A population genetics perspective on the evolutionary histories of three clonal, endemic, and dominant grass species of the Qinghai–Tibet Plateau: Orinus (Poaceae)

Abstract We performed analyses of amplified fragment length polymorphism (AFLP) in order to characterize the evolutionary history of Orinus according to its population genetic structure, as well as to investigate putative hybrid origins of O. intermedius and to provide additional insights into relationships among species. The genus Orinus comprises three clonal grasses that are dominant species within xeric alpine grasslands of the Qinghai–Tibet Plateau (QTP). Here, we used eight selectively obtained primer pairs of EcoRI/MseI to perform amplifications in 231 individuals of Orinus representing 48 populations and all three species. We compared our resulting data to genetic models of hybridization using a Bayesian algorithm within NewHybrids software. We determined that genetic variation in Orinus was 56.65% within populations while the among‐species component was 30.04% using standard population genetics statistics. Nevertheless, we detected that species of Orinus were clustered into three highly distinct genetic groups corresponding to classic species identities. Our results suggest that there is some introgression among species. Thus, we tested explicit models of hybridization using a Bayesian approach within NewHybrids software. However, O. intermedius likely derives from a common ancestor with O. kokonoricus and is probably not the result of hybrid speciation between O. kokonoricus and O. thoroldii. We suspect that recent isolation of species of Orinus in allopatry via vicariance may explain the patterns in diversity that we observed, and this is corroborated by a Mantel test that showed significant positive correlation between geographic and genetic distance (r = 0.05, p < 0.05). Recent isolation may explain why Orinus differs from many other clonal species by exhibiting the highest diversity within populations rather than among them.

sification of plant species within the QTP and have especially used population genetics methods to elucidate patterns of diversity and distributions and better understand the underlying mechanisms (Liu, Wang, Geng, et al., 2006;Ren, Conti, & Salamin, 2015;Wen et al., 2014). Recently, Wen et al. (2014) reviewed current evidence of mechanisms of speciation on the QTP using exemplar species within diverse vascular plant families, especially of Asteraceae, Crassulaceae, Ericaceae, Orobanchaceae, and Papaveraceae.
However, the mechanisms of speciation within alpine areas of the QTP (and beyond) remain poorly understood. These mechanisms likely include allopatric processes and, possibly, rapid genetic isolation due to increased mutation rates under high levels of ultraviolet light exposure (Davies, Savolainen, Chase, Moat, & Barraclough, 2004;Madriñán, Cortés, & Richardson, 2013;Willis, Bennett, & Birks, 2009). Within the QTP, studies of many plant species are needed to serve as models for diversification and speciation patterns and processes, especially to represent the numerous habits, life histories, environmental preferences, and other features of the rich botanical diversity of the region. Such studies are particularly urgent for regions, such as the alpine grasslands (Bowman, 2000;Li et al., 2014;Yi et al., 2011), that have become imperiled during the Anthropocene (Crutzen & Stoermer, 2000) especially due to climate change and pressures from intensive grazing by livestock (Han, Brierley, Cullum, & Li, 2016;Wilcox, Sorice, & Young, 2011).
Within the alpine grasslands of the QTP, the dominant vascular plants are three endemic species comprising the entirety of the genus, Orinus Hitchcock (Figure 1; Poaceae; Liu et al., 2018;Su, Wu, Li, & Liu, 2015). Orinus consists of clonal grasses and was estab- Orinus is especially characterized by long scaly rhizomes with numerous nodes, which serve as the basis for its clonal reproduction.
It also reproduces sexually via seeds borne on sparse panicles within pedicelled and laterally compressed spikelets that have 3-to 5-veined lemmas with short awns (Su, Liu, Wu, Luo, & Liu, 2017). Orinus thoroldii is distinguished from O. kokonoricus by having pubescent leaf by a Mantel test that showed significant positive correlation between geographic and genetic distance (r = 0.05, p < 0.05). Recent isolation may explain why Orinus differs from many other clonal species by exhibiting the highest diversity within populations rather than among them.

