Genetic Structure and Eco-Geographical Differentiation of Lancea tibetica in the Qinghai-Tibetan Plateau

The uplift of the Qinghai-Tibetan Plateau (QTP) had a profound impact on the plant speciation rate and genetic diversity. High genetic diversity ensures that species can survive and adapt in the face of geographical and environmental changes. The Tanggula Mountains, located in the central of the QTP, have unique geographical significance. The aim of this study was to investigate the effect of the Tanggula Mountains as a geographical barrier on plant genetic diversity and structure by using Lancea tibetica. A total of 456 individuals from 31 populations were analyzed using eight pairs of microsatellite makers. The total number of alleles was 55 and the number per locus ranged from 3 to 11 with an average of 6.875. The polymorphism information content (PIC) values ranged from 0.2693 to 0.7761 with an average of 0.4378 indicating that the eight microsatellite makers were efficient for distinguishing genotypes. Furthermore, the observed heterozygosity (Ho), the expected heterozygosity (He), and the Shannon information index (I) were 0.5277, 0.4949, and 0.9394, respectively, which indicated a high level of genetic diversity. We detected high genetic differentiation among all sampling sites and restricted gene flow among populations. Bayesian-based cluster analysis (STRUCTURE), principal coordinates analysis (PCoA), and Neighbor-Joining (NJ) cluster analysis based on microsatellite markers grouped the populations into two clusters: the southern branch and the northern branch. The analysis also detected genetic barriers and restricted gene flow between the two groups separated by the Tanggula Mountains. This study indicates that the geographical isolation of the Tanggula Mountains restricted the genetic connection and the distinct niches on the two sides of the mountains increased the intraspecific divergence of the plants.

the Herbarium of the Northwest Institute of Plateau Biology (HNWP), Chinese Academy of Sciences, Xining, Qinghai Province, China.

DNA Extraction and SSR Amplification
Total DNA was extracted from dried leaves of 456 samples using a modified cetyltrimethylammonium bromide (CTAB) method [36]. The quality of the DNA was checked by electrophoresis using 1.0% agarose gels and the quantity of the DNA was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Carlsbad, CA, USA).
Eight pairs of microsatellite markers developed by Tian et al. [35] were used in the present study ( Table 1). PCR reaction was performed in 20 µL reaction mixture containing 20 ng of template DNA, 2 µL of 10× PCR buffer (15 mM MgCl 2 ), 0.5 µL of each primer (5 pM), 0.2 µL of Taq DNA polymerase (TaKaRa Biotechnology Co., Dalian, China), and 0.5 µL of dNTP mix (10 mM), supplemented with ddH 2 O. The PCR reaction conditions included an initial denaturation (94 • C for 5 min), followed by quantification for 40 cycles (94 • C for 30 s, an appropriate primer specific annealing temperature for 35 s, 72 • C for 60 s), and a final extension (72 • C for 10 min). The amplified products were size separated by capillary electrophoresis on an Applied Biosystems Genetic Analyzer (ABI 3730) from Sangon Biotech Co., Ltd. (Shanghai, China) to obtain the size of microsatellite fragments. Table 1.
Genetic characteristics of eight microsatellite loci in Lancea tibetica populations.  In the current study, we hypothesized that the population patterns conformed with the island model, and the gene flow (N m ) can be calculated based on the following formula: To further analyze the genetic relationship among 31 populations, Nei's genetic distance matrix was calculated using POPULATIONS 1.2.28 (http://bioinformatics.org/populations/). A Neighbor-Joining (NJ) tree was constructed by MEGA 7.0 (https://www.megasoftware.net/) based on Nei's genetic distance matrix. Meanwhile, the principal coordinate analysis (PCoA) of 31 populations was carried out using the Multi-Variate Statistical Package (MVSP) 3.13 (Kovach Computing Services, Pentraeth, United Kingdom).
We evaluated the genetic structure using Bayesian model-based clustering in STRUCTURE 2.3.4 (https://web.stanford.edu/group/pritchardlab/structure.html). The population structure was detected under both the admixture model and no admixture model at the same time, with a burn-in period set at 100,000 and Markov Chain Monte Carlo (MCMC) repetitions after burn-in set at 100,000. The optimal K value was determined based on the change in slope of the plot of Ln Pr(X|K) versus ∆K, which is generally the corresponding K value at the inflection point of the curve.
Analysis of molecular variance (AMOVA) was performed using ARLEQUIN 3.5 (http://cmpg.unibe.ch/software/arlequin35/) after populations were grouped by STRUCTURE model-based and geographic regions of genetic diversity. Molecular variance was analyzed at two hierarchical divisions, within and among populations. Pairwise comparison of F st values between the populations was conducted in GENEPOP 4.0 (http://www.genepop.curtin.edu.au/). We investigated the historical barriers to gene flow among collection sites using Monmonier's maximum difference algorithm in the software BARRIER 2.2 (http://ecoanthropologie.mnhn.fr/software/barrier.html).
To investigate the correlation between genetic distance and geographical distance, a geographic distance matrix between 31 populations was calculated in Microsoft Excel. The obtained geographic distance was tested against genetic distance by the Mantel test in ARLEQUIN 3.5 (http://cmpg.unibe.ch/software/arlequin35/).

