A comparative study of the population genetics of wild and cultivated populations of Paris polyphylla var. yunnanensis based on amplified fragment length polymorphism markers

Abstract Paris polyphylla var. yunnanensis is one of the original plants used to make the traditional medicine Paridis Rhizoma. Wild individuals have been excessively collected in recent decades, and thus, it has become increasingly endangered. Cultivation is a major method for the conservation and sustainable utilization of its wild resources. In this study, amplified fragment length polymorphism markers were used in the genetic analysis of 15 wild and 17 cultivated populations of P. polyphylla var. yunnanensis. This study revealed that cultivated populations possessed higher genetic diversity than wild ones at the species level (H = 0.2636 vs. 0.2616, respectively). However, most of the genetic variation was found within populations for both of these groups (ΦST = 18.83% vs. 19.39%). In the dendrogram produced using UPGMA, the 32 populations were divided into three groups (I, II, and III). In group II, both wild and cultivated populations were included, but remained in distinct clusters within this group, which showed the significant separation between the cultivated and wild populations. Furthermore, wild populations were also clustered into three subgroups (W‐I, W‐II, and W‐III), with an obvious geographic structure. In contrast, although cultivated populations were similarly placed in three subgroups (C‐I, C‐II, and C‐III), the latter two of these were not separated based on geography. Finally, the wild populations in Guizhou, China (W‐I), possessed higher genetic diversity than those in Yunnan (W‐II and W‐III). As P. polyphylla var. yunnanensis is an endangered medicinal plant, the fact that it showed richer genetic diversity in its wild populations in Guizhou means that these should be regarded as priority areas for protection and used for provenance selection. Moreover, cultivated populations also showed high genetic variation, which might be attributed to them having originated from mixed provenances, indicating that screening for superior provenances should be carried out as soon as possible.


