4.2 Genetic differentiation and genetic structure
Genetic structure of a species is effectively the sum of genetic differentiation among and within populations (Hamrick & Godt, 1989). Overall, genetic structure occurring among populations results from the evolutionary history of the species in question; natural selection; genomic factors (e.g., mutations, reorganization, and genetic drift); and biological characteristics, including gene flow, mating system, mode of reproduction, and seed dispersal mechanisms (Slatkin, 1987; Zhen, 2010). Genetic differentiation is primarily controlled by aspects of gene flow, such as its rate and directionality (Hamrick & Godt, 1989). In plants, gene flow occurs primarily via the transmission of pollen and seeds during sexual reproduction. However, for clonal species, such asP. villosa , asexual propagules often have limited dispersal distance, and this restricts gene flow among populations (Xia, Li, & Li, 2002).
For P. villosa , we inferred that more than 56% of the genetic variation existed within populations, with average pairwiseF ST of 0.35841 for all 43 population and gene flow (N m) of 0.5761 (Table 4). According to Wright’s (1978) theory, genetic differentiation among populations might be large whenF ST > 0.25, but this can be mitigated by gene flow of N m > 1, which can reduce the effects of genetic drift and prevent genetic differentiation among populations. In P. villosa , we found highF ST but limited gene flow, which should yield high rates of between population differentiation. However, we found higher rates of within population differentiation for the species. This differs from findings in other studies of P. villosa that revealed greater genetic variation among populations, such as in Li & Ge (2001), who studied genetic diversity in P. villosa using and seven populations, of which two were from Shihuimiao Ecological Station and five were from the Shilongmiao Ecological Station. Similarly, Wang et al. (1999) assessed the genetic diversity of four populations ofP. villosa from mobile and fixed sand dunes in the Shihuimiao Ecological Station (two populations) and the Shilongmiao Ecological Station (two populations) and found greater genetic variation among populations. Nevertheless, population genetic structure was a comprehensive result of a variety of factors that were related to the evolutionary origins and modes of dispersal and sexual reproduction, which all might be unique to each population within a species (Wang, Ge, & Dong, 1999). Accordingly, we inferred this inconsistency may be related to the numbers of populations and geographic location. Our sampling, which included a larger number of populations and a greater portion of the geographic range ofP. villosa , might have yielded results that more robustly detect patterns among unique populations. In addition, the efficiency and accuracy of gene identification are also related to the number of polymorphic bands, and the results obtained by different experimental methods are different.
In this study, our SAMOVA analysis revealed two well-defined groups corresponding to Group1 and Group2 (Table S3). These groups are also consistent with our analyses using UPGMA, STRUCTURE, SplitsTree, and PCoA (Figure 2-5). Within populations of Groups 1 and 2, the average genetic variation was 64% with an average pairwiseF ST value of 0.43865. The gene flow (N m) in Groups 1and 2 was 0.9996, 0.6605, respectively (Table 4). In addition, the UPGMA tree revealed that populations with closer geographic distances did not always cluster together, and we observed a similar pattern in the results from STRUCTURE and SplitsTree. Similarly, the PCoA showed that individuals from the same population did not always group together, further indicating that, while gene flow may be limited in P. villosa , it occurs over long distances more often than between adjacent populations. This may suggest that some critical dispersal vector, such as birds, is yet-unknown for the species. Overall, P. villosa has undergone considerable genetic divergence and has a high level of genetic structure based on the combined results from our population genetics analyses.
Based on a Mantel test, we found that there was significant correlation between genetic distance and geographic distance, indicating that geographic distance is an important factor affecting the genetic structure of P. villosa . Therefore, we inferred that genetic structure might have been resulted mainly from geographic isolation imposed by mountains (e.g., Yin Mountains; Helan Mountains) and large deserts in northwestern China (e.g. Tengger Desert; Mu Us Sandy Land) as well as range contraction and population fragmentation induced by climatic oscillations (e.g., Gymnocarpos przewalskii Maxim.;Helianthemum songaricum Schrenk) (Liu, 1995; Su, Zhang, & Sanderson, 2011; Ma, Zhang, & Sanderson, 2012; Meng, Gao, Huang, & Zhang, 2014). In addition, founder effects and population bottlenecks might have also contributed to the genetic structures of the species (Birky, Fuerst, & Maruyama, 1989; Liu et al., 2015).
4.3 Demographic historyofP. villosa
The genetic diversity within the Group 1, identified according to SAMOVA, was lower than that of the populations in Group 2 (Table 3). Based on this, extant populations of this species originated from the genetic stock of Group 2, as geographic areas with both high genetic diversity and frequency of dominant genes usually represent centers of origins for source populations (Vavilov, 1926). However, our study design and results cannot discern the exact center of origin for the species nor the main migrational patterns of P. villosa , and accomplishing this will require additional molecular data and informatics approaches.
Climate oscillation during the Quaternary has often been hypothesized to be an important factor in influencing the current geographical distribution and demographic history of plant species (Hewitt, 2004; Su & Zhang, 2013). One widely utilized approach to comparing past and future distributions of plant species and determining the primary environment factors driving them is via ENM (e.g., Nabout, Magalhães, Ma, & Da, 2016; Bai et al., 2017; Huang et al., 2017; Noulèkoun, Chude, Zenebe, & Birhane, 2017; Swanti, Kusum, Dhruval, & Rajkanti, 2018; Wei, Wang, Hou, Wang, & Wu, 2018). Based on the neutral test, we found that the species did not expand its range during the Quaternary. Although these results were not statistically significant (Table 3), our ENMs, in general, show that the range of P. villosa contracted from the LGM to the present (Figure 6-7). Specifically, our models show that the range of P. villosa was the most extensive during the LIG period and included the northeast edge of the Qinghai-Tibet Plateau, Tarim Basin, Tianshan Mountains, Inner Mongolia Plateau, and the western regions of DaXinggan Ling. The range became limited to the Inner Mongolia Plateau, Ordos Plateau, and the Yinshan-Helanshan area during the LGM. The contraction of the range is likely the result of glaciation and climatic shifts within the Tianshan Mountains and Tarim Basin, where temperatures dropped significantly as glaciation developed on a large scale in the Northern Hemisphere during the early-Middle Pleistocene (Williams, Dunkerley, De, Dekker, Kershaw, & Stokes, 1993; Yi et al., 2004; Shi, Cui, & Su, 2005; Lehmkuhl & Owen, 2005; Xu et al., 2010; Meng, Gao, Huang, & Zhang, 2014). Nevertheless, it is surprising that the species range did not rebound as temperatures grew warmer following the LGM. This may be because of the onset of extreme aridity within the region during the Quaternary period, as this is widely-known to have played a significant role in determining the geographic distribution and evolutionary history of many plant species (Meng & Zhang, 2011; Su, Zhang, & Sanderson, 2011; Su & Zhang, 2013). For example, in a previous study of Helianthemum songaricum (Cistaceae), which occurs in Northern China and adjacent desert areas of central Asia (Yang & Gilbert, 2007), the worsening of the dry climate restricted the distribution range, and acceptable habitats for the species gradually became reduced and isolated (Su, Zhang, & Sanderson, 2011). Besides, it needs to explain that AFLP dataset used in this manuscript do not reveal too many informative sites, so more markers should be selected to discuss the evolutionary history of P. villosa in the future.