Genetic differentiation, structure, Barrier and Niche consistency detection
The AMOVA analysis based on cpDNA data and five low-copy nuclear genes showed that genetic differentiation between the two species was significant (F ST > 0.25) (Table 3). Genetic variation partitions revealed by different genes were inconsistent. cpDNA data showed that more percentage of variations were partitioned among populations within a species; four nuclear genes (PHYP , AC5 , PPRC and SAMS ) showed that more percentage of variations were partitioned among species; nuclear geneAAT showed that more percentage of variations were partitioned within species. The microsatellite data showed that there was more percentage of variation within species, and the genetic differentiation between the two speceis was relatively low (Tables 3). The STRUCTURE result showed that the optimal number of clusters across all individuals based on the 16 microsatellite loci was two (Figure 3a). One cluster was composed of the four populations of C. bifida , while the other cluster was constituted by the seven populations of C. micholitzii . When K = 3 or 4, C. micholitzii was divided into two or three clusters. The two-dimensional PCoA analysis (Figure 3b) showed the same result with the STRUCTURE analysis. The Barrier analysis showed that there was only one major genetic boundary with a 95% mean bootstrap value separated the two species, coinciding with the RRFZ (Figure 3c).
The AUC scores, for C. bifida and C. micholitzii , were 0.866 and 0.888 respectively, indicating the credibility of ENMs’ simulations. Niche identity tests showed that the observed niche overlaps were significantly lower than the null distribution of niche overlaps (P < 0.001) both for I and D, suggesting that niches of C. bifida and C. micholitzii were dissimilar (Figure 4), although we should consider that niche identity tests can be seriously biased by the environmental discrepancies that exist in the areas where the two species do not overlap (Warren et al., 2010).