| INTRODUC TI ON
More than 80% of world's population relies on plant-based medicines for their primary healthcare needs, most of which come from wild resources (Vines, 2004). Due to increasing demand for such medicines, wild individuals have been overexploited for the last few decades, and many medicinal species have been listed as endangered species, such as Phellodendron amurense (Yang et al., 2016) and Taxus wallichiana (Gao et al., 2007). To protect wild resources, many countries encourage the cultivation of wild medicinal plants so that they can be used sustainably (WHO, IUCN, & WWF, 1993). Cultivation is an effective means of conservation that is not only regarded as an important reservoir of genetic diversity (Miller & Schaal, 2005), but also as a means to meet the growing demand for traditional medicines (Guo, Lu, Wu, Chen, & Zhou, 2007). Previous studies showed that a long history of domestication in cultivated plants has generally resulted in reductions in their genetic variation, as well as bottleneck effects, such as those that are inferred to have affected soybean (Hyten et al., 2006) and wheat (Haudry et al., 2007). Until now, most studies have focused on economically valuable crops with a long-term history of cultivation (Doebley, Gaut, & Smith, 2006), but only a few studies have been performed on medicinal species domesticated over the past few decades (Shi, Yang, Chen, & Guo, 2008;Yuan et al., 2010). Comparative studies of the population genetics of wild and cultivated populations would be beneficial to the development of strategies for conserving wild resources, as well as to help in screening populations for superior provenances for use in cultivation.
There are obvious advantages and disadvantages to both of these marker types. Compared with SSR marker, AFLP markers, which are dominant, cannot distinguish heterozygotes from homozygotes (Mba & Tohme, 2005), but they can enhance the resolution achieved in population assignment, especially among weakly differentiated populations (Campbell, Duchesne, & Bernatchez, 2003). Further, primers for AFLP marker are universal among species, whereas primers for SSR marker need to be developed for each specific species or genus and are generally not suitable for use in other taxa.
In preliminary experiments done before this study, both AFLP and SSR markers were adopted to detect genetic diversity; however, the results revealed that AFLP markers were obviously better than SSR ones in terms of the success rates they achieved and the polymorphism detected with them (Song, Li, Xu, Zhao, & Chen, 2015). SNP markers are third-generation molecular markers that have high genetic stability and diversity, but their use requires advanced technology and has high costs (Xu, Wang, Hou, & Li, 2015). Thus, AFLP markers are still practical markers to use when evaluating the genetic diversity and population structure of endangered or medicinal plants.
Paris L. is an important genus in the family Melanthiaceae of the order Liliales, which includes 27 species and more than 10 varieties (Li, Su, Zhang, & Yang, 2015). Among these species, Paris polyphylla Smith is a complicated species that is composed of 10 varieties.
Paris polyphylla Smith var. yunnanensis (Franch). Hand.-Mazz. is one of these varieties, and is mainly distributed in southwestern China (Liang & Soukup, 2000). This variety generally grows under forests and shrubs at an altitude of 1,400-3,100 m .
Moreover, P. polyphylla var. yunnanensis is the main plant used in the original formulation of Paridis Rhizoma, a famous traditional Chinese medicine that is used for detoxification and pain relief (Chinese Pharmacopoeia Committee, 2015). This medicine is a major raw material included in multiple Chinese patent medicines, including Gongxuening capsules and Yunnan Baiyao. Due to the high sales of the aforementioned Chinese patent medicines, the demand for Paridis Rhizoma increased by 20% per year, which resulted in the overcollection of wild resources because there was no effective supply available from cultivation . In consideration of the endangered status of its natural populations, this variety has been listed as a national key protected wild plant of the second class in China (Huang, Xiao, & Wang, 2011). Since the 1980s, P. polyphylla var. yunnanensis has been transplanted from wild habitats, and its artificial cultivation gradually became an important way to satisfy the demand for medication derived from this plant .
According to a recent report, the total cultivated area of this variety was about 2 km 2 in Yunnan Province, China, but the cultivated supply was still seriously insufficient to meet market demands because of this plant's long growing period (Li, 2016). Furthermore, in field surveys we found that most of the cultivated populations came from provenances with two origins, namely the wild populations of nearby mountains in Yunnan and other cultivation bases. The provenances for cultivated populations are not currently screened, and might include other species or varieties easily confused with this one. Revealing the genetic diversity and structure of P. polyphylla var. yunnanensis populations would be beneficial to its effective conservation, as well as the breeding of superior provenances for its large-scale cultivation.
Since it is an important medicinal plant, ISSR, AFLP, and SSR markers have previously been adopted to investigate the genetic diversity and structure of wild populations of P. polyphylla var. yunnanensis (He, Wang, Li, & Chen, 2007;Li et al., 2018;Song et al., 2015). These studies showed that this variety possessed relatively high genetic diversity and low genetic differentiation among populations. However, the majority of these studies focused on wild populations, and their sample sizes were extremely limited, so they might have been unable to accurately capture the true genetic variation and population structure of P. polyphylla var. yunnanensis. Although He et al. (2007) compared the genetic characteristics of wild and cultivated populations, only six populations were collected from Yunnan Province, and no populations from other parts of the variety's distribution were included. Therefore, the samples examined in previous studies likely provided relatively poor representations of the variety's distribution, so they could not reveal whether genetic differences existed between wild and cultivated populations of P. polyphylla var. yunnanensis.
In our study, we collected P. polyphylla var. yunnanensis from 32 populations within its main distributional range in China, including 15 wild and 17 cultivated populations. We then employed AFLP markers to explore (a) spatial patterns in genetic variation within and among both wild and cultivated populations; (b) genetic differences between the wild and the cultivated populations; and (c) the mixed provenances on which cultivated populations were based.

| Collection of samples from wild and cultivated populations of P. polyphylla var. yunnanensis
In this study, a total of 413 individuals of P. polyphylla var. yunnanensis were sampled from 32 populations, including 15 wild and 17 cultivated populations, from throughout this variety's main geographic distribution in China (Table 1 and Figures S1 and S2). Fresh and healthy leaves were collected from mature individuals for DNA extraction, and dried using silica gel as soon after collection as possible.
To avoid repeated sampling from the same female parent, the minimum distance among individuals sampled was 15 m. Over the course of conducting fieldwork, we found that it was extremely difficult to collect enough individuals to satisfy the theoretical requirements for population genetics (generally 20-30 individuals per population), so the numbers of samples collected in some populations were <10.

