4.1 Phylogeographical patterns driven by river isolation
Through the examination of the ribotypes and chlorotypes of 217 individuals from 10 populations of the Hainan-endemic herbaceousP. heterotricha in the south-central mountain system of Hainan Island, low within-population genetic diversity and high genetic differentiation (Fst ) among populations were revealed (Figure 1). There was a significant phylogeographical structure and AMOVA showed high genetic variation mainly from inter-regions (Table 4). Three clades NW, SE and SW have evolved in prolonged isolation, as intimated by bayesian analysis (Figure 3), which is supported by results of STRUCTURE v.2.3.4. (Figure 4), and low gene flow (Nm ) among regions were detected. Two main genetic barriers were identified based on Monmonierʼs algorithm (Figure 5), which spatially concordant with the Changhua River and Wanglou River in south-central mountains system of Hainan Island (Figure 2, 3 and 4).
The average population genetic divergence detected in P. heterotricha is high overall (GST = 0.801 and 0.765 for nrITS and cpDNA gene respectively), considering that genetic differentiation among populations turns to be stronger whereGST > 0.25 (Buso et al. , 1998). Primulina heterotricha grows widely in rocky streamsides of forest valleys and humid or arid limestone and granites habitat, which are widespread in south-central Hainan Island. Yet, there was a slight IBD, although pollen or seed dispersal is very limited among most populations (Nm << 1, Table 3) owing to the discussed isolation effect of Changhua River, and the lack of adaptation to long-range seed dispersal. Furthermore, the distribution of effective pollinators (genus Glossamegilla ) is interrupted easily by rivers and mountains, and it even shows altitudinal differences within the same mountain system (Zhong et al. , 2014).
Geographic discontinuities are indeed a significant factor for population differentiation through weakening or blocking gene flow (Slatkin, 1985), and their impacts apparently have driven regional genetic differentiation within P. heterotricha populations. These results are in line with previous studies in other groups, showing that the geographic barrier associated with the course of Changhua River may have stimulated lineage diversification. Two examples are the multiple intra-specific genetic lineages detected in the endemic speciesMetapetrocosmea peltata (Merr. et Chun) W. T. Wang (Gesneriaceae) and genus Oreocharis (Gesneriaceae), which also presented a strong phylogeographical population structures (Li et al. , 2020; Ling et al. , 2020a).
The major role of mountains and valleys in the origins and evolution of plants has been widely reported in the Andes (Antonelli et al., 2009; Pennington et al. , 2010), Ghats (Robin et al. , 2015) and Hengduan Mountain (Liu et al. , 2013). And similar results of genetic structure associated with sharp geographical discontinuities have been reported in many Asian plants and animals. Regarding other families and phyla, Myricaria laxiflora (Franch.) P.Y. Zhang et Y.J. Zhang (Tamaricaceae) in the Three Gorges mountain region (Liu et al. , 2009), Taxus wallichiana Zucc. (Taxaceae) in the Himalaya-Hengduan Mountains region (Liu et al. , 2013), and several endemic montane birds from the Western Ghats (Robin et al. , 2015), all showed nested genetic patterns across complex mountain system isolated by valleys or rivers. The ensuing gene flow limitation, genetic drift enhancement and possibly natural selection, can likely have induced the genetic structure detected, that may in the long term lead to reproductive isolation and speciation.
Our genetic data pinpoint the lower Changhua River as the stronger gene-flow barrier between NW and SW clades (93% bootstrap) with minimumNm = 0.04 and 0.07 for nrDNA and cpDNA sequences, followed by a secondary influence of the Wanglou River, separating SE and SW clades with Nm = 0.10 and 0.07 for nrDNA and cpDNA sequences, then the third influence of the upper Changhua River dividing NW and SE clades with Nm = 0.40 and 0.41 for nrDNA and cpDNA sequences (Figure 5, 90% bootstrap). Notably, the Nansheng River splits nrITS clade SE into two subclades (BS below 80%, Figure 5A). Besides, according to the BARRIER analysis, the isolation magnitudes were proportional to the size of the rivers; the lower Changhua River features the greatest impact, possibly owing to its much larger width (1660 m, vs only 300 m of the upper Changhua River and 160 m of the Wanglou River), and bigger discharge volume.