K E Y W O R D S
alpine grassland, amplified fragment length polymorphism, genetic variation, hybridization, population biology blades and dark brown or purple spikelets with two to six flowers (Su et al., 2017). Leaf blades in O. kokonoricus are glabrous and spikelets are yellow or white and bear one to three flowers (Su et al., 2017). In a recent taxonomic revision of the genus, Su et al. (2017) Su et al. (2015) recognized O. intermedius as distinct on account of its rhizomes bearing sparse small scales compared to O. kokonoricus and O. thoroldii, which have many larger scales. However, Su et al. (2015) suspected that O. intermedius may have a hybrid origin with the other two species as progenitors. Nevertheless, O. intermedius appeared more likely to be an incompletely isolated sister of O.
kokonoricus than a hybrid based on a population-level phylogenetic study comprising chloroplast and nuclear ribosomal internal transcribed spacer (ITS; Liu et al., 2018). At present, the putative hybrid status of O. intermedius remains incompletely resolved.
Orinus represents an important model for evolution and biodiversity of vascular plants within the grasslands of the QTP for several reasons. As the dominant vascular plant species within the xeric, alpine grasslands, Orinus can provide a representative first glimpse into evolutionary diversification and diversity within this threatened habitat type (Ma et al., 2017;Sedlacek et al., 2016;Yang et al., 2004).
Moreover, few population genetics studies have targeted clonal species, which may exhibit different patterns of diversification than species that most often reproduce sexually. Finally, Orinus possesses an extensive system of roots and rhizomes (Cai, 2004;Su et al., 2015;Su, Yue, & Liu, 2013) that limit soil loss within the wind-swept alpine grasslands of the QTP (Figure 1; Yang et al., 2004). Thus, the diversity and diversification of the genus can also yield insights into the timing, mechanisms, and ecological consequences of regional desertification (Guo et al., 2002;Han, Fang, & Berger, 2012; see also Liu et al., 2018).
In this report, we investigated diversity and diversification in Orinus using analyses of amplified fragment length polymorphism (AFLP) markers. We specifically sought to address the following questions: (a) Are there three distinct species of Orinus, and do these exhibit recent or ongoing gene flow? and (b) Does O. intermedius have a hybrid origin? Additionally, we used our data to compare patterns of diversity and diversification in Orinus to other clonal plants, especially of alpine regions.

| Taxonomic sampling strategy and obtaining AFLPs
The AFLPs analyzed in this study were previously published in Liu et al. (2018) where they were used in a distance-based phylogenetic analysis complementary to phylogenetic reconstructions based on chloroplast and nuclear gene sequences. Here, we analyzed the AFLPs for the first time using population genetics methods and applied them to perform the first explicit test of the hybrid origin hypothesis for O. intermedius. Below, we describe obtaining the AFLPs, including taxonomic sampling, in brief, and refer to our prior work for greater detail (Liu et al., 2018).
We sampled a total of 231 individuals of the genus Orinus from 48 natural populations from 28°21′51.0 to N and 79°48′9.0 to 102°30′59.7E representing the distributional ranges of the species and including the type localities of each (Figures 1 and 2, Table 1).
As species of Orinus are dominant within the grasslands of the QTP, the boundaries among populations can be difficult to determine.
Thus, we sampled from localities at least 30 km apart to ensure, to the best of our abilities, the genetic independence of the sampling localities except via dispersal of pollen, seeds, or propagules. Per population, we collected fresh leaf blades from three to five vegetative units spaced at least 20 m apart in order to try and sample genetically unique individuals of this clonal species. Our sampling protocol was designed to detect the diversity of genotypes within and among populations covering a vast region, especially to capture rare alleles (e.g., as in Pluess & Stöcklin, 2004), and, notably, our objectives do not include determining the abundance of clonal genotypes within populations at this time. Nevertheless, we regard our within-population sampling as preliminary and acknowledge that greater depth of sampling will yield deeper insights into some aspects of diversity and diversification in the genus in future studies. We dried the leaf samples in silica gel. For each population, voucher specimens and geolocations are reported in Liu et al. (2018).
For the AFLP analyses of all individuals, we performed DNA digestion with DNAs obtained using standard methods (Doyle & Doyle, 1987;see Liu et al., 2018)
Subsequently, we separated and analyzed the fluorescently-labeled amplification products on an ABI PRISM 377 DNA Sequencer (Applied Biosystems) using GeneScan ROX-500 with an internal size standard. We scored the presence or absence of the resulting AFLP products (Figure 3) using GeneScan 3.1 (Applied Biosystems).