SSR Markers and Genetic Diversity
We employed eight microsatellite markers to genotype 456 individuals of L. tibetica from 31 populations. Overall, 55 alleles were amplified, ranging from 3 (LT7) to 11 (LT10, LT25) alleles per locus with an average of 6.875. We identified 34 rare alleles (RA) with a frequency less than 0.5% in seven microsatellites (except LT7), ranging from one to nine per locus with an average of

Population Structure
Based on the microsatellite data, the STRUCTURE group calculation was performed, and the change in slope of Ln Pr(X|K) value and ∆K value were plotted ( Figure S1). Both Ln Pr(X|K) and ∆K values had obvious inflection points with change in K value, and the best K value was 2. This supported the division of L. tibetica into two groups (K = 2). According to the recommended K values, STRUCTURE groupings are shown in Figure 1  For studying the genetic structure of L. tibetica, NJ clustering analysis was conducted to estimate the genetic relationships among the 31 populations using Nei's genetic distance matrix ( Figure 2). The results showed that these populations were classified into two main clusters, which were consistent with the results of STRUCTURE grouping (K = 2). To investigate the population structure, principal coordinate analysis (PCoA) was carried out and a scatter plot was generated based on the entire microsatellite dataset of 31 populations. The first and second principal coordinates explained 84.188% and 8.216% of the molecular variance. The PCoA For studying the genetic structure of L. tibetica, NJ clustering analysis was conducted to estimate the genetic relationships among the 31 populations using Nei's genetic distance matrix ( Figure 2). The results showed that these populations were classified into two main clusters, which were consistent with the results of STRUCTURE grouping (K = 2).  For studying the genetic structure of L. tibetica, NJ clustering analysis was conducted to estimate the genetic relationships among the 31 populations using Nei's genetic distance matrix ( Figure 2). The results showed that these populations were classified into two main clusters, which were consistent with the results of STRUCTURE grouping (K = 2). To investigate the population structure, principal coordinate analysis (PCoA) was carried out and a scatter plot was generated based on the entire microsatellite dataset of 31 populations. The first and second principal coordinates explained 84.188% and 8.216% of the molecular variance. The PCoA To investigate the population structure, principal coordinate analysis (PCoA) was carried out and a scatter plot was generated based on the entire microsatellite dataset of 31 populations. The first and second principal coordinates explained 84.188% and 8.216% of the molecular variance. The PCoA results indicated the division of 31 populations into two distinct groups ( Figure 3). LZ, AJL, YD, MLS, BX, DX, and LNZ populations gathered together on the left side of the plot, while the remaining 24 populations gathered together on the right side of the plot. This result was consistent with the STRUCTURE analysis (K = 2) and NJ tree.
Genes 2019, 10 FOR PEER REVIEW 6 results indicated the division of 31 populations into two distinct groups ( Figure 3). LZ, AJL, YD, MLS, BX, DX, and LNZ populations gathered together on the left side of the plot, while the remaining 24 populations gathered together on the right side of the plot. This result was consistent with the STRUCTURE analysis (K = 2) and NJ tree.

Genetic Differentiation
In order to estimate the partitioning of genetic variation, an AMOVA based on microsatellite data was conducted ( Table 2). The analysis revealed that the main genetic variation in L. tibetica was within the populations (81.6839%) rather than between populations (18.3161%). AMOVA was also performed on the two groups obtained by the population structure analysis. The results revealed that 20.9305% of the genetic variance occurred between the two groups, while 9.0187% of the genetic variance occurred within groups among populations. Furthermore, 70.0508% of the genetic variance occurred within populations. Genetic barriers among 31 populations were predicted using Monmonier's maximum difference algorithm on BARRIER 2.2. When the number of barriers was one, BX and AJL were separated from all other populations. When the number of barriers was two, YD, LZ, MLS, LNZ, and DX were further

