0.001; see Table 2 for more details).