Genetic Structure
The AMOVA generated the population structure based on both matR and nad5-1 markers. The results showed that 0.52% and 2.65% of the variation occurred among groups, 2.98% and 10.41% of the variation occurred among populations, while 96.50 and 86.95% of the variation occurred within populations for matR and nad5-1 genes respectively (Figure 4D, 5D, Supplementary Table S5, 6 ), suggesting that the variation of the five groups mainly occurred within populations, and nad5-1 marker supports a stronger population structure. The higher inter-population variation was detected at the nad5-1 marker, the comparison of the variation among populations within the groups was: LN*>SX*>HB*>LN>HB, which means the SX* group with the highest degree of variation among populations (left out LN* group) (Figure 5D, Supplementary Table S6 ).
To reveal the population structure precisely, the NJ tree, PCoA, and STRUCTURE were obtained from the sequence data of matR and nad5-1 markers (Figures 4, 5 ). Based on the matR marker, the results of the NJ tree (35 populations, 796 individuals, classified based on groups) (Figure 4A ) showed 8 similar topologies, and 5 groups were sorted in a fixed way, suggesting there was a fixed genetic relationship among the 5 groups, and their ancestral status relationship was as follows: SX*-HB*-HB-LN*-LN. The dispersion degree of LN populations in the NJ tree (Figure 4B ) was greater than that of HB, which indicated that there was higher variation among populations within the LN group. The discrete distribution of the SX* populations (Figure 4B ) suggesting there was also a high variation among populations. The PCoA results (Figure 4C) showed the distribution of all populations was relatively concentrated, indicating that the genetic differences of all populations at this locus were small. The STRUCTURE analysis (Figure 4E ) inferred the optimal number of the cluster was K=4, we found that the proportions of the four ancestral components were almost the same in all the populations, suggesting there was little difference in the genetic background among all populations at this locus. Based on the nad5-1 marker, the NJ tree (38 populations, 873 individuals, classified based on groups) showed a complicated result, and difficult to find some genetic regularity of five groups (Figure 5A ). NJ trees classified based on STRUCTURE results (Supplementary Figure 1 ) also showed the same result. The LN populations were dispersed in four topologies (Figure 5B ), which implies the LN group with a high variation among populations, was the same as that of matR marker. The high variation among populations within SX* group could also be reflected in the NJ tree (Figure 5B ). The PCoA results (Figure 5C ) showed the distribution of all populations was scattered relatively, indicating that there was a large genetic difference of all populations at this locus. The STRUCTURE analysis showed the K=4 (Figure 5C ), which was the same as that of matR marker, all populations showed higher heterozygosity in the ancestral component at this locus, suggesting there was an abundant difference in the genetic variation among all populations at this locus.