3.4.2 Population genetic structure
Three different analyses revealed that Japanese populations of S. lomentaria fall into two genetically differentiated groups, the Pacific coast populations and the Sea of Japan coast populations, regardless of the reproductive mode of the populations (Figures 5–7). Populations around Kyushu and the Tsugaru Strait were intermediates connecting the two groups of populations. Samples from Argentina and Europe were genetically close to those from parthenogenetic populations in Japan. Furthermore, judging from the results, sexual population p6 and p25 seemed to include parthenogenetic lineages (clonal lineages c5 and c28 in Table 1), in addition to sexual lineages.
In the STRUCTURE analysis based on the diploid dataset, the log-likelihood of the data, LnP(K ), increased with increasingK until K = 7 and the ideal number of clusters determined by the Evanno method (Evanno et al., 2005) was K = 2 (the ΔK value showed a single peak at K =2; Figure S3). AtK = 2, in all runs, clear genetic differentiation was observed between the Pacific coast populations and the Sea of Japan coast populations (although p33 was from the Pacific coast, it was clustered with the Sea of Japan coast populations; Figure 5). An admixture between the two genetic clusters was observed in the populations from Kyushu and the Tsugaru Strait, suggesting the occurrence of gene flow between the two groups. Although sex-check PCR, cetn -int2 data, and SNPs data suggested that p24 and p27 had two distinct genomes, an admixture of two clusters was not estimated in them, except for two runs at K = 3. In the two runs at K = 3, the cluster of the Sea of Japan coast populations in K = 2 were divided into the west coast of Hokkaido populations (p1–p6) and the Honshu populations (p7–p16), and an admixture between the west coast of Hokkaido populations and the Pacific coast populations were observed in p24 and p27 and also in p30 and p31. In other runs at K = 3, p24 and p27 formed a cluster with the sexual or parthenogenetic populations from the Pacific coast. In the runs at K > 3, p24 and p27 formed an independent cluster with or without p30. In the NJ trees constructed by STRUCTURE, the cluster was genetically intermediate between clusters of Pacific coast populations and those of Sea of Japan coast populations (e.g., Figure 5b, c). At K = 5, in five of 10 runs, the Sea of Japan populations were divided into two clusters (west coast of Hokkaido populations and the Honshu populations) and the Pacific coast populations were divided into three clusters (a cluster of sexual populations and two clusters of parthenogenetic populations; Figure 5).
In the PCA (Figure 6), PC1 roughly divided the Japanese populations into two groups: Pacific coast, and Sea of Japan coast, and explained 26.6% of the variance. The PC2 captured north-south variation along it in each group and explained 15.9% of the variance. The PCA appeared as if the individuals were plotted in a ring along the coast of the Japanese Islands; individuals from Kyushu (p16 and p17) and near the Tsugaru Strait (p5, p6, p30, p31) connected the edges of the two groups (Figure 6a). Each group was roughly divided into sexual and parthenogenetic populations along PC3 which explained 11.5% of the variance (Figure 6b). Samples from parthenogenetic populations p24 and p27, which were likely to be polyploid, were close to those from parthenogenetic populations around the Tsugaru Strait (p30–p32). In parthenogenetic populations (p1, p2, p4, p24, and p26–p33), the plots tended to be more densely clustered, relative to the sexual populations. This is probably due to the lower genetic variance of each parthenogenetic population.
The NeighborNet network showed polyphyly of parthenogenetic populations (Figure 7). The network also supported the genetic differentiation between the Pacific coast group and the Sea of Japan coast group. Populations around Kyushu (p15–p17) and the Tsugaru Strait (p5, p6, p30, p31) were intermediates connecting the two groups.
For the three analyses described above, no significant differences were observed between diploid and haploid datasets (Figures S4–S6).