| DNA extraction and amplified fragment length polymorphisms
Total genomic DNA was extracted from about 0.5 g of leaf tissue from each individual plant using the CTAB protocol (Doyle, 1987), with some modifications. The quality of the extracted DNA was checked on 2.0% agarose gels. The original total genomic DNA sample was diluted to 20-50 ng/μl for subsequent amplified fragment length polymorphism (AFLP) experiments, and stored in a refrigerator at −20°C until further use.
Amplified fragment length polymorphism analysis was performed as described below, according to the standard protocol (Vos et al., 1995), with a few modifications. First, the extracted DNA (200 ng) was digested at 37°C for 3 hr using two restriction enzymes (10 U/μl EcoRI and 10 U/μl MseI) in a 20 μl reaction mixture, after which it was kept at 65°C for 3 hr to ensure enzyme inactivation.
Next, the digested products were ligated to 50 μM EcoRI and 50 μM (with annealing initiated at a temperature of 65°C, which was then reduced for each of 12 subsequent cycles by 0.7°C), 1 min at 72°C; and 23 subsequent cycles of 30 s at 94°C, 30 s at 56°C, and 1 min at 72°C. Finally, the selective amplification products were 10-fold diluted and mixed with the GeneScan 500 LIZ size standard, and then, the mixed products were detected using an ABI 3730XL automated DNA sequencer (Applied Biosystems™). In fragment analysis, FAM was adopted to label primers of AFLP markers. Fragments were analyzed with GeneMarker v.1.70 (Soft Genetics, LLS). Fragments were scored as either present (1) or absent (0) when the intensity was up to 100.

| Data analysis
Genetic diversity parameters were calculated using PopGen32 (Yeh, Yang, & Boyle, 1999), including the effective number of alleles (N E ), Nei's genetic diversity (H), Shannon's information index (I), and the percentage of polymorphic bands (PPB). ArcGIS v. 10.3 (Esri) was used to draw maps of the geographic distribution of the sampled wild and cultivated populations of P. polyphylla var. yunnanensis.
PopGen32 was used to construct an unweighted pair group method with arithmetic mean (UPGMA) dendrogram (Michener & Sokal, 1957) at the population level for all of the populations sampled, including both the wild and the cultivated populations. In addition, NTSYS-pc v. 2.1 (Rholf, 2000) was used to construct UPGMA dendrograms for the populations in Yunnan at the individual level to explore the probable genetic promiscuity of the bases of these cultivated populations. Pairwise distances among individuals were calculated in terms of Jaccard's coefficient values (Jaccard, 1908), and then, a matrix of Jaccard distances was employed to cluster the individual samples.
STRUCTURE v. 2.3.4 was also used to infer the genetic structure of the 32 wild and cultivated populations (Falush, Stephens, & Pritchard, 2007;Pritchard, Stephens, & Donnelly, 2000). To predict the optimum number of clusters (K), an admixture model and correlated allele frequencies were chosen for use in this analysis. At the same time, ten replicates of each simulation with 2-10 clusters (K = 2-10) were simulated through 10,000 Markov chain Monte Carlo (MCMC) steps by sampling after a burn-in period of 10,000 iterations. STRUCTURE HARVESTER was used to extract the relevant data from the structure results files and to generate CLUMPP input files (Earl & vonHoldt, 2011). Replicates were then built with CLUMPP v. 1.1.2, using the FullSearch option to determine the optimal K value. The membership of different cluster groups was visualized using DISTRUCT v. 1.1 (Jakobsson & Rosenberg, 2007;Rosenberg, 2003).
GenAlEx v. 6.503 was used to perform a principal coordinates analysis (PCoA) to reveal the relationships among populations (Peakall & Smouse, 2012), as well as a Mantel test to assess the correlation between the geographic distance (in km) and the genetic distance among different wild and cultivated populations, and also among only the cultivated populations in Yunnan (Mantel, 1967).
To be comparable with the UPGMA results, PCoA was similarly performed for all of the sampled populations, including both the wild and the cultivated populations together, as well as separately. Analysis of molecular variance (AMOVA) was performed to quantify the genetic variation at different hierarchical levels with 1,000 permutations using Arlequin version 3.5 (Excoffier & Lischer, 2010).
Both cases with grouping and no grouping of populations were considered in the AMOVA.