| Genetic diversity and population genetic structure
For each population, we calculated the average standard deviation among markers. Thus, a population with all 1s or 0s for a particular  marker would have a standard deviation of zero for the marker, and clonal individuals should have an average deviation of zero. However, clonal individuals may vary in AFLP analyses due to errors in obtaining or processing the data or due to somatic mutations. Thus, we regarded any population with less than 0.05 average deviation as being comprised exclusively of clones, and we sought to exclude these populations from downstream analyses.
We assessed genetic diversity in Orinus, including natural breaks potentially corresponding to species, by analyzing binary matrix of AFLP bands. We analyzed the matrix in POPGENE 1.32 (Yeh, Yang, & Boyle, 1999) to calculate the following summary statistics: percentage of polymorphic loci (PPL), observed number of alleles (N a ), effective number of alleles (Ne), expected heterozygosity (H e ; Kimura & Crow, 1964), and Shannon's information index (I; Lewontin, 1972). We also analyzed the binary matrix using the NTSYS-pc 2.l statistical package (Rohlf, 2000). Specifically, in NTSYS, we generated a pairwise similarity matrix with a simple matching coefficient according to the SIMQUAL algorithm. We also used SAHN in NTSYS package to construct a UPGMA tree based on Nei's genetic distance for assessment of relationships among individuals and populations of Orinus, and we estimated support for the UPGMA tree using 2000 bootstrap replicates in Winboot software (Yap & Nelson, 1996; see also Liu et al., 2018).
We calculated a genetic similarity matrix from the AFLP data according to the method of Nei and Li (1979) and visualized genetic variation among individuals with a principal coordinate analysis (PCoA) performed in GENALEX 6.5 (Peakall & Smouse, 2012).
In addition, we constructed a similarity-based network using the Neighbor-Net algorithm based on Jaccard's distances within SplitsTree 4.13 (Huson & Bryant, 2006) to further depict relationships among individuals and populations and species based on the AFLP datasets.
We also sought to evaluate the genetic differentiation between and within populations of the three species of Orinus using average F ST , analysis of molecular variance (AMOVA; Excoffier, Smouse, & Quattro, 1992), and a Mantel test. We calculated F ST using Arlequin 3.11 (Excoffier, Laval, & Schneider, 2005) and determined significance of the pairwise F ST comparisons via permutation tests (n = 1,000) with a sequential Bonferroni correction. For the AMOVA, we tested significance with nonparametric permutation using 9,999 replications. We performed Mantel tests on the distance matrix of Jaccard's coefficients calculated in GENALEX 6.5 (Peakall & Smouse, 2012) in order to detect the correlations between genetic distances generated from each of the AFLP primer pairs, and geographic distances of populations derived from geographic coordinates using AFLP datasets (Ehrich, 2006). For the Mantel tests, we computed correlation coefficients and assessed the significance with 1,000 permutations.
We conducted a Bayesian analysis of the population structure in Orinus using STRUCTURE 2.3 (Falush, Stephens, & Pritchard, 2007;Hubisz, Falush, Stephens, & Pritchard, 2009;Pritchard, Stephens, & Donnelly, 2000) to determine whether the structure was consistent with species boundaries and to infer the relative amounts of gene flow between each species. We performed the analyses using an admixture model with independent allele frequencies for 10 independent runs for the number of clusters (K) ranging from 1 to 10.
We applied 1 × 10 6 Markov chain Monte Carlo repetitions with a burn-in rate of 25%. We summarized the outputs of all runs with the Web-based software Structure Harvester (Earl & von, 2012), and we calculated the average similarity coefficients among runs for each K. We determined the optimal K using two methods: the point of diminishing returns for adding additional K (i.e., elbow method) and the value representing the greatest change from the previous value

| Testing AFLP data against explicit genetic models of hybridization
Selective amplification primer

| Population genetic structure
Analysis of molecular variance (AMOVA) based on AFLP datasets and inbreeding coefficients (F ST ;  Figure 4). Similarly, the Mantel tests agreed that geography is positively correlated with genetic divergence (r = 0.05, p < 0.05).
Discrete cluster corresponding to species and geography was supported by the principal coordinate analysis (PCoA; Figure 5). The intermedius were clustered together (Figure 7d).

| Results of genetic models of hybridization
The results of NewHybrids show that models of a single genetic origin are best suited to most populations (Figure 8   However, this likely represents a level of introgression occurring contemporaneously with speciation processes, rather than backcrossing, as we did not detect any true hybrid individuals or populations.

