3.2 | Nuclear microsatellite genotypes
Among the 32 ncEST-SSR loci, we found three loci (CcC00610, FcC03095, and QmC01794), which had ≥ 0.173 estimated frequency of null alleles and showed significant (P < 0.001) deviation from the Hardy-Weinberg equilibrium in ≥ 8 populations (Data S1). Thus, we used remaining 29 out of the 32 loci in the following genetic analysis.
In PCA for allele frequencies of the 24 populations, the first and second PCs (PC1 and PC2) contributed to 22.5% and 17.0%, respectively, of variation among populations. The coordinates of PC1 and PC2 indicated that 13 Qc populations and eight Qch populations were grouped separately, except for three populations 05Hx, 06Hx, and 07Hx (Figure 3b). The three populations were identified as Qch in the field observation but were grouped to Qc in the ncEST-SSR variation. However, the separation of the genetic group of Qc (13Qc populations and three Qch populations 05Hx, 06Hx, and 07Hx) and the genetic group of Qch (eight Qch populations) was continuous, and three northern and southern marginal Qchpopulations 02H1, 02H2, and 11H, one Qc population 11C, and oneQch population 06Hx were located at intermediate positions in the PC coordinates (Figures 2, 3b). PCA of individual genotypes indicated that individuals of the genetic group of Qch tended to have higher values of PC1, which contributed to only 3.8% of variation among individuals, than individuals of the genetic group of Qc (Figure 3a). The PC1 values of individuals were overlapped between the genetic groups (Figure 3a).
The allelic richness (AR [32]) in the 24 populations was lower in higher latitudes (P = 0.011), was not related to elevation (P = 0.172), and was not different between the genetic groups of Qc and Qch (P = 0.449; Table 2, Figure S3a, b). The expected heterozygosity (H E) was neither dependent on latitude nor elevation (P ≥ 0.341) and was not different between the genetic groups of Qc and Qch (P = 0.416; Table 2, Figure S3c, d). The inbreeding coefficient (F IS) was not significantly (P < 0.05) positive, except for three populations 06C, 08C, and 06H (0.054 ≤ F IS ≤ 0.066), and were not different between the genetic groups of Qcand Qch (P = 0.259; Table 2).
The log probability of data (LnPD) in Bayesian clustering of individual genotypes showed a substantial increase from K = 1 to K = 2 and additional increases to K = 3 and to K = 4 (Figure S4). At K = 5, LnPD was unstable (Figure S4). At K = 2, two clusters were likely to represent the genetic groups of Qcand Qch (Figure 4a). Intermediate ancestry proportions of both clusters were frequently found, especially in some populations (for example, 11C and 11H; Figure 4a). At K = 3, the Qc cluster was divided into two clusters, one of which was dominant in population 05Hx (Figure 4b). At K = 4, the Qch cluster was divided into two clusters, one of which was frequent in northern populations 02H1, 02H2, and 04H of the genetic group of Qch (Figures 2, 4c).
The Weir and Cockerham’s pairwise F ST tended to be higher between the genetic groups (median F ST= 0.046, 0.026 ≤ F ST ≤ 0.107) than within the genetic group of Qc (median F ST = 0.019, 0.005 ≤ F ST ≤ 0.055) and the genetic group ofQch (median F ST = 0.039, 0.015 ≤F ST ≤ 0.079; Data S1). A neighbor-joining (NJ) tree of populations based on the Nei’s distanceD A showed divergence between the genetic groups, although the length of a branch between the genetic groups was relatively short (Figure 5). Jackknifing over loci (removing one out of the 29 loci) generated 29 NJ trees, and seven populations of the genetic group of Qch , except for population 11H, were always grouped in the generated NJ trees (Figure 5).