3.1 Historical demography
High genetic diversity was found in Primula ovalifolia (HT= 0.987), indicating that this species has a long evolutionary history of limited gene exchange among populations (Chiang et al., 2006; Varvio et al., 1986). This was corroborated by the TMRCA estimations. The TMRCA of all P. ovalifolia haplotypes could be tracked back to 2.44 Mya, coinciding with the early Pleistocene. It is difficult to trace the origin of P. ovalifolia due to the unresolved haplotype relationships. However, strong phylogeographic structure was found (NST > GST, P < 0.01), suggesting that closely related haplotypes occurred in the same geographic area. All populations were clustered into three main lineages (clades I, II, and III) (Fig. 2), which were disjunctly distributed in central and adjacent southwestern China around the Sichuan Basin. There are two scenarios that could have led to this disjunct distribution: long-distance dispersal to suitable habitats; or vicariance due to range contraction or fragmentation of a previously wider distribution (Collevatti et al., 2009). Chloroplast DNA is maternally inherited, so the cpDNA data only reflect the gene flow by seed dispersal in angiosperms (Petit et al., 2003). There are no long-distance dispersal mechanisms for P. ovalifolia seeds, and long-distance dispersal of its seeds has never been reported. Given the high species genetic diversity and interpopulation differentiation inferred from the acquired cpDNA data, the latter scenario seems a more likely explanation of the disjunctive distribution of this species. Thus, we assume that P. ovalifolia, which mainly grows in shaded habitats in temperate broad-leaved forests, may have had a broad range in central and southwestern China during the Quaternary (Yang et al., 1989; Xu et al., 1973; Guo, 1974).
Severe climatic oscillations in the Pleistocene epoch significantly influenced the distribution, diversity, and speciation of most species (Hewitt, 1996; Klicka & Zink, 1997; Avise & Walker, 1998). Many recent phylogeographic studies of alpine plants and birds in China have provided molecular evidence to support this hypothesis (Song et al., 2009; Li et al., 2009; Qiu et al., 2009a; Wang et al., 2010). According to the structure of the haplotype network (Fig. 3), these three main lineages of P. ovalifolia were connected by many missing haplotypes, indicating that there was a massive extinction of ancient haplotypes of this species in the Quaternary. During cooler glacial periods, P. ovalifolia could have shifted to lower elevations, expanding around the Basin, as did the temperate forests, with occasional gene exchange between fragment populations. After the glacial period, P. ovalifolia may quickly respond to the changed environment and migrate back to higher elevations for existing in the “refugia” (Qian & Ricklefs, 2000; Harrison et al., 2001).
The early divergence time (Table S1) meant that the three major lineages had distinct evolutionary histories during the Quaternary. Clade I (located west of the Sichuan Basin) probably survived in higher elevation refugia in the warm stage (interglacial), but shifted to lower elevations with gradual expansion during the glacial periods. Populations BX1 and HY shared an ancient haplotype, with more haplotypes found in population BX2, indicating that the Baoxing–Hongya region would be an important past refugium for this lineage. In clade III (from the eastern and southern Sichuan Basin), populations were separated from each other by distances exceeding 100 km. The haplotype network showed that H18 is an ancient haplotype, with all others being derived from it. Given the high polymorphism (Table 1) and ancient haplotype, the Hunan Sangzi region would be an important past interglacial refugium for clade III lineage. Following the rapid uplift of the Qinghai–Tibet Plateau, there was large-scale tectonic movement in central China, which created several mountainous regions between the Sichuan Basin and the Yunnan–Guizhou Plateau (Zheng & Li, 1990; Wang, 1991; Li et al., 2001). These mountain ranges probably isolated lineages in different locations and resulted in barrier of gene flow, leading to relatively high interpopulation differentiation (Table 2). The historical process of the clade II lineage is partly discussed above. The Emei Mountain region with its complicated topography was little affected by climate oscillation (Jiang & Wu, 1998; Shi et al., 1999). Many species might migrate to this region to survive the glacial period or interglacial period, leading to four forest vegetations living in this mountain (Zhuang, 1998).
Multiple glacial refugia during the Quaternary have also been suggested for two other plant species, Cathaya argyrophylla Chun & Kuang (Pinaceae) (Wang & Ge, 2006) and Saruma henryi Oliv. (Aristolochiaceae) (Zhou et al., 2010), which are also mainly distributed in central China. Some other plants and animals are thought to have had glacial refugia in central China and adjacent regions (Qiu et al., 2009a; Li et al., 2009). Several high mountain ranges, especially the Qinling Mountains, prevented cooler weather from the north reaching central China and kept the climatic conditions warm and moist for a long period of time (Ying et al., 1979; Sun, 2002), which was ideal for distribution and diversification of species.
In conclusion, P. ovalifolia had a complex evolutionary history during the Quaternary. Pleistocene climatic oscillations, combined with the complex local topography, were responsible for the strong phylogeographic pattern of P. ovalifolia. Our study provides further information to explain the high diversity of this region. As multiple refugia were inferred, our study supports the hypothesis that the central and adjacent southwestern China region was an important refugium for East Asia flora during the Quaternary. The stable environment and complex local topography best account for the extant species richness and abundance of primitive temperate species in this region. However, our understanding of historical biogeographic events in central and adjacent southwestern China is still incomplete, and more phylogeographic studies are needed to explore the ice age legacy in this region.
3.2 Relationship between Primula tardiflora and Primula ovalifolia
Primula tardiflora has similar morphology to P. ovalifolia, and is confined to a small area in the Emei Mountains. It can be distinguished by the attenuate base of leaf blade, sparsely pubescent veins, and blooming phenology (it flowers approximately 1 month later than P. ovalifolia). The parsimony tree and unrooted haplotype network analyses both showed that the cpDNA haplotypes (H10 and H11) of P. tardiflora were nested within those of P. ovalifolia, which revealed that P. tardiflora is genetically closed to the population EM of P. ovalifolia.
Both lineages of P. tardiflora and population EM of P. ovalifolia were sampled from Emei Mountain, located southwest of the Sichuan Basin. Emei Mountain was formed during the Himalayan orogeny in the mid-Tertiary, and became complicated after the Quaternary tectonic movement (Hu, 1964; Gu & Li, 2006). The vertical distribution vegetation (Li, 1984) indicates that there may be different habits in different altitudes on the Emei Mountain. Although both grow under the broad-leaved forest, the P. tardiflora lineage is found above 2000 m, higher than P. ovalifolia. Our data suggest that the genetic variability between P. tardiflora and P. ovalifolia seems not sufficient to support their status as two distinct species. However, we are aware of the fact that the present result was only derived from cpDNA data. As P. tardiflora has unique haplotypes and morphological and phenological features, it might represent an ideal example of ecological speciation. We therefore consider the species delimitation of P. tardiflora as an unresolved issue, and await future sampling of highly polymorphic nuclear markers (Hey et al., 2003; Duminil et al., 2006).