3.2 Phylogeographical patterns and population genetic structure
Overall ribotype and chlorotype diversities (hT ) were 0.908 and 0.970 respectively, and within-population ribotype and chlorotype diversity (hS ) were 0.181 and 0.228 (Table 2). Remarkable population differentiation was detected, with aGST = 0.801 and 0.765 andNST = 0.855 and 0.833. The test statisticU resulted in significant divergence betweenNST and GST(NST ˃ GST , P ˂ 0.05), showing a significant phylogeographical patterns of P. heterotricha in Hainan Island. The 10 populations of P. heterotricha clustered into three major groups according to the STRUCTURE analysis (Figure 4), the likelihood of the nrDNA dataset was also the highest when samples were clustered into three lineages (K=3, Figure S2).
The nrDNA and the cpDNA sequences respectively estimated an among-population genetic differentiation (Fst ) of 0.708 ± 0.319 and 0.826 ± 0.209, gene flow level (Nm ) of 1.44 ± 3.94 and 0.25 ± 0.73, the majority Nm values were less than one (Table 3). Furthermore, the estimated Nm of among-region was 0.40 between NW and SE, 0.10 between SE and SW and 0.04 between NW and SW for nrDNA sequences, and 0.41, 0.07, 0.07 for cpDNA sequences. According to the AMOVA, 72% and 55% of genetic variation among the three regions was respectively attributable to the nrITS and cpDNA differentiation (respectively 17% and 33% among populations, and 11% and 12% within populations, Table 4).
The Mantel tests showed a low but highly significant positive correlation between genetic and geographical distance matrices (respective R2 values of 0.2544 (P < 0.001) and 0.1479 (P < 0.001), Figure S1), indicating a slight IBD. The BARRIER analyses suggested two strong barriers to gene flow among the populations of P. heterotricha (Figure 5); one of them separated the NW and SW populations (red line, 93% BS), and the other separated the SE populations from the other (green lines, 90% BS). We detected a third, weaker barrier (blue line with BS below 80%) that separated of populations WZ and QX with XA and QL based on the nrITS dataset.