| Species limits in Orinus and introgression
Previously, we suggested that Orinus represents three species and noted that genetic isolation among all species of Orinus is nearly, but not entirely, complete (Liu et al., 2018;Su et al., 2017Su et al., , 2015. The present study is congruent with our prior work in showing that the three species of Orinus are largely distinct, especially according to the UPGMA (Figure 4), STRUCTURE (Figure 7), AMOVA (Table 4), and SplitsTree ( Figure 6). However, some gene flow does continue to occur among all species based on the results of these same analyses and may also explain the nonzero probabilities of backcrosses within some populations of O. kokonoricus and O. thoroldii according to NewHybrids (Figure 8; Appendix A).
Gene flow between O. intermedius and O. kokonoricus may enable them to maintain the higher levels of genetic diversity that we detected, compared with O. thoroldii, which is more genetically isolated. In contrast to our results, it is relatively common that more widely spread species, such as O. thoroldii, maintain greater genetic diversity than more geographically restricted species (Hamrick & Godt, 1989;Karron, 1987;Xue, Wang, Korpelainen, & Li, 2005) (Liu et al., 2018), in which barriers to gene flow remain incomplete.
Relatedly, due to earlier divergence time of O. thoroldii, it may have had more time in isolation to undergo some degree of genetic drift.
Orinus thoroldii is not only more genetically distant from its congeners ( Figure 5), but also more geographically distant. Thus, genetic differentiation in Orinus may be mediated by reduced gene flow over greater geographic distances as is consistent with an allopatric mode of speciation as has been observed in other plant species of the QTP (Ge, Zhang, Yuan, Hao, & Chiang, 2005;Hu et al., 2016;Liu, Wang, Geng, et al., 2006;Zhang, Chiang, George, Liu, & Abbott, 2005).

| Population and species diversification history in Orinus compared to other clonal species
Many clonal species exhibit a common pattern of diversity, which is low or intermediate within populations and very high among them (Ellstrand & Roose, 1987;Li & Ge, 2001). This pattern has been documented in other clonal grasses, such as Psammochloa villosa Hitchc.
(Poaceae; Li & Ge, 2001;Yu, Dong, & Krüsi, 2004). Overall, for clonal species, this pattern suggests that interpopulation movement of propagules is rare and that diversity within populations may be largely explained by the founder genotypes and, in some cases, outcrossing among genotypes (e.g., Carex curvula, Dryas octopetala L., this pattern of diversity probably due to ongoing gene flow, despite geographic isolation of populations on sky islands (Pluess & Stöcklin, 2004; see also sky islands in Hughes & Atchison, 2015;Körner, 2004).
In Orinus, gene flow is unlikely to account for this pattern of diversity, especially among species, because the species are, overall, distinct, and because the species probably experience limited gene flow by rare dispersals of rhizome sections and occasional pollen movement by wind, water, and animal visitors. Within Orinus, there is no obvious mechanism for seed dispersal. Therefore, alternatively to ongoing, regular gene flow, recency of isolation of species and populations of Orinus within the QTP may explain the limited genetic diversity at the interspecific and interpopulational levels, respectively (e.g., as in Cruickshank & Hahn, 2014). Indeed, Orinus may have begun diversifying within the QTP during the latter part of the Pliocene (2.85 million years ago; Liu et al., 2018), which represents the end of a global period of evolution of modern alpine species (Hughes & Atchison, 2015). This alternative also requires that the original populations possessed high genetic diversity that has been preserved, at least partially, to present times. High diversity within ancestral populations often results from isolation by vicariance, rather than dispersal, events (Mayr, 1942; see also Harris, Ickert-Bond, & Rodríguez, 2018;Kropf, Comes, & Kadereit, 2006). Vicariance within the QTP is often invoked to explain commonly observed patterns in the diversification of plant populations or species (e.g., Yang, Li, Ding, & Wang, 2008), especially the divergence of western lineages, such as O. thoroldii, from eastern ones, such as O.
kokonoricus and O. intermedius. Moreover, vicariance in the region may be attributed to either the topographic or climatic effects of recent geomorphism (Jia, Liu, Wang, Zhou, & Liu, 2011;Liu et al., 2013;Wen et al., 2014;Yang et al., 2008), and topology may be a better explanation for divergence in the case of Orinus, because the ecological niches of species are similar . Overall, the pattern of genetic diversity within Orinus could eventually come to resemble patterns overserved for other clonal species (Ellstrand & Roose, 1987;Li & Ge, 2001) given sufficient evolutionary time. However, as a caveat of the present study, we also cannot rule out that our limited sampling within populations accounts for some parts of the patterns in diversity that we observed, and expanded sampling is needed in the future.

ACK N OWLED G M ENTS
We

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
XS conceived and designed the study. XS, YL, and QG performed the laboratory work. YL, AJ-H, QG, XS, and ZR contributed to performing data analyses, interpreting results, and writing the manuscript.
All authors approved the manuscript as written.

DATA ACCE SS I B I LIT Y
All data are provided within the text, tables, appendix, and figures,