| Genetic diversity within wild and cultivated populations
In this study, a total of 669 loci were detected by five of the selected

| Population structure and genetic differentiation
In the present study, three UPGMA dendrograms were constructed for all 32 populations, the wild populations, and the cultivated populations, respectively, at the population level. In the first of these (Figure 1 and Figure S3), the 32 populations were roughly divided into three main groups (groups I, II, and III). Both the wild and cultivated populations in Guizhou formed group I, but they were clustered into two separate subgroups within this group. The remained populations, mainly those in Yunnan, were clustered into two groups (groups II and III). Group II was relatively complicated, including both wild and cultivated populations, whereas group III was only composed of six cultivated populations. Furthermore, the UPGMA dendrogram for the wild populations only also gathered these into three main subgroups, namely subgroups W-I, W-II, and W-III (Figure 2), which showed obvious genetic structuring that was also supported by STRUCTURE

| Genetic diversity of wild populations of P. polyphylla var. yunnanensis
Due to the impact of genetic drift and inbreeding, species with narrow geographic distributions and smaller-sized individuals generally show lower genetic diversity than more widespread ones (Li, Guan, Yang, Luo, & Chen, 2012;Willi, Van Buskirk, & Hoffmann, 2006). Therefore, endangered and rare species are regarded as being deficient in genetic variation (Hamrick & Godt, 1989 can likely be attributed to overcollecting and human disturbance, rather than to their biological characteristics. In this study, we mainly collected samples from wild populations in Yunnan and Guizhou, as well as one population in Sichuan (Table 1)

| Genetic diversity of cultivated populations and comparisons between wild and cultivated populations
Generally speaking, wild populations have higher genetic diversity than cultivated ones because of the bottleneck effects, founder effects, and genetic drift that occur during the process of demonstration and cultivation (Miller & Schaal, 2006;Wu, Li, & Huang, 2006). However, in this study, the genetic variation in cultivated populations was found to be higher than that in the yet resulted in any obvious loss of genetic variation. Our findings were consistent with those reported in a previous study (He et al., 2007). These results indicated that the seedlings of P. polyphylla var. yunnanensis currently used as the bases of most cultivation efforts have complicated genetic backgrounds; in other words, the individuals cultivated in the same place might have come from multiple wild source populations, or even different genetic lineages (He et al., 2007). This conclusion was also supported by our NTSYS analysis of the cultivated individuals in Yunnan ( Figure S4). The high genetic diversity found in the cultivated populations showed that the genetic resources of P. polyphylla var. yunnanensis could gain effective protection through cultivation, but did not satisfy the requirements for producing uniform and high-quality material for Paridis Rhizoma. Therefore, the next task for the cultivation of P. polyphylla var. yunnanensis should be screening and fostering superior cultivated varieties from the currently complicated, but rich, genetic resources available.