Genetic Differentiation
In order to estimate the partitioning of genetic variation, an AMOVA based on microsatellite data was conducted ( Table 2). The analysis revealed that the main genetic variation in L. tibetica was within the populations (81.6839%) rather than between populations (18.3161%). AMOVA was also performed on the two groups obtained by the population structure analysis. The results revealed that 20.9305% of the genetic variance occurred between the two groups, while 9.0187% of the genetic variance occurred within groups among populations. Furthermore, 70.0508% of the genetic variance occurred within populations. Genetic barriers among 31 populations were predicted using Monmonier's maximum difference algorithm on BARRIER 2.2. When the number of barriers was one, BX and AJL were separated from all other populations. When the number of barriers was two, YD, LZ, MLS, LNZ, and DX were further separated ( Figure 4). This was consistent with the results of STRUCTURE grouping, NJ tree, and PCoA. Furthermore, the first and the second barrier separated the populations in Tibet from populations in Qinghai, Gansu, and Sichuan.
Genes 2019, 10 FOR PEER REVIEW 7 separated (Figure 4). This was consistent with the results of STRUCTURE grouping, NJ tree, and PCoA. Furthermore, the first and the second barrier separated the populations in Tibet from populations in Qinghai, Gansu, and Sichuan. The pairwise Fst and Nm for 31 populations were also calculated ( Table S3). The results revealed the highest genetic differentiation (Fst = 0.5222) and lowest gene flow (Nm = 0.2287) between DL and AJL. Meanwhile, the lowest genetic differentiation (Fst = 0.0011) and highest gene flow (Nm = 227.0227) were between DX and LNZ.

Genetic Diversity Associated with Geography
We performed a Mantel test on all the populations that revealed a significant correlation between genetic distance and geographical distance (R 2 = 0.6702, p = 0; Figure 5A). When the Mantel test was performed on the northern branch, a weak correlation between genetic distance and geographical distance (R 2 = 0.5324, p = 0.004; Figure 5B) was detected, while a high correlation was found in the southern branch (R 2 = 0.2018, p = 0.012; Figure 5C). Sampling information of 31 L. tibetica populations projected onto a map using ArcGIS 10.2 and the results of genetic grouping are shown in Figure 6, which shows the division of 31 populations into two distinct branches. The populations from Tibet formed a southern branch and the populations from Qinghai, Gansu, and Sichuan formed a northern branch. The boundary between the northern and the southern branches was roughly located in the line of the Tanggula Mountains.

Genetic Diversity Associated with Geography
We performed a Mantel test on all the populations that revealed a significant correlation between genetic distance and geographical distance (R 2 = 0.6702, p = 0; Figure 5A). When the Mantel test was performed on the northern branch, a weak correlation between genetic distance and geographical distance (R 2 = 0.5324, p = 0.004; Figure 5B) was detected, while a high correlation was found in the southern branch (R 2 = 0.2018, p = 0.012; Figure 5C). Sampling information of 31 L. tibetica populations projected onto a map using ArcGIS 10.2 and the results of genetic grouping are shown in Figure 6, which shows the division of 31 populations into two distinct branches. The populations from Tibet formed a southern branch and the populations from Qinghai, Gansu, and Sichuan formed a northern branch. The boundary between the northern and the southern branches was roughly located in the line of the Tanggula Mountains.

Discussion
Genetic diversity is the total number of genetic characteristics in the genetic makeup of a species that serve as a means for populations to adapt to changing environments. Rich genetic diversity can help maintain species diversity and stability [37]. It can also slow down the extinction process caused by adaptation and evolution [38]. Microsatellite markers are efficient in examining genetic diversity and exploring genetic relationships in plants, and have been widely used to investigate the genetic diversity of Qinghai-Tibetan Plateau species [39][40][41][42][43][44][45].

Genetic Variation
In the current study, a rich genetic variation was detected in L. tibetica (Na = 6.875, Ho = 0.5277, He = 0.4949, I = 0.9394). The high level of genetic diversity within the populations of L. tibetica is similar to other species that are distributed in the QTP (e.g., Elymus nutans, He = 0.719 [44]; Armillaria luteovirens, He = 0.521 [45]; Sibiraea laevigata, He = 0.834 [46]; Sibiraea angustata, He = 0.832 [46]; Stipa purpurea, He = 0.585 [47]; Carex moorcroftii, He = 0.579 [48]). The high genetic diversity of L. tibetica may be due to its biological characteristics such as perenniality, pollination and seeds characteristics. In addition, a relatively low gene flow among populations (Nm = 0.8823) might also result in a high level of genetic diversity. Fis and Fit values indicated heterozygote excess in L. tibetica populations, suggesting a heterozygous advantage. This may be attributed to the fact that heterozygous genotypes grow faster and have lower mortality than homozygous genotypes, resulting in higher population heterozygosity [49]. The high genetic diversity of alpine species in the QTP gives rise to some degree of adaptation to their respective environmental conditions.