| Spatial genetic structure of wild and the cultivated populations
The spatial genetic structure of a species' populations reflects the interactions involved in their long-term evolutionary history, such as habitat fragmentation, genetic drift, type of mating system, and gene flow (Schaal, Haryworth, Olsen, Auscher, & Smith, 1998). In this study, the 15 wild populations sampled were divided into three subgroups (W-I, W-II, and W-III) based on the UPGMA dendrogram produced, which corresponded to an obvious geographic structure (Figure 2) that was in accordance with the results of PCoA and STRUCTURE ( Figure 4). Moreover, this geographic structure was also supported by the results of the Mantel test, which revealed that there was a significant, positive correlation between the genetic and geographic distances among populations (Figure 3). The W-I subgroup was composed of six populations located in Guizhou Province; meanwhile, the remaining two subgroups were comprised of populations that were mainly distributed in Yunnan Province, namely the central (W-II) and Yunnan. Second, P. polyphylla is an extremely complicated species that is composed of 10 varieties, including P. polyphylla var. yunnanensis (Liang & Soukup, 2000). Technicians working on cultivated populations are generally unable to discriminate the subtle differences in morphology among varieties due to a lack of taxonomic knowledge, and thus, they might introduce incorrect provenances, resulting in the production of cultivated lines with mixed provenances ( Figure   S4). Finally, compared with other well-demonstrated crops, P. polyphylla var. yunnanensis is still in the early stages of demonstration and cultivation, so it has not yet experienced a long history of selection and breeding (Brown, 1978;Ellstrand & Marshall, 1985;Hamrick & Godt, 1997;Hyten et al., 2006;Wang et al., 2017).
During fieldwork, we found that the cultivation of P. polyphylla var. yunnanensis was much more popular in Yunnan than in Guizhou, which was also reflected in our collection information F I G U R E 4 Principal coordinates analysis (PCoA) and Bayesian clustering by STRUCTURE of wild populations and the cultivated. (a1 and b1) A dot line circle represents a subgroup. "*" represents subgroup W-I; "+" represents subgroup W-II; "×" represents subgroup W-III; "▲" represents subgroup C-I; "•" represents subgroup C-II; "◆" represents subgroup C-III. (a2 and b2) Each color represents a subgroup  Table 1) According to the AMOVA results, there was no obvious overall genetic differentiation between the sampled wild and cultivated populations, as only 1.35% of the genetic variation found existed between these two types of populations (Table 3). Moreover, without grouping included in the AMOVA, the wild populations of P. polyphylla var. yunnanensis showed relatively low genetic differentiation (Φ ST = 18.83%) compared with that detected in previous studies of F. cirrhosa D. Don  and Camellia reticulata (Xin et al., 2017), as well as statistical data (Nybom, 2004).

| Conservation
Demand for traditional medicines is extremely high all over the world, and is still increasing (Hamilton, 2004). The supply of Paridis Rhizoma, an important herbal medicine, cannot currently meet this increasing demand of 20% each year  because the growth of the original plants from which it is made, namely P. polyphylla var. yunnanensis, is very slow; in fact, it takes more than 10 years for wild individuals to develop from the seed germination stage to harvest under natural growth conditions .
Artificial cultivation should be an effective method to protect the wild resources of this plant and guarantee their sustainable utilization (Allendorf & Luikart, 2007;Storfer, 1999). It was previously reported that in traditional cultivation methods, farmers collected seeds directly from wild populations and then cultivated them in similar habitats, which was a useful way to maintain the gene pools of these medicinal plants (Guo et al., 2007;He et al., 2009).  (Soldati, Fornes, Zonneveld, Thomas, & Zelener, 2013;Yao, Deng, & Huang, 2012). In the present study, the populations located in Guizhou Province showed higher genetic diversity compared with those in Yunnan, leading us to propose that the wild populations in Guizhou should be given priority protection and subjected to restricted harvest rules. Meanwhile, the HD population in Sichuan also possessed rich genetic variation at the population level (H = 0.2474), so southern Sichuan should be also be listed as a priority site for further study. Furthermore, reasonable manual intervention, such as removing seed coats and then sowing seeds in wild habitats, could improve the germination rates of seeds, and thus would benefit the renewal and maintenance of effective wild populations.

| CON CLUS IONS
In this study, AFLP markers were adopted as a means to explore and compare the genetic diversity and population structure of wild and although the cultivated populations were also clustered into three subgroups (C-I, C-II, and C-III), the latter two subgroups, which were mainly located in Yunnan, did not show significant correlations between the genetic and geographic distances among populations within them. This study indicated that cultivated seedlings had complicated and confusing origins, so the selection and breeding of varieties with good and homogenous quality should be listed as a priority for the future cultivation of P. polyphylla var. yunnanensis.
This study not only explored the genetic diversity and population structure of wild populations of P. polyphylla var. yunnanensis, but also revealed that cultivated populations presented some defective provenances. These findings could be beneficial to both the conservation of wild resources and the development of more robust cultivation approaches for this medicinal species.