Genetic Divergence and the Effects of the Tanggula Mountains
Mantel tests showed that isolation by distance had an important impact on the genetic divergence of L. tibetica. Meanwhile, Bayesian clustering (STRUCTURE), NJ cluster analysis, PCoA and genetic barrier prediction analysis demonstrated that the Tanggula Mountains played a significant role in intraspecific divergence. All the populations can be divided into two main groups: Figure 6. Geographic distribution map of L. tibetica (black dots represent herbarium records in the Chinese Virtual Herbarium, the red triangles represent the northern branch, the blue triangles represent the southern branch).

Discussion
Genetic diversity is the total number of genetic characteristics in the genetic makeup of a species that serve as a means for populations to adapt to changing environments. Rich genetic diversity can help maintain species diversity and stability [37]. It can also slow down the extinction process caused by adaptation and evolution [38]. Microsatellite markers are efficient in examining genetic diversity and exploring genetic relationships in plants, and have been widely used to investigate the genetic diversity of Qinghai-Tibetan Plateau species [39][40][41][42][43][44][45].

Genetic Variation
In the current study, a rich genetic variation was detected in L. tibetica (N a = 6.875, H o = 0.5277, H e = 0.4949, I = 0.9394). The high level of genetic diversity within the populations of L. tibetica is similar to other species that are distributed in the QTP (e.g., Elymus nutans, H e = 0.719 [44]; Armillaria luteovirens, H e = 0.521 [45]; Sibiraea laevigata, H e = 0.834 [46]; Sibiraea angustata, H e = 0.832 [46]; Stipa purpurea, H e = 0.585 [47]; Carex moorcroftii, H e = 0.579 [48]). The high genetic diversity of L. tibetica may be due to its biological characteristics such as perenniality, pollination and seeds characteristics. In addition, a relatively low gene flow among populations (N m = 0.8823) might also result in a high level of genetic diversity. F is and F it values indicated heterozygote excess in L. tibetica populations, suggesting a heterozygous advantage. This may be attributed to the fact that heterozygous genotypes grow faster and have lower mortality than homozygous genotypes, resulting in higher population heterozygosity [49]. The high genetic diversity of alpine species in the QTP gives rise to some degree of adaptation to their respective environmental conditions.

Genetic Divergence and the Effects of the Tanggula Mountains
Mantel tests showed that isolation by distance had an important impact on the genetic divergence of L. tibetica. Meanwhile, Bayesian clustering (STRUCTURE), NJ cluster analysis, PCoA and genetic barrier prediction analysis demonstrated that the Tanggula Mountains played a significant role in intraspecific divergence. All the populations can be divided into two main groups: the northern branch and the southern branch, roughly bounded by the line of the Tanggula Mountains. A similar pattern was also detected in previous studies of Rhodiola alsia [21] and Stuckenia filiformis [22] by using nuclear ITS sequence and chloroplast sequences. The pairwise F st and N m demonstrated that gene flow between populations on each side of the Tanggula Mountains is greater than across the Tanggula Mountains. Take MLS and JD for example, the two populations were located on different sides of the Tanggula Mountains, and the geographic distance was not very great, but a high genetic differentiation (G ST = 0.2710) and low gene flow (N m = 0.6722) was detected. The gene flow between them is less than that between populations on the same side of the mountains. The genetic divergence pattern strongly suggests that the Tanggula Mountains might be a geographic barrier that restricts the gene flow between the populations on the different sides of the Tanggula Mountains.
The genetic divergence of L. tibetica in the Tanggula Mountains can mainly be attributed to the snow-cover on the ridges of the mountains. High altitude snow-covered ridges can directly obstruct the genetic connection and reduce the distance of transmission. The differences in the ecological niches on the two sides (north and south) also play an important role. In terms of geomorphology, the northern side is relatively flat with an intact plateau environment and the headward erosion of modern rivers is not obvious, while the topography of the southern side is relatively fragmented forming mountains and gorges, and the rivers are severely cut [50]. Meanwhile, the climate in the northern region, influenced by the Asian monsoon is cold, dry, and has less precipitation, while in the southern region the climate is mainly influenced by the Indian Ocean monsoon, and is warmer and highly humid with relatively more precipitation [51][52][53]. In terms of vegetation, alpine shrub meadow, cold alpine steppe, and cold alpine desert steppe are mainly found on the northern side, while mountain forest, mountain shrub steppe, and alpine steppe are mainly distributed on the southern side [54]. Such distinct ecological niches reinforce the divergence of the two lineages following their initial spatial isolation.

Conclusions
This is the first study presenting the genetic diversity of L. tibetica based on microsatellite markers. The study revealed the significant role of the Tanggula Mountains in determining the genetic diversity and population structure of L. tibetica. Geographical barriers restricted gene flow among different populations and resulted in intraspecific genetic divergence. Genetic structure and eco-geographical differentiation can maintain adaptation to continuous geographical and environmental change.